Redox-sensitive GFP in Arabidopsis thaliana is a quantitative biosensor for the redox potential of the cellular glutathione redox buffer

Authors


(fax 49 6221 54 5859; e-mail ameyer@hip.uni-hd.de).

Summary

The cellular glutathione redox buffer is assumed to be part of signal transduction pathways transmitting environmental signals during biotic and abiotic stress, and thus is essential for regulation of metabolism and development. Ratiometric redox-sensitive GFP (roGFP) expressed in Arabidopsis thaliana reversibly responds to redox changes induced by incubation with H2O2 or DTT. Kinetic analysis of these redox changes, combined with detailed characterization of roGFP2 in vitro, shows that roGFP2 expressed in the cytosol senses the redox potential of the cellular glutathione buffer via glutaredoxin (GRX) as a mediator of reversible electron flow between glutathione and roGFP2. The sensitivity of roGFP2 toward the glutathione redox potential was tested in vivo through manipulating the glutathione (GSH) content of wild-type plants, through expression of roGFP2 in the cytosol of low-GSH mutants and the endoplasmic reticulum (ER) of wild-type plants, as well as through wounding as an example for stress-induced redox changes. Provided the GSH concentration is known, roGFP2 facilitates the determination of the degree of oxidation of the GSH solution. Assuming sufficient glutathione reductase activity and non-limiting NADPH supply, the observed almost full reduction of roGFP2 in vivo suggests that a 2.5 mm cytosolic glutathione buffer would contain only 25 nm oxidized glutathione disulfide (GSSG). The high sensitivity of roGFP2 toward GSSG via GRX enables the use of roGFP2 for monitoring stress-induced redox changes in vivo in real time. The results with roGFP2 as an artificial GRX target further suggest that redox-triggered changes of biologic processes might be linked directly to the glutathione redox potential via GRX as the mediator.

Introduction

The last two decades have seen an exponential increase of evidence for redox-dependent components involved in the signaling of environmental changes. Thioredoxins (TRXs) have been described as redox-active key players in regulating numerous processes, including carbon metabolism, DNA synthesis, sulfur assimilation, and regulation of transcription factors (Vieira Dos Santos and Rey, 2006). Recently, it was shown that NPR1, an essential regulator of plant systemic acquired resistance, needs to be reduced to enter the nucleus and to trigger the expression of pathogen-related genes (Mou et al., 2003). It has also been suggested that redox-dependent signaling pathways might interact with Ca2 +- and ABA-dependent signaling to further modulate signals, and thus help ensure specificity (Baier et al., 2004; Gomez et al., 2004).

One of the key metabolites involved in maintaining a reduced intracellular redox milieu is the tripeptide glutathione, which can be reversibly converted from the reduced form (GSH) to the oxidized form (GSSG). Glutathione has long been suggested to be part of signaling cascades transducing environmental signals to the nucleus, where the signal might result in altered gene expression (May et al., 1998). Several Arabidopsis mutants of the glutamate cysteine ligase (GSH1), the first enzyme of GSH biosynthesis, with significantly reduced GSH levels have been described. The mutants cad2 and rax1 contain only 15–30% (cad2) or 20–50% (rax1) of wild-type GSH (Ball et al., 2004; Cobbett et al., 1998). The concomitant change of gene expression has been linked to the key role of glutathione metabolism during stress (Ball et al., 2004). If the GSH level is further reduced as a result of more severe mutations of GSH1, plant development is arrested, as seen in the rml1 mutant (Vernoux et al., 2000), or the reduction can even result in embryonic lethality in gsh1 knock-outs (Cairns et al., 2006).

The accumulation of reactive oxygen species (ROS) is determined by the antioxidative system, and thus the plant antioxidant status plays a pivotal role in removing ROS and in setting thresholds for general defence responses provoked by abiotic and biotic stress (Foyer and Noctor, 2005). An increase in intracellular ROS might lead to altered transcriptional and translational events. To be specific, such redox signals need to be dynamic in their amplitude and in their temporal pattern. In the past it was only possible to extract the tissue and determine the redox status of certain metabolites. For ROS and the cellular redox environment, dynamic imaging has so far been hampered by the availability of suitable probes. Different ROS species have been visualized in situ, but because of the chemical nature of the dyes used these measurements are always static and provide very limited evidence about redox dynamics (Fricker et al., 2006). The recent development of redox-sensitive YFP (rxYFP, Østergaard et al., 2001) and GFP probes (roGFP, Hanson et al., 2004) might now overcome these limitations. Redox-dependent changes of the excitation efficiency of two excitation peaks renders roGFP ratiometric, providing a principle advantage over the single-excitation rxYFP (Björnberg et al., 2006a). Recently, Jiang et al. (2006) reported that roGFP1 expressed in Arabidopsis responds to reducing and oxidizing conditions in both the cytosol and the mitochondria. For the correct interpretation of the readout of the redox sensor, it is essential to know whether roGFP equilibrates with a strong preference for a specific cellular redox buffer compared with others. In the case of such a preference, roGFPs would effectively be sensors for this particular redox pool rather than the overall redox poise of the cell (Björnberg et al., 2006a,b). The kinetics for the response of roGFP2 expressed in HeLa cells to external application of H2O2 is much faster than the response in vitro (Dooley et al., 2004), suggesting that roGFP2 interacts with a yet unknown catalytic protein. When expressed in yeast cells rxYFP readily equilibrates with the glutathione redox buffer through the action of glutaredoxin (GRX) (Østergaard et al., 2004). The catalytic function of yeast GRX1 in vivo was unequivocally shown by using a double knock-out of the two cytosolic dithiol-type GRXs (Østergaard et al., 2004). In Arabidopsis the situation is far more complicated because of the presence of at least 30 GRXs (Lemaire, 2004; Rouhier et al., 2004, 2006) and several other candidate redox partners, such as thioredoxins and protein disulfide isomerases (Buchanan and Balmer, 2005). The redox potential of the glutathione buffer is dependent on both the total concentration of glutathione and the degree of oxidation (O×DGSH; Meyer and Hell, 2005). If roGFP interacts with one or more GRXs and equilibrates with the redox potential of glutathione, it would thus respond to both changes in the degree of oxidation and changes in the total quantity of glutathione in the subcellular compartment the probe is targeted to. This hypothesis was tested by characterization of roGFP2 both in vitro and in vivo after expression in the cytosol of Arabidopsis thaliana. By using pharmacologically and genetically modified glutathione levels it was shown that roGFP2 responded specifically to the glutathione redox buffer, and is highly sensitive to nanomolar concentrations of GSSG. Further in vitro experiments with recombinant GRX confirmed GRX as the missing link between glutathione and roGFP.

Results

Redox-dependent changes of roGFP fluorescence can be visualized by CLSM

Different versions of roGFPs have been generated and tested for their redox-dependent fluorescence (Dooley et al., 2004; Hanson et al., 2004). Although roGFP versions derived from wild-type GFP contain a cryptic intron preventing reliable expression in plants (Haseloff et al., 1997), roGFP versions based on mammalian codon usage circumvent this problem and can be expressed in plants at high levels. The versions roGFP1 and roGFP2 differ in the S65T mutation, which is present in roGFP2. This mutation causes a distinct shift in the excitation spectrum, whereas the general ratiometric properties are retained. The spectral difference has implications for the maximum achievable dynamic range, particularly when the probes are to be excited with distinct laser wavelengths. The excitation spectra reported by Dooley et al. (2004) indicate that for excitation at 405 and 488 nm on a confocal laser scanning microscopy (CLSM), roGFP2 offers a much larger dynamic range and is thus more appropriate than roGFP1. Arabidopsis roots expressing roGFP2 in the cytosol showed strong fluorescence when excited at 405 and 488 nm, respectively (Figure 1a). The ratio of the two images could not be significantly modified by perfusing the root with DTT solution, indicating that roGFP2 was almost fully reduced to about 95% in the undisturbed root (not shown). With the goal of investigating the feasibility of dynamic redox measurements plants were successively treated with sublethal concentrations of H2O2 and DTT. The addition of 1 mm H2O2 to the incubation solution resulted in immediate oxidation of roGFP2 in roots, reflected by an increase of the fluorescence ratio from 0.45 to 1.7 (Figure 1a,b,d). This oxidation in vivo was much faster than the response of isolated roGFP2 to 10 mm H2O2in vitro (Figure 1e). No further change of the fluorescence ratio could be observed between 2 and 10 min after the addition of H2O2. The addition of 2 mm DTT after H2O2 treatment led to a decrease of the fluorescence ratio indicating reduction of roGFP2 (Figure 1c,d). The kinetics of this reduction were dependent on the concentration of DTT used in the incubation medium, with higher concentrations of DTT causing faster reduction (not shown). Differences in absolute ratio values and the maximum achievable dynamic range of roGFP2 between in vivo and in vitro measurements result from the fact that different excitation wavelengths were used (see Experimental procedures). The use of different excitation wavelengths does not affect the quantitative analysis of ratio values and the calculation of redox potentials.

Figure 1.

 Redox-dependent fluorescence of redox-sensitive GFP2 (roGFP2) in the cytosol of Arabidopsis root cells is fully reversible. Stacks of images were taken by CLSM with excitation at 405 and 488 nm, respectively. Maximum projections of these stacks were used for the calculation of the ratio images.
(a) Root tip before the addition of H2O2. The color scale for the ratio values indicates reduced roGFP2 in blue and oxidized roGFP2 in yellow. Scale bar = 100 μm.
(b) 2 min after the addition of 1 mm H2O2.
(c) Ratio image taken 5 min after the addition of 2 mm DTT to the root previously oxidized with H2O2.
(d) Time course for the ratio values calculated from CLSM images during the course of successive oxidation and reduction (mean ± SD; n = 4).
(e) Response of purified reduced roGFP2 to oxidation by 10 mm H2O2in vitro. The ratio (black curve) was calculated through division of fluorescence at 390 nm (blue) by the fluorescence at 480 nm (red) (n = 5).

Glutaredoxin mediates the redox equilibration of roGFP2 with the glutathione redox buffer

The faster response of roGFP2 to H2O2in vivo compared with the response of isolated roGFP2 in vitro strongly suggested that in living cells roGFP2 is not oxidized by H2O2 directly, but rather by other proteins catalyzing the redox reaction. In search for such mediating factors, different proteins known to be active in the formation and cleavage of protein disulfides were tested for their ability to reduce roGFP2 in vitro. Recombinant poplar TRX h3, in conjunction with NADPH and NADPH-dependent TRX reductase (NTR), did not affect the oxidized roGFP2 (Figure 2a). In contrast to the very slow reduction of oxidized roGFP2 by GSH alone, the presence of 5 μmAtGRX C1 led to almost instantaneous reduction after the addition of GSH (Figure 2a,b). Within 4 min the initially oxidized roGFP2 was reduced to the maximum level, reflected by a decrease in the 390/480-nm excitation ratio from 1.9 down to about 0.5. The maximum reduction achieved here, however, was not equal to the complete reduction of roGFP2 because the GSH solution used contained some residual GSSG. The addition of 0.1–1.0 mm GSSG to the reaction buffer was reflected by a gradual increase of the 390/480-nm ratio for roGFP2 (Figure 2b).

Figure 2.

 Glutaredoxin catalyzes the transfer of electrons from GSH to roGFP2. The excitation ratio 390/480 nm with emission at 510 nm was followed over time. Reducing agents were added to a solution of oxidized roGFP2 and the respective catalyzing enzymes 3 min after the start of the measurement.
(a) Time course for reduction with 100 μm NADPH (bsl00001), 100 μm NADPH + 10 μm poplar TRX h3 + 10 μm NTR (bsl00000), 20 mm GSH (bsl00066) and 20 mm GSH + 5 μmAtGRX C1 (bsl00041). TRX and NTR were confirmed to be active in an independent 5,5′-dithiobis-(2-nitrobenzoic acid) (DTNB) assay (not shown).
(b) Time course for reduction of roGFP2 in the presence of 5 μmAtGRX C1 with 1 mm GSSG (bsl00000), 20 mm GSH (bsl00066) and with different GSH:GSSG ratios: (△) 20 mm GSH/1 mm GSSG, (bsl00046) 20 mm GSH/0.5 mm GSSG, (bsl00001) 20 mm GSH/0.1 mm GSSG and (bsl00041) 20 mm GSH.
(c) Time course for reduction of roGFP2 alone by addition of different GSH:GSSG ratios after 2 min: (bsl00041) 20 mm GSH, (bsl00000) 20 mm GSH/0.1 mm GSSG, (⋄) 20 mm GSH/0.2 mm GSSG, (△) 20 mm GSH/0.57 mm GSSG. The inset shows the initial part of the reduction at higher resolution. Ratios in panel (c) are slightly higher than in panels (a) and (b) because of different instrument settings.

The dependence of roGFP2 fluorescence on the degree of oxidation of the glutathione buffer was further examined in vitro (Figure 2c). In the absence of GRX, the addition of glutathione buffers with different GSH:GSSG ratios to roGFP2 very slowly reduced the calculated fluorescence ratio. Over a time course of several hours the fluorescence ratio was reduced to plateau values, dependent on the quantity of oxidized GSSG in the buffer solution, indicating gradual equilibration of the redox potential of roGFP2 with the redox potential of the glutathione buffer.

If roGFP2 is a true redox sensor for the redox potential of the glutathione redox buffer mediated by GRX, then the redox potentials of glutathione and roGFP2 should equilibrate according to Eqn 1 (see Experimental procedures), which describes the redox equilibrium for the reaction depicted in Figure 3a. Newly purchased GSH already contains significant quantities of GSSG, which prevents roGFP2 from full reduction (Figure 2; Østergaard et al., 2004). To overcome this limitation, freshly prepared GSH solution was treated with recombinant glutathione reductase (GR) and NADPH to drive the reduction so as to reduce the residual GSSG to the lowest possible concentration (Figure 3a). Incubation of roGFP2 with different concentrations of this GR-treated GSH solution resulted in an exponential decay of the fluorescence ratio, calculated from fluorescence excited at 390 and 480 nm (Figure 3b). Starting with 1 μm GSH roGFP2 was still almost fully oxidized, as seen from the high 390/480-nm ratio. With increasing levels of GSH the 390/480-nm ratio gradually decreased, indicating an almost complete reduction at 2.5 mm GSH.

Figure 3.

 The redox potential of roGFP2 equilibrates with the redox potential of the glutathione buffer. roGFP2 and GRX were incubated with GSH concentrations from 1 μm to 2.5 mm in the presence of glutathione reductase and 100 μm NADPH.
(a) The redox equilibrium between roGFP2 and glutathione is shifted toward more reduced roGFP2 because glutathione reductase continuously reduces GSSG. A stable equilibrium is reached when glutathione reductase can no longer reduce the residual O×DGSH.
(b) 390/480-nm ratios measured after equilibration of roGFP2 with different GSH concentrations plotted against the respective GSH concentration used.
(c) Theoretical curves for O×DroGFP2 calculated from the Nernst equation assuming different O×DGSH for the glutathione buffer (Eqn 5): bsl00041, 0.0001%; ⋄, 0.0005%; bsl00001, 0.001%; □, 0.002%; bsl00066, 0.003%; bsl00046, 0.01%; +, 0.1%; *, 1%.
(d) O×DroGFP2 calculated from the measured ratio values (bsl00000) (Eqn 6) and O×DroGFP2 calculated on the basis of hypothetical O×DGSH values (Eqn 5) plotted against the measured 390/480-nm ratio. bsl00046, O×DGSH = 0.01%; bsl00041, O×DGSH = 0.0001%; solid line, O×DGSH = 0.002%. Only the curve for an assumed O×DroGFP2 = 0.002% approximates the O×DroGFP2 calculated from the measured ratio values.
(e) Plotting O×DroGFP2 calculated from the measured ratio values (Eqn 6) against the redox potentials calculated for the used glutathione concentrations with an assumed O×DGSH of 0.002% (Eqn 4) results in the GSH-dependent titration curve for roGFP2.

Determination of the glutathione redox potential bears the problem that the redox potential in this case is dependent on the total concentration of glutathione and the degree of O×DGSH, which cannot be determined simultaneously. Provided that the redox potentials of roGFP2 and glutathione are in equilibrium, O×DroGFP2 can be calculated according to Eqn 5 as long as the total levels of GSH and O×DGSH are known. Given that in the in vitro experiments the total GSH concentrations are known exactly, a number of exponential decay curves can be calculated for assumed values of O×DGSH (Figure 3c). According to Eqn 6 the O×DroGFP2 is nonlinearly dependent on the measured 390/480-nm ratios. This nonlinear relationship is evident from the plot of O×DroGFP2 against the measured ratios (Figure 3d, □). Plotting the values for O×DroGFP2 calculated on the basis of assumed values for O×DGSH on top of the measured values immediately shows that the measured data could be approximated best by the curve calculated with O×DGSH = 0.002%. Curves calculated for higher or lower O×DGSH deviate significantly from the measured data. Even the best approximation of the measured data still shows some deviations. This might result from the fact that O×DGSH achieved by GR is not perfectly constant at all GSH concentrations used. These deviations, however, are so small that an average O×DGSH of 0.002% can be assumed for the experimental GSH solution in vitro. Finally, the plot of O×DroGFP2 calculated from the measured ratio values according to Eqn 6 against the redox potentials for glutathione calculated on the basis of an O×DGSH of 0.002% resulted in a typical sigmoidal titration curve (Figure 3e).

Depletion of glutathione causes oxidation of the cytosol

To test whether roGFP2 is also responsive to changes of the glutathione status in vivo, the cellular GSH pool was artificially depleted in different pharmacologic experiments. First, Arabidopsis seedlings were incubated with 1 mm 1-chloro-2,4-dinitrobenzene (CDNB) for 16 min. The electrophilic xenobiotic CDNB is detoxified via conjugation to GSH, resulting in fast and complete depletion of the cellular GSH pool (Hartmann et al., 2003). Within 16 min of CDNB treatment the fluorescence ratio observed in roots increased from 0.38 ± 0.03 to 1.08 ± 0.12, indicating progressive oxidation accompanying the depletion of GSH (Figure 4a–e). Oxidation of roGFP2 followed a progress curve with only very little change toward the end of the 16-min incubation period. Despite the fact that the ratio increased only to intermediate values, the addition of 10 mm H2O2 to CDNB-treated roots caused only a minor additional increase of the ratio to 1.3, implying that the probe could not be further oxidized at this stage (Figure 4e).

Figure 4.

 Fast depletion of GSH in Arabidopsis roots causes immediate oxidation of roGFP2. Arabidopsis roots expressing roGFP2 in the cytosol were treated with 1 mm CDNB for 16 min and then imaged for their redox response.
(a) Control before CDNB addition (t = 0 min).
(b) Redox response at 2.5 min after the addition of CDNB.
(c) Redox response at 5 min after the addition of CDNB.
(d) Redox response at 16 min after the addition of CDNB. Scale bar = 100 μm.
(e) Time course for the oxidation of roGFP2 after the addition of CDNB. After the ratio reached a plateau 10 mm H2O2 was added to test for further oxidation. Fluorescence ratios were calculated from images (a–d) collected with excitation at 405 and 488 nm, respectively (mean ± SD; n = 8).

In a second series of experiments, Arabidopsis seeds were germinated on agar plates supplemented with 1 mm l-buthionine-sulfoximine (BSO) as a specific inhibitor of GSH1 (Vernoux et al., 2000). Seven days after germination the seedlings showed arrested growth. This phenotype resembled the phenotype of the rml1 mutant that is known to contain severely reduced glutathione levels (Figure 5b; Vernoux et al., 2000). Inhibition of GSH biosynthesis resulted in the depletion of the cellular GSH pool in root cells close to zero, and thus the absence of fluorescent labeling by the GSH-specific dye MCB (Figure 5e; Cairns et al., 2006; Meyer et al., 2001). Ratio imaging of the roots clearly showed that under these conditions, i.e. in the absence of reduced glutathione, roGFP2 was fully oxidized with a 405/488-nm ratio of 1.8 ± 0.2 compared with 0.47 ± 0.08 in control seedlings germinated in the absence of BSO (Figure 5a,b,g). Simultaneous imaging of propidium iodide (PI) resulted in almost exclusive cell wall labeling, indicating full viability of the BSO-treated roots. The addition of H2O2 to these roots could not shift the fluorescence to a more oxidized ratio signal (data not shown).

Figure 5.

 Inhibition of glutathione biosynthesis causes oxidation of roGFP2. Arabidopsis seeds transformed with roGFP2 in the cytosol were germinated on agar plates with or without 1 mm BSO for 7 days and imaged for the redox status of roGFP2 and GSH levels. All images are maximum projections from stacks of serial optical sections.
(a–c) Ratiometric analysis of control roots (a), roots grown on BSO (b), or roots grown on BSO and then supplemented with 1 mm glutathione ethyl ester (GSH-OEt) at pH 5.8 for 14 h (c). The color scale shows the pseudocolor coding for reduced (blue) and oxidized roGFP2 (yellow); scale bars = 100 μm. Absence of nuclei labeling with propidium iodide (PI) clearly indicated that all cells remained viable during the imaging procedure.
(d–f) Labeling of intact roots with 100 μm MCB to visualize cytoplasmic GSH (green) and PI as counter stain (red). (d) Control. (e) Root grown on BSO. (f) Root grown on BSO and subsequently supplied with GSH-OEt for 14 h.
(g) Quantitative analysis of ratios determined in images shown in (a–c; mean ± SD, n = 3).

Rescue of seedlings germinated on BSO with glutathione ethyl ester (GSH-OEt) resulted in the uptake of GSH, and thus resulted in the ability to label the roots with MCB (Figure 5f). Strong fluorescence after incubation of GSH-OEt-treated roots with MCB indicated the restoration of the cellular glutathione pool (Figure 5d–f). The remaining difference in the labeling pattern compared with non-BSO-treated control roots can be attributed to the inhibitory effect of BSO on cell growth and a larger fraction of small, meristematic cells lacking a large central vacuole. GSB formed in these cells is transported into other cells and sequestered to the vacuole there (Fricker et al., 2000). Feeding of GSH-OEt to BSO-grown roots for 14 h concomitantly resulted in the partial recovery of the cellular redox state (Figure 5c,g).

In the GSH-deficient cad2 mutant roGFP2 is more oxidized than in wild type

Expression of roGFP2 in the cytosol of cad2 resulted in strong green fluorescence (raw data not shown). Ratiometric analysis of the fluorescence, however, revealed a ratio of 1.1 ± 0.12 compared with 0.56 ± 0.10 in wild-type plants grown and imaged under exactly the same conditions (Figure 6a–c). In the cad2 mutant the GSH content was decreased to about 20% of wild-type GSH, but because the GSSG content in cad2 was also decreased, the degree of oxidation, as measured in plant extracts, remained unchanged (Figure 6d). Thus, even in the absence of an altered redox ratio for the glutathione couple in plant extracts, roGFP2 was significantly more oxidized. We conclude that roGFP2 under these conditions is therefore sensing the cytosolic glutathione concentration.

Figure 6.

 Partial oxidation of roGFP2 expressed in the cytosol of the Arabidopsis cad2 mutant.
(a and b) Ratio images calculated from fluorescence images collected with excitation at 405 and 488 nm, respectively, of 6-week-old Arabidopsis leaves. Pseudocolor coding as in Figure 1.
(a)
Wild-type.
(b) cad2. Both images show expression of roGFP2 in cytosol and nucleus (arrowhead). Scale bar = 10 μm.
(c) Overall ratio values for the images shown in panels (a) and (b) (mean ± SD; n = 4).
(d) Contents of GSH, GSSG and the degree of oxidation of the glutathione pool in wild-type (blue) and cad2 (red) plants analyzed by HPLC (mean ± SD; n = 5).

roGFP2 is oxidized within the ER lumen

The glutathione redox potential is dependent on both the total concentration of glutathione and the degree of oxidation (Eqn 1). After observing partial oxidation of roGFP2 in the cytosol of cad2 mutants, in which only the total quantity of glutathione but not the degree of oxidation was changed (Figure 5), we also aimed to investigate the response of roGFP2 to altered oxidation of the glutathione redox buffer in vivo. To avoid possible artifacts resulting from uncontrollable oxidation based on treatment with oxidizing compounds, roGFP2 was expressed in the ER lumen, which has been reported earlier to contain a highly oxidized glutathione buffer (Hwang et al., 1992). Transient expression in the ER lumen of tobacco leaf cells resulted in a typical reticulate ER network including the nuclear ring visible in the images (Figure 7b). In this case roGFP2 was fully oxidized, consistent with the hypothesis that roGFP2 is sensing the glutathione redox potential (Figure 7b). When expressed in the cytosol of tobacco leaf cells roGFP2 was almost fully reduced, as previously shown for wild-type Arabidopsis cells (Figure 7a).

Figure 7.

 roGFP2 reports oxidizing conditions in the ER lumen. roGFP2 driven by a 35S promoter was transiently expressed in tobacco leaf cells.
(a) roGFP2 targeted to the cytosol.
(b) roGFP2 targeted to the ER lumen. Arrowheads point to the characteristic nuclear ring typical for ER labeling. The color scale shows the pseudocolor coding for reduced (blue) and oxidized roGFP2 (yellow). Scale bars = 50 μm.

Wound-induced oxidative bursts can be monitored by roGFP2

All results indicate that roGFP2 directly measures the redox potential of the cellular glutathione pool, with GRXs as mediating factors catalyzing the equilibration of the redox potentials. To further investigate whether roGFP2 can be used for the direct observation of redox changes elicited by abiotic stress, Arabidopsis plants expressing the sensor in the cytosol were mechanically wounded. Wounding is well known to elicit the massive production of H2O2 (Neill et al., 2002). H2O2 production in leaf cuttings were visualized histochemically through the formation of colored precipitates after incubation of leaf pieces with 3,3′-diaminobenzidine-tetrahydrochloride (DAB; Figure 8a). H2O2 production was restricted to cells close to the wound surface. To further test whether localized formation of H2O2 and H2O2-dependent redox changes can be monitored with roGFP2, transgenic plants with cytosolic expression of the biosensor were wounded in the same way. Ratiometric analysis of images taken from cells close to the wound surface and cells in the middle of a 9-mm2 leaf piece 2 min after wounding showed that the sensor was oxidized in cells close to the wounded area, whereas it remained in the reduced state in cells further away from the wound (Figure 8b). Incubation of leaf pieces with DTT rapidly reverted the sensor to the reduced state in all cells (Figure 8c).

Figure 8.

 Wounding causes oxidation of cells close to the wound surface. Small pieces (9 mm2) were cut from the leaves of transgenic plants expressing roGFP2 in the cytosol and were investigated for H2O2 production and response of the roGFP redox sensor.
(a) DAB staining for H2O2, scale bar = 1 mm; (b and c) redox response of cytosolic roGFP2 in leave tissue cuttings in the absence (b) or presence (c) of DTT. All images show single optical sections in which the nucleus is most prominent. The color scale for the ratio values indicates reduced roGFP2 in blue and oxidized roGFP2 in yellow. The image labels 1 and 2 refer to the respective positions at the wound surface and in the middle of the cut leaf piece, as indicated in panel (a).

Discussion

Glutaredoxin catalyses redox equilibration between glutathione and roGFP2

With increasing awareness about the importance of redox-dependent mechanisms for the regulation of normal plant metabolism and for stress-related signaling, the ability to directly monitor dynamic redox changes in living cells becomes an indispensable analytical tool. The recently developed GFP-based redox sensors now offer a new opportunity to exploit redox-dependent formation of an engineered GFP disulfide bridge to directly visualize the redox change non-invasively (Dooley et al., 2004; Hanson et al., 2004; Østergaard et al., 2001, 2004). These probes can be expressed in cells and targeted to all subcellular compartments. Full oxidation with H2O2 and subsequent reduction by DTT could be shown for roGFP1 and roGFP2 expressed in HeLa cells (Dooley et al., 2004; Hanson et al., 2004) and for roGFP1 expressed in Arabidopsis roots (Jiang et al., 2006). The kinetics of these redox reactions in vivo were found to be faster than the same reactions for purified protein in vitro, similar to the kinetic differences for roGFP2 reported in this work. This suggests that, when expressed in cells, roGFPs interact with redox-active enzymes that catalyze the transfer of electrons. For a full interpretation of fluorescence changes in living cells, especially under physiologic conditions, and to link the fluorescence signal to signaling cascades operating in living cells, it is thus mandatory to know which cellular redox component roGFP2 interacts with. For rxYFP expressed in yeast it has been shown that formation and release of the disulfide bridge is catalyzed by GRX (Østergaard et al., 2004). Our studies with recombinant AtGRX C1 show that plant GRXs are capable of transferring electrons from GSH to roGFP2. On the other hand, the TRX tested in this study (cytosolic poplar TRX h3) is unable to reduce this disulfide bridge on the same timescale when electrons are supplied via NADPH and the NADPH-dependent TRX reductase. It thus appears that in vitro in our experimental conditions and on our timescale, roGFP2 is specific for GRX and not affected by TRX. Similarly, it has been reported that the structurally similar rxYFP does not interact with any of the yeast TRXs (Østergaard et al., 2004). TRXs share structural features with GRX for correctly positioning substrate cysteine residues, but possess a unique structural element that allows recognition of protein disulfides (Maeda et al., 2006). Given that the structure of TRXs is highly conserved among species, it is likely that the inability to transfer electrons to roGFP2 is a general feature of TRXs. If GRXs are the only interaction partner for roGFP2 in vivo, it is then likely that roGFP2 expressed in the cytosol can reflect the GSH-GRX-mediated redox state of the cytosolic glutathione pool. Whether roGFP2 is as selective in other subcellular compartments containing GRXs (mitochondria, chloroplasts) remains to be determined experimentally by testing mitochondrial and chloroplastic GRXs in in vitro tests.

Glutathione oxidation can be determined by roGFP2

Characterization of roGFP2 and the measurement of the redox potential were originally performed with redox buffers consisting of DTTred and DTTox, the redox potentials of which are solely dependent on the degree of oxidation of DTT (Hanson et al., 2004). Calibration of roGFP2 against a glutathione buffer introduces the problem that the redox potential of glutathione is dependent on both the total level of glutathione and the degree of oxidation (Meyer and Hell, 2005). Although the concentration of glutathione can be exactly determined, the exact degree of oxidation remains unknown. Our results, however, indicate that because the redox potential of roGFP2 is already known the degree of oxidation of a glutathione solution can be determined by measuring the redox response of roGFP2 in a series of different glutathione dilutions. Because of its midpoint redox potential of −280 mV, the fluorescence of roGFP2 changes very quickly with minute quantities of GSSG. This high sensitivity can be exploited for redox measurements in vivo.

roGFP2 fluorescence indicates nanomolar GSSG concentrations in the cytosol

The cytosolic glutathione concentration in wild-type Arabidopsis root tips has been shown to be in the range of 2–3 mm (Fricker et al., 2000). In contrast, the glutathione concentration in root tips of rml1 seedlings is reduced to 40–200 μm (Cairns et al., 2006). From the Nernst equation for the glutathione redox buffer one can calculate that a decrease of the cytosolic glutathione pool from 2.5 mm down to 50 μm would cause the redox potential to become 50 mV less negative, even if the degree of oxidation of the glutathione pool is not changed (Meyer and Hell, 2005). Based on the Nernst equation it can also be calculated that roGFP2, with its redox potential of −280 mV at pH 7.0 (Dooley et al., 2004), will be in the 95% reduced dithiol state when the environmental redox potential is −318 mV, and that 95% of the protein will be in the oxidized disulfide state when the surrounding redox potential is −242 mV. As inhibition of GSH biosynthesis by 1 mm BSO affects GSH1 activity more severely than the rml1 point mutation, it makes sense that complete inhibition of GSH biosynthesis by BSO results in the observed full oxidation of roGFP2. Although depletion of GSH with BSO is a very slow process, depletion by incubation with CDNB is supposed to be very fast. CDNB treatment also resulted in oxidation of roGFP2. The observed intermediate ratio values might be explained in different ways. First of all it is possible that the GSH pool was not yet fully conjugated after 16 min of CDNB incubation. This, however, seems unlikely as lower CDNB concentrations resulted in similar intermediate ratio values. Another possibility is that because of very fast depletion of GSH within 16 min of CDNB incubation, insufficient quantities of GSSG were available to act as electron acceptors for the oxidation of roGFP2. In this case, however, further oxidation with H2O2 should have been possible, which was not the case. The third possibility might be that CDNB reacts with the cysteine residues of roGFP2 effectively blocking part of the protein in a ‘pseudo-reduced’ state. Irreversible alkylation of cysteine residues by CDNB has indeed been shown for thioredoxin reductase (Arnér et al., 1995; Nordberg et al., 1998). Cysteine alkylation of roGFP2 would block disulfide bridge formation, and thus explain why only very little further oxidation was achieved following the addition of H2O2 after the CDNB treatment.

The possibility that compounds supplied externally to deplete GSH might interfere with roGFP2 or impose any kind of oxidative stress was ruled out by expression of roGFP2 in the partially GSH-deficient mutant cad2. When expressed in the cytosol of cad2 mutants, roGFP2 fluorescence excited at 405 and 488 nm resulted in an increased fluorescence ratio, indicating that about 20% of the roGFP2 was oxidized under these conditions. This degree of oxidation of roGFP2 would thus indicate a glutathione-based redox potential of about −300 mV in cad2 at pH 7.0. Analysis of plant extracts confirmed that the degree of oxidation of the glutathione pool was not affected in cad2. A reduction of cytosolic glutathione from 2.5 mm in wild-type plants down to 0.5 mm in cad2 (a reduction to 20%), would shift the glutathione redox potential by about 20 mV to less negative values. Thus, this expected shift in the redox potential calculated from the Nernst equation is in very good agreement with the redox potential of −300 mV measured with roGFP2 in the mutant, compared with the resting redox potential of −320 mV measured in the wild type.

Almost complete reduction of roGFP2 by GSH would thermodynamically only be possible if the redox potential of the glutathione redox buffer is significantly more negative than the midpoint potential of the sensor of −280 mV. To achieve such negative redox potentials, with concentrations of total cytosolic glutathione between 1 and 5 mm, the concentration of GSSG needs to be in the low nanomolar range, and thus would be clearly below the GSSG concentrations generally determined by extraction and HPLC-based methods. In vitro, an almost complete reduction of roGFP2 was achieved with 2.5 mm GSH after prior incubation of the GSH solution with GR. The GR activity reduced the degree of oxidation of the GSH solution to 0.002%, which is equivalent to 25 nm GSSG. Assuming a similar activity of GR in vivo, a 2.5 mm cytosolic glutathione buffer would thus also contain 25 nm GSSG. According to the Nernst equation the redox potential of such a glutathione buffer would be −310 mV. Similarly, a glutathione concentration of 0.5 mm in cad2 would contain only 4 nm GSSG, and the redox potential of this glutathione buffer can be calculated to be −290 mV. Both values are slightly less negative than the in vivo measurements with roGFP2, but they are both calculated on theoretical assumptions for total GSH and O×DGSH. It is also conceivable that under non-stress conditions GR is working even more efficiently in vivo than in vitro, and thus might reduce the O×DGSH even further than 0.002%. Despite the ambiguity about the exact concentration of cytosolic GSH and its actual degree of oxidation, all measurements and theoretical calculations are consistent with a cytosolic concentration of GSSG in the low nanomolar range. The nanomolar concentration of GSSG is even considerably lower than the 4-μm concentration of GSSG determined in yeast cells expressing rxYFP (Østergaard et al., 2004). This measurement, however, was performed after permeabilization of the plasma membrane with digitonin. Although digitonin preferentially permeabilizes the cholesterol-rich plasma membrane, partial permeabilization of the ER membrane cannot be excluded. In the latter case GSSG would flow along its concentration gradient into the cytosol, and thus would necessarily cause an overestimation of the cytosolic GSSG concentration. Mixing of compartments with different degrees of oxidation may contribute to differences between roGFP and HPLC measurements. This discrepancy awaits further experimentation using compartment-targeted roGFPs and refined biochemical measurements of GSH/GSSG ratios.

Sensitivity of roGFP2 toward GSSG facilitates the monitoring of stress-induced redox changes

The high sensitivity of roGFP2 toward minute levels of GSSG mediated by GRX enables the use of roGFP2 for monitoring stress-induced redox changes. The generation of ROS during an oxidative burst is likely to feedback on the glutathione buffer, not least because H2O2 is at least in parts detoxified via the glutathione ascorbate cycle, which leads to intermittent oxidation of glutathione (Noctor and Foyer, 1998). Monitoring these changes with roGFP2 in vivo opens a window toward dynamic redox measurements in real time.

Redox-triggered switching of biologic processes mediated by GRX

The observation that a change in the glutathione redox potential by depletion of glutathione can lead to the formation of a disulfide bridge on the synthetic roGFP2, immediately suggests that similar changes or the formation of mixed disulfides of protein thiols with GSH might also occur on endogenous proteins. Redox-regulated target protein thiol groups would then effectively act as nanoswitches for biologic processes (Schafer and Buettner, 2001). In the absence of Cd2+ the cad2 mutant does not exhibit any phenotype (Cobbett et al., 1998). Nevertheless, the reduced GSH level is already sufficient to change the expression of a wide range of genes (Ball et al., 2004). It is now conceivable that this effect on gene expression might be mediated by GRXs, which directly reflects the redox potential of the cellular glutathione redox buffer (Figure 9). The essential role of GRX in switching biologic processes has been shown by the roxy1 mutant for flower development (Xing et al., 2005).

Figure 9.

 Glutaredoxins reversibly transfer electrons between the cellular glutathione redox buffer and target proteins. roGFP expressed in cells mimics native target proteins and thus functions as an artificial target.

The studies presented here for Arabidopsis establish the sensitivity of roGFP2 toward the cellular glutathione redox buffer, and indicate a very high degree of reduction of the glutathione buffer. Changes of the glutathione redox potential through oxidation or alteration of the pool size are rapidly transferred to roGFP2, and are thus converted into a reversible change of fluorescence. The option to target roGFPs to different subcellular compartments makes these fluorescent probes extremely useful for highly sensitive and non-invasive measurement of dynamic GSH-dependent redox signals.

Experimental procedures

Expression of roGFP in Arabidopsis and tobacco

The roGFP2 sequence (Hanson et al., 2004) was amplified by PCR with primers 5′-GGATCCCATGGTGAGCAAGGGCGAG-3′ and 5′-GTCGACTTACTTGTACAGCTCGTCC-3′ and cloned into the vector pCAPS (Roche, http://www.roche.com). After sequence confirmation the roGFP2 sequence was cut out with BamHI and SalI and cloned into the binary vector pBinAR (Höfgen and Willmitzer, 1990) behind a 35S promoter. pBinAR plasmids carrying 35S::roGFP2 were electroporated into Agrobacterium tumefaciens strain C58C1 (Deblaere et al., 1985). The resulting agrobacteria were used to transform Arabidopsis plants by the floral-dip method (Clough and Bent, 1998). Seeds were plated on agar plates containing 50 μg ml−1 kanamycin to select for transformants. The same agrobacteria were also used for transient expression of roGFP2 in the cytosol of tobacco cells. Well-watered tobacco plants were pressure infiltrated with an Agrobacterium suspension of OD600 0.5 and were imaged 2 days after transformation. For ER targeting roGFP2 was first PCR amplified to delete the start and stop codons and then cloned at NcoI and XbaI sites into pWEN81 between a chitinase targeting peptide and the HDEL retention signal. Subsequently, the entire cassette consisting of roGFP2 with N-terminal chitinase targeting signal and C-terminal HDEL-motif was subcloned into the binary vector pWEN22 behind a 35S promoter with XhoI and SacI. The resulting vector was electroporated into A. tumefaciens strain AGL-1 (Lazo et al., 1991). Leaf infiltration was performed as described previously (Sparkes et al., 2006).

Plant material and growth conditions

Arabidopsis (A. thaliana [L.] Heynh.) plants were grown on soil in controlled growth chambers with a diurnal cycle of 8-h light at 22°C and 16-h dark at 18°C. The light intensity was 120 μmol photons m−2 sec−1. For experiments on Arabidopsis roots, seeds from transgenic plants expressing roGFP2 were surface sterilized as described previously (Meyer and Fricker, 2000). Afterwards, sterilized seeds were washed twice with sterile water, dried on filter paper for 15 min, and then transferred to Arabidopsis growth medium solidified with 1% phytagel supplemented with 50 μg l−1 kanamycin for the selection of transformants where appropriate. Plates were kept at 4°C for 1 day before placing them in vertical orientation in a growth cabinet with 8-h light and 16-h dark at a reduced light intensity of 25–50 μmol photons m−2 sec−1. After germination all seedlings were additionally screened for good expression of roGFP2 by visual inspection of the plates under blue light on a stereomicroscope (Leica MZ FLIII; Leica, http://www.leica.com). Seedlings (5–7 days old) were used for experiments. Tobacco plants (Nicotiana tabacum L. cv. Samsun) were grown under long-day conditions in a greenhouse for 4 weeks. Plants were well watered before transfection to facilitate leaf infiltration with Agrobacterium suspensions.

Isolation of recombinant proteins

The roGFP2 sequence was cloned at NcoI and NotI sites into the expression vector pET-30a(+) (Novagen, http://www.emdbiosciences.com), which resulted in a C-terminal fusion with a His-tag. After transformation of Escherichia coli a pre-culture of 50-ml volume was grown at 37°C overnight. This pre-culture (4 ml) was then added to 200 ml Luria Bertani (LB) medium and grown to an OD600 of 0.6–0.8. Expression of roGFP2 was then induced by the addition of 1 mm isopropyl-β-d-thiogalactopyranoside (IPTG). To avoid misfolding, the temperature of the culture was reduced to 20°C. After 18-h expression bacteria were harvested and proteins were purified on a HiTrap chelating HP column (Amersham, http://www.amersham.com). Recombinant glutaredoxin (AtGRX C1; At5g63030; Rouhier et al., 2007), poplar TRX h3 and Arabidopsis NTR have been purified as described previously (Gelhaye et al., 2003; Jacquot et al., 1994).

In vitro characterization of roGFP2

Ratiometric time-course measurements with isolated roGFP protein were performed on a fluorescence plate reader (FLUOstar Optima; BMG, http://www.bmglabtech.com) with filter-based excitation at 390 and 480 nm and detection of emitted light at 510 ± 5 nm. Phosphate buffer (100 μl of 100 mm K2HPO4,/KH2PO4, 1 mm EDTA, pH 7.0) containing 1 μm roGFP2 and other proteins according to the information given in the text were pipetted into the wells of a 96-well plate with a clear bottom (Greiner Bio-One; http://www.gbo.com). Glutathione solutions (in 100 mm phosphate buffer, pH 7.0) were automatically injected using the built-in injectors. For maximum achievable reduction of the glutathione buffer 1 μg recombinant GR and 100 μm NADPH were added to each well. Stock solutions were always sufficiently highly concentrated to avoid inappropriate dilution of the roGFP2 solution during the injection. Similarly, H2O2 and DTT were automatically injected into the roGFP2 solution. Gain settings for detection of roGFP2 fluorescence after excitation with 390 and 480 nm were always adjusted before each experiment, and thus the absolute ratio values calculated might differ between the experiments. The relative ratio changes during a single experiment, however, were defined solely by the redox properties of roGFP2.

Wounding and detection of ROS

Small leaf pieces (3 × 3 mm) were cut out from mature leaves expressing roGFP2 in the cytosol and were immediately imaged for the redox state of roGFP. Duplicate samples were stained histochemically for H2O2 by using DAB as described (Thordal-Christensen et al., 1997). Leaf pieces were placed in a 1.68% (w/v), pH 3.8 DAB solution (AppliChem, http://www.applichem.de) for 12 h. After formation of brown precipitates in cells where H2O2 was present at high concentrations, leaf pieces were transferred to ethanol for removal of chlorophyll. Photographs were taken with a stereomicroscope (Leica MZ FLIII).

CLSM imaging and ratiometric analysis

Plants expressing roGFP2 were placed on a slide in a drop of 50 μm PI and imaged using a Zeiss confocal microscope LSM510META equipped with lasers for 405-, 488- and 543-nm excitation. Images were collected with a 10 ×  lens (Plan-Neofluar 0.3NA; Zeiss, http://www.zeiss.com) in multi-track mode with line switching and taking an average of four readings. Excitation of PI at 543 nm was performed in the same track as excitation of roGFP2 at 488 nm. In the second track roGFP2 was excited at 405 nm. For both excitation wavelengths roGFP2 fluorescence was collected with a bandpass filter of 505–530 nm. PI fluorescence was observed at 560–615 nm. Imaging of GSH in roots was performed according to the method described by Cairns et al. (2006) with excitation at 405 nm and collection of emitted fluorescence at 475–525 nm. Leaf images were collected as single optical sections. For root samples stacks of optical sections were collected and projected as maximum projections using the lsm software. Ratiometric analysis of fluorescence images was performed after background subtraction in imagej (v. 1.37v; http://rsb.info.nih.gov/ij/). In order to rule out the use of background pixels in the evaluation, the 405-nm fluorescence intensity image was corrected for its background. The corrected 405-nm image was divided by the 488-nm fluorescence intensity image to produce a ratio image on a pixel by pixel basis. The higher threshold of this ratio image was set to 3 to cancel out the rare cases in which a pixel with a significant intensity value in the 405-nm image had been divided by a background pixel in the 488-nm image. The edited image was the final 405/488-nm ratio image for which a histogram was calculated. A Gaussian fit was performed on this histogram to obtain the mean ratio value. To illustrate the evaluated 405/488-nm intensity ratio images, the grayscale was converted to color using the imagej look-up table ‘Fire’. The determined ratio values are very sensitive to changes of the hardware configuration. Thus, in the absence of an exact calibration, only ratios measured with identical settings can be compared in absolute terms.

Redox potential calculation

roGFP2 has an average consensus midpoint redox potential of E0′roGFP2 = −280 mV (Dooley et al., 2004). The actual redox potential depends on the degree of oxidation of the disulfide bridge following the Nernst equation. For calibration of roGFP2 against glutathione, a freshly made GSH solution was incubated with recombinant glutathione reductase and 100 μm NADPH. This maximally reduced GSH solution was then used for the determination of excitation ratios for roGFP2 in different GSH concentrations. All further calculations based on the Nernst equation were performed on the assumption of neutral pH conditions.

Reversible exchange of electrons between roGFP2 and glutathione leads to a redox equilibrium described by Eqn 1.

image(1)

In this equation R is the gas constant (8.315 J K−1 mol−1), T is the absolute temperature (298.15 K), z is the number of transferred electrons (2), and F is the Faraday constant (9.648 104 C mol−1). The standard redox potential of glutathione (E0′GSH) at pH 7 is −240 mV (Schafer and Buettner, 2001). The redox potentials of glutathione and roGFP2 are both dependent on the degrees of oxidation of the respective redox pair (O×DGSH and O×DroGFP2) according to Eqns 2 and 3.

image(2)
image(3)

Using these expressions for the degree of oxidation Eqn 1 can be transformed into Eqn 4. Provided the total quantity of GSH ([GSH]total) is known Eqn 4 contains only two unknown variables, O×DGSH and O×DroGFP2. Under equilibrium conditions these variables are interdependent and thus Eqn 4 can be transformed into Eqn 5 describing the dependency of O×DroGFP2 on O×DGSH.

image(4)
image(5)

At redox equilibrium between the glutathione solution and roGFP2 Eqn 5 then allowed us to calculate the degree of roGFP2 oxidation from the redox potential of the glutathione buffer with an assumed degree of oxidation. The degree of roGFP2 oxidation from fluorescence readings was calculated according to Eqn 6 with the fluorescence ratios R of 390/480 nm for in vitro experiments using a plate reader and 405/488 nm for in vivo experiments using a CLSM. In this equation Rred and Rox refer to the ratios of completely reduced and oxidized roGFP2, respectively. I480min and I480max refer to the fluorescence intensities measured with excitation at 480 nm for fully oxidized and fully reduced roGFP2. Because redox-dependent intensity changes at both wavelengths are linearly dependent on O×DroGFP2 the ratio R is not, unless the isosbestic point is used as a reference wavelength. In the latter case, the coefficient (I480min/I480max) would become 1, and thus Eqn 6 would simplify to the equation described by Hanson et al. (2004), with a linear relationship between O×DroGFP2 and the measured fluorescence ratio. In the case described here for two redox-dependent excitation wavelengths, the ratio values measured for different glutathione concentrations with a constant O×DGSH can be fitted nonlinearly according to Eqn 6 to extract O×DroGFP2.

image(6)

A plot of O×DroGFP2 calculated according to Eqn 6 against the redox potentials of the glutathione solutions resulted in a typical sigmoidal titration curve. Titration curves were fitted in sigmaplot 2001 (SPSS Inc., http://www.spss.com).

HPLC analysis of the cellular glutathione redox buffer

For determination of GSH and GSSG, 6-week-old plants were frozen in liquid N2, ground and extracted with 0.1 m PO4 buffer, pH 7.1 and 50% MeOH supplemented with 5 mm DTT for total glutathione or 5 mmN-ethylmaleimide for the determination of GSSG. Extracts were incubated at 60°C under vigorous shaking for 10 min. Fluorescent bimane-labeling and HPLC analysis was performed according to protocols described previously (Kranner et al., 2006; Wirtz et al., 2004).

Acknowledgements

We thank James Remington (University of Oregon) for providing roGFP constructs. We also thank Tobias Dick (DKFZ, Heidelberg) and Spencer Maughan (Biotechnology, University of Cambridge) for constructive comments and Markus Wirtz (University of Heidelberg) for his critical reading of the manuscript. This study was financially supported by grants from the University of Heidelberg and the Deutsche Forschungsgemeinschaft (DFG) to AJM (Grant ME1367/3-2).

Ancillary