Redox-sensitive GFP fusions for monitoring the catalytic mechanism and inactivation of peroxiredoxins in living cells.

Redox-sensitive green fluorescent protein 2 (roGFP2) is a valuable tool for redox measurements in living cells. Here, we demonstrate that roGFP2 can also be used to gain mechanistic insights into redox catalysis in vivo. In vitro enzyme properties such as the rate-limiting reduction of wild type and mutant forms of the model peroxiredoxin PfAOP are shown to correlate with the ratiometrically measured degree of oxidation of corresponding roGFP2 fusion proteins. Furthermore, stopped-flow kinetic measurements of the oxidative half-reaction of PfAOP support the interpretation that changes in the roGFP2 signal can be used to map hyperoxidation-based inactivation of the attached peroxidase. Potential future applications of our system include the improvement of redox sensors, the estimation of absolute intracellular peroxide concentrations and the in vivo assessment of protein structure-function relationships that cannot easily be addressed with recombinant enzymes, for example, the effect of post-translational protein modifications on enzyme catalysis.

Introduction

Genetically encoded fusion constructs between redox enzymes and redox-sensitive fluorescent proteins are commonly used to make non-invasive redox measurements in living cells [1], [2], [3], [4]. Fusion constructs between a peroxiredoxin (Prx), which serves as the peroxide sensor moiety, and redox-sensitive green fluorescent protein 2 (roGFP2), which serves as the reporter moiety, have recently been developed and allow real-time monitoring of intracellular hydroperoxide concentrations [4], [5], [6]. Here, we asked whether roGFP2 fusion constructs and hydroperoxide challenges can be used to deduce peroxidase properties and mechanisms in vivo. In other words, do classic enzyme kinetic parameters of peroxidases affect the roGFP2 readout in a predictable fashion? In particular, we were interested in understanding how roGFP2 readouts are affected by the peroxidase kcatapp values and catalytic efficiencies (kcatapp/Kmapp values reflecting second order rate constants) as well as inactivation kinetics due to hyperoxidation, i.e. the sulfinic and sulfonic acid formation of the active site cysteine residue.

To experimentally address these questions, we required a kinetically well-characterized peroxidase isoform and thus chose to use the Prx5-type model enzyme PfAOP from the malaria parasite Plasmodium falciparum. PfAOP localizes to the cytosol and plastid of the parasite, is dispensable for asexual blood stage development and accepts a variety of hydroperoxide substrates and electron donors in vitro [7], [8], [9], [10]. Steady-state kinetics in combination with site-directed mutagenesis, X-ray structures and gel filtration analyses have previously revealed that PfAOP requires only one cysteine residue for catalysis and predominantly forms stable homodimers [10], [11]. This is in contrast to GPx3 and typical 2-Cys Prx [12], [13], which have been used as ratiometric roGFP2-coupled peroxide sensors [3], [5], [6]. The latter peroxidases are either incompletely characterized regarding their kinetics or have more complicated reaction mechanisms due to the number of relevant cysteine residues and/or variable quaternary structures.

We have previously characterized the kinetic parameters of gain- and loss-of-function mutants of PfAOP and showed that the enzyme is rapidly inactivated by H2O2 but not by tert-butyl hydroperoxide (tBOOH) [10], [11]. Mutation of residue Leu109, which is situated at the bottom of the active site between the catalytic (peroxidatic) cysteine residue Cys117 and the buried non-catalytic cysteine residue Cys143, affects the catalytic as well as the inactivation properties of PfAOP [11]. For example, compared to recombinant wild type PfAOP, the gain-of-function mutant PfAOPL109M was shown to be less susceptible to H2O2-dependent inactivation and to have a 3-fold higher kcatapp value for tBOOH, a 2.3-fold higher kcatapp value for glutaredoxin (Grx) as an electron donor, and a 1.2- to 1.4-fold higher kcatapp value for reduced glutathione (GSH) as an electron donor [11]. Furthermore, the kcatapp/Kmapp values of PfAOPL109M for Grx and GSH increased 2.2- and 12-fold, respectively, whereas the kcatapp/Kmapp value for tBOOH was similar to the wild type enzyme. In contrast, the loss-of-function mutant PfAOPL109A was shown to be more susceptible to H2O2-dependent inactivation, to have 4-fold lower kcatapp values for tBOOH and GSH, a 7-fold lower kcatapp/Kmapp value for tBOOH and a 3-fold lower kcatapp/Kmapp value for GSH. The activating effects for PfAOPL109M depended on the presence of residue Cys143, although its exact role could not be resolved by steady-state kinetics or gel mobility shift assays. Based on our kinetic and structural data, we proposed a model according to which Leu109 and Cys143 together affect the equilibrium between the fully folded and locally unfolded conformation of PfAOP. Mutation of Leu109 to methionine was suggested to stimulate local unfolding of the active site helix α2, thereby preventing hyperoxidation and promoting the probably rate-limiting GSH-dependent reduction of the Cys117 sulfenic acid [11]. Here we used our gain- and loss-of-function mutants to compare the oxidative half-reaction using stopped-flow kinetics and to test the suitability of roGFP2 as a mechanistic reporter for redox catalysis inside living cells.

Materials and methods

Materials

H2O2, tBOOH, peroxynitrite, cumene hydroperoxide and 12(S)-hydroperoxy-5Z,8Z,10E,14Z eicosatetraenoic acid (12(S)HpETE) were purchased from Mallinckrodt Chemicals or Sigma. The concentration of H2O2 stock solutions was determined spectrophotometrically at 240 nm (ε240 nm = 43.6 M−1 cm−1). The peroxynitrite concentration was determined at alkaline pH at 302 nm (ε302 nm = 1.67 mM−1 cm−1) [14]. Concentrations of cumene hydroperoxide and 12(S)HpETE were calculated considering the manufactures specifications. Diamide, 1,4-dithiothreitol (DTT), diethylenetriaminepentaacetic acid (DTPA) and horseradish peroxidase were purchased from Sigma. Nickel-nitrilotriacetic acid agarose (Ni-NTA) was from Qiagen. HiTrap desalting columns were from Amersham Bioscience. All of the amino acids, glucose and yeast nitrogen base required for Hartwell's Complete (HC) yeast growth medium were purchased from Sigma. Flat-bottom 96 well microplates (product #353219) were from BD Biosciences.

Cloning of yeast expression vectors

The gene sequence for N-terminally truncated PfAOPΔ59 without its apicoplast-targeting sequence [7], [10] (herein after referred to as PfAOP) was optimized for expression in Saccharomyces cerevisiae, synthesized and cloned into a pUC57 vector (Genscript, Piscataway, USA). PfAOP was subcloned into roGFP2-GRX1/p416TEF [15] using EcoRI and HindIII restriction sites, thereby replacing GRX1 to generate wild type roGFP2-PFAOP/p416TEF. A standard site-directed mutagenesis protocol was employed to generate roGFP2-PFAOP/p416TEF, roGFP2-PFAOP/p416TEF and roGFP2-PFAOP/p416TEF. All mutations were confirmed by Sanger sequencing. The codon-optimized gene sequence encoding roGFP2-PfAOP has been deposited in GenBank (http://www.ncbi.nlm.nih.gov/genbank/) under accession number MF140392 and is listed in the Supplementary material.

Transformation, expression and fluorescence measurements in S. cerevisiae

Plasmids p416TEF, roGFP2/p416TEF and wild type and mutant forms of roGFP2-PFAOP/p416TEF were transformed into yeast strain BY4742 and roGFP2 measurements were conducted as described previously [3], [4]. Briefly, liquid cultures were grown to late exponential phase (OD600 = 3–4) in HC medium lacking uracil, in order to select for those cells retaining the p416TEF vectors. Cells were harvested by centrifugation for 3 min at 800×g and resuspended in buffer containing 100 mM NaCl, 100 mM sorbitol, 100 mM Tris-HCl, pH 7.4 to a final concentration of 7.5 OD600 units/mL. Aliquots of 200 µL of the cell suspension were transferred into the appropriate number of wells of a flat-bottom 96 well microplate. Two additional wells were utilized for controls and were supplemented with either diamide to a final concentration of 20 mM (fully oxidized control) or DTT to a final concentration of 100 mM (fully reduced control). The control wells are required for the determination of the degree of sensor oxidation (OxD). Cells in the experimental wells were treated with twelve increasing concentrations of H2O2 or tBOOH (10–500 µM) and responses were followed for up to 100 min at 30 °C using a CLARIOstar fluorescence plate reader (BMG Labtech). For each peroxide concentration the OxD was plotted against time. The area under the curve (AUC (OxD×min)) was subsequently calculated in Excel and plotted against the according peroxide concentration in SigmaPlot 13. All data were averaged from triplicate (tBOOH treatment) or quadruplicate (H2O2 treatment) measurements from independent yeast cultures. Statistical analyses were carried out in SigmaPlot 13 using the One-way ANOVA method.

Expression and purification of recombinant wild type and mutant PfAOP in Escherichia coli

Recombinant proteins were expressed and purified by affinity-chromatography as described previously [10], [11], [16]. Briefly, E. coli XL1-Blue cells were transformed with plasmid PFAOP/pQE30, PFAOP/pQE30, PFAOP/pQE30 or PFAOP/pQE30. Expression was induced with 0.5 mM isopropyl-β-D-1-thiogalactopyranoside for 4 h at 37 °C. Liquid cultures were harvested by centrifugation for 15 min at 4000×g and 4 °C. The bacteria were resuspended in buffer containing 20 mM imidazole, 300 mM NaCl, 50 mM NaxHyPO4, pH 8.0, incubated with lysozyme and disrupted by sonication. Proteins were affinity-purified on Ni-NTA agarose columns and eluted in buffer containing 200 mM imidazol, 300 mM NaCl, 50 mM NaxHyPO4, pH 8.0. Subsequently, samples were treated with 5 mM DTT for 30 min at 4 °C to fully reduce the protein. Remaining imidazole and DTT were removed using HiTrap desalting columns that were equilibrated with buffer containing 100 mM NaxHyPO4, 0.1 mM DTPA, pH 7.4. Protein elution was monitored at 280 nm using an Äkta FPLC system. The protein concentration was determined spectrophotometrically using the molar extinction coefficient ε280 nm = 21.43 mM−1 cm−1 as calculated for the primary sequence of the protein using the ProtParam ExPASy tool (http://web.expasy.org/protparam/). The thiol content of the proteins was analyzed with 5,5′-dithiobis-(2-nitrobenzoic acid) [17] revealing that 97% of the protein thiols were in a reduced state (data not shown).

Stopped-flow peroxidase measurements of recombinant wild type and mutant PfAOP

The oxidative half-reaction of 1 µM wild type PfAOP, PfAOPC117S, PfAOPC143S, and PfAOPL109M with different hydroperoxides and peroxynitrite was analyzed using a SX-20 stopped-flow spectrofluorometer (Applied Photophysics). All activities were determined at 25 °C in 100 mM NaxHyPO4 buffer containing 0.1 mM DTPA, pH 7.4. PfAOPC117S showed no change in fluorescence and served as a negative control. Two alternative methods were employed:

Direct assay: Changes of the intrinsic tryptophan fluorescence (λexc = 295 nm, total emission) of reduced wild type and mutant PfAOP were followed after mixing of the enzymes with 1–100 µM H2O2, peroxynitrite, cumene hydroperoxide or 12(S)HpETE. Experimental curves showed two main phases of fluorescence change, which were fitted to exponential curves using the Applied Photophysics Pro-data SX software. The first phase, which was recorded from 2 ms (mixing time of the apparatus) to 20–100 ms depending on the peroxide concentration, showed a decrease in fluorescence. The second phase was followed from 0.1 to 10 s and showed a slower increase in fluorescence. After 10 s there was a further small increase in fluorescence, particularly at higher peroxide concentrations, which was disregarded for the fitting. Rate constants kobs for the first and the second phase of fluorescence change were obtained from the exponential curves by analyzing the average of 5–6 runs. Rate constants k1-3 were calculated from linear fits by plotting the obtained k values against the peroxide concentrations [18], [19], [20]. Constants k1 and k2 were obtained from the slope of the first and the second plot, respectively. Rate constant k3 was obtained from the y-axis intercept of the second plot. Alternative global fitting approaches did not yield better results and complicated the data interpretation because of even more unknown variables and unexplained rate constants.

Competition assay: The rate constant for the reaction between the enzymes and H2O2 or peroxynitrite (k1*) was determined in a competition assay using horseradish peroxidase (HRP) as an alternative peroxide target as described previously [21], [22]. Briefly, the oxidation of 2 µM HRP by 1 µM peroxide to 'compound I′ was followed at 398 nm (ε398 nm = 42 mM−1 cm−1) [23] in the absence or presence of increasing PfAOP concentrations. The determined rate constant for the reaction of HRP with H2O2 and peroxynitrite was 1.4 × 107 M−1 s−1 and 3.5 × 106 M−1 s−1, respectively, in agreement with the previously reported values [22], [24]. The rate constants for the reactions of wild type and mutant PfAOP with the peroxide were calculated from the amount of compound I that was formed at different PfAOP concentrations.

Results

The gain-of-function mutant PfAOPL109M serves as a robust hydroperoxide sensor in yeast

Baker's yeast Saccharomyces cerevisiae is a well-established model organism for studying roGFP2-dependent redox sensing [3], [6]. Therefore, we chose to use it for the exemplary intracellular assessment of the catalytic mechanism and inactivation properties of PfAOP. First, we cloned codon-optimized fusion constructs encoding wild type roGFP2-PfAOP (Genbank accession number MF140392 and Supplementary material) as well as different variants containing PfAOP with Leu109 or Cys143 mutations into the vector p416TEF.

The p416TEF plasmids were transformed into wild type BY4742 yeast cells for cytosolic expression of the constructs. The yeast cells were subsequently treated with 10–500 µM H2O2 and the ratiometric degree of roGFP2 oxidation (OxD) was monitored over time. Striking differences were observed between the wild type roGFP2-PfAOP fusion construct (wt) and the L109M mutant (Fig. 1a,b). As a first readout parameter we assessed the maximum change in roGFP2 oxidation (ΔOxDmax). ΔOxDmax for L109M was higher than for the wild type fusion construct at most H2O2 concentrations. At H2O2 concentrations above 200 µM, the response of the wt construct decreased sharply, whilst, in contrast, the response of L109M was more robust. Comparable OxD responses were obtained following tBOOH treatment, although no decrease of ΔOxDmax was observed at higher tBOOH concentrations (Fig. 1c,d). In summary, PfAOPL109M is not only more active and robust than wild type PfAOP in vitro but its catalytic and inactivation properties with H2O2 and tBOOH are also reflected by the different roGFP2 responses of the L109M and wt fusion constructs in yeast cells.

Fig. 1

Dose-response curves for yeast cells with genetically encoded roGFP2-PfAOP fusion constructs after bolus treatments with hydroperoxides at 30 °C. (A) Time-course measurements of the ratiometric degree of oxidation (OxD) for the wild type roGFP2-PfAOP fusion construct (wt) at different initial H2O2 concentrations. (B) Time-course measurements of the OxD for the roGFP2-PfAOPL109M fusion construct (L109M) at different initial H2O2 concentrations. (C) and (D) Time-course measurements of the OxD for wt and L109M at different initial tBOOH concentrations. Data were averaged from four (H2O2) or three (tBOOH) independent biological replicates.

Intracellular comparison of PfAOP mutants and quantification of roGFP2 readouts

In order to enable a more statistically rigorous comparison between different constructs, we integrated the area below the OxD curve, henceforth referred to as Area Under Curve (AUC), including the mutants L109A and C143S (Fig. 2a). In contrast to L109M, the L109A construct exhibited a decreased AUC, relative to the wt construct, for all H2O2 concentrations tested. Replacement of the non-catalytic second cysteine residue of PfAOP in C143S also slightly decreased the AUC compared to the wt construct at some H2O2 concentrations. Most of the differences among the wild type and mutant PfAOP fusion constructs were statistically significant with p < 0.01 (Fig. 2b). Similar differences between the AUCs of wt, L109M, L109A and C143S were also observed for roGFP2 responses to challenges with tBOOH (Fig. 2c,d).

Fig. 2

Integrated dose-response curves for yeast cells with genetically encoded roGFP2-PfAOP fusion constructs after bolus treatments with hydroperoxides. Wild type roGFP2-PfAOP (wt) and roGFP2 alone (roGFP) served as positive and negative control, respectively, and confirmed that the OxD was PfAOP-dependent. Constructs roGFP2-PfAOPL109A (L109A), roGFP2-PfAOPL109M (L109M) and roGFP2-PfAOPC143S (C143S) carry previously characterized single point mutations of PfAOP [11]. (A) The area under the OxD curves (AUC) from Fig. 1 was determined between 0–60 min and plotted against the initial H2O2 concentration. All data were averaged from quadruplicate independent biological replicates. (B) Statistical analysis of the data from panel a. P-values were calculated using the One-way ANOVA method in SigmaPlot 13. (C) AUC from Fig. 1 plotted against the initial tBOOH concentration. All data were averaged from triplicate independent biological replicates. (D) Statistical analysis of the data from panel c. *p < 0.05; **p < 0.01; ***p < 0.001.

As a third readout parameter we analyzed the H2O2 concentration-dependence of the AUC. The H2O2 concentration at which the highest AUC was observed (cH2O2max) was noticeably lower for the L109A construct compared to the wt and L109M constructs (Fig. 2a,b). Furthermore, a decrease in AUC was observed for the L109A and wt constructs at H2O2 concentrations above 75 µM and 250 µM, respectively. At 400 and 500 µM H2O2 the AUC of these constructs was not significantly different to the roGFP2 negative control. In contrast, the AUC of the L109M construct indicated robust roGFP2 responses even at the highest H2O2 concentrations tested. Following treatment with tBOOH, the AUC was observed to either show an increase at all tested concentrations (L109M) or to reach a plateau only at much higher tBOOH concentrations relative to H2O2 (wt, L109A and C143S) (Fig. 2c,d). Please note that the AUC of the roGFP2 negative control increased with increasing tBOOH concentrations. Thus, the observed AUC plateau for wt, L109A and C143S at 300–500 µM tBOOH rather indicates a decrease of the peroxidase-dependent roGFP2 response of these constructs. In summary, enzymatic properties of gain- and loss-of-function mutants of PfAOP as determined in vitro correlate with changes in ΔOxDmax, the AUC and the cH2O2max of roGFP2 fusion constructs in yeast cells. Differences between mutants can be quantified and statistically analyzed using AUC values and allow us to monitor mechanistic aspects of peroxiredoxin catalysis in real-time in living cells.

In vitro oxidation and hyperoxidation kinetics of PfAOP

As outlined above, the role of residues Leu109 and Cys143 for H2O2-dependent PfAOP catalysis and hyperoxidation-dependent enzyme inactivation could not be resolved by steady-state kinetics or gel mobility shift assays [11]. In order to address a potential relevance of the oxidative half-reaction and/or the hyperoxidation of PfAOP for intracellular roGFP2 readouts, we performed stopped-flow kinetic experiments with recombinant PfAOPwt, PfAOPC143S and PfAOPL109M. The reaction between reduced PfAOP and H2O2 was directly monitored at variable peroxide concentrations by following changes in the intrinsic tryptophan fluorescence of the enzyme (Fig. 3). Two major phases were detected, one rapid decrease followed by a slower increase in fluorescence. Three rate constants were assigned to the reaction kinetics (Fig. S1), one [H2O2]-dependent rate constant for the first phase (k1) and one [H2O2]-dependent as well as one [H2O2]-independent rate constant for the second phase (k2 and k3, respectively). PfAOPwt and PfAOPC143S had similar kinetics with a second-order rate constant k1 of ~3.5 × 107 M−1 s−1 (Fig. 3a,b and Table 1). A spectrophotometric competition assay with horseradish peroxidase revealed a rate constant k1* of 2.1 × 107 M−1 s−1 and confirmed that k1 reflects the reaction between PfAOP and the peroxide (Fig. S2 and Table 1). The peroxidase competition assay also showed that the rate constant k1 for the sulfenic acid formation of PfAOPL109M is identical to PfAOPwt and PfAOPC143S. However, the rate constant k2 of PfAOPL109M at around 1.4 × 103 M−1 s−1 was more than one order of magnitude smaller than for PfAOPwt and PfAOPC143S. The smaller k2 value probably indicates a slower rate of hyperoxidation owing to reaction with a second molecule of H2O2 (see also Section 4). The relevance of rate constant k3 is unclear. Maybe it reflects a slow conformational change. Please note that we observed only a small change in fluorescence for PfAOPL109M compared to PfAOPwt (Fig. 3a,c), which might be explained if only a fraction of PfAOPL109M adopted the fully folded conformation (see also Discussion). Although this indicates a lower availability of the enzyme for the reaction with H2O2, this has no effect on the second order rate constants k1. Control experiments with PfAOPC117S revealed that all rate constants and phases depended on the presence of the peroxidatic cysteine of the enzyme (Fig. 3d).

Fig. 3

Stopped-flow kinetics of the oxidative half-reaction of reduced recombinant PfAOP after mixing with HO2 at pH 7.4 and 25 °C. Representative traces are shown for 1 µM enzyme and 1 µM substrate. (A) Kinetics for wild type PfAOP. (B) Kinetics for PfAOPC143S lacking the non-catalytic second cysteine residue. (C) Kinetics for the active site mutant PfAOPL109M. (D) PfAOPC117S without the peroxidatic cysteine residue served as a negative control.

Table 1

Rate constants for the oxidative half-reaction of 1 µM reduced recombinant wild type and mutated forms of PfAOP as determined by stopped-flow kinetic measurements with variable H2O2 concentrations at pH 7.4 and 25 °C.

	k1*a (M−1 s−1)	k1 (M−1 s−1)	k2 (M−1 s−1)	k3 (s−1)
PfAOPwt	(2.1 ± 0.8) × 107	(3.2 ± 0.5) × 107	(3.6 ± 0.6) × 104	0.28 ± 0.02
PfAOPC143S	(2.2 ± 0.6) × 107	(3.7 ± 1.5) × 107	(4.3 ± 0.0) × 104	0.16 ± 0.01
PfAOPL109M	(2.0 ± 0.8) × 107	n.d.	(1.4 ± 0.1) × 103	0.04 ± 0.00

n.d.: not determined.

Rate constant k1* was determined in a peroxidase competition assay.

Next, we tested the peroxide-dependence of the oxidative half-reaction using H2O2, peroxynitrite, cumene hydroperoxide, and the fatty acid hydroperoxide 12(S)HpETE (Table 2). Rate constant k1 of PfAOPwt was similar for all peroxide substrates tested except for aromatic cumene hydroperoxide, which was converted three- to sevenfold slower. Rate constants k2 and k3 did not appear to depend on the type of peroxide. In summary, stopped-flow kinetic measurements of the oxidative half-reaction showed that the reactivities of reduced PfAOPwt, PfAOPC143S and PfAOPL109M with H2O2 are very similar. PfAOPL109M differs from the other enzymes with regard to rate constant k2, which might reflect the hyperoxidation kinetics. Furthermore, reactivities of PfAOPwt with H2O2, peroxynitrite, and 12(S)HpETE are very similar, suggesting that there is no real enzyme-substrate complex and that the reaction proceeds as soon as the substrate enters the active site in a productive orientation.

Table 2

Rate constants for the oxidative half-reaction of 1 µM reduced recombinant wild type PfAOP as determined by stopped-flow kinetic measurements with variable peroxide concentrations at pH 7.4 and 25 °C.

	k1*a (M−1s−1)	k1 (M−1s−1)	k2 (M−1s−1)	k3 (s−1)
Peroxynitrite	(2.6 ± 1.9) × 107	(1.5 ± 0.5) × 107	n.d.	n.d.
Hydrogen peroxide	(2.1 ± 0.8) × 107	(3.2 ± 0.5) × 107	(3.6 ± 0.6) × 104	0.28 ± 0.02
Cumene hydroperoxide	n.d.	(4.8 ± 0.5) × 106	(4.5 ± 2.7) × 104	0.27 ± 0.17
12(S)HpETE	n.d.	(1.9 ± 0.6) × 107	(5.6 ± 1.2) × 104	0.45 ± 0.05

n.d.: not determined.

Rate constant k1* was determined in a peroxidase competition assay.

Discussion

How can we integrate and interpret the roGFP2 readouts and stopped-flow kinetic data with previous kinetic data in a comprehensive model? The ∆OxDmax and AUC values probably represent a metabolic flux comprising (i) the peroxide-dependent oxidation of the sensor moiety (yielding a Prx sulfenic acid species), (ii) the two-step reduction of the sensor by the reporter moiety (yielding a roGFP2 disulfide), and (iii) the reduction of the reporter moiety by GSH, Grx or another cytosolic thiol component (Fig. 4). Furthermore, there are side reactions that prevent the oxidation of the reporter moiety, including (iv) the potential reduction of the sulfenic acid by other thiols than the reporter moiety, and (v) the hyperoxidation of the sensor moiety. The stopped-flow kinetic data indicate that rate constant k1 reflects the sulfenic acid formation and that this constant is similar for PfAOPwt, PfAOPL109M, and PfAOPC143S. Furthermore, rate constant k1 is very similar for different peroxides (except for aromatic cumene hydroperoxide). Thus, step (i) did probably not account for the observed differences for wild type and mutant PfAOP fusion constructs after treatment with H2O2 and tBOOH. Similar k1 values of the oxidative half-reaction might in fact explain the similar sensitivity of the roGFP2 constructs in terms of the lowest detectable peroxide concentration. Step (iii) probably also did not account for the observed differences between the roGFP2 readouts, because the roGFP2 reporter moiety and the genetic background were identical for all analyzed constructs. Step (iv) includes several unknown variables, but the fact that the roGFP2-PfAOP fusion constructs were quite sensitive and responded well to peroxide challenges suggests that this bypass reaction was presumably not a major factor for the roGFP2 readouts.

Fig. 4

Model for the intracellular roGFP2-dependent assessment of PfAOP catalysis. The metabolic flux monitored by roGFP2 comprises (i) the peroxide-dependent oxidation of the PfAOP sensor moiety, (ii) the two-step reduction of the sensor moiety yielding the oxidized roGFP2 reporter moiety, (iii) the reduction of the reporter moiety, (iv) a potential roGFP2-independent bypass reaction between oxidized PfAOP and alternative reducing agents, and (v) the inactivation of the sensor moiety because of hyperoxidation.

The crucial remaining parameters in Fig. 4 are step (ii) and/or step (v). A rate-limiting sensor reduction by roGFP2, which is promoted in L109M and hampered in L109A, is in excellent agreement with our previous in vitro studies. In particular, the AUCs of L109M > wt > L109A correlate well with the corresponding kcatapp values of the rate-limiting reductive half-reaction with GSH in vitro [11]. Although step (ii) reflects the electron transfer between reduced roGFP2 and the sulfenic acid of PfAOP instead of the transfer between GSH and PfAOP, both processes should depend in a similar way on the Leu109-dependent local unfolding of the active site helix α2 (step iia). A correlation between the enzyme activity and the AUC is also supported for recombinant PfAOPC143S, which was shown to have a 25% lower activity in vitro [10], in accordance with the apparently slightly decreased roGFP2 response of C143S in yeast. Step (v), the hyperoxidation of the peroxidatic cysteine, inactivates the sensor and prevents the oxidation of the roGFP2 reporter moiety. Thus, Prx inactivation by hyperoxidation is probably the most crucial factor affecting cH2O2max and the decrease in AUC at higher H2O2 concentrations. Recombinant wild type PfAOP was shown to be rapidly inactivated by H2O2, whereas inactivation by tBOOH required much higher peroxide concentrations in vitro [11]. These findings correlate with the observation that the highest AUC of the wt construct was reached at much higher tBOOH concentrations compared to H2O2. Furthermore, recombinant PfAOPL109A and PfAOPL109M were more and less susceptible to H2O2-dependent inactivation, respectively [11], which correlates with the robustness of the roGFP2 responses of the L109A and L109M constructs. The more than ten times smaller rate constant k2 of recombinant PfAOPL109M determined by stopped-flow measurements also supports a slower enzyme inactivation in accordance with the increased robustness of the L109M construct in yeast. Again, the data is consistent with a catalytic model that suggests a slower hyperoxidation due to the shifted equilibrium towards the locally unfolded conformation for PfAOPL109M [11]. Overall, we observed a strong correlation between the sensitivity of the different roGFP2-PfAOP constructs to hyperoxidation with H2O2 in contrast to tBOOH and the known inactivation properties of recombinant PfAOP mutants in vitro. Importantly, the roGFP2 fusion constructs allow us to analyze the intracellular H2O2 responses of wild type and mutant PfAOP in vivo, whereas in vitro NADPH-coupled steady-state kinetic measurements require H2O2 concentrations ≥ 5 µM, which result in rapid enzyme inactivation and hamper in-depth analysis [11]. In summary, ratiometric roGFP2 measurements of Prx-fusion constructs in combination with site-directed mutagenesis are a powerful tool to gather mechanistic insights on Prx catalysis and inactivation in vivo.

What are the implications of our study? (i) The identification of correlations between roGFP2 readouts and kinetic parameters as well as inactivation properties of Prx might be useful for the optimization of redox sensors, e.g., by searching the literature for peroxidases that have highly efficient reductive half-reactions and that are not prone to inactivation (e.g., selected bacterial Prx-isoforms) [25]. (ii) Peroxidase fusion constructs with roGFP2 could be also useful for addressing inactivation properties that cannot be analyzed in steady-state kinetic assays, as shown for the reaction between H2O2 and wild type or mutant PfAOP [11]. Interesting candidates might be, for example, the Prx6-type 1-Cys peroxiredoxins [26], [27], which are almost inactive in standard assays (maybe owing to rapid inactivation) and remain to be analyzed in much more detail. Potential limitations of the method include genetic manipulations to remove much more active competing peroxidases in order to increase the OxD and AUC responses. (iii) Recombinant proteins often lack post-translational modifications (e.g., when purified from E. coli) or are obtained in rather low yields (e.g., when purified from insect cells). The analysis of roGFP2-coupled wild type and mutant enzymes might allow the in vivo evaluation of the relevance of post-translational modifications for Prx catalysis [28]. For example, serine/threonine to alanine/valine or glutamate mutants could be compared in order to directly monitor the role of phosphorylation sites, e.g., at the plasma membrane or the centrosome [29], [30], [31], [32], [33]. Potential limitations of the method comprise genetic manipulations to prevent the formation of heterooligomers with endogenous Prx copies as well as altered quaternary structures because of the bulky roGFP2-tag. (iv) Absolute values for the cytosolic H2O2 concentration in yeast and the diffusion-dependent ratio between extra- and intracellular H2O2 are, to the best of our knowledge, unknown. The latter ratio is of particular importance to judge the physiological relevance of bolus treatments with H2O2. Estimated steady-state and physiological peak concentrations for H2O2 in a variety of organisms are ~1 nM and 0.5–0.7 μM, respectively [34]. Based on the observation that the activity of recombinant PfAOPL109M peaked in steady-state kinetic measurements at 10 µM [11], we speculate that the intracellular H2O2 concentration in our experiments might have been roughly 20-fold lower than the extracellular H2O2 concentration because of the ∆OxDmax concentration for H2O2 in Fig. 1b at 200 µM. Although the yeast plasma membrane permeability for H2O2 was shown to be variable [35], [36], a 20-fold difference in H2O2 concentration across the plasma membrane is in good agreement with reports on 7- to 10-fold decreased intracellular H2O2 concentrations upon bolus treatments of E. coli or mammalian cells [37], [38], [39]. Hence, assuming a linear correlation between the extra- and intracellular H2O2 concentration, intracellular H2O2 concentrations in our experiments might have ranged from 0.5 to 25 µM. Thus, using the inactivation properties of PfAOPL109M for an estimation of the intracellular H2O2 concentration, we speculate that common bolus treatments in yeast research might result in intracellular H2O2 concentrations that significantly exceed a physiological threshold concentration around 0.7 µM [34].

Conclusions

Fusion constructs between roGFP2 and Prx have been commonly used to monitor intracellular hydroperoxide concentrations, but, so far, it was unknown how and to which extent kinetic parameters or inactivation properties of the Prx sensor are correlated with the measured roGFP2 readout. Here we showed for fusion constructs between roGFP2 and gain- and loss-of-function mutants of the model peroxiredoxin PfAOP that in vitro kcatapp values and inactivation properties of Prx correlate with the roGFP2 readout inside living cells. The findings of our proof-of-principle study open the door to a wide-range of potential future applications, including (i) the optimization of redox sensors, (ii) the noninvasive analysis of peroxidases, in particular, of enzymes that are labile or prone to inactivation in vitro, (iii) the evaluation of the relevance of post-translational protein modifications, and (iv) the estimation of absolute intracellular hydroperoxide concentrations.