Estimation of kinetic parameters related to biochemical interactions between hydrogen peroxide and signal transduction proteins

Introduction

Being higher reductions states of molecular dioxygen, reactive oxygen species are present in all aerobic organisms. Initially, these species were seen as harmful species that caused or participated in the etiology of many diseases through oxidative damage, but more recently physiological roles mediated by the modulation of the redox state of biomolecules were attributed to reactive oxygen species (Sies, 2014). Today redox biology is an established field. As Berzelius put it the venom is in the dose, and reactive oxygen species have different roles depending on their concentration. This work is centered on hydrogen peroxide (H₂O₂), a reactive oxygen species that has the properties of a second messenger (Forman et al., 2010) and participates in many pathways, including insulin (Mahadev et al., 2001; Haque et al., 2011), mitogenic (Irani et al., 1997), inflammatory, and apoptotic signaling (Oakley et al., 2009; Tschopp and Schroder, 2010). Having a relative low chemical reactivity, H₂O₂ reacts mainly with metal centers and with thiol compounds, such as cysteine residues in proteins (Marinho et al., 2014). Examples of H₂O₂ targets are PerR, a metal-dependent transcription factor that is inhibited by H₂O₂ in a Fenton-like reaction, and protein tyrosine phosphatases (PTPs), which are inhibited upon oxidation of cysteine residues in their active center (Tanner et al., 2011; Marinho et al., 2014; Sies, 2014). The list of proteins containing cysteine residues that were observed to be oxidized by H₂O₂ is vast, near 200 (Le Moan et al., 2006; Martínez-Acedo et al., 2012), and continues to increase as investigators find new targets for H₂O₂. In contrast, kinetic parameters concerning oxidation by H₂O₂ have been measured only for a few of these proteins (Ferrer-Sueta et al., 2011; Tanner et al., 2011), leading several researchers to point out the lack of proper quantitative data as a barrier to the development of the field (Brigelius-Flohé and Flohé, 2011; Buettner et al., 2013). Importantly, the triggering of biphasic responses by H₂O₂ in a narrow concentration range has important biological implications. For example, in H4IIEC hepatocytes H₂O₂ can either enhance or impair insulin signaling depending on its concentration (Iwakami et al., 2011). This dual role was attributed to the different sensitivity of PTP1B inhibition and JNK activation, two kinases that stimulate and inhibit insulin signaling, respectively. Thus, while H₂O₂ is an essential component of the insulin signaling pathway, it may also mediate the etiology of insulin resistance (Fisher-Wellman and Neufer, 2012). Although the underlying data is known for some time, such picture only emerged recently, probably because in absence of a quantitative framework, these biphasic responses were often interpreted as contradictory findings that were dependent of the biological model used or simply reflected non-reproducible experimental results. To study such complex responses it is advantageous to apply a quantitative and integrative approach typical of systems biology (Buettner et al., 2013), where the reactivity of targets toward H₂O₂ is determined to undercover which pathways operate in vivo under different conditions. In this work, we address how kinetic parameters can be determined from typical experiments performed in redox signaling.

Based on a simple reaction scheme representing H₂O₂ signaling, we started by deducing kinetic equations that are tailored to estimate kinetic parameters from experimental data. Next, to test the validity of the deduced equations, virtual experiments carried out under different conditions of H₂O₂ exposure were simulated with a kinetic model, and the results were fitted to the equations deduced. The agreement between the kinetic parameters obtained in these fittings and those used to obtain the virtual time courses was used as a criterion to decide on the accuracy of the kinetic equations deduced. Finally, to evaluate the applicability of kinetic equations to real data, experimental results described in the literature focusing on PTP-dependent signaling were fitted to the equations deduced. Two PTPs, PTP1B, and SHP-2, for which kinetic rate constants are known, were chosen as test cases. Our study demonstrates that insightful kinetic parameters related to biochemical interactions between H₂O₂ and signal transduction proteins can be estimated from typical H₂O₂-signaling experiments by applying the equations deduced here.

Theory and Methods

Master Equation

A minimal mathematical analytical model was set up to describe a signaling event triggered by H₂O₂, according to the following two reactions:

In the first reaction, the reduced form of a sensor protein target (Target_rd) is oxidized by H₂O₂, modifying its activity, which results in the modulation of a signaling pathway. In the second reaction, the oxidized target (Target_ox) is switched-off by being regenerated back to the reduced form. A specific example of these two reactions is the inhibition of PTPs by oxidation of cysteine residues in their active center, which are reactivated upon reduction of this site; the temporary inhibition of these phosphatases increases the level of phosphorylation of their targets, thus promoting the signaling process. For these two reactions rate laws were defined as follows:

• For the H₂O₂-dependent oxidation step (1) v₁ = k_activation × [Target_rd], where k_activation = k_{target + H2O2} × [H₂O₂]. k_{target + H2O2} is the rate constant for the direct oxidation of the target protein by H₂O₂.

• For the switch-off step (2), v₂ = k_switchoff × [Target_ox]. The total concentration of the target protein is assumed to be constant within the duration of the experiment ([Target]_total = [Target_ox] + [Target_rd]), and so v₂ = k_switchoff × ([Target]_total − [Target_rd]).

Based on these two chemical reactions, the following differential equation was set up, where Target_rd is the fraction of the target protein in the reduced state, t is time, and d/dt stands for the differential operator:

The master equation describing the time course of Target_rd is given by the analytical solution of Equation (3):

With

Target_rd |_t and Target_ox |_t are the fractions of the target protein in the reduced and oxidized state at time t, respectively. Once the experimental variation of these fractions with time is known and the fraction of reduced target at time 0 (Target_ox |₀) is measured, a non-linear fit to Equation (4) can be applied to estimate the kinetic parameters k_activation and k_switchoff. One possibility is to estimate these two unknown parameters from a two-parameter non-linear fitting. Alternatively, if one of these parameters is already known, only the remaining unknown parameter is estimated from a one-parameter non-linear fitting.

Next, we linearized Equation (4) so that kinetic parameters can be determined from linear plots, in which deviations from the master Equation (4) are easier to observe. To linearize Equation (4), the steady-state (ss) fraction of protein present in the reduced form, Target_ox |_ss, was obtained by letting t to tend to infinite, resulting in Equation (6):

Equation (6) was used to rewrite and linearize Equation (4) as Equations (7) and (8). If in the experimental time course this steady-state is not observed, Equation (4) cannot be linearized according to this procedure.

A plot of $\text{ln }\left(\frac{{\text{Target}}_{\text{rd}}{|}_t-{\text{Target}}_{\text{rd}}{|}_{ss}}{{\text{Target}}_{\text{rd}}{|}_0-{\text{Target}}_{\text{rd}}{|}_{ss}}\right)$ vs. time gives a linear relationship with slope = −(k_switchoff + k_activation).

Finally, combining this slope with Equation (6), kinetic parameters are estimated as:

Simplification of the Master Equation

Simplified forms of Equation (4) that apply to specific experimental conditions may constitute a useful alternative to estimate kinetic parameters.

Absence of H₂O₂

On the assumption that k_activation = 0, i.e., H₂O₂ is absent in the system, Equation (4) was simplified as Equation (10).

Or

This equation is applied to determine k_switchoff from time courses that follow the return of the sensor protein to its reduced form. Taking the logarithmic of both sides of Equation (10b):

A plot of ln (Target_ox|_t) vs. time produces a straight line with k_switchoff = −slope.

No target reduction

Equation (12), another simplified form of Equation (4), was obtained by ignoring target reduction, i.e., k_switchoff = 0.

This equation is used to estimate k_activation from short time courses when target reduction is still negligible. Taking the logarithmic of both sides of Equation (12):

A plot of ln (Target_rd|_t) vs. time produces a straight line with k_activation = −slope.

Concentration Studies

In all previous equations, H₂O₂ is a hidden variable that influences k_activation and kinetic parameters are estimated from experiments in which the time course of the oxidation state of the target protein is followed. If cells are exposed to various concentrations of H₂O₂, kinetic parameters may also be estimated by following the variation of the oxidation state of the target protein as a function of the H₂O₂ concentration at a given time point. To this end, k_activation was replaced by k_{target + H2O2} × [H₂O₂] in Equation (4), forming Equation (14):

For a known fixed t, a non-linear two-parameter fitting of Target_rd |_t vs. [H₂O₂] allows to estimate k_switchoff and k_{target + H2O2}. As before, if one of the two parameters is already known a one-parameter non-linear fitting may be used to determine the other parameter.

Concerning Equation (8), after specifying the H₂O₂ concentration explicitly this equation became:

A plot of $\text{ln }\left(\frac{{\text{Target}}_{\text{rd}}{|}_t-{\text{Target}}_{\text{rd}}{|}_{ss}}{{\text{Target}}_{\text{rd}}{|}_0-{\text{Target}}_{\text{rd}}{|}_{ss}}\right)$ vs. [H₂O₂] gives a linear relationship with slope = −k_{target + H2O2}× t and intercept = −k_switchoff × t. In order to use this equation, the fraction of reduced target reached at steady-state (Target_rd |_ss) must be known previously. Therefore, time courses are needed for each H₂O₂ concentration in order to obtain this value, which lessens the applicability of this equation.

In absence of target reduction, the equivalent of Equation (13) was deduced as:

A plot of ln (Target_rd|_t) vs. [H₂O₂] produces a straight line with slope = −k_{target + H2 O2} × t.

Importantly, the [H₂O₂] in these equations refers to the intracellular [H₂O₂] that reacts with the target. Therefore, in order to estimate k_{target + H2O2}, this concentration must be known. Intracellular [H₂O₂] attained when cells are exposed to extracellular H₂O₂ can be estimated from the gradient between extracellular and intracellular H₂O₂. If this gradient is unknown, then these equations may be applied with the extracellular H₂O₂ concentrations, but the value of k_{target + H2O2} obtained is referred to extracellular H₂O₂ concentrations, with the true value being higher. If instead k_{target + H2O2} is known a priori, equations may be used to estimate the gradient between extracellular and intracellular H₂O₂.

Validation of Equations

To test the validity of the equations, the following mathematical kinetic model was set up. This model simulates ideal experiments in which cells are exposed to extracellular H₂O₂ or are stimulated to produce endogenous H₂O₂ in a receptor-mediated process. A key characteristic of these virtual experiments is that the kinetic parameters concerning H₂O₂ signaling are known a priori, corresponding to the kinetic parameters introduced in the model. Thus, by fitting the virtual time courses to the equations deduced previously, the validity of the equations can be tested objectively. If the equations are valid, kinetic parameters obtained in these fittings should be similar to those used in the kinetic model. In addition, by varying several parameters of the model, namely those concerning the experimental set up describing H₂O₂ exposure, experimental conditions in which the equations are valid may be defined.

The model is described by the following differential equations, which take into account two compartments, one referring to the extracellular space (V_out) and the other to the cell volume (V_in). Multicompartmentation was implemented as described previously (Alves et al., 2006).

Reactions considered in the model (Table 1) were: extracellular production of H₂O₂ (H₂O₂_production_out), which simulates, for example, production of H₂O₂ by glucose oxidase added to the incubation medium; intracellular production of H₂O₂ (H₂O₂_production_in), which simulates the endogenous production triggered by a receptor-mediated process; permeation of H₂O₂ across the plasma membrane into (H₂O₂_import) and out of the cell (H₂O₂_export); consumption of H₂O₂ by an antioxidant enzyme (v_Gpx) and by a sensor protein target (v_Target); and finally, the switch-off mechanism of the target protein (v_switch_off). Table 2 shows the parameters used. Although kinetics and respective rate constants are based on published values, this model does not intend to model a particular cell or a specific signaling pathway. Reactivities of the target and the antioxidant enzyme toward H₂O₂ were based on that of PTP1B (Barrett et al., 1999) and glutathione peroxidase (GPx) (Flohe, 1979; Forstrom and Tappel, 1979), respectively. Levels of GPx, permeability constant for H₂O₂ across the plasma membrane, V_out and V_in were taken from Antunes and Cadenas (2000). k_switchoff was obtained from the lower range of values estimated in this work, based on previously published experiments. The resulting differential equations were solved numerically with PLAS (Voit, 1991). In the kinetic model, concentrations of Target_rd and Target_ox were used. The respective fractions were calculated subsequently so that simulation data could be analyzed with the equations deduced here. The parameter k_activation is not a rate constant in the numerical model, but it was calculated as k_{target + H2O2} × [H₂O_2in]; when [H₂O_2in] was not constant, for example when a bolus addition of H₂O₂ or the endogenous non-constant production of H₂O₂ was simulated, an average [H₂O_2in] was used.

Table 1. Reactions and respective rate laws included in the kinetic model.

Table 2. Parameter values used in the kinetic model.

Results

As a first step to test the equations derived here, redox signaling experiments were simulated to generate data that was introduced into the equations in order to determine kinetic parameters. The validity of the equations was checked by comparing the kinetic parameters obtained with those used in the simulations.

Validation of Equations with Simulated Experiments

The exposure of cell cultures to extracellular H₂O₂ initiates cellular responses that differ from those caused by the intracellular release of H₂O₂ triggered by receptor-mediated mechanisms (Forman, 2007), being the main difference the additional signal transduction pathways initiated in the first case. Nevertheless, the control of H₂O₂ delivery achieved by the extracellular exposure makes this approach more suitable for the purpose of estimating kinetic parameters. We simulated both approaches with the kinetic model described in Theory and Methods, starting with the extracellular exposure to H₂O₂.

Extracellular addition of H₂O₂

Cells may be exposed to extracellular H₂O₂ either by bolus additions or by incubation with steady-state concentrations of H₂O₂. In the bolus addition, a single dose of H₂O₂ is added to cells, constituting the most common method of exposing cell cultures to extracellular H₂O₂. It has the advantage of simplicity, but the results obtained are strongly dependent on the specific assay conditions (Marinho et al., 2013a). Among other factors, cell density and, for adherent-growing cells, the volume of incubation media dramatically affect the results.

In the steady-state methodology, exposure to H₂O₂ is calibrated so that cells are exposed for a known concentration of H₂O₂ that remains constant during the assay. Although more complex, this approach has much better experimental reproducibility with the actual H₂O₂ concentration in the assay being independent of experimental conditions. The implementation of this methodology is described in detailed in (Covas et al., 2013; Cyrne et al., 2013; Marinho et al., 2013a).

Steady-state. In the deduction of the kinetic equations, k_activation (k_{target + H2O2} × [H₂O₂]) was considered to be a constant parameter, i.e., H₂O₂ was assumed to be constant with time. So, as a positive control we started by analyzing the results simulated with a steady-state incubation, a case in which the equations tested should be valid.

In the first condition analyzed, cellular exposure to H₂O₂ (Figure 1A, curve 2) was long enough so that a balance between oxidation of the target that senses H₂O₂ and its regeneration was achieved. As observed in Figure 1A, curve 1, initially the target was oxidized until its oxidation state reached a near steady-state value as given by Equation (6). Simulated results were transformed according to Equation (8) (Figure 1B, blue line), with k_activation and k_switchoff being estimated from Equations (9A) and (9B). The estimations obtained, k_activation = 8.1 × 10⁻⁴ s⁻¹ and k_switchoff = 1.0 × 10⁻³ s⁻¹, matched closely the respective expected values of 7.9 × 10⁻⁴ s⁻¹ and 1.0 × 10⁻³ s⁻¹, which were used in the simulations to draw curve 1 in Figure 1A. Note that the fitting to Equation (8) departed from linearity for longer time points (Figure 1B), when Target_rd approached its steady-state value (Figure 1B). This behavior was caused by small uncertainties in this value, which must be known in order to plot data according to Equation (8). As an alternative, the two parameters were also obtained from a non-linear fitting to Equation (4) (Figure 1A, red dashed line), which neither requires data transformation nor knowing the value of the fraction of reduced target at steady-state. In this case, the estimated parameters, k_activation = 7.9 × 10⁻⁴ s⁻¹ and k_switchoff = 1.0 × 10⁻³ s⁻¹, matched exactly the expected values.

Figure 1. Determination of kinetic parameters from simulated experiments when extracellular H₂O₂ was delivered as a steady-state. H₂O₂ concentration (curve 2, green line) was maintained during the duration of the experiment (A) or was stopped at 15 min (C). (A) The profile of the fraction of the H₂O₂ target in the reduced form (Target_rd, blue line) was obtained by simulation of the kinetic model described in the Theory and Methods Section with v_H₂O_2out = 3.51 × 10⁻⁷ M s⁻¹ and [H₂O_2out] = 3 × 10⁻⁴ M at time = 0, (intracellular H₂O₂ production was absent, k_ H₂O_2in = 0); red dashed line is the two-parameter non-linear fitting to Equation (4) used to estimate the kinetic parameters k_activation and k_switchoff. (B) Fitting of the profile of Target_rd obtained in (A) to Equation (8) (curve 1, blue line) using Target_{rd_0} = 1 and Target_{rd_ss} = 0.56, (Tg in y-axis title means Target). k_activation and k_switchoff were estimated from the slope of the straight line by applying Equations (9A) and (9B). (C) The profile of the fraction of Target_rd (curve 1, blue line) was obtained as in (A) but v_H₂O_2out and [H₂O₂]_out were set to zero at 15 min; red dashed line is the one-parameter non-linear fitting to Equation (4) in which k_switchoff obtained in (D) curve 2 was used as an input to estimate k_activation. (D) Target_rd was fitted to Equation (13) (curve 1, left y-axis) while H₂O₂ was present, afterwards Target_ox was fitted to Equation (11) (curve 2, right y-axis); k_activation was estimated from the linear part of the fitting to Equation (13), k_switchoff from the fitting to Equation (11).

If the fraction of the reduced target does not reach a steady-state because, for example, a balance between its oxidation and regeneration is not attained before the exposure to H₂O₂ is terminated, Equation (8) cannot be applied. To illustrate this situation, a simulation was done under the exact same conditions as before with the exception that H₂O₂ exposure lasted only for 15 min (Figure 1C). Experimentally, this is equivalent to either replacing the external media to remove the H₂O₂ generating system, such as glucose oxidase, or by adding external catalase to the incubation media. This simulation was analyzed with Equation (13), an equation deduced ignoring the regeneration of the reduced form of the target. Good enough estimations were obtained by using only the first stage of the time course, when the degree of target oxidation was still low, and therefore the contribution of its regeneration to the time course of Target_rd, could be ignored. From the slope of the initial linear part of the curve (Figure 1D, curve 1), a k_activation = 6.8 × 10⁻⁴ s⁻¹ was obtained, which was close to the expected value of 7.9 × 10⁻⁴ s⁻¹. Concerning k_switchoff, this kinetic parameter was estimated by fitting to Equation (11) the part of the curve starting after removal of external H₂O₂. Note that Equation (11) was deduced assuming reduction of the oxidized target when H₂O₂ was absent. An excellent linear plot was observed (Figure 1D, curve 2) with the estimated k_switchoff of 1.0 × 10⁻³ s⁻¹ matching the expected value. Kinetic parameters were also obtained by non-linear fittings of Equation (4) to the first part of the curve when H₂O₂ was still present, either as a two-parameter non-linear fitting in which the two parameters – k_activation and k_switchoff – were determined, or as a one-parameter non-linear fitting, in which only one of the parameters was estimated, with the other being obtained from the linear plots of Figure 1B. If the conditions of the experimental assay fulfill all the assumptions applied to deduce Equation (4), a two-parameter non-linear fitting is the best choice, since both parameters are obtained without transformation of the experimental data. However, if the assumptions are not all fulfilled, which is the most common situation, we advise to apply a one-parameter non-linear fitting to Equation (4), inputting as a known parameter k_switchoff estimated from Equation (11), being k_activation the unknown parameter. As always whatever the option taken, the goodness of the fitting should be inspected. The dashed line in Figure 1C was obtained as a one-parameter non-linear fitting using k_switchoff = 1.0 × 10⁻³ s⁻¹ with the estimated k_activation value of 7.9 × 10⁻⁴ s⁻¹ matching the expected value.

Overall, kinetic parameters estimated from simulated experiments in which H₂O₂ was delivered as a steady-state matched the expected values, validating the equations applied. This could be anticipated because Equation (4) relies on the key assumption that H₂O₂ concentration is constant during the experiment.

Bolus addition. The bolus addition set up, the most common experimental approach to expose cells to H₂O₂, was simulated in Figure 2. Upon incubation with a 1 mM bolus addition, the H₂O₂ sensor was oxidized, the Target_rd fraction reached a minimum at approximately 12 min, then regeneration became more important than oxidation, and Target_rd increased (Figure 2A, curve 1). Kinetics of H₂O₂ consumption depends on the experimental set up, and under the conditions of this simulation, H₂O₂ was fully consumed after 60 min (Figure 2A, curve 2). Nevertheless, the general pattern observed in this simulation served as a test case to check how the non-constant H₂O₂ concentration impacts the estimation of kinetic parameters.

Figure 2. Determination of kinetic parameters from simulated experiments when H₂O₂ was delivered as an extracellular bolus addition. (A) Curve 1 is the profile of Target_rd fraction (blue line) obtained by simulation with v_H₂O_2out = 0 M s⁻¹ and [H₂O_2out] = 1 × 10⁻³ M at time = 0, (intracellular H₂O₂ production was absent, k_ H₂O_2in = 0). Curve 1′ is the profile of Target_rd (blue line) obtained as in curve 1 but [H₂O_2out] was set to zero at 12 min; red dashed line is the one-parameter non-linear fitting to Equation (4) in which k_switchoff obtained in (B) was used as an input to estimate k_activation. (B) Until 12 min results were fitted to Equation (13) (curve 1, left y-axis), afterwards were fitted to Equation (11) (curve 2, right y-axis); k_activation and k_switchoff were estimated from the linear part of the fittings to Equations (13) and (11), respectively.

Because the fraction of reduced target never reached a constant value, Equation (8) was not applied, and instead Equation (13) was used to estimate k_activation. Only the first part of the curve was considered (Figure 2B, curve 1), because shorter time courses minimize target regeneration, a process ignored by Equation (13). The k_activation estimation of 1.8 × 10⁻³ s⁻¹ was close to the expected value of 1.7 × 10⁻³ s⁻¹. Concerning k_switchoff, this parameter was estimated by fitting to Equation (11) the part of the curve after Target_rd reached its minimum (Figure 2B, curve 2). The presence of H₂O₂ in this part of the experiment promoted target oxidation, violating a key assumption behind Equation (11), and consequently deviations from linearity were observed. Even by using only the linear part of the curve, the k_switchoff estimation of 6.7 × 10⁻⁴ s⁻¹ underestimated the expected value of 1.0 × 10⁻³ s⁻¹. Removal of external H₂O₂ at 12 min, when Target_rd reached its minimum (Figure 2A, curve 2′), changed the regeneration profile of the H₂O₂ target (Figure 2A, curve 1′) and vastly improved the fitting to Equation (11) (Figure 2B, curve 2′) with the estimated k_switchoff of 1.0 × 10⁻³ s⁻¹ matching exactly the expected value.

Concerning the non-linear fitting to Equation (4), a k_activation = 1.8 × 10⁻³ s⁻¹ was obtained when k_switchoff = 6.7 × 10⁻⁴ s⁻¹ was used as an input. Alternatively, by inputting a k_switchoff = 1.0 × 10⁻³ s⁻¹ a k_activation = 2.0 × 10⁻³ s⁻¹ was obtained.

Overall, these results indicate that the proposed equations can be applied to experiments in which H₂O₂ is delivered as a bolus addition. The accuracy of parameter estimation improves if H₂O₂ is removed at the time when the reduced form of the target reaches its minimum.

Concentration studies. Besides time courses, studies often evaluate how the concentration of H₂O₂ affects the oxidation state of the sensor target. We started by simulating the dependency of Target_rd on external H₂O₂ concentration, delivered as a steady-state during 10 min (Figure 3A, curve 1). From the non-linear fitting to Equation (14), kinetic parameters that matched exactly the expected values were obtained (k_{target + H2O2} = 2.6 M⁻¹ s⁻¹ and k_switchoff = 1.0 × 10⁻³ s⁻¹). Note that k_{target + H2O2} estimated from the fitting was based on external H₂O₂ concentrations. By considering the gradient between these and the intracellular H₂O₂ concentrations—15 in the present simulation—the estimated value of the rate constant of 40 M⁻¹ s⁻¹ for the reaction between the target and H₂O₂ matched the value used in the simulation. Results were also linearized and fitted to Equation (16) (curve 2 in Figure 3A), which was deduced assuming absence of target reduction, i.e., k_switchoff = 0 s⁻¹. In this case, the estimated value of 1.7 M⁻¹ s⁻¹ for k_{target + H2O2}, which was converted to 26 M⁻¹ s⁻¹ when intracellular H₂O₂ concentrations were considered, underestimated the expected value.

Figure 3. Determination of kinetic parameters from simulated experiments when the concentration of extracellular H₂O₂ was changed. The reduced form of the target fraction (Target_rd) obtained at 10 min is plotted as a function of either H₂O₂ steady-state concentrations in (A) or initial bolus additions in (B) (curves 1, blue linen). Simulations were run with v_H₂O_2out varying in the range (0.35–23.4) × 10⁻⁷ M s⁻¹ and [H₂O_2out] in the range (0.03–1) × 10⁻³ M at time = 0 in A, while in B v_H₂O_2out was set to zero and [H₂O_2out] was changed in the range (0.03–1) × 10⁻³ M at time = 0; intracellular H₂O₂ production was absent (k_ H₂O_2in = 0) in both cases. In (A,B), results were analyzed with non-linear fits of Target_rd to Equation (14) (curve 1, red dashed line) in order to estimate k_{target + H2O2} and k_switchoff, or they were linearized according to Equation (16) (curve 2, blue line) with k_{target + H2O2} being estimated from the slopes of the red dashed lines.

To test how Equations (14) and (16) behave with data generated with bolus additions, the study of Figure 3A was repeated but now the H₂O₂ concentrations introduced in the equations were the initial bolus additions (Figure 3B, curve 1). Kinetic parameters obtained with the non-linear fitting to Equation (14) were k_{target + H2O2} = 2.0 M⁻¹ s⁻¹ (or 30 M⁻¹ s⁻¹ if referred to intracellular H₂O₂), and k_switchoff = 1.3 × 10⁻³ s⁻¹. Linearization according to Equation (16) (curve 2 in Figure 3B) gave a k_{target + H2O2} of 1.2 M⁻¹ s⁻¹, (or 18 M⁻¹ s⁻¹ if referred to intracellular H₂O₂). As expected, these estimations were less accurate than those obtained when H₂O₂ was delivered as a steady-state, but nevertheless they constitute satisfactory semi-quantitative estimations.

Receptor-mediated endogenous H₂O₂ production

The endogenous production of H₂O₂ upon cell stimulation by a ligand will give the best picture of the influence of H₂O₂ in a particular cell signaling pathway, as H₂O₂ production is both spatial and time restricted (Forman, 2007). Nevertheless, since the profile of H₂O₂ concentration generated is unknown this imposes potential problems to the determination of kinetic parameters. To test how the kinetic equations behave under such circumstances, we started by simulating a case where H₂O₂ intracellular production was rapidly triggered and then set at a near constant value. This scenario worked as positive control and was analyzed as described previously for the extracellular addition of steady-state H₂O₂. The kinetic parameters obtained matched exactly the expected values or were very close to these values depending on the fittings applied (not shown). Nevertheless, this scenario is seldom achieved when H₂O₂ is produced endogenously, and next we tested the kinetic equations under non-constant H₂O₂ intracellular production.

H₂O₂ endogenous production was simulated with a sine-like function: there was an initial increase in the H₂O₂ concentration, reaching its maximum at 10 min, and then a decrease until H₂O₂ production stopped at 20 min (Figures 4A,C, curve 2). In this context, two scenarios were simulated. In the first, H₂O₂ production was high enough so that a near constant level of reduced target was observed (Figure 4A, curve 1), and accordingly results were fitted to Equation (8). The presence of a non-constant H₂O₂ production caused deviations from linearity (Figure 4B, curve 1). Nevertheless results obtained from the near-linear intermediate portion of the curve gave estimations, k_switchoff = 7.8 × 10⁻⁴ s⁻¹ and k_activation = 1.2 × 10⁻² s⁻¹, that compared well with the expected parameters, k_switchoff = 1.0 × 10⁻³ s⁻¹ and k_activation = 1.0 × 10⁻² s⁻¹. As before, k_switchoff was also obtained from the second part of the curve by fitting data to Equation (11) (Figure 4B, curve 2), giving a k_switchoff = 7.6 × 10⁻⁴ s⁻¹, an underestimation of the expected value. As described for the bolus addition, removal of H₂O₂ from the system after Target_rd reached its minimum improved the estimations (not shown). In real experiments, the effect of this addition will be dependent on whether removal of extracellular H₂O₂ decreases the localized intracellular levels of H₂O₂. If this occurs, a change in the reduction profile of the oxidized target should be observed. In this simulation, the application of non-linear fittings did not improve the estimations of kinetic parameters: from a non-linear fitting where k_switchoff = 7.6 × 10⁻⁴ s⁻¹ was used as input (dashed line in Figure 4A) a k_activation of 0.51 × 10⁻² s⁻¹ was obtained (Figure 4A, dashed line) and a two-parameter non-linear fitting did not improve these estimations.

Figure 4. Determination of kinetic parameters from simulated experiments when H₂O₂ was produced endogenously. A non-constant sine-like H₂O₂ intracellular exposure (green lines) was simulated as kH₂O_2in × sine(time/1200 × 3.14), (extracellular production was absent, v_H₂O_2out = 0). (A) Profile of Target_rd fraction (blue line) was simulated with kH₂O_2in = 5 × 10⁻³ M s⁻¹; red dashed line is the one-parameter non-linear fitting to Equation (4) in which k_switchoff obtained in (B) curve 2 was used as an input to estimate k_activation. (B) The first part of the profile (until 11.3 min) of Target_rd was fitted to Equation (8) (curve 1, blue line) using Target_{rd_0} = 1 and Target_{rd_ss} = 0.06; the kinetic parameters k_activation and k_switchoff were obtained from the slope of near-linear intermediate portion of the curve by applying Equations (9A) and (9B). After 11.3 min, the profile of Target_ox was fitted to Equation 13, from which k_switchoff was estimated. (C) The profile of Target_rd fraction (blue line) was obtained as in (A), but with kH₂O_2in = 0.5 × 10⁻³ M s⁻¹; red dashed line is the one-parameter non-linear fitting to Equation (4) in which k_switchoff obtained in (D) curve 2 was used as an input to estimate k_activation. (D) Target_rd was fitted to Equation (13) (curve 1, left y-axis) until 16.0 min, afterwards, Target_ox was fitted to Equation (11) (curve 2, right y-axis); k_activation and k_switchoff were estimated from the linear part of the fittings to Equations (13) and (11), respectively.

In the second simulation in which endogenous production of H₂O₂ followed a sine-like function, a constant level of reduced target was not observed (Figure 4C, curve 1). The k_activation estimation of 1.0 × 10⁻³ s⁻¹, obtained from the linear portion of the plot according to Equation (13) (Figure 4D, curve 1), matched the expected value. The k_switchoff estimation of 9.3 × 10⁻⁴ s⁻¹, obtained from the second part of the curve after fitting data to Equation (11) (Figure 4D, curve 2), was close to the expected value of 1.0 × 10⁻³ s⁻¹. A one-parameter non-linear fitting (Figure 4C, dashed line) gave a k_activation of 1.0 × 10⁻³ s⁻¹, i.e., the expected value, when a k_switchoff of 9.3 × 10⁻⁴ was used as input.

Overall, when H₂O₂ production is not constant deviations from the equations derived here are expected. Nevertheless, estimated kinetic parameters are still satisfactory at a semi-quantitative level, and the deviations from linearity in the plots proposed here may be used as a useful tool to diagnose a non-constant H₂O₂ production.

Fits to Experimental Data

The use of simulation data was useful to test the validity of the equations deduced and to figure out how deviations from the assumptions behind their deduction affected the estimation of kinetic parameters. Nevertheless, simulation data points are virtually infinite and devoid of experimental error; in contrast real experiments contain a finite number of measurements with associated experimental error. To test how equations cope with these issues, they were applied to data obtained from the literature for two PTPs, PTP1B and SHP-2.

H₂o₂-external delivery

Two experiments in which H₂O₂ was added externally as a bolus addition (Figure 5) were analyzed. In the first experiment (Rinna et al., 2006), the time course of PTP1B oxidation was followed in a rat alveolar macrophage cell line after addition of a 100 μM H₂O₂ bolus dose for 15 min (Figure 5A, curve 1). By fitting data to Equation (4) with a two parameter non-linear fitting, a k_activation of 1.1 × 10⁻³ s⁻¹ and a k_switchoff of 2.6 × 10⁻³ s⁻¹ were estimated. Alternatively, when results were linearized according to Equation (13) (Figure 5A, curve 2) a k_activation of 0.59 × 10⁻³ s⁻¹ was obtained, after discarding the 15 min point. Note that also with simulation data a similar deviation from linearity at late time points was observed (Figure 2B, curve 1). By considering an external H₂O₂ concentration of 100 μM, the apparent first-order rate constant k_activation in the range (0.59–1.1) × 10⁻³ s⁻¹ was converted to a rate constant between the target and H₂O₂ (k_{target + H2O2}) of 5.9–11 M⁻¹ s⁻¹. This value refers to extracellular H₂O₂, and so if the gradient between extracellular and intracellular H₂O₂ was considered the value of k_{target + H2O2} for PTP1B would be higher.

Figure 5. Determination of kinetic parameters from experiments when H₂O₂ was delivered as an extracellular bolus addition. (A) Data (■) was taken from Figure 1 in Rinna et al. (2006). Gel was digitalized and analyzed with ImageJ (Rasband, 1997); fraction of oxidized target was calculated assuming a maximum level of oxidation of 31% (Rinna et al., 2006). Time course was fitted to Equation (4), with k_activation and k_switchoff as variables (curve 1, left y-axis), or was linearized (▲) and fitted to Equation (13) (curve 2, right y-axis). (B) Data (■) was taken from Figure 2A in Meng et al. (2002). Gel was digitalized and analyzed with ImageJ (Rasband, 1997); fraction of oxidized protein tyrosine phosphatase corresponding to 70 kDa (SHP-2) was calculated assuming that complete oxidation was achieved by 1 mM H₂O₂. Time course was fitted to Equation (14), with k_{target + H2O2} and k_switchoff as variables (curve 1, left y-axis), or was linearized (▲) and fitted to Equation (16) (curve 2, right y-axis).

In the second experiment, rat-1 fibroblasts were subjected to H₂O₂ bolus additions in the range 0–500 μM for 1 min, followed by the measurement of the oxidation level of SHP-2 (Meng et al., 2002) (Figure 5B, curve 1). After fitting data to Equation (14) (Figure 5B, curve 1) with a two parameter non-linear fitting, a k_{target + H2O2} of 60 M⁻¹ s⁻¹ and a k_switchoff of 1.3 × 10⁻³ s⁻¹ were estimated. Linearization according to Equation (16) gave a k_{target + H2O2} of 57 M⁻¹ s⁻¹ (Figure 5B, curve 2). Again, k_{target + H2O2} values refer to extracellular H₂O₂ concentrations. Even if a bolus addition was used, because short-term incubations of 1 min were done, the assumption of constant H₂O₂ behind the deduction of Equations (14) and (16) was verified.

Receptor-mediated signaling

To test how equations behave when analyzing receptor-mediated signaling, the following two experiments were considered. In the first, A431 human epidermoid carcinoma cells were stimulated by EGF, triggering H₂O₂ intracellular production that lead to PTP1B oxidation and inhibition (Figures 6A,B), while in the second experiment, rat-1 cells were stimulated with PDGF inducing SHP-2 oxidation (Figures 6C,D). In both cases, the profile of PTP oxidation did not reach a near steady-state, precluding the application of Equation (8). Concerning k_switchoff, estimations of 1.9 × 10⁻³ s⁻¹ and 8.7 × 10⁻³ s⁻¹ were obtained, respectively for PTP1B and SHP-2 reactivation, after fitting to Equation (11) the second part of the PTP oxidation curves (curves 2 in Figures 6B,D). For k_activation, estimations of 1.0 × 10⁻³ s⁻¹ and 9.3 × 10⁻³ s⁻¹ were obtained, respectively for PTP1B and SHP-2, after applying Equation (13) to linearize the first part of the PTP oxidation profile (curve 1 in Figures 6B,D). These k_activation values were close to those obtained from non-linear fittings, 2.0 × 10⁻³ s⁻¹ and 9.7 × 10⁻³ s⁻¹ for PTP1B and SHP-2, respectively (dashed lines in Figures 6A,C).

Figure 6. Determination of kinetic parameters from experiments where intracellular H₂O₂ production was triggered by cell stimulation. (A) Data concerning PTP1B oxidation following stimulation of A431 cells with EGF (■) was taken from Figure 2 in Lee et al. (1998). (C) Data concerning SHP-2 oxidation following stimulation of rat-1 cells by PDGF (■) was taken from gel in Figure 4 in Meng et al. (2002) after digitalization and analysis with ImageJ (Rasband, 1997), and assuming that 100% of oxidation of SHP-2 was achieved in absence of iodoacetic acid as described in Meng et al. (2002). The red dashed lines in (A,C) are one-parameter non-linear fittings to Equation (4) in which k_switchoff values obtained from curves 2 in (B,D), respectively, were used as input to estimate k_activation. (B,D) Linearization of data shown in (A,C), respectively; results in the first part of the experiment were fitted to Equation (13) (curves 1, left y-axis), afterwards were fitted to Equation (11) (curves 2, right y-axis); k_activation was estimated from the fitting to Equation (13), k_switchoff from the fitting to Equation (11).

Overall, data taken from literature fitted well to the equations deduced here, even if experiments analyzed were carried out without any special concern considering their application to estimate kinetic parameters.

Discussion

Herein, we deduced equations to determine kinetic parameters from typical redox signaling experiments in which H₂O₂ is either added externally to cells or is endogenously produced following receptor activation by diverse cellular stimuli. The equations were shown to be accurate after fitting them to data generated by simulations. We also performed simulations in which the assumption that H₂O₂ is constant during the experiment was not fulfilled, that is, H₂O₂ was delivered as a bolus addition, or the endogenous production of H₂O₂ was not constant. Under these conditions, deviations from linearity were observed when simulation results were plotted according to the linear equations we deduced. Nevertheless, the estimated kinetic parameters were close to the parameters introduced in the simulations. Finally, we tested the application of the equations to real experiments with published experimental data concerning the H₂O₂ signaling mediated by inhibition of PTPs, namely PTP1B and SHP-2. While in general excellent fittings were obtained, in some cases deviations as those observed when H₂O₂ was added as a bolus addition were observed. In general, the estimated kinetic parameters (Table 3) are consistent with the published rate constants.

Table 3. Kinetic parameters estimated in this work based on the analysis of published data.

Concerning the parameters describing redox signal switching-off (k_switchoff), which in the case of the PTPs analyzed here corresponds to their reactivation, the results summarized in Table 3 constitute, as far as we know, the first cell-based kinetic estimates for this process. This characterization is relevant because modulation of PTP reactivation regulates cell signaling (Dagnell et al., 2013). For PTP1B, k_switchoff values in the range (1.9–2.6) × 10⁻³ s⁻¹ were estimated, which are plausible taking into account the known data for PTP1B reactivation. In vitro, reduced thioredoxin (2 μM) reactivates oxidized PTP1B with an apparent rate constant of 1.4 × 10⁻³ s⁻¹ (Parsons and Gates, 2013), which corresponds to a rate constant of 700 M⁻¹ s⁻¹ for this reaction. Thus, considering this rate constant and the k_switchoff values determined here, we estimate the cellular concentration of reduced thioredoxin to be 2–3 μM. This range is close to the value observed experimentally in Jurkat T-cells, 0.43 μM (Adimora et al., 2010), with the difference observed being easily account for different cell lines used or by the participation of alternative thioredoxin-related proteins, like the redoxin TRP14, in the reactivation of PTP1B (Dagnell et al., 2013). This agreement further strengths the validity of the approach we purpose here to reveal kinetic information hidden in typical redox signaling experiments.

For SHP-2, k_switchoff was estimated in the range (1.3–8.7) × 10⁻³ s⁻¹ (Table 3), which is similar or higher than the range estimated for PTP1B. This is unexpected because the reactivity of SHP-2 toward thioredoxin is about 20 times lower than PTP1B (Parsons and Gates, 2013). Either the cell line rat-1, where SHP-2 reactivation data was obtained (Meng et al., 2002), contains much higher levels of thioredoxin or an alternative system other than thioredoxin is reactivating SHP-2. The second alternative is supported by the observation that in cells lacking thioredoxin reductase TrxR1, a key partner of thioredoxin that keeps it in the reduced state, SHP-2 oxidation remains unchanged (Dagnell et al., 2013).

Concerning the oxidation of PTPs by H₂O₂, rate constants estimated from experiments in which extracellular H₂O₂ was added were 5.9–11 M⁻¹ s⁻¹ for PTP1B and 57–60 M⁻¹ s⁻¹ for SHP-2 (Table 3). These values were calculated based on the external H₂O₂ concentrations added to cells. The actual H₂O₂ concentration that oxidizes these targets is lower as H₂O₂ gradients across the plasma membrane are established when cells are incubated with extracellular H₂O₂ (Antunes and Cadenas, 2000; Marinho et al., 2013b). Thus, the value of these rate constants will be higher if they are based on the actual intracellular H₂O₂ concentrations that oxidize PTPs. For PTP1B, rate constants obtained in kinetic studies with purified PTP1B are in the range 9–43 M⁻¹ s⁻¹ (Denu and Tanner, 1998; Barrett et al., 1999; Zhou et al., 2011; Marinho et al., 2014), and so a gradient between the extracellular and the intracellular concentration of H₂O₂ at the site of PTP1B oxidation is estimated to be in the range 2–7 for the experiments analyzed in this work, which matches the lower range of known gradients for human cell lines (Antunes and Cadenas, 2000; Makino et al., 2004; De Oliveira-Marques et al., 2007; Oliveira-Marques et al., 2013). However, gradients as high as 650 have been recently estimated taking into account the participation of peroxiredoxin (Huang and Sikes, 2014), whose role in the degradation of H₂O₂ is still an open issue (Benfeitas et al., 2014). Thus, the rate constants obtained for PTP1B fit the known quantitative data for the reactivity of this PTP with H₂O₂.

For SHP-2, the estimated rate constants of 57–60 M⁻¹ s⁻¹ (Table 3) for its oxidation by H₂O₂ were higher than those determined in vitro with purified SHP-2, which are in the range 9–15 M⁻¹ s⁻¹ (Chen et al., 2009; Zhou et al., 2011). Moreover, if the gradient of H₂O₂ across the plasma membrane is taken into account this difference will be even higher. Several possible explanations may account for this discrepancy. First, kinetic rate constants obtained in vitro with purified proteins may not reflect rate constants under in vivo conditions (Van Eunen et al., 2010, 2012). Second, peroxy-derivatives such as peroxymonocarbonate (Trindade et al., 2006; Zhou et al., 2011) and peroxymonophosphate (LaButti et al., 2007), which have higher reactivity with PTPs than H₂O₂, could be the actual species that oxidize SHP-2. Third, the primary sensor of H₂O₂ may not be SHP-2 but a high-reactive target that slowly relays the oxidation to SHP-2 (Winterbourn and Hampton, 2008; Forman et al., 2010; Brigelius-Flohé and Flohé, 2011; Ferrer-Sueta et al., 2011). Note that the models described here do not distinguish between a mechanism in which a low-reactive sensor is slowly oxidized by H₂O₂, from a mechanism in which a high-reactive sensor is rapidly oxidized by H₂O₂ and then, through a thiol-disulfide reshuffling transfer reaction, slowly oxidizes a low reactive sensor such as SHP-2. In general, known data about redox signaling pathways is consistent with either of these two scenarios (Marinho et al., 2014). Distinguishing between these possible alternative mechanisms will be possible after collecting rate constants in several cell lines upon the generalized application of the equations deduced here to redox signaling experiments.

The kinetic parameters estimated from experiments in which cells are activated by receptor-mediated pathways indicated that the apparent first-order rate constant for the oxidation of SHP-2 is about 5 times higher than that for PTP1B (Table 3). Because k_activation = k_{target + H2O2} × [H₂O₂], either the localized H₂O₂ intracellular concentration is higher in the experiment in which SHP-2 oxidation was observed, or k_{target + H2O2} is higher for SHP-2 than for PTP1B, or both. In this regard, the EGF receptor, the H₂O₂ producing enzyme NOX2, and SHP-2 immunoprecipitated all together (Paulsen et al., 2012), supporting the possibility of a highly localized H₂O₂ signaling pool. For PTP1B, from the k_activation estimation of (1.0–2.0) × 10⁻³ s⁻¹ the local intracellular H₂O₂ concentration reached locally in A431 cells, when stimulated by EGF under the experimental conditions described in Lee et al. (1998), is estimated to be in the range 23–220 μM, assuming a k_{target + H2O2} value in the range 9–43 M⁻¹ s⁻¹. Such local concentrations, particularly those in the low range of these values, can potentially be reached upon the concerted action of local production of H₂O₂ by NADPH oxidases (Chen et al., 2008; Mishina et al., 2011; Paulsen et al., 2012) and localized inhibition of H₂O₂ removing enzymes (Woo et al., 2010; Rawat et al., 2013). In addition, it can also be suggested that H₂O₂ diffusion out of membrane-entrapped signaling microcompartments may be constrained, because biomembranes constitute a regulable barrier for H₂O₂ diffusion (Antunes and Cadenas, 2000; Branco et al., 2004; Bienert et al., 2007; Miller et al., 2010).

While the equations deduced here were applied successfully to typical signaling experiments, a few alterations in the way experiments are carried out will improve the accuracy of parameter estimation. When cells are exposed to extracellular H₂O₂, we suggest a steady-state delivery so that H₂O₂ is constant during the experiment (Marinho et al., 2013a), a key assumption considered in the deduction of the equations. If the use of a bolus addition is absolutely needed, we suggest short-term experiments so that the H₂O₂ decay caused by its cellular consumption is less significant. Finally, removal of H₂O₂ by adding catalase or replacing extracellular incubation media without H₂O₂, during the second part of the experiment when target reduction starts to predominate, improves the estimation of k_switchoff values. This last suggestion may also be applied when H₂O₂ production is triggered by a receptor-mediated mechanism following cell stimulation.

In conclusion, the application of the equations deduced here to typical redox-signaling experiments reveals valuable quantitate kinetic information. Of note, the equations described require only measuring the relative levels of oxidation of a H₂O₂ sensor target and not absolute concentrations, thus facilitating their application to most experiments. While equations were tested with PTP signaling, they can be applied to other proteins that react with H₂O₂, such has thiol-proteins and those containing metal-centers. Being characterized by the presence of both multiple parallel pathways and biphasic effects, redox regulation is a field that will benefit from the widespread determination of kinetic parameters. Such knowledge is important to distinguish apparent contradictory biological effects of reactive oxygen species that are involved in pathological damaging pathways and, at the same time, are part of normal functional signaling pathways. In this way, the present knowledge on redox signaling and oxidative stress would be more efficiently translated into therapeutic applications.

Conflict of Interest Statement

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Acknowledgments

Supported by Fundação para a Ciência e a Tecnologia (FCT), Portugal (PEst-OE/QUI/UI0612/2013 and VIH/SAU/0020/2011).

Abbreviations

PTP1B, Protein tyrosine phosphatase 1; PTPs, protein tryrosine phosphatases; the Src homology 2 (SH2) domain containing phosphotyrosine phosphatase 2 (Shp-2).

References