Hyperoxidation of mitochondrial peroxiredoxin limits H₂O₂‐induced cell death in yeast

Abstract

Hydrogen peroxide (H₂O₂) plays important roles in cellular signaling, yet nonetheless is toxic at higher concentrations. Surprisingly, the mechanism(s) of cellular H₂O₂ toxicity remain poorly understood. Here, we reveal an important role for mitochondrial 1‐Cys peroxiredoxin from budding yeast, Prx1, in regulating H₂O₂‐induced cell death. We show that Prx1 efficiently transfers oxidative equivalents from H₂O₂ to the mitochondrial glutathione pool. Deletion of PRX1 abrogates glutathione oxidation and leads to a cytosolic adaptive response involving upregulation of the catalase, Ctt1. Both of these effects contribute to improved cell viability following an acute H₂O₂ challenge. By replacing PRX1 with natural and engineered peroxiredoxin variants, we could predictably induce widely differing matrix glutathione responses to H₂O₂. Therefore, we demonstrated a key role for matrix glutathione oxidation in driving H₂O₂‐induced cell death. Finally, we reveal that hyperoxidation of Prx1 serves as a switch‐off mechanism to limit oxidation of matrix glutathione at high H₂O₂ concentrations. This enables yeast cells to strike a fine balance between H₂O₂ removal and limitation of matrix glutathione oxidation.

Introduction

Reactive oxygen species (ROS) are an unavoidable consequence of life in an oxygen‐rich environment. Once considered solely as harmful molecules, which cells seek to remove as efficiently as possible, it is now accepted that some ROS have important physiological functions. In this regard, one of the best understood ROS is H₂O₂, which acts as a second messenger in several key cellular signaling pathways (Sundaresan et al, 1995; Delaunay et al, 2002; Sobotta et al, 2015; Stocker et al, 2018). On the other hand, it remains unequivocal that high concentrations of H₂O₂ are toxic and can lead to cellular dysfunction and cell death. Presumably therefore, cells strive to tightly regulate H₂O₂ to permit sufficiently large fluctuations in H₂O₂ concentration for signaling purposes, while simultaneously preventing accumulation of H₂O₂ to toxic levels.

Surprisingly, the exact mechanism(s) by which H₂O₂ leads to cell death remain poorly understood (Whittemore et al, 1995; Day et al, 2012; Uhl et al, 2015). The mediocre reactivity of H₂O₂ with most biological molecules argues against direct oxidation of cellular biomolecules being a principal driver of cell death, at least at physiologically relevant H₂O₂ concentrations (Winterbourn & Metodiewa, 1999; Stone, 2004; Winterbourn, 2008). Other possible triggers of H₂O₂‐induced cell death have been proposed, including induction of apoptosis (Greetham et al, 2013), depletion of reduced cytosolic thioredoxins (Day et al, 2012), and disruption of redox signaling pathways (Sies, 2017), but well‐defined molecular mechanisms remain largely elusive. Therefore, in this study we sought to address the molecular underpinnings of cellular H₂O₂ toxicity, using budding yeast as a model system.

We hypothesized that mitochondrial matrix‐localized processes may play an important role in H₂O₂ toxicity, for two reasons. First, the mitochondrial respiratory chain is thought to be a major source of cellular H₂O₂, and thus, the matrix is in close proximity to key sites of H₂O₂ production (Murphy, 2009; Quinlan et al, 2013). Specifically, the “leakage” of electrons from respiratory chain complexes to molecular oxygen leads to the generation of superoxide anions. Superoxide dismutases located in both the mitochondrial matrix (Sod2) and the cytosol/intermembrane space (Sod1) facilitate the subsequent rapid dismutation of superoxide to H₂O₂. Second, in comparison with the cytosol, the mitochondrial matrix appears to be poorly equipped with H₂O₂‐detoxifying enzymes. The only enzymes in the yeast mitochondrial matrix that are known to be able to react efficiently with H₂O₂ are the 1‐Cys peroxiredoxin, Prx1, and perhaps the glutathione peroxidase homolog, Gpx2 (Park et al, 2000; Ukai et al, 2011). Nonetheless, many aspects of H₂O₂ handling in the matrix remain poorly understood, for example, the efficiency of the matrix‐localized H₂O₂ removal systems. Furthermore, the crosstalk between matrix H₂O₂ and different matrix redox couples is unclear, and often, conflicting results are present in the literature. For example, the exact reductive mechanisms for both Prx1 and Gpx2 remain a matter of debate, particularly regarding the specific roles of glutathione, glutaredoxins, thioredoxins and thioredoxin reductase (Pedrajas et al, 2000, 2010, 2016; Avery & Avery, 2001; Tanaka et al, 2005; Greetham & Grant, 2009).

To gain further insight into matrix H₂O₂ handling in general and more specifically into possible mechanisms of H₂O₂‐mediated toxicity, we employed genetically encoded probes that enable subcellular compartment‐specific measurement of H₂O₂, the glutathione redox potential (E _GSH), and Prx1 oxidation, together with biochemical assessment of cysteine redox states, transcriptome analyses and cell death assays. We found that the matrix glutathione pool is significantly more sensitive to H₂O₂‐induced oxidation than the cytosolic glutathione pool. However, we found that H₂O₂‐induced matrix glutathione oxidation is completely dependent upon the presence of Prx1. Deletion of PRX1 eliminated H₂O₂‐induced oxidation of matrix glutathione and elicited a transcriptional response that increased levels of the cytosolic catalase, Ctt1, showing that cells can recognize, and respond to, impaired matrix redox homeostasis. The loss of glutathione oxidation in the matrix and the improved Ctt1‐dependent H₂O₂‐handling capacity of the cytosol synergistically rendered cells more resistant to an acute H₂O₂ treatment. We subsequently generated a range of matrix‐targeted thiol peroxidases and mutant variants thereof, with differing abilities to transfer oxidative equivalents from H₂O₂ to glutathione. By replacing endogenous Prx1 with these peroxiredoxin variants, we revealed a striking correlation between matrix glutathione oxidation and cell death. In wild‐type cells, we found that the degree of cell death is limited by hyperoxidation‐based inactivation of Prx1 at high H₂O₂ levels, which restricts oxidation of the matrix glutathione pool. In summary, Prx1 hyperoxidation allows cells to strike a fine balance between H₂O₂ removal and limitation of mitochondrial glutathione oxidation, which is strongly predictive of cell death.

Results

Exogenous H₂O₂ elicits compartment‐specific E _GSH and H₂O₂ responses

Little is known about the dynamic handling of H₂O₂ in the mitochondrial matrix of intact cells. We thus sought to characterize H₂O₂ flux inside the matrix and assess the impact of increased H₂O₂ on matrix reductive systems, particularly the glutathione pool. To this end, we made use of the genetically encoded fluorescent probes, roGFP2‐Tsa2∆C_R and Grx1‐roGFP2, which allow the real‐time monitoring of the basal H₂O₂ level and E _GSH, respectively, in specific subcellular compartments (Gutscher et al, 2008; Morgan et al, 2016; Fig 1A and B). Both probes comprise a redox‐sensitive green fluorescent protein (roGFP2; Hanson et al, 2004) genetically fused with a specific redox enzyme. For roGFP2‐Tsa2∆C_R, this is the Saccharomyces cerevisiae cytosolic typical 2‐Cys peroxiredoxin, Tsa2, from which the resolving cysteine has been removed to increase the sensitivity of the probe to H₂O₂. In the case of Grx1‐roGFP2, it is the human glutaredoxin, Grx1. The roGFP2‐Tsa2ΔC_R probe responds directly to H₂O₂, with the Tsa2ΔC_R moiety serving to efficiently transfer oxidative equivalents from H₂O₂ to roGFP2. This probe is predominantly reduced by endogenous GSH/glutaredoxins, which directly reduce the roGFP2 moiety. RoGFP2‐Tsa2ΔC_R oxidation is therefore determined by rapid H₂O₂‐driven oxidation and much slower GSH/glutaredoxin‐driven reduction (Morgan et al, 2016; Roma et al, 2018). Conversely, Grx1‐roGFP2 will predominantly only respond indirectly to H₂O₂, via H₂O₂‐induced glutathione disulfide (GSSG) production and the concomitant increase in E _GSH. The readout of roGFP‐based probes is commonly presented as degree of probe oxidation to allow comparison between different experiments (OxD; for calculation, see Materials and Methods and Gutscher et al, 2008). OxDs of 0 and 1 therefore indicate fully reduced and fully oxidized roGFP2 moieties, respectively (Appendix Fig S1A; for more detailed discussion of the probe mechanisms, see, e.g., Roma et al, 2018).

Figure 1. Matrix E_GSH is more sensitive to H₂O₂ than its cytosolic counterpart.

• A
  A scheme depicting the H₂O₂ sensing mechanism of the peroxiredoxin‐based H₂O₂ sensor roGFP2‐Tsa2ΔC_R.
• B
  A scheme depicting the mechanism of the E _GSH sensor, Grx1‐roGFP2.
• C, D
  The response of cytosolic and mitochondrial matrix‐localized roGFP2‐Tsa2ΔC_R (C) and Grx1‐roGFP2 (D) probes to bolus of exogenous H₂O₂ at the indicated concentrations. Probes were expressed in wild‐type BY4742 yeast cells grown to exponential phase in SGal (−Leu) medium.
• E, F
  The response of cytosolic and mitochondrial matrix‐localized roGFP2‐Tsa2ΔC_R (E) and Grx1‐roGFP2 (F) probes to 10 μM complex antimycin A (a complex III inhibitor). Probes were expressed in wild‐type BY4742 yeast cells grown to exponential phase in SGal (−Leu) medium.
• G
  The response of mitochondrial matrix‐localized (left panel) and cytosolic (right panel) roGFP2‐Tsa2ΔC_R probes, expressed in BY4742 wild‐type and Δtsa1Δtsa2 cells, to the addition of exogenous H₂O₂ at the indicated concentrations. Cells were grown to exponential phase in SGal (−Leu) medium. The lighter colored curves are controls, showing the probe response upon the addition of water.
• H
  The response of mitochondrial matrix‐localized (left panel) and cytosolic (right panel) roGFP2‐Tsa2ΔC_R probes, expressed in wild‐type and Δtsa1Δtsa2 cells, to the addition of 10 μM antimycin A. Cells were grown to exponential phase in SGal (−Leu) medium. The lighter colored curves are controls showing the probe responses upon addition of 0.1% (v/v) ethanol.
• I
  The response of a mitochondrial matrix‐localized roGFP2‐Tsa2ΔC_R probe, expressed in wild‐type and Δpor1 cells, to bolus at exogenous H₂O₂ at the indicated concentrations. Cells were grown to exponential phase in SGal (−Leu) medium. Lighter colored curves are controls showing the probe response upon the addition of water.
• J
  The response of a mitochondrial matrix‐localized roGFP2‐Tsa2ΔC_R probe, expressed in wild‐type and Δpor1 cells, to the addition of 10 μM antimycin A. Cells were grown to exponential phase in SGal (−Leu) medium. Lighter colored curves are controls showing the probe response upon the addition of 0.1% (v/v) ethanol.
• K
  Model. Cytosolic peroxiredoxins protect the mitochondrial matrix from internal (mitochondrial) and external H₂O₂.

Data information: In all panels, OxD refers to the degree of sensor oxidation. Error bars represent the standard deviation (n = 3 biological replicates, with cells obtained from three independent cultures for every strain and probe combinations. For each biological replicate, three technical replicates were performed). Significance was assessed with Student's 2‐tailed, unpaired, t‐test. *P < 0.05; **P < 0.01; ***P < 0.001; and ****P < 0.0001.

We targeted each probe to either the cytosol or the mitochondrial matrix and monitored the probe response to the addition of exogenous H₂O₂ using a fluorescence plate reader‐based system (Appendix Fig S1B, and Fig 1C and D). We observed that the matrix roGFP2‐Tsa2∆C_R probe exhibited a significantly smaller response to exogenous H₂O₂, applied at initial concentrations of 0.1 and 1 mM, than the cytosolic roGFP2‐Tsa2∆C_R probe (Fig 1C). In both subcellular compartments, the OxD of the roGFP2‐Tsa2∆C_R probe decreased over time in the absence of H₂O₂, an observation explained by the depletion of oxygen in our plate reader‐based assay (Morgan et al, 2016). In contrast, upon addition of exogenous H₂O₂, the matrix‐targeted Grx1‐roGFP2 probe exhibited a significantly larger response than the cytosol‐localized Grx1‐roGFP2 (Fig 1D). In summary, matrix E _GSH is apparently more sensitive to perturbation by exogenous H₂O₂, even though the roGFP2‐Tsa2∆C_R probe reports less H₂O₂ in the matrix than in the cytosol. We therefore sought to identify the cause(s) of these compartment‐specific differences.

Cytosolic enzymes protect the matrix against both exogenous and mitochondria‐derived H₂O₂

The smaller roGFP2‐Tsa2∆C_R response in the matrix compared to the cytosol could be explained by either (i) efficient cytosolic scavenging systems as well as the two mitochondrial membranes limiting the amount of H₂O₂ that reaches the matrix or (ii) by more efficient scavenging of H₂O₂ in the matrix limiting the amount of H₂O₂ available to react with the probe.

An initial insight into these two possibilities came from monitoring roGFP2‐Tsa2∆C_R and Grx1‐roGFP2 probe responses to antimycin A treatment. Antimycin A is an inhibitor of respiratory chain complex III that results in release of superoxide anions on the IMS side of the inner mitochondrial membrane. Superoxide anions will be rapidly dismutated, both enzymatically and spontaneously, to H₂O₂. Antimycin A treatment led to a larger roGFP2‐Tsa2∆C_R response in the matrix compared to the cytosol, i.e., the opposite of the situation following treatment with exogenous H₂O₂ (Fig 1E). Antimycin A treatment also induced a small deflection of E _GSH in both the matrix and the cytosol. The response in the matrix appeared to be slightly larger than in the cytosol, although the overall response was very limited in both compartments (Fig 1F). Currently, we do not understand why the comparatively strong roGFP2‐Tsa2∆C_R probe response upon antimycin A treatment did not result in a respective E _GSH deflection. A possible hint might be the antimycin A‐specific response dynamics. Compared to addition of external H₂O₂, antimycin A induced a comparatively late oxidation of the roGFP2‐Tsa2∆C_R probe without recovery. In conclusion, the observation of opposing compartment‐specific responses to antimycin A and exogenous H₂O₂ treatment indicates that the subcellular localization of H₂O₂ production/influx is an important determinant of subcellular compartment‐specific H₂O₂ levels. Likely, cellular H₂O₂ scavenging enzymes significantly limit the (sub)cellular diffusion of H₂O₂ leading to the generation of steep intracellular H₂O₂ gradients (Winterbourn, 2008; Lim et al, 2015; Travasso et al, 2017).

To further test this hypothesis, we monitored the response of cytosolic and matrix‐localized roGFP2‐Tsa2∆C_R probes to exogenous peroxide in either wild‐type cells or cells deleted for the genes encoding the two cytosolic typical 2‐Cys peroxiredoxins, Tsa1 and Tsa2. Tsa1, in particular, is a highly abundant protein and thought to be an important cytosolic scavenger of H₂O₂ (Iraqui et al, 2009). In ∆tsa1∆tsa2 cells, we saw that cytosolic and matrix roGFP2‐Tsa2∆C_R responses (although starting from a different initial steady state) to exogenous H₂O₂ were much more similar than in wild‐type cells (Fig 1G). These data thus further support the hypothesis that cytosolic H₂O₂ scavenging enzymes, including Tsa1 and Tsa2, limit the amount of exogenous H₂O₂ that can diffuse through the cytosol to ultimately reach the mitochondrial matrix. Interestingly, we also observed that Tsa1 and Tsa2 are important for the detoxification of mitochondria‐derived H₂O₂, as a matrix roGFP2‐Tsa2∆C_R probe in ∆tsa1∆tsa2 cells responded more rapidly to antimycin A treatment than in wild‐type cells (Fig 1H). Thus, release of H₂O₂ to the cytosol likely also constitutes a mitochondrial H₂O₂ detoxification pathway.

We next tested whether transfer over the mitochondrial membranes contributes to a decreased roGFP2‐Tsa2ΔC_R response in the matrix compared to the cytosol. In other systems, it has been demonstrated that the velocity of H₂O₂ transfer over membranes is increased by the presence of specific transporters, e.g., aquaporin 8 in the NADPH oxidase 2‐dependent signaling cascade (Bertolotti et al, 2016). The outer membrane of mitochondria (OMM) contains porins (in yeast Por1 and the less expressed Por2) that have been shown to facilitate small molecule transport (Kmita et al, 2004; Kojer et al, 2012). Indeed, we found that the response to 1 mM exogenous H₂O₂ of a matrix roGFP2‐Tsa2ΔC_R in a ∆por1 strain was decreased compared to a wild‐type strain (Fig 1I). Conversely, following antimycin A treatment, the matrix roGFP2‐Tsa2ΔC_R probe responded more readily in ∆por1 cells than in wild‐type cells, supporting the hypothesis that Por1 deletion decreases H₂O₂ transport to the cytosol (Fig 1J). In summary, Por1 appears to facilitate the bi‐directional transport of H₂O₂ across the OMM. Collectively, our data indicate that cytosolic peroxiredoxins and porins in the OMM contribute a major line of defense for mitochondria from external H₂O₂ and support the efficient removal of mitochondria‐derived H₂O₂ (Fig 1K).

Ctt1 upregulation constitutes an “adaptive response” in matrix redox enzyme mutants

We next investigated the contribution of matrix H₂O₂ scavenging enzymes toward regulation of matrix H₂O₂ level. The 1‐Cys peroxiredoxin, Prx1, is likely the most important H₂O₂ scavenger in the yeast mitochondrial matrix. However, surprisingly, when we deleted this enzyme, the matrix roGFP2‐Tsa2∆C_R response to exogenous H₂O₂ was decreased (Fig 2A). This is counterintuitive for an enzyme that has been implicated in efficient matrix H₂O₂ handling. Conversely, when we assessed the matrix roGFP2‐Tsa2∆C_R response toward antimycin A treatment, the differences in response between wild‐type and ∆prx1 cells were lost (Appendix Fig S2A). Collectively, these results suggest that defective matrix H₂O₂ handling might result in compensatory responses in the cytosol. To test this hypothesis, we monitored cytosolic H₂O₂ handling in strains lacking the matrix redox enzymes Prx1 and Trx3, which have both previously been linked to efficient H₂O₂ handling in the matrix (Pedrajas et al, 2000; Greetham & Grant, 2009; Gostimskaya & Grant, 2016). Interestingly, in both deletion strains cytosolic roGFP2‐Tsa2ΔC_R responses were attenuated compared to the wild‐type cells (Fig 2B and C). Complementation of gene loss by expressing the matrix redox enzymes from a plasmid saw cytosolic roGFP2‐Tsa2ΔC_R responses revert to a wild‐type‐like situation (Fig 2D).

Figure 2. Deletion of mitochondrial redox enzymes activates a cytosolic adaptive response.

• A
  The response of a mitochondrial matrix‐localized roGFP2‐Tsa2ΔC_R probe, expressed in wild‐type and Δprx1, to the addition of exogenous H₂O₂ at the indicated concentrations. Cells were grown in SGal (−Leu) medium and harvested at early exponential phase. Lighter colored curves are controls showing the probe response to the addition of water.
• B, C
  The response of a cytosolic roGFP2‐Tsa2ΔC_R probe, expressed in wild‐type, Δtrx3, and Δprx1 cells, to the addition of 0.1 mM (B) or 1 mM (C) exogenous H₂O₂. Cells were grown in SGal (−Leu) medium and harvested at early exponential phase.
• D
  The response of a cytosolic roGFP2‐Tsa2ΔC_R probe, in Δtrx3 cells transformed with a Trx3 plasmid and Δprx1 cells transformed with a Prx1 plasmid, to the addition of 0.1 mM exogenous H₂O₂. Cells were grown in SGal medium lacking the appropriate amino acids for plasmid selection and harvested at early exponential phase. The light gray curve in both panels represents the wild‐type control, treated with the same concentration of exogenous H₂O₂.
• E
  The profile of mRNA expression of the yeast redox or redox‐related enzymes in Δprx1 + empty vector cells, compared to Δprx1 + Prx1‐WT cells, both grown to an early exponential phase in SGal (−Ura) medium (n = 3 biological replicates, with cells obtained from three independent cultures). Cutoff was set a log₂(fold change (FC)) ± 0.32. Raw data are presented in Dataset EV1.
• F
  The response of a cytosolic roGFP2‐Tsa2ΔC_R probe, in Δprx1 cells or Δctt1Δprx1 cells transformed with either an empty plasmid or a Prx1‐WT plasmid to the addition of 0.1 mM exogenous H₂O₂. Cells were grown in SGal medium lacking the appropriate amino acids for selection and harvested in early exponential phase.
• G
  Model. Levels of the cytosolic catalase Ctt1 increase when activity of mitochondrial redox enzymes is impaired, leading to an increased cytosolic capacity for H₂O₂ removal.

Data information: In all panels, OxD refers to the degree of sensor oxidation. Error bars represent the standard deviation (n = 3 biological replicates, with cells obtained from three independent cultures for every strain and probe combinations. For each biological replicate, three technical replicates were performed). Significance was assessed with Student's 2‐tailed, unpaired, t‐test. **P < 0.01; ***P < 0.001; and ****P < 0.0001.

To understand this apparent cytosolic adaptation, we performed RNA‐Seq analysis on ∆prx1 cells transformed either with an empty p416TEF plasmid (∆prx1 + empty) or with a pEPT‐PRX1 plasmid (∆prx1 + Prx1‐WT) for the expression of wild‐type Prx1 (Fig 2E, and Appendix Fig S2B and C). We found that the transcript encoding the cytosolic catalase, Ctt1, was significantly upregulated, while transcripts encoding other redox enzymes were not significantly enriched above threshold (Fig 2E and Dataset EV1). In line with a role of Ctt1 in the cytosolic “adaptive response”, the deletion of CTT1 in a Δprx1 background (Δctt1Δprx1) ablated the decreased response of cytosolic roGFP2‐Tsa2ΔC_R that was observed in ∆prx1 cells (Fig 2F). In line with the loss of the adaptive response, ∆prx1∆ctt1 cells exhibit a prolonged lag phase compared to either wild‐type, ∆prx1, or ∆ctt1 cells during growth under chronic H₂O₂ stress (Appendix Fig S2D). Increased Ctt1 levels appear to be an “add‐on” response as deletion of PRX1 in ∆tsa1∆tsa2 cells only conferred a small, although significantly, improved cytosolic H₂O₂ handling (Appendix Fig S2E). Thus, not only cytosolic redox enzymes protect the matrix from external H₂O₂ under “unperturbed” conditions, but also there appears to be a cytosol‐based “adaptive response” if matrix redox enzyme systems are impaired. This Ctt1‐dependent, NADPH‐independent, response decreases the amount of exogenous H₂O₂ that reaches the mitochondrial matrix in cells with compromised matrix H₂O₂ detoxification systems (Fig 2G).

Glutathione reductase activity is limiting in the matrix

The experiments described above indicate that cytosolic redox enzymes significantly limit the amount of exogenous H₂O₂ that reaches the mitochondrial matrix. Nevertheless, matrix E _GSH is still more responsive to exogenous H₂O₂ than cytosolic E _GSH despite the lower concentration of H₂O₂ that reaches the matrix. We therefore wanted to gain a deeper understanding of the mechanistic basis of this difference. We reasoned that, in the matrix, either GSSG might be less efficiently reduced or H₂O₂ might more efficiently trigger glutathione oxidation in comparison with the situation in the cytosol. First, we assessed whether glutathione reductase (Glr1) levels are limiting in the matrix. Glr1 is dually localized to the cytosol and matrix. The two Glr1 variants are encoded by one gene and one mRNA that is translated from two different start codons, with the longer form encompassing a matrix‐targeting sequence (Outten & Culotta, 2004). Deleting GLR1 resulted in a higher steady‐state Grx1‐roGFP2 oxidation in both compartments and a much larger response to exogenous H₂O₂ compared to wild‐type control cells, confirming previous findings (Kojer et al, 2012; Morgan et al, 2013; Appendix Fig S3A). Notably, complementation of the Δglr1 strain with Glr1 expressed from a plasmid, under the control of a strong constitutive TEF promoter (Δglr1 + Glr1), led to a decreased matrix E _GSH response compared to that observed in wild‐type cells. However, maintenance of cytosolic E _GSH did not benefit to the same extent from Glr1 overexpression. Collectively, these data indicate that Glr1 is limiting for GSSG reduction in the matrix (Appendix Fig S3B). Nonetheless, even with Glr1 overexpression, matrix E _GSH still responds more readily to exogenous H₂O₂ than cytosolic E _GSH. We thus next asked whether H₂O₂ also more efficiently elicits GSSG production in the matrix than in the cytosol and if so, what enzymes are important for catalyzing the glutathione‐mediated reduction of H₂O₂.

Prx1 catalyzes the glutathione‐dependent reduction of H₂O₂ in the matrix

To gain further insight into how H₂O₂ affects E _GSH in the matrix, we monitored the response of matrix E _GSH to exogenous H₂O₂ in ∆trx3, ∆trr2, and ∆prx1 strains, and compared it with wild‐type, ∆tsa1∆tsa2, and ∆por1 cells (Fig 3A). As expected, ∆tsa1∆tsa2 cells exhibited an increased response of matrix Grx1‐roGFP2 compared to wild‐type because more H₂O₂ reaches the matrix in these cells (Fig 1G). Conversely, and in line with the roGFP2‐Tsa2∆C_R data (Fig 1I), ∆por1 cells exhibited a decreased response of the E _GSH sensor. Intriguingly, in the ∆trx3, ∆trr2, and ∆prx1 strains we observed either no E _GSH response (∆prx1) or an E _GSH response that was greatly decreased compared to wild‐type cells (∆trx3, ∆trr2), following the addition of exogenous H₂O₂ at an initial concentration of 1 mM (Fig 3A). To assess whether the cytosolic adaptive response described above (Fig 2) might explain the decreased matrix E _GSH response, we applied exogenous H₂O₂ at the higher initial concentration of 2.5 mM. While the matrix E _GSH response in the ∆trx3 and ∆trr2 strains increased to almost the same level as that observed in wild‐type cells, we still observed no matrix E _GSH response in ∆prx1 cells (Fig 3B). Importantly, at 2.5 mM exogenous H₂O₂, the matrix roGFP2‐Tsa2∆C_R response in ∆prx1 cells was clearly increased compared to the respective response in the wild‐type to 1 mM exogenous H₂O₂ (Fig 3C). This indicates that despite the increased H₂O₂‐handling capacity of the cytosol due to increased Ctt1 levels, Prx1 is crucial in mediating the E _GSH response to H₂O₂. Consistent with the enzymatic activity of Prx1 being required for the H₂O₂‐induced oxidation of glutathione, we observed no E _GSH response in ∆prx1 cells transformed with a plasmid encoding a Prx1 peroxidatic cysteine mutant, Prx1‐C91A (Fig 3D). However, transformation of a plasmid encoding a wild‐type Prx1 (Prx1‐WT) into ∆prx1 cells fully restored the E _GSH response (Fig 3D). In summary, the most likely explanation for our data is that glutathione is involved in the reduction of Prx1 following its reaction with H₂O₂, although it is not possible to say whether glutathione is directly or indirectly reducing Prx1 (Greetham & Grant, 2009; Pedrajas et al, 2016, 2010; Fig 3E).

Figure 3. Prx1 efficiently transfers oxidative equivalents from H₂O₂ to glutathione.

1. The response of a mitochondrial matrix‐localized Grx1‐roGFP2 sensor expressed in wild‐type, Δtsa1Δtsa2, Δpor1, Δtrr2, Δtrx3, or Δprx1 cells to a bolus of exogenous H₂O₂ at the indicated concentrations. Cells were grown in SGal (−Leu) medium and harvested at early exponential phase.
2. Left panel: The response of a mitochondrial matrix‐localized Grx1‐roGFP2 sensor expressed in wild‐type, Δtrr2, Δtrx3, or Δprx1 cells to a bolus of 2.5 mM H₂O₂. Cells were grown in SGal (−Leu) medium and harvested at early exponential phase. Right panel: Scheme depicting Prx1 and its peroxidatic cysteine at position 91.
3. The response of a mitochondrial matrix‐localized roGFP2‐Tsa2ΔC_R probe to exogenous 2.5 mM H₂O₂. The probe was expressed in wild‐type cells transformed with an empty vector or in ∆prx1 cells transformed with either an empty vector, a vector encoding wild‐type Prx1, or a vector encoding Prx1‐C91A, grown in SGal (−Leu, −Ura).
4. The response to of a mitochondrial matrix‐localized Grx1‐roGFP2 sensor expressed in either wild‐type cells transformed with an empty vector or in ∆prx1 cells transformed with either an empty vector, a Prx1‐WT, or a Prx1‐C91A plasmid to exogenous H₂O₂ at a concentration of 2.5 mM. Cells were grown to SGal (−Leu, −Ura) medium and harvested at early exponential phase.
5. Model illustrating the role of Prx1 in coupling H₂O₂ to oxidation of the mitochondrial glutathione pool.

Data information: In all panels, OxD refers to the degree of sensor oxidation. Error bars represent the standard deviation (n = 3 biological replicates, with cells obtained from three independent cultures for every strain and probe combinations. For each biological replicate, three technical replicates were performed). Significance was assessed with Student's 2‐tailed, unpaired, t‐test. ***P < 0.001 and ****P < 0.0001.

Prx1 hyperoxidation protects matrix glutathione from H₂O₂‐induced oxidation

Given that Prx1 appears to efficiently catalyze the transfer of oxidative equivalents from H₂O₂ to glutathione, we next asked about the consequences of acute H₂O₂ challenges for the mitochondrial matrix glutathione pool. To this end, we developed an acute stress‐washout assay. In this experiment, Δprx1 cells transformed with a plasmid for the expression of wild‐type Prx1 (Δprx1 + Prx1‐WT) and expressing matrix Grx1‐roGFP2 were incubated with increasing amounts of H₂O₂. The H₂O₂ was then removed, and in a subsequent readout experiment, the response of the matrix Grx1‐roGFP2 probe toward the addition of 1 mM exogenous H₂O₂ was monitored (Fig 4A and B). Counterintuitively, we observed a strong negative correlation between the concentration of H₂O₂ used in the pre‐treatment and the response of the matrix Grx1‐roGFP2 to the subsequent bolus of 1 mM exogenous H₂O₂ (Fig 4B; for data with wild‐type cells, see Appendix Fig S4A). When we repeated the experiment with ∆prx1 cells transformed with a plasmid encoding an enzymatically inactive Prx1‐C91A mutant, we observed no response to the 1 mM bolus of exogenous H₂O₂, irrespective of the concentration of H₂O₂ used in the initial pre‐treatment (Fig 4C). To test whether this attenuation of the E _GSH response after preceding H₂O₂ treatment could also be caused by matrix‐originating H₂O₂, we employed a matrix‐targeted D‐amino acid oxidase (DAO; Matlashov et al, 2014). Upon addition of d‐alanine but not l‐alanine, this enzyme locally produces H₂O₂. When we compared the E _GSH response toward 1 mM H₂O₂, we found that upon pre‐treatment with d‐alanine but not l‐alanine, the E _GSH response was indeed attenuated (Fig 4D and Appendix Fig S4B). Thus, both pre‐treatment of wild‐type cells with exogenous H₂O₂ and matrix‐specific generation of H₂O₂ induce a ∆prx1‐like mitochondrial matrix E _GSH response upon subsequent exogenous H₂O₂ treatment.

Figure 4. The active‐site cysteine of Prx1 can become hyperoxidized.

1. Experimental layout of the “acute stress‐washout” assay performed in (B) and in Appendix Fig S4A. Cells were pre‐treated with either 0, 1, 5, 10, 25, or 50 mM H₂O₂ for 10 mins after which H₂O₂ was removed, and cells were resuspended in fresh buffer. The response of a mitochondrial matrix‐localized Grx1‐roGFP2 probe to an exogenous bolus of 1 mM H₂O₂ was then measured.
2. “Acute stress‐washout” assay. Δprx1 cells containing a Prx1‐WT plasmid and expressing a mitochondrial matrix‐localized Grx1‐roGFP2 probe were grown to exponential phase in SGal (−Leu, −Ura) medium harvested and treated with either 0, 1, 5, 10, 25, or 50 mM H₂O₂ for 10 min. After removal of H₂O₂ and resuspension in fresh buffer, the response of a mitochondrial matrix‐localized Grx1‐roGFP2 probe to the addition of 1 mM H₂O₂ was measured. The middle panel is an enlargement of the left panel. The graph in the right panel shows the steady‐state Grx1‐roGFP2 oxidation following H₂O₂ pre‐treatment as well as the maximum Grx1‐roGFP2 oxidation in response to the subsequent second H₂O₂ treatment.
3. The same experiment as in (B) performed with Δprx1 cells transformed with a Prx1‐C91A plasmid.
4. ∆prx1 cells transformed with a vector encoding wild‐type Prx1, transformed with a vector encoding the matrix‐targeted D‐amino acid oxidase (su9‐DAO), and expressing a mitochondrial matrix‐localized Grx1‐roGFP2 were grown to exponential phase in SGal (−Leu, −Ura, −His) and then pre‐treated with 0.15 M d‐alanine for 0, 0.5, and 1.5 h. Afterward, the cells were washed and the response of Grx1‐roGFP2 to the addition of 1 mM H₂O₂ was measured. A redox shift assay to establish the redox state of Prx1 in these samples is present in Appendix Fig S4E.
5. Left panel: the response of mitochondrial matrix‐localized roGFP2‐Prx1 and roGFP2 probes expressed in wild‐type cells to exogenous H₂O₂ at the indicated concentrations. Cells were grown in SGal (−Leu) medium and harvested at early exponential phase. Right panel: model illustrating that hyperoxidation of the Prx1 moiety in the roGFP2‐Prx1 sensor leads to a roGFP2‐like behavior.
6. Scheme illustrating the workflow of the experiment preformed in (G, H). Dependent on the initial state of the single cysteine of Prx1, different outcomes of the modification approach and consequently different behavior on SDS–PAGE can be expected. In sample (1) “unmodified”, no shift expected, since no mmPEG₂₄ modification occurs. In sample (2) “TCEP‐reduced”, treatment with TCEP reduces only disulfide bonds and sulfenic acids to thiols and these are modified with mmPEG₂₄—hyperoxidized cysteine residues are not modifiable. In lane 3, “steady state”, only reduced thiols are directly modified with mmPEG₂₄—disulfide bonds, sulfenic acids, and hyperoxidized cysteine residues are not modifiable.
7. Left panel: a scheme illustrating the principle of the thiol modification‐based “redox shift” assay, for which reduced thiols (but not thiols oxidized to a disulfide bond or a sulfinic or sulfonic acid) can be irreversibly modified by the alkylating agent mmPEG₂₄ leading to a ˜ 2.4 kDa increase in mass and thus decreased mobility on an SDS–PAGE gel. Right panel: redox shift assay to assess the oxidation state of the Prx1 catalytic cysteine residue performed in Δprx1 + Prx1‐WT, wild‐type + empty vector, and Δprx1 + Prx1‐C91A.
8. Redox shift assays of Prx1. Δprx1 cells transformed with a plasmid encoding wild‐type Prx1 were grown to early exponential phase in SGal (−Ura) medium. Cells were subsequently treated with the indicated concentrations of H₂O₂ for 10 min. Exposure to H₂O₂ oxidizes the single cysteine of Prx1 to a state that cannot be reduced with TCEP.
9. Experimental layout of the “acute stress‐washout” assay in the presence or absence of the ribosome inhibitor cycloheximide (CHX) as performed in (J). Wild‐type cells expressing a mitochondrial matrix‐localized Grx1‐roGFP2 probe were grown to early exponential phase in SGal (−Leu). Cells were pre‐treated either with water or 50 mM H₂O₂ for 10 min and afterward washed and resuspended in fresh medium. During the next 2‐h recovery, cells were either treated with the translation inhibitor cycloheximide (CHX) or with DMSO as a vehicle control. The response of a mitochondrial matrix‐localized Grx1‐roGFP2 probe to the subsequent addition of 1 mM H₂O₂ was then measured.
10. The response of a mitochondrial matrix‐localized Grx1‐roGFP2, expressed in wild‐type cells grown, to the addition of exogenous 1 mM H₂O₂. Cells were grown in SGal (−Leu) and harvested at early exponential phase. Left panel: the pre‐treatment efficiently inhibits the ability of Prx1 to transfer oxidizing equivalents to E _GSH. Right panel: newly synthesized Prx1 is required to recover the transfer of oxidizing equivalents to E _GSH in the mitochondrial matrix.
11. Hyperoxidation of Prx1 can prevent H₂O₂‐induced oxidation of the matrix glutathione pool, and de novo synthesis is required to replace hyperoxidized Prx1.

Data information: In panels (B–E) and (J), OxD refers to the degree of sensor oxidation. Error bars represent the standard deviation (n = 3 biological replicates, with cells obtained from three independent cultures for every strain and probe combinations. For each biological replicate, three technical replicates were performed). Significance was assessed with Student's 2‐tailed, unpaired, t‐test. *P < 0.05; **P < 0.01; and ***P < 0.001.Source data are available online for this figure.

High levels of H₂O₂ can lead to hyperoxidation of peroxidatic cysteine thiol groups to sulfinic or sulfonic acids, which render the peroxiredoxin enzymatically inactive. We thus asked whether our H₂O₂ pre‐treatments lead to hyperoxidation of the Prx1 peroxidatic cysteine. We therefore employed two different assays to test whether Prx1 in the mitochondrial matrix can be hyperoxidized at the concentrations of H₂O₂ used in our acute stress assays. First, we made use of a fusion construct between Prx1 and roGFP2, roGFP2‐Prx1, which we targeted to the matrix. RoGFP2‐based fusion constructs have recently been shown to be well suited for monitoring peroxiredoxin activity and hyperoxidation in vivo (Staudacher et al, 2018). In the roGFP2‐Prx1 probe, Prx1 will directly interact with H₂O₂ and transfer oxidation onto roGFP2. Interaction of roGFP2 with matrix Grx2/glutathione will reduce roGFP2 again. We observed an increasing roGFP2‐Prx1 sensor response with increasing concentrations of exogenous H₂O₂ up to 2.5 mM (Fig 4E). However, at 5 mM exogenous H₂O₂, we observed no further increase in maximum probe oxidation and an initial more rapid recovery as compared to 2.5 mM H₂O₂ (Fig 4E). This is indicative of hyperoxidation of the Prx1 moiety (Staudacher et al, 2018). As a control, a matrix‐targeted unfused roGFP2 did not exhibit a lowered response with increasing H₂O₂ concentrations and showed a similar response to the roGFP2‐Prx1 probe at 5 mM H₂O₂, further supporting Prx1 inactivation (Fig 4E). Thus, hyperoxidation‐based inactivation of the fused peroxiredoxin results in a more reduced roGFP2 (Morgan et al, 2016; Roma et al, 2018; Staudacher et al, 2018). In line with these results, matrix Grx1‐roGFP2 responses in a ∆tsa1∆tsa2 strain became attenuated upon application of exogenous H₂O₂ at lower concentrations than those used in assays with the wild‐type strain (Appendix Fig S4C).

As a second approach to test for hyperoxidation, we used an SDS–PAGE‐based redox shift assay to monitor the redox state of the single cysteine in mature endogenous Prx1, i.e., the peroxidatic cysteine C91 (Fig 4F). We used the maleimide compound mmPEG₂₄, which upon modification of a cysteine thiol results in an increased mass of the modified protein that can be detected on SDS–PAGE (Kojer et al, 2015). Maleimides react with reduced cysteine thiols but not hyperoxidized cysteine residues or cysteine residues involved in disulfide bonds. Samples were either (i) treated with mmPEG₂₄ (steady state), (ii) treated with the reducing agent Tris(2‐carboxyethyl)phosphine (TCEP) (unmodified), or (iii) treated with both TCEP and mmPEG₂₄ (TCEP‐reduced). TCEP reduces disulfide bonds and sulfenic acids but not sulfinic or sulfonic acids (Reisz et al, 2013). Under unperturbed conditions, the cysteine of Prx1 was found to be predominantly in the thiol/thiolate form and can therefore be modified by maleimide (Fig 4G, lane 3, compare Prx1‐WT to Prx1‐C91A). Next, we assessed the Cys91 redox state following treatment of cells with H₂O₂. We observed that upon treatment with 1 mM H₂O₂, half of the Prx1 pool became unreactive toward mmPEG₂₄ (Fig 4H, lane 3). This can only be partially reverted by addition of TCEP (Fig 4H, lane 2) indicating partial hyperoxidation of Prx1 already at this H₂O₂ concentration. Upon treatment with 10 mM H₂O₂, Prx1 is rendered completely unreactive toward mmPEG₂₄ (Fig 4H, lane 3), even after subsequent treatment with TCEP, indicative of hyperoxidation of the peroxidatic cysteine. Treatment of matrix DAO‐expressing cells with d‐alanine but not l‐alanine also rendered Prx1 partially unreactive toward mmPEG₂₄ (Appendix Fig S4D and E). Thus, exposure of Prx1 to high concentrations of H₂O₂ renders its active‐site cysteine maleimide inaccessible (likely hyperoxidized) and results in a lower capacity to elicit an H₂O₂‐dependent E _GSH response.

In the cytosol, hyperoxidation in the form of a sulfinic acid (but not a sulfonic acid) can be reverted by sulfiredoxin. However, sulfiredoxin is not thought to be present in the matrix of yeast mitochondria. We thus tested whether recovery of Prx1 activity after oxidative shock relies on de novo Prx1 translation rather than reduction of Cys91. To this end, we allowed cells to recover after acute H₂O₂ challenge in the presence or absence of the translation inhibitor cycloheximide (Fig 4I and J). Under these conditions, we observed no recovery of the E _GSH response after acute H₂O₂ shock in the absence of cytosolic translation, while partial recovery of the E _GSH response was observed within 2 h in the absence of cycloheximide (Fig 4J and Appendix Fig S4F). In summary, these results strongly support the hypothesis that hyperoxidation of Prx1 prevents H₂O₂‐induced oxidation of the matrix glutathione pool. The logical next step was therefore to ask whether Prx1 hyperoxidation and the consequent protection of the mitochondrial glutathione pool have any influence on cell viability under H₂O₂ stress.

Prx1 increases cell sensitivity to acute H₂O₂ stress

We assessed the importance of Prx1 for cell viability under both chronic and acute H₂O₂ stresses. Prx1 has previously been shown to be important for maintaining cell viability under chronic oxidative stress (Greetham & Grant, 2009). We confirmed this result using halo assays (Appendix Fig S4G–I). We observed a significant detrimental effect of PRX1 deletion for growth in the continuous presence of H₂O₂ (Fig 4I, left panel), while we observed no impact of PRX1 deletion for viability of cells growing in the continuous presence of the thiol oxidant N,N,N′,N′‐tetramethylazodicarboxamide (diamide) (Fig 4I, right panel). Expression of the Zea mays aquaporin ZmPip2.5 wild‐type but not an inactive mutant, ZmPip2.5 H199K, leads to an increased zone of growth inhibition in our halo assays (Appendix Fig S4H). These results support the idea that aquaporin can mediate the transport of H₂O₂ across membranes and suggest that in laboratory yeast strains, which typically harbor inactive aquaporins (Laize et al, 2000; Sabir et al, 2017), H₂O₂ influx across the plasma membrane is limited. This may partly explain why high concentrations of H₂O₂ are frequently required to observe effects in yeast assays. We next asked whether PRX1 deletion also rendered cells more sensitive to an acute H₂O₂ stress. To this end, we determined the percentage of viable cells following 30‐min incubation with H₂O₂ at concentrations ranging from 0 to 25 mM (Appendix Fig S4J). As a control, we show that expression of the aquaporin ZmPip2.5 wild‐type but not the inactive mutant, ZmPip2.5 H199K, led to decreased viability upon acute H₂O₂ stress (Appendix Fig S4K). Surprisingly, when we compared wild‐type and ∆prx1 cells in these assays, we observed that ∆prx1 cells were not more sensitive than wild‐type cells at H₂O₂ concentrations below 10 mM, while at higher H₂O₂ concentrations, the presence of Prx1 was even found to be significantly detrimental (Appendix Fig S4L and M). Thus, Prx1 is not required for cell viability under acute H₂O₂ stress, and at higher H₂O₂ concentrations, its presence is detrimental.

The loss of Prx1 causes an adaptive response in the cytosol characterized by increased levels of Ctt1. To test whether these increased Ctt1 levels might explain the superior survival of ∆prx1 compared to wild‐type cells during acute stress, we analyzed a ∆prx1∆ctt1 strain in the presence or absence of Prx1 complementation (Appendix Fig S4N). Here, we also observed that cells lacking Prx1 performed better than cells containing Prx1 despite the absence of Ctt1 in both strains. Thus, the absence of Prx1 is specifically protective during acute H₂O₂ stress. Nonetheless, given that Prx1 hyperoxidation at high H₂O₂ concentration in wild‐type cells should effectively mimic a ∆prx1‐like state, it could be reasonably argued that at high H₂O₂ concentrations, the acute stress experiment is, in effect, comparing ∆prx1‐like cells with ∆prx1 cells. Thus, with these experiments it is not possible to determine the full extent of the protective effect of Prx1 hyperoxidation against acute H₂O₂ stress‐induced cell death.

Prx1 hyperoxidation protects against H₂O₂‐induced cell death

A rigorous assessment of the role of Prx1 hyperoxidation in acute H₂O₂ stress‐induced cell death requires the development of matrix‐targeted Prx1 variants, which are more resistant to hyperoxidation but nonetheless retain their capacity to oxidize glutathione. We were able to generate just such a Prx1 variant, a truncation mutant of Prx1, Prx1‐P233stop (Fig 5A and Appendix Fig S5A). Indeed, we observed an H₂O₂ concentration‐dependent increase in the response of a roGFP2‐Prx1‐P233stop probe, up to 5 mM exogenous H₂O₂, in contrast to the decreased response of a roGFP2‐Prx1 probe above 2.5 mM exogenous H₂O₂ (compare Figs 4E and 5B). We further tested this increased resistance of Prx1‐P233stop using our gel‐based redox shift assay to probe for hyperoxidation. This confirmed that Prx1‐P233stop was resistant to hyperoxidation after addition of up to 10 mM exogenous H₂O₂ (Fig 5C). In the same experiment, we observed hyperoxidation of wild‐type Prx1 from 1 mM exogenous H₂O₂ (Fig 5C). Importantly, by monitoring the response of a matrix‐localized Grx1‐roGFP2 probe, we observed that the Prx1‐P233stop variant was capable of transferring oxidation from H₂O₂ to glutathione to a similar extent as wild‐type Prx1 (Fig 5D, black lines). Furthermore, when we repeated this experiment after an initial pre‐treatment of our cells with 10 mM H₂O₂ for 10 min, we observed that the Grx1‐roGFP2 response in Prx1‐P233stop‐expressing cells was larger than without H₂O₂ pre‐treatment, opposite to what is observed in cells expressing wild‐type Prx1 (Fig 5D, red lines). These results are consistent with the strongly decreased hyperoxidation sensitivity of the Prx1P233stop variant, which means that it remains active following the H₂O₂ pre‐treatment. The increased Grx1‐roGFP2 response after H₂O₂ pre‐treatment is possibly due to hyperoxidation of cytosolic peroxiredoxins, meaning that more H₂O₂ reaches the mitochondrial matrix. Additionally, depletion of the matrix NADPH pool in the presence of a non‐hyperoxidizable peroxiredoxin might contribute to a stronger Grx1‐roGFP2 response. Importantly, we again observed that cells expressing functional Prx1 constructs were significantly more sensitive to acute H₂O₂ stress‐induced cell death than cells lacking Prx1 or expressing an inactive Prx1 variant. The Prx1‐P233stop variant also appeared to be more sensitive than Prx1‐WT although this difference was statistically significant only at 10 mM H₂O₂ (Fig 5E and Appendix Fig S5C).

Figure 5. Hyperoxidation of Prx1 promotes cell survival upon acute H₂O₂ stress.

1. Structure of ScPrx1 (PDB ID: 5YKJ). The positions of relevant amino acids are indicated, and the applied modifications in Prx1‐P233stop and Prx1‐C91A are shown. The C‐terminal region absent in the Prx1‐P233stop mutant is highlighted in green.
2. The response of a mitochondrial matrix‐localized roGFP2‐Prx1‐P233stop probe expressed in wild‐type cells to the indicated concentrations of exogenous H₂O₂. Cells were grown in SGal (−Leu) medium and harvested at early exponential phase. Error bars represent the standard deviation (n = 3 biological replicates, with cells obtained from three independent cultures for every strain and probe combinations. For each biological replicate, three technical replicates were performed).
3. Maleimide‐based gel shift assay to assess Prx1 cysteine redox state. Wild‐type cells transformed with an empty plasmid or ∆prx1 cells with either an empty plasmid or a plasmid encoding wild‐type Prx1, the hyperoxidation‐resistant Prx1‐P233stop, or the inactive Prx1‐C91A variants were grown to exponential phase in SGal (−Ura) medium. Cells were subsequently treated with the indicated concentrations of H₂O₂ for 10 min and then directly treated with the alkylating agent mmPEG₂₄.
4. The response of a mitochondrial matrix‐localized Grx1‐roGFP2 probe, expressed in Δprx1 cells containing a plasmid encoding either wild‐type Prx1, Prx1‐P233stop, or Prx1C91A, to the addition of 1 mM exogenous H₂O₂. Cells had been pre‐treated with either 10 mM H₂O₂ or water as a control. For these experiments, cells were grown in SGal (−Leu, −Ura) medium and harvested at early exponential phase. OxD refers to the degree of sensor oxidation. Error bars represent the standard deviation (n = 3 biological replicates, with cells obtained from three independent cultures for every strain and probe combinations. For each biological replicate, three technical replicates were performed).
5. Hydrogen peroxide “acute stress” assay. Δprx1 cells co‐transformed with empty vector or with a vector encoding either wild‐type Prx1, the P233stop, or the C91A variants were grown in SD (−Ura) and treated with the indicated concentrations of H₂O₂ for 30 min. Afterward, the cells were diluted, and a fixed volume was plated on YPD plates. The number of viable colonies was counted after 2 days growth at 30°C, here represented as a percentage relative to the 0 mM treatment. Error bars represent standard deviation (n = 5–8 biological replicates, with cells taken from independent cultures for each individual biological replicate).

Data information: In all relevant panels, significance was assessed with Student's 2‐tailed, unpaired, t‐test. *P < 0.05; **P < 0.01; and ***P < 0.001.Source data are available online for this figure.

Mitochondrial glutathione oxidation correlates with cell death under acute H₂O₂ stress

To further investigate the relationship between matrix glutathione oxidation and cell death, we turned to the peroxiredoxin, PfAOP, from the parasite Plasmodium falciparum. PfAOP is known to efficiently transfer oxidation to glutathione and has a well‐characterized mutant, L109M (Staudacher et al, 2015, 2018; Appendix Fig S6A). In vitro, L109M has an increased activity and decreased susceptibility to hyperoxidation compared to wild‐type PfAOP. Under the conditions of our assay, PfAOP‐L109M indeed more efficiently oxidized E _GSH than wild‐type PfAOP. However, like the wild‐type cells it became inactivated by an acute H₂O₂ pre‐incubation (Appendix Fig S6B). Consistent with our previous results, we observed that cells expressing PfAOP‐L109M were significantly more sensitive to acute H₂O₂ treatment‐induced cell death than cells expressing Prx1 and much more sensitive than cells with no matrix‐localized thiol peroxidase (Fig 6A and Appendix Fig S6C). PfAOP‐L109M‐expressing cells also appeared to be more sensitive to acute H₂O₂ treatment‐induced cell death than wild‐type PfAOP‐expressing cells although these differences were not statistically significant.

Figure 6. Modulating the coupling between H₂O₂ and matrix glutathione links matrix glutathione oxidation to cell death.

1. Hydrogen peroxide “acute stress” assay. Δprx1 cells co‐transformed with an empty plasmid or a plasmid encoding Prx1, Tsa1, PRDX3, PRDX5, PRDX6, PfAOP, or PfAOP‐L109M were grown in SD medium lacking the appropriate amino acids for plasmid selection and pre‐treated with 0, 1, 5, 10, and 25 mM H₂O₂ for 30 min. Afterward, the cells were diluted, and a fixed volume was plated on YPD plates. The number of viable colonies was counted after 2 days of growth at 30°C, here represented as a percentage relative to the 0 mM pre‐treatment. Error bars represent standard deviation (n = 3–8 biological replicates, with cells taken from independent cultures for each individual biological replicate). Significance was assessed with Student's 2‐tailed, unpaired, t‐test. *P < 0.05; **P < 0.01; and ***P < 0.001.
2. The efficiency of transfer of oxidation from H₂O₂ to E _GSH strongly correlates with cell death upon exposure to acute H₂O₂ stress. Data from viability and matrix Grx1‐roGFP2 responses during acute H₂O₂ stress were correlated.

Finally, we turned to targeting Tsa1, human PRDX3, and human PRDX5 (which are the two 2‐Cys peroxiredoxins residing in the matrix of human mitochondria) as well as human PRDX6 (a 1‐Cys peroxiredoxin found in the cytosol of human cells) to the mitochondrial matrix of ∆prx1 cells (Appendix Fig S6A). We found that Tsa1 only very inefficiently transferred oxidation to glutathione, consistent with thioredoxin being the preferred reductive partner for this protein (Tairum et al, 2012). A similar result was observed for PRDX3 and PRDX5 (Appendix Fig S6B). In all three cases, a slightly increased Grx1‐roGFP2 response was observed in cells pre‐treated with 10 mM H₂O₂, implying that at least a fraction of these proteins remains in a non‐hyperoxidized state (Appendix Fig S6B). In contrast, PRDX6 led to efficient oxidation of glutathione based on the matrix Grx1‐roGFP2 response. Furthermore, H₂O₂ pre‐treatment further increased this response, again likely due to hyperoxidation‐based inactivation of cytosolic peroxiredoxins (Appendix Fig S6B). Strikingly, cells expressing PRDX6 were significantly more sensitive to acute H₂O₂ stress‐induced cell death compared to cells expressing Prx1 and particularly compared to ∆prx1 cells transformed with PRDX5, Tsa1, or an empty plasmid (Fig 6A and Appendix Fig S6C). Overall, while cells expressing mitochondria‐localized Tsa1, PRDX3, PRDX5, or only containing an empty plasmid remained up to 60% viable after 30‐min treatment with 10 mM H₂O₂, cells expressing Prx1 were ~ 25% viable, while those expressing PfAOP‐L109M or PRDX6 were only ~ 10% viable.

E _GSH responses and acute stress assays indicated that PRDX6 is hyperoxidation‐resistant. To directly test this, we performed the maleimide shift assay with PRDX6 in comparison with Prx1 (Appendix Fig S6D–F). We observed that PRDX6 cysteines became inaccessible to mmPEG modification at comparatively low H₂O₂ concentrations. However, by treating with TCEP before maleimide modification, we found that a large fraction of PRDX6 oxidation was reversible and could therefore conclude that PRDX6 was not hyperoxidized, indicating that PRDX6 is strongly resistant against hyperoxidation. In conclusion, a lack of hyperoxidation sensitivity, which means that a peroxiredoxin remains active under acute H₂O₂ stress, leads to increased cell death probably due to oxidation of matrix glutathione (Fig 6B).

Mitochondrial glutathione oxidation is the predominant determinant of cell death following acute H₂O₂ stress

Is glutathione oxidation the main determinant for cell death or just a proxy for depletion of matrix NADPH levels? To answer this question, we first assessed the impact of GLR1 deletion on matrix E _GSH, with and without concomitant deletion of PRX1. In ∆glr1 cells, GSSG reduction is impaired, thereby preserving the NADPH pool (at the expense of a more oxidized E _GSH), while in ∆prx1 cells, GSH oxidation is impaired (preserving both the NADPH pool and E _GSH). In line with this, we observed a strongly increased matrix E _GSH response to exogenous H₂O₂ in ∆glr1 cells compared to wild‐type cells, which could not be rescued by expression of only the cytosolic form of Glr1. However, upon additional deletion of PRX1, to generate ∆glr1∆prx1 cells expressing only cytosolic Glr1, this increased matrix Grx1‐roGFP2 response was almost completely absent (Fig 7A). We next assessed the growth of these cells. Intriguingly, we observed that ∆glr1 cells had an extended lag phase, while the additional deletion of PRX1 (∆glr1∆prx1 cells) rescued this growth delay (Fig 7B). Lastly, we tested these cells in the acute H₂O₂ stress assay. We found that ∆glr1 cells were severely impaired in their survival after acute stress (Fig 7C, 1 and 5 mM H₂O₂). However, the additional deletion of PRX1 improved survival during acute H₂O₂ stress. Complementation of this double deletion strain with different Prx1‐WT, Prx1‐C91A, and Prx1‐P233stop variants confirmed that an increased capacity to promote glutathione oxidation during acute H₂O₂ stress resulted in decreased cell survival (Fig 7D). Collectively, these data thus support the conclusion that matrix glutathione oxidation specifically promotes cell death (Fig 7E).

Figure 7. Mitochondrial matrix GSSG accumulation drives cell death during acute H₂O₂ stress.

1. The response of a mitochondrial matrix‐localized Grx1‐roGFP2 probe to 1 mM H₂O₂ in BY4742 wild‐type cells with an empty vector, in Δglr1 cells transformed either with an empty vector or with a vector encoding wild‐type Glr1 or the cytosolic form of Glr1, where the MTS‐encoding region was removed, or in Δglr1Δprx1 cells transformed either with an empty vector or with a vector encoding the cytosolic form of Glr1. Cells were grown to exponential phase in SGal (−Leu, −Ura) medium. Error bars represent the standard deviation (n = 3 biological replicates, with cells obtained from three independent cultures for every strain and probe combinations. For each biological replicate, three technical replicates were performed).
2. Growth curve of wild‐type, Δprx1, Δglr1, and Δglr1Δprx1 cells in SD medium complemented with all amino acids (n = 3 biological replicates). Significance for the difference in the time the cultures reach 50% of their maximal OD₆₀₀ was assessed with the t‐test. Error bars (as ribbon) represent the standard deviation (n = 4 biological replicates, with cells taken from independent cultures for each individual biological replicate).
3. H₂O₂ “acute stress” assay. Wild‐type, Δprx1, Δglr1, and Δglr1Δprx1 cells pre‐grown in SGal medium complemented with all amino acids to early exponential phase. Cells were treated with H₂O₂ at the indicated concentrations for 30 min. Subsequently, cells were diluted, and a fixed volume was plated on YPD plates. The number of viable colonies was counted after 2 days growth at 30°C, here represented as a percentage relative to the 0 mM pre‐treatment. Error bars represent standard deviation (n = 3 biological replicates, with cells taken from independent cultures for each individual biological replicate). An inset with a different scale on the y‐axis is presented for the 1 and 5 mM concentration to allow better interpretation.
4. H₂O₂ “acute stress” assay. Δprx1Δglr1 cells co‐transformed with an empty plasmid or with a plasmid encoding either wild‐type Prx1, the P233stop or the C91A variants were grown in SGal (−Ura) medium to early exponential phase. Cells were treated with the indicated amounts of H₂O₂ for 30 min. Subsequently, the cells were diluted, and a fixed volume was plated on YPD plates. The number of viable colonies was counted after 2 days growth at 30°C, here represented as a percentage relative to the 0 mM treatment. Error bars represent standard deviation (n = 3 biological replicates, with cells taken from independent cultures for each individual biological replicate).
5. Model. Prx1 drives oxidation of glutathione upon exposure to hydrogen peroxide. During acute H₂O₂ stress, Prx1 becomes hyperoxidized, which effectively uncouples the glutathione pool from H₂O₂. Prx1 hyperoxidation thus helps to limit GSSG accumulation and contributes to cell survival.

Data information: Significance was assessed with Student's 2‐tailed, unpaired, t‐test. *P < 0.05; **P < 0.01; and ***P < 0.001.

Discussion

Prx1 activity leads to the oxidation of glutathione

Prx1 is a 1‐Cys peroxiredoxin in the Prx6 subfamily (Nelson et al, 2011). The physiological reductive mechanism of such peroxiredoxins remains unclear. Studies on Prx1 from S. cerevisiae have differentially concluded that: (i) GSH is part of the reductive mechanism of Prx1 but is not oxidized in the process, i.e., GSH serves as a “resolving cysteine”, forming a disulfide bond with the Prx1 peroxidatic cysteine, which is reduced by Trx3 (Pedrajas et al, 2016); (ii) GSH forms a transient mixed disulfide bond with the Prx1 peroxidatic cysteine that is subsequently reduced by Grx2, ultimately leading to GSSG formation (Pedrajas et al, 2010); (iii) GSH forms a transient mixed disulfide with the Prx1 peroxidatic cysteine that is subsequently reduced by Trr2, leading to the formation of a transient intermolecular disulfide between Prx1 and Trr2 that is reduced by GSH leading to GSSG formation (Greetham & Grant, 2009); and (iv) in vitro, the mitochondrial thioredoxin system, Trx3 and Trr2, can efficiently reduce Prx1 (Pedrajas et al, 2000).

While the study of the Prx1 reductive mechanism was not a primary objective of our study, our data unequivocally show that, inside living cells, Prx1 activity very efficiently drives GSSG formation. While it is not possible to conclude whether the involvement of GSH in Prx1 reduction is direct or indirect, our data nonetheless allow us to exclude models that suggest no involvement of GSH in Prx1 reduction. The involvement of GSH in the reduction of Prx6‐type peroxiredoxins appears not to be restricted to yeast Prx1 as human PRXD6, targeted to the yeast mitochondrial matrix, also very efficiently elicited GSSG formation upon treatment of cells with H₂O₂. This observation is interesting as previously PRDX6 was suggested to require a GST‐Pi to mediate its enzymatic activity (Manevich et al, 2004; Zhou et al, 2013). Our observations show that GST‐Pi cannot be essential for PRDX6 activity as there is no GST‐Pi in the yeast matrix, although it cannot be completely excluded that other enzymes in the yeast matrix may fulfill a similar role.

The observation that H₂O₂‐induced GSH oxidation requires Prx1 also supports the more general assertion that H₂O₂‐induced oxidation of most cellular thiols requires enzymatic catalysis. Indeed, cytosolic thioredoxin oxidation under acute H₂O₂ stress was also shown to be dependent on the presence of cytosolic peroxiredoxins (Day et al, 2012). Furthermore, it seems increasingly likely that H₂O₂‐induced oxidation of many, if not most, H₂O₂‐sensitive protein thiols requires enzyme catalysis, namely by thiol peroxidases (Delaunay et al, 2002; Bozonet et al, 2005; Jarvis et al, 2012; Sobotta et al, 2015; Stocker et al, 2018). Together, these studies underline the importance of kinetics and kinetic (un)coupling in determining the relative poise of different cellular redox species and redox couples and underline the conclusion that there is no such thing as a general cellular or subcellular “redox state”.

Mitochondrial glutathione oxidation leads to cell death

In this study, we have demonstrated that oxidation of the mitochondrial glutathione pool, driven almost exclusively by Prx1 activity, is an important determinant of cell death upon exposure to an acute H₂O₂ stress (Fig 7E). The groups of Elizabeth Veal and Chris Grant have previously reported that oxidation of cytosolic thioredoxins in fission yeast and oxidation of the mitochondrial thioredoxin in budding yeast, respectively, correlate with H₂O₂‐induced cell death (Day et al, 2012; Greetham et al, 2013). These studies are in line with our findings and suggest that depletion of cellular reductive systems is an important determinant of cell death upon an acute H₂O₂ stress and not, as might have previously been expected, an accumulation of “oxidative damage”. Nonetheless, the conclusions from these different studies raise the question of which, if any, redox couples are directly leading to cell death, i.e., is oxidation of thioredoxins or glutathione important, or are both perhaps simply markers of NADPH depletion. Furthermore, it may also be asked whether oxidation in the cytosol or mitochondrial matrix is more important for triggering cell death.

To address the above questions, we have employed several assays to allow us to distinguish between the relative importance of different matrix redox couples. First, by employing an matrix “redox engineering” approach in which we targeted different peroxiredoxins, with differing sensitivities to hyperoxidation and differing abilities to transfer oxidation to glutathione, we revealed a strong correlation between matrix glutathione oxidation and cell death upon acute H₂O₂ stress. The use of engineered peroxiredoxins (even from other species and compartments) in this study also argues against Prx1 exerting a specific intrinsic cell death signaling function. Second, to enable us to distinguish between glutathione oxidation and other possible causes of cell death, for example, NADPH depletion, we turned to strains deleted for glutathione reductase. Specifically, we observed that deletion of the matrix form of glutathione reductase leads to matrix glutathione oxidation, even under “unperturbed” conditions, and leads to a delayed entry to exponential growth. Intriguingly, we found that concomitant deletion of PRX1 in this background restored matrix E _GSH to a wild‐type‐like value and completely rescued the delayed cell growth. Furthermore, we found that Δglr1 cells fared much worse that wild‐type cells at 1 and 5 mM exogenous H₂O₂ in an acute H₂O₂ stress assay. Importantly however, the additional deletion of PRX1 at least partially rescued the increased sensitivity of a Δglr1 strain. As deletion of GLR1 would not be expected to decrease NADPH levels, these assays support the hypothesis that matrix glutathione oxidation specifically is detrimental for cell fitness. Nevertheless, it is important to emphasize that we cannot completely exclude that NADPH depletion may also be an important determinant of cell death upon acute H₂O₂ stress. Perhaps both matrix GSSG accumulation and NADPH depletion synergistically lead to cell death under many circumstances. However, our evidence does clearly support a role for matrix GSSG accumulation in driving cell death that is additional to, and independent of, any role of NADPH depletion. This conclusion is also consistent with a previous study that reported that the increased oxidant sensitivity of yeast cells lacking glutathione reductase can be ascribed to the loss of matrix glutathione reductase activity (Gostimskaya & Grant, 2016).

Is Prx1 hyperoxidation physiologically relevant?

In wild‐type cells, it was proposed that Prx1 hyperoxidation represents an important protective mechanism to mitigate against oxidation of the matrix glutathione pool in conditions of acute H₂O₂ stress. However, this leads to the question of whether peroxiredoxin oxidation is physiologically relevant, i.e., will cells ever encounter conditions severe enough to induce peroxiredoxin oxidation during their “normal” lifestyle?

It is our contention that the almost universal conservation of hyperoxidation sensitivity‐associated GGLG and YF motifs among eukaryotic typical 2‐Cys peroxiredoxins argues that there is a strong evolutionary selective pressure to evolve or maintain hyperoxidation sensitivity (Wood et al, 2003; Hall et al, 2009). Indeed, many bacterial typical 2‐Cys peroxiredoxins are robust against hyperoxidation, without any apparent detrimental impact upon catalytic efficiency (Ferrer‐Sueta et al, 2011; Perkins et al, 2014). One possible reason for maintaining hyperoxidation sensitivity is that hyperoxidized peroxiredoxins serve specific functions within the cell, for example, recruitment of chaperones to protein aggregates (Hanzen et al, 2016) or prevention of excessive oxidation of important cellular redox couples and therefore maintenance of cell viability, as shown in this study and in others (Day et al, 2012; Veal et al, 2018). Other conserved motifs were recently shown to further fine‐tune peroxiredoxin hyperoxidation sensitivity (Bolduc et al, 2018), likely serving to tailor each peroxiredoxin to its specific function(s), localization, and the typical H₂O₂ concentrations that it will encounter. In the specific case of Prx1, the possibility to generate a hyperoxidation‐insensitive variant, without any apparent loss in catalytic prowess, by removal of part of the C‐terminus (P233stop) implies that specific motifs also exist in Prx6‐family, 1‐Cys peroxiredoxins that confer sensitivity to hyperoxidation. However, the identity of such motifs was beyond the scope of this study and remains completely unclear. For Prx1, it remains to be demonstrated to what extent hyperoxidation occurs under “normal” physiological conditions in yeast cells grown in standard laboratory conditions. However, it is important to note that our data show a progressive increase in Prx1 hyperoxidation with increasing H₂O₂ concentration. We observed Prx1 hyperoxidation with as little as 1 mM exogenous H₂O₂: it is unclear what this means in terms of mitochondrial matrix H₂O₂ levels, but likely they are at most in the low micromolar range in these conditions. Thus, Prx1 hyperoxidation is not a “digital” all‐or‐nothing switch but rather an “analog” switch, implying that hyperoxidation of even a fraction of Prx1 may be beneficial for cell survival and therefore arguing strongly in favor of the physiological relevance of this mechanism. Investigation of the possible biological reasons for the wide range of peroxiredoxin hyperoxidation sensitivities will undoubtedly be an interesting area for future research.

As a final note, caution should always be applied when interpreting results obtained with laboratory yeast strains, grown in a carefully controlled environment. It is unclear what “normal physiological conditions” means for a yeast strain living in its wild environment. The impact of environmental stresses on cellular redox homeostasis, for example, heat, desiccation, UV radiation, limitation of key nutrients and exposure to xenobiotic compounds, is barely understood. Furthermore, certain potentially H₂O₂ relevant proteins, for example, aquaporins, are typically inactive in many laboratory yeast strains (Laize et al, 2000; Sabir et al, 2017). This adaptation, likely to frequent freezing and thawing, might explain an increased resistance to externally added H₂O₂. In line with this, artificial expression of aquaporins in the yeast plasma membrane (Appendix Fig S4) rendered these cells more sensitive to chronic and acute H₂O₂ stresses. It is thus tempting to speculate that hyperoxidation may be a highly relevant phenomenon for true wild‐type cells in their native environment.

Compartment‐specific NADPH‐independent systems to regulate E_GSH

E _GSH is robustly maintained in most investigated compartments (Meyer et al, 2007; Kojer et al, 2012; Morgan et al, 2013; Elbaz‐Alon et al, 2014). In the cytosol and the matrix, NADPH‐dependent glutathione reductase provides the major mechanism of E _GSH maintenance. Impairment of glutathione reductase in the cytosol is compensated by crosstalk between the glutathione and thioredoxin redox systems, and a conserved NADPH‐independent E _GSH maintenance pathway, i.e., the ABC transporter‐dependent removal of GSSG from the cytosol (Minich et al, 2006; Morgan et al, 2013). In the matrix, GSSG accumulation is prevented under conditions of acute stress by switching off Prx1 activity, which appears to be the exclusive source of GSSG upon H₂O₂ exposure. Inactivation of Prx1 is achieved by apparently irreversible hyperoxidation that can only be reversed by synthesizing and importing new Prx1. This indicates that under acute stress conditions, the prime imperative for the matrix seems to be avoiding excessive oxidation of the glutathione pool rather than the immediate removal of H₂O₂, thereby ensuring cell viability.

A cytosolic adaptive response occurs upon matrix redox enzyme perturbation

Restricting the amounts of H₂O₂, which reach the matrix by efficient cytosolic H₂O₂ removal, appears to be a further strategy of mediating compartment‐specific redox homeostasis. Under unperturbed conditions, the 2‐Cys peroxiredoxins Tsa1 and Tsa2 perform this task. By employing transcriptome analyses, we also found evidence for a cytosolic adaptive response, which is induced under conditions of impaired function of matrix redox enzymes. This adaptive response specifically involves upregulation of the catalase, Ctt1, thereby providing an additional NADPH‐independent system of H₂O₂ handling to further decrease the amount of H₂O₂ reaching the matrix. It will be extremely interesting to identify the signal that allows for a nuclear transcriptional response upon disruption of matrix redox enzymes. It is tempting to speculate that there is a specific target protein involved in the signaling mechanism whose activity is regulated by a Prx1‐dependent post‐translational redox modification.

Materials and Methods

All reagents were purchased from Sigma‐Aldrich unless otherwise stated.

Generation and growth of yeast strains

All experiments were performed in a BY4742 strain background (EUROSCARF, Frankfurt, Germany). Strains are listed in Appendix Table S1. Gene deletion strains were constructed using a PCR‐based standard homologous recombination technique (Janke et al, 2004). Gene deletions were confirmed by PCR using primers designed to anneal ~ 100 bp up‐ and downstream of the gene of interest. Yeast strains were grown as described previously (Kojer et al, 2012). Briefly, for the roGFP2‐based experiments and the redox shift experiments, the strains were grown in synthetic medium (S‐medium) lacking the appropriate amino acids for plasmid selection (given in the standard three‐letter code, e.g., S‐medium—Leu) with 2% galactose (Gal) as carbon source at 30°C. The medium used for growth of strains is given for each experiment in the Figure Legends. As an example, “SGal‐(Leu)” indicates synthetic medium with galactose as carbon source and lacking the metabolic marker amino acid leucine. For the acute stress experiments, the strains were grown in synthetic medium lacking the appropriate amino acids for plasmid selection with 2% glucose (D) or 2% galactose (Gal) as carbon source. After H₂O₂ exposure, cells were plated on rich media plates with glucose as carbon source (YPD). Growth to assess cell viability following the stress assays took place at 30°C for 2 days.

For 1 l synthetic medium (S‐medium), 1.7 g yeast nitrogen base (without amino acids) and 5 g ammonium sulfate were dissolved in water at pH 5.5. In the final medium, as carbon source either 2% glucose (D), 2% galactose (Gal), or 2% glycerol (G) was present. In addition, according to the auxotrophic selections, the following amino acids were either present or excluded: adenine (0.15 mM), lysine (0.20 mM), leucine (0.23 mM), histidine (0.10 mM), tryptophan (0.07 mM), and uracil (0.18 mM). For 1 l rich medium (yeast–peptone, YP‐medium), 10 g yeast extract and 20 g bacto‐peptone were dissolved in water at pH 5.5. In the final medium, as carbon source either 2% glucose (D), 2% galactose (Gal), or 2% glycerol (G) was present. All plates have been prepared following the recipes described above and adding agar 20 g/l.

Primers and plasmid construction

All roGFP2 sensors were constructed as previously described (Gutscher et al, 2008; Morgan et al, 2016). All roGFP2‐fusion proteins used in this study were constitutively expressed from the low copy‐number (CEN) p415 (LEU2 marker) or p416 plasmids (URA3 marker) under the control of a constitutive TEF promoter (from translation elongation factor 1α gene, TEF2, yeast). For mitochondrial matrix targeting, indicated constructs were genetically fused with the N‐terminal mitochondrial targeting sequence (MTS) from subunit 9 of the F₀‐ATPase (Su9) from Neurospora crassa (encoding amino acids 1–69).

The mitochondrial peroxiredoxin PRX1 gene was amplified by PCR including endogenous promoter and terminator from yeast genomic DNA preparations and cloned into the low copy‐number (CEN) p416 (URA3 marker) vector using standard molecular cloning procedures, removing the TEF promoter and CYC terminator from the plasmid (pEPT PRX1(WT)). The PRX1(P233stop) and the PRX1(C91A) constructs were generated from the PRX1(WT) construct by site‐directed mutagenesis.

The open reading frame of GLR1 and TSA1 gene was amplified by PCR from yeast genomic DNA and cloned into the p416 vector with TEF promoter and C‐terminal triple hemagglutinin (HA) tag. To allow mitochondrial targeting, the TSA1 gene was genetically fused with the N‐terminal Su9 MTS.

The AOP constructs from P. falciparum were constructed as previously described (Staudacher et al, 2018). They were genetically fused with the N‐terminal Su9 MTS and expressed in the low copy‐number p416 vector (URA3 marker) with TEF promoter.

The genes of Homo sapiens PRDX3, PRDX5, and PRDX6 were amplified by PCR from HEK293 cell cDNA preparations and cloned, respectively, into the low copy‐number p416 (URA3 marker), p415 (LEU2 marker), and p413 (HIS3 marker) vectors with TEF promoter and C‐terminal triple hemagglutinin (HA), hexahistidine (His6), or FLAG tags. Only the PRDX6 gene was genetically fused with the N‐terminal Su9 MTS, while for PRDX3 and PRDX5 genes, their endogenous MTS was used for mitochondrial targeting.

The constructs for the yeast‐codon‐optimized aquaporin PIP25 wild‐type and H199K variants from Z. mays were previously described (Bienert et al, 2014). They were expressed in the multicopy pRS426‐pTPIu (URA3 marker) vector with TPI promoter (from triosephosphate isomerase gene, TPI1, yeast).

The gene of D‐amino acid oxidase (DAO) from Rhodotorula gracilis (red yeast) was amplified by PCR from the “HyPer‐D‐amino acid oxidase” plasmid generated and characterized in Matlashov et al (2014). To allow mitochondrial targeting, the PCR product was genetically fused with the N‐terminal Su9 MTS and C‐terminal FLAG tag; subsequently, it was cloned into the low copy‐number p413 (HIS3 marker) vector under the control of the strong constitutive GPD promoter (from glyceraldehyde‐3‐phosphate dehydrogenase gene, TDH3/GPD, yeast). For a detailed list of primers and plasmids, see Appendix Table S2.

Antibodies

The antibody against yeast Prx1 was generated in this study. Mature Prx1 without the MTS was subcloned into pRSET‐A plasmid. Prx1 was expressed for 4 h at 30°C in Escherichia coli BL21 (DE3) cells, purified via its N‐terminal hexahistidine (His₆) tag and used for in‐house immunization of rabbits. The serum was confirmed in immunoblot analysis using the purified antigen as well as comparing yeast cells lacking Prx1 with wild‐type cells. The serum was used in a 1:1,000 dilution in 5% Milk‐TBS for Western blot detection (Milk powder 5%, NaCl 150 mM, Tris–HCl 10 mM, adjusted to pH 7.5).

The antibody against PRDX6 was purchased from Sigma (ID: P0058) and used 1:2,000 dilution in 5% milk–TBS for Western blot detection.

Fluorescence measurement of roGFP2‐sensor oxidation

Fluorescence measurements were performed with a CLARIOstar (BMG Labtech) fluorescence plate reader as described previously (Morgan et al, 2011). Fluorescence was recorded using filter optics at excitation wavelengths 410 ± 5 and 482 ± 8 nm and an emission wavelength of 530 ± 20 nm. For measurements, yeast cells were grown to mid‐log phase in synthetic medium lacking the appropriate amino acids for plasmid selection. The cells were harvested by centrifugation at 1,500 × g for 3 min at room temperature and subsequently resuspended in an isosmotic buffer (sorbitol 0.1 M, Tris–HCl pH 7.4 0.1 M, NaCl 0.1 M) to a final concentration of 1.5 OD₆₀₀ units/ml (where 1 OD₆₀₀ unit represents 1 ml of culture with an OD₆₀₀ = 1). This process was repeated one more time. Subsequently, each cell solution was transferred to a flat‐bottomed 96‐well imaging plate (BD Falcon) with 180 μl solution per well. To one well, 20 μl of the oxidant diamide was added (final concentration of 20 mM), serving as fully oxidized sensor control. To a second well, 20 μl of the reductant DTT was added (final concentration of 100 mM), serving as fully reduced sensor control. Cells not expressing the roGFP2‐based sensors, yet treated with diamide, DTT, or buffer, were used as blanks. The cells were pelleted by centrifugation at 20 × g for 5 min at room temperature and placed in the instrument, kept at 30°C. A “steady state” was measured for 10 min, then the sample cells wells were subjected to experimental treatments adding 20 μl of a 10× experimental solution (e.g., containing H₂O₂), and the response followed for 120 min. Water was used as negative control for H₂O₂, and ethanol with final concentration 0.1% was used as negative control for antimycin A treatments. The degree of sensor oxidation (OxD) was calculated as in Equation (1) (Meyer & Dick, 2010; Morgan et al, 2011):

$$\text{OxD}=\frac{R^{\mathrm{sample}}-R^{\mathrm{RED}}}{\frac{I_{482}^{\mathrm{OX}}}{I_{482}^{\mathrm{RED}}}\ast \left(R^{\mathrm{OX}}-R^{\mathrm{sample}})+(R^{\mathrm{sample}}-R^{\mathrm{RED}}\right)}\text{with}R=\frac{I_{410}}{I_{482}}$$ (1)

I _n = intensity at a given wavelength n; $I_{482}^{\mathrm{OX}}$, and $I_{482}^{\mathrm{RED}}$ = intensities at 482 nm upon complete oxidation by diamide or reduction by DTT. All experiments were performed in three independent biological replicates unless stated otherwise.

“Acute stress‐washout” experiment with genetically encoded sensors

Yeast cells were grown to mid‐log phase in synthetic medium lacking the appropriate amino acids for plasmid selection. The cells were harvested by centrifugation at 1,500 × g for 3 min at room temperature and resuspended in isosmotic buffer to a final concentration of 1.5 OD₆₀₀ units/ml. This process was repeated one more time. A bolus of water or H₂O₂ was added to bring the cell solution to the indicated final concentrations. After 10 min at 30°C, H₂O₂ was removed with two centrifugation steps at 1,500 × g for 3 min at room temperature and following resuspension in isosmotic buffer to a final concentration of 1.5 OD₆₀₀ units/ml. Subsequently, each cell solution was transferred to a flat‐bottomed 96‐well imaging plate, and the experiment followed the classical fluorescence measurement of roGFP2‐sensor oxidation.

“Acute stress‐washout” experiment with genetically encoded sensors and cytosolic translation inhibition experiment

Yeast cells were prepared as in the “Acute stress‐washout” experiment until the removal of H₂O₂ with two centrifugation steps at 1,500 × g for 3 min at room temperature. They were resuspended in the proper selective medium in the presence of either 50 μg/ml of the ribosome inhibitor cycloheximide (CHX) or dimethyl sulfoxide (DMSO) as control. After 2‐h shaking at 30°C, cells were harvested and underwent two centrifugation steps at 1,500 × g for 3 min at room temperature, followed by resuspension in isosmotic buffer to a final concentration of 1.5 OD600 units/ml. Subsequently, each cell solution was transferred to a flat‐bottomed 96‐well imaging plate, and the experiment followed the classical fluorescence measurement of roGFP2‐sensor oxidation. Controls at time zero, before the addition of cycloheximide or DMSO, but after the pre‐treatment with H₂O₂, were also measured.

Matrix DAO‐based “Acute stress‐washout” experiment with genetically encoded sensors

Yeast cells expressing Su9‐DAO were grown to mid‐log phase in synthetic medium lacking the appropriate amino acids for plasmid selection. The cells were harvested by centrifugation at 1,500 × g for 3 min at room temperature and resuspended in pre‐heated fresh medium to a final concentration of 1 OD₆₀₀ units/ml. This process was repeated one more time. A bolus of either d‐Alanine or l‐Alanine (final 0.15 M) was added to the cell suspension. The cells were kept shaking at 30°C for the indicated time. d‐/l‐Alanine was removed with two centrifugation steps at 1,500 × g for 3 min at room temperature and following resuspension in isosmotic buffer to a final concentration of 1.5 OD₆₀₀ units/ml. Subsequently, each cell solution was transferred to a flat‐bottomed 96‐well imaging plate, and the experiment followed the classical fluorescence measurement of roGFP2‐sensor oxidation.

RNA preparation, sequencing, and analysis

Yeast cells were grown to early‐log phase in synthetic medium lacking the appropriate amino acids for plasmid selection. 5 OD₆₀₀ units of cells were harvested by centrifugation at 1,500 × g for 3 min at room temperature and resuspended in isosmotic buffer, and a second centrifugation step was necessary to retrieve pellets. For the Illumina library preparations, total RNA was first extracted from collected pellets using the RiboPure™ RNA Purification Kit, yeast (Invitrogen), according to the manufacturer's protocol. RNA‐Seq libraries were prepared from total RNA using poly(A) enrichment of the mRNA (mRNA‐Seq) and later analyzed on an Illumina HiSeq 4000 with a read‐length of 2 × 75 base pairs. The sequencing data were uploaded to the Galaxy web platform (usegalaxy.org), and bioinformatics analysis was performed using the tools available on the public server (Afgan et al, 2016). Reads were aligned using the HISAT2 tool (Kim et al, 2015) and mapped using the S. cerevisiae R64‐1‐1.91 genome reference GTF file. The counting was performed using the featureCount tool. Statistical analysis was performed for three independent biological replicates, with the final differential expression between strains calculated using the package DESeq2 (Love et al, 2014). All tools were used with the default settings. The differential expression output data were represented in a volcano plot using the program OriginLab. The GO term analysis was performed using the online tool Gorilla (Eden et al, 2009).

In vivo redox state analyses of Prx1 using mmPEG₂₄‐based alkylation

Mature Prx1 without its MTS contains only one cysteine, the active‐site cysteine at position 91 (C91). To distinguish the redox states of this cysteine, a mmPEG ₂₄ ‐based alkylation was applied, yielding a shift of ~ 2.4 kDa in the migration behavior of reduced Prx1 in a Western blot analysis (Kojer et al, 2012). Briefly, yeast cells were grown to mid‐log phase in synthetic medium lacking the appropriate amino acids for plasmid selection. Cells were harvested by centrifugation at 1,500 × g for 3 min at room temperature and resuspended in isosmotic buffer to a final concentration of 1.5 OD₆₀₀ units/ml. This process was repeated. Pellets of 2 OD₆₀₀ units were resuspended in 100 μl SDS‐loading buffer containing either 10 mM mmPEG₂₄ (steady state), DMSO (unmodified), or 10 mM TCEP (maximally reduced) and boiled for 5 min at 96°C. The cells were then disrupted by vortexing with glass beads for 15 min in the dark. In the reduced sample, mmPEG₂₄ was added to a final concentration of 10 mM. After 40 min in the dark, all samples were analyzed in Western blots against Prx1.

Cells were also exposed to a pre‐treatment as explained for the “Oxidation shock” experiment. In that case, after the pre‐treatment 2 OD₆₀₀ units of cells were centrifuged at 20,000 × g for 30 s at room temperature and directly resuspended in 100 μl SDS‐loading buffer containing 10 mM mmPEG₂₄. They were boiled, disrupted, and analyzed by Western blot as described above.

To test for irreversibility and the type of inactivation occurring in Prx1 after the hydrogen peroxide pre‐treatment, 2 OD₆₀₀ units of cells were resuspended in 100 μl SDS‐loading buffer containing either DMSO (unmodified), 10 mM TCEP (TCEP reduces disulfides and sulfenic acids only), or 10 mM mmPEG₂₄ (steady state), and boiled for 5 min at 96°C. The cells were then disrupted vortexing with glass beads for 15 min in the darkness. In the TCEP‐reduced samples, mmPEG₂₄ was added to a final concentration of 10 mM. After 40 min in the darkness, all samples were analyzed in Western blots against Prx1.

To identify the redox state of Prx1 in the samples used in the Matrix DAO‐based “Acute stress‐washout” experiment, an aliquot of 2 OD₆₀₀ units of yeast for each time point was acquired at the step when d‐/l‐alanine was removed with two centrifugation steps at 1,500 × g for 3 min at room temperature and resuspended in isosmotic buffer to a final concentration of 1.5 OD₆₀₀ units/ml. The 2 OD₆₀₀ units of yeast were centrifuged at 20,000 × g for 30 s at room temperature and directly resuspended in 100 μl SDS‐loading buffer containing 10 mM mmPEG₂₄. They were boiled, disrupted, and analyzed by Western blot as described above.

“Acute stress” viability assay

Yeast cells were grown to mid‐log phase in synthetic medium lacking the appropriate amino acids (S‐medium) for plasmid selection or in YP‐medium if no plasmid was present both with either 2% glucose or 2% galactose as carbon source. 10 OD₆₀₀ units of cells were harvested by centrifugation at 1,500 × g for 3 min at room temperature and then resuspended in water to a final concentration of 10 OD₆₀₀ units/ml. Subsequently, the cell solution was split into five tubes and centrifuged at 1,500 × g for 3 min at room temperature. The five 2 OD₆₀₀ pellets were then resuspended in 1 ml of either water or H₂O₂ to the final concentration. After shaking at 30°C for 30 min, 3.5 μl of cells solution was added to 25 ml water to a final concentration of approximately 0.00025 OD₆₀₀/ml. Next, 200 μl was plated on YPD plate (ca. 500 cells), and after growth for 5 days at 30°C, the number of viable cells was determined by the counting the colonies using the software ImageJ.

“Halo” assay

Yeast cells were grown to mid‐log phase in YP‐medium with 2% glucose (YPD) as carbon source overnight and then diluted in YPD medium to a final concentration of 0.25 OD₆₀₀/ml. The culture was incubated at 30°C for 4 h and afterward diluted to a final concentration of 0.01 OD₆₀₀/ml in water. Subsequently, the suspension was used as inoculum and 100 μl spread on YP‐medium 1% agar plates with either 2% glucose (YPD), 2% galactose (YPGal), or 2% glycerol (YPG). A 6‐mm disk of cellulose loaded with 15 μl of 1 M diamide or 1 M H₂O₂ was placed at the center of the plate. From there, the reagent diffused into the plate. The plates were incubated at 30°C for 72 h, and then, the halo of growth around the disk was measured.

Statistical analysis

In all figures, error bars represent mean ± standard deviation. P‐values were determined using the t‐test.

Author contributions

GC, MD, BM, and JR designed the project, analyzed results, and wrote the article. GC, PSA, MNH, MM, TR, GPB, and EP performed the experiments and analyzed results.

Conflict of interest

The authors declare that they have no conflict of interest.

Supporting information

Appendix

Dateset EV1

Source Data for Appendix

Review Process File

Source Data for Figure 4

Source Data for Figure 5

The EMBO Journal (2019) 38: e101552