Redox requirements for ubiquitin-like urmylation of Ahp1, a 2-Cys peroxiredoxin from yeast

The yeast peroxiredoxin Ahp1, like related anti-oxidant enzymes in other species, undergoes urmylation, a lysine-directed conjugation to ubiquitin-like modifier Urm1. Ahp1 assembles into a homodimer that detoxifies peroxides via forming intersubunit disulfides between peroxidatic and resolving cysteines that are subsequently reduced by the thioredoxin system. Although urmylation coincides with oxidative stress, it is unclear how this modification happens on a molecular level and whether it affects peroxiredoxin activity. Here, we report that thioredoxin mutants decrease Ahp1 urmylation in yeast and each subunit of the oxidized Ahp1 dimer is modified by Urm1 suggesting coupling of urmylation to dimerization. Consistently, Ahp1 mutants unable to form dimers, fail to be urmylated as do mutants that lack the peroxidatic cysteine. Moreover, Ahp1 urmylation involves at least two lysine residues close to the catalytic cysteines and can be prevented in yeast cells exposed to high organic peroxide concentrations. Our results elucidate redox requirements and molecular determinants critical for Ahp1 urmylation, thus providing insights into a potential link between oxidant defense and Urm1 utilization in cells.

Accordingly, under non-reducing conditions, we detected a slower migrating HA-Urm1•Ahp1 conjugate roughly double in size (~72 kDa), which is absent from ahp1Δ cells and up-shifted (~90 kDa) upon c-Myc tagging (Fig. 1C). This conjugate corresponds to an oxidized dimer with both Ahp1 subunits urmylated and interlinked by disulfides that are sensitive to reduction by β-ME (Fig. 1C). Oxidized Ahp1 dimers detected by anti-Ahp1 EMSA lack urmylation (Fig. 1C), suggesting that attachment of HA-Urm1 (~17 kDa) blocks Ahp1 (~19 kDa) recognition and immune detection by the antiserum [39]. As unmodified Ahp1 remains detectable under our experimental conditions there seems to be an equilibrium of free and urmylated Ahp1 in vivo. In support of this notion, independent studies with TAP-tagged Urm1 show differential Ahp1 conjugates including Ahp1 disulfides, which under non-reducing conditions, have one TAP-Urm1 copy attached to each subunit (Suppl. Fig. S1). The occurrence of detectable disulfide-linked dimers modified by Urm1 implies that urmylation does not affect the redox-active thiols required for disulfide formation. Collectively, our data uncover disulfide bridged Ahp1•Urm1 conjugates that coexist with a pool of the Ahp1 peroxiredoxin that is not modified by Urm1.

Dose-dependent suppression of Ahp1 urmylation by t-BOOH
Prompted by data that exposure with the organic peroxide t-BOOH stimulates the formation of Ahp1 intersubunit disulfides [39], we next studied the impact of oxidative stress on Ahp1 urmylation in vivo. Using anti-HA and anti-Ahp1 EMSAs under reducing and non-reducing conditions, mild t-BOOH levels (0.3-0.6 mM) did not affect Ahp1 urmylation (Fig. 3). Intermediate t-BOOH doses (1.2-2.5 mM), however, progressively suppressed Ahp1 urmylation and formation of intersubunit disulfides, and highest doses (5-10 mM) eventually abolished both (Fig. 3). Thus, t-BOOH doses known to affect yeast cell growth in vivo [14] efficiently suppress Urm1 conjugation and Ahp1 disulfide formation (Fig. 3). Whether this involves t-BOOH interference with Ahp1, the thioredoxin system or Urm1-COSH, the thiocarboxylate critical for Ahp1 urmylation, is not known. In an effort to address these options, we found that lack of thioredoxin reductase Trr1 in the trr1Δ mutant counteracts the negative t-BOOH effect on urmylation seen with TRR1 wild-type cells (Fig. 4A). As a result, HA-Urm1•Ahp1 conjugates and Ahp1 disulfides reappeared and even increased in relation to the untreated trr1Δ control (Fig. 4A). Thus, inhibition of urmylation by t-BOOH apparently relies on thioredoxin function.
Our data suggest Ahp1 hyperoxidation by t-BOOH in vivo since the intersubunit disulfides that are locked in the trr1Δ mutant (Fig. 4A) ought to resist oxidation. In line with this notion, excess t-BOOH previously led to conversion of the C P thiol (Cys-62) into a sulfonate in vitro [12,41]. Hence, hyperoxidation may be ascribable, at least in part, for the inhibitory t-BOOH effects that we observe in vivo on urmylation of Ahp1 (Fig. 3). Alternatively, t-BOOH may interfere with Urm1-COSH, which is crucial for Ahp1 urmylation in vivo [19,21]. To study the latter, we exposed recombinant Urm1-COSH to t-BOOH and analyzed it by an APM-based gel retardation assay that distinguishes the starting material (Urm1-COSH) from the inactive form (Urm1-COOH) [42]. t-BOOH doses (1.2-2.5 mM) found to be effective in vivo (Fig. 3) gradually converted Urm1-COSH into mobile Urm1-COOH ( Fig. 4B). At t-BOOH doses (5-10 mM) that abolished urmylation in vivo (Fig. 3), Urm1-COOH exclusively accumulated in vitro (Fig. 4B). With the latter being unable to urmylate proteins (including Ahp1) [19,21,29], our in vitro data (Fig. 4B) suggest that the t-BOOH effect in vivo (Fig. 3) may involve inactivation of Urm1. Collectively, high organic peroxide doses appear to prevent urmylation in yeast cells through a combination of negative effects on Urm1 and Ahp1.

The dimer interface is required for peroxidase and Urm1 acceptor activity of Ahp1
To investigate redox requirements of Ahp1 for Urm1 conjugation we asked whether urmylation is linked to dimerization of Ahp1. To do so, we resorted to structural data [14,15] showing that two conserved phenylalanine residues (Phe-58, Phe-95) ( Fig. 1A and B) are located at the center of the dimer interface. When mutated (F58A; F95A; F58,95A), these were shown by native PAGE analysis to cause decreased dimerization [14]. To validate and extend the data, we estimated the molecular weights of wild-type Ahp1 and interface mutants via SEC-MALS (Fig. 5A), a technique coupling size exclusion chromatography with multiangle light scattering [43]. Despite a broad retention profile of wild-type Ahp1, the molecular weights for both distinguishable peaks via MALS analysis were in line with the value of an Ahp1 homodimer (46.7 kDa) (Fig. 5A).
In contrast, each of the interface mutants eluted at retention times comparable to wild-type Ahp1, yet as single peaks (Fig. 5A). However, all Ahp1 variants exhibited estimated molecular weights lower than that of a dimer, with most (except for F95A) approximately the value of a His-tagged Ahp1 monomer (~23.3 kDa) (Fig. 5A). In sum, our data confirm that substitution of subunit interface residues decreases the ability of Ahp1 to form dimers. Since the Ahp1 interface mutants disrupted oligomerization, we tested their peroxidase performance in a coupled activity assay with thioredoxin Trx2, thioredoxin reductase Trr1 and NADPH (Fig. 5B). Upon addition of t-BOOH, a sharp decrease in NADPH occurred for wild-type Ahp1 diagnostic for proper peroxide detoxification (Fig. 5B). In contrast, Ahp1 variants harboring single and double substitutions at the dimer interface resembled inactive enzymes lacking the crucial C P or C R thiols (C62S or C31S) (Fig. 6B).
Thus, our data show that the hydrophobic interface contributes to the ability of Ahp1 to form dimers and detoxify t-BOOH in vitro. In contrast to wild-type Ahp1, we observed by anti-HA and anti-Ahp1 EMSAs that the single (F58A; F95A) and double (F58,95A) interface mutants failed to be urmylated and lacked formation of Ahp1 intersubunit disulfides under non-reducing conditions (Fig. 5C). This shows an intact dimer interface is critical for anti-oxidant activity of Ahp1 and Urm1 conjugation. In line with this, the Phe-95 substitution alone or in tandem with the Phe-58 mutation were reported to enhance the sensitivity of yeast cells to growth inhibition by t-BOOH in vivo [14]. Together, our data demonstrate that dimerization and peroxidase activity are intimately linked with urmylation of Ahp1.

Mutagenesis of the redox-active thiol center in Ahp1 abolishes urmylation
Ahp1 was shown to be urmylated at Lys-32 close to its redox-active center (Cys-31 Cys-62) (Fig. 1B) [12,15,16,21]. Another cysteine residue (Cys-120) was assumed to be catalytic [16] before being refuted [15]. We examined whether serine substitution mutations at Cys-31, Cys-62 or Cys-120 would affect the sensitivity of an ahp1Δ null-mutant towards t-BOOH in vivo. Expression of C31S and C62S mutants failed to restore t-BOOH protection in ahp1Δ cells, while Ahp1 wild-type and the C120S mutant allowed for growth at t-BOOH doses of up to 2 mM (Suppl. Fig. S2). Yeast lacking Ahp1 and the oxidant-sensitive transcription factor Yap1 (yap1Δahp1Δ) are more sensitive to t-BOOH [39] than ahp1Δ cells. Expression of the C62S mutant in this background failed to protect against 0.9 mM t-BOOH, a dose tolerated by the C31S mutant (Fig. 6A). Hence, in the absence of Yap1, the importance of the C R and C P thiols for peroxidase activity apparently differs. Since t-BOOH tolerance was eliminated after substituting both redox-active thiols (C31,62S) in the double mutant (Fig. 6A), partial peroxidase activity seen with the C31S mutant alone depends on an active C P (Cys-62). Therefore, unlike Cys-31, Cys-62 is critical for anti-oxidant function of Ahp1. The C120S mutant, however, did not noticeably alter t-BOOH sensitivity (Fig. 6A) and combined with the C R or C P mutations (C31,120S or C62,120S), there are no additional growth defects compared to the single C31S or C62S mutants alone (Fig. 6A) (Suppl. Fig.  S2).
Our in vivo data are in agreement with in vitro Ahp1 peroxidase activity assays (Fig. 6B). Upon addition of t-BOOH, we observed a marked decrease in NADPH indicative for peroxide detoxification by wild-type Ahp1 and the C120S mutant, while peroxidase activity with C31S or C62S was negligible (Fig. 6B). In addition, we found wild-type Fig. 3. Suppression of Ahp1 urmylation and disulfide formation by t-BOOH in vivo. Shown are EMSAs under reducing (left panels) and non-reducing (right panels) conditions from strains treated with t-BOOH as indicated and expressing HA-URM1 (+) or not (−). NEMstabilized HA-Urm1 conjugation was studied by anti-HA blots (top panels) diagnostic for free HA-Urm1 and urmylated forms of Ahp1 (~36 kDa) and Ahp1 intersubunit disulfides (AID 72 kDa). anti-Ahp1 blots (middle panels) detect unmodified Ahp1 (~19 kDa) and AID (~38 kDa). anti-Cdc19 (bottom panels) served as internal standard. like urmylation levels including formation of urmylated (or non-modified) Ahp1 intersubunit disulfides in the Cys-120 mutant (Fig. 6C). Next, we asked whether C R (C31S) and/or C P (C62S) substitutions would affect Ahp1 oxidation and urmylation. As judged from anti-Ahp1 EMSA, the Cys-31 and Cys-62 substitutions alone (C31S; C62S) or in combination (C31,62S; C31,120S; C62,120S) all failed to form Ahp1 intersubunit disulfides under non-reducing conditions (Fig. 6C). This agrees with our data ( Fig. 6A and B) showing that the C R and C P thiols are key to the Ahp1 peroxidatic cycle (Fig. 1B). However, based on anti-HA EMSA, each mutation behaved different in terms of urmylation (Fig. 6C). While single C31S and double C31,120S mutants formed Urm1 conjugates under reducing and non-reducing conditions, C62S failed to do so. This indicates that disulfide formation upon oxidation is dispensable for urmylation, whereas the C P thiol (Cys-62) is essential for the conjugation reaction (Fig. 6C). Accordingly, when combined with C31S or C120S, the negative C62S effect dominates causing loss of urmylation in each double mutant (C31,62S or C62,120S) (Fig. 6C). Thus, Cys-62 is essential for urmylation even in the case of the C31S mutant, which lacks disulfide formation upon oxidation by the peroxide and cannot be recycled. This finding indicates that oxidation of Cys-62 rather than disulfide formation upon Cys-62 oxidation is critical for urmylation.

Analysis of lysine-based acceptor sites for urmylation of Ahp1
To monitor a possible link between the redox-active center in Ahp1 and lysine-directed Urm1 conjugation, we studied the impact of lysine substitutions in Ahp1 on urmylation. The proximity of Lys-32 and Lys-156 (Fig. 7A) to the catalytic thiols in the crystal structure of Ahp1 [15] prompted us to generate mutants with both replaced by arginine alone or in combination (K32R; K156R; K32,156R). We observed slightly decreased urmylation levels in the K156R mutant compared to wild-type suggesting a minor target role (Fig. 7B).
Unlike previously reported [21], urmylation in the K32R mutant was not entirely abolished, but was significantly decreased compared to wild-type or K156R cells (Fig. 7B). This indicates Lys-32 is targeted more easily by Urm1 than Lys-156, yet it is not essential to provide Ahp1 with full Urm1 acceptor activity. Strikingly, the absence of both residues in the double mutant (K32,156R) enhanced the urmylation defects of each single mutant (K32R or K156R) (Fig. 7B). As a result, Urm1 conjugation dropped to significantly low levels, albeit not as dramatic as with complete loss of urmylation seen in the peroxidasedead mutant (C62S) (Fig. 7B). This additive negative effect suggests that Lys-156 is an alternative urmylation site, particularly when Lys-32 is unavailable due to a substitution mutation. Moreover, based on low residual urmylation left in the double mutant (K32,156R) the existence of other Urm1 target sites has to be assumed.
Hence, we consider lysine-directed Urm1 conjugation to Ahp1 may be promiscuous and less specific to Lys-32 than originally [21] anticipated. Urmylation at lysine residues next to the catalytic thiols may interfere with the anti-oxidant activity of Ahp1. Therefore, we examined whether the K32R, K156R and K32,156R mutants would affect t-BOOH sensitivity in vivo in relation to peroxidase-minus (C62S) or URM1 and YAP1 URM1 deletion strains (Fig. 7C). In the yap1Δahp1Δ strain, the K32R, K156R and K32,156R mutants all restored t-BOOH tolerance similar to that of wild-type AHP1 (Fig. 7C). This is in contrast to the peroxidase-dead mutant (C62S), which fails to change the t-BOOH sensitivity and lacks urmylation ( Fig. 7B and C). Together with t-BOOH sensitivity of the yap1Δurm1Δ double mutant (Fig. 7C), our data thus indicate that the lysine substitution mutants do not significantly differ in their response to t-BOOH cytotoxicity from yap1Δ cells that express wild-type Ahp1 but cannot undergo protein urmylation due to URM1 gene deletion. Hence, lysine dependent urmylation defects (K32R; K32,156R) appear not to affect the anti-oxidant activity of

Ahp1.
In agreement with our mutational analysis of the redox-active cysteines above (Fig. 6), we confirmed that for lysine-directed Urm1 conjugation to occur, Ahp1 must be catalytically active (Fig. 7D). Thus, residual urmylation levels typical of the lysine substitution mutants (K32R; K156R; K32,156R) (Fig. 7B) were found to be abolished upon mutation of the C P thiol (C62S) rather than the C R thiol (C31S) (Fig. 7D). In further support of this view are studies with the human homolog of Urm1 (hURM1), which we had previously shown to modify Ahp1 in yeast [20]. Here, our data show that, as is the case with yeast Urm1, the ability of hURM1 to form lysine-directed conjugates depends on the integrity of the C P thiol (Cys-62) and hence, on the anti-oxidant activity of Ahp1 (Suppl. Fig. S3).
The redox requirements of Ahp1 and molecular determinants that influence its urmylation are largely unknown. Our analyses of unmodified and urmylated Ahp1 under reducing or non-reducing conditions show that only a fraction of Ahp1 is subject to urmylation in vivo. Thus, under our experimental conditions, Ahp1 urmylation is not limiting, a notion in accordance with quantitative proteomic studies showing that in budding yeast, Ahp1 is considerably more abundant than Urm1 [53]. In contrast to canonical ubiquitination [54], we find no evidence for oligo-or poly-urmylation, and among the urmylated pool are Ahp1 intersubunit disulfides, which carry one Urm1 copy attached to each subunit. Since we are not aware of a deurmylase activity from yeast (or other model systems), it remains to be elucidated whether Urm1 is permanently attached to Ahp1.
In principle, the thiolate that Cys-62 forms in its reduced state may react with the thiocarboxylate of Urm1 (Urm1-COSH) to produce a thioester. However, taking into account that in its reduced state, Ahp1 is fully folded [12,65], the thiol of Cys-62 may not be accessible without local unfolding for Urm1-COSH transfer. C P thiol (Cys-62) oxidation and sulfenylation (-SOH) by peroxide (Fig. 8) could facilitate unfolding and prime the formation of an acyl disulfide between Ahp1 and Urm1 (Ahp1-S-S-CO-Urm1) rather than the above thioester. From organic chemistry in vitro it is known that acyl disulfides, which form between the thiocarboxylate of one peptide and an activated thiol of a second carrying a free amino group, are short-lived and readily ligate via iso-peptide bonds [66][67][68]. In analogy, we envision that the acyl disulfide (Ahp1-S-S-CO-Urm1) formed in vivo is highly reactive and undergoes a nucleophilic attack on its carbonyl group by the ε-amino group of a nearby lysine residue (Fig. 8). This will generate an isopeptide bond with Urm1 (Ahp1-NH-CO-Urm1) and leave the peroxidatic cysteine persulfidated (Cys-S-SH) (Fig. 8). Whether Lys-based urmylation stabilizes this persulfide on Ahp1 for trans-persulfidation of other targets is not known (Fig. 8).
In line with our working model (Fig. 8), a previous report [21] showed that among ten out of fourteen lysines tested for Ahp1 target site function, Lys-32 (next to Cys-31) is necessary for iso-peptide linkage with Urm1. However, we find that Lys-32 substitution alone (K32R) or in tandem with Lys-156 (K32,156R), which maps proximal to the C P thiol (Cys-62) (Fig. 7A), still allows residual urmylation. So, rather than being essential, Lys-32 likely represents one of several target sites for Urm1 conjugation with Ahp1. In support of promiscuous lysine sites, which from conventional ubiquitylation substrates in yeast (e.g. Sic1, Rpn4) [69][70][71] are not unheard of, we observe a minor Urm1 target role for Lys-156, and two more lysine residues, Lys-102 and Lys-107, can be found proximal to the active site in Ahp1 (Fig. 7A).
Although Urm1 is conserved in eukaryotes [20,34], the precise role the modifier plays for its target proteins is ill-defined. As for Ahp1, we have shown here that substitutions of the lysine-based urmylation sites (K32R; K32,156R) exhibit negligible effects on the anti-oxidant activity of the enzyme in vivo while peroxidase-dead mutants (C62S; C31,62S; C62,120S) all fail to be urmylated. Nonetheless, Urm1 attachment occurs close to the redox-active (Cys-31 Cys-62) center ( Fig. 1A and B)   Fig. 7. Lysine-directed Ahp1 urmylation in vivo requires the catalytic C P thiol (Cys-62). (A) Overview of lysine residues (K32, K102, K107, K156) close to the redox-active center in the reduced form of the Ahp1 homodimer (see Fig. 1A and B). (B, D) Shown are EMSAs under reducing conditions from the indicated strains expressing HA-URM1 (+) or not (−). NEM-stabilized HA-Urm1 conjugation was studied by anti-HA blot (top panels) diagnostic for free HA-Urm1 and urmylated Ahp1 (~36 kDa). anti-Ahp1 Western blot (middle panels) detects unmodified Ahp1 (~19 kDa); anti-Cdc19 (bottom panels) served as internal standard. (C) t-BOOH cytotoxicity assay in vivo. Growth of ahp1Δ or yap1Δahp1Δ cells carrying empty vector (ev), wild-type peroxiredoxin gene (AHP1) and cysteine or lysine substitutions was monitored together with urm1Δ and yap1Δurm1Δ reference cells without or with the indicated t-BOOH doses. and near the Ahp1-Trx2 interface [15], which is why urmylation cannot be excluded to interfere with some aspect of the Ahp1 peroxidatic cycle (e.g. t-BOOH detoxification or thioredoxin reduction; Fig. 1B). In support, Ahp1 regeneration by Trx2 has been shown to be affected in Lys-32 substitution mutants (K32A; K32E) in vitro [15]. Bearing in mind that a bacterial sulfur carrier with an Urm1-like fold (CysO-COSH) is more resistant to oxidation than sulfide and upregulated by oxidative stress [72], the role for yeast Urm1-COSH may also relate to peroxides, in particular t-BOOH, the preferred substrate of Ahp1 [11,12]. The latter notion agrees with our in vivo studies showing that while Urm1 conjugation was detectable in response to mild and intermediate t-BOOH doses, higher peroxide levels gradually prevented urmylation and significantly decreased Ahp1 disulfide formation.
We cannot exclude that inactivation by t-BOOH of Ahp1 or Urm1-COSH itself may impede urmylation of the peroxiredoxin in vivo. However, Ahp1 has been co-purified with a sulfiredoxin (Srx1) capable to reduce hyperoxidized Tsa1 [73], a distantly related 2-Cys peroxiredoxin [43]. Moreover, in vitro we found indications for peroxide inactivation of Urm1-COSH (Fig. 4B) suggesting this may constitute a factor accounting for the negative t-BOOH effects on Ahp1 redox biology and urmylation (Fig. 3). In sum, our comprehensive urmylation analysis of the 2-Cys peroxiredoxin Ahp1 has laid the foundation to better understand the redox requirements for Urm1 conjugation in vivo and provide insight into the mechanism for urmylation of other protein targets.

Expression and production of thiocarboxylated Urm1
In order to obtain thiocarboxylated Urm1, the Urm1-Intein-CBD-His 6 fusion protein was overexpressed in E. coli and purified according to Refs. [29,86] with modifications. In brief, the bacterial pellet was resuspended in lysis buffer without reducing agent and lysed to homogeneity. The lysate was passed through a Ni-NTA column and, following washes, the fusion protein was eluted with elution buffer (30 mM Tris-HCl pH 7.5; 300 mM NaCl; 250 mM imidazole and 10% glycerol). The eluates were dialyzed overnight to chitin column buffer (30 mM Tris-HCl pH 8 and 500 mM NaCl) and applied on a chitin column. The column was washed with chitin column buffer and the cleavage of the tag was induced through incubation with cleavage buffer (30 mM Tris-HCl pH 8; 500 mM NaCl and 35 mM ammonium sulfide) for 16 h at room temperature. This procedure leads to the formation of Urm1 without additional residues at the N-terminus and with a thiocarboxylated C-terminal glycine (Urm1-COSH). The eluted Urm1-COSH was further purified by size-exclusion chromatography on a HiLoad 16/600 Superdex 75 column on ÄKTA™ start system and stored at −80°C in storage buffer (20 mM Tris pH 7.5 and 200 mM NaCl). The presence of thiocarboxylated C-terminus was confirmed by running the protein on a polyacrylamide gel containing 25 μM APM ([N-Acryloyl-amino] phenyl) mercuric chloride [21,42].

Bacterial protein expression and purification of Ahp1
Procedures for cloning bacterial expression constructs for Trx2 and Step 1: The peroxidatic thiol of Ahp1 (Cys-SH) reacts with a peroxide to form a sulfenic acid (Cys-SOH), which following the fully folded to locally unfolded (FF-LU) transition [59] may become surface exposed.
Step 3: The ε-amino group of a nearby Lys residue mounts a nucleophilic attack on the carbonyl group of the acyl disulfide generating an iso-peptide bond (Lys-NH-CO-Urm1) between Ahp1 and Urm1 and a persulfidated cysteine (Cys-S-SH). Whether Lys-based urmylation stabilizes the persulfide on Ahp1, triggers H 2 S-release or drives trans-persulfidation of other targets is not known. For simplicity, the mechanistic hypothesis involves only one subunit of the Ahp1 homo-dimer. Ahp1 in the vector pET45b have been reported previously [14,87]. The Trr1 was amplified out of S. cerevisiae genomic DNA (Primer listed in Table S2). The PCR product was digested with NdeI and XhoI and subsequently cloned into pET29a with a C-terminal His tag. All clones were validated by DNA sequencing. Procedures for expressing and purifying His-tagged Ahp1 and Trx2 proteins have been reported previously [14,88]. A similar procedure was followed for the expression and purification of Trr1. Briefly, E. coli Rosetta cells transformed with pET29a-Trr1 were grown to mid-log phase in 400 mL LB medium containing 100 μg/mL ampicillin. Trr1 expression was induced with 1 mM IPTG for 6 h at 37°C. Proteins were purified from cell pellets using the Qiagen NiNTA Fast-Start kit. Eluted proteins were desalted using PD Minitrap G25 column equilibrated with TDG buffer (50 mM Tris pH 7.5, 2 mM DTT, 10% glycerol and protease inhibitor cocktail (G Biosciences)). Proteins were estimated to be > 95% pure by reducing SDS-PAGE. Extinction coefficients for proteins were estimated from the protein coding sequences as follows: Ahp1 (37950 M −1 cm −1 ), Trx2 (18,020 M −1 cm −1 ), and Trr1 (24719 M −1 cm −1 ).

Analysis of oligomeric state of Ahp1-dimer interface variants
Size exclusion chromatography coupled with multiangle light scattering (SEC-MALS) was used to monitor oligomeric state of Ahp1 proteins. Briefly, proteins were reduced in TDG by adding additional DTT to a final concentration of 50 mM and incubating for 30 min at room temperature. Reduced proteins were exchanged into SEC-MALS buffer (20 mM HEPES pH 7.5, 100 mM NaCl and 1 mM TCEP) using a BioSpin 6 column and diluted to 125 μM. Proteins were resolved on a gel filtration column and analyzed by MALS as previously described [43].

4.7.
In vitro t-BOOH response assay 500 ng of thiocarboxylated Urm1 was mixed in reaction buffer (20 mM Tris pH 7.5 and 200 mM NaCl). 0-10 mM tBOOH, was included and excluded as indicated. The reaction mix was incubated for 30 min at 30°C, stopped by adding Lämmli sample buffer ± DTT and incubated for 5 min at 95°C. Subsequently the thiocarboxylated Urm1 samples were loaded on SDS-PAGE gel containing 25 μM APM or not. For protein visualization, the gels were stained with Coomassie Brilliant Blue.

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.