Redox Modification of the Iron-Sulfur Glutaredoxin GRXS17 Activates Holdase Activity and Protects Plants from Heat Stress1[OPEN]

Laura Martins,a,b,2 Johannes Knuesting,c,2 Laetitia Bariat,a,b Avilien Dard,a,b Sven A. Freibert,d Christophe H. Marchand,e David Young,f,g,h Nguyen Ho Thuy Dung,f,g,h Wilhelm Voth,i Anne Debures,a,b Julio Saez-Vasquez,a,b Stéphane D. Lemaire,e,j Roland Lill,d Joris Messens,f,g,h Renate Scheibe,d Jean-Philippe Reichheld ,a,b,3,4 and Christophe Riondeta,b Laboratoire Génome et Développement des Plantes, Université Perpignan Via Domitia, F–66860 Perpignan, France Laboratoire Génome et Développement des Plantes, Centre National de la Recherche Scientifique, F–66860 Perpignan, France Department of Plant Physiology, FB5, University of Osnabrück, D–49069 Osnabrueck, Germany Institut für Zytobiologie und Zytopathologie, Philipps-Universität, Marburg 35032, Germany Institut de Biologie Physico-Chimique, Unité Mixte de Recherche 8226, Centre National de la Recherche Scientifique, Sorbonne Université, F–75005 Paris, France VIB-VUB Center for Structural Biology, 1050 Brussels, Belgium Brussels Center for Redox Biology, 1050 Brussels, Belgium Structural Biology Brussels, Vrije Universiteit Brussel, 1050 Brussels, Belgium Department of Molecular, Cellular, and Developmental Biology, University of Michigan, Ann Arbor, Michigan 48109 Institut de Biologie Paris-Seine, Laboratory of Computational and Quantitative Biology, Unité Mixte de Recherche 7238, Centre National de la Recherche Scientifique, Sorbonne Université, F–75005 Paris, France

Glutaredoxins (GRX) are small ubiquitous thiol reductases belonging to the thioredoxin (TRX) superfamily (Rouhier et al., 2008;Meyer et al., 2009). Most living organisms harbor two GRX classes (I and II), whereas additional classes are found in specific genera, including plants (Couturier et al., 2009). GRX from class I exhibit oxidoreductase activities, controlling thiol redox homeostasis of many proteins, mostly through glutathionylation/deglutathionylation activities. This activity generally depends on glutathione (GSH) as a reductant and relies on an active site harboring at least one Cys . GRX from class II are involved in the regulation of iron (Fe) metabolism and in the maturation of iron-sulfur (Fe-S) proteins (Mühlenhoff et al., 2010;Couturier et al., 2013;Haunhorst et al., 2013). This function relies on their capacity to bind a labile Fe-S cluster and to transfer it to other Fe-S-containing proteins (Zhang et al., 2011;Banci et al., 2015;Frey et al., 2016) and interact with BolA proteins (for review, see Couturier et al., 2015;Rey et al., 2019). The Fe-S cluster in class II GRX is coordinated by a conserved monocysteinic active site (CGFS) and a GSH molecule, a feature largely conserved within eukaryotic organisms (Lillig et al., 2005;Wingert et al., 2005;Rouhier et al., 2007;Bandyopadhyay et al., 2008;Couturier et al., 2011;Riondet et al., 2012;Berndt and Lillig, 2017).
In the model plant Arabidopsis (Arabidopsis thaliana), class II GRX are called GRXS14 (At3g54900), GRXS15 (At3g15660), GRXS16 (At2g38270), and GRXS17 (At4g04950; Supplemental Fig. S1). They share the capacity to bind and exchange Fe-S clusters with different client proteins, like ferredoxins and aconitase (Bandyopadhyay et al., 2008;Moseler et al., 2015). Consistent with this activity, most plant type II GRX are able to rescue defects of the budding yeast (Saccharomyces cerevisiae) grx5 mutant deficient in mitochondrial Fe-S cluster biogenesis, underlying their involvement in the maturation of Fe-S proteins (Molina et al., 2004;Liu et al., 2013;Knuesting et al., 2015;Moseler et al., 2015). Nevertheless, evidence for their implication in the biogenesis of Fe-S proteins is poorly described in vivo. Arabidopsis plants deficient for the plastidic GRXS14 exhibit accelerated chlorophyll loss upon prolonged darkness, leading to a decreased abundance of proteins acting in Fe-S cluster metabolism (Rey et al., 2017).
PICOT (PKC-interacting cousin of thioredoxin)like GRX are multidomain type II GRX found in the cytosol of most eukaryotic organisms (Berndt and Lillig, 2017). In most organisms, PICOT proteins are composed of one or two adjacent GRX domains with a CGFS active-site sequence motif preceded by a TRX domain. The PICOT proteins of vascular plants, like GRXS17, contain a third CGFS-containing GRX domain (Supplemental Fig. S1; Couturier et al., 2009). In vitro, GRXS17 is a homodimer, and each GRX CGFS domain coordinates an Fe-S cluster in a GSH-dependent manner (Knuesting et al., 2015). Similar to other class II GRX, GRXS17 is capable of complementing the defect of the yeast grx5 deletion mutant, likely by transferring Fe-S clusters to mitochondrial acceptor proteins (Bandyopadhyay et al., 2008;Knuesting et al., 2015). Affinity chromatography approaches with crude extracts have shown that GRXS17 associates with known cytosolic Fe-S protein assembly components like ME-THIONINE REQUIRING18/METHYL METHANESUL-FONATE19, DEREPRESSED FOR RIBOSOMAL PROTEIN S14 EXPRESSION, and other cytosolic Fe-S clustercontaining proteins like XANTHINE DEHYDRO-GENASE1 and CYTOSOLIC THIOURIDYLASE SUB-UNIT1 (CTU1) and CTU2 (Iñigo et al., 2016;Wang et al., 2019). However, Arabidopsis grxs17 null mutants have only a minor decrease in cytosolic Fe-S enzyme activities, indicating that GRXS17 likely does not play a major role in de novo Fe-S cluster assembly in the cytosol (Knuesting et al., 2015). Genetic evidence from different plant species suggests a protective role of GRXS17 during thermal, genotoxic, and drought stress (Cheng et al., 2011;Wu et al., 2012Wu et al., , 2017Iñigo et al., 2016;Hu et al., 2017), in meristem development (Knuesting et al., 2015), in hormonal responses (Cheng et al., 2011), and in integrating redox homeostasis and Fe deficiency responses . Nevertheless, for most of these functions, the underlying mechanism and targets remain enigmatic.
GRXS17 interacts with BolA2 and other cytosolic or nuclear protein partners (i.e. NF-YC11) that do not contain Fe-S clusters, suggesting that GRXS17 may have other functions than its Fe-S cluster metabolism activity (i.e. oxidoreductase or chaperone activities; Couturier et al., 2014;Knuesting et al., 2015;Iñigo et al., 2016). Indeed, in their apo-form, other Fe-S-containing GRX, like the human Grx2 and the plant GRXC1, function as thiol reductases and are proposed to act as redox sensors for the activation of antioxidant enzymes in oxidative stress conditions (Lillig et al., 2005;Rouhier et al., 2007;Zaffagnini et al., 2008;Couturier et al., 2011;Riondet et al., 2012). Moreover, thiol reductases from different types exhibit moonlighting functions like endonuclease activities (Liu et al., 2013) and chaperone activities Park et al., 2009;Chae et al., 2013). Therefore, the labile character of the Fe-S cluster of GRX suggests that the holo-and apo-forms might perform different activities under stress conditions.
Here, we address this question by exploring GRXS17 properties under oxidative and heat stress conditions. We found that, when subjected to oxidative conditions, reconstituted holo-GRXS17 rapidly loses its Fe-S cluster and becomes a Cys-dependent holdase. The switch to holdase involves oligomerization. Moreover, GRXS17 exhibited also a weak foldase activity that does not depend on its active-site Cys residues. To confirm the in vivo importance of these chaperone activities of GRXS17, we analyzed a grxs17 knockout mutant highly sensitive to heat stress. Mutant complementation experiments with several Cys variants of GRXS17 indicated the crucial role of the active-site Cys residues in protecting proteins during heat stress. Finally, GRXS17 associated with a different set of partner proteins after heat stress, indicating a stress-dependent functional switch of GRXS17 that could contribute to protecting cells from heat stress damage.

GRXS17 Releases Its Fe-S Cluster and Oligomerizes under Combined Oxidative and Heat Stress
When purified under anaerobic conditions, GRXS17 adopts a dimeric form coordinating up to three Fe-S clusters together with six molecules of GSH (Knuesting et al., 2015). As Fe-S clusters are sensitive to oxidation, we sought to examine the structural modifications of GRXS17 upon transfer to aerobic conditions. As expected, the anaerobically reconstituted GRXS17 Fe-S cluster shows a circular dichroism (CD) spectrum typical for the presence of [2Fe-2S] clusters (Fig. 1A). When exposed to air, the spectral footprint of the Fe-S cluster was progressively destabilized by oxygen and finally lost, confirming the sensitivity of the Fe-S cluster to oxidation (Fig. 1A).
In parallel to the loss of the Fe-S cluster, a sizeexclusion chromatography (SEC) assay revealed a progressive shift of GRXS17 to a high-molecular-weight (HMW) form of greater than 250 kD after 3 h of exposure to air. After 12 h, an HMW complex of a higher range was observed (Fig. 1B). Similar data were observed by native gel electrophoresis and nonreducing SDS-PAGE (Fig. 1, E and F). As expected, holo-GRXS17 migrates, on nonreducing native gels, as a single band likely corresponding to a dimeric form. However, after storage under aerobic conditions for more than 12 h, oligomers were formed, underscoring the high sensitivity of holo-GRXS17 to oxygen (Fig. 1, E and F). All in all, we showed that the presence of the Fe-S cluster prevents the oligomerization of GRXS17.
To further examine the stability of the Fe-S cluster under oxidative conditions, we exposed the holo-GRXS17 to physiological concentrations (0.2-1 mM) of hydrogen peroxide (H 2 O 2 ) . As previously shown (Knuesting et al., 2015), the UV-visible spectrum of the reconstituted GRXS17 exhibited absorbance peaks at 320 and 410 nm, characteristic for [2Fe-2S] clusters. Upon H 2 O 2 treatment, the Fe-S cluster is lost in a dose-dependent manner (;50% loss in 12 and 5 min at 0.2 and 1 mM, respectively), confirming the sensitivity of the Fe-S cluster to oxidation (Fig. 1,C and D). As high temperature was previously shown to destabilize Fe-S clusters in other proteins (Yan et al., 2013), we also monitored its stability at high temperatures. Both 35°C and 43°C temperatures had little effect on the stability of the cluster after short-term treatments, indicating that the Fe-S cluster is not heat sensitive (Fig. 1,C and D;Supplemental Fig. S2). However, when combining H 2 O 2 treatment with a physiological high temperature of 35°C, an accelerated loss of the Fe-S cluster (;50% loss in 3 min at 0.2 mM H 2 O 2 ) was observed, showing that higher temperatures potentiate the oxidative effect on Fe-S cluster stability. As expected, extensive oligomerization of GRXS17 was observed by native gel and nonreducing SDS-PAGE after H 2 O 2 treatment, while high-temperature exposure had less impact on the oligomerization of GRXS17 (Fig. 1, E and F).

GRXS17 Oligomerizes through Intermolecular Disulfides and Noncovalent Interactions
We further explored GRXS17 oligomerization after loss of Fe-S clusters using SEC (Fig. 2). As expected, HMW complexes (;900-1,000 kD) were observed in the native apo-GRXS17 ( Fig. 2A, blue line). When separating these fractions on a nonreducing SDS-PAGE gel, a laddering profile was observed, suggesting that the HMW complexes are partially dissociated by SDS and that the remaining HMW bands contain disulfidelinked GRXS17 oligomers (Fig. 2D). These data indicate that native GRXS17 oligomerizes through noncovalent interactions and intermolecular disulfides. Consistently, incubation of the native GRXS17 with dithiothreitol (DTT) prior to SEC led to a decrease of the ;900to 1,000-kD peak and a concomitant increase of the 110-kD peak, likely corresponding to homodimeric GRXS17 ( Fig. 2A). Furthermore, nonreducing SDS-PAGE analyses indicated that the peaks corresponding to oligomeric and dimeric GRXS17 consist of noncovalently interacting monomers (Fig. 2D).
To analyze the role of the active-site Cys residues on oligomerization, mutated forms of GRXS17 were separated by SEC. Interestingly, the TRX domain C33S mutant behaved similarly to reduced wildtype GRXS17 ( Fig. 2B; Supplemental Fig. S1), exhibiting comparable peaks at ;900 to 1,000 kD and 110 kD. Moreover, the distribution between these peaks was not influenced by DTT treatment prior to SEC, indicating that Cys-33 plays a major role in the oligomerization, very likely by forming intermolecular disulfide bonds (Fig. 2B). Nonreducing SDS-PAGE analyses of the SEC fractions confirmed this statement, as only monomers were found in both conditions (Fig. 2E). Strikingly, mutation of all active-site Cys residues completely abolished oligomerization, as GRXS17-C33/179/309/416S exclusively eluted as a dimer ( Fig. 2C; Supplemental Fig. S1). Again, only monomers were found by nonreducing SDS-PAGE, indicating that the dimeric apo-GRXS17 is likely formed by noncovalently interacting (Fig. 2F).
To further investigate the reduction mechanism of the intermolecular disulfide bond-mediated oligomers, we used a physiological disulfide bond reduction system (NADPH/recombinant thioredoxin reductase A [NTRA]/TRXh5) located in the cytosol (Reichheld et al., 2007). Interestingly, incubating GRXS17 oligomers with the complete TRX system partially restored monomers, suggesting that at least in vitro, GRXS17 covalent oligomers are reduced by the TRX system (Fig. 2G).
Finally, to verify the impact of Cys mutations on the GRXS17 structure, we analyzed secondary structure changes of GRXS17 and GRXS17-C33/179/309/ 416S with CD spectroscopy (Supplemental Fig. S3). The general structure (a-helices, b-strands, and turn) remained similar between both proteins, indicating that the C33/179/309/416S mutations do not induce major secondary structure changes of GRXS17. The effect of Cys oxidation on the conformation of GRXS17 was also determined. Treatment with 400 mM H 2 O 2 resulted in a minor change in secondary structure composition of wild-type GRXS17 but not for the GRXS17-C33/179/ 309/416S variant (Supplemental Fig. S3A). This minor change in secondary structure composition of GRXS17 upon oxidation can correspond to either a local or a global conformational change, which is likely to involve the partial unfolding of a-helices and the formation of b-strands (Supplemental Fig. S3B). As this conformational change was not observed for the quadruple Cys variant GRXS17-C33/179/309/416S, we can conclude that this structural change depends on the oxidation of the catalytic Cys residues and the formation of disulfide bonds in wild-type GRXS17.
To summarize, we showed that recombinant GRXS17 is able to oligomerize by forming intermolecular disulfides Figure 1. Fe-S cluster stability and oligomerization of reconstituted GRXS17. A, Visible CD spectra of reconstituted holo-GRXS17 under reducing conditions (blue) and after oxygen treatment (3 h, brown; 12 h, red) at room temperature. B, SEC (Sephacryl S300 HR) under anaerobic conditions with reconstituted GRXS17 after the respective oxygen treatments. C, Absorption spectra of GRXS17 subjected to H 2 O 2 and heat treatments. UV-visible absorption spectra were measured after in vitro reconstitution under anaerobic conditions and after 20 min of treatment with 20°C (blue), 2.5 mM GSH at 20°C (orange), H 2 O 2 at 20°C (0.2 mM, green; 1 mM, red), 35°C in the presence of 0.2 mM H 2 O 2 (black), or without H 2 O 2 (light blue). The inset shows absorption at 410 nm of the same samples. Data are means of three to six biological repetitions 6 SE. The letters a, b, and c indicate significant differences compared with the control (blue; P . 0.1, P , 0.001, and P , 0.00001, respectively), by Student's t test. D, Kinetics monitoring the GRXS17 Fe-S cluster stability subjected to the same treatments as in C. E and F, Migration of 5 mg of recombinant GRXS17 with or without the cluster in native PAGE (E) or in nonreducing SDS-PAGE (F). M, Molecular marker. Holo-GRXS17 (lane 1) was maintained in air for 3 h (lane 2) or 12 h (lane 3) or treated for 10 min at 35°C (lane 4), with 0.2 mM H 2 O 2 (lane 5), or with a combination of both treatments (lane 6). Samples were subjected to nonreducing SDS-PAGE, and protein bands were stained by Coomassie blue. Each experiment is representative of three biological repetitions.
in which the active site Cys-33 of the TRX domain plays a major role, and which can be reduced in vitro by the TRX system. In addition to being involved in Fe-S cluster coordination, Cys-179, Cys-309, and Cys-416 seem to be important for maintaining noncovalent interactions between protomers. Finally, the active-site residues Cys-33, Cys-179, Cys-309, and Cys-416 are not involved in the dimer structure, which likely relies on noncovalent interactions.
A Holo-Apo Redox Switch of GRXS17 Activates Holdase Activity upon Oxidation Molecular chaperone activities are often associated with oligomerization and are related to thermotolerance . Therefore, the potential role of GRXS17 as a molecular chaperone was analyzed using different chaperone substrates. To monitor holdase activity, we used the Arabidopsis cytosolic malate dehydrogenase (MDH1; EC 1.1.1.37) and the porcine heart mitochondrial citrate synthase (CS; EC 4.1.3.7) as client proteins ( Fig. 3; Supplemental Fig. S4). We analyzed the effect of GRXS17 on the high temperature-induced aggregation of MDH and CS, a classical assay to monitor the activity of molecular chaperones in vitro (Voth et al., 2014). Incubation of MDH or CS with reduced or oxidized apo-GRXS17 prevented thermal aggregation at 55°C and 43°C, respectively ( Fig. 3A; Supplemental Fig. S4). Increasing the apo-GRXS17:MDH ratio progressively decreased MDH aggregation, which was fully prevented at a molar ratio of 1:1 apo-GRXS17:MDH (Fig. 3B). Notably, this effect was stronger than described for holdases from various organisms (e.g. HSP33, Get3, CnoX, and TDX; Jakob et al., 1999;Lee et al., 2009;Voth et al., 2014;Goemans et al., 2018). In contrast, in the absence of GRXS17 or in the presence of an excess of ovalbumin, MDH or CS significantly aggregated after 15 min (Fig. 3, A and B). Interestingly, a comparable holo-GRXS17:CS ratio did not prevent CS aggregation, indicating that the holdase activity of GRXS17 is specifically associated with the apo-GRXS17 form (Fig. 3A). Subsequently, we verified that the Fe-S cluster in the holo-GRXS17 is stable at 43°C (Supplemental Fig. S2). Notably, holo-GRXS17 holdase activity could not be measured using MDH because of the instability of the Fe-S cluster at 55°C (Supplemental Fig. S2). . GRXS17 forms HMW complexes. A to C, SEC analysis (Superose 6) of 740 mg of GRXS17 (A), GRXS17-C33S (B), and GRXS17-C33/179/309/ 416S (C) maintained in air after purification. The protein was reduced with 100 mM DTT (red) or without (blue) before loading on the column. Approximate molecular weights were calculated based on the elution volume of reference proteins (see "Materials and Methods"). D to F, Aliquots of elution fractions (30 mL) shown in A to C were subjected to nonreducing SDS-PAGE. G, GRXS17 oligomers are reduced by the TRX system. Recombinant GRXS17 (5 mM) was incubated with NADPH (150 mM) for 1 h at 30°C, with or without NTRA (0.8 mM), DTT (50 mM), and TRXh5 as indicated in the table. Afterward, proteins were separated by nonreducing SDS-PAGE.
We also monitored the foldase activity of GRXS17 using glucose-6-phosphate dehydrogenase (G6PDH) as a substrate. To distinguish between foldase and disulfide reductase activities, we used urea-denatured Cys-free G6PDH from Leuconostoc mesenteroides (EC 1.1.1.49). Increasing the apo-GRXS17:G6PDH ratio progressively restored the G6PDH activity (Fig. 3,C and D). This reactivation was faster in the presence of GRXS17, reaching nearly 36% recovery after 3 h at a quite high 10:1 GRXS17:G6PDH ratio (Fig. 3D). Reduced or oxidized holo-GRXS17 similarly restored G6PDH activity, indicating that the foldase activity of GRXS17 is independent of the presence or absence of the Fe-S cluster (Fig. 3, C and D). We also showed that these activities are ATP independent (Supplemental Fig. S5). Collectively, our data show that GRXS17 can adopt both a holdase and a slight foldase activity in vitro and that the holdase activity is specific for the apo-form.

Oligomerization Determines the Holdase Activity of GRXS17
The capacity of apo-GRXS17 to oligomerize depends on both disulfide bonds involving active-site Cys residues and noncovalent electrostatic interactions (Fig. 2). To uncover forms that are associated with chaperone activity, we tested the chaperone activity of the Cys/Ser variants of GRXS17 using MDH as substrate (Fig. 4A). We found that GRXS17(C33S), which does not abolish the oligomerization state, had holdase activity (Figs. 2C and 4A). However, in the oligomerization-abolished GRXS17-C33/179/ 309/416S variant (Fig. 2D), no holdase activity was observed (Fig. 4A). This suggested that the holdase activity depends on the presence of the active-site Cys residues. To further demonstrate that oligomerization is required for holdase activity, we incubated reduced and oxidized apo-GRXS17 with a low concentration of the chaotropic agent guanidinium hydrochloride, which abolishes weak bonds like hydrophobic and electrostatic interactions (Fig. 4B). We found that the holdase activity of reduced GRXS17 is inhibited with guanidinium hydrochloride while the holdase activity of oxidized GRXS17 is preserved. Finally, we tested the role of the Cys residues in GRXS17 foldase activity (Fig. 4C). We found that GRXS17 Cys mutations do not negatively affect GRXS17 foldase activity, suggesting that disulfide-linked oligomerization is not required for its foldase activity. CS or MDH were incubated alone (Control), with ovalbumin, or with the different GRXS17 proteins. A, Holo-GRXS17 or apo-GRXS17 in their reduced or oxidized form were incubated with CS at a 1:1 (0.5:0.5 mM) ratio at 43°C. The ovalbumin ratio was 1:1 (0.5:0.5 mM). B, Ratios of apo-GRXS17:MDH were 0.2:1 (0.1:0.5 mM), 0.5:1 (0.25:0.5 mM), and 1:1 (0.5:0.5 mM). The ovalbumin ratio was 1:1 (0.5:0.5 mM). C and D, GRXS17 foldase activity was measured by reactivation of the G6PDH activity after denaturation with urea. The zero time point represents the starting activity for the reactivation assay, with 100% being the activity before heat denaturation. AU, Arbitrary unit. C, Holo-GRXS17 or apo-GRXS17 in their reduced or oxidized form were incubated with G6PDH at a 5:1 (0.65:0.13 mM) ratio at 30°C. The ovalbumin ratio was 5:1 (0.65:0.13 mM). D, Ratios of apo-GRXS17:G6PDH were 2:1 (0.26:0.13 mM), 5:1 (0.65:0.13 mM), and 10:1 (1.3:0.13 mM). The ovalbumin ratio was 5:1 (0.65:0.13 mM). Data are means of at least three biological repetitions 6 SE. Control and ovalbumin curves are the same and represent means and SD of all experiments (n . 9).

GRXS17 Is Involved in Tolerance to Moderate Heat Stress
To reveal the physiological significance of the redoxdependent chaperone activity of GRXS17, we explored the possible role of GRXS17 in plant thermotolerance. Therefore, we subjected grxs17 mutant plants to different heat stress regimes (Yeh et al., 2012). The impact of heat stress on shoot meristem viability was first quantified by monitoring the appearance of new leaves. The viability of the grxs17 mutant was similar to that of wild-type plants for both short-term acquired thermotolerance (SAT) and long-term acquired thermotolerance (LAT) regimes (Fig. 5, A and B). In contrast, a mutant lacking the Heat Shock Protein101 (HSP101) chaperone was hypersensitive to SAT and LAT regimes (Yeh et al., 2012), indicating that GRXS17 is not involved in SAT and LAT (Fig. 5, A and B). However, compared with wild-type plants, grxs17 mutants were significantly more sensitive to a moderate heat stress regime (tolerance to moderated high temperature [TMHT]; Fig. 5C), indicating that GRXS17 is specifically involved in the response to this heat stress regime. The thermosensitive phenotype of grxs17 was fully complemented by constitutively overexpressing the fulllength GRXS17:GFP fusion under the control of the 35S cauliflower mosaic virus promoter, indicating that the TMHT phenotype of grxs17 is due to the loss of the GRXS17 function and that neither the attached GFP nor the constitutive overexpression under the control of a strong promoter perturbs the complementation functions ( To study the impact of high temperature on roots, we performed a similar experiment by following the root development after the TMHT regime (Fig. 6). Plants were grown at 20°C for 6 d before the temperature was shifted to 30°C. For root growth, 30°C was used instead of 35°C because this latter temperature affected the root growth of plants from all genotypes (Supplemental Fig.  S7). The primary root of wild-type seedlings grew similarly after the 30°C temperature shift compared with plants continuously grown at 20°C (Fig. 6, A and B). While root growth of the grxs17 mutant was slower than that of the wild type at 20°C, it almost completely stopped after the 30°C shift, confirming the role of GRXS17 in root growth, which is exacerbated at higher temperature. When the grxs17 mutant was complemented with the GRXS17-GFP fusion protein, the mutant behaved like the wild type in both conditions. However, GRXS17-C33/179/309/416S:GFP failed to complement the grxs17 mutant, pointing to the role of these Cys residues in root growth (Fig. 6, A and B). The size of the root apical meristem (RAM) was observed after 12 d of growth at 20°C or after the shift to 30°C (Fig. 6, C and D). The size of the RAM associated with the primary root growth was slightly smaller at 20°C in both in the grxs17 and the GRXS17-C33/179/309/416S:GFP-complemented lines and was . Role of GRXS17 active-site Cys residues and oligomerization in holdase and foldase activities. A, Holdase activity of GRXS17 and Cysmutated GRXS17 proteins was measured toward MDH heat-induced precipitation by light scattering at 55°C. MDH was incubated alone (Control), with ovalbumin, or with the different GRXS17 proteins at a 1:1 (0.5:0.5 mM) ratio. B, GRXS17 proteins reduced by 5 mM DTT (yellow) or oxidized with 1 mM H 2 O 2 (green) were treated with 20 mM guanidinium hydrochloride (Gua) as a chaotropic reagent (brown and light blue, respectively). MDH was incubated alone (Control) or with ovalbumin. AU, Arbitrary unit. C, Foldase activity of GRXS17 and Cys-mutated GRXS17 proteins was determined by measuring the G6PDH activity after denaturation with urea. The zero time point represents the starting activity for the reactivation assay, with 100% being the activity before heat denaturation. G6PDH was incubated alone (Control), with ovalbumin, or with the different GRXS17 proteins at a 5:1 (0.65:0.13 mM) ratio. Data are means of at least three biological repetitions 6 SE. Control and ovalbumin curves are the same and represent means and SD of all experiments (n . 9). strongly decreased at 30°C (Fig. 6, C and D; Supplemental  Fig. S8). This shows a high impact of heat stress on the meristematic activity of the mutant.
To characterize other root growth parameters, we also estimated the role of GRXS17 in lateral root (LR) development. While LR density was not affected at 20°C in any of the studied genotypes, it markedly decreased in the grxs17 and the GRXS17-C33/179/309/416S:GFPcomplemented lines at 30°C (Fig. 6E).The density of lateral root primordia (LRP; i.e. early stage of LR not emerged from the root cortex) increased in the same genotypes, suggesting that high temperature does not inhibit the initiation of LR but rather elongation in the mutant and GRXS17-C33/179/309/416S:GFPcomplemented lines (Fig. 6F). PI staining was used to monitor cell death in meristematic zones (Truernit and Haseloff, 2008). While we did not observe cell death in the RAM in any conditions and genotypes ( Fig. 6D; Supplemental Fig. S8), cell death was observed in the LRP in grxs17 (Fig. 6G). This suggests that GRXS17 prevents heat stress injuries at an early stage of LR development, supporting the LR elongation inhibition at 30°C. Indeed, Figure 5. GRXS17 is specifically involved in thermotolerance to moderate high temperature. A to C, Wild-type plants (Columbia-0 [Col-0]), the grxs17 knockout mutant (grxs17), the grxs17 mutant complemented with a Pr35S:GRXS17:GFP construct expressing the GRXS17:GFP fusion protein (1GRXS17), the GRXS17-C33S:GFP protein (1GRXS17-C33S), or the GRXS17-C33/179/309/416S:GFP protein (1GRXS17-C33/179/309/416S), Heat Shock Factor A1 quadruple knockout mutant (hsfA1QK), or HSP101 knockout mutant (hsp101) were subjected to different temperature stress regimes: LAT (A), SAT (B), or TMHT (C). do, Days old; rec, recovery. The viability of the plants was assessed by the recovery of shoot growth. Data are means of four to 10 biological repetitions 6 SE, n 5 20 to 25. The letters a, b, and c indicate significant differences compared with Col-0 (P . 0.1, P , 0.001, and P , 0.00001, respectively), by Student's t test. D, GRXS17 protein expression in complemented grxs17 lines. Total protein extracts were prepared from 2-week-old plantlets. Protein extracts were separated by SDS-PAGE and probed with antibodies directed against GRXS17. Plants expressing the transgenes at different levels are represented. In Col-0, the 50-kD signal corresponds to the native GRXS17 protein. In the complemented lines, the 80-kD signal corresponds to the GRXS17:GFP fusion protein.
the role of GRXS17 in early meristem establishment was also observed when the grxs17 mutant seeds were subjected to high temperature at 28°C. While the mutant seeds germinated properly, they stopped growing at the early plantlet stage and harbored a very short root. Importantly, the structure of the RAM looked disorganized and showed cell death (Supplemental Fig. S9).
Altogether, these results highlight the role of GRXS17 in the tolerance of plants to moderate heat stress, acting to protect both shoot and root meristems. Importantly, the active-site Cys residues of GRXS17 seem to play an essential role in root growth and thermotolerance.

GRXS17 Oligomerizes in Planta and Associates with a Different Set of Proteins upon Shifting to a Higher Temperature
In addition to in vitro assays, we also explored GRXS17 oligomerization in vivo. Total soluble protein Figure 6. GRXS17 is involved in root development under heat stress. Wild-type (Col-0), grxs17 knockout mutant (grxs17), and grxs17 mutant complemented with a Pr35S:GRXS17:GFP construct (1GRXS17) or the Pr35S:GRXS17-C33/179/309/ 416S:GFP construct (1GRXS17-C33/ 179/309/416S) plants were grown 12 d continuously at 20°C or 6 d at 20°C and shifted to 30°C for another 6 d. A and B, Primary root length of plants grown at 20°C (A) or shifted to 30°C (B). C, RAM of plants grown at 20°C (white box plots) or shifted to 30°C (gray box plots). D, RAM of plants shifted to 30°C and stained by propidium iodide (PI). RAM length (bidirectional arrows) was determined by the distance from the quiescent center to where endodermis cells were elongating. E and F, Density (number per cm of primary root) of emerged LR (E) and LRP (F) after growth at 20°C (white box plots) or shifted to 30°C (gray box plots). G, LRP of plants shifted to 30°C and stained by PI. White arrowheads pinpoint cell death in LRP, whereas unfilled arrowheads represent LRP where cell death did not occur. Data are means of three biological repetitions 6 SE, n 5 15. The letters a, b, and c indicate significant differences compared with Col-0 ( P . 0.1, P , 0.001, and P , 0.00001, respectively), by Student's t test. Bars 5 50 mm. extracts prepared from Arabidopsis seedlings grown at standard temperature (20°C) and exposed to heat stress (35°C for 2 h) were fractionated by SEC ( Fig. 7; Supplemental Fig. S10). Immunoblot analyses of eluted fractions revealed a peak containing GRXS17 eluting at ;100 to 150 kD (lanes 23-25), corresponding to the M r of a GRXS17 dimeric form (;110 kD; Knuesting et al., 2015). Moreover, at 20°C, a GRXS17 signal was also detected in fractions (lanes 18-21) corresponding to an HMW (;380 kD) form ( Fig. 7; Supplemental Fig. S10). Interestingly, plants subjected to heat stress accumulated higher levels of HMW forms eluting in the range from the SEC column (lanes 16-18), corresponding to 520 kD (Fig. 7). These data suggest that, similar to in vitro analyses ( Figs. 1 and 2), GRXS17 forms HMW complexes in vivo, although the sizes of these complexes are not fully similar. Also, we cannot rule out that GRXS17 oligomers associate with other proteins.
Finally, we sought to identify GRXS17 partners using immunoprecipitation coupled to mass spectrometry determination (IP-MS). To this end, grxs17 mutant plants were transformed with a Pr35S:GRXS17:FLAG-HA construct. The expression of the construct was verified by immunoblotting, and its functionality was assessed by complementation of the phenotype under TMHT (Supplemental Fig. S11). Total protein extracts from GRXS17:FLAG-HA-expressing plants grown at 20°C or shifted for 2 h to 35°C were collected for IP-MS analyses. Nonspecific GRXS17 contaminants were identified within the same experiments using grxs17 mutant plants expressing a GRXS17 protein devoid of the FLAG tag. We employed a label-free quantitative proteomic approach to identify proteins significantly enriched after immunoprecipitation by comparison with the control samples. Five biological replicates were analyzed by mass spectrometry after FLAG immunoprecipitation from total protein extracts of plants expressing GRXS17:FLAG-HA or nontagged GRXS17 (control) grown at 20°C or shifted for 2 h to 35°C. GRXS17 was the only protein significantly enriched in the tagged samples after immunoprecipitation (P , 0.05, Bonferroni-corrected Student's t test). In addition, we identified 23 proteins from plants grown at 20°C and 12 proteins from plants shifted to 35°C that were present and quantified in at least three of the five replicates from the tagged strain and not found in any of the five replicates from control nontagged plants (Table 1). Using this all-or-nothing approach (i.e. considering only proteins totally absent in control plants), several likely interactants, such as the redoxrelated enzymes protein disulfide isomerase (PDIL1 to PDIL4) and monodehydroascorbate reductase (MDAR2), were identified at 20°C. As suggested by the predicted subcellular localization, all partners identified are not presumably direct interactants of GRXS17 in the cell (Table 1). While protein domains common between BolA1 and BolA2 likely mediate the interaction with GRXS17 (Couturier et al., 2014), the interaction with BolA1 located in chloroplasts may not be physiologically relevant. A different set of interaction partners was identified after a 2-h temperature shift from 20°C to 35°C, and only four proteins were identified in both conditions (Table 1). Among them, the nucleocytoplasmic BolA2 was always present (Supplemental Table S1). This interaction is consistent with previous studies that report the interaction of GRXS17 with BolA2 (Couturier et al., 2014). Finally, to identify GRXS17 partners in HMW complexes, total protein extracts from plants grown at 20°C or shifted for 2 h to 35°C were first separated by SEC, and HMW fractions containing GRXS17:FLAG-HA were collected for IP-MS analysis. A simple subtractive analysis between the samples failed to identify a similar set of proteins as in the previous approach, suggesting that those proteins are not involved in HMW complexes (Supplemental Table S1). However, BolA2 was identified after IP-MS in HMW fractions, indicating that BolA2 is a major GRXS17 oligomer interactant possibly associated with the molecular chaperone activity of GRXS17.

Apo-GRXS17 Switches to Holdase Activity upon Stress
During periods of heat stress, living organisms need to rapidly adapt to cope with the damaging effects of increasing temperatures. The sessile nature of plants makes them particularly vulnerable to heat stress. As in other organisms, a set of heat shock proteins acts as molecular chaperones to prevent damage caused by protein inactivation and aggregation (Hartl et al., 2011;Hanzén et al., 2016). Here, we demonstrated that the Fe-S cluster-harboring GRXS17 becomes a holdase after exposure to heat in oxidizing conditions. This switch happens within minutes by a redox-and temperaturemediated destabilization of the Fe-S cluster and subsequent disulfide bond formation (Fig. 8). The redox switch of Fe-S cluster-containing GRX has previously been shown to occur under oxidizing conditions in other GRX. In these cases, the active-site Cys residues are involved in the Fe-S cluster, which inhibits the thiolreducing activity, with the loss of the Fe-S cluster Figure 7. GRXS17 forms HMW complexes in plant extracts. SEC analysis (Sephacryl S300 HR) of crude protein extracts from Arabidopsis seedlings cultivated for 14 d at 20°C and further treated for 2 h at 35°C is shown. The GRXS17 protein was detected after immunoblotting and immunodetection using an anti-GRXS17 serum. Numbered lines correspond to the protein fractions. Protein sizes of the respective fractions were obtained after calibration using marker proteins.
Interestingly, the active-site Cys residues of GRXS17 are crucial for the holdase redox switch, as they coordinate both the Fe-S cluster and respond to an oxidizing environment, such as aerobic conditions, H 2 O 2 exposure, or heat stress. As such, GRXS17 may act as a sensor for cellular redox changes. In fact, high temperature by itself does not destabilize the Fe-S cluster, but the simultaneous presence of H 2 O 2 oxidatively changes the conformation of the protein and induces oligomerization through noncovalent interactions and disulfides (Fig. 8). These observations indicate that the Fe-S cluster stabilizes the dimeric structure of GRXS17, and only after dissociation can the protein undergo oligomerization. A similar function has previously been shown in Archaea, where the loss of a [4Fe-4S] cluster causes oligomerization of the D subunit of the RNA polymerase (Hirata et al., 2008). In GRXS17, we showed that the active-site Cys residues (Cys-33, Cys-179, Cys-309, and Cys-416) are required for both noncovalently Binding partners of GRXS17 isolated from 10-d-old plantlets grown at 20°C or after 2 h of heat treatment at 35°C are listed. Putative/established subcellular localizations were determined using SUBA4 (Hooper et al., 2017). Common substrates located in the same (*) or in another (**) subcellular compartment as GRXS17 are indicated. For more details, see Supplemental Table S1. and covalently mediated oligomerization (Fig. 2). Although GRXS17 reduces in vitro a monomeric glutathionylated form and a dimeric disulfide-bridged form of BolA2 (Couturier et al., 2014), the oligomerization of the apo-GRXS17 protein under oxidative conditions may inhibit this activity (Ströher et al., 2016). Proteins that combine chaperone function with a redox activity have been described previously. For example, the plant thiol reductase proteins TRXh3, TDX, and NTRC exhibit chaperone activities triggered by oxidative stress and heat shock exposure, leading to a shift of the protein structure from a low-M r form to HMW complexes Park et al., 2009;Chae et al., 2013). Also, overoxidation of nucleophilic Cys of 2-Cys PRX triggers oligomerization and induces chaperone activity (Liebthal et al., 2018). GRXS17 is different, as it has no oxidoreductase activity but has a much more efficient holdase activity than the thiol oxidoreductases ( Fig. 3; Lee et al., 2009;Park et al., 2009;Chae et al., 2013). Remarkably, the key to induce holdase chaperone activity is the release of the redoxdependent Fe-S cluster. To our knowledge, this is the first example of an FeS protein that turns into a chaperone through a switch between holo-and apo-forms. In Escherichia coli, a similar example is Hsp33, which coordinates zinc via conserved Cys residues (Jakob et al., 1999(Jakob et al., , 2000Winter et al., 2008). Upon oxidation with H 2 O 2 , disulfide bonds are formed between the coordinating Cys residues and zinc is released, activating Hsp33 holdase activity. Here, zinc coordination, oxidation, and the activation state of Hsp33 are directly linked to the redox state of the environment (Jakob et al., 1999(Jakob et al., , 2000Winter et al., 2008). As another example in yeast, the cytosolic ATPase Get3 turns into an effective ATP-independent chaperone when oxidized (Voth et al., 2014). In yeast also, a redox switch in the peroxiredoxin Tsa1 is required for the recruitment of Hsp70 chaperones and the Hsp104 disaggregase to misfolded proteins formed upon H 2 O 2 exposure during aging (Hanzén et al., 2016). Furthermore, the holdase activity of the E. coli chaperedoxin CnoX is activated by chlorination upon bleach treatment to protect substrates from irreversible oxidation (Goemans et al., 2018). Whether GRXS17 acts by protecting its interactors from overoxidation or to recruit other chaperones to their target proteins remains to be determined.
Our in vitro experiments showed that GRXS17 has a slight constitutive foldase activity that is independent of ATP (Supplemental Fig. S5). Such activity clearly differs from most bona fide prokaryotic (e.g. GroEL-GroES) and eukaryotic (e.g. Hsp70 and Hsp90) foldases, which need ATP for their foldase activity (Mayer, 2010;Hartl et al., 2011). Nevertheless, a number of chaperones have been identified that promote folding in the absence of high-energy cofactors Horowitz et al., 2018). As an example, the small ATP-independent chaperone Spy from E. coli uses longrange electrostatic interactions and short-range hydrophobic interactions to bind to its unfolded client proteins and drive their folding .
PDI and cyclophilin, which catalyzes disulfide formation and isomerization of disulfides, exhibit folding activities through their active-site Cys (Park et al., 2013;Ali Khan and Mutus, 2014). GRXS17 functions differently, as the foldase activity of GRXS17 is independent of its active-site Cys residues (Fig. 4). Nevertheless, two additional Cys residues are present in the first and third GRXS17 domains, including Cys-470 located in the proximity of the potential GSH-interacting residue Asp-471 and that is also conserved in GRXS14 and GRXS16 but not in GRXS15 (Supplemental Fig. S1). Their involvement in GRXS17 foldase activity cannot be ruled out. Future experiments will have to be performed to decipher the mechanism of the foldase activity of GRXS17.

GRXS17 Associates with Different Proteins under High Temperature
In eukaryotic organisms grown under nonstress conditions, the cytosolic compartment is a highly reducing environment (Schwarzländer et al., 2008). However, in heat-stressed plants, intracellular oxidation triggered by Figure 8. Model of the GRXS17 redox switch. Under reducing conditions, the holo-GRXS17 forms a dimer that coordinates an Fe-S cluster (ISC). In this form, GRXS17 can transfer ISC to cytosolic or nuclear target proteins (represented as a green line in the middle of the oligomeric structures). Under oxidative conditions (H 2 O 2 , oxygen [O 2 ], and heat stress [HS]), ISC are released and GRXS17 switches to an apo-dimeric form with foldase activity. Dimers can spontaneously associate to oligomeric forms, and further oxidation induces oligomeric structure through the formation of intermolecular disulfide bonds at active-site Cys residues. In this form, GRXS17 acquires holdase activities and provides thermotolerance. The oligomeric form can be reduced to recover the apo-dimeric GRXS17, and the ISC can eventually be reassembled.
reactive oxygen species accumulation has been largely documented (Choudhury et al., 2017;Dickinson et al., 2018). Therefore, with regard to changes in redox status depending on the temperature, it is likely that the major GRXS17 form is the Fe-S cluster-containing dimeric form under standard growth temperatures, but a significant proportion of GRXS17 switches to HMW complexes in plants subjected to heat stress (Fig. 7). Although the differences in the size of the HMW complexes observed in recombinant GRXS17 and in plant extracts will require further exploration, the switch to HMW complexes appears to be necessary for plant survival under heat stress. As the grxs17 mutant is highly sensitive to moderate hightemperature regimes, the mutant can only be complemented by the GRXS17-Cys variants that are still able to oligomerize (GRXS17 and GRXS17-C33S) and not by GRXS17-C33/179/309/416S (Figs. 5 and 6; Cheng et al., 2011;Knuesting et al., 2015). While our data do not fully demonstrate that the holdase activity is needed for heat tolerance, they clearly show the importance of active-site Cys in this function. Moreover, the role of Fe-S clusters in thermotolerance will have to be further studied by generating mutant grxs17 plants complemented by GRXS17-C179/309/ 416S, in which only Cys residues involved in Fe-S cluster coordination (and not in chaperone activity and oligomerization) are mutated. We unfortunately failed to generate such plants yet.
GRXS17 has been previously proposed to improve thermotolerance, drought tolerance, and oxidative tolerance by enhancing reactive oxygen species-scavenging capacities and the expression of heat shock proteins. This might occur by modulation of gene expression, enzyme activity, or protection of antioxidant enzymes (Wu et al., 2012Hu et al., 2017). Our data suggest an additional function of GRXS17 through its chaperone activity. The GRXS17 partner protein identification resulted in a low overlap between proteins immunoprecipitated at 20°C and those identified at a higher temperature, suggesting that the set of interacting proteins changes under heat stress. Although it is tempting to suggest that the proteins we identified in immunoprecipitation experiments are protected by GRXS17, further experiments will need to validate these partners and the biological significance of these interactions. Importantly, our observations showing that GRXS17 maintains plant viability under moderate heat stress conditions and prevents cell death in founder cells of root primordia under high temperature might indicate a role in protecting key actors of meristem function. Possible candidates that we found in our immunoprecipitation are MPK4, which plays a role in temperature signaling pathways (Zhao et al., 2017), and casein kinase II, a ubiquitous Ser/Thr protein kinase in eukaryotes implicated in multiple developmental and stress-responsive pathways (Mulekar and Huq, 2014;Wang et al., 2014). Whether the GRXS17 chaperone activities are involved in the protection of these proteins in vivo will require further investigation.
Finally, we propose that the unique properties of GRXS17 among the redox-regulated chaperones (e.g. 2-Cys PRX, TRXh3, or TDX) come from its inducible holdase function (i.e. Fe-S cluster release under stress), which is not seen for other heat stress-related chaperones . Homologs of GRXS17 are present in all plants and in most other organisms (Couturier et al., 2009), and their functions in Fe metabolism are well documented (Encinar del Dedo et al., 2015;Berndt and Lillig, 2017). It is likely that the switch to holdase chaperone activity, which we described here, could be conserved among different kingdoms of life. Thus, like some other Fe-S cluster-containing GRX (Lillig et al., 2005;Klinge et al., 2007;Lill, 2009;Berndt and Lillig, 2017), GRXS17 uses this oxidation-induced switch to activate a different functionality, here protecting plants from heat-induced damage.

Plant Materials and Growth Conditions
The Arabidopsis (Arabidopsis thaliana) grxs17 (SALK_021301) mutant in ecotype Col-0) was previously characterized (Knuesting et al., 2015). Arabidopsis seeds were surface sterilized for 15 min with 70% (v/v) ethanol, rinsed with 95% (v/v) ethanol, and completely dried before use. Sterilized seeds were plated onto one-half strength Murashige and Skoog (MS) agar medium (2.2 g L 21 MS, 0.5 g L 21 MES, and 1% [w/v] plant agar, adjusted to pH 5.8) and stratified 48 h at 4°C in the dark. Seedlings were grown under 120 mmol m 22 s 21 photosynthetic flux at 20°C and a 16-h-light/8-h-dark cycle. Those growth conditions were systematically used unless otherwise indicated.

Gene Cloning, Mutagenesis, and Plasmid Construction
For the construction of a transgenic complemented line of the grxs17 mutant and the production of recombinant proteins, cDNA of GRXS17 (At4g04950) was inserted into the pGEM-T Easy plasmid (Promega). Different point mutations of Cys residues to Ser in GRXS17 were generated using the QuikChange II Directed Mutagenesis Kit (Agilent) using primers detailed in Supplemental  Table S2: GRXS17-C33S and GRXS17-C33S/C179S/C309S/C416S.

Transformation of Plants
Plasmids were transferred into Arabidopsis via Agrobacterium tumefaciens strain GV3101 using the described methods (Clough and Bent, 1998). Transformants were selected on one-half strength MS medium plates containing 50 mg mL -1 hygromycin and checked by PCR using appropriate primers (Supplemental Table S2).

Production and Purification of Recombinant Proteins
All proteins were produced by induction in Escherichia coli BL21(DE3). Cultures of 2 L of an ampicillin-resistant (100 mg mL -1 ) colony were grown at 37°C and induced by 100 mM isopropyl-b-D-galactopyranoside in the exponential phase (optical density 5 0.5). Bacteria were harvested by centrifugation 3 h after induction and stored at 280°C before protein extraction. Pellets were resuspended in buffer (50 mM Tris-HCl, pH 7, 100 mM NaCl, and one tablet of antiprotease cocktail [Roche] for 25 mL), extracted using a cell disrupter (Constant Systems), and purified with a commercial kit according to the manufacturer's instructions (His-bind buffer kit; Millipore).
For the GRXS17 reduction assay by the TRX system, 5 mM recombinant GRXS17 protein was incubated in Tris-HCl buffer (50 mM, pH 7) for 1 h at 30°C in the presence or absence of NADPH (150 mM), NTRA (0.8 mM) and TRXh5 (0.08 and 0.8 mM) from Arabidopsis (Huang et al., 2018) and DTT (50 mM). Reactions were stopped by the addition of nonreducing Laemmli buffer, followed by SDS-PAGE gel migration.

In Vitro Characterization of Holo-GRXS17
Chemical reconstitution of holo-GRXS17 was performed under strictly anaerobic conditions in an anaerobic vinyl tent (COY Laboratory Products) according to Freibert et al. (2018). In brief, 100 mM GRXS17 was reduced in reconstitution buffer (50 mM pH 8,150 mM NaCl,and 10% [v/v] glycerol) in the presence of 2 mM GSH and 2 mM DTT for 3 h at room temperature. For reconstitution, the GRXS17 concentration was diluted to 20 mM and ferric ammonium citrate was added to a final concentration of 400 mM. After 5 min of incubation, lithium sulfide (400 mM) was added followed by a 3-h incubation. Excess Fe ammonium citrate and lithium sulfide were removed by desalting using a NAP 25 column (GE Healthcare). Subsequently, the Fe-S cluster was characterized using a CD spectrophotometer (J-815; Jasco). The reconstituted GRXS17 was transferred to a sealed CD cuvette. Oxidative treatment of the holo-GRXS17 was performed with 21% (v/v) oxygen for the indicated time.
Analytical SEC was carried out under strict anaerobic conditions. Gel filtration was performed with the ÄKTA Prime Plus System (GE Healthcare) coupled to a Superdex200 10/300 GL column (GE Healthcare). In each case, 50 mg of GRXS17 was applied into the equilibrated column (50 mM pH 8,150 mM NaCl,and 10% [v/v] glycerol) and separated at a flow rate of 0.5 mL min 21 . The column was calibrated with an M r standard (29-669 kD; Sigma-Aldrich).

Holdase and Foldase Activities
The holdase chaperone was examined using MDH or CS assays. The cytMDH1 from Arabidopsis (Huang et al., 2018) was used instead of commercial MDH, and the protocol from Lee et al. (2009) was slightly adapted with a temperature of 55°C.

Confocal Laser-Scanning Microscopy, Cell Death Observation, and Root Architecture Measurements
Confocal microscopic observations were carried out using the Axio observer Z1 microscope with the LSM 700 scanning module and ZEN 2010 software (Zeiss). GFP and PI were excited using the 488-nm argon-ion laser and collected at 500 to 550 nm (GFP) or 600 to 656 nm (PI). Excitation of roGFP2 was performed at 488 and 405 nm, and a bandpass (BP 490-to 555-nm) emission filter was used to collect roGFP2 signal. For background subtraction, signal was recovered using a BP 420-to 480-nm emission filter during excitation at 405 nm. Image analyses and quantification were performed as previously described (Schwarzländer et al., 2008) using the public domain image-analysis program ImageJ 1.52i (https://imagej.net). For measurement of cell death in roots, seedlings were stained with 100 mg mL -1 PI (Sigma-Aldrich) for 5 min before imaging.
Measurements of RAM length and LRP counting were performed by confocal microscopic observations after PI staining. RAM length was determined by the distance from the quiescent center to the region where endodermis cells were elongating (Truernit and Haseloff, 2008). Measurements of primary root elongation was performed using the NeuronJ software as described (Trujillo-Hernandez et al., 2020). Counting of LR numbers was performed using a stereomicroscope (SMZ18; Nikon). LR and LRP density were reported for the entire root length.

SEC
Gel filtration experiments were performed using an ÄKTA fast-protein liquid chromatography system (Amersham Biosciences) at 4°C. Absorbance was monitored at 280 nm. The column was calibrated using an M r standard (29-669 kD). For recombinant GRXS17 (wild-type and mutated forms), a Superose 6 column was used, equilibrated with a 50 mM Tris-HCl (pH 7.5) buffer containing 150 mM NaCl and 5 mM MgCl 2 at a flow rate of 0.5 mL min -1 . All recombinant protein samples were dialyzed before application. For reducing conditions, samples were incubated with 100 mM DTT and dialyzed against the above-described buffer supplemented with 5 mM DTT to maintain the proteins in their reduced state.
For separation of total protein extracts, a HiPrep column (Sephacryl S-300 HR; GE Healthcare Life Sciences) was used and equilibrated with a 50 mM Tris-HCl (pH 7.5) buffer containing 150 mM NaCl, 5 mM MgCl 2 , 10% (v/v) glycerol, and 0.1% (v/v) Nonidet P-40 at a flow rate of 0.5 mL min -1 . Samples were extracted in the same buffer containing 10 mM MG132 and one tablet per 10 mL of antiprotease cocktail (Roche). Extraction was carried out with a ratio of 1 g of plant powder in 2 volumes of buffer.

Protein Separation by Gel Electrophoresis and Immunoblotting
Precast SDS-PAGE or native gels were used according to the manufacturer's instructions (Bio-Rad). For immunoblot analysis, gels were transferred to nitrocellulose membranes. Rabbit polyclonal antibodies against GRXS17 (1:25,000) were used for protein immunoblotting. Goat anti-rabbit antibodies conjugated to horseradish peroxidase (1:10,000) were used as secondary antibodies and revealed with enhanced chemiluminescence reagents (Immobilon Western Chemiluminescent HRP Substrate; Millipore).

Thermotolerance Assays
For the thermotolerance assays, three different protocols were applied: LAT, SAT, and TMHT, as defined by Yeh et al. (2012). The seeds were germinated and grown on horizontal plates containing MS medium (4.4 g L -1 MS and 1% [w/v] Suc, pH 5.8). The percentage of viability was calculated by counting the plants that recovered leaf growth after heat stress. For examining the impact of heat stress on roots, seeds were grown on vertical plates, and root growth was daily measured using ImageJ.

Immunoprecipitation
Proteins from plants expressing the Pr35S:GRXS17-Flag/HA or the Pr35S:GRXS17 construct were used for immunoprecipitation experiments. Seeds were germinated and grown on vertical plates containing one-half strength MS medium during 14 d at 20°C and treated or not for 2 h at 35°C. The proteins were extracted and submitted to immunoprecipitation. Agarose beads with covalently attached anti-Flag antibody were used according to the manufacturer's instructions, and bound proteins were eluted with 0.1 M Gly, pH 2.5. In another set of experiments, protein extracts were first submitted to SEC as described previously. Fractions 18 to 21 corresponding to oligomers for plants grown at 20°C and fractions 16 to 19 for plants treated for 2 h at 35°C (Fig. 2) were collected and subjected to immunoprecipitation.

Chemicals and Enzymes
Proteomics-grade trypsin and ProteaseMax surfactant were purchased from Promega. Reverse-phase C18 spin columns, precolumns, and analytical columns were all obtained from Thermo Fisher Scientific. Solvents and ion-pairing agents were certified liquid chromatography-mass spectrometry grade, and all other chemicals were purchased from Sigma-Aldrich with the highest purity available. carbamidomethylated by adding 11 mM iodoacetamide for 30 min at room temperature in darkness. Proteins were digested overnight at 35°C with modified porcine trypsin in a 1:50 (w/w) enzyme:substrate ratio. The digestion was stopped by the addition of 0.1% trifluoroacetic acid, and peptide mixtures were centrifuged for 10 min at 21,500g at 4°C. Tryptic peptides present in supernatants were then subjected to desalting using reverse-phase C18 spin columns as recommended by the supplier. After elution, desalted peptides were concentrated with a SpeedVac (Eppendorf), and 3% (v/v) acetonitrile and 0.1% (v/v) formic acid (solvent A) were added to obtain a final protein concentration of 0.27 mg mL -1 based on the initial concentration.

Mass Spectrometry
Peptide mixtures were analyzed on a Q-Exactive Plus (Thermo Fisher Scientific) coupled to a Proxeon Easy nLC 1000 reverse-phase chromatography system (Thermo Fisher Scientific) using a binary solvent system consisting of solvent A and solvent B (0.1% formic acid in acetonitrile). Three microliters (800 ng) of tryptic digests was loaded on an Acclaim Pepmap C18 precolumn (2 cm 3 75 mm internal diameter, 2 mm, 100 A) equilibrated in solvent A, and peptides were separated on an Acclaim Pepmap C18 analytical column (25 cm 3 75 mm internal diameter, 2 mm, 100 A) at a constant flow rate of 250 nL min -1 by two successive linear gradients of solvent B from 0% to 25% in 100 min, from 25% to 40% in 20 min, and then up to 85% in 2 min followed by an isocratic step at 85% for 7 min. The instrument was operated in positive and data-dependent acquisition modes with survey scans acquired at a resolution of 70,000 (at mass-to-charge ratio [m/z] 200) with a mass range of m/z 375 to 1,400. After each full-scan mass spectrometry step, up to 10 of the most intense precursor ions (except 11, 15 to 18, and unassigned charge state ions) were fragmented in the higher-energy collisional dissociation cell (normalized collision energy fixed at 27) and then dynamically excluded for 60 s. The automatic gain control target was fixed to 3 3 10 6 ions in mass spectrometry and 10 5 ions in tandem mass spectrometry, with a maximum ion accumulation time set to 100 ms for mass spectrometry and tandem mass spectrometry acquisitions. All other parameters were set as follows: capillary temperature, 250°C; S-lens radio frequency level, 60; isolation window, 2 D. Acquisitions were performed with Excalibur software (Thermo Fisher Scientific), and to improve the mass accuracy of full-scan MS spectra, a permanent recalibration of the instrument was allowed using polycyclodimethylsiloxane (m/z 445.12003 D) as lock mass.

Proteomic Data Processing
Mass spectrometry raw data of HMW complexes were processed with ProteomeDiscoverer 2.2 (Thermo Fisher Scientific) and searched against the UniProtKB Arabidopsis database (October 25, 2015; 31,480 entries) combined with a database of classical contaminants using an in-house Mascot search server (Matrix Science; version 2.4). Mass tolerance was set to 10 ppm for the parent ion mass and 20 mmu for fragments, and up to two missed cleavages per tryptic peptide were allowed. Met oxidation, deamidation of Asn, and N-terminal acetylation were taken into account as variable modifications and Cys carbamidomethylation as a fixed modification. Peptide and protein false discovery rates (FDRs) were determined by searching against a reversed decoy database. Peptide identifications were filtered at 1% FDR using the Percolator node. Proteins were filtered at 1% FDR, and reverse and contaminant proteins were removed.
Mass spectrometry raw data were processed with MaxQuant software (version 1.6.0.13) and with the associated Andromeda search engine using the UniProtKB Arabidopsis database (October 25, 2015; 31,480 entries) combined with the MaxQuant database of common contaminants. First and main searches were performed with precursor mass tolerances of 20 and 4.5 ppm, respectively. The minimum peptide length was set to seven amino acids, and specificity for protein digestion was restricted to trypsin/P cleavage with a maximum of two missed cleavages per peptide. Met oxidation and N-terminal acetylation were specified as variable modifications, and carbamidomethylation of Cys residues was specified as a fixed modification. The peptide and protein FDR was set to 1%. To increase protein identification, a match between runs was systematically performed (match time window, 0.7 min; alignment time window, 20 min), and for label-free quantification of immunoprecipitated proteins, the label-free quantification algorithm and normalization were chosen and unique and razor peptides were used.
MaxQuant output was further processed and analyzed using Perseus software (version 1.6.07). Protein quantification was performed using the Pro-teinGroups.txt file after removing reverse and contaminant proteins and the log 2 transformation of quantitative data. Proteins showing at least two unique and razor peptides and having quantitative data in at least three biological replicates were considered for statistical analysis of each temperature condition using a Benjamini-Hochberg-corrected Student's t test. Proteins showing FDR , 0.05 were considered significantly enriched. For both temperature conditions, proteins having quantitative data in biological samples expressing the tagged GRXS17 but not in the five corresponding control samples were treated separately. In this case, only proteins showing quantitative data in three out of the five biological replicates with at least two unique and razor peptides from samples expressing tagged GRXS17 were considered as significant.

CD
In preparation for CD analysis, samples of GRXS17 were buffer exchanged into 20 mM sodium phosphate and 100 mM NaF, pH 7.3, using Bio-Spin 6 columns, adjusted to 4 mM, and treated with 400 mM either tris(2-carboxyethyl) phosphine or H 2 O 2 for 1 h at 25°C. Samples were centrifuged and vacuum degassed for 10 min prior to analysis by CD. CD spectra were collected at 20°C using a J-715 spectropolarimeter (Jasco) with a 0.1-mm path-length cuvette, from a range of 250 to 185 nm in 1-nm sampling, and spectrum averaged over eight repeats. The spectra of the buffer alone with tris(2-carboxyethyl)phosphine/H 2 O 2 were subtracted prior to the conversion of data to units of mean residue ellipticity. Secondary structure content was estimated using the online server BeStSel (Micsonai et al., 2015). For monitoring change in secondary structure over time, CD spectra of 250 to 190 nm were collected at 100 nm min 2 1 . The secondary structure change of GRXS17 following the addition of a 100fold excess of H 2 O 2 was measured as the change in molar ellipticity at 222 nm over a course of 65 min, with spectra collected every 41 s. In Prism (GraphPad), a one-phase decay model was fitted to these data, and a rate of conformational change was determined.

Supplemental Data
The following supplemental materials are available.
Supplemental Figure S1. Alignment of type II GRX of Arabidopsis and yeast.
Supplemental Figure S2. Fe-S cluster stability under increasing temperatures.

Supplemental
Supplemental Figure S5. GRXS17 foldase chaperone activity is independent of ATP.
Supplemental Figure S7. Primary root growth under 35°C heat stress.
Supplemental Figure S8. Determination of the RAM size at 20°C.
Supplemental Figure S9. Cell death in the RAM of grxs17.
Supplemental Table S1. Full list of proteins identified by mass spectrometry after coimmunoprecipitation experiments using either HMW complexes or total extracts from Arabidopsis plants as starting materials.
Supplemental Table S2. Primers used in this study.