The role of glutathione in periplasmic redox homeostasis and oxidative protein folding in Escherichia coli

The thiol redox balance in the periplasm of E. coli depends on the DsbA/B pair for oxidative power and the DsbC/D system as its complement for isomerization of non-native disulfides. While the standard redox potentials of those systems are known, the in vivo redox potential imposed onto protein thiol disulfide pairs in the periplasm remains unknown. Here, we used genetically encoded redox probes (roGFP2 and roGFP-iL), targeted to the periplasm, to directly probe the thiol redox homeostasis in this compartment. These probes contain two cysteine residues, that are virtually completely reduced in the cytoplasm, but once exported into the periplasm, can form a disulfide bond, a process that can be monitored by fluorescence spectroscopy. Even in the absence of DsbA, roGFP2, exported to the periplasm, was fully oxidized, suggesting the presence of an alternative system for the introduction of disulfide bonds into exported proteins. However, the absence of DsbA shifted the periplasmic thiol-redox potential from -228 mV to a more reducing -243 mV and the capacity to re-oxidize periplasmic roGFP2 after a reductive pulse was significantly decreased. Re-oxidation in a DsbA strain could be fully restored by exogenous oxidized glutathione (GSSG), while reduced GSH accelerated re-oxidation of roGFP2 in the WT. In line, a strain devoid of endogenous glutathione showed a more reducing periplasm, and was significantly worse in oxidatively folding PhoA, a native periplasmic protein and substrate of the oxidative folding machinery. PhoA oxidative folding could be enhanced by the addition of exogenous GSSG in the WT and fully restored in a ΔdsbA mutant. Taken together this suggests the presence of an auxiliary, glutathione-dependent thiol-oxidation system in the bacterial periplasm.


Introduction 33
Extracellular proteins are stabilized by structural disulfide bonds that are introduced during 34 oxidative protein folding. Thus, the correct folding of most of the proteins translocated into 35 the periplasm and beyond in the model bacterium Escherichia coli depends on the formation 36 of native disulfide bonds. The major pathway for introducing disulfide bonds in the periplasm 37 of E. coli involves the DsbA-DsbB thiol oxidase system (Bardwell et al., , 1991Collet 38 and Bardwell, 2002;Missiakas et al., 1993;. DsbA 39 contains a thioredoxin-like domain including an active site CxxC motif with a standard redox 40 potential of around -122 mV . After disulfide 41 introduction, DsbA itself is re-oxidized by the integral membrane protein DsbB, which in turn 42 is connected to the ubiquinone pool of the bacterial plasma membrane (Christensen et  proteins contain multiple cysteine residues, cells require quality control systems to isomerize 45 non-native disulfide bonds. In E. coli, this system includes the disulfide isomerase DsbC/G 46 and the reductive power from DsbD, which itself is coupled to the cytoplasmic NADPH pool 47 through Thioredoxin A (TrxA) (Andersen et al., 1997;Cho and Collet, 2013;Missiakas et al., 48 1994;Rietsch et al., 1997;Shevchik et al., 1994;Sone et al., 1997). 49 In eukaryotes, oxidative folding in the endoplasmic reticulum (ER) is catalyzed in a similar 50 manner by protein disulfide isomerase(s) (PDI), which also contains thioredoxin-like 51 domains. Unlike DsbA, PDI not only acts as thiol oxidase but also functions as a disulfide 52 reductase/isomerase (Ali Khan and Mutus, 2014; Ushioda and Nagata, 2019). PDI is re-53 oxidized for example by the sulfhydryl oxidase Endoplasmic reticulum oxidoreductin 1 54 (ERO1-Lα) (Tu and Weissman, 2004). 55 The tripeptide glutathione (GSH) is the most abundant low molecular weight thiol in many 56 domains of life ranging from bacteria to eukaryotes (Fahey et al., 1978;Smirnova and 57 Here, we explore the role of GSH in oxidative protein folding and redox homeostasis in the 84 periplasm of E. coli by targeting genetically encoded redox-sensitive probes with an 85 engineered disulfide bond to this compartment and analyzing the periplasmic redox dynamics 86 in the presence or absence of DsbA and/or GSH. We provide evidence that GSH indeed plays 87 a key role in disulfide bond formation and redox homeostasis in the periplasm: We observed 88 that GSSG can complement for the loss of DsbA and, unexpectedly, that GSH accelerates 89 disulfide bond formation in the presence of DsbA. In line, the periplasmic redox balance of 90 GSH-deficient cells was slightly shifted to a more reducing redox potential and the oxidative 91 folding of native DsbA substrates in these cells is impaired. 92

Strains, Plasmids and growth conditions 94
Bacterial strains and plasmids used in this study are listed in Supplementary tables S1 and S2. 95 Escherichia coli DH5α served as host for plasmid construction and storage and E. coli BL21 96 (DE3) was used for DsbA and E. coli MG1655 for roGFP2 recombinant protein production. 97 All E. coli strains used in this study were routinely cultivated at 37 °C in Luria-Bertani (LB) 98 medium, supplemented with antibiotics when required for plasmid maintenance and marker 99 selection (ampicillin 200 µg/mL or kanamycin 100 µg/mL), if not stated differently.      1A). We verified translocation of roGFP2 into 322 the periplasm using fluorescence microscopy (Fig. 1B, left panel). Determination of the 323 roGFP2 oxidation state (OxD) revealed complete oxidation of the probe in the periplasm of 324 E. coli WT (Fig. 1C, left panel). As the DsbA/DsbB pair is the major system for oxidative 325 protein folding in the periplasm of E. coli (Collet and Bardwell, 2002), roGFP2 was also 326 targeted to the periplasm of cells lacking DsbA (Fig. 1B, right panel). Surprisingly however, 327 periplasmic roGFP2 was still fully oxidized in the absence of DsbA (Fig. 1C, right panel), 328 suggesting the presence of an alternative mechanism for the introduction of disulfide bonds in 329 this compartment. In both cases, due to the complete oxidation of the probe, we were not able 330 to determine the redox potential for the roGFP2 dithiol disulfide couple. 331 The redox-potential of the periplasm is significantly more reducing in cells lacking DsbA 332 In order to determine the periplasmic redox potential exerted on proteins containing cysteines, 333 we turned to roGFP variants with a more oxidizing standard redox potential ( Thus, roGFP-iL with a midpoint potential of -229 mV was expressed and targeted to the 341 periplasm of the WT and ΔdsbA strain (Fig. 1A, B, right panel). In contrast to roGFP2, 342 roGFP-iL was only around 50 % oxidized in the WT periplasm (Fig. 1B, left panel), allowing 343 the calculation of the redox-potential using the Nernst equation as described before (Xie et al., 344 2020). In the WT the redox-potential of roGFP-iL in the periplasm was -228 mV (Fig. 1E,  345 left panel). Not surprisingly, the lack of DsbA resulted in a more reduced roGFP-iL probe 346 (Fig. 1E, right panel). The shift in the redox-potential was around 14 mV to -243 mV, 347 supporting the role of DsbA in oxidative protein folding, however, the relatively small size of 348 the shift suggests the presence of an alternative mechanism for the introduction of disulfide 349

bonds. 350
Periplasmic thiol oxidation is significantly impaired but not abrogated in a 370 periplasmic roGFP2 after a reductive pulse. To explore the role of DsbA in this process, we 371 also analyzed the re-oxidation of roGFP2 in cells lacking this oxidoreductase. In these 372 experiments, we treated WT and cells lacking DsbA cells, both expressing roGFP2 in the 373 periplasm with a pulse of DTT. This was followed by the removal of the reductant and the 374 recording of the oxidation state of roGFP2 over time ( Fig. 2A). In the WT, periplasmic 375 roGFP2 oxidation starts immediately after DTT removal and restores maximum oxidation 376 within 150 min. Cells that lack DsbA also showed sensor re-oxidation, however, roGFP2 was 377 not restored to maximum oxidation in the time frame of our measurement (Fig. 2B). 378 Additionally, the re-oxidation rate, calculated as the linear slope after DTT removal, was 379 significantly lower in DsbA-deficient cells, however, there was still re-oxidation capacity 380 present (Fig. 2C). These findings suggest that there is an additional factor playing a role in 381 periplasmic redox homeostasis, other than DsbA. To further explore our hypothesis, we used 382 cell lysates derived from E. coli WT and DsbA-deficient cells and explored their capacity to 383 re-oxidize DTT-reduced, purified roGFP2 (Fig. 2D). Similar to the cell re-oxidation assay, we 384 calculated the oxidation rate from the linear slope within the first two hours of measurement. 385 Although roGFP2 oxidation by cell lysate was slow, when compared to intact cells, lysate 386 from WT was significantly faster in catalyzing roGFP2 re-oxidation than a ∆ dsbA lysate. 387 However, in line with our previous findings, indicating another factor capable of catalyzing 388 disulfide bond formation in the periplasm, the ∆ dsbA lysate was still significantly faster than a 389 buffer control (Fig. 2E, F). The rather slow oxidation of roGFP2 in vitro may be explained by 390 the fact that DsbA preferably introduces disulfide bonds in unfolded proteins entering the 391 periplasm over folded ones (Kadokura et al., 2004). 392   We thus analyzed the periplasmic capacity for re-oxidizing roGFP2 in the presence of 419 glutathione after a reductive pulse. E. coli WT and DsbA-deficient cells producing 420 periplasmic roGFP2 were thus reduced with DTT, and, after reductant removal, roGFP2 421 oxidation was recorded in the presence of 5 mM GSH or GSSG (Fig. 3A). Surprisingly, 422 addition of GSH to WT cells resulted in an accelerated roGFP2 re-oxidation in the periplasm, 423 whereas GSSG supplementation had no discernible influence (Fig. 3B). In contrast, in cells 424 that lack DsbA, GSH and GSSG acted in a more expected way; while GSSG supplementation 425 rescued roGFP2 re-oxidation, GSH had no influence on the periplasmic redox dynamics in 426 ΔdsbA (Fig. 3C). Calculation of the re-oxidation rate confirmed that while GSH significantly 427 speeds up re-oxidation of periplasmic roGFP2 in the WT, GSSG significantly accelerates the 428 re-oxidation rate of roGFP2 in DsbA-deficient cells, essentially to WT level (Fig. 3D). DsbA is rather slow compared to in vivo re-oxidation, although significantly higher compared 450 to buffer alone (Fig. 4A, B). The oxidation rate of roGFP2 in the presence of GSSG alone was 451 slightly slower, but still comparable to roGFP2 oxidation by DsbA. Adding both GSSG and 452 DsbA at the same time did not even double the probe's oxidation rate, indicating an additive 453 effect of GSSG and DsbA and not a GSSG-driven catalytic action of DsbA. 454 To further confirm the incapability of DsbA to perform GSSG-dependent roGFP2 oxidation, 455 we used the cytosol of genetically manipulated yeast cells as "cellular test tubes". This assay 456 is performed in yeast cells, lacking the glutathione reductase (Glr1) and both cytosolic class I 457 dithiol glutaredoxins (Grx1, Grx2), while simultaneously expressing Opt1, a glutathione 458 transporter . This GSSG. In this assay, GSSG-driven roGFP2 oxidation did not depend on the oxidase DsbA 465 (Fig. 4C, D). However, we cannot exclude that roGFP2 is an inappropriate substrate of DsbA 466 in the fusion construct and therefore roGFP2 may not be oxidized by DsbA in this assay. In 467 contrast, the roGFP2-HsGrx fusion probe strongly reacted to the addition of GSSG (Fig. 4E). 468 Both the in vitro and the "cellular test tube" approach strongly suggest that DsbA does not 469 interact with glutathione itself, indicating that the role of glutathione in the periplasmic redox 470 homeostasis is independent of the known mechanism for disulfide bond formation in E. coli. 471 While we excluded the interaction of GSSG with DsbA, we showed that the addition of GSH 487 accelerated the re-oxidation rate of periplasmic roGFP2 after a reductive pulse in a DsbA-488 dependent manner (Fig. 3B, D). To investigate whether this effect is limited to GSH, we 489 tested the influence of other reduced monothiols and dithiols. Cystein, β-mercaptoethanol, 490 and DTT were added to E. coli WT and ΔdsbA cells and roGFP2 re-oxidation dynamics in the 491 periplasm were measured as described above. The addition of the monothiols cysteine and β-492 mercaptoethanol to WT significantly accelerated roGFP2 oxidation similar to GSH, as 493 measured by the time after which the probe reached full oxidation. In contrast to monothiols, 494 addition of the dithiol DTT to WT completely inhibited roGFP2 re-oxidation. We also 495 included cystine in our assay, the oxidized form of the amino acid cysteine. In contrast to 496 GSSG, which had no impact on the roGFP2 oxidation rate in WT, cystine massively reduced 497 the time until full oxidation was reached (Fig. 5A, B). In the ΔdsbA mutant, the presence of 498 monothiols had no significant impact, while cystine supplementation caused a drastically 499 increased re-oxidation state and rate, similar to WT (Fig. 5C, D). Overall, these findings 500 suggest a DsbA-dependent effect of monothiols, accelerating thiol oxidation in the periplasm, 501 even though we did not observe direct interaction of oxidized glutathione with DsbA itself. 502

521
Endogenous glutathione is involved in stabilizing and maintaining the periplasmic redox state 522 All our experiments thus far were performed with exogenous glutathione. But we also 523 wondered about the role of endogenous glutathione, synthesized in E. coli's cytoplasm. For 524 this, we determined the oxidation state of periplasmic roGFP-iL in cells lacking GshA, the 525 first enzyme of E. coli's glutathione biosynthesis pathway. In glutathione-free media, these 526 cells do not contain GSH (Apontoweil and Berends, 1975;Carmel-Harel and Storz, 2000). 527 Confirming our observation with exogenous GSH, our experiment revealed a slightly, but 528 significantly lower oxidation of roGFP-iL in the periplasm of cells lacking GSH. The redox 529 potential shifted from around -228 mV (WT) to -233 mV, counterintuitively a more reducing 530 state, in the absence of GSH (Fig. 6A, B). Nevertheless, the shift was not as pronounced as in 531 cells lacking DsbA, which had a roGFP-iL redox potential of around -243 mV (Fig. 1E). 532 We also asked whether the presence of endogenous GSH leads to faster re-oxidation of 533 roGFP2. To address this question, we analyzed the capacity to restore the redox balance after 534 reductive challenge in GSH-depleted cells producing periplasmic roGFP2 as described before. 535 Our data indicates that roGFP2 oxidation rate and end oxidation state in the periplasm of 536 GSH-deficient cells was comparable to WT (Fig. 6C, D) suggesting GSH is not essential for 537 recovery after a reductive challenge. 538 Next, we asked what happens when both, DsbA and GSH are missing. The redox state of 539 roGFP-iL in a ∆ gshA∆dsbA strain showed a periplasmic redox potential imposed on the 540 sensor (ca. -244 mV) similar to the redox potential in the dsbA single mutant (ca. -243 mV) 541 ( Fig. 1, Fig. 6B). Intriguingly however, using a ∆ gshA∆dsbA strain in a periplasmic roGFP2 542 re-oxidation assay revealed that exogenous GSSG did not completely restore WT oxidation 543 rate and final oxidation state (Fig. 6E, F) contrary to cells lacking solely DsbA (Fig. 3). 544 Adding GSH to a ∆ gshA∆dsbA strain even decelerated periplasmic roGFP2 oxidation, 545 something we did not observe in the ∆ dsbA single mutant (Fig 3).

Endogenous glutathione is involved in disulfide bond formation in E. coli's own periplasmic 573
proteins 574 The observed influence of GSH on growth phenotypes (Supplementary figure 3) suggests that 575 our previous observations of disulfide bond formation in heterologously expressed roGFP-576 based redox sensors also apply to endogenous DsbA substrates. One well-characterized 577 substrate of DsbA is alkaline phosphatase PhoA. PhoA is only active upon formation of 578 intramolecular disulfide bonds essential for correct protein folding. The activity of this 579 enzyme can be measured in a colorimetric assay using the substrate para-580 nitrophenolphosphate ( Fig. 7A) (Berg, 1981;Brickman and Beckwith, 1975;Sone et al., 581 1997). We thus tested the activity of alkaline phosphatase PhoA in different E. coli deletion 582 strains grown in MOPS minimal medium. As described above, the lack of GSH, although 583 without effect on re-oxidation of roGFP2 in the periplasm, slightly shifted the periplasmic 584 redox homeostasis to more reducing conditions, as seen in the lowered redox potential of 585 roGFP-iL in those cells (Fig. 6). In accordance with this, PhoA activity was slightly, but 586 significantly lowered in GSH-deficient cells compared to WT. Also, in line with our previous 587 results (Fig. 1), cells lacking DsbA showed only poor PhoA activity, and the same was true 588 for cells lacking both DsbA and GSH (Fig. 7B). 589 To test whether the observed reduction in PhoA activity is dependent on glutathione levels in 590 the periplasm, we also measured PhoA activity in a cydD mutant. This mutant can still 591 synthesize glutathione, but is not able to produce the inner membrane ABC-transporter 592 CydDC that transports glutathione from the cytosol into the periplasm and hence suffers from 593 reduced periplasmic glutathione levels (Mironov et al., 2020;Pittman et al., 2005;Shepherd, 594 2015). Again, we observed a slight, but significant reduction of PhoA activity in this strain, 595 although less pronounced than in cells completely lacking glutathione biosynthesis (Fig. 7B). 596 Conversely, blocking glutathione transport from the periplasm into the cytosol by deleting 597 GsiC, the inner membrane component of the GsiA-D ABC Transporter (Suzuki et al., 2005;598 Wang et al., 2017598 Wang et al., , 2018 did not result in significant changes in PhoA activity compared to 599 the WT. 600 Next, we supplemented the growth medium with exogenous GSSG to test if, in accordance 601 with the re-oxidation assays (Fig. 3)  addition restored PhoA activity back to WT level (Fig. 7C). Furthermore, addition of 604 exogenous GSSG also increased PhoA activity in the WT, contrary to what we observed with 605 roGFP2, suggesting that the role of periplasmic glutathione in oxidative folding differs, 606 depending on the protein in question (Fig. 3, Supplementary figure 3). 607 Proteins that are not correctly folded are usually unstable in the periplasm (Hiniker and 608 , and thus no PhoA protein could be detected on a western blot in DsbA-609 deficient cells (Fig. 7D). External GSSG addition prevented PhoA from degradation in a 610 ∆ dsbA mutant. 611 Taken together, our data indicates a significant role for glutathione, not only in oxidative 612 protein folding of periplasmic proteins, but also by balancing and maintaining the periplasmic 613 redox homeostasis. 614 The DsbA/B system is the major thiol oxidation system in the periplasm of E. coli and 631 together with the DsbC/D thiol disulfide isomerase system forms the oxidative folding 632 machinery (Collet and Bardwell, 2002;Manta et al., 2019). The standard redox potentials of 633 those systems are known Zapun et al., 1995), but the in vivo redox 634 potential seen by protein thiol disulfide pairs in the periplasm is unknown. Here, we used the 635 genetically encoded proteins roGFP2 and roGFP-iL targeted to the periplasm to directly probe 636 the thiol redox homeostasis in this compartment. When expressed in the cytosol of E. coli, the 637 probe roGFP2 with a midpoint redox potential of -280 mV is almost completely reduced 638 (Degrossoli et al., 2018). However, we found that it is virtually fully oxidized, when 639 expressed in the periplasm, suggesting that its engineered cysteine residues are completely 640 oxidized by the oxidative folding machinery. 641 To our surprise, roGFP2 was also fully oxidized in cells lacking the major thiol oxidase in the 642 periplasm, DsbA. We thus also analyzed the re-oxidation capacity after a reductive pulse, as 643 the oxidation state in an unperturbed cell solely reflects the steady state. And in the absence of 644 DsbA we did indeed find a significantly diminished re-oxidation velocity underlining the 645 importance of DsbA for oxidative folding. 646 Using roGFP-iL, we were able to determine that the redox potential imposed onto a thiol pair 647 in the periplasm is -229 mV. In the absence of DsbA, this redox potential shifted to a more 648 reducing -243 mV. In a previous study, Messens et al. used an Ag/AgCl electrode to measure 649 the redox potential in E. coli WT periplasmic extracts (Messens et al., 2007). Typically, an 650 electrode will measure the redox potential of all species it can chemically interact with and in 651 this case the redox potential was determined to be -165 mV. Unexpectedly, in their setting, 652 removal of DsbA shifted the redox potential to an even more oxidizing redox potential. Our 653 finding that the thiol disulfide redox potential is significantly below the overall redox 654 potential observed by Messens et al. with an electrode could explain their seemingly 655 paradoxical finding, since in a ∆ dsbA strain a component that is reducing in comparison to the 656 overall redox potential measured by the electrode is removed from the overall redox pool. In 657 our case, removal of DsbA did indeed result in an overall more reducing redox potential 658 imposed on thiol pairs. 659 In concordance with our results in the periplasm, roGFP2 targeted to the ER is fully oxidized, 660 as well. Similarly, a roGFP2-iL-Grx1 fusion redox probe targeted to the ER of Arabidopsis 661 thaliana cells lacking Ero1/2, the functional homolog of DsbB in eukaryotes was more 662 reduced, and re-oxidation after a reductive pulse was inhibited by the lack of Ero1/2 (Ugalde 663 et al., 2022). 664 While roGFP2 re-oxidation was diminished in ∆ dsbA, it was not absent, suggesting to us the 665 presence of an alternative pathway for disulfide bond formation. As glutathione is one of the 666 major redox buffers in cells, we assumed a possible role for the small molecule in periplasmic 667 redox homeostasis. Glutathione's cytosolic functions have been studied extensively and, until 668 a few years back, GSH was thought to be absent from the periplasm (Pittman et al., 2005). WT. Surprisingly, adding exogenous GSH to the WT accelerated roGFP2 re-oxidation, but 679 not in a ∆ dsbA strain. We also observed this seemingly paradoxical acceleration of thiol re-680 oxidation in WT by other reducing monothiols like cysteine or β-mercaptoethanol. However, 681 we did not observe direct interactions of DsbA with glutathione in vitro and in our yeast cell 682 experiments. 683 We next assessed the role of endogenous glutathione in the redox balance of the periplasm. 684 And in line with our observations with exogenous GSH, the oxidation state of roGFP-iL was 685 slightly shifted to a more reducing state in ∆ gshA, indicating a role for endogenous 686 glutathione in the periplasmic redox homeostasis as well. It should be noted, however, that 687 roGFP2 re-oxidation in a GSH-deficient mutant was comparable to WT. Taken together, the 688 presence of GSH and GSSG in the periplasm, driven by the availability of exogenous and 689 endogenous glutathione, seems to be important for the fine-tuning of the periplasmic redox 690 potential. 691 Expression of a non-native redox sensor might not reflect the natural redox state in the 692 periplasm or might even influence it by diverting oxidative power available for the formation 693 of native disulfide bonds. We thus analyzed the activity of PhoA, a native E. coli protein, 694 which depends on oxidation by DsbA for its activity, in different mutants. This approach 695 revealed that reduced periplasmic glutathione levels indeed resulted in significantly lower 696 PhoA activity, in line with the more reduced roGFP-iL redox state in ΔgshA. While PhoA's 697 oxidative folding is clearly influenced by the presence or absence of GSH, not all DsbA 698 substrates seem to be influenced by glutathione. RNaseI folding and isomerization by 699 DsbA/DsbC e.g., was not influenced by the loss of GSH (Messens et al., 2007). 700 In order to understand the role of glutathione in the periplasm, it is helpful to have a look at 701 the role of GSH in the eukaryotic ER, for which, in contrast to bacteria, more studies are 702 available. In the ER, the glutathione concentration is around 15 mM, higher than in whole cell 703 lysates with around 7 mM (Birk et al., 2013). Similar to the periplasm, the GSH:GSSG ratio 704 in the ER is lower (3:1 to 1:1) compared to the cytosol (100:1), resulting in a more oxidizing 705 environment (Hwang et al., 1992). It is discussed whether GSH is the reductive power for 706 disulfide isomerization by PDI (Vitu et al., 2010) and high GSSG levels could act as oxidant 707 reservoir; however, it is still unclear if GSSG itself is able to oxidize PDI (Lappi and 708 Ruddock, 2011; Ushioda and Nagata, 2019). In the periplasm of Gram-negative bacteria, 709 DsbB recycles DsbA, but in contrast to Ero1, it uses the respiratory chain as electron sink. As 710 aforementioned, we observed accelerated roGFP2 re-oxidation in the WT, but not in ∆ dsbA, 711 when adding monothiols to the cells and the opposite effect for GSSG, compensating the lack 712 of DsbA, raising the question whether DsbB is somehow regulated by glutathione. However, 713 analyzing a dsbB mutant strain regarding its oxidation state and re-oxidation capacity in 714 presence or absence of GSH or GSSG revealed that GSSG was still able to complement for 715 the loss of DsbB, indicating a DsbB-independent mechanism (Supplementary figure 4). 716 Overall, we showed that oxidized glutathione can compensate for the loss of DsbA by an 717 unknown mechanism. One possibility is that oxidized glutathione directly oxidizes roGFP2 718 and other reduced proteins, however previous studies  and the current in 719 vitro and yeast data show that direct roGFP2 oxidation by glutathione is very inefficient. 720 Another possibility is the presence of a yet unidentified redox factor in the periplasm that can 721 catalyze oxidative protein folding in the absence of DsbA. One possibility is DsbC and it has 722 been suggested that reduced glutathione can react with the isomerase, especially by providing 723 reductive power when DsbD is missing (Pittman et al., 2005;Smirnova et al., 2012). 724 However, (Messens et al., 2007) could also show that DsbC alone was not able to substitute 725 for DsbA in folding of RNaseI. In the ER up to 20 different oxidoreductases, for example the 726 peroxiredoxin Prx4 or the glutathione peroxidase-like enzymes Gpx7 and Gpx8 are found 727 besides PDI and for at least some of them it has been proposed that they possibly oxidize PDI 728 (Nguyen et al., 2011;Wang et al., 2018;Zito et al., 2010). We think it is possible, that E. coli 729 has a similar backup system for DsbA, presumably a glutaredoxin-like protein, which is 730 coupled to the periplasm's glutathione pool, providing either oxidizing or reducing power. 731 Taken together, our data underlines the importance of glutathione as a player in redox 732 homeostasis not only in the cytosol, but also in oxidative cellular compartments and it shows 733 that its role in oxidative protein folding did already evolve in bacteria. 734