Activation of RidA chaperone function by N-chlorination

Escherichia coli RidA is a member of a structurally conserved, yet functionally highly diverse protein family involved in translation inhibition (human), Hsp90-like chaperone activity (fruit fly) and enamine/imine deamination (Salmonella enterica). Here, we show that E. coli RidA modified with HOCl acts as a highly effective chaperone. Although activation of RidA is reversed by treatment with DTT, ascorbic acid, the thioredoxin system and glutathione, it is independent of cysteine modification. Instead, treatment with HOCl or chloramines decreases the amino group content of RidA by reversibly N-chlorinating positively charged residues. N-chlorination increases hydrophobicity of RidA and promotes binding to a wide spectrum of unfolded cytosolic proteins. Deletion of ridA results in an HOCl-sensitive phenotype. HOCl-mediated N-chlorination thus is a cysteine-independent post-translational modification that reversibly turns RidA into an effective chaperone holdase, which plays a crucial role in the protection of cytosolic proteins during oxidative stress.

H ypochlorous acid (HOCl) is not only a very potent disinfectant, but functions as a physiological oxidant released by neutrophils to fight infections (for a comprehensive overview see ref. 1). Once activated, neutrophils phagocytize pathogens and internalize them into the phagosome. During the oxidative burst, neutrophils then generate large amounts of superoxide anions (O 2 À ) and hydrogen peroxide (H 2 O 2 ) at the expense of oxygen by NADPH-oxidases. Myeloperoxidase converts this H 2 O 2 and chloride ions into HOCl, which together with the other products of the respiratory burst, is highly effective in killing microorganisms.
The mode of action of HOCl-dependent microbial killing involves reactions with a variety of biological molecules, including DNA, lipids, NADH, cholesterol, proteins and non-protein thiols such as glutathione 1 . Due to their high cellular abundance, however, proteins are considered to be the major targets of HOCl-mediated modifications. Damage caused by HOCl includes protein fragmentation, dimerization and side chain modification, typically leading to protein inactivation and degradation by the cellular protein quality control [2][3][4] . Methionine and cysteine residues react most rapidly with HOCl, while its reactivity with side chains of histidine, tryptophan, lysine, tyrosine and arginine is two to five orders of magnitude lower 4 .
As HOCl does not only target microbial proteins, but readily halogenates organic and inorganic primary amines present in the phagosome, chloramines such as monochloramine (NH 2 Cl) and N-chlorotaurine accumulate there as well 4,5 . Although being less reactive and more stable than HOCl, chloramines share the oxidizing and chlorinating properties of HOCl, making them equally useful as disinfectants in drinking water supplies, medicine and household settings 6 .
Pathogens have developed numerous strategies to successfully escape the immune response during infection (see ref. 7 for a comprehensive overview). Yet, HOCl-response mechanisms in bacteria remain largely elusive. In Firmicutes, such as Bacillus subtilis and Staphylococcus carnosus, inactivation of conserved metabolic enzymes and essential proteins such as methionine synthase and the redox-sensing repressor OhrR by S-bacillithiolation mediates the HOCl stress response 8,9 . Salmonella enterica specifically reduces the influx of HOCl through ArcA-dependent downregulation of the outer membrane porine OmpW in response to hypochlorite stress 10 . Escherichia coli responds to hypochlorite stress using a number of complementary strategies. Hydrogen peroxide and heat-shock response genes were shown to be regulated in response to hypochlorite stress 11 . The E. coli transcription factors NemR, HypT and RclR are involved in hypochlorite tolerance by the regulation of genes involved in electrophile detoxification, sulfur metabolism and iron acquisition, and survival upon chlorine stress, respectively [12][13][14][15] . Furthermore, the redox-regulated chaperone Hsp33 is activated by oxidative unfolding mediated by HOCl. The activated state of Hsp33 protects cellular proteins from HOCl-induced aggregation by direct protein-protein interaction under conditions that prevent ATP-dependent chaperones from working properly 16 .
In this study, we describe a novel mechanism, which confers protection of E. coli against HOCl stress-mediated protein aggregation. This mechanism is dependent on RidA, a member of the functionally highly diverse YjgF/YER057c/UK114 protein family. Although highly conserved across kingdoms, members from this protein family display very different functions. They are, for instance, involved in translation inhibition (rat), activation of calpain protease (cow), adenine-mediated repression of purine biosynthesis (Bacillus subtilis) and mitochondrial maintenance (Saccharomyces cerevisiae) [17][18][19][20] . The Salmonella enterica homologue RidA has recently been described as an enamine/imine deaminase, playing a crucial role in the synthesis of branched-chain amino acids 21 .
Here, we show that E. coli RidA acts additionally as redoxregulated chaperone that is activated on exposure to HOCl stress. Although many proteins are inactivated on reaction with HOCl and precipitate, RidA turns into a potent chaperone, preventing highly abundant cellular proteins from HOCl-induced aggregation. Surprisingly, unlike in the other known redox-regulated chaperone Hsp33, where activation involves reversible disulfide bond formation 22 , activation of RidA's chaperone activity is mediated by N-chlorination, a modification that has previously been thought to be deleterious to protein function. We also observe activation of chaperone function via N-chlorination in the Drosophila melanogaster RidA homologue DUK114, as well as in a few unrelated proteins, indicating that this is a novel posttranslational modification that induces affinity for aggregationsensitive folding intermediates. In addition to reduction by glutathione, thioredoxin is capable of reducing N-chlorinated RidA, a hitherto unknown function of this well-characterized thiol-disulfide oxidoreductase.

Results
HOCl-treated RidA inhibits threonine dehydratase activity. Salmonella enterica RidA has previously been reported to act as deaminase of reactive enamine/imine intermediates, stimulating IlvA-mediated degradation of threonine to 2-ketobutyrate 21 . We recently discovered that modification of a single cysteine residue in RidA by peroxynitrite treatment inhibits this stimulatory function 23 . We tested the effect of several other nitrosative and oxidative stressors on RidA's enamine/imine deaminase activity. Intriguingly, when treated with the physiological disinfectant sodium hypochlorite (HOCl), RidA not only lost its stimulatory effect in the threonine deaminase assay, but inhibited the overall rate of the reaction to levels below the ones observed in the absence of RidA (Fig. 1a). These results could no longer be explained by a simple inhibition of RidA alone but suggested that the threonine dehydratase IlvA is directly inhibited by HOCltreated RidA (RidA HOCl ), potentially through a tight direct protein-protein interaction of the two proteins. We, therefore, hypothesized that RidA, once treated with HOCl, might be a chaperone that effectively binds other proteins.
RidA is a chaperone activated by HOCl. To test RidA HOCl for chaperone activity, we performed aggregation assays using chemically denatured citrate synthase and untreated and HOCltreated RidA. RidA was incubated with a 10-fold molar excess of HOCl, a concentration that is sufficient to fully activate the welldescribed HOCl-activated chaperone Hsp33 (ref. 16). Aggregation of chemically denatured citrate synthase was prevented by RidA HOCl in a dose-dependent manner (Fig. 2a,b). Other oxidants including H 2 O 2 or diamide did not activate RidA's chaperone activity, suggesting that the activation was indeed HOCl-specific (Fig. 1b). A fivefold excess of RidA HOCl was sufficient to completely inhibit citrate synthase aggregation, suggesting that HOCl is a potent activator of RidA chaperone activity (Fig. 2b). These concentration ratios are similar to other chaperones, such as Hsp33. Analysis of the activation kinetics revealed that a 10:1 molar ratio of HOCl over RidA was sufficient for full chaperone activity within the first 5 s of incubation (Fig. 2c,d). The addition of ATP did not enhance chaperone activity, consistent with a holdase function of HOCl-treated RidA (Fig. 3j).
To determine the impact of HOCl treatment of RidA on its oligomeric state of RidA, we performed analytical gel filtration (Fig. 1c,d). For this purpose, untreated RidA and RidA HOCl were loaded on a Superdex 75 10/300 GL column to monitor the molecular weight of the protein complex. While untreated RidA eluted as a trimer as reported previously 23 (Fig. 1c), treatment of RidA with HOCl partially coincides with the formation of higher molecular weight oligomers (Fig. 1d). This points towards a tendency of RidA to homo-oligomerize when treated with HOCl. Oligomerization thus may be a requirement for binding of unfolding proteins under these conditions. RidA is activated by monochloramine. Within the phagosome, most of the produced HOCl readily reacts with free ammonia or amino acids resulting in the formation of chloramines such as monochloramine or N-chlorotaurine 1 . In fact, HOCl-derived chloramines are discussed to be the major contributors for the antimicrobial activity of HOCl 24,25 . Thus, we investigated whether monochloramine (NH 2 Cl; MCA) is also capable of activating the chaperone function of RidA. Indeed, monochloramine-treated RidA (RidA MCA ) also prevented citrate synthase aggregation (Fig. 2e,f). RidA activation was maximal with a 10:1 molar ratio of MCA over RidA (Fig. 2g). In contrast to HOCl, however, RidA activation by MCA proceeded at much lower rates and required at least 20 min incubation to achieve full chaperone activity (Fig. 2h). This finding may be explained by the lower absolute rate constants of the reaction of chloramines with biomolecules when compared with HOCl 26,27 .
Unfolded IlvA is a direct substrate of RidA. To assess whether RidA HOCl is capable of also interacting with IlvA, we established aggregation assays substituting citrate synthase with IlvA. We found that chemically denatured IlvA readily aggregates upon dilution into assay buffer (Fig. 2i). RidA HOCl prevented aggregation of IlvA in a concentration-dependent manner (Fig. 2i,j). In contrast, incubation of IlvA with a 10-fold molar excess of untreated RidA had no effect on IlvA's aggregation behaviour (Fig. 2i). IlvA thus is a bacterial protein suitable to be used as a substrate in chaperone assays.
RidA activation is reversible by reduction. To test whether the activation of RidA's chaperone activity is a reversible process, we tested chaperone activity of RidA MCA after treatment with the thiol-based reductant dithiothreitol (DTT). Indeed, DTT-reduced RidA HOCl and RidA MCA were no longer able to prevent citrate synthase aggregation (Fig. 3a-c). To analyse if the reduced RidA MCA can be activated again, we did another series of citrate synthase aggregation assays (Fig. 3b,c). RidA activity could be recovered by subsequent treatment with MCA (Fig. 3b,c). This demonstrates that modification by MCA followed by reduction with DTT does not compromise integrity and functionality of the protein. RidA thus is a novel redox-regulated protein activated by HOCl and chloramines.
We also performed thermal aggregation assays at 43°C using citrate synthase to analyse whether reduction of activated RidA results in substrate release (Fig. 3d). Citrate synthase did aggregate significantly slower in the presence of RidA HOCl . However, when DTT was added to the reaction mixture after 250 s, citrate synthase started to aggregate, indicating that reduction of RidA HOCl is accompanied by the release of its substrate.
RidA activation is reversible by thioredoxin and glutathione. In bacteria, the reductive redox state of cytosolic proteins is maintained by thioredoxins and glutaredoxins 28,29 . Both proteins have been shown to reduce disulfide-containing substrate proteins by direct thiol-disulfide exchange reactions. The disulfide bond, which is subsequently formed in thioredoxin and glutaredoxin is then reduced by the NADPH-dependent thioredoxin reductase or  glutathione, respectively 28 . In E. coli, the thioredoxin system is composed of thioredoxin A (TrxA) and thioredoxin reductase (TrxB).
To test whether E. coli TrxA is involved in the inactivation of RidA, we reconstituted the complete thioredoxin system comprising TrxA, TrxB and NADPH in vitro. Activity of TrxA ARTICLE was monitored by the consumption of NADPH (measured at 340 nm). In the absence of RidA HOCl , we did not observe any significant decrease in absorption (Fig. 3e). In contrast, however, addition of 50 mM of RidA HOCl to the reaction mixture resulted in the significant consumption of NADPH. Moreover, thioredoxindependent reduction of RidA caused a loss of chaperone activity as observed in citrate synthase aggregation assays (Fig. 3f,g). The overall velocity of thioredoxin-dependent reduction of RidA HOCl observed could be substantially different from a reduction velocity in vivo, as the amount of NADPH used in our experimental setup limits the reduction rates. To further elaborate if reduction by other thiol-based reducing systems present in the cell may occur, we incubated RidA HOCl with excess glutathione (Fig. 3h). While oxidized glutathione (GSSG) did not have any impact on RidA HOCl activity, the addition of reduced glutathione (GSH) blocked its holdase function.
Loss of free amino groups by HOCl treatment. On the basis of our previous observation that RidA is oxidized at its single cysteine residue (C107) upon peroxynitrite treatment 23 , we assumed that HOCl-dependent activation of RidA is also mediated by modification of C107. Strikingly, however, we found that the cysteine-free variant RidA-C107S was still as active as the wild-type protein after HOCl treatment, and that activation of the treated mutant variant was reversible by DTT as well (Fig. 3i). This result strongly points towards a novel mechanism of functional activation by HOCl, independent of cysteine.
Apart from the oxidation of cysteine thiols, HOCl is also known to react with a number of other amino acid side chains, including methionine, histidine, lysine, arginine, tryptophan and tyrosine 4 . Reaction of HOCl with side chain amines of lysine, for instance, results in N-chlorination to mono-and dichloramines. To analyse whether HOCl-mediated activation of RidA involves the chlorination of side chain amines, RidA HOCl was treated with ascorbic acid. In contrast to DTT, ascorbic acid does not significantly reduce cysteine oxidation products but readily reacts with chloramines allowing the discrimination of these modifications 30,31 . As shown in Fig. 3a, incubation of RidA HOCl with ascorbic acid caused a complete loss of chaperone activity supporting the idea that RidA is a chaperone activated by N-chlorination.
N-chlorination of lysine and arginine side chains inevitably decreases the total amount of detectable amino groups. RidA possesses 14 potential N-chlorination targets: five arginine residues, eight lysine residues and one terminal amino group. We monitored the accessibility of these potential targets upon HOCl or monochloramine treatment using fluorescamine 32 . The amino group content of RidA decreased to o40% within seconds of HOCl treatment and was not further reduced upon longer treatment (Fig. 4a). Monochloramine treatment led to a much slower loss of amino groups but reached the same final percentage (Fig. 4b). Disappearance of amino groups, therefore, appeared to parallel the appearance of RidA's chaperone function (compare Fig. 2d,h). The observed 60% reduction of amino group content corresponds to approximately eight residues that no longer contain a free amino group. DTT treatment leads to a full recovery of accessible amino groups, which coincides with the inactivation of RidA (compare Fig. 4c and Fig. 3a).
HOCl treatment increases hydrophobicity of RidA. The abovedescribed loss of positive charges prompted us to conclude that binding of substrate proteins may be enabled by an increase in hydrophobicity. We thus used the uncharged hydrophobic dye Nile red to directly monitor the polarity of RidA 33 . The fluorescent signal of 0.2 mM Nile red was compared after mixing with either 50 mM RidA or 50 mM RidA HOCl (Fig. 4d). We observed both a wavelength shift from 613 to 600 nm and an increase in absolute fluorescence for RidA HOCl when compared with untreated RidA (Fig. 4d). The RidA and RidA HOCl concentrations leading to half-saturation of the proteins with the dye were determined (Fig. 4e,f). This concentration was significantly higher for untreated RidA (121.1 mM; Fig. 4e) when compared with RidA HOCl (30.7 mM; Fig. 4f). These results indicate that RidA HOCl has an increased hydrophobicity when compared with untreated RidA, which most likely enables binding of unfolding substrates to RidA HOCl . RidA is N-chlorinated by HOCl and monochloramine. To identify the exact type of modification that activates RidA, we performed mass spectrometry of full-length RidA (Fig. 5a). After  ARTICLE appeared. In addition to chlorine adducts, we also observed the irreversible addition of up to five oxygen atoms to methionine and cysteine residues. These oxidation products seemed less prominent after monochloramine treatment of RidA. To exclude the oxidizing effects of chlorinating agents on cysteine, we used the cysteine-free RidA variant C107S for the analysis of a timeresolved monochloramine treatment (Fig. 5b). The chlorination of RidA-C107S observed in this experiment paralleled the activation of chaperone activity (compare Fig. 2k,l and Fig. 5b). The number of chlorines added to RidA-C107S consecutively increased with incubation time until B20 min, when the maximum number of added chlorine atoms was observed. Reduction with DTT resulted in the complete removal of chlorine residues, while oxygen, presumably added to methionine, remained (Fig. 5b).
A ridA mutant is sensitive towards HOCl in vivo. To address the question whether RidA protects cells against the consequences of HOCl-induced damage, we compared the growth of wild-type, ridA deletion mutant and a RidA overexpression strain in the absence and presence of HOCl (Figs 6 and 7). In the absence of HOCl, all the three strains showed a comparable growth behaviour (Fig. 6a). However, in the presence of medium HOCl concentrations (1-2 mM), the ridA mutant displayed a significantly prolonged lag phase when compared with the wild type (Figs 6b and 7b,d) and a DhslO mutant (deficient in the known HOCl-activated chaperone Hsp33; Fig. 6c). Overexpression of RidA in the ridA mutant resulted in an intermediate phenotype, indicating that plasmid-encoded RidA partially restores the growth phenotype of the ridA mutant. HOCl concentrations 42 mM led to a complete growth arrest, while concentrations o1 mM did not influence growth of either strain (Fig. 7). Our results indicate that RidA has a role in protecting E. coli cells against the consequences of HOCl-mediated stress and thus might be an integral component of the bacterial defense strategies against the mammalian immune system.
Mass spectrometry identifies cytosolic clients of RidA. To identify cytosolic RidA substrate proteins, we co-purified proteins from cell lysates bound to His-tagged RidA HOCl by Ni 2 þ -NTA chromatography and elution by de-chlorination (Fig. 8a-d). For this purpose, His-tagged RidA was added to E. coli MG1655-DridA cell lysates. HOCl was added in a 10-fold excess over the final RidA His concentration. Controls were left untreated. After loading the mixture onto Ni 2 þ -NTA chromatography columns and extensive washing, DTT-containing buffer was used to reduce RidA and thereby elute bound substrate proteins (Fig. 8b,d).
No significant elution was observed in the HOCl-free controls (Fig. 8a,b), as well as the DTT-free control (Fig. 8c).
Substrate proteins were identified by mass spectrometry after tryptic digest (Supplementary Tables 2,3). A total of 143 proteins were identified, 18 of which eluted exclusively in samples with DTT (Supplementary Table 3). The abundance of proteins was much higher in samples after DTT elution, except for the protein FklB. Sixty-five of the identified proteins belong to the most abundant proteins in E. coli MG1655 measured after growth in minimal medium 34 . Identified proteins from Supplementary  Tables 2 and 3 were categorized according to their biological function using the program Cytoscape (Fig. 8e). This analysis confirmed that the majority of proteins binding to RidA HOCl were high abundant proteins associated with the primary metabolism of E. coli. It is, therefore, most likely that RidA acts as an unspecific chaperone that binds any unfolding protein in its direct vicinity at the onset of chlorine stress. However, we also identified proteins that are not included in the dataset from Lu et al. and may therefore be of lower abundance, such as thioredoxin 1 (TrxA), uracil phosphoribosyltransferase Upp and iron sulfur cluster assembly scaffold protein IscU 34 . Whether such proteins are preferentially bound by N-chlorinated RidA remains subject to further studies.

Discussion
RidA from E. coli belongs to the YjgF/YER057c/UK114 family of proteins with members in all domains of life. Although highly conserved, different functions have been assigned to members of this family in different species. In Enterobacteriaceae, RidA is an enamine/imine deaminase 21 . We recently reported that RidA is important to overcome nitrosative stress in E. coli 23 . Treatment of cells with peroxynitrite led to oxidation of the single cysteine residue of RidA. In this oxidized state, RidA was no longer able to accelerate the release of ammonia from an enamine/imine intermediate generated by threonine dehydratase IlvA. We then tested the influence of various other oxidative stressors on RidA's enamine/imine deaminase function in vitro. To our surprise, HOCl treatment of RidA did not result in its inactivation but instead RidA bound tightly to IlvA, leading to the inhibition of IlvA's dehydratase activity.
In the present study, we show that this inhibition is the result of a chaperone function of HOCl-activated RidA. Activation of RidA's chaperone function by HOCl occurs rapidly within sample mixing time, similar to previous reports on Hsp33, another bleach-activated chaperone 16 . Activation was maximal at a 10-fold molar excess of HOCl over RidA. However, in contrast to the activation of other redox-regulated chaperones, such as Hsp33 or 2-cys peroxiredoxins, which involve cysteine  oxidation 35 , activation of RidA is cysteine-independent. A cysteine-free RidA variant can still be fully activated by bleach. Activation of chaperone function coincides with a decrease in amino group content, an increase in hydrophobicity and the appearance of multiply chlorinated RidA species as shown by mass spectrometry (Fig. 5). We suspect that N-chlorination most likely occurs at the e-amino groups of lysine and/or the guanidinium groups of arginine as well as the protein amino (N) terminus. The formation of chloramines has previously been regarded as a damaging amino acid modification, which could result in protein inactivation, unfolding, aggregation and even fragmentation due to halogen transfer and formation of crosslinks 3,36-40 . However, we could show that chloramines of RidA are fully removed by treatment with DTT and ascorbic acid. This restoration is accompanied by the inactivation of RidA's chaperone function. A new cycle of activation is possible after The mixtures were applied to four separate Ni 2 þ -NTA chromatography columns. After intensive washing, DTT-containing buffer was used to elute proteins that were bound to RidA in b and d. Controls were washed with the same volumes of DTT-free washing buffer (shown in a and c). Elution of RidA His was performed using buffer containing 250 mM of imidazole. Samples from each fraction were applied to non-reducing SDS gels. (e) Proteins were categorized into biological processes according to GO terms using the Cytoscape/BiNGO interface. The number of proteins found for each category is given. The P value of the enrichment of a certain category is indicated by colour. reduction, demonstrating that the reaction of RidA with HOCl and chloramines is fully reversible and does not alter the structural integrity of the protein.
In vivo, the lack of RidA makes E. coli more susceptible to chlorine stress. A knockout mutant in a BL21 strain background shows a significant growth deficiency in the presence of 2 mM HOCl. Expression of RidA from a plasmid could partially restore this phenotype. A full rescue is probably not achieved due to gene dosage effects and an overall increased stress level due to protein overexpression. In this specific assay, the BL21-DhslO strain (Hsp33 deficient) behaved wild-type-like and was less sensitive towards HOCl as compared with the BL21-DridA mutant. This was surprising, as a DhslO strain has been shown before to be hypersensitive towards HOCl 16 . However, these experiments have been performed in the already chlorine-sensitive E. coli MC4100. Various attempts to construct a DridA strain in the MC4100 strain, as well as in a DhslO background, (DhslO-DridA double mutant) failed, indicating that the lack of ridA in these strains might be lethal.
Purification of RidA HOCl substrate proteins from E. coli cell lysates after HOCl treatment revealed that highly abundant cytosolic proteins, such as ribosomal subunit proteins, are bound by RidA HOCl . A pathway analysis assigned the majority of these proteins to the primary metabolism. This finding, although performed in cell lysates, points towards an unspecific binding mechanism of N-chlorinated RidA in vivo. The identification of numerous cysteine-free proteins, such as the trigger factor, demonstrates that binding of N-chlorinated RidA does not depend on cysteine disulfide bond formation.
Looking for an in vivo mechanism of inactivation of RidA's chaperone function, we found that catalytic quantities of a reconstituted E. coli thioredoxin system could inactivate RidA HOCl NADPH-dependently in vitro. Rates observed in our assay were 10-to 20-fold lower than rates observed in a classic thioredoxin activity assay, which uses insulin as substrate but substantially higher than rates obtained with oxidized glutathione as substrate 41 . We assume that this inactivation occurs when the cell is no longer exposed to HOCl and once again enough reduction equivalents are available to the Trx system. The exact mechanism of the thioredoxin-dependent reduction has to be investigated in future experiments. It might be that chlorines from the amino groups in RidA are transferred to the thiol group of the attacking cysteine in TrxA's active site. This would result in the formation of a sulfenyl chloride intermediate. The sulfenyl chloride would then be reduced by the resolving cysteine of TrxA's active site and H þ and Cl À will be released. The disulfide bond formed in this reaction may then become reduced NADPHdependently by thioredoxin reductase TrxB. To our knowledge, the reduction of chloramines by thioredoxin has not been reported to date and is a new function of this ubiquitous reducing system. However, we also observed glutathione-based inactivation of RidA HOCl . As glutathione concentrations are in the millimolar range in the bacterial cytoplasm 42 , it may be assumed that inactivation by glutathione is a major mechanism of inactivation in vivo.
We asked whether the observed chaperone activity is a shared characteristic of the whole YjgF/YER057c/UK114 family of proteins. We therefore analysed the Drosophila melanogaster RidA homologue DUK114. This protein has previously been shown to act as a chaperone, albeit no prior HOCl treatment was reported 43 . In our hands, untreated DUK114, heterologously expressed and purified from E. coli, could not inhibit citrate synthase aggregation (Fig. 9). However, treatment with HOCl induced holdase activity in DUK114, comparable to RidA. These seemingly contradictory findings might be explained by a different purification protocol used in our study, which included DTT during the preparation of the protein. DTT reduces the modifications that might have been introduced in vivo. However, DUK114 HOCl remained active after DTT treatment and still inhibited citrate synthase aggregation, indicating distinct mechanisms of activation and deactivation of YjgF/YER057c/UK114 family members.
We also found that a number of other proteins, including b-lactoglobulin, a-amylase and bovine serum albumin (BSA) showed HOCl-dependent holdase activity (Fig. 9). Chaperone activity of BSA has been reported before and we show that this activity is further increased by HOCl treatment 44 . Similarly, chaperone activity of human a 2 -macroglobulin has been shown to be largely increased by HOCl treatment 45 . Like BSA, a 2 -macroglobulin is a circulating blood plasma protein, indicating that a set of chaperones might be responsible for the protection of host proteins upon infection and inflammation.
Other proteins such as RNaseA and porcine pancreas lipase did not respond to HOCl treatment (Fig. 9), while lysozyme was highly prone to precipitation during treatment as has been reported before 46 . N-chlorination by HOCl has been investigated in more detail for various proteins including low-density lipoprotein, lysozyme and insulin 38,47,48 . The extent of Nchlorination differed significantly depending on parameters such as surface accessibility, presence of thiol-containing amino acids, the presence of histidine, and the overall tertiary structure of the protein 48 . Recently, the BSA homologue human serum albumin was shown to inhibit the major high-density lipoprotein receptor SR-BI 49 . Interestingly, this inhibition depends on Nchlorination of lysine residues and is irreversible. Moreover, HOCl-modified human serum albumin was shown to bind and inhibit proteins from West Nile virus and HIV 50,51 . These findings suggest a critical role of HOCl-modified plasma proteins during infection and inflammatory processes.
We compared the predicted physicochemical properties of these proteins (protein localization, number of specific amino acids, predicted hydrophobicity and isoelectric point) to identify possible characteristics that are favourable for turning a protein into a HOCl-mediated holdase (Supplementary Table 4). However, on the basis of these properties, a clear commonality between the HOCl-mediated holdase proteins could not be established.
In bacteria, together with other proteins, such as Hsp33, which is activated by disulfide bond formation and the novel HOClresponsive transcription factors HypT, which is activated by methionine oxidation 12,13 , and NemR and RclR, which are activated by cysteine oxidation 14,15 , N-chlorinated RidA may protect cells during host invasion from the deleterious effect of the strong physiological oxidant hypochlorous acid.
N-chlorination of proteins constitutes a novel activitymodulating post-translational switch that reversibly turns RidA into a holdase. N-chlorinated RidA binds a wide spectrum of proteins and can prevent protein aggregation of model substrates, such as citrate synthase or IlvA. Removal of the chloramine modifications, which can be performed NADPH-dependently by catalytic amounts of the thioredoxin system, leads to substrate release by RidA. Foldases can then refold these proteins (Fig. 10).
Protein purification. Overexpression and purification of RidA, RidA mutant and IlvA was performed as previously described using strains and plasmids listed in Supplementary Table 1 (refs. 23,52). Protein concentrations were determined using a JASCO V-650 spectrophotometer. Extinction coefficients used were e 280 ¼ 2,980 M À 1 cm À 1 for the monomer of RidA, e 280 ¼ 4,470 M À 1 cm À 1 for DUK114, and e 280 ¼ 31,860 M À 1 cm À 1 for IlvA.
Oxidation and reduction of proteins. Oxidation and reduction was performed by incubation of 700 mM of the respective protein with varying ratios of HOCl, monochloramine, DTT or ascorbic acid for incubation times indicated at 30 and 37°C, respectively. Removal of stressors was carried out with Micro Bio-Spin Chromatography Columns according to the manufacturer's instructions (Bio-Rad, München, Germany).
Aggregation assays with citrate synthase and IlvA. To monitor aggregation, 12 mM citrate synthase or IlvA were chemically denatured in 4.5 M GdnCl, 40 mM HEPES, pH 7.5 at room temperature overnight. Aggregation was induced by the addition of 0.15 mM citrate synthase or 0.3 mM IlvA to 1,600 ml 40 mM HEPES, pH 7.5. Increase of light scattering was monitored with a JASCO FP-8500 fluorescence spectrometer equipped with a Peltier thermo-holder 'EHC-813' at 30°C for 240 s under continuous stirring. Measurement parameters were set to 360 nm (Ex/Em), 2.5 nm slit width (Ex/Em) and medium sensitivity. The activity of chaperones was determined by the addition of varying amounts of untreated or treated proteins before the addition of citrate synthase or IlvA. Calculation of relative chaperone  To measure the influence of ATP on chaperone activity, a twofold or eightfold excess of RidA over citrate synthase was used. ATP was added in a 10-fold excess over RidA.
Thioredoxin-and glutathione-based inactivation of RidA HOCl . Reduction of RidA by thioredoxin was followed spectrometrically at the absorption maximum of NADPH using a JASCO V-650 spectrophotometer equipped with the temperaturecontrolled cell holder PSC-718. An amount of 50 mM RidA were mixed with 5 mM TrxA (Sigma-Aldrich, St Louis, USA) and 100 mM NADPH (Sigma-Aldrich, St. Louis, USA) in 50 mM Tris-buffer, pH 7.5. After 30 s of stirring, 80 nM TrxB were added. Consumption of NADPH was measured at 340 nm for 2 h.
Reduction by glutathione was carried out by incubating RidA HOCl with a 1-fold or 10-fold molar excess of glutathione (GSH or GSSG) for 30 min at 37°C before the measurement of chaperone holdase activity.
Detection of accessible amino groups-fluorescamine assay. The detection of protein amino groups was carried out as previously described 32 . Briefly, 3 mg ml À 1 of fluorescamine (Sigma-Aldrich, St Louis, USA) were dissolved in acetone. One millilitre of 100 mg ml À 1 RidA solutions in phosphate-buffered saline were mixed with 334 ml of fluorescamine solution. Emission of fluorescamine was measured from 400 to 600 nm with the excitation set to 388 nm. Relative amino group content was calculated by setting the fluorescence of fully reduced RidA to 100%.
Nile red hydrophobicity assays of RidA and RidA HOCl . Nile red (Sigma-Aldrich, St. Louis, USA) was dissolved in DMSO to 0.25 mM. Varying concentrations of RidA and RidA HOCl (1-200 mM) were mixed in 40 mM HEPES-KOH, pH 7.4 with 0.2 mM Nile red. The fluorescence was monitored using a JASCO FP-8500 fluorescence spectrometer equipped with a Peltier thermo-holder 'EHC-813'. Measurement parameters were set to 550 nm excitation, 570-700 nm emission with 5 nm slit width (Ex/Em) and medium sensitivity. RidA and RidA HOCl concentrations resulting in half-saturation were calculated from these data using Igor Pro Version 6.34A (Wave Metrics, Lake Oswego, USA) and Excel for Mac 2011 version 14.3.9 (Microsoft, Redmond, USA). Data were fitted with Igor Pro using the 'Sigmoid' fit function. To obtain the negative inverse of the half-saturation, the reverse protein concentration was plotted against the reverse fluorescence intensity and the 'Trendline' linear regression function of Excel was used.
Mass spectrometry of full-length RidA. RidA treated with HOCl and MCA, and/or the reductant DTT was adjusted for pipette tip-based purification to 0.1% TFA. Purification was performed using OMIX C18 Tips (Varian, Walnut Creek, USA) according to the manufacturer's instructions. Elution of RidA was performed with 100 ml 0.1% FA, 70% acetonitrile. RidA was then directly injected into an Orbitrap Elite mass spectrometer (Thermo Scientific, Waltham, USA) using an API Source with a static nanospray probe and a 1 mm picoTip emitter from EconoTips (NewObjective, Woburn, USA). Measurements were performed with a source voltage of 1.2 kV and a capillary temperature of 275°C. Spectra were deconvoluted and further analysed using Xcalibur software (Thermo Scientific, Waltham, USA).
Threonine deaminase assay. Analysis of a-ketobutyrate generation was performed as described previously 21,23 . Briefly, RidA was treated with HOCl and/or DTT as described in the main body of the manuscript. After the removal of reagents, 1.8 mM RidA were mixed with 0.9 mM IlvA in 1 Â reaction buffer (50 mM MES, 50 mM HEPES, 50 mM TAPS pH 7.5). The reaction was started by the addition of 10 mM threonine. a-Ketobutyrate generation was followed at 230 nm in a JASCO V-650 spectrophotometer equipped with the temperature-controlled cell holder PSC-718 (JASCO, Tokyo, Japan) using a 3.5 ml QS-macro-cuvette (10 mm) with a magnetic stir bar at 20°C. entries (including sequences of the PhosB standard, trypsin, keratin and RidA-His 6 ) was used for protein identification using the following parameters: peptide tolerance, automatic; fragment tolerance, automatic; minimal fragment ion matches per peptide, 2; minimal fragment ion matches per protein, 2; minimal peptide matches per protein, 1; maximum protein mass, 300 kDa; primary digest reagent, trypsin; secondary digest reagent, none; missed cleavages, 2; fixed modifications, carbamidomethyl C; variable modifications, deamidation N and Q, oxidation M, oxidation C, sulfinic acid C, cysteic acid C; false positive rate, 4%; calibration protein, PhosB standard. Only proteins identified with at least three peptides in at least two of three replicates were considered for evaluation.
Gene ontology of RidA interaction partners. Proteins that were eluted after DTT treatment were categorized according to biological functions. The EcoCyc GO annotation of E.coli 56