Excess copper catalyzes protein disulfide bond formation in the bacterial periplasm but not in the cytoplasm

Copper avidly binds thiols and is redox active, and it follows that one element of copper toxicity may be the generation of undesirable disulfide bonds in proteins. In the present study, copper oxidized the model thiol N‐acetylcysteine in vitro. Alkaline phosphatase (AP) requires disulfide bonds for activity, and copper activated reduced AP both in vitro and when it was expressed in the periplasm of mutants lacking their native disulfide‐generating system. However, AP was not activated when it was expressed in the cytoplasm of copper‐overloaded cells. Similarly, this copper stress failed to activate OxyR, a transcription factor that responds to the creation of a disulfide bond. The elimination of cellular disulfide‐reducing systems did not change these results. Nevertheless, in these cells, the cytoplasmic copper concentration was high enough to impair growth and completely inactivate enzymes with solvent‐exposed [4Fe‐4S] clusters. Experiments with N‐acetylcysteine determined that the efficiency of thiol oxidation is limited by the sluggish pace at which oxygen regenerates copper(II) through oxidation of the thiyl radical–Cu(I) complex. We conclude that this slow step makes copper too inefficient a catalyst to create disulfide stress in the thiol‐rich cytoplasm, but it can still impact the few thiol‐containing proteins in the periplasm. It also ensures that copper accumulates intracellularly in the Cu(I) valence.

Not surprisingly, microbes have evolved defensive measures ( Figure 1a). Bacteria commonly possess cytoplasmic chaperones that tightly bind errant copper and deliver it to efflux pumps.
These systems aim to keep the cytoplasm devoid of loose copper (Andrei et al., 2020;Drees et al., 2017;Rensing et al., 2000). This arrangement is tolerable because while bacteria possess a handful of copper-dependent enzymes, they are either localized in the periplasm or they are integral membrane proteins whose copper cofactors are loaded from the periplasm. In E. coli, CopZ is the cytoplasmic chaperone that traps copper, and CopA is the ATP-dependent exporter that expels it. CopA transfers the copper directly to the CusF periplasmic chaperone. CusF in turn passes the copper to a second exporter, the CusCBA complex, which pumps excess copper across the outer membrane and out of the cell entirely. CusF is also likely to bind excess loose periplasmic copper on its own and deliver it straight to the efflux system; however, its affinity for copper is much lower than that of CopZ (Andrei et al., 2020), indicating that the periplasm is more accommodating of free copper than is the cytoplasm.
An additional defensive protein, CueO, is a periplasmic copper oxidase that converts Cu(I) to Cu(II). It seems plausible that Cu(I) can leak into the cytoplasm via monocation channels; if so, by keeping periplasmic copper in its divalent valence, CueO may provide an extra layer of cytoplasmic defense. All of these defensive systems are induced when cells are exposed to copper, and mutants that lack them can be inhibited by even submicromolar copper (Macomber & Imlay, 2009).
These observations raise an obvious question: How exactly does copper poison cells? An early notion was that copper reacted with endogenous hydrogen peroxide in a Fenton-like event, generating hydroxyl radicals that would damage key biomolecules, including DNA; this chemistry is observable in vitro and is commonly denoted as a Fenton-like reaction (Chevion, 1988;Stohs & Bagchi, 1995).
However, once E. coli mutants were available that could be overloaded with copper, experiments did not support the Fenton-like model in vivo. When copZ copA cus cueO mutants were treated with copper, they showed no unusual sensitivity to hydrogen peroxide.
Further, DNA repair mutants were not particularly sensitive to copper (Macomber et al., 2007). Instead, it was found that copper inactivates key cytoplasmic dehydratases in pathways of amino acid biosynthesis and the TCA cycle (Macomber & Imlay, 2009). These enzymes employ [4Fe-4S] clusters to bind their substrates; because these clusters are solvent exposed, copper can adhere to their coordinating cysteine residues, displace the catalytic iron atoms, and eradicate activity. It was subsequently determined that excess copper can also bind and inhibit IscA, a protein involved in the de novo assembly of iron-sulfur clusters (Tan et al., 2014). More recently, two groups determined that copper disrupts heme and chlorophyll biosynthesis causing excretion of coproporphyrin III, which they attributed to likely inhibition by copper of the HemN coproporphyrinogen III oxidase (Azzouzi et al., 2013;Djoko & McEwan, 2013;Steunou et al., 2020). This enzyme belongs to the family of radical S-adenosylmethionine enzymes (RSEs), which also have exposed iron-sulfur clusters, and recent work with purified enzymes confirmed that these clusters are directly damaged by copper (Rohaun & F I G U R E 1 (a) Overview of copper metabolism in E. coli. Environmental copper passively enters the periplasm through porins. (Green) This copper metallate cuproenzymes in the periplasm or the cytoplasmic membrane. Some copper may adventitiously enter the cytoplasm. (Red) Cytoplasmic copper threatens iron-sulfur dehydratases and the Isc cluster assembly system. (Blue) Entry of copper into the cytoplasm may be suppressed by its oxidation by CueO to the divalent form, which is less permeant. Copper(I) in the cytoplasm is captured by CopZ, which delivers it to the CopA pump. CopA extrudes it directly to CusF, which in turn hands it to the CusABC efflux system. The CueR and CusSR regulatory systems are activated by cytoplasmic and periplasmic copper, respectively. Some bacteria charge cuproenzymes using copper that is pumped into and out of the cytoplasm (Andrei et al., 2020). (b) Model compounds used in this study. Because free cysteine coordinates copper in a bidentate fashion, N-acetylcysteine is a better model for cysteine residues in polypeptides. Imlay, 2022) The upshot is that copper may poison pathways dependent on RSEs, including those that synthesize a variety of enzyme cofactors.
However, copper still displays some toxicity in growth media that do not require the function of these enzymes, suggesting that additional targets may exist. The facility of copper at redox reactions suggested another possibility: that copper might poison cells by catalyzing the oxidation of protein cysteine residues. This hypothesis is suggested by the familiar observation that trace metals in cell extracts will oxidize protein thiols, which has prompted generations of biochemists to include chelators like EDTA and/or reductants such as dithiothreitol in their lysis buffers (Cleland, 1964).
Indeed, copper is the most thiophilic biological metal, and its redox potential is compatible with mediating electron transfer from thiolates to molecular oxygen. Studies with free cysteine and glutathione confirmed that copper can oxidize these molecules in vitro (Carrasco-Pozo et al., 2008;Pecci et al., 1997;Rigo et al., 2004;Saez et al., 1982;Smith et al., 1994;Taylor et al., 1966). Moreover, two reports have indicated that copper can oxidize protein thiols in the periplasm (Hiniker et al., 2005;Lippa & Goulian, 2012). However, the idea that copper might additionally oxidize the thiols of cytoplasmic proteins has not been cleanly tested. If copper were able to do so, its oxidation of catalytic cysteine residues could directly inactivate enzymes, and intermolecular disulfide bonds might produce protein aggregates that physically impair the cell.
In this study, we used copper efflux mutants to impose elevated levels of copper upon the E. coli cytoplasm, and we employed alkaline phosphatase (AP) as a sensitive proxy of disulfide bond formation. We determined that copper can effectively oxidize AP cysteine residues both in vitro and in the E. coli periplasm. In contrast, copper was unable to do so when AP was located in the cytoplasm, even at copper doses that fully inactivated other cytoplasmic targets. Our data suggest that cytoplasmic copper stalls in Cu(I)-thiyl complexes, precluding the creation of disulfide stress.

| RE SULTS
The ability of copper to catalyze the oxidation of thiols has most commonly been explored using the biomolecules cysteine and glutathione. However, both of these molecules form bidentate complexes with metals, leading to multiple metal species and complicated kinetics that elude easy interpretation. Moreover, binding by the non-thiolate ligand-the amine in the case of cysteine-strongly influences copper redox behavior (Carrasco-Pozo et al., 2008;Davis et al., 1983;Munday et al., 2004;Ngamchuea et al., 2016;Smith et al., 1994;Taylor et al., 1966). To circumvent these effects, we employed a derivative of cysteine-N-acetylcysteine (NAC)-as a simple model to define the events and rates underlying thiol oxidation. The acetylation of the NAC amine blocks its ability to bind metals (Korshunov et al., 2020), and in this regard, NAC resembles the cysteine residues in polypeptides, where the amine is occupied by an amide bond (see Figure 1b).
The expectation was that thiols would univalently reduce copper(II) to copper(I), which would then transfer the electron to molecular oxygen. Indeed, in test tube experiments, copper steadily catalyzed NAC oxidation, as evidenced by the disappearance of thiol moieties ( Figure 2). The process was biphasic. The immediate thiol oxidation was rapid, did not require oxygen, and was stoichiometric with the copper concentration. This phase presumably reflects direct electron transfer from thiol to copper. A second phase, also proportionate to copper concentration, was catalytic but far slower.
One thiol was oxidized per copper atom equivalent every 5 min ( Figure 2b). This phase depended upon the presence of oxygen and accelerated proportionately when the oxygen concentration was elevated ( Figure 2c). We infer that this phase involves electron transfer from Cu(I) to oxygen and that the rate of this step determines the rate of the overall process.
Accordingly, the second phase of the reaction could also be tracked by monitoring the consumption of oxygen (Figure 3a; Figure S1A). The ratio of thiol consumption to oxygen consumption was 1.9 ± 0.3. We believe that superoxide was the immediate product of copper oxidation, but this could not be directly verified because copper complexes interfere with standard assays for superoxide. Instead, we demonstrated the accumulation of its dismutation product, hydrogen peroxide: The level of dissolved oxygen rose when catalase was added to the reaction products ( Figure 3b).
Still, neither superoxide nor hydrogen peroxide was needed for thiol oxidation to continue, as scavenging enzymes did not impair the process ( Figure S1B).
Collectively, these data align with a straightforward scheme: Reactions 4 and 5 have been studied in the context of glutathione oxidation chemistry and are known to be very rapid (Winterbourn, 1993) (Koppenol, 1993). Those reactions, like reaction 3, may occur while the reactants remain in a sulfur-copper complex. This basic scheme is consistent with observations made by other workers in other copper/thiol reaction systems (Carrasco-Pozo et al., 2008;Pecci et al., 1997;Rigo et al., 2004;Saez et al., 1982;Smith et al., 1994;Taylor et al., 1966).
Our data indicate that electron transfer from copper(I) to oxygen (reaction 3) comprised the rate-limiting step. Hanna and Mason (1) reported that free, dissociated copper(I) oxidizes in a phosphate buffer with a half-time of ~5 s (Hanna & Mason, 1992), and we observed a similar rate in the Tris buffer used in our system ( Figure 3c). That rate sharply contrasts with the time frame of ~5 min for each copper atom to complete the oxidation of a thiol (Figure 2b). This difference indicates that the residual association of Cu(I) with the thiyl radical slows its reaction with oxygen by substantially more than an order of magnitude.
Importantly, the second phase of thiol oxidation was further suppressed when NAC concentrations were elevated, reflecting the progressive binding of copper atoms by multiple NAC molecules  & Stoppani, 1996;Hanna & Mason, 1992;Ngamchuea et al., 2016;Pecci et al., 1997;Smith et al., 1994;Taylor et al., 1966). We minimized this complication in the experiments of Figures 2 and 3 by working at lower levels of NAC. However, this issue will need to be considered in the context of the bacterial cytoplasm, which contains small-molecule thiols such as cysteine and glutathione.

| Copper catalyzes the oxidation of cysteinyl side chains of alkaline phosphatase in vitro
To probe the ability of copper to oxidize the cysteine residues of proteins, we selected alkaline phosphatase as a target. This approach can provide greater sensitivity than proteomic analyses, which are F I G U R E 2 Copper catalyzes thiol oxidation in vitro, and electron transfer to oxygen is the slow step. (a) N-acetylcysteine (NAC, 100 μM) was incubated with Cu(II) in aerobic buffer, and the thiol content was tracked. The immediate drop in thiol content reflects rapid electron transfer to copper. (b) The steady-state rate of NAC (100 μM) oxidation was proportional to Cu(II) concentration. The data indicate that each copper atom completes an oxidation cycle approximately once per 5 min. (c) NAC (100 μM) was incubated with 20 μM Cu(II) in buffer that was anoxic, air saturated, or bubbled with 100% oxygen, showing that the oxidation of Cu(I) by molecular oxygen is the slow step in overall process. In this figure and all others that follow, experiments were performed in triplicate, and error bars represent the standard error of the mean. Small error bars may be obscured by symbols.

F I G U R E 3
Oxygen consumption in NAC oxidation experiments. (a) Molecular O 2 was consumed when Cu(II) (40 μM) oxidized NAC (100 μM), but not when NAC or Cu(II) was omitted. Oxygen levels were monitored with a Clark electrode. Figure S1A demonstrates that the rate of oxygen consumption depended upon Cu(II) concentration. (b) When catalase was added at 5 min, O 2 levels rebounded, reflecting the presence of accumulated hydrogen peroxide. The reaction was initiated with 40 μM Cu(II) and 100 μM NAC. (c) Copper(I) is immediately oxidized when it is not coordinated by sulfur species. Copper(I) chloride in anoxic acetonitrile was injected into the standard air-saturated Tris reaction buffer (final copper concentration: 160 μM). The 1:2 stoichiometry of oxygen consumed/copper added, and the rebound in oxygen concentration upon catalase addition, demonstrate that H 2 O 2 was the accumulated product. This curve is representative of four replicates. more global but can acquire some background signal through incomplete modification of reduced protein thiols or chemical oxidation events during sample handling. More importantly, AP allowed us to compare the ability of copper to oxidize thiols of the same protein in three circumstances: in vitro, in the bacterial periplasm, and the bacterial cytoplasm. Alkaline phosphatase (AP) is a periplasmic enzyme whose activation requires the oxidation of cysteine residues to form two disulfide bonds. This process is normally catalyzed by the periplasmic disulfide bond (Dsb) system (Smith et al., 1994). DsbA is a periplasmic protein whose own disulfide bond undergoes electron exchange with the dithiol moieties of secreted proteins, thereby transferring the disulfide bond to the client protein while DsbA itself becomes reduced ( Figure S2). The reduced DsbA is then reoxidized by its partner protein, DsbB, which is embedded in the cytoplasmic membrane. DsbB then transfers the electron pair to the quinone pool, which ultimately delivers it to the respiratory oxidant.
Purified, active AP was denatured with guanidinium hydrochloride and reduced with β-mercaptoethanol. The denaturant and reductant were then removed by spin columns under anoxic conditions in an anaerobic chamber, rendering AP with reduced cysteine residues. Accordingly, the enzyme was inactive.
The activation of reduced AP was readily achieved by incubation with either cystine or oxidized glutathione ( Figure S3). This effect was driven by disulfide exchange reactions that mirror those with F I G U R E 5 Cu(II)/O 2 can activate AP by oxidizing its cysteine residues in vitro. (a) Reduced AP was incubated with Cu(II) or 0.5 mM cystine at 37°C in aerobic buffer prior to assay. (b) The activation of reduced AP required both Cu(II) (30 μM) and oxic conditions. (c) AP activation required that copper (30 μM) be in its Cu(II) valence and did not require reactive oxygen species. SOD: added superoxide dismutase; Cat: catalase. (d, e) High levels of NAC impaired the activation of AP by Cu(II). Reduced AP was incubated with 30 μM Cu and 400 μM NAC. DsbA ( Figure S2), and it indicated that the reduced AP was competent for reactivation. It also confirmed that the key cysteine residues of the reduced protein are accessible to solutes.
Reduced AP was not activated by oxygen alone. However, activation was triggered by the addition of Cu(II) ( Figure 5a). As expected (rxns 1-7), reactivation did not happen in the absence of oxygen (Figure 5b) or when copper was provided only in the Cu(I) valence ( Figure 5c). Neither catalase nor SOD diminished the rate of activation, indicating that although reactive oxygen species are presumably formed during the process, they are not responsible for thiol oxidation (Figure 5c). We note that, as with NAC oxidation, the activation of AP was not fast. Further, activation was blocked when N-acetylcysteine was supplied at a concentration that substantially exceeded that of copper (Figure 5d,e)-a result that mirrored what we had observed for the oxidation of N-acetylcysteine itself. Therefore, although copper has the capacity to oxidize cysteine residues, this process will be hindered in a thiol-rich environment. Hiniker et al. (2005) reported that mutants lacking DsbA and DsbB proteins are defective at periplasmic AP activity, as the enzyme accumulates in its reduced form. They found that exogenous copper could reverse this defect. Lippa and Goulian (2012) detected a similar event. However, in our initial experiments, the AP activity of a dsbA mutant unexpectedly was as high as that of Dsb + cells-even without the addition of copper. This outcome resembled observations made by Meehan et al. (2017). Ultimately this inconsistency was resolved when we determined that the redox status of AP in dsbA mutants depends upon the composition of the growth medium ( Figure S4). In amino-acid-free medium-which was also used by Meehan et al.-aeration was sufficient to activate AP. However, when cultures were supplemented with amino acids-as in refs. (Hiniker et al., 2005) and (Lippa & Goulian, 2012)-the aerated cultures had minimal AP activity. Dissection of the medium revealed that histidine was the key component: When histidine was included in the growth medium, Dsb-independent AP activation was prevented. Histidine is a capable chelator of metals, suggesting that the oxidation of AP in unsupplemented dsbA cultures was catalyzed by trace metals in the medium. The identity of the putative metal(s) was not determined. Standard lab media contain both copper and iron, as contaminants of medium components; both metals can catalyze thiol oxidation (Carrasco-Pozo et al., 2008;Rigo et al., 2004;Taylor et al., 1966).

| Copper can activate periplasmic AP
When copper was then added to histidine-containing dsbA cultures, it activated AP ( Figure 6). As little as 5 μM was enough to exert a significant effect. This result reproduced the observations of (Hiniker et al., 2005) and (Lippa & Goulian, 2012), and it demonstrates that periplasmic copper capably oxidizes the cysteine residues of secreted AP. The result matches the effect of copper upon AP in vitro. Moreover, it supports the proposal (Hiniker et al., 2005) that one avenue of copper toxicity may be through the inappropriate oxidation of periplasmic protein thiols.

| However, copper does not activate AP inside the cytoplasm
Removal of the AP leader sequence causes AP to be retained in the cytoplasm (Derman & Beckwith, 1991). Its activity is very low, as bacterial cytoplasms lack a dedicated system for disulfide-bond formation. In fact, the small amount of AP activity that we detected was primarily due to a minor degree of leader-independent AP secretion, as the addition of a dsbA mutation suppressed it ( Figure S5).
Our previous work showed that imported cystine can activate this leaderless cytoplasmic AP (hereafter denoted cAP), presumably through the same disulfide exchange reactions that had allowed cystine to activate the enzyme in vitro (Korshunov et al., 2020).
This effect occurred only in strains possessing an active cystine importer, confirming that activation took place in the cytoplasm. Therefore, we regard this approach as a good one to identify agents or stresses that can oxidize cytoplasmic cysteine residues. The method detects the appearance of cAP activity starting at a baseline near zero, and therefore the approach is much more sensitive than would be a method that tracks the disappearance of activity in a thiol-dependent enzyme. Further, once disulfide bonds are formed, the cAP polypeptide folds around them and prevents any reduction during subsequent growth or in vitro handling (Akiyama & Ito, 1993).
We reproduced the phenomenon of cAP activation during cystine import (Figure 7b). The cAP expression plasmid was then transferred to a copA cueO cus strain that lacks the cytoplasmic CopZ chaperone, the periplasmic CueO copper oxidase, and the CopA and Cus copper efflux systems. Through translational frameshifting, F I G U R E 6 Reduced periplasmic AP is activated by copper. Wildtype (SE50) and dsbA (SE60) strains were grown with full aeration in minimal M9-glucose medium containing 0.5 mM histidine. Periplasmic AP activity was restored to dsbA strains by the addition of either CuSO 4 or 0.5 mM cystine.
In the present experiments, we employed media containing glucose and branched-chain amino acids, so that growth would not be impaired by the damage to iron-sulfur dehydratases in the TCA cycle and branched-chain biosynthetic pathways; by doing so, we were able to apply especially high doses of copper. We identified copper concentrations that substantially inhibited growth (Figure 7a).
Notably, even at these concentrations, no increase in AP activity occurred (Figure 7b) [A decline in basal activity occurs, as copper disrupts maturation of the small amount of AP secreted to the periplasm ( Figure S9)]. At the same time, these copper doses were sufficient to completely inactivate isopropylmalate isomerase through destruction of its iron-sulfur cluster (Figure 7c).
E. coli possesses thioredoxin and glutaredoxin systems dedicated to eradicating cytoplasmic protein disulfide bonds (Ritz & Beckwith, 2001). Our earlier work indicated that during cystine import, the yield of active cAP was elevated in strains lacking thioredoxin A (Korshunov et al., 2020). We inferred that this thioredoxin could sometimes reduce newly oxidized cAP before the protein folded, and we wondered whether this countervailing action might have obscured some cAP oxidation by copper. However, the addition of a trxA mutation did not improve cAP activity during copper stress ( Figure 7d). Similarly, copper did not boost cAP activity in mutants lacking glutathione, which would be devoid of glutaredoxin activity ( Figure S7).

| Cytoplasmic copper did not activate the OxyR system
The OxyR transcription factor uses a sensory cysteine residue to detect stresses that lead to thiol oxidation (Choi et al., 2001;Zheng et al., 1998). The activated form of the protein possesses a disulfide bond that locks it into an activated form. OxyR is regarded primarily as a sensor of hydrogen peroxide, and many of the F I G U R E 7 Copper does not activate cytoplasmic AP even when internal Cu concentrations are sufficient to abolish IPMI activity. (a) Copper inhibited the growth of a copA cueO cus strain (SE71) in M9-glucose medium (containing Ile, Val, and Leu). Growth was monitored at 500 nm. Cells were harvested at the indicated time point for enzyme analysis (panel b). (b) cAP activity was not boosted by copper exposure (A) (SE71). Data have been normalized to untreated sample. (c) The copA cueO cus strain (SE85) with pLEUCD, encoding isopropylmalate isomerase (IPMI), was grown in M9-glucose medium with copper prior to IPMI assay. Data have been normalized to untreated sample. (d) Copper exposure was repeated in a copA cueO cus strain (SE77) additionally lacking the trxA-encoded thioredoxin 1. Copper again failed to activate cAP. Data have been normalized to untreated sample. genes induced by oxidized OxyR have functions explicitly suited to protecting the cell against H 2 O 2 , including catalase and NADH peroxidase (Sen & Imlay, 2021). However, its use of a cysteine residue to sense stress makes OxyR responsive to any stress that creates disulfide bonds. It can respond to diamide (Zheng et al., 1998), which is a synthetic chemical designed to create disulfide bonds in proteins (Kosower et al., 1969), as well as to imported cystine (Korshunov et al., 2020). Therefore, we tested whether copper stress leads to OxyR activation.
We found that excess copper did not activate OxyR (Figure 8). In the positive control, a strain that lacks H 2 O 2 -scavenging enzymes, 1 micromolar H 2 O 2 accumulated and strongly induced the fusion, making it clear that copper did not furnish even a small fraction of activity. The failure of copper to activate OxyR-either by direct oxidation or indirectly by the generation of disulfide metabolites-is consistent with its failure to activate cAP.

| Cuprous copper is present inside the cell
EPR analysis was conducted on the efflux mutant after growth in copper(II)-supplemented medium. Treatment with a bolus of hydrogen peroxide, which oxidizes Cu(I) to Cu(II), substantially elevated the Cu(II) EPR signal of the cells (Figure 9). This result shows that cell-associated copper tends to accumulate in the Cu(I) valence. The EPR analysis does not specify whether these signals reflect cytoplasmic or periplasmic copper. Our previous work, under slightly different conditions, also showed that copper-loaded cells contain predominantly Cu(I) (Macomber & Imlay, 2009). The presence of Cu(I) complexes in vivo mirrors the behavior that we observed with NAC in vitro, where Cu(I) oxidation was rate limiting and oxygen concentration dictated the rate of thiol oxidation. This barrier would explain why cytoplasmic protein thiols are not easily oxidized.
The dependence of thiol oxidation upon oxygen concentration means that oxidation would be further depressed in most microbial habitats, including the hypoxic gut margin where E. coli lives. In that region, oxygen concentrations are less than 4% of the level in our air-saturated cultures (Kelly & Colgan, 2016).

| Do other thiol-oxidizing agents create cytoplasmic disulfide stress?
Diamide itself was examined for its ability to activate cAP. As expected, we observed a progressive increase in cAP activity with diamide concentration, with maximum activity coinciding with a dose that also briefly slowed cell growth (Figure 10a,b).
Diamide is not a natural chemical, but tetrathionate is a disulfide compound found in mammalian intestines (Rogers et al., 2021).
Reactive oxygen species produced by the cellular immune response can oxidize thiosulfate to tetrathionate, and Salmonella species have even evolved the ability to employ tetrathionate as a terminal electron acceptor. We found that very high concentrations were needed to slow the growth of E. coli (Figure 10c). Somewhat lower doses triggered activation of cAP (Figure 10d). Unlike Salmonella, E.
coli lacks a dedicated tetrathionate importer, which likely explains why high doses were needed to exert these effects.
Hypochlorous acid (HOCl) is a potent cysteine oxidant, and we examined its action upon AP because of the substantial interest it carries in the context of the immune response (Davis & Hawkins, 2020). However, HOCl did not activate reduced AP, either in vivo or in vitro ( Figure S8). Its reaction with cysteine residues initially generates a sulfenyl chloride species, which progresses to disulfide, cysteic, or sulfonamide end products depending upon the local environment (Hawkins et al., 2003); therefore, we infer that HOCl either reacts with other critical amino acids of AP more readily than it does with its cysteine residues or that the reaction of HOCl with the latter residues leads to products other than disulfide bonds.

| Why does copper not oxidize the cysteine residues of cytoplasmic proteins?
Copper overload is a phenomenon that can occur in a range of natural environments, and copper-resistance responses are common among microbes. The present investigation was undertaken F I G U R E 8 Cytoplasmic copper does not oxidize the sensory thiol of OxyR. A copZ/A cueO cus strain (SE71) and a katE katG ahpCF strain (Hpx − ) (AL494), both containing katG-lacZ transcriptional fusions, were grown in M9-glucose medium. Where indicated, copper was added to the medium of the copZ/A cueO cus strain 45 min before harvesting. The katE katG ahpCF strain accumulates ~1 μM H 2 O 2 , which provides a positive control for katG-lacZ induction by activated OxyR.
to identify the cellular targets that these defenses are protecting ( Figure 1a). Over the years, a variety of ideas have been suggested.
Much literature was devoted to the hypothesis that copper could catalyze a Fenton-type reaction that generates toxic hydroxyl radicals, which might form lethal DNA lesions (Chevion, 1988;Stohs & Bagchi, 1995). This chemistry works efficiently in vitro. However, studies with copper-overloaded E. coli indicated that copper did not stimulate DNA damage, even when cells were perfused with hydrogen peroxide (Macomber et al., 2007). This result implied that cytoplasmic copper must be chelated by biomolecules away from DNA.
The present work supports that idea.
We determined that copper can catalyze AP thiol oxidation in the test tube and in the periplasm, but not in the cytoplasm.
Molecular oxygen instantaneously equilibrates across biological membranes (Imlay & Fridovich, 1991;Kihara et al., 2014), and experiments verify that the oxygen concentration inside E. coli is not lower than that outside the cell (Becker et al., 1996;Imlay & Fridovich, 1991). Therefore, the lack of oxidation in the cytoplasm was not due to a dearth of oxygen. Instead, we focus on the relatively slow turnover of the oxidation reaction, coupled with the large difference in the ratios of copper atoms to protein thiols in these two compartments.
The whole E. coli cell contains only 10-30 micromolar copper, very little of which is likely to be loose in the cytoplasm rather than integrated into cuproenzymes, whereas the cytoplasm contains approximately 40 mM protein thiols (calculated from data in Imlay et al., 2015;Neidhardt & Umbarger, 1996). Given the slow rate at which copper turns over (Figures 2 and 3), it is implausible that these few copper atoms could oxidize a significant fraction of any cytoplasmic protein on the time scale of a cell generation. In contrast, the periplasmic compartment is much more accessible to environmental copper, while its proteins average 30-fold fewer free thiol residues than do cytoplasmic proteins (Table S1). This disparity is extraordinary. The consequence is that exogenous copper atoms can narrowly target the few periplasmic thiols so that even slow copper turnover can have a discernible impact on them. In addition, copper may oxidize the short-lived cysteine residues as nascent proteins are secreted. For this reason, E. coli mutants lacking DsbC and DsbD are sensitive to copper: These two proteins resolve incorrectly ordered disulfide bonds in periplasmic proteins, apparently including those whose cysteine residues were chemically oxidized by copper before DsbA could process them correctly (Gupta et al., 1997;Hiniker et al., 2005).
One reason that copper is a poor oxidant of cysteine residues is that apparent thiyl-Cu(I) complexes transfer electrons to oxygen only sluggishly. Beyond that, the cytoplasm is also likely to feature the complexation of Cu(I) by multiple small-molecule thiols in a form that will block oxidation almost entirely. In vitro studies, both here and in other work (Carrasco-Pozo et al., 2008;Correa & Stoppani, 1996;Hanna & Mason, 1992;Ngamchuea et al., 2016;Pecci et al., 1997;Smith et al., 1994;Taylor et al., 1966), observed that a range of thiol compounds produce Cu(I)-dithiol chelates that are almost fully stable to oxygen. These two effects would ensure that cytoplasmic copper accumulates in the Cu(I) valence-which would not be true were the oxidation of these complexes fast.
Indeed, the prevalence of the cuprous valence in the cytoplasm explains why cytoplasmic copper defense mechanisms focus on Cu(I) rather than Cu(II). It is the Cu(I) form that is bound by CopZ, the chaperone that patrols the cytoplasm for copper (Hearnshaw et al., 2009); by CopA, the transporter that exports it (Fan & Rosen, 2002;Zhou et al., 2012); and by CueR, the transcription factor that senses copper overload (Chen et al., 2003). Each of these proteins captures Cu(I) in an oxygen-stable complex in which the metal is liganded in linear S-Cu-S fashion by two cysteine residues. F I G U R E 9 EPR analysis of the valence of cell-associated copper. (a) An EPR peak of standard Cu + H 2 O 2 . Arrows indicate how peak height was measured to quantify the signal. (b) A copA cueO cus strain (LEM29) additionally lacking cellular catalases was grown in LB medium. Where indicated 2 mM copper was included in the medium. Hydrogen peroxide was added to some samples to convert Cu(I) to Cu(II), immediately before EPR analysis. The rise in the EPR signal indicates that much cell-associated copper was originally in the Cu(I) valence. This analysis does not indicate whether the Cu signals are associated with the cytoplasm or periplasm/cell surface.
This arrangement mimics the (RS − ) 2 -Cu(I) complexes that are formed by thiol metabolites; the zeptomolar (10 −21 ) dissociation constant of CueR:Cu(I) demonstrates the strength of such bonds and explains their resistance to oxidation (Chen et al., 2003). Thiolate trapping of copper(I) by small molecules in the cytoplasm would quench any protein-oxidizing behavior, and it may have been responsible for blocking DNA oxidation, too. In contrast, thiol species are expected to be more scarce in the periplasm, so that copper retained its ability to oxidize AP cysteine residues there.

| How does copper poison cells?
The Cu(I) species remain a potent poison of solvent-exposed ironsulfur clusters. This action is not a redox event, as copper simply displaces iron atoms from their coordinating sulfur atoms; in this behavior, Cu(I) is more potent than Cu(II) (Macomber & Imlay, 2009).
The ability of copper to inhibit E. coli growth in a minimal glucose medium stemmed from its destruction of clusters in enzymes that are requisite for amino acid synthesis, and this inhibition was pronounced even under rigorously anaerobic conditions. However, our growth studies suggest that additional mechanisms of copper toxicity may exist. The affinity of copper for protein thiols allows it to directly inhibit select enzymes by complexing active-site cysteine residues. Glutathione reductase is an example of such an enzyme, and we show in Figure 11 that nanomolar concentrations of Cu(I) suffice to inhibit it in vitro. Other workers have similarly observed that copper inhibits analogous enzymes (Ahmed et al., 2016;Bandyopadhyay et al., 1997;Gutierrez-Correa & Stoppani, 1997;Hou et al., 2015;Murakami et al., 2014)-although those experiments were invariably performed in oxic buffers, and thiol oxidation rather than binding was sometimes proposed to be the mode of copper action. The experiment of Figure 11 was performed in the absence of oxygen, showing that inhibition depended only on Cu(I) binding. This route of toxicity seems plausible in the periplasm, where the copper/thiol ratio can be high; May et al. observed that F I G U R E 1 0 Added diamide and tetrathionate each activate cytoplasmic AP. (a) Diamide was added to a WT/pcAP strain (SE17) growing in M9-glucose medium, and biomass was tracked by optical density. A representative experiment is shown. (b) Cytoplasmic AP activity was measured in the WT/pcAP strain (SE17) after 15 min of diamide exposure. Data have been normalized to untreated sample. Asterisk denotes p-value is less than 0.05. (c) Sodium tetrathionate (Na 2 S 4 O 6 ) was added to a WT/pcAP strain (SE17) growing in M9-glucose medium, and optical density was monitored. (d) Cytoplasmic AP activity was measured in the WT/pcAP strain (SE17) after 15 min of tetrathionate exposure. Data have been normalized to untreated sample. Asterisk denotes p-value is less than 0.05. millimolar levels of environmental copper inhibit the maturation of lipoproteins, presumably by physically blocking the derivatization of their cysteine residues (May et al., 2019). An analogous problem would be less likely in the cytoplasm, as the copper/thiol ratio is inadequate to occupy a large fraction of any enzyme population.

| How do disulfide bonds form inside the cytoplasm?
Genomic analyses have shown that evolution sharply selects against the exposure of cysteine residues on protein surfaces, apparently because of their propensity for adventitious chemistry-most obviously, redox reactions (Marino & Gladyshev, 2010). E. coli possesses two thioredoxins and three glutaredoxins (Ritz & Beckwith, 2001), which is compelling evidence that undesirable oxidations somehow occur, even in the cytoplasm. However, at the end of the present study, we remain ignorant of the major routes by which this might happen. Molecular oxygen itself is a very poor direct oxidant of thiols; in our experiments, protein disulfide bonds did not detectably accumulate in AP under aerobic growth conditions even when redoxins were absent. This remained true in cells severely overloaded with copper, a thiophilic redox catalyst. We did observe that imported disulfide compounds-cystine and tetrathionate-can transfer their own disulfide bonds to proteins, but this would seem to be a specialized circumstance that is unlikely to warrant the presence of so many reducing systems. Similarly, while hypochlorous acid is an established oxidant of protein thiols, E. coli is not routinely exposed to it.
A reasonable hypothesis is that we examined the wrong transition metal. Iron, like copper, can catalyze thiol oxidation in the presence of molecular oxygen in vitro (Taylor et al., 1966). Indeed, iron may have been the histidine chelatable contaminant that catalyzed periplasmic AP oxidation in dsb mutants. Unlike copper, iron is naturally abundant in the cytoplasm, where it is needed to activate scores of enzymes (Andrei et al., 2020;Outten & O'Halloran, 2001)-and it is a harder metal (Irving & Williams, 1948) that is not immobilized in non-oxidizable thiol complexes. Even more striking is the fact that both thioredoxin 2 and glutaredoxin 1 are directly induced when hydrogen peroxide activates the OxyR transcription factor (Ritz et al., 2000;Zheng et al., 1998). While H 2 O 2 is not a good direct oxidant of thiols (Cardey & Enescu, 2007;Radi et al., 1991), in collaboration with iron it excels (Radi et al., 1991). These ideas are under investigation.

| Strains and plasmids
The full list of strains and plasmids can be found in the supplementary tables (Tables S2 and S3). Null mutations were made using the lambda red recombinase method to replace the open reading frame (ORF) with a Cam resistance cassette amplified from the pKD3 template (Datsenko & Wanner, 2000). P1 transduction was used to introduce mutations into new strains. All mutations were verified by PCR and gel analysis.

| Growth conditions
All strains were grown in M9 medium (Miller, 1972) at 37°C containing 0.2% glucose, 0.5 mM branched-chain amino acids (L-isoleucine, L-valine, and L-leucine), 0.5 mM thiamine, 0.01% MgSO 4 , and 0.01% CaCl 2 . In initial experiments tracking AP activity in the periplasm, the M9-glucose medium was supplemented with 0.2% casamino acids or with subsets of amino acids at 0.5 mM each, as discussed in the text.

F I G U R E 11
Glutathione reductase is inhibited by nanomolar concentrations of Cu(I). The activity of purified glutathione reductase from yeast (GOR, 3.9 nM) was determined under anoxic conditions in the presence of Cu(I) at 37°C.
The branched-chain amino acids were supplied because copper poisons isopropylmalate isomerase and dihydroxy acid dehydratase, two [4Fe-4S]-dependent dehydratases in the branched-chain biosynthetic pathways (Macomber & Imlay, 2009). Similarly, the use of glucose as a carbon source allows the cell to generate energy with minimal flux through the TCA cycle, whose aconitase and fumarase are also inhibited when copper damages their clusters (Macomber & Imlay, 2009) The growth medium was supplemented with 200 μg/mL ampicillin to maintain pAID135 and 50 μg/mL ampicillin for pLEUCD3. Most cultures were grown aerobically with vigorous shaking.
Where indicated, cultures were bubbled with 100% oxygen from a 100% O 2 gas tank with the aid of a frit.
Some experiments were conducted under anoxic conditions inside an anaerobic chamber. The reagents were moved into the chamber while still hot (from autoclaving) to minimize dissolved oxygen, and further degassed overnight in the chamber.
Experiments were conducted with log-phase cultures. Cells were grown overnight, subcultured to low densities (~0.005 OD 600 ), and grown for at least four generations before dilution into stress conditions for subsequent measurements. Cell density was tracked by absorbance at 600 nm. Cell viability was determined by dilution and plating on LB plates.

| In vitro oxidation of N-acetylcysteine and glutathione by copper
Reactions between N-acetylcysteine or reduced glutathione and copper(II) sulfate were performed in 50 mM Tris, pH 8.0, at RT. Unless explicitly noted, the reagents and buffer were equilibrated with room air. At time points, aliquots were removed to tubes containing 1 mM EDTA (to sequester copper and stop the reaction) and 250 μM DTNB.
The mixture was then incubated for 5 min at RT, and then the absorbance at 412 nm was determined and converted to molarity using an extinction coefficient of 14.15 M −1 cm −1 (Riddles et al., 1983). Where indicated, superoxide dismutase (20 U/mL final concentration) and/or catalase (9 U/mL final concentration) were included in the reaction mix.
In some experiments with glutathione, the production of oxidized glutathione (GSSG) was verified by quantifying the total change in NADPH concentration, using A 340 = 6.22 M −1 s −1 , after the copper-GSH reaction product was incubated for 2 min with 100 μM NADPH and 0.32 U/mL units of glutathione reductase.

| Preparation of reduced alkaline phosphatase in vitro
To prepare E. coli alkaline phosphatase (AP) that was reduced and inactive, the purified enzyme (200 U, 4 mg) (Sigma catalog # P5931) in 50 mM Tris pH 8.0 was denatured with 3.6 M guanidinium HCl and reduced with 25 mM β-mercaptoethanol under aerobic conditions, as described (Walker & Gilbert, 1994). The AP solution was then transferred to the anaerobic chamber and held in an incubator at 37°C overnight. Activity was checked the next day. If not fully eliminated, incubation was continued at 37°C. The guanidinium HCl and β-mercaptoethanol were then removed using an Amicon Ultra-0.5 (Millipore) Centrifugal Filter Device. The denatured APase (400 μL) was loaded onto a 30 kD spin column, and the samples were centrifuged at 14,000× g at RT for 15 min, leaving 40 μL of sample. The filtrate was discarded, and 400 μL of anoxic folding buffer was added (1 mM ZnAc 2 , 1 mM MgCl 2 , and 100 mM Tris, pH 8). The column was recentrifuged as above. For protein collection, the column was then inverted and centrifuged into a fresh tube, 1000× g for 2 min. The volume was adjusted to 400 μL with folding buffer. The reduced, inactive enzyme typically retained ~0.1% of the original activity, with some variation from one preparation to the next. It was stored on ice in the anaerobic chamber to avoid activation in air; it was stable in this form for at least 1 month.

| Activation of purified AP
Stock solutions of copper sulfate, tetrathionate, diamide, and GSSG were prepared in water. A 50 mM stock of cystine was prepared in ddH 2 O; the cystine was dissolved by addition of a couple of drops of 6 M HCl. Sodium hypochlorite was diluted in 10 mM NaOH, and the molar concentration was determined by measuring absorbance at 292 nm [extinction coefficient: 350 M −1 cm −1 (Floris & Wever, 1992)].
AP (diluted 1:10) was exposed to oxidizing agents in 100 mM Tris, pH 8, at 37 °C. This reaction mixture also included 10 mM MgCl 2 and 10 mM Zn(II) diacetate, as zinc is a cofactor of AP. Unless otherwise indicated, reactions were conducted in aerated buffer. Where indicated, reactions included 20 U/mL superoxide dismutase and 30 U/ mL catalase. At time points, 50 μL aliquots were removed to 1 mL 1 mM p-nitrophenylphosphate in 1 M Tris, pH 8. AP activity was measured based on its ability to hydrolyze p-nitrophenylphosphate to p-nitrophenol, a chromogenic product that absorbs at 405 nm (Bessey et al., 1946).

| Activation of alkaline phosphatase in the cytoplasm
A leaderless phoA construct (phoA ∆2-22) encodes a non-exported form of alkaline phosphatase (AP) that has been used to detect disulfide bond formation in the E. coli cytoplasm (Derman et al., 1993). Without its signal sequence, AP accumulates in the cytoplasm and is minimally exported to the periplasm. AP is only active when it acquires disulfide bonds; once those bonds are formed, AP folds around them, stabilizing them against potential reductants during continued culture, cell harvesting, and extract preparation (Akiyama & Ito, 1993). AP is expressed from a pBR322-based plasmid [pAID135 (Derman et al., 1993)] which is ampicillin resistant, and the phoA gene is under the control of a tac promotor. We observed that it was not necessary to induce the tac promoter to establish sufficient AP synthesis for our purpose, and so we did not do so, preferring steady AP production. All strains used lack the chromosomal phoA gene.
Copper or other potential oxidative stressors were added to the growth medium as described in experimental captions. To quantify AP activity inside cells, 20 mL of culture (ca. 0.1 OD 600 ) was centrifuged, resuspended in ice-cold lysis buffer (20 mM aerobic Tris pH 8.0 and 10 mM EDTA), recentrifuged, resuspended in 1 mL lysis buffer, and lysed by passage through a French press. The extract was clarified by centrifugation (17,000× g for 20′) and then diluted 1:10 into post-lysis buffer (10 mM MgCl 2 , 10 mM Zn(II) diacetate, and 1 M Tris, pH 8.0). The extract was incubated for ~10 min at RT for remetallation. Then, 100 μL was assayed for AP activity as above. A Bradford assay (Thermo Scientific) was used to determine the total protein concentration, with ovalbumin as the protein standard.

| Activation of alkaline phosphatase in the periplasm
AP activation was also tracked in the periplasm. The strains that were used contain the WT phoA allele and do not contain the cAP plasmid. These strains constitutively express phoA at a higher level due to a phoR null mutation, and one strain (SE60) carries a dsbA mutation that prevents the insertion of disulfide bonds into secreted AP, rendering it inactive. The base medium used was the same as described under "Growth conditions," with the amino acid content manipulated and oxidative stressors added as described in experimental captions. In general, overnight cultures were diluted 1:1000, and ~20 mL cells were harvested at an OD 600 of 0.2-0.3. Cells were lysed and AP was assayed as described for cytoplasmic AP.

| Assay of isopropylmalate isomerase
The pLEUCD3 plasmid (Jang & Imlay, 2007) was transformed into the copper efflux mutant to provide enough isopropylmalate isomerase (IPMI) for assays in crude extracts. In this plasmid, the leuCD genes are controlled by a tac promoter. Cells were grown aerobically from OD 600 0.005 to 0.1 and then induced with 1 mM IPTG. Cupric sulfate was then added, and cells were grown for one more doubling. Cells were centrifuged aerobically; the pellet was then moved to the anaerobic chamber, and all subsequent steps were performed in the anaerobic chamber with anaerobic solutions. The cell pellet was resuspended in ice-cold anaerobic 100 mM Tris pH 7.6, centrifuged again, and finally, resuspended in 1 mL 100 mM Tris. Cells were lysed by sonication, and the cell debris was removed by centrifugation at 17,530× g for 10 min. IPMI was promptly assayed at 235 nm anaerobically at room temperature. The enzyme gradually loses activity during storage (Jang & Imlay, 2007), so assays were performed shortly after cell lysis. Citraconate (20 mM) was used as a substrate (Jang & Imlay, 2007); it was always freshly prepared, and it was dissolved anaerobically in 100 mM Tris pH 7.

| Monitoring activity of the OxyR transcription factor
To test whether copper stress activates the OxyR protein, the expression of a katG-lacZ transcriptional fusion (Liu et al., 2011) was monitored. The fusion was transferred from strain AL441 into the copper efflux strain SE73 by P1 transduction, creating SE87.
Selection was achieved on LB plates containing 20 μg/mL chloramphenicol. Transductants were verified via colony PCR. Cells were grown aerobically in M9-glucose medium at 37°C from 0.01 OD 600 to OD 0.1-0.15, exposed to copper, and harvested after an additional 45 min. Twenty mL of cells were centrifuged, suspended in 10 mL ice-cold 50 mM Tris pH 8.0, centrifuged, suspended in 1 mL of the same cold buffer, and lysed by French press. Cell debris was removed by centrifugation for 20 min at 4°C. β-galactosidase activity was assayed at 420 nm (Miller, 1972).

| EPR determination of copper valence
The EPR analysis was adapted from previous work in our lab (Macomber & Imlay, 2009). Overnight cultures were diluted to OD 600 0.005 and grown to OD 600 0.1 in 1 L in LB or M9 with branched-chain amino acids and glucose at 37°C. Where indicated, copper was included in the medium. Cells were harvested by centrifugation (8000 rpm, 5 min). The cell pellet was washed twice with 5 mL of ice-cold 20 mM Tris-HCl (pH 7.4), and finally, resuspended in 300 μL of ice-cold 20 mM Tris-HCl-10% glycerol (pH 7.4). An aliquot of the cell suspension (250 μL) was loaded into a quartz EPR tube, frozen on dry ice, and stored at −80°C. EPR signals were measured with a Varian Century E-112 X-band spectrophotometer equipped with a Varian TE102 cavity and temperature controller. Cuprous ion cannot be directly detected by EPR; however, redox-active cuprous ion can be detected if it is first oxidized to cupric ion by reaction with H 2 O 2 . Where indicated, 31 μL of 98 mM hydrogen peroxide was added to the cell suspension, creating a final concentration of 10 mM. The cell suspension was immediately added to a quartz EPR tube and frozen on dry ice. The spectra were measured at a temperature of 50 K with the following settings: field center, 2900G; field sweep, 2000G; modulation frequency, 100 kHz; modulation amplitude, 5G; time constant, 0.032; receiver gain, 10,000; and power, 0.6 mW.

| Sample preparation for ICP-MS analysis of copper content
Strains (WT and copA cueO cus mutant) were grown in M9-glucose medium overnight at 37°C. Cells were then precultured for three generations, from OD 600 of 0.0125 to ~0.1, and subcultured to OD 600 0.0125 and grown again to 0.1. Finally, 50 mL was harvested by centrifugation. Pellets were frozen on dry ice and shipped to the Laboratory for Environmental analysis, Center for Applied Isotope Studies, at the University of Georgia.

| Inhibition of glutathione reductase by copper
The ability of copper(I) to inhibit glutathione reductase was quantified in an anaerobic chamber, with all solutions being anoxic. A stock solution of Cu(I) was prepared by incubating 30 μM of copper(II) sulfate with 1 mM ascorbate in 50 mM Tris, pH 8. Spectroscopic scans of Cu(II) confirmed that all copper was reduced within 5 min. To the same Tris buffer at 37°C were added (in order) 100 μM NADPH, 150 μM oxidized glutathione, the desired concentration of Cu(I), and 3.9 nM of yeast glutathione reductase (GOR, Sigma). The mixture was assembled in an anaerobic spectroscopic cuvette that was then capped and transferred out of the chamber to a spectrophotometer.
The amount of GOR activity was determined by monitoring the rate of NADPH oxidation at 340 nm.

| Calculation of cysteine thiol content in periplasmic proteins
The cysteine content of the whole proteome is 1.2% (Rosato et al., 2002). To evaluate the frequency of cysteine residues in periplasmic proteins, the coding sequences were examined for each of the 186 proteins identified by the EcoCyc website (https://ecocyc.org/) as being unambiguously localized to the periplasm. Our goal was to estimate the number of cysteine residues that might be unmodified in the mature periplasmic proteins; therefore, we subtracted from the total count any cysteine residues that belong to leader sequences, that bind iron-sulfur clusters or form covalent linkages to c-type hemes, or that exist in disulfide bonds. In some cases, biochemical or structural data were not available; if so, the presence of disulfide bonds was inferred from the AlphaFold solutions, which are available for all E. coli ORFs (and are accessible through the EcoCyc website). The AlphaFold heuristic was derived from machine learning technology, and we considered that it could be predisposed to predict disulfide bonds for proteins with recognizable leader sequences. Nevertheless, when crystal structures were available and examined, AlphaFold correctly called the cysteine/disulfide status with a single exception (AslA), leading us to conclude that these predictions are reliable. The cysteine status for all 186 proteins is listed in Table S1. The assignments that are based on AlphaFold are labeled.

ACK N OWLED G M ENTS
We thank Jon Beckwith (Harvard Medical School) for providing pAID135, which enabled us to track the activity of alkaline phosphatase that is restricted to the cytoplasm, and Chris Rensing (Fujian Agriculture and Forestry University), who provided the parental copA cueO cus strain. This work was supported by grant GM49640 from the National Institutes of Health.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.

E TH I C S S TATEM ENT
None.