Copper Induces Protein Aggregation, a Toxic Process Compensated by Molecular Chaperones

The presence of excess Cu²⁺ in the growth medium has been shown to trigger protein aggregation in bacteria grown under aerobic conditions (22). However, it is difficult to untangle whether protein aggregation is caused by Cu²⁺, Cu-mediated ROS production, or directly by Cu⁺. We therefore decided to monitor growth in the presence of copper either under anaerobic conditions to avoid the Cu-induced Fenton reaction or under aerobic conditions. In addition, we evaluated two other parameters: we quantified the amount of intracellular copper content via inductively coupled plasma optical emission spectrometry (ICP-OES) and analyzed intracellular protein aggregation by SDS-PAGE. For all of these experiments, we exposed cells grown under either anaerobic or aerobic conditions to various CuSO₄ concentrations for 20 min (Fig. 1A). We then washed the cells to remove any adventitiously associated copper and serially diluted the bacteria onto LB agar plates to determine cell survival (Fig. 1A and B). As expected, we found that copper is highly toxic to bacteria, yet the copper toxicity was significantly more pronounced under anaerobic conditions. While we observed a significant 4-log decrease in cell survival in the presence of 5 mM CuSO₄ under aerobic conditions, 0.75 mM CuSO₄ was sufficient to cause the same effects under anaerobic conditions (Fig. 1B). We also found massive differences in the degree of protein aggregation between aerobic and anaerobic copper stress (Fig. 1C), concomitant with substantial differences in their intracellular copper contents (Fig. 1D). E. coli cells, stressed with copper under anaerobic conditions, accumulated higher levels of intracellular copper and revealed a disproportionally high accumulation of insoluble protein aggregates (Fig. 1C and D). These results revealed that E. coli is sensitive toward copper-induced protein aggregation under anaerobic growth.

To further investigate the differences observed between cells exposed to copper in the presence or absence of oxygen, we performed in vitro studies using cellular extracts. We reasoned that by using ex vivo assays, we will be able to tightly control the Cu redox status and directly investigate the impact of Cu²⁺ and Cu⁺ ions on the stability of proteins. Therefore, we prepared soluble protein extracts of E. coli cells and exposed them to different Cu treatments, which we will refer to as “Cu⁺ stress” or “Cu²⁺ stress.” Cu⁺ stress [i.e., addition of tetrakis(acetonitrile) copper(I) hexafluorophosphate under strictly controlled anaerobic conditions to prevent any ROS production] mimics the accumulation of intracellular cytoplasmic copper under anaerobic conditions. In contrast, Cu²⁺ stress (i.e., addition of CuSO₄ under aerobic conditions) represents reactions that typically occur when cells grow under aerobic conditions in the presence of a high level of CuSO₄. We then treated the cell lysates with previously established in vitro concentrations (19, 20) of 100 or 500 μM Cu²⁺ stress under aerobic conditions or Cu⁺ stress under anaerobic conditions for 30 min, separated the soluble and insoluble proteins by centrifugation, and monitored the proteins by reducing SDS-PAGE (Fig. 2A). To our surprise, we did not observe any significant differences in the extent of protein aggregation between Cu⁺ stress- or Cu²⁺ stress-exposed cell lysates, suggesting that the differences observed in vivo are not due to the Cu redox state but might be due to the higher levels of intracellular copper concentrations that accumulate under anaerobic conditions (Fig. 1C and D). Moreover, these results clearly show that Cu⁺-induced protein aggregation is not the result of oxidative damage but must be triggered directly by Cu⁺. To independently confirm this conclusion, we treated the cell lysates with AgNO₃ under anaerobic conditions. Ag⁺ has a similar coordination chemistry to Cu⁺, yet in contrast to Cu⁺, Ag⁺ is not redox active and does not catalyze the Fenton reaction. Like Cu⁺, we found that Ag⁺ induces the aggregation of numerous E. coli proteins (Fig. 2A).

The bacterial cytosol contains high levels of reduced glutathione (GSH), which is able to bind to Cu⁺ and hence protects cytoplasmic proteins from copper-induced damage (13, 21). To test the impact of GSH on protein-induced aggregation in vitro, we incubated CuSO₄ (i.e., Cu²⁺) with physiological concentrations of GSH (5 mM). Under these conditions, Cu²⁺ is immediately reduced to Cu⁺ and tightly bound by the tripeptide (from here on “Cu-GSH”) (23). To prevent any Cu⁺ or cysteine oxidation by dioxygen or ROS formation, we conducted these experiments also under fully anaerobic conditions. Incubation of cell lysates with Cu-GSH for 30 min under the conditions of our experiment caused substantial protein aggregation (Fig. 2A). This observation is consistent with the above result (Fig. 1) showing the formation of aggregates in vivo induced by copper.

However, GSH is known in vivo to protect cells against intracellular copper toxicity (21). To better understand the role played by GSH in vivo under copper stress (aerobic versus anaerobic), we measured cell viability in the presence of copper of two mutant strains affected in their ability to produce GSH. These strains lack either gshA (encoding a γ-glutamylcysteine synthetase) or gshB (encoding a glutathione synthetase). After exposure to copper stress, the two mutant strains exhibited a significant decrease in cell survival compared to the wild-type (WT) strain under anaerobic conditions (see Fig. S1 in the supplemental material). This result confirms that GSH protects cells against copper damage. The effect was observed mainly under anaerobic conditions, suggesting that the high level of intracellular copper measured (Fig. 1B) accumulates in the cytosol and that cell death under this condition is due to cytoplasmic damage. Copper complexation to GSH could limit copper binding to proteins; however, a slight cytoplasmic increase in copper could modify the equilibrium, and depending on Cu⁺ affinity to some proteins, its binding could induce protein aggregation, as observed in vivo (Fig. 1B) and in vitro (Fig. 2A).

To obtain more detailed insights into the impact of Cu⁺ and Cu²⁺ on protein stability, we performed similar experiments but used purified proteins instead. We tested citrate synthase (CS) and firefly luciferase, two well-established aggregation-prone proteins (24), as well as purified EF-Tu from E. coli, which we found to be highly Cu sensitive in cell lysates (see the asterisk in Fig. 2A). As indicated above, we treated the proteins with either Cu²⁺ or Cu⁺ under aerobic or anaerobic conditions, respectively, and monitored protein precipitation after 30 min of incubation. As shown in Fig. 2B, independent of the redox form of copper, all three proteins aggregated within the given time frame of the experiment. Control experiments showed that neither the counterions nor the buffer alone caused any protein aggregation (see Fig. S2 in the supplemental material). Again, we used Ag⁺ to further exclude the involvement of ROS in our Cu⁺-treated samples and found that CS aggregation also occurred in the presence of this non-redox metal (Fig. 2C). Importantly, none of the other metals that we tested, including FeSO₄, NiCl₂, ZnSO₄, and CoCl₂, caused any significant aggregation of CS (Fig. 2C).

To determine the minimum amount of Cu⁺ required to induce CS aggregation, we monitored CS aggregation by light scattering, which is more sensitive than a solubility test, and used rubber-sealed cuvettes to maintain anaerobic conditions. We found that under our experimental conditions, CS began to aggregate when we reached a CS/Cu⁺ ratio of 1:40 (see Fig. S3A in the supplemental material). To test whether lower Cu⁺ concentrations can change the secondary structure of CS without triggering visible protein aggregation, we monitored far-UV circular dichroism (CD) spectra of Cu⁺-treated CS incubated in rubber-sealed cuvettes (Fig. 2D). We observed that a 4-fold molar excess of Cu⁺ ions to CS is sufficient to cause a significant loss of signal from the initial circular dichroism spectrum, followed by the inactivation of CS (Fig. 2E). We obtained very similar results when we treated CS with Cu²⁺ under aerobic conditions (Fig. 2E; Fig. S3). We concluded from these results that both redox states of Cu cause a loss of protein structure and function in vitro.

We next asked whether proteins differ in their aggregation sensitivity following Cu²⁺ or Cu⁺ stress. To address this issue, we conducted a comparative proteomic analysis of aggregates from lysates treated with 100 μM Cu⁺ under anaerobic conditions or with 100 μM Cu²⁺ under aerobic conditions (Fig. 2A). As controls, we analyzed the protein aggregates from lysates incubated for 30 min at 30 or 45°C. To identify and quantify changes in protein abundance in the different samples, we applied label-free quantification (LFQ) using the MaxQuant software (25). We identified 1,144 proteins across all three treatments that did not appear in the 30°C control treatment (see Table S1 in the supplemental material). As shown in the Venn diagram (Fig. 3A), all three treatments caused the aggregation of a common set of 668 proteins, which comprises about 58% of the aggregated proteome. Many of these proteins have been previously shown to be aggregation sensitive ( (26, 27). Of the remaining 476 aggregation-prone proteins, most proteins (i.e., 403) were sensitive to Cu⁺ stress, and 182 proteins were exclusively found to aggregate upon Cu⁺ treatment. In response to Cu²⁺ treatment, 78 additional proteins aggregated, of which only 20 proteins were uniquely sensitive to Cu²⁺ treatment. The heat treatment resulted in 41 (from 228) distinctly aggregated proteins.

To quantify stress-specific changes in the abundance of the aggregated proteins, we performed a Student's t test analysis and defined the subset of proteins that was significantly enriched under either heat, Cu²⁺, or Cu⁺ treatment (Fig. 3B; see Fig. S4 in the supplemental material). This protein profiling clearly revealed the differential sensitivity of some proteins toward the type of stress treatment. We compiled a data set of proteins that were uniquely sensitive (Fig. 3A) or significantly enriched (Fig. 3B, more than 4-fold change) relative to the other two conditions (Table S1). A functional enrichment analysis did not reveal any statistically significant preference for any specific pathway or biological function in the aggregated proteome in either of the treatments (see Table S2A in the supplemental material). However, we observed the inactivation of key proteins involved in metabolic pathways (for example, enzymes of the central carbon metabolism) that could explain the global impact on cells. We also noticed that metal binding proteins seemed to be more impacted by Cu⁺ than Cu²⁺ treatment, as illustrated by the overrepresentation of zinc binding proteins, iron transport/storage proteins, or oxidoreductases among the Cu⁺-sensitive proteins (Table S2A). These results are in agreement with previous in vivo reports, suggesting that Cu⁺ targets metalloproteins (13, 14, 17).

The next step in our data analysis was to identify the intrinsic properties that make proteins sensitive to copper stress. The strength of our in vitro approach using cellular extracts is to avoid modifications of the proteome that would occur in vivo in response to copper, such as transcriptional regulation or protein degradation. We compared sequence compositions among our previously selected groups of proteins (Fig. 3C; see Table S1, Table S2B, and Fig. S5 in the supplemental material). This analysis showed that heat treatment affects preferentially larger proteins (i.e., average size of 445 amino acids [aa]) compared to proteins precipitated by Cu²⁺ or Cu⁺ (Fig. S5C). In contrast, Cu⁺-sensitive proteins were found to be slightly enriched in cysteine and histidine residues relative to the Cu²⁺-sensitive proteins (Fig. 3C; Table S2B). As cysteine and histidine residues are known to have a high binding affinity for Cu⁺ at neutral pH (28, 29), this analysis indicates that Cu⁺ might promote protein aggregation by preferentially interacting with these residues. In contrast, Cu²⁺ appears to affect preferentially charged proteins, suggesting that Cu²⁺ might interfere with electrostatic interactions by a mechanism that remain to be determined. Interestingly, the distribution of predicted secondary structural elements was found to be similar in all three protein groups, suggesting that the sequence rather than the structure of proteins defines their sensitivity to unfold upon specific stress. These results demonstrate that although both Cu states can induce protein aggregation in vitro, Cu²⁺ and Cu⁺ treatments target different subsets of proteins.

Our results showed that both Cu⁺ and Cu²⁺ trigger the aggregation of a large number of proteins. We next monitored the expression level of genes known to be involved in the protection of cells against stress conditions that cause protein misfolding, particularly genes involved in the heat shock response (30). We monitored the expression levels of dnaK, dnaJ, and htpG because the three genes encode different classes of canonical molecular chaperones (Hsp70, Hsp40, and Hsp90, respectively). The expression profiling was conducted in E. coli cells treated with different concentrations of CuSO₄ for 20 min under either aerobic or anaerobic conditions (Fig. 4A). All three selected heat shock genes were significantly upregulated in cells growing under anaerobic conditions at the lowest Cu concentration tested (0.3 mM) (Fig. 4B; see Fig. S6A in the supplemental material). Under aerobic conditions, induction of these three hsp genes reached a significant induction level upon the presence of 4 mM CuSO₄ in the medium (Fig. 4B; Fig. S6A). As a positive control, we measured the expression levels of copA (Fig. 4B) and cusF (Fig. S6B) under Cu⁺ and Cu²⁺ stress. Both genes are known to be induced by copper stress and responsible for maintaining copper homeostasis by exporting excess copper into the medium (31, 32). Accordingly, an increase of their expression levels is clearly monitored from a 10- and up to 1,000-fold increase under both stress conditions (Fig. 4B; Fig. S6B). In addition to the heat shock genes, we also monitored the expression levels of tig, a constitutively expressed gene that encodes the well-known chaperone trigger factor as well as cpxP, an indicator of the envelope stress response (33). Whereas the expression levels of tig did not change upon either of the copper treatments (Fig. 4B), cpxP was induced upon the addition of the lowest concentration of copper under both aerobic and anerobic conditions (Fig. S6C). Overall, our results revealed that copper treatment triggers the heat shock and envelope stress response in an apparent effort to mitigate misfolding and protein aggregation. To validate our hypothesis that cells growing anaerobically in the presence of copper do not undergo oxidative stress in contrast to aerobically growing cells, we monitored the expression level of the gene sodA, encoding an antioxidant superoxide dismutase, known to be upregulated by the SoxR regulator in response to oxidative stress (34). We observed a marked overexpression of sodA when cells are exposed to CuSO₄ under aerobic conditions but not under anaerobic conditions (Fig. 4B). This result supports our conclusion that under anaerobic growth conditions, copper stress does not lead to ROS production. We did an additional experiment to confirm the absence of oxidative stress under our anaerobic growth conditions. We used an E. coli strain lacking the periplasmic disulfide isomerase DsbC, a protein known to rearrange incorrect disulfide bonds in periplasmic proteins. This strain has previously been shown to be highly sensitive toward copper stress under aerobic conditions (also observed in Fig. S7 in the supplemental material), presumably due to its inability to repair oxidative stress-induced nonnative disulfide bonds (35). When we treated the cells with copper under anaerobic conditions, however, the deletion of dsbC had no impact on cell survival (Fig. S7), further supporting the conclusion that oxidative stress is not involved in the toxicity of anaerobic copper stress.

Our results raised the question as to the precise role that cytoplasmic molecular chaperones play under Cu⁺ and Cu²⁺ stress in bacteria. The E. coli DnaKJE chaperone machinery assists in protein folding and protects against stress-induced protein unfolding through an ATP-dependent cycle of client protein binding and release (36–38). This chaperone complex has been extensively studied over the years, yet its role under copper stress conditions has not been investigated. We therefore treated E. coli wild-type or ΔdnaK strains for 20 min with different concentrations of CuSO₄ under aerobic or anaerobic conditions and monitored growth on LB agar plates after serial dilutions (Fig. 4C). Under anaerobic conditions, we observed that the ΔdnaK strain is highly sensitive to copper stress and does not survive exposure to 0.75 mM CuSO₄, a concentration that the WT strain tolerates (Fig. 4C). In contrast, under aerobic growth conditions, WT and ΔdnaK strains behave very similarly, and both strains significantly grow in the presence of 5 mM CuSO₄ (Fig. 4C). Similar results were obtained when we tested the Δtig strain, which lacks trigger factor (Fig. 4C). Our results demonstrate a key role of trigger factor and the DnaK system to maintain proteostasis upon short-term exposure to Cu⁺ stress, which leads to widespread protein unfolding and aggregation. They furthermore imply that the toxicity of a short-term treatment with copper under aerobic conditions is independent of protein aggregation and hence cannot be prevented by molecular chaperones. In contrast, when we spotted cells on LB plates containing increasing concentrations of CuSO₄ (long-term exposure) (Fig. 5A), we observed that the growth of both ΔdnaK and Δtig strains is strongly affected under both aerobic and anaerobic conditions (Fig. 5B and C). We also noted that in the presence of copper, the ΔdnaK strain grew more slowly than the WT strain in liquid cultures under aerobic conditions (Fig. S7). Ectopic expression of dnaK or tig restored the copper-dependent growth defect in the respective deletion strains (Fig. 5B and C).

Overall, these results demonstrate that molecular chaperones play central roles in protecting bacteria against exposure to extracellular copper excess.

Our in vitro work demonstrates that both copper redox states trigger protein aggregation, clearly excluding ROS formation as the primary proteotoxic culprit. In vivo, however, we observed protein aggregation when E. coli cells were grown under anaerobic conditions in the presence of copper. Analysis of the intracellular copper concentrations revealed that anaerobically grown bacteria accumulate up to 3-fold more copper intracellularly than aerobically grown bacteria, which might be one explanation of the in vitro versus in vivo discrepancy in our results. Proteomic analysis of the in vitro-formed aggregates, however, also revealed some qualitative differences and demonstrated that Cu⁺-sensitive proteins contain on average a higher number of Cys and His residues than Cu²⁺-sensitive proteins. This result is in agreement with the notion that Cu⁺ (like Ag⁺) is a soft Pearson's acid, which exerts a much higher thiophilicity than the intermediate Pearson's acid Cu²⁺ (44). It is likely that Cu⁺ coordination to Cys and His residues in proteins stabilizes potentially nonnative conformations, triggers local unfolding events, and leads to protein aggregation. In contrast, Cu²⁺ might interfere with protein folding by other means, such as its ability to directly oxidize side chains and/or by indirectly causing side chain modifications through ROS production.

A high level of cytoplasmic GSH protects cells from metal-related damage (45–47). Indeed, previous in vitro studies showed that Cu²⁺ triggers the aggregation or the inactivation of purified proteins, such as bovine serum albumin or fumarase A, which can be effectively prevented by GSH (13, 21). In addition, the deletion of gshA or gshB genes renders cells more sensitive to short-term exposure to copper stress under the anaerobic condition. Under the aerobic condition, we did not observe difference in the viability of the WT and ΔgshA or ΔgshB strains in the presence of copper, most likely because cells might die before any cytoplasmic copper increase. In agreement with our results, a previous report (45) showed, under aerobic conditions, a high sensitivity of the ΔgshA and ΔgshB strains to copper in E. coli strains with the copA gene deleted. Based on our results, we propose that under their conditions, the absence of CopA artificially allows cytoplasmic copper increase, whereas under anaerobic conditions, cytoplasmic copper accumulates without the need to delete copA. Although we cannot rule out that some proteins are being protected by GSH against Cu-induced aggregation, our data showed no significant protective effect of GSH when we tested the impact of Cu-GSH complexes on cellular extracts. Moreover, proteins aggregated in vivo in the presence of a high level of endogenous GSH. These results suggest that a substantial number of proteins can be sensitive to Cu⁺ treatment even in the presence of intracellular GSH. This finding expands on recent work that suggested a gradual impact of copper on cells, depending on the amount of intracellular copper, the first state, which corresponds to the induction of copper efflux systems to maintain intracellular copper homeostasis, the intermediate state, in which copper is buffered by the GSH (a strain deficient in GSH production is more sensitive to increased intracellular copper), and a third state, in which the ability of GSH to cope with copper is limited by increasing intracellular copper concentrations, subsequently leading to protein mismetallation (48) (Fig. 6). Based on our data, we now propose a fourth phase, where high intracellular copper levels cause nonspecific protein binding followed by protein aggregation (Fig. 6).

Intracellular copper measurements revealed that the oxygenation levels during the copper treatment strongly affected the intracellular copper concentrations (Fig. 6). Consistent with previous reports, we found that under anaerobic conditions, cells treated with CuSO₄ accumulate significantly higher levels of copper than cells grown under aerobic conditions (49, 50). Although cells are treated with Cu²⁺ (CuSO₄), intracellular copper is rapidly reduced to Cu⁺, which is therefore the predominant form in the bacterial cytoplasm under all growth conditions (32).

We now consider several possibilities to explain these results. The periplasmic protein CueO, which converts Cu⁺ into Cu²⁺ by reducing oxygen into water, is therefore no longer active under anaerobic conditions (31, 51). Bacteria would thus accumulate periplasmic Cu⁺, which might cross the membrane more easily than the highly charged Cu²⁺ (32). In agreement, Tree et al. (52) observed an increase in intracellular copper in a cueO mutant strain. It is also feasible that Cu transport is affected under anaerobic conditions, either by reducing the activity of the Cu-ATPase pump or by other alternative pathways. In any case, our work underscores the importance of measuring the intracellular amount of copper to clarify the reason for cell death.

Our finding that the absence of the chaperones’ trigger factor or DnaKJE increases copper toxicity supports our conclusion that both short- and long-term exposure to copper stress under anaerobic conditions cause cell death through an imbalance in proteostasis and increased protein aggregation. In contrast, short-term copper exposure under aerobic conditions does not lead to massive accumulation of intracellular copper and causes cell death, even without significant protein aggregation. The reason for aerobic copper toxicity might be related to the oxidative inactivation of critical periplasmic/membrane proteins. A combination of several factors such as oxidative stress but also protein aggregation could explain the toxicity when cells are exposed for a long time to a lower concentration of copper since we observed a requirement for cytoplasmic molecular chaperones (ΔdnaK or Δtig mutant), and periplasmic proteins (ΔdsbC mutant).

Proteostasis imbalance induced by copper could explain the broad impact of copper on cells by the inactivation of several pathways. Our mass spectrometry analysis of copper-sensitive proteins in vitro highlights numerous pathways that could be affected by copper in vivo. For instance, we found that Cu⁺ induces Zur aggregation, which could result in an inability of cells to respond to zinc starvation. In agreement with this result, Kershaw et al. (53) found a decrease in expression of the zur-dependent operon (znuABC) in response to an increase in copper concentration. Such cross talk between different metal homeostasis systems has been shown by others (54) and will represent attractive research topics for future studies.

When using copper as an antibacterial agent, several parameters must be taken into consideration: the main redox species, the complexation of the metal, the exposure time, and the concentration, as well as the oxygen level. Depending on these nonexclusive parameters, cells will react differently. On the other hand, their metabolism, growth phase, and/or gene expression represent other parameters to be taken into account in the future in order to overcome or optimize the death of bacteria. For instance, it was previously shown that strains growing on different carbon sources react differently to copper stress (45).

Protecting cells against proteotoxic agents involves expelling or scavenging the culprits and repairing the damage by refolding proteins and/or preventing their aggregation. We speculated that if the copper-induced death was indeed due to protein aggregation, bacteria lacking molecular chaperones should be more sensitive to copper. Indeed, we were able to demonstrate that cytoplasmic molecular chaperones play a crucial role in protecting bacteria against copper toxicity (Fig. 6). We found that the DnaKJE chaperone machine as well as the trigger factor play a key role in protecting cells against copper stress. Interestingly, other essential metals found in cells (iron, zinc, nickel, etc.) did not induce protein aggregation under the conditions tested, unlike highly toxic nonessential heavy metals (such as arsenic and cadmium), which also trigger protein misfolding and aggregation (55, 56). Although essential for most cells, copper has properties similar to those of highly toxic metals, emphasizing the importance for cells to tightly control the amount as well as the impact of intracellular copper. For these reasons, cellular mechanisms exist to protect the cell against the proteotoxicity of this metal. Indeed, we show in our study that the molecular chaperones tested appear to be active under copper stress, whereas it has been shown that arsenic inhibits chaperone activity (55). Therefore, because the chaperone proteins behave differently in the presence of metals, we can speculate that cells have adapted to survive in the presence of a low concentration of intracellular copper.

Overall, we show that under anaerobic condition (i) intracellular copper accumulates, (ii) copper induces protein aggregation, even in the presence of intracellular GSH, and (iii) major molecular chaperones like DnaKJE and trigger factor protect cells against this threat (Fig. 6). As copper stress reflects an environment often encountered during microbial infection, molecular chaperones could therefore be regarded as potential pharmacological targets. Indeed, inhibition of these molecular chaperones will lower bacterial defense mechanisms and help the immune system to kill invading pathogens. In addition, understanding the molecular mechanism responsible for the Cu⁺/Cu²⁺-mediated cytotoxicity is also of widespread interest in neurodegenerative diseases. Cu⁺/Cu²⁺ has been shown to play a critical role in the progression of diseases, such as Parkinson’s disease, Alzheimer’s disease, or prion disease (57–59). These diseases are associated with protein misfolding and the deposition of aggregated proteins. Our work opens new avenues concerning (i) the development of copper chelating agents that could help prevent Cu-induced cellular damage and (ii) the role molecular chaperones might achieve to protect the cells from this metal.

The Escherichia coli strains and the plasmids used in this study are listed in Table S3 in the supplemental material. Cells were grown either under aerobic conditions in LB medium with shaking at 160 rpm or under anaerobic conditions in LB supplemented with 20 mM MOPS (morpholinepropanesulfonic acid) and 45 mM glucose in Hungate culture tubes in order to maintain an anaerobic environment, without agitation.

The following metal solutions were used. For Cu²⁺, CuSO₄ (Sigma, no. C3036) was resuspended in H₂O. For Cu⁺, tetrakis(acetonitrile) copper(I) hexafluorophosphate (Sigma, no. 346276) was resuspended in ultrapure water with 10% acetonitrile (ACN) (Sigma, no. 271004), stored, and used in the anaerobic chamber. Acetonitrile forms complexes with Cu⁺, which stabilizes it in water, preventing its disproportionation in Cu⁰ and Cu²⁺ (60). As a control, ultrapure water with 10% acetonitrile was used. Cu-GSH is a mixture of 10 mM CuSO₄ with 100 mM freshly prepared GSH (Sigma, no. G4251) used in the anaerobic chamber. This mixture was added to the assay mixture to obtain a final concentration of 500 μM CuSO₄ and 5 mM GSH (when mentioned in the experimental setup, 100 μM CuSO₄ was also prepared with another CU-GSH mixture in order to get a final concentration of 100 μM CuSO₄ in a final concentration of 5 mM GSH). For Ag⁺, AgNO₃ (Normapur, no. 21572) in H₂O was used under anaerobic conditions and stored in the dark. FeSO₄, ZnSO₄, NiSO₄, CoSO₄, and MnSO₄ were dissolved in H₂O and used under aerobic conditions. All the experiments in anaerobic conditions were carried out in a Jacomex glove box under a nitrogen atmosphere.

E. coli cells were grown at 30 or 37°C under anaerobic or aerobic conditions until they reached an optical density at 600 nm (OD₆₀₀) of 0.7. Cells were then exposed to increasing CuSO₄ concentrations for 20 min. Cells were either directly spotted on plates as described below (see also Fig. 4A), or cells were then harvested by centrifugation at 4,000 rpm for 10 min at 4°C and washed (see the scheme in Fig. 1A). For the anaerobic samples, the following washing steps were performed with buffers degassed with argon and in a glove box to maintain an anaerobic atmosphere. The cell pellets were resuspended with 10 mL of phosphate-buffered saline (PBS) buffer (137 mM NaCl, 2.7 mM KCl, 10 mM Na₂HPO₄, 1.8 mM KH₂PO₄) containing 50 mM EDTA. These steps (centrifugation and the washing step) were repeated three times. Three microliters of 10-fold serial dilutions was spotted on LB agar plates under aerobic conditions or on plates supplemented with 20 mM MOPS and 45 mM glucose under anaerobic conditions. For O₂-deprived conditions, the plates in the glove box were directly transferred in anaerobic jars with an anaerobic filter (Anaerocult A). The plates were placed either at 30 or 37°C for 36 h under anaerobic conditions or for 24 h under aerobic conditions.

The E. coli strains mentioned in Table S3 (WT and ΔdnaK::Cm^r or Δtig::Cm^r mutant) were freshly transformed when needed with plasmid pSE380, pSE380(dnaK), or pSE380(tig), respectively. Cells were grown at 30°C in LB medium supplemented with glucose (45 mM), MOPS (20 mM), and, when necessary, with ampicillin (1 mM) until the OD₆₀₀ reached 0.7 under either aerobic or in anaerobic conditions. Ten-fold serial dilutions were prepared, and 3 μL was spotted on LB agar plates containing increasing CuSO₄ concentrations supplemented with 1 mM ampicillin (when cells contained a plasmid), with or without isopropyl-β-d-thiogalactopyranoside (IPTG) (0 to 5 μM). The plates were incubated in an aerobic or anaerobic atmosphere at 30 or 37°C.

The E. coli strains mentioned in Table S3 (MC4100 WT and ΔdnaK::Cm^r mutant) were grown at 30°C in LB liquid medium under aerobic conditions until an OD₆₀₀ of 0.7 was reached. The cells were diluted to a final OD₆₀₀ of 0.1 and were incubated with increasing concentrations of CuSO₄. The growth of these strains was followed by measuring the absorbance at 600 for almost 7 h.

E. coli cells were stressed and washed as described in the section “Cell survival after short-term exposure to copper.” After the third wash, each pellet was resuspended in 500 μL of PBS buffer and 500 μL of 69% nitric acid. The samples were boiled for 40 min, and 4 mL of 3% nitric acid was added to each sample. Copper analyses were performed on an ICAP 6000 series optical emission spectrometer (Thermo Scientific). Serial dilutions of copper standard solution were used to calibrate the inductively coupled plasma optical emission spectrometry (ICP-OES) system.

E. coli cells were exposed to stress under anaerobic or aerobic conditions and then washed as described in “Cell survival after short-term exposure to copper.” Note that the following steps were performed under either aerobic or anaerobic conditions (glove box). After the washing steps, each pellet was resuspended in 120 μL of buffer A (10 mM KH₂PO₄ [pH 6.5], 1 mM EDTA, 20% [wt/vol] sucrose, 1 mg/mL lysozyme) and incubated for 30 min in ice. A total of 1,080 μL of buffer B (10 mM KH₂PO₄ [pH 6.5], 1 mM EDTA) was added to the cells, which were subsequently lysed by sonication, using either a Branson Sonifier 450 (50% duty, level 2, 10 cycles) for aerobic samples or the Ultrasonic processor UP100H (Hielscher) for samples in the anaerobic glove box. After lysis, the cells were centrifuged at 4,000 rpm for 15 min at 4°C to remove unbroken cells. To isolate the insoluble cellular fraction (containing membrane and aggregated proteins), centrifugation at 13,000 rpm for 35 min at 4°C was performed. The pellets were resuspended in 1 mL of buffer B by sonication and centrifuged at 13,000 rpm for 25 min at 4°C. The pellets were resuspended in 960 μL of buffer B by brief sonication, and 240 μL of 10% (vol/vol) NP-40 was added to solubilize the membrane proteins. After homogenization, centrifugation at 13,000 rpm for 35 min at 4°C was performed to isolate the aggregated proteins. This washing and sonication steps were repeated twice to remove most of the membrane proteins, which for an unknown reason, can sometimes be detected in all samples. The pellets were suspended in 60 μL of 6 M urea, then loaded on SDS-PAGE to visualize aggregated proteins. The gels were stained with Coomassie protein stain (Instant Blue).

First, cell extracts were prepared by the following steps. E. coli cells were grown in LB medium under aerobic conditions until they reached an OD₆₀₀ of 0.6. (Note that the following steps were performed under either aerobic or anaerobic conditions [glove box].) Cells were then harvested by centrifugation at 4,000 rpm for 10 min at 4°C. Cells were resuspended in buffer C (40 mM MOPS, 0.2 M KCl [pH 7.5]) and lysed (Ultrasonic processor UP100H [Hielscher], 2 cycles, 4°C, 160,000 Pa). After ultracentrifugation (45,000 rpm, 1 h 30 min), the protein concentration of the supernatant containing the soluble proteins was determined by bicinchoninic acid (BCA) assay (Sigma), and samples were frozen at −80°C. These cell extracts were then incubated with different stressed agents, and aggregated proteins were isolated as described below. The cell extracts (1 mg/mL) were incubated with or without 100 or 500 μM Cu²⁺ (under aerobic conditions), Cu⁺, Cu-GSH, or Ag⁺ (under anaerobic conditions) at 30°C during 30 min. The aggregates (A) and the soluble proteins (S) were separated by centrifugation (11,000 rpm, 40 min). The proteins from the supernatant were precipitated with trichloroacetic acid (10% [vol/vol] final concentration). Both pellets were resuspended in 8 M urea and analyzed by 12% SDS–PAGE or by mass spectrometry.

Citrate synthase (CS) (Sigma), luciferase (Promega), or purified E. coli EF-Tu was diluted into 40 mM MOPS (pH 7.5) at 30°C to a final concentration of 2 μM in the absence or presence of the indicated ratio of Cu²⁺ (under aerobic conditions) or Cu⁺ (under anaerobic conditions). Light scattering at 360 nm was followed using a Cary spectrophotometer for 15 min at 30°C. Alternatively, the samples were taken after 30 min of incubation at 30°C, and protein aggregates were separated from the soluble fraction by centrifugation (11,000 rpm, 40 min) and analyzed by 12% SDS–PAGE.

Far-UV circular dichroism spectroscopy of 2.5 μM CS was recorded in 40 mM KH₂PO₄ (pH 7.5) using a Jasco-815 spectropolarimeter at 25°C. When indicated, a 4-, 8-, 20-, 40-, or 160-fold molar excess of Cu⁺ was added under anaerobic conditions using rubber-sealed cuvettes. All spectra were buffer corrected.

The activity of CS was determined according to the method described by Jakob et al. (61). A 0.15 μM concentration of citrate synthase was incubated with 0, 4, 8, 20, 40, 80, or 160 molar equivalents of Cu²⁺ or Cu⁺ under aerobic or anaerobic conditions, respectively, at 25°C. After 30 min of incubation, the activity of CS was determined. The activity of CS in the absence of copper was set to 100%.

The protein pellets (also called “A” for “aggregates”) obtained as described in the section “Purification of protein aggregates from cell lysate” were resuspended in 100 μL of a mixture of 6 M urea, 50 mM Tris-HCl (pH 8), and 3 mM dithiothreitol (DTT) and incubated for 2 h at 37°C. After cooling and centrifugation, the thiol groups were alkylated by incubation with 0.1 M iodoacetamide in the dark for 1 h at 350 rpm at 25°C. Then, the urea buffer was diluted with the digestion buffer (50 mM Tris buffer [pH 8] plus 10% acetonitrile) to 1 M, and the digestion was carried out overnight at 37°C at 350 rpm using 1.5 μg of Trypsin Gold of mass spectrometry grade (Promega). The peptide concentration was determined, after which the peptides were loaded onto C₁₈, in-house, stage tips in equal amounts as described in reference 62. Three biological replicates for each of the four conditions were subsequently analyzed by liquid chromatography mass spectrometry.

For nanoscale liquid chromatography-tandem mass spectrometry (nano-LC-MS/MS), the peptides (1.5 μg) of each sample were injected and washed with 4% acetonitrile and 0.1% formic acid for 45 min at a flow rate of 300 nL/min and separated on a C₁₈ reverse-phase column coupled to the nano-electrospray EASY-spray device (PepMap, 75 mm by 50 cm; Thermo Scientific) using an Dionex Nano-HPLC (i.e., high-performance liquid chromatography) system (Thermo Scientific) coupled online to an Orbitrap mass spectrometer, Q Exactive HF (Thermo Scientific). The following linear gradient was applied with a flow rate of 150 nL/min at 45°C: from 1% to 28% in 90 min, from 28% to 50% in 17 min, and from 50% to 80% in 10 min, followed by being held at 80% for an additional 13 min and then equilibrated at 1% for 20 min (solvent A is 0.1% formic acid, and solvent B is 80% acetonitrile plus 0.1% formic acid). Additional column washes with 80% ACN for 40 min were carried out between each sample run to avoid potential carryover of the peptides. The Q Exactive HF was operated in a data-dependent mode. The survey scan range was set to 300 to 1,650 m/z, with a resolution of 60,000 at m/z. Up to the 15 most abundant isotope patterns with a charge of ≥2 were subjected to higher-energy collisional dissociation with a normalized collision energy of 27, an isolation window of 1.6 m/z, and a resolution of 15,000 at m/z. To limit repeated sequencing, dynamic exclusion of sequenced peptides was set to 20 s. Thresholds for ion injection time and ion target value were set to 20 ms and 3 × 10⁶ for the survey scans and to 25 ms and 10⁵ for the tandem mass spectrometry (MS/MS) scans. Only ions with the “peptide preferable” profile were analyzed for MS/MS. Data were acquired using Xcalibur software (Thermo Scientific). The raw mass spectrometry data were uploaded to the PRIDE public site (accession no. PXD019288).

For protein identification and quantification, we used the MaxQuant software version 1.6.3.3 (25). We used Andromeda search incorporated into MaxQuant to search MS/MS spectra against the UniProtKB database of the E. coli K-12 proteome (2018 Uniprot release). The enzyme specificity was set to trypsin, allowing cleavage N terminal to proline and a maximum of two miscleavages. Peptides had to have a minimum length of seven amino acids to be considered for identification. Carbamidomethylation was set as a fixed modification, and methionine oxidation was set as a variable modification. A false-discovery rate (FDR) of 0.05 was applied at the peptide and protein levels. An initial precursor mass deviation until 4.5 ppm and fragment mass deviation until 20 ppm were allowed. Only proteins identified by more than two peptides were considered (Table S1). To quantify changes in protein expression, we utilized LFQ using the MaxQuant default parameters (25). The Perseus software was used for statistical analysis, defining significantly different protein profiles as well as for visualization (63). For functional annotation analysis, the DAVID web server (https://david.ncifcrf.gov) was used. Statistical analysis of the sequence features (amino acid propensity, hydrophobicity, net charge, and others) was done using in-house scripts. The hydrophobicity and instability were calculated using the GRAVY index scale (64) and as described by Guruprasad et al. (65), respectively. The frequencies of the disorder and amyloid fragments were calculated using the PASTA 2.0 (26). The sequence features were normalized to the length of a given sequence.

For quantitative real-time reverse transcription-PCR (RT-PCR), the E. coli WT MC4100 strain was cultivated at 37°C under either aerobic or anaerobic conditions until the cells reached an OD₆₀₀ of 0.7. The cultures were incubated with 0, 1, 2, or 4 mM CuSO₄ and 0, 0.3, 0.5, or 0.75 mM CuSO₄, under aerobic or anaerobic conditions, respectively, for 20 min at 37°C without agitation. The cells (around 2 × 10⁹ cells) were centrifuged at 8,300 rpm for 15 min at 4°C, frozen with liquid nitrogen, and stored at −80°C. Total RNAs were extracted from cells using a Maxwell 16 LEV microRNA (miRNA) tissue kit (Promega) and quantified by spectrophotometry at 260 nm (NanoDrop 1000; Thermo Fisher Scientific). One microgram of total RNA and 0.5 μg of primers (Promega) were used with GoScript reverse transcriptase (Promega) to perform DNAc synthesis. To determine the amplification kinetic, the fluorescence of the EvaGreen dye incorporated into the PCR product was measured at the end of each cycle using SoFast EvaGreen Supermix 2× kit (Bio-Rad, France). The gene gapA was used as reference for normalization. For each point, a technical triplicate was performed, and the amplification efficiency for each primer pair was approximately 80 to 100%.