Role of Glutathione in Buffering Excess Intracellular Copper in Streptococcus pyogenes

The control of intracellular metal availability is fundamental to bacterial physiology. In the case of copper (Cu), it has been established that rising intracellular Cu levels eventually fill the metal-sensing site of the endogenous Cu-sensing transcriptional regulator, which in turn induces transcription of a copper export pump. This response caps intracellular Cu availability below a well-defined threshold and prevents Cu toxicity. Glutathione, abundant in many bacteria, is known to bind Cu and has long been assumed to contribute to bacterial Cu handling. However, there is some ambiguity since neither its biosynthesis nor uptake is Cu-regulated. Furthermore, there is little experimental support for this physiological role of glutathione beyond measuring growth of glutathione-deficient mutants in the presence of Cu. Our work with group A Streptococcus provides new evidence that glutathione increases the threshold of intracellular Cu availability that can be tolerated by bacteria and thus advances fundamental understanding of bacterial Cu handling.

in cytoplasmic Cu buffering to supplement the transcriptionally responsive Cu sensing and efflux system. This additional buffering extends the range of intracellular Cu concentrations that can be tolerated by bacteria and thus prevents a sudden or abrupt transition from Cu homeostasis to Cu stress upon exposure to an excess of this metal ion.

RESULTS
Initial characterization of a ⌬copA mutant. The copYAZ operon in GAS has been previously shown to resemble other Cop systems in Gram-positive bacteria (29) (see Fig. S1A in the supplemental material). Consistent with a role in Cu efflux, expression of this operon functionally complemented a heterologous Escherichia coli ΔcopA mutant strain (29). In silico analyses found one additional open reading frame downstream of copZ (see Fig. S1A). It encodes a small, uncharacterized protein (56 amino acids) with an N-terminal transmembrane domain, a putative metal-binding C-X 3 -M-H motif at the C terminus, and no characterized homologue. This gene is absent from copYAZ operons in other Gram-positive bacteria and its function in Cu homeostasis is unknown.
For the present study, a non-polar ΔcopA mutant of GAS M1T1 strain 5448 was constructed. This mutation did not alter basal expression of downstream cop genes (see Fig. S1Bi). As anticipated, the ΔcopA mutant was more susceptible to growth inhibition by added Cu than was the wild type (see Fig. S1C). This mutant also accumulated more intracellular Cu (see Fig. S1D), leading to increased expression of the other cop genes compared to the wild type (see Fig. S1Bii). Marker rescue (copA ϩ ) restored the expression of both copA and wild-type phenotypes (see Fig. S1B to D).

Deletion of copA does not lead to a loss of virulence in a mouse model of infection.
To determine whether the Cop system and its interactions with host Cu have an effect on GAS pathogenesis, an established invasive disease model using transgenic human-plasminogenized mice was used (33). Mice subcutaneously infected with wildtype GAS developed ulcerative skin lesions at the site of injection after 1 day. These lesions were excised 3 days post-infection and were found to contain more Cu than adjacent healthy skin or skin from uninfected mice (Fig. 1A). Consistent with these results, the copYAZ operon was upregulated in GAS isolated from infected mouse tissues compared to those grown in THY medium (34). There was also an increase in Cu levels in mouse blood after 3 days of infection (Fig. 1B). Notably, these Cu levels in the blood are comparable to those measured in the sera of mice infected with the fungal pathogen Candida albicans or the parasite Plasmodium berghei (5). These observations support a model where redistribution of host Cu is a feature of the general immune response to infection (5).
Comparing the survival of mice post-infection, no statistically significant difference was observed whether mice were infected with the wild type or the ⌬copA mutant (P ϭ 0.0991; Fig. 1C). Although no single animal model can fully represent the complex features of human streptococcal diseases (35), consistent with in vivo findings, the ΔcopA mutant was no more susceptible to killing by human neutrophils compared with the wild-type or copA ϩ mutant strains in an ex vivo infection assay (Fig. 1D). In addition, recent reports did not identify the cop genes to be fitness determinants during ex vivo infection of human blood (36) or in vivo soft tissue infection in mice (37). These results imply that, despite the systemic and niche-specific elevated levels of host Cu, the Cu efflux pump CopA is not essential for GAS virulence in this model.
Cu treatment leads to defects in the late exponential phase of growth. The lack of a virulence defect for the ⌬copA mutant in vivo prompted us to examine the impact of Cu treatment on GAS physiology in vitro. Addition of Cu (up to 10 M) to the culture medium did not affect the doubling time of the ΔcopA mutant during the exponential phase of growth, but it did reduce the final culture yield ( Fig. 2A; see Fig. S2A and B). This phenotype was reproduced during growth in the presence of glucose or alternative carbon sources (see Fig. S2C). Under each condition, growth of Cu-treated cultures ceased upon reaching approximately the same optical density at 600 nm (OD 600 ; ϳ0.35) regardless of growth rate, indicating that the growth defect was related to bacterial cell numbers and/or growth stage. Consistent with this interpretation, Cu treatment did not affect growth in the presence of mannose (see Fig. S2C) or limiting amounts of glucose (see Fig. S2D), since neither experimental condition supported growth of GAS beyond an OD 600 of ϳ0. 35.
Parallel assessments of plating efficiency and total ATP levels confirmed that differences between Cu-treated and untreated cultures appeared only in the late exponential or early stationary phase of growth (after ϳ4 h when grown in the presence of glucose; Fig. 2B and C). There were clear decreases in the plating efficiency and ATP production by Cu-treated ΔcopA cultures during this period compared to the untreated control.
Cu treatment leads to metabolic arrest in the late exponential phase of growth. GAS is a lactic acid bacterium. Under our experimental conditions, this organism carried out homolactic fermentation and generated lactic acid as the major end product (see Fig. S3A and B). However, Cu-treated ΔcopA cultures did not acidify the growth medium (see Fig. S3C), leading us to hypothesize that Cu treatment impairs fermentation in GAS.
Consistent with this proposal, Cu-treated ΔcopA cultures produced ϳ50% less lactic acid and consumed ϳ50% less glucose compared to the untreated control ( Fig. 3A; see also Fig. S3Bi and ii). Pyruvate production remained unchanged (see Fig. S3Biii). There is no evidence of a shift toward mixed-acid fermentation since the reduction in lactate FIG 1 Changes in Cu levels during GAS infection and the effect of a copA mutation on GAS virulence in host infection models. (A) Cu levels in mouse lesions. Mice were infected subcutaneously with GAS wild-type strain or left uninfected (n ϭ 10 each). After 3 days, skin from uninfected mice, and both skin lesions and healthy skin adjacent to the lesions from infected mice were excised. Total Cu levels were measured by ICP-MS and normalized to the weight of the tissues. Cu levels in infected lesions were higher than those in adjacent healthy skin (P Ͻ 0.0001) or skin from uninfected mice (P Ͻ 0.0001). (B) Cu levels in mouse blood. Mice were infected subcutaneously with GAS wild-type strain or left uninfected (n ϭ 10 each). Blood was collected and total Cu levels were measured by ICP-MS. Values below the detection limit were represented as zero. Cu levels in the blood of infected mice on days 1 and 3 were higher from those in the blood of uninfected mice (***, P ϭ 0.0001; ****, P Ͻ 0.0001). ns, P ϭ 0.81 (versus uninfected mice). (C) Virulence in an in vivo mouse model of infection. Mice were infected subcutaneously with GAS wild-type (WT) or ΔcopA mutant strains (n ϭ 10 each). The number of surviving mice was counted daily up to 10 days post-infection. Differences in survival curves were analyzed using the Mann-Whitney test, which found no statistical difference (P ϭ 0.099). (D) Virulence in an ex vivo human neutrophil model of infection. Human neutrophils were infected with GAS wild-type (WT), ΔcopA, or copA ϩ strains (n ϭ 3 each). Survival of bacteria relative to the input was measured after 0.5 h. There was no difference between survival of the ΔcopA mutant compared with the WT (P ϭ 0.87) or copA ϩ (P ϭ 0.35) strains.

Stewart et al.
® levels was not accompanied by a concomitant increase in acetate levels (see Fig. 3Biv). Ethanol levels were undetectable (detection limit, ϳ0.2 mM).
Differences in lactate production between Cu-treated and untreated ΔcopA cultures appeared, again, only after ϳ4 h of growth (Fig. 3A). While our methods are not sufficiently sensitive to detect small changes in glucose levels at earlier time points, it is clear that Cu-treated ΔcopA cultures did not consume glucose beyond t ϳ 4 h (see Fig. S3Di). Pyruvate production was, again, not affected at any time point (see Fig. S3Dii). These results suggest that Cu treatment leads to defects in metabolism but only after entry into the late exponential phase of growth.
Cu treatment results in a loss of GapA activity in the late exponential phase of growth. The loss of lactate production, but not pyruvate, implies that lactate dehydrogenase (Ldh) is inactivated (Fig. 3B). To test this proposal, we cultured GAS in the absence or presence of added Cu for 4 h, prepared whole-cell extracts, and measured Ldh activity. Figure 3Ci and 3Di show that Ldh remained active in all strains, regardless of Cu treatment.
What, then, is the target of Cu intoxication in GAS? This bacterium does not possess a tricarboxylic acid cycle or the biosynthesis pathways for multiple amino acids, vitamins, and cofactors (e.g., heme). Thus, it lacks obvious candidate iron-sulfur cluster enzymes that are destabilized by excess Cu ions in other systems (13). In an attempt to develop a molecular explanation for the loss of fermentation, the activity of the two Copper and Glutathione ® GAPDH (glyceraldehyde-3-phosphate dehydrogenase) enzymes in GAS, namely, the classical, phosphorylating, ATP-generating GapA and the alternative, nonphosphorylating GapN, was examined (Fig. 3B). GapA has been identified as a target of Ag and Cu poisoning in E. coli (38) and Staphylococcus aureus (39), respectively, and as such, it is a likely candidate for Cu poisoning in GAS. As expected, Cu treatment led to a decrease in GapA activity in ΔcopA mutant cells ( Fig. 3Cii and Dii), which would explain the reduction in lactate secretion (Fig. 3A) and ATP production (Fig. 2C). The reduction in GapA activity would also cause upstream glycolytic precursors to accumulate, with consequent feedback inhibition of downstream enzymes (40), as well as glucose phosphorylation and uptake (see Fig. S3Bii and S3Di) (41,42).
This Cu-dependent inhibition is specific to GapA since there was no reduction in GapN activity ( Fig. 3Ciii and Diii). Given that there was no detectable change in GapA protein levels in cell extracts (see Fig. S3E), these observations are consistent with mismetalation of GapA, as established recently for the GapA homologue in S. aureus (39). The excess Cu ions likely bind to the conserved Cys and His residues at the catalytic site, as suggested previously for the binding of Ag ions to GapA from E. coli (38). Cu treatment suppressed lactate production in the ΔcopA cultures (P Ͻ 0.0001). (B) Fermentation pathway in GAS. Enzymes of interest, namely, GapA (NAD ϩ -dependent GAPDH, M5005_SPy_0233), GapN (NADP ϩ -dependent GAPDH, M5005_SPy_1119), and Ldh (lactate dehydrogenase, M5005_SPy_0873) are shown. (C and D) Activity of glycolytic enzymes Ldh (i), GapA (ii), and GapN (iii). GAS strains were cultured for t ϭ 4 h with 0 or 1 M added Cu (n ϭ 3) (C) or 0 or 5 M added Cu (n ϭ 4) (D). Enzyme activities were determined in cell extracts. Cu treatment decreased GapA activity in ΔcopA cultures (***, P ϭ 0.0004). ns, P ϭ 0.14. (E) GapA activity over time. GAS ΔcopA mutant strain was cultured with added Cu as indicated for t ϭ 2 h (n ϭ 3), 3 h (n ϭ 7), or 4 h (n ϭ 7). Enzyme activities were determined in cell extracts. Cu treatment did not have an effect on GapA activity at t ϭ 2 h (P ϭ 0.99) or 3 h (ns, P ϭ 0.18), but it strongly inhibited GapA activity at t ϭ 4 h (P Ͻ 0.0001). All statistical analyses were versus 0 M Cu.
Remarkably, when cultures were sampled earlier (at t ϭ 2 and 3 h), no difference was observed between GapA activity in Cu-treated and control ΔcopA cells (Fig. 3E). The timing of GapA inhibition, i.e., at the onset of the late exponential phase of growth (at t ϭ 4 h; Fig. 3E), coincided with the arrest in bacterial growth and metabolism, supporting the hypothesis that GapA is a key target of Cu intoxication in GAS.
Cu treatment leads to misregulation of metal homeostasis in late exponential phase of growth. The puzzling but consistent, 4-h delay in the onset of all observable phenotypes led us to hypothesize that there was a time-dependent shift in Cu handling by GAS. To test this proposal, the response of the Cu sensor CopY was measured by monitoring expression of copZ during growth in the presence of the lowest inhibitory concentration of added Cu (0.5 M; see Fig. S2Aii). The results show that copZ transcription was upregulated ϳ4-fold immediately upon Cu exposure (t ϭ 0 h, in which ϳ12 min passed between the addition of Cu into the culture, centrifugation, and the addition of lysis buffer; Fig. 4A). This level of upregulation remained largely unchanged during growth (measured up to 5 h; Fig. 4A), even though intracellular Cu levels continued to rise (see Fig. S4). These results suggest that the CopY sensor became fully metalated and expression of copZ reached its maximum at t ϭ 0 h post-challenge with added Cu. These data also establish that the copYAZ operon is transcriptionally induced before the onset of observable growth defects (hereafter referred to as Cu "stress").
We concurrently measured the expression of genes that are controlled by other metalloregulators, namely adcAII (regulated by AdcR, a MarR-family Zn-sensing transcriptional corepressor [43]), siaA (controlled by MtsR, a DtxR-family Mn/Fe-sensing corepressor [44]), and cadD (regulated by CadC, an ArsR-family Zn/Cd-sensing derepressor [45]). Clear changes in the expression levels of all three genes were detected in response to Cu treatment. While adcAII and siaA were downregulated, cadD was upregulated ( Fig. 4B to D). Each of these transcriptional responses indicates metalation of the corresponding metallosensor ( Fig. 4B to D), but whether by the cognate metal or by Cu cannot be distinguished. These observations were further corroborated by results from genome-wide RNA sequencing (RNA-seq) analyses. Multiple AdcR-and MtsR-controlled genes were negatively regulated, while both the CadC-controlled genes were positively regulated in response to 5 M added Cu (Table 1; see Data Set S1). Interestingly, no clear effect on gczA or czcD expression was detected, suggesting that the metalation status of GczA, a TetR-family Zn-sensing derepressor (46), is not altered by Cu treatment.
Crucially, changes in the expression of adcAII, siaA, and cadD appeared only after ϳ4 h of growth ( Fig. 4B to D). These transcriptional changes were not accompanied by increases in total intracellular Zn, Mn, or Fe levels (see Fig. S4). Thus, the simplest model that accounts for the sudden metalation (or mismetalation) of multiple metallosensors, as well as GapA, is that excess Cu is released from an intracellular buffer, leading to mislocation of Cu to adventitious binding sites and/or redistribution of intracellular metals.
The onset of the Cu stress phenotype coincides with depletion of GSH. What comprises the intracellular buffer for excess Cu in GAS? This organism does not possess a homologue of the metallothionein MymT (17) or the Cu storage protein Csp (47). Instead, this buffer likely consists of a polydisperse mixture of cytoplasmic small molecules or metabolites (48). Noting that GAS is auxotrophic for most nutrients, including multiple amino acids, vitamins, nucleobases, and GSH, we hypothesized that: (i) one or more of these nutrients constitute the intracellular Cu buffer, either directly by coordinating Cu or indirectly by acting as a synthetic precursor to the buffer, and that (ii) these nutrients become exhausted from the extracellular medium during bacterial growth, leading to the observable effects of Cu stress.
The above hypothesis was tested using two complementary approaches and the results identified GSH as the key limiting nutrient. First, mass spectrometry was employed to measure consumption of nutrients from the growth medium. Several amino acids, the nucleobases adenine and uracil, as well as GSH (and/or its disulfide GSSG) were nearly or completely spent after ϳ4 h of growth (see Fig. S5). Cys and its disulfide were below detection limits. Next, the culture medium was supplemented with each or a combination of the spent or undetected extracellular nutrients. Their ability to restore growth of Cu-treated ΔcopA mutant cultures was subsequently examined. Only supplementation with GSH was strongly protective against Cu intoxication (see Fig. S6).
The GAS genome encodes neither the common pathway for GSH biosynthesis (GshAB) nor the bifunctional glutathione synthetase (GshF [49]). Instead, an uncharacterized homologue of the GSH-binding solute-binding protein GshT is present (M5005_Spy0270, 59% sequence identity, 74% sequence similarity with the characterized homologue from S. mutans) (50). GshT, in conjunction with the endogenous cystine importer TcyBC, likely allows GAS to import extracellular GSH (␥-Glu-Cys-Gly) into the cytoplasm (50). This system may also import ␥-Glu-Cys or Cys-Gly (50), but addition of these dipeptides, or Cys alone, or a mixture of the amino acids Glu, Cys, and Gly did not improve growth of Cu-treated ΔcopA mutant cultures (Fig. 5A). Altogether, these results suggest that: (i) the protective effect of GSH is unlikely to result from chelation of extracellular Cu ions by free thiols, (ii) extracellular GSH is depleted during growth of GAS, and (iii) this depletion is responsible for the observable Cu stress phenotypes. Consistent with propositions ii and iii, addition of GSH completely suppressed the effects of Cu treatment and restored plating efficiency, as well as glucose consumption, lactate secretion, and ATP production beyond the late exponential phase of growth (Fig. 5B to E).
GSH contributes to buffering of excess intracellular Cu. To test that the protective effect of GSH is not linked to chelation of extracellular Cu ions and decreased uptake of Cu into the cytoplasm, total intracellular Cu levels were measured in ΔcopA mutant cultures that were supplemented with Cu and/or GSH. As expected, GSH supplementation did not suppress total intracellular Cu levels when ΔcopA cultures a GAS ΔcopA mutant strain was cultured with or without 5 M added Cu for t ϭ 5 h (n ϭ 3). Total RNA was extracted, rRNA was depleted, and cDNA was generated and finally sequenced by Illumina. Differential gene expression was determined using DeSeq2 and is presented as the fold change (FC) in gene expression in the Cutreated cultures relative to that in the untreated control. Only genes of interest are listed. These are genes regulated by metal-sensing transcriptional regulators CopY, CadC (45), AdcR (43), MtsR (44), and GczA (46), as well as those that encode components of the putative GSH uptake system (50). A complete list of differentially regulated genes is provided in Data Set S1.
were challenged with low concentrations of added Cu (500 nM; Fig. 6A). Surprisingly, at high concentrations of added Cu (5 M), GSH-replete cultures appeared to accumulate higher, rather than lower, intracellular Cu levels (Fig. 6A). Yet, these cultures did not display an observable Cu stress phenotype (Fig. 5). These findings are discussed below. The time-dependent reduction in extracellular GSH levels (see Fig. S5D) was mirrored by a decrease in intracellular GSH (Fig. 6Bi). Both the wild-type and ΔcopA mutant strains contained ϳ4 mM intracellular GSH (and GSSG) at t ϭ 0 h (Fig. 6Bi). This amount was likely already present in the inoculum, which was cultivated in the complex medium THY ([GSH] THY ϳ 30 M [51]). Intracellular GSH levels in both strains reduced to ϳ0.1 mM at t ϭ 4 h, regardless of Cu treatment (Fig. 6Bi and ii). This decrease occurred presumably as a consequence of bacterial growth and replication in a chemically defined medium with a limited GSH supply ([GSH] CDM ϳ 0.5 M; see Fig. S5D). The low amount of intracellular GSH coincided with the onset of the observable Cu stress phenotypes. It might also explain why cultures that grew to low OD 600 values displayed no sign of Cu stress (see Fig. S2C and D); these cultures likely had not depleted their intracellular GSH supply.  (Table 1). This result supports previous transcriptomic studies in several Gram-positive and Gram-negative bacteria, none of which identified GSH biosynthesis or uptake as a key transcriptional response to Cu treatment (15,20,22,23).
Copper and Glutathione ® However, it did allow ΔcopA cells to maintain intracellular concentrations of this tripeptide at ϳ1 mM (one log unit higher than unsupplemented cells) beyond the late exponential growth phase, regardless of Cu treatment (Fig. 6Bii). As mentioned earlier, these GSH-treated cells were Cu-tolerant (Fig. 5). In fact, these cells accumulated more intracellular Cu compared with the GSH-untreated control (Fig. 6A). The simplest explanation for this finding is that the rise in intracellular GSH levels leads to an increased ability to buffer intracellular Cu. A more detailed examination of GSHsupplemented ΔcopA cells confirmed that GapA was protected from inactivation by added Cu (Fig. 6C). In addition, the Cu-induced, time-dependent changes in cadD and adcAII expression were abolished (Fig. 6D), suggesting that CadC and AdcR did not become mismetalated. Some downregulation of siaA transcription was observed, albeit to a lesser magnitude compared with GSH-deplete cultures (Fig. 6D versus Fig. 4D). In general, these results support a model whereby GSH constitutes the major buffer for excess intracellular Cu in GAS and protects potential noncognate binding sites from becoming (mis)metalated by Cu.
Importantly, GSH supplementation did not affect expression of copZ at low concentrations of added Cu (0 to 500 nM; Fig. 6E). This observation further strengthens the proposal that GSH does not rescue the ΔcopA mutant simply by chelating extracellular Cu ions. However, GSH treatment did partially suppress copZ expression in response to a high concentration of added Cu (1,000 nM; Fig. 6E). This observation indicates the relative buffering strengths of GSH and CopY, which are discussed below.

DISCUSSION
Role of GSH in buffering excess cytoplasmic Cu. GSH has been proposed to bind Cu by assembling a stable, tetranuclear Cu 4 GS 6 cluster (52). In such a model, when present at low millimolar concentrations (e.g., ϳ4 mM in GAS at t ϭ 0 h; Fig. 6A), GSH would bind Cu with an apparent affinity of K D ϭ 10 Ϫ16.7 M and thus would impose a threshold of Cu availability at 10 Ϫ16.7 M (see Fig. S7A). This threshold is above the range of Cu availability set by most bacterial Cu sensors (see Fig. S7B) (53)(54)(55). Therefore, GSH contributes to Cu buffering only when the transcriptionally responsive Cu homeostasis system is impaired (e.g., in a ΔcopA mutant [25,26]) or overwhelmed (e.g., when intracellular Cu levels rise above the responsive range of the Cu sensors). Figure 6E shows that supplementation with GSH had little impact on metalation of CopY (and thus expression of copZ) when the amounts of added Cu were low. However, GSH appeared to dampen CopY response at higher concentrations of added Cu, indicating that this thiol competes with CopY for binding Cu when intracellular Cu levels are high. Hence, the thresholds of intracellular Cu availability set by GSH and CopY may overlap, at least partially, with GSH being the weaker buffer (52,55,56). The thermodynamic model in Fig. S7B is compatible with these experimental data, but it will need refinement. This model was estimated using known parameters (Cu affinity, DNA affinity, and number of DNA targets) for CopY from S. pneumoniae (CopY Spn ) (55), but CopY Spn differs from CopY GAS in several key aspects. CopY Spn lacks one of the two Cys-X-Cys motifs found in other CopY homologues such as CopY GAS and CopY from E. hirae (CopY Eh ) (see Fig. S7D). CopY Spn binds two Cu atoms per dimer in a solventexposed center while CopY Eh binds four Cu atoms per functional dimer and assembles a solvent-occluded center (55,57). In addition, two cop boxes are present in S. pneumoniae (58), while only one is found in GAS. How these differences shift the threshold model will need to be examined using careful in vitro studies with purified proteins and DNA. In the simplest scenario, an increase in the stability (affinity) of the bound Cu atoms in CopY, which may occur as a consequence of coordination by extra Cys ligands, would lower the threshold of Cu availability set by CopY (see Fig. S7C) and thus better fit our experimental data.
Depletion of intracellular GSH to 0.1 mM at the late exponential phase of growth would weaken its buffering capacity by at least 2 log units (see Fig. S7A). Figure 4 shows that Cu is then able to metalate nonspecific binding sites in non-cognate metallosensors or metalloenzymes. These results further suggest that AdcR, CadC, and MtsR can allosterically respond to Cu and differentially regulate expression of their target genes in vivo. Precisely how this occurs will need to be confirmed with purified proteins and DNA in vitro. Cu-responsive regulation of genes under the control of non-cognate metallosensors has indeed been reported both in vivo and in vitro, although not for the families of regulators described here (15,28,(59)(60)(61).
Not all bacteria use GSH as the major cytoplasmic thiol. Some bacilli, such as B. subtilis and S. aureus, produce the glycoside bacillithiol (BSH) instead. The affinity of BSH to Cu is at least 2 orders of magnitudes tighter than that of GSH (56,62). Hence, BSH likely imposes a lower limit on cytoplasmic Cu availability than does GSH, but it is worth noting that its intracellular level is ϳ30 times lower than that of GSH (63). Importantly, the relative order with the endogenous Cu sensor CsoR still holds, with BSH binding Cu at least 3 log units more weakly than does CsoR (54). Indeed, this thiol is also thought to contribute to Cu homeostasis by buffering excess Cu. Deletion of the B. subtilis bshC gene for BSH biosynthesis led to a slight increase in copZ expression in response to added Cu. This result mirrors the finding in Fig. 6E and suggests that the Cu sensor CsoR is more readily metalated by Cu in the absence of the major buffering thiol (64). It is also notable that the identification of GapA as a major reservoir of excess Cu ions in the cytoplasm was in a strain of S. aureus that does not synthesize BSH (39).
In summary, this study provides a new line of evidence that Cu handling in the bacterial cytoplasm, when formulated using the threshold model, comprises two components (Fig. 7). The transcriptionally responsive component, which includes the Cu sensor, Cu efflux pump, and additional Cu-binding metallochaperones, functions in housekeeping or homeostasis and sets a low limit of Cu availability in the cytoplasm. Rising Cu levels can saturate this homeostasis system and sudden Cu shock can overwhelm it, but the transcriptionally unresponsive component, in this case GSH, buffers the excess Cu and confers additional Cu tolerance. This second system acts as the final layer of protection before cells experience widespread mismetalation and, therefore, Cu stress (Fig. 7). This additional buffering essentially extends the range of cytoplasmic Cu availability that can be tolerated by bacteria, allows bacteria to maintain key cellular functions, and thus prevents an abrupt transition from Cu homeostasis to Cu stress upon exposure to an excess of this metal ion. While this study focused exclusively on a Gram-positive bacterial organism, this concept is likely to apply to other bacterial systems and mammalian models (65).
Role of GSH in buffering bacterial Cu during host-pathogen interactions. This study was conducted originally to examine the role of the Cop Cu homeostasis system in GAS pathogenesis. Although GAS occupied a Cu-rich environment in mice (Fig. 1A), Threshold model for bacterial Cu homeostasis, tolerance, and stress. As Cu levels in the cytoplasm increase, this metal ion binds to the allosteric site of the Cu-sensing transcriptional regulator, which subsequently induces expression of the Cu efflux pump. Together, the Cu sensor and efflux pump impose a low limit of Cu availability and maintain Cu homeostasis. A further rise in cytoplasmic Cu levels saturates the Cu homeostasis system and begins to fill binding sites in GSH. Since there are no observable defects in bacterial metabolism or growth at this stage, GSH can be considered to confer Cu tolerance. GSH depletion or a further increase in cytoplasmic Cu levels saturates this tolerance capacity. Cu now binds to noncognate metal-binding sites, leading to inhibition of bacterial metabolism and growth. These conditions are considered as Cu stress.
inactivation of the copA gene did not lead to a reduction in GAS virulence (Fig. 1C). Our in vitro investigations now suggest that GAS may withstand host-imposed increases in Cu levels, as long as it has access to a source of GSH in vivo. Indeed, GSH was detected in the skin ulcers of infected mice, but interestingly, the amount was ϳ25-fold less compared to skin from healthy mice or healthy skin from infected mice (see Fig. S8). Whether this depletion of GSH is a feature of the general host immune response, a consequence of inflammation and/or host tissue necrosis, or a consequence of GAS metabolism is not known. Nevertheless, the virulence of the ΔcopA mutant implies that the level of host GSH, albeit reduced, can support Cu buffering inside the GAS cytoplasm. Alternatively, the level of host Cu (Fig. 1A) may not be sufficient to overwhelm the Cop homeostasis system, since the copYAZ operon was only slightly upregulated in bacteria isolated from mouse ulcers (average log 2 FC ϭ 1.13 versus THY) (34).
Link between the failure to buffer Cu and redox stress. Under our experimental conditions, untreated ΔcopA cells contained 17,000 to 23,000 Cu atoms when sampled at t ϭ 3 h (before the onset of Cu stress). Cu treatment increased this number to 78,000 to 330,000 atoms (see Fig. S4). The intracellular GSH concentrations at the same time point ([GSH] i ϭ 0.76 mM; Fig. 6B) would translate to ϳ500,000 molecules of GSH, which are clearly insufficient to buffer all of the intracellular Cu ions. However, there was no observable Cu stress phenotype at this time point, suggesting that the excess Cu is also bound to other cytoplasmic component(s). These components may include CopZ and/or the novel, uncharacterized protein CopX (see Fig. S1A). This idea will be the focus of future studies.
Finally, the GSH/GSSG couple is the major redox buffer of the cell. Assuming that the GSH/GSSG ratio remains unchanged, depletion of intracellular GSH in GAS from ϳ4 to ϳ0.1 mM would raise the cytoplasmic redox midpotential by ϳ46 mV. This relatively more oxidizing environment, when combined with a lack of Cu buffering, may promote the Cu-catalyzed generation of reactive oxygen species (66) or the formation of disulfides (67). Yet, our RNA-seq results do not suggest widespread oxidative stress (see Data Set S1). In E. coli, deletion of gshA did not accelerate DNA damage in Cu-replete cells, even in the presence of added H 2 O 2 (68). Similarly, proteomic analyses of a non-BSH-producing strain of S. aureus indicated that Cu treatment does not induce a strong oxidative stress response in this organism (39).
Regardless of the relative importance of mismetalation versus redox stress, our work demonstrates that excess Cu is not bacteriotoxic as long as cytoplasmic GSH is abundant and thus able to buffer the excess of this metal ion (Fig. 7). In GAS, a GSH auxotroph, this intracellular buffer is dynamic; its levels change during bacterial growth and/or in response to extracellular GSH availability. Future studies should take these effects into account when examining the impact of Cu treatment on bacterial cultures. Had our work not identified the 4-h time point as metabolically relevant, sampling cultures 1 h earlier would have led to a different conclusion.

MATERIALS AND METHODS
Data presentation and statistical analyses. We follow recent recommendations regarding transparency in data representation (69,70). Except for growth curves, individual data points from independent experiments are plotted, with shaded columns representing the means and error bars representing standard deviations. Growth curves show the means of independent experiments, with shaded regions representing standard deviations. The number of independent experiments is stated clearly in each figure legend. Statistical analyses have been performed on all numerical data, but notations of statistical significance are displayed on plots only if they aid in rapid, visual interpretation. Otherwise, P values for key comparisons are stated in the figure legends. Unless otherwise stated, statistical tests used two-way analysis of variance using the statistical package in GraphPad Prism 8.0. All analyses were corrected for multiple comparisons.
Ethics statement. Animal experiments were conducted according to the Guidelines for the Care and Use of Laboratory Animals (National Health and Medical Research Council, Australia) and were approved by the University of Queensland Animal Ethics Committee (Australia). Human blood donation for use in neutrophil killing studies was conducted in accordance with the National Statement on Ethical Conduct in Human Research and in compliance with the regulations governing experimentation on humans, and was approved by the University of Queensland Medical Research Ethics Committee (Australia).