Copper stress upregulates oxidative stress response, histidine production, and iron acquisition genes in Escherichia coli

Abstract

Copper is widely used as a fast-acting antimicrobial, yet the strategies that allow bacteria to survive copper stress remain incompletely understood. Here, we characterize the transcriptional responses of Escherichia coli MG1655 to excess ionic copper using RNA sequencing and a genome-wide GFP-based promoter library. We applied 2 mM copper, which slows growth, and 8 mM copper, a near-lethal concentration. RNA-seq revealed extensive transcriptome remodeling, with 487 genes upregulated at 2 mM and 364 at 8 mM. Both concentrations strongly induced canonical copper-responsive systems, oxidative stress defenses, histidine biosynthesis, and multiple iron acquisition pathways—including enterobactin biosynthesis and transport—despite external iron failing to reduce copper toxicity. At 2 mM copper, additional pathways were activated, including heat-shock and protein-folding functions as well as lipid A, methionine and arginine biosynthesis. Copper exposure also repressed large gene sets: 486 genes at 2 mM, enriched for biofilm formation and pH elevation, and 217 genes at 8 mM, enriched for anaerobic metabolism. In contrast to the robust RNA-seq results, we investigated the Horizon Discovery E. coli genome-wide GFP based promoter library as an alternative screening tool. However, in our experiments it showed low signal-to-noise ratios, limiting its suitability for large scale gene expression screening.

RNA-seq but not a promoter-GFP library detects large-scale transcriptomic rearrangements in response to copper stress in Escherichia coli.

Introduction

Copper (Cu) is an essential transition metal required for correct folding and function of a variety of proteins in bacteria, yet it becomes highly toxic when present in excess. Even the mammalian immune system exploits copper’s toxicity by mobilizing it to kill invading bacteria (Focarelli et al. 2022). The antibacterial activity of copper has been used for centuries by humans and even now, copper is considered the most effective compound to prevent healthcare-associated infections (Salgado et al. 2013, Sifri et al. 2016, Zerbib et al. 2020, Aillon-Garcia et al. 2023). The precise mechanisms by which copper damages bacteria still remain under debate, but several modes of action have been proposed.

One of the main copper toxicity mechanisms is thought to be mis-metallation of proteins by replacing essential metals such as Fe, Mn, and Zn, thereby rendering enzymes nonfunctional (Barwinska-Sendra and Waldron 2017). This includes the disruption of iron-sulfur cluster-containing proteins (Macomber and Imlay 2009, Chillappagari et al. 2010), which are used in a broad range of cellular processes such as central metabolism, respiration and DNA replication. Another proposed mechanism is the generation of reactive oxygen species (ROS) through redox cycling between Cu⁺ and Cu²⁺ in the presence of hydrogen peroxide. Although doubted by some authors (Macomber et al. 2007, Saenkham et al. 2020, Casanova-Hampton et al. 2021), this type of cycling between the two copper oxidation states can generate highly reactive free radicals, leading to oxidative damage of lipids, proteins, and DNA (Sies 1993, Yoshida et al. 1993, Vieira et al. 1997, Nan et al. 2008). In addition to reactive oxygen species generated by ionic copper, the physical interaction of copper surfaces or nanoparticles with bacterial membranes can also cause oxidative damage to membrane phospholipids and ultimately compromise membrane integrity, leading to cell lysis and death (Santo et al. 2008, Hong et al. 2012).

Given the toxicity of copper, bacteria must employ efficient defense strategies to survive in the presence of this metal. Under normal physiological conditions, the level of free copper in the cytoplasm is kept extremely low (Outten et al. 2000). In Gram-negative bacteria, copper detoxification occurs primarily in the periplasm, where most excess intracellular copper is directed (Changela et al. 2003, Macomber et al. 2007, Parmar et al. 2018). During copper stress, the transcriptional regulator CueR senses cytoplasmic Cu⁺ and activates expression of the copper export and detoxification genes copA and cueO (Outten et al. 2000, Yamamoto and Ishihama 2005). CopA is a P-type ATPase that pumps Cu⁺ from the cytoplasm to the periplasm, where the multicopper oxidase CueO oxidizes Cu⁺ into the less reactive and less membrane-permeable Cu²⁺ (Djoko et al. 2010). Simultaneously, the CusS/R two-component system detects periplasmic Cu⁺ and induces the CusCFBA efflux pump, which expels copper from the periplasm to the extracellular environment (Rensing and Grass 2003, Yamamoto and Ishihama 2005). The periplasmic space in Escherichia coli also harbors several proteins that either bind metals or facilitate their transport. Recent research has highlighted the role of periplasmic methionine sulfoxide reductases (MsrPQ), also induced by CusS/R, which repair oxidized methionine residues under copper stress (Vergnes et al. 2022). This repair mechanism contributes to the overall cellular health and function during copper exposure, indicating that E. coli uses multiple strategies to cope with metal-induced stress.

Copper-induced transcriptional responses in E. coli have been characterized previously using DNA microarray analyses. Yamamoto and Ishihama (2005) identified 28 genes that were markedly upregulated and one gene that was downregulated following copper exposure. Their study reported the induction of genes associated with copper homeostasis (cusB, cusC, cusF, yedV, yedW, copA, cueO), molybdenum cofactor biosynthesis (moaB, moaC, moaD, moaE), and envelope stress and protein quality control (cpxP, ebr, spy, yccA, ycfS, yqjA, yebE, htpX), along with several additional genes that could not be clearly assigned to functional categories. A parallel study by Kershaw et al. (2005) compared a near-lethal concentration (2 mM Cu-glycine) and a well-tolerated level (0,75 mM) to a no copper control and counted a total of 340 genes as having significantly altered expression. However, while raw data is available for re-analysis, the publication presents a selected list of functionally grouped genes and not a full list of differentially expressed genes. According to their results, copper stress induced key copper efflux genes (copA, cusF), oxidative stress genes belonging to the SoxRS regulon (soxS, fldA, fur, zwf) and envelope stress response genes controlled by the CpxAR system (cpxP, degP, ompC, ppiD, yccA). Additionally, many genes involved in iron uptake, storage, or Fe–S cluster assembly (fiu, cirA, exbD, fecI, fepA, fepB, fes, fhuF, tonB, yecI, ydiC, ydiE, ynh) were upregulated, including enterobactin biosynthesis genes (entA, entB, entC, entD, entF). The remaining upregulated genes did not cluster into well-defined pathways. Kershaw et al. also reported that most of the downregulated genes encoded proteins involved in motility and respiration. Although informative, these DNA microarray-based studies either identified only a modest number of copper-responsive genes or focused on the alterations in specific pathways and regulons. We therefore sought to revisit copper-induced transcriptional remodeling using RNA sequencing that could reveal additional transcriptomic changes thanks to a more sensitive direct readout and to provide an unbiased view of the cellular response to acute copper stress.

To that end, we exposed mid-log-phase E. coli cells to both growth-slowing and near-lethal copper concentrations and performed RNA sequencing. To validate the RNA-seq data, we tested the possibility to use the Horizon Discovery E. coli genome-wide promoter collection with ∼1900 promoter-GFP fusions.

Material and methods

Media and growth conditions

Bacterial strains were stored as 50% glycerol stocks at −80°C and streaked onto LB agar plates containing the appropriate antibiotic(s) where applicable. Precultures were grown overnight in 5 ml of medium in 50 ml culture tubes at 37°C with orbital shaking at 150 rpm. Throughout this paper we use MOPS medium [0.4 M potassium MOPS, 40 mM Tricine, 100 µM FeSO₄, 95 mM NH₄Cl, 2.76 mM K₂SO₄, 5 µM CaCl₂, 5.28 mM MgCl₂, 500 mM NaCl, 30 nM (NH₄)₆Mo₇O₂₄, 4 µM H₃BO₃, 300 nM CoCl₂, 100 nM CuSO₄, 800 nM MnCl₂, 100 nM ZnSO₄ (Goormaghtigh and Van Melderen 2016)] supplemented with 0.4% glucose, 0.4% casamino acids, and 0.02 mg/ml tryptophan. For experiments involving the copper biosensor strain, the medium was supplemented with double the amount of glucose, casamino acids and tryptophan (see below). For the Horizon Discovery E. coli promoter collection assay and for pH measurements, LB medium [1% tryptone, 0.5% yeast extract, 1% NaCl (Bertani 1951)] was used in parallel to MOPS.

Analysis of CopA induction using copper biosensor

Escherichia coli MG1655 harboring two plasmids, a high-copy plasmid pSLcueR (ampicillin 100 µg/ml), encoding the copper-responsive transcriptional regulator CueR, and a medium-copy plasmid pDNPcopA:: lux (tetracycline 10 µg/ml), in which the copA promoter is transcriptionally fused to the luxCDABE operon, was used. In this strain, bioluminescence is induced by bioavailable copper binding to the CueR regulatory protein. The system has been described earlier in (Ivask et al. 2009) and in this study the original host strain E. coli MC1061 was exchanged to MG1655.

The copper biosensor was grown in MOPS medium (5 ml in 50 ml tubes, 150 rpm) supplemented with double the standard concentrations of glucose, casamino acids, and tryptophan (0.8%, 0.8%, and 0.04 mg/ml, respectively) as well as selective antibiotics (10 µg/ml tetracycline and 50 µg/ml ampicillin). This enriched medium was used to improve growth, as the strain exhibits poor growth in standard MOPS due to the combined burden of the high-copy plasmid pSLcueR and dual antibiotic selection. Overnight cultures were diluted to an OD₆₀₀ of 0.05 in fresh medium and 140 µl of the culture was dispensed into each well of a black, clear-bottom 96-well plate (BRANDplates, catalog no. 781671). The culture was then grown to mid-logarithmic phase (OD₆₀₀ ≈ 0.4) in a BioTek Synergy H1 plate reader (Agilent) and subsequently treated with 10 µl of serial dilutions of CuSO₄ (Biotop) prepared in water, or water alone for the negative control. OD₆₀₀ and luminescence were recorded every 5 min after copper addition.

Statistical analysis was performed using a two-way ANOVA followed by Dunnett’s multiple comparisons test, with correction for multiple comparisons, comparing the control group to each copper concentration at each time point individually.

Determination of bacterial growth inhibition by copper

Escherichia coli MG1655 overnight cultures were diluted to an OD₆₀₀ of 0.05, and 140 µl was transferred to clear 96-well plates unless otherwise specified. Bacteria were grown at 37°C in a BioTek Synergy H1 plate reader (Agilent) with double-orbital shaking at 282 rpm (3 mm orbit) to an OD₆₀₀ of 0.4. For copper treatment, 10 µl of different concentrations of CuSO₄ (Biotop) solution or water was added and cultures were incubated at 37°C with shaking. OD₆₀₀ was recorded every 10 min.

For MIC determination using E. coli BW25112 and the Keio mutants BW25113 hisA:: Km, BW25113 hisH:: Km, and BW25113 hisF:: Km in the MOPS glucose CAA medium, overnight cultures were diluted to an OD₆₀₀ of 0.00073 to conform to a standard MIC assay protocol. For copper treatment, 10 µl of twofold serial dilutions of CuSO₄ (Biotop), starting from 16 mM, or water as a control, was added to the 140 μl cultures. Cells were grown at 37°C in a BioTek Synergy H1 plate reader (Agilent) with double-orbital shaking at 282 rpm (3 mm orbit). Optical density at 600 nm (OD₆₀₀) was recorded every 10 min.

Analysis of copper-induced transcriptional effects

Bacterial exposure to copper for RNA sequencing

Escherichia coli MG1655 (a plasmid-free bacterium from Horizon Discovery E. coli promoter collection, see below) was grown overnight in MOPS medium (5 ml in 50 ml tubes, 150 rpm). Then, the culture was diluted to an OD₆₀₀ of 0.05, and 140 µl was transferred to each well of a clear 96-well plate. Cells were grown at 37°C with shaking in a plate reader to an OD₆₀₀ of 0.4. At OD₆₀₀ = 0.4, 10 µl of water (control) or CuSO₄ (Biotop) was added to final concentrations of 2 or 8 mM, and cultures were further incubated for 30 min at 37°C with shaking. Total 1200 µl of control, 2, or 8 mM CuSO₄ exposed culture was pooled from microplate wells for RNA extraction and immediately stabilized using RNA protect (Qiagen). Samples were stored at −20°C for up to one week prior to RNA extraction using RNeasy Mini Kit (Qiagen). Altogether, four samples from three independent experiments were sequenced. The growth of the technical replicates corresponding to the sequenced samples is shown in Fig. 1A. RNA sequencing was performed by Novogene, including RNA quality control (RIN measurement), DNase I treatment, rRNA depletion, and library preparation. Sequencing output was ≥2 Gb per sample, with ≥85% of bases achieving Q30 quality. Reads were aligned to the reference genome, and gene expression quantification and differential expression analysis were performed by Novogene. Differential expression was defined as P_adj < 0.01 and a ≥2-fold change. The data has been deposited in ENA under accession number PRJEB108196.

Figure 1.

(A) Growth of cultures used for RNA-seq. Growth curves of E. coli MG1655 in MOPS with 2 and 8 mM CuSO₄. Cultures shown are the same ones sampled at 30 min for RNA extraction. The error bars represent mean ± 95% CI of three independent experiments with one to two replicates. (B) Determining an appropriate copper exposure time using a bioluminescent biosensor. Escherichia coli MG1655 carrying pSLcueR and pDNPcopA:: lux (Ivask et al. 2009) responding to 2 and 8 mM CuSO₄ over time. Means ± 95% confidence intervals (CI) of two independent experiments with seven technical replicates each are shown.

Functional Enrichment Analyses

Gene Ontology (GO) biological process enrichment analysis was performed using the ShinyGO web application (http://bioinformatics.sdstate.edu/go/, version 0.85.1) (Ge et al. 2020) with the following parameters: source EnsemblBacteria (E. coli K-12 MG1655 genes GCF_000005845.2), minimal pathway size: 5 and redundancy removal enabled. To assess GO term overrepresentation, sets of upregulated and downregulated transcripts were compared against the background of all detected transcripts. The statistical analysis is built into the ShinyGO application using false discovery rate cutoff of P < 0.001.

Horizon Discovery E. coli promoter collection assay

The E. coli promoter library was obtained from Horizon Discovery (https://horizondiscovery.com/en/non-mammalian-research-tools/products/e-coli-promoter-collection). This genome–scale collection consists of fluorescent reporter strains in which individual E. coli K–12 MG1655 promoters are fused to a bright, fast–folding GFPmut2 on a low–copy plasmid referred to as pUA66 in academic literature (Zaslaver et al. 2006) and pMSs201 commercially. The collection includes over 1900 promoters to enable high–resolution monitoring of promoter activity across the genome. Promoters were defined as intergenic regions longer than 40 bp, and during library construction 50–150 bp from each flanking gene were included upstream of gfp to preserve native regulatory context (Zaslaver et al. 2006). Selected strains from the Horizon Discovery E. coli promoter collection—corresponding to the 12 most strongly up- and downregulated genes identified in our RNA-seq analysis–were sequenced using the pUA66-fw primer (CAACCTTACCAGAGGGCG). Two genes, tnaC and ybcV, were excluded from further testing due to unsuccessful sequence confirmation for the E. coli promoter collection construct. The promoter collection strains were grown in MOPS medium supplemented with 0.4% glucose, 0.4% casamino acids and 0.02 mg/ml tryptophan or LB medium with 25 µg/ml kanamycin. OD₆₀₀ and GFP fluorescence (excitation: 485 nm, emission: 528 nm) were recorded every 10 min. Background was subtracted from fluorescence and OD600 measurements. OD₆₀₀-normalized fluorescence of Cu-treated cultures relative to the no-copper control was calculated for each time point:

Measurement of growth-media pH

Escherichia coli MG1655 was grown in 10 ml MOPS or LB medium at 37°C until OD₆₀₀ ≈ 0.4. Cultures were then treated with CuSO₄ to final concentrations of 2 mM or 8 mM for 30 min, pelleted at 4430 × g for 7 min, and the supernatants were passed through 0.2 µm Minisart syringe filters (Sartorius). pH was measured with SevenDirect SD20 (Mettler Toledo).

Statistical analysis was performed using two–way ANOVA followed by Dunnett’s multiple comparisons test, comparing untreated controls to each copper concentration within each medium.

GFP quenching experiment

Escherichia coli MG1655 was grown overnight in a plate reader as explained in “Determination of bacterial growth inhibition by copper” for 18 h. The stationary phase bacteria were challenged with different concentrations of CuSO₄, AgNO₃ and benzalkonium chloride. OD₆₀₀ and GFP fluorescence (excitation: 485 nm, emission: 528 nm) were recorded every 2 min.

Results

Low millimolar copper affects E. coli within minutes

To determine the growth conditions for investigating the E. coli copper stress response in MOPS medium, we conducted a preliminary minimal inhibitory concentration (MIC) assay. We found that E. coli tolerates 2–8 mM copper with increasing difficulty and 16 mM copper already inhibits growth completely (Supplementary Figure 1). Therefore, we chose 2 mM CuSO₄ as a concentration that slows growth, and 8 mM CuSO₄ as a near-lethal level that only allows initial growth before cell numbers dwindle. Growth curves of the actual cultures used for RNA extraction grown with 2 and 8 mM Cu are shown in Fig. 1A.

As previous transcriptional analyses have used widely different copper treatment times–5 min (Yamamoto and Ishihama 2005) and 10 h (Kershaw et al. 2005)—we then sought to determine an appropriate treatment duration for our RNA sequencing experiments. To do this, we used a copper-responsive biosensor (Ivask et al. 2009) that produces increased bioluminescence in the presence of elevated intracellular copper. Exposure to 8 mM CuSO₄ caused a significant rise in bioluminescence as early as 5 min (P_adj < 0.0001), while 2 mM CuSO₄ produced a significant increase at 20 min (P_adj < 0.0001) (Fig. 1B). Both concentrations continued to induce an increasing signal over time. Based on these results, we selected 30 min as the treatment duration for the subsequent RNA sequencing experiment.

Copper exposure upregulates oxidative stress response, histidine production, and iron acquisition genes

To investigate how E. coli MG1655 responds to copper stress, we performed transcriptional profiling of exponentially growing cells in MOPS medium treated with growth-inhibitory 2 mM copper and near-lethal 8 mM copper (Fig. 1A). With a cutoff of P_adj < 0.01 and a minimum two-fold change, 2 mM copper upregulated 487 genes, whereas 8 mM Cu upregulated 364 genes, with an overlap of 296 genes (Supplementary file 1). This confirmed that the chosen Cu treatment time was relevant and sufficient to induce transcriptomic rearrangements. A heatmap of all genes shows that the replicates are consistent with each other and the variability is due to copper concentration in the experiment (Supplementary Figure 2).

To identify biological functions affected by copper stress, we performed Gene Ontology (GO) analysis to determine which terms were enriched in the Cu-exposed bacteria. Treatment with 8 mM Cu resulted in 13 enriched GO terms, while 2 mM Cu yielded 19 terms, including all those found at 8 mM (Fig. 2A, B, Supplementary file 2). Categories enriched at both Cu concentrations included response to copper ion, response to oxidative stress, histidine biosynthetic process, and multiple GO terms related to iron acquisition, such as E. coli siderophore (enterobactin) biosynthesis. Additional terms enriched only at 2 mM Cu were response to heat and protein folding as well as biosynthetic processes for methionine, arginine and lipid A (Fig. 2A, Supplementary file 2). Below, we describe in detail the genes assigned to each of the major upregulated functional categories.

Figure 2.

Transcriptional profiling of E. coli under copper stress. GO enrichment analysis of E. coli MG1655 genes upregulated at 2 mM Cu (A) and 8 mM Cu (B) and downregulated at 2 mM Cu (C) and 8 mM Cu (D) compared to no Cu control. Only statistically significantly up- or downregulated GO terms that contain at least five genes are shown.

As expected, several known copper responsive genes were upregulated upon elevated Cu exposure. Exposure to 2 mM Cu upregulated 13 genes associated with copper stress response and homeostasis. Of these, nine were identified through GO enrichment analysis (Fig. 2, orange background, Supplementary file 2), and four additional genes (indicated with an asterisk) were identified by manual inspection of the differentially expressed gene set (supplementary file 1). The identified genes are responsible for transporting excess copper out (copA*, zntA*, cusA, cusB, cusC, cusF), regulating the copper response (cusS*/cusR, yedV/yedW*), detoxifying copper in the periplasm (cueO) and possibly trying to keep copper from entering the bacterium (bhsA). Curiously the expression of comR (ycfQ), the repressor for bhsA, has gone up at the same time as bhsA itself. The expression of comR is only modestly upregulated, 2.24-fold at 2 mM Cu and 2.02-fold at 8 mM Cu, whereas bhsA shows much stronger induction, increasing 13.79-fold at 2 mM Cu and 10.91-fold at 8 mM Cu. This is however consistent with the fact that copper causes ComR to disassociate from DNA, releasing repressed promoters (Mermod et al. 2012).

Copper stress strongly induces theiron acquisition response genes (Fig. 2, grey background, Supplementary file 2). In total, 59 iron-related genes are upregulated in the presence of 2 mM Cu, and 55 genes show upregulation at 8 mM Cu (Supplementary file 2). This response includes components involved in the uptake of both ferrous (Fe²⁺) and ferric iron (Fe³⁺). The whole feo system, responsible for Fe²⁺ uptake (feoA, feoB, feoC), is induced. Fe³⁺ acquisition pathways are more complex, yet also strongly upregulated during copper exposure. This includes the induction of the full fec operon (fecA, fecB, fecC, fecD, fecI, fecR), mediating ferric-citrate transport, as well as enterobactin biosynthesis genes (entA, entB, entC, entD, entE, entF, entH, entS, ybdZ) and their corresponding transport and processing genes (fepA, fepB, fepC, fepD, fepG, fes, fhuF).

GO enrichment analysis also revealed induction of amino acid biosynthesis genes in copper-exposed bacteria (Fig. 2, violet background, Supplementary file 2). Histidine biosynthesis genes (hisF, hisA, hisH, hisI, hisG, hisC, hisB, hisD) were enriched at both copper concentrations, while methionine (metA, metF, metH, metE, mmuP, metC, metL, metB, metR, thrA) and arginine (argE, argI, argH, argF, argA, argB, argC, argD) biosynthesis genes were additionally enriched at 2 mM Cu.

Exposure to both 2 mM and 8 mM Cu resulted in a marked enrichment of oxidative stress response genes (Fig. 2 yellow background, Supplementary file 2). A total of 12 genes were upregulated at both copper concentrations, while an additional three genes were induced exclusively at 2 mM Cu and five genes exclusively at 8 mM Cu. Collectively, they are linked with protection against peroxides (oxyR, katG, ahpC, ahpF, hslO, degP, azoR), defence against superoxide O₂⁻ (soxR, sodA, ydbK, ytfK) and Fe-S cluster repair (sufA, sufD, sufE, nfuA, fumC, acnA). The prominence of Fe-S cluster damage under Cu stress is further supported by the strong enrichment of the Gene Ontology term “Fe-S cluster assembly,” encompassing 16 genes involved in Fe-S cluster biogenesis.

Three GO terms associated with heat response and protein folding: “response to heat”, “protein folding” and “chaperone cofactor-dependent protein refolding” are significantly enriched at 2 mM Cu, encompassing 27 unique genes (Fig. 2, pink background, Supplementary file 2). Bulk of these genes encode proteins that reverse or prevent protein misfolding (dnaK, dnaJ, grpE, groL, groS, hscA, hscB, clpB, ibpA, ibpB, htpG, hslO, ppiA) or degrade misfolded proteins (clpP, hslU, hslV, lon, degP, hflK, hflC). A smaller subset represents general or heat-induced stress-response genes (rpoH, rpoD, marR, pspA, sodA, moaA, hslR). Although rpoD functions primarily as the housekeeping sigma factor, it is known to be upregulated under heat-stress conditions (Park et al. 2024).

Exposure to 2 mM Cu resulted in a 3.9-times enrichment of lipid A biosynthesis genes (Fig. 2A, Supplementary file 2). Lipid A is the molecule that anchors the lipopolysaccharide molecule into the outer membrane. In total, 10 lipid A biosynthesis genes (arnT, arnD, arnC, arnE, eptA, lpxC, arnB, arnA, eptB) are upregulated.

Copper exposure represses genes related to adhesion and anaerobic respiration

Exposure to 2 mM Cu downregulated 486 genes while 8 mM Cu repressed 217 genes, with an overlap of 129 genes (Supplementary file 1). Interestingly the enriched GO terms for downregulated genes between 2 and 8 mM Cu do not overlap.

Exposure to 2 mM Cu downregulated 19 genes involved in bacterial adhesion to surfaces belonging to GO categories “cell adhesion” and “cell adhesion involved in single-species biofilm formation” (Fig. 2C, Supplementary file 2). Most of these genes encode confirmed or predicted adhesins (fimH, elfA, elfG, sfmA, yadC, yadL, yadM, ybgO, ycgV, ygiL, ypjA, yqiI). Additional downregulated genes include type 1 fimbriae genes (fimA, fimF, fimG, fimI) known to be important for attachment and biofilm formation (Lasaro et al. 2009). The remaining three genes encode a flagellar biofilm exopolysaccharide PGA synthase pgaD, filament capping protein fliD, and putative c-di-GMP binding protein cdgI.

Additionally, intracellular pH elevation (gadB, adiA, gadC, gadA, kefC, kefF) stood out as a downregulated category in response to 2 mM Cu (Fig. 2C, Supplementary file 2).

Exposure to 8 mM Cu downregulated 12 genes associated with anaerobic energy metabolism, falling into the GO categories “anaerobic respiration,” “glucose catabolic process,” “formate oxidation,” and “anaerobic electron transport chain” (Fig. 2D, Supplementary file 2). These include genes encoding subunits of hydrogenase 1 (hyaA, hyaC) which mediates H₂ oxidation, hydrogenase 2 (hybA, hybB, hybO), which mediates hydrogen oxidation and production (Laurinavichene and Tsygankov 2001, Lukey et al. 2010), as well as hydrogenase 3 (hycB, hycC, hycD, hycF, hycG), responsible for H₂ generation from formate as part of the formate hydrogenlyase complex (Maeda et al. 2007). Additional downregulated genes include fdhF, which provides electrons to hydrogenase 3, and dcuB, encoding the fumarate/succinate antiporter required for fumarate respiration.

Cu-stressed E. coli are not relieved by external iron supplementation

Because our RNA-seq data showed strong induction of iron-acquisition genes during copper stress, we considered the possibility that the cells were experiencing functional iron limitation. To test whether supplemental iron could alleviate copper-induced stress, we measured the copper MIC in the presence of 100 or 200 µM FeSO₄ (Fig. 3). The addition of FeSO₄ did not improve the growth of copper-treated cells, indicating that external iron does not relieve the effects of copper stress under these conditions.

Figure 3.

Influence of iron supplementation on copper-induced growth inhibition. Growth curves of E. coli MG1655 in MOPS medium supplemented with 2, 4, or 8 mM CuSO₄, with or without added FeSO₄. The error bars represent mean ± 95% CI of three independent experiments with three replicates.

Limited utility of the Horizon Discovery E. coli promoter collection for screening copper-responsive promoters

Before measuring the effect of copper on the transcriptome via RNA sequencing, we attempted to study cellular response to copper exposure using a genome-wide E. coli promoter collection from Horizon Discovery. This library contains 1900 promoters from E. coli MG1655, each individually cloned upstream of a fast-folding GFP in a kanamycin-resistant plasmid [pMS201, also referred to as pUA66 in the literature, (Zaslaver et al. 2006)]. This setup enables screening for genes that are up- or downregulated when treated with a chemical by measuring the GFP fluorescence in strains carrying different promoters. When the library strains were exposed to 1 mM copper, the signal-to-noise ratio was extremely low (data not shown), making this approach unsuitable for a reliable primary screening. The same issue of low signal-to-noise ratio was observed also when an attempt was made to study the response of the promoter library to 12.5 µM silver and 2.5 µM benzalkonium chloride (data not shown) referring to the fact that the low signal issue was not specifically related to copper exposure. We therefore explored whether the promoter collection could be used to validate RNA-seq results. From the RNA-seq dataset, we selected 12 genes that were among the most strongly upregulated or downregulated and were present in the promoter collection. The 24 promoter regions in the library were verified by sequencing and two downregulated genes were excluded due to unsuccessful sequence confirmation. Copper concentrations of 1 mM and 0.5 mM were chosen for this experiment as the highest concentrations that did not affect growth of MG1655 carrying the promoter collection library plasmid pUA66 (Supplementary Figure 3). MOPS medium was used to ensure consistency with the RNA-seq experiments, whereas LB medium was selected because it is the manufacturer-recommended growth medium for the Horizon Discovery E. coli promoter collection.

Of the 12 promoters highly upregulated in the RNA-seq dataset, only two exhibited more than a twofold increase in GFP fluorescence [log₂fold change (log₂FC) > 1] compared to the control without copper. The cusC promoter-GFP fusion showed the strongest response among the tested strains and was induced by copper in both used media, LB and MOPS medium (Fig. 4). Induction of this promoter construct started within the first minutes of copper exposure and reached its maximum after 30 min. In the presence of 0.5 mM and 1 mM of copper, fluorescence increased on average 2.9-fold (log₂FC = 1.5) and 2.5-fold (log₂FC = 1.3), respectively. In LB medium those copper concentrations induced the fluorescence by 5.6- and 6.0-fold (log₂FC = 2.5 and 2.6), respectively. Considering that the cusC gene showed a dramatic 2293-fold (log₂FC = 11.2, Supplementary file 2) upregulation in RNA-seq assay, the promoter construct upregulation was truly modest. For the yebE promoter-GFP fusion, no induction was observed in LB medium and in MOPS medium only a modest 2.1-fold increase (log₂FC = 1.1) was observed (Fig. 4). Interestingly, in the presence of 1 mM Cu in MOPS medium, no induction was observed. Analogously to cusC, also yebE was highly upregulated in the RNA-seq assay with 135-fold (log₂FC = 7.1) induction. Moreover, the rest of the 10 promoters selected due to their upregulation in RNA-seq assay did not seem to be activated at all (Fig. 4).

Figure 4.

Using Horizon Discovery E. coli promoter collection to study gene expression changes due to copper stress. Dynamics of OD₆₀₀-normalized fluorescence for GFP-based Horizon Discovery E. coli promoter collection strains relative to the no-copper control over time (log₂ scale). Higher relative fluorescence indicates activation and negative relative fluorescence indicates repression of each given gene promoter. Data represent means of four independent experiments in LB and five in MOPS.

Among the 10 promoters strongly downregulated in RNA-seq, only gadX consistently showed repression in the GFP-based assay (Fig. 4). Excluding the first 6 h, GFP signal from the gadX promoter decreased on average by 2.5-fold (log₂FC = -1.3) with 0.5 mM Cu and 2.7-fold (log₂FC = −1.4) with 1 mM Cu in MOPS. In LB, downregulation was less pronounced, remaining below twofold.

In conclusion, of the 12 promoters identified as highly upregulated by RNA-seq, only two (cusC and yebE) showed any measurable activation, and in both cases the response was modest, with yebE exhibiting only borderline induction. Similarly, among the 10 promoters classified as downregulated in the RNA-seq dataset, only gadX displayed consistent repression in the reporter assay. These results indicate that the Horizon Discovery E. coli promoter collection is not well suited for assessing copper-dependent transcriptional responses under the conditions tested.

One possible explanation for the weak performance of the promoter collection in copper-exposure assays is fluorescence quenching. Both Cu⁺ and Cu²⁺ ions are known to strongly quench ST-Kaede and ST-GFP fluorescent proteins, resulting in near-complete loss of fluorescence in plant cells at around 0.4 mM CuSO₄ (Bourge et al. 2015). To assess whether copper quenches GFP under our experimental conditions, we monitored GFP fluorescence in an E. coli strain carrying a promoter library construct with the strong rpsT promoter. Cultures were grown to stationary phase and then treated with Cu or one of three controls: H₂O, Ag, or benzalkonium chloride (BAC). Aside from a brief drop in fluorescence associated with pausing plate-reader shaking, none of the treatments: copper, silver or benzalkonium chloride, caused a substantial decrease in fluorescence relative to the water control (Fig. 5). Therefore, the low signal-to-noise ratio observed in the Horizon Discovery E. coli promoter collection is unlikely to be attributable to copper-induced quenching of GFP fluorescence.

Figure 5.

Quenching of GFP signal. OD₆₀₀ normalized fluorescence of Horizon Discovery E. coli promoter collection strain MG1655 with pUA66 rpsT in the presence of copper (A), silver (B) or benzalkonium chloride (BAC)(C). Means ± 95% CI of four independent experiments with two to three technical replicates each are shown. “Shaking paused” indicates time points when the plate reader was briefly stopped to add other chemical(s), while nothing was added to this sample.

Copper-induced acidification is minimal in MOPS medium

From previous studies it is known that copper may acidify bacterial growth media (Gunawan et al. 2011). To assess whether the transcriptomic responses to copper exposure could be instead caused by copper-induced growth media acidification, the pH of E. coli cultures after 30 min exposure to 2 and 8 mM copper was measured in the same buffered MOPS medium used for RNA-seq, and LB medium was included as an unbuffered control. In LB, the pH significantly decreased from 6.5 to 5.6 with 2 mM Cu and further to 3.9 with 8 mM Cu (Fig. 6). In the buffered MOPS medium it remained more stable: the pH without added Cu was 6.7, in the presence of 2 mM Cu the pH was also 6.7, and in the presence of 8 mM Cu the pH was 6.1.

Figure 6.

Effect of copper exposure on culture pH. Changes in pH after 30-min copper exposure in E. coli cultures grown in MOPS or LB medium. Means ± 95% CI of three independent experiments with three technical replicates each are shown.

Discussion

Our RNA-seq analysis showed that acute copper stress triggers a broad and coordinated remodeling of E. coli gene expression. This was evident under both concentrations: 2 mM copper, which slows bacterial growth, and 8 mM copper, a near-lethal concentration that only allows limited growth before cell density declines (Fig. 1A; Supplementary Figure 1). Exposure to 2 mM copper elevated the expression of 487 genes and 8 mM copper upregulated 364 genes with a substantial overlap between the two conditions (Supplementary file 1).

The large number of differentially regulated genes detected indicates that RNA-seq does offer a large sensitivity increase compared to the previous microarray study by Yamamoto and Ishihama that detected 28 upregulated genes and a slight advantage over the parallel work by Kershaw et al. that uncovered a total of 340 genes with altered expression (Kershaw et al. 2005, Yamamoto and Ishihama 2005).

Firstly, all canonical copper–stress response genes were activated at both copper concentrations (Fig. 2, orange background, Supplementary file 2). These include the Cu⁺ exporter CopA, the CusABCF efflux system, the periplasmic multicopper oxidase CueO, and the copper–responsive two–component systems CusS/CusR and YedV/YedW (Supplementary file 2). The induction of all major components of the known copper–response network indicates that our treatment conditions were appropriate for eliciting a robust copper stress response.

Our RNA-sequencing data also revealed upregulation of histidine, methionine and arginine biosynthesis genes in response to copper exposure (Fig. 2, violet background, Supplementary file 2). Notably, the growth medium already contains these three amino acids as part of the added casamino acid solution. It has been previously speculated that elevated histidine biosynthesis may be important in protection against copper toxicity (Sullivan et al. 2024) as histidine is a known amino acid to bind both Cu²⁺ and Cu⁺ (Mehlenbacher et al. 2015, Brander et al. 2020). Here we show that in E. coli copper stress upregulates histidine biosynthesis genes with both copper concentrations: growth-slowing 2 mM and near-lethal 8 mM (Fig. 2, violet background, Supplementary file 2). Histidine can in principle bind copper ions that could reduce free copper levels in the cell. However, measuring the copper MIC for E. coli wild-type and three independent histidine biosynthesis deficient mutants yielded an uniform value of 4 mM for all tested strains (Supplementary Figure 2). Therefore, any Cu-protective role that histidine may play is small and can at best be viewed as part of a multilayer defense. Moderate 2 mM copper exposure also upregulated methionine and arginine biosynthesis genes (Fig. 2, violet background, Supplementary file 2). Methionine can also bind both Cu²⁺ and Cu⁺ (Contaldo et al. 2025) but as it does so less effectively than histidine at physiological pH (Rubino et al. 2011), we did not pursue the mechanisms further at this point. Why arginine biosynthesis gets upregulated is even less straightforward as it binds copper very weakly (Wu et al. 2010, Bottari et al. 2019).

In the field of copper toxicity, there is a divide between authors who believe that oxidative stress is an important copper toxicity mechanism and those who believe it not to be. Here we observed that 15 oxidative stress response genes are upregulated during moderate copper stress and 17 under high copper stress (Fig. 2, yellow background). These include genes involved in protection against peroxides, defense against superoxide O₂⁻, and Fe-S cluster (Supplementary file 2). Although the upregulation of oxidative stress response genes suggests that oxidative stress is involved in bacterial copper response, this alone does not establish it as the dominant effect of copper exposure but at the same time does not allow it to be discarded either.

Our RNA-seq data indicates that copper stress robustly induces genes involved in iron acquisition (Fig. 2, grey background, Supplementary file 2). This encompasses genes involved in the uptake of both ferrous Fe²⁺ and ferric Fe³⁺, including those required for enterobactin biosynthesis, processing, and transport. At first glance, this transcriptional profile resembles a state of severe iron starvation. It is known that Cu mis-metallates proteins by replacing Fe (Barwinska-Sendra and Waldron 2017), which suggests that additional Fe may be needed to produce new enzymes. Previous studies have shown that even though the total measurable Fe does not significantly decrease in E. coli cells during copper exposure, the levels of bioavailable Fe can drop (Casanova-Hampton et al. 2021). This is consistent with the activation of iron scavenging pathways. Notably, we show that external iron supplementation does not relieve the copper stress (Fig. 3), which indicates that iron is not the limiting factor during Cu stress.

As mentioned above, we show here that among other iron acquisition genes, many enterobactin biosynthesis, transport and processing genes are upregulated in copper stress (Fig. 2A and B, Supplementary file 2). The link between enterobactin and copper stress has been drawn before, however it is a little complicated. Mutants defective in enterobactin production are more sensitive to Cu (Casanova-Hampton et al. 2021, Peralta Daiana Romina et al. 2022), indicating that enterobactin has a protective role. This has been linked with reducing copper-induced oxidative stress: strains lacking enterobactin are more susceptible to copper-induced oxidative stress, and physiological enterobactin concentrations confer protective effects by reducing ROS levels (Adler et al. 2014, Peralta Daiana et al. 2016, Peralta Daiana Romina et al. 2022). However, high enterobactin levels may promote Cu²⁺ reduction to Cu⁺ and increase toxicity (Peralta Daiana Romina et al. 2022). The copper-siderophore chemistry is context-dependent: enterobactin can act as an antioxidant under certain redox conditions and cellular localizations, but it can also participate in redox cycling that enhances copper toxicity if present in excess or not properly processed (Kim et al. 2001, Grass et al. 2004, Peralta Daiana Romina et al. 2022).

Under moderate copper stress (2 mM), we observed also upregulation of heat–shock-associated genes, which largely overlap with protein folding and refolding pathways (Fig. 2, pink background, Supplementary file 2). This suggests that proteins in Cu–stressed cells experience substantial damage or misfolding. Such disruption is expected, given that copper can mis–metallate proteins (Barwinska-Sendra and Waldron 2017) and destroy Fe-S clusters (Macomber and Imlay 2009, Chillappagari et al. 2010), leaving enzymes inactive and destabilized. Together, these effects likely create a substantial burden of misfolded or damaged proteins, triggering a pronounced heat–shock response.

Under moderate copper stress, we also observed an enrichment of lipid A biosynthesis genes (Fig. 2A, Supplementary file 2), including the key enzyme LpxC (Supplementary file 1). However, since excessive lipid A synthesis is harmful for bacteria (Hummels 2025), we cannot at this moment distinguish whether the upregulation is a beneficial reaction to, or an unintended consequence of copper stress.

Copper exposure also caused broad gene repression across the transcriptome. At 2 mM Cu, 486 genes were downregulated (Supplementary file 1); however, enrichment analysis only showed enrichment in categories related to biofilm formation and, unexpectedly, pathways linked to pH elevation (Fig. 2C, Supplementary file 2). The reduction in biofilm-related genes fits with the idea that copper-stressed bacteria are less inclined to settle on surfaces, which is consistent with previous reports showing that copper inhibits biofilm formation of E. coli and Listeria monocytogenes at both MIC and sub-MIC concentrations (Dehkordi et al. 2023). The repression of genes involved in pH elevation at 2 mM Cu is more difficult to interpret. Even though copper can acidify the growth medium (Gunawan et al. 2011), the buffered MOPS medium maintains a stable pH at 2 mM copper (Fig. 6) and should not require a cellular pH response. We can only put forth a hypothesis linking the observed downregulation of glutamate decarboxylase genes gadA and gadB to the effects of copper. Excess copper may damage glutamate synthase—a Fe-S-containing enzyme essential for synthesizing glutamate (Djoko et al. 2017). Because GadA and GadB convert glutamate into GABA to help raise intracellular pH, continuing to express these enzymes would consume glutamate. If copper stress would limit glutamate production, reducing GadAB expression may therefore help the cell conserve its remaining glutamate, especially when the external pH remains relatively stable and the need for GadAB-mediated pH regulation is lower.

At 8 mM Cu, 217 genes were repressed, and enrichment analysis showed enrichment only in anaerobic respiration and hydrogenase-based energy metabolism. The repression of anaerobic respiration and hydrogenase-linked pathways at 8 mM Cu is consistent with copper’s known ability to interfere with Fe–S cluster assembly, which both processes depend on (Tan et al. 2017). However, as the experiments were performed in aerobic conditions, these pathways are expected to be generally expressed lower. Furthermore, 8 mM copper strongly inhibits growth (Fig. 1A) and therefore, metabolism repression could be linked to the growth arrest.

As an alternative approach to expression profiling, we evaluated the Horizon Discovery promoter library, but it proved unsuitable due to poor signal-to-noise ratios during copper exposure. Of the highly copper-responsive genes identified by RNA-seq, only cusC showed strong activation in the GFP-based assay (Fig. 4). Another gene, yebE, exhibited only a modest expression from its promoter in MOPS at 0.5 mM Cu, and failed to reach the twofold induction threshold in all other conditions. The remaining ten strongly upregulated genes detected by RNA-seq showed no measurable activation in the promoter collection. A similar pattern was observed for copper–repressed genes: of the ten promoters strongly downregulated in RNA–seq, only gadX promoter consistently showed repression in MOPS. The other nine promoters displayed no detectable decrease in GFP signal.

The underlying cause of the poor performance of the collection remains unresolved. We considered fluorescence quenching by copper, since in some conditions it practically removes GFP fluorescence (Bourge et al. 2015) by causing FRET (Hötzer et al. 2011, Jensen et al. 2024). Our experiments indicate, however, that under our tested conditions, the GFP signal is already low and copper does not inhibit it further (Fig. 5). In essence, compared to the direct readout of transcriptional activity by RNA-seq, a GFP-based system is more convoluted. Any change in fluorescence levels requires not only transcription but also translation and maturation of the GFP. In the case of downregulation, GFP stability creates another issue that precludes detection of rapid effects. Altogether, this particular promoter library seems to be suboptimal for our experimental setup.

Conclusions

In this work, we describe the acute transcriptional changes in E. coli during moderate and severe copper stress by RNA-seq. We detected the upregulation of known copper response pathways, histidine biosynthesis, oxidative stress response and iron acquisition pathways. During moderate copper stress, the bacteria retain better capacity to mount protective responses as we saw additional defensive pathways: biosynthesis of additional amino acids methionine and arginine, heat response and protein folding, lipid A biosynthesis activated compared to severe Cu stress. Additionally, we explored the use of a promoter-GFP library from Horizon Discovery to screen for transcriptional responses to copper. However, at least in our experimental setup, poor signal-to-noise ratio made its use impractical and therefore we consider the RNA-seq results as a detailed view of copper stress response in E. coli.

Author contributions

Hanna Ainelo (Conceptualization, Study design, Investigation, Data analysis, Interpretation and management, Writing the manuscript), Karola Jõearu (Investigation, Formal data analysi), Andres Ainelo (Data analysis and interpretation, Review and editing of the manuscript, Funding acquisition), and Angela Ivask (Conceptualization, Study design, Review and editing of the manuscript, Project administration, Funding acquisition)

Conflicts of interest

None declared.

Funding

This work was supported by the Estonian Research Council grants PRG1496, TemTA55 and TK210 (AI), and STP39 (AA), and Horizon Europe project 101159721 (AI). This research is conducted by partially using the research infrastructure “Experimental Studies and Applications of Cellular Processes—RAKERA” funded by the Estonian Research Council (TARISTU24-TK14).