Mutant Strains of Escherichia coli and Methicillin-Resistant Staphylococcus aureus Obtained by Laboratory Selection To Survive on Metallic Copper Surfaces

Microbes are rapidly killed on solid copper surfaces by contact killing. Copper surfaces thus have an important role to play in preventing the spread of nosocomial infections. Bacteria adapt to challenging natural and clinical environments through evolutionary processes, for instance, by acquisition of beneficial spontaneous mutations. We wish to address the question of whether mutants can be selected that have evolved to survive contact killing on solid copper surfaces. We isolated such mutants from Escherichia coli and methicillin-resistant Staphylococcus aureus (MRSA) by artificial laboratory evolution. The ability to survive on solid copper surfaces was a stable phenotype of the mutant population and not restricted to a small subpopulation. As a consequence, standard operation procedures with strict hygienic measures are extremely important to prevent the emergence and spread of copper-surface-tolerant persister-like bacterial strains if copper surfaces are to be sustainably used to limit the spread of pathogenic bacteria, e.g., to curb nosocomial infections.

ABSTRACT

Artificial laboratory evolution was used to produce mutant strains of Escherichia coli and methicillin-resistant Staphylococcus aureus (MRSA) able to survive on antimicrobial metallic copper surfaces. These mutants were 12- and 60-fold less susceptible to the copper-mediated contact killing process than their respective parent strains. Growth levels of the mutant and its parent in complex growth medium were similar. Tolerance to copper ions of the mutants was unchanged. The mutant phenotype remained stable over about 250 generations under nonstress conditions. The mutants and their respective parental strains accumulated copper released from the metallic surfaces to similar extents. Nevertheless, only the parental strains succumbed to copper stress when challenged on metallic copper surfaces, suffering complete destruction of the cell structure. Whole-genome sequencing and global transcriptome analysis were used to decipher the genetic alterations in the mutant strains; however, these results did not explain the copper-tolerance phenotypes on the systemic level. Instead, the mutants shared features with those of stressed bacterial subpopulations entering the early or “shallow” persister state. In contrast to the canonical persister state, however, the ability to survive on solid copper surfaces was adopted by the majority of the mutant strain population. This indicated that application of solid copper surfaces in hospitals and elsewhere has to be accompanied by strict cleaning regimens to keep the copper surfaces active and prevent evolution of tolerant mutant strains.

IMPORTANCE Microbes are rapidly killed on solid copper surfaces by contact killing. Copper surfaces thus have an important role to play in preventing the spread of nosocomial infections. Bacteria adapt to challenging natural and clinical environments through evolutionary processes, for instance, by acquisition of beneficial spontaneous mutations. We wish to address the question of whether mutants can be selected that have evolved to survive contact killing on solid copper surfaces. We isolated such mutants from Escherichia coli and methicillin-resistant Staphylococcus aureus (MRSA) by artificial laboratory evolution. The ability to survive on solid copper surfaces was a stable phenotype of the mutant population and not restricted to a small subpopulation. As a consequence, standard operation procedures with strict hygienic measures are extremely important to prevent the emergence and spread of copper-surface-tolerant persister-like bacterial strains if copper surfaces are to be sustainably used to limit the spread of pathogenic bacteria, e.g., to curb nosocomial infections.

INTRODUCTION

The growth of pathogenic bacteria in hospitals and other hygiene-sensitive areas, such as food production facilities, needs to be strictly controlled to ensure human health. Nevertheless, a multitude of processes and characteristics allow bacteria to adapt to such stress-inducing environments. Among these characteristics is phenotypic heterogeneity of bacterial populations, which allows them to subvert the trade-off between reproduction and survival in bacteria (1). Persister subpopulations survive diverse challenging conditions while the majority of cells succumb to stress (2). Mutant subpopulations accumulate beneficial mutations following initial, random point mutations or recombination events, and new genes may be acquired via horizontal gene transfer or gene duplication (3–8). When these factors are taken into consideration, it is unsurprising that the use of antibiotics has resulted in rapid development of resistance against these common antimicrobial drugs (e.g., methicillin; reviewed in reference 9), and the use of disinfectants and antiseptics has also led to bacterial strains with decreased susceptibility (e.g., to formaldehyde or iodophor antiseptics; reviewed in reference 10).

Because of these microbial challenges, innovative, hygiene-improving methods and regimens need to be adopted. One approach is the application of dry, antimicrobial surface materials that may aid, but not make obsolete, standard hygiene strategies such as hand sanitization or liquid surface disinfection. Among such dry antimicrobial surfaces, those made from massive metallic copper (alloys) have shown great promise in laboratory studies (11–15) and hospital trials (16–19) to drastically reduce or even eliminate the microbial surface burden within a few minutes. A multiple-institution clinical study found that infection rates decreased by 58% in rooms fitted with appliances made of massive copper compared to their noncopper counterparts (19).

The overall molecular mode of action exerted by metallic copper on diverse microbes, such as Gram-negative (13, 20, 21) or Gram-positive (12, 14, 16) bacteria, viruses (22–24), fungi (25), and yeasts (15), initiates through a process called contact killing (13), which is now generally understood. Use of contact killing could be an important step in the struggle against the spread of nosocomial infections, whereby 2.6 million infections per year occur in Europe, resulting in 91,000 deaths (26). Several overlapping stresses accumulate to result in inactivation of copper surface-exposed microbes. Damage to the membranes or to DNA accrues by oxidative stress events, depending on the presence of glutathione (27), or by massive overload of the cell by dissolved copper ions (11, 28). Physiologically, respiration ceases due to depletion of concentration gradients across the compromised membranes (29), and replication is prevented because genomes become fragmented beyond repair (28). These events may be followed by a complete breakdown of cellular structural integrity (11, 12).

As in the case of other strategies used in microbial growth control, such as the use of antibiotics, concerns have been raised that over time microbes might develop resistance against contact killing on surfaces made of metallic copper, despite the material’s superior antimicrobial properties (30). Indeed, metabolically active bacteria were successfully isolated from general circulation copper (alloy) coinage (30–32). Some of these isolates were able to withstand contact killing considerably longer than their cognate species type strain counterparts (30–32). Though the genomes of some of these isolates were sequenced and analyzed, no genetic basis of this resistance trait could so far be identified (33). In this study, we applied artificial in vitro evolution of Gram-negative model organism Escherichia coli and Gram-positive model organism methicillin-resistant Staphylococcus aureus (MRSA), which are both metallic copper sensitive. Evolved mutants from E. coli and MRSA were physiologically characterized and subjected to whole-genome and whole-transcriptome sequencing.

RESULTS

Isolation of mutants tolerant to metallic copper surfaces.

E. coli and MRSA mutant strains that survived on metallic copper surfaces were generated in an adaptive laboratory evolution experiment (34–37). To that end, short sequential rounds of cultivation and selection were performed (Fig. 1). The bacteria were cultivated in LB broth, concentrated, incubated on metallic copper surfaces (99.9% Cu), washed from these surfaces, and incubated on solid LB agar medium, and then single colonies were randomly picked and cultivated for the next round of selection. For each iteration, the cultured survivors were thus the founder population for the next round of adaptation/evolution. The growth step without selection pressure allowed cellular repair processes and increased the genetic diversity due to mutations occurring during DNA replication (38).

FIG 1.

Schematic diagram depicting the procedure for generating copper-tolerant mutants.

Since E. coli and S. aureus have been reported to be completely contact-killed after 1 (13) and 7 (12) min, respectively (see also Fig. 2), the initial incubation time tested on solid copper surfaces was 30 s for E. coli and 3 min for S. aureus. This time interval increased between 30 s and 1 min per exposure cycle until, after 41 or 38 iterative rounds of artificial evolution, respectively, the cells survived for 1 h on the metallic copper surface, resulting in the mutant strains termed E. coli mCu60 and MRSA mCu60.

FIG 2.

Evolved metallic copper-tolerant mutants of E. coli and MRSA survive contact killing on massive copper surfaces. (A and B) Parental strain (diamonds) or evolved mCu60 mutants (squares) of E. coli (A) and MRSA (B) were applied on copper (filled symbols) or stainless steel (open symbols) coupons, incubated for different time periods, removed from the coupons, and spread on solid medium, and the survivors were counted as CFU per coupon after growth for 24 h (13). The panels show results as averages from 3 or 4 independent experiments with standard deviations (error bars).

On stainless-steel coupons used as controls, parent and mCu60 mutant strains of both bacterial species did not show any difference in survival rates (Fig. 2). The CFU of both E. coli strains decreased on stainless steel by a factor of about 10, possibly caused by general desiccation stress. On copper, the CFU of the E. coli mCu60 mutant decreased by a factor of about 100, while the parental strain was killed after 1 min (Fig. 2A). The decrease of the CFU on both metals was smaller for the MRSA strains, about 10-fold of the MRSA mCu60 mutant on copper (Fig. 2B). Thus, the ability to survive on solid copper surfaces was a material-specific phenotype and could be generated in a Gram-negative proteobacterium as well as a Gram-positive firmicute.

In order to investigate the stability of the mutant phenotype, the mCu60 mutant strains of both bacterial species were stored at –80°C as stock cultures and recultivated in LB broth. Subsequently, both mutant strains were repeatedly cultivated without selection pressure in LB broth and diluted 4,000-fold on a daily basis for 21 days (about 250 generations). There was no difference between the growth of the mCu60 mutant and its parent in time-dependent growth experiments in LB broth without selection pressure (Fig. 3).

FIG 3.

Artificial laboratory evolution of metallic copper tolerance in E. coli and MRSA did not alter the mutants’ growth characteristics. (A and B) Growth of cultures of metallic copper-tolerant (squares) and parental strains (diamonds) of E. coli (A) or MRSA (B) in LB broth was monitored until the stationary phase. Samples were taken every 20 min, and the OD₆₀₀ was determined. The averages from 3 independent experiments with standard deviations (error bars) are shown.

When the cultures, which were maintained without selection pressure, were incubated on solid copper or stainless-steel surfaces, the mCu60 mutant strains of E. coli and MRSA survived for 1 h on the copper surface (Fig. 4). The mCu60 mutant strains from both bacterial species showed no difference in growth from their respective parental strains under nonselective conditions in LB or on stainless steel surfaces. Thus, the ability to survive on solid copper surfaces was a stable phenotype.

FIG 4.

The evolved copper surface-tolerant phenotype is stable. To test if the evolved copper-tolerant mCu60 bacteria retained their ability to withstand metallic copper-mediated contact killing, bacteria were first stored at –80°C, recultivated, and then repeatedly grown on solid medium without stress for the duration of 3 weeks and, finally, challenged on copper coupons. (A and B) Cultures of E. coli (A) or MRSA (B) mCu60 mutants of these passaged cultures were challenged on dry metallic copper (filled symbols) and stainless-steel control surfaces (open symbols) as described for Fig. 2. The averages from 3 or 4 independent experiments with standard deviations (error bars) are shown.

Phenotypes of the mCu60 mutant strains.

Since copper released from metallic copper surfaces is an important factor for contact killing of bacteria on metallic copper surfaces, resistance of the mCu60 mutant strains to ionic Cu(II) sulfate was compared to that of the respective parental strains. The MIC values against copper ions for both the E. coli mCu60 mutant and its parent were identical, 3.0 ± 0 mM CuSO₄ (n = 3), as were the MIC values for the evolved MRSA strain and its parent, 1.0 ± 0 mM CuSO₄ (n = 3). This indicated that tolerance of the mCu60 mutant strains to metallic copper was not due to changes in the cellular copper ion homeostasis system, e.g., increased production of copper ion efflux pumps (27).

On solid copper surfaces, the E. coli parent and mCu60 mutant cells accumulated about 5 × 10⁶ Cu atoms/cell after the first 30 s of the experiment (not shown in Fig. 5A) and 10 × 10⁶ Cu atoms/cell within 5 min (Fig. 5A). In the mCu60 mutant strain, this value decreased slowly to 5 × 10⁶ Cu atoms/cell within 1 h. In contrast, the parental strain reproducibly exhibited a peak of (51 ± 4) × 10⁶ Cu atoms/cell after 15 min. Since the E. coli parent cells were killed after 5 min, but the mCu60 mutant cells were not (Fig. 2), and both strains contained a similar number of Cu atoms per cell at this time point (Fig. 5A), the mutant had acquired the ability to withstand the initial copper shock on copper surfaces between 0 and 15 min. Moreover, the mCu60 mutant escaped the second copper shock peak after 15 min, which would equate to a cellular quota of 94 mM copper if a volume of 0.9 fl is assumed (39). The 15-min cellular copper peak could be a consequence of the surface killing of the parental strain by the initial copper shock. As the mCu60 mutant was not killed, it did not show this second copper peak (Fig. 5A).

FIG 5.

ICP-MS analysis of metal accumulation into E. coli mCu60 and its parent on copper surfaces. The evolved metallic copper-tolerant E. coli mCu60 mutant (squares) and its parent (diamonds) were incubated in buffer on either copper (filled symbols) or stainless steel (open symbols) coupons for the indicated time periods, withdrawn from the surfaces, and washed to remove loosely bound copper from the cells. (A) The metal contents of whole cells were measured by ICP-MS from mineralized samples and calculated as cellular copper quota. (B and C) For cell fraction analysis, 10 parallel preparations (cytoplasmic [B] and membrane [C] protein fractions) of E. coli mCu60 (squares) and the parent strain (diamonds) were exposed and washed as in panel A. Cells were lysed by sonication, cell debris was removed by centrifugation, and the soluble, cytoplasmic proteins (B) were separated from the membrane proteins (C) prior to ICP-MS analysis. Protein concentrations were determined and used to calculate Cu/mg protein. The averages from 2 or 3 independent experiments with standard deviations (error bars) are shown.

The E. coli mutant and parental cells were incubated on the copper surface, the cells were harvested and lysed by ultrasonication, and the soluble and membrane fractions were separated by ultracentrifugation. The soluble fraction of both strains initially contained 5.0 ± 1.5 Cu atoms/mg protein in unchallenged cells. This value increased to 3,277 ± 462 Cu atoms/mg protein within 15 min of incubation on solid copper surfaces in both strains (Fig. 5B) and decreased during the following 45 min, with no difference between the mCu60 mutant and parental cell (Fig. 5B). The mCu60 cells did not accumulate less copper in the cytoplasm than the parental cells.

In contrast, the ultracentrifugation sediment of the parental cells contained 5,038 ± 1,055 Cu atoms/mg protein after 15 min, while only 2,000 Cu atoms per mg protein were present in the ultracentrifugation sediment of the mutant cells. Both values changed to about 3,000 Cu atoms per mg protein present after 1 h (Fig. 5C). The copper peak present in parent cells after 15 min (Fig. 5A) could also be demonstrated for the ultracentrifugation sediment (Fig. 5C). While the copper content of the soluble fractions of mutant and parent were similar, the copper content of the ultracentrifugation sediment of 15-min cells was 2.5-fold higher in the parental than the mutant cells. Ultracentrifugation sediments contain small membrane vesicles and membrane-bound proteins but also ribosomes (40). Since ribosomes are located in the cytoplasm, an increased ribosomal copper content should be paralleled by an increased copper content of the cytoplasmic fraction, which was not the case (Fig. 5B). The increased copper content of the ultracentrifugation sediment derived from the parental cells (Fig. 5C) should thus reflect an increased copper content of small membrane vesicles but not of the cytoplasmic fraction. Copper released from the solid copper surfaces did not accumulate in the cytoplasm of parental and mutant cells (Fig. 5B). An overexpression of inner-membrane efflux pumps was not involved in resistance to killing, in agreement with a similar MIC of copper sulfate in mutant and parent cells.

The mCu60 mutant prevents damage to its cell membrane upon contact with metallic copper surfaces.

To test if damage of the cytoplasmic membrane was responsible for the strongly increased copper accumulation and cell death of the parental strain but not of the mCu60 mutant (Fig. 6), mCu60 mutants and parental cells of E. coli (Fig. 6A) or S. aureus MRSA (Fig. 6B) were incubated on copper (Cu) or stainless steel (SS) control surfaces, and the contact-killing process was followed by Live/Dead staining and microscopy. When E. coli or S. aureus was incubated on stainless steel for up to 60 min, no obvious membrane damage was observed (Fig. 6, top two rows in both panels). Most cells stained green, which indicated undamaged membranes. Only a few cells fluoresced red, which was probably caused by general stress during desiccation on the dry metal surface.

FIG 6.

Evolved mutant strains retain their membrane integrity upon prolonged challenge to metallic copper. Bacteria were exposed to copper (Cu, bottom two lines) or stainless steel (SS, top two lines) coupons, removed after the indicated time periods and live/dead-stained. Membrane-damaged cells fluoresce red, whereas intact cells fluoresce green. (A and B) Panels are micrographs of the evolved mCu60 mutant (mCu60) or parental (Wt) E. coli (A) or MRSA (B) cells taken after the indicated time periods of exposure on stainless steel (SS) or metallic copper (Cu), respectively. Differences in overall cell numbers in the field of vision are due to cell losses during the numerous centrifugation and washing steps prior to microscopy. Representative results from replicate experiments with similar results are shown.

E. coli parent cells on dry copper surfaces exhibited damage to their membrane (red fluorescence) after 15 min; however, they were structurally still intact (Fig. 6A; parental [Wt] Cu 15 min). These parental cells were killed (Fig. 2) by the initial copper shock after 1 min by damage to their membrane, which subsequently resulted in accumulation of large amounts of intracellular copper, the copper peak, after 15 min in the dead but structurally intact E. coli cells (Fig. 5A). After further incubation of the E. coli parent cells for 30 or 60 min, only cellular debris remained (Fig. 6A), which explained the decrease in the cellular copper content after the maximum at 15 min of incubation (Fig. 5). The formerly structurally intact, but dead, cells disintegrated and released their copper ions. In contrast to the parent, the E. coli mCu60 mutant cells retained their membrane integrity (Fig. 6A, bottom rows), despite the high copper content (Fig. 5A).

Due to technical reasons, the copper content of cells on solid copper surfaces could not be determined for the MRSA strains. Nevertheless, the mCu60 MRSA strain also survived for 1 h on copper surfaces, while the parental cells disintegrated after 30 min (Fig. 6B). The mCu60 mutants of both species were able to survive the initial copper shock and to maintain their membrane integrity for up to 1 h of incubation. Although the percentage of mCu60 cells that was still able to form colonies decreased 10 to 100-fold on solid copper surfaces (Fig. 2), all recovered mCu60 mutant cells retained their membrane integrity. The ability to survive on solid copper surfaces was a capability of the whole population and not just of a subpopulation, possibly due to persisters.

The mCu60 mutant cells showed a lower content of fatty acids with >17 carbons.

Since the membranes of the mCu60 mutants of both species were able to withstand the initial copper shock without losing membrane integrity, the composition of the membrane fatty acids (FA) was determined in mutant and parental cells by a fatty acid methyl ester (FAME) analysis.

E. coli and MRSA strains possessed very distinct FAME profiles, making cross-comparison difficult. However, differences between mCu60 mutants and their parental strains were evident when analyzing the proportion of each individual FA, as well as when certain chemical groups sharing common characteristics were examined. Compared to their parents, the mCu60 mutants contained an increased portion of small fatty acids (estimated chain length, ≤FA 17:0) and a decrease of some longer ones (Tables 1 and 2). For E. coli mCu60, this was evidenced by an increase (∼3.20%) of short linear fatty acids (saturated or mono-unsaturated) 13:0, 15:0, 15:1ω8c, 16:1ω5c, and 17:0 and a decrease (∼3.29%) of long fatty acids (cyclic or linear saturated) cy17:0ω7c, 18:0, and cy19:0ω8c (Table 1). The average decrease of cyclic fatty acids, usually associated with a response or adaptation to a diversity of stresses (salinity, pH, temperature, or detergents [41, 42]) was particularly pronounced (3.16%), being the most affected fatty acid chemical group. The total amounts of unsaturated fatty acids, which are the primary targets of lipid peroxidation, were very low for both the mutant (~3.08%) and the parental strain (∼2.44%). The ability of the mutants to survive the initial copper shock did not result from a smaller amount of these peroxidation targets.

TABLE 1.

Average FAME composition of E. coli mCu60 and its parent strain^a

ID	Avg FAME composition (%)		Mean difference (%)	t test (P < 0.05)
	mCu60	Parent
10:0	0.02 ± 0.04	0.00 ± 0.00	0.02	No
11:0	0.02 ± 0.02	0.00 ± 0.00	0.02	No
12:0	3.28 ± 0.27	3.47 ± 0.13	–0.19	No
13:0	0.53 ± 0.04	0.27 ± 0.02	0.26	Yesc
12:0 3OH	0.05 ± 0.08	0.00 ± 0.00	0.05	No
14:0	6.84 ± 0.55	6.88 ± 0.23	–0.04	No
13:0 3OH/i15:1 H	0.16 ± 0.28	0.00 ± 0.00	0.16	No
15:1ω8c	0.15 ± 0.05	0.00 ± 0.00	0.15	Yesb
15:0	4.52 ± 0.22	2.46 ± 0.03	2.06	Yesc
14:0 3OH/i16:1 I/12:0	7.56 ± 0.58	7.99 ± 0.24	–0.43	No
16:1ω9c	0.05 ± 0.09	0.05 ± 0.09	0.00	No
16:1ω7c/ω6c	1.52 ± 0.17	1.25 ± 0.08	0.27	No
16:1ω5c	0.20 ± 0.00	0.15 ± 0.01	0.05	Yesb
16:0	33.13 ± 2.62	36.17 ± 1.12	–3.04	No
15:0 3OH/16:0 12Me	0.17 ± 0.01	0.04 ± 0.07	0.13	No
cy17:0ω7c	22.18 ± 1.83	24.02 ± 0.71	–1.84	Yesb
17:0	1.92 ± 0.22	1.24 ± 0.04	0.68	Yesb
18:1ω7c/ω6c	0.98 ± 0.07	0.79 ± 0.03	0.19	No
i17:0 2OH	0.31 ± 0.03	0.24 ± 0.01	0.07	No
18:1ω5c	0.18 ± 0.02	0.17 ± 0.03	0.00	No
18:0	0.39 ± 0.02	0.52 ± 0.01	–0.13	Yesc
18:1ω7c 11Me	0.00 ± 0.00	0.03 ± 0.05	–0.03	No
17:0 2OH	1.26 ± 2.18	0.00 ± 0.00	1.26	No
18:0 12Me	0.71 ± 0.04	0.71 ± 0.06	0.00	No
i19:0	0.22 ± 0.19	0.35 ± 0.35	–0.14	No
cy19:0ω8c	6.63 ± 0.25	7.95 ± 0.21	–1.32	Yesc
20:0	2.38 ± 2.24	1.31 ± 2.26	1.08	No

^aValues presented are averages of three independent GC runs, with standard deviations. Mean differences and t test results are indicated; significantly different values are shown in boldface.

^bP ≤ 0.05.

^cP ≤ 0.01.

TABLE 2.

Average FAME composition of MRSA mCu60 and its parent strain^a

ID	Avg FAME composition (%)		Mean difference (%)	t test (P < 0.05)
	mCu60	Parent
10:0	0.03 ± 0.03	0.02 ± 0.03	0.01	No
i14:0	1.10 ± 0.01	1.06 ± 0.01	0.05	Yesb
14:0	0.15 ± 0.01	0.16 ± 0.01	–0.01	No
i15:0	4.31 ± 0.04	4.27 ± 0.02	0.03	No
a15:0	48.09 ± 0.36	47.32 ± 0.18	0.78	Yesb
i16:0	3.59 ± 0.04	3.26 ± 0.04	0.33	Yesd
16:0	0.89 ± 0.02	0.99 ± 0.02	–0.10	Yesc
i17:0	3.59 ± 0.06	3.48 ± 0.04	0.11	Yesb
a17:0	24.99 ± 0.34	24.08 ± 0.23	0.90	Yesc
i18:0	1.86 ± 0.04	1.86 ± 0.02	0.00	No
18:1ω9c	0.11 ± 0.01	0.15 ± 0.02	–0.04	No
18:0	2.95 ± 0.05	3.32 ± 0.03	–0.37	Yesd
i19:0	1.16 ± 0.08	1.29 ± 0.02	–0.13	No
a19:0	5.09 ± 0.04	5.87 ± 0.04	–0.78	Yesd
i20:0	0.24 ± 0.02	0.31 ± 0.02	–0.07	Yesb
20:0	1.19 ± 0.06	1.88 ± 0.27	–0.69	No

^aValues presented are averages of three independent GC runs with standard deviations. Mean differences and t test results are indicated; significantly different values are shown in boldface.

^bP ≤ 0.05.

^cP ≤ 0.01.

^dP ≤ 0.001.

For the mCu60 mutant of S. aureus MRSA, there was an increase (∼2.16%) of short-chain saturated branched fatty acids i14:0, a15:0, i16:0, i17:0, and a17:0 and a decrease (∼1.32%) of long-chain branched or linear fatty acids 18:0, a19:0, and i20:0 (Table 2). The total amounts of branched-chain saturated fatty acids in the parents and mutants were approximately the same, with the mutants showing a small increase of the iso/anteiso FA ratio (iso FA: mutant 15.85% and parent 15.53%; ante iso: mutant 78.17% and parent 77.27%).

Overall, the difference in fatty acid composition between the mCu60 mutants and their parental strains was similar in both bacterial species, with shorter fatty acids present in the mutants having a possible impact on membrane fluidity.

Genome-wide mutation analysis.

To identify the mutation(s) responsible for the observed copper tolerance in E. coli mCu60 or MRSA mCu60, respectively, the genomes of these two mutant strains and those of their parents were sequenced, and the resulting reads were mapped to the corresponding reference genomes. All four genomes were almost completely covered, with average coverages of 202- to 310-fold. Bioinformatic analysis revealed four single nucleotide polymorphisms (SNPs) between E. coli mCu60 and its parent, three SNPs located in an insertion element (IS) and one in a gene, a C-T transition in the glgB gene at position 4,069,049 (GenBank accession number NC_007779.1, 1,4-α-glucan branching enzyme), leading to a silent mutation converting a CTG codon to a TTG codon, both encoding leucine.

Similarly, the two SNPs found in MRSA mCu60 compared to its parent were a G-T transversion at position 523,575 and a G-A transition at position 1,593,558 (GenBank accession number NC_002745.2). The G-T transversion was located downstream of the stop codon of the rsmA gene (position 522,679 to 522,572), and the G-A transition was in the gene SA_RS07850/SA1388 (position complement 1,593,470 to 1,594,570), changing an ACG threonine codon into an ATG methionine codon (T338M). Gene SA1388 encodes a protein with unknown function present in a hexameric structure (see Fig. S1 in the supplemental material). Each monomer is composed of two Nif3-like domains separated by a PII-like domain, and each monomer contains a binuclear zinc site. Comparison with the Conserved Domain Database (43) revealed that SA1388 is a member of the YbgI/SA1388 protein family. This family was shown to consist of dinuclear metal center proteins of unknown function (44). The gene is upregulated by SarA (staphylococcal accessory regulator) (45) and downregulated in acid-shocked cells (46). The related protein YbgI from E. coli is important for survival of ionizing radiation and resistance to some antibiotics and may be involved in formation of the bacterial cell wall (47, 48). Based on the available structural data, T338 (PDB: 3LNL; Fig. S1) is located in the center of an α-helix at the surface, as well of the monomer of the hexamer. The hydropathy index of threonine (–0.7), according to Kyte and Doolittle (49), is not much different from that of methionine (1.9), compared to isoleucine (4.5), on the one hand, and arginine (–4.5), on the other hand, so that this change may not affect the tertiary structure of the hexamer. The threonine residue is also not close to the PII domain (Fig. S1B), which may interact with other factors in a signal transduction chain. In total, the amino acid exchange could not be connected to the phenotype of the MRSA mCu60 strain.

Similarly, no simple mutational explanation for the observed mCu60 phenotype was apparent from the mapping data for E. coli. Therefore, a de novo assembly of the sequencing data was performed. Manual inspection of the resulting assembled contigs revealed the insertion of an IS2 element into the ycbX gene (Y75_p0919), which encodes a 2Fe-2S cluster-containing protein, of unknown function (Fig. 7). As defined mutants for all nonessential E. coli genes are available (50), the knockout mutant JW5126-1 (CGSC 11193) was obtained from the CGSC strain collection and tested for increased tolerance to metallic copper surfaces but was found to be as susceptible as the parental E. coli strain (data not shown).

FIG 7.

Changes in transcription and DNA sequence in E. coli mCu60 compared to the parent E. coli strain due to insertion of IS2 in ycbX. Depicted is a condensed view of the chromosomal region containing ycbX (Y75_p0919) and ycbY (Y75_p0920) taken from ReadXplorer, with the accumulated reads derived from E. coli (top) or from E. coli mCu60 (bottom). The y and x axes represent coverage and genomic position, respectively. Below is an expanded view at the nucleotide level, depicting the native sequence in E. coli and the change in E. coli mCu60 due to the insertion of IS2 (only the 3′ end is displayed). The putative –35 and –10 regions of the resulting hybrid promoter are boxed in red, with bases corresponding to the σ⁷⁰ consensus promoter motif shown in green. Please note that due to the library preparation method used, the transcript does not appear to start at +1 but at 8 to 11 bases further downstream.

Comparative transcriptomics using comparative cDNA sequencing (RNASeq).

Whole-transcriptome cDNA libraries were prepared from RNA isolated from cells of the tolerant strains as well as their parents grown in LB medium. After mapping the reads per kilobase per million mapped reads (RPKM values), the logarithms to the basis 2 of all RPKM values were derived. The relative expression values for each gene were visualized as M versus A plots to identify differentially transcribed genes (Fig. 8; Table S1 and S2). For E. coli, the data points were additionally annotated using the KEGG (Kyoto Encyclopedia of Genes and Genomes) pathway orthology (KO) and BRITE (hierarchical classification of biological entities) search paths (51, 52).

FIG 8.

Changes in gene expression between the mCu60 strains and their parents. (A and B) The transcriptomes of the mCu60 and parent strains of E. coli (A) and S. aureus MRSA (B) were determined, and the Log₂ value of the RPKM values (reads per kilobase per million mapped reads) was calculated. The difference “Log₂(mutant) minus Log₂(parent)” for each species was plotted against the mean values so that a 4-fold increase or decrease (horizontal bars) indicates a difference value above 2 or below –2. Values above 2 are shown in greens, below –2, in red, and in between in gray. Data points identified within the KEGG orthology (KO) reference hierarchy are in red, green, or large gray spots; those within the br (BRITE) reference hierarchy are in orange, light green, or small gray spots. Moreover, the following KO groups were additionally indicated: copper resistance (dark blue), transcription (black), translation (purple small open circles), DNA recombination and repair (blue open squares), energy metabolism (open red squares), and ure genes other than ureB (light blue) and SA1388 (black diamond). Additional genes are labeled. The downregulated genes in E. coli are all metabolic genes on both hierarchy systems; other downregulated genes are shown as gray symbols.

In E. coli mCu60, a total of 18 genes were found with at least 4-fold increased mRNA pools, while a total of 317 genes had mRNA pools decreased by at least 4-fold (Fig. 8A). Among the upregulated genes, two (Fig. 8A, dark green) were assigned to the metabolism of carbohydrates (sgcB) and amino acids (ygaT). They encoded a putative component of a phosphotransferase system and a hypothetical protein, respectively. Six upregulated genes were not annotated (small gray dots), and eight were associated with the BRITE categories “unclassified” (7 genes) or “poorly characterized” (1 gene). BglH could be a cryptic carbohydrate-specific outer membrane protein, and YcbY, a methyltransferase involved in ribosome biosynthesis (53, 54). The ycbY gene was upregulated, maybe due to the IS2 insertion in ycbX (Fig. 7).

As shown in Fig. 8A, genes involved in copper homeostasis (blue circles), transcription (black dots), translation (purple small open circles), or DNA repair and recombination (blue open squares) were not changed in expression in the E. coli mCu60 mutant, but genes involved in metabolism as assigned by KO (red filled circles) or BRITE (orange filled circles) made up the majority of the downregulated genes, especially genes involved in conservation of energy (red open squares), such as the F₁F_o-ATPase. Compared to the parent, the mCu60 mutant exhibited strongly downregulated genes implicated in different aspects of metabolic activity. The strongest downregulation (nearly 1,000-fold) was determined for the gene encoding the tryptophanase TnaA, the tryptophan importer TnaB, and the leader of the tna operon, tnaC. Additional strongly downregulated genes were those involved in pyrimidine biosynthesis (Fig. 8A).

To allow stress resistance, energy-consuming activities of the exponential growth phase may be downregulated, while stationary-phase genes and actions may be upregulated by the sigma factor RpoS, the alarmone ppGpp, or the regulator DksA (55–58). Defense and resistance genes are upregulated, and damaged proteins and DNA sequences are repaired by genes or regulons controlled by RpoH, RpoE, or regulators interacting with RpoD-dependent RNA polymerase, such as genes/regulons involved in the SOS response. These mechanisms allow increased survival (59–62). A possible contribution of genes encoding proteins involved in stress resistance or small regulatory RNAs was additionally analyzed in detail. These were genes for alternative stress sigma factors, ppGpp handling, shock proteins, the SOS response, synthesis and repair of iron-sulfur clusters, handling reactive oxygen species or glutathione, the stress-sensitive aconitase protein, chaperones, universal stress proteins, two-component regulatory systems, or nucleoid-associated proteins (Table S3). None of these genes was upregulated with a difference value (M) greater than 2 (more than 4-fold), while 12 genes were downregulated with M <–2 (Table 3), similar to the majority of the regulated genes (Fig. 8A). Regulatory genes that were downregulated were those involved in the cold shock response, those that encoded the iron-containing superoxide dismutase SodB, the two-component systems involved in drug resistance (EvgAS) or low magnesium response (RstBA), small RNAs involved in regulation of carbon storage response or amino acid availability, and a subunit of the nucleoid-binding HU protein (63). The genes showing the strongest upregulation were soxS (M = 1.62, 3.1-fold) and soxR (M = 1.54, 2.9-fold), which are involved in redox stress (64, 65). None of these data explained the mCu60 phenotype of the E. coli mutant by a rearrangement of a global regulatory cycle.

TABLE 3.

Regulation of genes involved in stress tolerance^a

Feature	Locus	M	Product
ymcE	Y75_p0963	–2.74	Cold shock protein
cspB	Y75_p1533	–2.57	Cold shock protein
cspI	Y75_p1528	–2.17	Cold shock protein
sodB	Y75_p1633	–2.36	Superoxide dismutase, Fe
ccmE	Y75_p2158	–2.25	Periplasmic heme chaperone
clpB	Y75_p2541	–2.12	Protein disaggregation chaperone
evgS	Y75_p2337	–2.76	Hybrid sensory histidine kinase with EvgA
evgA	Y75_p2336	–1.18	DNA-binding response regulator with EvgS
rstB	Y75_p1585	–2.31	Sensory histidine kinase with RstA
rstA	Y75_p1584	–2.43	DNA-binding response regulator with RstB
dcuS	Y75_p4012	–1.68	Sensory histidine kinase with DcuR of fumarate respiration
dcuR	Y75_p4011	–1.93	DNA-binding response regulator with DcuS
uhpB	Y75_p3506	–1.52	Sensory histidine kinase with UhpA
uhpA	Y75_p3505	–1.36	DNA-binding response regulator with UhpB
csrC	Y75_s0043	–2.73	sRNA
gcvB	Y75_s0032	–2.98	sRNA
hupA	Y75_p3196	–2.00	HU, DNA-binding transcriptional regulator subunit alpha

^aThe transcriptome of the mCu60 and parent (WT) strains of E. coli was determined, and the Log₂ value of the RPKM values (reads per kilobase per million mapped reads) calculated. The difference value M, defined as M = Log₂(mutant) minus Log₂(WT), was plotted against the mean values so that a 4-fold increase or decrease indicates a difference value, M, above 2 or below –2. From these, genes for proteins involved in stress-resistance and for small RNAs were analyzed in detail (Table S3). Shown are genes downregulated more than 4-fold (M values in bold) and some genes for two-component regulatory systems that were downregulated more than 2-fold.

For MRSA mCu60, five genes were found with at least 4-fold increased mRNA pools, whereas 37 genes had at least 4-fold decreased mRNA pools (Fig. 8B). Overall, the number of downregulated genes and the extent of their downregulation were much smaller in the MRSA mCu60 mutant compared to the E. coli mCu60 mutant. Among the upregulated genes was ureB for a component of the urease (Fig. 8B, dark green), but other genes involved in urease biosynthesis were not affected (light blue). As with the E. coli mutant, genes involved in transcription (black), translation (purple), DNA replication and repair (blue open squares) and the copper-exporting P-type ATPase CopA (blue) were not altered in expression. In contrast to the E. coli mCu60 mutant, genes involved in conservation of energy (red open squares) were not downregulated. The downregulated genes (BRITE, orange; KO, red) concerned transport systems (triangles), pyrimidine and amino acid biosynthesis (closed red circles), and other genes (red triangles). The gene with the locus tag SA1388 that carries a mutation in the mutant strain was not changed in expression (Fig. 8B, black diamond).

DISCUSSION

Shortly after the widespread use of early antibiotics began in the first half of the previous century, strains of S. aureus that were no longer susceptible to penicillin were discovered. These resistant strains spread with increased administration of the drug (66), and similar observations were made for other antimicrobial compounds. Today, between 90% and 95% of all clinical S. aureus strains are thought to be resistant toward penicillins (67).

Recently, the use of massive copper appliances in lieu of their noncopper counterparts in hospitals and other health care-associated facilities has aided in improving surface-related hygiene (17–19, 68), and copper appliances are likely responsible for diminishing the incidence of hospital-acquired infections (17, 19, 68). The experiences regarding emerging antibiotic resistances in microbes led to concern that copper-surface-resistant bacteria may evolve, so that antimicrobial copper would lose its efficacy. Earlier studies presented ambiguous results (30, 69). Bacteria were isolated from copper (alloy) coins that were in general circulation in Germany and Portugal (30). These bacteria withstood the effects of contact killing when tested on copper surfaces in the laboratory. On the other hand, the randomly collected copper coins showed low-level contamination, with an average of about 6 CFU of culturable heterotrophic aerobic bacteria collected from their surfaces (30). Unsurprisingly, these bacteria mostly represented a diverse collection of the human skin microbiota, including some potential facultative pathogenic species (30). Others found that MRSA only survived on coins if there were organic compounds soiling the coin surface and thus protecting the bacteria from direct copper contact (69).

These considerations led to the concern that mutants can be generated that are able to withstand the toxic effects exerted by contact with massive metallic copper. In the current study, mutants from two different bacterial taxa, proteobacteria and firmicutes, were isolated that survived for 1 h on a solid copper surface. This phenotype was stable in both species used, E. coli and MRSA. The E. coli K-12 parent and mutant cells contained 5 million Cu atoms per cell after 30 s of incubation on the copper surface. Contact with solid copper surfaces thus resulted in an enormous accumulation of copper atoms per cell, and these copper ions seemed to be located mainly in the cellular envelope (Fig. 5). For comparison, E. coli cells and cells of the metal-resistant bacterium Cupriavidus metallidurans contained about 60,000 Cu atoms per cell when cultivated in nonamended mineral salts medium (70), and C. metallidurans accumulated up to 200,000 Cu atoms per cell under conditions of extreme copper stress due to gold-mediated inhibition of copper efflux (71).

Upregulation of the copA gene for the copper-exporting P-type ATPase CopA in E. coli requires 2 min, and that of the CusCBA periplasmic efflux system requires even longer (72), so that increased production of copper ion resistance systems is too slow to protect the cell against the overwhelming number of 5 million copper atoms. In agreement with this, neither of the two mCu60 mutant strains exhibited any upregulation of copper ion resistance systems (Fig. 8) or increased copper resistance as phenotype.

Copper released from the solid copper surface killed the parental cells of both species within minutes (Fig. 2) by compromising membrane integrity followed by cell disintegration (Fig. 6). In contrast, all mCu60 mutant cells, and not just a subpopulation, e.g., persisters, survived and kept their membrane integrity (Fig. 6). The genotypes of both mutants were not much different from their parents. Phenotypically, the E. coli mCu60 mutant expressed the ycbY gene at 4-fold higher levels due to insertion of IS2 in the adjacent ycbX gene (Fig. 7 and 8). YcbY (RlmL) is a 23S rRNA methylase involved in ribosome maturation (53, 54). The S. aureus MRSA mCu60 mutant contains a mutated product of the SA1388 gene for a protein that could be involved in cell wall biosynthesis and radiation resistance (47, 48). The few identified differences in the genomes of both mCu60 strains, nevertheless, were accompanied by multiple changes to the transcriptomes of the mCu60 mutants compared to their parents (Fig. 8), indicating a change in some global regulatory pathway; however, a connection between this pathway change and the identified mutations was not evident.

In the E. coli mCu60 mutant, the tnaA, tnaB, and tnaC genes for synthesis of the tryptophanase, tryptophan-proton symporter, and tnaAB operon leader peptide that enables attenuation control of its expression were the most downregulated genes (Fig. 8A). TnaB imports tryptophan and TnaA degrades it to pyruvate, ammonium, and indole, using exclusively external tryptophan. Indole is an extracellular and intracellular signal in E. coli (73) and other indole-producing bacteria. It may regulate adhesion and biofilm promotion (74). Antibiotics promoting formation of reactive oxygen species may inhibit development of biofilms, and this process depends on indole (75). For contamination of dry copper surfaces, the ability to form biofilms does not play any role, as there is no growth under these conditions. With respect to the downregulated tnaAB operon, the mCu60 mutant resembles a tnaA mutant, which shows increased resistance to isobutanol (76).

Indole produced by TnaA has an effect on persister formation (77). Two different studies came to different conclusions concerning the role of indole in persister formation. On the one hand, indole triggers formation of the persister subpopulation in a bacterial population by activating the stress response (78), so that diminished production of indole due to downregulation of tnaA should decrease persister formation. The respective stress-responsive genes (oxyR, pspBCA) were not upregulated in the mCu60 mutant strain (Data Set S1). In a study by Vega et al. (78), the inability to import tryptophan into Δmtr mutant cells resulted in increased survival in LB broth-grown cells, while the ΔtnaA mutant exhibited decreased survival. Both genes are downregulated in the mCu60 mutant (Data Set S1). Decreasing the tnaA-mRNA concentration by the action of the toxin YafQ of the YafQ/DinJ toxin-antitoxin-type regulatory system clearly increased persister formation by a decreased indole level (79). The yafQ gene did not differ in expression in the E. coli mCu60 mutant compared to its parent (Data Set S1). Since tnaA is needed for “deep persistence,” survival for 24 h, but not for “shallow persistence,” survival for a shorter period of time (80), indole may be required to induce formation of persisters to allow long-term survival by deep persistence but prevents maintenance of a persister subpopulation in exponentially growing or early stationary-phase cells. This subpopulation may be able to withstand a sudden attack of a toxic substance because they are metabolically less active, leading to the phenomenon of shallow persistence. Interestingly, swimming E. coli cells are repelled by low concentrations of indole but attracted by high concentrations (81), which may sort E. coli populations into spatially separated shallow and deep persister subpopulations.

The S. aureus parental strain N315 does not contain the gene for a tryptophanase (BLAST search not shown) and should also be unable to produce indole. In S. aureus, halogenated or 5-methyl-indole (82, 83) inhibits persister formation, so that the function of indole in preventing early but allowing deep persistence may be similar. Indeed, indole production is widespread in Gram-positive and Gram-negative bacteria (84), so that its role in formation of shallow and deep persisters may also be widespread.

Indole alone is not responsible for survival on metallic copper surfaces. Both mCu60 strains also displayed a change in their fatty acid composition, which was also connected to persister formation (85). Both mutants also displayed a downregulated thymine biosynthesis (Fig. 8), for instance, of carB encoding the large subunit of the carbamoyl phosphate synthetase. A deletion of carB resulted in 2,500-fold decreased survival after antibiotic treatment in Pseudomonas aeruginosa (86), so that decreased pyrimidine biosynthesis may decrease persister formation. A genetic difference between the E. coli parent and its mCu60 mutant involved the insertion of IS2 into ycbX (Fig. 7), leading to upregulation of the adjacent, convergently transcribed gene, ycbY (Fig. 8A). This gene encodes a putative methyltransferase that, based on the presence of a THUMP domain (87), is expected to act on RNA. Modifications of the ribosomal 23S rRNA are involved in persister resuscitation (88), and ribosomes seem to play a central role in persister formation and resuscitation (89). This all links the mCu60 mutants by multiple threads to persisters.

The persister phenotype is not, however, the result of a genetic change (89), whereas the mCu60 phenotype in both bacteria clearly is. Second, only a small part of the population of stationary-phase cells enter the persister stage, while the whole mCu60 mutant population survived the initial copper shock on solid copper surfaces. This shock occurs within minutes, not allowing the cells to change gene expression. Moreover, downregulation of pyrimidine synthesis counteracts persister formation, and the role of indole is ambiguous. The mCu60 mutant cells seemed to have changed their global regulatory patterns. They are constantly in a defense-ready, albeit easy to resuscitate, shallow persister-like state, allowing survival due to a generally downregulated metabolism (Fig. 8), but are nevertheless able to resume rapid growth when conditions become favorable again (Fig. 3). Persisters are a form of dormancy, and its establishment seems to be a progressive, stepwise process (2). With a double-shock treatment that stops, or dramatically slows, metabolism of E. coli cells with rifampin and subsequently ampicillin treatment to kill any remaining nonpersister cells, nearly the whole bacterial cell population can be converted into persister cells (88). Similarly, the artificial laboratory evolution procedure used here shocked at each round the cells on copper surfaces and killed all nonpersisters, selecting mutants that may be permanently in such a shallow persister-like regulatory state.

In bacterial populations, only a small subpopulation of cells enter the persister state (2). The majority of cells attain resistance, tolerance, or resilience by changes in gene expression, which requires an active metabolism. The mCu60 mutants may have gained the ability to tolerate the copper shock occurring on solid copper surfaces at the cost of a lost metabolic flexibility. The fact that mutant strains able to survive on solid copper surfaces can evolve does not render copper surfaces useless as antibacterial material. Instead, this challenge highlights the necessity to enforce the use of copper surfaces with a routine regimen of strict cleaning. Copper surfaces alone do not prevent the transfer and spread of pathogenic microbes, but these solid antimicrobials are nevertheless an additional tool to increase the efficacy of a hygiene plan.

MATERIALS AND METHODS

Bacterial strains and general growth conditions.

The strains used in this study were E. coli K-12 (W3110) and S. aureus N315. All bacteria were grown on LB agar or in LB broth (Merck, Darmstadt, Germany) with shaking at 37°C for 16 h.

Preconditioning of metal coupons.

Copper (C11000, Cu 99.9%) and stainless steel (U2A AISI 32 control coupons; Wieland-Werke, Ulm, Germany) were used. The contact surface area of the coupons was 2.5 by 2.5 cm. Surfaces were prepared with sodium hydroxide and sulfuric acid as described previously (90) and kept in sterile petri dishes until further use.

Laboratory evolution of copper surface-tolerant mutants.

In order to select copper-tolerant bacteria, we modified adaptive laboratory evolution (ALE) by introducing short sequential rounds of artificial selection in a mutation-accumulation experiment (MAE). ALE is an in vitro approach for addressing questions on the basic mechanisms of molecular evolution and adaptive changes accumulating in microbial populations during selection pressure. MAE involves periodically bottlenecking a population such that evolution proceeds in an almost random manner, whereby selection is weak and mutations originate from errors in DNA copying, unrepaired DNA damage, or genome rearrangements (34–37). To create copper-tolerant bacteria, overnight cultures of E. coli or MRSA were concentrated in phosphate-buffered saline (PBS, pH 7.2) to approximately 2.5 × 10⁹ cells/ml. A volume of 40 μl (1 × 10⁸ cells) of this suspension was applied to a sterile cotton swab and spread three times across the coupons. After different time periods, the coupons were transferred to centrifuge tubes, washed, and plated on solid growth medium as described previously (13). Parental cells of Escherichia and Staphylococcus were completely contact killed after 1 (13) or 7 (12) minutes, respectively. To create mutants that survived longer on solid copper, E. coli was challenged for 30 s and MRSA for 3 min on copper surfaces, washed as described above, and spread on solid medium. Cultures were incubated at 37°C overnight. Some surviving colonies were picked, spread on solid medium, and incubated again at 37°C overnight. Subsequently, cells were again concentrated and copper surface-exposed as described above with the modification that the incubation period on copper was increased by 30 to 60 s for every iteration of exposure (Fig. 1). The growth step without selection pressure allowed for cellular repair processes and increased the genetic diversity due to mutations occurring during DNA replication (38). This in vitro evolution procedure was terminated once both E. coli and MRSA had acquired the capability to survive on metallic copper surfaces for at least 60 min. The final copper-tolerant bacteria were named E. coli mutant mCu60 and MRSA mutant mCu60 for “mutant surviving 60 min.”

Determination of growth kinetics.

Bacteria were grown in LB broth until the stationary phase, diluted 1:100 into fresh broth, and grown again at 37°C with shaking at 200 rpm. Aliquots were taken every 20 min, and their cell densities measured at an optical density at 600 nm (OD₆₀₀) until the cultures reached the stationary phase.

Quantification of MIC against CuSO₄.

Bacteria grown in LB broth until the stationary phase (24 h) were diluted 1:100 into fresh broth and grown for 2 h to the early exponential phase. Suspensions were spread on LB agar containing 0.25 mM to 4.0 mM CuSO₄ concentrations and incubated for 24 h at 37°C.

Inductively coupled plasma mass spectrometry (ICP-MS) analysis of cells and cell fractions.

To determine the metal content of copper surface-exposed cells, 40-μl aliquots of a culture of 2.5 × 10⁹ cells/ml of E. coli parental strain or its evolved mutant strain, mCu60, were spread on either copper or stainless-steel coupons and incubated for up to 90 min. As a control, 40 μl PBS buffer without cells was incubated on the coupons. At different time points, cells were rinsed with 100 μl ice-cold PBS, and the suspension was centrifuged at 4,500 × g for 30 min at 4°C. Cell pellets were washed twice with 50 mM ice-cold Tris-HCl buffer (pH 7) containing 10 mM EDTA to remove copper loosely bound to the cell surface. The supernatant was discarded, and any residual liquid was carefully removed at each step. The pellet was suspended in concentrated 67% (wt/vol) HNO₃ (trace metal grade; Normatom/PROLABO, Radnor, PA, USA) and mineralized at 70°C for 2 h. To determine the metal content in cytoplasmic or membrane cell fractions, 10 parallel preparations were mixed and mechanically extracted on ice via sonication (UW60; UniEquip GmbH, Martinsried, Germany) by 6 cycles of 30 s (range, 70%; power, 80 W). Cell debris and crude extract were separated by centrifuging at 20,817 × g for 30 min at 4°C. The crude extract was centrifuged again at 98,000 × g for 30 min at 4°C to separate the membrane fraction from the soluble, cytoplasmic fraction. The protein concentration of the fractions was quantified (91), and defined volumes were mineralized with 67% (wt/vol) HNO₃ as described above. Samples were diluted to a final concentration of 2% (wt/vol) nitric acid. Indium was added as an internal standard at a final concentration of 10 ppb. Elemental analyses were performed via ICP-MS using ESI-sampler SC-2 (Elemental Scientific, Inc., Omaha, NE, USA) and an X-Series II ICP-MS instrument (Thermo Fisher Scientific; Bremen, Germany) operating with a collision cell and flow rates of 5 ml/min of helium/H₂ (93%/7%), with an argon carrier flow rate of 0.76 liters/min and an argon make-up flow rate of 15 liters/min. An external calibration curve was recorded with ICP-multielement standard solution XVI (Merck, Darmstadt, Germany) in 2% nitric acid. Each sample was introduced via a peristaltic pump and analyzed for its metal content. PBS buffer was used for blank measurements and quality/quantity thresholds, and calculations were based on DIN32645 (97). The results were transformed from ppm, ppb, or ppt via molar units into atoms per sample and divided by the number of cells per sample, which had been determined before as CFU or for the cell fractions as mg/ml protein. The intracellular metal concentration was determined by using an E. coli cell volume of 2 μm³ (for E. coli cells growing in LB medium until the middle of the exponential phase) (92).

Live/dead staining of surface-exposed cells.

Live/dead staining (Live/Dead BacLight bacterial viability kit; Life Technologies, Darmstadt, Germany) differentiates between undamaged and damaged bacterial membranes by employing two fluorescence dyes, which intercalate into DNA. Live, undamaged bacteria fluoresce green, while membrane-damaged bacteria fluoresce red. This is because the green stain SYTO 9 penetrates the membrane of both undamaged and damaged cells, while the red stain propidium iodide can only penetrate damaged membranes and, after intercalating into the bacterial DNA, reduces the green fluorescence of SYTO 9. The bacteria and coupons were pretreated as described above for contact-killing assays, and staining was performed as previously described (11, 90). Cells were examined (λ_Ex 488/543 nm, λ_Em 522/590 nm) on an inverted confocal fluorescence microscope (Carl Zeiss MicroImagine, Jena, Germany). The ZEN 2008 program was used for analysis.

Fatty acid methyl ester (FAME) analysis.

Cultures for fatty acid analysis were grown on LB medium plates at 37°C for 24 h. FAMEs were extracted as described in MIDI technical note 101 for FAME extraction and analysis (MIDI, Inc., Newark, DE, USA). FAMEs were identified using the standard MIS library generation software (Microbial ID). Samples were run on an 6890N gas chromatograph (GC) (Agilent Technologies, Wilmington, DE, USA) equipped with autosampler, 7683B injector, split-splitless inlet, and flame ionization detector. The system was controlled with MIS Sherlock (MIDI, Inc., Newark, DE, USA) and Agilent ChemStation software. FAMEs were separated on an Agilent Ultra 2 column, 25 m long by 0.2 mm internal diameter by 0.33 μm film thickness. A sample volume of 10 μl was injected into the capillary column of the GC. The injector temperature was 170°C, and the detector temperature was 300°C. The oven temperature ramped from an initial 170°C to 270°C, at 5°C/minute. Following analysis, a rapid temperature increase to 300°C allowed cleaning of the column, which was maintained for 2 min. A split ratio of 30:1 was employed, and hydrogen was used as the carrier gas at a 1.2 ml/min constant flow rate; nitrogen was the “makeup” gas, and air was used to support the flame. Identification of fatty acids was achieved by their equivalent chain lengths (ECL), calculated from their retention times using the MIDI peak identification software. FAME standards were used for calibration (9:0, 10:0, 11:0, 10:0 2OH, 12:0, 13:0, 14:0, 15:0, 14:0 2OH, 14:0 3OH, 16:0, 17:0, 16:0 2OH, 18:0, 19:0, 20:0).

FAMEs were designated according to the omega-reference system convention X:YωZ, where “X” indicates the total number of carbon atoms in the molecule (except for molecules with a midchain branch), “Y” indicates the number of double bonds, “Z” indicates the position of the 1st double bond or cyclopropane ring, and “ω” indicates the position counted from the methyl end of the molecule. The prefix “i” indicates iso branching, “a” indicates anteiso branching, “10Me” refers to methyl branching on the 10th carbon from the carboxyl end, and “cy” stands for cyclopropane ring. The suffixes “c” and “t” indicate the cis and trans configuration, respectively. The number before an OH refers to the location of a hydroxyl-group relative to the carboxyl end of the molecule.

DNA and RNA extraction for whole-genome sequencing and transcription analysis.

For DNA and RNA preparation E. coli, E. coli mCu60, MRSA, and MRSA mCu60 were grown overnight on solid LB medium. Total DNA of the isolates was extracted with a QIAmp DNA minikit (Qiagen, Hilden, Germany) according to the manufacturer’s protocol. For RNA preparation, a few single colonies were resuspended in 100 μl RNase-free water (5Prime, Hamburg, Germany) and centrifuged for 3 min at 8,000 × g. The pellet was resuspended in 100 μl RNase-free water and vortexed for 3 min. A volume of 100 μl of a 1:1 chloroform/phenol solution (Roth, Karlsruhe, Germany) was added to the suspension and incubated for 30 min at 70°C with shaking. The phases were separated at 10,000 × g for 10 min, and 100 μl of the supernatant was removed and transferred to a fresh tube. The RNA extraction was prepared with an RNeasy kit (Qiagen) as described in the manufacturer’s instructions. The elution volume was 50 μl for each preparation.

Genome sequencing, mapping, and assembly.

Genomic DNA extracted from the four strains was used to create whole-genome sequencing (WGS) libraries using the Illumina-compatible Nextera DNA sample prep kit (Epicentre, WI, USA) according to the manufacturer’s protocol. All four libraries were sequenced on an Illumina MiSeq sequencer in a 2 × 250 bp run, yielding 4,686,804 (E. coli), 5,252,652 (E. coli mCu60), 4,749,404 (MRSA), and 4,825,760 (MRSA mCu60) reads in total. The reads were mapped against the appropriate reference sequences (E. coli W3110 [GenBank accession number NC_007779.1] and MRSA N315 [NC_002745.2 and NC_003140.1]) using the exact alignment program SARUMAN (93), resulting in mapping coverage of 202-fold to 310-fold on average. Genomic coverage by at least one read ranged from 99.68% (E. coli) to 100% (MRSA N315). Single nucleotide polymorphisms (SNPs) were extracted from mapped reads with customized Perl scripts using a minimum coverage of 10 reads and a minimum allele frequency of 80% as thresholds for detection. For the de novo assembly, the Newbler assembler v2.8 (Roche, Mannheim, Germany) was used. Transposon (TNP) insertion sites were searched by manual inspection of the data in CONSED (94, 95). The data were submitted to the SRA (accession number SRP286723).

Comparative cDNA sequencing (RNASeq).

For the comparison of the transcriptomes of the mutant and the respective parent strains, 10 μg total RNA per sample was used. As the first step, the rRNA was depleted using the Ribo-Zero rRNA removal kit for bacteria (Epicentre, Madison, WI, USA). Subsequent purification of the rRNA-depleted sample by ethanol precipitation was done according to the manufacturer’s instructions. From each of the resulting rRNA-depleted RNA samples, one Illumina-compatible cDNA library was created using the TruSeq stranded total RNA sample prep kit (Illumina, Berlin, Germany). All four libraries were sequenced on an Illumina MiSeq sequencer in a 2 × 50-bp run. The reads were mapped against the appropriate reference sequences (E. coli W3110 [GenBank accession number NC_007779.1] and MRSA N315 [NC_002745.2 and NC_003140.1]) using the exact alignment program SARUMAN (93) and were then filtered, retaining only read pairs with at least one uniquely matching partner. This yielded a total of 2,086,908 (E. coli), 2,011,609 (E. coli mCu60), 3,749,266 (MRSA), and 3,752,574 (MRSA mCu60) mapped read pairs. Afterward, the mapped reads were visualized, and the relative expression values were calculated using ReadXplorer (96). The KEGG database (51, 52) was used for further analysis.

Data availability.

The genomic sequencing data have been submitted to the SRA (accession number SRP286723). Additional data, i.e., the RNASeq sequencing data, are available via BioProject number PRJNA667899.

File S1 Changes in transcription in E. coli mCu60 compared to E. coli.

Copyright © 2020 American Society for Microbiology.

All Rights Reserved.

File S2 Changes in transcription in MRSA mCu60 compared to MRSA.

Copyright © 2020 American Society for Microbiology.

All Rights Reserved.

Copyright © 2020 American Society for Microbiology.

All Rights Reserved.