Experimental evolution of bacterial survival on metallic copper

Abstract

Antimicrobial copper‐containing surface materials have a great potential of reducing the risks of healthcare‐associated infections (HAIs), but their increased use in hospital facilities may select copper‐resistant strains, causing concerns to antimicrobial resistance management. Here, we describe a long‐term bacterial evolution experiment wherein a non‐pathogenic Pseudomonas strain was subjected to daily transfer in laboratory media with and without copper‐mediated contact killing. The copper treatment sequentially involved two surface materials differing in Cu content and thus contact killing effectiveness: first on brass (Cu 63.5%) and then on pure copper (Cu 99.9%). A gradual increase in bacterial survival rate (or a decrease of killing effectiveness) was observed over time on the related copper surfaces. For the final evolved populations after 320 transfers, 37.8% cells of the copper‐evolved populations were able to survive 60 min on pure copper, whereas populations in the control lines remained sensitive with a survival rate of 0.09% under the same contact killing condition. Genome re‐sequencing revealed ~540 mutations accumulated in the copper lines but only 71, on average, in the control lines (variant frequency > 0.5). The mutagenic activities of Cu⁺ ions were confirmed by measuring spontaneous mutation rate in a laboratory medium supplemented with copper sulfate at a non‐inhibitory concentration. The copper‐evolved populations have acquired increased resistance to Cu⁺ ions and tobramycin (an aminoglycoside antibiotic), but showed decreased production of biofilm, exoprotein, and pyoverdine. Together, our data demonstrate the potential of bacteria to evolve prolonged survival on metallic copper, and the long‐term impacts should be considered with increased copper usage in hospital environments.

INTRODUCTION

Copper in its ionic form (Cu²⁺ or Cu⁺) is an essential micronutrient for cellular organisms. It functions as a co-factor for oxidation–reduction reactions mediated by copper-containing oxidase, including those in the electron-transport chain. However, too much copper is harmful. This is largely due to its ability to produce hydroxyl radicals (OH⁻) in a Fenton-like reaction. Cu⁺ ion toxicity is much higher for lower organisms such as bacteria and fungi, when compared to plants, animals, and humans (Ladomersky & Petris, 2015). When the toxicities of 17 metals were compared with soil bacteria, copper ranks the fifth highest, behind silver, mercury, chromium, and cadmium (Drucker et al., 1979). Therefore, antimicrobial copper compounds have long been used in agriculture for the control of plant diseases (La Torre et al., 2018; Lamichhane et al., 2018).

In recent years, however, the potential application of solid metallic copper (Cu) in public health has received increasing research attention. Healthcare-associated infections (HAIs) are a major health problem worldwide, affecting more than 10% patients admitted to hospitals for the treatments of other diseases (Weber et al., 2013). A major source of HAIs is microbial pathogens present on commonly touched items in hospitals (e.g., doorknobs, push plates, toilet seats, bed rails, and IV poles), which are normally made of surface materials lacking any intrinsic antimicrobial properties (such as aluminum, stainless steel, plastics, or woods). Therefore, introducing self-sanitizing surface materials like Cu and its alloys can potentially reduce the transmission of microbial pathogens via contacts between hospital patients, visitors, and healthcare workers. A wide range of microorganisms, including coronaviruses, can be killed within minutes of contact with a Cu surface [see reviews (Govind et al., 2021; Grass et al., 2011; Santo et al., 2008)]. The process is generally termed as Cu-mediated contact killing. Much attention has been focused on the roles of Cu in controlling various HAI-associated pathogens and its effectiveness in maintaining hospital environmental hygiene (Colin et al., 2018; Michels et al., 2015; Schmidt et al., 2012). The potential health risks remain largely unexplored (Maertens et al., 2021; Santo et al., 2010).

Cu products have been domestically used for centuries without any major health concerns, and thus, they are generally considered safe. The US Environmental Protection Agency has approved the registration of more than 500 types of Cu and its alloys as antimicrobial materials (Vincent et al., 2016). However, with an increased long-term use of Cu in hospital environment, pathogenic bacteria may develop Cu resistance, which is likely linked to microbial pathogenicity and antibiotic resistance (Virieux-Petit et al., 2022). Many field studies showed that copper amendment of agricultural soil selects for antibiotic-resistant bacteria in the field (Berg et al., 2005; Hu et al., 2016; Knapp et al., 2011). At the molecular level, antibiotic resistance is linked to the cellular Cu⁺ metabolism under the control of global regulators such as MarR as revealed in E. coli (Hao et al., 2014). More importantly, accumulation of toxic Cu⁺ ions is a natural defense mechanism employed by human cells against bacterial infection (Chaturvedi & Henderson, 2014; Hodgkinson & Petris, 2012; Ladomersky & Petris, 2015). Therefore, in the worst-case scenario, Cu could select for highly virulent pathogenic bacteria with overlapping resistance to antibiotics, thus producing detrimental effects on the current antibiotic resistance management.

The mechanisms of Cu-mediated contact killing are not fully understood, but it has been generally accepted that toxic Cu⁺ ions are the major cause of cell death (Vincent et al., 2018). Copper ions disrupt bacterial membrane integrity and further damage DNAs in the cytoplasm (Hong et al., 2012; Molteni et al., 2010; Warnes et al., 2012). However, the free Cu²⁺ ion concentrations on metallic Cu are not high enough to kill bacteria, and direct physical contact between bacteria and copper was shown to be crucial for the antimicrobial activities (Mathews et al., 2013; Molteni et al., 2010). Liu and Zhang (2016) proposed and experimentally tested a copper ion burst releasing (CIBR) model, which posits that bacteria are rapidly killed by toxic Cu⁺ ions continuously produced from chemical reactions between atomic copper and organic compounds in the bacterial cell surface structures, for example, exopolysaccharides, lipopolysaccharides, and pili. According to the CIBR model, no bacteria would be able to survive from direct physical contact with a Cu surface, and it is only a matter of time how long it takes to be killed.

Bacterial cells have evolved three major types of mechanisms to survive in high copper environments: (1) copper efflux systems that act to pump excess Cu⁺ ions outside of the cell; (2) sequestration systems that detoxify Cu⁺ by specific high-affinity binding with proteins (e.g., copper chaperones); and (3) oxidation systems that convert Cu⁺ into less toxic Cu²⁺ form (Giachino & Waldron, 2020). These systems collectively determine the adaptive capability of a bacterium to environmental stress imposed by ionic copper (Santo et al., 2008). However, as bacterial cells continuously absorb and accumulate more and more Cu⁺ ions on metallic Cu surfaces, these resistance systems will eventually be outpaced, resulting in cell death. The only chance of survival is to escape Cu surfaces before being killed. This may occur in the hospital environments by means of human contacts. For example, oily substances from hands can physically separate bacteria from Cu surface and prevent further damage of the bacterial cell (de Carvalho & Caramujo, 2014).

Here, we hypothesized that optimization of the Cu⁺ resistance determinants (and other yet-unknown genetic factors) can enhance bacterial survival on metallic copper, and thus, selection can potentially occur for mutants that survive longer on Cu surfaces. To test this hypothesis, we first performed a long-term evolution experiment by subjecting Pseudomonas fluorescens SBW25 to daily transfers in a laboratory medium with and without Cu contact killing. P. fluorescens SBW25 is a non-pathogenic model bacterium commonly used in experimental evolution (Zhang & Rainey, 2013). It belongs to the same genus as P. aeruginosa, one of the most significant nosocomial pathogens (Moradali et al., 2017). Having observed an increase in bacterial survival rate on Cu surfaces, we subjected the evolved populations to genome re-sequencing, followed by characterization of phenotypes selected on the basis of the detected mutations. The data enable us to discuss the potential evolutionary impacts associated with an increased use of Cu-containing surface materials in hospitals for reducing the risks of HAIs.

MATERIALS AND METHODS Bacterial cultivation and evolution

Pseudomonas fluorescens SBW25 and its naturally evolved derivatives were routinely cultured in lysogeny broth (LB) or King's B (KB) medium at 28°C as previously described (Zhang & Rainey, 2013). When grown in the minimal M9 medium, glucose (0.4% or 22.2 mM) and NH₄Cl (1 mg/ml or 18.7 mM) were supplemented as the carbon and nitrogen sources. For biofilm assays, P. fluoresces was grown in the minimal M9 salt medium supplemented with succinate (20 mM) and histidine (10 mM) as the carbon and nitrogen sources. Tobramycin was purchased from Sigma-Aldrich and added at the indicated concentrations. Growth kinetics were determined in a microtiter plate using a Synergy 2 plate reader installed with Gen5 software (Bio-Tek Instruments). Strains were inoculated from cells stored at −80°C freezer to ensure that bacteria in comparison were physiologically equivalent (Zhang & Rainey, 2007b). Absorbance data were collected at 10-min intervals over a period of 24 h.

Experimental evolution was performed with SBW25-lacZ, which was derived from P. fluorescens SBW25 with the insertion of a promoterless lacZ gene at a neutral prophage site (Zhang & Rainey, 2007a). This strain forms distinctive blue colonies on LB agar plate supplemented with 5-bromo-4- chloro-3- indolyl-β-D-galactopyranoside (X-gal, 60 μg/ml). This property was used as a marker for detecting any potential contaminations. Strain identity was additionally checked by strain-specific PCR amplification of xut genes using primers xutA-compF and xutR-lacZF (Liu et al., 2015). As outlined in Figure 1, the work involved two treatments with and without Cu contact killing, and each was performed with eight independent evolutionary lines (Liu & Zhang, 2016). For Cu treatment, cells from overnight culture were washed once with sterile water, and 20 μl was then dropped onto a Cu coupon (1 × 1 cm) placed in a Petri dish. After 60 min at room temperature (~20°C), the brass coupon was transferred into a 30-ml plastic tube containing 2 ml LB medium. Survived cells on the coupon (~10,000) were released into the fresh medium by vigorous vortexing, and the coupon was subsequently removed with the help of a sterile inoculation loop. In the control line without Cu treatment, bacterial cultures were subjected to serial daily transfer by inoculating 5 μl culture into 5 ml fresh medium (thus, 10 generations per transfer). The evolved bacterial populations were stored at −80°C freezer every 10 transfers for future assays of contact killing on Cu surfaces. The metal sheets of appropriately 1 mm thick were purchased from Wakefield Metals: brass (UNS number C27200, 63.5% Cu plus 36.5% Zn), pure copper (UNS number C11000, 99.9% Cu).

FIGURE 1. Increase in bacterial survival on solid Cu surfaces during the experimental evolution. (a) An outline of the experimental procedures. Bacteria grown in 2 ml LB broth in eight replicates were subjected to bottleneck every 24 h with Cu treatment. For each transfer, 20 μl of washed cells from overnight culture was inoculated onto the surface of a metal coupon. After 60 min at room temperature, cells on the coupon were suspended into a fresh LB broth. Cells survived from the contact killing would form a new population for the next round of contact killing. The Cu material was shifted from bass to pure copper when the maximum survival rates were reached after 200 transfers. Bacteria in the eight control lines were propagated by normal daily transfers without Cu treatment (not shown in the schematic map). (b) Rates of bacterial contact killing on the surfaces of brass (left) and pure copper (right). The evolved populations were stored in −80°C freezer at every 10 transfers, and their resistance levels were determined in parallel on the brass and copper surfaces. Data are means and standard errors of eight replicate cultures for each evolutionary line.

Copper resistance assays

Bacterial survival on Cu surfaces was measured using the so-called “wet” inoculation method (Molteni et al., 2010). Briefly, bacterial cells in overnight LB culture were spun down and re-suspended into the same amount of sterile water, and then 20 μl was inoculated onto the surface of a Cu coupon (1 × 1 cm). The coupons were placed beforehand in a Petri dish that contains a few drops of sterile water for keeping moisture. After a given period of treatment at room temperature (~20°C), bacteria on coupon were transferred into 2 ml H₂O in a 30-ml tube. Cells were released by vortexing, and colony formation units (cfu) were subsequently determined by dilution plating onto LB agar plates. Bacterial resistance to ionic copper was assessed in LB agar plate (or LB broth medium in 96-well microtiter plate) supplemented with varying concentrations of CuSO₄. The minimum inhibitory concentration (MIC) is defined as the lowest CuSO₄ concentration at which no colonies were observed after 3-day incubation.

Biofilm quantification

The ability of P. fluorescens to form biofilm was quantitatively assessed using a standard method previously described by O'Toole (2011). Briefly, bacteria in an overnight culture were inoculated into a test medium by 100 times dilution and then aliquoted into a 96-well microtiter plate (100 μl per well) placed in a 28°C static incubator for 1, 2, or 3 days. However, only the 3-day data were presented for clarity, and comparison was made between the evolved populations and their ancestral strain. Biofilm formed on the inner wall of a well was stained with 125 μl of 0.1% Crystal Violet (CV) for 15 min. The dye was then dissolved using 125 μl of 30% acetic acid and spectrophotometrically measured at the wavelength of 550 nm (A₅₅₀). Meanwhile, cell density in the broth was estimated by measuring absorbance at the wavelength of 450 nm (A₄₅₀). Data in comparison were obtained from independent cultures grown in the same 96-well plates.

Bacterial fitness and mutation rate

Given that the evolved strains carry a neutral lacZ marker gene, their performance in laboratory medium was examined by direct competition with wild-type P. fluorescens SBW25. Competitive growth was initiated by mixing the two competitors at a 1:1 ratio, and the mixed bacteria were inoculated into a fresh medium by 100 times dilution. The initial and final population densities (N_i) were determined at time t = 0 and at t = T by dilution plating on LB + X-Gal plates. The average rate of population increase (i.e., Malthusian parameter m_i) is calculated for each strain using the formula of m_i = ln[N_i(T)/N_i(0)]. Relative fitness is expressed here as the selection rate constant (SRC): r_ij = m_i − m_j, resulting in a fitness of zero when the two strains or populations are equally fit (Lenski, 1991).

Mutation rate was estimated using a modified procedure of fluctuation test as previously described (Pal et al., 2007). For each strain, at least six microcosms were set up and each was inoculated with ~500 cells. After grown in 6 ml LB broth for 24 h in a shaken (160 rpm) incubator at 28°C, final cell density and the frequency of antibiotic resistance mutant were counted by dilution plating in LB and LB supplemented with nalidixic acid (75 μg/ml). Rate of spontaneous mutation was calculated using an integrated web tool named as bz-rates, which is based on Luria–Delbrück distribution-generating algorithm (Gillet-Markowska et al., 2015).

Quantify pyoverdine and exoproteins

Pyoverdine production was quantified by measuring fluorescence of the supernatant at 460 nm with an excitation wavelength of 365 nm in a Synergy 2 multimode microplate reader (Bio-Tek Instruments). Data were expressed as relative fluorescence units (RFU; Zhang & Rainey, 2013). Exoproteins were visualized by 12% sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE), wherein 20 μl cell-free supernatant prepared from overnight LB culture was loaded into each well. Protein concentration in the LB culture supernatant was estimated using the Quick Start™ Braford Dye Reagent (Bio-Rad) following the manufacturer's recommendation.

Genome sequencing and statistical analysis

Bacterial DNAs were prepared from the evolved populations using the Promega Wizard Genomic DNA purification Kit (In Vitro Technologies). Sequencing was performed on an Illumina HiSeq 4000 platform using services provided by Novogene Technology Co. Ltd. Libraries were constructed from sheared DNA fragments of ~300 bp. The obtained 150 bp paired-end reads were processed in the Geneious 9.0.5 program (Biomatters Ltd). Sequences were mapped to the SBW25 reference genome (NC_012660.1) using the “Medium-Low Sensitivity/Fast” option. Mutations were identified by “Find Variations/SNPs” analysis with a minimum coverage of 10, minimum variant frequency of 0.5, maximum variant p-value of 10⁻⁶, minimum strand-bias p-value 10⁻⁵, and find polymorphism inside and outside of coding sequences (CDS). Statistical analyses were performed using related tools available in GraphPad Prism v9.

RESULTS Experimental evolution led to an increase in bacterial resistance to metallic copper

As outlined in Figure 1a, P. fluorescens SBW25 was subjected to daily transfer in eight replicates, with and without Cu contact killing. The Cu treatment was first performed on brass and then shifted to pure copper after 200 transfers when the survival rate reached ~30% within a 60-min contact. Dynamic changes in copper resistance were assayed in parallel, using cells stored at −80°C freezer every 10 transfers. Results showed gradual increases in Cu resistance on the surfaces of brass and pure copper over the course of experimental evolution on Cu. For the final resultant 320th populations, an average of 37.8% cells can survive from contact killing on pure copper for 60 min. In contrast, cells in the eight control lines remained sensitive to metallic copper with a mean survival rate of 0.09% (Figure 1).

Cu⁺ ion toxicity is the major mechanism underlying the Cu-mediated contact killing. Thus, we hypothesized that the Cu-evolved strains may have developed higher resistance to ionic copper. To test this, we first determined MIC of the final 320th populations on LB agar plates supplemented with varying concentrations of CuSO₄. Parallel to our expectation, the control lines and wild-type P. fluorescence SBW25 had the same MIC value of 3.25 mM, whereas the Cu-evolved lines had an increased MIC value of 3.75 mM for 320D3 and 320D4, and 3.5 mM for the other six populations (Figure S1). Next, Cu⁺ ion resistance was assessed in a 96-well microtiter plate containing 200 μl LB broth per well. The control lines (320C1 to 320C8) grew similarly as wild-type SBW25 in LB broth, and their growth was completely ceased with the supplementation of 3.5 mM CuSO₄ (data not shown). However, the Cu-evolved populations showed significant growth in LB + CuSO₄ (3.5 mM) broth with an average final cell density (A₆₀₀) of 0.39 (Figure 2b). Of note, a growth defect was observed for all Cu-evolved populations when grown in LB broth without CuSO₄ supplementation (Figure 2a). Together, our data indicate that the Cu-evolved strains have achieved increased resistance to both metallic copper and ionic copper.

FIGURE 2. Cu-evolved Pseudomonas populations showed an increase in Cu+ resistance. Growth kinetics were measured with bacteria grown in 96-well microtiter plate in LB broth (a) and LB broth supplemented with 3.5 mM CuSO4 (b). Absorbance at the wavelength of 450 nm (A450) was recorded every 10 min after inoculation. Data are means of five independent cultures (standard errors are small and not shown for clarity).

Genome re-sequencing of the evolved bacterial populations

To determine mutations accumulated over the course of experimental evolution, total DNAs were prepared from bacterial populations of the 320th and 200th transfers on copper (named 320D1 to 320D8 and 200D1 to 200D8, respectively). Additionally, final populations of the eight control lines (named 320C1 to 320C8) were also included in total DNA isolation and subsequent genome re-sequencing. Sequencing reads were aligned to the reference genome of P. fluorescens SBW25 (NC_012660.1) with ~99.5% pairwise identity. The average sequence depths were 81x for 320D1-D8, 165x for 200D1-D8, and 170× for 320C1-C8 populations. Mutations were identified in Geneious R9 with a minimal variant frequency of 0.5. In the control lines, an average of 71 sequence variations were detected, and 21–24 mutations were fixed in the population (Table 1). Most variants are single nucleotide polymorphisms (SNPs, 75.7%). In contrast, about 540 variations were accumulated in the final copper-evolved populations (320D1 to 320D8), and the number of mutations fixed in the population (100%) ranges from 240 to 408. Even after 200 transfers, the number of mutations in the Cu lines (130.6 on average) is significantly higher than that of the control lines (Figure 3a). Similar trends were obtained with regard to the total number of mutations when the analysis was done using minimum variant frequency of 0.5, 0.7, and 1.0 (Table 1). Together, the experimental evolution data implicate higher mutation rates associated with the Cu treatment.

TABLE 1 Number of mutations occurred at frequencies larger than 50% in a population.

Populationa	Transversion (0.5)	Transition (0.5)	SNPs (0.5)	Indel (0.5)	Mutation (0.5)	Mutation (0.7)	Mutation (1.0)
320C1	2	42	44	15	59	50	23
320C2	2	47	49	18	67	50	23
320C3	4	64	68	21	89	56	24
320C4	4	52	56	19	75	51	21
320C5	3	50	53	16	69	50	24
320C6	3	48	51	16	67	50	21
320C7	3	57	60	17	77	50	23
320C8	3	47	50	16	66	50	21
200D1	2	100	102	22	124	78	14
200D2	3	101	104	22	126	91	22
200D3	2	108	110	25	135	119	16
200D4	1	133	134	26	160	79	23
200D5	2	101	103	23	126	119	21
200D6	2	101	103	23	126	120	8
200D7	2	97	99	19	118	77	31
200D8	1	98	99	31	130	84	9
320D1	11	436	447	83	530	506	240
320D2	15	434	449	98	547	490	313
320D3	12	406	418	83	501	461	238
320D4	14	471	485	92	577	518	301
320D5	14	366	380	80	460	333	206
320D6	15	380	395	88	483	339	211
320D7	26	461	487	106	593	557	399
320D8	28	500	528	102	630	600	408

^aPopulations named 320C1 to 320C8 are the control lines after 320 transfers in LB broth. Populations evolved on copper after 200 and 320 transfers are designated 200D1 to 200D8 and 320D1 to 320D8, respectively. Minimum variant frequency is indicted in parenthesis.

FIGURE 3. Higher mutation rates induced by solid Cu and Cu2+ ions. (a) Total number of mutations accumulated in the control lines after 320 transfer (320C1–320C8) and the Cu treatment lines after 200 transfers (200D1–200D8) and 320 transfer (320D1–320D8). The number of indel mutations (open bars) is stacked on top of SNPs (solid bars). Data are means and 95% CI of eight replicate lines. (b) Mutation rates estimated by fluctuation test in LB broth supplemented with varying concentrations of copper sulfate. Data are means and 95% CI of eight independent cultures calculated using the bz-rates Web tool.

Next, we tested the hypothesis that Cu exposure can enhance bacterial mutation rate as free Cu⁺ ions cause direct damage to DNAs. The mutation rates of P. fluorescens SBW25 were examined in LB medium with CuSO₄ being supplemented below the maximum non-inhibitory concentration. Results of the fluctuation test show that mutation rate was about five times greater when CuSO₄ (2 mM) was added into the LB broth (Figure 3b). Additionally, it is interesting to note that mutations in DNA mismatch repairing systems were detected in both the copper-evolved lines and the control lines. Consequently, their spontaneous mutation rates were significantly higher than that of the wild-type SBW25 (Figure S2).

Phenotypic characterization of the Cu-evolved populations

Mutations were detected in a total of 1392 open reading frames across the eight Cu-evolved lines from 320D1 to 320D8. These genes can be classified into eight functional categories (Figure 3a). While functional characterization of these mutations was beyond the scope of this study, we used this information to investigate important phenotypes that are functionally relevant to copper homeostasis and general stress responses. Results are summarized below.

First, we examined their ability to form biofilms in comparison with the wild-type ancestor. Mutations were detected in 50 genes involved in the production of extracellular polymeric substances (EPS) and the determination of cell envelop structures (Figure 4a). These include the well-characterized wss genes and the associated wsp regulators for the synthesis of acetylated cellulose polymers; algL and algK for biosynthesis of alginate polymer. Of particular note are multiple unique mutations accumulated in three putative adhesin-encoding loci (Figure 4b). Adhesins are a group of cell surface proteins that assist bacterial attachment to biotic and abiotic surfaces, and surface attachment is the crucial initial step of biofilm formation (Palmer et al., 2007). Together, all the genetic data suggest possible defects in biofilm formation. Indeed, biofilm formation was significantly lower for all eight Cu-evolved populations when compared to wild-type SBW25 in LB medium (Figure 4c). The difference was much more pronounced when grown in the minimal M9 medium (Figure 4d). Such differences were not observed for the densities of cells (A₄₅₀) present in the corresponding broth cultures (Figure S3). Moreover, we quantified biofilms formed by strains in the control lines (320C1 to 320C8), and the data are presented in Figure S4. No significant differences were found in LB medium between the control lines and the wild-type ancestor (p < .05, Figure S4A). However, significant differences in biofilm were detected when the control lines were grown in the minimal M9 medium (p < .05, Figure S4B). The data are consistent with the finding that mutations associated with cellulose (wssE and wspF) and alginate (algF) production were also found in the eight control lines (Data S1).

FIGURE 4. Functional overview of gene mutations and the phenotypic effects on biofilm formation. (a) Functional distribution of mutated genes in the eight copper evolution lines. (b) Schematic maps of three adhesion-encoding loci. Red dots denote the approximate positions of the observed mutations. (c) Biofilm formation in LB medium. (d) Biofilm formation in minimal M9 medium with glucose and NH4Cl. Biofilms on the inner surfaces of a 96-well microtiter plate were stained with 0.1% crystal violet (CV) at 3 days after inoculation, and the dye was then solubilized with 30% acetic acid, followed by measurement of absorbance at the wavelength of 550 nm (A550). The line in the middle of the box represents the median of 10 independent cultures, whereas the bottom and top boxes represent the 25th and 75th percentiles, respectively. The lower and upper whiskers indicate the smallest and largest values, respectively. One-way ANOVA multiple comparisons indicate significant differences (p [less than] .05) with the related wild-type (WT) controls. D1 to D8 represent the 320th populations 320D1 to 320D8, respectively.

Next, we tested if there are any changes in proteins that are secreted into the laboratory medium, as mutations in exoprotein genes were clearly enriched during evolution on copper (Figure 4a). Proteins in the supernatant of LB cultures were quantified using the Bradford method (Figure S5A). Results showed that the amount of exoproteins was significantly lower in 320D6, 320D7, and 320D8 cultures when compared to the wild type, and cultures 320D1 to 320D5 were at the similar level as the negative control, that is, the un-inoculated LB broth. Subsequent SDS-PAGE analysis detected a clear band for the wild-type strain, whereas the Cu-evolved populations produced no visible bands (Figure S5B).

Various nonsynonymous mutations were detected for pyoverdine synthesis genes in all eight Cu lines. These include five and seven unique mutations in pvdI and pvdJ, respectively. Both genes encode non-ribosomal peptide synthetase (NRPS) involved in the biosynthesis of the peptide backbone (Guillon et al., 2013). Additionally, three unique fpvA mutations were fixed in three populations (320D2, 320D6, and 320D8). FpvA is an outer membrane ferric pyoverdine receptor, which regulates the expression of pyoverdine synthesis genes through the FpvA-FpvR-PvdS signaling pathway (Moon et al., 2008). To test the potential effects of these mutations, pyoverdine production was measured by growing bacteria in KB and LB broth media. Pyoverdine was pronouncedly produced in KB but not in LB as previously reported for SBW25 (Figure S6). Pyoverdine production was almost abolished for 320D3 and 320D4 and occurred at significantly lower levels for other six populations in KB medium (Figure S6A). Of note, pyoverdine was produced at significantly lower levels for the eight control lines in KB, despite the observation that no mutations of the previously identified pyoverdine genes were detected at a frequency larger than 50% (Figure S6A). However, the control lines produced pyoverdine at similarly low levels as the wild type when grown in the LB broth medium (Figure S6B).

Finally, we explored the potential effects on bacterial resistance to tobramycin, a commonly used antibiotic for treating Pseudomonas infections in humans (Yeung et al., 2011). Resistance was assayed in broth medium supplemented with varying concentrations of tobramycin. The assay was first performed in a minimal medium with succinate and histidine as the source of carbon and nitrogen, respectively. The related catabolic genes have genetically been characterized (Naren & Zhang, 2020, 2021). As shown in Figure 5a, growth of wild-type SBW25 was completely inhibited by 2 μg/ml tobramycin, while significant growth was detected for 320D1-D6. However, growth defects of varying degrees were detected in the same medium without antibiotic supplementation (Figure 5a), and two populations (320D7 and 320D8) showed no growth within 2 days of incubation at 28°C. This finding is not surprising because of mutations in DctB/DctD required for succinate utilization and CbrAB/NtrBC genes involved in the activation of histidine utilization (hut) genes. Additionally, mutations occurred in hut catabolic genes such as hutI and hutT_h (Naren & Zhang, 2021). Next, we assayed tobramycin resistance in the nutrient-rich LB medium (Figure 5c). All eight Cu-evolved lines showed enhanced growth to varying extents when compared to wild type in the presence of tobramycin (2 μg/ml).

FIGURE 5. Cu-evolved populations showed slower growth but an increase in tobramycin resistance. (a) Growth dynamics in M9 salt medium supplemented with 20 mM succinate (Suc) and 10 mM histidine (His). (b) Growth in the same succinate and histidine medium with the addition of 2 μg/ml tobramycin. (c) Growth dynamics in LB containing 2 μg/ml tobramycin. Data are means of three independent cultures (errors bars are small and not shown for clarity). (d) Relative fitness of the Cu-evolved bacterial populations to wild type in LB medium and LB with 2 μg/ml tobramycin. WT—wild-type P. fluorescens SBW25; D1–D8, 320D1 to 320D8.

To examine the growth effects more closely, the evolved populations were subjected to direct competition with wild type in LB medium supplemented with and without tobramycin (2 μg/ml). A fitness zero indicates that the fitness of the tested population is identical to wild-type SBW25 (Figure 5d). Results confirmed the finding that the Cu-evolved populations exhibit reduced growth in LB medium (Figure 2a) and their fitness was significantly less than zero (p < .05, Figure 5d). In contrast, they were more fit than wild type in the presence of tobramycin (2 μg/ml; Figure 5d). Together, our data indicate that the Cu-evolved populations displayed slower growth but higher tobramycin resistance phenotypes.

DISCUSSION

In the present study, we subjected a non-pathogenic model bacterium P. fluorescens SBW25 to daily transfer with and without a contact killing treatment on the surfaces of brass and pure copper. The experiment was designed to test the potential health concerns associated with the common use of antimicrobial Cu-containing surface materials for controlling hospital environmental hygiene, as a new strategy for reducing the risks of HAIs. Results of this long-term evolution experiment demonstrate the evolutionary potentials of bacterial pathogens developing resistance to metallic Cu in terms of prolonged survival. Furthermore, we showed that the Cu-evolved strains had increased resistance to Cu²⁺ ions when grown in laboratory media. This result is generally consistent with the established role of Cu²⁺ toxicity in Cu-mediated contact killings (Liu & Zhang, 2016). Pathogens that can survive longer will clearly have a greater chance to escape Cu surfaces and persist in the hospital environments. Soil and oily substances on hands can protect bacteria from direct contact with Cu surfaces (de Carvalho & Caramujo, 2014; Santo et al., 2010). Therefore, our data suggest that natural selection may occur when Cu-containing surface materials are commonly used in hospital facilities, and proper regular cleaning is thus required to minimize the selection of copper-resistant strains (Bleichert et al., 2021).

The antimicrobial efficacy of Cu surfaces is determined by many factors (Grass et al., 2011). In general, the antimicrobial activities of Cu alloys are proportional to the Cu content of the surface materials. For example, brass (Cu 63.5%) is less efficient than pure copper (Cu 99.9%; Liu & Zhang, 2016). After 200 transfers, ~70% cells were killed by contact killing on brass. However, when the same bacterial populations were subjected to contact killing on pure copper, more than 99.99% cells were killed (only 10,000 cells survived) under the similar condition (Figure 1b). Apart from the material itself, environmental conditions such as temperature and moisture also produce significant influences on the efficacy of contact killing. It generally takes a shorter time for bacteria to be killed on dry surfaces relative to moist ones (Santo et al., 2011). The so-called “wet” inoculation method is commonly used in laboratory assays of bacterial contact killing, and it was adopted in this study in the process of experimental evolution. However, the moist surfaces may not reflect the complex situations in the healthcare environments.

Copper has long been used as a common coinage metal, including the New Zealand $1 coin made of copper, aluminum, and nickel (Cu 92%, Al 6%, and Ni 2%). Thus, these coins have intrinsic antimicrobial activities, and bacteria were rapidly killed once inoculated onto their surfaces (Vriesekoop et al., 2016). In a previous study, Santo et al. (2010) conducted a microbiological survey of Cu alloy coins and identified bacteria that are resistant against the toxic properties exerted by dry Cu surfaces. Most of these resistant isolates were Gram-positive staphylococci and micrococci, but also included Gram-negative species such as Pseudomonas oleovorans and Pseudomonas putida. Interestingly, these dry-surface-resistant strains seemingly had the similar levels of resistance to wet Cu surfaces and Cu⁺ ion resistance in laboratory medium, when compared to phylogenetically related reference strains (Santo et al., 2010). These data implicate the complexity of bacterial survival on antimicrobial surfaces. The “sensitive” strains survived from contact killing likely due to soiling by organic matters (de Carvalho & Caramujo, 2014).

We showed that Cu⁺ ions were mutagenic as we would expect based on their toxic effects on genomic DNAs (Warnes et al., 2012). Higher spontaneous mutation rates were observed even when CuSO₄ was added at a non-inhibitory concentration (Figure 3b). This finding explains the significantly higher numbers of sequence variations observed in the Cu-evolved lines when compared to the non-Cu controls (Figure 3a). However, no mutations were detected in the cueAR locus. This result is surprising because the cueAR operon is known to encode the primary Cu⁺ efflux system in P. fluorescence SBW25 (Zhang & Rainey, 2007c, 2008). CueA is a copper-transporting P1-type ATPase whose expression is induced by copper under the control of a MerR-type transcriptional activator CueR. In addition to cueA, CueR regulates the expression of many other genes involved in copper homeostasis. These include cueZ (pflu0660) encoding a copper chaperone of 65 amino acids in length. Only one cueZ mutation (C56T) was detected in one of the eight Cu-evolved lines, and this mutation was not fixed in the population 320D6 (Data S1). However, mutations were detected in the chromosome-encoded CopRS two-component system in seven out of the eight Cu-evolved populations (pflu1575 and flu1576). Interestingly, deletion of copS gene resulted in an increased resistance to Cu⁺ (Zhang & Rainey, 2008). Thus, the new evolution data support the previous conclusion that the CopRS-regulated genes (specifically, copC and copD) act as a Cu⁺ uptake system, and their loss of function would benefit bacterial survival in high Cu environments.

The Cu-evolved populations have severe growth defects in LB broth medium. This strongly suggests that selection was dominated by adaptation to copper stress rather than growth enhancement in the laboratory medium. Many mutations occurred in catabolic genes (31.75%, 442/1392) and their regulators (Table S1). These include two two-component systems CbrAB and NtrBC, which play global regulatory roles in cellular carbon and nitrogen metabolisms (Naren & Zhang, 2021). Both CbrAB and NtrBC are involved in the regulation of bacterial growth in biofilms for Pseudomonas (Alford et al., 2020; Amador et al., 2016). The CbrA kinase is functionally linked to antibiotic resistance. Inactivation of cbrA gene in P. aeruginosa PA14 caused an increase in tobramycin minimal inhibitory concentration (MIC) from 2 μg/ml for wild type to 8 μg/ml for the ΔcbrA mutant (Yeung et al., 2011). These mutations thus partially explain the observed phenotypes with regard to a decrease in biofilm formation but an increased resistance to tobramycin.

It has been increasingly recognized that copper contamination in soil results in the enrichment of bacterial genes associated with multiple heavy metal resistance and antibiotic resistance. For example, Chen et al. (2019) examined bacterial communities in a copper tailing dam area in Northern China, and their results of network analysis revealed co-selection of copB, a copper resistance gene, and antibiotic resistance genes. In this work, we detected two mutations in the CzcS sensor kinase gene, which is involved in the regulation of bacterial resistance to heavy metals (zinc, cadmium, and cobalt) and to the carbapenem antibiotic in P. aeruginosa (Perron et al., 2004). Copper homeostasis determines resistance to gold and silver (Graves et al., 2015; Zhang & Rainey, 2008), and it is also linked to the cellular iron metabolism. Pyoverdine is known as an iron-chelating siderophore that shows high binding affinities with copper ions. Thus, an enhanced production of pyoverdine would benefit bacteria to cope with the environmental stress provoked by metallic copper. However, pyoverdine non-producing mutants were surprisingly selected in Cu-evolved lines. These results suggest the complex interactions among traits associated with heavy metal and antibiotic resistance (Rainey et al., 2014).

All eight replicate populations in the Cu evolution line carried a mutation in mutL gene, which encodes a DNA mismatch repair protein. Inactivation of mutL is known to produce a mutator phenotype with a higher spontaneous mutant rate (Matic, 2019). The specific mutL mutation involved one guanine deletion at a position that contains five guanine residues (1174_G5 to 1174_G4). Thus, this is likely a mutational hotspot. The same mutation was detected but not fixed in the 200th populations (200D1–200D8). However, among the 320th populations, this particular mutL mutation was fixed in three out of the eight populations (320D1, 320D6, and 320D7). An additional new mutL mutation (T875C) was found in population 320D5. This finding suggests that higher mutation rate is advantageous for bacterial adaptation to Cu stress. Finally, we noted that the same mutations were found in multiple replicate lines, which cannot be explained simply by parallel evolution due to their high frequencies of occurrence. In this work, we used a lacZ marker to detect potential contamination by other bacteria from the environments. However, strains in each replicate line were not individually barcoded; and consequently, we are unable to exclude the possibility of crossline contaminations. This may affect the absolute independence of the eight replicates within a treatment, but not the significant differences between Cu-treated and Cu-untreated control lines.

In conclusion, the experimental evolution data presented here show that bacteria were capable of evolving prolonged survival on the surfaces of metallic copper. A significantly higher number of mutations (540 on average) were accumulated in the final copper-evolved lines relative to an average of 71 mutations found in the control lines. The resulting resistant strains showed increased resistance to Cu⁺ ions, but no mutations were detected in the previously identified Cu⁺-efflux system CueA and its regulator CueR. Thus, Pseudomonas must possess some yet-unknown mechanisms that act to maintain cellular copper homeostasis. Genes in the CopRS regulon are the likely candidates for further characterization. With regard to phenotypic changes, the Cu-evolved strains showed decreased abilities to grow in laboratory media, biofilm formation, pyoverdine production, and exoprotein secretion; but more importantly, they have evolved increased resistance to tobramycin. Together, our data suggest that the long-term evolutionary impacts should be considered if copper were to become commonly used in the hospital environments.

AUTHOR CONTRIBUTIONS

Feng Xu: Data curation (lead); funding acquisition (supporting); investigation (supporting); methodology (supporting). Sha Liu: Conceptualization (supporting); data curation (supporting); formal analysis (supporting); investigation (supporting); methodology (lead); writing – review and editing (supporting). Naran Naren: Data curation (supporting); investigation (supporting); methodology (supporting). Lily Li: Data curation (supporting); investigation (supporting); methodology (supporting). Luyan Z. Ma: Investigation (supporting); writing – review and editing (supporting). Xue-Xian Zhang: Conceptualization (lead); data curation (supporting); formal analysis (lead); funding acquisition (lead); investigation (lead); project administration (lead); resources (lead); writing – original draft (lead); writing – review and editing (lead).

ACKNOWLEDGMENTS

We thank Stephen Ritchie and Tim Cooper for helpful discussions. This work was financially supported by the Maurice and Phyllis Paykel Trust (MPPT) and the New Zealand MBIE Catalyst Fund (project number 92846082). F. Xu acknowledges support of an international visitor's scholarship from Nanchang University.

CONFLICT OF INTEREST

The authors declare no conflict of interest.

DATA AVAILABILITY STATEMENT

DNA sequences are available in GenBank with a BioProject accession number PRJNA838214.