All heavy metal(loid)s used (e.g., AsCl₃, CdCl₂, CrSO₄, HgCl₂, PbSO₄, MnCl₂, AuCl, and ZnCl₂) were purchased from Sigma-Aldrich (St. Louis, MO, USA). Novagen pCDF-Duet and pET-21(a) vectors (Sigma-Aldrich) were used to carry the sensing domain, zntR and mutants, and the reporter domain, zntA promoter fused with egfp, respectively (Yoon et al., 2016). PfuTurbo DNA polymerase (Agilent, Santa Clara, CA, USA) was used for site-directed mutagenesis of zntR, and all restriction enzymes and T4 DNA ligase were purchased from Takara Korea Biomedical (Seoul, Korea). E. coli DH5α and BL21(DE3) strains were used for plasmid constructions and as host cells of the biosensors, respectively. E. coli strains, BL21-copA/zntR and BL21-zntA/zntR used in previous studies (Kang et al., 2018a; Kim et al., 2020b) and BL21-copA/zntA/zntR, newly generated in this study, were used as host cells for the biosensors. Endogenous gene deletion was performed using a Quick and Easy E. coli Gene Deletion Kit (Gene Bridges, Heidelberg, Germany). All strains used in the study are listed in Online Resource 1 with a description of their genetic properties.

Triple-gene-deficient E. coli was generated by deleting copA from BL21-zntA/zntR using a Quick & Easy E. coli Gene Deletion Kit. Briefly, the FRT-flanked PGK-gb2-neo cassette and copA sequences were amplified by a polymerase chain reaction and electroporated into E. coli BL21-zntA/zntR cells harboring the pRedET plasmid, which has the recombinase-encoding gene. Electroporation was performed with an Electroporator 2510 (Eppendorf, Hamburg, Germany) at 1,350 V, 10 μF, and 600 Ω (Kang et al., 2018b). The cells were then induced by adding 10% of arabinose to replace copA with the kanamycin resistance gene (kan). Gene deletion was further verified by polymerase chain reaction.

Engineered ZntRs were generated by site-directed mutagenesis of the metal-binding loop (MBL) region, consisting of 12 amino acids. The nucleotide sequences for the amino acid residues surrounding the MBL were mutated to encode specific amino acids using designed primers, which was confirmed by DNA sequencing. The mutation targets were selected upon analysis of the three-dimensional structure of the ZntR. The engineered ZntRs used (Table 1) were tested as sensing domains for metal ion detection in E. coli BL21-zntR, BL21-copA/zntR, BL21-zntA/zntR, and BL21-copA/zntA/zntR. The primers for gene cloning and point mutations are listed in Online Resource 2.

Escherichia coli cell-based biosensors were generated by introducing a pair of plasmids carrying zntAp::egfp and engineered zntR in genetically modified E. coli BL21 strains, including E. coli-zntR, E. coli-zntR/copA, E. coli-zntR/zntA, and E. coli-zntR/copA/zntA. The biosensor assay was conducted as previously described with some modifications (Kang et al., 2018b; Yoon et al., 2018). Briefly, the biosensor cells were grown at 37°C in a shaking incubator overnight, and 1 ml of the cultures was then inoculated in 50 ml of fresh Luria broth until it reached an OD₆₀₀ value of 0.10. Next, the cells were placed in test tubes and directly exposed to metal ions. After 2 h of incubation, the E. coli cells were harvested, and the cell pellet was resuspended in Tris buffer (50 mM Tris-HCl, 160 mM KCl, pH 7.4) before measuring fluorescence intensity.

The intensity of the fluorescence signals generated by the biosensors after heavy metal exposure was measured using fluorescence spectroscopy (FC-2; Scinco, Seoul, Korea). The expression of the enhanced green fluorescent protein (eGFP) was measured at 480 and 510 nm for excitation and emission wavelength, respectively. The fluorescence signals induced by heavy metal exposure were represented as the induction coefficient values defined as (fluorescence intensity of metal-exposed biosensors)/(fluorescence intensity of non-exposed biosensors). The detection limit (DL) for biosensor assay was calculated by the following: DL = 3 s/m (s: standard deviation obtained from multiplicated measurements; m: slope of the standard curve).

The crystal X-ray structure of ZntR was obtained from the Protein Data Bank (PDB ID:1Q08) and modified amino acids by PyMol (Schrödinger, New York, NY, USA). Sybyl 7.3 (Tripos, St. Louis, MO, USA) was used for energy minimization processing. In the protein structures, water molecules and phosphate ions were removed, and hydrogens were added. After modification, they were minimized using the Tripos force field. The energy minimization process was terminated at the convergence criteria of the total energy (0.05 kcal/molÅ). Chimera 1.11 software (https://www.cgl.ucsf.edu/chimera/download.html) was used for protein visualization and analysis.

R v4.1.0 (https://cran.r-project.org) and DescTools package were used for the statistical analysis of all data when the minimum difference between the control and treatment groups was statistically significant (Team, 2014; Signorell et al., 2019).

Since biosensors based on zntA-deficient strains showed strong fluorescence signals even without metal exposure, the experimental conditions used in this study for heavy metal selectivity were modified from those previously reported (Yoon et al., 2016; Kang et al., 2018a). The metal selectivity of the biosensors using E. coli-zntR, E. coli-zntR/zntA, and E. coli-zntR/copA/zntA with wild-type (WT) ZntR was tested under optimized experimental conditions. Since further incubation was required before exposition to metal ions, the responses indicated as induction coefficient values and background fluorescent signals were decreased. Various metal ion concentrations (0.5, 1.0, 5.0, and 10 μM) were tested; the induction coefficient values induced by 1 μM metal ions are shown in Figure 1. As reported previously, E. coli WT-based biosensors showed Cd(II)-specific response although the induction coefficient values were decreased (Figure 1A; Yoon et al., 2016). Hence, the optimized experimental conditions were reliable to test the detection potential of the biosensors. Interestingly, Pb(II) responses were enhanced by copA or zntA deletions, whereas Cd(II) responses were slightly decreased (Figures 1B,C). Moreover, the order of strength was inversed as Pb(II) > Cd(II). In turn, the Pb(II) responses were clearly enhanced, and the responding order to heavy metals was changed to Pb(II) > Cd(II) > Hg(II) when both copA and zntA were deleted (Figure 1D). In addition, it was necessary to mention that the responses toward Zn(II) were not significant. Although the znt-operon had been identified as a Zn(II)-responsive genetic system, the regulatory protein, ZntR, prefers Cd(II) and Hg(II) rather than Zn(II). It was consistent with our previous data (Yoon et al., 2018). Thus, deleting zntA in E. coli led to increased Pb(II) responses, although the induction coefficient values varied. The effects of copA deletion on metal responses were less significant than those obtained upon zntA deletion, which suggests that zntA plays more relevant roles in divalent metal homeostasis in E. coli cells. In fact, it was unclear how metal-translocating channels affect the metal selectivity of E. coli-based biosensors. Nevertheless, metal selectivity and sensitivity can be modulated by disrupting metal homeostasis in the host cells.

Since the E. coli biosensors employed the znt operon, the target selectivity and sensitivity were determined by using ZntR, a known regulatory protein of the znt operon (Brocklehurst et al., 1999). Thus, the amino acids in the MBL regions of ZntR were selected as targets for genetic engineering to modulate its metal selectivity, and the engineered ZntRs were introduced into E. coli-zntR/copA/zntA strains with zntAp::egfp to generate the enhanced biosensors (Table 1).

The biosensor cells were exposed to 1 μM of heavy metal ions, including As(III), Cd(II), Cr(II), Hg(II), Pb(II), Mn(II), Au(I), and Zn(II), and the fluorescence signals were measured after 2 h of exposure (Figure 2). Using ZntR WT as regulatory protein produced similar responses to Cd(II) and Pb(II), reaching ~3-fold induction coefficient values, and the same strength to Hg(II) and Au(I) as observed in the E. coli-zntR/copA/zntA strains (Figure 2A). Moreover, the orders of response strength to heavy metals were changed to Pb(II) > Cd(II) > Hg(II) in the systems engineered with mutated ZntR (Cys115Ile, Cys115Ser, His119Arg, and Cys141Ser) (Figure 2). Although it was unclear why the target selectivity of the biosensors shifted due to the point mutations, it was hypothesized that mutations on the metal-binding sites altered the ability of the cell to interact with the metal ions and induce conformational changes to selectively prefer Pb(II) over Cd(II). Since Cys115 and His119 directly mediate the conformational interaction with the metal ions, the selectivity of the engineered ZntRs was modulated by these amino acids. In addition, the Cys141 mutation, which is located in the C-terminal region of ZntR, also shifted the target selectivity from Cd(II) to Pb(II). Although this outcome was unexpected, the Cys141Ser mutation may induce conformational changes that affect metal selectivity. However, the biosensors employing ZntR Thr117 deletion and Cys124Ser lost their capacity to respond to the presence of the metal ions (Figures 2D,F).

Most interestingly, it was noticed that biosensors having ZntR Cys115Ile as a regulatory protein showed extremely strong responses to Pb(II) at induction coefficient values of over 35 (Figure 2B). Although it showed strong responses to Cd(II) and Hg(II) compared with ZntR WT, the response to Pb(II) was significant. In this regard, the E. coli-zntR/copA/zntA biosensors having ZntR C115I were investigated to further elucidate the potential as Pb(II)-specific biosensors and to understand the mechanisms of superior responses to Pb(II).

The biosensors generated with the engineered ZntRs and E. coli-zntR/copA/zntA showed modulated metal selectivity and responses to heavy metals (1 μM) upon 2 h of exposure. To elucidate the sensitivity to each heavy metal tested, the responses of the biosensors exposed to 0–10 μM of Pb(II), Cd(II), and Hg(II) were determined under the same experimental conditions (Figure 3). The biosensors employing ZntR WT, Cys115Ser, His119Arg, and Cys141Ser showed responses to all three tested metal ions in the order of induction coefficient values of Pb(II) ≈ Cd(II) > Hg(II). Noteworthy, the biosensor employing ZntR Cys115Ile showed superior signals toward Pb(II), Cd(II), and Hg(II); however, it had superior specificity toward Pb(II) over the other heavy metals at 0.1 μM and below (Figure 3). The induction coefficient value for Pb(II) was ~8, whereas it was less than 2 for Cd(II) and Hg(II). Thus, E. coli-zntR/copA/zntA biosensors with ZntR Cys115Ile as the regulatory protein may be Pb(II)-specific biosensors within certain concentration ranges.

Since the newly generated biosensor employing ZntR Cys115Ile showed responses toward Pb(II) only at 0.1 μM, the properties of the biosensors were further characterized. The responses of the biosensor to Pb(II) were determined at concentrations below 100 nM (Figure 4). Overall, the biosensor responded in a concentration-dependent manner, with its detection limit being calculated at ~5 nM Pb(II) and statistical significance being observed at 1 nM Pb(II) (Figure 4, inbox). Moreover, the relationship between the induction coefficient values and Pb(II) concentration showed a linear correlation, thereby allowing for quantitative analysis.

Interference of the other heavy metals, such as Cd(II) and Hg(II), on Pb(II) detection was also investigated to elucidate the robustness of the biosensor to serve as a Pb(II) detector. The biosensors were exposed to 10 nM of Pb(II), Cd(II), and Hg(II) alone or in combination (Figure 5). Biosensors exposed to Cd(II) and Hg(II) showed 0.12- and 0.25-fold increased induction coefficient values, respectively, whereas Pb(II) increased the fluorescence signal by 2.2-fold. When exposed to a combination of heavy metal ions, such as Cd(II)/Pb(II) or Hg(II)/Pb(II), the biosensor showed 2.4-fold induction coefficient values, which were comparable to those obtained with Pb(II) alone. Although the signals were slightly increased by Cd(II) and Hg(II) presence, the new biosensor was shown to have superior Pb(II) selectivity and specificity at certain concentration ranges.

As described above, the engineered ZntR granted improved target selectivity and specificity to the biosensors while using E. coli-zntR/copA/zntA as host cells. To answer why and how the specificity and selectivity were modulated, it was necessary to analyze the fluorescence intensity, since the responses were determined as induction coefficient values. Hence, the fluorescence signals from Pb(II)-exposed and Pb(II)-non-exposed biosensors were screened (Figure 6). In the case of Thr117 deletion and Cys124Ser variant, the signals were similar regardless of the presence or absence of the metal ion, resulting in no increase of the induction coefficient value. Nonetheless, the other biosensors showed similar selectivity toward Pb(II), Cd(II), and Hg(II) although the induction coefficient values varied (Figure 2). As shown in Figure 6, the ZntR WT in E. coli-zntR/copA/zntA showed a strong signal even without Pb(II) exposure, which may be explained by the increased level of endogenous metal ions due to the deletion of the metal exporters that consequently turned on the egfp transcription. However, this assumption would not be accepted because the background signals of ZntR Cys15Ile were much lower even if they responded to metal ions better. In addition, the mutations on the MBL, such as His119Arg and Cys115Ser, showed less background and induced signals compared with ZntR WT in E. coli-zntR/copA/zntA. Furthermore, Cys141Ser located at the C-terminal region of ZntR showed no contribution for weaker background signals. Therefore, it was assumed that conformational changes on ZntRs were responsible for the variable strength observed to initiate the transcription of the reporter gene.

To elucidate this assumption, structural analysis was performed by computational three-dimensional modeling and simulation. Overall, the MBL of ZntR C115I was relatively looser compared with that of the WT because of the hydrophobic nature of isoleucine (Figures 7A,B). It was clear when only MBLs were shown as stick representation that three cysteines and one histidine were placed closer to two zinc ions in the WT compared with those amino acids of the mutant C115I (Figures 7C,D). Additionally, the distances between the zinc ions and residues around the MBL, including three cysteines and one histidine, were measured (Table 2). Taken together, the enlargement of the metal-binding site may be the reason why ZntR Cys115Ile showed enhanced Pb(II) specificity. However, it was still unclear why the different mutations led to variable transcription levels. In fact, we expected that the mutations on the metal-binding site could lead to conformational changes of the ZntR DNA-binding domains, but no significant differences were observed in the model structures.