Development of Whole-Cell Biosensors for Screening of Peptidoglycan-Targeting Antibiotics in a Gram-Negative Bacterium

ABSTRACT

There is an urgent need to develop novel antibiotics since antibiotic resistance is an increasingly serious threat to global public health. Whole-cell biosensors are one of the promising strategies for new antibiotic discovery. The peptidoglycan (PG) of the bacterial cell wall is one of the most important targets for antibiotics. However, the biosensors for the detection of PG-targeting antibiotics in Gram-negative bacteria have not been developed, mainly because of the lack of the regulatory systems that sense and respond to PG stress. Recently, we identified a novel two-component signal transduction system (PghKR) that is responsible for sensing and responding to PG damage in the Gram-negative bacterium Shewanella oneidensis. Based on this system, we developed biosensors for the detection of PG-targeting antibiotics. Using ampicillin as an inducer for PG stress and the bacterial luciferase LuxCDABE as the reporter, we found that the PghKR biosensors are specific to antibiotics targeting PG synthesis, including β-lactams, vancomycin, and d-cycloserine. Deletion of genes encoding PG permease AmpG and β-lactamase BlaA improves the sensitivity of the biosensors substantially. The PghKR biosensor in the background of ΔblaA is also functional on agar plates, providing a simple method for screening bacteria that produce PG-targeting antibiotics.

IMPORTANCE The growing problem of antibiotic resistance in Gram-negative bacteria urgently needs new strategies so that researchers can develop novel antibiotics. Microbial whole-cell biosensors are capable of sensing various stimuli with a quantifiable output and show tremendous potential for the discovery of novel antibiotics. As the Achilles’ heel of bacteria, the synthesis of the peptidoglycan (PG) is targeted by many antibiotics. However, the regulatory systems that sense and respond to PG-targeting stress in Gram-negative bacteria are reported rarely, restricting the development of biosensors for the detection of PG-targeting antibiotics. In this study, we developed a highly sensitive and specific biosensor based on a novel two-component system in the Gram-negative bacterium Shewanella oneidensis that is responsible for the sensing and responding to PG stress. Our biosensors have great potential for discovering novel antibiotics and determining the mode of action of antibiotics.

INTRODUCTION

Antibiotics are used widely for the treatment of infectious diseases. However, the irrational use and abuse of antibiotics accelerate the emergence and spread of antibiotic resistance, especially for Gram-negative bacteria (1, 2). In the priority list of antibiotic-resistant bacteria published by the World Health Organization, 8 out of 12 bacteria or bacterial families are Gram-negative bacteria (3). Thus, there is an urgent need to develop novel antibiotics against Gram-negative bacteria using new strategies.

Many antibiotics specifically target the synthesis of peptidoglycan (PG), which is the major component of the bacterial cell wall and is essential for the maintenance of cell viability and morphology (4–6). For example, β-lactams (e.g., ampicillin) covalently bind to the active site of PG synthases, and the glycopeptides (e.g., vancomycin) bind to the d-Ala-d-Ala of PG precursors, with both leading to the prevention of the formation of PG cross-links. d-cycloserine inhibits the d-Ala ligase and alanine racemase, which are two essential enzymes involved in the synthesis of PG precursors. Upon exposure to these antibiotics, the synthesis of PG is inhibited, resulting in PG damage and then cell lysis or death (7). Recently, two novel antibiotics (corbomycin and complestatin) were found to inhibit PG hydrolysis rather than PG synthesis (8), indicating that there are more PG-targeting antibiotics than we might think.

Whole-cell biosensors are genetically engineering microorganisms that have tremendous potential for the discovery of novel antibiotics (9–11). They sense and respond to specific input with a quantifiable output, such as fluorescence or luminescence signal (12). A great number of biosensors have been developed widely for the detection of environmental contaminants (13–15) and the monitoring of intermediate metabolites (16, 17). Both regulatory proteins and RNA are commonly used for the development of biosensors, mainly including transcriptional regulators, two-component systems (TCSs), and riboswitches (18). Among them, TCSs are one of the most important signal transduction regulatory systems that sense and respond to changes in the external environment and the homeostasis of the intracellular environment (19–21). The prototypical TCSs consist of a histidine kinase (HK) and a response regulator (RR). The HK detects a specific signal and activates its kinase activity, causing autophosphorylation of a conserved histidine residue. The phosphoryl group is then transferred to an aspartate residue of the cognate RR, resulting in the phosphorylation of RR. Finally, the phosphorylated RR binds to the promoter region of downstream genes, thus modulating gene expression in response to a variety of stimuli.

Based on the regulatory systems for sensing and responding to antibiotics, several biosensors have been developed. Some biosensors specifically detect one class of antibiotics, including β-lactams (9, 22), tetracyclines (23, 24), macrolides (25–28), and amphotericin-like polyenes (29). There are still some biosensors specific to one of the major biosynthetic pathways of bacteria, such as cell wall synthesis and cell envelope synthesis (30–32), which are more suitable for the initial screening of antimicrobial compounds. However, the biosensor specific to PG stress has not been developed in Gram-negative bacteria since the regulatory systems for sensing and responding to PG stress are rarely reported in these microorganisms.

Recently, we identified a novel TCS that is responsible for sensing and responding to PG damage in the Gram-negative bacterium Shewanella oneidensis (J. Yin, H. Gao, and Z. Yu, submitted for publication). This TCS consists of the histidine kinase PghK (GenBank identifier [ID] AAN55234.2) and the response regulator PghR (GenBank ID AAN55235.1). Upon exposure to PG-targeting antibiotics (e.g., ampicillin), PG damage is induced due to the unbalance between PG synthesis and hydrolysis, leading to the activation of PghKR. Subsequently, PghR binds to the promoter region of its target genes and then regulates gene expression (Fig. 1A). The blaA gene is one of the target genes of PghR, which encodes the β-lactamase that hydrolyzes β-lactam antibiotics, resulting in β-lactam resistance (33–36).

FIG 1.

Construction of the whole-cell biosensors for the detection of PG-targeting antibiotics. (A) Schematic illustration of the biosensors based on the PghKR TCS. H508 and D53 represent the conserved phosphorylation sites of PghK and PghR, respectively. OM, outer membrane; CM, cytoplasmic membrane. (B) Effects of reporter genes (sfgfp and luxCDABE) on the output signals of biosensors in the presence of 50 and 100 μg/mL ampicillin. The RFU and RLU values over OD₆₀₀ (normalized RFU and RLU, respectively) of the biosensor strains WT/P_blaA-sfgfp and WT/P_blaA-lux were calculated after treated for 3 h and 1 h, respectively. Data are shown as means ± standard deviations. Asterisks indicate statistically significant differences (**, P < 0.01; and ***, P < 0.001).

In this study, we developed whole-cell biosensors based on the PghKR TCS in S. oneidensis. The biosensors are specific to antibiotics that target PG synthesis. By changing the genetic background strains, PghKR biosensors display high sensitivity in response to antibiotics, which is capable of detecting PG-targeting antibiotics not only in liquid cultures but also on agar plates. In addition, the PghKR biosensors were used for screening bacteria producing PG-targeting antibiotics from environmental samples, implying that our biosensors have great potential for discovering novel antibiotics.

RESULTS

Construction of whole-cell biosensors based on the PG damage sensing and response system.

Given that the PghKR TCS in S. oneidensis is capable of sensing and responding to damage to PG, we supposed that this regulatory system could be employed to develop biosensors for screening antibiotics targeting PG synthesis and hydrolysis. To this end, the promoter of blaA (P_blaA) was placed in front of the reporter gene luxCDABE or sfgfp, encoding the bacterial luciferase and the superfolder green fluorescent protein (sfGFP), respectively. The resulting plasmids were conjugated into the wild-type (WT) of S. oneidensis MR-1, yielding biosensors WT/P_blaA-lux and WT/P_blaA-sfgfp (Fig. 1A). Using ampicillin as a representative for PG-targeting antibiotics, we determined the relative fluorescence units (RFUs) or relative luminescence units (RLUs) of the biosensors upon exposure. As shown in the Fig. 1B, both normalized RFU and RLU values were significantly increased in a concentration-dependent manner. However, the output signals of the WT/P_blaA-lux were remarkably higher than those of the WT/P_blaA-sfgfp. Compared with the untreated control groups, WT/P_blaA-lux displayed ~15-fold and ~90-fold increases in output signals when treated with ampicillin at 50 μg/mL and 100 μg/mL, respectively, while WT/P_blaA-sfgfp had only ~3-fold and ~6-fold increases in output signals (Fig. 1B). Therefore, the reporter gene luxCDABE is more suitable than sfgfp for the development of biosensors detecting PG-targeting antibiotics.

Factors affecting the response of the biosensor.

Given that antibiotics have a detrimental effect on bacterial cell growth, the initial cell density of the biosensor strain should affect the response to antibiotics. Thus, we measured the output signals of the biosensor WT/P_blaA-lux with different initial cell densities (optical density at 600 nm [OD₆₀₀] of 0.05, 0.1, and 0.2). Results showed that the growth of the biosensor strain was not affected by 50 μg/mL ampicillin after incubation for 1 h, but severe growth defects were found after incubation for 3 h, which showed an inverse correlation with the initial cell densities (Fig. 2A; see Fig. S1 in the supplemental material). In the presence of ampicillin, the biosensor with initial OD₆₀₀ of 0.05 and 0.1 had significant increases in output signal after incubation for 1 h, while the output signal was barely changed at an initial OD₆₀₀ of 0.2 (Fig. 2B). These results suggest that the biosensor is capable of sensing ampicillin at relatively low cell densities. However, the biosensor with an extremely low initial cell density (e.g., OD₆₀₀ of 0.05) displayed an increased basal fluorescence value, leading to a reduction of the dynamic range of the biosensor (Fig. 2B). Taken together, we can conclude that an initial OD₆₀₀ of 0.1 is appropriate for the measurement of the response of the biosensor.

FIG 2.

Factors affecting the response of the biosensor. (A and B) Effects of initial cell densities (OD₆₀₀ of 0.05, 0.1, and 0.2) on cell growth (A) and output signals (B) of the biosensor with or without 50 μg/mL ampicillin after treatment for 1 h. (C) The time-dependent response of the biosensor with or without 50 μg/mL ampicillin. (D) The dose-dependent response of the biosensor in the presence of ampicillin at different concentrations (0, 0.4, 3.2, 12.8, 25.6, 51.2, and 102.4 μg/mL) after treatment for 1 h. Data are shown as means ± standard deviations. Asterisks indicate statistically significant differences (ns, not significant; *, P < 0.05; ***, P < 0.001; and ****, P < 0.0001).

Then, the time- and dose-dependent response of the biosensor was measured. Upon exposure to 50 μg/mL ampicillin, the output signals of the biosensor were significantly increased during the first 2 h and then fell rapidly after exposure for 3 h (Fig. 2C). These results are in agreement with the fact that the bacterial luciferase can serve as a real-time reporter (37). Combined with the data for initial cell densities, the luminescence signals of the biosensor were measured after 1 h of treatment in the following studies. The dose-response study for the biosensor showed that the output signals were positively correlated with the concentration of ampicillin, and the detection limit for ampicillin is about 25.6 μg/mL (Fig. 2D).

Specificity of the biosensor.

To explore the specificity of the biosensor WT/P_blaA-lux, the output signals of this biosensor were determined in response to other PG-targeting antibiotics, including β-lactams (penicillin G and carbenicillin), vancomycin, and d-cycloserine. These antibiotics inhibit PG synthesis at different steps (5, 6). In the presence of these antibiotics, the output signals of the biosensor were increased in a manner dependent on concentration, which resembles the effects of ampicillin (Fig. 3). We also determined the response of the biosensor after exposure to antibiotics targeting protein synthesis (gentamicin and kanamycin) and the cell membrane (polymyxin B). Results showed that none of them was sufficient to induce a significant increase in the output signals of the biosensor (Fig. 3). Therefore, the response of the biosensor based on the PghKR TCS is specific to antibiotics targeting bacterial PG synthesis.

FIG 3.

The biosensor is specific to PG-targeting antibiotics. The normalized RLU values were determined in response to various PG-targeting antibiotics at concentrations related to MICs, including carbenicillin (MIC, 16 μg/mL), penicillin G (MIC, 8 μg/mL), ampicillin (MIC, 4 μg/mL), vancomycin (MIC, 64 μg/mL), d-cycloserine (MIC, 9.76 μg/mL), gentamicin (MIC, 2 μg/mL), kanamycin (MIC, 4 μg/mL), and polymyxin B (MIC, 0.125 μg/mL). The biosensor was treated with antibiotics at different concentrations for 1 h, and then the normalized RLU values were determined. Data are shown as means ± standard deviations. Statistical significance was shown as ns, not significant; *, P < 0.05; **, P < 0.01; and ***, P < 0.001.

The expression level of PghK modulates the response of the biosensor.

The HK of TCS is the sensor responsible for signal perception, and therefore, its expression level could affect the response of the biosensor. To test this idea, the pghK gene was expressed under the control of three synthetic promoters, including the strong promoter J23106, the medium-strength promoter J23114, and the weak promoter J23113. Meanwhile, the expression of pghK by the leaky strong promoter P_tac was used as a positive control, which produces a functional TCS sensor in the absence of an inducer (Fig. 4A). While strains expressing pghK driven by P_J23113 and P_J23114 grew well in the presence of ampicillin, the strain harboring pghK controlled by the strong promoter P_J23106 had a moderate growth defect in liquid culture containing ampicillin (Fig. 4A), which may have resulted from the decreased expression of β-lactamase. Consistently, the expression of pghK controlled by P_J23106 resulted in a significant decrease in output signal upon exposure to 128 μg/mL ampicillin compared with that controlled by P_J23114 and P_tac (Fig. 4B). Moreover, the response of PghKR was slightly reduced in the strain harboring P_J23113 for PghK expression. These data suggest that strains expressing a very high and very low level of PghK have negative effects on the output signals.

FIG 4.

Effects of the expression levels of PghK on the function of PghKR TCS (A) and the response of the biosensor (B). The expression of pghK was under the control of promoters with different strengths, including the leaky strong promoter Ptac, the weak promoter J23113, the medium-strength promoter J23114, and the low strong promoter J23106. The function of PghKR was evaluated by the growth of the bacterial cells (OD₆₀₀) with 50 μg/mL ampicillin for 6 h (A). (B) The normalized RLU values of the biosensor strains harboring different levels of PghK were determined in response to ampicillin at different concentrations (0, 16, 64, and 128 μg/mL) after treatment for 1 h. Data are shown as means ± standard deviations. Statistical significance was shown as ns, not significant difference; *, P < 0.05; and ***, P < 0.001.

Loss of PG permease AmpG and β-lactamase BlaA remarkably improve the sensitivity of the biosensor.

The biosensor constructed in the background of the WT strain had poor sensitivity in response to antibiotics. To improve the sensitivity of the biosensor, two strategies were employed. On the one hand, the promoter activity of blaA has been shown to be more sensitive to ampicillin in the strain lacking AmpG (33), a permease involved in the transportation of PG fragments. We supposed that deletion of ampG would increase the sensitivity of the biosensor. On the other hand, the WT strain of S. oneidensis is naturally resistant to β-lactams (e.g., ampicillin) due to the production of β-lactamase BlaA (33–36). We hypothesized that lowering the resistance to antibiotics should increase PG stress induced by the same concentration of antibiotics, thus increasing the sensitivity of the biosensor. Although deletion of blaA from S. oneidensis WT abolished the ability to grow with ampicillin on the solid LB plate, the deletion mutant can grow well in liquid culture containing ampicillin within incubation for 4 h. After that step, cell lysis occurred, accompanied by a decrease in OD₆₀₀ (34). Given that the luminescence values of the biosensor were measured after incubation for 1 h (Fig. 2C), the strain lacking blaA should be a good candidate as the background strain for the biosensor. To test these hypotheses, the genetic circuit harboring P_blaA and luxCDABE was introduced into strains lacking ampG or blaA, resulting in ΔampG/P_blaA-lux or ΔblaA/P_blaA-lux, respectively.

The dose-response curves of these biosensors were determined after exposure to ampicillin, and then the data were fitted to the Hill equation (Fig. 5A and Table 1). The response of the biosensor was remarkably increased in strains lacking ampG or blaA compared with that in the WT. The minimal detection limit of the biosensor in the ΔampG and ΔblaA backgrounds was 1.6 and 0.2 μg/mL, respectively, which were 16- and 128-fold lower than that in the WT. Deletion of ampG or blaA also increased the dynamic range from 158-fold to ~240-fold. Moreover, we also measured the output signals of these biosensors in response to vancomycin and d-cycloserine (Fig. 5B and C). While deletion of blaA did not affect the luminescence values, deletion of ampG resulted in significant increases in biosensor sensitivity, especially for vancomycin (Fig. 5B). Collectively, these data suggest that deletion of ampG and blaA is capable of improving the sensitivity of the PghKR biosensor.

FIG 5.

Effects of ampG and blaA deletion on the response of the biosensors. (A) Dose-response curve of the biosensors. The normalized RLU values were determined in response to ampicillin at different concentrations (0, 0.1, 0.2, 0.4, 3.2, 6.4, 12.8, 25.6, 51.2, 102.4, and 204.8 μg/mL). The data were fitted to the Hill equation. (B and C) The normalized RLU values of the biosensors in response to vancomycin (B) and d-cycloserine (C) at concentrations indicated. Data are shown as means ± standard deviations. Statistical significance was shown as ns, not significant difference; *, P < 0.05; **, P < 0.01; and ***, P < 0.001.

TABLE 1.

Performance features of the PghKR biosensors in different background strains

Biosensor	K1/2 (μg/mL)a	nb	R2	Dynamic rangec	Minimal detection limit (μg/mL)d
WT/PblaA-lux	164	3.739	0.9990	158	25.6
ΔampG/PblaA-lux	348.1	0.8815	0.9911	231.8	1.6
ΔblaA/PblaA-lux	32.89	1.927	0.9900	240	0.2

^aK_1/2 is the concentration of ampicillin that gives rise to half-maximal biosensor activation.

^bn is the Hill coefficient.

^cDynamic range = (MAX − MIN)/MIN.

^dMinimal detection limit refers to the lowest concentration with a significant difference.

Mutation of the second GXGXG residue in PghK does not further improve the detection threshold of the biosensor.

It has been reported that mutation of the second position of the GXGXG motif within the catalytic ATP-binding (CA) domain of HK (such as Shewanella TtrS and ThsS) is capable of tuning the detection threshold of TCSs (18, 38). Multiple sequence alignment analysis revealed that the GXGXG motif within the CA domain of PghK (HLGLG) is quite similar to that of TtrS and ThsS (GLGLG), with only one different residue at the first position (Fig. 6A and B). To explore whether this method can be used to improve the detection threshold of PghKR, we mutated the second position of the HLGLG motif within PghK (L681) using maturation mutagenesis in the background of ΔampG. Then, the native and 19 mutant TCSs were subjected an ampicillin susceptibility assay and output signal determination (Fig. 6C; see Fig. S2 in the supplemental material). Among them, 10 mutants failed to respond to ampicillin and cannot grow on LB plates supplemented with ampicillin (Fig. S2), suggesting that these mutations abolish the function of PghKR. In contrast, substituting V, I, F, P, G, and C at L681 increased the basal luminescence values to levels that are comparable to the output signals after exposure to ampicillin (Fig. 6C and D). Replacing L681 with A and W resulted in functional TCS sensors, but both showed a weak response to ampicillin. The L681M mutation had minor effects on the function of PghKR and the response to ampicillin (Fig. 6C and D). In conclusion, these data suggest that the phosphatase tuning method reported previously does not further improve the detection threshold of the PghKR biosensor in the background of ΔampG.

FIG 6.

Effects of the phosphatase tuning method on the response of the biosensor. (A) Schematic diagram of the structure of PghK. The GXGXG motif is within the CA domain in the cytoplasm, and Leu681 is the second position of the GXGXG motif. (B) A sequence logo of the GXGXG motif of PghK, Shewanella TtrS, and Shewanella ThsS. (C and D) The output signals (C) and the fold changes (D) of the biosensor with native PghK or PghK variants as indicated upon exposure to 50 μg/mL ampicillin for 1 h. Data are shown as means ± standard deviations.

Agar plate-based bioassay for the biosensor.

To explore whether the PghKR biosensors are functional on agar plates, we determined the output signals of the biosensors using a simple spot-on lawn assay, where antibiotics or antibiotic producers are spotted onto a lawn of the biosensor strain (39). First, the output signals of the biosensors in different backgrounds (WT, ΔampG, and ΔblaA) were compared (see Fig. S3 in the supplemental material). In the presence of ampicillin, only the biosensor in the ΔblaA background displayed a ring-shaped luminescence signal, suggesting that this biosensor is responsive to ampicillin on agar plates. The dose-response study demonstrated that the luminescence signals of this biosensor on agar plates were also positively correlated with the concentrations of ampicillin, and the detection threshold under this condition is 10 μg/mL (Fig. 7A). Next, we investigated the specificity of this biosensor on agar plates (Fig. 7B). In agreement with the results from liquid cultures, the ring-shaped luminescence signals were observed with the addition of PG-targeting antibiotics (ampicillin, penicillin G, carbenicillin, vancomycin, and d-cycloserine) but not other classes of antibiotics (kanamycin, gentamicin, and polymyxin) (Fig. 7B). Based on these data, we can conclude that the biosensor in the ΔblaA background is capable of detecting PG-targeting antibiotics not only in liquid cultures but also on agar plates.

FIG 7.

The response of the biosensor in the background of ΔblaA determined by the spot-on-lawn method. The agar plate is covered by a layer of soft agar (0.75% agar) containing the ΔblaA biosensor. A total of 3 μL of antibiotics (A and B) or bacterial cultures (C) was spotted onto the double-layered plate. After 12 h of culture at 30°C, the white light image (top) and the luminescence image (bottom) of the same agar plate were captured. (A) The response to ampicillin at different concentrations (0, 0.625, 1.25, 2.5, 5, 10, 25, 50, and 100 μg/mL). ddH₂O, double-distilled water. (B) The response to different antibiotics, including ampicillin (AMP, 10 μg/mL), carbenicillin (CAR, 10 μg/mL), penicillin G (PEN, 200 μg/mL), vancomycin (VAN, 2.5 mg/mL), d-cycloserine (DCS, 2.5 mg/mL), gentamicin (GM, 50 μg/mL), kanamycin (KM, 100 μg/mL), and polymyxin B (PMB, 2.5 μg/mL). (C) Detection of bacteria producing PG-targeting antibiotics. Two Bacillus strains (B8 and B9) isolated from soil resulted in a ring-shaped luminescence signal of the biosensor surrounding the colony (bottom). B. subtilis 168 was used as a negative control. The images are representative.

Biosensor-guided screening of bacteria producing PG-targeting antibiotics.

Finally, we used the biosensor ΔblaA/P_blaA-lux to screen bacteria that produce PG-targeting antibiotics from environmental samples. To this end, bacterial strains were isolated from soil and then subjected to the spot-on lawn assay. If PG-targeting antibiotics are produced by bacterial cells, they will induce a significant increase in the output signals of the biosensor, resulting in a ring-shaped luminescence signal of the biosensor surrounding the producer colony. In total, seven strains (e.g., B8 and B9) were found to form weak but evident ring-shaped luminescence signals surrounding the bacterial colonies (Fig. 7C; see Fig. S4 in the supplemental material), indicating that these strains may produce antibiotics that target PG synthesis. In comparison, the negative-control strain Bacillus subtilis 168 does not produce any antibiotics that induce the PghKR biosensor, so that the ring-shaped luminescence signal that surrounds the bacterial colony was not observed (Fig. 7C). To identify these strains, the 16S rRNA gene sequences were amplified by PCR using universal primers 27-F and 1492-R. After sequencing and the phylogenetic analysis, we found that four out of seven strains belong to the Bacillus genus (see Fig. S5A in the supplemental material), which has been recognized widely as a source of antimicrobial compounds (40). Since the 16S rRNA gene sequences share high similarities among Bacillus species (41, 42), the rpoB gene sequence of the four Bacillus strains was also analyzed. Results showed that three strains were identified as Bacillus velezensis, and one strain was identified as Bacillus amyloliquefaciens (Fig. S5A).

DISCUSSION

Whole-cell biosensors have been developed widely for monitoring environmental pollutants and intracellular metabolites (13–17). They also have great potential for antimicrobial compound discovery (11). At present, several biosensors have been developed in Gram-positive bacteria to detect either one class of antibiotics (9, 28, 29) or antibiotics targeting one of the major biosynthetic pathways (30–32). However, the biosensors for the detection of PG-targeting antibiotics have not been constructed in Gram-negative bacteria, mainly because they lack the regulatory systems for sensing and responding to these antibiotics. Based on the finding that the PghKR TCS in the Gram-negative bacterium S. oneidensis is capable of sensing and responding to PG stress, we developed the biosensors for the specific detection of antibiotics targeting PG synthesis. To our knowledge, the PghKR biosensor is the only example for this purpose in Gram-negative bacteria.

The expression level of HK or RR is crucial for the response of the TCS-based biosensors. Several previous studies have shown that a very high or very low expression of HKs often degrades TCS responses (18, 38, 43). In agreement with this idea, our results demonstrated that the expression of PghK by the weak promoter or the strong promoter impairs the response of PghKR in a different degree. Like many canonical HKs, PghK is also bifunctional and possesses both kinase and phosphatase activities. Therefore, a low level expression of PghK may decrease the rate at which PghR is phosphorylated, while overexpression of PghK may dephosphorylate the phosphorylated PghR (PghR~P). In both cases, the response of PghKR should be affected.

The PghKR biosensor in the background of WT exhibits low sensitivity to ampicillin. One of the most important reasons is that S. oneidensis is naturally resistant to β-lactams by producing β-lactamase to hydrolyze antibiotics (34). Because the PghK sensor senses PG damage rather than the antibiotics per se, the antibiotic-resistant bacteria typically need a higher concentration of antibiotics than the sensitive bacteria to induce PG damage. Consistent with this information, our results showed that deletion of the β-lactamase gene blaA remarkably increases the sensitivity of the biosensor not only in liquid cultures but also on agar plates. The ΔblaA strain is extremely susceptible to β-lactam antibiotics (34) so that the presence of very low levels of antibiotics is sufficient to trigger PG damage, thereby activating the response of the PghKR biosensor. A study for the β-lactam-specific biosensor in Bacillus subtilis also demonstrated that removal of penP encoding β-lactamase increases the signal intensity in response to β-lactams on agar plates (9). Therefore, changing the resistance of the host organism to antibiotics should be a promising strategy for tuning the sensitivity of biosensors for antibiotics detection.

Another strategy to improve the sensitivity of the PghKR biosensor is blocking PG recycling. In contrast to most canonical sensors that directly interact with their cognate ligands (18), the PghK sensor is unlikely to bind to structurally diverse PG-targeting antibiotics directly. Besides, the majority of PG-targeting antibiotics (such as β-lactams and vancomycin) are difficult to penetrate into the cytoplasm, implying that the signals sensed by PghK exist in the periplasm (36). Although the signals remain to be identified, they are most likely the glycan fragments of PG (36), which is highly produced upon exposure to PG-targeting antibiotics. As AmpG is responsible for the transportation of PG fragments (mainly GlcNAc-1,6-anhydromuropeptides) from the periplasm to cytoplasm (44), deletion of ampG results in the accumulation of sensor signals in the periplasm, thereby reducing the minimal detection limit of the biosensor. Given that PG damage can be induced by various PG-targeting antibiotics, one can imagine that the biosensor in the ΔampG would display high sensitivity to antibiotics more than β-lactams. Consistently, our results showed that the deletion of ampG increases the sensitivity not only to β-lactams but also to vancomycin and d-cycloserine, which is in contrast to the effects of blaA deletion.

In addition, the sensitivity of the PghKR biosensor was also optimized by the phosphatase tuning method that mutated the second position of the GXGXG motif within the CA domain of HKs. The GXGXG motif is present in 64% of HKs and is involved in regulating its phosphatase activity (38, 45). It has been reported that this method is a simple strategy for altering the detection threshold of diverse TCSs, even in the absence of known phosphatase-altering mutations (18, 38). However, our results indicate that the saturation mutagenesis on L681 is insufficient to further improve the sensitivity of the PghKR biosensor in the background of the ΔampG strain. Two reasons may account for the unexpected results. First, although the GXGXG motif of PghK is highly conserved with previously studied HKs (Shewanella TtrS and ThsS) (38), the first amino acid residue is histidine instead of glycine. This difference may affect the roles of the second GXGXG residue in each HK. In our study, most of mutations at L681 in PghK make the TCS inactivated (10 out of 19) or always activated (6 out of 19), leading to the biosensors being unresponsive to antibiotics. These results indicate that L681 is crucial for the function of PghK. Second, the mutation for PghK is performed in the background of ΔampG rather than the WT. We cannot exclude the possibility that some mutations may improve the detection threshold of PghKR in the WT background since the deletion of ampG increases the basal output signal of the biosensor and makes the biosensor more sensitive in response to PG stress. Further characterization is needed to determine the specific reason.

PG-targeting antibiotics are among the most important and widely used antibiotics for the treatment of bacterial infections (4, 5). A recent study based on an evolution-guided method discovered two PG-targeting antibiotics (corbomycin and complestatin) that block the activity of PG hydrolases, which is distinct from other antibiotics affecting PG synthesis (8). In addition, the shape, elongation, division, and sporulation (SEDS) proteins were found to be a widespread family of bacterial PG polymerases, thus providing the exceptional potential as targets for new antibiotics (46–48). These latest findings suggest that many unknown PG-targeting antibiotics remain to be uncovered. In fact, PG is considered the Achilles’ heel of bacteria (6). Therefore, the PghKR biosensors developed in this study can serve as an initial screening platform to discover novel PG-targeting antibiotics.

MATERIALS AND METHODS

Bacterial strains, plasmids, primers, and growth conditions.

All strains and plasmids used in this study are listed in Table 2. All primers used in this study are shown in Table 3. Escherichia coli and S. oneidensis strains were grown in Luria-Bertani (LB) medium at 37°C and 30°C, respectively. When needed, the media were added with chemicals at the following concentrations: 5.7 μg/mL 2,6-diaminopimelic acid (DAP), 10 μg/mL kanamycin (KAN), 25 μg/mL chloramphenicol (CHL), and 10 μg/mL spectinomycin (SPT).

TABLE 2.

Strains and plasmids used in this study

Strain or plasmid	Description	Source or reference
Strains
E. coli
DH5α	Host strain for plasmids	Lab stock
DH5α λπ	Host strain for plasmids	Lab stock
WM3064	Donor strain for conjugation; ΔdapA	W. Metcalf, UIUCa
S. oneidensis
WT	Wild type	Lab stock
ΔampG	ampG deletion strain	33
ΔblaA	blaA deletion strain	34
ΔpghK	pghK deletion strain	This study
ΔampGΔpghK	ampG and pghK deletion strain	This study
Plasmids
pHGM01	Apr, Gmr, Cmr, suicide vector	52
pHGEI01	Integrative lacZ reporter vector	51
pHG101	Promoter-less vector for complementation	50
pHGE-Ptac	Kmr, IPTG-inducible Ptac expression vector	53
pBBR-cre	Specr; expressing Cre recombinase	35
pBBR-lux	Cmr; containing the luciferase gene luxCDABE	49
pBBR-PblaA-lux	Expression of luxCDABE controlled by the PblaA promoter	This study
pHG101-PblaA-sfgfp	Expression of sfgfp controlled by the PblaA promoter	This study
pHGEI01-PblaA-lux	Integrative vector containing luxCDABE controlled by PblaA	This study
pHGE-Ptac-pghK	Expression of sfgfp controlled by the Ptac promoter	This study
pHGE-PJ23113-pghK	Expression of sfgfp controlled by the weak promoter PJ23113	This study
pHGE-PJ23114-pghK	Expression of sfgfp controlled by the medium-strength promoter PJ23114	This study
pHGE-PJ23106-pghK	Expression of sfgfp controlled by the strong promoter PJ23106	This study

^aUIUC, University of Illinois at Urbana-Champaign.

TABLE 3.

Primers used in this study

Primer by use	Sequences (5′–3′)
PblaA-F1	GGTGGTACCTGAATTCAGTCGTAGCGATTGACCACTTCCGCCA
PblaA-R1	ACGCATCTAGTATTTCTCCTCTTTAATATCCCCTTGCTTAATGCAATATG
sfgfp-F1	CATATTGCATTAAGCAAGGGGATATTAAAGAGGAGAAATACTAGATGCGT
sfgfp-F2	GGTGGTACCTGAATTCAGTCGAAAGAGGAGAAATACTAGATGCGT
sfgfp-R1	GGTCGTTAAATAGCCGCTTATGTCATCATTTGTACAGTTCATCCA
Vector-pHG101-F	CATAAGCGGCTATTTAACGACC
Vector-pHG101-R	CGACTGAATTCAGGTACCACC
PblaA-F2	GGATCCGTCGACAAGCTTGGACATCCCCTAGGAGTTGCAC
lux-F1	GGGATCCGTCGACAAGCTTGGATGACTAAAAAAATTTCATTC
lux-R1	CATGGCCTGCCCGGTTATTATCAACTATCAAACGCTTCGGT
Vector-pHGEI01-F	TAATAACCGGGCAGGCCATG
Vector-pHGEI01-R	CCAAGCTTGTCGACGGATCC
J23113-sbrK-F	TTCCGGTAGTCAATAAACTGATGGCTAGCTCAGTCCTAGGGATTATGCTAGCATGACACGTTTTCCTTTCG
J23114-sbrK-F	TTCCGGTAGTCAATAAATTTATGGCTAGCTCAGTCCTAGGTACAATGCTAGCATGACACGTTTTCCTTTCG
J23106-sbrK-F	TTCCGGTAGTCAATAAATTTACGGCTAGCTCAGTCCTAGGTATAGTGCTAGCATGACACGTTTTCCTTTCG
sbrK-term-R	GGTTTACCGGTGTGGTGAATCACCGACAAACAACAGATAA
RBS-F	ACCATGTTAAGTAGAAGATTAGCCGCCAGCCTATGACACGTTTTCCTTTC
J23113+RBS-R	TTCTACTTAACATGGTGCTAGCATAATCCCTAGGAC
J23114+RBS-R	TTCTACTTAACATGGTGCTAGCATTGTACCTAGGACT
J23106+RBS-R	TTCTACTTAACATGGTGCTAGCACTATACCTAGGACT
Vector-pHGE-F	ATTCACCACACCGGTAAACC
Vector-pHGE-R	TTTATTGACTACCGGAAGCAGT
In-frame gene deletion
pghK-5O	GGGGACAAGTTTGTACAAAAAAGCAGGCTTGGCAATGAGATCGACGGTG
pghK-5I	AATAGGCGCTAAGCTGATAGTGGTACCCCAGCCAAGGTAG
pghK-3I	CTATCAGCTTAGCGCCTATTCCAGGAGAAACCCCACCTTG
pghK-3O	GGGGACCACTTTGTACAAGAAAGCTGGGTCTTCACCAATCAGGGGATCGT
Site saturation mutagenesis
L681-F	ACCCAGGAGAAACCCCACNNNGGCCTAGGTTTGTA
L681-R	TTGATTATCCCTCACCGAGACCATTGAGTCGAAGA
Phylogenetic tree construction
27-F	AGAGTTTGATCCTGGCTCAG
1492-R	TACGGCTACCTTGTTACGACTT
rpoB-F	AGGTCAACTAGTTCAGTATGGAC
rpoB-R	AAGAACCGTAACCGGCAACTT

Construction of the biosensors.

For the biosensor in the WT background, a fragment of the blaA promoter (P_blaA) was amplified from the genome of S. oneidensis, digested with restriction endonucleases SpeΙ and BamHΙ, and then cloned into the vector harboring luxCDABE (pBBR-lux) (49) or sfgfp (pHG101) (50), yielding the recombinant plasmid pBBR-P_blaA-lux or pHG101-P_blaA-sfgfp, respectively. The correct recombinant plasmids were maintained in E. coli WM3064 (DAP auxotroph) and subsequently transferred into the WT of S. oneidensis via conjugation, resulting in the biosensors WT/P_blaA-lux and WT/P_blaA-sfgfp. For the biosensors in the ΔampG and ΔblaA backgrounds, the plasmid pBBR-P_blaA-lux was transferred into the ΔampG and ΔblaA of S. oneidensis, respectively.

For the biosensors with PghK mutations, the genetic circuit harboring P_blaA and luxCDABE was integrated into the genome of S. oneidensis. Briefly, the lacZ gene sequence of the integrative plasmid pHGEI01 (51) was replaced by the fragment containing P_blaA and luxCDABE amplified from the plasmid pBBR-P_blaA-lux, resulting in the recombinant plasmid pHGEI01-P_blaA-lux. The resultant plasmid was transferred into relevant S. oneidensis strains by conjugation for integration into the degenerated nrfCD locus. The kanamycin resistance gene was removed as described previously (35).

In-frame deletion mutagenesis.

In-frame deletion strains for S. oneidensis were constructed by the att-based fusion PCR method as described previously (52). Briefly, two fragments flanking the gene of interest were amplified by PCR and then joined by the fusion PCR method. The fusion fragments were introduced into the suicide plasmid pHGM01 by the attB/attP (BP) recombination reaction using the Gateway BP clonase II enzyme mix (Invitrogen) according to the manufacturer’s instruction. The resulting mutagenesis constructs were maintained in E. coli WM3064 (DAP auxotroph) and subsequently transferred into relevant S. oneidensis strains via conjugation. Integration of the mutagenesis constructs into the chromosome was selected by resistance to gentamicin and was sensitive to sucrose and confirmed by PCR. Verified transconjugants were grown in LB without NaCl and plated onto LB agar supplemented with 10% sucrose. Gentamycin-sensitive and sucrose-resistant colonies were screened and examined by PCR for deletion of the target gene. Finally, the deleted mutants were verified by sequencing.

Expression of pghK by synthetic promoters.

The pghK gene was amplified from the genome of S. oneidensis and then cloned into the isopropyl-β-d-thiogalactopyranoside (IPTG)-inducible expression vector pHGE-P_tac (53), resulting in the recombinant plasmid pHGE-P_tac-pghK. The tac promoter (P_tac) was replaced with three synthetic promoters (BBa_J23113, BBa_J23114, and BBa_J23106) from the Anderson promoter library (https://parts.igem.org/Promoters/Catalog/Anderson) using a one-step cloning method according to the manufacturer’s instructions (Vazyme, Nanjing, China). After sequencing verification, the resulting plasmids were extracted and transformed into E. coli WM3064. Then, the resulting plasmids were transferred into the S. oneidensis strain lacking the pghK gene and harboring a genetic circuit harboring P_blaA and luxCDABE in the genome.

Site saturation mutagenesis of PghK^L681.

The recombinant plasmid pHGE-P_tac-pghK was linearized by reverse PCR using the primers containing random bases as listed in Table 3. The PCR products were digested with DpnI and transformed into E. coli DH5α. After sequencing verification, the resulting plasmids were extracted and transformed into E. coli WM3064. Then, the resulting plasmids were transferred into the S. oneidensis strain lacking the pghK gene and harboring a genetic circuit harboring P_blaA and luxCDABE in the genome.

Determination of the RLU and RFU values.

The biosensor strains were grown overnight in LB broth. The overnight cultures were subcultured 1:100 in 50 mL fresh LB broth and grown to an OD₆₀₀ of 0.05, 0.1, or 0.2. An aliquot of 3-mL bacterial culture was mixed with various antibiotics at different concentrations as indicated in the figure legends. After incubation for 1 h or 3 h at 30°C with 200 rpm shaking, the luminescence or fluorescence units were determined, respectively. For the determination of luminescence units, 200-μL cultures were transferred to 96-well microplates (white plate, clear bottom) and then measured by using a SynergyH1 microplate reader (BioTek, Winooski, VT) for the luminescence intensity and the absorbance. The determination of fluorescence units was performed as described previously (32). In brief, 200-μL cultures were transferred to 96-well microplates and 96-well black microplates for measurements of absorbance and fluorescence, respectively. The absorbance was measured at 600 nm. The relative luminescence units and fluorescence units normalized OD₆₀₀ (normalized RLU or RFU) were calculated.

Dose-response analysis of the biosensors.

The biosensor strains were grown overnight in LB broth. Then, the overnight cultures were diluted 1:100 in 50 mL fresh LB broth and grown to an OD₆₀₀ of 0.1. An aliquot of 3-mL bacterial culture was mixed with ampicillin at different concentrations (0, 0.4, 3.2, 12.8, 25.6, 51.2, 102.4, and 204.8 μg/mL). After incubation for 1 h at 30°C with 200 rpm shaking, 200-μL cultures were transferred to 96-well microplates (white plate, clear bottom) for the measurement of luminescence and OD₆₀₀. The normalized RLU values were fitted to the Hill equation using GraphPad software:

$$y=\mathrm{MIN}+\frac{{\mathrm{x}}^{\mathrm{n}}\times \mathrm{MAX}-\mathrm{MIN}}{{\mathrm{x}}^{\mathrm{n}}+{\mathrm{K}}_{1/2}^{\mathrm{n}}}$$

where MIN is normalized RLU value in the absence of antibiotic, MAX is the maximum normalized RLU value in the presence of antibiotic, x is the concentration of the inducer, K_1/2 is the concentration of antibiotic that yielded a half-maximal induction, and n is the Hill coefficient describing biosensor cooperativity.

Plate-based bioassay.

A spot-on-lawn method was employed to determine the response of the biosensors as described previously (32, 39). The overnight cultures of biosensor strains were diluted to an OD₆₀₀ of 0.05 and were mixed with the soft Mueller-Hinton (MH) agar (1:10, vol/vol). Then, the mixture was poured onto MH agar plates and swirled gently. After the plates solidified, 3 μL of various concentrations of ampicillin (0, 0.625, 1.25, 2.5, 5, 10, 25, 50, and 100 μg/mL), penicillin G (200 μg/mL), carbenicillin (10 μg/mL), vancomycin (2.5 mg/mL), d-cycloserine (2.5 mg/mL), gentamicin (50 μg/mL), kanamycin (100 μg/mL), and polymyxin B (2.5 μg/mL) were spotted onto the double-layered agar plates. After incubation for 12 h at 30°C, the ring-shaped luminescence signals of the biosensor strains were recorded by an iBright 1500 imaging system (Thermo Fisher Scientific).

Screening for bacteria producing PG-targeting antibiotics from soil.

Bacterial strains were isolated from soil samples obtained from different sources. Cells from each colony were inoculated in 5 mL LB broth and grown for 48 h. A total of 3 μL of each culture was spotted onto the double-layered plates containing the biosensor in the ΔblaA background. After 12 h of culture at 30°C, the luminescence of the agar plates was captured on an iBright 1500 imaging system (Thermo Fisher Scientific).

Bioinformatics analyses.

The sequences of TtrS (Sbal195_3859; GenBank ID ABX51019.1) from Shewanella baltica OS195, ThrS (Shal_3128; GenBank ID ABZ77675.1) from Shewanella halifaxensis HAW-EB4, and PhgK (SO_2192; GenBank ID AAN55234.2) from S. oneidensis MR-1 were downloaded from the NCBI database (https://www.ncbi.nlm.nih.gov). Sequence alignments were performed by MEGA X software, and a sequence logo was made by WebLogo3 (http://weblogo.threeplusone.com/create.cgi).

Statistical analyses.

All experiments were performed in biological triplicates. Experimental values are presented as the mean ± standard deviations. The Student’s t test was used to compare statistical differences between the groups of experiment data (ns, no significant difference; *, P < 0.05; **, P < 0.01; ***, P < 0.001; and ****, P < 0.0001).