Klebsiella pneumoniae Expressing VIM-1 Metallo-β-Lactamase Is Resensitized to Cefotaxime via Thiol-Mediated Zinc Chelation

Antibiotic-resistant Klebsiella pneumoniae isolates constitute a great clinical challenge. One important resistance mechanism in K. pneumoniae is the metallo-β-lactamases (MBLs), which require zinc for their function. Thus, zinc chelation could be a strategy to resensitize K. pneumoniae to β-lactams. However, the potential role for endogenous zinc chelators for this purpose remains to be explored.

ABSTRACT

Antibiotic-resistant Klebsiella pneumoniae isolates constitute a great clinical challenge. One important resistance mechanism in K. pneumoniae is the metallo-β-lactamases (MBLs), which require zinc for their function. Thus, zinc chelation could be a strategy to resensitize K. pneumoniae to β-lactams. However, the potential role for endogenous zinc chelators for this purpose remains to be explored. The aim was to search for endogenous factors that could resensitize MBL-expressing K. pneumoniae to cefotaxime (CTX). Clinical K. pneumoniae isolates expressing different MBLs were screened for sensitivity to CTX in supernatants from human HT-29 colonic epithelial cells. Factors influencing CTX susceptibility were isolated and identified with chromatographic and biochemical methods. Free zinc was measured with a Zinquin assay, the thiol content was assessed with a fluorometric thiol assay, and the reducing ability of the supernatant was measured with a fluorescent l-cystine probe. Urine samples from healthy volunteers were used to validate findings ex vivo. VIM-1-expressing K. pneumoniae regained susceptibility to CTX when grown in supernatants from HT-29 cells. This effect was mediated via free thiols in the supernatant, including l-cysteine, and could be prevented by inhibiting thioredoxin reductase activity in the supernatant. Free thiols in urine samples appeared to have a similar function in restoring CTX activity against VIM-1-expressing K. pneumoniae in a zinc-dependent manner. We have identified l-cysteine as an endogenous zinc chelator resulting in the resensitization of VIM-1-expressing K. pneumoniae to CTX. These results suggest that natural zinc chelators in combination with conventional antibiotics could be used to treat infections caused by VIM-1-expressing pathogens.

INTRODUCTION

The rapid increase of antimicrobial resistance is of high concern, with ESKAPE pathogens (Enterococcus faecium, Staphylococcus aureus, Klebsiella pneumoniae, Acinetobacter baumannii, Pseudomonas aeruginosa, and Enterobacter species) posing a particular threat (1). Hospital-acquired infections by these pathogens are considered especially problematic, where the development of antibiotic resistance limits treatment options, increases mortality, and contributes to increased costs for the health care system (2). To address this problem, research efforts have focused on the development of new antimicrobials and combination treatments and modification of contemporary antibiotics (3–7). Unfortunately, the rate of development of these new drugs is low, while bacterial resistance mechanisms are evolving rapidly. In particular, Gram-negative bacteria, such as K. pneumoniae, show extensive drug resistance against many clinically used antibiotics and thus constitute significant clinical challenges (8, 9).

One of the most important resistance mechanisms utilized by K. pneumoniae is the expression of specific β-lactamase enzymes that hydrolyze, and thereby inactivate, β-lactam antibiotics. These β-lactamases can be categorized according to the Ambler system, with classes A, C, and D comprising serine β-lactamases and class B comprising metallo-β-lactamases (MBLs) (10). MBLs are a group of β-lactamases that require zinc ions in the enzyme’s binding pocket to catalyze the hydrolysis of the amide bond in the β-lactam ring, resulting in inactivation of the antibiotic. The MBL group includes New Delhi metallo-β-lactamase (NDM), Verona integron-borne metallo-β-lactamase (VIM), and IMP-type metallo-β-lactamase (IMP). The genes encoding these β-lactamases can either be integrated into the bacterial chromosome or be present on plasmids that can be transferred between bacteria, spreading antibiotic resistance rapidly (11, 12).

The zinc dependence of MBLs has been studied with zinc chelators, such as EDTA or dipicolinic acid, which are able to resensitize bacteria to β-lactam antibiotics upon chelation of zinc (5, 13). One problem hampering the use of these inhibitors in a therapeutic setting is their tendency to chelate not only zinc but also a broad set of divalent metals, which can lead to toxic effects on human cells. Nevertheless, zinc chelation is an interesting approach against MBL-mediated antibiotic resistance and has shown efficacy in vitro and in a mouse model (14). However, effective and safe synthetic zinc chelators have not yet been developed or evaluated for human use.

Given the potential benefit of zinc chelation as a novel treatment option against MBL-producing bacteria, it is relevant to search for endogenous molecules with this capacity. Therefore, we set out to search for an endogenous inhibitor of metallo-β-lactamases by screening cell culture supernatants for factors that could restore susceptibility of MBL-producing K. pneumoniae to cefotaxime (CTX). First, an assay was established to determine the susceptibility of K. pneumoniae strains with different resistance mechanisms to CTX in the cell culture supernatant of HT-29 cells. Next, size exclusion and reverse-phase fractionations were utilized to isolate the factor(s) in the supernatant that resensitized VIM-1-producing K. pneumoniae to CTX. Finally, by a candidate approach, we identified a redox-sensitive zinc chelator that could restore the susceptibility of VIM-1-producing K. pneumoniae to CTX.

RESULTS

A secreted component in the supernatant of human epithelial cells sensitizes VIM-1-producing K. pneumoniae to CTX.

To search for endogenous inhibitors of β-lactamases, K. pneumoniae strains producing different β-lactamases (KPC, VIM, OXA-48, or NDM) (Table 1) were cultured in either RPMI medium with 5% Luria broth (LB) or the supernatant from HT-29 colon epithelial cells with 5% LB in the presence or absence of CTX, with growth monitored with a Bioscreen instrument (Fig. 1A to L). Analysis of the bacterial growth showed that all strains were resistant to CTX in RPMI medium with 5% LB, while two strains producing VIM-1 metallo-β-lactamases displayed high susceptibility to CTX when incubated in the supernatant with 5% LB (Fig. 1A and B). These results were verified by a CFU survival assay, where the VIM-1-producing strain (AO15200) showed a 4-log reduction in CFU counts when cultured in the supernatant with 5% LB in the presence of CTX compared to culturing it in RPMI medium with 5% LB together with CTX (Fig. 1L). To determine whether this resensitization was specific for CTX, VIM-1-producing strain AO15200 was cultured in the supernatant with different antibiotics (ciprofloxacin, azithromycin, fosfomycin, imipenem, amoxicillin, doxycycline, cefotaxime, and ceftriaxone) (Fig. 1M to T). Out of the antibiotics tested, only the third-generation cephalosporins cefotaxime and ceftriaxone showed regained activity against the bacteria when cultured in the supernatant with 5% LB, compared to RPMI medium with 5% LB alone.

TABLE 1.

K. pneumoniae strains used in the study and resistance profiles for susceptibility testing^a

K. pneumoniae strain	β-Lactamase (ResFinder)	MIC (zone diameter [mm]) (sensitivity)
		CTX	CAZ	IPM	CIP	TZP	ETP	MEM	SXT	GN	AK	CAT	FOX
AO15200	VIM-1	6 (R)	6 (R)	12 (R)	6 (R)	6 (R)	13 (R)	6 (R)	6 (R)	19 (S)	15 (I)
VPKP229	VIM-1	6 (R)	6 (R)	16 (I)	6 (R)	8 (R)	19 (S)	18 (I)	6 (R)	16 (I)	14 (I)
71076	KPC-2	6 (R)	6 (R)	20 (I)	6 (R)	12 (R)	13 (R)	15 (R)	6 (R)	18 (S)	21 (S)
RS044	KPC	6 (R)	6 (R)	10 (R)	6 (R)	6 (R)	6 (R)	6 (R)	6 (R)	18 (S)	12 (R)	6 (R)	9 (R)
RS116	KPC	6 (R)	6 (R)	13 (R)	6 (R)	6 (R)	10 (R)	17 (I)	6 (R)	18 (S)	15 (I)	6 (R)	6 (R)
RS092	NDM	6 (R)	6 (R)	9 (R)	6 (R)	6 (R)	6 (R)	6 (R)	6 (R)	6 (R)	6 (R)	6 (R)	6 (R)
RS104	NDM	6 (R)	6 (R)	15 (R)	6 (R)	6 (R)	6 (R)	6 (R)	6 (R)	6 (R)	6 (R)	6 (R)	6 (R)
RS086	NDM	6 (R)	6 (R)	6 (R)	6 (R)	6 (R)	6 (R)	6 (R)	6 (R)	6 (R)	6 (R)	6 (R)	6 (R)
RS098	NDM	6 (R)	6 (R)	16 (I)	6 (R)	6 (R)	11 (R)	13 (R)	6 (R)	6 (R)	6 (R)	6 (R)	6 (R)
RS095	OXA-48	6 (R)	6 (R)	15 (R)	6 (R)	6 (R)	7 (R)	12 (R)	6 (R)	6 (R)	6 (R)	6 (R)	6 (R)
RS083	OXA-48	6 (R)	8 (R)	22 (S)	11 (R)	8 (R)	21 (R)	23 (S)	6 (R)	19 (S)	19 (S)	18 (S)	21 (S)

^aThe antibiotics tested on the different strains were cefotaxime (CTX), ceftazidime (CAZ), imipenem (IPM), ciprofloxacin (CIP), piperacillin-tazobactam (TZP), ertapenem (ETP), meropenem (MEM), trimethoprim-sulfamethoxazole (SXT), gentamicin (GN), amikacin (AK), faropenem (CAT), and cefoxitin (FOX). Boldface type indicates resistance (R). I, intermediate, S, susceptible.

FIG 1.

Screening of β-lactamase-producing K. pneumoniae in the cell culture supernatant of HT-29 cells reveals altered susceptibility to CTX. (A to L) K. pneumoniae strains with the indicated resistance mechanisms (Table 1) were screened for susceptibility to 16 mg/liter CTX (in the resistant range) in RPMI medium plus 5% LB or in the cell supernatant (Sup) plus 5% LB by measuring growth curves in a Bioscreen C device and by viable CFU counts. (M to T) VIM-1-producing K. pneumoniae (AO15200) was screened for susceptibility to ciprofloxacin, azithromycin, fosfomycin, imipenem, amoxicillin, doxycycline, cefotaxime, and ceftriaxone in RPMI plus 5% LB or in the supernatant plus 5% LB. T0 represents initial inoculum. Data are presented as the means of results from three or more independent experiments. CFU counts are presented, with statistical analysis of log-transformed CFU per milliliter with an independent t test. Statistical significance, presented as P values of <0.05, are indicated (****, P < 0.0001).

The restored susceptibility of VIM-1-producing K. pneumoniae to CTX is zinc dependent.

The observation that the HT-29 cell-conditioned supernatant changed the resistance pattern of VIM-1-producing K. pneumoniae, specifically against cephalosporins, suggested a direct involvement of the β-lactamase enzyme. As VIM-1 activity has been shown to be zinc dependent, we first aimed to determine whether zinc chelation could resensitize VIM-1 K. pneumoniae to CTX in our system (Fig. 2A). Growth of the AO15200 VIM-1-producing K. pneumoniae strain in RPMI medium plus 5% LB with the addition of EDTA and CTX indeed showed that divalent metal chelation by EDTA could resensitize VIM-1 K. pneumoniae to CTX in our system. Importantly, the effect of EDTA addition resembled the effect of incubation of the K. pneumoniae strain in the supernatant with CTX, suggesting that metal chelation might also explain the effect of the supernatant on the CTX susceptibility of the VIM-1 K. pneumoniae strains. To further investigate this, different divalent metal ions were added to the system to study whether the restored susceptibility in the supernatant was influenced by specific metals. This showed that the addition of ZnSO₄, but not FeSO₄, MgSO₄, and CaCl₂, reinstated bacterial resistance to CTX (Fig. 2B). Of note, EDTA did not have any direct effects on the permeability of the cell membrane, as shown with a Sytox green assay (Fig. 2C). In addition, a reduced amount of free zinc was detected in the HT-29 supernatant compared to RPMI medium when fixed concentrations of ZnSO₄ were added (see Fig. S1 in the supplemental material). Together, these findings suggested that an active zinc chelator in the cell supernatant was responsible for the resensitization of VIM-1-expressing K. pneumoniae to CTX.

FIG 2.

The altered susceptibility of VIM-1 metallo-β-lactamase to CTX is a result of zinc chelation in the supernatant. (A) VIM-1-producing K. pneumoniae strain (AO15200) cultured in RPMI medium plus 5% LB, with the addition of EDTA (0.1 mM) and CTX (16 mg/liter). (B) AO15200 was cultured in RPMI medium plus 5% LB or in the supernatant (Sup) plus 5% LB, with supplementation with the divalent metals FeSO₄, ZnSO₄, MgSO₄, and CaCl₂ (0.5 mM). T0 represents initial inoculum. (C) Sytox green cell permeability measured with EDTA and compared to that of heat-killed bacteria. Graphs represent means of data from three or more independent experiments, with statistical significance presented as P values of <0.05, as indicated (**, P < 0.01; ***, P < 0.001; ****, P < 0.0001; ns, not significant).

The zinc binding factor was identified as l-cysteine, and the activity was dependent on free thiols.

To identify factors in the cell supernatant that could contribute to the observed resensitization, the HT-29 supernatant was first filtered through a 3-kDa size exclusion column. This showed that the resensitizing activity was retained in the fraction below 3 kDa (Fig. 3A). Next, the filtrate was passed through an Oasis reverse-phase column and calibrated in 0.1% trifluoroacetic acid (TFA) to enrich for peptides and proteins. Notably, the activity was found in the flowthrough, while no activity in fractions obtained after elution with 0.1% TFA in 80% acetonitrile was found, suggesting that the compound was strongly hydrophilic (Fig. 3A). When the RPMI cell culture medium formulation was examined, it was observed that the medium contained l-cystine (intact disulfide bridge), which can be reduced to l-cysteine during cell culture growth and is known to be involved in the zinc binding properties of proteins (15, 16). Given this information, the presence of l-cysteine in the supernatant was studied, as were the effects of l-cysteine on VIM-1 K. pneumoniae susceptibility to CTX. First, while the addition of l-cystine to RPMI medium (5% LB) did not affect the susceptibility of the bacteria to CTX, the addition of the reduced form, l-cysteine (free thiol group), to RPMI medium (5% LB) restored the susceptibility of VIM-1 K. pneumoniae to CTX. This suggested that the reduction of l-cystine to l-cysteine in the supernatant could be important for the inhibition of VIM-1 activity (Fig. 3B). The addition of l-cysteine was also studied in the Bioscreen system, which indicated that the lowest concentration of l-cysteine that affected antibiotic resistance was 125 μM (Fig. S2). To further validate the role of l-cysteine in VIM-1 K. pneumoniae susceptibility to CTX, either complete Dulbecco’s modified Eagle’s medium (DMEM) or DMEM without l-cystine, methionine, and glutamine (DMEM/−) was added to HT-29 cells. The different supernatants (DMEM, DMEM/−, and l-cysteine in sterile distilled water [dH₂O]) were subjected the same sample purification, and the flowthrough of the Oasis column was lyophilized and passed through a reverse-phase column equilibrated in 0.1% TFA and fractionated under isocratic flow at 1 ml/min. As l-cysteine has an active thiol group, the thiol content of the supernatants was examined in each fraction. No thiols were present in the DMEM/− supernatant, whereas the DMEM supernatant with l-cystine resulted in a clear thiol content in fractions 6 and 7 (Fig. 3C). Importantly, pure l-cysteine in sterile water appeared in the same fractions (fractions 6 and 7) as the free thiols detected in the DMEM supernatant. Furthermore, fractions 6 and 7 from pure l-cysteine as well as the DMEM supernatant were able to resensitize K. pneumoniae to CTX, while fractions 6 and 7 from the DMEM/− supernatant did not cause any resensitization of VIM-1-producing K. pneumoniae (Fig. 3D). To determine whether other thiols besides l-cysteine could have a similar impact on CTX susceptibility, glutathione (GSH) was added to RPMI medium. Oxidized glutathione (intact disulfide bridge) did not show any activity on bacterial resistance to CTX, similarly to l-cystine; however, the reduced form of glutathione showed the same activity as l-cysteine, restoring susceptibility to CTX (Fig. 3E). Finally, to confirm that l-cystine was indeed reduced in the supernatant in a cell-dependent manner, fluorescently labeled l-cystine (BODIPY FL–l-cystine) was used, which increases fluorescence upon reduction to l-cysteine. Incubation of labeled l-cystine in the cell supernatant resulted in an increase in the relative fluorescence, whereas no reduction occurred in normal RPMI medium (Fig. 3F). Importantly, when the thioredoxin pathway was inhibited with the specific inhibitor PX-12 (10 to 100 μM) in the fresh supernatant, the conversion of l-cystine to l-cysteine did not occur using BODIPY FL–l-cystine, identifying the role of the thioredoxin reductase (TRX) pathway in generating free thiols in the cell supernatant (Fig. 3G). To confirm that l-cysteine had a direct effect on VIM-1 activity, an enzyme kinetic assay was performed on the closely related VIM-2 enzyme, and the hydrolysis rate of CTX was determined in the presence of EDTA, l-cysteine, and l-cystine (Fig. S3). Both EDTA and l-cysteine inhibited the hydrolysis of CTX, while l-cystine did not, further supporting the hypothesis that zinc chelation is a potent enzyme-inhibiting mechanism.

FIG 3.

The altered susceptibility of VIM-1-producing K. pneumoniae results from the reduction of l-cystine to l-cysteine in the cell culture supernatant of HT-29 cells. (A) The supernatant was subjected to a size exclusion filter where the low-molecular-weight fraction (<3 kDa) was loaded onto an Oasis column, enriching for peptides and proteins. K. pneumoniae (AO15200) was cultured in the different fractions in the presence of 16 mg/liter CTX. (B) The K. pneumoniae strain was cultured in the presence of l-cystine and l-cysteine (0.5 mM) with and without 16 mg/liter CTX, after which viability was determined by a colony count assay. T0 represents initial inoculum. (C) The Oasis column flowthrough fractions of supernatants from HT-29 cells cultured in either DMEM or DMEM/− were loaded onto a reverse-phase column (RPC Resource column) and fractionated with an isocratic flow of 0.1% TFA. Fractions were collected for 15 min (flow rate, 1 ml/min). The thiol content was measured in the fractions with a GSH standard, and fractions where l-cysteine was loaded onto the column constituted a positive control. (D) Colony count assay of the K. pneumoniae strain (AO15200) cultured in the presence of fractions 6 and 7 with or without 16 mg/liter CTX. (E) Colony count assay of the K. pneumoniae strain cultured with glutathione (oxidized [ox] and reduced [red]) or l-cysteine in the presence or absence of 16 mg/liter CTX. (F) The supernatant collected from HT-29 cells was analyzed for reducing properties with BODIPY FL-Cys conversion of l-cystine to l-cysteine. Samples include cells with fresh RPMI medium, cells with RPMI medium after 24 h, and the supernatant alone. Values are given relative to the value for the RPMI control. (G) Measurement of l-cystine-to-l-cysteine conversion with BODIPY FL-Cys in the HT-29 cell supernatant in the presence of the indicated concentrations of the thioredoxin antagonist PX-12. Means were compared to the value for the supernatant with 0 μM PX-12. Graphs represent means of data from three independent experiments, with statistical significance presented as P values of <0.05, as indicated (*, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001). Statistical significance was measured on the means of log-transformed CFU per milliliter with an independent t test, and for relative fluorescence, one-way ANOVA was performed with Tukey’s multiple-comparison test.

Expression of VIM-1 and NDM-1 in a neutral background confirms that CTX susceptibility can be restored for both MBLs by l-cysteine.

An NDM-1-producing Escherichia coli DH5α isolate was tested for CTX susceptibility in the presence of l-cysteine, resulting in a 1,000-fold reduction of the MIC after 10 h (Fig. 4B), which is comparable to the reduction observed in the AO15200 strain (Fig. 4A). To study the change in the susceptibility of VIM-1 in a neutral background, a two-step strain transconjugation was performed, transferring the VIM-1 plasmid from AO15200 to E. coli ATCC 25922 and MG1655 through a camR ΔdapA intermediate strain (DA42859). The final recipient strains were tested in the Bioscreen instrument for susceptibility to CTX with and without l-cysteine, confirming that the VIM-1 enzyme activity was affected by the free thiol in l-cysteine (Fig. 4C). The transfer was confirmed by colony PCR and gel electrophoresis (Fig. 4D), showing that the VIM enzyme was successfully transferred to the recipient strains.

FIG 4.

MBL expression in a neutral background confirms the thiol-mediated effect on VIM and NDM. (A and B) The MICs of CTX for AO15200 and an NDM-1-expressing E. coli DH5α strain with and without l-cysteine were measured with a Bioscreen growth curve, and the OD₆₀₀ after 10 h was evaluated. (C) The growth of the VIM-1 transconjugants ATCC 25922_VIM-1 and MG1655_VIM-1 was measured with a Bioscreen assay in the presence of l-cysteine and CTX. (D) PCR and gel electrophoresis confirm plasmid transfer from AO15200 to ATCC 25922 and MG1655. Graphs represent means of data from three independent experiments.

Zinc-dependent thiols in urine determine CTX susceptibility of VIM-1-producing K. pneumoniae.

While our findings showed that endogenous thiols could resensitize VIM-1-expressing K. pneumoniae strains to CTX in vitro, we aimed to investigate whether this effect could also be detected in a more physiological setting. Since K. pneumoniae is a known urinary pathogen, the resistance of the VIM-1 strain to CTX was studied ex vivo in human urine collected from healthy volunteers. Interestingly, when incubated in urine, the susceptibility of VIM-1-expressing K. pneumoniae to CTX was restored (Fig. 5A). To characterize whether free thiols were also responsible for the observed resensitization in this setting, the urine was passed through a 3-kDa size exclusion filter and an Oasis column and applied to a reverse-phase column, and fractions were collected. This procedure of fractionation was performed in the same manner as for the supernatant from HT-29 cells. In line with the findings in the supernatant, fractions 6 and 7 were able to resensitize the bacteria to CTX (Fig. 5B). In addition, fractions 6 and 7 also contained free thiols (Fig. 5C), suggesting the presence of l-cysteine or other small thiol-containing molecules in the urine with the capacity to inhibit VIM-1 metallo-β-lactamases. Finally, we assessed whether the inhibition of VIM-1 activity by the thiols in urine was dependent on zinc. Urine from different donors was diluted from 100% to 5%, and bacterial growth in the presence of 16 mg/liter CTX (MIC value in the resistant range) was studied. The urine concentration required to resensitize K. pneumoniae to CTX was documented for each donor sample and ranged from 20 to 50% urine depending on the thiol content of the sample (Fig. 5D). Finally, each urine sample was diluted to a concentration that still showed resensitization to CTX, after which it was supplemented with ZnSO₄ (10 μM) and again tested for susceptibility to CTX. This showed that, similar to the findings with the cell supernatant, the addition of ZnSO₄ prevented resensitization of the AO15200 VIM-1 K. pneumoniae strain to CTX (Fig. 5E).

FIG 5.

Zinc-chelating thiols in urine resensitize VIM-1-producing K. pneumoniae to CTX. (A) Colony count assay to determine the susceptibility of the K. pneumoniae strain (AO15200) to CTX in both urine and the HT-29 supernatant. (B) The urine samples were subjected to the same size exclusion, peptide/protein enrichment (Oasis column), and reverse-phase fractionation as the HT-29 cell culture supernatant. The fractions obtained after the reverse-phase step were used in a colony count assay with the K. pneumoniae strain (AO15200) in RPMI medium plus 5% LB in the presence or absence of CTX. (C) The thiol content in the fractions was measured with a fluorometric thiol assay and compared to a GSH standard. (D) Urine samples were diluted from 100% to 5% in PBS and tested for inhibition of growth of K. pneumoniae (AO15200) with CTX. The dilution before resensitization was lost (evaluated for each donor sample) was used in panel E (green circle for a representative sample). (E) Pure or diluted urine was used to culture the K. pneumoniae strain in the presence of CTX and with or without the addition of ZnSO₄ (10 μM). T0 represents initial inoculum. Statistical significance was measured on the means of data from three or more independent experiments of log-transformed CFU per milliliter with an independent t test. P values are indicated (**, P < 0.01; ***, P < 0.001; ****, P < 0.0001).

DISCUSSION

Here, we found that free thiols regulated the susceptibility of a metallo-β-lactamase-producing K. pneumoniae strain to CTX, a commonly used cephalosporin, while susceptibility to other β-lactam antibiotics was unaffected. The effect was the result of zinc chelation by free thiols present in the supernatant of HT-29 cells and in human urine restoring bacterial susceptibility to CTX and ceftriaxone. The free thiols were originally not present in the cell culture medium but were the result of extracellular reduction of l-cystine to l-cysteine in the cell supernatant. This reduction, exerted through the thioredoxin pathway, exposed the free thiol group in cysteine, which is known to be crucial for the interaction between cysteine and zinc (15, 16). Notably, out of the tested strains, only VIM-1-expressing isolates regained susceptibility to CTX upon zinc chelation by EDTA and the supernatant of HT-29 cells. To rule out that the effect was related to the genetic background of the bacteria, the VIM-1 plasmid was transferred via transconjugation to one ATCC strain and to one MG1655 wild-type strain, where resensitization to CTX in the presence of l-cysteine was acquired by the recipient strain. This supported the notion that the effect was indeed directly related to the VIM-1 plasmid.

In contrast, and somewhat unexpectedly, the NDM-expressing clinical isolates did not exhibit a similar restored susceptibility (compared to the CTX susceptibility in supernatant or with EDTA), despite the fact that NDM is also a metallo-β-lactamase. This is most likely caused by the presence of additional resistance mechanisms in these strains, where the presence of an extended-spectrum-β-lactamase (ESBL) plasmid can compensate for the inhibition of NDM by EDTA and explain why the NDM-expressing strains in this study did not respond to zinc chelation. This idea was further strengthened by the fact that a strain expressing NDM-1 in a neutral background regained susceptibility to CTX in the presence of l-cysteine.

l-Cysteine and other thiols have been shown to have zinc-chelating abilities, and thiols have been suggested as possible inhibitors of metallo-β-lactamases (17, 18). Together with our data, this suggests that redox regulation of endogenous molecules, such as l-cysteine, could be an interesting target for resensitizing resistant bacteria to cephalosporins. When human urine samples were analyzed with the same purification procedure as the one used for the HT-29 cell supernatant, the same fractions improved CTX susceptibility of the VIM-1-producing strain as the supernatant. This suggests that the same compound (l-cysteine or other free thiols) plays a major role in regulating the susceptibility of VIM-1-expressing bacteria to cephalosporins. Notably, the finding that the cellular supernatant has an impact on antibiotic susceptibility via reduction of l-cystine to l-cysteine has been reported previously, but the underlying mechanisms have not yet been identified (19). For example, the inactivation of meropenem by cellular supernatants has been reported (19), as has the inactivation of penicillins by l-cysteine (20).

In this study, the bacteria regained susceptibility to cephalosporins only (cefotaxime and ceftriaxone), suggesting an unknown step in thiol-mediated zinc chelation and MBL inactivation. The degradation of cephalosporins has been shown to produce thiol intermediates that can inhibit the MBL enzyme by reacting to the binding pocket (5, 21), which could be a factor contributing to the zinc chelation effect observed in our study.

To the best of our knowledge, this is the first time that cysteine has been shown to resensitize multidrug-resistant (MDR) K. pneumoniae to CTX. In fact, these results suggest that redox regulation of not only l-cysteine but also many cysteine-containing proteins could be involved in the chelation of zinc and thus constitutes a novel way to regulate metallo-β-lactamase activity in other bacteria, including Acinetobacter and Pseudomonas.

Here, we show that PX-12, an inhibitor of the thioredoxin pathway, blocked the conversion of l-cystine to l-cysteine, indicating a role of the thioredoxin reductase in our system. This could potentially be relevant since this enzyme is upregulated by, e.g., inflammation or infection (22).

Despite the strengths of these novel findings, the findings also had some limitations. First, the effect appears to be limited to MBLs in relation to cephalosporins but is not apparent for other antibiotics of the β-lactam class. It would be reasonable that a similar mechanism would also apply for carbapenems. However, we failed to observe an effect for imipenem, which points to a more complex mechanism in MBL resistance against carbapenems. To connect the observations to biological relevance, urine was chosen for its relevance to the bacteria, where K. pneumoniae is known to cause urinary tract infections. Urine contains free thiols, which can be regulated by various factors, including inflammation, drugs, and diet (23). Notably, urine per se did not have the ability to reduce cystine to cysteine (see Fig. S4 in the supplemental material) but contains a large amount of thiols, which have the capacity to restore bacterial susceptibility to CTX. The addition of ZnSO₄ again counteracted the susceptibility, similar to the results observed in the supernatant of HT-29 cells, indicating that the change in the resistance pattern of the MBL-producing bacteria was caused by limitation of zinc, which could be reversed with ZnSO₄ supplementation.

Previous research efforts have focused on new chemical entities as inhibitors of MBL enzymes (13), but little has been studied in relation to innate defenses. The contribution of the host has often been neglected when resistant strains are studied. The data presented here may be used to develop novel adjunctive treatment options for infections, especially for VIM-1-producing K. pneumoniae but possibly also for other bacteria with zinc-dependent multidrug-resistant enzymes.

Notably, the importance of zinc for the function of metallo-β-lactamases has been well established; however, broad metal chelation, such as with EDTA, is not a viable option for treating antibiotic-resistant bacterial infections due to toxic side effects. Applying host factors, such as cysteine or other sources of free thiols, may be a key solution in combination with conventional antibiotics to reduce the amount of antibiotics needed and regain their activity against strains that have developed resistance.

In conclusion, we have demonstrated the importance of free thiols in the extracellular environment for the activity of CTX against VIM-1-expressing K. pneumoniae. These findings open up novel avenues to develop therapies against clinical infections caused by MDR bacteria. However, the results need to be validated in a clinical context. Possible ways to increase the thiol content in biological settings, including the urinary tract, the intestine, and the lung, could be investigated as a novel way to resensitize bacteria to clinically relevant antibiotics.

MATERIALS AND METHODS

Cell culture and supernatant preparation.

HT-29 colon epithelial cells were cultured and maintained in RPMI cell culture medium (Thermo Fisher Scientific, USA) supplemented with 2 mM l-glutamine and 10% fetal bovine serum (FBS) (Thermo Fisher Scientific, USA). For supernatant isolation, cells were allowed to reach 80% confluence, washed with phosphate-buffered saline (PBS), and incubated in FBS-free RPMI medium. After 24 h of culturing, the supernatant was collected, centrifuged at 1,500 rpm for 5 min, and used directly or stored at −20°C for short-term storage. Alternatively, FBS-free DMEM or FBS-free DMEM without cystine, methionine, and glutamine (DMEM/−) (Thermo Fisher Scientific, USA) was used for supernatant collection.

Collection of urine from healthy volunteers.

Urine from healthy donors was collected and centrifuged at 4,500 rpm for 10 min at 4°C. Next, samples were sterile filtered (Filtopur S 0.2; Sarstedt, Germany), aliquoted, and stored at −20°C. Thawed aliquots were vortexed prior to use. The collection of urine samples was performed under ethical permit number 2019-02519 (Swedish Ethical Review Authority).

Bioscreen growth measurements.

For all bacterial strains, colonies from a culture grown overnight on blood agar plates were picked and cultured to exponential phase (optical density at 600 nm [OD₆₀₀] of 0.3 to 0.5) in Luria broth (LB) (pH 7.5) at 37°C on a shaking incubator. The bacterial suspension was centrifuged at 5,000 rpm for 3 min, washed once with PBS, and diluted to an OD of 0.05 in the desired culture medium. For all growth conditions, RPMI medium, the supernatant, or PBS was supplemented with 5% LB for bacterial growth (24, 25). For each well, the total starting inoculum was fixed to 5 μl from an OD of 0.05 (∼1 × 10⁵ CFU) in a final volume of 150 μl RPMI medium, including various treatments, in a Honeycomb plate (Bioscreen C; Oy Growth Curves Ab Ltd., Finland). The following antibiotics were used for screening of susceptibility: cefotaxime, ceftriaxone, ciprofloxacin, azithromycin, fosfomycin, amoxicillin, doxycycline, and imipenem (Sigma-Aldrich, Germany). The Bioscreen instrument (Bioscreen C) registered OD₆₀₀ measurements every 10 min for 24 h at 37°C, thus producing a growth curve from each well.

Viable CFU counts.

The viable CFU counts were performed similarly to the Bioscreen assay. Bacteria (∼1 × 10⁵ CFU) were suspended in RPMI, DMEM, DMEM/−, the supernatant, or urine in a final volume of 150 μl in combination with various treatments in 96-well round-bottom plates (Costar; Corning, USA). Samples were incubated at 37°C on a shaking incubator for 4 h. Serial 10-fold dilutions of each sample were made, and every dilution was plated onto the same blood agar plate, where each plate was divided into 8 parts. The plates were incubated overnight at 37°C, and colonies were counted the next day. Treatments (CTX, EDTA, l-cystine, l-cysteine, and divalent metals) were prepared in a 96-well plate (Costar; Corning, USA). EDTA (Substrate Unit, Karolinska University Hospital, Sweden) was used at 0.1 mM, and the addition of various metal salts (ZnSO₄, FeSO₄, MgSO₄, and CaCl₂) (all from Merck, Germany) were used at a 5 μM concentration. The final concentration of CTX (Sigma-Aldrich, USA) was 16 mg/liter unless stated otherwise. l-Cystine and l-cysteine (Sigma-Aldrich, USA) were used at 0.5 mM, and glutathione (GSH) (Sigma-Aldrich, USA) was used at 0.25 mM.

Sample fractionation.

After collection of the cell culture supernatant or urine, samples were acidified with 0.1% trifluoroacetic acid (TFA) (Sigma-Aldrich, Germany) and loaded onto Oasis HLB 1-ml extraction columns (Waters, Ireland) equilibrated in 0.1% TFA. The samples were eluted with 80% acetonitrile (Sigma-Aldrich) in 0.1% TFA (Sigma-Aldrich), and the flowthrough from the columns was also collected. For size exclusion, 3-kDa size exclusion Amicon Ultra 15 3K spin columns (Merck Millipore, Ireland) were used. All samples were lyophilized (Maxi dry lyo; Gemini, Netherlands) and concentrated 10 times in sterile dH₂O. Next, samples were loaded onto an RPC Resource reverse-phase column (GE Healthcare, Sweden), utilizing the Äkta Pure system (GE Healthcare, Sweden) (500-μl volume). The column was equilibrated in 0.1% TFA with a flow rate of 1 ml/min. The first 15 fractions were collected with an isocratic run of 0.1% TFA, and the absorbance was recorded at 214 nm to detect peptide bonds in the samples. Finally, the fractions were lyophilized, resuspended in 250 μl sterile dH₂O, and subsequently used for further analysis.

Zinquin free zinc measurement.

To measure free zinc in the cell culture supernatant, a protocol reported previously by Zalewski et al. was followed, with the use of a Zinquin stock dye solution (Sigma-Aldrich, USA) (26). A zinc standard was made with serial dilutions of ZnSO₄ (Merck, Germany). The Zinquin reagent was diluted in Hanks’ balanced salt solution (HBSS) buffer (Thermo Fisher Scientific, USA) to a final concentration of 10 μM Zinquin in a white clear-bottom 96-well plate (Corning, USA). Samples were transferred to each well and incubated for 40 min at room temperature (RT). The fluorescence was measured with an excitation filter of 365 nm and an emission filter of 490 nm with a Tecan Infinite M200 reader (Tecan, Switzerland).

Sytox green viability.

A Sytox green viability assay was performed to assess whether EDTA addition had any effect on the permeability and viability of K. pneumoniae. The AO15200 strain was cultured in LB to exponential phase and diluted to an OD₆₀₀ of 0.1. Samples with 10 μl of bacteria in 500 μl broth with or without 0.1 mM EDTA were incubated further for 1 h. A positive-control sample was prepared by heat killing the bacteria at 70°C for 10 min. Samples were centrifuged at 10,000 × g for 2 min, washed twice in PBS, and finally resuspended in 2 μM Sytox green (Thermo Fisher Scientific, USA) in PBS. After a 5-min incubation in the dark, the fluorescence was measured at an excitation wavelength of 488 nm and an emission wavelength of 523 nm in a Tecan Infinite M200 reader.

Thiol measurement assay.

To measure thiol content, a fluorometric thiol assay kit (Sigma-Aldrich, USA) was used according to the manufacturer’s instructions. The fractions from the reverse-phase column were analyzed, and the thiol content was determined based on a glutathion (GSH) standard curve. The fluorescence was measured with a Tecan Infinite M200 plate reader with an excitation filter of 490 nm and an emission filter of 520 nm.

Cystine-cysteine conversion.

HT-29 colon epithelial cells were seeded in a 96-well plate (Sarstedt, Germany), grown to 80% confluence in RPMI medium with 10% FBS, washed with PBS, and incubated in FBS-free RPMI medium, DMEM, or DMEM/− for 24 h. Next, the supernatants (with or without cells present) were incubated with 5 μM BODIPY FL–l-cystine (Thermo Fisher Scientific) for 1 h at 37°C, and the fluorescence emitted by the breakdown of l-cystine to l-cysteine was observed with a 485-nm excitation filter and a 535-nm emission filter with a Tecan Infinite M200 reader. To inhibit the reduction of l-cystine, the thioredoxin inhibitor PX-12 (Sigma-Aldrich, Germany) was added to the freshly collected supernatant (10 to 100 μM) before incubation with BODIPY FL–l-cystine.

Bacterial plasmid transconjugation.

To transfer the VIM-1 plasmid from the AO15200 clinical isolate to a neutral E. coli background (ATCC 25922 and MG1655), a two-step transfer was performed. An intermediate recipient strain was used for transconjugation, a ΔdapA camR auxotroph E. coli MG1655 strain (DA42859) (provided by Linus Sandegren and Fredrika Rajer, Uppsala University), giving a double-positive selection where the intermediate strain is unable to grow without 2,6-diaminopimelic acid (DAP) (Sigma-Aldrich, Darmstadt, Germany) and is chloramphenicol resistant. The conjugation of the donor (AO15200) and the recipient was conducted on Whatman Protran BA85 papers (GE Healthcare Life Sciences, Germany) on LB agar plates containing 20 μg/ml DAP. The strains were incubated together for 2 h in LB at 37°C prior to transfer to the filters on LB agar plates containing DAP for overnight incubation at 37°C. Next, the filters were washed in PBS and streaked onto a plate containing 20 μg/ml DAP, 15 μg/ml chloramphenicol (Sigma-Aldrich, Germany), and 100 μg/ml ampicillin (Sigma-Aldrich, Germany) to select for DA42859 that acquired resistance (transconjugant). CFU from a culture incubated overnight were restreaked onto the same type of selection plate for confirmation. The steps were repeated to transfer the plasmid from the intermediate strain to both E. coli ATCC 25922 and MG1655. The selection for the final recipient transconjugant was done using 100 μg/ml ampicillin with no DAP present.

Plasmid transfer confirmation by colony PCR amplification and gel electrophoresis.

Colonies were picked from the agar plate and dissolved in 50 μl sterile dH₂O. The bacterial suspension was heated at 96°C for 10 min, and PCR mixtures were prepared from the bacterial lysates. The following primers were used to amplify VIM-1: forward primer 5′-AGTGGTGAGTATCCGACAG-3′ and reverse primer 5′-ATGAAAGTGCGTGGAGAC-3′. DreamTaq hot-start green PCR master mix (Thermo Fisher Scientific, USA) was used, and the following PCR heat cycle was used: 94°C for 2 min; 25 cycles of 94°C for 15 s, 59°C for 15 s, and 72°C for 15 s; and 72°C for 2 min. The product was loaded onto a 1% agarose gel with a 1:10,000 dilution of GelRed dye (Biotium, USA) for 40 min at 100 V.

VIM-2 enzymatic assay.

To determine the effects of cysteine on enzymatic activity, a VIM-2 vector was constructed to overexpress the protein, as follows: (i) the primer BamHI-TEV-VIM-2-D-F (5′-CGGGATCCGAAAACCTGTATTTCCAGTCCGTAGATTCTAGCGG-3′) with His tags and the corresponding reverse primer, XhoI-VIM-2-R (5′-CCGCTCGAGCTACTCAACGACTGAGCGATT-3′), were used to amplify the corresponding gene; (ii) the amplification products were cloned into the pET-28a vector, and the resulting plasmids were transformed into E. coli BL21(DE3) (TransGen Biotech, Beijing, China); and (iii) all transformants were selected on a Mueller-Hinton agar (MHA) plate supplemented with 50 mg/liter kanamycin and further confirmed by PCR and sequence analysis. Proteins were expressed in E. coli BL21(DE3) cells and purified by using Ni-nitrilotriacetic acid agarose according to the manufacturer’s instructions. His tags were cleaved using Turbo tobacco etch virus (TEV) protease (Accelagen, San Diego, CA, USA), and SDS-PAGE was employed to assess protein purity. The protein concentration was determined by using a Pierce bicinchoninic acid (BCA) protein assay kit (Thermo Scientific, Waltham, MA, USA). The initial enzyme hydrolysis rate of CTX was determined in different buffers (PBS, 0.1 mM EDTA, 0.5 mM l-cystine, and 0.5 mM l-cysteine) at 28°C with 2 μM Zn²⁺ at pH 8.0 by using a SpectraMax M2e multidetection microplate reader (Molecular Devices, Sunnyvale, CA, USA), and a 260-nm wavelength was adopted to measure the absorbance of CTX.

Statistical analysis.

For all CFU data, the CFU per milliliter were log transformed and subjected to an unpaired t test or one-way analysis of variance (ANOVA) followed by Tukey’s multiple-comparison test. P values of <0.05 were considered statistically significant. Statistical analysis was performed with GraphPad Prism 7.01 software.