I. D. Kerr, School of Biomedical Sciences, University of Nottingham, Queen’s Medical Centre, Nottingham NG7 2UH, UK Fax: +44 115 8230142 Tel: +44 115 8230122 E-mail: firstname.lastname@example.org
The acquisition of multidrug resistance in bacteria underlies the failure of antimicrobial therapy, and the emergence of pathogens that are resistant to almost the entire armoury of antibiotics. Among the proteins that can mediate or contribute to the drug-resistance profile in Gram-positive bacteria is a subset of ATP-binding cassette proteins that are comprised of a tandem-repeated nucleotide-binding domain. In this study, we expressed one of these NBD2 proteins, LmrC, in an antibiotic-sensitive Gram-positive host strain (Lactococcus lactis) and demonstrated the acquisition of resistance to ribosomally active antibiotics. Mutation of key catalytic residues suggested that the resistance profile was the result of a cellular response, rather than being a function of the NBD2 protein itself. This observation was confirmed by 2D SDS/PAGE, which demonstrated that the expression of the NBD2 protein induced a stress response in L. lactis. A model combining this stress response induction and the acquisition of antibiotic resistance is proposed.
The emergence of drug resistance in bacteria, and in particular multidrug resistance, has become increasingly prominent over the past 30 years, and is a growing cause for concern in public health . In Gram-positive bacteria, a major class of clinically relevant antibiotics remains the macrolides, lincosamides and streptogramins-B (MLSB) antibiotics. These agents all target the ribosome, preventing the continued production of nascent protein by inhibiting protein synthesis . Structural studies of 50S ribosomal subunits have shown that the three classes of antibiotics have overlapping binding sites, with the 23S rRNA providing the contacts on this interaction surface . The major contributory factors to resistance to MLSB antibiotics are modification of the ribosome by methylation of bases, enzymatic modification and drug efflux [4–6]. Clinically, acquisition of methyltransferase is probably the most significant resistance mechanism , but the ability to export antibiotics is also a significant challenge to effective antimicrobial therapy, particularly as the efflux transporters often display broad substrate specificity enabling them to pump out chemically unrelated drugs [7,8].
One group of proteins with a less well-characterized role in antibiotic resistance is the NBD2 proteins or class II proteins , members of the ATP-binding cassette (ABC) family . These genes have been isolated from MLSB-resistant Gram-positive bacteria (best exemplified by the MsrA gene in Staphylococcus aureus), but have also been identified in the genomes of antibiotic-producing Streptomyces bacteria. The mechanism of action of these proteins remains unknown. ABC transporters are typically characterized by a functional unit including two transmembrane domains that bind and mediate substrate transport, and two nucleotide-binding domains (NBD) which provide free energy for transport, through binding and hydrolysis of ATP. In prokaryotes, ABC proteins are often encoded as multiple components that assemble either post- or co-translationally to form a membrane-spanning transporter. However, NBD2 proteins are somewhat different. These genes are either not part of an operon structure or, if they are located within an operon, there are no convincing candidate genes for the transmembrane domains in that operon. In other words, the NBD2 proteins are not positively identified as being part of a membrane-spanning transporter complex [10,11].
The mechanism of action of these NBD2 proteins remains unresolved. By extrapolation from their sequence similarity to eukaryotic elongation factors  it has been proposed that they may mediate translational control [12,13]. Indeed, the crystallographic structure of yeast elongation factor 3, and electron microscopic analysis of its interaction with the ribosome, describes a mechanism for translational control by these NBD2-containing eukaryotic proteins . Alternatively, it is possible that the bacterial NBD2 proteins confer antibiotic resistance by an export mechanism, and that the partner transmembrane proteins simply have yet to be identified. The localization of the NBD2 protein Vga(A) with a component of the FoF1-ATPase, rather than with ribosomes is compatible with this argument .
In this study we employed a Gram-positive expression system (L. lactis) to examine the NBD2 protein LmrC, from the lincomycin producer Streptomyces lincolnensis. This protein has previously been shown to confer eightfold resistance to lincomycin when expressed in the related, but lincomycin-sensitive, Streptomyces lividans . Here, we show that LmrC expression in L. lactis was associated with a gain of resistance to antibiotics which act at the 50S ribosome, but that this resistance is not dependent on the canonical Walker-A motifs of LmrC. Moreover, it appeared that L. lactis underwent a stress response upon being induced to express LmrC that was associated with an increase in the expression of chaperones and other heat shock proteins. The data are compatible with a model in which the preinduction of stress-response systems primes the cell to be able to resist subsequent addition of antibiotic.
LmrC expression in L. lactis is associated with a nonspecific resistance to antibiotics
Previous analysis of microbial genomes has revealed the presence of NBD2 proteins of the ABC family that are likely to be involved in events other than membrane transport [10,11]. We investigated whether expression of a NBD2 protein from an antibiotic-producing organism could confer antibiotic resistance on an antibiotic-sensitive host strain. Expression of LmrC from S. lincolnensis, in the antibiotic-sensitive L. lactis, was investigated for dependence on the duration of induction, inducer concentration and culture absorbance at the time of induction (Fig. 1). Addition of culture supernatant from the nisin-producing strain NZ9000 at dilutions from 1 : 250 to 1 : 10 000 enabled expression of LmrC tagged with an N-terminal His6ubiquitin tag (Fig. 1A,B,D). Dilutions of 1 : 1000 were routinely used thereafter. Detection of LmrC expression required western blotting with either anti-His6 IgG1 (Fig. 1B) or anti-ubiquitin IgG (data not shown), and the protein was not visible in Coomassie Brilliant Blue-stained 1D SDS/PAGE. Protein expression in the nisin system usually commences 1–2 h after induction. However, in this set of experiments, we wished to determine the effect of LmrC expression on antibiotic resistance, for which standard experimental protocols employ overnight assays. Consequently, it was important to demonstrate that the expression of LmrC was still detectable for > 20 h post induction. Indeed, the LmrC expression level after overnight culture was still 60–70% of that obtained after 2–4 h duration (Fig. 1C,E). LmrC expression was not highly sensitive to the absorbance at the time of induction, and comparable levels of protein expression were obtained with starting absorbance of 0.3, 0.6 or 0.9 (data not shown).
Expression of LmrC for > 16 h post induction enabled us to determine whether this twin-NBD protein was associated with acquisition of resistance to ribosomally active antibiotics. Broth microdilution assays, employing twofold dilution series of the macrolide tylosin demonstrated that LmrC-expressing cultures of L. lactis had a threefold greater resistance to tylosin compared with empty vector controls (IC50 4.6 ± 0.3 μg·mL−1, cf. IC50 1.5 ± 0.3 μg·mL−1; P < 0.001; Fig. 2A). Although this is lower than the eightfold resistance observed when LmrC was expressed in a related Streptomyces species , it is similar to the magnitude of increased resistance to macrolides documented in clinical isolates . In addition, significant differences in the resistance to the macrolide erythromycin, whose site of action overlaps with that of tylosin, but not tetracycline, whose site of action is distinct, were observed upon LmrC expression (data not shown).
Twin-NBD proteins, like related ABC transporters, are characterizd by the presence of Walker-A and Walker-B motifs which are essential for ATPase activity and protein function, and are likely to adopt a classical interlocking NBD dimer structure [17,18]. We constructed two mutant isoforms of LmrC in which the Walker-A lysine residues, involved in ATP coordination , were either singly (K44A) or doubly (K44A, K380A) mutated to alanine. In other ABC transporters, replacement of the Walker-A lysine is associated with a loss of function [19–21]. We were surprised to observe that expression of the Walker-A lysine mutant isoforms of LmrC at levels comparable with wild-type protein (Fig. 2B) was still associated with resistance to tylosin (Fig. 2C). Analysis of variance (ANOVA) demonstrated that all three isoforms were significantly more resistant to tylosin than the negative control (empty vector transformants), and were not significantly different from one another. Thus, it appears that the conferral of tylosin resistance upon LmrC expression does not require the functional integrity of the ATPase domains of the protein.
Induction of a stress response in L. lactis associated with LmrC expression
To investigate the basis of this presumably ATPase-independent effect of LmrC induction upon antibiotic resistance, the proteome was analysed to identify global changes in protein expression. L. lactis transformed with pNZ8048_His6Ub-Lmrc was compared with empty vector controls, and total protein was separated on pH 4–7 IEF strips with the second dimension being a 5–20% gradient acrylamide gel. Analysis of the resulting 2D gels (representative gel images in Fig. 3A,B) showed that ∼ 400 protein spots were detectable representing approximately one-sixth of the published genome , and a comparable number to that identified in previous proteomic studies of L. lactis [23–26]. These 400 protein spots probably represent the most highly expressed proteins in L. lactis as they are observed with Coomassie Brilliant Blue staining. Comparison of triplicate gels identified 86 protein spots whose intensities (normalized to the total spot volume) were significantly different between the LmrC-expressing and empty vector control cells (P < 0.05). Of these, 16 protein spots that also showed a > 1.8-fold change in expression level (staining intensity) were identified by trypsin digestion and peptide mass fingerprinting (Fig. 4, Table 1). Of the 16 protein spots analysed, four were associated with search scores of the MASCOT database that fell short of significance (> 78 in this case). The identification of these protein spots was supported in each case by further analysis. For example, for the spot assigned as GroES, three peptides were identified representing 32% of the protein sequence, and the MASCOT score was 67. The position of the spot on the 2D gel is perfectly in accordance with the pI and molecular mass of GroES, and a spot in this location has been previously identified as GroES in other L. lactis proteomic studies [25,27]. A comparable argument supports the identification of 6-phosphogluconolactonase . In addition, although the two protein spots identified as isoforms of ClpP both had MASCOT scores just below significance, the dual identification adds confidence that they do represent this protease. Finally, and consistent with the need to use western blotting to validate LmrC expression, none of the protein spots analysed had a peptide mass fingerprinting matching LmrC.
Table 1. Identification of protein expression changes in Lactococcus lactis upon LmrC expression. Proteins showing a > 1.8-fold change in expression when LmrC is expressed were identified by analysis of 2D SDS/PAGE. Fold change represents change in expression of protein when LmrC is expressed compared with nonexpressing L. lactis pNZ8048 (control). M, predicted molecular mass in kDa; pI, predicted isoelectric point; MASCOT, measure of how likely the peptides analysed correctly identify the named protein; P, probability that match is a random event, when MASCOT score > 78, P < 0.05; NA, not applicable.
Fold change ± SD
a,b,c Pairs of spots representing isoforms of the same protein.
Heat shock protein
3.2 ± 1.1
Heat shock protein
1.9 ± 0.4
Heat shock protein
2.8 ± 0.1
Heat shock protein
2.9 ± 0.4
Heat shock protein
2.8 ± 0.5
Heat shock protein
2.9 ± 0.8
Heat shock protein
2.0 ± 0.4
Heat shock protein
2.2 ± 0.3
Heat shock protein
−1.8 ± 0.1
3.7 ± 0.9
−2.6 ± 0.2
10 + 11
1.6 ± 0.4
E3 subunit of PDH
1.9 ± 0.2
−3.1 ± 0.7
−2.1 ± 0.1
−7.3 ± 0.8
−7.3 ± 0.2
We went on to consider the functional significance of the identified proteins. Notably, of the 16 protein spots showing a > 1.8-fold change in expression, nine correspond to members of stress response pathways in L. lactis. Six protein spots were assigned to the Clp protease family, the chaperone complex components dnaK and GroES were both identified and finally a universal stress response protein was identified, although its expression level was seen to decrease upon LmrC induction. The latter protein remains poorly characterized in any prokaryote , so the implication of the reduced expression remains unanswered. Notwithstanding this, it is clear that the expression of LmrC is associated with a stress response in L. lactis and that this is potentially the basis for the antibiotic resistance (see Discussion).
In this study, we examined the expression of the twin NBD protein LmrC in L. lactis and identified that it induces resistance to antibiotics whose site of action is the ribosome. Expression of Walker-A mutant isoforms of LmrC was still associated with antibiotic resistance, indicating either that the Walker-A mutant isoforms are functional or that the effect observed was not LmrC specific. We prefer the latter option for several reasons. First, the majority of Walker-A lysine → alanine substitutions made in a single NBD of an ABC transporter inhibit ATPase activity [19–21]. Second, we have made a double lysine → alanine mutation in this study and the structures of NBD dimers clearly demonstrate this being incompatible with ATP hydrolysis at either NBD . Third, although the mutant isoforms might still in theory bind ATP, albeit with weaker affinity as evidenced by comparable studies on other ABC transporters , it is the hydrolysis of nucleotide that powers antibiotic resistance in the class II ABC proteins exemplified here . Finally, we have shown in related studies that the twin NBD protein CD2593 can induce a protein-specific antibiotic resistance in the same lactococcal system, and that this is abrogated by the same Walker-A mutation (JM Dorrian and ID Kerr, unpublished results). We thus propose that the resistance elicited here is mediated by stress-response pathways induced by the heterologous expression of LmrC. How though is the stress-response pathway connected to the acquisition of antibiotic resistance?
The proteins of the heat shock response in Bacillus subtilis, the Gram-positive paradigm, can be split into three classes based on their amino acid sequence and the identity of their master regulators, although class II chaperone genes are believed to be absent in the lactococcal genome precluding further discussion . Class I chaperone genes (dnaK, dnaJ, grpE, groES, groEL) are controlled by negative regulators such as hrcA which bind to heat shock elements designated as CIRCE sequences. The induction of dnaK expression on heat shock has been documented in L. lactis  and in this study the expression of dnaK increased threefold (Table 1 and Fig. 4) when LmrC was expressed in L. lactis. DnaK has been shown to have at least two roles related to protein folding; in association with trigger factor it mediates the normal folding of proteins as they are translated, and it participates in the folding of damaged or misfolded proteins with the cooperation of ClpB [34,35]. Upon LmrC expression both ClpB and trigger factor were also upregulated, although only the former reached statistical significance, suggesting that dnaK acts here at the levels of both nascent and misfolded protein. In this study, expression of the chaperone groES increased by 2.9-fold when LmrC was expressed compared with control cultures, although GroEL was not identified in our 2D gel analysis. Previous data has also shown the induction of groES by salt stress in L. lactis [25,36].
Class III genes are controlled by the negative regulator CtsR which binds the specific repeat CtsR-box and regulates the expression of genes such as Clp (caseinolytic peptidase) family of proteins. This group contains the ATPases ClpC, ClpB and ClpE and the protease ClpP. In this study the expression of two isoforms of ClpE, ClpP and ClpB were all increased significantly upon LmrC expression (Table 1) and this upregulation is consistent with a misfolded protein response induced by antibiotic treatment, because previous data have shown that Clp mutants are sensitive to puromycin treatment . In parallel with the class I genes, ClpP induction has been demonstrated in response to heat shock  in L. lactis.
The data presented demonstrate a general increase in two stress pathways upon heterologous protein expression, a phenomenon that has been observed in other bacterial expression systems , even though here the expression level of LmrC is low enough to require western blots to identify it. A possible mechanism to explain the upregulation of stress-pathway proteins and antibiotic resistance when LmrC is heterologously expressed is presented in Fig. 5. It is proposed that expression of LmrC sequesters basal levels of heat shock proteins (dnaK, groELS, ClpC) and this acts to upregulate the cellular stress proteins (Table 1) through transcriptional control mechanisms documented in other Gram-positive organisms [39,40]. The mechanism of this regulation in L. lactis is incompletely understood but both dnaK and ClpE are documented to regulate expression of lactococcal stress-response genes [41,42]. Following upregulation of the class I and class III stress-response pathways it is predicted that exposure to some ribosomally active antibiotics is more tolerated (i.e. a degree of resistance is conferred) because the cell is ‘primed’ to resolve any protein misfolding and aggregation resulting from antibiotic treatment. Future exploration of a connection between cellular stress and antibiotic resistance will reveal whether the proteomic changes we have seen here, and which others have documented in response to mild environmental stress [23,27,32], present a general mechanism for the acquisition of antibiotic resistance.
Materials and methods
Organisms and culture conditions
Lactococcuslactis strains NZ9000 and NZ9700 were grown at 30 °C without shaking in GM17 media (M17 supplemented with 0.5% w/v d-glucose). Where transformed with pNZ8048 derived plasmids, cultures were supplemented with 5 μg·mL−1 chloramphenicol.
The cDNA of LmrC was amplified using Phusion Polymerase (New England Biolabs, Hitchin, Herts, UK) from genomic S. lincolnensis DNA using primers (forward 5′-CATCGATGAATCCATGGCTGCAGCGAGTA TTC and reverse 5′-GCGGACCACGCGTTCTAGAGCC GCCTCACTTATC), which include NcoI and XbaI sites (underlined), respectively. Purified PCR product was digested and ligated into an Escherichia coli mobilizable vector pHue  to acquire an N-terminal His6–ubiquitin tag, and thence to the lactococcal expression plasmid pNZ0848 . For site-directed mutagenesis of the Walker-A lysine residues, the template DNA was the pHue–His6–Ub–LmrC plasmid because the L. lactis plasmid DNA is incompatible with mutagenesis strategies requiring digestion of parental DNA with DpnI. Mutations (K44A, K44A/K380A) were introduced with pairs of primers, the sequence of the coding strand of which was:
K44A: 5′-CCAATGGCGCGGGAGCGAGTACTCTGT TGCGGCTCG
K380A: 5′-CCAACGGCTCGGGCGCGAGTACTCTG TTGAAGCTGATC
Mutagenesis was verified by DNA sequencing, and following this, mutant isoforms of LmrC were constructed in pNZ8048 by subcloning the entire coding region.
Antibiotic resistance assay in LmrC expressing L. lactis
Strain NZ9000 was transformed with pNZ8048–His6–Ub–LmrC by electroporation and selected by plating on GM17 (5 μg·mL−1 chloramphenicol) agar plates. Single colonies of NZ9000 transformants were used to inoculate 5 mL cultures which were incubated at 30 °C without shaking until an absorbance of 0.3 was attained. Protein expression was induced by the addition of culture supernatant from the nisin-secreting strain NZ9700 , diluted by a factor of 1/1000 for 1 h. Subsequently, culture aliquots (200 μL) were pipetted into clear-bottomed sterile 96-well plates containing antibiotic concentrations (up to 256 μg·mL−1), and incubated at 30 °C overnight. Absorbance at 600 nm was plotted as a function of antibiotic concentration and fitted using a nonlinear regression to the general dose–response equation to identify the IC50, the concentration of antibiotic to produce a 50% inhibition of growth:
where Y is the absorbance, x is the antibiotic concentration, and top and bottom represent the maximum and minimum absorbance values. All data analysis was performed in prism (Graph Pad Software, La Jolla, CA, USA).
SDS/PAGE and western blotting
Unless stated otherwise, proteins were resolved on 10% w/v polyacrylamide gels using standard methodologies. For detection of the His6 epitope a horseradish peroxidase-conjugated anti-His IgG1 (R&D Systems, Abingdon, Oxon, UK) was used at a 1/5000 dilution in NaCl/Pi supplemented with 0.1% v/v Tween. Western blots were visualized using enhanced chemiluminescence (Pierce, Rockford, IL, USA), scanned and analysed by scion image (NIH, Bethesda, MD, USA).
GM17 (5 mL) was inoculated with a colony of either L. lactis pNZ8048–His6–Ub–LmrC or L. lactis pNZ8048 (negative control) and incubated for 16 h, prior to dilution 50-fold into fresh GM17 (up to 1 L). This culture was incubated at 30 °C until an absorbance of 0.3 was attained. Nisin-containing supernatant was added to a dilution of 1/1000 and cultures were subsequently incubated for 16 h before centrifugation at 4000 g at 4 °C for 15 min. Pellets were resuspended in 50 mL ice-cold 100 mm potassium phosphate buffer, pH 7.0 and centrifuged as above. Four rounds of resuspension and pelleting were performed before a final resuspension in 1 mL 10 mm potassium phosphate buffer pH 7.0, supplemented with 1 mm MgCl2, and protease inhibitors (Type III; Merck Biosciences, Nottingham, Notts, UK). The resuspended bacteria were sonicated on ice (Jencons Vibracell; VWR International, Lutterworth, Leics, UK) at 40% power for 4 × 30 s with a 30 s cooling period. Benzonase (250 units; Promega, Southampton, Hants, UK) was added and the lysate was incubated for 37 °C for 30 min to digest chromosomal DNA. Lysates were clarified by centrifugation at 18 000 g in a microcentrifuge 4 °C for 20 min and total soluble protein was subsequently acetone precipitated from the supernatant. The protein pellet was solubilized by adding 100–200 μL of DeStreak (GE Healthcare, Amersham, Bucks, UK) buffer supplemented with 0.5% (v/v) pH 4–7 ampholyte solution (GE Healthcare) with gentle rocking for at least 2 h at room temperature. When protein was fully solubilized it was assayed for protein concentration (2D Quant kit; GE Healthcare).
Total protein (100 μg) was cup-loaded at the anode and isoelectrically focused overnight using a linear gradient across 11 cm immobilized pH gradient 4–7IPG) strips (GE Healthcare). IEF strips were washed for 3 × 5 min in 1 mL equilibration buffer (0.38 m Tris, 6 m urea, 2% (w/v) SDS, 30% (v/v) glycerol pH 8.8 HCl) containing 3.33% (w/v) dithiothreitol and 3 × 5 min in 1 mL equilibration buffer supplemented with 3.33% (w/v) iodoacetamide. After equilibration, strips were washed in 1× Laemmli protein running buffer and mounted on a 5–20% gradient acrylamide gel. A 1% w/v agarose solution in 1× Laemmli buffer with 0.1% (w/v) bromophenol blue was pipetted onto the gel strip to allow visualization of the dye front and to seal the strip in place. Proteins were resolved by applying a constant current of 40 mA per gel for ∼ 2.5 h or until the dye front had left the gel. Gels were fixed in 40% (v/v) methanol, 10% (v/v) acetic acid for 1 h at room temperature with gentle rocking, washed twice in distilled water and then stained in colloidal Coomassie Brilliant Blue stain [0.0125% (w/v) Coomassie G250, 3% (v/v) phosphoric acid, 8% (w/v) ammonium sulfate, 20% (v/v) ethanol]. Following destaining in distilled water gels were scanned using a Duoscan T1200 (Agfa, London, UK) scanner.
Analysis of 2D gels and mass spectrometric methods
Images of gels were analysed with Progenesis and SameSpots (both Nonlinear Dyamics, Newcastle, UK), enabling average gel images to be constructed from multiple repeat experiments. Protein spots showing a > 1.8-fold change in expression (equivalent to a 1.8-fold change in normalized spot volume), and a P-value of < 0.05, were excised, digested in-gel with trpysin and then analysed by MALDI-TOF MS (Micromass; Waters, Manchester, UK). Peptide mass fingerprint analysis was performed using the MASCOT interface to the SwissProt database, with a score of > 8 (equivalent to a P < 0.05) being used to identify protein spots.