Expression of OXA-Type and SFO-1 β-Lactamases Induces Changes in Peptidoglycan Composition and Affects Bacterial Fitness

Abstract

β-Lactamases and penicillin-binding proteins (PBPs) have evolved from a common ancestor. β-Lactamases are enzymes that degrade β-lactam antibiotics, whereas PBPs are involved in the synthesis and processing of peptidoglycan, which forms an elastic network in the bacterial cell wall. This study analyzed the interaction between β-lactamases and peptidoglycan and the impact on fitness and biofilm production. A representative set of all classes of β-lactamases was cloned in the expression vector pBGS18 under the control of the CTX-M promoter and expressed in Escherichia coli MG1655. The peptidoglycan composition of all clones was evaluated, and quantitative changes were found in E. coli strains expressing OXA-24, OXA-10-like, and SFO-1 (with its upstream regulator AmpR) β-lactamases; the level of cross-linked muropeptides decreased, and their average length increased. These changes were associated with a statistically significant fitness cost, which was demonstrated in both in vitro and in vivo experiments. The observed changes in peptidoglycan may be explained by the presence of residual dd-endopeptidase activity in these β-lactamases, which may result in hydrolysis of the peptide cross bridge. The biological cost associated with these changes provides important data regarding the interaction between β-lactamases and the metabolism of peptidoglycan and may provide an explanation for the epidemiology of these β-lactamases in Enterobacteriaceae.

INTRODUCTION

Antibiotic resistance is currently one of the main problems in public health. The most widespread and threatening mechanism of antibiotic resistance is the production of β-lactamases, enzymes that degrade β-lactam antibiotics. β-Lactamases exhibit diversity in both structure and function. The simplest classification (described by Ambler), based on the protein sequence, divides β-lactamases into four molecular classes (A to D) (1). Members of classes A, C, and D are serine-dependent enzymes, while members of class B are metalloenzymes that use zinc to facilitate β-lactamase hydrolysis. Moreover, the classification scheme described Bush and Jacoby (13), which is based on functional similarities (assessed by substrate and inhibitor profile), is of greater significance to physicians and microbiologists in diagnostic laboratories because it considers clinically relevant β-lactams and inhibitors.

The targets of β-lactam antibiotics are penicillin-binding proteins (PBPs). PBPs are involved in the synthesis and processing of peptidoglycan (PG) (41), which forms an elastic network in the cell wall. The main functions of this network are to resist the intracellular pressure of the cell and to maintain a defined cell shape (44).

All β-lactamases and PBPs evolved from a common ancestor, the primordial PBP, in independent and perhaps parallel processes (27), and each is specialized in different functions; PBPs display transpeptidase, carboxypeptidase, or endopeptidase activity (41), and β-lactamases inactivate β-lactam antibiotics by hydrolyzing the β-lactam ring. However, PBP5 from Pseudomonas has been described as a weak β-lactamase (23). Conversely, mutation of certain PBPs modifies the expression of β-lactamases (such as PBP4 and AmpC in Pseudomonas aeruginosa) (32). The AmpC-type β-lactamase from Citrobacter has also been suggested to display residual dd-carboxypeptidase activity in vivo, with an effect on fitness when expressed in Salmonella (31). The interaction between resistance mechanisms and bacterial fitness is a subject of increasing interest. The effect of these mechanisms on fitness is a key parameter for evaluating the epidemiological implications for a given microorganism, which depend on its ability to persist in bacterial populations once the selective pressure exerted by antibiotics disappears.

Measurement of bacterial fitness competition between bacteria expressing different β-lactamases provides a sensitive method for detecting phenotypic differences (34). Furthermore, the acquisition of plasmids with antibiotic resistance genes may have several effects on bacterial fitness (2, 25). Hence, the objectives of the present study were (i) to determine the interaction between the expression of β-lactamases and the composition and/or structure of PG and (ii) to analyze the influence of these changes in PG on fitness and biofilm formation.

MATERIALS AND METHODS

β-Lactamases: cloning experiments and determination of specific activity.

A set of isogenic clones expressing representative members of all classes of β-lactamases was constructed (Table 1). All β-lactamases were cloned between BamHI and EcoRI restriction sites into the pBGS18-pCT vector (pB) under the control of a bla_CTX-M promoter, an expression vector harboring a kanamycin resistance gene. Escherichia coli was chosen as an Enterobacteriaceae model for further studies. The E. coli MG1655 (MG) (14) electrocompetent strains were transformed with these plasmid constructions. To confirm the accuracy of the nucleotide sequences, all cloned genes were sequenced following standard procedures.

Table 1.

β-Lactamases expressed in this study

B-lactamase	Ambler class	Hydrolysis antibiotics	Host(s) most frequently encountered	Reference(s)	Oligonucleotide sequences used for cloning (5′ → 3′)a
TEM-1	A	Penicillins, early cephalosporins	Escherichia coli	8, 12	F, ggatccATGAGTATTCAACATTTTCGTGTC; R, gaattcTTACCAATGCTTAATCAGTGAGG
TEM-29	A	ESBLb	E. coli and Klebsiella pneumoniae	8, 12	F, ggatccATGAGTATTCAACATTTCCGTGT; R, gaattcTTACCAATGCTTAATCAGTGAGG
CTX-M-32 (cluster 1)	A	ESBL (100-fold more active against ceftazidime than the other CTX-Ms)	Enterobacteriaceae	15, 24, 35	F, ggatccATGGTTAAAAAATCACTGCGTCA; R, gaattcTTACAAACCGTTGGTGACGAT
CTX-M-8 (cluster 8)	A	ESBL	Enterobacteriaceae	6, 24, 35	F, ggatccATGATGAGACATCGCGTTAAG; R, gaattcTTAATAACCGTCGGTGACGAT
CTX-M-2 (cluster 2)	A	ESBL	Enterobacteriaceae	24, 35; this work	F, ggatccATGATGACTCAGAGCATTCG; R, gaattcTCAGAAACCGTGGGTTACG
CTX-M-14 (cluster 9)	A	ESBL	Enterobacteriaceae	7, 24, 35	F, ggatccATGGTGACAAAGAGAGTGCAA; R, gaattcTTACAGCCCTTCGGCGAT
SFO-1	A	ESBL	Enterobacter cloacae	16, 28	F, ggatccATGGTTAAAAATACATTACGTCAAAC; R, gaattcTCAAAGCCCTTCGGTGACAA
AmpR-SFO-1	A	ESBL	Enterobacter cloacae	16, 28	F, ggatccCTACGTATTATCTTCTGCCGC; R, gaattcTCAAAGCCCTTCGGTGACAA
VIM-1	B	Most β-lactams, except aztreonam, including carbapenems	Pseudomonas, Enterobacteriaceae	30, 45	F, ggatccATGTTAAAAGTTATTAGTAGTTTATTGG R, gaattcCTACTCGGCGACTGAGC
FOX-4	C	Penicillins, narrow-, expanded-, and broad-spectrum cephalosporins, including cephamycins	Enterobacteriaceae	10, 20, 26	F, ggatccATGCAACAACGACGTGCG; R, gaattcTCACTCGGCCAACTGACTC
OXA-10-like	D	Oxacillin, cloxacillin	Pseudomonas aeruginosa	12; this work	F, ggatccATGAAAACATTTGCCGCATATGTA; R, gaattcTTAGCCACCAATGATGCCCT
OXA-24	D	Oxacillin, cloxacillin, carbapenems	Acinetobacter baumannii	9, 29	F, ggatccATGAAAAAATTTATACTTCCTATATTCA; R, gaattcTTAAATGATTCCAAGATTTTCTAGCG

^aF, forward; R, reverse. Lowercase indicates target site for restriction enzymes.

^bESBL, hydrolysis profile that confers resistance to the penicillins, broad-spectrum cephalosporins, and monobactams but not the cephamycins or carbapenems.

To study the specific activity, the bacterial strains were grown overnight in 50 ml of Luria-Bertani (LB) broth at 37°C with shaking. The cells were harvested by centrifugation, washed once with potassium-sodium phosphate (phosphate-buffered saline [PBS]), and resuspended in 1 ml of the same buffer. To obtain the cell extract, cells were broken by ultrasonic treatment at 4°C. Cell debris was removed by centrifugation at 10,000 rpm for 10 min. The specific enzyme activity of each extract was determined by measuring the hydrolysis of a 100 μM nitrocefin solution prepared in dimethyl sulfoxide (DMSO), monitored at 25°C, in a spectrophotometer at a wavelength of 485 nm. Protein concentration was measured by the Bradford method (11).

Peptidoglycan analysis.

PG was purified from bacterial cells grown in LB broth to an optical density at 600 nm (OD₆₀₀) of 1. We added imipenem, at a concentration of 0.06 μg/ml (ca. 1/4 of the MIC), at the beginning of the culture. The bacteria were collected, suspended in 3 ml of PBS, mixed immediately in a 1:1 (vol/vol) proportion with a boiling solution of 10% sodium dodecyl sulfate (SDS) (Bio-Rad), and maintained at 100°C for 18 h. The SDS-insoluble material, containing PG, was washed until it was free of SDS by successive suspensions in distilled water and centrifuged at high speed at room temperature. PG was treated with α-amylase (100 μg/ml) at 37°C for 90 min and with pronase E (100 μg/ml) at 60°C for 60 min. The reaction was stopped by adding SDS and boiling the mixture for 20 min. The PG was further digested in 50 mM phosphate buffer (pH 4.9) with Cellosyl (Hoechst AG, Frankfurt, Germany) at a 100-μg/ml final concentration at 37°C overnight. The insoluble material was removed by centrifugation, and soluble muropeptides were reduced with sodium borohydride and frozen at −70°C. The muramidase-digested samples were analyzed by high-performance liquid chromatography (HPLC) as previously described (43). The subunits which comprise the polymer murein and were resolved by HPLC have been described previously (19, 44). The basic disaccharide peptide subunit (monomer) consists of β-1,4-linked GlcNAc and MurNAc with a peptide stem of three, four, or five residues. Cross-linked monomers with a single bond are dimers, and those with two bonds are trimers. Monomers, dimers, and trimers ending with a 1,6-anhydro-MurNAc residue are termed anhydro, and they correspond to chain ends. Subunits having a peptide bond with the Lpp protein are termed Lpp. Pentapeptides are any subunit having a five-residue peptide stem. Glycan strand chain length (chain length) was calculated as the inverse of the total proportion of anhydro muropeptides. Cross-linking was calculated as the percentage of dimers or trimers (×2) in relation to the total amount of muropeptides. In some cases (MG alone and MG with the TEM-1, VIM-1, AmpR-SFO-1, OXA-10-like, and OXA-24 β-lactamases), experiments were also done without imipenem (see Table 2).

Table 2.

Muropeptide compositions of peptidoglycans from E. coli MG1655 carrying pBGS18-pCT (plasmid alone, control strain) and different β-lactamases^a

Sampleb	Relative abundance (mol%)c						Cross-linkage (%)	Chain length
	Monomers	Dimers	Trimers	Lpp	Anhydrous forms	Pentapeptides
MG(pBGS18-pCT)	61.0 ± 0.8	36.1 ± 0.8	2.9 ± 0.1	18.0 ± 0.4	8.6 ± 1.4	0.3 ± 0.2	42.0 ± 0.9	11.9 ± 2
MG(pBTEM-1)	62.9 ± 2.8	34.9 ± 3.2	2.8 ± 0.3	16.1 ± 2.9	8.4 ± 1.1	0.8 ± 0.06	40.5 ± 3.7	12 ± 1.5
MG(pBTEM-29)	62.8 ± 2.6	33.5 ± 3.4	3.8 ± 0.8	14.8 ± 4.1	10.9 ± 1.1	0.5 ± 0.02	41.1 ± 1.8	9.2 ± 0.9
MG(pBCTX-M-32)	63.1 ± 0.2	34 ± 0.3	2.7 ± 0.03	14.1 ± 0.9	7.0 ± 0.04	0.2 ± 0.02	39.6 ± 0.1	14.3 ± 0.07
MG(pBCTX-M-8)	64 ± 0.4	33.6 ± 0.5	2.4 ± 0.03	13.1 ± 1.8	5.4 ± 0.3	0.23 ± 0.01	38.4 ± 0.4	18.3 ± 1.0
MG(pBCTX-M-2)	63.3 ± 0.5	33.7 ± 0.4	2.9 ± 0.9	14.9 ± 1.7	6.4 ± 0.2	0.3 ± 0.1	39.7 ± 1.3	15.7 ± 0.5
MG(pBCTX-M-14)	62.5 ± 0.5	34.8 ± 0.3	2.7 ± 0.2	15.6 ± 2.1	6.9 ± 0.7	0.1 ± 0.001	40.3 ± 0.7	14.7 ± 1.5
MG(pBSFO1)	61.3 ± 1.3	35.1 ± 0.7	3.7 ± 0.6	15.4 ± 0.9	12.1 ± 3	0.5 ± 0.3	42.4 ± 1.9	8.8 ± 2.8
MG(pBAmpR-SFO1)	68.7 ± 0.2	29.4 ± 0.2	1.9 ± 0.02	12.7 ± 0.01	3.3 ± 0.1	0.1 ± 0.02	33.3 ± 0.2	29.8 ± 1.2
MG(pBVIM-1)	61.2 ± 2.1	35.8 ± 1.5	2.9 ± 0.6	15.5 ± 2.3	6.4 ± 1.0	0.2 ± 0.1	41.8 ± 2.7	15.9 ± 2.5
MG(pBFOX-4)	63.7 ± 0.5	33.8 ± 0.7	2.4 ± 0.2	12.2 ± 1.6	5.8 ± 0.1	0.3 ± 0.04	38.6 ± 0.4	17.2 ± 0.3
MG(pBOXA-10-like)	65.9 ± 0.1	31.8 ± 0.1	2.3 ± 0.02	12.2 ± 0.1	4.9 ± 0.8	0.2 ± 0.2	36.3 ± 0.1	20.7 ± 3
MG(pBOXA-24)	68.9 ± 1	28.8 ± 0.9	2.2 ± 0.2	12.2 ± 2.3	3.6 ± 0.2	0.1 ± 0.03	33.3 ± 1.2	28.1 ± 0.5
MG(pBGS18-pCT)d	69.5 ± 0.7	28.3 ± 0.6	2.3 ± 0.12	11.8 ± 1.35	4.3 ± 0.1	0.1 ± 0.1	32.8 ± 0.8	23.4 ± 0.7
MG(pBTEM-1)d	68.1 ± 0.2	29.4 ± 0.4	2.5 ± 0.2	13.3 ± 0.6	4.9 ± 0.3	0.1 ± 0.1	34.4 ± 0.1	20.5 ± 1.2
MG(pBAmpR-SFO1)d	69.1 ± 1.6	28.7 ± 1.5	2.3 ± 0.2	12.1 ± 1.3	4.9 ± 0.7	0.1 ± 0.04	33.2 ± 1.8	20.5 ± 2.8
MG(pBVIM-1)d	69.2 ± 1.0	28.4 ± 0.6	2.35 ± 0.4	11.9 ± 0.9	4.5 ± 1.8	0.1 ± 0.01	33.2 ± 1.3	20.1 ± 3.7
MG(pBOXA-10-like)d	74.2 ± 5.9	25.9 ± 2.8	2.4 ± 0.5	11.6 ± 0.6	4.8 ± 2.2	0.07 ± 0.01	32.6 ± 0.8	21.5 ± 3.5
MG(pBOXA-24)d	69.2 ± 0.8	28.5 ± 0.5	2.3 ± 0.3	9.9 ± 3.0	3.8 ± 0.2	0.2 ± 0.1	33.1 ± 1.1	26.5 ± 1.2

^aValues are means and standard deviations from three independent experiments. Boldface indicates the isolates in which changes in the PG composition have been described.

^bPeptidoglycan of E. coli MG1655 transformed with the indicated plasmid.

^cMuropeptides are grouped according to structural similarities, differences in PG composition, and structure of strains.

^dData were obtained without the presence of imipenem.

TEM.

For analysis by transmission electron microscopy (TEM), the strains with the different plasmid constructions were cultured overnight in the presence of 0.06 μg/ml of imipenem. Cells were then harvested and resuspended in 4% cold glutaraldehyde diluted in 0.1 M phosphate buffer (pH 7.4), allowed to stand for 30 min at room temperature, and fixed in 1% osmium tetroxide in the same buffer for 1 h at 4°C. Pellets were then dehydrated in an acetone gradient and embedded in Spurr resin. Ultrathin sections (70 nm) were examined and photographed using a JEOL JEM 1010 TEM (80 kV).

Bacterial growth.

All bacterial strains were grown in LB broth at 37°C and 180 rpm. Growth was monitored by measuring the OD₆₀₀. The growth rate (μ) was calculated on the basis of the exponential segment of the growth curve and defined as ln₂ g⁻¹, where g is the doubling time of an exponentially growing culture (3). Experiments were done in triplicate.

In vitro competition experiments (in vitro fitness).

Experiments were performed to measure the in vitro competition between β-lactamase clones and E. coli MG1655 with the pBGS18-pCT vector [MG(pB)]. Exponentially growing cells of the corresponding transformant and control strain were mixed in a 1:1 proportion and resuspended in 0.9% saline solution. Approximately 10³ cells from each mixture were inoculated into 10-ml flasks of LB broth and grown at 37°C and 180 rpm for 16 to 18 h, which corresponds to approximately 20 cell generations. Serial 10-fold dilutions were plated in duplicate onto LB agar (LBA) with 50 μg/ml of kanamycin (LBA-K) and LBA with 50 μg/ml of kanamycin and ampicillin (LBA-K/A) in order to determine, respectively, the total number of CFU and the CFU of the clones harboring β-lactamases, after overnight incubation at 37°C. The competition index (CI) was defined as the ratio between the CFU of the strain harboring a β-lactamase and the MG(pB) strain. The CI values were calculated for each independent competition assay, and the median values were calculated too. Statistical analysis of the distribution of the CI values was performed with the Mann-Whitney U test. Differences were considered statistically significant at a P value of <0.05 (33).

Fitness in the mouse model of systemic infection (in vivo fitness).

In vivo fitness was assessed by competition experiments with a mouse model of systemic infection. For this purpose, mixtures of each of the strains containing approximately 5 × 10⁶ exponentially growing cells were inoculated intraperitoneally into five female C54 mice, of about 18 to 20 g in weight, previously treated with cyclophosphamide (100 mg/kg) for 3 days. The mice were sacrificed at 24 h after inoculation, and their spleens were aseptically extracted and homogenized in 1 ml of 0.9% saline solution in a Retsch MM200 mixer mill. The number of CFU of each strain and the CI values were determined as described for in vitro competition experiments.

Static biofilm assays.

Static assays of biofilm formation, by determining adherence to polystyrene, were performed as described previously (39) with some modifications. Isolates of E. coli were grown overnight in 5 ml LB broth (37°C, 180 rpm). After overnight incubation, E. coli cultures were standardized to an optical density of 0.2 at 600 nm. All standardized cultures were inoculated 1/20 in 150 μl of M9 with glucose (0.2%) in a 96-well polystyrene microtiter plate (U bottom; Soria Genlab S.A.) Test plates were transferred to plastic bags to avoid evaporation of medium and incubated at 37°C for 48 h without shaking. Eight wells were inoculated for each strain. After removal of planktonic cells, the wells were rinsed with 0.9% sodium chloride and biofilms were stained with 200 μl of 0.2% crystal violet (CV). After 15 min, the plate was rinsed with 0.9% sodium chloride. The amount of biofilm formed in each well was determined spectrophotometrically after solubilization of the CV retained with 200 μl of an 80:20 mixture of ethyl alcohol and acetone. The absorbance of CV was measured at 570 nm in an automated 96-well plate reader.

RESULTS

Clones carrying the different β-lactamases were obtained (Table 2); MG carrying pB, without any β-lactamase, was used as a negative control. Although the OXA-10-like enzyme has two amino acid changes compared to the wild-type OXA-10 (Met99Tyr and Gly212Met), kinetic parameters, such as K_cat and K_m, toward ampicillin and oxacillin were the same for the OXA-10-like enzyme and wild-type OXA-10 (data not shown). No imipenem hydrolysis was detected with this OXA-10-like variant. The relative specific activities of MG carrying pB-AmpR-SFO-1 or pB-SFO-1 against nitrocefin were 520 ± 9 μmol min⁻¹ mg⁻¹ and 140 ± 16 μmol min⁻¹ mg⁻¹, respectively. The results revealed higher hydrolytic activity against nitrocefin when the SFO-1-β-lactamase gene was cloned with the upstream ampR gene (pB-AmpR-SFO-1 construction), which is consistent with previous findings (28).

The expression of all β-lactamases was checked by determination of the MICs of specific antibiotics, which depend on the characteristic hydrolytic profile of each β-lactamase. All clones showed appropriate susceptibility antibiotic profile according to previously reported data, thus demonstrating proper expression of all β-lactamases. Imipenem MICs were identical with all strains (0.25 μg/ml) except MG expressing the VIM-1 and OXA-24 β-lactamases, which showed a 1-dilution increase in MIC.

Effect of β-lactamase production on E. coli PG composition.

The steady-state length distribution of glycan strands in E. coli is extremely broad and depends on the strain, conditions, and growth phase (44). Representative β-lactamases of all classes were therefore cloned into a vector with identical genetic environment, and extraction and analysis of PG were performed under identical experimental conditions in three independent experiments. The PG compositions of E. coli producing different β-lactamases are detailed in Table 2.

No major qualitative differences were found in the PG from any of the strains harboring different β-lactamases. However, some quantitative changes in MG1655 strains carrying pB-OXA-10-like, pB-OXA-24, and pB-AmpR-SFO-1 were found, showing a decrease in the levels of dimers and trimers, anhydrous muropeptides, and pentapeptides and an increase in monomers and lipoprotein-bound muropeptides (Table 2). In addition, the level of cross-linked muropeptides decreased and the average length increased in these strains. The muropeptide compositions of PG in the remaining strains were almost identical to that of the control E. coli MG1655 with pBGS18-pCT vector [MG-(pB)]. For comparative purposes, data for with MG1655 expressing the TEM-1, OXA-10-like, VIM-1, AmpR-SFO-1, and OXA-24 enzymes without imipenem are also shown (Table 2).

TEM.

Transmission electron microscopy (TEM), in which an electron beam is passed through ultrathin sections and produces an image of the sample, has been used to investigate the effects of antibiotics on the cell wall, and morphological and structural changes have been observed (42). The morphologies of MG1655 strains carrying recombinant vectors pB-AmpR-SFO-1, pB-OXA-10-like, pB-OXA24, and pBGS18-pCT were analyzed by TEM. Despite the quantitative compositional changes in the PGs of these strains, it was not possible to observe any ultrastructural changes in the cell wall by TEM (Fig. 1).

Fig 1.

Transmission electron microscopy analysis of E. coli MG1655 strains harboring pBOXA-10-like and pBAmpR-SFO1 recombinant plasmids and of control strain MG(pBGS18-pCT).

Growth curves.

A growth curve experiment was performed to assess the growth rates under noncompetitive conditions and therefore to check the effects of carrying the plasmids on the E. coli growth rate. The curve patterns were similar for all strains (Fig. 2). All strains showed a similar growth rate (μ), with values ranging from 0.021 to 0.023, regardless of the expressed β-lactamases.

Fig 2.

Growth curves for E. coli MG1655 harboring different β-lactamases in LB medium. Data are mean values of three measurements. For simplification, error bars have been omitted, although in all cases the standard deviation was <10%.

In vitro competition experiments (in vitro fitness).

In an antibiotic-free environment, bacterial strains carrying plasmids harboring resistance genes compete with those strains that do not possess such resistance genes. The outcome of the competition process depends on the relative fitness, defined as the efficacy of multiplication of the resistant cell compared with that of the susceptible cell. The presence of plasmids encoding CTX-M β-lactamases, such as pB-CTX-M-32 and pB-CTX-M-8, pB-TEM-1, pB-VIM-1, and pB-FOX-4, was not associated with a significant fitness cost with respect to E. coli MG(pB) in vitro. However, we observed a marked and statistically significant decrease in fitness competition for E. coli MG1655 carrying recombinant plasmids pB-AmpR-SFO-1, pB-OXA-10-like, and pB-OXA-24, as shown by the competition index (CI) results in Fig. 3 (P < 0.05 in all cases). The greatest decrease was observed for the MG(pB-OXA-10-like)/MG(pB) competition, with a median CI of 0.3, followed by the MG(pB-OXA-24)/MG(pB) and MG(pB-AmpR-SFO-1)/MG(pB) competitions, with median CIs of 0.43 and 0.55, respectively.

Fig 3.

Results of in vitro competition experiments. The CI values obtained for each of the eight independent experiments are plotted, with the median CI values shown by horizontal lines.

It is remarkable that those β-lactamases that showed changes in the quantitative composition of PG were clearly and directly associated with a reduction in fitness cost (Table 2 and Fig. 3). Overall, the data showed that the median CI values for the MG1655 strains carrying plasmids pB-AmpR-SFO-1, pB-OXA-10-like, and pB-OXA-24 were associated with a statistically significantly decreased fitness in the in vitro experiments.

Fitness in the mouse model of systemic infection (in vivo fitness).

Since results obtained in vitro and in vivo do not always correspond and since an animal model is much closer to what occurs in the human body at the infection site, we assessed the competition assays in a mouse model of systemic infection. The results were similar to those obtained in vitro (Table 3). MG(pB-OXA-10-like) and MG(pB-AmpR-SFO-1) were associated with a significantly increased (P < 0.001) biological cost, and MG(pB-TEM-1) and MG(pB-SFO-1) did not show any biological cost effect since fitness was the same as in the MG(pBGS18-pCT) control strain. In the case of OXA-24, an effect similar to that obtained in the in vitro assays was observed (data not shown). It is important to emphasize that the effect on biological cost obtained with SFO-1 with its upstream AmpR regulator is abrogated when the β-lactamase gene lacks its AmpR positive regulator, thus demonstrating that the effect is directly related to the β-lactamase gene expression.

Table 3.

Results of in vivo competition experiments in the mouse model of systemic infection

Competition	Median CI (range)
TEM-1/pBGS18	1.15 (1.01–1.35)
SFO-1/pBGS18	1.08 (0.8–1.5)
AmpR-SFO1/pBGS18	0.3 (0.2–0.51)a
OXA-10–like/pBGS18	0.32 (0.25–0.68)a

^aSignificant (P < 0.001).

Biofilm production.

In order to assess whether the changes in PG composition had any effect on biofilm formation, those strains in which changes in PG composition had been observed were assayed for biofilm formation. All the studied strains formed less biofilm than the control strain carrying the pBGS18-pCT vector (Fig. 4).

Fig 4.

Effects of β-lactamases on biofilm formation. Static biofilm formation by E. coli MG1655 on polystyrene is shown. Results are the mean values from eight experiments in the presence of 0.06 μg/ml imipenem.

DISCUSSION

Understanding the physiological basis of fitness costs in isolates carrying β-lactamases is clinically relevant, since β-lactams are the most frequently used treatment for bacterial infections and β-lactamases constitute the main mechanism of resistance to β-lactams in Gram-negative bacilli. The main aim of this study was to assess biological changes in PG composition in members of Enterobacteriaceae expressing β-lactamases in relation to their fitness. In this respect, there is some information on changes in the PG composition of Salmonella strains producing AmpC and their effect on virulence (31), although their relationship to fitness has not been described. It is difficult to address this topic because of the diversity of genetic factors, which may affect the final result. The experiments presented here were therefore performed on isogenic strains and under identical conditions, so that the results reflect the precise relationship between β-lactamases, PG composition, and fitness.

An effect of β-lactamases on PG composition was observed when imipenem was added at subinhibitory concentrations. This low imipenem concentration was sufficient to increase the sensitivity of the assay induced by the expression of β-lactamases and to observe its effect on PG composition. As a consequence, the interaction between different β-lactamases and PG was magnified, and changes were observed when the recombinant strains carrying β-lactamases were compared with the E. coli control strain without any β-lactamase.

The binding of the cell wall donor strand (-d-Ala-d-Ala) to the active-site serine of the PBP can be mimicked by the interactions of β-lactams with the same enzymes. Therefore, the β-lactamases, which bind the β-lactam antibiotics, could also bind the donor strand of the cell wall analogously to a dd-peptidase. Among all β-lactamases tested, only OXA types and AmpR-Sfo1 yielded any apparent effect, possibly due to a cell wall donor binding activity, which could be the explanation for the reduction of cross-linking, due to the lack of terminal d-Ala that is needed for the cross-linking. However, a direct dd-endopeptidase activity on the mature PG or dd-carboxypeptidase activity on the cell wall donor strand cannot be excluded (41).

Production of the different β-lactamases had no apparent effect on the growth rate of E. coli under the noncompetitive experimental conditions (rich medium, LB broth, and no antibiotics), with similar growth rates being observed. Competition in vitro and in a mouse model showed that fitness decreased with the OXA-type (OXA-10-like and OXA-24) and AmpR-SFO-1 β-lactamases. The effect on fitness and on PG composition was observed only when the SFO-1 β-lactamase was cloned under the control of the ampR gene, since greater hydrolytic activity was detected in strains carrying AmpR-SFO-1 than in clones harboring only SFO-1. Although the regulation of the AmpR-mediated expression of AmpC in E. coli seems to be simpler than that described in Pseudomonas (21), it is not possible to exclude completely a putative effect due the presence of AmpR in strain MG(pB-AmpR-Sfo1), but neither there is any evidence for that. Therefore, the effect of AmpR must be due essentially to the increase in expression, acting as an upregulator of SFO-1 β-lactamase (28).

These data showed that expression of some OXA-type and AmpR-SFO-1 β-lactamases in an E. coli strain entails a fitness cost, and they also provide some knowledge about potential weakness of selection of these resistant bacteria when expressing these genetic determinants of antibiotic resistance. The biological cost is thought to be the main driving force for reduction in the frequency of resistant bacteria in the absence of antibiotics (22).

To our knowledge, no clear explanation has been given as to why some β-lactamases are often isolated from a specific genus, whereas others are scarcely detected in a specific microorganism. Identification of differences in the impacts of different β-lactamases expressed in members of the Enterobacteriaceae may help in understanding why some β-lactamases are more widespread in some Gram-negative bacilli than in others.

While most extended-spectrum β-lactamases (ESBLs) have been found in E. coli, Klebsiella pneumoniae, and other Enterobacteriaceae, most of the OXA-type ESBLs and carbapenemases have been mainly found in P. aeruginosa and Acinetobacter spp. Carbapenem-hydrolyzing OXA-type lactamases are often found in Acinetobacter baumannii (46), OXA-type lactamases with an extended spectrum in P. aeruginosa, and the OXA-48 carbapenem-hydrolyzing enzyme in members of the Enterobacteriaceae (37). Until now, the only OXA-10-like lactamase described in Enterobacteriaceae was OXA-101 (38). The SFO-1 β-lactamase has been isolated only from Enterobacter cloacae (16, 28), and because of its plasmid location, it could be transmitted to other bacteria. To our knowledge, no OXA-24 has so far been described in Enterobacteriaceae.

The SFO-1 is the only class A β-lactamase with the ampR gene upstream (28). Few reports have described codified class C plasmids (4, 17, 36, 40) with the upstream ampR gene acting to induce expression. With all these data, it could be interesting to propose a hypothesis in which the low dissemination of these enzymes in Enterobacteriaceae may be related to their effects on changes in PG and therefore to the decrease in fitness.

Although resistance traits encoded by plasmids are usually associated with a fitness cost (5), we observed that some β-lactamases did not affect fitness, which means that this relationship depends on the type of β-lactamase expressed, at least as assayed in our isogenic model. Hypothetically and based on the decrease in biological fitness caused by OXA-10-like, OXA-24, and AmpR-SFO-1 β-lactamases in the absence of antibiotics, the strains expressing these enzymes probably would not be able to dominate the bacterial population. According to the above-described hypothesis, in the absence of antibiotics these strains will be displaced and the frequency of isolation will be lower. Interestingly, the in vitro data were confirmed in an in vivo model of mouse systemic infection.

No relationship between biofilm production and PG modifications was observed, since the expression of AmpR-SFO-1, OXA-24, and OXA-10-like genes decreased biofilm production in the same way as the TEM-1 β-lactamase, with which no changes in the PG were observed. The decreased in biofilm formation has previously been attributed to the presence of class A and class D β-lactamases, specifically TEM-1 and OXA-3 (18). It has been hypothesized that the biofilm phenotypes of class A and class D β-lactamase-expressing transformants may be brought about by the ability of these β-lactamases to bind to or to modify PG (18), although this was not demonstrated in our model.

In conclusion, the expression of some specific β-lactamases may be associated with changes in the cell wall of the bacterial host, which are translated into a loss of fitness in vivo and in vitro. This may explain the “natural” selection of β-lactamases in a specific bacterial genus and may provide important data regarding the epidemiology of β-lactamases and the low rate of isolation in specific genus. Furthermore, the interaction with the PG observed with some of the β-lactamases studied, such as the AmpR-SFO-1, OXA-10-like, and OXA-24 β-lactamases, supports the hypothesis that the PBPs and β-lactamases have evolved from a common gene ancestor (27).