Cryptic β-Lactamase Evolution Is Driven by Low β-Lactam Concentrations

Abstract

Our current understanding of how low antibiotic concentrations shape the evolution of contemporary β-lactamases is limited. Using the widespread carbapenemase OXA-48, we tested the long-standing hypothesis that selective compartments with low antibiotic concentrations cause standing genetic diversity that could act as a gateway to developing clinical resistance. Here, we subjected Escherichia coli expressing bla_OXA-48, on a clinical plasmid, to experimental evolution at sub-MICs of ceftazidime. We identified and characterized seven single variants of OXA-48. Susceptibility profiles and dose-response curves showed that they increased resistance only marginally. However, in competition experiments at sub-MICs of ceftazidime, they demonstrated strong selectable fitness benefits. Increased resistance was also reflected in elevated catalytic efficiencies toward ceftazidime. These changes are likely caused by enhanced flexibility of the Ω- and β5-β6 loops and fine-tuning of preexisting active site residues. In conclusion, low-level concentrations of β-lactams can drive the evolution of β-lactamases through cryptic phenotypes which may act as stepping-stones toward clinical resistance.IMPORTANCE Very low antibiotic concentrations have been shown to drive the evolution of antimicrobial resistance. While substantial progress has been made to understand the driving role of low concentrations during resistance development for different antimicrobial classes, the importance of β-lactams, the most commonly used antibiotics, is still poorly studied. Here, we shed light on the evolutionary impact of low β-lactam concentrations on the widespread β-lactamase OXA-48. Our data indicate that the exposure to β-lactams at very low concentrations enhances β-lactamase diversity and drives the evolution of β-lactamases by significantly influencing their substrate specificity. Thus, in contrast to high concentrations, low levels of these drugs may substantially contribute to the diversification and divergent evolution of these enzymes, providing a standing genetic diversity that can be selected and mobilized when antibiotic pressure increases.

Keywords: Escherichia coli; OXA-48; carbapenem; carbapenemase; catalytic efficiency; ceftazidime; cryptic evolution; resistance development; structural flexibility; sub-MIC.

FIG 1

Population-level effects of sub-MIC ceftazidime exposure. (a) Experimental design. (b) IC₅₀ fold change for populations evolved without (gray) and under sub-MIC ceftazidime conditions (violet), relative to wild-type OXA-48. Bands represent the standard deviation around the geometric mean. (c) The top section shows the fraction of clones able to grow on ceftazidime, 1 mg/liter (>2-fold MIC); the bottom section displays MIC fold change distributions of preselected clones. Box plots represent quartiles and the median of the distributions.

FIG 2

Phenotypic and structural investigation of OXA-48 allele variants. (a) Relative genotype frequencies of bla_OXA-48 variants within Pop4 to 6, at 50 and 300 generations. (b) Relative growth advantage of OXA-48 variants expressed at sub-MICs of ceftazidime versus their ceftazidime IC₅₀. Despite marginal changes in their ceftazidime susceptibility (IC₅₀ increased by 2- to 4-fold), the expression of these alleles displays large fitness benefits at sub-MIC ceftazidime. Error bars represent the standard deviation. (c) Ribbon structure of OXA-48 including the amino acid changes close to the active site. (d) Representative structures from molecular dynamics simulations of wild type, P68S, F72L, and L158P performed with ceftazidime covalently bound to the active site S70. In short, S68 in P68S displays an H-bond with the tyrosine in the conserved Y¹⁴⁴GN motif of OXA-48. F72L lacks the aromatic stacking interaction between F72 and F156/W157. L158P disrupts the H-bond network within the Ω-loop.

FIG 3

Head-to-head competitions. (a) E. coli MG1655 mal⁺ competed with MG1655 ΔmalF expressing wild-type and allele variants of OXA-48, respectively, without (gray) and at sub-MICs (violet) of ceftazidime. While expression without selection pressure was neutral for all alleles, at sub-MICs, all allele variants showed fitness benefits over the wild-type allele. (b) E. coli MG1655 mal⁺ expressing F72L versus ΔmalF F72L/G131S and mal⁺ L158P versus ΔmalF N146S/L158P. G131S and N146S did not improve bacterial fitness at sub-MIC ceftazidime. The dots represent biological replicates, and significantly different averages compared to OXA-48 in the presence of ceftazidime (0.06 mg/liter) are marked with * (P < 0.05), ** (P < 0.01), and *** (P < 0.001).

FIG 4

Differences in dynamics between wild-type OXA-48 and variants P68S, F72L, and L158P (as well as OXA-163). Loop flexibility as measured by backbone root mean square fluctuations (RMSFs) from molecular dynamics simulations for residues in the Ω-loop (a) and in the b5-b6 loop (b). Principal component (PC) analysis of the Cα atoms from molecular dynamics simulations of OXA-48 and the P68S, F72L, and L158P variants, with PC1 (c, top) and PC5 (d, top) illustrated on ribbon structures using arrows indicating the direction of the eigenvectors and the magnitude of the corresponding eigenvalue (for clarity, arrows are only shown for atoms with eigenvalues of >2.5 Å). Normalized histograms (using 200 bins per enzyme) for PC1 (c, bottom) and PC5 (d, bottom) indicate differences of the range of the PC sampled in different variants.

FIG 5

Ceftazidime binding pocket (top panel), the 1st shell residues interacting with S70 (middle) and the mutational site (bottom panel) of L67F (red) compared to the wild-type structure of OXA-48, shown in gray (PDB no. 3HBR) (15). The crystal structure of L67F was solved to 1.9 Å and displayed hydrolyzed ceftazidime ∼9 Å away from the active site S70. (Top) Binding pocket of L67F without (left, chain C) and with (right, chain A) ceftazidime compared to wild-type OXA-48. For ceftazidime, no 2Fo-Fc electron density was detected for the R2 group. (Middle) Superimposition of the first shell residues of L67F (chain A) around the active site S70 compared the wild-type structure. (Bottom) Investigation of the mutational site, shown as the first shell residues around L67F (both chain A and C), compared to wild-type OXA-48. Displacements of L158 (1 Å) and I215 (2 Å) in the L67F structure are indicated with arrows. W221 was flipped 180° in the L67F structure.

FIG 6

Schematic representation of hydrolyzed ceftazidime in front of the active site of the OXA-48 variant L67F (based on PDB no. 7ASS). The ceftazidime side chains R1 and R2 are labeled and marked. For R2, no electron density was observed and therefore no interactions were detected. Hydrogen bonds from ceftazidime to D101, Q124, T213, and R214 are represented with dashed lines. The ionic interactions with R214 are indicated with arrows.