Nasal Resistome Development in Infants With Cystic Fibrosis in the First Year of Life

Abstract

Polymicrobial infections of the respiratory tract due to antibiotic resistant bacteria are a great concern in patients with cystic fibrosis (CF). We therefore aimed at establishing a functional metagenomic method to analyze the nasal resistome in infants with CF within the first year of life. We included samples from patients before antibiotic treatment, which allowed obtaining information regarding natural status of the resistome. In total, we analyzed 130 nasal swabs from 26 infants with CF and screened for β-lactams (ampicillin, amoxicillin-clavulanic acid, and cefuroxime) and other classes of antibiotic resistances (tetracycline, chloramphenicol and trimethoprim-sulfamethoxazole). For 69 swabs (53% of total), we found at least one non-susceptible phenotype. Analyses of the inserts recovered from non-susceptible clones by nanopore MinION sequencing revealed a large reservoir of resistance genes including mobile elements within the antibiotic naïve samples. Comparing the data of the resistome with the microbiota composition showed that the bacterial phyla and operational taxonomic units (OTUs) of the microbiota rather than the antibiotic treatment were associated with the majority of non-susceptible phenotypes in the resistome. Future studies will reveal if characterization of the resistome can help in clinical decision-making in patients with CF.

Keywords: antibiotics; beta-lactamases; functional metagenomics; nasal swabs; resistance.

FIGURE 1

Overview of the functional metagenomic approach, samples collection and patient history with antibiotic therapy. (A) Schematic representation of the experimental workflow. (B) We used 130 nasal swabs from 26 infants with CF enrolled in a cohort study. We chose samples containing a decent amount bacterial DNA (≥40 ng/μL 16S rRNA PCR product) for functional metagenomic analysis. Samples are represented according to patient antibiotic history at the time of sampling; circles: antibiotic naïve, squares with red border: during an antibiotic treatment and squares: after one or more antibiotic therapy.

FIGURE 2

Network of all proteins identified conferring antitiotic resistance. We identified 171 resistance proteins and grouped them according to the main classes of antibiotics observed: β-lactamases (with Ambler classification), dihydrofolate reductases (DHFR, folA or dfrG), thymidylate synthase, ribosomal protection protein, acetyltransferase, transporters and efflux pumps (ABC, MSF or others) and non-described proteins. The respective non-susceptible phenotypes were recorded and a network was built using Cytoscape (3.5.1). Each node represents a protein type identified and is connected to the antibiotics for which it presented non-susceptibility. Size of the node represents the frequency of identification.

FIGURE 3

Phylogenetic tree of predicted β-lactamases identified. Predicted protein sequences were obtained using ORFfinder (NCBI) (≥50 amino acids in length) using the genes received from functional metagenomics sequencing. β-lactamases (n = 64) were assigned to an Ambler class according the best hit from blast against NCBI. Sequences with less than 85% identity over 90% of sequence to known β-lactamase were classified as non-described proteins. All sequences were aligned with MUSCLE (software MEGA, 5.2.1) and an approximately-maximum-likelihood phylogenetic tree was built using FastTree (http://www.microbesonline.org/fasttree/, July 2017).

FIGURE 4

Non-susceptible phenotypes were not associated with AB treatment history. After functional metagenomics analysis of all samples (n = 130), proportions of non-susceptible phenotypes observed for all AB tested were calculated according to the AB status of the patient: (A) AB naïve vs. non-naïve or (B) AB naïve (group 0), collected during an AB treatment (group 1), collected after a single AB treatment (group 2), or collected after more than one AB therapy (group 3). A χ² test for a 2 × 2 or 2 × 4 contingency tables was applied for statistical significance. (A) Non-susceptible phenotypes were more frequently recovered from patient with AB history. (B) We observed an increasing trend for β-lactams resistance in patients who received one or more AB therapies. Standard deviations are indicated.

FIGURE 5

Antimicrobial resistance genes (ARG) and the related phyla. We related all identified ARG (n = 171) to the closest bacterial phyla using RAIphy on the contigs sequences obtained. Protein sequences were grouped according the main following families: β-lactamases (with Ambler classification), dihydrofolate reductases, thymidylate synthase, ribosomal protection protein, acetyltransferase, transporters/efflux pumps (ABC, MSF or others) and non-described proteins. We reveal the Proteobacteria, Firmicutes, Bacteroidetes, Actinobacteria, while other phyla are grouped into the “others.” (A) The majority of ARG came from Proteobacteria and Firmicutes. (B) Classes A and B β-lactamases were related to Proteobacteria, Firmicutes or Bacteroidetes while class C and D were exclusively related to Proteobacteria.

FIGURE 6

Beta-diversity analyses of all samples (n = 130) from infants with CF for ampicillin resistance using non-metric multidimensional scaling (nMDS). Illustrated are (A) results for abundance based distance matrix (permutational multivariate ANOVA P = 0.04) and (B) binary based distance matrix (permutational multivariate ANOVA P = 0.001). Ampicillin-non-susceptible samples were more similar to each other than compared with susceptible sample. Red; Sample is ampicillin-non-susceptible. Black; Sample is susceptible toward ampicillin. Ordispider plots are illustrated which combine items to their centroid.

FIGURE 7

Comparison of the functional resistome with shotgun metagenomics sequencing for three samples. Functional resistome analyses revealed 18 genes for sample ID CF1 (2025.1–2025.18), 1 gene for CF2 (2026.1) and 5 genes for CF3 (2031.1-2031.5). The plots illustrate the sequence coverage from the shotgun sequencing for the genes identified by functional genomics. We recovered reads from shotgun metagenomics sequencing for 18/18 (for ID 2025), 0/1 (for ID 2026), and 5/5 (for ID 2031) resistance genes. Certain resistance genes (e.g., 2025.1, 2025.2, and 2025.3) were found in multiple samples [indicated as CF1 (ID 2025), CF2 (ID 2026) and CF3 (ID 2031)].