Patient recruitment

Healthy infants attending a suburban, demographically diverse pediatric practice were enrolled at birth and prospectively followed at the 2-, 6-, and 12-month well-child visits. At birth and at each visit, a stool sample was collected and refrigerated for up to 3 days before being stored at −80°C for batch testing. These handling procedures resulted in <10% decrease in cultivatable C. difficile colony numbers compared with samples cultured on the day the stool was produced.

A contemporaneous set of sequential adult stool samples submitted to the clinical microbiology laboratory at University Hospitals Cleveland Medical Center, located ~9.65 km (~6 miles) from the pediatric practice, was used for comparison. Samples were identified as C. difficile–positive by commercial polymerase chain reaction assay (PCR) and stored at −80°C. A subset of these samples was categorized by the hospital infection prevention service as “community onset” according to National Healthcare Surveillance Network case definitions ^{Reference McDonald, Gerding and Johnson9} ; these were deidentified and selected for further analysis.

Microbiology and ribotyping

Infant stool samples were tested for presence of free toxin using a commercially available enzyme immunoassays (EIAs) for toxins A and B (Abbot Laboratories, Abbot Park, IL). Infant and adult samples were cultivated for C. difficile using previously defined methods. ^{Reference Riggs, Sethi, Zabarsky, Eckstein, Jump and Donskey10} In short, frozen stool was transferred to an anaerobic chamber and thawed for 10 minutes at 37°C. A 10-mg aliquot of thawed stool suspension was plated onto prereduced C. difficile brucella agar containing 0.1% taurocholic acid and 5 mg/L lysozyme. ^{Reference Nerandzic and Donskey11} Putative C. difficile colonies were selected based on typical morphology and identity was confirmed by a latex agglutination assay (Microgen Bioproduct, Surrey, UK). We performed PCR amplification of the tcdC and cdtB genes to identify toxigenic C. difficile isolates and binary toxin-positive isolates, respectively. ^{Reference Riggs, Sethi, Zabarsky, Eckstein, Jump and Donskey10}

The ribotype of each isolate was determined by fluorescent PCR ribotyping according to previously published methods. ^{Reference Martinson, Broadaway and Lohman12} In short, genomic DNA was extracted by heating an overnight culture to 95°C for 15 minutes and amplified by PCR using the primers, conditions, and program reported by Martinson et al. ^{Reference Martinson, Broadaway and Lohman12} Fluorescent amplicons were then sequenced for fragment analysis by Capillary Electrophoresis at the University of Michigan Biomedical Research Core facility. Fluorescent ribotypes were designated using the Montana State University Fluorescent PCR Ribotyping Pipeline reference library. ^{Reference Martinson, Broadaway and Lohman12}

Whole-genome sequencing

Samples from the infant and adult CA-CDI sets that expressed the same ribotypes were further subjected to WGS. DNA was extracted using the QIAamp DNA Micro Kit (Hilden, Germany). The preparation of the libraries was performed using the Nextera DNA Flex Library Prep Kit (Illumina, San Diego, CA), and paired-end reads (2 × 150 bp) were generated using the Illumina MiniSeq reagent kit and MiniSeq instrument (Illumina, San Diego, CA). Whole-genome single-nucleotide polymorphism (WgSNP) analysis was performed using the cloud based BioNumerics version 7.6 software platform (Applied Maths NV, Sint-Martens-Latem, Belgium, available at http://www.applied-maths.com) after dynamic trimming and mapping of the raw reads to a reference genome isolate (National Center for Biotechnology Information [NCBI] Cdiff Refseq 630). The reference genome de novo assembly was completed using SPAdes assembler on the Bionumerics calculation engine using the Bowtie2 mapping algorithm.

We used the following settings for the Bowtie mapping algorithm: minimum total coverage of 3, minimum forward coverage of 1, minimum reverse coverage of 1, single-base threshold of 0.75, double-base threshold of 0.85, triple-base threshold of 0.95, and gap threshold of 0.5. An SNP analysis was performed using the patient isolates (adult and infant). Using SNP filtering, positions with at least 1 unreliable (N) or ambiguous (non-ACGT) base was removed. Each retained SNP position has minimum 5× coverage, at least covered once in both forward and reverse directions. Further, advanced cluster analysis (similarity matrix) was constructed using similarity coefficient for categorical (differences). A dendrogram was created using global closest pair method (UPGMA) with no resampling. All of the analyses were performed within the BioNumerics version 7.6 software platform.

Whole-genome multilocus sequence typing (wgMLST) with assembly-free and assembly-based calls was performed using the calculation engine on the Bionumerics version 7.6 platform, employing the sequences derived from the WGS analysis. Quality assurance was performed to ensure accuracy of the calls. Isolates so analyzed were assigned both to their MLST Pasteur sequence type (ST) and their public ST.