Settings and Study Design

During August 2016, we conducted a prospective follow-up study in Uppsala, Sweden. All 71 municipal preschools in the central parts of the city were invited to participate. In 2015, a total of 80% of all children 1–5 years of age in Uppsala County attended preschool. Of these children, 74% were enrolled in municipal preschools (mean 16.7 children in each group). The remaining children attended private preschools or other child care providers or stayed at home (14). Most attendees spent 6–8 hour/workday in the preschools.

Before the study was conducted, an advisory statement was obtained from the Regional Ethics Committee that established diapers as biologic waste. Therefore, no ethics consent was needed. Information about the study was thereafter sent to the directors of the participating preschools for further distribution to the staff and parents. In preschools in which staff and parents gave their verbal consent, 1 diaper/child was collected the morning after defecation. The time span from defecation to processing of the samples was 16–25 hours. Each diaper was marked with the age of the child to avoid duplicates. Approximately 1 month later, we revisited the preschools and collected environmental samples from water traps of washbasins, toilet flush buttons, and lunch tables.

Statistics on antimicrobial drug consumption in Uppsala County were provided by the Swedish Strategic Programme for the Rational Use of Antimicrobial Agents and Surveillance of Resistance (https://isim.ku.dk/staff/vip/?pure=en%2Fpublications%2Fstrama-the-swedish-strategic-programme-for-the-rational-use-of-antimicrobial-agents-and-surveillance-of-resistance(95ae88b0-93b5-11dd-86a6-000ea68e967b).html). Information for prescriptions of antimicrobial drugs used in Uppsala County and drug requisitions to Uppsala University Hospital in 2015 was also provided by this program.

Bacteria and Media

After collection, diapers were transported to the Department of Clinical Microbiology at Uppsala University Hospital. Stool was streaked onto MacConkey agar plates (Acumedia, Lansing, MI, USA) containing a cefotaxime disc (5 μg; Oxoid Ltd., Basingstoke, UK). Stool was also inoculated into 2 tubes containing Luria–Bertani broth (Becton, Dickinson and Co., Sparks, MD, USA) supplemented with cefpodoxime (5 μg/mL) or ertapenem (1.25 μg/mL) (Oxoid Ltd.). After incubation at 37°C for 24 hours, we semiquantified growth in the inhibition zones on MacConkey agar plates as follows: light, 1–10 CFU; moderate, 11–100 CFU; and heavy, >100 CFU. The broth was plated onto cysteine–lactose electrolyte-deficient agar plates (Becton, Dickinson and Co.) that contained cefotaxime (5 μg) and ceftazidime (10 μg) discs for selection of ESBL producers and imipenem (10 μg) and ceftazidime (10 μg) discs for selection of carbapenemase producers. We identified colonies growing in the inhibition zones after 24 hours of incubation to the species level by using matrix-assisted laser desorption/time-of-flight mass spectrometry.

Environmental samples were collected with cotton swabs and inoculated on site into Luria–Bertani broth. Subsequent handling was identical with that for fecal isolates.

Antimicrobial Drug Susceptibility Testing

We determined antimicrobial drug susceptibility for all isolates that showed growth in the inhibition zones of cefotaxime, ceftazidime, or imipenem by using the disc diffusion method and breakpoints recommended by the European Committee on Antimicrobial Susceptibility Testing (15). We tested mecillinam, amoxicillin/clavulanic acid, piperacillin/tazobactam, cefoxitin, cefotaxime, ceftazidime, cefepime, aztreonam, meropenem, gentamicin, tobramycin, amikacin, nalidixic acid, ciprofloxacin, trimethoprim, cotrimoxazole, nitrofurantoin, and tigecycline. We defined multidrug resistance as resistance to >2 antimicrobial drug classes, in addition to resistance to cephalosporins, penicillins, and monobactams.

We performed phenotypic confirmation of ESBL production as described (16). Isolates resistant to cephalosporins but not inhibited by clavulanic acid or cefoxitin were tested by using a multiplex PCR described by Pérez-Pérez and Hanson to detect plasmid-mediated AmpC (17).

Repetitive Element Palindromic PCR

To explore genetic relatedness between Escherichia coli isolates from the same preschool, we performed a previously described repetitive element palindromic PCR (rep-PCR) (11) with a slight modification: only primer ERIC2 (5′-AAGTAAGTGACTGGGGTGAGCG-3′) was used. We analyzed 86 isolates from 18 preschools with the largest number of diapers with susceptible E. coli. A total of 45 isolates from 13 preschools were ESBL producers; the remaining 41 isolates were fully susceptible to cephalosporins and meropenem. We collected these susceptible isolates from 4 preschools that had no ESBL-positive isolates (n = 6, n = 12, n = 6, and n = 10) and 1 preschool with 1 ESBL-positive isolate (n = 7). These isolates were included as controls to determine whether transmission occurred between children independent of ESBL production.

Whole-Genome Sequencing

To determine ESBL types and genetic relatedness, we subjected 43 ESBL-producing E. coli isolates from 25 preschools to whole-genome sequencing. Isolates representative of each DNA pattern determined by the rep-PCR were chosen for each preschool. For confirmation of the performance of the rep-PCR, all ESBL producers from 2 preschools (n = 3 and n = 4) and all ESBL producers from 2 children (n = 2 and n = 2) were also completely sequenced.

We prepared DNA by using the MagAttract DNA mini M48 Kit (QIAGEN, Solna, Sweden) and the BioRobot M48 (GenoVision, West Chester, PA, USA). We sequenced genomes of selected isolates by using IonTorrent (Life Technologies, Carlsbad, CA, USA) and a read length of 400 bp, according to the manufacturer’s instructions. We assembled reads into a draft genome using the AssemblerSPAdes Plugin in TorrentSuite Version 4.2 (Thermo Fisher Scientific, Waltham, MA, USA) and the recommended settings of Life Technologies. We constructed databases for each of the assembled genomes (18). Sequence data were edited by using the Emboss suite (19) and the E. coli Multilocus Sequence Typing website (20).

We aligned whole-genome sequences with a common reference sequence (GenBank accession no. NC_010473.1) by using ABACAS (21) with gaps omitted. These sequences were separated by strain type according to multilocus sequence typing. Within each strain type, differences between samples were further refined by counting the number of single-nucleotide polymorphisms (SNPs) between each pair. SNP determination was performed by using MUMmer (22).

Statistical Analysis

We used the 2-tailed Fisher exact test to compare groups. Differences were considered significant if p<0.05.

Study Population

A total of 58 (82%) of 71 municipal preschools agreed to participate in the study; the preschools were evenly distributed across the city. Four of the preschools were excluded because of no delivery of diapers on the day of collection, and 4 others were excluded because of a sudden relocation of children after a fire, making it impossible to follow the epidemiology. Thus, of the 71 invited municipal preschools, 50 (70%) delivered samples that were included in the study.

Figure 1

Figure 1. Age distribution of healthy preschool children in study of rapid increase in carriage rates of Enterobacteriaceae producing extended-spectrum β-lactamases, Sweden.

The number of diapers collected from each preschool ranged from 1 to 18 (median 6 diapers/preschool). The collection yielded 334 stool samples and 204 environmental samples. The age range of participating children was 13–45 months (median 25 months) (Figure 1); 8 samples had no information about the age of the child.

Cefotaxime-Resistant Isolates

Figure 2

Figure 2. Age distribution of extended-spectrum β-lactamase carriers in study of rapid increase in carriage rates of Enterobacteriaceae producing extended-spectrum β-lactamases, Sweden.

Of 334 children, we found 67 (20.1%) from 32 preschools who had cefotaxim-resistant Enterobacteriaceae isolates in their stool. Of these 67 children, 56 (16.8%) had >1 ESBL producer (Figure 2) and 15 (4.5%) had >1 AmpC producer. We isolated 62 ESBL-producing and 15 AmpC-producing Enterobacteriaceae strains.

The number of carriers of ESBL-producing isolates from a single preschool ranged from 1 to 7, yielding a carriage rate of 0%–80% for the included preschools. For 6 of the preschools, the carriage rate was >40%, and in 18 preschools, none of the children were ESBL carriers. The median detection rate for the preschools was 13% positive samples.

Most (n = 47) of the ESBL-positive children had 1 ESBL producer, but 5 children had 2 ESBL-producing strains at the same time, either 2 E. coli strains (n = 3) or a combination of E. coli and Klebsiella pneumoniae (n = 2). For 4 children, both ESBL- and AmpC-type enzymes were detected. One child had 2 ESBL producers (E. coli and K. pneumoniae) and 1 AmpC producer (CITM-positive E. coli).

Eleven children had only an AmpC-producing isolate. Thus, the total number of detected cefotaxime-resistant AmpC-producing Enterobacteriaceae isolates was 15 (4.5%). These isolates had the following species distribution: E. coli (n = 7), Enterobacter cloacae (n = 6), Citrobacter freundii (n = 1), and Citrobacter brakii (n = 1). For the E. coli isolates, plasmid-mediated resistance predominated; 4 isolates were DHAM positive and 1 isolate was CITM positive.

Thirty (69.8%) of the 43 fully sequenced ESBL-producing E. coli isolates harbored CTX-M-15, five had CTX-M-14 or CTX-M-14–like β-lactamases, four had CTX-M-55, three had CTX-M-57, and one had CTX-M-27. Eleven of the isolates harbored concomitant TEM enzymes: four had TEM-2, two had TEM-55, two had TEM-56, two had TEM-109, and one had TEM-4.

The 3 most common multilocus sequence types (STs) for the ESBL producers for which whole-genome sequencing was performed were ST131 (n = 11, 25.6%), followed by ST10 and ST354 (both n = 5, 11.6%). ST131 was present in 9 preschools, ST10 in 4, and ST354 in 3. Other STs represented were ST38 or the clonal complex containing ST58, ST127, ST155, ST328, ST349, ST362, ST450, ST602, ST998, ST1423, ST1634, ST2064, ST3877, and ST6448.

The 204 environmental samples yielded only 1 AmpC-producing E. coli isolate from a washbasin in 1 of the diaper-changing rooms of 1 preschool. This preschool had 3 fecal samples with ESBL/AmpC production, which was unrelated to the AmpC-producing E. coli.

Transmission of E. coli Isolates

We detected identical strains in both preschools with and without ESBL producers. If there were 2–3 ESBL-positive children in a preschool, they, as a rule, had the same strain. With increasing numbers of ESBL-positive children in a school, the likelihood increased for >1 strain being involved. In the preschool that had the highest total number of ESBL-positive children (n = 7), 4 ESBL strains were identified, of which 3 were E. coli. All E. coli strains had been transmitted to 1 or 2 additional children, but the ESBL-positive K. pneumoniae strain was detected in only 1 child. We found no correlation between the quantity of growth in the inhibition zone of cefotaxime and the tendency to spread.

The 3 most common STs were further analyzed to rule out transmission between preschools. ST10 isolates were unambiguously different (SNPs 1,200–15,000). ST38 isolates were generally more closely related than were ST10 isolates but were insufficient to support direct transmission. The closest pair of STs showed a difference of ≈300 SNPs. The most frequent ST group, ST131, held several moderately to closely related isolates. Three isolates formed the closest group (SNPs <45). Another group of 3 isolates was moderately related to the other 3 samples and to each other (SNPs ≈100–350). The remaining isolates formed a separate group with SNPs >400.

In preschools that had susceptible E. coli strains, <2 children had the same strain, and <3 strains were present in the same children. Only 1 preschool showed no spread of E. coli strains between the children. In this preschool, there were no children who had ESBL producers.

Antimicrobial Drug Resistance

Figure 3

Figure 3. Antimicrobial drug resistance for extended-spectrum β-lactamase–producing Enterobacteriaceae isolates from 62 healthy preschool children, Sweden. Values in parentheses along the x-axis are percentages. AMC, amoxicillin/clavulanic acid; AMK, amikacin; ATM, aztreonam; CAZ, ceftazidime;...

Figure 4

Figure 4. Antimicrobial drug resistance for AmpC-producing Enterobacteriaceae isolates from 15 healthy preschool children, Sweden. Values in parentheses along the x-axis are percentages. AMC, amoxicillin/clavulanic acid; AMK, amikacin; ATM, aztreonam; CAZ, ceftazidime; CIP,...

Thirty-six (58%) of the 62 ESBL-producing isolates and 3 (20%) of the 15 AmpC-producing isolates were multidrug resistant. For ESBL producers, the most frequent drug resistance was resistance to trimethoprim (71.7%) or cotrimoxazole (65.0%), followed by the quinolones (nalidixic acid 68.3% and ciprofloxacin 35%) (Figure 3). For AmpC-positive isolates, the pattern was similar but the overall resistance rates were lower, with the exception of nitrofurantoin (Figure 4).

Consumption of Antimicrobial Drugs

Uppsala County had the second highest number of prescriptions and sales on hospital requisitions in Sweden in 2015: a total of 13.9 defined daily doses/1,000 person-days. Of these doses, 4.87 defined daily doses/1,000 person-days were prescribed to children <1–17 years of age and 5.58 defined daily doses/1,000 person-days to children 1–5 years of age. Most (93.5%) of the antimicrobial drugs were prescribed for outpatients. The most commonly used drug by children was penicillin V, followed by amoxicillin with or without clavulanic acid and flucloxacillin (Table).