Abstract

Objectives

Antimicrobial susceptibility testing of bacterial isolates is essential for clinical diagnosis, to detect emerging problems and to guide empirical treatment. Current phenotypic procedures are sometimes associated with mistakes and may require further genetic testing. Whole-genome sequencing (WGS) may soon be within reach even for routine surveillance and clinical diagnostics. The aim of this study was to evaluate WGS as a routine tool for surveillance of antimicrobial resistance compared with current phenotypic procedures.

Methods

Antimicrobial susceptibility tests were performed on 200 isolates originating from Danish pigs, covering four bacterial species. Genomic DNA was purified from all isolates and sequenced as paired-end reads on the Illumina platform. The web servers ResFinder and MLST (www.genomicepidemiology.org) were used to identify acquired antimicrobial resistance genes and MLST types (where MLST stands for multilocus sequence typing). ResFinder results were compared with phenotypic antimicrobial susceptibility testing results using EUCAST epidemiological cut-off values and MLST types.

Results

A total of 3051 different phenotypic tests were performed; 482 led to the categorizing of isolates as resistant and 2569 as susceptible. Seven cases of disagreement between tested and predicted susceptibility were observed, six of which were related to spectinomycin resistance in Escherichia coli. Correlation between MLST type and resistance profiles was only observed in Salmonella Typhimurium, where isolates belonging to sequence type (ST) 34 were more resistant than ST19 isolates.

Conclusions

High concordance (99.74%) between phenotypic and predicted antimicrobial susceptibility was observed. Thus, antimicrobial resistance testing based on WGS is an alternative to conventional phenotypic methods.

Introduction

Large amounts of antimicrobial agents are given worldwide to livestock for treatment and non-therapeutic purposes or to humans for treatment of infectious diseases.1,2 This provides favourable conditions for the selection, spread and persistence of antimicrobial-resistant bacteria capable of causing infections in animals and humans. Furthermore, with increased travel and trade of food animals and food products worldwide, problems emerging in one country can soon become a worldwide problem. This has emphasized the need for global initiatives and the establishment of standardized monitoring for antimicrobial resistance.2 Reliable, reproducible and standardized procedures for susceptibility testing are essential, not only for predicting clinical treatment, but also to detect the emergence of resistance and to compare data between reservoirs and over time.2–5

In 1995, Denmark was the first country to establish an integrated surveillance of antimicrobial resistance based on susceptibility testing of indicator and pathogenic bacteria and to compare isolates from food animals, food and humans.6,7 Several other countries have since established similar programmes,3 and recommendations for antimicrobial agents to be included, methodology to use and suggestions for epidemiological cut-off (ECOFF) values and clinical breakpoints have been published.4,5,8

Since its first development by Alexander Fleming during his work on the purification of penicillin, phenotypic testing by MIC determination or disc diffusion has been the cornerstone of antimicrobial susceptibility testing.9 However, these currently used methodologies are still associated with time delays and economic cost, especially for organisms that are difficult to grow. The results obtained in different laboratories are not entirely comparable and problems with phenotypic testing continue to result in mistakes.10–12 Current routine surveillance programmes are often accompanied by a need for further genetic characterization of isolates,13 such as sub-typing and identification of resistance genes, often requiring the involvement of specialized or reference laboratories. This further adds to the cost and time delays, reducing the possibility of a timely response.

During recent years there has been a dramatic reduction in cost and an increase in the quality of whole-genome sequencing (WGS), making this technology economically feasible as a routine tool, not only for scientific research14–17 but also for clinical diagnostics18–20 and surveillance.21 One obstacle to its implementation has been the lack of easy-to-use bioinformatics tools allowing analysis in real time of the data produced. We have recently developed methods for simple multilocus sequence typing (MLST)22 and the identification of acquired resistance genes23 in WGS data. However, especially with regard to antimicrobial susceptibility, there is insufficient knowledge regarding the concordance between phenotypic testing and the presence of resistance genes in different isolates.

This study provides, to our knowledge, the first attempt to compare WGS for the surveillance of antimicrobial resistance with the current phenotypic procedures. The study was performed using isolates from Danish national surveillance. In addition, the population structures of the bacteria included are described.

Materials and methods

Bacterial isolates

Two hundred isolates originating from Danish pigs, covering four bacterial species, Salmonella Typhimurium (n = 50), Escherichia coli (n = 50), Enterococcus faecalis (n = 50) and Enterococcus faecium (n = 50), were included in the study. All isolates were collected randomly during the first half of 2011.

Whole-genome sequencing

Genomic DNA was purified from the isolates and sequenced on the Illumina platform (paired-end reads). Reads were assembled de novo prior to prediction of the resistance profiles. N50 is the length of the smallest contig in the set that contains the fewest (largest) contigs whose combined length represents at least 50% of the assembly. N50 values of the assembly were used for quality control of the DNA sequences.

Identification of resistance genes and phenotypic susceptibility testing

The ResFinder web server (www.genomicepidemiology.org)23 was used to identify acquired antimicrobial resistance genes in the WGS data, using a threshold of 98.00% identity (ID). ResFinder will detect the presence of whole resistance genes, but not functional integrity and expression or resistance due to acquired variation in housekeeping genes. Based on the ResFinder results, a predicted phenotype was determined using phenotypes from original published studies of the genes found. The predicted resistances were compared with phenotypic antimicrobial susceptibility testing (MICs determined by the microdilution method, as previously described).10 The isolates were tested for susceptibility to 14–17 different antimicrobial agents depending on the species (see Table 1). Results were interpreted using current EUCAST (www.eucast.org) ECOFF and European Food Safety Authority (EFSA) epidemiologic breakpoints,4 except that a cut-off value of ≤8 mg/L was used for streptomycin in E. coli.8

Table 1.

Overview of the phenotypic susceptibility of the 197 isolates

Antimicrobial agent Number of phenotypically resistant isolates (%)a
 
Salmonella Typhimurium (n = 49) E. coli (n = 48) E. faecalis (n = 50) E. faecium (n = 50) 
Apramycin 0 (0) 0 (0) — — 
Gentamicin 0 (0) 0 (0) 9 (18) 0 (0) 
Kanamycin — — 17 (34) 10 (20) 
Neomycin 4 (8.6) 2 (4.2) — — 
Spectinomycin 8 (16.3) 14 (29.2) — — 
Streptomycin 25 (51.0) 25 (52.1) 19 (38) 19 (38) 
Amoxicillin/clavulanate 0 (0) — — — 
Ampicillin 23 (46.9) 13 (27.1) 0 (0) 2 (4) 
Cefotaxime 0 (0) 1 (2.1) — — 
Ceftiofur 0 (0) 1 (2.1) — — 
Penicillin — — 0 (0) 14 (28) 
Erythromycin — — 25 (50) 15 (30) 
Quinupristin/dalfopristin — — 50 (100) 0 (0) 
Chloramphenicol 2 (4.3) 5 (10.4) 10 (20) 0 (0) 
Florfenicol 2 (4.3) 0 (0) — — 
Sulfamethoxazole 26 (53.1) 13 (27.1) — — 
Tetracycline 22 (44.9) 15 (31.3) 43 (86) 31 (62) 
Trimethoprim 2 (4.3) 12 (25) — — 
Teicoplanin — — 0 (0) 1 (2) 
Vancomycin — — 0 (0) 1 (2) 
Linezolid — — 0 (0) 0 (0) 
Salinomycin — — 0 (0) 0 (0) 
Tigecycline — — 0 (0) 0 (0) 
Ciprofloxacin 0 (0) 0 (0) — 0 (0) 
Nalidixic acid 0 (0) 1 (2.1) — — 
Colistin 0 (0) 0 (0) — — 
Antimicrobial agent Number of phenotypically resistant isolates (%)a
 
Salmonella Typhimurium (n = 49) E. coli (n = 48) E. faecalis (n = 50) E. faecium (n = 50) 
Apramycin 0 (0) 0 (0) — — 
Gentamicin 0 (0) 0 (0) 9 (18) 0 (0) 
Kanamycin — — 17 (34) 10 (20) 
Neomycin 4 (8.6) 2 (4.2) — — 
Spectinomycin 8 (16.3) 14 (29.2) — — 
Streptomycin 25 (51.0) 25 (52.1) 19 (38) 19 (38) 
Amoxicillin/clavulanate 0 (0) — — — 
Ampicillin 23 (46.9) 13 (27.1) 0 (0) 2 (4) 
Cefotaxime 0 (0) 1 (2.1) — — 
Ceftiofur 0 (0) 1 (2.1) — — 
Penicillin — — 0 (0) 14 (28) 
Erythromycin — — 25 (50) 15 (30) 
Quinupristin/dalfopristin — — 50 (100) 0 (0) 
Chloramphenicol 2 (4.3) 5 (10.4) 10 (20) 0 (0) 
Florfenicol 2 (4.3) 0 (0) — — 
Sulfamethoxazole 26 (53.1) 13 (27.1) — — 
Tetracycline 22 (44.9) 15 (31.3) 43 (86) 31 (62) 
Trimethoprim 2 (4.3) 12 (25) — — 
Teicoplanin — — 0 (0) 1 (2) 
Vancomycin — — 0 (0) 1 (2) 
Linezolid — — 0 (0) 0 (0) 
Salinomycin — — 0 (0) 0 (0) 
Tigecycline — — 0 (0) 0 (0) 
Ciprofloxacin 0 (0) 0 (0) — 0 (0) 
Nalidixic acid 0 (0) 1 (2.1) — — 
Colistin 0 (0) 0 (0) — — 

aPercentages were determined by dividing the number of phenotypically resistant isolates by the total number of isolates (per species).

The association between genotype and phenotypic susceptibility to β-lactam antibiotics was further investigated in E. faecium by comparing the pbp5 gene from the E. faecium strain DO (accession number CP003585) with the pbp5 gene from the isolates.

MLST

MLST was performed on all isolates using the MLST web server (www.genomicepidemiology.org).22 The allele sequences were aligned with MUSCLE v3.8.31. From the alignments, 1000 bootstrap samples were created with Seqboot, distance matrices calculated with DNAdist, trees were created with FastME, and the final trees were created by CompareToBootstrap.

Investigation of the flanking DNA regions of aadA1 genes

In order to investigate the genetic environment flanking the aadA1 gene in aadA1-positive E. coli isolates, the individual contigs (fragments) carrying the aadA1 genes were subjected to BLASTx analysis using the CLCbio Genomic Workbench 5.51 software (CLCbio, Aarhus).

Results

Bacterial isolates and whole-genome sequencing

Of the 200 isolates included in the study, 3 were excluded: one Salmonella Typhimurium because of poor quality of the WGS data (N50 = 197), and two E. coli, which turned out to be Escherichia fergusonii. The remaining isolates were assembled to draft genomes with high N50 values (average 178 272 ± 17 457, 95% CI).

Antimicrobial susceptibility testing and resistance genes

Tables 1 and 2 show an overview of the phenotypic occurrences of resistance and the resistance genes detected by ResFinder. See Table S1 (available as Supplementary data at JAC Online) for data on individual isolates.

Table 2.

Overview of resistance genes detected in the isolates by ResFinder with ID ≥98.0%

Antimicrobial agent Resistance gene Number of isolates (%)a
 
Salmonella Typhimurium (n = 49) E. coli (n = 48) E. faecalis (n = 50) E. faecium (n = 50) 
Aminoglycoside str 0 (0) 0 (0) 5 (10.0) 1 (2.0) 
 ant(6)-Ia 0 (0) 0 (0) 18 (36.0) 18 (36.0) 
 ant(6′)-Ii 0 (0) 0 (0) 0 (0) 49 (98.0) 
 aph(3′)-Ia 2 (4.1) 2 (4.2) 0 (0) 0 (0) 
 aph(3′)-Ic 2 (4.1) 0 (0) 0 (0) 0 (0) 
 aph(3′)-III 0 (0) 0 (0) 17 (34.0) 10 (20.0) 
 aac(6′)-aph(2″) 0 (0) 0 (0) 10 (20.0) 0 (0) 
 strA/strB 19 (38.8) 10 (20.8) 0 (0) 0 (0) 
 aadA1 5 (10.2) 19 (39.6) 0 (0) 0 (0) 
 aadA2 2 (4.1) 4 (8.3) 0 (0) 0 (0) 
 aadA4 0 (0) 1 (2.1) 0 (0) 0 (0) 
 aadA13 1 (2.0) 0 (0) 0 (0) 0 (0) 
β-Lactam pbp5 0 (0) 0 (0) 0 (0) 49 (98.0) 
 blaTEM-1 21 (42.9) 12 (25.0) 0 (0) 0 (0) 
 blaTEM-117 0 (0) 1 (2.1) 0 (0) 0 (0) 
 blaCTX-M-14 0 (0) 1 (2.1) 0 (0) 0 (0) 
 blaCARB-2 2 (4.1) 0 (0) 0 (0) 0 (0) 
MLS erm(B) 0 (0) 1 (2.1) 25 (50.0) 15 (30.0) 
 lsa(A) 0 (0) 0 (0) 50 (100.0) 0 (0) 
 lnu(B) 0 (0) 0 (0) 11 (22.0) 15 (30.0) 
 msr(C) 0 (0) 0 (0) 0 (0) 44 (88.0) 
 mph(A) 0 (0) 1 (2.1) 0 (0) 0 (0) 
Phenicol catA1 0 (0) 2 (4.2) 0 (0) 0 (0) 
 floR 2 (4.1) 0 (0) 0 (0) 0 (0) 
 cmlA1 0 (0) 3 (6.3) 0 (0) 0 (0) 
 cat(pC194) 0 (0) 0 (0) 0 (0) 1 (2.0) 
Sulphonamide sul1 9 (18.4) 8 (16.7) 0 (0) 0 (0) 
 sul2 20 (40.8) 7 (14.6) 0 (0) 0 (0) 
 sul3 0 (0) 3 (6.3) 0 (0) 0 (0) 
Tetracycline tet(A) 1 (2.0) 11 (22.9) 0 (0) 0 (0) 
 tet(B) 19 (38.8) 4 (8.3) 0 (0) 0 (0) 
 tet(G) 2 (4.1) 0 (0) 0 (0) 0 (0) 
 tet(M) 0 (0) 0 (0) 34 (68.0) 27 (54.0) 
 tet(L) 0 (0) 0 (0) 24 (48.0) 5 (10.0) 
 tet(S) 0 (0) 0 (0) 0 (0) 1 (2.0) 
 tet(O) 0 (0) 0 (0) 0 (0) 1 (2.0) 
Trimethoprim dfrA1 1 (2.0) 9 (18.8) 0 (0) 0 (0) 
 dfrA12 0 (0) 2 (4.2) 0 (0) 0 (0) 
 dfrA14 1 (2.0) 0 (0) 0 (0) 0 (0) 
 dfrA21 0 (0) 1 (2.1) 0 (0) 0 (0) 
 dfrD 0 (0) 0 (0) 0 (0) 1 (2.0) 
 dfrG 0 (0) 0 (0) 17 (34.0) 0 (0) 
Glycopeptide vanA 0 (0) 0 (0) 0 (0) 1 (2.0) 
Antimicrobial agent Resistance gene Number of isolates (%)a
 
Salmonella Typhimurium (n = 49) E. coli (n = 48) E. faecalis (n = 50) E. faecium (n = 50) 
Aminoglycoside str 0 (0) 0 (0) 5 (10.0) 1 (2.0) 
 ant(6)-Ia 0 (0) 0 (0) 18 (36.0) 18 (36.0) 
 ant(6′)-Ii 0 (0) 0 (0) 0 (0) 49 (98.0) 
 aph(3′)-Ia 2 (4.1) 2 (4.2) 0 (0) 0 (0) 
 aph(3′)-Ic 2 (4.1) 0 (0) 0 (0) 0 (0) 
 aph(3′)-III 0 (0) 0 (0) 17 (34.0) 10 (20.0) 
 aac(6′)-aph(2″) 0 (0) 0 (0) 10 (20.0) 0 (0) 
 strA/strB 19 (38.8) 10 (20.8) 0 (0) 0 (0) 
 aadA1 5 (10.2) 19 (39.6) 0 (0) 0 (0) 
 aadA2 2 (4.1) 4 (8.3) 0 (0) 0 (0) 
 aadA4 0 (0) 1 (2.1) 0 (0) 0 (0) 
 aadA13 1 (2.0) 0 (0) 0 (0) 0 (0) 
β-Lactam pbp5 0 (0) 0 (0) 0 (0) 49 (98.0) 
 blaTEM-1 21 (42.9) 12 (25.0) 0 (0) 0 (0) 
 blaTEM-117 0 (0) 1 (2.1) 0 (0) 0 (0) 
 blaCTX-M-14 0 (0) 1 (2.1) 0 (0) 0 (0) 
 blaCARB-2 2 (4.1) 0 (0) 0 (0) 0 (0) 
MLS erm(B) 0 (0) 1 (2.1) 25 (50.0) 15 (30.0) 
 lsa(A) 0 (0) 0 (0) 50 (100.0) 0 (0) 
 lnu(B) 0 (0) 0 (0) 11 (22.0) 15 (30.0) 
 msr(C) 0 (0) 0 (0) 0 (0) 44 (88.0) 
 mph(A) 0 (0) 1 (2.1) 0 (0) 0 (0) 
Phenicol catA1 0 (0) 2 (4.2) 0 (0) 0 (0) 
 floR 2 (4.1) 0 (0) 0 (0) 0 (0) 
 cmlA1 0 (0) 3 (6.3) 0 (0) 0 (0) 
 cat(pC194) 0 (0) 0 (0) 0 (0) 1 (2.0) 
Sulphonamide sul1 9 (18.4) 8 (16.7) 0 (0) 0 (0) 
 sul2 20 (40.8) 7 (14.6) 0 (0) 0 (0) 
 sul3 0 (0) 3 (6.3) 0 (0) 0 (0) 
Tetracycline tet(A) 1 (2.0) 11 (22.9) 0 (0) 0 (0) 
 tet(B) 19 (38.8) 4 (8.3) 0 (0) 0 (0) 
 tet(G) 2 (4.1) 0 (0) 0 (0) 0 (0) 
 tet(M) 0 (0) 0 (0) 34 (68.0) 27 (54.0) 
 tet(L) 0 (0) 0 (0) 24 (48.0) 5 (10.0) 
 tet(S) 0 (0) 0 (0) 0 (0) 1 (2.0) 
 tet(O) 0 (0) 0 (0) 0 (0) 1 (2.0) 
Trimethoprim dfrA1 1 (2.0) 9 (18.8) 0 (0) 0 (0) 
 dfrA12 0 (0) 2 (4.2) 0 (0) 0 (0) 
 dfrA14 1 (2.0) 0 (0) 0 (0) 0 (0) 
 dfrA21 0 (0) 1 (2.1) 0 (0) 0 (0) 
 dfrD 0 (0) 0 (0) 0 (0) 1 (2.0) 
 dfrG 0 (0) 0 (0) 17 (34.0) 0 (0) 
Glycopeptide vanA 0 (0) 0 (0) 0 (0) 1 (2.0) 

MLS, macrolide–lincosamide–streptogramin B.

aPercentages were determined by dividing the number of isolates harbouring the gene by the total number of isolates (per species).

The E. coli and Salmonella Typhimurium isolates were tested for susceptibility to 16 and 17 different antimicrobial agents belonging to 7 different antimicrobial classes, E. faecalis for susceptibility to 14 antimicrobial agents belonging to 9 different classes and E. faecium for susceptibility to 15 antimicrobial agents belonging to 10 different classes. In total, 3051 different phenotypic tests were performed for the entire dataset. For 23 tests in 20 isolates there was disagreement between the phenotypic and predicted susceptibility. Repeating the susceptibility tests for these isolates led to a match between predicted and tested resistance for 16 of these.

For the entire dataset, 482 of the 3051 tests showed resistance, while the remaining 2569 showed susceptibility. Excluding data for β-lactams in enterococci and ciprofloxacin and nalidixic acid in Salmonella Typhimurium and Escherichia, disagreements between tested and predicted susceptibility were only observed in seven cases (six E. coli and one E. faecalis, all false positive), corresponding to a concordance of 99.74%. For the Salmonella Typhimurium and E. faecium isolates complete agreement between tested and predicted susceptibility was observed (Table S1).

For E. coli, all observed disagreements were related to spectinomycin resistance. The discrepancy consisted of the presence of an aadA1 gene with 100% identity, whereas the isolates tested phenotypically susceptible to spectinomycin (five with MIC 64 mg/L and one with MIC 32 mg/L). To investigate if the genetic environment flanking the aadA1 genes in these six isolates could give an explanation for this phenotype, the corresponding contigs were subjected to BLASTx analysis. This revealed that the aadA1 genes in all six cases were associated with class 1 integrons. In contrast to this, strains with MIC >64 mg/L all either had the aadA1 gene associated with class 2 integrons or had the aadA1 gene associated with a class 1 integron but then also carried a secondary aminoglycoside resistance gene. ENTfs47 contained aac(6)-aph(2) (ID 100%), but tested phenotypically susceptible to gentamicin (MIC 32 mg/L). In addition to these seven cases, ENTfs6 contained ant(6)-Ia (ID 100%), but lacked the first 109 bp of the gene, consistent with the fact that ENTfs6 tested phenotypically susceptible to streptomycin.

In addition, 14 E. faecium isolates tested resistant to penicillin and 2 of these also to ampicillin, and 1 E. coli isolate tested resistant to ciprofloxacin and nalidixic acid, but no predicted resistance was identified with the current version of ResFinder.

All except 1 of the 50 E. faecium isolates harboured the pbp5 gene with ID ≥95.19%. When the nucleotide sequences of the isolates' pbp5 genes were compared with the pbp5 gene from the reference strain DO, a correlation between the mutations A401S and T499I and resistance to penicillin was observed (Table S2, available as Supplementary data at JAC Online). In addition, two strains with reduced susceptibility to penicillin were identified with changes (D325E, D372E and L610I) not previously observed.

Forty-three different resistance genes were observed among the 197 isolates, covering eight different antimicrobial classes (Table 2). The most commonly observed genes in Salmonella Typhimurium were blaTEM-1 (21 isolates), sul2 (20 isolates), strA/strB (19 isolates) and tet(B) (19 isolates). In E. coli they were aadA1 (19 isolates), blaTEM-1 (12 isolates) and tet(A) (11 isolates). Among E. faecalislsa(A) (50 isolates), tet(M) (34 isolates), erm(B) (25 isolates), tet(L) (24 isolates) ant(6)-Ia (18 isolates), aph(3)-III (17 isolates) and dfrG (17 isolates) were the most commonly observed and among E. faeciummsr(C) (44 isolates), tet(M) (27 isolates), ant(6)-Ia (18 isolates), erm(B) (15 isolates) and lnu(B) (15 isolates) were the most commonly observed. The bifunctional aac(6)-aph(2) gene encoding high-level gentamicin resistance (HLGR) was detected in 10 E. faecalis isolates. A single E. coli isolate harbouring erm(B) and mph(A) was also identified.

MLST and phylogenetic analysis

The MLST web server detected previously identified MLST types for 154 of the 197 isolates, while the remaining 43 isolates were characterized as unknown sequence type (ST) (Table 3). Three different MLST types were observed in Salmonella Typhimurium, ST19 (42.9%), ST34 (44.9%) and ST376 (2%). For E. coli, 19 different MLST types were observed, but many isolates had unknown STs (41.7%). For E. faecalis and E. faecium 12 and 17 different MLST types, respectively, were observed. The most prevalent MLST types in E. faecalis were ST58 (46%) and ST16 (16%) and those in E. faecium were ST5 (14%) and ST133 (14%). In addition, 26% of the E. faecium isolates belonged to unknown STs. Comparing the MLST trees for E. faecium, E. faecalis and E. coli with resistance profiles gave no correlation (data not shown). For Salmonella Typhimurium a tendency was observed: isolates with ST34 were more resistant than ST19 isolates. Sixteen (76.2%) of the isolates with ST19 were phenotypically susceptible to all antimicrobial agents tested. In contrast, 21 (95%) of the 22 isolates with MLST type ST34 were phenotypically resistant to at least one antimicrobial agent, 10 of them with the same resistance profile: streptomycin (strA/strB), ampicillin (blaTEM-1), sulphonamide (sul2) and tetracycline [tet(B)].

Table 3.

Number of isolates with a specific MLST type (%)a

MLST Salmonella Typhimurium (n = 49) MLST E. coli (n = 48) MLST E. faecalis (n = 50) MLST E. faecium (n = 50) 
ST19 21 (42.9) ST10 4 (8.3) ST9 1 (2) ST5 7 (14) 
ST34 22 (44.9) ST23 1 (2.1) ST16 8 (16) ST6 3 (6) 
ST376 1 (2) ST58 1 (2.1) ST32 2 (4) ST22 2 (4) 
unknown ST 5 (10.2) ST59 2 (4.2) ST40 2 (4) ST32 2 (4) 
  ST117 2 (4.2) ST47 1 (2) ST60 1 (2) 
  ST131 1 (2.1) ST58 23 (46) ST133 7 (14) 
  ST154 1 (2.1) ST73 2 (4) ST150 1 (2) 
  ST156 1 (2.1) ST100 2 (4) ST156 1 (2) 
  ST206 2 (4.2) ST242 1 (2) ST185 1 (2) 
  ST367 1 (2.1) ST314 1 (2) ST214 1 (2) 
  ST453 2 (4.2) ST317 1 (2) ST217 3 (6) 
  ST542 2 (4.2) ST363 1 (2) ST253 1 (2) 
  ST763 1 (2.1) unknown ST 5 (10) ST272 1 (2) 
  ST847 1 (2.1)   ST327 1 (2) 
  ST1034 1 (2.1)   ST361 2 (4) 
  ST1112 2 (4.2)   ST536 2 (4) 
  ST1665 1 (2.1)   ST540 1 (2) 
  ST1721 1 (2.1)   unknown ST 13 (26) 
  ST1800 1 (2.1)     
  unknown ST 20 (41.7)     
MLST Salmonella Typhimurium (n = 49) MLST E. coli (n = 48) MLST E. faecalis (n = 50) MLST E. faecium (n = 50) 
ST19 21 (42.9) ST10 4 (8.3) ST9 1 (2) ST5 7 (14) 
ST34 22 (44.9) ST23 1 (2.1) ST16 8 (16) ST6 3 (6) 
ST376 1 (2) ST58 1 (2.1) ST32 2 (4) ST22 2 (4) 
unknown ST 5 (10.2) ST59 2 (4.2) ST40 2 (4) ST32 2 (4) 
  ST117 2 (4.2) ST47 1 (2) ST60 1 (2) 
  ST131 1 (2.1) ST58 23 (46) ST133 7 (14) 
  ST154 1 (2.1) ST73 2 (4) ST150 1 (2) 
  ST156 1 (2.1) ST100 2 (4) ST156 1 (2) 
  ST206 2 (4.2) ST242 1 (2) ST185 1 (2) 
  ST367 1 (2.1) ST314 1 (2) ST214 1 (2) 
  ST453 2 (4.2) ST317 1 (2) ST217 3 (6) 
  ST542 2 (4.2) ST363 1 (2) ST253 1 (2) 
  ST763 1 (2.1) unknown ST 5 (10) ST272 1 (2) 
  ST847 1 (2.1)   ST327 1 (2) 
  ST1034 1 (2.1)   ST361 2 (4) 
  ST1112 2 (4.2)   ST536 2 (4) 
  ST1665 1 (2.1)   ST540 1 (2) 
  ST1721 1 (2.1)   unknown ST 13 (26) 
  ST1800 1 (2.1)     
  unknown ST 20 (41.7)     

aPercentages were determined by dividing the number of isolates having a specific MLST type by the total number of isolates (per species).

Discussion

To our knowledge, this study describes the first large comparison of WGS and phenotypic susceptibility testing for the prediction of antimicrobial susceptibility. We observed a very high concordance between predicted and observed phenotypes. Especially for the 49 Salmonella Typhimurium isolates, complete agreement was found between the result of phenotypic susceptibility testing and that predicted from the observed genes. Overall, the concordance between phenotypic testing and susceptibility predicted by ResFinder was 99.74%. This is a much higher concordance than results previously obtained during phenotypic performance testing and ring trials (external quality assurance exercises), even involving national reference laboratories, where performances as low at ∼90% correct results have been considered to be acceptable.10–12 Furthermore, of the 23 original disagreements observed between the phenotypic susceptibility tests and the prediction, repeated testing showed that the prediction was correct in 16 cases. Most remaining disagreements were observed for susceptibility to spectinomycin in E. coli. It is well established that phenotypic detection of streptomycin and spectinomycin resistance, especially in E. coli, is problematic10 and that the correlation between resistance and the presence or absence of genes is not always perfect.8 Using the sequencing data generated through this project we were able to show that differences in susceptibility were related to the class of integrons with which the aadA1 genes were associated. Testing susceptibility to spectinomycin and streptomycin is rarely done for clinical purposes, but is often included in surveillance programmes as an important epidemiological marker for the presence of genes encoding the phenotype.11 Thus, for this specific purpose WGS would give more relevant information and thus be superior to phenotypic testing, especially if information on integron structures is also included in the analysis.

The high concordance might be due partly to a low frequency of acquired resistance as well as the low complexity of resistance genes present in the bacterial population under study. However, it does prove the value of WGS in predicting susceptibility among isolates that have tested susceptible in vitro. Furthermore, in a previous study by us,23 as well as in two other studies18,19 focusing on S. aureus and a limited number of resistance genes (two and nine, respectively), a manual comparison between resistance genes found using WGS and phenotypic susceptibility was performed. In these studies complete agreement between the resistance genes that were present and in vitro susceptibility testing was observed, providing further support for the value of WGS in predicting antimicrobial susceptibility.

Genes showing high homologies to lsa(A) were found in all 50 E. faecalis isolates. It has previously been shown that lsa(A) is required for intrinsic resistance of this species to quinupristin/dalfopristin.24 We did not observe any isolates with 100% homology to the lsa(A) gene, but all 50 strains showed similar levels of resistance to quinupristin/dalfopristin (Table S1).

Most E. faecium isolates harboured msr(C), which is observed in most E. faecium and is associated with the intrinsically lower susceptibility of E. faecium to macrolide antibiotics, but does not by itself give full resistance.25,26 In the present study, no association between the presence of msr(C) and susceptibility to erythromycin could be observed.

In the present study, all enterococcal isolates that tested phenotypically resistant to erythromycin harboured the erm(B) gene. Interestingly, a single E. coli isolate also contained the erm(B) and mph(A) genes, but was not tested for susceptibility to macrolide antibiotics as E. coli is considered intrinsically resistant to these drugs. These genes have previously been detected in E. coli27,28 and may constitute a reservoir for further spread to Gram-positive organisms.

A number of E. faecium isolates were resistant to penicillin and ampicillin. Mutations in the penicillin-binding-protein (pbp5) have previously been found to be associated with reduced susceptibility to β-lactam antibiotics,29,30 although it has also been observed that this does not directly explain the observed MIC values.31 Most studies have focused on susceptibility to ampicillin and only rarely on benzylpenicillin. In the present study, only two isolates were found to be resistant to ampicillin and no obvious correlation to changes in pbp5 was identified. However, the pbp5 mutations A401S and T499I were strongly correlated with benzylpenicillin resistance. Changes at position 499 have previously been implicated in resistance to β-lactam antibiotics,29 whereas position 401 has not previously been mentioned. The importance of different combinations of mutations within pbp5 and in regulatory genes awaits further studies. Until a stronger association of genotypes with observed phenotypic susceptibility to β-lactam antibiotics is established, it is not advisable to use WGS for predicting β-lactam susceptibility of E. faecium.

We found nine gentamicin-resistant E. faecalis isolates; three of these belonged to ST16 and one to ST40. This is in agreement with previous studies in Denmark that also found HLGR E. faecalis isolates from pigs belonging to ST16, ST40 and ST97.32,33

In the present study, the different resistance genes were not distributed equally among the Gram-negative species. Thus, strA/strB most commonly encoded resistance to streptomycin in Salmonella Typhimurium, whereas this resistance was encoded mainly by aadA variants in E. coli. sul1 and sul2 were equally distributed among the E. coli isolates, while sul2 was most common in Salmonella Typhimurium. tet(A) was prevalent in E. coli, whereas tet(B) was prevalent in Salmonella Typhimurium. A previous study in Denmark observed that strA/strB was more common than aadA variants among Salmonella Typhimurium,34 which is in contrast to a number of other studies on Salmonella, where aadA variants have been most commonly found.8,35,36 Frequent occurrence of aadA variants has previously been detected in E. coli.8,37 The almost equal distribution of sul1 and sul2 among E. coli, but predominance of sul2 among Salmonella, is in agreement with other studies.33,35,37 Previous studies have also found that the gene that most commonly encodes tetracycline resistance is tet(A) in E. coli but tet(B) in Salmonella.36,37

As the cost of WGS continues to decrease and the necessary instrumentation becomes more widely available both in reference and diagnostic laboratories, it must be expected that this technology will increasingly be used either alone or in combination with conventional methods.18–21,38 The advantage of phenotypic testing compared with WGS is that phenotypic testing will detect new resistance mechanisms or resistance caused by point mutations that might be missed using an entirely genetic approach. However, problems with standardization of phenotypic testing, especially when comparing results between laboratories, also demonstrate the limitations of an entirely phenotypic method,13–15 while it must be expected to be easier to standardize a procedure for detection of resistance genes in WGS data. It should also be noted that the presence of a resistance gene does not necessarily assure expression or sufficiently high expression to produce resistance, as might be the case for the aadA1 genes present in class 1 integrons in the present study. With our current knowledge it is not advisable to completely replace phenotypic susceptibility testing with WGS in the clinical setting. However, our results do suggest that this might already be a feasible procedure for surveillance purposes.

Another benefit of WGS is that it will provide a lot of additional information. In this study we only performed MLST of the isolates, which can yield important information on changing patterns of bacterial populations, but WGS might also give a continuous background of data for detection of outbreaks or for comparison in putative outbreak situations.

In conclusion, this study showed a very high concordance between phenotypic antimicrobial susceptibility and that predicted from WGS data. This suggests that WGS might eventually replace or be used with great benefit in combination with phenotypic methods initially for surveillance purposes, but eventually also for rapid clinical diagnosis.

Funding

This study was supported by the Center for Genomic Epidemiology (www.genomicepidemiology.org) grant 09-067103/DSF from the Danish Council for Strategic Research.

Transparency declarations

None to declare.

Supplementary data

Tables S1 and S2 are available as Supplementary data at JAC Online (http://jac.oxfordjournals.org/).

Acknowledgements

We are grateful to Martin Vestergaard and Inge Marianne Hansen for excellent technical assistance.

References

1
Aarestrup
FM
Occurrence, selection and spread of resistance to antimicrobial agents used for growth promotion for food animals in Denmark
APMIS Suppl
 , 
2000
, vol. 
101
 (pg. 
1
-
48
)
2
WHO
The Evolving Threat of Antimicrobial Resistance—Options for Action
 
3
Aarestrup
FM
Monitoring of antimicrobial resistance among food animals: principles and limitations
J Vet Med B Infect Dis Vet Public Health
 , 
2004
, vol. 
51
 (pg. 
380
-
8
)
4
European Food Safety Authority-Working Group on Developing Harmonised Schemes for Monitoring Antimicrobial Resistance in Zoonotic Agents
Harmonised monitoring of antimicrobial resistance in Salmonella and Campylobacter isolates from food animals in the European Union
Eur J Clin Microbiol Infect Dis
 , 
2008
, vol. 
14
 (pg. 
522
-
33
)
5
Cornaglia
G
Hryniewicz
W
Jarlier
V
, et al.  . 
European recommendations for antimicrobial resistance surveillance
Eur J Clin Microbiol Infect Dis
 , 
2004
, vol. 
10
 (pg. 
349
-
83
)
6
Aarestrup
FM
Bager
F
Jensen
NE
, et al.  . 
Resistance to antimicrobial agents used for animal therapy in pathogenic-, zoonotic- and indicator bacteria isolated from different food animals in Denmark: a baseline study for the Danish Integrated Antimicrobial Resistance Monitoring Programme (DANMAP)
APMIS
 , 
1998
, vol. 
106
 (pg. 
745
-
70
)
7
Hammerum
AM
Heuer
OE
Emborg
H-D
, et al.  . 
Danish integrated antimicrobial resistance monitoring and research program
Emerg Infect Dis
 , 
2007
, vol. 
13
 (pg. 
1632
-
9
)
8
Garcia-Migura
L
Sunde
M
Karlsmose
S
, et al.  . 
Establishing streptomycin epidemiological cut-off values for Salmonella and Escherichia coli
Microb Drug Resist
 , 
2012
, vol. 
18
 (pg. 
88
-
93
)
9
Fleming
A
On the antibacterial action of cultures of a Penicillium, with special reference to their use in the isolation of B. influenzae
Br J Experiment Pathol
 , 
1929
, vol. 
10
 (pg. 
226
-
36
)
10
Hendriksen
RS
Seyfarth
AM
Jensen
AB
, et al.  . 
Results of use of WHO Global Salm-Surv external quality assurance system for antimicrobial susceptibility testing of Salmonella isolates from 2000 to 2007
J Clin Microbiol
 , 
2009
, vol. 
47
 (pg. 
79
-
85
)
11
Lo Fo Wong
DM
Hendriksen
RS
Mevius
DJ
, et al.  . 
External quality assurance system for antibiotic resistance in bacteria of animal origin in Europe (ARBAO-II), 2003
Vet Microbiol
 , 
2006
, vol. 
115
 (pg. 
128
-
39
)
12
Bronzwaer
S
Buchholz
U
Courvalin
P
, et al.  . 
Comparability of antimicrobial susceptibility test results from 22 European countries and Israel: an external quality assurance exercise of the European Antimicrobial Resistance Surveillance System (EARSS) in collaboration with the United Kingdom National External Quality Assurance Scheme (UK NEQAS)
J Antimicrob Chemother
 , 
2002
, vol. 
50
 (pg. 
953
-
64
)
13
Giske
CG
Cornaglia
G
Supranational surveillance of antimicrobial resistance: the legacy of the last decade and proposals for the future
Drug Resist Updat
 , 
2010
, vol. 
13
 (pg. 
93
-
8
)
14
Hendriksen
RS
Price
LB
Schupp
JM
, et al.  . 
Population genetics of Vibrio cholerae from Nepal in 2010: evidence on the origin of the Haitian outbreak
mBio
 , 
2011
, vol. 
2
 (pg. 
e00157
-
11
)
15
Price
LB
Stegger
M
Hasman
H
, et al.  . 
Staphylococcus aureus CC398: host adaptation and emergence of methicillin resistance in livestock
mBio
 , 
2012
, vol. 
3
 (pg. 
e00305
-
11
)
16
McAdam
PR
Templeton
KE
Edwards
GF
, et al.  . 
Molecular tracing of the emergence, adaptation, and transmission of hospital-associated methicillin-resistant Staphylococcus aureus
Proc Natl Acad Sci USA
 , 
2012
, vol. 
109
 (pg. 
9107
-
12
)
17
He
M
Sebaihia
M
Lawley
TD
, et al.  . 
Evolutionary dynamics of Clostridium difficile over short and long time scales
Proc Natl Acad Sci USA
 , 
2010
, vol. 
107
 (pg. 
7527
-
32
)
18
Köser
CU
Holden
MTG
Ellington
MJ
, et al.  . 
Rapid whole-genome sequencing for investigation of a neonatal MRSA outbreak
N Engl J Med
 , 
2012
, vol. 
366
 (pg. 
2267
-
75
)
19
Török
ME
Peacock
SJ
Rapid whole-genome sequencing of bacterial pathogens in the clinical microbiology laboratory—pipe dream or reality?
J Antimicrob Chemother
 , 
2012
, vol. 
67
 (pg. 
2307
-
8
)
20
Eyre
DW
Golubchik
T
Gordon
NC
, et al.  . 
A pilot study of rapid benchtop sequencing of Staphylococcus aureus and Clostridium difficile for outbreak detection and surveillance
BMJ Open
 , 
2012
, vol. 
2
 pg. 
e001124
 
21
Aarestrup
FM
Brown
EW
Detter
C
, et al.  . 
Integrating genome-based informatics to modernize global disease monitoring, information sharing, and response
Emerg Infect Dis
 , 
2012
22
Larsen
MV
Cosentino
S
Rasmussen
S
, et al.  . 
Multilocus sequence typing of total genome sequenced bacteria
J Clin Microbiol
 , 
2012
, vol. 
50
 (pg. 
1355
-
61
)
23
Zankari
E
Hasman
H
Cosentino
S
, et al.  . 
Identification of acquired antimicrobial resistance genes
J Antimicrob Chemother
 , 
2012
, vol. 
67
 (pg. 
2640
-
4
)
24
Singh
KV
Weinstock
GM
Murray
BE
An Enterococcus faecalis ABC homologue (Lsa) is required for the resistance of this species to clindamycin and quinupristin-dalfopristin
Antimicrob Agents Chemother
 , 
2002
, vol. 
46
 (pg. 
1845
-
50
)
25
Singh
KV
Malathum
K
Murray
BE
Disruption of an Enterococcus faecium species-specific gene, a homologue of acquired macrolide resistance genes of staphylococci, is associated with an increase in macrolide susceptibility
Antimicrob Agents Chemother
 , 
2001
, vol. 
45
 (pg. 
263
-
6
)
26
Werner
G
Hildebrandt
B
Witte
W
The newly described msrC gene is not equally distributed among all isolates of Enterococcus faecium
Antimicrob Agents Chemother
 , 
2001
, vol. 
45
 (pg. 
3672
-
3
)
27
Arthur
M
Andremont
A
Courvalin
P
Distribution of erythromycin esterase and rRNA methylase genes in members of the family Enterobacteriaceae highly resistant to erythromycin
Antimicrob Agents Chemother
 , 
1987
, vol. 
31
 (pg. 
404
-
9
)
28
Phuc Nguyen
MC
Woerther
P-L
Bouvet
M
, et al.  . 
Escherichia coli as reservoir for macrolide resistance genes
Emerg Infect Dis
 , 
2009
, vol. 
15
 (pg. 
1648
-
50
)
29
Zorzi
W
Zhou
XY
Dardenne
O
, et al.  . 
Structure of the low-affinity penicillin-binding protein 5 PBP5fm in wild-type and highly penicillin-resistant strains of Enterococcus faecium
J Bacteriol
 , 
1996
, vol. 
178
 (pg. 
4948
-
57
)
30
Rice
LB
Bellais
S
Carias
LL
, et al.  . 
Impact of specific pbp5 mutations on expression of β-lactam resistance in Enterococcus faecium
Antimicrob Agents Chemother
 , 
2004
, vol. 
48
 (pg. 
3028
-
32
)
31
Galloway-Peña
JR
Rice
LB
Murray
BE
Analysis of PBP5 of early U.S. isolates of Enterococcus faecium: sequence variation alone does not explain increasing ampicillin resistance over time
Antimicrob Agents Chemother
 , 
2011
, vol. 
55
 (pg. 
3272
-
7
)
32
Larsen
J
Schønheyder
HC
Lester
CH
, et al.  . 
Porcine-origin gentamicin-resistant Enterococcus faecalis in humans, Denmark
Emerg Infect Dis
 , 
2010
, vol. 
16
 (pg. 
682
-
4
)
33
Hammerum
AM
Sandvang
D
Andersen
SR
, et al.  . 
Detection of sul1, sul2 and sul3 in sulphonamide resistant Escherichia coli isolates obtained from healthy humans, pork and pigs in Denmark
Int J Food Microbiol
 , 
2006
, vol. 
106
 (pg. 
235
-
7
)
34
Madsen
L
Aarestrup
FM
Olsen
JE
Characterisation of streptomycin resistance determinants in Danish isolates of Salmonella Typhimurium
Vet Microbiol
 , 
2000
, vol. 
75
 (pg. 
73
-
82
)
35
Aarestrup
FM
Lertworapreecha
M
Evans
MC
, et al.  . 
Antimicrobial susceptibility and occurrence of resistance genes among Salmonella enterica serovar Weltevreden from different countries
J Antimicrob Chemother
 , 
2003
, vol. 
52
 (pg. 
715
-
8
)
36
Peirano
G
Agersø
Y
Aarestrup
FM
, et al.  . 
Occurrence of integrons and antimicrobial resistance genes among Salmonella enterica from Brazil
J Antimicrob Chemother
 , 
2006
, vol. 
58
 (pg. 
305
-
9
)
37
Guerra
B
Junker
E
Schroeter
A
, et al.  . 
Phenotypic and genotypic characterization of antimicrobial resistance in German Escherichia coli isolates from cattle, swine and poultry
J Antimicrob Chemother
 , 
2003
, vol. 
52
 (pg. 
489
-
92
)
38
Didelot
X
Bowden
R
Wilson
DJ
, et al.  . 
Transforming clinical microbiology with bacterial genome sequencing
Nat Rev Genet
 , 
2012
, vol. 
13
 (pg. 
601
-
12
)

Supplementary data