J Antimicrob Chemother 2017; 72: 700–704 doi:10.1093/jac/dkw511 Advance Access publication 30 December 2016
Prediction of antibiotic resistance from antibiotic resistance genes detected in antibiotic-resistant commensal Escherichia coli using PCR or WGS Downloaded from https://academic.oup.com/jac/article-abstract/72/3/700/2762720 by Biomedical Library user on 04 April 2019
Robert A. Moran1, Sashindran Anantham1, Kathryn E. Holt2,3 and Ruth M. Hall1*
1School of Life and Environmental Sciences, The University of Sydney, NSW 2006, Australia; 2Centre for Systems Genomics, University of Melbourne, Parkville, Victoria 3010, Australia; 3Department of Biochemistry and Molecular Biology, Bio21 Molecular Science and Biotechnology Institute, University of Melbourne, Parkville, Victoria 3010, Australia
*Corresponding author. School of Life and Environmental Sciences, Molecular Bioscience Building G08, The University of Sydney, NSW 2006, Australia. Tel: þ61-2-9351-3465; Fax: þ61-2-9351-5858; E-mail: [email protected]
Received 22 August 2016; accepted 27 October 2016
Objectives: To assess the effectiveness of bioinformatic detection of resistance genes in whole-genome se- quences in correctly predicting resistance phenotypes. Methods: Genomes of a collection of well-characterized commensal Escherichia coli were sequenced using Illumina HiSeq technology and assembled with SPAdes. Antibiotic resistance genes identified by PCR, SRST2 ana- lysis of reads and ResFinder analysis of SPAdes assemblies were compared with known resistance phenotypes. Results: Generally, the antibiotic resistance genes detected using bioinformatic methods were concordant, but only ARG-ANNOT included sat2. However, the presence or absence of genes was not always predictive of the phenotype. In one strain, trimethoprim resistance was due to a known mutation in the chromosomal folA gene. In cases where the copy number was low, the aadA5 gene downstream of dfrA17 did not confer streptomycin or spectinomycin resistance. Resistance genes were found in the genomes that were not detected previously by PCRs targeting a limited gene set and gene cassettes in class 1 or class 2 integrons. In one isolate, the aadA1 gene cassette in the estX-aadA1 cassettes pair was outside an integron context and was not expressed. The qnrS1 gene, conferring reduced susceptibility to fluoroquinolones, and the blaCMY-2 gene, encoding an ESBL, were each detected in a single isolate and mphA (macrolide resistance) was present in six isolates surrounded by IS26 and IS6100. Conclusions: WGS analysis detected more genes than PCR. Some were not expressed, causing inconsistencies with the experimentally determined phenotype. An unpredicted chromosomal folA mutation causing trimetho- prim resistance was found.
Introduction volunteers using a number of measures of strain diversity, includ- Bacteria resistant to therapeutic antibiotics represent a significant ing the antibiotic resistance phenotype of each isolate, phylogen- global health challenge as infections caused by multiply, exten- etic group and random amplified polymorphic DNA (RAPD) 3–6 sively or pan antibiotic-resistant Gram-negative and Gram-positive profiling. A single representative of each strain type detected bacteria continue to increase. As WGS becomes more affordable was retained. For the resistant strains, PCR was used to detect class and searchable databases of acquired antibiotic resistance genes 1 and class 2 integrons and the gene cassettes they harbour, as have been made available,1,2 predicting the antibiotic resistance well as a limited set of other resistance genes. The plasmid content 6 profile by identifying antibiotic resistance genes in WGS data has was determined recently and a few plasmids that carry resistance 5,7 become feasible. However, studies that compare experimentally genes have been studied or completely sequenced. determined resistance profiles with resistance gene content are Here, the genomes of the antibiotic-resistant isolates in the needed in order to assess the reliability of WGS-based approaches. collection were sequenced via Illumina and both reads-based We have established a non-redundant collection of commensal analysis with SRST2 using the ARG-ANNOT database and Escherichia coli recovered from healthy Australian adults by exam- assembly-based analysis with ResFinder were used to determine ining the population structure of E. coli from the colons of the resistance gene content. The outputs of this analysis were
VC The Author 2016. Published by Oxford University Press on behalf of the British Society for Antimicrobial Chemotherapy. All rights reserved. For Permissions, please email: [email protected]. 700 Resistance genes in commensal Escherichia coli JAC
Table 1. Antibiotic-resistant commensal E. coli collection
Resistance Additional genes a b Strain phenotype Phylogroup blaTEM strAB aadA sul dfrA tetA Other WGS
14.3-R4 TET A0 A— —
24.16-R4 AMP STR SMX TMP A0 11225 2 ——
24.20-R5 TET A0 22222A— — Downloaded from https://academic.oup.com/jac/article-abstract/72/3/700/2762720 by Biomedical Library user on 04 April 2019 1-R1 AMP STR SMX TET A1 þþ 1, 27 A— — TMP
2.2-R2 NAL TET A1 A— —
3.6-R4 AMP STR SMX A1 þþ 2 ——
3.6-R5 AMP CHL GEN NAL A1 þþ51, 217 A catA1, aacC2d mphA STR SMX TET TOB TMP
21.1-R1 AMP STR SMX TET A1 þþ 2 A— —
1.4-R4 SMX TET TMP A1 17 A— —
15.1-R1 AMP STR SMX TMP A1 þþ 25 ——
1.2-R2 TET A1 A— —
1.2-R3 SMX TET TMP A1 þ 214A— —
1.9-R7 AMP TET A1 þ A — qnrS1
14.2-R3 SMX TET TMP A1 2 A— aadA5, dfrA17
24-R3 AMP STR SMX TMP A1 11225 2 —— 3.5-R3 STR SPT TMP B1c 1 1 sat2 — 19.1-R1a STR SPT TMP B1 1 1 sat2 — 19.1-R1 TMP B1 1 sat2 — 1.10-R8 SMX TET TMP B2 22215 B— — 2.1-R1 AMP B2 þ —— 3-S1R AMP B2 þ —— 3.3-R2 AMP B2 þ —— 13.1-R2 AMP SMX B2 þ 2 —— 13.1-R2a AMP SMX TET B2 þ 2 B— — 14.2-R2 AMP B2 þ —— 22.1-R1 AMP CHL STR SPT B2 þ 1, 23 12B cmlA1 mef(B) SMX TET TMP 10.1-R1 AMP STR SMX TET B2 þþ 21 B sat2 — TMP 11.1-R1 AMP TMP B2 þ —— 11.3-R3 AMP TET B2 þ A— — 19.1-R3 AMP STR SPT SMX B2 þþ11, 21 B— — TET TMP 1.9-R6 AMP CIP GEN NAL B2 þþ51, 217 A aacC2d mphA STR SMX TET TOB TMPd 5.1-R1 AMP STR SMX B2 þþ 2 —— 2.3-R3 STR SMX TMP D þ 1, 25 —— 2.3-R4 SMX TET TMP D 15 C— — 2.3-R5 STR SMX TET TMP D þ 1, 25 B— — 3-R1 AMP TET D þ B— — 4-R1 AMP TET D þ B— — 4.2-R3 TET D B— — 5.2-R2 AMP CHL STR SMX D þþ 25e D catA1 — TET TMP 11.4-R4 SMX TMP D 22215 2 —— 18.1-R1 AMP TET D þ B— — 4-R2 AMP STR SPT SMX D þþ51, 217 A — mphA TET TMP
Continued
701 Moran et al.
Table 1. Continued
Resistance Additional genes a b Strain phenotype Phylogroup blaTEM strAB aadA sul dfrA tetA Other WGS
4.3-R2a AMP D þ —— 4.4-R2b AMP STR SMX TET D þþ 2 A — mphA 11.2-R2 AMP STR SPT SMX D þþ51, 217 A — mphA TET TMP Downloaded from https://academic.oup.com/jac/article-abstract/72/3/700/2762720 by Biomedical Library user on 04 April 2019 14.1-R1 AMP STR SPT SMX D þþ51, 217 A— mphA TET TMP 6.2-R1 SMX TMP D þ 214 —— 9.1-R1 SMX D 2 — aadA1 13.1-R1 TET D B— —
24.1-R1 AMP CAZ CTX TET D 22222B— blaCMY-2 24.1-R2 AMP CHL STR SPT D 111, 23 12A cmlA1 — SMX TET TMP aNew strains and ciprofloxacin and nalidixic acid resistance are bold. bAMP, ampicillin; CAZ, ceftazidime; CTX, cefotaxime; CHL, chloramphenicol; CIP, ciprofloxacin; GEN, gentamicin; NAL, nalidixic acid; STR, streptomycin; SPT, spectinomycin; SMX, sulfamethoxazole; TET, tetracycline; TOB, tobramycin; TMP, trimethoprim. c 6 Previously reported as phylogroup A0 in Moran et al. (2015). dPreviously reported as AMP (CHL) STR SMX TET TMP in Anantham and Hall5 (2012). eIncorrectly reported as dfrA1 in Anantham and Hall5 (2012). reconciled with the resistance phenotype and resistance gene JAC Online). Resistance genes in assembled contigs were detected using determination using PCR-based methods. ResFinder (https://cge.cbs.dtu.dk//services/ResFinder/)1 and raw reads were used to query ARG-ANNOT (http://en.mediterranee-infection.com/article. php?laref¼283%26titre¼arg-annot)2 using SRST2 with default settings.11 Materials and methods The coverage of resistance genes relative to the average coverage for genes E. coli isolates used for MLST was used to assess copy number. Assembled sequences were compared with those found in the GenBank non-redundant DNA The strains used were either derived from a published collection of com- database using the BLAST alignment facility (http://blast.ncbi.nlm.nih.gov). mensal E. coli strains recovered from the faeces of 22 healthy human sub- Gene Construction Kit version 2.5 (Textco, West Lebanon, NH, USA) was jects between 2008 and 20108 or were isolated from further samples used to draw figures to scale. collected from the same subjects and from an additional subject over the time frame 2008–14.4–6 Sample collection followed protocols approved by the University of Sydney Human Research Ethics Committee (HREC) (04- GenBank accession numbers
2008/10778) with informed consent from subjects. The protocols for isolation Sequences of fragments containing blaCMY-2 and the estX-aadA1 cassettes and analysis are described elsewhere.4 The 51 unique isolates that are resist- have been deposited in GenBank with the accession numbers KX462017 ant to at least 1 of the 12 antibiotics (ampicillin, ceftazidime, cefotaxime, and KX462014, respectively. The aacC2d-containing fragments are under chloramphenicol, gentamicin, kanamycin, neomycin, streptomycin, spectino- accession numbers KX462015 for 3.6-R5 and KX462016 for 1.9-R6. WGS mycin, sulfamethoxazole, tetracycline and trimethoprim) tested initially were data are available under BioProject PRJNA335932. analysed. However, four are phenotypic variants of another strain found in the same faecal sample.3,6 All isolates were further tested for resistance to tobramycin, amikacin and netilmicin and to nalidixic acid and ciprofloxacin. Results and discussion Detection of resistance genes using PCR PCR amplification, DNA sequencing and sequence The 51 antibiotic-resistant E. coli strains isolated from the com- analysis mensal flora of healthy adults are listed in Table 1. Ciprofloxacin Genomic DNA was extracted and PCR amplification and product identifica- and nalidixic acid resistance has been added to the resistance pro- tion were carried out using primers and conditions described previously.8 files reported previously (bold in Table 1) and the resistance gene 9 Amplification of the aacC2 gene was performed using published primers: content had been reported previously for most isolates. Seven 0 0 0 aacC2-F (5 -GGCAATAACGGAGGCAATTCGA-3 ) and aacC2-R (5 - additional strains (bold lines in Table 1) were examined for the CTCGATGGCGACCGAGCTTCA-30). presence of genes that could account for the observed resistance Genome sequencing was performed at the Australian Genome phenotype or for the presence of class 1 or class 2 integrons and Research Facility. Most (n ¼ 47) strains were sequenced using Illumina HiSeq, yielding mean read depths of 34–120 (average 66 ); 4 strains associated gene cassettes, as described previously. Two of the 51 were sequenced using Illumina MiSeq technology and yielded much lower isolates were resistant to gentamicin (3.6-R5 and 1.9-R6) and were read depths (5–10 ). Reads were assembled using SPAdes,10 generating tested for resistance to further aminoglycosides and were found to contig sets with an N50 in the range 63 849–299 249 for HiSeq data and be resistant to tobramycin. Primers that detect the aacC2 amino- 19 106–114 025 for MiSeq (Table S1, available as Supplementary data at glycoside resistance gene and its variants, which are known to
702 Resistance genes in commensal Escherichia coli JAC
(a)
(b) Downloaded from https://academic.oup.com/jac/article-abstract/72/3/700/2762720 by Biomedical Library user on 04 April 2019
(c)
Figure 1. Structure of regions surrounding resistance genes identified. (a) Context of the aacC2d gene; (i) is in 3.6-R5 and (ii) is in 1.9-R6. Arrows below the central line represent ORFs. Flanking ISs are labelled and an arrow indicates the orientation of the tnp gene. (b) Context of the estX- aadA1 gene cassettes in 9.1-R1. ORFs are shown as labelled arrows and the attC sites of the estX (pink) and aadA1 (cyan) gene cassettes as black, filled boxes. Sequence from CR2 is indicated by thicker blue boxes with ori and ter ends labelled. The extent of this region that matches three
entries in GenBank is indicated below. (c) blaCMY-2 insertion in 24.1-R1. The insertion is represented as filled boxes: red for ISEcp1, yellow for sequence derived from C. freundii, grey for E. coli chromosomal sequence and black for unnamed IS. The surrounding chromosome is shown as a thin line. ORFs are shown as arrows and labelled. The sequence and location of the direct repeat is indicated above. This figure appears in colour in the online version of JAC and in black and white in the print version of JAC. confer a consistent phenotype, yielded an amplicon of the correct promoter upstream of dfrA17-aadA5 was intermediate (a weak size. The resistance phenotype and resistance genes detected promoter)14 in all isolates that carried these cassettes. Some had using this approach are listed in Table 1. Only trimethoprim resist- slightly reduced susceptibility to spectinomycin (scored as resist- ance in two isolates, 14.2-R3 and 11.1-R1, was not accounted for. ant in Table 1) and the copy number of the region was significantly In 14.2-R3, a gene cassette is likely to be responsible, as the intI1 higher (4–10-fold) in these three isolates. gene was detected, but the cassette array could not be amplified. Additional resistance genes detected using WGS analysis are An aadA5 gene cassette was detected downstream of dfrA17 in listed in the last column of Table 1. They include an aadA1 gene cas- two isolates that were not significantly resistant to spectinomycin. sette in isolate 9.1-R1 that is located downstream of an estX cas- In 19.1-R1, a variant of the more abundant 19.1-R1a, the aadA1 sette,12 but the sequence data show that the two cassettes are not gene is not expressed and was found to be interrupted by an IS.8 in an integron (Figure 1b). The isolate is not resistant to streptomycin or spectinomycin, indicating that the aadA1 gene is not expressed. The mphA macrolide resistance gene was found in six isolates Detection of resistance genes using WGS (see Table 1) and was located in a 3232 bp segment between The isolates were sequenced. All of the resistance genes detected IS6100 and IS26. This context has been described previously by PCR were found using SRST2 and all except the sat2 streptothri- (e.g. GenBank accession numbers CP015160, CP015131 and 13 cin resistance gene12 were found using ResFinder, which does not LO017738). Here, the mphA gene was associated with the same include this gene in its database. However, the aadA1 gene, which additional genes including the dfrA17-aadA5 cassettes in five iso- is interrupted in 19.1-R1 and hence not expressed, was recorded lates and four of these, recovered from three different subjects, be- only using ARG-ANNOT. Failure to detect interruptions in genes will long to ST69 and may be related. Macrolide resistance was not limit the utility of WGS analysis in predicting phenotype. Similarly, assessed, as members of this antibiotic family are not used to treat the incomplete mef(B) macrolide resistance gene in 22.1-R1, Gram-negative infections. 7 which lies within the sul3-associated class 1 integron in pCERC3 The blaCMY-2 gene with ISEcp1 upstream (Figure 1c) was present was reported by ResFinder, but could not confer resistance. in the only isolate found to be resistant to ceftazidime and cefotax- The aacC2 gene found in 3.6-R5 and 1.9-R6 was identified as ime. It lies within a novel region of 3709 bp that is inserted in an aacC2d.13 Though the length of the segment downstream of ORF downstream of phoU in the chromosome of strain 24.1-R1 aacC2d differed, the context of the gene was the same with IS10 and flanked by a 5 bp direct repeat indicating transposition upstream and IS26 downstream (Figure 1a). (Figure 1c). The insertion is made up of ISEcp1 (1656 bp), 1422 bp Isolate 14.2-R3 carried dfrA17-aadA5 followed by only 41 bp of derived from a Citrobacter freundii chromosome that includes the the 30-conserved segment and hence does not include the R1 pri- blaCMY-2 gene, 502 bp from the pilT gene of an E. coli chromosome mer site, explaining the failure to detect the cassettes by PCR amp- and bases 75–204 of an IS200C-like IS (96% identical). lifying complete cassette arrays. The strAB streptomycin The genome of isolate 1.9-R7 contained the qnrS1 gene, which 15 resistance genes are not present in 14.2-R3 and neither strepto- can reduce susceptibility to fluoroquinolones, in a segment of mycin nor spectinomycin resistance was detected. The Pc 2824 bp located between a copy of ISKpn19 and sequence derived
703 Moran et al.
from the tnpA end of Tn2. This segment has been seen in several supported by NHMRC Fellowship 1061409 and S. A. and R. A. M. were sup- plasmids in GenBank (e.g. accession numbers LC056646, ported by Australian Postgraduate Awards. CP014757 and CP012734). However, 1.9-R7 was not fluoroquino- lone resistant, exhibiting only slightly reduced susceptibility to ciprofloxacin and nalidixic acid as it lacked appropriate mutations Transparency declarations in gyrA and parC. None to declare.
Trimethoprim resistance in isolate 11.1-R1 Downloaded from https://academic.oup.com/jac/article-abstract/72/3/700/2762720 by Biomedical Library user on 04 April 2019 As no acquired trimethoprim resistance gene was detected in Supplementary data 11.1-R1 by PCR or in the genome sequence, the chromosomal folA Table S1 is available as Supplementary data at JAC Online (http://jac. gene was examined. It differed from that found in the DH5a gen- oxfordjournals.org/). ome (GenBank accession number NZ_JRYM01000008.1) by two single base changes that introduce amino acid substitutions in FolA. One of these substitutions, L28R, has been shown to be asso- References 16 ciated with high-level trimethoprim resistance and this is likely 1 Zankari E, Hasman H, Cosentino S et al. Identification of acquired anti- the cause of resistance rather than the presence of an unknown microbial resistance genes. J Antimicrob Chemother 2012; 67: 2640–4. acquired resistance gene. Trimethoprim resistance mediated by 2 Gupta SK, Padmanabhan BR, Diene SM et al. ARG-ANNOT, a new bioinfor- mutations in folA is relatively rare in clinical samples. However, re- matic tool to discover antibiotic resistance genes in bacterial genomes. sistance could be missed if only acquired resistance genes are Antimicrob Agents Chemother 2014; 58: 212–20. assessed. 3 Bailey JK, Pinyon JL, Anantham S et al. Distribution of human commensal Escherichia coli phylogenetic groups. JClinMicrobiol2010; 48: 3455–6. Quinolone and fluoroquinolone resistance 4 Bailey JK, Pinyon JL, Anantham S et al. Distribution of the blaTEM gene and blaTEM-containing transposons in commensal Escherichia coli. J Antimicrob Quinolone resistance is most often caused by a specific mutation Chemother 2011; 66: 745–51. or mutations in the chromosomal gyrA gene and fluoroquinolone 5 Anantham S, Hall RM. pCERC1, a small, globally disseminated plasmid car- resistance is gained when the chromosomal parC gene is also rying the dfrA14 cassette in the strA gene of the sul2-strA-strB gene cluster. mutated. However, these mutations are not documented by the Microb Drug Resist 2012; 18: 364–71. ResFinder or ARG-ANNOT databases used to search the WGS data. 6 Moran RA, Anantham S, Pinyon JL et al. Plasmids in antibiotic susceptible 1.9-R6 was resistant to ciprofloxacin and nalidixic acid and manual and antibiotic resistant commensal Escherichia coli from healthy Australian inspection revealed that, consistent with this, it encodes Leu adults. Plasmid 2015; 80: 24–31. replacing Ser at amino acid 83 and D87N in GyrA and S80I in ParC. 7 Moran RA, Holt KE, Hall RM. pCERC3 from a commensal ST95 Escherichia 2.2-R2 and 3.6-R5 are both resistant to nalidixic acid, but not cipro- coli: a ColV virulence-multiresistance plasmid carrying a sul3-associated class floxacin (Table 1). 2.2-R2 has the substitution S83L in GyrA and in 1 integron. Plasmid 2016; 84–85: 11–9. 3.6-R5 the substitution is S83A. 8 Bailey JK, Pinyon JL, Anantham S et al. Commensal Escherichia coli of healthy humans: a reservoir for antibiotic-resistance determinants. JMed Conclusions Microbiol 2010; 59: 1331–9. 9 Frana TS, Carlson SA, Griffith RW. Relative distribution and conservation of Overall, WGS was useful for the detection of the full spectrum of genes encoding aminoglycoside-modifying enzymes in Salmonella enterica antibiotic resistance genes in each isolate. However, some genes serotype typhimurium phage type DT104. Appl Environ Microbiol 2001; 67: were not complete or were interrupted and this would lead to inac- 445–8. curacies in prediction. Some genes were not expressed due to ab- 10 Bankevich A, Nurk S, Antipov D et al.SPAdes:anewgenome sent or weak promoters combined with low copy number. Hence, assembly algorithm and its applications to single-cell sequencing. JComput direct determination of phenotype remains most accurate. Biol 2012; 19: 455–77. Occasionally, resistance to relevant antibiotics such as trimetho- 11 Inouye M, Dashnow H, Raven LA et al. SRST2: rapid genomic surveillance prim can be conferred by a mutation in a chromosomal gene and for public health and hospital microbiology labs. Genome Med 2014; 6: 90. these could easily be missed in the absence of phenotypic data. 12 Partridge SR, Hall RM. Correctly identifying the streptothricin resistance However, identification of quinolone and fluoroquinolone resist- gene cassette. JClinMicrobiol2005; 43: 4298–300. ance mutations is readily achieved using WGS. 13 Partridge SR. Analysis of antibiotic resistance regions in Gram-negative bacteria. FEMS Microbiol Rev 2011; 35: 820–55. 14 Collis CM, Hall RM. Expression of antibiotic resistance genes in the inte- Acknowledgements grated cassettes of integrons. Antimicrobial Agents Chemother 1995; 39: 155–62. We thank all our subjects for generously providing samples. We thank Dr Steven Nigro and Mr Ryan Wick for assistance with assemblies. 15 Poirel L, Leviandier C, Nordmann P. Prevalence and genetic analysis of plasmid-mediated quinolone resistance determinants QnrA and QnrS in Enterobacteriaceae isolates from a French university hospital. Antimicrob Agents Chemother 2006; 50: 3992–7. Funding 16 Toprak E, Veres A, Michel JB et al. Evolutionary paths to antibiotic resist- This work was supported by National Health and Medical Research ance under dynamically sustained drug selection. Nat Genet 2011; 44: Council (NHMRC) project grants 512434 and 1043830. K. E. H. was 101–5.
704