Comparison of two DNA microarrays for detection of plasmid-mediated antimicrobial resistance and virulence factor genes in clinical isolates of Enterobacteriaceae and non-Enterobacteriaceae

A DNA microarray was developed to detect plasmid-mediated antimicrobial resistance (AR) and virulence factor (VF) genes in clinical isolates of Enterobacteriaceae and non-Enterobacteriaceae. The array was validated with the following bacterial species: Escherichia coli ( n = 17); Klebsiella pneumoniae ( n = 3); Enterobacter spp. ( n = 6); Acinetobacter genospecies 3 ( n = 1); Acinetobacter baumannii ( n = 1); Pseudomonas aeruginosa ( n = 2); and Stenotrophomonas maltophilia ( n = 2). The AR gene profiles of these isolates were identified by polymerase chain reaction (PCR). The DNA microarray consisted of 155 and 133 AR and VF gene probes, respectively. Results were compared with the commercially available Identibac AMR-ve Array Tube. Hybridisation results indicated that there was excellent correlation between PCR and array results for AR and VF genes. Genes conferring resistance to each antibiotic class were identified by the DNA array. Unusual resistance genes were also identified, such as bla SHV-5 in a bla OXA-23 -positive carbapenem-resistant A. baumannii . The phylogenetic group of each E. coli isolate was verified by the array. These data demonstrate that it is possible to screen simultaneously for all important classes of mobile AR and VF genes in Enterobacteriaceae and non-Enterobacteriaceae while also assigning a correct phylogenetic group to E. coli isolates. Therefore, it is feasible to test clinical Gram-negative bacteria for all known AR genes and to provide important information regarding pathogenicity simultaneously. validate capacity of the customised ERV array to identify a representative collection nosocomial pathogens to detect transferable AR and VF genes in these strains. The capacity of the Identibac AMR-ve Tube detect genes these strains also for comparison. bla OXA-23-like , bla OXA-24-like , bla OXA-51-like and bla OXA-58 . Full-length PCR amplicons used for sequencing were generated using PCR primers or primers designed to flank the entire gene. Ten E. coli isolates from blood cultures were screened for 15 VFs associated with extraintestinal pathogenic E. coli pathogenesis by PCR.


Introduction
Levels of antimicrobial resistance in some developed countries are reaching crisis point, severely limiting treatment options for an increasing number of nosocomially acquired infections [1]. Rapid and reliable identification on nosocomial Gramnegative pathogens, along with characterisation of their resistance and virulence mechanisms, are highly desirable for effective management of infections. Early diagnosis and appropriate antimicrobial treatment are essential to decrease the acquisition and spread of antimicrobial resistance (AR) and/or virulence factor (VF) genes in these pathogens and thus improve patient survival and reduce healthcare costs [2].

Identification and characterisation of the genes responsible for AR and virulence in
Gram-negative pathogens have been typically limited to gene-specific multiplex polymerase chain reaction (PCR) and sequencing. However, these methods have many common drawbacks. They are both labour intensive and time consuming and thus are not ideal for routine use in microbiology laboratories. Moreover, they only screen for a relatively small number of bacterial determinants, thereby overlooking numerous other AR and VF genes that may be present. There is therefore a demand for practical and cost-effective diagnostic methods that rapidly and simultaneously detect all AR and VF genes in any given bacterial strain.
Introduction of DNA microarray technology offers a viable alternative.

Page 5 of 24
A c c e p t e d M a n u s c r i p t virulence (ERV) array. Recently, a commercial microarray has become available, the Identibac AMR-ve TM Array Tube (Veterinary Laboratories Agency, Addlestone, UK). This is a microtube DNA array that detects up to 58 AR genes of clinical importance in Salmonella spp. and Escherichia coli. The aim of this study was to validate the capacity of the customised ERV array to identify a representative collection of nosocomial Gram-negative pathogens and to detect all transferable AR and VF genes in these strains. The capacity of the Identibac AMR-ve Array Tube to detect all AR genes in these strains was also investigated for comparison.
Results of both microarray systems were validated by PCR amplification and, where necessary, by direct sequencing.

PCR detection of resistance genes
Enterobacteriaceae isolates were screened for the presence of bla TEM , bla SHV , bla OXA , bla CTX-M , transferable ampC, aac(6')-Ib-cr and qnr by PCR. Pseudomonas aeruginosa and Acinetobacter spp. isolates were screened for the presence of bla VIM , bla IMP , bla SPM and bla GIM . Acinetobacter isolates were also screened for Page 6 of 24 A c c e p t e d M a n u s c r i p t bla OXA-23-like , bla OXA-24-like , bla OXA-51-like and bla OXA-58 . Full-length PCR amplicons used for sequencing were generated using PCR primers or primers designed to flank the entire gene. Ten E. coli isolates from blood cultures were screened for 15 VFs associated with extraintestinal pathogenic E. coli pathogenesis by PCR.

ERV oligonucleotide probe design and array construction
Generation of the custom ERV oligonucleotide probe-based array and selection criteria of the genes were as described by Cooke et al. [8]. All target genes used to generate oligonucleotide probes and their corresponding nucleotide accession numbers are documented in Supplementary Table 1.

Genomic DNA labelling and microarray hybridisation
Genomic DNA was extracted using the Archive Pure DNA Cell/Tissue Kit (5PRIME, Hamburg, Germany) and labelled with Cy3-dCTP using the BioPrime ® DNA Labeling System (Invitrogen-BioSciences Ltd., Dun Laoghaire, Ireland). After labelling, probes were purified and applied to the microarray slide as outlined previously [9]. Following incubation in a sealed humid chamber at 42 C for 16-24 h, slides were washed twice in a series of wash steps, dried and scanned immediately [9].

Identibac AMR-ve Array Tube
The Identibac AMR-ve Array Tube was recently validated to identify AR genes (n = 58) in Gram-negative bacteria, including E. coli and Salmonella [3]. Amplification reactions, hybridisation, washing and scanning of the array tube were performed according to the manufacturer's instructions. Data normalisation, using the ihfA control gene probe, and subsequent analysis was carried out using IconoClust Software (AT-Version; CLONDIAG GmbH, Jena, Germany). The mean signal value for three replicate spots per probe was used for analysis. Probes with intensity values of ≥0.4 were consider positive and those with intensity values of <0.3 were considered negative, whilst those between 0.3 and <0.4 were considered ambiguous.

Page 8 of 24
A c c e p t e d M a n u s c r i p t

Detection and discrimination of antimicrobial resistance and virulence factor
genes with the ERV array DNA from all isolates hybridised with between 5 and 39 AR probes. The test group included isolates that were negative for all genes investigated by PCR or were positive for only one gene. The phenotypic resistance profile of all isolates is described in Table 1. DNA from E. coli isolates hybridised with between 10 and 80 of the VF probes. DNA from other isolates hybridised with between 1 and 33 VF probes. However, it must be noted that the VF probes were designed from E. coli and as such many would not necessarily be present in non-E. coli isolates.

Correlation between PCR and ERV array hybridisation data
A good correlation was observed between the two methods for bla TEM , bla SHV , bla OXA , bla CMY and aac(6')-Ib-cr AR genes ( Table 2). The microarray was more sensitive in detecting bla TEM in one E. coli isolate and bla SHV in one K. pneumoniae isolate, which were negative by PCR. One Enterobacter isolate was positive for bla OXA by PCR and negative by microarray, but the opposite was the case for one Acinetobacter resistance genotypes showed 100% identity between the two methods in discrimination between A. baumannii and Acinetobacter genospecies 3 using the presence of bla OXA-51 as a positive marker for A. baumannii identification.
Identification of bla OXA-51 in A. baumannii could also be used as a species

Page 9 of 24
A c c e p t e d M a n u s c r i p t identification marker. However, bla OXA-27 was also identified in both isolates of Acinetobacter. One P. aeruginosa isolate was positive by PCR for bla VIM but was positive by microarray for bla SPM . There were discrepancies between the PCR and microarray results for the detection of bla DHA in K. pneumoniae and Enterobacter spp.
Seven isolates were positive for bla CTX-M by microarray (CTX-M-9, n = 4) and negative by PCR. Of these seven isolates, four were phenotypically confirmed as extended-spectrum -lactamase (ESBL)-producers, a further two were resistant to cefotaxime and ceftazidime, and the final isolate was susceptible to both antimicrobials ( Table 1). All isolates were negative for other ESBL genes. The bla TEM genes identified in these isolates were bla TEM-1 .
The bla CMY gene was detected by the microarray and there were no falsepositives. There was 100% correlation between the two methods with regards to aac(6')-Ib-cr-positive isolates. Most of the additional isolates that were positive by microarray for aac(6')-Ib were gentamicin-resistant. There was no correlation between the two methods for detecting qnr genes.
A good correlation was also observed between the two methods for the 10 E. coli isolates that were screened for VF genes (Table 2a). Of the fimbrial VFs screened for, there was a 100% correlation for papA (n = 5), papG allele II (n = 4) and sfa/focDE (n = 3). PCR was more sensitive in detecting papG allele III and afa/draBC, whilst the microarray was more sensitive in detecting fimH. Amongst the fimbrial VFs screened, papG allele I gave the most varying correlations Page 10 of 24 A c c e p t e d M a n u s c r i p t between the two methods. It was detected in four E. coli isolates by microarray but not by PCR.
There was a 100% correlation between the two methods for all E. coli isolates that encoded hlyA (n = 4) and cnf1 (n = 3) toxin genes, and also for traT (n = 5) and ibeA (n = 1) genes. The microarray was more sensitive in detecting fyuA and iutA siderophore genes compared with PCR (8 vs. 7 and 10 vs. 6 positives, respectively). Similarly, the kpsMT II capsule biosynthesis gene was detected in six E. coli isolates by microarray but in four by PCR.
Nine E. coli isolates displayed a 100% correlation between the two methods with regards to classification of phylogenetic group. The microarray detected TSPE4.C2 in the final isolate that was not detected by PCR. However, because this isolate was also positive for chuA and negative for yjaA, the presence or absence of TSPE4.C2 would not have an overall effect on the phylogenetic group of the isolate, since either way it would be classified as belonging to group D.
Whilst the number of VFs identified in the non-invasive E. coli is greatly reduced compared with the bloodstream isolates, the overall patterns of the main VFs within the E. coli isolated from urine, swab and sputa samples corresponds to those identified within the bloodstream isolates. The urine isolates belonged to the phylogenetic groups A and B2, the swab isolate to group B2 and the sputum isolate to group B1.

Page 11 of 24
A c c e p t e d M a n u s c r i p t

Identibac AMR-ve Array Tube
Eleven E. coli, two K. pneumoniae and five Enterobacter spp. isolates that were employed to validate the ERV array were subsequently analysed by the Identibac AMR-ve Array Tube Array for comparison.
In total, nine E. coli isolates, two K. pneumoniae and four Enterobacter spp. and their duplicates generated successful hybridisation results. The remaining isolates and their duplicates failed to generate successful hybridisation results and therefore were not included in further analysis. The Identibac array showed good correlation with PCR results for the detection of -lactamases and plasmidmediated quinolone resistance determinants, except for the detection of aac(6')-Ibcr genes. Overall, both of the microarray systems detected more AR genes compared with PCR amplification analysis (Supplementary Table 2).

Discussion
In this study, we studied two microarray systems compared with conventional multiplex PCR methods, which rapidly provide comprehensive information about the genotypic profile of clinical bacterial isolates.
The ERV array detected all bla TEM , bla SHV , bla OXA , bla CMY-2 and aac (6' aiding epidemiological studies [10][11][12]. The ERV array also correctly detected a high proportion of VFs, especially with probes that corresponded to the adhesin and toxin genes.
Validation of this ERV array is unique in that it also included non-Enterobacteriaceae. This demonstrates the lack of false-positive crosshybridisation between the plasmid resistance genes on the array and any chromosomal genes in these bacteria. One of the major advantages of screening isolates for all known AR genes is the identification of genes that may be colocated or co-transferred on mobile elements. By screening the Acinetobacter spp.
isolates in this manner, we identified the presence of bla SHV-5 and aac(6')-Ib in both isolates. This is a similar resistance gene profile to an ESBL-producing A.
baumannii in the USA that was susceptible only to colistin and rifampicin [13]. The difference is that the carbapenem resistance in the isolates from our study could be accounted for by the presence of bla OXA-23 carbapenemase genes in both isolates. The microarray also detected a wide variety of aminoglycoside, chloramphenicol, tetracycline and trimethoprim resistance genes in each species of Enterobacteriaceae. As expected, few of these resistance genes were detected in the non-Enterobacteriaceae.
The original validation study of the Identibac AMR-ve Array Tube system reported 98.8% correlation between microarray and PCR results, therefore verifying the high specificity of the microarray probes [3]. However, the array was not able to detect resistance genes in a number of isolates that were resistant to the Page 13 of 24 A c c e p t e d M a n u s c r i p t aminoglycosides, -lactams, streptomycin, tetracycline, trimethoprim and sulphonamides. Taking into account the number of discrepancies reported by Batchelor et al. [3], some of which were also observed in this study, it is possible that this array format needs to undergo more stringent quality control checks, as there appears to be a high degree of batch variability leading to misdetection of a number of AR genes.
In conclusion, the ERV array proved more successful than the Identibac AMR-ve Array Tube. It not only contains more AR probes but also a comprehensive range of VF probes. There was a better concordance between the genes detected by PCR, or with their phenotypic results where PCR had not been performed, and those detected by the ERV array hybridisations. Also, all ERV array hybridisations successfully generated genotypic data using this system, unlike the Identibac AMR-ve Array Tube. Another important feature of the ERV array is that it is a very flexible system. It allows for future addition of supplemental target genes for further AR and VF genes as they are identified and characterised. Similarly, it allows for future refinement by adding or removing existing probes to improve the range, sensitivity and specificity. Plasmid-mediated resistance transfer in non-Enterobacteriaceae is an important factor in the proliferation and spread of antibiotic resistance in hospitals. Most microarray studies to date have used E. coli and Salmonella spp. for validation. This study is the first to include non-Enterobacteriaceae for validation of the array.
If the ERV array were to be combined with methodologies to amplify pathogen DNA directly from patient blood, this would provide for a powerful and effective Page 14 of 24 A c c e p t e d M a n u s c r i p t diagnostic tool. We are currently examining whether whole-genome amplification methodologies such as OmniPlex may be suited to this aspiration [14]. M a n u s c r i p t M a n u s c r i p t