Genomic investigation of an outbreak of carbapenemase-producing Enterobacter cloacae: long-read sequencing reveals the context of blaIMP4 on a widely distributed IncHI2 plasmid

Background We describe whole genome sequencing (WGS) to analyse a cluster of blaIMP-4 carbapenemase-producing Enterobacter cloacae. Methods A cluster of carbapenemase-producing E. cloacae were identified over a two month period in 2015 within an Intensive Care Unit (ICU)/Burns Unit in Brisbane, Australia. Phylogenetic relationships based on core single nucleotide polymorphisms (SNPs) were determined using WGS. Genomic comparisons were made to IMP-producing Enterobacteriaceae from neighbouring hospitals and to publicly available genomes to contextualise the isolates in the broader community. Pacific Biosciences Single Molecule Real-Time (SMRT) sequencing of one IMP-4-producing E. cloacae strain was used to resolve the full context of the resistance genes. Results All outbreak strains were sequence type 90 and differed by only four core SNPs. WGS analysis unequivocally linked all 10 isolates to a 2013 isolate from the same ward, confirming the hospital environment as the most likely original source of infection in the 2015 cases. No clonal relationship was found to IMP-4-producing isolates identified from other local hospitals. However, all IMP-4-producing strains were found to possess an identical blaIMP-4 carried on a large IncHI2 plasmid. Conclusions During the course of an outbreak investigation, WGS revealed the transmission dynamics of a carbapenemase-producing E. cloacae cluster, linking it to a historical isolate from the same Unit and revealing the full context of blaIMP-4 on a multi-drug resistant IncHI2 plasmid that appears to be widely distributed in Australia. 40-word summary Whole genome sequencing of blaIMP-4-producing Enterobacter cloacae detected an unknown persistent source of infection within the hospital. All isolates were found to carry multiple antibiotic resistance genes, located in a large multidrug resistant region on a 330,060 bp IncHI2 plasmid.

6 antibiogram with susceptibility to third generation cephalosporins, meropenem and gentamicin. 122 She received 3 days of oral norfloxacin 400mg twice daily with microbiological resolution. 123 124 Patient 3, a 39-year old woman, was admitted with 66% total body surface area burns to the same 125 ICU 5 weeks after Patient 1 and 2 were admitted and 20 days after they had been discharged from 126 the ICU (Figure 1). MDR-E. cloacae was cultured from the ETT of Patient 3 on day 12 of ICU 127 admission. She had frequent brief admissions to several hospitals since 2010 (never to ICU), and no 128 Brisbane area (referred to as Hospital A and B), were obtained from the Central Laboratory of 146 Pathology Queensland for comparison (Table S3).  isolates collected from the outbreak were resistant to ceftriaxone, ceftazidime, ticarcillin-207 clavulanate, piperacillin-tazobactam, meropenem, gentamicin and trimethoprim-sulphamethoxazole 208 by Vitek 2 testing (Table 1) and demonstrated carbapenemase production by Carba-NP. The MICs 209 for meropenem were considerably lower when tested by Etest [29], often falling below the 210 clinically susceptible breakpoint defined by EUCAST, but above the epidemiological cut-off 211 (ECOFF) [30]. MS7889 was fully susceptible to carbapenems (meropenem MIC=0.032 by Etest) 212 and was negative for IMP-4-like genes by PCR (Table 1). 213 214 Whole genome sequencing identifies a link to a previous IMP-producing isolate 215 WGS of 10 isolates from patients 1, 2 and 3 was initiated after microbiological confirmation of a 216 bla IMP-4 E. cloacae isolate from a third patient from the RBWH ICU ( Figure 1). In silico MLST 217 showed all belonged to sequence type (ST) 90 with the majority exhibiting the same resistance gene 218 profile, including a 100% identical bla IMP-4 gene ( Table 1). The exception was the carbapenem 219 susceptible isolate MS7889, which was confirmed by WGS to have lost the bla IMP-4 gene as well as 220 several additional resistance genes conserved in the other E. cloacae isolates (Table 1). All ten 221 isolates contained an IncHI2 plasmid. Sequence analysis suggests that AmpC derepression is 222 unlikely to contribute to carpabenemase activity in these strains (further details are given in the 223 supplementary appendix). 224

225
Comparison of the E. cloacae genomes to publicly available draft assemblies identified a close 226 match to E. cloacae Ecl1 (GenBank: JRFQ01000000), an ST90 strain isolated from a burns patient 227 at the RBWH ICU almost two years prior to the 2015 outbreak [13,31]. Antibiotic resistance 228 profiling of the Ecl1 genome revealed an identical resistance profile compared to the majority of the 229 2015 isolates (Table 1). 230

231
The 2015 outbreak isolates were near identical at the core genome level to an isolate from 232

233
To investigate the relationship between the isolates at single-nucleotide resolution, reads from the 234 2015 RBWH isolates were mapped to E. cloacae draft assembly for Ecl1. All 2015 RBWH isolates 235 differed by fewer than five core SNPs (4,934,357 bp core genome), consistent with a direct 236 ancestral relationship (Figure 2). Two isolates from Patient 1 and two isolates from Patient 3 were 237 indistinguishable at the core genome level (Figure 2), although all of the isolates from Patient 3 had 238 lost a prophage region (refer supplementary appendix). Ecl1 (isolated in 2013) was very closely 239 related to these isolates, differing by only one core SNP. All four isolates from Patient 2 contained a 240 discriminatory single-nucleotide deletion, thereby ruling out Patient 2 to Patient 3 transmission 241  The finding that the outbreak strains were virtually indistinguishable from an IMP-4-producing E. 295 cloacae isolated two years previously from the same unit was unexpected and highlighted the need 296 to consider environmental sources and potential person-to-person transmission, as has been 297 previously described in Australian ICU and burns units [14]. Although we were unable to isolate 298 any IMP-producing Enterobacter spp. from environmental sampling, it is possible that this may 299 have been due to enhanced cleaning and additional infection control measures. Healthcare workers 300 are also a possible reservoir, with previous studies confirming carriage of a range of clinically 301 important bacteria [36][37][38]. 302 303 Using SMRT sequencing technology, we determined the full context of bla IMP-4 and its location 304 within a large, complex and highly repetitive MDR region harbouring two integrons: In37 and 305 In809. In37 is a widespread class 1 integron that has been found in many bacterial species [39,40]. 306 In809, which carries bla IMP-4 , has previously been described from Klebsiella pneumoniae 307 IncHI2 MDR plasmid (pIMP4-SEM1) [43]. Remarkably, we found that pIMP4-SEM1 was near 312 identical to pMS7884A ( Figure S5). This finding highlights the role of domestic animals (or the 313 food they eat) as a reservoir for antibiotic resistance genes. 314 315 Analysis of several CPE in this study suggested that a common plasmid or integron carrying 316 multiple antibiotic resistance genes is likely the major driver of antibiotic resistance dissemination 317 across a broad range of Enterobacteriaceae. In addition to the presence of bla IMP-4 , four resistance 318 genes (bla TEM-1b , bla IMP-4 , qnrB, and aac(6')-Ib) carried by these isolates were previously detected 319 by PCR in the majority of 29 IMP-4-producing E. cloacae isolates surveyed from Queensland 320 hospitals between June 2009 to March 2014 [13]. Only one of these isolates was ST90, suggesting 321 lateral transfer of these genes to different Enterobacter clones in Queensland before 2013. 322 323 There were significant discrepancies between meropenem MICs according to the testing modality 324 used, with the Etest consistently testing as "susceptible/intermediate" (MIC ≤4 mg/L; range 0.5-4 325 mg/L) and Vitek2 as "resistant" (usually with MICs ≥16 mg/L). According to 326 pharmacokinetic/pharmacodynamic (PK/PD) principles, provided the MIC to a carbapenem falls 327 within a susceptible range, the agent may still be effective despite the presence of a carbapenemase 328 [44]. Robust clinical data to help guide therapy are lacking and many clinicians rely on combination 329 14 therapy to optimize efficacy against carbapenemase-producers, largely based on observational 330 studies suggesting benefit [45,46]. The presence of carbapenemase genes may be missed if clinical 331 breakpoints for carbapenem MICs are used [30], however it can be rapidly ascertained by WGS, 332 without a priori assumptions of which genes are likely to be present. A wealth of additional 333 information that may influence clinical decisions can be obtained, such as the presence of other β-334 lactamases, factors that may regulate resistance gene expression (e.g. IS elements), mutations in 335 outer-membrane proteins, or other known resistance genes. 336 337

Conclusions 338
We used WGS to help elucidate genetic relationships between bla IMP-4 carbapenemase-producing E. 339 cloacae identified from our ICU and Burns facility. Real-time application of this technology 340 revealed an unexpected clonal relationship with a strain isolated from the same unit two years 341 previously.
Comparison with other Enterobacteriaceae containing bla IMP-4 isolated from 342 surrounding hospitals revealed its carriage on a broad host range IncHI2 plasmid, assumed to be 343 circulating via lateral gene transfer across different E. cloacae clones and also E. coli. SMRT 344 sequencing enabled the genetic context of all resistance genes within this plasmid to be resolved 345 and revealed the mechanism of loss of resistance genes in one E. cloacae strain that reverted to a 346 fully carbapenem-susceptible phenotype. As WGS technologies become increasingly available, 347 they are likely to prove an essential tool for the clinical microbiology laboratory to respond to 348 emergent infection control threats, and can be used in real-time to provide clinically meaningful 349 information. 350 351  Strain  7884  7885  7886  7887  7888  7889  7890  7891  7892  7893   Source  ETT  urine  ETT  urine  blood  urine  ETT  blood   Leg  swab  blood   ST  90  90  90  90  90  90  90  90  90  90   Plasmid  IncHI2  IncHI2  IncHI2  IncHI2  IncHI2  IncHI2  IncHI2  IncHI2  IncHI2  IncHI2 MIC (  shows specific core single nucleotide variant (SNV) differences identified between strains. Strains 508 within the same circle have identical core SNV profiles. Lines connecting circles represent 509 accumulating SNV differences between strains (not-to-scale), where each line represents one SNV 510 (including nucleotide deletion). Specific nucleotide differences between isolates are given in the 511 Dataset S1. All 11 isolates differed by 5 SNVs overall. RBWH isolates (n=10) as they were found to be near identical at the core genome level. 63,861 519 core SNPs were identified and used to generate a ML tree with RAxML (1000 bootstrap replicates), 520 which determined no relationship between the RBWH isolates (pink) and the Hospital B 521 (blue)/Hospital A (orange) isolates. Four closely related strains were identified from Hospitals A 522 and B (red box). Alignment of trimmed reads from MS8077, MS8079 and MS7926 to MS7924 523 identified 117 core SNPs, however, a number of these SNPs were removed as they were identified 524 as residing within transposon or phage regions. The remaining 58 core SNPs were used to generate 525 a ML tree (1000 bootstrap replicates), showing that Hospital B strains differ by less than 20 SNPs. 526 527 Figure 4: Large IncHI2 plasmid with ~55 kb multidrug resistance region containing IMP-4 528 carbapenemase: A. A 330,060 bp IncHI2 plasmid carrying multiple resistance operons, including a 529 large ~55 kb multidrug resistance (MDR) region, was fully recovered and assembled using Pacific 530 Biosciences (PacBio) SMRT sequencing of strain MS7884 (patient 1, isolate 1). The multidrug 531 resistance region was found to contain two class 1 integrons (In809, In37) along with several other 532 antibiotic resistance genes, as indicated. Comparison of this MDR region to publicly available 533 genomes found a close match to pEl1573, isolated in 2012 from an E. cloacae isolate in Sydney, 534 Australia. B. A predicted model of homologous recombination between two aac(6')-Ib (aac6) genes 535 (red asterisks) within the ~55 kb MDR region in MS7889 (patient 2, isolate 4, IMP-, carbapenem-536 susceptible) leading to the loss of a ~34 kb region containing bla IMP-4 as well as several other 537 antibiotic resistance genes.