Mechanisms of colistin resistance in Escherichia strains isolated from bloodstream infections

Abstract Infections by multidrug-resistant Gram-negative bacteria are increasingly common, prompting the renewed interest in the use of colistin. Colistin specifically targets Gram-negative bacteria by interacting with the anionic lipid A moieties of lipopolysaccharides, leading to membrane destabilization and cell death. Here, we aimed to uncover colistin resistance mechanisms in ten colistin-resistant Escherichia strains out of 1140 bloodstream isolates, originating from patients hospitalised in a tertiary hospital over a ten-year period (2006 - 2015). Core genome phylogenetic analysis showed that each patient was colonised by a unique strain, suggesting that colistin-resistant strains were acquired independently in each case. All colistin-resistant strains had lipid A that was modified with phosphoethanolamine. One strain carried the mobile colistin resistance gene mcr-1.1. Through construction of chromosomal transgene integration mutants, we experimentally determined that mutations in basRS, encoding a two-component signal transduction system, led to colistin resistance in four strains. While colistin resistance in E. coli can be acquired through mcr-1.1, sequence variation in basRS is another, potentially more prevalent but underexplored, cause of colistin resistance.

interacting with the anionic lipid A moieties of lipopolysaccharides, leading to membrane 23 destabilization and cell death. Here, we aimed to uncover colistin resistance mechanisms in ten 24 colistin-resistant Escherichia strains out of 1140 bloodstream isolates, originating from patients 25 hospitalised in a tertiary hospital over a ten-year period (2006 -2015). Core genome phylogenetic 26 analysis showed that each patient was colonised by a unique strain, suggesting that colistin-resistant 27 strains were acquired independently in each case. All colistin-resistant strains had lipid A that was 28 modified with phosphoethanolamine. One strain carried the mobile colistin resistance gene mcr-1.1. 29 Through construction of chromosomal transgene integration mutants, we experimentally determined 30 that mutations in basRS, encoding a two-component signal transduction system, led to colistin 31 resistance in four strains. While colistin resistance in E. coli can be acquired through mcr-1.1, 32 sequence variation in basRS is another, potentially more prevalent but underexplored, cause of colistin 33 resistance. 34 Introduction other Enterobacteriaceae, which effectively negates the possible activating effects of mutations in 61 phoPQ or mgrB, through PmrD, on the levels of phosphorylated BasR. This may explain why not all 62 of the previously described mutations reported to confer colistin resistance in Salmonella and 63 Klebsiella confer resistance in E. coli 14,[20][21][22] . In addition, phoPQ expression in E. coli is not only 64 controlled by MgrB but also by the sRNA MicA, adding to the mechanisms controlling PhoPQ 65 activation and making it less likely that the deletion or inactivation of mgrB can contribute to colistin 66 resistance in E. coli 14,23 . This may explain why colistin resistance in clinical E. coli strains has only 67 been linked to mutations in basRS 24-28 , although experimental validation of the role of these mutations 68 in colistin resistance is currently mostly lacking. 69 The PmrAB (BasRS) two-component system plays a crucial role in mediating the 70 modification of LPS that lead to colistin resistance in Gram-negative bacteria 14,17 . Normally, this 71 two-component system is activated by environmental stimuli, such as the presence of antimicrobial 72 peptides or a low pH. Activation can increase virulence and survival through evasion of the host 73 immune system by upregulating genes associated with modification of LPS, which is the predominant 74 immunogenic molecule of Gram-negative bacteria 29,30 . In E. coli, the activation of BasRS leads to 75 increased expression of various operons, including its own. This operon also includes eptA, which 76 encodes a lipid A-specific phosphoethanolamine transferase 11,14,31 . 77 Relatively little is known about colistin resistance mechanisms in E. coli, other than the 78 acquisition of mcr-genes 32 . Therefore, we studied a collection of colistin-resistant E. coli strains from 79 bloodstream infections by a combination of whole genome sequencing and matrix-assisted laser

Determination of minimal inhibitory concentration 110
Minimal inhibitory concentrations (MICs) to colistin were determined as previously 111

Isolation and analysis of lipid A 163
Isolation of lipid A molecules and subsequent analysis by negative-ion MALDI-TOF mass 164 spectrometry was performed as previously described 19,44,45 . Briefly, Escherichia strains were grown in 165 LB (Oxoid) and the lipid A was purified from stationary cultures using the ammonium 166 hydroxide/isobutyric acid method described earlier 46 . Mass spectrometry analyses were performed on 167 a Bruker autoflex™ speed TOF/TOF mass spectrometer in negative reflective mode with delayed 168 extraction using as matrix an equal volume of dihydroxybenzoic acid matrix (Sigma-Aldrich) 169 dissolved in (1:2) acetonitrile-0.1% trifluoroacetic acid. The ion-accelerating voltage was set at 170 20 kV. Each spectrum was an average of 300 shots. A peptide calibration standard (Bruker) was used 171 to calibrate the MALDI-TOF. Further calibration for lipid A analysis was performed externally using 172 lipid A extracted from E. coli strain MG1655 grown in LB medium at 37°C. 173 174

Prediction of functional impact of sequence variation in basRS 175
We aimed to predict the functional impact of sequence variation in basR and basS using 176 ConSurf 47 , Scorecons 48 , and PROVEAN 49 . Multiple sequence alignments (MSAs) for BasR and BasS 177 amino acid sequences were generated using ConSurf with standard settings. Using the BasR and BasS 178 MSAs, ConSurf was used to determine the rate of conservation of each amino acid residue using the 179 Bayesian calculation method, and JTT as evolutionary substitution model 47,50 . Scorecons was used to 180 quantify residue conservation, with consideration of the characteristics of each residue, using standard 181 settings 48 . These methods were used to score conservation of the specific residues, with highly 182 conserved residues being more likely to importantly contribute to protein function. In addition to 183 The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/864983 doi: bioRxiv preprint

Construction of chromosomal basRS transgene insertions 188
Chromosomal transgene insertions of basRS were constructed in BW27848 by utilizing the 189 The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/864983 doi: bioRxiv preprint

Statistical analysis 217
Statistical significance was determined using the non-parametric Kruskal-Wallis one-way, 218 two-tailed ANOVA test. Correction for multiple comparison testing was performed using Dunn's 219 correction. Family-wise significance was defined as a p-value < 0.05. 220 221 . CC-BY-NC 4.0 International license author/funder. It is made available under a The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/864983 doi: bioRxiv preprint

Low prevalence of colistin resistance in invasive Escherichia bloodstream isolates 223
A total of 1140 bloodstream isolates (collected from January 2006 to December 2015), for 224 which automated antibiotic susceptibility profiles were available, and for which species identification 225 had been performed, were available for this study. Twelve isolates were deemed resistant to colistin 226 through routine diagnostic procedures. Two of those isolates were isolated from the same patient, on 227 the same day, and were thus considered duplicates, and only one of these was included in this study. 228 In ten of the eleven remaining isolates, colistin resistance, defined as an MIC > 2 µg/ml colistin, was 229 confirmed through broth microdilution ( Fig. 1). Strain A783 was a false positive for colistin resistance 230 during automated susceptibility testing in routine diagnostic procedures, and was excluded from 231 further analysis, leaving ten isolates for further investigation ( patients received colistin to treat infections, but all three patients were also administered colistin as 236 part of selective digestive or oropharyngeal decontamination (SDD/SOD), a prophylactic antibiotic 237 treatment widely used in Dutch intensive care units 53 . The ten colistin-resistant strains were analyzed 238 further in this study to determine their relatedness and mechanism through which they had developed 239 colistin resistance. 240 241

Colistin-resistant bloodstream E. coli isolates are not clonal 242
To assess the phylogenetic relationships between the colistin-resistant strains, a phylogenetic 243 tree was generated based on the assembled contigs of the colistin-resistant strains and 210 publicly 244 available complete genome sequences (Supplemental Table 1). Based on a core genome alignment of 245 874 kbp, we did not observe direct transmission of colistin-resistant strains between patients (Fig. 2a). 246 Three colistin-resistant strains (strains I1121, H2129, and G821) belonged to the globally 247 . CC-BY-NC 4.0 International license author/funder. It is made available under a The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/864983 doi: bioRxiv preprint disseminated ST131 clone (Fig. 2a), and all three were dispersed throughout the multidrug-resistant 248 clade C of ST131 3,54 . This indicates that these ST131 strains have independently acquired colistin 249 resistance (Fig. 2b). Strain A2361 clustered among E. albertii, although it had been typed as E. coli in 250 routine diagnostic procedures. 251 By screening for acquired antibiotic resistance genes through ResFinder 3.2, we found that 252 only strain E3090 carried the mcr gene mcr-1.1 (0.086% of all bloodstream isolates; Supplemental 253 effects. We were unable to generate the construct for strain E650, presumably due to the toxicity of 296 the insert. 297 The colistin MIC determination of the generated basRS chromosomal transgene insertion 298 mutants from strains I1121, H2129, G821, and Z821 had significantly higher colistin MIC values than 299 . CC-BY-NC 4.0 International license author/funder. It is made available under a The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/864983 doi: bioRxiv preprint the BW27848::Tn7-empty strain (adjusted p-values of 0.0195, 0.0094, 0.0008, and 0.008 300 respectively), with observed MIC values higher than the 2 µg/ml cut-off for resistance as set by 301 EUCAST (Fig. 4). As expected, the basRS allele of the mcr-1.1 positive strain E3090 did not lead to 302 colistin resistance. We were unable to show the contribution of basRS to colistin resistance in the 303 additional four colistin-resistant strains (F2745, E2372, D2373, A2361) that lacked mcr-1.1. To assess the possible role of observed variations in basRS in colistin resistance, we used 317 ConSurf 47 and Scorecons 48 to score conservation of the specific residues. In addition, we used 318 PROVEAN 49 to predict the impact of the mutations on protein function. As input sequences, we used 319 the BasR and BasS amino acid sequences of the genomes that most closely matched to the 320 colistin-resistant strains (Fig. 2), as this allowed the scoring of the identified substituted residues. The 321 rate of conservation of an amino acid indicates its importance in the protein, with higher conserved 322 residues being more likely to be important for correct protein functioning. ConSurf results indicated 323 that one of the three substituted residues, A159 in BasS, was highly conserved (Fig. 5). In contrast, 324 Scorecons shows that the specific characteristics of all three affected residues were conserved. The 325 . CC-BY-NC 4.0 International license author/funder. It is made available under a The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/864983 doi: bioRxiv preprint PROVEAN results suggested that the G53S substitution in BasR, and the duplication of the 326

HAMP-domain and A159P substitution in BasS could impact protein function. ConSurf or Scorecons 327
could not be used for the analysis of the observed duplication in strain G821. Thus, for all observed 328 mutations, an impact of the substitution of the residues on protein functioning was supported by at 329 least one method. 330 Using these in silico prediction methods, we hypothesised that the observed mutations were The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/864983 doi: bioRxiv preprint

Discussion 341
In the present study, we identified the mechanisms through which E. coli bloodstream isolates 342 can develop colistin resistance. We did not find evidence for transfer of colistin-resistant strains 343 between patients, suggesting that colistin resistance has been acquired independently in all cases. In 344 seven patients colistin-resistant strains were isolated without the patients being previously exposed to 345 the drug. All colistin-resistant strains had LPS that was modified by the addition of 346 phosphoethanolamine to the lipid A moiety of LPS. Resistance in one of the bloodstream isolates 347 could be explained by the acquisition of mcr-1.1. In four other strains, we identified mutations in 348 basRS that contribute to colistin resistance. Although colistin-susceptible strains that were isogenic to 349 the resistant strains were not available, we were able to pinpoint the mutations in basRS leading to 350 resistance in these strains by matching the genomic sequences of our nosocomial isolates with 351 publicly available genomes, none of which were reported to be colistin-resistant, and subsequent 352 construction of chromosomally integrated basRS transgene alleles in the ∆basRS strain BW27848. 353 The mechanisms of colistin resistance in the remaining five strains remain to be characterized.    . CC-BY-NC 4.0 International license author/funder. It is made available under a The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/864983 doi: bioRxiv preprint   colistin-resistant strains from this study were inserted into the attTn7 site of BW27848. ECOFF for 625 colistin resistance is defined as an MIC value higher than 2 µg/ml by EUCAST, and is depicted with a 626 dashed line. The values presented represent mean (± standard deviation) of three independent 627 replicate experiments performed in triplicate. Statistical testing was performed by comparing the 628 strains carrying a basRS allele, BW27848::Tn7-empty strain, that harbours an insertion of the 629 unmodified Tn7 transposon at the attTn7 site in the BW27848 background. Statistical testing was 630 performed using the non-parametric Kruskall-Wallis, one-way, two-tailed ANOVA test, with Dunn's 631 correction for multiple comparisons. Significance was defined as a P-value < 0.05 (*), < 0.01 (**), or 632 <0.001 (***). Family-wise significance was defined as a p-value < 0.05. 633 . CC-BY-NC 4.0 International license author/funder. It is made available under a The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/864983 doi: bioRxiv preprint 634 . CC-BY-NC 4.0 International license author/funder. It is made available under a The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/864983 doi: bioRxiv preprint The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/864983 doi: bioRxiv preprint