A prospective matched case-control study on the genomic epidemiology of colistin-resistant Enterobacterales from Dutch patients.

Background: Colistin is a last-resort treatment option for infections with multidrug-resistant Gram-negative bacteria. However, colistin resistance is increasing.
Methods: A six-month prospective matched case-control study was performed in which 22 Dutch laboratories with 32 associated hospitals participated. Laboratories were invited to send a maximum of five colistin-resistant Escherichia coli or Klebsiella pneumoniae (COLR-EK) isolates and five colistin-susceptible isolates (COLS-EK) to the reference laboratory, matched for patient location, material of origin and bacterial species. Epidemiological/clinical data were collected and included in the analysis. Characteristics of COLR-EK/COLS-EK isolates were compared using logistic regression with correction for variables used for matching. Forty-six ColR-EK/ColS-EK pairs were analysed by next-generation sequencing (NGS) for whole-genome multi-locus sequence typing and identification of resistance genes, including mcr genes. To identify chromosomal mutations potentially leading to colistin resistance, NGS reads were mapped against gene sequences of pmrAB, phoPQ, mgrB and crrB.
Results: In total, 72 COLR-EK/COLS-EK pairs (75% E. coli and 25% K. pneumoniae) were included. Twenty-one percent of COLR-EK patients had received colistin, in contrast to 3% of COLS-EK patients (OR > 2.9). Of COLR-EK isolates, five contained mcr-1 and two mcr-9. One isolate lost mcr-9 after repeated sub-culturing, but retained colistin resistance. Among 46 sequenced COLR-EK isolates, genetic diversity was large and 19 (41.3%) isolates had chromosomal mutations potentially associated with colistin resistance. Conclusions: Colistin resistance is present but uncommon in the Netherlands and caused by the mcr gene in a minority of COLR-EK isolates. There is a need for surveillance of colistin resistance using appropriate susceptibility testing methods.

Introduction

Multidrug-resistant Gram-negative bacteria are rapidly emerging worldwide[1-3]. The mean resistance percentage for carbapenem among Klebsiella pneumoniae isolates in Europe is 7.9%, with some countries reporting resistance percentages between 25 to 50% or ≥50%. It is observed in only 0.3% of Escherichia coli isolates[3]. The polymyxin colistin is a last resort treatment option against severe infections by multi-drug resistant Gram-negative organisms (MDRO) and is increasingly used. However, colistin is potentially neuro- and nephrotoxic when administered parenterally.

Colistin has been used for decades for the prevention and treatment of infections caused by Enterobacterales in livestock[4,5]. In humans in the Netherlands, colistin is mainly used as part of the treatment of infections with Pseudomonas aeruginosa in nebulised form in patients with pulmonary diseases such as cystic fibrosis, as well as for topical treatment of otitis externa and ophthalmic infections[4,6]. In addition, colistin is used in prophylactic antibiotic regimens as a component of selective decontamination of the digestive tract (SDD) or selective oropharyngeal decontamination (SOD) with the aim to reduce infections and mortality in intensive care unit (ICU)-admitted patients and in neutropenic patients with a haematological disease[7]. Colistin is also used as last-resort treatment for MDRO.

Colistin resistance is increasing worldwide[8-10] and this poses problems in treatment of infections with MDRO. K. pneumoniae is the species most commonly involved in the development of colistin resistance[11]. Among 646 carbapenem‑resistant K. pneumoniae found in Europe in 2013–2014, 28% had tested colistin-resistant[12]. Outbreaks of carbapenemase (blaNDM-1 and blaOXA-48)-producing and colistin-resistant K. pneumoniae have been reported in Europe and highlight the emerging threat that humans are currently facing[13]. The prevalence and incidence of colistin resistance is difficult to assess, as colistin susceptibility testing is usually not part of the initial routine testing panel for Enterobacterales and is methodologically challenging with several methods producing unreliable results. Broth microdilution is the gold standard method, but is labour-intensive and time-consuming. Methods such as disk diffusion and agar dilution produce unreliable results due to the large molecular size of colistin making it poorly diffusible through agars[14]. Furthermore, many laboratories use automated antimicrobial susceptibility testing (AST) systems with high very major error rates (producing false susceptible results)[15,16].

Several chromosomal mutations in bacteria can lead to colistin resistance. For K. pneumoniae, mutations in the chromosomally located pmrAB, phoPQ, mgrB and crrB genes have been intensively studied. Mutations in these genes lead to the upregulation of the modification of lipid A in lipopolysaccharide (LPS). This modification leads to decreased negative charge of the bacterial membrane impairing the interaction between colistin and LPS[17]. In E. coli, evidence on the role of chromosomal mutations in colistin resistance is scarce[18]. Colistin resistance in E. coli strains has been linked to phoPQ and pmrAB genes, but experimental validation is mostly lacking[18].

The risk for spread of colistin resistance is further increased by transferable plasmid-mediated colistin resistance (mcr) genes that can transmit colistin resistance more easily between bacteria, including bacteria from different species[19]. Until now, mcr genes 1 to 10 have been discovered. Notably, E. coli is the most abundant mcr-containing species[20,21]. A study that examined 457 mcr-1-positive Enterobacterales isolates from 31 different countries, found 411 E. coli isolates (89.9%)[20]. Colistin resistance by chromosomal mutations and mcr genes is mostly caused by adding cationic groups to LPS[22]. Colistin resistance may be triggered directly by selection during treatment with colistin[11] or indirectly during treatment with other antibiotics by co-transfer of the mcr gene with other resistance genes on the same plasmid or different plasmids[23,24].

Little is known about the genomic epidemiology of colistin resistance in the Netherlands. One outbreak with six patients from a hospital and nursing home with a colistin-resistant carbapenemase-producing K. pneumoniae in 2013 has been described[25].

The objectives of this study were to determine the incidence and risk factors of patients colonised or infected with colistin-resistant E. coli or K. pneumoniae (COLR-EK) and to characterise the isolates. This study shows that colistin resistance is present but uncommon in the Netherlands and is plasmid-mediated in a minority of isolates.

Methods

A prospective matched case-control study with density-based sampling was performed. This project took place between May 2019 and February 2020 and was part of a pan-European multicentre study on colistin- and carbapenem-resistant Enterobacterales (CCRE survey) of the European Centre for Disease Prevention and Control (ECDC)[26].

Participating laboratories

Twenty-two Dutch medical microbiology laboratories (MMLs) providing services for 32 hospitals, participated in this project using the infrastructure of a web-based laboratory network, called Type-Ned, which is used for the national carbapenemase-producing Enterobacterales surveillance in the Netherlands[27]. Laboratories were selected based on NUTS-2 regions (nomenclature of territorial units for statistics) of the associated hospitals. In the Netherlands, the provinces represent the NUTS-2 regions. At least one hospital site per NUTS-2 region had to be included[28]. Participating hospitals had to offer acute care services.

Study population and isolates

MMLs were requested to send a maximum of five COLR-EK isolates with a minimum inhibitory concentration (MIC) > 2 mg/L for colistin and/or a mcr gene that were collected in a 6-month period. In line with the European CCRE survey guidelines, only isolates not producing carbapenemases and with a meropenem MIC ≤ 0.25 mg/L were included. Controls were selected with density-based sampling: for each COLR-EK, the first following colistin-susceptible E. coli or K. pneumoniae (COLS-EK) with a colistin MIC ≤ 2 mg/L, no mcr gene and a meropenem MIC ≤ 0.25 mg/L matched with the COLR-EK for patient location (sender of the isolate: from community or hospital), patient material and bacterial species, was requested. Only a single isolate per patient was included in the study. MMLs were asked to send isolates which they classified as COLR-EK and COLS-EK based on their routine susceptibility testing.

Detection of resistance genes and antimicrobial susceptibility testing

Microbiological confirmation of all submitted isolates was performed at the Dutch National Institute for Public Health and the Environment (RIVM). Species assignments were confirmed by MALDI-TOF (Microflex LT System; Bruker, Leiderdorp, Netherlands). The absence of carbapenem resistance was assessed by the meropenem Etest (BioMérieux Inc., Marcy L’Étoile, France). The Carbapenem Inactivation Method (CIM)[29] was used to determine whether phenotypical carbapenemase production was absent. An in-house developed multiplex PCR assay was used to assess the presence of carbapenemase-encoding genes using primers that target the allelic variants of the following genes: blaNDM, blaKPC, blaVIM, blaIMP, and blaOXA-48. The presence of mcr genes was assessed by using two specific in-house multiplex PCRs for mcr-1 to mcr-5 genes and for mcr-6 to mcr-8 genes. Colistin resistance was confirmed using a standardised broth microdilution (BMD; Micronaut MIC strip colistin, Merlin) using an ECDC-recommended protocol[30] including one positive control (NCTC 13846) and three negative controls (ATCC 25922, ATCC 27853 and ATCC 700603). AST of colistin and meropenem was performed according to EUCAST detection guidance and breakpoints[31,32].

COLR-EK and COLS-EK isolates were classified as MDRO or non-MDRO, based on antibiograms received by participating laboratories (mostly results of VITEK automated testing). MDRO was defined according to definitions of the Dutch Working Group on Infection Prevention[33]: E. coli or K. pneumoniae that are extended-spectrum beta-lactamase (ESBL)-producing, that are resistant to both a fluoroquinolone and an aminoglycoside or that produce carbapenemases. ESBL-production was defined as resistance to ceftazidime and/or either cefotaxime or ceftriaxone with additional resistance to cefepime or susceptibility to cefoxitin. When results of Etest ESBL strips or combination disc methods[34] were available, these were used to define ESBL-producers.

Genomic analysis

Forty-six COLR-EK/COLS-EK pairs were subjected to next-generation sequencing (NGS) to assess genetic relatedness of strains, presence of chromosomal mutations leading to colistin resistance, antimicrobial resistance (AMR) genes, serotypes, plasmid replicons and virulence factors. The selection of isolates for NGS was based on a minimal time between sample dates of a colistin-resistant and the matched colistin-susceptible isolate, a sufficient number of both E. coli and K. pneumoniae isolates and a diverse selection of geographic locations. Isolates were subjected to NGS using the Illumina HiSeq 2500 (BaseClear). Genetic relatedness was assessed with NGS data by classical and whole-genome multilocus sequence typing (wgMLST). A minimum spanning tree was created in Bionumerics version 7.6.3 (Applied Maths, Sint-Martens-Latem, Belgium) using an in-house wgMLST scheme in SeqSphere software version 6.0.2 (Ridom GmbH, Münster, Germany). The categorical coefficient was used to calculate the MST. For K. pneumoniae, this in-house wgMLST scheme was comprised of 4978 genes (3471 core-genome and 1507 accessory-genome targets) using K. pneumoniae MGH 78,578 (NC_009648.1) as a reference genome. For E. coli, 4503 genes (3199 core-genome and 1304 accessory-genome targets) were used with E. coli 536 (CP000247.1) as a reference genome[27]. Genetic clusters were defined as collections of isolates with a maximum of 25 alleles differences for E coli and 20 for K. pneumoniae. For classical MLST, the existing schemes available via SeqSphere were used. AMR genes were identified via ResFinder software[35] and only AMR genes with ≥97% sequence identity with the reference sequences were included. The presence of mcr genes 1 to 10 was also analysed using BLAST (CLC Genomics Workbench version 20.0.3; Qiagen Bioinformatics, Aarhus, Denmark). The presence of plasmid replicons was assessed using PlasmidFinder software[36], including only replicons with 100% sequence identity and that were completely present. For the identification of serotypes and virulence factors, VirulenceFinder[37], SerotypeFinder[38] and Kleborate (https://github.com/katholt/Kleborate) were used. All NGS-derived data were imported into BioNumerics for subsequent analyses. Raw NGS sequence data of all sequenced isolates were deposited in the Sequence Read Archive and plasmids with mcr genes in GenBank of the National Centre for Biotechnology Information (NCBI) under BioProject ID PRJNA754858.

Isolates carrying mcr genes were subjected to third-generation sequencing (TGS; Oxford Nanopore, Oxford, United Kingdom). For TGS, an in-house protocol was used to isolate high-molecular-weight DNA[39]. Bacteria were grown overnight in 1.5 ml Brain heart infusion broth. Subsequently, the culture was spun down for 2 min at 13,000 × g. We washed the pellet and resuspended it in 500 µl of NaCl (150 mM). The resulting suspension was spun down for 5 min at 5000 × g. We resuspended the pellet in 100 µl QuickExtract DNA Extraction Solution (Lucigen) and 0.1 µl Ready-Lyse Lysozyme solution (Epicentre). This was incubated at 37 °C for 1 h and then we added 85 µl 10 mM Tris 1 mM EDTA pH = 8 (1 × TE), 10 µl proteinase K (> 600 mAU/mL, Qiagen) and 5 µl 20% sodium dodecyl sulfate solution. The mixture was incubated for 30 min at 56 °C. Subsequently, 0.1 × volume 3 M sodium acetate pH = 5.2 and 2.5 × volume ice-cold 100% ethanol were added and DNA was allowed to precipitate overnight at −20 °C. This was spun down for 15 min at 13,000 × g. The resulting pellets were washed with 1 ml 70% ethanol and this was spun down again for 5 min at 13,000 × g. The pellet was dried, dissolved in 200 µl 1 × TE and diluted with Nuclease-free water to 1 µg.

After DNA isolation, we used the Oxford Nanopore protocol SQK-LSK109 (https://community.nanoporetech.com) and the expansion kit for native barcoding EXP-NBD104 (Oxford Nanopore Technologies)[39]. No optional shearing of DNA was performed. FFPE and end-repair kits (New England BioLabs) were used to repair the DNA. Barcodes were ligated with 1× bead clean up using AMPure XP (Beckman Coulter Nederland) after each step. Sequencing adaptors were added to pooled barcoded isolates by ligation. The final library was loaded onto a GridION flow cell (MIN-106 R9.4.1). Subsequently, GridION was used for live base calling using the MinKNOW GUI and afterwards de-multiplexing with the Guppy algorithm (ONT Guppy barcoding software version 3.5.1) on a Red Hat Enterprise Linux Server. Reads with length <5000 base pairs were omitted using NanoFilt (version 2.2.0), and then, 50 base pairs of both sides were trimmed using headcrop and tailcrop settings. In addition, FiltLong (version 0.2.0) was used to filter the reads with the 90% highest Qscore and make a subset up to a maximum of 500 Mb. The resulting FASTQ was used as input for assembly. The NGS and TGS data were used in Unicycler v0.4.8 for hybrid assembly[40] to reconstruct chromosomes and plasmids, which were annotated by Prokka v1.14.6 and loaded into BioNumerics for further analyses[41]. NanoPlot (v1.31.0) was used for quality control of the TGS data, where most isolates had coverages of at least 80x chromosome size with only three isolates 10–20x coverage. The read length N50 range was between 35 and 55 kb. NGS data were not trimmed before running Unicycler. Unicycler was run with default settings and verbosity 2. The default depth_filter setting of 0.25 was used. However, when the mcr gene that was found in the NGS data could not be retrieved in the TGS data, a depth_filter of 0.1 was used. Only contigs of >2.5 kb were analysed in this study. Plasmids containing mcr genes were compared with each other using chromosome comparison in BioNumerics and with other previously found mcr-containing plasmids in NCBI BLAST.

Isolates with the mcr-9 gene were repeatedly subcultured, with the aim to cure the isolates from the mcr-9 gene. The absence of the mcr-9 gene was examined using PCR. In case absent, BMD was performed simultaneously on the cured and non-cured isolate and both isolates were subjected to NGS and TGS.

To identify mutations that may potentially lead to colistin resistance, NGS reads were mapped in CLC Genomics Workbench version 20.0.3 against the gene sequences of pmrAB, phoPQ, mgrB and crrB, present in the NCBI sequence database (references for K. pneumoniae: crrB—KY587106, mgrB—MN187248, phoQ—KY587110, pmrA—MG243721, pmrB—KJ626267, phoP—KY587067; E. coli: phoP—NZ_CP038353-Eco, phoQ—NZ_CP038353-Eco, pmrA—NZ_CP038353-Eco, pmrB—NZ_CP038353 Eco).

Metadata collection

Clinical and epidemiological data were extracted from the electronic patient medical records by the participating laboratories and entered into Type-Ned. Microbiological data (including no of detected, tested and positive K. pneumoniae and E. coli isolates and the test on which submission of isolates was based), extracted from the local laboratory information system, and general hospital data (denominator data and use of colistin in SDD/SOD) were entered into web-based questionnaires. In addition, a questionnaire on local laboratory testing policies was composed in collaboration with the Dutch Infectious Disease Surveillance Information System-Antibiotic Resistance (ISIS-AR)[42] prior to start of this study and these data were extracted from ISIS-AR for participating laboratories after this study[42].

The minimal rate of carbapenem-susceptible COLR-EK isolates was calculated as the number of colistin-resistant E. coli and K. pneumoniae isolates, confirmed for this study, divided by the number of patient days or person years, provided by the participating laboratories (cases of hospitals without provided denominator data were subtracted).

Ethical permissions and privacy

The medical ethical committee of the University Medical Centre Utrecht has defined this study (19/262) as not falling under the scope of the Dutch law ‘Wet medisch-wetenschappelijk onderzoek met mensen’ (“Niet WMO-plichtig”). This means that no further medical ethical evaluation by the committee is needed since no additional individual patient data or isolates/materials were collected specifically for this study and no actions were requested from patients. Furthermore, the data collected in this study do not include personally identifiable information. Written or verbal informed consent was therefore not required. The collection and storage of data complied to the General Data Protection Regulation (EU 2016/679).

Statistics and reproducibility

Data are presented as n (%) for categorical variables and mean (standard deviation) or, for variables that have a skewed distribution, median and interquartile range (IQR) [first quartile-third quartile] for numerical variables. Categorical variables with characteristics of matched COLR-EK and COLS-EK isolates/patients were compared using matching-adjusted logistic regression (LR) with correction for variables used for matching (patient location, material of origin and bacterial species) and the odds ratio (OR) with 95% confidence interval (CI) were calculated. When necessary, a matching- and confounding-adjusted logistic regression was performed with additional covariates. For numerical variables with non-normal distribution, the Wilcoxon signed-rank test was used. For the comparison of virulence factors between colistin-resistant and -susceptible E. coli, data was corrected with a Bonferroni multiple testing correction. A two-sided p-value of <0.05 was considered statistically significant. No data were excluded from the analyses. In all analyses, a complete case analysis was performed. The number of patients with available data per variable are mentioned. Our hypothesis is that missing data were mostly missing completely at random, due to the substantial workload or difficulties to find certain information in the electronic patient files (most missing data were observed in variables, such as antibiotic use or colistin use in the previous 6 months and a profession with direct patientcare). However, persons that filled in the questionnaire were not blinded for colistin susceptibility testing results and therefore missing not at random cannot be ruled out. The researcher that analysed the data was also not blinded for colistin-susceptibility testing results. Randomization was not applicable. This study is set up with strict inclusion criteria and a highly structured protocol. Experimental data were generated using well established, reproducible, robust and validated methodology. Where applicable, positive and negative controls were included, which showed highly reproducible results. STATA SE version 15.1 (StataCorp, College Station, TX, USA) was used for data-analysis.

Table 1

Characteristics of patients carrying COLR-EK or COLS-EK isolates.

	COLR-EK (N = 72)			COLS-EK (N = 72)
	N	N total	%	N	N total	%
Median age (interquartile range)		72	73.5 (IQR 56.0-83.0)		72	72.0 (IQR 51.5–78.0)
Female	53	72	73.6%	53	72	73.6%
Species
E. coli	54	72	75.0%	54	72	75.0%
K. pneumoniae	18	72	25.0%	18	72	25.0%
Material
Urine	51	72	70.8%	52	72	72.2%
Urine in case of bladder catheter	4	72	5.6%	3	72	4.2%
Rectum/perineum swab	10	72	13.9%	10	72	13.9%
Faeces	3	72	4.2%	3	72	4.2%
Wound secretion	1	72	1.4%	1	72	1.4%
Blood	2	72	2.8%	2	72	2.8%
Throat swab	1	72	1.4%	1	72	1.4%
Sender of isolate
Hospital	42	72	58.3%	42	72	58.3%
Inpatient	27	42	64.3%	36	42	85.7%
Outpatient	15	42	35.7%	6	42	14.3%
General practitioner	24	72	33.3%	24	72	33.3%
Nursing home/Elderly home/Care centre	6	72	8.3%	6	72	8.3%
Infection	54	69	78.3%	51	69	73.9%
Comorbidity
Renal insufficiency	2	45	4.4%	3	58	5.2%
Immunosuppression	3	45	6.7%	5	58	8.6%
Type 2 diabetes mellitus	3	45	6.7%	5	58	8.6%
Chronic obstructive pulmonary disease	3	45	6.7%	1	58	1.7%
Malignancy	10	45	22.2%	6	58	10.3%
Comorbidity
Previous antibiotic use
Use of colistin in past 6 monthsa	9	43	20.9%	1	40	2.5%
Other antibiotic use in past 6 months	30	41	73.2%	19	34	55.9%
Nursing home/elderly home/rehabilitation centre resident	16	71	22.5%	9	69	13.0%

Characteristics are mentioned per category (in bold).

aAll selective intestinal/oropharyngeal decontamination with colistin.

COLR-EK colistin-resistant E. coli or K. pneumoniae, COLS-EK colistin-susceptible E. coli or K. pneumoniae, IQR interquartile range.

Table 2

Colistin susceptibility testing policies and methods for Enterobacterales.

Policy for colistin susceptibility testing	Colistin susceptibility testing method		Num. of labs
	Initial test	Confirmation test
Always	VITEK	BMDa–c	7
		Etestd	1
		Unknownc	1
Only when considering colistin as treatment	BMD	None	1
	Unknown	Unknownc	1
In case of a combination of factors	BMD	None	2
	Etest or BMD	None	1
	VITEK or BMD	BMD if VITEK was useda	1
		BMD and when ColR NGSa	1
Only for this study			1
No data available			5
Total			22

Confirmation test is performed when: acolistin is considered as treatment, bin case of a combination of criteria, cthe isolate is colistin-resistant in the initial test and has certain characteristics or dthe isolate is colistin-resistant in the initial test.

BMD broth microdilution, ColR colistin-resistant, NGS next-generation sequencing.

Table 3

Colistin susceptibility testing policies and confirmation of submitted isolates.

Used test result for isolate selection	Num. of labs	COLR-EK				COLS-EK
		Colistin susceptibility testing methoda	Submitted	Confirmed	%	Colistin susceptibility testing methoda	Submitted	Confirmed	%
Screening test	15	Automatedb	106	61	57.5%	Automatedc	80	61	76.3%
Confirmation test	7	BMD/Etest	13	11	84.6%	Automated	12	11	91.7%
Total	22		119	72	60.5%		92	72	78.3%

Of note: This tabe also includes isolates that are rejected based on other reasons than an incorrect colistin MIC.

aUnknown method for the five laboratories with no submitted isolates.

bAlso two submitted isolates with disk diffusion and two isolates with an Etest.

cAlso one submitted isolate with disk diffusion.

BMD broth microdilution, COLR-EK colistin-resistant E. coli or K. pneumoniae, COLS-EK colistin-susceptible E. coli or K. pneumoniae.

Table 4

Overview of isolates with mcr genes or non-silent chromosomal mutations, potentially involved in colistin resistance.

Isolate ID	mcr	pmrA	pmrB	mgrB	crrB	phoP	phoQ
K. pneumoniae
RIVM_C019776	No			First 39 bp are absent
RIVM_C019778	No			P74: insertion ISEc68
RIVM_C019785	No			G37S(G109A)
RIVM_C019837	No		T157P(A469C)
RIVM_C000156	No			Absent	955 bp deletion
RIVM_C000119	No			P46: insertion IS-like el
RIVM_C019878	No			P70: insertion IS-like el
RIVM_C019770	mcr-9			C39Y(G116A)
RIVM_C019752	mcr-1.1
RIVM_C000164	mcr-9
E. coli
RIVM_C019737	No		V128E(T382A)
RIVM_C019749	No		T159M(C475T)
RIVM_C019767	No		V91E(T272A)
RIVM_C019769	No						E464D(G1392T)
RIVM_C019789	No		P22: extra GCG(aa A)
RIVM_C019808	No		L105Q(T314A)
RIVM_C019825	No	N67K(C200A + C201A); D68E(C204A)	M4I(G12C)
RIVM_C019864	No		V91E(T272A + A273G)
RIVM_C019866	No		First 48 bp are absent			V88A(T263C)
RIVM_C028932	No		M4I(G12C)
RIVM_C029515	mcr-1.1					I175F(A523T)
RIVM_C000121	mcr-1.1
RIVM_C019762	mcr-1.1
RIVM_C019792	mcr-1.1

Point mutations are indicated by the amino acid substitution (nucleotide substitution) with the number representing the location. Silent mutations and mutations that were present in (both colistin-resistant and) colistin-susceptible isolates were not included in this table.

aa amino acid, bp base pairs, el element, MIC minimum inhibitory concentration, p position.