Co-occurrence of aminoglycoside and ฀-lactam resistance mechanisms in aminoglycoside- non-susceptible Escherichia coli isolated in the Zurich area, Switzerland

: The co-occurrence of aminoglycoside and ฀-lactam resistance was assessed in 3358 consecutive Escherichia coli clinical isolates collected in 2014 in the greater Zurich area, Switzerland. Non-susceptibility to at least one of the tested aminoglycosides was observed in 470/3358 E. coli strains (14%). In strains categorized as broad-spectrum ฀-lactamase (BSBL)-producers (1241/3358 isolates), extended-spectrum ฀-lactamase (ESBL)-producers (262/3358) and AmpC-producers (66/3358), resistance to aminoglycoside was found in 23%, 52% and 20% of the isolates, respectively. In contrast, aminoglycoside-susceptible strains were rarely resistant to ฀-lactams (33/1777, 1.9%). The genomes of 439 aminoglycoside-resistant E. coli were sequenced and aminoglycoside and ฀-lactam genotypes were analysed. The most prevalent aminoglycoside resistance genes were aph(3’)-Ia (133 strains, 30.3%), aac(3)-IId (100 strains, 22.8%), and aac(6’)-Ib-cr (52 strains, 11.8%). The most frequent associations with ฀-lactam resistance genes were aph(3’)-Ia or aac(3)-IId with blaTEM-1 (94 and 72 strains, respec-tively), and aac(3)-IIa/aac(6’)-Ib-cr with blaCTX-M-15/blaOXA-1 (23 strains). These results indicate a frequent association of aac(3) and aph(3’) genotypes with BSBL production, and a frequent co-occurrence of aac(6’) genes with ESBL production. The high rate of co-occurrence of aminoglycoside resistance and ฀-lactamase production must be considered in combination therapy. The co-occurrence of aminoglycoside and β -lactam resistance was assessed in 3358 consecutive Es- cherichia coli clinical isolates collected in 2014 in the greater Zurich area, Switzerland. Non-susceptibility to at least one of the tested aminoglycosides was observed in 470/3358 E. coli strains (14%). In strains categorized as broad-spectrum β -lactamase (BSBL)-producers (1241/3358 isolates), extended-spectrum β -lactamase (ESBL)-producers (262/3358) and AmpC-producers (66/3358), resistance to aminoglycoside was found in 23%, 52% and 20% of the isolates, respectively. In contrast, aminoglycoside-susceptible strains were rarely resistant to β -lactams (33/1777, 1.9%). The genomes of 439 aminoglycoside-resistant E. coli were sequenced and aminoglycoside and β -lactam genotypes were analysed. The most prevalent aminoglycoside resistance genes were aph(3’)-Ia (133 strains, 30.3%), aac(3)-IId (100 strains, 22.8%), and aac(6’)-Ib-cr (52 strains, 11.8%). The most frequent associations with β -lactam resistance genes were aph(3’)-Ia or aac(3)-IId with bla TEM-1 (94 and 72 strains, respectively), and aac(3)-IIa / aac(6’)-Ib-cr with bla CTX-M-15 /bla OXA-1 (23 strains). These results indicate a frequent association of aac(3) and aph(3’) geno- types with BSBL production, and a frequent co-occurrence of aac(6’) genes with ESBL production. The high rate of co-occurrence of aminoglycoside resistance and β -lactamase production must be considered in combination therapy.


Introduction
Aminoglycosides are an important class of bactericidal antibiotics that are frequently used, mostly in combination with β-lactams, to treat severe infections caused by Gram-negative bacteria [1] . Resistance to aminoglycosides has been increasingly reported, including, most worryingly, in association with that to other antibiotic classes, such as β-lactams and fluoroquinolones [2][3][4][5] .
Resistance to aminoglycosides in Gram-negative bacteria is mainly due to the production of aminoglycoside-modifying enzymes (AMEs) [1 , 6] or modification of the ribosome by acquired 16S rRNA methyltransferases (RMTases) [6 , 7] . AME production is by far the most frequent resistance mechanism in E. coli . AMEs are divided into three classes according to the reaction they catalyse: (i) aminoglycoside N -acetyltransferases (AAC), (ii) aminoglycoside O-phosphotransferases (APH) and (iii) aminoglycoside Onucleotidyltransferases (ANT) [1] . AMEs can modify aminoglycosides at various sites of the drug scaffold and the enzymes are classified into subclasses and types according to different substrate profiles. For example, AAC(3) acetylates the amino group at position 3 of the central 4,6-di-substituted deoxystreptamine ring II, AAC(6') acetylates the amino group at position 6' of ring I and APH(3') phosphorylates the hydroxyl group at position 3' of ring I of the aminoglycoside. AMEs frequently modify more than one aminoglycoside and the same aminoglycoside can be affected by several enzymes. Lastly, aminoglycoside modification may not always result in recognizable phenotypic resistance as determined in vitro by assessment of minimal inhibitory concentrations (MICs) [1 , 8] .
The structural genes for AMEs and β-lactamases are often part of mobile genetic elements carried by a variety of plasmids in combination with resistance genes to other drug classes, resulting in multidrug-resistant isolates [9] . Particularly worrisome is the frequent co-occurrence of RMTases and metalloβ-lactamases [7] .
The ever-growing problem of multidrug resistance and the need for carbapenem-sparing regimens to treat infections caused by ESBLproducing Enterobacteriaceae have revived interest in aminoglycosides and efforts to detect and identify the resistance mechanisms against this drug class [8 , 12] . Although there is abundant literature on the microbiological, clinical and epidemiological aspects of βlactam resistance in Enterobacteriaceae, there is little information about aminoglycoside resistance. The aim of this study was to assess the aminoglycoside and β-lactam resistance rates in clinical E. coli isolated in the Zurich metropolitan area in 2014, and to investigate the co-occurrence of aminoglycoside and β-lactam resistance genes.

Clinical isolates
All 5765 E. coli collected during 2014 from various clinical materials in the diagnostic laboratory of the Institute of Medical Microbiology, University of Zurich were included in the study ( Figure S1). When more than one E. coli was isolated from the same patient, only the first strain was included. If aminoglycosidesusceptible and -resistant strains were recovered from the same patient, only the first aminoglycoside-resistant isolate was studied. Each patient was included in the analysis only once. Of the resulting 3358 non-duplicate E. coli , 470 had growth inhibition diameters below the cut-off of at least one of the tested aminoglycosides, gentamicin, tobramycin and kanamycin. A total of 461 aminoglycoside non-wild-type strains were sequenced ( Figure S1) and screened for the presence of aminoglycoside or β-lactam resistance genes.

Antibiotic susceptibility testing (AST)
AST was performed by disk diffusion according to EUCAST guidelines [14] . Aminoglycoside susceptibility profiles were evaluated by disk diffusion according to EUCAST epidemiological cutoffs (ECOFFs, gentamicin and tobramycin = 16 mm [15] ) or local ECOFF (kanamycin = 15 mm) [16] . A cefpodoxime cut-off of 21 mm was used for screening of ESBL production. Carbapenemase production was suspected if the meropenem inhibition zone was below 25 mm, or if the meropenem inhibition zone was between 25 and 28 mm and the piperacillin/tazobactam inhibition zone diameter was below 17 mm according to EUCAST [17] . Screening for AmpC, ESBL and carbapenemase production was performed according to phenotype-based algorithms described previously [18][19][20] . In brief, AmpC production was suspected if cefoxitin inhibition zones were below 19 mm. Results were confirmed using combination disk testing: for AmpC the difference between cefoxitin with and without cloxacillin was measured; for ESBL the difference was determined between cefotaxime/clavulanic acid vs. cefotaxime, and ceftazidime/clavulanic acid vs. ceftazidime. Strains resistant to any β-lactam, but not producing an AmpC, ESBL or carbapenemase, were classified as BSBL. During 2014, all 3358 E. coli were tested for gentamicin and tobramycin susceptibility. In addition, 3011 strains were tested for kanamycin susceptibility.

Whole-genome sequencing (WGS)
DNA libraries were prepared following the Illumina Nextera protocol (Illumina, San Diego, CA, USA) or the QIAseq FX DNA Library Kit (QIAGEN AG, Hombrechtikon, Switzerland). Quality control of the library was performed using capillary electrophoresis (Fragment Analyzer Automated CE System by Advanced Analytical). Sequencing was done on either the HiSeq 150 0 or MiSeq platform (Illumina).

Detection of resistance genes
The fastq sequence files were processed by the ARIBA pipeline [21] and Resistance Gene Identifier 4.0.1 [22] . Resistance genes were identified using ARG-ANNOT (Antibiotic Resistance Gene-ANNOTation) [23] and CARD (Comprehensive Antibiotic Resistance Database) [22] .

Sequence type (ST) analysis
Sequence typing was performed according to the Warwick scheme [24] with RidomSeqsphere software 4.1.9 (Ridom GmbH, Muenster, Germany).

Statistical analysis
R (version 3.6.1) was used for statistical analysis [25] . Fisher's exact test, R base version, was used. A P -value below 0.05 was considered statistically significant.

Phenotypic aminoglycoside resistance
ECOFFs were used to screen for aminoglycoside-non-susceptible E. coli as ECOFFS separate wild-type from non-wild-type populations more accurately than clinical breakpoints (CBPs) [6 , 8 , 26] . A total of 5765 E. coli strains collected from various clinical materials during 2014 were analysed. When more than one E. coli strain was isolated from the same patient, only the first aminoglycosideresistant strain, if available, was included. Of 3358 clinical isolates, 2888 (86%) were susceptible to gentamicin, tobramycin and kanamycin. The remaining 470 strains (14%) were resistant to at least one aminoglycoside ( Fig. 1 and Table 1 ). Seven aminoglycoside resistance phenotypes were observed (Table S1). The resistance rate observed is somewhat higher than that reported in 2015 by ANRESIS (8.9%), the Swiss centre for antibiotic resistance for Switzerland [27] , and this may be due to several reasons. First, the ANRESIS report includes strains isolated from hospitalized patients and outpatients, whereas most of the E. coli analysed in the current study were collected from the University Hospital of Zurich, a  Table 1 Aminoglycoside non-susceptibility rates grouped by β-lactam resistance mechanism in 3358 unique E. coli .  [28] . Non-susceptibility to gentamicin, tobramycin or kanamycin individually was found in 270 (8%), 311 (9.3%) and 301 (9.0%) strains, respectively. These rates are comparable to those reported by the ECDC for Europe in 2014; e.g., France, Germany and Austria have resistance rates to tobramycin and/or gentamicin of 7.7%, 6.9% and 6.9% in invasive E. coli isolates, respectively [29] .

Identification of aminoglycoside resistance genes
The aminoglycoside resistance genes present in the current study isolates were then investigated. Altogether, the resistance mechanisms in 439 E. coli clinical isolates were determined by WGS and are given in Table 2 . Unfortunately, nine strains were no longer available and a further 22 strains were excluded for phenotype-genotype discrepancies. Overall, 31 resistance genotypes were found. Fourteen consisted of a single determinant specifying an AME and 17 consisted of various gene combinations of AMEs and 16S rRNA methyltransferases. The most prevalent genes for individual AMEs were aph(3')-Ia (133/439, 30.3%), aac(3)-IId (100/439, 22.8%) and aac(6')-Ib-cr (52/439, 11.8%) ( Table S2).
In a study conducted by Miró et al . , 264 aminoglycoside nonsusceptible E. coli clinical isolates collected in a Spanish hospital in 2006 were analysed and the most prevalent AME genes conferring resistance to gentamicin, tobramycin and amikacin were aph(3')-Ia (13.9%), aac(3)-IIa (12.4%), and aac(6')-Ib (4.2%) [32] . Interestingly, despite the different criteria used to select the strains (ECOFF vs. CBP), the relative prevalence of the resistance mechanisms was similar to that in the current study. In another work in 2009, the most frequent AMEs in a collection of 105 E. coli resistant or intermediately resistant to gentamicin and/or tobramycin were AAC(3)-II (66.7%) and AAC(6')-Ib (10.2%) [33] . Of note, in the latter work neither the susceptibility to kanamycin nor the presence of the aph(3')-I gene were investigated.

Co-occurrence of aminoglycoside and β-lactam resistance genes
The presence of β-lactam resistance genes was investigated in the 439 aminoglycoside non-susceptible E. coli ( Table 2 ). Due to their small numbers, 11 AmpC and 4 ESBL/AmpC-producers were not included in Fig. 2 and will not be discussed further. Fig. 2 shows the co-occurrence of aminoglycoside and β-lactam resistance genes in 424 strains classified as β-lactam wild-type, or BSBL-or ESBL-producers (Fisher's exact test P = 0.0 0 04998). Based on genotypic data, the isolates were grouped by aminoglycoside resistance mechanism conferring the same inferred resistance phenotype ( Table 3 ). Thus, all strains only carrying an aph(3') gene were classified as kanamycin-resistant [34] . Strains harbouring an aac(3') gene were considered as resistant to gentamicin, kanamycin and tobramycin [8] . Resistance to kanamycin, tobramycin and amikacin was inferred from the presence of an aac(6') gene [6 , 8] .
In a previous a study, 105 of 257 E. coli isolates resistant to amoxicillin/clavulanic acid were also resistant to at least one aminoglycoside [2] . Of 15 strains producing TEM-1 alone, eight carried aph(3')-Ia , four carried aac(3)-IIa and one isolate harboured ant(2")-Ia . Among 30 OXA-1 producing strains, four harboured an aac(6')-Ib gene and three a combination of aac(6')-Ib and aph(3')-Ia . Contrary to the current findings, in 21/30 strains the OXA-1 enzyme was associated with an ANT(2")-Ia. Curiously, in strains carrying a combination of ESBL and OXA-1, co-occurrence with AAC(6')-Ib was found in 23/24 strains. Although these observations point to an association of β-lactam and aminoglycoside resistance mechanisms, the full extent of this association can only be assessed by studying large numbers of corresponding isolates by WGS, as in the current study.

Co-localization of aminoglycoside and β-lactam resistance genes
To address whether the described resistance mechanisms are located on a common mobile element, long-read sequencing is necessary. Although we do not have this data, the carriage of aac(6')-Ib-cr, bla CTX-M-1/9/15 , bla TEM-1 , and bla OXA-1 on IncF plasmids has been well established, particularly in association with FII, FIA and FIB replicons [37] . Incorporation of aac(6')-Ib-cr / bla CTX-M-15 / bla OXA-1 in an IncFII plasmid is common [36] . This probably explains the high co-occurrence of CTX-M-15/OXA-1/AAC(6')-Ib-cr observed in the current study. Similar multidrugresistance plasmids, encoding combinations of TEM-1 with AAC(3) and/or APH(3'), may be responsible for most co-occurrences in this study.

Co-evolution of aminoglycoside and β-lactam resistance
The pattern of co-resistance described here is puzzling. In general, the prevalence and co-occurrence of resistance mechanisms most likely reflects the history of antibiotic usage. AMEs Table 2 Aminoglycoside and β-lactam resistance genes in 439 E. coli  active against the first marketed aminoglycosides (gentamicin, tobramycin and kanamycin), such as APH(3')-I [34 , 38] and AAC(3)-II [34 , 38] , were often found in association with BSBLs, which confer resistance towards first-generation β-lactams [6 , 39] . Indeed, strains resistant to first-generation BSBLs, such as cephalotin, were frequently resistant to gentamicin and tobramycin [40][41][42] .
Thus, the evolution of co-occurrence of aminoglycoside and βlactam resistance mechanisms is ongoing. Beginning in the 1960s with APH(3') and AAC(3) associated with BSBLs, this was followed by AAC(6') associated with ESBLs in the 1980s, and more recently by RMTases associated with carbapenemases [7 , 44 , 45] . The latter two mechanisms confer resistance to virtually all β-lactam and aminoglycoside antibiotics currently available in clinical practice, including plazomicin, the most recently developed aminoglycoside antibiotic [46] .

Sequence types
To determine whether clonal spread is involved in the cooccurrence of resistance described, the sequence types (STs) of the E. coli genomes were analysed. This revealed a wide diversity of 76 STs (Table S3). The most abundant, ST131, was found in 124 strains associated with several β-lactam and aminoglycoside resistance genes. The most frequent genotypes were aac(3)-IId /

Conclusions
In conclusion, aminoglycoside resistance and the prevalence of AMEs in E. coli in the greater Zurich area are comparable to reports from other countries, such as Spain, Poland [4] and  Table S4. Norway. To the best of our knowledge this study is the first to examine the co-occurrence of β-lactamase and AME genes by WGS. Non-susceptibility to aminoglycosides was caused by a remarkable variety of AMEs and was predominantly due to aph(3')-Ia, aac(3)-II and aac(6')-Ib-cr , which are mostly associated with various types of β-lactamases. Non-susceptibility to aminoglycosides was rarely found in β-lactam-susceptible clinical isolates of E. coli. The frequent co-occurrence of AMEs/RMTases conferring resistance to all aminoglycosides available needs careful consideration, particularly for ESBL/carbapenemase-producing Enterobacteriaceae .