Alyssa R Golden1, James A Karlowsky1,2, Andrew Walkty1,2, Melanie R Baxter1, Andrew J Denisuik1, Melissa McCracken3, Michael R Mulvey1,3, Heather J Adam1,2, Denice Bay1, George G Zhanel1. 1. Department of Medical Microbiology and Infectious Diseases, Max Rady College of Medicine, Rady Faculty of Health Sciences, University of Manitoba, 727 McDermot Avenue, Winnipeg, Manitoba R3E 3P5, Canada. 2. Department of Clinical Microbiology, Shared Health Manitoba, MS673-820 Sherbrook Street, Winnipeg, Manitoba R3A 1R9, Canada. 3. National Microbiology Laboratory-Public Health Agency of Canada, 1015 Arlington Street, Winnipeg, Manitoba R3E 3R2 Canada.
Abstract
OBJECTIVES: To determine whether the genotypic resistance profile inferred from WGS could accurately predict phenotypic resistance for ESBL-producing Escherichia coli isolated from patient samples in Canadian hospital laboratories. METHODS: As part of the ongoing CANWARD study, 671 E. coli were collected and phenotypically confirmed as ESBL producers using CLSI M100 disc testing criteria. Isolates were sequenced using the Illumina MiSeq platform, resulting in 636 high-quality genomes for comparison. Using a rules-based approach, the genotypic resistance profile was compared with the phenotypic resistance interpretation generated using the CLSI broth microdilution method for ceftriaxone, ciprofloxacin, gentamicin and trimethoprim/sulfamethoxazole. RESULTS: The most common genes associated with non-susceptibility to ceftriaxone, gentamicin and trimethoprim/sulfamethoxazole were CTX-M-15 (n = 391), aac(3)-IIa + aac(6')-Ib-cr (n = 121) and dfrA17 + sul1 (n = 169), respectively. Ciprofloxacin non-susceptibility was most commonly attributed to alterations in both gyrA (S83L + D87N) and parC (S80I + E84V), with (n = 187) or without (n = 197) aac(6')-Ib-cr. Categorical agreement (susceptible or non-susceptible) between actual and predicted phenotype was 95.6%, 98.9%, 97.6% and 88.8% for ceftriaxone, ciprofloxacin, gentamicin and trimethoprim/sulfamethoxazole, respectively. Only ciprofloxacin results (susceptible or non-susceptible) were predicted with major error (ME) and very major error (VME) rates of <3%: ciprofloxacin (ME, 1.5%; VME, 1.1%); gentamicin (ME, 0.8%-31.7%; VME, 4.8%); ceftriaxone (ME, 81.8%; VME, 3.0%); and trimethoprim/sulfamethoxazole (ME, 0.9%-23.0%; VME, 5.2%-8.5%). CONCLUSIONS: Our rules-based approach for predicting a resistance phenotype from WGS performed well for ciprofloxacin, with categorical agreement of 98.9%, an ME rate of 1.5% and a VME rate of 1.1%. Although high categorical agreements were also obtained for gentamicin, ceftriaxone and trimethoprim/sulfamethoxazole, ME and/or VME rates were ≥3%.
OBJECTIVES: To determine whether the genotypic resistance profile inferred from WGS could accurately predict phenotypic resistance for ESBL-producing Escherichia coli isolated from patient samples in Canadian hospital laboratories. METHODS: As part of the ongoing CANWARD study, 671 E. coli were collected and phenotypically confirmed as ESBL producers using CLSI M100 disc testing criteria. Isolates were sequenced using the Illumina MiSeq platform, resulting in 636 high-quality genomes for comparison. Using a rules-based approach, the genotypic resistance profile was compared with the phenotypic resistance interpretation generated using the CLSI broth microdilution method for ceftriaxone, ciprofloxacin, gentamicin and trimethoprim/sulfamethoxazole. RESULTS: The most common genes associated with non-susceptibility to ceftriaxone, gentamicin and trimethoprim/sulfamethoxazole were CTX-M-15 (n = 391), aac(3)-IIa + aac(6')-Ib-cr (n = 121) and dfrA17 + sul1 (n = 169), respectively. Ciprofloxacin non-susceptibility was most commonly attributed to alterations in both gyrA (S83L + D87N) and parC (S80I + E84V), with (n = 187) or without (n = 197) aac(6')-Ib-cr. Categorical agreement (susceptible or non-susceptible) between actual and predicted phenotype was 95.6%, 98.9%, 97.6% and 88.8% for ceftriaxone, ciprofloxacin, gentamicin and trimethoprim/sulfamethoxazole, respectively. Only ciprofloxacin results (susceptible or non-susceptible) were predicted with major error (ME) and very major error (VME) rates of <3%: ciprofloxacin (ME, 1.5%; VME, 1.1%); gentamicin (ME, 0.8%-31.7%; VME, 4.8%); ceftriaxone (ME, 81.8%; VME, 3.0%); and trimethoprim/sulfamethoxazole (ME, 0.9%-23.0%; VME, 5.2%-8.5%). CONCLUSIONS: Our rules-based approach for predicting a resistance phenotype from WGS performed well for ciprofloxacin, with categorical agreement of 98.9%, an ME rate of 1.5% and a VME rate of 1.1%. Although high categorical agreements were also obtained for gentamicin, ceftriaxone and trimethoprim/sulfamethoxazole, ME and/or VME rates were ≥3%.