Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2020 Dec 1.
Published in final edited form as: J Glob Antimicrob Resist. 2019 Apr 18;19:136–143. doi: 10.1016/j.jgar.2019.04.006

The use of Online Tools for Antimicrobial Resistance Prediction by Whole Genome Sequencing in MRSA and VRE

Ahmed Babiker 1,2, Mustapha M Mustapha 2, Marissa P Pacey 2, Kathleen A Shutt 1,2, Chinelo D Ezeonwuka 2, Sara L Ohm 2, Vaughn S Cooper 3, Jane W Marsh 2, Yohei Doi 1, Lee H Harrison 1,2
PMCID: PMC6800622  NIHMSID: NIHMS1527347  PMID: 31005733

Abstract

Objectives:

The antimicrobial resistance (AMR) crisis represents a serious threat to public health and has resulted in concentrated efforts to accelerate development of rapid molecular diagnostics for AMR. In combination with publicly-available web-based AMR databases, whole genome sequencing (WGS) offers the capacity for rapid detection of antibiotic resistance genes. Here we studied the concordance between WGS-based resistance prediction and phenotypic susceptibility testing results for methicillin-resistant Staphylococcus aureus (MRSA) and vancomycin resistant Enterococcus (VRE) clinical isolates using publicly-available tools and databases.

Methods:

Clinical isolates prospectively collected at the University of Pittsburgh Medical Center between December 2016 and December 2017 underwent WGS. Antibiotic resistance gene content was assessed from assembled genomes by BLASTn search of online databases. Concordance between WGS-predicted resistance profile and phenotypic susceptibility as well as sensitivity, specificity, positive and negative predictive values (NPV, PPV) were calculated for each antibiotic/organism combination, using the phenotypic results as the gold standard.

Results

Phenotypic susceptibility testing and WGS results were available for 1242 isolate/antibiotic combinations. Overall concordance was 99.3% with a sensitivity, specificity, PPV, NPV of 98.7% (95% CI, 97.2-99.5%), 99.6% (95 % CI, 98.8-99.9%), 99.3% (95% CI, 98.0-99.8%), 99.2% (95% CI, 98.3-99.7%), respectively. Additional identification of point mutations in housekeeping genes increased the concordance to 99.4% and the sensitivity to 99.3% (95% CI, 98.2-99.8%) and NPV to 99.4% (95% CI, 98.4-99.8%).

Conclusion

WGS can be used as a reliable predicator of phenotypic resistance for both MRSA and VRE using readily-available online tools.

Keywords: methicillin-resistant Staphylococcus aureus, vancomycin resistant Enterococcus, whole genome sequencing, antimicrobial resistance

1. Introduction

The antimicrobial resistance (AMR) crisis represents a serious threat to the public health and economy, claiming an estimated 23,000 deaths in the United States each year (1). Methicillin-resistant Staphylococcus aureus (MRSA) and vancomycin-resistant Enterococcus (VRE) are among the most common healthcare-associated antimicrobial-resistant pathogens and cause significant morbidity and mortality. The impact of increasing AMR has resulted in concerted efforts to develop rapid molecular diagnostics of resistant pathogens, as current culture-based phenotypic susceptibility assays require up to 48–72 hours for the results to become available (2,3).

With the current advances in sequencing technology, there has been increasing utilization of whole genome sequencing (WGS) for species identification, antimicrobial susceptibility prediction and outbreak detection (4). WGS has the potential to deliver sequencing data from clinical samples in a short time frame, allowing for earlier tailored therapy (5, 6). Previous studies have revealed high concordance between WGS and conventional methods of detecting antimicrobial resistance among Gram-positive pathogens (7-10).

In parallel to the advancement of sequencing methods, development of bioinformatics tools and web-based databases with acquired resistance genes provides user-friendly methods for identification of resistance genes in whole genome data (11, 12). These web-based databases require little formal training and can become practical tools for the clinical microbiology laboratories. In addition, there have been recent breakthroughs in developing platforms which can create automatic genotype-based AMR reports from raw sequencing data, providing further proof of concept (9).

The purpose of this study was to examine the concordance of AMR predicted by WGS and web-based resistance databases with conventional phenotypic susceptibility testing methods among contemporary MRSA and VRE clinical isolates.

2. Materials and Methods

2.1. Study setting and population

The study isolates were prospectively collected at the University of Pittsburgh Medical Center Presbyterian Hospital (UPMC) between December 2016 and December 2017. The isolates were sequenced in batch routinely as part of a larger study to assess the utility of prospective WGS surveillance linked with electronic medical record mining in early detection of hospital outbreaks (13, 14). Isolates included represent unique patients isolates. The study was approved by the institutional review board at the University of Pittsburgh.

2.2. Whole-genome sequencing, assembly, and resistance gene detection

Genomic DNA was extracted from pure overnight cultures of single bacterial colonies and extracted using Qiagen DNAeasy Tissue Kit according to manufacturer’s instructions (Qiagen, Germantown, MD). Library construction and sequencing were conducted using the Illumina Nextera DNA Sample Prep Kit with 150bp paired-end read length and sequenced on the NextSeq whole-genome sequencing platform (Illumina, San Diego, CA). Reads were trimmed using CutAdapt in Trim Galore v0.4.1(15) and then de novo assembled using SPAdes v3.10 (16) from filtered short-read sequences. Sequence types were identified using MLST (github.com/tseeman/mlst) and new sequence types were submitted to the PubMLST website (pubmlst.org).

All MRSA genomes and majority (96/100, 96%) of VRE isolates underwent core-genome SNP analyses. Four E. faecalis genomes were excluded from core-genome analyses due to small sample size. SNPs were identified using Snippy (17) with default parameters using the best available genome assembly from our hospital for each species as reference genomes. VRE reference genome, VRE32553 was assembled from Pacific Biosciences long sequencing reads while the MRSA reference, MRSA10173 was assembled from short read Illumina sequences. A phylogenetic tree based on the core SNP alignment was generated using RAxML v8.2.9 (18) by running 100 bootstrap replicates under the generalized time-reversible model of evolution, a categorical model of rate heterogeneity (GTR-CAT) and Lewis correction for ascertainment bias and visualized using Interactive Tree Of Life (ITOL) v4 (19).

2.3. In silico prediction of AMR using online tools

AMR gene content was assessed by BLASTn of assembled contigs against downloaded ResFinder, and the Comprehensive Antibiotic Resistance Database (CARD) databases with 80% sequence identity and 90% sequence coverage cut-off (11, 12). These online platforms analyze WGS data directly from uploaded sequence files (in FASTQ format) and/or or assembled genome sequences (contigs in FASTA format). Isolates were considered resistant by genotyping if they contained at least one AMR gene known to confer resistance to that class of antibiotic (Table 1).

Table 1:

Frequency of Mechanisms of Genotypic Resistance among clinical MRSA and VRE isolates

Antimicrobial Agent No. (%) MRSA isolates 1 No. (%)VRE isolates 1
Erythromycin/Clindamycin
ermA 42 (42.0) 0 (0)
ermC 6 (6.0) 0 (0)
ermB 0 (0) 48 (85.7)
ermG 0 (0) 1 (1.8)
efmA 0 (0) 39 (69.6)
emeA 0 (0) 3 (5.4)
msrA 55 (55.0) 1 (1.7)
msrC 0 (0) 54 (90.0)
mphC 55 (55.0) 0 (0)
Gentamicin
aph(2″)-I 2 (1.9) n/a
Linezolid
23S rRNA mutation2 0 (0) 2 (2.1)
Oxacillin
mecA 105 (100) 0 (0)
Tetracycline
tetK 6 (5.6) 0 (0)
tetM 3 (2.8) 28 (58.3)
tetU 0 42 (81.3)
tetL 0 25 (52.1)
Sulfamethaxozole-trimethoprim 7 (6.5) n/a
dfrG 6 (5.7) 0 (0)
drfC 1 (1.0) 0 (0)
Rifampin
rpoB gene mutations3 4 (3.9)
Vancomycin
vanA 0 (0) 96 (96.0)
Open in a new tab
1

Percentages were determined by dividing the number of isolates harboring the gene by the total number of isolates (per species) with phenotypic testing for drug of interest

2

Linezolid 23S rRNA gene mutations: T1547C and T1245C

3

Rifampin rpoB gene mutation: A477D R484H and H481N

2.4. Phenotypic susceptibility testing

Routine susceptibility testing was performed by Microscan WalkAway™ (Siemens Healthcare Diagnostics) and susceptibility determined using reference Clinical and Laboratory Standards Institute (CLSI) breakpoints (20). Intermediate results were considered resistant for the purpose of this study. Antibiotics included in the analysis were those routinely tested at our institution, deemed of highest clinical relevance and have established genetic resistance determinants that can be readily identified from the online databases mentioned above. These included erythromycin, clindamycin, gentamicin, linezolid, methicillin, rifampin, trimethoprim-sulfamethoxazole, tetracycline and vancomycin. Not all isolates where tested against all drugs, as susceptibility panels used on isolates depend on specimen types.

2.5. Resolution of discordance between phenotypic and genotypic resistance testing

For isolates that showed discordance between genotypic and phenotypic methods, repeat susceptibility testing was performed by the Kirby Bauer disk diffusion method. Susceptibility was determined using reference Clinical and Laboratory Standards Institute (CLSI) breakpoints (20). For isolates with discordant results for antibiotics known to have resistance conferred by amino acid substitutions in particular housekeeping genes (Table 1) comparison to reference strains Staphylococcus aureus MSSA476 (GenBank accession number: BX571857.1) or Enterococcus faecalis ATCC 29212 (GenBank accession number: CP008816.1) using CLC Genomics Workbench v11 (Qiagen) and Molecular Evolutionary Genetics Analysis (MEGA) software (21) was undertaken. Raw sequencing reads for all isolates were deposited in GenBank under accession numbers SAMN09400893-SAMN09401217.

2.6. Statistical analysis

Concordance between WGS-predicted and phenotypic methods was determined. The sensitivity, specificity, positive predictive values (PPV) and negative predictive values (NPV) for WGS-predicted resistance were calculated for each antibiotic/organism combination with the phenotypic results as the gold standard using Wilson Brown Method (22). In cases of discordance, disk diffusion results were considered the gold standard. The statistical calculations were performed with GraphPad Prism v7 (GraphPad Software, San Diego, California).

3. Results

Phenotypic susceptibility testing and WGS results were available for 108 and 100 unique MRSA and VRE isolates, respectively (Figure 1 and 2). Of 1242 isolate/antibiotic combinations, overall concordance was 99.3% with a sensitivity, specificity, PPV, NPV of 98.7% (95% CI, 97.2-99.5%), 99.6% (95 % CI, 98.8-99.9%), 99.3% (95% CI, 98.0-99.8%), 99.2% (95% CI, 98.3-99.7%), respectively. Additional identification of point mutations in housekeeping genes increased the concordance to 99.4% and the sensitivity to 99.3% (95% CI, 98.2-99.8%) and NPV to 99.4% (95% CI, 98.4-99.8%).

Figure 1:

Figure 1:

Open in a new tab

Genetic relationships and antimicrobial resistance (AMR) gene content among 108 methicillin resistant S. aureus strains. Maximum-likelihood phylogenetic tree constructed from aligned core genome SNPs (left); color strip indicates multi-locus sequence type (ST), while gray boxes (right panel) indicate presence of corresponding AMR gene.

Figure 2:

Figure 2:

Open in a new tab

Genetic relationships and antimicrobial resistance (AMR) gene content among 96 vancomycin resistant E. faecium strains. Maximum-likelihood phylogenetic tree constructed from aligned core genome SNPs (left); color strip indicates multi-locus sequence type (ST), while gray boxes (right panel) indicate presence of corresponding AMR gene.

3.1. Concordance in MRSA

Phenotypic susceptibility testing and WGS results were available for 108 unique MRSA isolates. (Figure 1). Of 942 isolate/antibiotic combinations, initial concordance between the two methods was 96.7%. There were 31 discrepancies between WGS-predicted resistance and automated phenotypic susceptibility testing. Of the 31 discrepant results, 23 results were reconciled by disk diffusion testing, increasing concordance to 99.5%. As a result, WGS-based prediction had an overall sensitivity, specificity, PPV and NPV of 98.5% (95% CI, 96.2-99.6%), 99.9% (95% CI, 99.2-100%), 99.7% (95% CI 97.4-100%) and 99.4% (95% CI, 98.5-99.8%), respectively (Table 2).

Table 2:

Genotypic and phenotypic concordance among clinical MRSA isolates

Antibiotic No.
isolates		 Phenotypic resistance Genotypic
resistance		 Concordance
with
Disk
Diffusion
(%)		 Sensitivity
(95%
CI) (%)		 Specificity
(95%
CI)
(%)		 PPV
(95%
CI)
(%)		 NPV
(95% CI)
(%)
Automated
No. (%)		 Disk
Diffusion
No. (%)		 Inferred resistant
by WGS
No. (%)
Methicillin 107 107
(100)		 105
(98.1)		 105
(98.1)		 100 100
(96.6-100)		 100
(15.8-100)		 100
(96.6-100)		 100
(15.8-100)
Erythromycin 100 88
(88.0)		 87
(87.0)		 84
(84.0)		 96.0 96.6
(90.3-99.3)		 92.3
(64.0-99.8)		 98.8
(92.7-99.8)		 80.0
(56.6-92.5)
Clindamycin 100 34
(34.0)		 52
(52.0)		 51
(51.0)		 99.0 98.1
(89.9-100)		 100
(92.5-100)		 100
(92.5-100)		 97.9
(87.1-99.7)
Tetracycline 105 10
(9.5)		 9
(8.6)		 9
(8.6)		 100 100
(66.4-100)		 100
(96.2-100)		 100
(66.4-100)		 100
(96.2-100)
Gentamicin 108 2
(1.9)		 2
(1.9)		 2
(1.9)		 100 100
(15.8-100)		 100
(96.6-100)		 100
(15.8-100)		 100
(96.6-100)
Trimethopri m-Sulfamethoxa zole 105 7
(6.7)		 6
(5.7)		 6
(5.7)		 100 100
(59.0-100)		 100
(96.3-100)		 100
(59.0-100)		 100
(96.3-100)
Rifampin 103 6
(5.8)		 4
(3.9)		 2
(1.9)		 100 100
(37.7-100)		 100
(96.3-100)		 100
(0.5-100)		 100
(96.3-100)
Vancomycin 107 0 0 0 100 n/a 100
(96.6-100)		 n/a 100
(96.6-100)
Linezolid 107 0 0 0 100 n/a 100
(96.6-100)		 n/a 100
(96.6-100)
Total 1 942 255
(27.1)		 267
(28.3)		 263
(27.9)		 99.5 98.5
(96.2-99.6)		 99.9
(99.2-100)		 99.6
(97.4-100)		 99.4
(98.5-99.8)
Open in a new tab
1

Total results for all MRSA isolate/antibiotic combinations

3.1.1. Methicillin:

mecA encodes penicillin binding protein 2a, which has a low affinity for beta-lactam antibiotics and confers methicillin resistance (23). Initial susceptibility testing revealed that 108 (100%) isolates were resistant to methicillin. The mecA gene was identified in 105 (98.1%) MRSA isolates (Table 1). Repeat testing by disk diffusion revealed that 2 isolates without mecA were rather methicillin-susceptible S. aureus (MSSA). Therefore, WGS showed 100% concordance with phenotypic methicillin susceptibility testing results, with two instances of initial discordance resolved by repeat susceptibility testing (Table 2).

3.1.2. Erythromycin:

Resistance to the macrolides, lincosamides and streptogramin B ( MLSB) group antibiotics can result either from the N6-dimethylation of an adenine residue in the 23S rRNA by erm genes, causing reduced affinity with the antibiotic, enzymatic modification (phosphorylation) of the antibiotic mediated by acquisition of mph genes, or by active transport of the antibiotic out of the cell via efflux proteins encoded by msr genes (24). Initial susceptibility testing revealed that 88 (88.0%) MRSA isolates were resistant to erythromycin. By WGS, 169 macrolide resistance genes (Table 1) were identified, with the presence of at least one gene detected in 84 (84.0%) of isolates. There were 5 isolates that showed discordance. Four isolates were erythromycin-resistant with no macrolide resistance gene identified. Repeat susceptibility testing by disk diffusion resulted in one isolate being reclassified as susceptible. Of the discordant isolates, one initially tested as susceptible despite the presence of macrolide resistance genes, msrA (98.0% identity) and mph(25) (99.3% identity) remained susceptible upon repeat disk diffusion testing. Overall, WGS was 96.0% concordant with phenotypic erythromycin susceptibility testing results with one instance of discordance resolved by disk diffusion (Table 2).

3.1.3. Clindamycin:

Resistance to the lincosamide antibiotics is mediated either through inactivation by lincosamide nucleotidyltransferase enzymes encoded by lnu genes, methylation of 23S ribosomal RNA through erm genes, or by active transport of the antibiotic out of the cell by efflux pumps encoded by the lsa genes or ABC transporter proteins encoded by vga genes (24).Initial susceptibility testing revealed 34 (34.0%) MRSA isolates to be resistant to clindamycin. By WGS, 53 lincosamide resistance genes were identified (Table 2) with 51 (51.0%) isolates having at least one lincosamide resistance gene. There were 18 discordant isolates. Seventeen isolates were susceptible with the presence of at least one resistance gene. All 17 isolates tested positive for inducible clindamycin resistance by disk diffusion. One isolate was resistant without the presence of a lincosamide resistance gene and repeat susceptibility testing confirmed it as resistant. WGS was therefore 99.0% concordant with phenotypic clindamycin susceptibility testing results (Table 2).

3.1.4. Tetracycline:

Resistance to tetracycline antibiotics is conferred by ribosomal protection proteins which alter the binding of tetracycline to bacterial ribosomes, or through active transport out of the cell by efflux pumps, both of which are encoded by tet genes (26, 27). Initial susceptibility testing revealed that 10 (9.5%) MRSA isolates were resistant to tetracycline. By WGS, tetracycline resistance genes were identified in 9 (8.6%) MRSA isolates. One isolate initially tested as tetracycline-intermediate without the presence of a tetracycline resistance gene. Repeat susceptibility testing determined the isolate to be tetracycline-susceptible. WGS was therefore 100% concordant with phenotypic tetracycline susceptibility testing results (Table 2).

3.1.5. Gentamicin:

Aminoglycoside resistance is mediated by aac, ant, and aph genes which encode aminoglycoside N-acetyltransferases, O-adenylyltransferases and O-phosphotransferases, respectively (28). Initial susceptibility testing revealed that 2 (1.9%) MRSA isolates were resistant and both possessed aph(2″)-I gene. Thus, WGS was 100% concordant with phenotypic gentamicin susceptibility testing results (Table 2).

3.1.6. Trimethoprim-Sulfamethoxazole (SXT):

Resistance to trimethoprim is conferred either through the production of dihydrofolate reductase (DHFR) variants encoded by dfr genes or by an amino acid substitution in the dfrB housekeeping gene (29). Initial susceptibility testing revealed that 7 (6.7%) MRSA isolates were SXT-resistant. WGS revealed 6 isolates with the SXT resistance genes (5.7%). There was one discordant isolate with a dfrG gene which was susceptible on initial testing but tested resistant on repeat testing. Thus, WGS was 100% concordant with phenotypic SXT susceptibility testing results (Table 2).

3.1.7. Rifampin:

Resistance to rifampin is mediated either through point mutations in the chromosomal gene rpoB which encodes the bacterial RNA polymerase (30) or less commonly by arr enzymes that catalyze ADP-ribosylation of rifamycins (31). Initial susceptibility testing revealed that 5 (4.8%) MRSA isolates were resistant to rifampin. WGS revealed 4 (4.0%) isolate with resistance-conferring mutation on rpoB genes (H481N, R484H, A477D) (32, 33). Repeat susceptibility testing revealed that 4 (3.9%) MRSA isolates were rifampin-resistant. WGS was 100% concordant with phenotypic rifampin results (Table 2).

3.1.8. Vancomycin and Linezolid:

No MRSA isolates were resistant to vancomycin or linezolid, and none harbored known vancomycin or linezolid resistance genes.

3.2. Concordance in VRE

Phenotypic susceptibility and WGS data were available for 100 unique VRE isolates (Figure 2). Out of a total of 300 isolate/antibiotic combinations, initial genotypic/phenotypic concordance was 96.3%. There were 11 discrepancies between WGS-predicted resistance and automated phenotypic susceptibility testing. Of the 11 discrepant results, 7 could be reconciled by disk diffusion testing, increasing concordance to 98.7%, with a sensitivity, specificity, PPV and NPV of 99.0% (CI 95% 96.4-99.9%), 98.0% (CI 95% 93.0-99.8%), 99.0% (CI 95% 96.1-99.7%), 98.0%. (CI 95% 92.6-99.5%), respectively. Additional identification of point mutations in housekeeping genes increased the concordance to 99.3%, sensitivity to 100% (95% CI, 99.1-100%) and NPV to 100% (95% CI, 96.8-100%).

3.2.1. Vancomycin:

Resistance to vancomycin is encoded by different clusters of genes referred to as the van gene clusters which results in the replacement of D-Ala-D-Ala–ending peptidoglycan precursors with D-alanyl-D-lactate termini, to which vancomycin binds with substantially lower affinity (34). Initial susceptibility testing revealed that all 100 (100%) VRE isolates, by definition, were resistant to vancomycin. WGS revealed the presence of the vanA gene in 96 (96.9%) isolates. Repeat susceptibility of 4 isolates without a van gene revealed the 4 isolates to be vancomycin-susceptible. Thus, WGS was 100% concordant with vancomycin susceptibility testing results (Table 3).

Table 3:

Results of WGS for Prediction of AMR among clinical VRE isolates

Antibiotic No.
Isolates		 Phenotypic
Resistance		 Genotypic
Resistance		 Concordance
with
Disk
Diffusion
(%)		 Sensitivity
(95%
CI) (%)		 Specificity
(95%
CI)
(%)		 PPV
(95%
CI)
(%)		 NPV
(95%
CI)
(%)
Automated
No,
(%)		 Disk
Diffusion,
No, (%)		 Inferred resistant
by WGS
No. (%)
Vancomycin 100 100
(100)		 96
(96.0)		 96
(96.0)		 100 100
(96.2-100)		 100
(39.8-100)		 100
(96.2-100)		 100
(39.8-100)
Linezolid1 96 3
(3.1)		 2
(2.1)		 0 97.9 0
(0-84.2)		 100
(96.2-100)		 n/a 97.9
(97.9-97.9)
Erythromycin 56 56
(100)		 56
(100)		 56
(100)		 100 100.00
(93.6-100)		 n/a 100
(93.6-100)		 n/a
Tetracycline 48 44
(91.7)		 45
(93.8)		 47
(97.9)		 95.8 100
(92.1-10)		 33.3
(0.82-90.6)		 95.7
(91.0-98.0)		 100
(5.0-100)
Total2,3 300 203
(67.7)		 199
(66.3)		 199
(66.3)		 98.7 99.0
(96.4-99.9)		 98.0
(93.0-99.8)		 99.0
(96.1-99.7)		 98.0
(92.6-99.5)
Open in a new tab
1

With the additional step of examining for point mutations in housekeeping genes concordance and NPV were increased to 100% (95 CI, 96.2-100%)

2

Total results for all VRE isolate/antibiotic combinations

3

Additional identification of point mutations in housekeeping genes increased the concordance to 99.3%, sensitivity to 100% (95% CI, 99.1-100%) and NPV to 100% (95% CI, 96.8-100%).

3.2.2. Linezolid:

Resistance to linezolid is most frequently caused by a point mutation within the 23S rRNA, with the 23S rRNA G2576T mutation being frequent (35). Additionally resistance may occur by acquisition of the multidrug resistance (MDR) gene cfr which encodes an rRNA methyltransferase that adds a methyl group at the C-8 position of 23S rRNA nucleotide A2503, thus preventing the drug from binding to the target site (36). Alterations in the ribosomal proteins L3, L4 and L22, encoded by rplC, rplD and rplV, respectively, have also been associated with increased resistance to linezolid (37). Initial susceptibility testing revealed 3 (3.1%) isolates that were resistant to linezolid. WGS did not reveal the presence of any cfr genes. Repeat testing revealed 1 VRE isolate to be linezolid-susceptible. The 2 remaining discordant VRE isolates were found to have point mutations in the 23S rRNA (T1547C and T1245C), which may have resulted in functional alteration in 23S rRNA. WGS (not including additional examination for housekeeping gene mutation) was 97.9% concordant with linezolid susceptibility testing results. With identification of point mutations in housekeeping genes, sensitivity, specificity, PPV and NPV were 100% (Table 3).

3.2.3. Erythromycin:

Initial susceptibility testing revealed that all 56 VRE isolates were resistant to erythromycin. By WGS 145 macrolide resistance genes were identified (Table 2), with the presence of at least one resistance gene in all VRE isolates. WGS was 100% concordant with erythromycin phenotypic susceptibility testing with sensitivity, specificity, PPV and NPV of 100% (Table 3).

3.2.4. Tetracycline:

Initial susceptibility testing revealed that 44 (91.7%) VRE isolates were tetracycline-resistant. By WGS, tetracycline resistance genes were identified in 47 isolates (97.9%). Three initial discordant results were found. After repeat susceptibility testing, two instances of discordance remained with two isolates susceptible to tetracycline despite the presence of tetU (99.4% identity) gene in one isolate and tetM (97.3% identity) and tetL (96.6% identity) genes in another isolate. WGS was 95.8% concordant with phenotypic tetracycline susceptibility (Table 3).

4. Discussion

Infections caused by resistant organisms are associated with increased morbidity, mortality and economic burden (38, 39). Our results show high concordance between WGS-predicted resistance and phenotypic susceptibility testing by use of publicly available online tools. Rapid determination of AMR profiles can lead to decrease in time to appropriate therapy, as was recently demonstrated by Tamma et al. and hence minimize the unintended consequences of antimicrobials and improve patient outcomes (2, 40). Although manual upload of genomes to online databases may be sufficient, significant improvement in throughput and workflow could be obtained by implementing command line versions of these tools using programmed analysis pipelines. With falling costs, reduced turnaround times and increased sequence quality, WGS has the potential to become a more routinely used tool in clinical microbiology laboratories (10). And while rapid diagnostics of resistance detection are having decreasing turnaround times (as little as 8h), with sequencing one gains a host of valuable information which can provide high discriminatory power for prediction of antibiotic resistance and molecular epidemiology making it a valuable tool not only for clinical management but also for infection control and surveillance (25).

While similar results have previously been demonstrated among MRSA isolates (8, 9, 41), to our knowledge, validation of WGS for AMR prediction among human clinical isolates of VRE is lacking. Most previous investigations have used in-house curated databases of resistance determinants (8, 9, 42). Given the large body of literature on the genetic basis of resistance for MRSA and VRE, AMR determinants are well documented and represented in online databases (11, 12). In our study, we used publicly available web-based AMR databases and achieved high concordance with gold standard phenotypic methods. One notable exception was low concordance rates for erythromycin resistance in MRSA. This may be due to the instability of ermC-containing plasmids which may be lost during isolate passage in the laboratory (42, 43). In most instances of discrepancy between automated phenotypic susceptibility results and sequencing data, repeat susceptibility with disk diffusion validated the WGS data. Discrepancy between automated susceptibility testing and gold standard methods such as broth microdilution and disk diffusion among S. aureus and Enterococcus has been reported with higher MICs reported by automated methods (44, 45). These results can lead to increased use of broad-spectrum antibiotics as confirmatory testing with broth microdilution and disk diffusion is an extra step which may be foregone.

A major barrier to the wide-spread adoption of genomic methods is the lack of bioinformatics expertise. In a recently developed software package “Mykrobe Predictor”, a known panel of AMR genes and point mutation sites are used in analysis of raw sequencing data is run to generate a user friendly report, circumventing the need for bioinformatics expertise (9). Similarly, even without the use of such a platform, with implementation of an automated pipeline, one can receive a simple file format and with basic computer navigational skills upload genome sequences directly to the different web-based databases and receive easy to interpret results (11, 12).

There are several limitations of our study. First, although we had high overall levels of sensitivity, specificity, PPV and NPV, our process may fail to identify resistance conferred by point mutations in chromosomal housekeeping genes. With the additional step of manual sequence inspection, point mutations associated with resistance could be identified in a few instances and increase the sensitivity and specificity of WGS for detection of AMR. Nevertheless, this process can be timeconsuming and laborious and may requires additional bioinformatics software and skill. The more recently developed ARG ANOT is the first database to include detection of point mutations in chromosomal genes associated with antimicrobial resistance (46). Furthermore, there was no vancomycin or linezolid resistance and only a low level of aminoglycoside resistance amongst our MRSA isolates so we could not confidently evaluate resistance prediction for those agents. Additionally, while we attempted to include most relevant antibiotics, daptomycin was not included in the analysis due to the current gap in the knowledge of genetic mechanisms of resistance for daptomycin, and therefore these mechanisms are not fully represented in online databases. (47). Lastly, while we focused on MRSA and VRE due to their clinical significance, further validation should be undertaken in susceptible organisms.

5. Conclusion

By using WGS and online AMR databases we are able to achieve high concordance with phenotypic susceptibility testing among clinical isolates of MRSA and VRE. With the increasing investment in furthering genomic analysis and investigation of genetic basis of resistance, additional prediction tools may soon become available. Nonetheless, there remains a current need for a central comprehensive open source, curated database containing all validated AMR genes as well as point mutations in housekeeping genes known to be associated with AMR.

Highlights.

  • Whole genome sequencing can rapidly deliver genetic antimicrobial resistance data

  • Using genetic antimicrobial resistance data phenotypic resistance can be predicted

  • This predication can be performed using easy to use online resistance databases

  • High concordance exits between genetic and phenotypic resistance amongst MRSA and VRE

Acknowledgements

We thank Ryan Shields for his expertise and input, the members of the Center for Innovative Antimicrobial Therapy, the Microbial Genomic Epidemiology and the Cooper Laboratories for their assistance with sequencing and phenotypic susceptibility testing and Lloyd Clarke for his outstanding database support.

Funding: This study was funded in part by the National Institute of Allergy and Infectious Diseases (R21Al109459 and R01AI127472).

Footnotes

Competing Interests: None declared

Ethical aApproval: Not required

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

References

  • 1.CDC. Antibiotic resistance threats in the United States, 2013.; 2013. [Google Scholar]
  • 2.Caliendo AM, Gilbert DN, Ginocchio CC, Hanson KE, May L, Quinn TC, et al. Better tests, better care: improved diagnostics for infectious diseases. Clinical infectious diseases : an official publication of the Infectious Diseases Society of America. 2013;57 Suppl 3:S139–70. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3.Evans SR, Hujer AM, Jiang H, Hujer KM, Hall T, Marzan C, et al. Rapid Molecular Diagnostics, Antibiotic Treatment Decisions, and Developing Approaches to Inform Empiric Therapy: PRIMERS I and II. Clinical infectious diseases : an official publication of the Infectious Diseases Society of America. 2016;62(2):181–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4.Didelot X, Bowden R, Wilson DJ, Peto TEA, Crook DW. Transforming clinical microbiology with bacterial genome sequencing. Nature reviews Genetics. 2012;13(9):601–12. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5.Grumaz S, Stevens P, Grumaz C, Decker SO, Weigand MA, Hofer S, et al. Next-generation sequencing diagnostics of bacteremia in septic patients. Genome medicine. 2016;8(1):73. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6.Schmidt K, Mwaigwisya S, Crossman LC, Doumith M, Munroe D, Pires C, et al. Identification of bacterial pathogens and antimicrobial resistance directly from clinical urines by nanopore-based metagenomic sequencing. The Journal of antimicrobial chemotherapy. 2017;72(1):104–14. [DOI] [PubMed] [Google Scholar]
  • 7.Zankari E, Hasman H, Kaas RS, Seyfarth AM, Agerso Y, Lund O, et al. Genotyping using whole-genome sequencing is a realistic alternative to surveillance based on phenotypic antimicrobial susceptibility testing. The Journal of antimicrobial chemotherapy. 2013;68(4):771–7. [DOI] [PubMed] [Google Scholar]
  • 8.Gordon NC, Price JR, Cole K, Everitt R, Morgan M, Finney J, et al. Prediction of Staphylococcus aureus antimicrobial resistance by whole-genome sequencing. Journal of clinical microbiology. 2014;52(4):1182–91. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9.Bradley P, Gordon NC, Walker TM, Dunn L, Heys S, Huang B, et al. Rapid antibiotic-resistance predictions from genome sequence data for Staphylococcus aureus and Mycobacterium tuberculosis. Nature communications. 2015;6:10063. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10.Ellington MJ, Ekelund O, Aarestrup FM, Canton R, Doumith M, Giske C, et al. The role of whole genome sequencing in antimicrobial susceptibility testing of bacteria: report from the EUCAST Subcommittee. Clinical microbiology and infection : the official publication of the European Society of Clinical Microbiology and Infectious Diseases. 2017;23(1):2–22. [DOI] [PubMed] [Google Scholar]
  • 11.Zankari E, Hasman H, Cosentino S, Vestergaard M, Rasmussen S, Lund O, et al. Identification of acquired antimicrobial resistance genes. The Journal of antimicrobial chemotherapy. 2012;67(11):2640–4. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12.McArthur AG, Waglechner N, Nizam F, Yan A, Azad MA, Baylay AJ, et al. The comprehensive antibiotic resistance database. Antimicrobial agents and chemotherapy. 2013;57(7):3348–57. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13.Miller JK, Chen J, Sundermann A, Marsh JW, Saul MI, Shutt KA, et al. Statistical outbreak detection by joining medical records and pathogen similarity. Journal of Biomedical Informatics. 2019;91:103126. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14.Sundermann AJ, Miller JK, Marsh JW, Saul MI, Shutt KA, Pacey M, et al. Automated data mining of the electronic health record for investigation of healthcare-associated outbreaks. Infection Control & Hospital Epidemiology.1–6. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15.Krueger F Babraham Bioinformatics - Trim Galore! 2016. [Available from: https://github.com/FelixKrueger/TrimGalore.
  • 16.Bankevich A, Nurk S, Antipov D, Gurevich AA, Dvorkin M, Kulikov AS, et al. SPAdes: a new genome assembly algorithm and its applications to single-cell sequencing. Journal of computational biology : a journal of computational molecular cell biology. 2012;19(5):455–77. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17.Seemann T Snippy: fast bacterial variant calling from NGS reads 2015. [Available from: https://github.com/tseemann/snippy.
  • 18.Stamatakis A RAxML version 8: a tool for phylogenetic analysis and post-analysis of large phylogenies. Bioinformatics (Oxford, England). 2014;30(9):1312–3. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19.Letunic I, Bork P. Interactive tree of life (iTOL) v3: an online tool for the display and annotation of phylogenetic and other trees. Nucleic acids research. 2016;44(W1):W242–5. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20.Wayne PA. M100-S27, Performance standards for antimicrobial susceptibility testing, 27th informational supplement. Clinical and Laboratory Standards Institute; 2017. [Google Scholar]
  • 21.Kumar S, Stecher G, Tamura K. MEGA7: Molecular Evolutionary Genetics Analysis Version 7.0 for Bigger Datasets. Molecular biology and evolution. 2016;33(7): 1870–4. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22.Wilson EB. Probable inference, the law of succession, and statistical inference. Journal of the American Statistical Association. 1972;22:209–12. [Google Scholar]
  • 23.Utsui Y, Yokota T. Role of an altered penicillin-binding protein in methicillin- and cephem-resistant Staphylococcus aureus. Antimicrobial agents and chemotherapy. 1985;28(3):397–403. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24.Leclercq R Mechanisms of resistance to macrolides and lincosamides: nature of the resistance elements and their clinical implications. Clinical infectious diseases : an official publication of the Infectious Diseases Society of America. 2002;34(4):482–92. [DOI] [PubMed] [Google Scholar]
  • 25.Charnot-Katsikas A, Tesic V, Love N, Hill B, Bethel C, Boonlayangoor S, et al. Use of the Accelerate Pheno System for Identification and Antimicrobial Susceptibility Testing of Pathogens in Positive Blood Cultures and Impact on Time to Results and Workflow. Journal of clinical microbiology. 2018;56(1). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26.McMurry L, Petrucci RE Jr., Levy SB. Active efflux of tetracycline encoded by four genetically different tetracycline resistance determinants in Escherichia coli. Proceedings of the National Academy of Sciences of the United States of America. 1980;77(7):3974–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27.Connell SR, Tracz DM, Nierhaus KH, Taylor DE. Ribosomal protection proteins and their mechanism of tetracycline resistance. Antimicrobial agents and chemotherapy. 2003;47(12):3675–81. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 28.Ramirez MS, Tolmasky ME. Aminoglycoside modifying enzymes. Drug resistance updates : reviews and commentaries in antimicrobial and anticancer chemotherapy. 2010;13(6):151–71. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29.Huovinen P Resistance to trimethoprim-sulfamethoxazole. Clinical infectious diseases : an official publication of the Infectious Diseases Society of America. 2001;32(11):1608–14. [DOI] [PubMed] [Google Scholar]
  • 30.Floss HG, Yu TW. Rifamycin-mode of action, resistance, and biosynthesis. Chemical reviews. 2005;105(2):621–32. [DOI] [PubMed] [Google Scholar]
  • 31.Baysarowich J, Koteva K, Hughes DW, Ejim L, Griffiths E, Zhang K, et al. Rifamycin antibiotic resistance by ADP-ribosylation: Structure and diversity of Arr. Proceedings of the National Academy of Sciences of the United States of America. 2008;105(12):4886–91. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32.Aubry-Damon H, Soussy CJ, Courvalin P. Characterization of mutations in the rpoB gene that confer rifampin resistance in Staphylococcus aureus. Antimicrobial agents and chemotherapy. 1998;42(10):2590–4. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 33.Villar M, Marimon JM, Garcia-Arenzana JM, de la Campa AG, Ferrandiz MJ, Perez-Trallero E. Epidemiological and molecular aspects of rifampicin-resistant Staphylococcus aureus isolated from wounds, blood and respiratory samples. The Journal of antimicrobial chemotherapy. 2011;66(5):997–1000. [DOI] [PubMed] [Google Scholar]
  • 34.Courvalin P Vancomycin resistance in gram-positive cocci. Clinical infectious diseases : an official publication of the Infectious Diseases Society of America. 2006;42 Suppl 1:S25–34. [DOI] [PubMed] [Google Scholar]
  • 35.Mendes RE, Deshpande LM, Jones RN. Linezolid update: stable in vitro activity following more than a decade of clinical use and summary of associated resistance mechanisms. Drug resistance updates : reviews and commentaries in antimicrobial and anticancer chemotherapy. 2014;17(1-2):1–12. [DOI] [PubMed] [Google Scholar]
  • 36.Giessing AM, Jensen SS, Rasmussen A, Hansen LH, Gondela A, Long K, et al. Identification of 8-methyladenosine as the modification catalyzed by the radical SAM methyltransferase Cfr that confers antibiotic resistance in bacteria. RNA (New York, NY). 2009;15(2):327–36. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 37.Locke JB, Hilgers M, Shaw KJ. Novel ribosomal mutations in Staphylococcus aureus strains identified through selection with the oxazolidinones linezolid and torezolid (TR-700). Antimicrobial agents and chemotherapy. 2009;53(12):5265–74. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 38.DiazGranados CA, Zimmer SM, Klein M, Jernigan JA. Comparison of mortality associated with vancomycin-resistant and vancomycin-susceptible enterococcal bloodstream infections: a meta-analysis. Clinical infectious diseases : an official publication of the Infectious Diseases Society of America. 2005;41(3):327–33. [DOI] [PubMed] [Google Scholar]
  • 39.Cosgrove SE. The relationship between antimicrobial resistance and patient outcomes: mortality, length of hospital stay, and health care costs. Clinical infectious diseases : an official publication of the Infectious Diseases Society of America. 2006;42 Suppl 2:S82–9. [DOI] [PubMed] [Google Scholar]
  • 40.Tamma PD, Fan Y, Bergman Y, Pertea G, Kazmi AQ, Lewis S, et al. Applying Rapid Whole-Genome Sequencing To Predict Phenotypic Antimicrobial Susceptibility Testing Results among Carbapenem-Resistant Klebsiella pneumoniae Clinical Isolates. Antimicrobial agents and chemotherapy. 2019;63(1). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 41.Koser CU, Holden MT, Ellington MJ, Cartwright EJ, Brown NM, Ogilvy-Stuart AL, et al. Rapid whole-genome sequencing for investigation of a neonatal MRSA outbreak. The New England journal of medicine. 2012;366(24):2267–75. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 42.Holden MT, Hsu LY, Kurt K, Weinert LA, Mather AE, Harris SR, et al. A genomic portrait of the emergence, evolution, and global spread of a methicillin-resistant Staphylococcus aureus pandemic. Genome research. 2013;23(4):653–64. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 43.Weisblum B Insights into erythromycin action from studies of its activity as inducer of resistance. Antimicrobial agents and chemotherapy. 1995;39(4):797–805. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 44.Kruzel MC, Lewis CT, Welsh KJ, Lewis EM, Dundas NE, Mohr JF, et al. Determination of vancomycin and daptomycin MICs by different testing methods for methicillin-resistant Staphylococcus aureus. Journal of clinical microbiology. 2011;49(6):2272–3. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 45.Palavecino EL, Burnell JM. False daptomycin-nonsusceptible MIC results by Microscan panel PC 29 relative to Etest results for Staphylococcus aureus and enterococci. Journal of clinical microbiology. 2013;51(1):281–3. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 46.Gupta SK, Padmanabhan BR, Diene SM, Lopez-Rojas R, Kempf M, Landraud L, et al. ARG-ANNOT, a new bioinformatic tool to discover antibiotic resistance genes in bacterial genomes. Antimicrobial agents and chemotherapy. 2014;58(1):212–20. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 47.Tran TT, Munita JM, Arias CA. Mechanisms of drug resistance: daptomycin resistance. Annals of the New York Academy of Sciences. 2015;1354:32–53. [DOI] [PMC free article] [PubMed] [Google Scholar]

RESOURCES