Tracking the Emergence of Azithromycin Resistance in Multiple Genotypes of Typhoidal Salmonella

ABSTRACT The rising prevalence of antimicrobial resistance in Salmonella enterica serovars Typhi and Paratyphi A, causative agents of typhoid and paratyphoid, have led to fears of untreatable infections. Of specific concern is the emerging resistance against azithromycin, the only remaining oral drug to treat extensively drug resistant (XDR) typhoid. Since the first report of azithromycin resistance from Bangladesh in 2019, cases have been reported from Nepal, India, and Pakistan. The genetic basis of this resistance is a single point mutation in the efflux pump AcrB (R717Q/L). Here, we report 38 additional cases of azithromycin-resistant (AzmR) Salmonella Typhi and Paratyphi A isolated in Bangladesh between 2016 and 2018. Using genomic analysis of 56 AzmR isolates from South Asia with AcrB-R717Q/L, we confirm that this mutation has spontaneously emerged in different Salmonella Typhi and Paratyphi A genotypes. The largest cluster of AzmR Typhi belonged to genotype 4.3.1.1; Bayesian analysis predicts the mutation to have emerged sometime in 2010. A travel-related Typhi isolate with AcrB-R717Q belonging to 4.3.1.1 was isolated in the United Kingdom, increasing fears of global spread. For real-time detection of AcrB-R717Q/L, we developed an extraction-free, rapid, and low-cost mismatch amplification mutation assay (MAMA). Validation of MAMA using 113 AzmR and non-AzmR isolates yielded >98% specificity and sensitivity versus phenotypic and whole-genome sequencing assays currently used for azithromycin resistance detection. With increasing azithromycin use, AcrB-R717Q/L is likely to be acquired by XDR strains. The proposed tool for active detection and surveillance of this mutation may detect pan-oral drug resistance early, giving us a window to intervene.

resistant Salmonella Typhi is imminent. We developed a low-cost, rapid PCR tool to facilitate real-time detection and prevention policies. KEYWORDS typhoid, Salmonella Typhi infection, paratyphoid fever, antimicrobial drug resistance, azithromycin, Bangladesh, AMR, Paratyphi, Typhi, paratyphoid, typhoid S almonella enterica serovars Typhi and Paratyphi A, which cause typhoid and paratyphoid, respectively, are estimated to be responsible for 14.3 million illnesses and 136,000 deaths globally each year (1). The majority of the burden lay disproportionately on low-and middle-income countries (LMICs), specifically South Asia and Sub-Saharan Africa. Typhoidal Salmonella has exhibited resistance to all antimicrobials widely used to treat typhoid and paratyphoid, leaving limited options for treatment and raising fears of untreatable infections.
In the early 1900s, mortality due to typhoid/paratyphoid exceeded 30% in many areas, but ampicillin, chloramphenicol, and trimethoprim-sulfamethoxazole were instrumental in reducing it to ,1% (2)(3)(4). Simultaneous resistance to all three first line of drugs, defined as multidrug resistance (MDR), started emerging in the 1980s; MDR strains carried an IncHI1 plasmid with multiple resistance genes (5)(6)(7). As a result, the primary treatment for enteric fever shifted to fluoroquinolones, but soon there were reports of decreasing fluoroquinolone susceptibility due to the rise of point mutations in the gyrase and topoisomerase genes (8,9). Finally, in addition to a few sporadic reports of ceftriaxone resistance (10,11), in 2016, an outbreak of extensively drug-resistant (XDR) Salmonella Typhi (resistant to chloramphenicol, ampicillin, trimethoprimsulfamethoxazole, streptomycin, fluoroquinolones, and third-generation cephalosporins) was identified in Pakistan; to date, more than 11,000 cases have been confirmed (12). Cephalosporin resistance of the XDR strains is caused by the acquisition of a broad-spectrum beta-lactamase resistance gene on a plasmid (13). For patients with uncomplicated XDR typhoid in Pakistan, azithromycin is the last oral option. With the increasing use of this antimicrobial in South Asia, the number of reports of azithromycin-resistant (AzmR) Salmonella Typhi is on the rise.
Our group demonstrated that the underlying mechanism of resistance is a point mutation at amino acid position 717 (R717Q/L) in the efflux pump encoded by the acrB gene (14). The presence of this single nucleotide polymorphism (SNP) raises the MICs of Salmonella Typhi and Paratyphi A for azithromycin to $32 mg/ml, the Clinical and Laboratory Standards Institute (CLSI) breakpoint of resistance (15). Since the report from Bangladesh in 2019, at least six AzmR Salmonella Typhi cases mediated by this mutation have been reported from Nepal, India, and Pakistan, the epicenter of the XDR typhoid outbreak (16)(17)(18). However, no further AzmR Salmonella Paratyphi A has been described. This makes it imperative to gain deeper insights into the evolution and spread of this mutation in typhoidal Salmonella. Low-cost and rapid diagnostic methods to identify this mutation need to be developed, since it can currently only be identified using whole-genome sequencing (WGS).
Leveraging our ongoing enteric fever surveillance system (19), which yielded 3,025 Salmonella Typhi and Paratyphi A isolates since 2016 in Bangladesh, and the available public data of other reports of AzmR strains from the region, we used WGS and comparative genomics to understand the emergence, evolution, and spread of azithromycin resistance conferred for the AcrB-R717 mutation. In addition, we developed a PCRbased mismatch amplification mutation assay (MAMA) to detect AcrB-R717 mutations in typhoidal Salmonella serovars, which is a low-cost and straightforward tool to rapidly identify and track this mutation.

RESULTS
Azithromycin resistance in Bangladesh. Between October 2016 and July 2018, we screened 2,519 isolates of Salmonella Typhi and 506 Salmonella Paratyphi A for azithromycin resistance using disc diffusion and identified 59 Salmonella Typhi and 45 Salmonella Paratyphi A potentially resistant isolates (zone diameters of #12 mm [15]).
Further assessment of these 104 isolates by Etest identified 32 Salmonella Typhi and 6 Salmonella Paratyphi A isolates with MIC $32 mg/ml, confirming them as AzmR isolates (20). In addition, we identified two Salmonella Typhi and three Salmonella Paratyphi A isolates with MICs of 24 mg/ml; these isolates are not considered AzmR by the CLSI guidelines, but the MIC is much higher than for other nonresistant isolates (see Table S1 in the supplemental material).
We conducted WGS analyses for all 32 AzmR Salmonella Typhi and 6 AzmR Paratyphi A isolates, as well as the 5 non-AzmR isolates with high MICs. All 32 AzmR Salmonella Typhi isolates had the AcrB-R717Q/L mutation; 29 had AcrB-R717Q, and three had AcrB-R717L. Five Paratyphi A isolates contained AcrB-R717Q. Of the five non-AzmR isolates with a high MIC (24 mg/ml), one Salmonella Typhi also contained an AcrB-R717Q mutation, whereas the other four did not. For downstream analysis of Bangladeshi isolates, we included the 38 isolates with the AcrB-R717Q/L mutation identified here (including the one with an MIC of 24 mg/ml) and another 13 isolates from our previous study on the emergence of azithromycin resistance in Bangladesh during 2009 to 2016, taking the total AzmR-resistant isolates identified in Bangladesh to 51.
At least two instances of the independent emergence of AcrB-R717 mutations were also observed in the phylogenetic tree of Salmonella Paratyphi A (Fig. 3). This tree included five AzmR and 15 randomly selected sensitive Salmonella Paratyphi A isolates, contextualized against 121 publicly available genomes (see Table S3) (14,23,24). The five Salmonella Paratyphi A with AcrB-R717Q isolated between 2016 and 2018 clustered together, whereas the only genome with AcrB-R717L belonged to a distant and much smaller Bangladesh-specific cluster (Fig. 3).
Emergence and transmission of AcrB-R717Q mutation within the predominant genotype 4.3.1.1. As mentioned above, the largest cluster of AzmR strains in Bangladesh belonged to genotype 4.3.1.1 (n = 29). In addition, the genome of Salmonella Typhi isolated in the United Kingdom that contained the AcrB-R717Q mutation also belonged to the same cluster, and was related to travel to Bangladesh (14,25). To gain insights into the history of the emergence of this mutation, we performed evolutionary analysis using the Bayesian Evolutionary Analysis Sampling Trees (BEAST) program, focusing on these 30 Bangladeshi Salmonella Typhi genomes containing the AcrB-R717Q mutation and eight closely related genomes with AcrB-WT as identified from the phylogenetic analysis. Salmonella Typhi strain P-stx-12 that belongs to genotype 4.3.1.1 and has a complete annotated genome was also included. The maximum clade credibility tree obtained indicated that 4.3.1.1 AcrB-R717Q cluster emerged between 2010 and   (Fig. 4), not long before it was first detected in our passive typhoid surveillance in 2013. All AzmR isolates of this cluster have four unique SNPs in genes STY2741, STY0519, and STY1399, which were also described previously (14).
Rapid and low-cost PCR detection of AcrB-R717 mutation. For rapid and lowcost detection of azithromycin resistance in typhoidal Salmonella, we designed a mismatch amplification mutation assay (MAMA) to detect the SNP of the acrB gene that leads to the AcrB-R717Q/L mutation. The primal focus of this rapid mismatch assay was nucleotide 2150 (part of the codon for amino acid 717) of the acrB gene. A universal forward primer AcrB-UFP, targeting the conserved upstream region (nucleotides 1770 to 1789) of the mutation, was designed. The reverse primer AcrB-MAMA-R was designed so that it partially matches the wild-type allele of the acrB gene with a mismatch at the third nucleotide (A to C) from the 39 end; this increased allelic discrimination ( Fig. 5A and B). A single nucleotide discrepancy in the AcrB-MAMA-R primer has minimal impact on the PCR yield but terminates the reaction if an additional neighboring mismatch is present, such as the SNP at position 2150 ( Fig. 5B and C) (26,27). A control set of primers (ParC-F and ParC-R) targeting the highly conserved parC (topoisomerase IV) gene of Salmonella Typhi and Salmonella Paratyphi A was also designed as a PCR control so that it can be run in a multiplex fashion with the AcrB primers described above (Fig. 5A). Further details of this PCR are provided in Materials and Methods. The entire process of detection of the mutation using this procedure with an isolate takes ;3.5 h; the step of boiling and centrifugation takes ;25 min, reaction mix preparation and PCR takes ;120 min, and gel preparation and electrophoresis takes ;60 min.
To validate the assay, 51 Salmonella Typhi and Paratyphi A AcrB-R717Q/L mutants (38 found in this study and 13 reported by Hooda et al. [14]) and 62 randomly selected isolates (47 Salmonella Typhi and 15 Salmonella Paratyphi A) with AcrB-WT were chosen from our collection (see Table S1). PCR using AcrB-UFP and AcrB-MAMA-R primers generated a 397-bp amplicon in wild-type (WT) strains. As expected, there was no band for AcrB mutants due to PCR inhibition. In all cases, ParC control primers yielded a 758-bp band indicating successful PCR amplification (see Fig. S2). The test set of 113 Salmonella Typhi and Paratyphi A AcrB R717Q/L mutants and WT isolates detected via MAMA exhibited 100% sensitivity and 100% specificity compared to WGS data (Fig. 5D) and 98% sensitivity and 98.4% specificity compared to Etest results (Fig. 5E). All PCRs for validation were conducted using boiled DNA, bypassing the step of DNA extraction to reduce time and cost and to save resources.

DISCUSSION
The rising prevalence of resistance and the drying pipeline of new antimicrobials have left us with a few options of oral drugs to treat typhoidal Salmonella. In this study, we describe the increasing resistance of Salmonella Typhi and Paratyphi A against azithromycin, the last oral antimicrobial against uncomplicated typhoid fever, in Bangladesh and other countries of endemicity. Since the first confirmed identification of AzmR Salmonella Typhi, conferred by a SNP, from Bangladesh, there has been an increasing number of reports from surrounding countries in South Asia. With the largest collection of AzmR isolates, our data, in concordance with other publications, show that azithromycin Tracking Azithromycin-Resistant Typhoidal Salmonella ® resistance has arisen independently multiple times in different genotypes in Salmonella Typhi. We further go on to show similar spontaneous emergence in Paratyphi A isolates. In order to vigilantly track this mutation and guide treatment options and decisions, we developed a low-cost PCR tool for quick detection of the mutation, AcrB-R717Q/L, responsible for azithromycin resistance.
Since 2013, we have identified over 50 Typhi and Salmonella Paratyphi A isolates in Bangladesh that carried the AcrB-R717Q/L mutation and were resistant to azithromycin. Phylogenetic analysis showed the presence of this SNP in several different genotypes, indicating independent emergence of the AcrB-R717 mutation in both Salmonella Typhi and Paratyphi A ( Fig. 2A and 3). In Salmonella Typhi, we also noted that the diversity of genotypes carrying the AzmR mutation increases over time, most likely due to the strong selection pressure placed by increasing use of azithromycin. Use of this antimicrobial is common in many countries of endemicity, including Pakistan and Bangladesh, where the drug is sold over-the-counter without a prescription (28).
In Bangladesh, the AcrB-R717Q/L mutation is most frequently observed in Salmonella Typhi genotype 4.3.1.1 (see Fig. S3), which is also the predominant circulating genotype in the country ( Fig. 2A) (22). Using Bayesian analysis, we predict that the first AcrB-R717 mutation in this genotype appeared between 2010 and 2012, which subsequently led to the detection of 29 cases in our surveillance in Dhaka, Bangladesh (14). A travel-related AzmR isolate identified in the United Kingdom belonged to the same cluster, confirming the fear of global transmission of this mutation. In addition, the XDR Salmonella Typhi outbreak strain of Pakistan also belongs to genotype 4.3.1.1 (13), and recently, the same AcrB mutation was identified in this genotype in Pakistan (17). Fortunately, the Pakistani AzmR strain was not XDR, but the gradual takeover of the entire Salmonella Typhi population by this genotype heralds the emergence of pan-oral drug-resistant (PoDR; resistant to chloramphenicol, ampicillin, trimethoprim-sulfamethoxazole, fluoroquinolones, thirdgeneration cephalosporins, and azithromycin) Salmonella Typhi unless new effective oral drugs are widely introduced (29).
Monitoring this AcrB mutation is essential to track its spread and forecast PoDR typhoid outbreaks. Currently, the only method of detecting this mutation is by WGS, which is cost and resource extensive and not possible to continuously conduct in LMICs in real-time to guide rapid public health action. To overcome this barrier, we developed a mismatch amplification mutation assay that can reliably detect the AcrB-717 mutation using a simple step of conventional PCR, without the requirement of the time-consuming and expensive steps of DNA extraction. It costs ,USD 1.5, requires a conventional PCR machine, demands little expertise, and can be performed in a basic laboratory, delivering results in ;3.5 h. We validated this method by using a mix of 113 resistant and sensitive Salmonella Typhi and Paratyphi A isolates, and this assay showed 100% concordance with the WGS data. Compared to Etest results, it showed more than 98% sensitivity and specificity.
Considering the quick turnaround times, high sensitivity, and specificity of this test, it may be used in accompaniment with phenotypic susceptivity tests for azithromycin, which takes .16 h after isolation of the bacteria (;18 h). In addition, some laboratories are unable to conduct MIC tests as E-strips are expensive and, as shown in this and previous studies, the results of disk diffusion assays of azithromycin are unreliable (only 36.5% are true resistant isolates) (14). The interpretation of the PCR assay is simple, requiring no additional training: the presence or absence of a 397-nucleotide band indicates the absence or presence of the mutation. We have also incorporated an internal control using the conserved gene parC to confirm valid PCR runs. It must be noted, however, that (i) this assay cannot distinguish between the two mutations (R717Q and R717L), which are also phenotypically indistinguishable, and (ii) new modes of azithromycin resistance will not be detected through this method.
This study highlights the increasing and spontaneous occurrence of azithromycin resistance in both Salmonella serovars Typhi and Paratyphi A. Although no azithromycin-resistant XDR isolate has been reported to date, (i) the increasing use of azithromycin, (ii) the spontaneous acquisition of AcrB-R717 mutations, (iii) the clear historical record of widespread dissemination of resistance to all previously widely used antimicrobials by Salmonella Typhi, and (iv) the global spread of XDR strains (3,4,44) suggest that strains resistant to almost all oral antimicrobials is only a matter of time. Acquisition of the plasmid that confers cephalosporin resistance in XDR strains by the Bangladeshi AzmR strains or acquisition of the AcrB mutation in the XDR strains in Pakistan could be the end of oral treatment for typhoid. After fluoroquinolone-nonsusceptible isolates appeared, they became the dominant strains within a decade. Similarly, XDR Salmonella Typhi isolates now account for .70% of Salmonella Typhi isolated in Pakistan, less than 5 years after their first appearance (30,31). This underscores the risk of PoDR Salmonella Typhi becoming the dominant type in the near future once it appears. This would pose serious threats to the health system of LMICs, where typhoid is endemic. Institution of immediate preventative measures, such as improved water, sanitation, and hygiene, and active AMR surveillance to track the resistant genotypes are imperative to interrupt this trend toward the emergence and spread of PoDR.

MATERIALS AND METHODS
Study site and patient enrollment. In this study, we report data from enteric fever surveillance conducted in the inpatient departments of the two largest pediatric hospitals of Bangladesh-Dhaka Shishu (Children) Hospital and Shishu Shasthya (Child Health) Foundation Hospital-and three branches of an Tracking Azithromycin-Resistant Typhoidal Salmonella ® January/February 2021 Volume 12 Issue 1 e03481-20 mbio.asm.org 9 outpatient-based clinic, the Popular Diagnostic Center. These sites include sentinel sites of the World Health Organization-supported Invasive Bacterial Vaccine Preventable Diseases surveillance platform (32) and the Surveillance of Enteric Fever Project of Sabin Vaccine Institute (33). Etiology detection and antimicrobial susceptibility testing. Identification of blood culture positive Salmonella Typhi/Paratyphi A isolates was first confirmed by standard biochemical tests. Next, all Salmonella species were confirmed using polyvalent Salmonella O antisera (A-S MAST Assure; Mast Group, Ltd., Liverpool, UK). All Typhi and Paratyphi A serovars were confirmed using the monovalent antisera factor-d and factor-O2 (MAST Assure), respectively, accompanied by Salmonella-specific antisera. For antimicrobial susceptibility testing (AST), the Kirby-Bauer disk diffusion method was followed utilizing 15-mg azithromycin discs (Oxoid, Thermo Scientific, Waltham, MA) (34). To determine azithromycin MIC, E-strips were used (bioMérieux, France). Since there is no defined azithromycin CLSI breakpoint for any Salmonella serovars other than Salmonella Typhi, the following definition of resistance (adopted from Salmonella Typhi) was used: zone # 12 mm and MIC $ 32 mg/ml (20). ASTs were performed in a blinded fashion without any prior knowledge of sequence data to minimize bias.
DNA extraction and whole-genome sequencing. Isolates were grown on MacConkey agar (Oxoid, UK) overnight and DNA was extracted from a suspension of the overnight culture using the QIAamp DNA minikit (Qiagen, Hilden, Germany). Sequencing was conducted in an Illumina HiSeq 4000 pairedend (150-bp) system at the Wellcome Sanger Institute. Raw reads of all the isolates sequenced in this study can be found under ENA study accession number PRJEB30334.
Maximum-likelihood trees (MLT) were built from the chromosomal SNP alignments using RAxML version 8.2.12 (with the Generalized Time-Reversible model and a Gamma distribution to model site-specific rate variation [GTRGAMMA] in RAxML) (41). Support for the MLT was calculated using 100 bootstrap pseudoanalyses of the alignment. The MLT was outgroup rooted by including the pseudoalleles from Salmonella Paratyphi A AKU_12601 (for Salmonella Typhi) or Salmonella Typhi CT18 (for Salmonella Paratyphi A) in the alignment. Tree visualization was performed using iTol v5.5 (42). The same tool was used to prune 10 Salmonella Typhi with low genotyping support (,0.9) from the MLT of Salmonella Typhi, eventually making it an MLT of 825 isolates.
Evolutionary analysis with BEAST. Bayesian phylogenetic analyses were conducted to investigate the emergence of AcrB-R717Q mutation in isolates belonging to genotype 4.3.1.1. The 30 AzmR isolates from Bangladesh and 8 closely related azithromycin-sensitive Salmonella Typhi genomes belonging to 4.3.1.1 were selected (see Table S4), and SNPs were identified using the ParSNP tool (1phipack package) (43) with Salmonella Typhi P-stx-12 strain as the reference genome (44). The SNPs were subsequently used for Bayesian analysis with BEAST v1.10.4 (45). GTR1G 4 substitution model, an uncorrelated lognormal relaxed-clock model, and the exponential coalescent tree prior were used. Three independent analyses were performed with 5 Â 10 8 steps, recording samples every 5 Â 10 4 steps. The results from the three independent analyses gave similar estimates for the emergence of the 4.3.1.1 AzmR cluster. To calibrate the molecular clock, we used the sampling year of all sequences. The selected model combinations, including the molecular clock model and tree priors, have previously been used by Park et al. (39) for Salmonella Typhi evolutionary analysis.
Validation of MAMA. (i) DNA preparation. To validate the assay, we selected an additional 38 AzmR (Salmonella Typhi, n = 32; Salmonella Paratyphi A, n = 06) and 62 randomly selected non-AzmR (Salmonella Typhi, n = 48; Salmonella Paratyphi A, n = 14) isolates isolated between 2016 and 2018 in Bangladesh. All the isolates (n = 113, including 13 from Hooda et al. [14]) selected for validation (see Table S1) were subjected to cell lysis by a simple boiling method. In brief, a single colony from MacConkey agar (Oxoid, Thermo Scientific) was suspended in a 1.5-ml microcentrifuge tube containing 200 ml of sterile DNase-free water. The tube was placed in a heating block at 100°C for 20 min and then centrifuged at 12,000 rpm for 5 min. The supernatant was used for downstream applications.
(ii) Duplex PCR. The total volume of the duplex PCR assay was 25 ml, with 5Â master mix (Hot FIREPol; Solis BioDyne, Tartu, Estonia). The final concentrations of the primers were 0.1 and 0.8 mM for the parC and acrB genes, respectively. Amplification was performed in a thermal cycler (ProFlex 3 Â 32; Thermo Scientific) (Fig. 5A). The thermocycling conditions are shown in Fig. 5A. Amplified PCR products were run on 2% agarose gel (Invitrogen, Carlsbad, CA) at 100 V for 60 min and visualized on a Bio-Rad Gel Doc XR1 (Bio-Rad, Richmond, CA).
Statistical analysis. R 4.0.0 base functions and ggplot2, dplyr, epiR, and map packages were used for the sensitivity and specificity tests and statistical analyses. A Wilcoxon rank sum test was used to compare the mean MICs of wild-type and mutant isolates.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only.