Abstract
With rising fluoroquinolone and ceftriaxone-resistant Salmonella Typhi, azithromycin, a macrolide, has become the last oral drug available against typhoid. Between 2009-2016, we isolated 1,082 Salmonella Typhi and Paratyphi A strains in Bangladesh, 13 (12 Typhi and 1 Paratyphi A) of which were azithromycin-resistant. When compared to 462 previously sequenced Typhi strains, the genomes of the 12 azithromycin-resistant Typhi strains (4.3.1 sub-clade, H58) harbored an exclusive non-synonymous single-point mutation R717Q in AcrB, an RND-efflux pump. Expression of AcrB-R717Q in E. coli and Typhi strains increased its minimum inhibitory concentration (MIC) for azithromycin by 11- and 3-fold respectively. The azithromycin-resistant Paratyphi A strain also contained a mutation at R717 (R717L), whose introduction in E. coli and Paratyphi A strains increased MIC by 7- and 3-fold respectively, confirming the role of R717 mutations in conferring azithromycin resistance. With increasing azithromycin use, strains with R717 mutations may spread leading to treatment failures, making antibiotic stewardship and vaccine introduction imperative.
Introduction
Typhoid and paratyphoid, collectively known as enteric fever, are among the most common bacterial causes of morbidity worldwide, with the greatest burden in low- and middle-income countries (GBD 2017 Typhoid and Paratyphoid Collaborators, 2019). Salmonella enterica subspecies enterica serovars Typhi (Salmonella Typhi) and Paratyphi (A, B and C), etiologies of enteric fever, cause an estimated 14 million illnesses and 136,000 deaths annually.
In the pre-antibiotic era, enteric fever mortality rates exceeded 20% in many areas, but ampicillin, chloramphenicol and co-trimoxazole were instrumental in reducing the rates to <1%. Resistance to all three antibiotics (referred to as multidrug resistance, MDR) emerged in late 1980’s (Mirza et al., 1996), predominantly due to the rise and subsequent continental migration of H58 haplotype (now referred to as 4.3.1), which contained the resistance genes either on IncH1 plasmids or integrated within the chromosome (Holt et al., 2011; Wong et al., 2015, 2016). Fluoroquinolones soon became the most-commonly prescribed antibiotic (White et al., 1996), but since the 2000’s there have been increasing reports of decreased fluoroquinolone susceptibility due to the acquisition of chromosomal mutations in the DNA gyrase and topoisomerase IV genes (Roumagnac et al., 2006; Chau et al., 2007, Dimitrov et al., 2007; Pham Thanh et al., 2016). In Bangladesh, >99% of all Typhi and Paratyphi strains exhibit decreased susceptibility to ciprofloxacin (Saha et al., 2018b). In 2011, WHO recommended ceftriaxone or azithromycin for treating Salmonella Typhi non-susceptible to fluroquinolones (Balasegaram et al., 2012).
There have been sporadic reports of ceftriaxone-resistant Salmonella Typhi strains, (Saha et al., 1999; Djeghout et al., 2018), but in 2016, an outbreak of extensively drug-resistant (XDR) Salmonella Typhi, resistant to chloramphenicol, ampicillin, cotrimoxazole, fluoroquinolones, and third-generation cephalosporins was recognized in Pakistan and to date >1000 cases have been confirmed (Andrews et al., 2018). Cephalosporin resistance of the XDR strains was caused by the acquisition of a broad-spectrum beta-lactamase (bla-CTX-M-15) on an IncY plasmid found in other enteric species. Typhoid patients in Pakistan are primarily being treated with the last available oral option, the macrolide azithromycin, resistance to which is rare (Klemm et al., 2018, Parry et al., 2015). This increasing use of azithromycin places selective pressure for the emergence and spread of azithromycin-resistant isolates, raising concerns of untreatable infections and increased mortality rates. Little is known about azithromycin resistance in typhoidal Salmonella; while there are some sporadic reports on azithromycin treatment failures (Molloy et al., 2010; Sjölund-Karlsson et al., 2011; Wong et al., 2015; Patel et al., 2017), there are no data on the molecular mechanism of resistance.
In Bangladesh, Salmonella Typhi and Paratyphi A are the most common causes of bloodstream infections in children >2 months of age and comprise of two-third of blood-culture positive isolates in microbiology laboratories (Saha et al., 2017). Leveraging our surveillance system in place for enteric fever, here we describe the emergence of azithromycin resistance among typhoidal Salmonella in Bangladesh and identify the molecular basis behind this resistance.
Result and Discussion
Emergence of azithromycin-resistant Salmonella Typhi and Paratyphi A
Between 2009 and 2016, through our enteric fever surveillance (Saha et al., 2017) in the inpatient departments of the two largest pediatric hospitals of Bangladesh, we isolated 939 Salmonella Typhi and 143 Paratyphi A strains. Twelve of the Typhi and one of the Paratyphi A strains were resistant to azithromycin, with disc diameters of ≤12 mm, and minimum inhibitory concentration (MIC) of ≥32 µg/ml (Parry et al., 2015). All 12 azithromycin-resistant Salmonella Typhi strains were also MDR and were increasingly isolated since 2013 (Fig 1A), marking gradual emergence of azithromycin-resistant Salmonella Typhi in Bangladesh. All patients lived in Dhaka city, known to be endemic for typhoid (Figure S1).
Azithromycin resistant Salmonella Typhi harbors a mutation in the AcrB efflux pump
We sequenced the 12 azithromycin-resistant Typhi strains and found that all azithromycin-resistant strains belonged to genotype 4.3.1 (H58), the most common genotype found in South Asia (Tanmoy et al., 2018; Wong et al., 2015). In a whole-genome single nucleotide polymorphism (SNP) tree, the 12 strains clustered together indicating that they are genetically similar to one another and potentially arose for a single common parental strain (Figure 1B). To identify the genetic basis of azithromycin resistance, we used three bioinformatic tools: SRST2 (Inouye et al., 2014), Resfinder (Zankari et al., 2012) and CARD (Jia et al., 2017) and to evaluate the results obtained from these tools, we tested antimicrobial susceptibility against a panel of nine other antibiotics (Figure 1C). While the tools successfully predicted the observed susceptibility patterns for the nine antibiotics, no known azithromycin resistance mechanism was identified (Figure 1C). Using PlasmidFinder (Carattoli et al., 2014) we identified two plasmids found in these Salmonella Typhi strains: (i) IncQ1 (12/12 strains), containing genes for ampicillin, co-trimoxazole and chloramphenicol resistance, and (ii) InFIB (12/12 strains), a plasmid commonly found in Salmonella Typhi strains (Kidgell et al., 2002; Park et al., 2018). Both these plasmids were also present in closely related azithromycin-sensitive strains. The lack of known azithromycin-resistance genes indicated a novel mechanism of azithromycin resistance in these strains.
We compared the sequences of these 12 azithromycin-resistant strains to that of 462 Typhi strains that we had previously sequenced and genetically characterized (Tanmoy et al., 2018). In the WGS SNP tree, we identified four unique SNPs, present only in the 12 azithromycin-resistant strains, three of which were non-synonymous: STY2741 (codes for purN, a glycinamidine ribonucleotide transformyltransferase), STY1399 (codes for a hypothetical protein) and STY0519 (codes for AcrB, an inner membrane permease) (Figure 2A, Figure S2). For the first two candidates, there is no evidence of their involvement in mediating antimicrobial resistance in the literature. However, the third gene, acrB is part of a trans-envelope resistance-nodulation-division (RND) efflux pump that has been previously reported to transport macrolides including azithromycin across the bacterial cell envelope, making it the most promising candidate (Nakashima et al., 2011). Mutations affecting expression of AcrB have been implicated in macrolide resistance in Neisseria gonorrhoeae (Wadsworth et al., 2018). Furthermore, laboratory mutagenesis studies in Escherichia coli have shown that mutations in acrB can lead to higher macrolide efflux thereby contributing to resistance (Ababou and Koronakis, 2016). The SNP observed in the 12 azithromycin-resistant Salmonella Typhi strains changed the arginine residue (R) at position 717 to a glutamine (Q) (Figure 2B). R717 is a conserved residue on the periplasmic cleft that acts as the entry portal for most drugs in AcrB (Figure 2C). In a previous mutagenesis study, substitution of the arginine residue with an alanine (R717A) was found to partially increase efflux of the macrolide erythromycin in E. coli (Yu et al., 2005).
R717 mutations in AcrB confer azithromycin resistance
We cloned acrB from azithromycin susceptible and resistant Salmonella Typhi strains into an E. coli plasmid and introduced them into an E. coli strain that lacks the endogenous acrB (E. coli ΔacrB). Compared to E. coli strains containing empty plasmid or wild type acrB, the strain expressing AcrB-R717Q showed a smaller zone of disc clearance for both azithromycin (26.3 mm vs 16.7 mm, p = 0.0013) and erythromycin discs (22.3 mm vs 11.7 mm, p value = 0.0009) and exhibited a 11-fold increase in azithromycin MIC (0.22 µg/ml vs 2.7 µg/ml, p = 0.0002) (Figure 3A, B). Certain AcrB mutations have previously been shown to effect transport of other antibiotics such as ciprofloxacin (Blair et al., 2015), but the R717Q mutation did not change the susceptibility patterns for any other nine antibiotics we tested (Figure S3). For further confirmation of the effects of this mutation in Salmonella Typhi, we introduced the plasmids in an azithromycin-sensitive Typhi strain and observed a 3-fold increase in MIC (5 µg/ml vs 13 µg/ml, p < 0.0001) in the presence of AcrB-R717Q (Figure 3C). The difference here is lower compared to that seen in E. coli ΔacrB plausibly because the Typhi strain contains endogenous wild-type AcrB competing against the exogenous AcrB-R717Q that we artificially introduced. Taken together, these results confirm that AcrB-R717Q leads to increased macrolide resistance.
We conducted an extensive BLAST search to identify other typhoidal Salmonella strains with mutations in AcrB the NCBI database and found only one Salmonella Typhi strain isolated in Oceania in 2008 that contained the same R717Q mutation, however no AMR data were available for this strain (Wong et al., 2015). Interestingly, whole genome sequencing of the one azithromycin-resistant Salmonella Paratyphi A strain identified during our surveillance showed that this strain also contained a mutation in acrB which changed R717 to a leucine (L) (Figure 4A). This mutation was absent in the genomes of the Paratyphi A strains in the NCBI database. To determine to effect of R717L mutation, we expressed Paratyphi wild-type AcrB and AcrB-R717L in E. coli ΔacrB (Figure 4B, C). As seen for Typhi AcrB R717Q, Paratyphi AcrB-R717L leads to a smaller disc clearance for azithromycin (26.3 mm vs 16.3 mm, p = 0.0001) and erythromycin (22.4 mm vs 11.4 mm, p = 0.0001) and 10-fold higher azithromycin MIC (0.22 µg/ml vs 2.5 µg/ml, p = 0.003). When these plasmids were introduced in a sensitive Paratyphi A strain, we observed 4-fold change in MIC (7 µg/ml vs 28 µg/ml, p = 0.0001) in the presence of the R7171L mutation, confirming that mutations in R717 lead to macrolide resistance.
Rising antimicrobial resistance threatens the progress made so far in management of enteric fever. The rate of azithromycin resistance of typhoidal Salmonella in Bangladesh is low and the genetic basis is a chromosomal SNP. However, in light of the outbreak in Pakistan, increasing azithromycin use can place selective pressure on strains such as the ones with R717 mutations to spread. Although no azithromycin resistant XDR isolate has been reported to date, the increasing use of azithromycin and the clear historical record of widespread dissemination of resistance to all previously widely used antimicrobials by Salmonella Typhi and Paratyphi, suggest we will soon face strains resistant to all oral antibiotics. An azithromycin-resistant XDR strain would shift enteric fever treatment from outpatient departments, where patients are currently treated with oral azithromycin, to inpatient departments to be treated with injectable antibiotics like carbapenems, thereby weighing down already struggling health systems in endemic regions (Andrews et al., 2018; Saha et al., 2018a). Moreover, with the dearth of novel antimicrobials in the horizon, we risk losing our primary defense against widespread mortality from enteric fever. In addition to continued surveillance and antimicrobial stewardship, it is imperative to roll-out the recent WHO prequalified typhoid conjugate vaccine in endemic areas and decrease the overall burden of typhoid.
Materials and Methods
Study site and population
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, DSH, and Shishu Shasthya (Child Health) Foundation Hospital, SSFH. These are sentinel sites of the World Health Organization supported Invasive Bacterial Vaccine Preventable Diseases surveillance platform in Bangladesh.
Patient enrollment, etiology detection and antibiogram
Blood culture was performed at the discretion of the treating physicians. We enrolled patients with positive blood cultures for Salmonella Typhi or Paratyphi A. Blood cultures were performed using standard methods (Saha et al., 2017). We aseptically obtained 2–3 milliliters of blood, which was inoculated into trypticase soy broth supplemented with sodium polyanethole sulphonate (0.25%) and isovitalex (1%). Incubated blood culture bottles were sub-cultured on the second, third, and fifth days of incubation. Identification of Salmonella Typhi/Paratyphi isolates was confirmed using standard biochemical tests and agglutination with Salmonella species and serovar-specific antisera (Ramel, Thermo Fisher Scientific). Laboratory methods for blood culture and organism identification were consistent over the reporting period.
We used disc diffusion methods for determining antibiotic susceptibility patterns for azithromycin, ampicillin, co-trimoxazole, chloramphenicol, ciprofloxacin, levofloxacin, ceftriaxone, cefepime, cefixime and ceftazidime (Oxoid, Thermo Scientific, MA, USA). Azithromycin e-strips (bioMérieux, France) were used to determine the minimum inhibitory concentration (MIC) and confirm azithromycin resistance for strains that exhibited zone of clearance ≤12 mm in the presence of azithromycin discs. All results were interpreted according to the latest Clinical and Laboratory Standards Institute guidelines 2018.
DNA extraction and whole genome sequencing
We conducted whole genome sequencing on all identified azithromycin-resistant strains (12 Salmonella Typhi and 1 Paratyphi A). Isolates were grown in MacConkey agar (Oxoid, UK) overnight and DNA was extracted from a suspension of the overnight culture using the QIAamp DNA minikit (Qiagen, Hilden, Germany). Whole genome sequencing was performed on the Illumina HiSeq 4000 platform to generate 150 bp paired-end reads (Novogene Co. Ltd., Beijing, China). We used SPAdes (Bankevich et al., 2012) to assemble the short paired-end reads into contigs for downstream analyses. All the sequences have been submitted to EnteroBase and NCBI (BioProject ID: PRJNA528114).
Bioinformatics analysis
For comparative genomic analysis, we compared the 12 azithromycin-resistant Salmonella Typhi strains with 462 strains that were previously isolated and genetically characterized by our group in Bangladesh (Tanmoy et al., 2018). Using the ParSNP tool (Treangen et al., 2014), we constructed whole-genome SNP tree using all 475 genomes and determined the genotypes using srst2 package (Inouye et al., 2014). Ggtree was used to make the phylogenetic tree and overlay the genotype data (Yu et al., 2017). Srst2, ResFinder (Zankari et al., 2012) and CARD (Jia et al., 2017) were used to predict antimicrobial resistance markers, and PlasmidFinder (Carattoli et al., 2014) to identify the putative plasmids present in these strains. Finally, we compared the resistant strains to all sensitive Salmonella Typhi strains manually to find SNPs exclusive to the resistant strains (comparison to the most closely related 5 genomes are shown in Figure 1C) using the GinGR tool from the Harvest suite (Treangen et al., 2014). To predict the function of the SNPs on protein function, we examined the protein sequence and conducted structural analyses (Figure 2).
Macrolide susceptibility test in E. coli and Salmonella Typhi
We amplified the acrB genes from azithromycin-resistant Salmonella Typhi strain 5003 and Paratyphi A strain 3144 and azithromycin-sensitive Typhi strain 4119 and Paratyphi A strain 4071 for downstream cloning. The genes were inserted into the multiple cloning site of pHERD30T using Gibson assembly (Gibson et al., 2009). We verified the sequences of all inserted genes through Sanger sequencing. The plasmids were chemically transformed into CaCl2-competent E. coli BW25112 ΔacrB strains, and electroporated into Typhi strain 4119 and Paratyphi A strain 4071. E. coli strains with plasmids (with or without an insert) were tested for susceptibility patterns for erythromycin, azithromycin, ampicillin, co-trimoxazole, chloramphenicol, ciprofloxacin, levofloxacin, ceftriaxone, cefepime, cefixime and ceftazidime using the disc diffusion method, and MIC was determined using azithromycin E-strips. Typhi and Paratyphi strains with plasmids (with or without an insert) were tested for susceptibility patterns for azithromycin using E-strips.
Funding
No external funding was attained for this study. The enteric fever surveillance in Dhaka Shishu (Children) Hospital was supported by Gavi, the Vaccine Alliance, through the World health Organization-supported Invasive Bacterial Vaccine Preventable Diseases study (grant numbers 201588766, 201233523, 201022732, 200749550, 201686542)
Conflict of interest
Authors declare no conflict of interest pertaining to the work presented here.
Acknowledgements
We are thankful to Dr. Shweta Karambelkar, Dr. Balint Csorgo, and Beatriz A. Osuna of University of California, San Francisco, and Sunita Patil of Stanford University for technical assistance with laboratory work and to Arif M. Tanmoy of the Child Health Research Foundation for bioinformation guidance.