Global diversity and antimicrobial resistance of typhoid fever pathogens: insights from 13,000 Salmonella Typhi genomes

Curation statements for this article:
  • Curated by eLife

    eLife logo

    eLife assessment

    Although largely descriptive, this meta-analysis of 13,000 published Typhi genomes is hugely important to public health. The dataset and presented analysis represents the first wholesale analysis of all available Typhi genomes from the last 21 years. The findings are of great significance to tracking the emergence and maintenance of AMR in Typhi and include novel insights into XDR strain emergence in Pakistan as well as the relationship between MDR maintenance and chromosomal integration.

This article has been Reviewed by the following groups

Read the full article See related articles

Abstract

The Global Typhoid Genomics Consortium was established to bring together the typhoid research community to aggregate and analyse Salmonella enterica serovar Typhi (Typhi) genomic data to inform public health action. This analysis, which marks twenty-one years since the publication of the first Typhi genome, represents the largest Typhi genome sequence collection to date (n=13,000), and provides a detailed overview of global genotype and antimicrobial resistance (AMR) distribution and temporal trends, generated using open analysis platforms (GenoTyphi and Pathogenwatch). Compared with previous global snapshots, the data highlight that genotype 4.3.1 (H58) has not spread beyond Asia and Eastern/Southern Africa; in other regions, distinct genotypes dominate and have independently evolved AMR. Data gaps remain in many parts of the world, and we show potential of travel-associated data to provide informal “sentinel” surveillance for such locations. The data indicate ciprofloxacin non-susceptibility (>1 resistance determinant) is widespread across geographies and genotypes, with high-level resistance (≥3 determinants) reaching 20% prevalence in South Asia. Extensively drug-resistant (XDR) typhoid has become dominant in Pakistan (70% in 2020), but has not yet become established elsewhere. Ceftriaxone resistance has emerged in eight non-XDR genotypes, including a ciprofloxacin-resistant lineage (4.3.1.2.1) in India. Azithromycin resistance mutations were detected at low prevalence in South Asia, including in two common ciprofloxacin-resistant genotypes. The Consortium’s aim is to encourage continued data sharing and collaboration to monitor the emergence and global spread of AMR Typhi, and to inform decision-making around the introduction of typhoid conjugate vaccines (TCVs) and other prevention and control strategies.

Article activity feed

  1. eLife assessment

    Although largely descriptive, this meta-analysis of 13,000 published Typhi genomes is hugely important to public health. The dataset and presented analysis represents the first wholesale analysis of all available Typhi genomes from the last 21 years. The findings are of great significance to tracking the emergence and maintenance of AMR in Typhi and include novel insights into XDR strain emergence in Pakistan as well as the relationship between MDR maintenance and chromosomal integration.

  2. Reviewer #1 (Public Review):

    This is the most complete genomic overview of the epidemiology of Salmonella enterica serovar Typhi including close to 13,000 genmoes from multiple countries, clearly demonstrating the geographical differences in molecular epidemiology and antibiotic resistance traits. This database could serve as the global reference for the future with constant addition of new information.

    This is a descriptive study, not providing fundamentally new mechanistic insights of the disease, but providing an overview of the global epidemiology of this bacterium.
    Open-ended questions remain the generalizability of the findings, which is linked to the completeness of the surveillance systems, as well as the linkage of genotypes to clinical disease presentation (severity) and of linkage of local antibiotic use and the prevalence of the different resistance traits.

    Publication of these data will be very helpful for all those interested in the molecular epidemiology of Salmonella and may stimulate not-yet participating institutes to add information for future analyses. It may also stimulate investigators to use the data for deriving more insights in clinical disease presentations, associations with antibiotic use and input for mathematical modelling.

    The (lenghty) introduction is textbook epidemiology of the emergence of antimicrobial resistance in Typhi.

  3. Reviewer #2 (Public Review):

    The authors present a thorough and comprehensive analysis of 13000 Typhi genomes sampled globally over the last 21 years. The paper is an important example of how to perform meta-analysis of large numbers of published genomes while keeping credit equitable and including all original investigators as authors. This should be commended and maintained by the genomics community as the correct protocol when performing meta-analyses of this kind.

    The study presents important findings on the emergence, maintenance and dynamics of AMR in different Typhi lineage backgrounds globally. This is extremely important for surveillance and appropriate adjustments to empirical therapy guidelines.

    The study was also able to deduce new findings on the emergence of XDR Typhi in Pakistan and to date the first case to much earlier than previously thought. This is a good demonstration of why collating and re-analysing data in this fashion can be so valuable.

    The authors present interesting evidence that settings where MDR is chromosomally integrated has remained at high prevalence whereas it seems to be declining in settings where MDR is plasmid-borne. I found Figure S11 particularly interesting. As noted by the authors, this is consistent with the hypothesis that the IncHI1 MDR plasmid is associated with a fitness cost that is removed when the MDR transposon becomes chromosomally-integrated.

    This study also represents a good demonstration of why patient travel information can be such a useful metadata field for genomic studies and the potential for its use in helping to survey areas where no genomic studies have taken place yet. Other studies (e.g. https://www.medrxiv.org/content/10.1101/2022.08.23.22279111v1) have used this information from UKHSA to similarly represent the phylogeography of a different serovar of Salmonella and have found that data collected in this way can provide broader global coverage and more uniform sampling than what is currently available on NCBI. This data should be encouraged to be shared and this study goes a long way in proving its general utility for surveillance studies in public health.

  4. Reviewer #3 (Public Review):

    The authors present a study of 13000, Salmonella Typhi genomes from across the globe. Here, they present an overview of the global genomic epidemiology of Salmonella Typhi, in the context of the evolution of antimicrobial resistance. The authors present the temporal trends in the prevalence of Salmonella Typhi genotypes in select regions/ countries as well as the prevalence and antimicrobial resistance. The authors cite travel isolates of Salmonella Typhi as a useful proxy for surveillance in high burden settings where there exists a paucity of genomic data. While the authors acknowledge the limitations of their study, there remain major concerns over sampling bias and representativeness that question the generalizability of their findings.

    Based on the methods section, the authors did not make mention of adjusting their prevalence estimates for outbreak investigations. When conducting a population analysis, including outbreak samples can lead to an overestimation of the prevalence of the outbreak strain. First, outbreaks tend to be sampled more densely than isolates from routine surveillance of endemic disease, secondly in an outbreak, you are essentially sampling the same strain multiple times. This needs to be taken into consideration when estimating the prevalence of genotypes in the population. Treating outbreak investigations and routine surveillance equally in calculating prevalence can be misleading if the proportion of outbreak isolates sequenced is greater than the proportion of isolates in the surveillance area that are sequenced.

    There are concerns regarding the validity of the results presented in Figures 1-3. These results require a nuanced assessment of the factors that are likely to influence genotypic diversity including type of study, duration of sampling and total number of genomes sequenced. In big Countries like Nigeria and India, where can be heterogeneity in different regions of the country and this needs to also be considered in inferring the prevalence of genotypes.

    This heterogeneity in prevalence of genotypes was observed in countries with multiple laboratories. In India for example, the prevalence of lineage 4.3.1.2 ranged from 39% to 82%, in different cities/ regions. The authors have not provided sufficient context on the underlying source of this variation in prevalence. In order to understand the reason for observing these differences there needs to be a discussion around when the samples in each place/ region were conducted, how long the study was conducted, how many isolates were collected and whether this was a routine surveillance, outbreak investigation or other type of study. Similar variability is observed in Nigeria where most isolates were from Abuja (Zankli Medical Center, n=105, 2010-2013) and other sources included Ibadan (University of Ibadan, n=14, 2017-2018), and reference laboratories in England (n=15, 2015-2019) and the USA (n=10, 2016-2019). Given the small sample sizes and the fact that the time periods for sample collection varied, using this dataset to get a snapshot of the prevalence of genotypes in Nigeria can be potentially misleading.

    Moreover, the authors cite that 70% of cases in Pakistan are caused by XDR. Is this based on the proportions of isolates that are XDR in this dataset? Klemm et al 2018 sequenced primarily XDR isolates, therefore that dataset is not representative of the wider population. Rasheed et al included on 27 genomes, which were isolated at hospitals. Hospitals isolates may give an overestimated XDR burden because susceptible isolates are likely to get treated successfully with antibiotics alleviating the need for hospitalization. Similarly, Yousafzai et al 2019 was an investigation into an outbreak of ceftriaxone-resistant Salmonella Typhi in Hyderabad, which is a densely sample dataset and not necessarily a representation of the wider population. Aggregating these data may lead to an accumulation of bias that gives a distorted snap shot of the diversity on genotypes. Also, it is unclear whether the number of isolates collected from each of these studies was consistent with time. Thus, changes in the prevalence may be representative of a change in the proportion of genomes that were sampled from individual studies.

    One of the major recommendations from this study was that travel associated isolates can be a proxy for surveillance in high burden regions where there is paucity of data. The authors have not demonstrated a rigorous test for representativeness of the travel associated samples. The test conducted by the authors looked at how well the travel isolates correlated with the isolates from other studies conducted in the source population. However, they have not factored in potential biases associated with the studies conducted in the host countries. Also, travel is more likely to encompass a specific socio-economic demography of people who can afford to travel. This leads to underrepresentation of low income individuals and communities, especially in low-income countries. Moreover, the authors have not shown that the phylogenetic placement of the travel isolates supports the claim that they originated from that country. Conclusions drawn from travel associated isolates need to be tempered, while it can be a useful tool for early detection of potentially virulent lineages or lineages that have novel resistance mechanisms, using it to determine prevalence can be misleading.

    Other minor observations include:

    Introduction needs to trim significantly to be more concise. The authors can demonstrate that Salmonella typhi accumulates resistance genotypes over time and as new antibiotics are introduced resistance mutations become selected for and fixed in the population.

    Figure 4 is very similar to Figure 1 of Klemm et al 2018, does not add any new insights.