Genomic epidemiology of Staphylococcus aureus isolated from bloodstream infections in South America during 2019 supports regional surveillance

Abstract

Staphylococcus aureus remains one of the leading causes of infections worldwide and a common cause of bacteraemia. However, studies documenting the epidemiology of S. aureus in South America using genomics are scarce. We hereby report on the largest genomic epidemiology study to date of both methicillin-resistant S. aureus (MRSA) and methicillin-susceptible S. aureus (MSSA) in South America, conducted by the StaphNET-SA network. We characterised 404 genomes recovered from a prospective observational study of S. aureus bacteraemia in 58 hospitals from Argentina, Bolivia, Brazil, Paraguay and Uruguay between April and October 2019. We show that a minority of S. aureus isolates are phenotypically multi-drug resistant (5.2%), but more than a quarter are resistant to macrolide-lincosamide-streptogramin B (MLSb). MSSA were more genetically diverse than MRSA. Lower rates of associated antimicrobial resistance in community-associated(CA)-MRSA versus hospital-associated (HA)-MRSA were found in association with three S. aureus genotypes dominating the MRSA population: CC30-MRSA-IVc-t019-lukS/F-PV+, CC5-MRSA-IV-t002-lukS/F-PV- and CC8-MRSA-IVc-t008-lukS/F-PV+-COMER+. These are historically from a CA origin, carry on average fewer antimicrobial resistance determinants, and often lack key virulence genes. Surprisingly, CC398-MSSA-t1451-lukS/F-PV- related to the CC398 human-associated lineage is widely disseminated throughout the region, and is described here for the first time as the most prevalent MSSA lineage in South America. Moreover, CC398 strains carrying ermT (largely responsible for the MLSb resistance rates of MSSA strains: inducible iMLSb phenotype) and sh_fabI (related to triclosan resistance) were recovered from both CA and HA origin. The frequency of MRSA and MSSA lineages differed between countries but the most prevalent S. aureus genotypes are high-risk clones widely distributed in the South American region without a clear country-specific phylogeographical structure. Therefore, our findings underline the need for continuous genomic surveillance by regional networks such as StaphNET-SA. This article contains data hosted by Microreact.

Keywords: CC30; CC398; CC5; CC8; MRSA; MSSA; S. aureus; South America.

Fig. 1.

Distribution of clonal complexes (CCs) by country. (a) Frequency of CCs by country and MRSA or MSSA. STs comprising fewer than three genomes are grouped under ‘Others’. (b) Geographical localization of major CCs in the region. The pie charts on the maps depict the distribution of CCs at each sampling site; the pie size is scaled by the number of genomes collected at each site. Project views are available in microreact: https://microreact.org/project/staphnet-sa-1st-survey.

Fig. 2.

(a) Genotypes of the main clonal complexes in the study. ‘Other’ includes genotypes present in fewer than three genomes. (b) ML tree of 404 genomes inferred from 156 868 SNP sites identified on 2182 core genes (Panaroo) with RAxML: midpoint rooted; 500 bootstrap replicates. Tree nodes and blocks are coloured as described in the key. MDR phenotype was defined as previously described [21] for all antibiotics tested including: gentamicin, cefoxitin, erythromycin, clindamycin, fluoroquinolones, trimethoprim-sulfamethoxazole, and rifampin. ‘>3 antibiotic classes’: genome with AMR determinants for more than three antibiotic classes including those detailed above plus phenicols, fusidic acid, fosfomycin, mupirocin and bleomycin (not tested phenotypically). Project view is available in microreact: https://microreact.org/project/staphnet-sa-1st-survey.

Fig. 3.

(a) CC30 global phylogeny (the divergent ST34 clade is not represented). Branches are coloured by clade: EMSSA-ST30 (orange), EMRSA-16 (blue), MRSA-ST30 (red), phage type 80/81 (aqua). Coloured nodes represent StaphNET-SA genomes, detailing number and country of origin. AR: Argentina, BO: Bolivia, BR: Brazil, PY: Paraguay, UY: Uruguay. Project is available in microreact: https://microreact.org/project/cc30-global-context. (b) Detailed ST30 phylogeny including 61 genomes from our collection. Outgroup-rooted phylogenetic tree inferred from 61 666 SNP sites obtained after mapping the genomes to the complete genome of strain TCH60 and masking regions of recombination and MGEs. Leaf nodes are coloured by collection: StaphNET-SA (violet), global context (grey). Coloured blocks represent the presence of an intact genetic determinant (purple) or feature as described in the key. For both trees, bars represent the number of SNPs per variable site. Project is available in microreact: https://microreact.org/project/emrsa-st30-global-context.

Fig. 4.

(a) CC5 global phylogeny. Outgroup-rooted phylogenetic tree inferred from 66 837 SNP sites obtained after mapping the genomes to the complete genome of strain JH1 (ST105) and masking regions of recombination and MGEs. Branches are coloured by clade: CC5-Basal (blue), CC5-I (violet), CC5-II-A (light violet), CC5-II-B (salmon). AR: Argentina, BO: Bolivia, BR: Brazil, PY: Paraguay, UY: Uruguay. (b) CC5-Basal clades. Same tree from the phylogeny in (a) but CC5-I and CC5-II clades are collapsed as purple (263 genomes) and salmon (436 genomes) circles, respectively. Leaf nodes are coloured by collection: global context (grey), StaphNET-SA (violet). Coloured blocks represent the presence of an intact genetic determinant: virulence gene (violet). Country and SCCmec type colours are described in the key. For both trees, bars represent the number of SNPs per variable site. Project is available in microreact: https://microreact.org/project/cc5-global-context.

Fig. 5.

(a) CC8 global phylogeny (more divergent Iberian and ST239 clades are not represented). Outgroup-rooted phylogenetic tree inferred from 66 899 SNP sites obtained after mapping the genomes to the complete genome of strain USA300-FPR3757 (ST8) and masking regions of recombination and MGEs. Coloured nodes represent StaphNET-SA genomes, detailing number and country of origin. AR: Argentina, BO: Bolivia, BR: Brazil, PY: Paraguay, UY: Uruguay. (b) Detail of the USA300 subtree from (a). The USA300-NAE clade (760 genomes) is collapsed as a blue circle. Leaf nodes are coloured by collection: StaphNET-SA (violet), global context (grey). Coloured blocks represent the presence of an intact genetic determinant: virulence gene (violet), COMER (brown). Country and SCCmec type colours are described in the key. For both trees, the outgroup is omitted, and bars represent the number of SNPs per variable site. Project is available in microreact: https://microreact.org/project/cc8-global-context.

Fig. 6.

(a) CC398 global phylogeny. Outgroup-rooted phylogenetic tree inferred from 71 097 SNP sites obtained after mapping the genomes to the complete genome of strain S0385 (ST398) and masking regions of recombination and MGEs. Branches are coloured by the host association: animal (blue), human (red). Coloured nodes represent StaphNET-SA genomes, detailing number and country of origin. AR: Argentina, BO: Bolivia, BR: Brazil, PY: Paraguay, UY: Uruguay. (b) Human-associated CC398 subtree from (a). Leaf nodes are coloured by collection: StaphNET-SA (violet), global context (grey). Coloured blocks represent the presence of an intact genetic determinant: AMR gene (red), virulence gene (violet), MGE (aqua), triclosan resistance gene (orange). Country and continent colours are described in the key. For both trees, bars represent the number of SNPs per variable site. Project is available in microreact: https://microreact.org/project/cc398-global-context.

Fig. 7.

Distributions of core genome SNP differences between pairs of genomes belonging to the same clade from either the same (blue) or different (orange) countries. Pairwise SNP distance is represented as boxplots (median and interquartile range) and a histogram of frequency for each clade. The number of genomes included is detailed in the key for each of the main lineages analysed. Text in the bottom of each figure is an interpretation of the difference between each pair of distributions, obtained using the R package ‘effsize’, which applies the parametric and non-parametric effect size estimators Cohen’s d and Cliff’s δ to the results of a Wilcoxon test (P). Subtrees of each main clade are represented next to each plot. The bar next to the tree represents the number of SNPs. Nodes (genomes included in the analysis) are coloured by country as described in the key.