Machine learning to attribute the source of Campylobacter infections in the United States: A retrospective analysis of national surveillance data.

OBJECTIVES: Integrating pathogen genomic surveillance with bioinformatics can enhance public health responses by identifying risk and guiding interventions. This study focusses on the two predominant Campylobacter species, which are commonly found in the gut of birds and mammals and often infect humans via contaminated food. Rising incidence and antimicrobial resistance (AMR) are a global concern, and there is an urgent need to quantify the main routes to human infection. METHODS: During routine US national surveillance (2009-2019), 8856 Campylobacter genomes from human infections and 16,703 from possible sources were sequenced. Using machine learning and probabilistic models, we target genetic variation associated with host adaptation to attribute the source of human infections and estimate the importance of different disease reservoirs. RESULTS: Poultry was identified as the primary source of human infections, responsible for an estimated 68% of cases, followed by cattle (28%), and only a small contribution from wild birds (3%) and pork sources (1%). There was also evidence of an increase in multidrug resistance, particularly among isolates attributed to chickens. CONCLUSIONS: National surveillance and source attribution can guide policy, and our study suggests that interventions targeting poultry will yield the greatest reductions in campylobacteriosis and spread of AMR in the US. DATA AVAILABILITY: All sequence reads were uploaded and shared on NCBI's Sequence Read Archive (SRA) associated with BioProjects; PRJNA239251 (CDC / PulseNet surveillance), PRJNA287430 (FSIS surveillance), PRJNA292668 & PRJNA292664 (NARMS) and PRJNA258022 (FDA surveillance). Publicly available genomes, including reference genomes and isolates sampled worldwide from wild birds are associated with BioProject accessions: PRJNA176480, PRJNA177352, PRJNA342755, PRJNA345429, PRJNA312235, PRJNA415188, PRJNA524300, PRJNA528879, PRJNA529798, PRJNA575343, PRJNA524315 and PRJNA689604. Contiguous assemblies of all genome sequences compared are available at Mendeley data (assembled C. coli genomes doi: 10.17632/gxswjvxyh3.1; assembled C. jejuni genomes doi: 10.17632/6ngsz3dtbd.1) and individual project and accession numbers can be found in Supplementary tables S1 and S2, which also includes pubMLST identifiers for assembled genomes. Figshare (10.6084/m9.figshare.20279928). Interactive phylogenies are hosted on microreact separately for C. jejuni (https://microreact.org/project/pascoe-us-cjejuni) and C. coli (https://microreact.org/project/pascoe-us-ccoli).

Rapid geographical source attribution of Salmonella enterica serovar Enteritidis genomes using hierarchical machine learning.

Salmonella enterica serovar Enteritidis is one of the most frequent causes of Salmonellosis globally and is commonly transmitted from animals to humans by the consumption of contaminated foodstuffs. In the UK and many other countries in the Global North, a significant proportion of cases are caused by the consumption of imported food products or contracted during foreign travel, therefore, making the rapid identification of the geographical source of new infections a requirement for robust public health outbreak investigations. Herein, we detail the development and application of a hierarchical machine learning model to rapidly identify and trace the geographical source of S. Enteritidis infections from whole genome sequencing data. 2313 S. Enteritidis genomes, collected by the UKHSA between 2014-2019, were used to train a 'local classifier per node' hierarchical classifier to attribute isolates to four continents, 11 sub-regions, and 38 countries (53 classes). The highest classification accuracy was achieved at the continental level followed by the sub-regional and country levels (macro F1: 0.954, 0.718, 0.661, respectively). A number of countries commonly visited by UK travelers were predicted with high accuracy (hF1: >0.9). Longitudinal analysis and validation with publicly accessible international samples indicated that predictions were robust to prospective external datasets. The hierarchical machine learning framework provided granular geographical source prediction directly from sequencing reads in <4 min per sample, facilitating rapid outbreak resolution and real-time genomic epidemiology. The results suggest additional application to a broader range of pathogens and other geographically structured problems, such as antimicrobial resistance prediction, is warranted.

Host ecology regulates interspecies recombination in bacteria of the genus Campylobacter.

Horizontal gene transfer (HGT) can allow traits that have evolved in one bacterial species to transfer to another. This has potential to rapidly promote new adaptive trajectories such as zoonotic transfer or antimicrobial resistance. However, for this to occur requires gaps to align in barriers to recombination within a given time frame. Chief among these barriers is the physical separation of species with distinct ecologies in separate niches. Within the genus Campylobacter, there are species with divergent ecologies, from rarely isolated single-host specialists to multihost generalist species that are among the most common global causes of human bacterial gastroenteritis. Here, by characterizing these contrasting ecologies, we can quantify HGT among sympatric and allopatric species in natural populations. Analyzing recipient and donor population ancestry among genomes from 30 Campylobacter species, we show that cohabitation in the same host can lead to a six-fold increase in HGT between species. This accounts for up to 30% of all SNPs within a given species and identifies highly recombinogenic genes with functions including host adaptation and antimicrobial resistance. As described in some animal and plant species, ecological factors are a major evolutionary force for speciation in bacteria and changes to the host landscape can promote partial convergence of distinct species through HGT.

Quantifying bacterial evolution in the wild: A birthday problem for Campylobacter lineages.

Measuring molecular evolution in bacteria typically requires estimation of the rate at which nucleotide changes accumulate in strains sampled at different times that share a common ancestor. This approach has been useful for dating ecological and evolutionary events that coincide with the emergence of important lineages, such as outbreak strains and obligate human pathogens. However, in multi-host (niche) transmission scenarios, where the pathogen is essentially an opportunistic environmental organism, sampling is often sporadic and rarely reflects the overall population, particularly when concentrated on clinical isolates. This means that approaches that assume recent common ancestry are not applicable. Here we present a new approach to estimate the molecular clock rate in Campylobacter that draws on the popular probability conundrum known as the 'birthday problem'. Using large genomic datasets and comparative genomic approaches, we use isolate pairs that share recent common ancestry to estimate the rate of nucleotide change for the population. Identifying synonymous and non-synonymous nucleotide changes, both within and outside of recombined regions of the genome, we quantify clock-like diversification to estimate synonymous rates of nucleotide change for the common pathogenic bacteria Campylobacter coli (2.4 x 10-6 s/s/y) and Campylobacter jejuni (3.4 x 10-6 s/s/y). Finally, using estimated total rates of nucleotide change, we infer the number of effective lineages within the sample time frame-analogous to a shared birthday-and assess the rate of turnover of lineages in our sample set over short evolutionary timescales. This provides a generalizable approach to calibrating rates in populations of environmental bacteria and shows that multiple lineages are maintained, implying that large-scale clonal sweeps may take hundreds of years or more in these species.

Campylobacter jejuni genotypes are associated with post-infection irritable bowel syndrome in humans.

Campylobacter enterocolitis may lead to post-infection irritable bowel syndrome (PI-IBS) and while some C. jejuni strains are more likely than others to cause human disease, genomic and virulence characteristics promoting PI-IBS development remain uncharacterized. We combined pangenome-wide association studies and phenotypic assays to compare C. jejuni isolates from patients who developed PI-IBS with those who did not. We show that variation in bacterial stress response (Cj0145_phoX), adhesion protein (Cj0628_CapA), and core biosynthetic pathway genes (biotin: Cj0308_bioD; purine: Cj0514_purQ; isoprenoid: Cj0894c_ispH) were associated with PI-IBS development. In vitro assays demonstrated greater adhesion, invasion, IL-8 and TNFα secretion on colonocytes with PI-IBS compared to PI-no-IBS strains. A risk-score for PI-IBS development was generated using 22 genomic markers, four of which were from Cj1631c, a putative heme oxidase gene linked to virulence. Our finding that specific Campylobacter genotypes confer greater in vitro virulence and increased risk of PI-IBS has potential to improve understanding of the complex host-pathogen interactions underlying this condition.

Genome evolution and the emergence of pathogenicity in avian Escherichia coli.

Chickens are the most common birds on Earth and colibacillosis is among the most common diseases affecting them. This major threat to animal welfare and safe sustainable food production is difficult to combat because the etiological agent, avian pathogenic Escherichia coli (APEC), emerges from ubiquitous commensal gut bacteria, with no single virulence gene present in all disease-causing isolates. Here, we address the underlying evolutionary mechanisms of extraintestinal spread and systemic infection in poultry. Combining population scale comparative genomics and pangenome-wide association studies, we compare E. coli from commensal carriage and systemic infections. We identify phylogroup-specific and species-wide genetic elements that are enriched in APEC, including pathogenicity-associated variation in 143 genes that have diverse functions, including genes involved in metabolism, lipopolysaccharide synthesis, heat shock response, antimicrobial resistance and toxicity. We find that horizontal gene transfer spreads pathogenicity elements, allowing divergent clones to cause infection. Finally, a Random Forest model prediction of disease status (carriage vs. disease) identifies pathogenic strains in the emergent ST-117 poultry-associated lineage with 73% accuracy, demonstrating the potential for early identification of emergent APEC in healthy flocks.

Genomic epidemiology of Campylobacter jejuni associated with asymptomatic pediatric infection in the Peruvian Amazon.

Campylobacter is the leading bacterial cause of gastroenteritis worldwide and its incidence is especially high in low- and middle-income countries (LMIC). Disease epidemiology in LMICs is different compared to high income countries like the USA or in Europe. Children in LMICs commonly have repeated and chronic infections even in the absence of symptoms, which can lead to deficits in early childhood development. In this study, we sequenced and characterized C. jejuni (n = 62) from a longitudinal cohort study of children under the age of 5 with and without diarrheal symptoms, and contextualized them within a global C. jejuni genome collection. Epidemiological differences in disease presentation were reflected in the genomes, specifically by the absence of some of the most common global disease-causing lineages. As in many other countries, poultry-associated strains were likely a major source of human infection but almost half of local disease cases (15 of 31) were attributable to genotypes that are rare outside of Peru. Asymptomatic infection was not limited to a single (or few) human adapted lineages but resulted from phylogenetically divergent strains suggesting an important role for host factors in the cryptic epidemiology of campylobacteriosis in LMICs.

A Hopeful Sea-Monster: A Very Large Homologous Recombination Event Impacting the Core Genome of the Marine Pathogen Vibrio anguillarum.

Vibrio anguillarum is the causative agent of vibriosis in many species important to aquaculture. We generated whole genome sequence (WGS) data on a diverse collection of 64 V. anguillarum strains, which we supplemented with 41 publicly available genomes to produce a combined dataset of 105 strains. These WGS data resolved six major lineages (L1-L6), and the additional use of multilocus sequence analysis (MLSA) clarified the association of L1 with serotype O1 and Salmonidae hosts (salmon/trout), and L2 with serotypes O2a/O2b/O2c and Gadidae hosts (cod). Our analysis also revealed a large-scale homologous replacement of 526-kb of core genome in an L2 strain from a con-specific donor. Although the strains affected by this recombination event are exclusively associated with Gadidae, we find no clear genetic evidence that it has played a causal role in host specialism. Whilst it is established that Vibrio species freely recombine, to our knowledge this is the first report of a contiguous recombinational replacement of this magnitude in any Vibrio genome. We also note a smaller accessory region of high single nucleotide polymorphism (SNP) density and gene content variation that contains lipopolysaccharide biosynthesis genes which may play a role in determining serotype.

Agricultural intensification and the evolution of host specialism in the enteric pathogen Campylobacter jejuni.

Modern agriculture has dramatically changed the distribution of animal species on Earth. Changes to host ecology have a major impact on the microbiota, potentially increasing the risk of zoonotic pathogens being transmitted to humans, but the impact of intensive livestock production on host-associated bacteria has rarely been studied. Here, we use large isolate collections and comparative genomics techniques, linked to phenotype studies, to understand the timescale and genomic adaptations associated with the proliferation of the most common food-born bacterial pathogen (Campylobacter jejuni) in the most prolific agricultural mammal (cattle). Our findings reveal the emergence of cattle specialist C. jejuni lineages from a background of host generalist strains that coincided with the dramatic rise in cattle numbers in the 20th century. Cattle adaptation was associated with horizontal gene transfer and significant gene gain and loss. This may be related to differences in host diet, anatomy, and physiology, leading to the proliferation of globally disseminated cattle specialists of major public health importance. This work highlights how genomic plasticity can allow important zoonotic pathogens to exploit altered niches in the face of anthropogenic change and provides information for mitigating some of the risks posed by modern agricultural systems.

PIRATE: A fast and scalable pangenomics toolbox for clustering diverged orthologues in bacteria.

BACKGROUND: Cataloguing the distribution of genes within natural bacterial populations is essential for understanding evolutionary processes and the genetic basis of adaptation. Advances in whole genome sequencing technologies have led to a vast expansion in the amount of bacterial genomes deposited in public databases. There is a pressing need for software solutions which are able to cluster, catalogue and characterise genes, or other features, in increasingly large genomic datasets. RESULTS: Here we present a pangenomics toolbox, PIRATE (Pangenome Iterative Refinement and Threshold Evaluation), which identifies and classifies orthologous gene families in bacterial pangenomes over a wide range of sequence similarity thresholds. PIRATE builds upon recent scalable software developments to allow for the rapid interrogation of thousands of isolates. PIRATE clusters genes (or other annotated features) over a wide range of amino acid or nucleotide identity thresholds and uses the clustering information to rapidly identify paralogous gene families and putative fission/fusion events. Furthermore, PIRATE orders the pangenome using a directed graph, provides a measure of allelic variation, and estimates sequence divergence for each gene family. CONCLUSIONS: We demonstrate that PIRATE scales linearly with both number of samples and computation resources, allowing for analysis of large genomic datasets, and compares favorably to other popular tools. PIRATE provides a robust framework for analysing bacterial pangenomes, from largely clonal to panmictic species.

Gene Composition as a Potential Barrier to Large Recombinations in the Bacterial Pathogen Klebsiella pneumoniae.

Klebsiella pneumoniae (Kp) is one of the most important nosocomial pathogens worldwide, able to cause multiorgan infections and hospital outbreaks. One of the most widely disseminated lineage of Kp is the clonal group 258 (CG258), which includes the highly resistant "high-risk" sequence types ST258 and ST11. Genomic investigations revealed that very large recombination events have occurred during the emergence of Kp lineages. A striking example is provided by ST258, which has undergone a recombination event that replaced over 1 Mb of the genome with DNA from an unrelated Kp donor. Although several examples of this phenomenon have been documented in Kp and other bacterial species, the significance of these very large recombination events for the emergence of either hypervirulent or resistant clones remains unclear. Here, we present an analysis of 834 Kp genomes that provides data on the frequency of these very large recombination events (defined as those involving >100 kb), their distribution within the genome, and the dynamics of gene flow within the Kp population. We note that very large recombination events occur frequently, and in multiple lineages, and that the majority of recombinational exchanges are clustered within two overlapping genomic regions, which have been involved by recombination events with different frequencies. Our results also indicate that certain lineages are more likely to act as donors to CG258. Furthermore, comparison of gene content in CG258 and non-CG258 strains agrees with this pattern, suggesting that the success of a large recombination depends on gene composition in the exchanged genomic portion.

Gene pool transmission of multidrug resistance among Campylobacter from livestock, sewage and human disease.

The use of antimicrobials in human and veterinary medicine has coincided with a rise in antimicrobial resistance (AMR) in the food-borne pathogens Campylobacter jejuni and Campylobacter coli. Faecal contamination from the main reservoir hosts (livestock, especially poultry) is the principal route of human infection but little is known about the spread of AMR among source and sink populations. In particular, questions remain about how Campylobacter resistomes interact between species and hosts, and the potential role of sewage as a conduit for the spread of AMR. Here, we investigate the genomic variation associated with AMR in 168 C. jejuni and 92 C. coli strains isolated from humans, livestock and urban effluents in Spain. AMR was tested in vitro and isolate genomes were sequenced and screened for putative AMR genes and alleles. Genes associated with resistance to multiple drug classes were observed in both species and were commonly present in multidrug-resistant genomic islands (GIs), often located on plasmids or mobile elements. In many cases, these loci had alleles that were shared among C. jejuni and C. coli consistent with horizontal transfer. Our results suggest that specific antibiotic resistance genes have spread among Campylobacter isolated from humans, animals and the environment.

Patterns of cross-resistance and collateral sensitivity between clinical antibiotics and natural antimicrobials.

Bacteria interact with a multitude of other organisms, many of which produce antimicrobials. Selection for resistance to these antimicrobials has the potential to result in resistance to clinical antibiotics when active compounds target the same bacterial pathways. The possibility of such cross-resistance between natural antimicrobials and antibiotics has to our knowledge received very little attention. The antimicrobial activity of extracts from seaweeds, known to be prolific producers of antimicrobials, is here tested against Staphylococcus aureus isolates with varied clinical antibiotic resistance profiles. An overall effect consistent with cross-resistance is demonstrated, with multidrug-resistant S. aureus strains being on average more resistant to seaweed extracts. This pattern could potentially indicate that evolution of resistance to antimicrobials in the natural environment could lead to resistance against clinical antibiotics. However, patterns of antimicrobial activity of individual seaweed extracts vary considerably and include collateral sensitivity, where increased resistance to a particular antibiotic is associated with decreased resistance to a particular seaweed extract. Our correlation-based methods allow the identification of antimicrobial extracts bearing most promise for downstream active compound identification and pharmacological testing.

WGS of 1058 Enterococcus faecium from Copenhagen, Denmark, reveals rapid clonal expansion of vancomycin-resistant clone ST80 combined with widespread dissemination of a vanA-containing plasmid and acquisition of a heterogeneous accessory genome.

OBJECTIVES: From 2012 to 2015, a sudden significant increase in vancomycin-resistant (vanA) Enterococcus faecium (VREfm) was observed in the Capital Region of Denmark. Clonal relatedness of VREfm and vancomycin-susceptible E. faecium (VSEfm) was investigated, transmission events between hospitals were identified and the pan-genome and plasmids from the largest VREfm clonal group were characterized. METHODS: WGS of 1058 E. faecium isolates was carried out on the Illumina platform to perform SNP analysis and to identify the pan-genome. One isolate was also sequenced on the PacBio platform to close the genome. Epidemiological data were collected from laboratory information systems. RESULTS: Phylogeny of 892 VREfm and 166 VSEfm revealed a polyclonal structure, with a single clonal group (ST80) accounting for 40% of the VREfm isolates. VREfm and VSEfm co-occurred within many clonal groups; however, no VSEfm were related to the dominant VREfm group. A similar vanA plasmid was identified in ≥99% of isolates belonging to the dominant group and 69% of the remaining VREfm. Ten plasmids were identified in the completed genome, and ∼29% of this genome consisted of dispensable accessory genes. The size of the pan-genome among isolates in the dominant group was 5905 genes. CONCLUSIONS: Most probably, VREfm emerged owing to importation of a successful VREfm clone which rapidly transmitted to the majority of hospitals in the region whilst simultaneously disseminating a vanA plasmid to pre-existing VSEfm. Acquisition of a heterogeneous accessory genome may account for the success of this clone by facilitating adaptation to new environmental challenges.

Real time monitoring of Aeromonas salmonicida evolution in response to successive antibiotic therapies in a commercial fish farm.

Our ability to predict evolutionary trajectories of pathogens in response to antibiotic pressure is one of the promising leverage to fight against the present antibiotic resistance worldwide crisis. Yet, few studies tackled this question in situ at the outbreak level, due to the difficulty to link a given pathogenic clone evolution with its precise antibiotic exposure over time. In this study, we monitored the real-time evolution of an Aeromonas salmonicida clone in response to successive antibiotic and vaccine therapies in a commercial fish farm. The clone was responsible for a four-year outbreak of furunculosis within a Recirculating Aquaculture System Salmo salar farm in China, and we reconstructed the precise tempo of mobile genetic elements (MGEs) acquisition events during this period. The resistance profile provided by the acquired MGEs closely mirrored the antibiotics used to treat the outbreak, and we evidenced that two subclonal groups developed similar resistances although unrelated MGE acquisitions. Finally, we also demonstrated the efficiency of vaccination in outbreak management and its positive effect on antibiotic resistance prevalence. Our study provides unprecedented knowledge critical to understand evolutionary trajectories of resistant pathogens outside the laboratory.

Genomic epidemiology of the commercially important pathogen Renibacterium salmoninarum within the Chilean salmon industry.

Renibacterium salmoninarum is the causative agent of bacterial kidney disease (BKD), which is a commercially important disease of farmed salmonids. Typing by conventional methods provides limited information on the evolution and spread of this pathogen, as there is a low level of standing variation within the R. salmoninarum population. Here, we apply whole-genome sequencing to 42 R. salmoninarum isolates from Chile, primarily from salmon farms, in order to understand the epidemiology of BKD in this country. The patterns of genomic variation are consistent with multiple introductions to Chile, followed by rapid dissemination over a 30 year period. The estimated dates of introduction broadly coincide with major events in the development of the Chilean aquaculture industry. We find evidence for significant barriers to transmission of BKD in the Chilean salmon production chain that may also be explained by previously undescribed signals of host tropism in R. salmoninarum. Understanding the genomic epidemiology of BKD can inform disease intervention and improve sustainability of the economically important salmon industry. This article contains data hosted by Microreact.

Piggy: a rapid, large-scale pan-genome analysis tool for intergenic regions in bacteria.

Background: The concept of the "pan-genome," which refers to the total complement of genes within a given sample or species, is well established in bacterial genomics. Rapid and scalable pipelines are available for managing and interpreting pan-genomes from large batches of annotated assemblies. However, despite overwhelming evidence that variation in intergenic regions in bacteria can directly influence phenotypes, most current approaches for analyzing pan-genomes focus exclusively on protein-coding sequences. Findings: To address this we present Piggy, a novel pipeline that emulates Roary except that it is based only on intergenic regions. A key utility provided by Piggy is the detection of highly divergent ("switched") intergenic regions (IGRs) upstream of genes. We demonstrate the use of Piggy on large datasets of clinically important lineages of Staphylococcus aureus and Escherichia coli. Conclusions: For S. aureus, we show that highly divergent (switched) IGRs are associated with differences in gene expression and we establish a multilocus reference database of IGR alleles (igMLST; implemented in BIGSdb).

Draft Genome Sequences of Salinivibrio proteolyticus, Salinivibrio sharmensis, Salinivibrio siamensis, Salinivibrio costicola subsp. alcaliphilus, Salinivibrio costicola subsp. vallismortis, and 29 New Isolates Belonging to the Genus Salinivibrio.

The draft genome sequences of 5 type strains of species of the halophilic genus Salinivibrio and 29 new isolates from different hypersaline habitats belonging to the genus Salinivibrio have been determined. The genomes have 3,123,148 to 3,641,359 bp, a G+C content of 49.2 to 50.9%, and 2,898 to 3,404 open reading frames (ORFs).

Comparative Analyses of Selection Operating on Nontranslated Intergenic Regions of Diverse Bacterial Species.

Nontranslated intergenic regions (IGRs) compose 10-15% of bacterial genomes, and contain many regulatory elements with key functions. Despite this, there are few systematic studies on the strength and direction of selection operating on IGRs in bacteria using whole-genome sequence data sets. Here we exploit representative whole-genome data sets from six diverse bacterial species: Staphylococcus aureus, Streptococcus pneumoniae, Mycobacterium tuberculosis, Salmonella enterica, Klebsiella pneumoniae, and Escherichia coli We compare patterns of selection operating on IGRs using two independent methods: the proportion of singleton mutations and the dI/dS ratio, where dI is the number of intergenic SNPs per intergenic site. We find that the strength of purifying selection operating over all intergenic sites is consistently intermediate between that operating on synonymous and nonsynonymous sites. Ribosome binding sites and noncoding RNAs tend to be under stronger selective constraint than promoters and Rho-independent terminators. Strikingly, a clear signal of purifying selection remains even when all these major categories of regulatory elements are excluded, and this constraint is highest immediately upstream of genes. While a paucity of variation means that the data for M. tuberculosis are more equivocal than for the other species, we find strong evidence for positive selection within promoters of this species. This points to a key adaptive role for regulatory changes in this important pathogen. Our study underlines the feasibility and utility of gauging the selective forces operating on bacterial IGRs from whole-genome sequence data, and suggests that our current understanding of the functionality of these sequences is far from complete.