A multicenter genomic epidemiological investigation in Brazil, Chile, and Mexico reveals the diversity and persistence of Salmonella populations in surface waters

ABSTRACT This study examined the diversity and persistence of Salmonella in the surface waters of agricultural regions of Brazil, Chile, and Mexico. Research groups (three in 2019–2020 and five in 2021–2022) conducted a long-term survey of surface water across 5–8 months annually (n = 30 monthly). On-site, each team filtered 10-L water samples with modified Moore Swabs to capture Salmonella, which were then isolated and identified using conventional microbiological techniques. Salmonella isolates were sequenced on Illumina platforms. Salmonella was present in 1,493/3,291 water samples (45.8%), with varying isolation rates across countries and years. Newport, Infantis, and Typhimurium were the most frequent among the 128 different serovars. Notably, 22 serovars were found in all three countries, representing almost half of the 1,911 different isolates collected. The resistome comprised 72 antimicrobial resistance (AMR) genes and six point mutations in three genes. At least one AMR determinant was observed in 33.8% (646/1,911) of the isolates, of which 47.4% (306/646) were potentially multidrug resistant. Phylogeny based on core genome multilocus sequence typing (cgMLST) showed that most isolates clustered according to sequence type and country of origin. Only 14 cgMLST multi-country clusters were detected among the 275 clusters. However, further analysis confirmed that close genetic relatedness occurred mostly among isolates from the same country, with three exceptions. Interestingly, isolates closely related phylogenetically were recovered over multiple years within the same country, indicating the persistence of certain Salmonella in those areas. In conclusion, surface waters in these regions are consistently contaminated with diverse Salmonella, including strains that persist over time. IMPORTANCE Salmonella is a leading foodborne pathogen responsible for millions of illnesses, hospitalizations, and deaths annually. Although Salmonella-contaminated water has now been recognized as an important contamination source in the agrifood chain, there is a lack of knowledge on the global occurrence and diversity of Salmonella in surface water. Moreover, there has been insufficient research on Salmonella in surface waters from Latin American countries that are major producers and exporters of agricultural products. Incorporating genetic profiling of Salmonella isolates from underrepresented regions, such as Latin America, enhances our understanding of the pathogen’s ecology, evolution, antimicrobial resistance, and pathogenicity. Moreover, leveraging genomic data derived from pathogens isolated from diverse geographical areas is critical for assessing the potential public health risk posed by the pathogen and expediting investigations of foodborne outbreaks. Ultimately, global efforts contribute significantly to reducing the incidence of foodborne infections.

IMPORTANCE Salmonella is a leading foodborne pathogen responsible for millions of illnesses, hospitalizations, and deaths annually.Although Salmonella-contaminated water has now been recognized as an important contamination source in the agrifood chain, there is a lack of knowledge on the global occurrence and diversity of Salmo nella in surface water.Moreover, there has been insufficient research on Salmonella in surface waters from Latin American countries that are major producers and exporters of agricultural products.Incorporating genetic profiling of Salmonella isolates from underrepresented regions, such as Latin America, enhances our understanding of the pathogen's ecology, evolution, antimicrobial resistance, and pathogenicity.Moreover, leveraging genomic data derived from pathogens isolated from diverse geographical areas is critical for assessing the potential public health risk posed by the pathogen and expediting investigations of foodborne outbreaks.Ultimately, global efforts contribute significantly to reducing the incidence of foodborne infections.KEYWORDS water microbiological contamination, WGS, Salmonella persistence, Latin America S almonella is a major global health concern, responsible for over 95 million cases and more than 50,000 deaths annually, predominantly in developing countries (1).The pathogen causes disease in a wide range of hosts, survives in various environments for extended periods, and is mainly transmitted to humans through contaminated food and water (2), leading to multiple salmonellosis outbreaks each year (2)(3)(4).While traditionally associated with meat and poultry, recent outbreaks increasingly implicate produce, such as onions, alfalfa sprouts, peaches, and papayas (5).One of the primary sources of produce contamination is irrigation water (6)(7)(8)(9).
In Latin America, as in other regions, surface waters such as rivers, canals, dams, and lakes are used extensively for crop irrigation.Moreover, studies have reported these water sources are vulnerable to contamination with pathogens and other hazards (9,10).These water sources are critical in understanding the interconnected health of humans, animals, and the environment, as per the One Health approach.They serve as a reliable indicator of regional health and potential reservoirs of various pathogens, including Salmonella.
Global studies have revealed widespread Salmonella contamination in surface waters (11)(12)(13)(14)(15).A meta-analysis examining published reports from several countries estimates the presence of Salmonella in surface water to be 31% (16).Recently, high rates of Salmonella contamination in the surface waters of California (56.4%) and Pennsylvania (49%) have also been reported (17).However, not all Salmonella serovars are epidemio logically relevant, and some seem to pose higher risks to humans (2).Studies have identified both high-risk serovars, such as Newport, Typhimurium, and Enteritidis, and less common types in surface water (11,(17)(18)(19)(20).
Research specific to surface waters in Latin America is limited, with existing studies often focusing on small areas or single countries and are of a cross-sectional nature (18,19,21,22).Comprehensive and longitudinal research across the region could reveal the diversity and commonality of Salmonella serovars, their resistome, and their ability to persist and disseminate across different regions in Latin America.This investigation aimed to understand the broader picture of Salmonella's population in agricultural waters in different regions of Latin America to obtain further insights into its role in food safety and its global diversity.

Samples and isolates
Water samples were obtained in Mexico (n = 689) and Chile (n = 2,036) from 2019 to 2022.Additionally, through 2021 and 2022, 542 samples were obtained in Brazil (Table 1; Fig. 1).Reports for Salmonella isolation rates in 2019 and 2020 were published by Toro et al. for Chile and Ballesteros-Nova et al. for Mexico (21,22).The sampling procedure for four of the teams involved on-site filtration of 10 L of surface water (rivers, canals, streams, ponds, and dams, among others) through a modified Moore swab as described by Sbodio et al. (23), using a peristaltic pump.Samples from Paraíba State in Brazil were taken in triplicates, filtering a total of 30 L per sample.Samples were transported on ice to one of the five participant local teams' laboratories: Laboratorio de Microbiología y Probióticos, Universidad de Chile, Santiago, Chile; Laboratorio de Inocuidad Veterinaria, Pontificia Universidad Católica de Chile, Santiago, Chile; Labo ratorio de Epidemiologia Veterinaria, Universidad Autónoma de México, Mexico City, Mexico; Laboratório de Avaliação de Produtos de Origem Animal, Universidade Federal da Paraíba, Areia, Paraíba, Brazil; Laboratório de Investigação em Microbiologia Médica, Universidade Federal do Rio de Janeiro, Rio de Janeiro, Brazil.Samples were stored in refrigeration until processing within 7 days.The samples were processed follow ing the U.S. Food and Drug Administration (FDA)'s Bacteriological Analytical Manual methodology for Salmonella spp. with minor modifications (https://www.fda.gov/food/laboratory-methods-food/bam-chapter-5-salmonella), as described by Toro et al. (21).Presumptive colonies (1-10 per sample) were confirmed as Salmonella through invA PCR with primers F 5′ GAATCCTCAGTTTTTCAACGTTTC 3′ and R 5′ TAGCCGTAACAACCAATAC AAATG 3′ and kept in brain heart infusion (BHI) broth with 20% glycerol at −80°C for further analysis (24).

Whole-genome sequencing, assembly, and quality control
Over 4,255 Salmonella isolates from 1,169 samples were sequenced at the Center for Food Safety and Applied Nutrition (CFSAN/FDA).All isolates were confirmed as Salmonella with VitekMS (Biomerieux, Marcy-l'Étoile, France) before DNA extraction.Then, individual colonies were grown overnight in TSB broth (BD, Sparks, MD, USA) at 37°C.DNA was purified on a Maxwell RSC-48 Instrument using the Cultured Cells DNA Kit (Promega, Madison, WI, USA).Libraries were prepared with the Illumina DNA Prep kit (Illumina, San Diego, CA, USA) on the Sciclone G3 NGSx iQ Workstation (Perkin Elmer, MA, USA), and sequencing was performed on the Illumina NextSeq 2000 using the NextSeq 1000/2000 P2 reagents 300 cycles with the 150 paired-end chemistry (Illumina).
Read quality control procedures were performed following the CFSAN MicroRunQC pipeline in GalaxyTrakr at CFSAN, FDA (26).The pipeline includes Trimmomatic for read trimming (27), skesa (V2.4.0) for assembly (28), mlst (V2.23.0) to identify species and determine sequence type (29,30), and fastq-scan (V1.0.1) for assembly statistics (31).Genomes that passed quality control (32) were uploaded as SRA files to the NCBI (Dataset S1) in batches and made publicly available through the NCBI Pathogen Detection browser (PD).NCBI-curated assembled genomes, public metadata, assembly metrics, and antimicrobial resistance (AMR) gene results from the PD browser were downloaded for the current data analysis (https://www.ncbi.nlm.nih.gov/pathogens/).Metadata were curated using each participant's laboratory database.No genomes were available from 324 positive samples due to contamination, unavailability, low sequencing quality, and other unforeseen circumstances.Genomic data from isolates collected in Mexico during 2019 and 2020 were published in Ballesteros-Nova et al. (22).

Genomic characterization of Salmonella enterica isolates (MLST, in silico serotyping, and AMR detection)
Achtman multilocus sequence typing (MLST) types with genes aroC, dnaN, hemD, hisD, purE, sucA, and thrA were used for Salmonella MLST determination with the Ridom Seqsphere software (Ridom GmbH, Germany) (33).For serovar definition, we downloa ded the serovar prediction from PD and analyzed each selected genome with SISTR (34) in Ridom.Disagreements between PD and SISTR's serovar predictions were resolved with manual curation based on PD clustering.AMRFinderPlus results from the PD browser were used to identify the presence of AMR genes and assign AMR class to each AMR gene detected as NCBI definition for every gene (35).Genomes carrying AMR determi nants against three or more antimicrobial classes were identified as multidrug-resistant.Kruskal-Wallis and Dunn's post hoc (adjusted for multiple comparisons) tests were run in Past v4.03 (36) to investigate the potential differences in the number of AMR genes between isolates from different countries.

Phylogenomic relatedness among Salmonella isolates
In the first phase, we cleaned our database by removing clonal isolates from the same sample (i.e., isolates obtained from the same day and location).First, available genomes (n = 4,255) were retrieved from the PD browser.Then, we examined each single nucleotide polymorphism (SNP) cluster comprising genomes from the project and determined the SNP count for all genomes from the same sample.Clones were defined as genomes with 20 or fewer SNP and monophyletic relationships (37).The genome with the greatest quality (defined by N 50 and the number of contigs) was chosen to represent the clonal group in each sample.A total of 1,911 genomes were selected for further studies.
Then, we analyzed the relatedness among selected genomes (n = 1,911) through a core genome multilocus sequence typing (cgMLST) scheme in Ridom SeqSphere v9.0.2 (Ridom GmbH, Germany), which comprises 3,002 core genes for Salmonella species (33).A Minimum Spanning Tree (MST) was crafted from the resulting distance matrix.We removed 1,887 columns with missing values in at least one sample and excluded one genome that was missing more than 10% of alleles in distance columns (CLA-161-1, Chile) for a total of 1,910 genomes used for this analysis.Additionally, we examined the phylogenetic relatedness among genomes from the same serovar and different countries that resulted in mixed country clusters in the previous MST using the same methodology.

Worldwide genomic comparison
We investigated the relatedness among the genomes in our collection and genomes from isolates collected elsewhere as per PD.For that, we searched all selected genomes in the PD cluster and downloaded the SNP cluster information into a spreadsheet (22 January 2024).Each PD cluster was examined, and clustering with genomes of other origins was examined.Information was added to each record, such as the country where the genome came from, serovar, the presence of genomes from clinical isolates in the cluster, and other notes associated with the origin of the isolates sequenced.

Salmonella in surface water
Our study revealed a notable occurrence of Salmonella in surface water across three Latin American countries, with an overall isolation rate of 45.4% (1,493/3,291) over the period 2019-2022 (Table 1).The isolation rates varied significantly by country, with Brazil reporting the highest at 68.2%, followed by Mexico at 62.9% and Chile at 33.1% (Table 1).In samples testing positive for Salmonella, 1-10 isolates per sample underwent genomic sequencing, characterization, and phylogenomic analysis.

Genomic diversity and analysis of Salmonella from surface water
The number of non-clonal isolates per sample ranged from 1-6, and two samples carried 10 (Dataset S1; Table S1).Most samples (58.4% or 683 out of 1,169 examined) contained only a single clonal group (Fig. 2; Table S2).Notably, the samples from Brazil carried an average of 2.01 non-clonal isolates per sample, and two samples carried 10 different isolates representing 10 different serovars (Fig. 2; Table S2).Meanwhile, the average for Mexican samples was 1.77, while the average for non-clonal isolates from Chile was 1.34.Statistical analysis revealed that Chile carried the least number of non-clonal isolates per sample (P < 0.001), but no differences were found between the number of isolates per sample collected in Brazil and Mexico.In general, non-clonal isolates from the same sample were of different serovars (Dataset S1).Mexico (n = 686) contributed the most genomes to the collection, followed by Chile (n = 684) and Brazil (n = 541) (Tables S1 and  S2).

Sequence types and serovars of Salmonella from surface water
We identified 201 distinct sequence types (STs) among 1,911 representative isolates, but 12 were not attributed to any ST due to the incomplete sequence of one gene required for complete typing.The most prevalent STs were ST32 (7.4%; Salmonella Infantis), ST19 (4.8%; Salmonella Typhimurium), and ST13 (4.6%; Salmonella Agona).Chile exhibited the least diversity with 86 STs, while Mexico and Brazil had slightly higher diversity with 91 and 98 STs, respectively.A notable finding was that only 17 STs were present in all three countries with most STs detected in a single country (147/201; Table S3).
Similarly, the sequenced isolates were classified into 128 serovars based on the genomics analysis.The predominant serovars across the study were Newport (9.6%), Infantis (7.7%), and Typhimurium (6.7%) (Table 2).The same three serovars were the most common for the Chilean genomes, while Brazil and Mexico showed different top three serovars (Newport, Saintpaul, and Rubislaw for Brazil; Newport, Senftenberg, and Anatum for Mexico) (Table 2).Most serovars (77 out of 128) were unique to a single country, with only 22 serovars shared among all three nations (Fig. 3).Interestingly, Salmonella Enteritidis was identified predominantly in Chile (47/49 S. Enteritidis isolates) but was absent in samples from Brazil (Table 2).
Quinolone resistance determinants were the most frequently detected in each country (22.8%) (Fig. 4).The gene qnrB19, linked to reduced susceptibility to quinolones and previously described in plasmids, was the most prevalent ARG (14.3%) (Table S6).Additionally, four different point mutations in the gyrA gene, often associated with quinolone resistance, were detected in 5.8% of the genomes (110/1,911; Table S6).Tetracycline resistance was predicted in 15.8% (301/1,911) of the genomes (Fig. 4), with tet(A) as the second most frequent ARG (13.7%;Table S6).The resistance to this antimicrobial class was among the three most frequent in the three countries.The floR gene conferring resistance to phenicols ranked third in the prevalence (10.5%),   although phenicol resistance was only the fifth most common in the AMR resistance prediction (Fig. 4; Table S6).Moreover, 17 different genes linked to aminoglycosidemodifying enzymes (AME) were identified, and 129 genomes carried simultaneously three or more AME genes (Table S6).Resistance to aminoglycosides was among the four most important in each country.Notably, the mobile gene mcr-9.1,a newly described colistin resistance gene, was found in three genomes (two S. Agona from Mexico and one from S. 1,4, [5],12:d:-from Chile) (Dataset S2; Fig. 4).Finally, resistance to fosfomycin was the third most frequently detected in Salmonella from Brazil (5.0% of Brazilian genomes), but it only reached seventh place in Chile and eighth in Mexico (Fig. 4).Data analysis revealed the genomes from Mexico generally carried the highest number of AMR determinants and had the most frequent occurrence of ARG per antimicrobial class (P < 0.001) (Fig. 4; Table S6).For instance, 35.3% of genomes from Mexico carried AMR determinants against quinolones, compared to only 16.1% from Chile and 15.4% from Brazil (Fig. 4).This trend was consistent across all analyzed antimicrobial classes.Moreover, Mexican genomes collectively carried AMR determinants against 14 antimicrobial classes, whereas the Chilean and Brazilian genomes had AMR determinants against 13 and 10 families, respectively (Dataset S2).

Genomic diversity of Salmonella from surface water through core genome MLST analysis
The cgMLST analysis grouped 1,910 Salmonella genomes into 275 clusters with ≤7 gene differences (Fig. 5).Out of the 3,002 genes in the cgMLST scheme for Salmonella, only 1,115 were ubiquitously present across all genomes and were used for distance  Although most clusters with seven or fewer allele differences were composed of genomes of the same serovar and country, we detected 17 multi-country clusters (6.2%) (Fig. 5; Table S7).Specifically, only five clusters contained related genomes from all three countries, representing serovars Agona (n = 28; Cluster Interestingly, we observed that closely related genomes from the same country were isolated over long periods.For example, S. Braenderup from Mexico isolated from 2019 to 2021 formed a cluster of nine genomes and S. Enteritidis from Chile (n = 12) were isolated from 2019 to 2021 (Fig. 6a; Fig. S1b).S. Infantis from Chile (Cluster 1; Fig. 5) were collected throughout every year from 2019 to 2022 and formed a cluster of more than 40 highly related genomes (Fig. 6b).Moreover, subclusters were formed by undistinguishable S. Infantis from Chile collected over 3 (April-June) and 6 (July through December) months in 2019, and some of them were isolated in the same sampling site (Fig. 5 and 6b).

DISCUSSION
In this study, we described the presence of a large diversity of Salmonella serovars in surface waters of Latin America, with genomes genetically closely related within each country but more distant or unrelated when comparing among countries.First, Salmonella isolation rates across different regions varied.This was expected because of the regional differences in ecosystem characteristics and climatic conditions, which could affect Salmonella spread and survival in the environment.For instance, sampled regions in Central Chile have a temperate Mediterranean climate; Central Mexico has arid climates; and Brazil has tropical climates that are usually humid.This variability highlights the complexity of Salmonella's ecological presence, affected by factors such as temperature and rainfall patterns (38).The research teams from Chile, using data from the first 2 years of the project, already explored environmental and anthropogenic factors that could influence the likelihood of isolating Salmonella locally.They found that the sampling month was the factor with more impact on Salmonella isolation and that the anthropogenic variables evaluated only marginally contributed to this result (21).The other teams in the project are currently exploring local factors influencing the phenomena.
Co-occurrence of multiple Salmonella serovars was observed in almost 40% of samples.Samples from Brazil contained the highest number of serovars per sample (n = 10).This can be explained by the fact that the group in Paraiba State, Brazil, used a triplicate-sampling approach, resulting in the filtering of 30 L of water per sample compared with the 10-L samples used by the other groups, which used uniplicate sampling (10 L).In the remaining groups, we observed up to six isolates per sample.This could also have affected the isolation rate comparability among groups as a larger sample volume resulted in higher detection rates and greater diversity across samples.Other studies have described the presence of more than one Salmonella serovar in a single sample.For instance, a study in the Susquehanna River watershed in Pennsylvania reported the presence of up to 10 different Salmonella serovars per sample, with an average of three (17), which is higher than the average of 1.6 from this study.However, Deaven et al. (17) used CRISPR-SeroSeq, an amplicon-based sequencing tool that allows for all Salmonella present in a sample to be identified.In our study, we picked colonies from plates and sequenced up to 10 isolates per sample.Since one serovar might be more abundant than others by several orders of magnitude, picking colonies represent ing different populations becomes challenging, and it might underestimate the actual Salmonella diversity in our samples.This has been observed in other matrices, such as chicken carcasses and droppings (39,40).Consequently, our diversity per sample could be even more substantial than reported.
The frequent isolation of S. Newport, S. Typhimurium, and S. Infantis in our samples is noteworthy, given their high relevance in foodborne salmonellosis outbreaks and the emergence of antimicrobial-resistant strains belonging to these serovars.S. Newport has been frequently implicated in foodborne outbreaks linked to produce in the U.S.: alfalfa sprouts in 2010, cantaloupe in 2012, cucumbers in 2014, papayas in 2017, and onions in 2020 (41)(42)(43).S. Newport has also been frequently isolated from tomatoes grown on the East Coast in the U.S. (44).This serovar has been reported among the top five serovars causing human clinical cases in Brazil and Mexico for several years (45,46).Accordingly, this was the most common serovar found in surface water samples from both countries.Although S. Newport is not among the top five most prevalent human serovars in the last Chilean national report (47), it was among the most frequently identified serovars from Chile.S. Typhimurium has been reported as the first or second most frequent cause of salmonellosis in the three participant countries (45)(46)(47)(48), and it has been isolated from several other sources in Latin America, such as domestic animals (49), food of animal origin (46), and water (18,19).This serovar has also caused foodborne outbreaks in the U.S. linked to tomatoes in 2006, cantaloupes in 2012, and alfalfa sprouts in 2022 (50)(51)(52).S. Typhimurium high frequency in surface water indicates that this serovar is widespread in the environments surveyed.Interestingly, S. Typhimurium was detected in only one of the two participating groups in Brazil, probably indicating potential risk factors associated with this serovar.S. Infantis is an emerging serovar worldwide, characterized by the presence of a plasmid carrying multiple AMR genes (53).In Chile, it was the second leading cause of salmonellosis in 2018 (47), and multi-resistant isolates have been isolated from poultry facilities (32), chicken meat (33), pet foods (34), and water (19), among other sources.In Brazil, it has also been isolated from diverse sources, including the poultry industry (32), and it has been among the top 10 clinical human serovars since 2013 (45).In Mexico, the presence of this serovar has been associated with beef (46).Moreover, this serovar was linked to an outbreak involving papayas in the U.S. in 2017 (54).Altogether, the presence of these three serovars among the most frequently isolated across the whole study indicates that high-risk salmonellae contaminate surface waters used for irrigation in Latin America and are widespread in the surveyed areas, potentially becoming a risk to consumers.
Other clinically relevant serovars were found in this study.For instance, serovars that have caused outbreaks linked to produce, such as Agona 4.6% (n = 87), Anatum 2.9% (n = 56), Panama 3.8% (n = 73), Muenchen 2.6% (n = 49), Braenderup 2.5% (n = 42), Adelaide 1.8% (n = 34), Oranienburg 1.4% (n = 26), and Javiana 1.2% (n = 22), were present in all three countries in variable frequencies (Table 2).On the other hand, other relevant serovars were not collected from all three countries.For example, Salmonella Enteritidis has been the top serovar linked to human disease in Chile for several years, and it represents a top disease-causing Salmonella in Brazil and Mexico (45)(46)(47).Interestingly, almost all S. Enteritidis isolates were collected in Chile, and none came from Brazil.Salmonella Heidelberg has been consistently reported as a top serovar causing salmonellosis in Brazil (45,48), and it is linked with several foodborne outbreaks in the U.S. (55)(56)(57); this serovar was only isolated in samples from Brazil.These variations were expected and might be due to natural and anthropogenic conditions such as climates, fauna, local policies, and sanitation conditions.For example, in Brazil, vaccination in poultry against S. Enteritidis is widespread (58).Since poultry is the main reservoir for this serovar, this practice could restrict the spread of S. Enteritidis into the environment.
It is estimated that by 2050, AMR will be responsible for the deaths of 10 million people annually (59).Enterobacteriaceae is among the bacterial families most significantly impacted by the rise in AMR rates, and Salmonella is no exception (60).In our study, almost one-third of the Salmonella isolates contained AMR determinants, indicating that AMR genetic elements are widespread in the surveyed areas.Moreover, some isolates carried over 10 AMR determinants and were predicted to be resistant to nine antimicrobial families simultaneously.These findings showed that water environ ments act as a reservoir for AMR determinants, facilitating their dissemination in the environment and possible transfer to other bacteria.Antimicrobial residues in surface waters might increase this problem, which has been described in at least one of the watersheds sampled in this study (61).These findings encourage us to further investigate the phenotypic antimicrobial resistance traits of Salmonella from this study in the future.
The phylogenetic analysis showed intra-serovar diversity.Isolates from the same serovar but collected in different countries were generally not clonal as shown by the multiple cgMLST analysis, including genomes of single serovars, incrementing the technique's resolution.This highlights the relevance of performing additional analysis in front of apparently closely related genomes to detect whether the phylogenetic relatedness observed is actual or an artifact of the analysis.Some exceptions include some isolates of worldwide spread serovars Enteritidis, Infantis, and Typhimurium; large PD clusters of these serovars contained not only water isolates from multiple countries in the study but also hundreds of genomes of diverse origins, including clinical, animal, and environmental isolates.Numerous plausible hypotheses can be made about the international transmission of these isolates; however, in the absence of a trace-back investigation, these remain speculative at best.This makes it exceedingly challenging to draw definitive conclusions, as the causes may include international food trade, travel, wildlife migration (e.g., avian migration), biofilms in transoceanic boats, or other factors that remain undefined.Furthermore, we observed that multiple isolates from the same country and serovar were clonal or closely related, even from samples collected 3 years apart, suggesting the long-term persistence of specific strains in water sources.This persistence could be attributed to various factors, including the ability of Salmonella to form biofilms and survive in diverse hosts, among others (62).This is interesting since surface waters can integrate the environment with both animal production and human activities, serving as a valuable sample in the One Health context.
Interestingly, over 50% of the PD clusters comprised genomes generated in this study did not include genomes from other countries, suggesting that other unique, undiscovered Salmonella may exist in unexplored environments and that diversity is widely more extensive than expected.However, some genomes originated from clinical isolates were phylogenetically close to water genomes of this study with differences as close as 0 SNPs.The closeness between some of the water isolates and isolates from clinical cases, especially in the U.S. and the UK, cannot be dismissed, but the epidemio logical link between these genomes has yet to be determined.More importantly, this finding highlights the presence of high-risk Salmonella in surface waters in Latin America.
A significant portion of the study was conducted during the COVID-19 pandemic.The disruptions caused by this worldwide event in this study included changes in sampling schedules and locations, and it could have further impacted the representativeness of our findings.As a result, it is important to note that the Salmonella isolation rates obtained from this research are only applicable to the study design and sampled areas and should be interpreted as representative of the pathogen's prevalence in surface waters across Latin America or an individual country.
Foodborne outbreaks associated with produce have been linked to irrigation water and, specifically, to surface waters (9,63).In Latin America, surface waters are one of the primary water sources for crop irrigation (64); therefore, the presence of high-risk foodborne pathogens in these waters could represent a risk to consumers-the three countries in this study, Brazil, Chile, and Mexico, are prominent food exporters world wide.Mexico is a leading exporter of produce to its neighbor country, the United States, and Chile exports fruits in counter-season to the northern hemisphere.Brazil mainly exports meats, corn, and soybeans, and it is one of the leading producers of fruit juices worldwide.To date, some vegetable-origin foods from these countries have been linked to foodborne outbreaks in the U.S. For example, Mexican onions contaminated with Salmonella Oranienburg were attributed to causing an outbreak in 2021 (65).Accord ingly, our study provides valuable information showing that Salmonella isolates present in surface waters in Latin America are highly genetically related to Salmonella found in different food products and to others that have caused clinical cases of salmonellosis.Therefore, it is crucial to investigate foodborne diseases using the One Health perspec tive to better understand the dynamics of these types of pathogens in the environment and prevent more cases of foodborne diseases.
Surface waters in agricultural regions of Brazil, Chile, and Mexico were investigated in this study.The genomic information gleaned from our sampling efforts sheds light on the persistence and diversity of Salmonella in these regions.The extensive sampling and sequencing employed in our study distinguish it from previous investigations.For a more comprehensive understanding of the ecological dynamics of Salmonella in agricultural surface waters, we integrated these diverse components.While we observed distinctions and similarities in the distribution of Salmonella serovar prevalence among different countries and time periods, we also examined the resistome and evolutionary connec tions of the isolates.Even though antibiotic resistance, serovar risks, and epidemics are well-known topics, this study provides current data and context for the agricultural regions surveyed.Furthermore, we highlight the persistence of certain Salmonella strains over time, which has important implications for public health interventions.

Conclusion
This study highlights an important public health issue regarding surface waters contaminated with clinically significant Salmonella serovars in Latin America, some carrying multiple antimicrobial-resistance determinants.The findings underline the increased risks the pathogen poses to human and animal health and reveal the remarkable diversity of Salmonella in the region.We observed long-term contamination in surface waters within countries, which could be explained either by the persistence of strains in the aquatic environments or arising from repeated contamination events from other constantly disseminating sources.These findings suggest that most strains are rather restricted to specific regions in Latin America.The presence of many high-risk Salmonella serovars in surface waters, a key source of irrigation for produce production, poses a significant risk not only to those in direct contact with contaminated waters but also to the broader population through the consumption of produce irrigated with these waters.Improved, coordinated surveillance, control, and prevention strategies are needed to mitigate the spread and impact of Salmonella on public health to ensure food safety and protect public health across the region and beyond.
This study represents a comprehensive analysis of Salmonella diversity in surface waters across agricultural regions of Brazil, Chile, and Mexico, offering significant insights into the genomic diversity and persistence of Salmonella in Latin America that has been relatively underrepresented in previous studies.

FIG 2
FIG 2 Non-clonal Salmonella isolates per sample (%).Clonality was examined in the Pathogen Detection browser (NCBI).Clonal isolates had less than 20 SNP differences and a monophyletic clustering.Genomes available for analysis were from 269 samples from Brazil, 513 from Chile, and 387 from Mexico, for a total of 1,169 samples.
calculations.The largest cluster (Cluster 1) consisted of 61 S. Infantis isolates from Chile, 6 from Mexico, and 1 from Brazil.The next two largest clusters comprised S. Newport (Cluster 2) and S. Enteritidis (Cluster 3) from Chile.The fourth largest cluster (Cluster 4) included 29 S. Newport from Mexico collected between 2019 and 2021 and 1 from Chile in 2021.The most numerous clusters formed by Brazilian genomes included 20 Salmonella Saintpaul (Cluster 11) collected in 2021 and 2022 (Fig. 5).

FIG 4
FIG 4 Genomes (%) with predicted antimicrobial resistance to 14 antimicrobial classes by country.The number of available genomes per country: Brazil n = 541, Chile n = 685, and Mexico n = 686.

FIG 5
FIG 5 Minimum spanning tree (MST) of 1,910 Salmonella isolates obtained from the surface waters in Latin America.The MST was crafted with 1,115 genes present in all isolates.Each circle represents one isolate or a group of clonal isolates.Colors represent isolation countries: Brazil in red, Chile in green, and Mexico in purple.Gray background linking two or more circles indicates clonal groups of seven or fewer core genome gene difference.The number next to the lines between circles indicates cg gene differences.Clusters 1, 2, 3, 4, 5, 11, 12, 38, and 65 mentioned in the Results section are highlighted with a larger cluster name and a red circle around the genomes.

FIG 6
FIG 6 Minimum spanning tree (MST) of phylogenetically related isolates from a single serovar isolated in the three countries.MST generated from core genome MLST of individual serovars clustering together from different countries in the general analysis cgMLST.Each circle represents one genome or a group of clonal genomes.Each circle color represents isolation countries: Brazil in red, Chile in green, and Mexico in purple.Gray background linking two or more circles indicates clonal groups of seven or fewer core genome gene difference.The number next to the lines between circles indicates cg gene differences.The number of isolates and genes considered for distance calculation are as follows: (a) Salmonella Enteritidis genomes (n = 49) had 2,826 common genes.(b) The Salmonella Infantis collection (n = 147) had 2,656 common genes.(c) Salmonella Newport genomes (n = 183) had 2,509 common genes.Dark boxes highlight clusters comnprised of genomes from different countries.

FIG 7
FIG 7 Minimum spanning tree (MST) of non-closed phylogenetically related genomes from a single serovar isolated in the three countries.(a) Agona, (b) Corvallis, (c) Mbandaka, and (d) Soerenga.MSTs generated from core genome MSLT of genomes of individual serovars clustering together from different countries in the general analysis cgMLST.Each circle represents one genome or a group of clonal genomes.Each circle color represents isolation countries: Brazil in red, Chile in green, and Mexico in purple.Gray background linking two or more circles indicates clonal groups of seven or fewer core genome gene difference.The number next to the lines between circles indicates cg gene differences.The number of isolates and genes considered for distance calculation are as follows: (a) Salmonella Agona (n = 87) had 2,786 common genes.(b) Salmonella Corvallis genomes (n = 37) had 2,872 common genes.(c) The Salmonella Mbandaka (n = 26) group had 2,616 genes in common.(d) Salmonella Soerenga (n = 9) had 2,896 common genes.

TABLE 1
Salmonella isolation from surface water by country and year a Sample sizes varied by country and year, and samples were collected in different seasons.Brazil joined the project in 2021, so there are no results for 2019 or 2020 for Brazil.Values in parentheses indicate samples in which Salmonella was isolated/total number of samples. a

TABLE 2
Number of genomes of each Salmonella serovar in surface water samples in Latin America (n = 1,911) (Continued on next page) Research Article mBio July 2024 Volume 15 Issue 7 10.1128/mbio.00777-247

TABLE 2
Number of genomes of each Salmonella serovar in surface water samples in Latin America (n = 1,911) (Continued) (Continued on next page)

TABLE 2
Number of genomes of each Salmonella serovar in surface water samples in Latin America (n = 1,911) (Continued)

TABLE 3
Number of antimicrobial resistances predicted by isolate and country a a Prediction was based on the presence of AMR elements (genes and mutations) in each genome and phenotype associated with each element in the NCBI AMRFinderPlus database.