Global spread of *Salmonella Enteritidis* via centralized sourcing and international trade of poultry breeding stocks

Poultry products are major sources of dietary protein globally, with industrialized poultry production organized as a breeding pyramid in which genetic stocks at the top supply the entire industry. Since the 1940s, selective breeding and consolidation have transformed the sector, with significant mergers and acquisitions from the early 1980s leading to centralized sourcing of breeding stocks for both broilers and layers. In the 1980s, Salmonella enterica serotype Enteritidis (SE) infections linked to poultry increased simultaneously in North and South America and Europe, later spreading to Asia and Africa. Although incidence declined in the US and UK by the late 1990s, SE remains a major public health concern, exemplified by a large 2015–2018 European outbreak linked to eggs. A historical puzzle persists regarding how SE spread so rapidly and widely. Observing the coincidence between the onset of the SE pandemic and consolidation of breeding stocks, the authors hypothesize that centralized sourcing and international trade of SE-infected breeding stocks explains the synchronized, expansive spread. Public surveillance data for SE in breeding stocks are sparse and detailed provider-level trade data are unavailable, but large-scale whole-genome sequencing of SE and comprehensive international live poultry trade records exist. The study aims to integrate global SE genomic data with international poultry trade data to evaluate the likelihood, scale, and duration of SE dissemination via breeding stocks.

Circumstantial and regional evidence suggests a role for breeding stocks in SE spread. SE or other Salmonella serotypes were detected in primary breeding flocks in the UK (1988) and US (1999). Sweden’s control of Salmonella in imported grandparent stocks was credited with preventing a national SE pandemic, and a similar approach aided clearance in the Netherlands. In Brazil, SE was isolated from transport boxes of imported day-old chicks (1997–1998), and national subtyping (1986–2010) suggested a common origin of circulating strains. In Japan, SE capable of transovarian infection was found in chicks imported from England (1988–1989), coinciding with a surge in SE cases and clone replacement. In Egypt, SE was isolated from imported day-old chicks in the early 2000s, and the country imported over 110 million live birds in 2006. Poultry has repeatedly seen emergence of other S. enterica serotypes (Pullorum, Gallinarum, Enteritidis, Heidelberg, Kentucky), and more recently a multidrug-resistant Infantis clone spread through poultry in Europe and the US. Prior traceback investigations have linked SE spread via eggs in Europe and avian influenza via live poultry in China at outbreak scale, but a global chain of evidence connecting SE spread to breeding stock trade has been lacking.

Data sources and genomic analyses: The study retrieved 33,142 SE genomes from EnteroBase as of November 2020; after excluding genomes lacking source information, cgST, unreleased entries, and applying quality filters, 30,015 genomes from 98 countries (1949–2020) were analyzed using cgMLST to assess population structure and genomic diversity across 32 source categories. Minimum spanning trees were built (GrapeTree, MSTree V2). Intra-source genomic diversity was measured by pairwise allelic differences (PADs), selecting one representative per cgST to reduce redundancy. Phylogeny and phylodynamics: For high-resolution SNP-based phylogeny, a representative subset was selected from the 30,015 genomes, including US poultry isolates (historical ≤2009 and post-2009 spanning major clades) and diverse non-US poultry and non-poultry isolates worldwide. Sequencing reads were collected from SRA/ENA; additional Mauritius poultry isolates were sequenced on Illumina MiSeq. Quality control removed genomes with assembly N50 <100 kb or coverage <30x and outliers not in the main EnteroBase cluster. Reads were trimmed (Trimmomatic), assembled (SPAdes), assessed (QUAST). SNPs were called with SnapperDB against reference P125109 after masking repetitive and phage regions (MUMmer, PHASTER) and removing recombination (Gubbins). A maximum-likelihood tree (PhyML) was built. To mitigate sampling bias, clusters of near-duplicate genomes (same source, year, country, <10 SNPs) were collapsed to one representative, yielding 914 genomes from 46 countries (1954–2020) for phylodynamic analyses. Temporal signal screening and BEAST analyses: A custom Python script screened all internal nodes for temporal signal (R^2 > 0.4 between sampling year and root-to-tip distances). Subtrees with strong temporal signal were refined with TempEst. After removing five outliers, 317 genomes (Global lineage) and 54 genomes (Atlantic lineage) underwent BEAST v1.10.4 analyses to infer MRCA dates, phylogeography (ancestral trait reconstruction), and effective population size dynamics (Bayesian skyline). Multiple clock and demographic models were compared via marginal likelihood (path sampling/stepping-stone); Bayesian skyline with relaxed log-normal clock was favored. Maximum clade credibility trees were annotated (TreeAnnotator), and parameter estimates summarized (Tracer). Trade data compilation: International live poultry trade data were compiled from FAO (live chickens, 1986–2019; earlier years lack importer info), OEC (days-old birds <185 g, 1995–2017; eggs under HS 040700), and USDA FAS (US exports distinguishing breeding stocks and egg types, 1989–2020). Historical values were inflation-adjusted to constant 2019 USD. Analyses quantified intercontinental trade volumes and exporter shares. GLM phylogeographic modeling: The generalized linear model extension of Bayesian phylogeography in BEAST assessed predictors of international spread for lineages with strong temporal signal (Global and Atlantic). Predictors included trade of eggs (hatching and unfertilized combined in OEC), live chickens <185 g, live chickens >185 g, whole chickens, chicken cuts, geographic distance (capital-to-capital), exporter sample size, and importer sample size. Aggregated bilateral trade values (1995–2017) from OEC were used. Analyses were performed for the Global lineage (275 isolates from 17 countries with ≥5 isolates each) and Atlantic lineage (29 isolates from 4 countries). A sensitivity analysis for the Global lineage used only chicken isolates (n=107), excluding egg isolates (n=23).

- Intercontinental genetic links: Six US post-harvest chicken isolates (2015–2020) were 0–7 SNPs (1 cgMLST allele) from a pre-harvest broiler isolate from Suriname (2016), indicating a common source consistent with breeding stock import from the US. Additional links included Netherlands–Brazil, Denmark–Brazil, Turkey–China, and Colombia–USA (1–4 cgMLST alleles, 2006–2018). Eight instances connected a human isolate on one continent to a poultry isolate on another. - Other serotypes: Pre-harvest chicken isolates of Salmonella Ohio in Suriname and Salmonella Kentucky in Barbados and Trinidad and Tobago (2016–2019) were closely related (1–5 cgMLST alleles) to US poultry isolates, consistent with poultry-mediated spread. - Genomic diversity and population structure: Poultry isolates (from 28 countries) clustered tightly into two major lineages and showed the lowest intra-source genomic diversity (median PAD 110) compared with other sources (medians 152–357). Three poultry sub-populations had median PADs of 34, 59, and 107. Poultry and human PAD distributions were parallel, suggesting shared populations. - Chicken–egg unity: Chicken and egg isolates were intermingled in phylogeny; for each egg isolate, the closest chicken isolate was 0–125 cgMLST alleles (median 17) or 0–101 SNPs (median 15) away. Ancestral state inference commonly showed egg isolates with chicken ancestors and vice versa within <5 years. - Lineages and geography: Three major lineages were identified: Global (all inhabited continents), Atlantic (dominated by Europe and US, including the 2015–2018 European egg-linked outbreak), and US (largely US, includes Suriname isolate). Country-specific clades (e.g., Chile, Brazil in Global lineage) showed signs of domestic clonal expansion and diversification, whereas Mauritius isolates formed a recent, less-diverse cluster within the Atlantic lineage. - Temporal signals and dating: Strong temporal signals (R^2 > 0.4) enabled robust phylodynamics for Global and Atlantic lineages. MRCA of the Global lineage dated to the late 1970s (95% HPD 1970–1979), preceding major global consolidation of breeding stocks. MRCA of the Atlantic lineage dated to the 1940s (95% HPD 1918–1970), aligning with early modern poultry breeding in Europe and the US. Phylogeographic reconstruction suggested European origins for Atlantic and South American origins for Global among sampled isolates (noting potential sampling biases). - Population dynamics: Effective population size of Global lineage expanded rapidly in the 1980s and plateaued since the mid-1990s. The Atlantic lineage expanded in the 1990s and declined from the early 2000s, consistent with epidemiological declines in US/Europe. - Trade concordance: From 1962–2019, intercontinental live poultry trade increased >8-fold (inflation-adjusted), with Europe and the US together supplying 84–98% of intercontinental exports annually. All 46 countries represented in the phylogeny imported breeding stocks/live poultry from Europe and/or the US. - GLM results: For the Global lineage, international egg trade (combined category including hatching and table eggs) had high inclusion probability (0.89) with positive GLM coefficients, indicating strong association with SE dispersal. A chicken-only sensitivity analysis produced similar results (with higher contribution of exporter sample size due to sampling imbalance), reinforcing the role of hatching egg trade rather than unfertilized eggs. USDA FAS data corroborated substantial shares of exported eggs being hatching eggs used as breeding stocks and documented Salmonella on eggshells in primary breeder hatcheries in US surveys (1991, 1998). For the Atlantic lineage, no variable reached meaningful inclusion probability, likely due to small sample size (29 isolates across 4 countries).

Multiple lines of evidence converge to support breeding stock–mediated, global-scale dissemination of Salmonella Enteritidis from centralized origins. Near-identical SE strains in domestically raised Suriname poultry and US retail chicken strongly indicate intercontinental transmission via breeding stocks and demonstrate that an imported strain can traverse an importing country’s production pyramid to pre-harvest flocks. Similar patterns in Mauritius and country-specific clades (Chile, Brazil) suggest introductions followed by domestic establishment and diversification, indicating protracted spread that may continue to the present. The tight genomic clustering and reduced diversity of poultry SE compared to other sources are consistent with a limited number of centralized origins rather than numerous unrelated introductions. Phylodynamics place the emergence of the Global lineage just before industry consolidation in the 1980s, with population expansion coinciding with the pandemic era and a subsequent plateau; the Atlantic lineage appears linked to early industrial breeding in the US/Europe, with a decline paralleling observed epidemiology but persistence evidenced by recent large outbreaks. The overwhelming dominance of Europe/US in intercontinental breeding stock exports and the GLM inference implicating egg (hatching egg) trade as a key driver quantitatively tie trade dynamics to SE dispersal. Potential confounders were examined. Although alternative routes such as feedstuff or migratory birds could, in principle, disseminate Salmonella, surveillance and prior studies found little support for SE in poultry feeds and limited prevalence in wild birds, and established wild-bird Salmonella lineages differ from poultry-associated ones. The polyclonal nature of the pandemic by phage typing is reconciled by WGS, which groups diverse phage types within a few common evolutionary lineages. Phylogeographic origin inferences may be affected by dataset imbalances and historical isolates, yet remain consistent with unidirectional trade patterns from Europe/US to other regions. Given the massive amplification at the top of the breeding pyramid, even rare contamination events in breeding stocks can have outsized public health impacts despite downstream control programs.

By integrating global SE genomes with international live poultry and egg trade data, the study provides multifaceted evidence that centralized sourcing and international trade of poultry breeding stocks have driven the global spread and long-term persistence of Salmonella Enteritidis. Findings include intercontinental near-identity of poultry isolates, reduced genomic diversity indicating few origins, phylodynamic timing consistent with industry consolidation, and quantitative support for hatching egg trade as a key dispersal driver. These results highlight a critical, highly concentrated, and relatively opaque segment of the food chain as a major foodborne risk. Future efforts should focus on enhanced transparency, surveillance, and targeted interventions at primary breeding and hatchery levels, including monitoring of breeding flocks and hatching eggs, to prevent introduction and amplification of pathogens. The framework demonstrated here can be applied to other emergent poultry-associated pathogens.

- Lack of direct SE isolates from primary breeding stocks limits definitive source attribution; reliance on downstream poultry and product isolates may introduce inference uncertainty. - Sparse public surveillance and high biosecurity around primary breeding, along with emphasis on intellectual property, restrict data availability and transparency. - Potential sampling biases and unbalanced datasets (e.g., overrepresentation of certain regions, few historical isolates) can skew phylogeographic reconstructions and lineage origin inferences. - GLM analysis for the Atlantic lineage was underpowered due to small sample size and limited country coverage; results are inconclusive for that lineage. - Trade data limitations include inability to distinguish hatching versus table eggs in OEC data (addressed partially by FAS data) and lack of provider-level trade flow details. - Some intercontinental genetic links could conceivably reflect processed poultry trade or human travel in human–poultry pairs; while specific cases (e.g., Suriname) minimize these possibilities, they cannot be universally excluded. - Limited genomic surveillance in certain countries (e.g., Suriname) constrains assessment of domestic dissemination and persistence. - Alternative routes (feed, wildlife) cannot be entirely excluded, though available evidence suggests they are less likely drivers for SE spread in industrial poultry.