Centralized industrialization of pork in Europe and America contributes to the global spread of Salmonella enterica

Abstract

Salmonella enterica causes severe food-borne infections through contamination of the food supply chain. Its evolution has been associated with human activities, especially animal husbandry. Advances in intensive farming and global transportation have substantially reshaped the pig industry, but their impact on the evolution of associated zoonotic pathogens such as S. enterica remains unresolved. Here we investigated the population fluctuation, accumulation of antimicrobial resistance genes and international serovar Choleraesuis transmission of nine pig-enriched S. enterica populations comprising more than 9,000 genomes. Most changes were found to be attributable to the developments of the modern pig industry. All pig-enriched salmonellae experienced host transfers in pigs and/or population expansions over the past century, with pigs and pork having become the main sources of S. enterica transmissions to other hosts. Overall, our analysis revealed strong associations between the transmission of pig-enriched salmonellae and the global pork trade.

Fig. 1. Summary of the pig-enriched ceBGs in the Salmonella database in EnteroBase.

a, Histogram of the numbers of Salmonella strains for each non-human source in EnteroBase. b, Histogram of the numbers of pig-associated Salmonella strains per year in 1885–2022. c, Bubble plot of the 61 pig-associated ceBGs, each with ≥20 strains. Each ceBG is proportional in size to the number of strains in it and placed according to the numbers and percentages of its pig-associated strains. d, Hierarchical bubble plot of the 61 pig-associated ceBGs as in c. The sizes of the circles are proportional to the number of strains, and the three levels in the plot represent (from outer to inner) clusters at the levels of subspecies (HC2850), super-lineage (HC2000) and ceBG (HC900), as described previously. Pie charts represent the proportions of strains from different sources. The 9 pig-enriched ceBGs, each with >40% pig strains, are labelled in c and d. Source data

Fig. 2. Population dynamics, ARGs and global transmissions of ceBG1272 (Choleraesuis).

a, The maximum-likelihood phylogeny (left), metadata (middle) and predicted ARGs for all strains in ceBG1272 (right). The predicted lineages, clades and clusters are labelled near the associated branches. The associations between the numbers of predicted ARGs and the sampling years, source categories and countries are also visualized. b,c, The associations between the numbers of predicted ARGs and the sampling years in Paratyphi C lineages (b) and the association between the number of ARGs per strain and host sources (c). d, Visualization of the correlations between the numbers of predicted ARGs and countries in Choleraesuis. e–g, Bayesian inferences of the population dynamics of ceBG1272 over the past ~2,500 years. e, Global transmission of ceBG1272. Pie charts show the proportional composition of clades in each major country, and the arrows show the transmissions reconstructed based on the tree in f, with the transmission dates shown nearby. The pie charts and arrows were colour coded based on the associated clades. Inset: the Simpson diversity of clades in each geographic region in the world. f, The MCC tree of ceBG1272 by BEAST 2. The branches were colour coded based on the most probable ancestral geographic origins (as in the key). Pie charts of all possible geographic origins are shown over certain nodes where the most probable origins had <90% posterior supports. The dates of origin for some branches are shown together with the 95% confidence intervals in brackets. g, The fluctuation of effective population sizes with time by the ‘skygrowth’ package in R. Arrows point to the time of three major developments in the modern pig industry. Credit: Map in c, Santiago H. Cardona (https://github.com/hrcarsan/world-map/blob/master/LICENSE). Source data

Fig. 3. Host transfers for the pig-enriched ceBGs.

a, The curves show the dynamic changes of the proportional host sources with time for each ceBG. The ancestral host associations were predicted by TreeTime. The predicted median effective population sizes are also shown as black curves for ceBG1272 (Choleraesuis) and ceBG3 (Derby). The period for host transfers into pigs (red) or population expansions (yellow) is shown above each plot. A detailed prediction of the population dynamics for all nine ceBGs can also be found in Extended Data Fig. 5. b, Proportional source of host transfers summarized for all nine ceBGs in the past 50 years. Detailed host transfer data for each ceBG can be found in Extended Data Fig. 6. c, Proportional source of host transfers in the past 50 years summarized for eight pig-containing ceBGs that have 5–35% of pig strains, including ceBGs of 5, 8, 22, 125, 276, 709 and 1,898 (Fig. 1c). b,c, The arrows show the direction of transfers and are colour coded by average frequencies. The sources with the most contributions are indicated with asterisks. d, Median numbers of disrupted coding sequences (CDSs) per genome with 95% confidence intervals in all pig-enriched and pig-containing ceBGs. Source data

Fig. 4. Association between the transmission of the pig-enriched ceBGs and the global trade of pig-related products.

a,b, Visualization of the international trade of all pork-related products (a) and the transmissions of pig-enriched ceBGs (b). The pie charts show the relative proportions of source continents for the products or pathogens to the target continents. c, UMAP plot of Pearson’s correlations among the trade and the transmission of pig-enriched salmonellae. Each coloured dot in the plot shows an animal-related product as in the Harvard database, and the grey dots are other, non-animal, products. The triangle shows intercontinental transmission data summarized from all pig-enriched salmonellae. The insert highlights the dashed box in the plot, with the arrows specifying the correlation coefficient (R) between the transmissions of pathogens and the trade of pig-related products. d–g, Linear regressions of different categories of pig-related products (x-axis) and the intercontinental transmissions of pig-enriched ceBGs (y-axis). The correlation coefficient values for the linear regressions are also shown. Credit: Maps in a,b, Santiago H. Cardona (https://github.com/hrcarsan/world-map/blob/master/LICENSE). Source data

Fig. 5. The influences of modern agriculture on the population dynamics of S. enterica serovars.

This image provides a visual summary of the paper. Agricultural production has become increasingly modernized over the past half century. On the one hand, the pattern of large-scale intensive pig farming has led to the emergence and population expansion of pig-enriched Salmonella; on the other hand, globalized trade exchanges concerning pigs have similarly increased the probability of global transmission of pig-enriched S. enterica serovars. In addition, with the use of antibiotics in the process, more and more pig-enriched Salmonella has obtained new antibiotic resistance genes, including many of the previously reported human-specific antibiotic resistance genes. The impact of the development model of modern agriculture on pig-enriched Salmonella is comprehensive and far reaching. Credit: pill, pig and globe icons adapted from The Noun Project under a Creative Commons license CC0 1.0.

Extended Data Fig. 1. Circular presentation of the maximum-likelihood phylogeny in Fig. 1a.

Outer rings: The clade, source, and geographic origin of each strain. Colored arcs underneath the tree show the cluster assignments as in the key. Source data

Extended Data Fig. 2. Evaluation of the presence of temporal signal in S. enterica serovar Choleraesuis.

The analysis was performed on 587 genomes. a, Linear regression between root-to-tip distances of strains and the sampling years with a coefficient of determination (R²) of 0.67. b, Substantially lower R² values were obtained for ten date-randomisation datasets. c, The average (dots) and standard deviation (error bars) of the substitution rates for actual data (black) and ten date-randomization datasets (red), estimated by BactDating. Source data

Extended Data Fig. 3. Assessing the existence of temporal signal randomization test in pig-enriched ceBGs.

a, Coefficients of determination (R²) for the ten date-randomisation tests were obtained by linear regression between root-to-tip distances of strains and the sampling years in pig-enriched ceBGs. b, The average (dots) and standard deviations (error bars) of the substitution rates for actual data (black) and ten date-randomization datasets (red) by BactDating. Pig-enriched ceBGs including ceBG3 (n = 3136), ceBG10 (n = 622), ceBG17 (n = 155), ceBG35 (n = 516), ceBG37 (n = 1012), ceBG40 (n = 176), ceBG459 (n = 1441), ceBG621 (n = 787), ceBG1272 (n = 586). The ancient sample was not included in ceBG1272 in both a and b. Source data

Extended Data Fig. 4. The geographic states of the ancestral nodes in the Choleraesuis population after downsamplings.

a, Downsampling tests of up to ten strains per country/region. b, Downsampling tests of up to five strains per country/region. Both: the geographic states were predicted using TreeTime. Each test was run in 100 parallels. The pie charts illustrate the proportions of the best-supported geographic states in the 100 parallels for the corresponding ancestral nodes. Source data

Extended Data Fig. 5. Reconstructed ancestral host associations for all nodes in nine pig-enriched ceBGs.

The corresponding ceBGs for each panel are: a, ceBG3 (Derby), b, ceBG1272 (Choleraesuis), c, ceBG17 (Chailey), d, ceBG35 (Worthington), e, ceBG37 (London), f, ceBG40 (Cerro), g, ceBG459 (Johannesburg), h, ceBG621 (Ohio), i, ceBG10 (Adelaide). The mean values of effective population sizes with time were also shown for a and b, with 95% confidence intervals in grey shapes. The orange and grey boxes show the periods of population expansion (a,b) or the periods of host transfers into pigs (c-i). Source data

Extended Data Fig. 6. The host transmission frequency of nine pig-enriched ceBGs.

The corresponding ceBGs for each panel are: a, ceBG10 (Adelaide), b, ceBG1272 (Choleraesuis), c, ceBG17 (Chailey), d, ceBG3 (Derby), e, ceBG35 (Worthington), f, ceBG37 (London), g, ceBG 40 (Cerro), h, ceBG 459 (Johannesburg), i, ceBG 621 (Ohio). Different arrows represent the direction of host transmission, with darker colors indicating higher transfer frequency. “*” marks the most contributing host sources for the transmissions. Source data

Extended Data Fig. 7. A 5 ×5 table showing the summarised international transmission events of all nine pig-enriched ceBGs in Salmonella in the past 50 years.

The maximum-likelihood phylogeny was reconstructed based on SNPs in the core genome of each ceBG and the date of origin was estimated using BactDating. TreeTime was applied to reconstruct the country sources for all nodes in the tree. A transmission was recorded when the ancestral node and descending node of a branch were different. All transmissions were then summarized and grouped based on their associated continents. Source data

Extended Data Fig. 8. Pearson’s correlations analysis of sub-classification for pig-related products.

The Sub classifications involved include a, Pork offal, frozen (021011); b, Pork, frozen (020649); c, Pork, frozen cut (020329); d, Pork, preserved (020322); e, Pork offal, fresh (020630); f, Pig fat (020910); g, Pork, fresh cuted (020912); h, Pork, fresh (020319); i, Pig carcasses (020310); j, Pig, live breeding (010391); k, Pig, live, less 50 kg (010392); l, Pig, live, over 50 kg (020311). The codes in parentheses represent the Harmonized System Codes of the products. Analyzing the correlation between the intercontinental dissemination of each pig-related commodity and the intercontinental transmission of pig-enriched ceBGs. R: Pearson’s correlation coefficient. Source data