Destination shapes antibiotic resistance gene acquisitions, abundance increases, and diversity changes in Dutch travelers

Abstract

Background: Antimicrobial-resistant bacteria and their antimicrobial resistance (AMR) genes can spread by hitchhiking in human guts. International travel can exacerbate this public health threat when travelers acquire AMR genes endemic to their destinations and bring them back to their home countries. Prior studies have demonstrated travel-related acquisition of specific opportunistic pathogens and AMR genes, but the extent and magnitude of travel's effects on the gut resistome remain largely unknown.

Methods: Using whole metagenomic shotgun sequencing, functional metagenomics, and Dirichlet multinomial mixture models, we investigated the abundance, diversity, function, resistome architecture, and context of AMR genes in the fecal microbiomes of 190 Dutch individuals, before and after travel to diverse international locations.

Results: Travel markedly increased the abundance and α-diversity of AMR genes in the travelers' gut resistome, and we determined that 56 unique AMR genes showed significant acquisition following international travel. These acquisition events were biased towards AMR genes with efflux, inactivation, and target replacement resistance mechanisms. Travel-induced shaping of the gut resistome had distinct correlations with geographical destination, so individuals returning to The Netherlands from the same destination country were more likely to have similar resistome features. Finally, we identified and detailed specific acquisition events of high-risk, mobile genetic element-associated AMR genes including qnr fluoroquinolone resistance genes, bla_CTX-M family extended-spectrum β-lactamases, and the plasmid-borne mcr-1 colistin resistance gene.

Conclusions: Our results show that travel shapes the architecture of the human gut resistome and results in AMR gene acquisition against a variety of antimicrobial drug classes. These broad acquisitions highlight the putative risks that international travel poses to public health by gut resistome perturbation and the global spread of locally endemic AMR genes.

Keywords: Antibiotic resistance; ESBL; Resistome; Travel; mcr-1; β-Lactamases.

Fig. 1

Destinations for Dutch travelers. A total of 190 Dutch individuals’ gut microbiomes were sampled before and after traveling (380 total samples) to 4 different subregions (Northern Africa, Eastern Africa, Southern Asia, and Southeastern Asia)

Fig. 2

AMR gene abundance and α-diversity increases with travel and AMR gene β-diversity decreases. a The left panel shows the AMR gene abundance in RPKM. Each point is a sample, and the boxes are the medians with interquartile ranges for the pre-travel samples in blue and the post-travel samples in red. The p value (paired-sample t test) for the comparison is given at the top of the panel. The right panel shows the difference between the bootstrapped distributions of the post- and pre-travel samples. The red line gives the 95% confidence interval for the difference, and the point gives the estimate. b AMR gene α-diversity is measured by richness (top left panel), and Shannon Index (bottom left panel) is compared between the pre-travel (blue) and post-travel (red) samples. Each point corresponds to a given sample, and each box gives the median and interquartile range for the distribution. The p value (paired-sample t test) for the comparison is given at the top of each panel. The panels to the right of the boxplots show the difference between the bootstrapped distributions of the post- and pre-travel samples. The red line gives the 95% confidence interval for the difference and the point gives the estimate. c AMR gene β-diversity measured by Bray-Curtis dissimilarity is compared between the pre-travel (blue) and post-travel (red) samples. Each point is a comparison between two samples within the same time point group. The distributions are shown to the right of the points, and boxplots showing the median and interquartile ranges are overlaid on top of the points. The p value by paired Wilcoxon test for the comparison is shown near the top. In the right panel, the lines show the 95% confidence intervals, and points show the mean values for the pre- (blue) and post-travel (red) Bray-Curtis dissimilarity distributions. Source data is provided in the source data file (Additional file 3)

Fig. 3

Travel outweighs subject effects in shaping resistome architecture. a 95% confidence interval (red line segment), odds ratio (red point), and p value calculated by Fisher’s exact test for samples with the same time point being drawn from the same metaresistome. The black vertical line at 1 shows the expected result under the null. b Each row in this plot corresponds to a metaresistome (m1–m8) in a Dirichlet multinomial mixture model of all traveler samples. The pie charts on the left are proportional in size to the number of samples in each metaresistome. The fill of the chart corresponds to the number of individuals in the time point (pre-travel in blue and post-travel in red). The network shows the number of individuals that transition from any model to any other model following their return from abroad. The black lines indicate staying within the same model, and the green lines indicate transition from one model to another model. The thickness and opacity of the lines correspond to the number of people following that transition path. Node label sizes correspond to the number of individuals in the model from the time point. Nodes filled in blue are significantly enriched in pre-travel samples, and nodes in red are significantly enriched in post-travel samples. The right panel shows the estimates (points) and 95% confidence intervals (lines) for binomial tests of bias for pre- or post-travel samples. p values for the comparison (FDR-corrected binomial test) are given above the lines. The expected estimate under the null model is given by the dark black line at 0.5. Source data for all panels is provided in the source data file (Additional file 3)

Fig. 4

Travelers to different destinations cluster separately by resistome composition but show similar trends by abundance and diversity metrics. a The bottom panel shows the comparisons of AMR gene abundance before and after travel to the four subregions in this study. Points correspond to samples, and boxes give the median and interquartile ranges. Pre-travel is shown in blue, and post-travel is shown in red. The p values (FDR-corrected paired Wilcoxon tests) for comparisons within the region between the pre- and post-travel samples are shown above each comparison. The top panel shows the difference between the bootstrapped distributions of the post- and pre-travel samples. The red line gives the 95% confidence interval for the difference, and the point gives the estimate. b AMR gene α-diversity is measured by richness (left), and Shannon Index (right) is compared by region between the pre-travel (blue) and post-travel (red) samples. Each point corresponds to a given sample, and each box gives the median and interquartile range for the distribution. The p values (FDR-corrected paired Wilcoxon test) are above each comparison. The panels above show the difference between the bootstrapped distributions of the post- and pre-travel samples for each destination. The red line gives the 95% confidence interval for the difference, and the point gives the estimate. c The left panel compares the β-diversity for pre-travel (blue) and post-travel (red) collections for the four travel destinations. The points are pairwise Bray-Curtis dissimilarity between two samples, and the boxes represent the median and interquartile ranges of the distributions. p values (paired Wilcoxon test) are given above each comparison. The right panel shows the difference between the bootstrapped dissimilarities of the pre- and post-travel groups. The lines give the 95% confidence interval for the difference, and the point gives the estimate. Source data for all panels is provided in the source data file (Additional file 3)

Fig. 5

Travelers’ resistomes group significantly by region after travel, and Southeastern Asia and Eastern Africa have the strongest signature. a 95% confidence intervals, odds ratios, and p values for the samples with the same destination being drawn from the same metaresistome. Fisher’s exact tests were done for this comparison within the time point (y-axis). The black vertical line at 1 shows the expected result under the null. Source data for all panels is provided in the source data file (Additional file 3). b The left of this Sankey diagram has models built from the pre-travel samples, and the right has models built from the post-travel samples. Each model has a pie chart that shows the number of samples in the model (total of 190 for each time point), and these pies are divided by destination. The lines connecting the pre- and post-travel models are colored according to region (dark blue is Northern Africa, light blue is Eastern Africa, orange is Southern Asia, and red is Southeastern Asia), and their thickness is proportional to the number of samples that follow that path

Fig. 6

AMR gene abundance changes and acquisitions are unequal across AMR mechanisms. a AMR mechanism abundance is compared between pre-travel (blue) and post-travel (red) samples. Each point is a sample, and the boxes represent the median and interquartile range. p values (FDR-corrected paired Wilcoxon test) for the comparisons are given near the top of the panel. The top panel shows the difference between the bootstrapped distributions of the post- and pre-travel samples. The lines give the 95% confidence interval for the difference, and the point gives the estimate. AMR classes where the 95% confidence interval does not cross 0 are red. b AMR class abundance is compared between pre-travel (blue) and post-travel (red) samples. Each point is a sample, and the boxes represent the median and interquartile range. p values (FDR-corrected paired Wilcoxon test) for the comparisons are given near the top of the panel. The top panel shows the difference between the bootstrapped distributions of the post- and pre-travel samples. The lines are the 95% confidence interval for the difference, and the point is the estimate. AMR classes where the 95% confidence interval does not cross 0 are red. c AMR gene acquisitions or losses after travel. Each point is an AMR gene, and points are filled in according to their AMR mechanism. The x-axis is the number of individuals that had the gene in the pre-travel time point, but not in the post-travel time point. The y-axis is the number of individuals that had the gene in the post-travel time point, but not in the pre-travel time point. The red-shaded region spans significantly acquired AMR genes, the blue-shaded region spans significantly lost AMR genes, and the gray-shaded region spans genes that were not significantly acquired or lost. The diagonal line is the null of equal losses and gains for an AMR gene. The inset panel shows which AMR mechanisms were significantly acquired during travel by permutation testing. The colored histograms show the expected distribution according to 10,000 permutations, and the black vertical lines show the observed value (points in the red-shaded region of the main plot). The z-score and the FDR-corrected p value for the comparison of observations to their expected distribution are given in the top left of each plot. Source data for all panels is provided in the source data file (Additional file 3)

Fig. 7

AMR gene acquisitions and mobile genetic elements differed by travel destination. a Significance of AMR gene acquisitions by travel destination. The lines show the 95% confidence intervals, and the points show the estimates of binomial tests for bias. Binomial tests were conducted by region for the number of acquired AMR genes and the number of lost AMR genes. Both acquisitions and losses were normalized by the number of individuals traveling to the region. p values (FDR-corrected) from this test are shown just below the dotted line at 0.5 indicating the null. Numbers lower than 0.5 indicate AMR gene loss, and numbers greater than 0.5 indicate AMR gene gain. b Genes that showed significant region-specific bias following multinomial testing. Points indicate their number of acquisitions normalized by the number of travel subjects, and p values are given in the top left. c Sankey diagram of AMR gene acquisitions by travel region. Black nodes are when the gene was not found, and bright red nodes indicate the gene was present. The width of all lines is proportional to the number of individuals following that path. d, e The number of MGE elements detected from the functional metagenomic libraries is plotted on the y-axis, and the number of input d libraries and e reads is on the x-axis. p values calculated by the FDR-corrected multinomial test are in the bottom left of each panel. Most p values in b, d, and e hit underflow and have been set to p<0.001. Source data for all panels is provided in the source data file (Additional file 3)

Fig. 8

Quinolone resistance genes were acquired in regions with equal frequency, while β-lactam resistance genes had destination-specific acquisition. a AMR genes acquired or lost after travel detected by qPCR. Each point is an AMR gene. The x-axis is the number of individuals that had the gene in the pre-travel time point, but not in the post-travel time point. The y-axis is the number of individuals that had the gene in the post-travel time point, but not in the pre-travel time point. Significant acquired AMR genes are in red. The number of significant genes is tabulated in the top right. Non-significant genes are in black. The diagonal line is the null of equal losses and gains for an AMR gene. The inset panel gives the results from binomial tests of bias for AMR gene acquisition for the post-travel time point. The lines are 95% confidence intervals, and the points are estimates. p values (FDR-corrected binomial test) are given at the bottom of the plot for each gene. The dotted line is the expected value under the null. The lines and points are red if significantly acquired. b Sankey diagrams of significant gene acquisitions by travel region detected by qPCR. Black nodes are when the gene was not found, and bright red nodes indicate the gene was present. Ribbon colors correspond to the destination countries (dark blue is Northern Africa, light blue is Eastern Africa, orange is Southern Asia, and red is Southeastern Asia). The width of all lines is proportional to the number of individuals following that path. Source data for both panels is provided in the source data file (Additional file 3)

Fig. 9

mcr-1 containing contig from a Dutch traveler matched a plasmid sequenced from a gastroenteritis patient in the destination region. a Sankey network showing region-specific acquisition for mcr-1. b Map showing where the reference genome was isolated from a gastroenteritis patient. c Alignment between a plasmid from an E. coli isolated from a gastroenteritis patient in Chiang Mai and a contig assembled from a Dutch traveler’s gut microbiome. Source data for all panels is provided in the source data file (Additional file 3)