Transmission pathways of Campylobacter jejuni between humans and livestock in rural Ethiopia are highly complex and interdependent

Transmission pathways of Campylobacter jejuni between humans and livestock in rural Ethiopia are highly complex and interdependent | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Transmission pathways of Campylobacter jejuni between humans and livestock in rural Ethiopia are highly complex and interdependent Nitya Singh, Cecilie A.N. Thystrup, Bahar Mummed Hassen, Menuka Bhandari, and 9 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-5735672/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 03 May, 2025 Read the published version in Gut Pathogens → Version 1 posted 8 You are reading this latest preprint version Abstract Background Campylobacter jejuni and C. coli are the most common causes of bacterial enteritis worldwide whereas symptomatic and asymptomatic infections are associated with stunting in children in low- and middle-income countries. Little is known about their sources and transmission pathways in low- and middle-income countries, and particularly for infants and young children. We assessed the genomic diversity of C. jejuni in Eastern Ethiopia to determine the attribution of infections in infants under 1 year of age to livestock (chickens, cattle, goats and sheep) and other humans (siblings, mothers). Results Among 287 C. jejuni isolates, 48 seven-gene sequence types (STs), including 11 previously unreported STs were identified. Within an ST, the core genome STs of multiple isolates differed in fewer than five alleles. Many of these isolates do not belong to the most common STs reported in high-resource settings, and of the six most common global STs, only ST50 was found in our study area. Isolates from the same infant sample were closely related, while those from consecutive infant samples often displayed different STs, suggesting rapid clearance and new infection. Four different attribution models using different genomic profiling methods, assumptions and estimation methods predicted that chickens are the primary reservoir for infant infections. Infections from chickens are transmitted with or without other humans (mothers, siblings) as intermediate sources Model predictions differed in terms of the relative importance of cattle vs. small ruminants as additional sources. Conclusions The transmission pathways of C. jejuni in our study area are highly complex and interdependent. While chickens are the most important reservoir of C. jejuni , ruminant reservoirs also contribute to the infections. The currently nonculturable species Candidatus C. infans is also highly prevalent in infants and is likely anthroponotic. Efforts to reduce the colonization of infants with Campylobacter and ultimately stunting in low-resource settings are best aimed at protecting proximate sources such as caretakers’ hands, food and indoor soil through tight integration of the currently siloed domains of nutrition, food safety and water, sanitation and hygiene. Campylobacter jejuni Attribution Transmission Pathways Zoonosis Diversity Persistence Spatial Distribution Sequencing Typing Figures Figure 1 Figure 2 Figure 3 Figure 4 Background Campylobacter jejuni and C. coli are the most common causes of bacterial enteritis worldwide whereas symptomatic and asymptomatic infections are associated with stunting in children in low- and middle-income countries 1 . The vaccination of rhesus macaques against C. coli not only reduced the incidence of overt diarrhea but also improved their linear growth 2 . We previously reported a high prevalence of bacteria of the Campylobacter genus in humans and livestock in smallholder households in Haramaya woreda, East Hararghe Zone, Oromia State, Ethiopia 3 . The prevalence and load in infants increased significantly with age with logistic regression predicting a prevalence of approximately 90% at 1 year of age. Most infections are asymptomatic, but the bacterial load is positively correlated with the risk of diarrhea 4 . The load of Campylobacter in infant stools was greater in girls than in boys and increased with increasing food insecurity, different feeding practices (prelacteal feeding, early introduction of complementary foods, consumption of any solid foods, drinking of raw milk, household ownership of cattle and sheep but not chickens or goats), hygiene-related factors (improper disposal of infant stools, contact with animals or their feces, mouthing soil) and treatment with antibiotics in the previous month. Mothers’ handwashing with soap and drinking from bottles with nipples were associated with lower loads 4 , 5 . Two main species were found in infants by real-time polymerase cChain reaction (qPCR) and shotgun metagenomic sequencing: Candidatus C. infans ( C. infans , 60% at 1 year of age) and C. jejuni (50% at 1 year of age). C. upsaliensis was also detected but less frequently (20% at 1 year of age). C. coli was not detected in infants by shotgun sequencing and was detected infrequently (1%) by qPCR 6 , 7 . In high-income countries, animals, specifically livestock, are considered the primary reservoirs of human infections with the well-studied species C. jejuni/coli , with foodborne transmission 8 , mainly through poultry meat 8 as the main pathway. However, less is known about their sources and transmission pathways in low- and middle-income countries, particularly for infants and young children. The load of C. infans was greater in girls than in boys and was also elevated for infants who drank raw milk or crawled in areas contaminated with animal feces, whereas the load of C. jejuni was greater for infants who put soil in their mouths. At both the genus and species level, there were mixed and often counterintuitive signals related to keeping animals in the home, whether confined or not 4 , 6 . These results suggest a complex contamination network among humans, animals and their environment. Source attribution of C. jejuni/coli has largely been based on legacy (seven-gene) Multi Locus Sequence Typing (MLST) 9 , using frequency matching models such as the Dutch model 10 , Hald model 11 and variants 12 , or population genetic models such as the asymmetric Island model 13 or STRUCTURE 14 . More recently, attribution models based on whole-genome sequencing (WGS) data have been developed. WGS data enhance the ability to distinguish genetic variations and potentially more accurately determine the origin of infection-causing isolates 15 . Both core genome MLST (cgMLST) and k -merization have been used for taxonomic profiling and are particularly effective for large genomes 16 – 18 . K -mer counting involves the use of short oligonucleotides to compare a sequence to either a reference genome or against genome of interest without needing an alignment 19 . The use of cgSTs or k- mers for differentiating Campylobacter genomes is based on the concept of genomic signatures and builds on the premise that infections originating from the same source are genetically more similar than those from different sources, facilitating the tracking of infections across various sources. Random forest models may use cgMLST data following the numerical encoding of alleles 20 . Encoding genes with the PCO-encoding method 21 incorporates information that quantifies the similarity between each pair of alleles and addresses issues related to missing alleles and new genotypes in observations for prediction. All models typically assume unidirectional flow from sources to sinks. However, a model to include intermediate nodes, which may act as both a source and a sink was developed to explore the role of water in the transmission of bacteria from livestock and water birds 22 . This study aims to assess the genomic diversity of Campylobacter spp. in infants, humans and livestock in the Haramaya woreda, and to determine the attribution of Campylobacter infections in infants to livestock (chickens, cattle, goats and sheep) and other humans (siblings, mothers) on the basis of the genetic population structure of Campylobacter spp. circulating in these reservoirs, using four different attribution models. Given the observed species distribution in infants, and the fact that Candidatus C. infans is not yet culturable, this study focused on source attribution of C. jejuni . Methods Isolation of Campylobacter from human stool and livestock fecal samples Direct culturing was applied to approximately 15% (326/2,1833) of the total samples processed. For the isolation of thermotolerant species by direct plating, one gram of fresh stool/animal feces was suspended in 9 ml buffered peptone water (pH 7; BD Difco). 100 µl of homogenized samples were spread on CHROMagar Campylobacter (CaC, DRG International, Springfield, New Jersey USA) using sterile glass beads and incubated for 48 hours at 42°C in microaerobic condition (85% nitrogen, 10% carbon dioxide, 5% oxygen) in anaerobic jars with GasPak EZ Campy Container System Sachets (ThermoFisher Scientific, Waltham, MA, USA). Similarly, for non-thermotolerant species, the same volumes of samples (100 µl) were plated on Columbia agar supplemented with 5% defibrinated sheep blood, Skirrow supplement (2 µL/mL), amphotericin B (5µg/ mL), cefoperazone (8µg/mL) and Campylobacter growth supplement (ThermoFisher Scientific, Waltham, MA, USA). The plates were incubated at 37°C for 48 hours in microaerobic condition. In parallel, samples were also enriched in Preston and Bolton broth with a proportion of 1 g feces in 9 ml of broth and incubator at 42°C and 37°C for 48 hours, respectively, as described above. After incubation, 100 µl of Preston broth enriched samples were plated onto CaC and Bolton broth enriched samples on Colombia agar and plates were incubated at either 37 °C or 42 °C 23 . We could not directly culture almost 85% (1,857/2,183) of the fresh feces/stools due to global supply issues during the COVID-19 pandemic. We stored these samples in 20% (w/v) glycerol at -80 °C. Preliminary analysis indicated that up to 99% of the Campylobacter population in the feces could not be recovered on CaC within the first month of storage; therefore, samples were pre-enriched in Bolton broth before plating. Typical Campylobacter colonies (up to 5 per plate) were sub-cultured onto a CaC plate and confirmed by genus-specific qPCR 24 . Potential thermotolerant and non-thermotolerant Campylobacter were characterized by streaking the confirmed pure isolate on to two fresh CaC plates and incubating at 37°C and 42°C in microaerophilic conditions for 48 hrs. The isolates growing at 42°C and 37°C were recorded as potentially thermotolerant while the isolates growing only at 37°C were recorded as potentially non-thermotolerant. Despite our efforts to isolate non-thermotolerant Campylobacter species, we were only able to culture one non-thermotolerant isolate, later determined to be C. hyointestinalis . All isolates were stored in glycerol at -80 °C. For genomic DNA extraction, Campylobacter isolates from the freezer stock were grown on a CaC agar plate for approximately 36 hours under microaerophilic condition at 42 ℃. A loopful of growth was collected from the CaC plate, resuspended in 1 ml of Mueller Hinton broth (ThermoFisher Scientific, Waltham, MA, USA) and genomic DNA was extracted using Promega Wizard genomic DNA purification kit (Promega, Madison, WI, United States) following the manufacturer’s instructions. The concentration and quality of the DNA were determined using NanoDrop 2000 C Spectrophotometer (ThermoFisher Scientific, MA, USA). Purified DNA was shipped to eight GenomeTrakr Laboratories (FDA, USA) for sequencing. Short-read genomic DNA libraries were prepared with the Illumina DNA prep kit, following the PulseNet Sequencing Protocol PNL35 25 . Samples were sequenced using either paired-end 2x150 bp or 2x250 bp reads, which vary between sequencing laboratories (see Supplementary file MLST_profiles.xlsx for Biosample IDs). The paired-end reads were assessed for quality and contamination and trimmed using BBMerge (v.38.90) 26 and BBDuk (v.38.90) 27 with the following parameters: hammering distance 1, optimal k -mers 23, quality cutoff Q14 and minimum read length 30 bp, with end-trimming of a maximum 1of 0 bp. Species assignment was performed using KMC (version 3.0) 28 resulting in the identification of 380 Campylobacter jejuni isolates for analysis in this study. MLST assignment Legacy MLST profiles (STs) for all C. jejuni isolates were determined using the mlst tool 29 , which utilizes the most recent update of the PubMLST database 30 (updated December 14, 2024), which incorporates the seven housekeeping loci scheme, as previously described 31 . We identified 11 novel STs and submitted these new schemes to PubMLST for the assignment of new STs. Core genome MLST (cgMLST) profiles were assigned using a 1,343-loci scheme (Cody et al., 2017), implemented through the cgMLST tool 32 . Missing alleles, which were unassigned, were identified using a custom R script and were assigned unique identifiers within the dataset. This approach ensured that all 1,343 loci were included in the analysis. The R script is publicly available at https://github.com/jmarshallnz/cgmlst. The sample set of 380 WGS samples included sequences collected from the same household and time point, representing different colonies obtained during subculturing in the pure isolation process. To create a unique representative dataset, we selected the isolates with the highest genomic coverage for each cgST type from each household and time point. This filtering resulted in a final WGS dataset comprising 287 isolates. MS tree construction and map A minimum spanning tree (MST) was constructed using the GrapeTree plugin with the MSTreeV2 algorithm, which is designed to handle missing data more effectively than classical MST methods do 33 . The process begins by calculating a directed minimal spanning arborescence using Edmonds' algorithm from asymmetric distances, with tie-breaking based on allelic distances. Local branch recrafting was then performed to remove spurious branches. PERMANOVA To explore the transmission of C. jejuni at different levels within the sampling hierarchy we performed a nested, permutational multivariate analysis of variance (PERMANOVA) 34 to estimate the proportion of variance in infant cgST profiles attributable to each level, namely sample (i.e., multiple isolates from the same infant sample), infant (i.e., multiple isolates from the same infant at different time points), ganda (village) and kebele (the smallest administrative unit in Ethiopia, a set of gandas). PERMANOVA models were constructed using a customised R script (https://github.com/jmarshallnz/permanova). Pairwise genetic distances were calculated from the cgST profiles to create a distance matrix with values in the matrix corresponding to the Gower distance calculated using the vegdist() function in the R package vegan 35 . Multiple two-level nested models were considered: infant within ganda , ganda within kebele , infant within kebele and infant time point within infant . It was not possible to fit models considering higher-level nested structures, so only two-level nested models are presented. Univariate PERMANOVA models were performed for each factor with p-value s obtained using 100000 unrestricted permutations of raw data. Diversity and persistence of C. jejuni infections The isolate set included up to four isolates from the same infant sample, while 25 sets of repeat samples at different timepoints (approximately monthly intervals) were also available. We quantified the diversity of C. jejuni isolates in these samples by counting the number of isolates by ST/cgST using a bespoke coding system in Excel (Supplementary file MLST_profiles.xlsx). Similarly, the occurrence of the same or different STs/cgSTs in sequential samples from the same infant was quantified and summarized in relation to the time interval between two samples using pivot tables in Excel. Attribution While livestock are commonly recognized as the main sources of human infections, humans other than infants are exposed to the same contaminated environment as the infants are, and they can be considered either as independent receiving hosts or as receiving, amplifying hosts that transmit infection to infants. Exclusive human-to-human transmission cycles cannot be excluded a priori . We therefore fitted models with only livestock sources as well as models with other humans as additional sources of infections in infants. A summary of all fitted models is provided in Table 1. All analyses were performed in the statistical language R version 4.3.0 or later 36 or Excel (Microsoft Corporation, Redmond, WA), using dedicated R packages or other software as indicated in the text. Table 1. Overview of source attribution models Source attribution model Input data Summary data Attribution pathways Model fitting procedure Asymmetric Island Assembled contigs cgST * profiles Direct MCMC & k -mer Raw reads 9-mers Direct Random Forest with feature reduction PCO Assembled contigs cgST profiles + allele sequences Direct Random Forest without feature reduction Source-sink Assembled contigs cgST profiles Direct and indirect MCMC * core genome Multi Locus Sequence Typing & Markov Chain Monte Carlo Asymmetric Island model The asymmetric-island source attribution model 13 was applied to the cgMLST data for 287 isolates using the islandR package (https://github.com/jmarshallnz/islandR). This method assigns sources to isolates by modeling both recombination and mutation processes as distinct events. In our analysis, the recombination and mutation probabilities were assumed to be constant across all sources, resulting in pooled estimates for both processes. The model incorporates genetic differences between isolates to estimate the most likely source, accounting for multiple potential sources. From the model estimates, the mean and 95% confidence intervals of the posterior distribution were calculated to determine the attribution percentages for each source. This approach is particularly useful for including recombination and mutation in source attribution, enhancing the accuracy of source estimation based on cgST profiles. PCO model The PCO source attribution model is a random forest model that uses a principal coordinates approach to overcome the problem of missing levels in the data used for prediction. This method uses a target-agnostic approach to encode cgMLST predictor variables by using both the cgMLST allele profiles and Hamming distances between allele sequences to determine the similarity between pairs of isolates 21 . Analyses were carried out using the ranger package 37 and the PCO-encoding method (https://github.com/smithhelen/LostInTheForest). Estimates of uncertainty are calculated using a probability forest with the same set of parameters as the original random forest. For each tree in the forest, the probability of each human isolate being attributed to each source is calculated. These probabilities are then averaged over the set of human isolates, giving an average probability of attribution to each source for each tree. The 2.5% and 97.5% quantiles are then determined from this set of mean probabilities to give a 95% uncertainty interval. Source/sink model The role of humans as intermediate hosts for zoonotic infections was further explored using a model in which they can act both as sources and as a sink. The IslandR model was reparametrized to consider the mothers and siblings as both receiving infection from the animal reservoirs and being a source of infection for infants. The source/sink model was based on an approach developed to examine the contribution of water as both a source and a sink for human campylobacteriosis in New Zealand 22 . In essence, we assume that mother and sibling isolates arise from a mix of the animal reservoirs, whereas infant isolates arise from a mix of both the animal reservoirs and mothers and siblings. This model is fit using the attribution_intermediate() function in the islandR package (https://github.com/jmarshallnz/islandR). k-mer model The tool KMC (version 3.0) 28 was used to extract k- mers with length of k = 9 for each of the samples using the short-read sequences. All k -mer frequencies were then combined into one matrix using an in-house Python script. A recently developed 20 source attribution model was applied to the two datasets using the k -mers to predict the sources of human campylobacteriosis cases. Feature reduction was carried out on the matrix to reduce the number of k -mers in the final model using the caret package (version 6.0-94) 38 and the Boruta package (version 8.0.0) 39 . The near-zero-variance method was used to reduce the number of 9-mers. The Boruta algorithm was then applied to select important attributes in the matrix using a random forest classifier. To account for the uneven distribution of sources in the samples, all sources were upsampled to the highest number of samples available within a source, so that all sources had the same number of samples. Two machine-learning algorithms previously applied successfully in sequencing studies were evaluated 40–43 . For the evaluation, the data containing k -mers from sources were split into test- and training data sets. The training data were then used to randomly generate smaller sets of test and training data once again to determine which of the two selected machine-learning algorithms fit the data best. Each smaller test- and training data set was split 70% and 30%, respectively, and the test-data were used to evaluate the performance of the model using seven-fold cross-validation. After 10 iterations, the accuracy of each algorithm was assessed, and the algorithm with the highest accuracy was selected for model construction. The model with the highest accuracy was constructed again following the same steps as for the model selection, described previously, and the performance of the model was evaluated based on the accuracy of the cross-validation step, the kappa value and the confusion matrix, which determines the model’s ability to predict the sources of the samples in the source-data. The sensitivity and specificity were also reported. Finally, the model was applied to isolates from infants. This was done by estimating the probability of each human case being attributed to each of the sources included in the model. Tree-level predictions were pooled together to calculate the mean probability for each source, with 2.5 th and 97.5 th percentiles providing uncertainty intervals for each case. To estimate the uncertainty in the overall mean attribution probabilities, we performed 1,000 bootstrap resampling iterations. In each iteration, case isolates were sampled with replacement, and mean attribution probabilities were recalculated. The 2.5 th and 97.5 th percentiles of the bootstrap distributions provided 95% uncertainty intervals for the mean probabilities. Results We included WGS data from 380 isolates for attribution from different sources analysis by k -mers (Table 1) and of these, assigned cgST profiles to 287 isolates. The majority of human samples other than those from infants were collected from siblings and the majority of livestock samples were collected from chickens with fewer samples from ruminant species. Owing to the small number of isolates from several sources, we ran attribution models of infections in infants with two merged source groups: other humans (mothers and siblings) and small ruminants (sheep and goats). Cattle are often recognized as a major reservoir for Campylobacter transmission to humans, particularly C. jejuni . By keeping cattle separate, their unique role as a distinct source is emphasized. Additionally, other studies and risk assessments use the "small ruminant" classification for sheep and goats because of their shared characteristics. This convention supports comparability and consistency across studies. Table 2. C. jejuni isolates included in the k-mer and cgMLST attribution models Source group Source k-mer data set Grouped k-mer data set cgMLST data set Grouped cgMLST data set Infants 229 229 174 174 Other Humans Mothers 3 29 3 20 Siblings 26 17 Chickens 86 86 71 71 Cattle 13 13 6 6 Small Ruminants Sheep 2 24 4 16 Goats 22 12 Total 380 380 287 287 Genomic diversity of C. jejuni from different sources The set of 287 isolates included 48 STs. The population structure based on cgMLST is presented as a minimum spanning tree (MSTree) in Figure 1. The tree is fully structured according to seven-gene STs and isolates within the same ST are highly related, differing by fewer than five alleles. There were 11 newly assigned STs (37 isolates; see Table S1). Among these, one type was common to infants, chickens, cattle and sheep; one was common to infants and chickens; six were unique to infants; and three were unique to chickens (Table S2). Twelve STs were represented by more than 10 isolates, which together accounted for almost two-thirds (185/287) of all the isolates. Isolates from six of these STs were shared among infants, livestock and other humans, whereas six were shared only between infants and livestock. Detailed data on these isolates are available in the Supplementary file MLST_profiles.xlsx. Diversity and persistence of C. jejuni infections We obtained multiple isolates of C. jejuni from 81 samples. Of these, one sample yielded four isolates, 34 (28+5+1) samples yielded three isolates, and 46 (*39+7) samples yielded two isolates. All isolates from the same sample were of the same seven-gene ST in 84% (68/71) of the samples (Table 3 and supplementary file MLST_profiles.xlsx), thus not providing evidence of the diversity of STs within the host. Two or three different STs were found in 9 samples, suggesting sequence type diversity within the host. The cgST profiles of all the isolates from the same sample within the same ST were fewer than five alleles different. Table 3. Multi-Locus Sequence Type diversity in human and livestock samples Isolate pattern * Infant Sibling Mother Cattle Goat Chicken Sheep Total A1234 & 1 0 0 0 0 0 0 1 A123 & 25 1 1 0 0 1 0 28 A12 & 22 5 0 2 2 7 1 39 A12B # 5 0 0 0 0 0 0 5 AB # 3 0 0 0 0 4 0 7 ABC # 1 0 0 0 0 0 0 1 Evidence of sequence type diversity No & 48 6 1 2 2 8 1 68 Yes # 9 0 0 0 0 4 0 13 Total 57 6 1 2 2 12 81 * Letters indicate different seven-gene STs, and numbers indicate different cgSTs among isolates from one ST. The number of isolates differs between samples and is represented by the number of different letter/number combinations. For example, A1234 indicates four isolates from one sample with the same ST, but all different cgSTs, whereas A12B indicates three isolates from one sample of which two with the same ST but different cgSTs, and one isolate with a different ST (and consequently also a different cgST). Subscripts indicate patterns with # or without & evidence of sequence type diversity. Persistence of C. jejuni infections (i.e., isolation of the same ST from two sequential samples) was observed in 8% (2/25) of sample pairs from the same infant (supplementary file MLST_profiles.xlsx). These pairs were taken approximately 1 or 2 months apart. Different STs were observed in 84% of the sample pairs, the majority of these STs were also 1-2 months apart, but five pairs had 3-months intervals and 1 pair had a 4-month interval. Eight percent (2/25) of sample pairs (1- or 2-month intervals) provided inconclusive evidence with the same STs being isolated in both samples, accompanied by one or more different STs. Spatial clustering Figure 2 shows the spatial distribution of STs from infants. Common STs, such as ST50, ST883 and ST2042 were found in multiple gandas (villages) and kebeles (the smallest administrative unit in Ethiopia). We analyzed the distribution of cgST types at different nested levels of spatial and temporal sampling using PERMANOVA, estimating the contribution of the variation in cgMLST alleles attributable to kebeles, gandas, infants at different time points and infant samples. We tested for significant clustering at each level, using multiple two-level nested models (Table S3). Considering infants within kebeles , some 14% of the total variation in cgST profiles was between kebeles. Most of the total variation was between infants within kebeles (53%), and 34% of the variation was within infants . There was marginally significant clustering at the kebele level (p = 0.03) and highly significant clustering at the infant level (p < 0.001). Considering gandas within kebeles , some 14% of the total variation in cgST profiles was between kebeles. Most of the total variation was between households within kebeles (46%), and 40% of the variation was within households . There was a tendency toward clustering at the kebele level (p =0.053) and highly significant clustering at the ganda level (p < 0.001). Considering infants within gandas , some 58% of the total variation in cgST profiles was between gandas. The total variation between infants within gandas was 9%, and 33% of the variation was within infants . There was no significant clustering at the ganda level (p = 0.29) but there was highly significant clustering at the infant level (p < 0.001). Considering samples within infants , some 68% of the total variation in cgST profiles was between infants. The total variation between samples within infants was 22%, and 10% of the variation was within infants . There was significant clustering at the infant level (p < 0.01) and highly significant clustering at the sample level (p < 0.001). The highly significant clustering at the sample level confirms observations in Table 2 of highly related isolates from one sample. We, therefore, repeated the analysis, using a reduced dataset including only unique STs per sample (97 isolates). The results of the partitioning of variance at all levels of clustering were very similar to those of the full dataset. However, the significance of clustering at the infants within gandas changed markedly to no significant clustering at both the ganda and infant levels. Attribution The attribution results for the Asymmetric Island, PCO and k -mer models are summarized in Figure 3 and Table 4. The PCO random forest model was trained on the set of source isolates using all 1343 cgMLST genes (as nominal predictors). The genes were encoded using the PCO-encoding method together with a dissimilarity matrix of Hamming distances of the nucleotide sequencing information between each pair of alleles for each gene. Any new alleles in the set of human isolates for prediction were encoded using the method of principal coordinates based on pairwise Hamming distance from the new alleles to the alleles in the set of source isolates. The original sources of the infant isolates were then predicted. The k -mer models were built on 131,073 attributes (or 9-mers), which were then reduced further by the near-zero variance method which removed few attributes from the model, depending on which dataset was modeled. The Boruta algorithm further reduced the matrix by selecting only those that are confirmed to provide enough important information to be included in the model. For the dataset including both humans and animals as sources, the near-zero variance method identified two attributes with low variance, which were removed from the data set. The Boruta algorithm further reduced the dataset to include 69 important attributes used for further modeling. For the dataset including only animal sources, the near zero variance method removed one attribute with low variance, whereas the Boruta algorithm further reduced the number of attributes to 23 used for further modeling. For both datasets, the performances of the random forest and the logit-boost algorithms were compared (Table S4). The average accuracies obtained from taking the average across ten iterations showed that the random forest and the logit boost algorithms performed very similarly in terms of accuracy for both data sets. The logit boost performed marginally better but could not provide uncertainty intervals comparable to the other approaches. Consequently, we decided to use the random forests algorithm. The final models predicted probabilities for each of the C. jejuni infections of infants to originate from each of the sources (Table 4). The results were similar for all models with most cases being attributed to chickens. The asymmetric Island model attributed most infections among ruminant sources to cattle, whereas the PCO and k -mer models (both machine-learning models) considered small ruminants more likely. The percentage of cases attributed to chickens was lower for the k -mer model than for the other two models, both of which use cgMLST for genomic characterization. Although both the k -mer and the PCO models are random forest models, the k -mer method uses feature reduction to substantially decrease the number of variables in the model. The smaller pool of predictors means there is less variation in each tree of the forest, which potentially explains the smaller uncertainty intervals When other humans are included as putative sources, the models estimate that some 16-22% of all infant isolates originate from other humans, reducing mainly the estimate for chickens. Table 4: Attribution estimates for C. jejuni infections of infants to livestock and human sources using Asymmetric Island, PCO and k -mer models. Models and source categories Attribution percentage Chickens Cattle Small Ruminants (goats and sheep) Other Humans (mothers and siblings) Asymmetric Island model Livestock and humans 60.6 * (44.4-74.6) 15.2 (4.7-28.5) 6.9 (0.1-14.5) 17.3 (10.0-26.4) Livestock 78.1 (61.6-90.1) 17.8 (0.1-30.3) 3.6 (0.0-16.3) - PCO model Livestock and humans 60.2 (47.9-74.0) 4.6 (0.1-10.7) 16.5 (0.1-29.6) 16.3 (0.1-28.3) Livestock 72.7 (59.4-86.0) 4.9 (0.1-13.5) 21.9 (0.1-35.1) - k -mer model Livestock and humans 52.3 (48.8-55.9) 8.4 (7.5-9.4) 17.2 (15.4-19.0) 22.0 (19.1-25.0) Livestock 62.8 (59.5-65.9) 12.1 (10.4-14.0) 25.0 (22.4-27.9) - * Mean (95% uncertainty interval) The source/sink model was run for different source combinations (Figure 4 and Table 5. When considering infections of infants from all putative sources independently, the results were similar to those of the Island model, as expected. When considering infections in mothers and siblings from livestock sources, the attribution was almost exclusively to chickens with very low percentages attributed to cattle or small ruminants. Likewise, if mothers and siblings were considered as intermediate sources between livestock and infants, chickens were identified as the main source which then gets passed on to infants. This results in a lack of identifiability for the attribution to infants when aiming to separate the direct chicken route from the indirect chicken route through mothers and siblings (Figure 4D). Table 5: Attribution estimates for C. jejuni infections of infants to livestock and human sources using the source-sink model. Models and source categories Attribution percentage Chickens Cattle Small Ruminants (goats and sheep) Other Humans (mothers and siblings) To infants from all sources, no intermediate source Livestock and humans 58.4 * (42.6-73.5) 17.2 (3.9-31.1) 7.0 (1.1-14.4) 17.4 (10.3-26.3) To mothers and siblings from livestock, ignoring infants Livestock 92.2 (67.4-99.9) 2.9 (0.0-18.5) 5.1 (0.0-25.1) - To infants from livestock via mother and siblings Livestock 92.2 (69.3-99.9) 3.3 (0.0-20.1) 4.5 (0.0-21.6) - To infants from all sources, mothers and siblings as intermediate sources Livestock and humans 50.9 (0.3-88.0) 16.2 (2.6-29.0) 2.8 (0.0-14.7) 30.0 (0.0-86.8) * Mean (95% uncertainty interval) Discussion In the study region, 9 out of 10 infants were colonized by one or more Campylobacter species by the age of one year of which 6 out of 10 were colonized by C. jejuni 3 , 6 . Most infections were asymptomatic, although the odds of diarrhea increased with increasing bacterial load 6 . We used whole-genome-sequencing data to analyze the genomic diversity of C. jejuni , and to attribute infections to putative livestock and human sources. Among the 287 isolates, 48 STs were identified, among which 11 were previously unreported types. In a meta-analysis of global genotypes of C. jejuni , Poorrashidi et al. 44 reported that the six most common STs globally are ST21, ST45, ST50, ST48, and ST257. Among these STs, only ST50 was found in our study as the most frequent ST. Many of the isolates recovered in this study do not belong to the most common STs reported in high-resource settings, and many have never been reported before. This is likely to be due to sampling bias and limited application of WGS in low-resource settings 45 . Of the 87,542 genomes of Campylobacter in the PubMLST database, only 386 are from Africa ( https://pubmlst.org/bigsdb?db=pubmlst_campylobacter_isolates&page=query&genomes=1 , accessed December 12, 2024). C. jejuni ST50 isolates have been frequently reported globally from environmental, food and clinical sources. Poultry isolates from Oceania, Europe and North America tended to cluster on the basis of the continent where the sample was collected 46 . No isolates from Africa were included in this study. Other frequent STs in our study are ST883, ST19, ST20242 and ST2031. The two former STs were found in infants and goats in our study, have previously been found in humans (children and mothers) in Ethiopia 47 and have been associated with a raw cow’s milk outbreak in Finland (ST883) 48 and Denmark (ST19) 49 , respectively. Several studies have detected a high prevalence of C. jejuni in human, livestock and food samples from Ethiopia, but only two studies have included the genomic characterization of isolates. Among 14 isolates from dairy in the Ethiopian regions of Amhara, Oromia, and the Southern Nations, Nationalities, and Peoples (SNNPR) between 2020 and 2021, two STs (ST51 and ST2084) were detected 50 ; neither of these types was identified in our study. Another study from the Harar town and Kersa district identified 8 distinct STs for 19 Campylobacter strains isolated from children, caretakers and potential exposure sources 47 . The most abundant STs were ST353, ST19 and ST1365, all of which were also found in this study. French et al. 51 have studied the genomic structure of poultry isolates from Tanzania and human isolates from Kenya, and reported that ST353, ST8043, ST2122 and ST1932 were the dominant STs (in this order). Of these, only ST353 was identified in the current study as one of the twelve most frequently occurring types. In both studies, these types were isolated from both humans and livestock. These authors also tabulated STs as identified in other studies from Africa. Among the twelve most common STs identified in our study, ST362 was identified frequently in South Africa, whereas some other STs occurred sporadically in the database. While the current results suggest little overlap between types in geographically adjacent regions, the genomic diversity of C. jejuni and other foodborne pathogens in Africa is largely uncharacterized. More generally, WGS-based surveillance is poorly developed in low-resource countries 45 . Our isolate collection included several samples from which multiple isolates were obtained, offering a unique opportunity to assess genomic diversity in single hosts. Among stool samples from infants, we detected up to four isolates with the same seven-gene ST and only a few samples with more than one ST. Within one ST, cgSTs had fewer than five alleles difference. Bloomfield et al. 52 studied sixteen isolates from an immunodeficient patient who was colonized by C. jejuni for 10 years after an episode of diarrhea. All isolates shared a common ancestor, coinciding with the onset of symptoms for the patient and evidence was found for genetic bottlenecks due to antimicrobial treatment. Djeghout et al. 53 sequenced ninety-two C. jejuni isolates from four different patients with gastroenteritis and used Single Nucleotide Polymorphisms (SNP) to assess phylogenetic relationships. Three patients yielded a single seven-gene ST, whereas one patient yielded two different STs. Isolates from one patient were genetically diverse, even within one ST (12–43 core non-recombinant SNPs and 0–20 frame-shifts). These authors concluded that this diverse population was unlikely to have evolved from a single isolate at the time point of initial patient infection and that patients were likely infected with a heterogeneous C. jejuni population. However, even upon exposure to a heterogeneous population, multiple barriers in the host create a strong evolutionary bottleneck, resulting in the selection of one single variant causing colonization except when the host is exposed to high doses 54 . We therefore suggest that in our population of young infants, infections are caused mainly by a single transmission event, followed by microevolution within the host. Further phylogenetic analysis of our isolates including Average Nucleotide Identity and Single Nucleotide Polymorphisms may provide more detailed information but was beyond the scope of this study. A reanalysis of data from the MAL-ED birth cohort study suggested that persistent infections with Campylobacter were associated with poorer 9-month linear growth. Persistent infections were defined as three or more consecutive Campylobacter positive monthly stools by qPCR 55 . However, in the main text the authors state that “persistent Campylobacter infections cannot be differentiated from recurrent reinfections with epidemiologic data alone” and suggest the term persistent carriage would better describe the qPCR results. Because of the longitudinal nature of our study, we were able to evaluate temporal patterns of genotype diversity in infants. Repeat isolates of the same ST at two different time points (one or two months apart) were observed but in most cases, we observed different STs. This suggests a dominant pattern of clearance of one type and new infection by other types within a time interval of one to four months, which is consistent with findings from a Markov Chain model applied to the same MAL-ED dataset 56 . These findings are further supported by PERMANOVA, indicating a high level of clustering of isolates from the same sample, but no significant clustering of isolates from the same infant at different time points. The clustering at the kebele level suggests that there is localized, within kebele transmission which is independent of the clustering observed within samples and between samples from the same individual over time. We used two different methods (cgMLST and k -merization) to characterize the genomic diversity of our isolates and two types of models (population genetics and machine learning) to attribute infections in infants to putative sources. Chickens were the main source of infection with model uncertainty in the proportion of infections attributed to cattle or small ruminants. When other humans are included as possible sources, attribution to livestock decreases but chickens remain the main source. Neither the Asymmetric Island method using cgMLST nor the random forest methods using cgMLST or k -mers can determine directionality. The assumption is that there is unidirectional flow from any of the included sources to the receiving host (sink). The “source-sink” model relaxes this assumption in the sense that some compartments are considered both as sources and as sinks. This model was originally applied to evaluate the role of water as a sink for contamination by water birds and as a source for human infections 22 . We applied this model to study the role of other humans as sinks from infections from livestock and as sources for infections in infants. The model suggested that other humans are largely infected by chickens, and that these infections may be transmitted to infants. We cannot conclude whether other humans act as mechanical vectors or if they are amplifying hosts. Our source attribution results may be biased towards chickens because of the substantially greater number of isolates than those from other livestock and human sources. For example, isolates from goats and siblings frequently occurred among the top twelve STs but less frequently among rare STs. Nevertheless, the use of high-resolution genome sequencing data, and the similar findings from multiple models based on different underlying assumptions, suggest that any bias attributed to unbalanced sampling is unlikely to affect the conclusion that chickens are the most important animal reservoir for infant infections in this study. Nevertheless, as a more important role of ruminants cannot be ruled out, control strategies aimed at eliminating transmission from the chicken reservoir only may not be effective. Additionally, the degree of transmission between chickens and ruminants is unknown but is likely to occur, particularly in areas where the animals are not confined to barns or other housing facilities. Notable, most of the isolates were from older infants, because of increasing prevalence with age. Furthermore, this study was undertaken during the global COVID-19 pandemic and early samples could not be cultured because of global supply chain issues and samples were preserved by freezing. Even though we stored the samples in glycerol, recovery was strongly affected. Additionally, species-specific qPCR results indicated that Candidatus C. infans is more prevalent in infants than C. jejuni is (70% at one year of age). This species has mainly been detected in other humans, with low levels of detection in livestock 6 , suggesting this species is anthroponotic in nature with occurrence in livestock as a reverse zoonosis, or even merely passing through the animal gut from a highly contaminated environment, where open defecation is common 4 . The transmission pathways of C. jejuni in our study area are highly complex and interdependent. The EXCAM study employed behavioral observations, microbiological analysis and mathematical modeling to create an agent-based exposure model framework to quantify the exposure to generic Escherichia coli through different pathways in the first and second half years of life of the infants included in this study. The major sources of exposure to E. coli were food and breastfeeding in the first half year of life and food and soil in the second half year of life. Caretakers’ hands are the main sources of contamination of both food and breast surfaces 57 . Considering all the evidence, we conclude that efforts to reduce the colonization of infants with Campylobacter in the study area, and most likely in many other similar settings, are best aimed at protecting proximate sources such as caretakers’ hands, food and indoor soil. Contributing to the ultimate goal of reducing stunting through controlling exposure of infants and young children to enteric pathogens calls for a tight integration of currently siloed domains of nutrition, food safety and water, sanitation and hygiene. Conclusions Many of the C. jejuni isolates identified in this study do not belong to the most common STs reported in high-resource settings. Among the six most common global STs, only one was found in our study area. Isolates from the same infant sample were highly related, isolates from consecutive infant samples usually had different STs, suggesting rapid clearance and new infection. The transmission pathways of C. jejuni in our study area are highly complex and interdependent. While chickens are the most important reservoir of C. jejuni in infants, ruminant reservoirs also contribute to the infections. Model predictions differed in terms of the relative importance of cattle vs. small ruminants as additional sources. Infections from chickens are transmitted with or without other humans (mothers, siblings) as intermediate sources. To reduce the colonization of infants with Campylobacter and ultimately mitigate stunting in low-resource settings, the best approach is to protectproximate sources. This includes ensuring the cleanliness of caretakers’ hands food and indoor soil through tight integration of the currently siloed domains of nutrition, food safety and water, sanitation and hygiene Abbreviations CaC Chromagar Campylobacter cgMLST Core genome Multi Locus Sequence Typing cgST Core genome Sequence Type k -mer Substring of length k contained within a DNA sequence PCO Ordinal encoding of predictor variables using Principal Components MST Minimum Spanning Tree PERMANOVA Nested, permutational multivariate analysis of variance PubMLST Public databases for molecular typing and microbial genome diversity qPCR Quantitative (real-time) Polymerase Chain Reaction SNP Single Nucleotide Polymorphism ST Legacy (seven-gene) Multi Locus Sequence Type Declarations Ethical approval and consent to participate Ethical approval was obtained from the University of Florida Internal Review Board (IRB201903141) and the Ethiopia National Research Ethics Review Committee (SM/14.1/1059/20). Written informed consent was obtained from all participating households (husband and wife), with mothers consenting to infant participation, using forms in the local language (Afan Oromo). Consent for publication Not applicable Availability of data All whole-genome sequences supporting the conclusions of this article are available under Bioproject PRJNA1015272 at https://www.ncbi.nlm.nih.gov/bioproject/?term=PRJNA1015272. Biosample IDs for selected Campylobacter jejuni isolates, post-processing, are provided in the supplementary file MLST_profiles.xlsx. Code on GitHub is indicated in the text. Competing interests The authors declare no competing interests. Funding This project was funded by the United States Agency for International Development Bureau for Food Security under Agreement #AID-OAA-L-15-00003 as part of Feed the Future Innovation Lab for Livestock Systems and by the Bill & Melinda Gates Foundation OPP#1175487. Under the grant conditions of the Foundation, a Creative Commons Attribution 4.0 Generic License has already been assigned to the Author Accepted Manuscript version that might arise from this submission. Any opinions, findings, conclusions, or recommendations expressed here are those of the authors alone. Author contributions Conceptualization: NPF, JYH, AHH, MJM, SLM, GR. Data curation: LD, BMH, AO, NS, CANT. Formal Analysis: NS, CANT, NPF, JCM, HLS, AHH. Funding acquisition: TH, JYH, AHH, SLM, GR. Investigation: BHM. Methodology: NPF, TMH, JCM, HLS. Project administration: JYH, AHH, SLM. Resources: NPF, TMH, JYH, AHH, GR. Software: JCM, HLS. Supervision: NPF, TMH, JYH, AHH, SLM, GR. Validation: NPF, TMH, AHH. Visualization: NPF, AHH, NS. Writing–original draft: NPF, TMH, AHH, CANT, NS. Writing–review & editing: All. All authors approved the final version for submission. Acknowledgments The CAGED Research Team Members consisted of Abadir Jemal Seran, Abdulmuen Mohammed Ibrahim, Amanda E. Ojeda, Bahar Mummed Hassen, Belisa Usmael Ahmedo, Cyrus Saleem, Dehao Chen, Efrah Ali Yusuf, Getnet Yimer, Ibsa Abdusemed Ahmed, Ibsa Aliyi Usmane, Jafer Kedir Amin, Kedir Abdi Hassen, Kedir Teji Roba, Kunuza Adem Umer, Karah Mechlowitz, Loic Deblais, Mahammad Mahammad Usmail, Mark J. Manary, Mawardi M. Dawid, Mussie Bhrane, Nur Shaikh, Wondwossen Gebreyes, Xiaolong Li, Yang Yang, Yenenesh Demisie Weldesenbet, Zelalem Hailu Mekuria. We thank Tina Lusk Pfeffer, Kelli Hiett, Kannan Balan, Hyein Jang, Marianne Sawyer (Office of Applied Research and Safety Assessment), and Ruth Timme (Office of Regulatory Science, Division of Microbiology), Center for Food Safety and Applied Nutrition, Food and Drug Administration, USA) for their support in acquiring funding and project management for Whole Genome Sequencing. We acknowledge the sequencing efforts by the GenomeTrakr laboratories at the Colorado State Department of Public Health and Environment , New Hampshire Department of Health and Human Services / Public Health Laboratories, Virginia Division of Consolidated Laboratory Services, Minnesota Department of Health, Missouri State Public Health Laboratory, Maryland Department of Health, Ohio Department of Agriculture /Animal Disease Diagnostic Laboratory, Rhode Island Department of Health. This publication made use of the PubMLST website (https://pubmlst.org/) developed by Keith Jolley and cited at the University of Oxford. The development of that website was funded by the Wellcome Trust. We thank Xiaolong Li for assistance with mapping the sequence types. Map data copyrighted OpenStreetMap contributors and available from https://www.openstreetmap.org. References Rogawski, E. T. et al. Use of quantitative molecular diagnostic methods to investigate the effect of enteropathogen infections on linear growth in children in low-resource settings: longitudinal analysis of results from the MAL-ED cohort study. The Lancet Global Health 6 , e1319–e1328 (2018). Hendrickson, S. M. et al. Campylobacter vaccination reduces diarrheal disease and infant growth stunting among rhesus macaques. Nat Commun 14 , 3806 (2023). Deblais, L. et al. Prevalence and Load of the Campylobacter Genus in Infants and Associated Household Contacts in Rural Eastern Ethiopia: a Longitudinal Study from the Campylobacter Genomics and Environmental Enteric Dysfunction (CAGED) Project. Applied and Environmental Microbiology 89 , e00424-23 (2023). Chen, D. et al. Campylobacter colonization and undernutrition in infants in rural Eastern Ethiopia: a longitudinal community-based birth cohort study. medRxiv (2024) doi:10.1101/2024.05.21.24307707. Li, X. et al. Geospatial Analysis of Multilevel Socio-environmental Factors Impacting the Campylobacter Burden Among Infants in Rural Eastern Ethiopia: A One Health Perspective. 2024.07.03.24309853 Preprint at https://doi.org/10.1101/2024.07.03.24309853 (2024). Ojeda, A. et al. Determinants of Campylobacter species diversity in infants and association with family members, livestock, and household environments in rural Eastern Ethiopia. Submitted to Gut Pathogens . Mekuria, Z. et al. Host Clustering of Campylobacter Species and Other Enteric Pathogens in a Longitudinal Cohort of Infants, Family Members and Livestock in Rural Eastern Ethiopia. Submitted for publication . Hald, T. et al. World Health Organization Estimates of the Relative Contributions of Food to the Burden of Disease Due to Selected Foodborne Hazards: A Structured Expert Elicitation. PLoS One 11 , e0145839 (2016). Cody, A. J., Maiden, M. C., Strachan, N. J. & McCarthy, N. D. A systematic review of source attribution of human campylobacteriosis using multilocus sequence typing. Eurosurveillance 24 , 1800696 (2019). Van Pelt, W. et al. Origin, extent and costs of human salmonellosis. Part 1 Origin of human salmonellosis with respect to pig, cattle, chicken, eggs and other sources (in Dutch). Infectieziekten Bulletin 10 ,. Hald, T., Vose, D., Wegener, H. C. & Koupeev, T. A Bayesian Approach to Quantify the Contribution of Animal-Food Sources to Human Salmonellosis. Risk Analysis 24 , 255–269 (2004). Mullner, P. et al. Source Attribution of Food-Borne Zoonoses in New Zealand: A Modified Hald Model. Risk Analysis 29 , 970–984 (2009). Wilson, D. J. et al. Tracing the Source of Campylobacteriosis. PLoS Genetics 4 , (2008). Pritchard, J. K., Stephens, M. & Donnelly, P. Inference of Population Structure Using Multilocus Genotype Data. Genetics 155 , 945–959 (2000). Arning, N., Sheppard, S. K., Bayliss, S., Clifton, D. A. & Wilson, D. J. Machine learning to predict the source of campylobacteriosis using whole genome data. PLoS Genet 17 , e1009436 (2021). Bernard, G., Chan, C. X. & Ragan, M. A. Alignment-free microbial phylogenomics under scenarios of sequence divergence, genome rearrangement and lateral genetic transfer. Sci Rep 6 , 28970 (2016). Panyukov, V. V., Kiselev, S. S. & Ozoline, O. N. Unique k-mers as Strain-Specific Barcodes for Phylogenetic Analysis and Natural Microbiome Profiling. International Journal of Molecular Sciences 21 , 944 (2020). Cody, A. J., Bray, J. E., Jolley, K. A., McCarthy, N. D. & Maiden, M. C. J. Core Genome Multilocus Sequence Typing Scheme for Stable, Comparative Analyses of Campylobacter jejuni and C. coli Human Disease Isolates. J Clin Microbiol 55 , 2086–2097 (2017). Zielezinski, A., Vinga, S., Almeida, J. & Karlowski, W. M. Alignment-free sequence comparison: benefits, applications, and tools. Genome Biology 18 , 186 (2017). Munck, N., Njage, P. M. K., Leekitcharoenphon, P., Litrup, E. & Hald, T. Application of Whole‐Genome Sequences and Machine Learning in Source Attribution of Salmonella Typhimurium. Risk Analysis 40 , 1693 (2020). Smith, H. L., Biggs, P. J., French, N. P., Smith, A. N. H. & Marshall, J. C. Lost in the Forest: Encoding categorical variables and the absent levels problem. Data Min Knowl Disc 38 , 1889–1908 (2024). Liao, S.-J. Statistical modelling for zoonotic diseases : a thesis presented in partial fulfilment of the requirements for the degree of Doctor of Philosophy in Statistics at Massey University, Palmerston North, New Zealand. (Massey University, 2020). Havelaar, A. H. et al. Unravelling the reservoirs for colonisation of infants with Campylobacter spp. in rural Ethiopia: protocol for a longitudinal study during a global pandemic and political tensions. BMJ Open 12 , e061311 (2022). Platts-Mills, J. A. et al. Detection of Campylobacter in Stool and Determination of Significance by Culture, Enzyme Immunoassay, and PCR in Developing Countries. J Clin Microbiol 52 , 1074–1080 (2014). Davedow, T. et al. PulseNet International Survey on the Implementation of Whole Genome Sequencing in Low and Middle-Income Countries for Foodborne Disease Surveillance. Foodborne Pathogens and Disease (2022) doi:10.1089/fpd.2021.0110. Bushnell, B., Rood, J. & Singer, E. BBMerge – Accurate paired shotgun read merging via overlap. PLOS ONE 12 , e0185056 (2017). BBDuk Guide. DOE Joint Genome Institute https://jgi.doe.gov/data-and-tools/software-tools/bbtools/bb-tools-user-guide/bbduk-guide/. Kokot, M., Długosz, M. & Deorowicz, S. KMC 3: counting and manipulating k-mer statistics. Bioinformatics 33 , 2759–2761 (2017). Seemann, T. mlst. https://github.com/tseemann/mlst. Jolley, K. A. & Maiden, M. C. BIGSdb: Scalable analysis of bacterial genome variation at the population level. BMC Bioinformatics 11 , 595 (2010). Dingle, K. E. et al. Multilocus Sequence Typing System forCampylobacter jejuni. Journal of Clinical Microbiology 39 , 14–23 (2001). Zomer, A. aldertzomer/cgmlst. (2022). Zhou, Z. et al. GrapeTree: visualization of core genomic relationships among 100,000 bacterial pathogens. Genome Res. 28 , 1395–1404 (2018). Anderson, M. J. Permutational Multivariate Analysis of Variance (PERMANOVA). in Wiley StatsRef: Statistics Reference Online 1–15 (John Wiley & Sons, Ltd, 2017). doi:10.1002/9781118445112.stat07841. Oksanen, J. et al. Vegan: Community Ecology Package . (2024). R Core Team. R: A Language and Environment for Statistical Computing . (R Foundation for Statistical Computing, Vienna, Austria, 2023). Wright, M. N. & Ziegler, A. ranger: A Fast Implementation of Random Forests for High Dimensional Data in C++ and R. Journal of Statistical Software 77 , 1–17 (2017). Kuhn & Max. Building Predictive Models in R Using the caret Package. Journal of Statistical Software 28 , 1-26 (2008). Kursa, M. B. & Rudnicki, W. R. Feature Selection with the Boruta Package. Journal of Statistical Software 36 , 1–13 (2010). Brinch, M. L. et al. Comparison of Source Attribution Methodologies for Human Campylobacteriosis. Pathogens 12 , 786 (2023). Njage, P. M. K., Leekitcharoenphon, P. & Hald, T. Improving hazard characterization in microbial risk assessment using next generation sequencing data and machine learning: Predicting clinical outcomes in shigatoxigenic Escherichia coli . International Journal of Food Microbiology 292 , 72–82 (2019). Ogutu, J. O., Piepho, H.-P. & Schulz-Streeck, T. A comparison of random forests, boosting and support vector machines for genomic selection. BMC Proceedings 5 , S11 (2011). Machado, G., Mendoza, M. R. & Corbellini, L. G. What variables are important in predicting bovine viral diarrhea virus? A random forest approach. Veterinary Research 46 , 85 (2015). Poorrashidi, M., Hitchcock, M. & Xu, J. Meta-analyses of the global multilocus genotypes of the human pathogen Campylobacter jejuni. Genome 67 , 189–203 (2024). Mather, A. E., Gilmour, M. W., Reid, S. W. J. & French, N. P. Foodborne bacterial pathogens: genome-based approaches for enduring and emerging threats in a complex and changing world. Nat Rev Microbiol (2024) doi:10.1038/s41579-024-01051-z. Wallace, R. L. et al. Campylobacter jejuni ST50, a pathogen of global importance: A comparative genomic analysis of isolates from Australia, Europe and North America. Zoonoses and Public Health 68 , 638–649 (2021). Belina, D. et al. Occurrence and diversity of Campylobacter species in diarrheic children and their exposure environments in Ethiopia. PLOS Glob Public Health 4 , e0003885 (2024). Jaakkonen, A., Kivistö, R., Aarnio, M., Kalekivi, J. & Hakkinen, M. Persistent contamination of raw milk by Campylobacter jejuni ST-883. PLoS One 15 , e0231810 (2020). Joensen, K. G. et al. Whole-Genome Sequencing to Detect Numerous Campylobacter jejuni Outbreaks and Match Patient Isolates to Sources, Denmark, 2015–2017 - Volume 26, Number 3—March 2020 - Emerging Infectious Diseases journal - CDC. Emerging Infectious Diseases 26 , 523–532 (2020). Admasie, A. et al. Genomic diversity of Campylobacter jejuni and Campylobacter coli isolated from the Ethiopian dairy supply chain. PLOS ONE 19 , e0305581 (2024). French, N. P. et al. Population Structure and Antimicrobial Resistance in Campylobacter jejuni and C. coli Isolated from Humans with Diarrhea and from Poultry, East Africa - Volume 30, Number 10—October 2024 - Emerging Infectious Diseases journal - CDC. Emerging Infectious Diseases 30 , 2079–2089 (2024). Bloomfield, S. J. et al. Long-term Colonization by Campylobacter jejuni Within a Human Host: Evolution, Antimicrobial Resistance, and Adaptation. The Journal of Infectious Diseases 217 , 103–111 (2018). Djeghout, B. et al. Comparative genomics of Campylobacter jejuni from clinical campylobacteriosis stool specimens. Gut Pathogens 14 , 45 (2022). Moxon, E. R. & Murphy, P. A. Haemophilus influenzae bacteremia and meningitis resulting from survival of a single organism. Proceedings of the National Academy of Sciences 75 , 1534–1536 (1978). Schiaffino, F. et al. The epidemiology and impact of persistent Campylobacter infections on childhood growth among children 0–24 months of age in resource-limited settings. eClinicalMedicine 76 , 102841 (2024). Chen, D., Havelaar, A. H., Platts-Mills, J. A. & Yang, Y. Acquisition and clearance dynamics of Campylobacter spp. in children in low- and middle-income countries. Epidemics 46 , 100749 (2024). Wang, Y. et al. Quantitative Multi-pathway Assessment of Exposure to Fecal Contamination for Infants in Rural Ethiopia. 2024.08.29.24312786 Preprint at https://doi.org/10.1101/2024.08.29.24312786 (2024). Additional Declarations No competing interests reported. Supplementary Files Supplementarytablesfigures.docx MLSTprofiles.xlsx Cite Share Download PDF Status: Published Journal Publication published 03 May, 2025 Read the published version in Gut Pathogens → Version 1 posted Editorial decision: Revision requested 05 Mar, 2025 Reviews received at journal 22 Feb, 2025 Reviewers agreed at journal 11 Jan, 2025 Reviewers agreed at journal 04 Jan, 2025 Reviewers invited by journal 02 Jan, 2025 Editor assigned by journal 30 Dec, 2024 Submission checks completed at journal 30 Dec, 2024 First submitted to journal 30 Dec, 2024 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-5735672","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":395826444,"identity":"d5efbf06-23f0-4a40-a18c-5f629498986a","order_by":0,"name":"Nitya Singh","email":"","orcid":"","institution":"University of Florida","correspondingAuthor":false,"prefix":"","firstName":"Nitya","middleName":"","lastName":"Singh","suffix":""},{"id":395826445,"identity":"80c10abd-f532-4f65-9f94-3789dea16a00","order_by":1,"name":"Cecilie A.N. Thystrup","email":"","orcid":"","institution":"Technical University of Denmark","correspondingAuthor":false,"prefix":"","firstName":"Cecilie","middleName":"A.N.","lastName":"Thystrup","suffix":""},{"id":395826446,"identity":"153f32da-e0cd-42f5-af05-4fe2a331ddf1","order_by":2,"name":"Bahar Mummed Hassen","email":"","orcid":"","institution":"Haramaya University","correspondingAuthor":false,"prefix":"","firstName":"Bahar","middleName":"Mummed","lastName":"Hassen","suffix":""},{"id":395826447,"identity":"b5803eca-64b7-4717-9621-2ffe5ce340b0","order_by":3,"name":"Menuka Bhandari","email":"","orcid":"","institution":"Virginia Tech","correspondingAuthor":false,"prefix":"","firstName":"Menuka","middleName":"","lastName":"Bhandari","suffix":""},{"id":395826448,"identity":"21f657f1-9b6e-4000-8f83-0efb3fe1245c","order_by":4,"name":"Gireesh Rajashekara","email":"","orcid":"","institution":"University of Illinois Urbana-Champaign","correspondingAuthor":false,"prefix":"","firstName":"Gireesh","middleName":"","lastName":"Rajashekara","suffix":""},{"id":395826449,"identity":"2e6c9b6f-0aa1-4e7e-bce5-018c38dd2abd","order_by":5,"name":"Tine M. Hald","email":"","orcid":"","institution":"Technical University of Denmark","correspondingAuthor":false,"prefix":"","firstName":"Tine","middleName":"M.","lastName":"Hald","suffix":""},{"id":395826450,"identity":"c040c492-6b9f-4357-bc29-d240400da49b","order_by":6,"name":"Mark J. Manary","email":"","orcid":"","institution":"Washington University in St. Louis","correspondingAuthor":false,"prefix":"","firstName":"Mark","middleName":"J.","lastName":"Manary","suffix":""},{"id":395826451,"identity":"b68c3f6b-7b63-46e6-b0b0-17cd2485f038","order_by":7,"name":"Sarah L. McKune","email":"","orcid":"","institution":"University of Florida","correspondingAuthor":false,"prefix":"","firstName":"Sarah","middleName":"L.","lastName":"McKune","suffix":""},{"id":395826452,"identity":"a5c81a0f-2b9b-4ec1-8868-dfaa563b65a0","order_by":8,"name":"Jemal Yusuf Hassen","email":"","orcid":"","institution":"Haramaya University","correspondingAuthor":false,"prefix":"","firstName":"Jemal","middleName":"Yusuf","lastName":"Hassen","suffix":""},{"id":395826453,"identity":"fc1ebc4d-a0fb-4ece-9ecb-deb509cd63b0","order_by":9,"name":"Helen L. Smith","email":"","orcid":"","institution":"Massey University","correspondingAuthor":false,"prefix":"","firstName":"Helen","middleName":"L.","lastName":"Smith","suffix":""},{"id":395826454,"identity":"4d74208e-b9ce-4a3d-a9e7-28a86ea1ede1","order_by":10,"name":"Jonathan C. Marshall","email":"","orcid":"","institution":"Massey University","correspondingAuthor":false,"prefix":"","firstName":"Jonathan","middleName":"C.","lastName":"Marshall","suffix":""},{"id":395826457,"identity":"0cbf97be-43ed-4f87-944c-9216e3c48200","order_by":11,"name":"Nigel P. French","email":"","orcid":"","institution":"Massey University","correspondingAuthor":false,"prefix":"","firstName":"Nigel","middleName":"P.","lastName":"French","suffix":""},{"id":395826459,"identity":"5ba7bd5e-5e07-4864-8c69-30af2b104c4d","order_by":12,"name":"Arie H. Havelaar","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA6ElEQVRIiWNgGAWjYFACxgZmEMUHIj4wyMkAKQPitLCBmDMYjHmI0MLAANfCzEOMFv7ZhxsfF9QclmNjP/zss22bAQ8De/M2CXxaJM4lNhvPOHbYmI0nzXh2LkgLz7EyvFoYzjC2SfOwpSW2MSQYM+e2/eFhkMgxw6tF/gxj+2+ef2n1bfzPPzNbgmyRf4NfiwHQFmbeNpsENokcY2ZGkBYJHvxaDM8wNkvz9tkYtkm8KWbsOWfAA/RUsQU+LXJn2B9+5vkmIc/Pn76Z4UeZgRw/++GNN/BpwQRspCkfBaNgFIyCUYANAACJ0zp8dH4mwQAAAABJRU5ErkJggg==","orcid":"","institution":"University of Florida","correspondingAuthor":true,"prefix":"","firstName":"Arie","middleName":"H.","lastName":"Havelaar","suffix":""}],"badges":[],"createdAt":"2024-12-30 12:53:19","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-5735672/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-5735672/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1186/s13099-025-00691-7","type":"published","date":"2025-05-03T15:56:59+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":72739744,"identity":"4d499696-c599-4982-980d-9e0ec0a03429","added_by":"auto","created_at":"2025-01-01 09:25:42","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":917378,"visible":true,"origin":"","legend":"\u003cp\u003eMinimum spanning tree illustrating the distribution of \u003cem\u003eCampylobacter jejuni\u003c/em\u003e core genome MLST types among human and livestock isolates from Ethiopia. Nodes are color-coded according to host type (human or livestock), and each node represents a unique isolate. Solid lines between nodes indicate phylogenetic relatedness, and scale bar represents a 600-loci distance. Clusters of nodes sharing identical ST types are labeled with the corresponding ST sequence type in black text.\u003c/p\u003e","description":"","filename":"Figure1.png","url":"https://assets-eu.researchsquare.com/files/rs-5735672/v1/9746e45f79a5f9d22f437bfd.png"},{"id":72739748,"identity":"2787b36f-1198-46df-8f96-4ad811ef1e99","added_by":"auto","created_at":"2025-01-01 09:25:42","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":4606430,"visible":true,"origin":"","legend":"\u003cp\u003eSpatial distribution of STs from infants in Haramaya woreda, Ethiopia. Solid blue lines indicate kebele boundaries, dots indicate geographic position of STs. Shading indicates vegetation density. Haramaya University is in the center of the map, just north of Haro Maya city. Map constructed with \u003cem\u003eggmap\u003c/em\u003e , based on OpenStreetMaps.\u003c/p\u003e","description":"","filename":"Figure2.png","url":"https://assets-eu.researchsquare.com/files/rs-5735672/v1/66ef9ae48e6385f42e122666.png"},{"id":72739758,"identity":"7fa6e258-1ced-43ac-9147-a0c1d9ee0abb","added_by":"auto","created_at":"2025-01-01 09:25:43","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":274583,"visible":true,"origin":"","legend":"\u003cp\u003eAttribution estimates for \u003cem\u003eC. jejuni\u003c/em\u003e infections of infants to livestock and human sources using Asymmetric Island, PCO and \u003cem\u003ek\u003c/em\u003e-mer models.\u003c/p\u003e","description":"","filename":"Figure3.png","url":"https://assets-eu.researchsquare.com/files/rs-5735672/v1/2373caaf5a51da3a59cfbe9a.png"},{"id":72739753,"identity":"254594ad-af54-49b3-9812-c0ffbbbb228f","added_by":"auto","created_at":"2025-01-01 09:25:43","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":1319597,"visible":true,"origin":"","legend":"\u003cp\u003eAttribution estimates for \u003cem\u003eC. jejuni\u003c/em\u003e infections of infants to livestock and human sources using the source-sink model.\u003c/p\u003e","description":"","filename":"Figure4.png","url":"https://assets-eu.researchsquare.com/files/rs-5735672/v1/b19692a1fbaf2bb204ba298b.png"},{"id":81987702,"identity":"bf0d6a27-5a4c-40fc-99d3-ab9f3d27892b","added_by":"auto","created_at":"2025-05-05 16:05:01","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":8096412,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-5735672/v1/b2abbd50-7aea-43c9-b512-82022faf1846.pdf"},{"id":72739746,"identity":"38d28de7-fa66-47c3-982b-f314d2b449d0","added_by":"auto","created_at":"2025-01-01 09:25:42","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":357295,"visible":true,"origin":"","legend":"","description":"","filename":"Supplementarytablesfigures.docx","url":"https://assets-eu.researchsquare.com/files/rs-5735672/v1/0ab167da273ddd6885768598.docx"},{"id":72740623,"identity":"ac7ec868-f7a4-4da5-9dfe-1386f45988c6","added_by":"auto","created_at":"2025-01-01 09:33:42","extension":"xlsx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":56222,"visible":true,"origin":"","legend":"","description":"","filename":"MLSTprofiles.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-5735672/v1/e378383e86f13b774dea4eca.xlsx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Transmission pathways of Campylobacter jejuni between humans and livestock in rural Ethiopia are highly complex and interdependent","fulltext":[{"header":"Background","content":"\u003cp\u003e \u003cem\u003eCampylobacter jejuni\u003c/em\u003e and \u003cem\u003eC. coli\u003c/em\u003e are the most common causes of bacterial enteritis worldwide whereas symptomatic and asymptomatic infections are associated with stunting in children in low- and middle-income countries \u003csup\u003e\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e\u003c/sup\u003e. The vaccination of rhesus macaques against \u003cem\u003eC. coli\u003c/em\u003e not only reduced the incidence of overt diarrhea but also improved their linear growth \u003csup\u003e\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/sup\u003e. We previously reported a high prevalence of bacteria of the \u003cem\u003eCampylobacter\u003c/em\u003e genus in humans and livestock in smallholder households in Haramaya woreda, East Hararghe Zone, Oromia State, Ethiopia \u003csup\u003e\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e\u003c/sup\u003e. The prevalence and load in infants increased significantly with age with logistic regression predicting a prevalence of approximately 90% at 1 year of age. Most infections are asymptomatic, but the bacterial load is positively correlated with the risk of diarrhea \u003csup\u003e\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e\u003c/sup\u003e. The load of \u003cem\u003eCampylobacter\u003c/em\u003e in infant stools was greater in girls than in boys and increased with increasing food insecurity, different feeding practices (prelacteal feeding, early introduction of complementary foods, consumption of any solid foods, drinking of raw milk, household ownership of cattle and sheep but not chickens or goats), hygiene-related factors (improper disposal of infant stools, contact with animals or their feces, mouthing soil) and treatment with antibiotics in the previous month. Mothers\u0026rsquo; handwashing with soap and drinking from bottles with nipples were associated with lower loads \u003csup\u003e\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e,\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e\u003c/sup\u003e. Two main species were found in infants by real-time polymerase cChain reaction (qPCR) and shotgun metagenomic sequencing: \u003cem\u003eCandidatus\u003c/em\u003e C. infans (\u003cem\u003eC. infans\u003c/em\u003e, 60% at 1 year of age) and \u003cem\u003eC. jejuni\u003c/em\u003e (50% at 1 year of age). \u003cem\u003eC. upsaliensis\u003c/em\u003e was also detected but less frequently (20% at 1 year of age). \u003cem\u003eC. coli\u003c/em\u003e was not detected in infants by shotgun sequencing and was detected infrequently (1%) by qPCR \u003csup\u003e\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e,\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eIn high-income countries, animals, specifically livestock, are considered the primary reservoirs of human infections with the well-studied species \u003cem\u003eC. jejuni/coli\u003c/em\u003e, with foodborne transmission \u003csup\u003e\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u003c/sup\u003e, mainly through poultry meat \u003csup\u003e\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u003c/sup\u003e as the main pathway. However, less is known about their sources and transmission pathways in low- and middle-income countries, particularly for infants and young children.\u003c/p\u003e \u003cp\u003eThe load of \u003cem\u003eC. infans\u003c/em\u003e was greater in girls than in boys and was also elevated for infants who drank raw milk or crawled in areas contaminated with animal feces, whereas the load of \u003cem\u003eC. jejuni\u003c/em\u003e was greater for infants who put soil in their mouths. At both the genus and species level, there were mixed and often counterintuitive signals related to keeping animals in the home, whether confined or not \u003csup\u003e\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e,\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e\u003c/sup\u003e. These results suggest a complex contamination network among humans, animals and their environment.\u003c/p\u003e \u003cp\u003eSource attribution of \u003cem\u003eC. jejuni/coli\u003c/em\u003e has largely been based on legacy (seven-gene) Multi Locus Sequence Typing (MLST) \u003csup\u003e\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e\u003c/sup\u003e, using frequency matching models such as the Dutch model \u003csup\u003e\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e\u003c/sup\u003e, Hald model \u003csup\u003e\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e\u003c/sup\u003e and variants \u003csup\u003e\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u003c/sup\u003e, or population genetic models such as the asymmetric Island model \u003csup\u003e\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e\u003c/sup\u003e or STRUCTURE \u003csup\u003e\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e\u003c/sup\u003e. More recently, attribution models based on whole-genome sequencing (WGS) data have been developed. WGS data enhance the ability to distinguish genetic variations and potentially more accurately determine the origin of infection-causing isolates \u003csup\u003e\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e\u003c/sup\u003e. Both core genome MLST (cgMLST) and \u003cem\u003ek\u003c/em\u003e-merization have been used for taxonomic profiling and are particularly effective for large genomes \u003csup\u003e\u003cspan additionalcitationids=\"CR17\" citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e\u003c/sup\u003e. \u003cem\u003eK\u003c/em\u003e-mer counting involves the use of short oligonucleotides to compare a sequence to either a reference genome or against genome of interest without needing an alignment \u003csup\u003e\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e\u003c/sup\u003e. The use of cgSTs or \u003cem\u003ek-\u003c/em\u003emers for differentiating \u003cem\u003eCampylobacter\u003c/em\u003e genomes is based on the concept of genomic signatures and builds on the premise that infections originating from the same source are genetically more similar than those from different sources, facilitating the tracking of infections across various sources. Random forest models may use cgMLST data following the numerical encoding of alleles \u003csup\u003e\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e\u003c/sup\u003e. Encoding genes with the PCO-encoding method \u003csup\u003e\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e\u003c/sup\u003e incorporates information that quantifies the similarity between each pair of alleles and addresses issues related to missing alleles and new genotypes in observations for prediction. All models typically assume unidirectional flow from sources to sinks. However, a model to include intermediate nodes, which may act as both a source and a sink was developed to explore the role of water in the transmission of bacteria from livestock and water birds \u003csup\u003e\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eThis study aims to assess the genomic diversity of \u003cem\u003eCampylobacter\u003c/em\u003e spp. in infants, humans and livestock in the Haramaya woreda, and to determine the attribution of \u003cem\u003eCampylobacter\u003c/em\u003e infections in infants to livestock (chickens, cattle, goats and sheep) and other humans (siblings, mothers) on the basis of the genetic population structure of \u003cem\u003eCampylobacter\u003c/em\u003e spp. circulating in these reservoirs, using four different attribution models. Given the observed species distribution in infants, and the fact that \u003cem\u003eCandidatus\u003c/em\u003e C. infans is not yet culturable, this study focused on source attribution of \u003cem\u003eC. jejuni\u003c/em\u003e.\u003c/p\u003e"},{"header":"Methods","content":"\u003ch3\u003eIsolation of \u003cem\u003eCampylobacter\u003c/em\u003e from human stool and livestock fecal samples\u003c/h3\u003e\n\u003cp\u003eDirect culturing was applied to approximately 15% (326/2,1833) of the total samples processed. For the isolation of thermotolerant species by direct plating, one gram of fresh stool/animal feces was suspended in 9 ml buffered peptone water (pH 7; BD Difco). 100 \u0026micro;l of homogenized samples were spread on CHROMagar Campylobacter (CaC, DRG International, Springfield, New Jersey USA) using sterile glass beads and incubated for 48 hours at 42\u0026deg;C in microaerobic condition (85% nitrogen, 10% carbon dioxide, 5% oxygen) in anaerobic jars with GasPak EZ Campy Container System Sachets (ThermoFisher Scientific, Waltham, MA, USA). Similarly, for non-thermotolerant species, the same volumes of samples (100 \u0026micro;l) were plated on Columbia agar supplemented with 5% defibrinated sheep blood, Skirrow supplement (2 \u0026micro;L/mL), amphotericin B (5\u0026micro;g/ mL), cefoperazone (8\u0026micro;g/mL) and \u003cem\u003eCampylobacter\u003c/em\u003e growth supplement (ThermoFisher Scientific, Waltham, MA, USA). The plates were incubated at 37\u0026deg;C for 48 hours in microaerobic condition.\u003c/p\u003e\n\u003cp\u003eIn parallel, samples were also enriched in Preston and Bolton broth with a proportion of 1 g feces in 9 ml of broth and incubator at 42\u0026deg;C and 37\u0026deg;C for 48 hours, respectively, as described above. After incubation, 100 \u0026micro;l of Preston broth enriched samples were plated onto CaC and Bolton broth enriched samples on Colombia agar and plates were incubated at either 37 \u0026deg;C or 42 \u0026deg;C \u003csup\u003e23\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003eWe could not directly culture almost 85% (1,857/2,183) of the fresh feces/stools due to global supply issues during the COVID-19 pandemic. We stored these samples in 20% (w/v) glycerol at -80 \u0026deg;C. Preliminary analysis indicated that up to 99% of the \u003cem\u003eCampylobacter\u003c/em\u003e population in the feces could not be recovered on CaC within the first month of storage; therefore, samples were pre-enriched in Bolton broth before plating.\u003c/p\u003e\n\u003cp\u003eTypical \u003cem\u003eCampylobacter\u003c/em\u003e colonies (up to 5 per plate) were sub-cultured onto a CaC plate and confirmed by genus-specific qPCR \u003csup\u003e24\u003c/sup\u003e. Potential thermotolerant and non-thermotolerant \u003cem\u003eCampylobacter\u003c/em\u003e were characterized by streaking the confirmed pure isolate on to two fresh CaC plates and incubating at 37\u0026deg;C and 42\u0026deg;C in microaerophilic conditions for 48 hrs. The isolates growing at 42\u0026deg;C and 37\u0026deg;C were recorded as potentially thermotolerant while the isolates growing only at 37\u0026deg;C were recorded as potentially non-thermotolerant. Despite our efforts to isolate non-thermotolerant \u003cem\u003eCampylobacter\u003c/em\u003e species, we were only able to culture one non-thermotolerant isolate, later determined to be \u003cem\u003eC. hyointestinalis\u003c/em\u003e. All isolates were stored in glycerol at -80 \u0026deg;C.\u003c/p\u003e\n\u003cp\u003eFor genomic DNA extraction, \u003cem\u003eCampylobacter\u003c/em\u003e isolates from the freezer stock were grown on a CaC agar plate for approximately 36 hours under microaerophilic condition at 42 ℃. A loopful of growth was collected from the CaC plate, resuspended in 1 ml of Mueller Hinton broth (ThermoFisher Scientific, Waltham, MA, USA) and genomic DNA was extracted using Promega Wizard genomic DNA purification kit (Promega, Madison, WI, United States) following the manufacturer\u0026rsquo;s instructions. The concentration and quality of the DNA were determined using NanoDrop 2000 C Spectrophotometer (ThermoFisher Scientific, MA, USA). Purified DNA was shipped to eight GenomeTrakr Laboratories (FDA, USA) for sequencing.\u003c/p\u003e\n\u003cp\u003eShort-read genomic DNA libraries were prepared with the Illumina DNA prep kit, following the PulseNet Sequencing Protocol PNL35 \u003csup\u003e25\u003c/sup\u003e. Samples were sequenced using either paired-end 2x150 bp or 2x250 bp reads, which vary between sequencing laboratories (see Supplementary file MLST_profiles.xlsx for Biosample IDs). The paired-end reads were assessed for quality and contamination and trimmed using BBMerge (v.38.90) \u003csup\u003e26\u003c/sup\u003e and BBDuk (v.38.90) \u003csup\u003e27\u003c/sup\u003e with the following parameters: hammering distance 1, optimal \u003cem\u003ek\u003c/em\u003e-mers 23, quality cutoff Q14 and minimum read length 30 bp, with end-trimming of a maximum 1of 0 bp. Species assignment was performed using KMC (version 3.0) \u003csup\u003e28\u003c/sup\u003e resulting in the identification of 380 \u003cem\u003eCampylobacter jejuni\u003c/em\u003e isolates for analysis in this study.\u003c/p\u003e\n\u003ch3\u003eMLST assignment\u003c/h3\u003e\n\u003cp\u003eLegacy MLST profiles (STs) for all \u003cem\u003eC. jejuni\u003c/em\u003e isolates were determined using the \u003cem\u003emlst\u003c/em\u003e tool \u003csup\u003e29\u003c/sup\u003e, which utilizes the most recent update of the PubMLST database \u003csup\u003e30\u003c/sup\u003e (updated December 14, 2024), which incorporates the seven housekeeping loci scheme, as previously described \u003csup\u003e31\u003c/sup\u003e. We identified 11 novel STs and submitted these new schemes to PubMLST for the assignment of new STs. Core genome MLST (cgMLST) profiles were assigned using a 1,343-loci scheme (Cody et al., 2017), implemented through the \u003cem\u003ecgMLST\u003c/em\u003e tool\u003csup\u003e32\u003c/sup\u003e. Missing alleles, which were unassigned, were identified using a custom R script and were assigned unique identifiers within the dataset. This approach ensured that all 1,343 loci were included in the analysis. The R script is publicly available at https://github.com/jmarshallnz/cgmlst. The sample set of 380 WGS samples included sequences collected from the same household and time point, representing different colonies obtained during subculturing in the pure isolation process. To create a unique representative dataset, we selected the isolates with the highest genomic coverage for each cgST type from each household and time point. This filtering resulted in a final WGS dataset comprising 287 isolates.\u003c/p\u003e\n\u003ch3\u003eMS tree construction and map\u003c/h3\u003e\n\u003cp\u003eA minimum spanning tree (MST) was constructed using the GrapeTree plugin with the MSTreeV2 algorithm, which is designed to handle missing data more effectively than classical MST methods do \u003csup\u003e33\u003c/sup\u003e. The process begins by calculating a directed minimal spanning arborescence using Edmonds\u0026apos; algorithm from asymmetric distances, with tie-breaking based on allelic distances. Local branch recrafting was then performed to remove spurious branches.\u003c/p\u003e\n\u003ch3\u003ePERMANOVA\u003c/h3\u003e\n\u003cp\u003eTo explore the transmission of \u003cem\u003eC. jejuni\u003c/em\u003e at different levels within the sampling hierarchy we performed a nested, permutational multivariate analysis of variance (PERMANOVA)\u0026nbsp;\u003csup\u003e34\u003c/sup\u003e to estimate the proportion of variance in infant cgST profiles attributable to each level, namely sample (i.e., multiple isolates from the same infant sample), infant (i.e., multiple isolates from the same infant at different time points), ganda (village) and kebele (the smallest administrative unit in Ethiopia, a set of gandas). PERMANOVA models were constructed using a customised R script (https://github.com/jmarshallnz/permanova). Pairwise genetic distances were calculated from the cgST profiles to create a distance matrix with values in the matrix corresponding to the Gower distance calculated using the vegdist() function in the R package \u003cem\u003evegan\u003c/em\u003e \u003csup\u003e35\u003c/sup\u003e. Multiple two-level nested models were considered: \u003cem\u003einfant within ganda\u003c/em\u003e, \u003cem\u003eganda within kebele\u003c/em\u003e, \u003cem\u003einfant within kebele\u003c/em\u003e and \u003cem\u003einfant time point\u003c/em\u003e \u003cem\u003ewithin infant\u003c/em\u003e. It was not possible to fit models considering higher-level nested structures, so only two-level nested models are presented. Univariate PERMANOVA models were performed for each factor with \u003cem\u003ep-value\u003c/em\u003es obtained using 100000 unrestricted permutations of raw data.\u003c/p\u003e\n\u003ch3\u003eDiversity and persistence of \u003cem\u003eC. jejuni\u003c/em\u003e infections\u003c/h3\u003e\n\u003cp\u003eThe isolate set included up to four isolates from the same infant sample, while 25 sets of repeat samples at different timepoints (approximately monthly intervals) were also available. We quantified the diversity of \u003cem\u003eC. jejuni\u003c/em\u003e isolates in these samples by counting the number of isolates by ST/cgST using a bespoke coding system in Excel (Supplementary file MLST_profiles.xlsx). Similarly, the occurrence of the same or different STs/cgSTs in sequential samples from the same infant was quantified and summarized in relation to the time interval between two samples using pivot tables in Excel.\u003c/p\u003e\n\u003ch3\u003eAttribution\u003c/h3\u003e\n\u003cp\u003eWhile livestock are commonly recognized as the main sources of human infections, humans other than infants are exposed to the same contaminated environment as the infants are, and they can be considered either as independent receiving hosts or as receiving, amplifying hosts that transmit infection to infants. Exclusive human-to-human transmission cycles cannot be excluded \u003cem\u003ea priori\u003c/em\u003e. We therefore fitted models with only livestock sources as well as models with other humans as additional sources of infections in infants.\u003c/p\u003e\n\u003cp\u003eA summary of all fitted models is provided in Table 1. All analyses were performed in the statistical language R version 4.3.0 or later \u003csup\u003e36\u003c/sup\u003e or Excel (Microsoft Corporation, Redmond, WA), using dedicated R packages or other software as indicated in the text.\u003c/p\u003e\n\u003cp\u003eTable 1. Overview of source attribution models\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"624\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 145px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eSource attribution model\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 100px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eInput data\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 170px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eSummary data\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 95px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eAttribution pathways\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 113px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eModel fitting procedure\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 145px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eAsymmetric Island\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 100px;\"\u003e\n \u003cp\u003eAssembled contigs\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 170px;\"\u003e\n \u003cp\u003ecgST\u003csup\u003e*\u003c/sup\u003e profiles\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 95px;\"\u003e\n \u003cp\u003eDirect\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 113px;\"\u003e\n \u003cp\u003eMCMC\u003csup\u003e\u0026amp;\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 145px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u003cem\u003ek\u003c/em\u003e-mer\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 100px;\"\u003e\n \u003cp\u003eRaw reads\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 170px;\"\u003e\n \u003cp\u003e9-mers\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 95px;\"\u003e\n \u003cp\u003eDirect\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 113px;\"\u003e\n \u003cp\u003eRandom Forest with feature reduction\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 145px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u003cem\u003ePCO\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 100px;\"\u003e\n \u003cp\u003eAssembled contigs\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 170px;\"\u003e\n \u003cp\u003ecgST profiles + allele sequences\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 95px;\"\u003e\n \u003cp\u003eDirect\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 113px;\"\u003e\n \u003cp\u003eRandom Forest without feature reduction\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 145px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eSource-sink\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 100px;\"\u003e\n \u003cp\u003eAssembled contigs\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 170px;\"\u003e\n \u003cp\u003ecgST profiles\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 95px;\"\u003e\n \u003cp\u003eDirect and indirect\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 113px;\"\u003e\n \u003cp\u003eMCMC\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u003csup\u003e*\u003c/sup\u003ecore genome Multi Locus Sequence Typing\u003c/p\u003e\n\u003cp\u003e\u003csup\u003e\u0026amp;\u003c/sup\u003eMarkov Chain Monte Carlo\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eAsymmetric Island model\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eThe asymmetric-island source attribution model \u003csup\u003e13\u003c/sup\u003e was applied to the cgMLST data for 287 isolates using the \u003cem\u003eislandR\u003c/em\u003e package (https://github.com/jmarshallnz/islandR). This method assigns sources to isolates by modeling both recombination and mutation processes as distinct events. In our analysis, the recombination and mutation probabilities were assumed to be constant across all sources, resulting in pooled estimates for both processes. The model incorporates genetic differences between isolates to estimate the most likely source, accounting for multiple potential sources. From the model estimates, the mean and 95% confidence intervals of the posterior distribution were calculated to determine the attribution percentages for each source. This approach is particularly useful for including recombination and mutation in source attribution, enhancing the accuracy of source estimation based on cgST profiles.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003ePCO model\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eThe PCO source attribution model is a random forest model that uses a principal coordinates approach to overcome the problem of missing levels in the data used for prediction. This method uses a target-agnostic approach to encode cgMLST predictor variables by using both the cgMLST allele profiles and Hamming distances between allele sequences to determine the similarity between pairs of isolates\u0026nbsp;\u003csup\u003e21\u003c/sup\u003e.\u0026nbsp;Analyses were carried out using the \u003cem\u003eranger\u003c/em\u003e package \u003csup\u003e37\u003c/sup\u003e and the PCO-encoding method (https://github.com/smithhelen/LostInTheForest). Estimates of uncertainty are calculated using a probability forest with the same set of parameters as the original random forest. For each tree in the forest, the probability of each human isolate being attributed to each source is calculated. These probabilities are then averaged over the set of human isolates, giving an average probability of attribution to each source for each tree. The 2.5% and 97.5% quantiles are then determined from this set of mean probabilities to give a 95% uncertainty interval.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eSource/sink model\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eThe role of humans as intermediate hosts for zoonotic infections was further explored using a model in which they can act both as sources and as a sink. The IslandR model was reparametrized to consider the mothers and siblings as both receiving infection from the animal reservoirs and being a source of infection for infants. The source/sink model was based on an approach developed to examine the contribution of water as both a source and a sink for human campylobacteriosis in New Zealand\u0026nbsp;\u003csup\u003e22\u003c/sup\u003e. In essence, we assume that mother and sibling isolates arise from a mix of the animal reservoirs, whereas infant isolates arise from a mix of both the animal reservoirs and mothers and siblings. This model is fit using the\u0026nbsp;attribution_intermediate()\u0026nbsp;function in the \u003cem\u003eislandR\u003c/em\u003e package (https://github.com/jmarshallnz/islandR).\u003c/p\u003e\n\u003cp\u003e\u003cem\u003ek-mer model\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eThe tool KMC (version 3.0) \u003csup\u003e28\u003c/sup\u003e was used to extract \u003cem\u003ek-\u003c/em\u003emers with length of k = 9 for each of the samples using the short-read sequences. All \u003cem\u003ek\u003c/em\u003e-mer frequencies were then combined into one matrix using an in-house Python script. A recently developed \u003csup\u003e20\u003c/sup\u003e source attribution model was applied to the two datasets using the \u003cem\u003ek\u003c/em\u003e-mers to predict the sources of human campylobacteriosis cases. Feature reduction was carried out on the matrix to reduce the number of \u003cem\u003ek\u003c/em\u003e-mers in the final model using the \u003cem\u003ecaret\u003c/em\u003e package (version 6.0-94) \u003csup\u003e38\u003c/sup\u003e and the \u003cem\u003eBoruta\u003c/em\u003e package (version 8.0.0) \u003csup\u003e39\u003c/sup\u003e. The near-zero-variance method was used to reduce the number of 9-mers. The Boruta algorithm was then applied to select important attributes in the matrix using a random forest classifier. To account for the uneven distribution of sources in the samples, all sources were upsampled to the highest number of samples available within a source, so that all sources had the same number of samples. Two machine-learning algorithms previously applied successfully in sequencing studies were evaluated \u003csup\u003e40\u0026ndash;43\u003c/sup\u003e. For the evaluation, the data containing \u003cem\u003ek\u003c/em\u003e-mers from sources were split into test- and training data sets. The training data were then used to randomly generate smaller sets of test and training data once again to determine which of the two selected machine-learning algorithms fit the data best. Each smaller test- and training data set was split 70% and 30%, respectively, and the test-data were used to evaluate the performance of the model using seven-fold cross-validation. After 10 iterations, the accuracy of each algorithm was assessed, and the algorithm with the highest accuracy was selected for model construction.\u003c/p\u003e\n\u003cp\u003eThe model with the highest accuracy was constructed again following the same steps as for the model selection, described previously, and the performance of the model was evaluated based on the accuracy of the cross-validation step, the kappa value and the confusion matrix, which determines the model\u0026rsquo;s ability to predict the sources of the samples in the source-data. The sensitivity and specificity were also reported. Finally, the model was applied to isolates from infants. This was done by estimating the probability of each human case being attributed to each of the sources included in the model. Tree-level predictions were pooled together to calculate the mean probability for each source, with 2.5\u003csup\u003eth\u003c/sup\u003e and 97.5\u003csup\u003eth\u003c/sup\u003e percentiles providing uncertainty intervals for each case. To estimate the uncertainty in the overall mean attribution probabilities, we performed 1,000 bootstrap resampling iterations. In each iteration, case isolates were sampled with replacement, and mean attribution probabilities were recalculated. The 2.5\u003csup\u003eth\u003c/sup\u003e and 97.5\u003csup\u003eth\u003c/sup\u003e percentiles of the bootstrap distributions provided 95% uncertainty intervals for the mean probabilities.\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003eWe included WGS data from 380 isolates for attribution from different sources analysis by \u003cem\u003ek\u003c/em\u003e-mers (Table 1) and of these, assigned cgST profiles to 287 isolates. The majority of human samples other than those from infants were collected from siblings and the majority of livestock samples were collected from chickens with fewer samples from ruminant species. Owing to the small number of isolates from several sources, we ran attribution models of infections in infants with two merged source groups: other humans (mothers and siblings) and small ruminants (sheep and goats). Cattle are often recognized as a major reservoir for \u003cem\u003eCampylobacter\u003c/em\u003e transmission to humans, particularly \u003cem\u003eC. jejuni\u003c/em\u003e. By keeping cattle separate, their unique role as a distinct source is emphasized. Additionally, other studies and risk assessments use the \u0026quot;small ruminant\u0026quot; classification for sheep and goats because of their shared characteristics. This convention supports comparability and consistency across studies.\u003c/p\u003e\n\u003cp\u003eTable 2. \u003cem\u003eC. jejuni\u003c/em\u003e isolates included in the k-mer and cgMLST attribution models\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"530\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 90px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eSource group\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eSource\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 68px;\"\u003e\n \u003cp\u003e\u003cstrong\u003ek-mer data set\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 102px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eGrouped k-mer data set\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 84px;\"\u003e\n \u003cp\u003e\u003cstrong\u003ecgMLST data set\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eGrouped cgMLST data set\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"2\" valign=\"top\" style=\"width: 165px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eInfants\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 68px;\"\u003e\n \u003cp\u003e229\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 102px;\"\u003e\n \u003cp\u003e229\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 84px;\"\u003e\n \u003cp\u003e174\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e174\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"2\" valign=\"top\" style=\"width: 90px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eOther Humans\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003eMothers\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 68px;\"\u003e\n \u003cp\u003e3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"2\" valign=\"top\" style=\"width: 102px;\"\u003e\n \u003cp\u003e29\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 84px;\"\u003e\n \u003cp\u003e3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e20\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003eSiblings\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 68px;\"\u003e\n \u003cp\u003e26\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 84px;\"\u003e\n \u003cp\u003e17\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 90px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003eChickens\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 68px;\"\u003e\n \u003cp\u003e86\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 102px;\"\u003e\n \u003cp\u003e86\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 84px;\"\u003e\n \u003cp\u003e71\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e71\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 90px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003eCattle\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 68px;\"\u003e\n \u003cp\u003e13\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 102px;\"\u003e\n \u003cp\u003e13\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 84px;\"\u003e\n \u003cp\u003e6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e6\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"2\" valign=\"top\" style=\"width: 90px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eSmall Ruminants\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003eSheep\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 68px;\"\u003e\n \u003cp\u003e2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"2\" valign=\"top\" style=\"width: 102px;\"\u003e\n \u003cp\u003e24\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 84px;\"\u003e\n \u003cp\u003e4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e16\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003eGoats\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 68px;\"\u003e\n \u003cp\u003e22\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 84px;\"\u003e\n \u003cp\u003e12\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 90px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eTotal\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 68px;\"\u003e\n \u003cp\u003e380\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 102px;\"\u003e\n \u003cp\u003e380\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 84px;\"\u003e\n \u003cp\u003e287\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 111px;\"\u003e\n \u003cp\u003e287\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003ch3\u003eGenomic diversity of \u003cem\u003eC. jejuni\u003c/em\u003e from different sources\u003c/h3\u003e\n\u003cp\u003eThe set of 287 isolates included 48 STs. The population structure based on cgMLST is presented as a minimum spanning tree (MSTree) in Figure 1. The tree is fully structured according to seven-gene STs and isolates within the same ST are highly related, differing by fewer than five alleles. There were 11 newly assigned STs (37 isolates; see Table S1). Among these, one type was common to infants, chickens, cattle and sheep; one was common to infants and chickens; six were unique to infants; and three were unique to chickens (Table S2). Twelve STs were represented by more than 10 isolates, which together accounted for almost two-thirds (185/287) of all the isolates. Isolates from six of these STs were shared among infants, livestock and other humans, whereas six were shared only between infants and livestock. Detailed data on these isolates are available in the Supplementary file MLST_profiles.xlsx.\u003c/p\u003e\n\u003ch3\u003eDiversity and persistence of \u003cem\u003eC. jejuni\u003c/em\u003e infections\u003c/h3\u003e\n\u003cp\u003eWe obtained multiple isolates of \u003cem\u003eC. jejuni\u003c/em\u003e from 81 samples. Of these, one sample yielded four isolates, 34 (28+5+1) samples yielded three isolates, and 46 (*39+7) samples yielded two isolates. All isolates from the same sample were of the same seven-gene ST in 84% (68/71) of the samples (Table 3 and supplementary file MLST_profiles.xlsx), thus not providing evidence of the diversity of STs within the host. Two or three different STs were found in 9 samples, suggesting sequence type diversity within the host. The cgST profiles of all the isolates from the same sample within the same ST were fewer than five alleles different.\u003c/p\u003e\n\u003cp\u003eTable 3. Multi-Locus Sequence Type diversity in human and livestock samples\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"540\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 88px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eIsolate pattern\u003csup\u003e*\u003c/sup\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 53px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eInfant\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 59px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eSibling\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 58px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eMother\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 50px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eCattle\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 58px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eGoat\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 66px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eChicken\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 54px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eSheep\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 54px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eTotal\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 88px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eA1234\u003csup\u003e\u0026amp;\u003c/sup\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 53px;\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 59px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 58px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 50px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 58px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 66px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 54px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 54px;\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 88px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eA123\u003csup\u003e\u0026amp;\u003c/sup\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 53px;\"\u003e\n \u003cp\u003e25\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 59px;\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 58px;\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 50px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 58px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 66px;\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 54px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 54px;\"\u003e\n \u003cp\u003e28\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 88px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eA12\u003csup\u003e\u0026amp;\u003c/sup\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 53px;\"\u003e\n \u003cp\u003e22\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 59px;\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 58px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 50px;\"\u003e\n \u003cp\u003e2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 58px;\"\u003e\n \u003cp\u003e2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 66px;\"\u003e\n \u003cp\u003e7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 54px;\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 54px;\"\u003e\n \u003cp\u003e39\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 88px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eA12B\u003csup\u003e#\u003c/sup\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 53px;\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 59px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 58px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 50px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 58px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 66px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 54px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 54px;\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 88px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eAB\u003csup\u003e#\u003c/sup\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 53px;\"\u003e\n \u003cp\u003e3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 59px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 58px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 50px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 58px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 66px;\"\u003e\n \u003cp\u003e4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 54px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 54px;\"\u003e\n \u003cp\u003e7\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 88px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eABC\u003csup\u003e#\u003c/sup\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 53px;\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 59px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 58px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 50px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 58px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 66px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 54px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 54px;\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"9\" valign=\"top\" style=\"width: 540px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eEvidence of sequence type diversity\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 88px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eNo\u003csup\u003e\u0026amp;\u003c/sup\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 53px;\"\u003e\n \u003cp\u003e48\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 59px;\"\u003e\n \u003cp\u003e6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 58px;\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 50px;\"\u003e\n \u003cp\u003e2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 58px;\"\u003e\n \u003cp\u003e2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 66px;\"\u003e\n \u003cp\u003e8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 54px;\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 54px;\"\u003e\n \u003cp\u003e68\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 88px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eYes\u003csup\u003e#\u003c/sup\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 53px;\"\u003e\n \u003cp\u003e9\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 59px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 58px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 50px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 58px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 66px;\"\u003e\n \u003cp\u003e4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 54px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 54px;\"\u003e\n \u003cp\u003e13\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 88px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eTotal\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 53px;\"\u003e\n \u003cp\u003e57\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 59px;\"\u003e\n \u003cp\u003e6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 58px;\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 50px;\"\u003e\n \u003cp\u003e2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 58px;\"\u003e\n \u003cp\u003e2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 66px;\"\u003e\n \u003cp\u003e12\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 54px;\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 54px;\"\u003e\n \u003cp\u003e81\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u003csup\u003e*\u003c/sup\u003eLetters indicate different seven-gene STs, and numbers indicate different cgSTs among isolates from one ST. The number of isolates differs between samples and is represented by the number of different letter/number combinations. For example, A1234 indicates four isolates from one sample with the same ST, but all different cgSTs, whereas A12B indicates three isolates from one sample of which two with the same ST but different cgSTs, and one isolate with a different ST (and consequently also a different cgST). Subscripts indicate patterns with\u003csup\u003e#\u003c/sup\u003e or without\u003csup\u003e\u0026amp;\u003c/sup\u003e evidence of sequence type diversity.\u003c/p\u003e\n\u003cp\u003ePersistence of \u003cem\u003eC. jejuni\u003c/em\u003e infections (i.e., isolation of the same ST from two sequential samples) was observed in 8% (2/25) of sample pairs from the same infant (supplementary file MLST_profiles.xlsx). These pairs were taken approximately 1 or 2 months apart. Different STs were observed in 84% of the sample pairs, the majority of these STs were also 1-2 months apart, but five pairs had 3-months intervals and 1 pair had a 4-month interval. Eight percent (2/25) of sample pairs (1- or 2-month intervals) provided inconclusive evidence with the same STs being isolated in both samples, accompanied by one or more different STs.\u003c/p\u003e\n\u003ch3\u003eSpatial clustering\u003c/h3\u003e\n\u003cp\u003eFigure 2 shows the spatial distribution of STs from infants. Common STs, such as ST50, ST883 and ST2042 were found in multiple gandas (villages) and kebeles (the smallest administrative unit in Ethiopia). We analyzed the distribution of cgST types at different nested levels of spatial and temporal sampling using PERMANOVA, estimating the contribution of the variation in cgMLST alleles attributable to kebeles, gandas, infants at different time points and infant samples. We tested for significant clustering at each level, using multiple two-level nested models (Table S3).\u003c/p\u003e\n\u003cp\u003eConsidering \u003cem\u003einfants within kebeles\u003c/em\u003e, some 14% of the total variation in cgST profiles was between kebeles. Most of the total variation was between \u003cem\u003einfants within kebeles\u003c/em\u003e (53%), and 34% of the variation was \u003cem\u003ewithin infants\u003c/em\u003e. There was marginally significant clustering at the kebele level (p = 0.03) and highly significant clustering at the infant level (p \u0026lt; 0.001).\u003c/p\u003e\n\u003cp\u003eConsidering \u003cem\u003egandas within kebeles\u003c/em\u003e, some 14% of the total variation in cgST profiles was between kebeles. Most of the total variation was between \u003cem\u003ehouseholds within kebeles\u003c/em\u003e (46%), and 40% of the variation was \u003cem\u003ewithin households\u003c/em\u003e. There was a tendency toward clustering at the kebele level (p =0.053) and highly significant clustering at the ganda level (p \u0026lt; 0.001).\u003c/p\u003e\n\u003cp\u003eConsidering \u003cem\u003einfants within gandas\u003c/em\u003e, some 58% of the total variation in cgST profiles was between gandas. The total variation between \u003cem\u003einfants within gandas\u003c/em\u003e was 9%, and 33% of the variation was \u003cem\u003ewithin infants\u003c/em\u003e. There was no significant clustering at the ganda level (p = 0.29) but there was highly significant clustering at the infant level (p \u0026lt; 0.001).\u003c/p\u003e\n\u003cp\u003eConsidering \u003cem\u003esamples within infants\u003c/em\u003e, some 68% of the total variation in cgST profiles was between infants. The total variation between \u003cem\u003esamples within infants\u003c/em\u003e was 22%, and 10% of the variation was \u003cem\u003ewithin infants\u003c/em\u003e. There was significant clustering at the infant level (p \u0026lt; 0.01) and highly significant clustering at the sample level (p \u0026lt; 0.001).\u003c/p\u003e\n\u003cp\u003eThe highly significant clustering at the sample level confirms observations in Table 2 of highly related isolates from one sample. We, therefore, repeated the analysis, using a reduced dataset including only unique STs per sample (97 isolates). The results of the partitioning of variance at all levels of clustering were very similar to those of the full dataset. However, the significance of clustering at the \u003cem\u003einfants within gandas\u003c/em\u003e changed markedly to no significant clustering at both the ganda and infant levels.\u003c/p\u003e\n\u003ch3\u003eAttribution\u003c/h3\u003e\n\u003cp\u003eThe attribution results for the Asymmetric Island, PCO and \u003cem\u003ek\u003c/em\u003e-mer models are summarized in Figure 3 and Table 4.\u003c/p\u003e\n\u003cp\u003eThe PCO random forest model was trained on the set of source isolates using all 1343 cgMLST genes (as nominal predictors). The genes were encoded using the PCO-encoding method together with a dissimilarity matrix of Hamming distances of the nucleotide sequencing information between each pair of alleles for each gene. Any new alleles in the set of human isolates for prediction were encoded using the method of principal coordinates based on pairwise Hamming distance from the new alleles to the alleles in the set of source isolates. The original sources of the infant isolates were then predicted.\u003c/p\u003e\n\u003cp\u003eThe \u003cem\u003ek\u003c/em\u003e-mer models were built on 131,073 attributes (or 9-mers), which were then reduced further by the near-zero variance method which removed few attributes from the model, depending on which dataset was modeled. The Boruta algorithm further reduced the matrix by selecting only those that are confirmed to provide enough important information to be included in the model.\u003c/p\u003e\n\u003cp\u003eFor the dataset including both humans and animals as sources, the near-zero variance method identified two attributes with low variance, which were removed from the data set. The Boruta algorithm further reduced the dataset to include 69 important attributes used for further modeling. For the dataset including only animal sources, the near zero variance method removed one attribute with low variance, whereas the Boruta algorithm further reduced the number of attributes to 23 used for further modeling. For both datasets, the performances of the random forest and the logit-boost algorithms were compared (Table S4). The average accuracies obtained from taking the average across ten iterations showed that the random forest and the logit boost algorithms performed very similarly in terms of accuracy for both data sets. The logit boost performed marginally better but could not provide uncertainty intervals comparable to the other approaches. Consequently, we decided to use the random forests algorithm.\u003c/p\u003e\n\u003cp\u003eThe final models predicted probabilities for each of the \u003cem\u003eC. jejuni\u0026nbsp;\u003c/em\u003einfections of infants to originate from each of the sources (Table 4). The results were similar for all models with most cases being attributed to chickens. The asymmetric Island model attributed most infections among ruminant sources to cattle, whereas the PCO and\u0026nbsp;\u003cem\u003ek\u003c/em\u003e-mer models (both machine-learning models) considered small ruminants more likely. The percentage of cases attributed to chickens was lower for the\u0026nbsp;\u003cem\u003ek\u003c/em\u003e-mer model than for the other two models, both of which use cgMLST for genomic characterization. Although both the\u0026nbsp;\u003cem\u003ek\u003c/em\u003e-mer and the PCO models are random forest models, the\u0026nbsp;\u003cem\u003ek\u003c/em\u003e-mer method uses feature reduction to substantially decrease the number of variables in the model. The smaller pool of predictors means there is less variation in each tree of the forest, which potentially explains the smaller uncertainty intervals When other humans are included as putative sources, the models estimate that some 16-22% of all infant isolates originate from other humans,\u0026nbsp;reducing\u0026nbsp;mainly the estimate for chickens.\u003c/p\u003e\n\u003cp\u003eTable 4: Attribution estimates for \u003cem\u003eC. jejuni\u003c/em\u003e infections of infants to livestock and human sources using Asymmetric Island, PCO and \u003cem\u003ek\u003c/em\u003e-mer models.\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"589\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"2\" valign=\"top\" style=\"width: 140px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eModels and source categories\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd colspan=\"4\" valign=\"top\" style=\"width: 446px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eAttribution percentage\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 3px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eChickens\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eCattle\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 131px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eSmall Ruminants (goats and sheep)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 162px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eOther Humans (mothers and siblings)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 3px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"6\" valign=\"top\" style=\"width: 589px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eAsymmetric Island model\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 140px;\"\u003e\n \u003cp\u003e\u0026nbsp; Livestock and humans\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003e60.6\u003csup\u003e*\u003c/sup\u003e\u003c/p\u003e\n \u003cp\u003e(44.4-74.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e15.2\u003c/p\u003e\n \u003cp\u003e(4.7-28.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 131px;\"\u003e\n \u003cp\u003e6.9\u003c/p\u003e\n \u003cp\u003e(0.1-14.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 162px;\"\u003e\n \u003cp\u003e17.3\u003c/p\u003e\n \u003cp\u003e(10.0-26.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 3px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 140px;\"\u003e\n \u003cp\u003e\u0026nbsp; Livestock\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003e78.1\u003c/p\u003e\n \u003cp\u003e(61.6-90.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e17.8\u003c/p\u003e\n \u003cp\u003e(0.1-30.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 131px;\"\u003e\n \u003cp\u003e3.6\u003c/p\u003e\n \u003cp\u003e(0.0-16.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 162px;\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 3px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"6\" valign=\"top\" style=\"width: 589px;\"\u003e\n \u003cp\u003e\u003cstrong\u003ePCO model\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 140px;\"\u003e\n \u003cp\u003e\u0026nbsp; Livestock and humans\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003e60.2\u003c/p\u003e\n \u003cp\u003e(47.9-74.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e4.6\u003c/p\u003e\n \u003cp\u003e(0.1-10.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 131px;\"\u003e\n \u003cp\u003e16.5\u003c/p\u003e\n \u003cp\u003e(0.1-29.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 162px;\"\u003e\n \u003cp\u003e16.3\u003c/p\u003e\n \u003cp\u003e(0.1-28.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 3px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 140px;\"\u003e\n \u003cp\u003e\u0026nbsp; Livestock\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003e72.7\u003c/p\u003e\n \u003cp\u003e(59.4-86.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e4.9\u003c/p\u003e\n \u003cp\u003e(0.1-13.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 131px;\"\u003e\n \u003cp\u003e21.9\u003c/p\u003e\n \u003cp\u003e(0.1-35.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 162px;\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 3px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"6\" valign=\"top\" style=\"width: 589px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u003cem\u003ek\u003c/em\u003e\u003c/strong\u003e\u003cstrong\u003e-mer model\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 140px;\"\u003e\n \u003cp\u003e\u0026nbsp; Livestock and humans\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003e52.3\u003c/p\u003e\n \u003cp\u003e(48.8-55.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e8.4\u003c/p\u003e\n \u003cp\u003e(7.5-9.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 131px;\"\u003e\n \u003cp\u003e17.2\u003c/p\u003e\n \u003cp\u003e(15.4-19.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 162px;\"\u003e\n \u003cp\u003e22.0\u003c/p\u003e\n \u003cp\u003e(19.1-25.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 3px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 140px;\"\u003e\n \u003cp\u003e\u0026nbsp; Livestock\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003e62.8\u003c/p\u003e\n \u003cp\u003e(59.5-65.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e12.1\u003c/p\u003e\n \u003cp\u003e(10.4-14.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 131px;\"\u003e\n \u003cp\u003e25.0\u003c/p\u003e\n \u003cp\u003e(22.4-27.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 162px;\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 3px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u003csup\u003e*\u0026nbsp;\u003c/sup\u003eMean (95% uncertainty interval)\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe source/sink model was run for different source combinations (Figure 4 and Table 5. When considering infections of infants from all putative sources independently, the results were similar to those of the Island model, as expected. When considering infections in mothers and siblings from livestock sources, the attribution was almost exclusively to chickens with very low percentages attributed to cattle or small ruminants. Likewise, if mothers and siblings were considered as intermediate sources between livestock and infants, chickens were identified as the main source which then gets passed on to infants. This results in a lack of identifiability for the attribution to infants when aiming to separate the direct chicken route from the indirect chicken route through mothers and siblings (Figure 4D).\u003c/p\u003e\n\u003cp\u003eTable 5: Attribution estimates for \u003cem\u003eC. jejuni\u003c/em\u003e infections of infants to livestock and human sources using the source-sink model.\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"589\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"2\" valign=\"top\" style=\"width: 140px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eModels and source categories\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd colspan=\"4\" valign=\"top\" style=\"width: 446px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eAttribution percentage\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 3px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eChickens\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eCattle\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 131px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eSmall Ruminants (goats and sheep)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 162px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eOther Humans (mothers and siblings)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 3px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"6\" valign=\"top\" style=\"width: 589px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eTo infants from all sources, no intermediate source\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 140px;\"\u003e\n \u003cp\u003e\u0026nbsp; Livestock and humans\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003e58.4\u003csup\u003e*\u003c/sup\u003e\u003c/p\u003e\n \u003cp\u003e(42.6-73.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e17.2\u003c/p\u003e\n \u003cp\u003e(3.9-31.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 131px;\"\u003e\n \u003cp\u003e7.0\u003c/p\u003e\n \u003cp\u003e(1.1-14.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 162px;\"\u003e\n \u003cp\u003e17.4\u003c/p\u003e\n \u003cp\u003e(10.3-26.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 3px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"6\" valign=\"top\" style=\"width: 589px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eTo mothers and siblings from livestock, ignoring infants\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 140px;\"\u003e\n \u003cp\u003e\u0026nbsp; Livestock\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003e92.2\u003c/p\u003e\n \u003cp\u003e(67.4-99.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e2.9\u003c/p\u003e\n \u003cp\u003e(0.0-18.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 131px;\"\u003e\n \u003cp\u003e5.1\u003c/p\u003e\n \u003cp\u003e(0.0-25.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 162px;\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 3px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"6\" valign=\"top\" style=\"width: 589px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eTo infants from livestock via mother and siblings\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 140px;\"\u003e\n \u003cp\u003e\u0026nbsp; Livestock\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003e92.2\u003c/p\u003e\n \u003cp\u003e(69.3-99.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e3.3\u003c/p\u003e\n \u003cp\u003e(0.0-20.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 131px;\"\u003e\n \u003cp\u003e4.5\u003c/p\u003e\n \u003cp\u003e(0.0-21.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 162px;\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 3px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"5\" valign=\"top\" style=\"width: 586px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eTo infants from all sources, mothers and siblings as intermediate sources\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 3px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 140px;\"\u003e\n \u003cp\u003eLivestock and humans\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003e50.9\u003c/p\u003e\n \u003cp\u003e(0.3-88.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e16.2\u003c/p\u003e\n \u003cp\u003e(2.6-29.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 131px;\"\u003e\n \u003cp\u003e2.8\u003c/p\u003e\n \u003cp\u003e(0.0-14.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 162px;\"\u003e\n \u003cp\u003e30.0\u003c/p\u003e\n \u003cp\u003e(0.0-86.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 3px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u003csup\u003e*\u0026nbsp;\u003c/sup\u003eMean (95% uncertainty interval)\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eIn the study region, 9 out of 10 infants were colonized by one or more \u003cem\u003eCampylobacter\u003c/em\u003e species by the age of one year of which 6 out of 10 were colonized by \u003cem\u003eC. jejuni\u003c/em\u003e \u003csup\u003e\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e,\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e\u003c/sup\u003e. Most infections were asymptomatic, although the odds of diarrhea increased with increasing bacterial load \u003csup\u003e\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e\u003c/sup\u003e. We used whole-genome-sequencing data to analyze the genomic diversity of \u003cem\u003eC. jejuni\u003c/em\u003e, and to attribute infections to putative livestock and human sources. Among the 287 isolates, 48 STs were identified, among which 11 were previously unreported types. In a meta-analysis of global genotypes of \u003cem\u003eC. jejuni\u003c/em\u003e, Poorrashidi et al. \u003csup\u003e\u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e44\u003c/span\u003e\u003c/sup\u003e reported that the six most common STs globally are ST21, ST45, ST50, ST48, and ST257. Among these STs, only ST50 was found in our study as the most frequent ST. Many of the isolates recovered in this study do not belong to the most common STs reported in high-resource settings, and many have never been reported before. This is likely to be due to sampling bias and limited application of WGS in low-resource settings \u003csup\u003e\u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e45\u003c/span\u003e\u003c/sup\u003e. Of the 87,542 genomes of \u003cem\u003eCampylobacter\u003c/em\u003e in the PubMLST database, only 386 are from Africa (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://pubmlst.org/bigsdb?db=pubmlst_campylobacter_isolates\u0026amp;page=query\u0026amp;genomes=1\u003c/span\u003e\u003cspan address=\"https://pubmlst.org/bigsdb?db=pubmlst_campylobacter_isolates\u0026amp;page=query\u0026amp;genomes=1\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e, accessed December 12, 2024).\u003c/p\u003e \u003cp\u003e \u003cem\u003eC. jejuni\u003c/em\u003e ST50 isolates have been frequently reported globally from environmental, food and clinical sources. Poultry isolates from Oceania, Europe and North America tended to cluster on the basis of the continent where the sample was collected \u003csup\u003e\u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e46\u003c/span\u003e\u003c/sup\u003e. No isolates from Africa were included in this study. Other frequent STs in our study are ST883, ST19, ST20242 and ST2031. The two former STs were found in infants and goats in our study, have previously been found in humans (children and mothers) in Ethiopia \u003csup\u003e\u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e47\u003c/span\u003e\u003c/sup\u003e and have been associated with a raw cow\u0026rsquo;s milk outbreak in Finland (ST883) \u003csup\u003e\u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e48\u003c/span\u003e\u003c/sup\u003e and Denmark (ST19)\u003csup\u003e\u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e49\u003c/span\u003e\u003c/sup\u003e, respectively.\u003c/p\u003e \u003cp\u003eSeveral studies have detected a high prevalence of \u003cem\u003eC. jejuni\u003c/em\u003e in human, livestock and food samples from Ethiopia, but only two studies have included the genomic characterization of isolates. Among 14 isolates from dairy in the Ethiopian regions of Amhara, Oromia, and the Southern Nations, Nationalities, and Peoples (SNNPR) between 2020 and 2021, two STs (ST51 and ST2084) were detected \u003csup\u003e\u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e50\u003c/span\u003e\u003c/sup\u003e; neither of these types was identified in our study. Another study from the Harar town and Kersa district identified 8 distinct STs for 19 Campylobacter strains isolated from children, caretakers and potential exposure sources \u003csup\u003e\u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e47\u003c/span\u003e\u003c/sup\u003e. The most abundant STs were ST353, ST19 and ST1365, all of which were also found in this study. French et al. \u003csup\u003e\u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e51\u003c/span\u003e\u003c/sup\u003e have studied the genomic structure of poultry isolates from Tanzania and human isolates from Kenya, and reported that ST353, ST8043, ST2122 and ST1932 were the dominant STs (in this order). Of these, only ST353 was identified in the current study as one of the twelve most frequently occurring types. In both studies, these types were isolated from both humans and livestock. These authors also tabulated STs as identified in other studies from Africa. Among the twelve most common STs identified in our study, ST362 was identified frequently in South Africa, whereas some other STs occurred sporadically in the database. While the current results suggest little overlap between types in geographically adjacent regions, the genomic diversity of \u003cem\u003eC. jejuni\u003c/em\u003e and other foodborne pathogens in Africa is largely uncharacterized. More generally, WGS-based surveillance is poorly developed in low-resource countries\u003csup\u003e\u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e45\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eOur isolate collection included several samples from which multiple isolates were obtained, offering a unique opportunity to assess genomic diversity in single hosts. Among stool samples from infants, we detected up to four isolates with the same seven-gene ST and only a few samples with more than one ST. Within one ST, cgSTs had fewer than five alleles difference. Bloomfield et al. \u003csup\u003e\u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e52\u003c/span\u003e\u003c/sup\u003e studied sixteen isolates from an immunodeficient patient who was colonized by \u003cem\u003eC. jejuni\u003c/em\u003e for 10 years after an episode of diarrhea. All isolates shared a common ancestor, coinciding with the onset of symptoms for the patient and evidence was found for genetic bottlenecks due to antimicrobial treatment. Djeghout et al. \u003csup\u003e\u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e53\u003c/span\u003e\u003c/sup\u003e sequenced ninety-two \u003cem\u003eC. jejuni\u003c/em\u003e isolates from four different patients with gastroenteritis and used Single Nucleotide Polymorphisms (SNP) to assess phylogenetic relationships. Three patients yielded a single seven-gene ST, whereas one patient yielded two different STs. Isolates from one patient were genetically diverse, even within one ST (12\u0026ndash;43 core non-recombinant SNPs and 0\u0026ndash;20 frame-shifts). These authors concluded that this diverse population was unlikely to have evolved from a single isolate at the time point of initial patient infection and that patients were likely infected with a heterogeneous \u003cem\u003eC. jejuni\u003c/em\u003e population. However, even upon exposure to a heterogeneous population, multiple barriers in the host create a strong evolutionary bottleneck, resulting in the selection of one single variant causing colonization except when the host is exposed to high doses \u003csup\u003e\u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e54\u003c/span\u003e\u003c/sup\u003e. We therefore suggest that in our population of young infants, infections are caused mainly by a single transmission event, followed by microevolution within the host. Further phylogenetic analysis of our isolates including Average Nucleotide Identity and Single Nucleotide Polymorphisms may provide more detailed information but was beyond the scope of this study.\u003c/p\u003e \u003cp\u003eA reanalysis of data from the MAL-ED birth cohort study suggested that persistent infections with \u003cem\u003eCampylobacter\u003c/em\u003e were associated with poorer 9-month linear growth. Persistent infections were defined as three or more consecutive \u003cem\u003eCampylobacter\u003c/em\u003e positive monthly stools by qPCR \u003csup\u003e\u003cspan citationid=\"CR55\" class=\"CitationRef\"\u003e55\u003c/span\u003e\u003c/sup\u003e. However, in the main text the authors state that \u0026ldquo;persistent Campylobacter infections cannot be differentiated from recurrent reinfections with epidemiologic data alone\u0026rdquo; and suggest the term persistent carriage would better describe the qPCR results. Because of the longitudinal nature of our study, we were able to evaluate temporal patterns of genotype diversity in infants. Repeat isolates of the same ST at two different time points (one or two months apart) were observed but in most cases, we observed different STs. This suggests a dominant pattern of clearance of one type and new infection by other types within a time interval of one to four months, which is consistent with findings from a Markov Chain model applied to the same MAL-ED dataset \u003csup\u003e\u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e56\u003c/span\u003e\u003c/sup\u003e. These findings are further supported by PERMANOVA, indicating a high level of clustering of isolates from the same sample, but no significant clustering of isolates from the same infant at different time points. The clustering at the kebele level suggests that there is localized, within kebele transmission which is independent of the clustering observed within samples and between samples from the same individual over time.\u003c/p\u003e \u003cp\u003eWe used two different methods (cgMLST and \u003cem\u003ek\u003c/em\u003e-merization) to characterize the genomic diversity of our isolates and two types of models (population genetics and machine learning) to attribute infections in infants to putative sources. Chickens were the main source of infection with model uncertainty in the proportion of infections attributed to cattle or small ruminants. When other humans are included as possible sources, attribution to livestock decreases but chickens remain the main source. Neither the Asymmetric Island method using cgMLST nor the random forest methods using cgMLST or \u003cem\u003ek\u003c/em\u003e-mers can determine directionality. The assumption is that there is unidirectional flow from any of the included sources to the receiving host (sink). The \u0026ldquo;source-sink\u0026rdquo; model relaxes this assumption in the sense that some compartments are considered both as sources and as sinks. This model was originally applied to evaluate the role of water as a sink for contamination by water birds and as a source for human infections\u003csup\u003e\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e\u003c/sup\u003e. We applied this model to study the role of other humans as sinks from infections from livestock and as sources for infections in infants. The model suggested that other humans are largely infected by chickens, and that these infections may be transmitted to infants. We cannot conclude whether other humans act as mechanical vectors or if they are amplifying hosts.\u003c/p\u003e \u003cp\u003eOur source attribution results may be biased towards chickens because of the substantially greater number of isolates than those from other livestock and human sources. For example, isolates from goats and siblings frequently occurred among the top twelve STs but less frequently among rare STs. Nevertheless, the use of high-resolution genome sequencing data, and the similar findings from multiple models based on different underlying assumptions, suggest that any bias attributed to unbalanced sampling is unlikely to affect the conclusion that chickens are the most important animal reservoir for infant infections in this study. Nevertheless, as a more important role of ruminants cannot be ruled out, control strategies aimed at eliminating transmission from the chicken reservoir only may not be effective. Additionally, the degree of transmission between chickens and ruminants is unknown but is likely to occur, particularly in areas where the animals are not confined to barns or other housing facilities.\u003c/p\u003e \u003cp\u003eNotable, most of the isolates were from older infants, because of increasing prevalence with age. Furthermore, this study was undertaken during the global COVID-19 pandemic and early samples could not be cultured because of global supply chain issues and samples were preserved by freezing. Even though we stored the samples in glycerol, recovery was strongly affected. Additionally, species-specific qPCR results indicated that \u003cem\u003eCandidatus C. infans\u003c/em\u003e is more prevalent in infants than \u003cem\u003eC. jejuni\u003c/em\u003e is (70% at one year of age). This species has mainly been detected in other humans, with low levels of detection in livestock\u003csup\u003e\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e\u003c/sup\u003e, suggesting this species is anthroponotic in nature with occurrence in livestock as a reverse zoonosis, or even merely passing through the animal gut from a highly contaminated environment, where open defecation is common \u003csup\u003e\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eThe transmission pathways of \u003cem\u003eC. jejuni\u003c/em\u003e in our study area are highly complex and interdependent. The EXCAM study employed behavioral observations, microbiological analysis and mathematical modeling to create an agent-based exposure model framework to quantify the exposure to generic \u003cem\u003eEscherichia coli\u003c/em\u003e through different pathways in the first and second half years of life of the infants included in this study. The major sources of exposure to \u003cem\u003eE. coli\u003c/em\u003e were food and breastfeeding in the first half year of life and food and soil in the second half year of life. Caretakers\u0026rsquo; hands are the main sources of contamination of both food and breast surfaces \u003csup\u003e\u003cspan citationid=\"CR57\" class=\"CitationRef\"\u003e57\u003c/span\u003e\u003c/sup\u003e. Considering all the evidence, we conclude that efforts to reduce the colonization of infants with \u003cem\u003eCampylobacter\u003c/em\u003e in the study area, and most likely in many other similar settings, are best aimed at protecting proximate sources such as caretakers\u0026rsquo; hands, food and indoor soil. Contributing to the ultimate goal of reducing stunting through controlling exposure of infants and young children to enteric pathogens calls for a tight integration of currently siloed domains of nutrition, food safety and water, sanitation and hygiene.\u003c/p\u003e"},{"header":"Conclusions","content":"\u003cp\u003eMany of the \u003cem\u003eC. jejuni\u003c/em\u003e isolates identified in this study do not belong to the most common STs reported in high-resource settings. Among the six most common global STs, only one was found in our study area. Isolates from the same infant sample were highly related, isolates from consecutive infant samples usually had different STs, suggesting rapid clearance and new infection. The transmission pathways of \u003cem\u003eC. jejuni\u003c/em\u003e in our study area are highly complex and interdependent. While chickens are the most important reservoir of \u003cem\u003eC. jejuni\u003c/em\u003e in infants, ruminant reservoirs also contribute to the infections. Model predictions differed in terms of the relative importance of cattle vs. small ruminants as additional sources. Infections from chickens are transmitted with or without other humans (mothers, siblings) as intermediate sources. To reduce the colonization of infants with \u003cem\u003eCampylobacter\u003c/em\u003e and ultimately mitigate stunting in low-resource settings, the best approach is to protectproximate sources. This includes ensuring the cleanliness of caretakers\u0026rsquo; hands food and indoor soil through tight integration of the currently siloed domains of nutrition, food safety and water, sanitation and hygiene\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003cp\u003eCaC Chromagar Campylobacter\u003c/p\u003e\n\u003cp\u003ecgMLST Core genome Multi Locus Sequence Typing\u003c/p\u003e\n\u003cp\u003ecgST Core genome Sequence Type\u003c/p\u003e\n\u003cp\u003e\u003cem\u003ek\u003c/em\u003e-mer Substring of length k contained within a DNA sequence\u003c/p\u003e\n\u003cp\u003ePCO Ordinal encoding of predictor variables using Principal Components\u003c/p\u003e\n\u003cp\u003eMST Minimum Spanning Tree\u003c/p\u003e\n\u003cp\u003ePERMANOVA Nested, permutational multivariate analysis of variance\u003c/p\u003e\n\u003cp\u003ePubMLST Public databases for molecular typing and microbial genome diversity\u003c/p\u003e\n\u003cp\u003eqPCR Quantitative (real-time) Polymerase Chain Reaction\u003c/p\u003e\n\u003cp\u003eSNP Single Nucleotide Polymorphism\u003c/p\u003e\n\u003cp\u003eST Legacy (seven-gene) Multi Locus Sequence Type\u003c/p\u003e"},{"header":"Declarations","content":"\u003ch2\u003eEthical approval and consent to participate\u003c/h2\u003e\n\u003cp\u003eEthical approval was obtained from the University of Florida Internal Review Board (IRB201903141) and the Ethiopia National Research Ethics Review Committee (SM/14.1/1059/20). Written informed consent was obtained from all participating households (husband and wife), with mothers consenting to infant participation, using forms in the local language (Afan Oromo).\u003c/p\u003e\n\u003ch2\u003eConsent for publication\u003c/h2\u003e\n\u003cp\u003eNot applicable\u003c/p\u003e\n\u003ch2\u003eAvailability of data\u003c/h2\u003e\n\u003cp\u003eAll whole-genome sequences supporting the conclusions of this article are available under Bioproject PRJNA1015272 at https://www.ncbi.nlm.nih.gov/bioproject/?term=PRJNA1015272. Biosample IDs for selected \u003cem\u003eCampylobacter jejuni\u003c/em\u003e isolates, post-processing, are provided in the supplementary file MLST_profiles.xlsx. Code on GitHub is indicated in the text.\u003c/p\u003e\n\u003ch2\u003eCompeting interests\u003c/h2\u003e\n\u003cp\u003eThe authors declare no competing interests.\u003c/p\u003e\n\u003ch2\u003eFunding\u003c/h2\u003e\n\u003cp\u003eThis project was funded by the United States Agency for International Development Bureau for Food Security under Agreement #AID-OAA-L-15-00003 as part of Feed the Future Innovation Lab for Livestock Systems and by the Bill \u0026amp; Melinda Gates Foundation OPP#1175487. Under the grant conditions of the Foundation, a Creative Commons Attribution 4.0 Generic License has already been assigned to the Author Accepted Manuscript version that might arise from this submission. Any opinions, findings, conclusions, or recommendations expressed here are those of the authors alone.\u003c/p\u003e\n\u003ch2\u003eAuthor contributions\u003c/h2\u003e\n\u003cp\u003eConceptualization: NPF, JYH, AHH, MJM, SLM, GR. Data curation: LD, BMH, AO, NS, CANT. Formal Analysis: NS, CANT, NPF, JCM, HLS, AHH. Funding acquisition: TH, JYH, AHH, SLM, GR. Investigation: BHM. Methodology: NPF, TMH, JCM, HLS. Project administration: JYH, AHH, SLM. Resources: NPF, TMH, JYH, AHH, GR. Software: JCM, HLS. Supervision: NPF, TMH, JYH, AHH, SLM, GR. Validation: NPF, TMH, AHH. Visualization: NPF, AHH, NS. Writing\u0026ndash;original draft: NPF, TMH, AHH, CANT, NS. Writing\u0026ndash;review \u0026amp; editing: All. All authors approved the final version for submission.\u003c/p\u003e\n\u003ch2\u003eAcknowledgments\u003c/h2\u003e\n\u003cp\u003eThe CAGED Research Team Members consisted of Abadir Jemal Seran, Abdulmuen Mohammed Ibrahim, Amanda E. Ojeda, Bahar Mummed Hassen, Belisa Usmael Ahmedo, Cyrus Saleem, Dehao Chen, Efrah Ali Yusuf, Getnet Yimer, Ibsa Abdusemed Ahmed, Ibsa Aliyi Usmane, Jafer Kedir Amin, Kedir Abdi Hassen, Kedir Teji Roba, Kunuza Adem Umer, Karah Mechlowitz, Loic Deblais, Mahammad Mahammad Usmail, Mark J. Manary, Mawardi M. Dawid, Mussie Bhrane, Nur Shaikh, Wondwossen Gebreyes, Xiaolong Li, Yang Yang, Yenenesh Demisie Weldesenbet, Zelalem Hailu Mekuria.\u003c/p\u003e\n\u003cp\u003eWe thank Tina Lusk Pfeffer, Kelli Hiett, Kannan Balan, Hyein Jang, Marianne Sawyer (Office of Applied Research and Safety Assessment), and Ruth Timme (Office of Regulatory Science, Division of Microbiology), Center for Food Safety and Applied Nutrition, Food and Drug Administration, USA) for their support in acquiring funding and project management for Whole Genome Sequencing. We acknowledge the sequencing efforts by the GenomeTrakr laboratories at the Colorado State Department of Public Health and Environment , New Hampshire Department of Health and Human Services / Public Health Laboratories, Virginia Division of Consolidated Laboratory Services, Minnesota Department of Health, Missouri State Public Health Laboratory, Maryland Department of Health, Ohio Department of Agriculture /Animal Disease Diagnostic Laboratory, Rhode Island Department of Health.\u003c/p\u003e\n\u003cp\u003eThis publication made use of the PubMLST website (https://pubmlst.org/) developed by Keith Jolley and cited at the University of Oxford. The development of that website was funded by the Wellcome Trust. We thank Xiaolong Li for assistance with mapping the sequence types. Map data copyrighted OpenStreetMap contributors and available from https://www.openstreetmap.org.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n \u003cli\u003eRogawski, E. T. \u003cem\u003eet al.\u003c/em\u003e Use of quantitative molecular diagnostic methods to investigate the effect of enteropathogen infections on linear growth in children in low-resource settings: longitudinal analysis of results from the MAL-ED cohort study. \u003cem\u003eThe Lancet Global Health\u003c/em\u003e \u003cstrong\u003e6\u003c/strong\u003e, e1319\u0026ndash;e1328 (2018).\u003c/li\u003e\n \u003cli\u003eHendrickson, S. M. \u003cem\u003eet al.\u003c/em\u003e Campylobacter vaccination reduces diarrheal disease and infant growth stunting among rhesus macaques. \u003cem\u003eNat Commun\u003c/em\u003e \u003cstrong\u003e14\u003c/strong\u003e, 3806 (2023).\u003c/li\u003e\n \u003cli\u003eDeblais, L. \u003cem\u003eet al.\u003c/em\u003e Prevalence and Load of the Campylobacter Genus in Infants and Associated Household Contacts in Rural Eastern Ethiopia: a Longitudinal Study from the Campylobacter Genomics and Environmental Enteric Dysfunction (CAGED) Project. \u003cem\u003eApplied and Environmental Microbiology\u003c/em\u003e \u003cstrong\u003e89\u003c/strong\u003e, e00424-23 (2023).\u003c/li\u003e\n \u003cli\u003eChen, D. \u003cem\u003eet al.\u003c/em\u003e Campylobacter colonization and undernutrition in infants in rural Eastern Ethiopia: a longitudinal community-based birth cohort study. \u003cem\u003emedRxiv\u003c/em\u003e (2024) doi:10.1101/2024.05.21.24307707.\u003c/li\u003e\n \u003cli\u003eLi, X. \u003cem\u003eet al.\u003c/em\u003e Geospatial Analysis of Multilevel Socio-environmental Factors Impacting the Campylobacter Burden Among Infants in Rural Eastern Ethiopia: A One Health Perspective. 2024.07.03.24309853 Preprint at https://doi.org/10.1101/2024.07.03.24309853 (2024).\u003c/li\u003e\n \u003cli\u003eOjeda, A. \u003cem\u003eet al.\u003c/em\u003e Determinants of Campylobacter species diversity in infants and association with family members, livestock, and household environments in rural Eastern Ethiopia. \u003cem\u003eSubmitted to Gut Pathogens\u003c/em\u003e.\u003c/li\u003e\n \u003cli\u003eMekuria, Z. \u003cem\u003eet al.\u003c/em\u003e Host Clustering of Campylobacter Species and Other Enteric Pathogens in a Longitudinal Cohort of Infants, Family Members and Livestock in Rural Eastern Ethiopia. \u003cem\u003eSubmitted for publication\u003c/em\u003e.\u003c/li\u003e\n \u003cli\u003eHald, T. \u003cem\u003eet al.\u003c/em\u003e World Health Organization Estimates of the Relative Contributions of Food to the Burden of Disease Due to Selected Foodborne Hazards: A Structured Expert Elicitation. \u003cem\u003ePLoS One\u003c/em\u003e \u003cstrong\u003e11\u003c/strong\u003e, e0145839 (2016).\u003c/li\u003e\n \u003cli\u003eCody, A. J., Maiden, M. C., Strachan, N. J. \u0026amp; McCarthy, N. D. A systematic review of source attribution of human campylobacteriosis using multilocus sequence typing. \u003cem\u003eEurosurveillance\u003c/em\u003e \u003cstrong\u003e24\u003c/strong\u003e, 1800696 (2019).\u003c/li\u003e\n \u003cli\u003eVan Pelt, W. \u003cem\u003eet al.\u003c/em\u003e Origin, extent and costs of human salmonellosis. Part 1 Origin of human salmonellosis with respect to pig, cattle, chicken, eggs and other sources (in Dutch). \u003cem\u003eInfectieziekten Bulletin\u003c/em\u003e \u003cstrong\u003e10\u003c/strong\u003e,.\u003c/li\u003e\n \u003cli\u003eHald, T., Vose, D., Wegener, H. C. \u0026amp; Koupeev, T. A Bayesian Approach to Quantify the Contribution of Animal-Food Sources to Human Salmonellosis. \u003cem\u003eRisk Analysis\u003c/em\u003e \u003cstrong\u003e24\u003c/strong\u003e, 255\u0026ndash;269 (2004).\u003c/li\u003e\n \u003cli\u003eMullner, P. \u003cem\u003eet al.\u003c/em\u003e Source Attribution of Food-Borne Zoonoses in New Zealand: A Modified Hald Model. \u003cem\u003eRisk Analysis\u003c/em\u003e \u003cstrong\u003e29\u003c/strong\u003e, 970\u0026ndash;984 (2009).\u003c/li\u003e\n \u003cli\u003eWilson, D. J. \u003cem\u003eet al.\u003c/em\u003e Tracing the Source of Campylobacteriosis. \u003cem\u003ePLoS Genetics\u003c/em\u003e \u003cstrong\u003e4\u003c/strong\u003e, (2008).\u003c/li\u003e\n \u003cli\u003ePritchard, J. K., Stephens, M. \u0026amp; Donnelly, P. Inference of Population Structure Using Multilocus Genotype Data. \u003cem\u003eGenetics\u003c/em\u003e \u003cstrong\u003e155\u003c/strong\u003e, 945\u0026ndash;959 (2000).\u003c/li\u003e\n \u003cli\u003eArning, N., Sheppard, S. K., Bayliss, S., Clifton, D. A. \u0026amp; Wilson, D. J. Machine learning to predict the source of campylobacteriosis using whole genome data. \u003cem\u003ePLoS Genet\u003c/em\u003e \u003cstrong\u003e17\u003c/strong\u003e, e1009436 (2021).\u003c/li\u003e\n \u003cli\u003eBernard, G., Chan, C. X. \u0026amp; Ragan, M. A. Alignment-free microbial phylogenomics under scenarios of sequence divergence, genome rearrangement and lateral genetic transfer. \u003cem\u003eSci Rep\u003c/em\u003e \u003cstrong\u003e6\u003c/strong\u003e, 28970 (2016).\u003c/li\u003e\n \u003cli\u003ePanyukov, V. V., Kiselev, S. S. \u0026amp; Ozoline, O. N. Unique k-mers as Strain-Specific Barcodes for Phylogenetic Analysis and Natural Microbiome Profiling. \u003cem\u003eInternational Journal of Molecular Sciences\u003c/em\u003e \u003cstrong\u003e21\u003c/strong\u003e, 944 (2020).\u003c/li\u003e\n \u003cli\u003eCody, A. J., Bray, J. E., Jolley, K. A., McCarthy, N. D. \u0026amp; Maiden, M. C. J. Core Genome Multilocus Sequence Typing Scheme for Stable, Comparative Analyses of Campylobacter jejuni and C. coli Human Disease Isolates. \u003cem\u003eJ Clin Microbiol\u003c/em\u003e \u003cstrong\u003e55\u003c/strong\u003e, 2086\u0026ndash;2097 (2017).\u003c/li\u003e\n \u003cli\u003eZielezinski, A., Vinga, S., Almeida, J. \u0026amp; Karlowski, W. M. Alignment-free sequence comparison: benefits, applications, and tools. \u003cem\u003eGenome Biology\u003c/em\u003e \u003cstrong\u003e18\u003c/strong\u003e, 186 (2017).\u003c/li\u003e\n \u003cli\u003eMunck, N., Njage, P. M. K., Leekitcharoenphon, P., Litrup, E. \u0026amp; Hald, T. Application of Whole‐Genome Sequences and Machine Learning in Source Attribution of Salmonella Typhimurium. \u003cem\u003eRisk Analysis\u003c/em\u003e \u003cstrong\u003e40\u003c/strong\u003e, 1693 (2020).\u003c/li\u003e\n \u003cli\u003eSmith, H. L., Biggs, P. J., French, N. P., Smith, A. N. H. \u0026amp; Marshall, J. C. Lost in the Forest: Encoding categorical variables and the absent levels problem. \u003cem\u003eData Min Knowl Disc\u003c/em\u003e \u003cstrong\u003e38\u003c/strong\u003e, 1889\u0026ndash;1908 (2024).\u003c/li\u003e\n \u003cli\u003eLiao, S.-J. Statistical modelling for zoonotic diseases : a thesis presented in partial fulfilment of the requirements for the degree of Doctor of Philosophy in Statistics at Massey University, Palmerston North, New Zealand. (Massey University, 2020).\u003c/li\u003e\n \u003cli\u003eHavelaar, A. H. \u003cem\u003eet al.\u003c/em\u003e Unravelling the reservoirs for colonisation of infants with Campylobacter spp. in rural Ethiopia: protocol for a longitudinal study during a global pandemic and political tensions. \u003cem\u003eBMJ Open\u003c/em\u003e \u003cstrong\u003e12\u003c/strong\u003e, e061311 (2022).\u003c/li\u003e\n \u003cli\u003ePlatts-Mills, J. A. \u003cem\u003eet al.\u003c/em\u003e Detection of Campylobacter in Stool and Determination of Significance by Culture, Enzyme Immunoassay, and PCR in Developing Countries. \u003cem\u003eJ Clin Microbiol\u003c/em\u003e \u003cstrong\u003e52\u003c/strong\u003e, 1074\u0026ndash;1080 (2014).\u003c/li\u003e\n \u003cli\u003eDavedow, T. \u003cem\u003eet al.\u003c/em\u003e PulseNet International Survey on the Implementation of Whole Genome Sequencing in Low and Middle-Income Countries for Foodborne Disease Surveillance. \u003cem\u003eFoodborne Pathogens and Disease\u003c/em\u003e (2022) doi:10.1089/fpd.2021.0110.\u003c/li\u003e\n \u003cli\u003eBushnell, B., Rood, J. \u0026amp; Singer, E. BBMerge \u0026ndash; Accurate paired shotgun read merging via overlap. \u003cem\u003ePLOS ONE\u003c/em\u003e \u003cstrong\u003e12\u003c/strong\u003e, e0185056 (2017).\u003c/li\u003e\n \u003cli\u003eBBDuk Guide. \u003cem\u003eDOE Joint Genome Institute\u003c/em\u003e https://jgi.doe.gov/data-and-tools/software-tools/bbtools/bb-tools-user-guide/bbduk-guide/.\u003c/li\u003e\n \u003cli\u003eKokot, M., Długosz, M. \u0026amp; Deorowicz, S. KMC 3: counting and manipulating k-mer statistics. \u003cem\u003eBioinformatics\u003c/em\u003e \u003cstrong\u003e33\u003c/strong\u003e, 2759\u0026ndash;2761 (2017).\u003c/li\u003e\n \u003cli\u003eSeemann, T. mlst. https://github.com/tseemann/mlst.\u003c/li\u003e\n \u003cli\u003eJolley, K. A. \u0026amp; Maiden, M. C. BIGSdb: Scalable analysis of bacterial genome variation at the population level. \u003cem\u003eBMC Bioinformatics\u003c/em\u003e \u003cstrong\u003e11\u003c/strong\u003e, 595 (2010).\u003c/li\u003e\n \u003cli\u003eDingle, K. E. \u003cem\u003eet al.\u003c/em\u003e Multilocus Sequence Typing System forCampylobacter jejuni. \u003cem\u003eJournal of Clinical Microbiology\u003c/em\u003e \u003cstrong\u003e39\u003c/strong\u003e, 14\u0026ndash;23 (2001).\u003c/li\u003e\n \u003cli\u003eZomer, A. aldertzomer/cgmlst. (2022).\u003c/li\u003e\n \u003cli\u003eZhou, Z. \u003cem\u003eet al.\u003c/em\u003e GrapeTree: visualization of core genomic relationships among 100,000 bacterial pathogens. \u003cem\u003eGenome Res.\u003c/em\u003e \u003cstrong\u003e28\u003c/strong\u003e, 1395\u0026ndash;1404 (2018).\u003c/li\u003e\n \u003cli\u003eAnderson, M. J. Permutational Multivariate Analysis of Variance (PERMANOVA). in \u003cem\u003eWiley StatsRef: Statistics Reference Online\u003c/em\u003e 1\u0026ndash;15 (John Wiley \u0026amp; Sons, Ltd, 2017). doi:10.1002/9781118445112.stat07841.\u003c/li\u003e\n \u003cli\u003eOksanen, J. \u003cem\u003eet al.\u003c/em\u003e \u003cem\u003eVegan: Community Ecology Package\u003c/em\u003e. (2024).\u003c/li\u003e\n \u003cli\u003eR Core Team. \u003cem\u003eR: A Language and Environment for Statistical Computing\u003c/em\u003e. (R Foundation for Statistical Computing, Vienna, Austria, 2023).\u003c/li\u003e\n \u003cli\u003eWright, M. N. \u0026amp; Ziegler, A. ranger: A Fast Implementation of Random Forests for High Dimensional Data in C++ and R. \u003cem\u003eJournal of Statistical Software\u003c/em\u003e \u003cstrong\u003e77\u003c/strong\u003e, 1\u0026ndash;17 (2017).\u003c/li\u003e\n \u003cli\u003eKuhn \u0026amp; Max. Building Predictive Models in R Using the caret Package. \u003cem\u003eJournal of Statistical Software\u003c/em\u003e \u003cstrong\u003e28\u003c/strong\u003e, 1-26 (2008).\u003c/li\u003e\n \u003cli\u003eKursa, M. B. \u0026amp; Rudnicki, W. R. Feature Selection with the Boruta Package. \u003cem\u003eJournal of Statistical Software\u003c/em\u003e \u003cstrong\u003e36\u003c/strong\u003e, 1\u0026ndash;13 (2010).\u003c/li\u003e\n \u003cli\u003eBrinch, M. L. \u003cem\u003eet al.\u003c/em\u003e Comparison of Source Attribution Methodologies for Human Campylobacteriosis. \u003cem\u003ePathogens\u003c/em\u003e \u003cstrong\u003e12\u003c/strong\u003e, 786 (2023).\u003c/li\u003e\n \u003cli\u003eNjage, P. M. K., Leekitcharoenphon, P. \u0026amp; Hald, T. Improving hazard characterization in microbial risk assessment using next generation sequencing data and machine learning: Predicting clinical outcomes in shigatoxigenic \u003cem\u003eEscherichia coli\u003c/em\u003e. \u003cem\u003eInternational Journal of Food Microbiology\u003c/em\u003e \u003cstrong\u003e292\u003c/strong\u003e, 72\u0026ndash;82 (2019).\u003c/li\u003e\n \u003cli\u003eOgutu, J. O., Piepho, H.-P. \u0026amp; Schulz-Streeck, T. A comparison of random forests, boosting and support vector machines for genomic selection. \u003cem\u003eBMC Proceedings\u003c/em\u003e \u003cstrong\u003e5\u003c/strong\u003e, S11 (2011).\u003c/li\u003e\n \u003cli\u003eMachado, G., Mendoza, M. R. \u0026amp; Corbellini, L. G. What variables are important in predicting bovine viral diarrhea virus? A random forest approach. \u003cem\u003eVeterinary Research\u003c/em\u003e \u003cstrong\u003e46\u003c/strong\u003e, 85 (2015).\u003c/li\u003e\n \u003cli\u003ePoorrashidi, M., Hitchcock, M. \u0026amp; Xu, J. Meta-analyses of the global multilocus genotypes of the human pathogen Campylobacter jejuni. \u003cem\u003eGenome\u003c/em\u003e \u003cstrong\u003e67\u003c/strong\u003e, 189\u0026ndash;203 (2024).\u003c/li\u003e\n \u003cli\u003eMather, A. E., Gilmour, M. W., Reid, S. W. J. \u0026amp; French, N. P. Foodborne bacterial pathogens: genome-based approaches for enduring and emerging threats in a complex and changing world. \u003cem\u003eNat Rev Microbiol\u003c/em\u003e (2024) doi:10.1038/s41579-024-01051-z.\u003c/li\u003e\n \u003cli\u003eWallace, R. L. \u003cem\u003eet al.\u003c/em\u003e Campylobacter jejuni ST50, a pathogen of global importance: A comparative genomic analysis of isolates from Australia, Europe and North America. \u003cem\u003eZoonoses and Public Health\u003c/em\u003e \u003cstrong\u003e68\u003c/strong\u003e, 638\u0026ndash;649 (2021).\u003c/li\u003e\n \u003cli\u003eBelina, D. \u003cem\u003eet al.\u003c/em\u003e Occurrence and diversity of Campylobacter species in diarrheic children and their exposure environments in Ethiopia. \u003cem\u003ePLOS Glob Public Health\u003c/em\u003e \u003cstrong\u003e4\u003c/strong\u003e, e0003885 (2024).\u003c/li\u003e\n \u003cli\u003eJaakkonen, A., Kivist\u0026ouml;, R., Aarnio, M., Kalekivi, J. \u0026amp; Hakkinen, M. Persistent contamination of raw milk by Campylobacter jejuni ST-883. \u003cem\u003ePLoS One\u003c/em\u003e \u003cstrong\u003e15\u003c/strong\u003e, e0231810 (2020).\u003c/li\u003e\n \u003cli\u003eJoensen, K. G. \u003cem\u003eet al.\u003c/em\u003e Whole-Genome Sequencing to Detect Numerous Campylobacter jejuni Outbreaks and Match Patient Isolates to Sources, Denmark, 2015\u0026ndash;2017 - Volume 26, Number 3\u0026mdash;March 2020 - Emerging Infectious Diseases journal - CDC. \u003cem\u003eEmerging Infectious Diseases\u003c/em\u003e \u003cstrong\u003e26\u003c/strong\u003e, 523\u0026ndash;532 (2020).\u003c/li\u003e\n \u003cli\u003eAdmasie, A. \u003cem\u003eet al.\u003c/em\u003e Genomic diversity of Campylobacter jejuni and Campylobacter coli isolated from the Ethiopian dairy supply chain. \u003cem\u003ePLOS ONE\u003c/em\u003e \u003cstrong\u003e19\u003c/strong\u003e, e0305581 (2024).\u003c/li\u003e\n \u003cli\u003eFrench, N. P. \u003cem\u003eet al.\u003c/em\u003e Population Structure and Antimicrobial Resistance in Campylobacter jejuni and C. coli Isolated from Humans with Diarrhea and from Poultry, East Africa - Volume 30, Number 10\u0026mdash;October 2024 - Emerging Infectious Diseases journal - CDC. \u003cem\u003eEmerging Infectious Diseases\u003c/em\u003e \u003cstrong\u003e30\u003c/strong\u003e, 2079\u0026ndash;2089 (2024).\u003c/li\u003e\n \u003cli\u003eBloomfield, S. J. \u003cem\u003eet al.\u003c/em\u003e Long-term Colonization by Campylobacter jejuni Within a Human Host: Evolution, Antimicrobial Resistance, and Adaptation. \u003cem\u003eThe Journal of Infectious Diseases\u003c/em\u003e \u003cstrong\u003e217\u003c/strong\u003e, 103\u0026ndash;111 (2018).\u003c/li\u003e\n \u003cli\u003eDjeghout, B. \u003cem\u003eet al.\u003c/em\u003e Comparative genomics of Campylobacter jejuni from clinical campylobacteriosis stool specimens. \u003cem\u003eGut Pathogens\u003c/em\u003e \u003cstrong\u003e14\u003c/strong\u003e, 45 (2022).\u003c/li\u003e\n \u003cli\u003eMoxon, E. R. \u0026amp; Murphy, P. A. Haemophilus influenzae bacteremia and meningitis resulting from survival of a single organism. \u003cem\u003eProceedings of the National Academy of Sciences\u003c/em\u003e \u003cstrong\u003e75\u003c/strong\u003e, 1534\u0026ndash;1536 (1978).\u003c/li\u003e\n \u003cli\u003eSchiaffino, F. \u003cem\u003eet al.\u003c/em\u003e The epidemiology and impact of persistent \u003cem\u003eCampylobacter\u003c/em\u003e infections on childhood growth among children 0\u0026ndash;24 months of age in resource-limited settings. \u003cem\u003eeClinicalMedicine\u003c/em\u003e \u003cstrong\u003e76\u003c/strong\u003e, 102841 (2024).\u003c/li\u003e\n \u003cli\u003eChen, D., Havelaar, A. H., Platts-Mills, J. A. \u0026amp; Yang, Y. Acquisition and clearance dynamics of \u003cem\u003eCampylobacter\u003c/em\u003e spp. in children in low- and middle-income countries. \u003cem\u003eEpidemics\u003c/em\u003e \u003cstrong\u003e46\u003c/strong\u003e, 100749 (2024).\u003c/li\u003e\n \u003cli\u003eWang, Y. \u003cem\u003eet al.\u003c/em\u003e Quantitative Multi-pathway Assessment of Exposure to Fecal Contamination for Infants in Rural Ethiopia. 2024.08.29.24312786 Preprint at https://doi.org/10.1101/2024.08.29.24312786 (2024).\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"gut-pathogens","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"gutp","sideBox":"Learn more about [Gut Pathogens](http://gutpathogens.biomedcentral.com/)","snPcode":"13099","submissionUrl":"https://submission.nature.com/new-submission/13099/3","title":"Gut Pathogens","twitterHandle":"@GutPathogens","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"BMC/SO AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Campylobacter jejuni, Attribution, Transmission Pathways, Zoonosis, Diversity, Persistence, Spatial Distribution, Sequencing Typing","lastPublishedDoi":"10.21203/rs.3.rs-5735672/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-5735672/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eBackground\u003c/h2\u003e \u003cp\u003e \u003cem\u003eCampylobacter jejuni\u003c/em\u003e and \u003cem\u003eC. coli\u003c/em\u003e are the most common causes of bacterial enteritis worldwide whereas symptomatic and asymptomatic infections are associated with stunting in children in low- and middle-income countries. Little is known about their sources and transmission pathways in low- and middle-income countries, and particularly for infants and young children. We assessed the genomic diversity of \u003cem\u003eC. jejuni\u003c/em\u003e in Eastern Ethiopia to determine the attribution of infections in infants under 1 year of age to livestock (chickens, cattle, goats and sheep) and other humans (siblings, mothers).\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e \u003cp\u003eAmong 287 \u003cem\u003eC. jejuni\u003c/em\u003e isolates, 48 seven-gene sequence types (STs), including 11 previously unreported STs were identified. Within an ST, the core genome STs of multiple isolates differed in fewer than five alleles. Many of these isolates do not belong to the most common STs reported in high-resource settings, and of the six most common global STs, only ST50 was found in our study area. Isolates from the same infant sample were closely related, while those from consecutive infant samples often displayed different STs, suggesting rapid clearance and new infection. Four different attribution models using different genomic profiling methods, assumptions and estimation methods predicted that chickens are the primary reservoir for infant infections. Infections from chickens are transmitted with or without other humans (mothers, siblings) as intermediate sources Model predictions differed in terms of the relative importance of cattle vs. small ruminants as additional sources.\u003c/p\u003e\u003ch2\u003eConclusions\u003c/h2\u003e \u003cp\u003eThe transmission pathways of \u003cem\u003eC. jejuni\u003c/em\u003e in our study area are highly complex and interdependent. While chickens are the most important reservoir of \u003cem\u003eC. jejuni\u003c/em\u003e, ruminant reservoirs also contribute to the infections. The currently nonculturable species \u003cem\u003eCandidatus\u003c/em\u003e C. infans is also highly prevalent in infants and is likely anthroponotic. Efforts to reduce the colonization of infants with \u003cem\u003eCampylobacter\u003c/em\u003e and ultimately stunting in low-resource settings are best aimed at protecting proximate sources such as caretakers\u0026rsquo; hands, food and indoor soil through tight integration of the currently siloed domains of nutrition, food safety and water, sanitation and hygiene.\u003c/p\u003e","manuscriptTitle":"Transmission pathways of Campylobacter jejuni between humans and livestock in rural Ethiopia are highly complex and interdependent","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-01-01 09:25:38","doi":"10.21203/rs.3.rs-5735672/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2025-03-05T11:18:53+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-02-22T08:07:36+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"33346352647386268206103503336950645693","date":"2025-01-11T13:51:24+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"245269172645779177915777610269469393303","date":"2025-01-04T09:22:15+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-01-02T09:05:35+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2024-12-30T13:01:42+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2024-12-30T13:00:11+00:00","index":"","fulltext":""},{"type":"submitted","content":"Gut Pathogens","date":"2024-12-30T12:48:57+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"gut-pathogens","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"gutp","sideBox":"Learn more about [Gut Pathogens](http://gutpathogens.biomedcentral.com/)","snPcode":"13099","submissionUrl":"https://submission.nature.com/new-submission/13099/3","title":"Gut Pathogens","twitterHandle":"@GutPathogens","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"BMC/SO AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"ec14f274-8084-447d-bc92-b5d29936760e","owner":[],"postedDate":"January 1st, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[],"tags":[],"updatedAt":"2025-05-05T15:59:43+00:00","versionOfRecord":{"articleIdentity":"rs-5735672","link":"https://doi.org/10.1186/s13099-025-00691-7","journal":{"identity":"gut-pathogens","isVorOnly":false,"title":"Gut Pathogens"},"publishedOn":"2025-05-03 15:56:59","publishedOnDateReadable":"May 3rd, 2025"},"versionCreatedAt":"2025-01-01 09:25:38","video":"","vorDoi":"10.1186/s13099-025-00691-7","vorDoiUrl":"https://doi.org/10.1186/s13099-025-00691-7","workflowStages":[]},"version":"v1","identity":"rs-5735672","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-5735672","identity":"rs-5735672","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}