Aldert Zomer

Background: Campylobacter is the leading cause of bacterial gastroenteritis, with retail chicken meat representing a major source of infection. Campylobacter isolation from retail meat can be difficult due to the complex growth requirements of the organism. Metagenomic sequencing may be a useful alternative approach for the direct identification of pathogens like Campylobacter from food samples, while also allowing characterisation of the overall sample microbiome. Results: Here, metagenomes of 58 retail chicken samples cultured for Campylobacter were assessed. Campylobacter relative abundance within the metagenome reads varied between samples (0.00086-0.47%), with overlap between culture positive and negative samples. Characterisation of Campylobacter to sequence type (ST) level was possible in one sample. There were no significant differences in alpha or beta diversity of taxa between the Campylobacter culture positive and negative samples, and individual differentially abundant taxa were not identified using ALDEx2 based on Benjamini-Hochberg adjusted Wilcoxon p-values. Preliminary analysis of metabolic reaction relative abundances indicated that several individual reactions differed in abundance between the Campylobacter culture positive and negative samples. However, these were found to not be statistically significant following multiple testing correction, nor through beta diversity and ALDEx2 analysis. Differences in taxa diversity metrics between chicken sample cuts and types were observed, which may be relevant for associations with Campylobacter but a larger study is required to investigate this fully. Conclusions: Metagenomic analysis of the retail chicken meat microbiome indicates that the identification of Campylobacter and microbial signatures associated with Campylobacter presence and absence on retail chicken meat is difficult, at least with current methods. Alternatively, such signatures may be absent altogether. While the small sample size in this study may have restricted our ability to identify important taxa or metabolic reactions, we outline an approach that future studies can use to investigate differences in microbial communities.

Escherichia coli is a highly diverse bacterial species that includes avian pathogenic E. coli (APEC), one of the most prevalent causative agents of disease in poultry worldwide. Rapid and accurate discrimination of E. coli strains is essential for outbreak management, antimicrobial resistance surveillance, and vaccine development. In this study, we compared the performance of Fourier Transform Infrared (FTIR) spectroscopy using the IR Biotyper system with Nanopore and Illumina whole-genome sequencing (WGS) for typing 200 E. coli isolates, originating from four poultry rearing farms in the Netherlands. From each farm, we sampled 10 one-day-old meat type rearing chicks, and from every chick, we isolated 5 E. coli strains. FTIR clustering showed strong concordance with WGS-based classifications, particularly serotyping and core-genome similarity determined by PopPUNK analysis (Adjusted Rand Index 0.75–0.92). While Nanopore and Illumina sequencing provided the highest genetic resolution, FTIR offered a faster (max 6 vs 12–28 days for 200 isolates) and more cost-effective alternative for assessing clonality. Across all methods, multiple strains were detected per farm, whereas most birds carried a single dominant E. coli strain. Our findings demonstrate that FTIR provides a reliable and scalable phenotypic method for rapid strain discrimination in E. coli , complementing WGS in diagnostic, surveillance, and epidemiological settings where speed and throughput are critical. IMPORTANCE: Escherichia coli is a major pathogen in poultry and a potential zoonotic risk for humans. Rapid and accurate discrimination of avian pathogenic E. coli (APEC) strains is critical for outbreak management, antimicrobial resistance surveillance, and the design of effective autogenous vaccines. In this study, we compared Fourier Transform Infrared (FTIR) spectroscopy with Nanopore and Illumina whole-genome sequencing for strain typing of E. coli isolates originating from poultry. The results show that FTIR provides comparable clustering accuracy to genomic approaches at a fraction of the time and costs. This work demonstrates that FTIR can serve as a practical, high-throughput alternative for routine monitoring of E. coli in veterinary diagnostics and food safety of poultry meat, enabling faster decision-making and more targeted interventions across the poultry production chain.

Background: Extraintestinal pathogenic Escherichia coli (ExPEC) can colonize and infect body sites outside the intestines, which can lead to invasive E. coli disease potentially resulting in sepsis or death. We evaluated O-serotype distribution, antimicrobial resistance profiles, and sequence types (STs) of ExPEC bloodstream isolates collected during 17 global surveillance studies conducted from 2011 through 2023. Methods: E. coli clinical blood isolates from an adult population were analyzed for O-serotype, ST, and multidrug-resistance (MDR) using data obtained by agglutination, whole-genome sequencing, and determination of minimal inhibitory concentrations of antimicrobial drugs. Results: Globally, 10 098 E. coli isolates were collected; 4925 (48.8%) from men and 5136 (50.9%) from women, median (range) age was 72 (18–104) years. The most prevalent ExPEC O-serotypes were O25 (18.7%), O2 (8.2%), O6 (8.2%), and O1 (7.0%). Highest MDR rates were observed in India (83.3%), Mexico (76.6%), and China (69.4%); lowest rates were in Sweden (10.6%), New Zealand (11.6%), and Netherlands (12.1%). Among MDR isolates (n=3343), serotype O25 was the most prevalent in all countries (range: 11.3% China to 56.2% Mexico). We observed a resistance rate of 1.1% and 0.7% to last-resort carbapenem and colistin antibiotics, respectively, distributed over diverse STs and O-serotypes. Conclusions: Overall, serotypes O25, O2, O6, and O1 were the most prevalent. MDR levels showed large regional variation. O25 was dominant within the subset of MDR isolates across all countries. O153-ST648, O101/O162-ST744, O102-ST405, and O25-ST131 were associated with MDR.

Motivation: Oxford Nanopore sequencing enables long-read analysis for diverse applications, but artefacts introduced by Nanopore barcoding are poorly characterized and can compromise demultiplexing accuracy and downstream analyses. Results: Using a rapid barcoding experiment on 66 diagnostic samples, we found that only 83% of reads followed the expected single-barcode configuration, while 17% showed complex barcode attachments. We observed similar patterns in public datasets, and also in native barcoding datasets where only 30%–70% of the reads had barcodes on both ends. Widely used demultiplexers, including Dorado, fail to resolve these cases, leaving ∼10% of our rapid barcoding reads partially trimmed and contaminated with adapter fragments. We developed Barbell, a pattern-aware demultiplexer that is designed to detect complex barcode configurations. Barbell reduced contaminated reads from >400 000 (Dorado/Flexiplex) to 166 (99.96% reduction), minimized barcode bleeding, and supports custom experimental designs such as dual-end barcodes and shorter barcodes (e.g. Illumina barcodes). We further show that such contamination is widespread in public databases, with Nanopore sequences detected in hundreds of NCBI entries, some of which are responsible for artificial taxonomic connections. Availability and implementation: Barbell is open source and available at https://github.com/rickbeeloo/barbell.

Pangenome graphs are increasingly used to represent population-scale bacterial diversity, yet construction methods span fundamentally different representation paradigms whose outputs and sensitivities to assembly quality remain poorly quantified. We systematically reviewed microbial pangenome graph tools and benchmarked six representative methods spanning gene-cluster, compacted coloured de Bruijn graph (ccDBG), multiple sequence alignment, and hybrid approaches. Using a repeat-rich Escherichia coli O157:H7 dataset with complete genomes and matched short-read data, we constructed graphs from identical inputs and observed orders-of-magnitude differences in graph size and fragmentation, indicating that global topology is driven by representation strategy. Varying completeness composition revealed that assembly fragmentation is a first-order determinant of graph structure: gene-cluster graphs contracted as draft assemblies replaced complete genomes, whereas unitig graphs expanded, with distinct degree–prevalence fingerprints across tools. Computational cost mirrored these shifts and depended strongly on completeness composition, including a pronounced runtime penalty for one ccDBG implementation on all-draft inputs. Finally, analysis of Shiga toxin loci showed that pangenome-level reconciliation does not reliably correct assembly artefacts at challenging multi-copy genes and that performance varies by locus. Together, these findings show that pangenome graphs are representation-dependent models of bacterial diversity, and that assembly completeness is a primary determinant of their topology, scalability, and locus-level accuracy. Authors Summary: Bacterial populations are often described using “pangenome graphs,” which aim to capture all genetic variation across many genomes in a single structure. However, different tools build these graphs in fundamentally different ways, and little is known about how those differences affect the results. In this study, we systematically compared several widely used approaches using a clinically important strain of Escherichia coli that is rich in repeated and mobile DNA. We found that the size, connectivity, and overall structure of the resulting graphs varied dramatically depending on the method used. Importantly, we also show that incomplete genome assemblies (common in large sequencing studies) strongly alter graph structure, and that different tools respond to incomplete data in different ways. In some cases, this affects the detection of medically relevant genes, including Shiga toxin genes linked to severe disease. Our results demonstrate that pangenome graphs are not interchangeable representations of bacterial diversity. Instead, their structure depends on both the method and the quality of the input data. We argue that researchers should choose graph-building tools carefully and report structural properties explicitly to ensure reproducible and interpretable results.

In this study we employed nanopore whole genome sequencing to analyze the resistance genes, genomic islands and prophage DNA in two multidrug resistant E. rhusiopathiae strains, i.e., 670 and 147, isolated from domestic geese. MLST profiles and core-genome phylogeny were determined to assess strain relatedness. In strain 670 (serotype 8, ST 113), a novel 53 kb prophage φEr670 carrying the lnuB and lsaE resistance genes was identified. Regions highly homologous to the φEr670 prophage were detected in 36 of 586 (6.14%) publicly available E. rhusiopathiae genomes, as well as in some other Gram-positive bacteria, and usually contained resistance genes. E. rhusiopathiae strain 147 (serotype 5, ST 243) was found to contain a composite 98 kb genomic island (GI_Er147) carrying the ant(6)-Ia and spw genes, as well as gene encoding a putative lincosamide nucleotidyltransferase designated lnu(J) and a vat family gene encoding a putative streptogramin A O-acetyltransferase. The lnu(J) gene exhibited 83.6% homology to the lnu(D) gene, and lnu(J)-positive E. rhusiopathiae strains displayed intermediate susceptibility to lincomycin. Vat-positive strain 147 and vat-negative E. rhusiopathiae strains showed similar susceptibility to quinupristin/dalfopristin. The presence of the Tn916 transposon carrying the tetM gene was confirmed in the genomes of both E. rhusiopathiae strains; in strain 147, however, Tn916 was located within ICEEr1012. Based on analyses of additional E. rhusiopathiae genomes, the integration sites of Tn916, ICEEr1012, and GI_Er147 were identified as genomic “hot spots,” contributing to the genome plasticity of E. rhusiopathiae. Prophage φEr670 and GI_Er147 as well as the Tn916 transposon and ICEEr1012 are most likely responsible for the dissemination of resistance genes in E. rhusiopathiae. Prophages highly homologous to φEr670 act as carriers of resistance genes in various Gram-positive bacteria. However, the transferability of the identified genetic elements and the functional role of the lnu(J) gene require further investigation. This study provides new insights into the diversity of MGEs in E. rhusiopathiae and advances understanding of the genomic mechanisms driving antimicrobial resistance in Gram-positive bacteria.

Background: Many clinically important antimicrobial resistance (AMR) phenotypes such as fluoroquinolones and rifamycins are driven by antimicrobial resistance-conferring mutations (ARMs) in conserved chromosomal loci (e.g., gyrA , parC, rpoB ). Resistome profiling by metagenomics sequencing is often proposed as an ideal AMR surveillance tool as it is organism-agnostic, but currently, existing metagenomic AMR surveillance pipelines are only able to identify acquired AMR genes but not point mutations associated with AMR. This is a serious gap in metagenomics-based AMR surveillance, as the true extent of AMR may be underestimated. Methods: We developed MetaPointFinder (v1), a read-based method that can process both long and short metagenomic reads. The tool identifies resistance-determining regions in reads using DIAMOND (translated protein) and KMA (nucleotide), and classifies known resistant versus wild-type variants by aligning the sequences using pwalign and assessing the observed mutations based on known resistance mutations in the AMRFinderPlus database. The tool outputs ARMs per read, per gene and per antibiotic class. Results: In proof-of-concept analyses, MetaPointFinder identified known AMR-associated mutations and quantified resistant/susceptible read counts and ratios from metagenomic samples with corroborating phenotypic resistance data when available. We benchmark our tool using simulated reads from DNA and from reverse-translated protein references with read lengths of 100-5000 bp and error rates between 0 and 30%, simulating both Illumina and Nanopore error rates. We show that MetaPointFinder outperforms any available method for detection of ARMs in metagenomics data. Conclusion: MetaPointFinder complements gene-centric resistome profiling by capturing chromosomal mutation-based AMR directly from metagenomes.

Background: Oxford Nanopore sequencing enables long-read sequencing across diverse applications, yet the experimental artifacts introduced by Nanopore barcoding are not well characterized. These artifacts can affect demultiplexing accuracy and downstream analyses. Results: We performed a rapid barcoding experiment on 66 diagnostic samples and found that 83% of reads carried the expected single-barcode pattern, while 17% contained multiple barcodes or other artifacts. Current demultiplexers, including the widely used Dorado , fail to correctly handle these complex cases, leaving approximately 7% of reads partially trimmed and contaminated with adapter fragments. Additional issues include the presence of two barcodes at the same read end—either identical, originating from the same sample, or different, introduced after pooling. The latter can lead to barcode bleeding when the outer barcode is incorrectly selected. To address these challenges, we developed Barbell , a pattern-aware demultiplexer that detects all barcode configurations. Barbell reduces trimming errors by three orders of magnitude, minimizes barcode bleeding, and supports custom experimental setups such as shorter barcodes, dual-end barcodes, and custom flank sequences. Conclusions: Our results highlight the impact of complex barcode attachments in Nanopore sequencing and demonstrate that Barbell drastically reduces their effects on downstream analyses. Barbell is open source and available at https://github.com/rickbeeloo/barbell .

Lactobacillaceae is a large family of bacteria from which probiotic strains often originate. Microorganisms used as feed additives in the EU must meet a number of formal criteria, some of which concern antimicrobial susceptibility. In this study, we determined the susceptibility of 19 reference strains and 121 wild-type strains of Lactobacillaceae to aminoglycoside antibiotics using the broth microdilution method based on the ISO 10932:2010/IDF 223:2010 standard. Strains were categorized as resistant or susceptible according to European Food Safety Authority (EFSA) guidelines. Resistance genes were detected by whole genome sequence (WGS) analysis or by PCR. The MICs read after 48 h of incubation showed that 36.8% of reference strains were resistant to kanamycin, 26.3% to streptomycin, and 5.3% to gentamicin, with no aminoglycoside resistance genes detected in any genome. As many as 93.2% of field isolates of Ligilactobacillus salivarius, 85% of Ligilactobacillus agilis, and 58.8% of Lactiplantibacillus plantarum were classified as resistant to kanamycin, with the aac(6)-Ie-aph(2)-Ia gene detected only in two isolates. In six of 12 streptomycin-resistant strains, the ant(6)-Ia gene was identified, which usually coexisted with the spw gene. Three isolates with high neomycin MICs harbored the ant(4′)-Ia gene. In Lactobacillus gallinarum strain LMG 9435, characterized by streptomycin MIC value > 1024 µg/mL, a potential resistance-causing mutation in the rpsL gene (Lys56 → Arg) was detected. The results of the study indicate that some genera of Lactobacillaceae, in particular L. salivarius and L. agilis, exhibit natural resistance to aminoglycoside antibiotics, mainly kanamycin. Therefore, there is a need to update the EFSA guidelines on antimicrobial susceptibility testing of Lactobacillaceae, so that strains lacking resistance genes and/or chromosomal mutations are not considered to be resistant.