Integrated Phenotypic and Genomic Profiling of Antimicrobial Resistance and Virulence-Associated Determinants in Poultry-Derived Enterococcus spp. from Hungary

Simple Summary

Enterococci are common gut bacteria in animals and humans, but some lineages can also cause difficult-to-treat infections. Their public health relevance increases when antimicrobial resistance and virulence-associated traits co-occur, because such combinations may support persistence, colonization, and onward dissemination across the animal–food–environment–human interface. Here, we investigated poultry-derived Enterococcus isolates using paired phenotypic microdilution susceptibility testing and whole-genome sequencing for a defined subset. We focused on two complementary genomic layers: (i) acquired antimicrobial resistance genes (resistome) and (ii) virulence-associated determinants (virulome). Analyses were performed in a species-stratified manner because Enterococcus faecalis and Enterococcus faecium differ in their typical virulence repertoires and population structures. In the sequenced subset, E. faecalis carried a broad set of virulence-associated genes linked to adhesion and biofilm-related functions, whereas E. faecium showed a more limited high-confidence virulence gene repertoire. Acquired resistance determinants were common and supported the observed multidrug non-susceptibility patterns. Importantly, we interpret virulence genes as genetic potential rather than proof of pathogenicity, because gene presence does not imply expression or clinical disease. Overall, our results highlight that poultry-associated enterococci can co-harbor determinants relevant to persistence and antimicrobial resistance, supporting their use as One Health surveillance indicators and motivating targeted functional follow-up where genotype–phenotype discordance is observed.

Abstract

Background: Poultry-associated Enterococcus spp. are widespread commensals but may serve as One Health indicators when virulence-associated determinants and antimicrobial resistance co-occur. We characterized paired phenotypic and genomic profiles to delineate species-stratified virulome and resistome patterns. Methods: Isolates originated from a previously established poultry collection with MIC testing. Genotype–phenotype analyses were restricted to the whole-genome sequenced subset (n = 31). The acquired antimicrobial resistance genes were identified using the Comprehensive Antibiotic Resistance Database (CARD), and virulence-associated determinants were screened using the Virulence Factors Database (VFDB). Results were summarized as isolate-level presence/absence matrices and integrated with MIC-derived susceptible/intermediate/resistant categories. Results: The WGS subset comprised E. faecalis (n = 23) and E. faecium (n = 8) with diverse sequence types. Virulome architecture was strongly species-dependent: E. faecalis carried a broad repertoire of adhesion/biofilm-associated determinants, whereas E. faecium showed a limited set of high-confidence virulence-associated hits. Acquired resistance determinants were common across isolates, and resistome profiles displayed structured co-occurrence. Integrated analyses suggested only a modest overall association between virulence-gene burden and acquired resistome size, largely driven by species-level differences. Genotype–phenotype concordance was class-dependent, with incomplete alignment in several antimicrobial classes, consistent with mechanisms beyond the screened acquired gene set. The acquired resistance determinants detected in the WGS subset predominantly mapped to antimicrobial classes commonly used in food-producing animals (e.g., tetracyclines, macrolides, lincosamides, aminoglycosides, and phenicols), supporting interpretation in the context of production-associated antimicrobial selection rather than implying last-line clinical resistance by default. Conclusions: Poultry-derived enterococci may combine genetic features compatible with persistence/colonization and acquired antimicrobial resistance, with co-occurrence patterns shaped primarily by species/lineage background. These findings support risk-stratified One Health surveillance and targeted functional and mechanism-focused follow-up. This integrated virulome–resistome view highlights species-specific risk signatures in poultry-associated Enterococcus and identifies discordant high-level phenotypes that merit targeted mechanistic follow-up.

1. Introduction

Antimicrobial resistance (AMR) is widely recognized as a major global health threat that undermines the effective prevention and treatment of infections across human and veterinary medicine. Recent estimates indicate that bacterial AMR was directly responsible for 1.27 million deaths in 2019 and associated with 4.95 million deaths, highlighting that AMR is already a contemporary, not merely projected, burden. Consequently, AMR is best understood through a One Health lens, as selection and dissemination are shaped by interconnected human, animal, food, and environmental systems [1].

Within this continuum, Enterococcus spp. represent a particularly informative and challenging genus. Enterococci are common gastrointestinal commensals of humans and animals, yet they are exceptionally resilient, with broad stress tolerance that supports persistence and transmission across diverse niches, including food-production chains and environmental reservoirs [2,3]. This ecological versatility is central to their epidemiological relevance [4]. Enterococci can move between compartments and maintain adaptive traits under multiple, often simultaneous, selective pressures.

At the same time, enterococci are major opportunistic pathogens and among the leading causes of healthcare-associated infections. Clinically, Enterococcus faecalis has historically accounted for a substantial fraction of enterococcal infections, whereas Enterococcus faecium has become a high-priority nosocomial concern because of its strong association with multidrug resistance and the global expansion of hospital-adapted lineages [5]. Importantly, E. faecium is not a homogeneous population: large-scale comparative genomics has demonstrated distinct phylogenetic groupings/lineages with differential adaptation to community, animal, and hospital ecosystems, commonly framed as hospital-associated versus non-hospital-associated clades (and subclades), reflecting evolutionary responses to ecological separation and antimicrobial exposure [6,7,8]. This population structure provides a critical framework for interpreting One Health signals because lineage context influences the likelihood that resistance and accessory traits are concentrated, maintained, and disseminated.

Poultry production systems constitute a relevant node in this network. Enterococci are abundant in the poultry gut microbiota and can be recovered along the production chain, including from carcasses and retail meat. Their persistence is facilitated by tolerance to stressors encountered in food processing and production environments, and poultry-associated enterococci have been repeatedly highlighted as useful indicator organisms for antimicrobial susceptibility monitoring at the population level (e.g., standardized surveillance approaches in healthy food-producing animals) [9,10,11,12]. These features position poultry-derived enterococci as a practical lens for understanding how resistance determinants are maintained and potentially disseminated via the food chain. In poultry-associated enterococci, acquired resistance determinants often reflect long-standing selection by antimicrobial classes widely used in food-producing animals and thus provide important ecological and surveillance signals even when they do not correspond to last-line resistance determinants of greatest clinical concern.

Beyond resistance, virulence-associated determinants add a second dimension of One Health relevance. Enterococcal pathogenicity is multifactorial and strongly context-dependent; therefore, the presence of a virulence-associated gene does not imply expression or clinical virulence. Nevertheless, virulence-associated determinants, particularly those implicated in adhesion/colonization, biofilm formation, immune interaction, and tissue damage, are biologically meaningful as proxies for colonization capacity, persistence, and host interaction potential. In the context of AMR, such traits may increase the probability of establishment and onward transmission, thereby amplifying risk without requiring direct inference of pathogenicity from gene presence alone [13].

Whole-genome sequencing (WGS) provides a coherent framework to integrate these dimensions—resistome, virulome, and population structure—especially when anchored to robust phenotypic antimicrobial susceptibility testing such as minimum inhibitory concentration (MIC) determination. WGS enables lineage-resolved interpretation of accessory gene content and supports cautious, mechanistic inference regarding the genetic basis of phenotypes and the potential role of mobile genetic elements in dissemination [14,15,16]. Despite the recognized One Health relevance of poultry-associated enterococci, integrated MIC–WGS datasets linking resistome, virulome, and population structure remain limited for Hungary, restricting baseline, surveillance-ready interpretation at the national level.

In this study, we apply a combined phenotype–genotype approach to characterize Enterococcus isolates from Hungarian poultry within a One Health context. Using phenotypic MIC data alongside WGS on a targeted subset, we (i) describe species composition and population structure (including MLST context where applicable), (ii) map the genomic landscape of virulence-associated determinants across functionally meaningful categories, (iii) characterize the resistome and key multidrug-resistance patterns, and (iv) explore associations (not causation) between virulence gene profiles and resistance phenotypes/classes. Our goal is to provide a rigorous baseline for surveillance-oriented interpretation while explicitly separating gene presence from inferred pathogenicity.

2. Materials and Methods

2.1. Study Design, Sampling Frame, and Isolate Collection

The isolates used in this study were obtained retrospectively from an existing strain repository established between 2022 and 2023 from clinically healthy poultry flocks (domestic chicken and turkey) kept under intensive farming conditions in Hungary. Importantly, no additional sampling, animal handling, or animal procedures were performed specifically for the present study. The archived isolates had originally been derived from two anatomical sites (cloaca and respiratory tract) and covered all seven Hungarian NUTS-1 regions (Közép-Magyarország, Közép-Dunántúl, Nyugat-Dunántúl, Dél-Dunántúl, Észak-Magyarország, Észak-Alföld, and Dél-Alföld). For each isolate, county-level origin, production type, and age group were recorded and used for downstream analyses.

Phenotypic antimicrobial susceptibility data (MICs) were available retrospectively for a total of n = 218 Enterococcus isolates from our existing strain collection; MICs had been generated previously in our laboratory by broth microdilution following CLSI methodology (Section 2.3 for details).

2.2. Isolation, Presumptive Identification, and Strain Preservation

Swabs were streaked onto selective m-Enterococcus agar (Merck KGaA, Darmstadt, Germany) and incubated at 37 °C for 18–24 h. Presumptive Enterococcus colonies were subcultured onto tryptone soya agar (Biolab Zrt., Budapest, Hungary) for purity and incubated under the same conditions. Genus-level identification was performed using standard microbiological criteria matrix-assisted laser desorption ionization–time of flight (MALDI-TOF) mass spectrometry (Flextra-LAB Kft., Budapest, Hungary) and the Biotyper software version 12.0 (Bruker Daltonics GmbH, Bremen, Germany; 2024 release). MALDI-TOF MS-based identification was performed as routinely applied in clinical microbiology and as previously validated for rapid, reliable bacterial species identification [17]. For the WGS subset, species assignment was additionally confirmed in silico (Section 2.6, Section 2.7 and Section 2.8). Pure cultures were stored at −80 °C using a cryopreservation system in Microbank cryovials (Pro-Lab Diagnostics, Richmond Hill, ON, Canada).

2.3. Antimicrobial Susceptibility Testing

MICs were determined using a broth microdilution workflow following Clinical and Laboratory Standards Institute (CLSI)-aligned procedures for inoculum preparation and quality control. Broth microdilution was performed following CLSI M07 (12th edition, 2024) [18]. Briefly, isolates were revived from −80 °C storage on Mueller–Hinton broth (CAMHB; VWR International, Debrecen, Hungary) and incubated at 37 °C for 18–24 h. A standardized bacterial suspension corresponding to 0.5 McFarland was prepared using a nephelometer (ThermoFisher Scientific, Budapest, Hungary), and the working inoculum was adjusted to achieve the target cell density for microdilution testing. Microdilution plates (Brand-plates, VWR International, Debrecen, Hungary) were inoculated and incubated at 37 °C for 18–24 h, after which MIC endpoints were read visually.

The MIC panel included imipenem, vancomycin, azithromycin, linezolid, tilmicosin, oxytetracycline, amoxicillin, amoxicillin–clavulanic acid (2:1 ratio), ceftriaxone, neomycin, spectinomycin, doxycycline, florfenicol, tylosin, tiamulin, lincomycin, enrofloxacin, colistin, and potentiated sulfonamide (trimethoprim and sulfamethoxazole 1:19). E. faecalis ATCC 29212 was used as the quality-control reference strain.

2.4. Interpretation of MIC Values and Resistance Classification

In cases where clinical breakpoints were available, isolates were categorized as susceptible (S), intermediate (I), or resistant (R) using CLSI. MICs were interpreted using CLSI interpretive criteria (VET06, 1st Edition, 2017), and the CLSI version applied in this study is explicitly stated here for transparency [19]. In cases where veterinary-specific interpretive criteria were unavailable for a given Enterococcus species–antimicrobial combination, CLSI M100 (34th edition, 2024) was used for categorical interpretation, and remaining MICs were reported descriptively [20].

EUCAST epidemiological cut-off values (ECOFFs) were not applied in this study because our primary objective was a harmonized breakpoint-based S/I/R categorization aligned with CLSI criteria for genotype–phenotype comparisons; moreover, ECOFF availability is incomplete across Enterococcus spp. and drug–species combinations in veterinary contexts, and ECOFFs define wild-type/non-wild-type separation rather than clinical susceptibility.

Multidrug resistance (MDR) was defined using a standard operational framework (non-susceptibility to ≥1 agent in ≥3 antimicrobial classes), and additional resistance categories (XDR: extensively drug-resistant; PDR: pan drug-resistant) were explored, where the available panel supported defensible class-based classification [21]. Because breakpoint availability differs between compounds/classes, MDR categorization was applied consistently using only antimicrobial classes with interpretable endpoints in the compiled scheme.

The WGS subset (n = 31) was purposively selected from the MIC-tested strain collection to enrich for multidrug-resistant phenotypes and to maximize diversity across host species and geographic origin. Accordingly, the results derived from the WGS subset are intended for mechanistic/genomic profiling and exploratory surveillance interpretation rather than population-level prevalence inference.

2.5. DNA Extraction, Library Preparation, and Sequencing

Genomic DNA was extracted from overnight cultures using the Zymo Quick-DNA Fungal/Bacterial Miniprep Kit (Zymo Research, Irvine, CA, USA) according to the manufacturer’s protocol [22]. Mechanical disruption was performed using a TissueLyzer LT (Qiagen GmbH, Hilden, Germany) at 50 Hz for 5 min, and lysates were stored at −20 °C until library preparation.

For library preparation, the Illumina Nextera XT DNA Library Preparation Kit (Illumina, San Diego, CA, USA) was used [23]. DNA fragmentation and indexing were performed with the Nextera XT Index Kit v2 Set A using i5 and i7 index primers (Illumina, San Diego, CA, USA). Genomic DNA was diluted to a final concentration of 0.2 ng/µL in a 2.5 µL volume, followed by the addition of 5 µL Tagment DNA Buffer and 2.5 µL Amplicon Tagment Mix (Illumina, San Diego, CA, USA). Tagmentation was carried out at 55 °C for 6 min using an Eppendorf Mastercycler nexus GX2 thermal cycler (Eppendorf SE, Hamburg, Germany), followed by cooling to 10 °C. The reaction was neutralized by adding 2.5 µL Neutralize Tagment Buffer (Illumina, San Diego, CA, USA) and incubating at room temperature for 5 min. The tagmented DNA was amplified using 7.5 µL Nextera PCR Master Mix (Illumina, San Diego, CA, USA) and 2.5 µL each of i5 and i7 index primers with the following PCR conditions: 95 °C for 30 s; 12 cycles of 95 °C for 10 s, 55 °C for 30 s, and 72 °C for 30 s; and a final extension at 72 °C for 5 min, followed by a hold at 10 °C.

The indexed libraries were purified using the Geneaid Gel/PCR DNA Fragments Extraction Kit (Geneaid Biotech, New Taipei City, Taiwan) via column-based cleanup and quantified fluorometrically using the Qubit dsDNA HS Assay Kit (Thermo Fisher Scientific, Waltham, MA, USA). Libraries were pooled in equimolar ratios. Paired-end sequencing was performed by Novogene (Beijing, China) on an Illumina NovaSeq X Plus platform (2 × 150 bp).

2.6. Read Processing, De Novo Assembly, and Assembly Quality Assessment

Raw reads were quality-checked using FastQC v0.11.9 [24] and processed using fastp v0.23.2-3 [25] for adapter/quality trimming. Additional k-mer-based QC/cleaning tools were applied as needed, with Bloocoo v1.0.7 [26]. High-quality reads were assembled de novo using SPAdes v4.0.0 [27] and MEGAHIT v1.2.9 [28], and assemblies were combined using GAM-NGS v1.1b [29] to improve contiguity in cases where this enhanced assembly quality.

Assembly quality was evaluated using QUAST v5.2 [30], and genome completeness was assessed using BUSCO v5 [31] where applicable. Predicted coding sequences were called using Prodigal, and assemblies were annotated using Prokka v1.14.5 [32], generating standardized GFF/TSV outputs for downstream gene-content analyses.

2.7. In Silico Species Confirmation and Population Structure

For the WGS subset, taxonomic assignment was cross-validated in silico using read/assembly-based classifiers Kraken2 v1.1.1 [33] and CheckM v1.2.2 [34]. Population structure was characterized using multilocus sequence typing (MLST) with the mlst pipeline, applying the appropriate species-specific schemes for E. faecalis and E. faecium. Sequence types (STs) were reported where complete allele profiles were available; isolates with incomplete profiles were retained but flagged as “ST not assigned”.

2.8. Resistome and Virulome Profiling

The acquired antimicrobial resistance genes were identified using a Comprehensive Antibiotic Resistance Database (CARD)-based workflow [35]. Hits were summarized at the isolate level, and gene presence/absence matrices were generated for downstream comparisons with phenotypic MIC categories. For transparency and reproducibility, alignment metrics (percent identity and percent coverage) were retained in the supplementary dataset, while interpretation in the main text focused on high-confidence calls.

Virulence-associated genes were screened using the Virulence Factors Database (VFDB) [36]. Results are summarized as (i) gene-level hit tables with alignment metrics and (ii) isolate-level presence/absence matrices. Virulence determinants were interpreted as genetic potential and were not used to infer clinical virulence, consistent with the principle that gene presence does not imply gene expression or disease causation. Screening thresholds and full hit metrics are reported in the Supplementary Material. For resistome and virulome screening, hits were retained for downstream presence/absence matrices if they met predefined alignment thresholds of ≥95% identity and ≥80% query coverage.

2.9. Data Integration and Statistical Analysis

All analyses were performed in R (v4.1.0). Descriptive phenotypic analyses (MIC distributions and S/I/R categorization where applicable) were conducted using the available MIC dataset (n = 218) to provide an overall susceptibility landscape of poultry-derived Enterococcus isolates. Genotype-based analyses—including resistome/virulome profiling and genotype–phenotype comparisons—were restricted to isolates with whole-genome sequencing data (WGS subset, n = 31). Hits were retained using predefined alignment thresholds (≥95% sequence identity and ≥60% query coverage), and full hit-level metrics are provided in the Supplementary Material. Associations between categorical variables (e.g., gene presence versus S/I/R category) were assessed using Fisher’s exact test, and multiple testing was controlled using the Benjamini–Hochberg false discovery rate (FDR) correction where appropriate. Differences in MIC distributions between genotype-defined groups were evaluated using non-parametric tests (Mann–Whitney U), with effect sizes reported alongside p-values.

3. Results

3.1. Whole-Genome Sequenced Subset and Metadata Overview

The present study focuses on the whole-genome sequenced subset (n = 31) to enable integrated virulome–resistome–phenotype analyses (Table 1). The subset comprised isolates recovered from clinically healthy poultry (chicken and turkey) sampled from two anatomical sites (cloaca and respiratory tract) and represented multiple Hungarian regions. Species assignment based on the MLST scheme selection identified 23 isolates as E. faecalis and 8 isolates as E. faecium. The WGS subset included isolates from both host species and sampling sites, supporting species-stratified and metadata-informed interpretation of genomic determinants. For all WGS-profiled isolates, phenotypic MIC values were available and used for downstream genotype–phenotype comparisons. Detailed isolate-level metadata and linked MIC values for the WGS subset are provided in Supplementary Tables S1 and S2.

Table 1.

Overview of the whole-genome sequenced Enterococcus subset (n = 31) and linked metadata. MDR, multidrug-resistant. ST, sequence type (multilocus sequence typing).

Species	n	Chicken n (%)	Turkey n (%)	Cloaca n (%)	Respiratory n (%)	Regions n	ST Assigned n (%)	Distinct STs n	MDR n (%)
Enterococcus faecalis	23	10 (43.5)	13 (56.5)	13 (56.5)	10 (43.5)	6	18 (78.3)	10	23 (100.0)
Enterococcus faecium	8	8 (100.0)	0 (0.0)	7 (87.5)	1 (12.5)	3	7 (87.5)	6	8 (100.0)
Total	31	18 (58.1)	13 (41.9)	20 (64.5)	11 (35.5)	6	25 (80.6)	16	31 (100.0)

3.2. Phenotypic Antimicrobial Susceptibility Profile of the WGS Subset

To enable direct genotype–phenotype comparisons, MIC values were interpreted for antimicrobials (Figure 1) with available interpretive criteria and summarized as susceptible (S), intermediate (I), or resistant (R). Across the 12 compounds with S/I/R categorization, isolates exhibited a high overall non-susceptibility burden (median number of non-susceptible results per isolate: 7; interquartile range: 6–7; range: 3–9).

Figure 1.

Distribution of S/I/R categories across antimicrobials with applicable interpretive criteria in the whole-genome sequencing (WGS) subset (n = 31).

Non-susceptibility was most frequent for lincomycin (31/31; 100%), florfenicol (28/31; 90.3%), azithromycin (25/31; 80.6%), and oxytetracycline (23/31; 74.2%). Doxycycline and tylosin non-susceptibility were observed in 19/31 (61.3%) and 18/31 (58.1%) isolates, respectively. Intermediate-to-high non-susceptibility rates were recorded for enrofloxacin (15/31; 48.4%) and amoxicillin (14/31; 45.2%). Lower non-susceptibility rates were observed for linezolid (9/31; 29.0%), vancomycin (8/31; 25.8%), amoxicillin–clavulanic acid (4/31; 12.9%), and imipenem (3/31; 9.7%). Detailed isolate-level MIC values and categorical interpretations for the WGS subset are provided in Supplementary Table S2.

Species-stratified summaries indicated distinct phenotypic patterns. Azithromycin and oxytetracycline non-susceptibility predominated among E. faecalis (23/23 and 21/23, respectively), whereas enrofloxacin non-susceptibility was enriched among E. faecium (7/8). Imipenem resistance was observed only in E. faecium (3/8), while linezolid and vancomycin non-susceptibility were detected only among E. faecalis (9/23 and 8/23, respectively). Detailed isolate-level MIC values and categorical interpretations for the WGS subset are provided in Supplementary Table S2.

3.3. Population Structure of the WGS Subset Based on MLST

MLST analysis was performed using species-specific schemes (E. faecalis and E. faecium) to contextualize virulome and resistome profiles within the population structure of the WGS subset. A sequence type (ST) could be assigned for 25/31 isolates (80.6%), including 18/23 E. faecalis (78.3%) and 7/8 E. faecium (87.5%). Overall, 16 distinct STs were identified across the dataset, comprising 10 STs in E. faecalis and 6 STs in E. faecium, indicating substantial genetic diversity rather than the dominance of a single clonal lineage.

Within E. faecalis, the most frequent STs were ST4, ST207, ST300, and ST407 (each 3/18 among ST-assigned isolates), while the remaining STs were singletons. Within E. faecium, ST1866 was detected in 2/7 ST-assigned isolates, with the other STs occurring once each. For six isolates, an ST could not be determined (incomplete MLST profile), and these isolates were retained for downstream analyses but reported as “ST not assigned”. Detailed isolate-level MLST assignments are provided in Supplementary Table S3, and the distribution of sequence types is shown in Supplementary Figure S1.

Because isolates were purposively selected for WGS, the MLST sequence type (ST) distribution reported here is descriptive of the WGS subset only and should not be interpreted as reflecting population structure in the underlying collection.

3.4. Virulence-Associated Gene Repertoire in the WGS Subset

Virulence-associated determinants showed a pronounced species-dependent distribution in the WGS subset (Figure 2). Under stringent screening criteria, E. faecalis isolates (n = 23) carried a broad repertoire of virulence-associated genes, whereas E. faecium isolates (n = 8) yielded only a limited set of high-confidence hits.

Figure 2.

Presence and absence heatmap of virulence-associated determinants (virulome) across the whole-genome sequenced subset (n = 31), stratified by species. Rows represent isolates (ordered by species and MLST sequence type), and columns represent virulence-associated genes screened using VFDB. Blue squares indicate gene detection (presence), and blank cells indicate no hit under the applied screening criteria. This species-stratified view highlights a broader virulome repertoire in Enterococcus faecalis compared with Enterococcus faecium, supporting interpretation in the context of lineage ecology rather than clinical virulence.

In E. faecalis, a conserved “core” virulence-associated signature was observed, including the pilus/biofilm-associated locus ebpB/ebpC and srtA (each 23/23; 100%), as well as gelE (23/23; 100%) and its regulator fsrB (21/23; 91.3%). Several pheromone/plasmid-associated determinants (cCF10, cOB1, cad, and camE) were also frequent, with cCF10, cOB1, and cad present in all E. faecalis isolates (23/23; 100%) and camE detected in 22/23 (95.7%). Additional adhesion-associated genes were common, including efaAfs (22/23; 95.7%), ElrA (19/23; 82.6%), and ace (15/23; 65.2%). The hyaluronidase-associated genes hylA and hylB were detected in 16/23 (69.6%) and 16/23 (69.6%) E. faecalis isolates, respectively. In contrast, cytolysin-associated genes were comparatively infrequent (cylA 4/23, 17.4%; cylL 5/23, 21.7%; cylB and cylM each 1/23, 4.3%), and the aggregation substance gene agg was detected in 5/23 (21.7%).

Among E. faecium, only the collagen-binding adhesin acm was detected under stringent criteria (5/8; 62.5%), while the remaining screened determinants were not identified as high-confidence hits. Consistent with these distributions, the virulence gene burden (number of detected virulence-associated genes per isolate) was markedly higher in E. faecalis (median 15 genes per isolate; range 13–19). In contrast, E. faecium isolates displayed a highly restricted virulence gene profile. A species-associated efaA allele consistent with efaA_fm was detected across all E. faecium genomes, while acm was present in a subset of isolates. Accordingly, the virulence gene burden in E. faecium remained low (median 2 genes per isolate; range 1–2), consistent with Figure 2.

The isolate-level presence/absence matrix for all screened virulence-associated determinants is provided in Supplementary Table S4.

3.5. Resistome Profiles of the WGS Subset

Acquired antimicrobial resistance gene profiles showed clear species-dependent patterns in the WGS subset (Figure 3). Overall, E. faecalis isolates harbored a broader resistome compared with E. faecium, consistent with their higher phenotypic non-susceptibility burden across multiple antimicrobial classes. Across the dataset, macrolide–lincosamide resistance determinants and tetracycline resistance genes predominated, in line with the high phenotypic non-susceptibility rates observed for azithromycin/tylosin and tetracyclines.

Figure 3.

Presence and absence heatmap of acquired antimicrobial resistance determinants (resistome) across the whole-genome sequenced subset (n = 31), stratified by species. Rows represent isolates (ordered by species and MLST sequence type), and columns represent acquired resistance genes identified using CARD. Blue squares indicate gene detection (presence), and blank cells indicate no hit under the applied screening criteria. The resistome patterns underscore a high overall acquired resistance burden in the whole-genome sequencing (WGS) subset and facilitate direct visual comparison with the virulome architecture shown in Figure 2.

In E. faecalis, a conserved set of acquired resistance determinants was frequently detected (Figure 3), with recurrent combinations spanning macrolide–lincosamide, tetracycline, aminoglycoside, and phenicol resistance. In contrast, E. faecium isolates displayed a more restricted but lineage-consistent resistome, with fewer total acquired genes per isolate and a narrower set of recurrent determinants. Across both species, resistome profiles exhibited marked co-occurrence structure, indicating that non-susceptibility burdens were typically driven by multi-gene constellations rather than single determinants (Supplementary Table S1).

3.6. Integrated Virulome–Resistome Burden in the WGS Subset

Across the WGS subset (n = 31), the number of virulence-associated genes per isolate (unique hits) ranged from 1 to 19 (median: 15), whereas the number of acquired antimicrobial resistance (AMR) genes ranged from 9 to 32 (median: 13). Virulence gene burden differed markedly by species: E. faecalis showed substantially higher virulence gene counts than E. faecium (median 15 vs. 2; Mann–Whitney U test p = 2.54 × 10⁻⁵), consistent with the species-stratified virulome patterns shown in Figure 2. In contrast, the acquired AMR gene burden was comparable between species (median 14 in E. faecalis vs. 11.5 in E. faecium; p = 0.0966), indicating that the most pronounced separation between species in this dataset was driven by virulence-associated determinants rather than by the overall size of the acquired resistome (Figure 4).

Figure 4.

Relationship between virulence-associated gene burden and acquired antimicrobial resistance (AMR) gene burden in the whole-genome sequencing (WGS) subset (n = 31). Each point represents one isolate; symbols indicate species. Gene burdens are expressed as the number of unique virulence-associated genes (y-axis) and unique acquired AMR genes (x-axis) detected per isolate. The overall association is summarized using Spearman correlation.

When considering all isolates together, virulence and resistance gene burdens showed a modest positive association (Spearman ρ = 0.36, p = 0.0499). However, this relationship was not retained within species (Spearman ρ = 0.18, p = 0.411 in E. faecalis; ρ = 0.39, p = 0.340 in E. faecium), suggesting that the overall correlation largely reflects between-species differences rather than within-species co-accumulation of virulence and AMR determinants. Notably, one E. faecalis isolate (ID 759) exhibited an expanded acquired resistome (32 unique AMR gene hits), representing an upper-end outlier in the dataset.

3.7. Genotype–Phenotype Concordance for Selected Antimicrobial Classes

Genotype–phenotype comparisons were restricted to the WGS subset (n = 31), linking phenotypic S/I/R calls to the presence/absence of acquired resistance determinants detected in the CARD-based workflow. Because resistance phenotypes in enterococci may arise from both acquired genes and chromosomal variation, genotype–phenotype comparisons were interpreted mechanistically. Higher concordance was expected for tetracyclines and MLS agents, where transferable determinants (tet, erm, and lsa) commonly drive non-susceptibility, whereas discordance was anticipated for phenotypes predominantly mediated by mutations or regulatory changes beyond the scope of acquired-gene screening (Table 2). Because the resistome screen captures primarily acquired gene determinants, phenotypic resistance without a corresponding acquired gene signal was interpreted conservatively as potentially reflecting (i) chromosomal mutations not captured here, (ii) allele divergence below the high-confidence calling threshold, and/or (iii) methodological/breakpoint-related effects. Notably, canonical transferable determinants mediating clinically significant resistance (e.g., vanA/vanB for vancomycin; optrA/poxtA/cfr-family determinants for linezolid) were not detected in the WGS subset, despite phenotypic non-susceptibility in a limited number of isolates.

Table 2.

Genotype–phenotype concordance for selected antimicrobial classes in the whole-genome sequencing (WGS) subset (n = 31). For each antimicrobial, concordance was calculated as the proportion of isolates in the specified non-susceptible category carrying at least one corresponding acquired resistance determinant.

Antimicrobial (Phenotype)	Gene(s) Evaluated (Acquired)	Phenotype Group	n Phenotype	n with Gene(s) in Phenotype Group	Concordance (%)
Oxytetracycline	any-tet (tet(M/L/O))	I/R	23	14	60.9
Doxycycline	any-tet (tet(M/L/O))	I/R	19	10	52.6
Azithromycin	erm(B)	R	24	5	20.8
Tylosin	erm(B)	R	18	5	27.8
Lincomycin	lsaA, lsaE	R	31	22	71.0
Linezolid	optrA, poxtA, cfr, cfr(B), cfr(D)	R	5	0	0.0
Vancomycin	vanC-type regulators	R	3	2	66.7

Tetracycline-associated genes (tet(M), tet(L), and tet(O)) were detected exclusively in E. faecalis (15/23), while E. faecium carried none under the high-confidence calls. Phenotypically, oxytetracycline resistance was common (22/31 R; 23/31 I/R), and 60.9% (14/23) of oxytetracycline non-susceptible isolates (I/R) carried at least one tet gene (any-tet). For doxycycline, 52.6% (10/19) of non-susceptible isolates (I/R) carried any-tet. At the species level, E. faecalis showed a tighter—but not complete—alignment between tet carriage and tetracycline non-susceptibility, whereas E. faecium displayed doxycycline non-susceptibility despite the absence of detected tet genes, suggesting additional determinants beyond the high-confidence acquired gene set.

The methylase gene erm(B) was detected in 5/31 isolates (4 E. faecalis, 1 E. faecium). Azithromycin resistance was frequent overall (24/31 R) and near-universal in E. faecalis, but erm(B) accounted for only 20.8% (5/24) of azithromycin-resistant isolates at the acquired-gene level, indicating that most macrolide resistance in this dataset is not explained by erm(B) alone. For tylosin, erm(B) showed better specificity: all erm(B)-positive isolates were tylosin-resistant, yet only 27.8% (5/18) of tylosin-resistant isolates carried erm(B), again pointing to additional macrolide resistance mechanisms not represented among the high-confidence acquired calls.

All isolates were categorized as resistant to lincomycin (31/31 R). The lincosamide/streptogramin-associated determinant lsaA was present in 22/31 isolates (22/23 E. faecalis; 0/8 E. faecium), consistent with a strongly species-structured background. Because the phenotype lacked variability, formal association testing was not informative; nevertheless, the pattern supports that lincomycin resistance is driven largely by species-associated intrinsic/constitutive mechanisms, with acquired determinants contributing variably.

Linezolid (5/31 R; 9/31 I/R) and vancomycin (3/31 R; 8/31 I/R) non-susceptibility occurred only in E. faecalis in this subset, while canonical acquired determinants corresponding to these phenotypes were not detected in the present acquired-gene panel. These findings are therefore treated as phenotypic observations requiring cautious interpretation and discussed as potential chromosomal/allelic or methodological/breakpoint-related phenomena rather than being over-attributed to unobserved acquired genes.

4. Discussion

Enterococci occupy a paradoxical ecological niche: they are ubiquitous gastrointestinal commensals in humans and animals, yet they are also among the most important opportunistic pathogens in modern medicine. Their ability to persist in diverse environments, acquire mobile genetic elements, and tolerate antimicrobial and host-derived stressors makes them highly informative “One Health” sentinels that bridge animal production, the food chain, and healthcare settings [15]. In this context, our integrated analysis of a genomically characterized poultry-associated Enterococcus subset provides two complementary layers of evidence: (i) species-stratified virulence-associated genetic potential (virulome architecture) and (ii) the accompanying resistome and its relationship to phenotypic MIC categories. Importantly, our interpretation is explicitly restricted to genomic potential; the presence of a putative virulence gene does not imply expression, fitness advantage in vivo, or clinical virulence, which is shaped by host context, regulatory state, and ecological interactions [37]. By integrating MIC-derived susceptibility categories with isolate-level resistome and virulome profiles in a species-stratified framework, our study delineates distinct One Health-relevant ‘risk signatures’ in poultry-associated enterococci. Moreover, the observed genotype–phenotype discordances for selected high-impact agents are best interpreted as surveillance flags that prioritize targeted mechanistic follow-up rather than as evidence for novel resistance pathways.

A central observation was the pronounced species-level separation of virulome profiles. This is biologically expected because E. faecalis and E. faecium differ substantially in their canonical virulence repertoires, population structures, and host adaptation pathways [37]. In E. faecalis, multiple determinants implicated in colonization, adhesion, and biofilm development frequently co-occur in natural populations. The endocarditis- and biofilm-associated pilus (Ebp; ebpA–ebpB–ebpC) is broadly distributed and experimentally linked to adherence to host extracellular matrix components and to virulence in relevant infection models [38]. Biofilm formation in E. faecalis is additionally modulated by secreted factors, including the gelatinase GelE, which can enhance biofilm formation under specific conditions [39]. GelE expression is regulated by the Fsr quorum-sensing system (fsrA–fsrB–fsrC), which coordinates protease production and contributes to multiple physiological and host-interaction phenotypes [40]. Together, these mechanistic links provide a strong biological rationale for interpreting common E. faecalis adhesion/biofilm-associated gene constellations as markers of colonization competence and persistence potential, rather than as a direct proof of pathogenicity.

In contrast, E. faecium often displays a narrower “classic” virulence gene panel while relying on distinct adhesins and hospital-adaptive traits in specific lineages. A well-characterized example is the collagen-binding adhesin Acm, widely detected among E. faecium isolates and implicated in adherence and, in some contexts, infection-associated phenotypes [41]. Separately, the enterococcal surface protein Esp has been highlighted as a determinant associated with biofilm formation in hospital-adapted E. faecium lineages (notably within CC17-associated populations) [42]. Therefore, the comparatively lower virulence-gene counts we observed in E. faecium should not be misconstrued as “avirulence”; rather, it likely reflects (i) true biological differences in the repertoire captured by common VF panels and (ii) the fact that many clinically relevant E. faecium traits are lineage-dependent and may be underrepresented in animal-associated populations [37,42]. This reinforces why species-stratified and ideally lineage-aware interpretation is essential when translating virulome screening into risk narratives.

From a One Health perspective, the poultry setting matters for two reasons. First, enterococci can behave as opportunists in birds and have been implicated in poultry health problems, underscoring that host context can shift the commensal–pathogen balance [43]. Second, poultry production ecosystems provide selective and ecological conditions that may enrich for antimicrobial resistance determinants and persistence traits, facilitating maintenance and dissemination across compartments, from farm to processing, then to food, and ultimately human exposure [15]. Recent One Health syntheses and meta-analyses support the notion that resistant enterococci, including vancomycin-resistant enterococci (VRE), can occur across the food chain and environmental interfaces, motivating surveillance frameworks that explicitly integrate animal and human data streams [44]. This aligns with global prioritization efforts: vancomycin-resistant E. faecium has been repeatedly listed among high-priority antibiotic-resistant bacteria in WHO prioritization initiatives [45]. Within the broader European One Health surveillance landscape, genomic datasets from poultry-associated enterococci remain comparatively heterogeneous across countries and production systems; therefore, regionally anchored WGS-linked phenotype datasets, such as the present Hungarian collection, can help refine risk interpretation and support cross-study comparability.

The resistome results in our genomically profiled subset are consistent with the well-documented plasticity of enterococcal genomes and their propensity to accumulate acquired resistance determinants in addition to intrinsic traits. A key interpretive nuance is the role of species-intrinsic resistance phenotypes that can complicate genotype–phenotype mapping if not explicitly accounted for. For example, E. faecalis typically exhibits intrinsic resistance to lincosamides and streptogramin A via the chromosomally encoded ABC protein Lsa(A), which has been experimentally linked to this characteristic phenotype [46]. Thus, “gene presence” may function differently across drug classes: some phenotypes are best explained by intrinsic determinants (expected across most isolates of a species), whereas others reflect horizontally acquired genes whose presence/absence is more informative for comparative risk assessment.

Our integrative virulome–resistome view suggests that any apparent positive association between total virulence-gene counts and ARG burden must be interpreted cautiously. In mixed-species datasets, such correlations are often dominated by differences between-species (i.e., E. faecalis tending to carry a broader set of classical adhesion/biofilm loci than E. faecium), rather than indicating within-species coupling of virulence and resistance. This distinction is more than statistical hygiene: it directly affects One Health inference. Over-interpreting a global correlation risks implying a mechanistic link that is not supported at the within-species level. A more defensible conclusion is that poultry-associated enterococci can co-harbor (i) genetic determinants compatible with persistence/colonization and (ii) acquired resistance determinants, but the degree of co-occurrence is structured by species and lineage ecology rather than by a universal virulence–resistance coupling mechanism. To maintain conceptual consistency with our virulome interpretation, we interpret acquired resistance determinants primarily as markers of ecological adaptation and persistence under antimicrobial exposure in production settings. In this framework, ARG detection is not used to infer clinical treatment failure, but rather to characterize selection signatures and potential transmission-relevant genetic payloads in a One Health surveillance context.

Genotype–phenotype discordance deserves explicit, conservative interpretation, particularly for high-impact drug classes. For linezolid, resistance in enterococci can arise through (i) mobile genes such as optrA or poxtA (ribosomal protection) and cfr variants (rRNA methylation) and/or (ii) chromosomal mutations, especially in the 23S rRNA (classically at positions corresponding to G2576 in E. coli numbering), among other loci [47]. Consequently, an incomplete match between the MIC category and the detection of canonical mobile genes does not invalidate either dataset; it typically indicates that the responsible mechanism may be mutational, regulatory, copy-number related, or outside the screened gene set [47]. For vancomycin, the epidemiologically most important high-level resistance in enterococci is commonly mediated by van operons (e.g., vanA, vanB) in E. faecium populations, which is precisely why VRE remains a high-priority global concern [45]. If phenotypic non-susceptibility occurs in the absence of detected van genes, the most parsimonious explanations are methodological and interpretive rather than sensational: borderline MICs near interpretive thresholds, species- or lineage-specific intrinsic traits (in non-faecium/faecalis enterococci), heteroresistance/tolerance phenomena, or technical differences in gene-calling sensitivity and assembly context. Importantly, our data do not demonstrate vanA/vanB-mediated VRE in this poultry-associated WGS subset. The observed phenotypic non-susceptibility without canonical acquired determinants should therefore be interpreted cautiously as borderline or unexplained findings that warrant confirmatory testing and/or deeper mechanistic follow-up rather than evidence of transferable last-line resistance.

From a veterinary One Health perspective, the resistome patterns observed here are epidemiologically plausible because antimicrobial consumption in food-producing animals in Europe is dominated by classes such as penicillins and tetracyclines, with additional contributions from macrolides and other groups depending on country and production sector. This usage landscape provides sustained selection pressure that can enrich tetracycline- and macrolide-associated determinants and shape the ecological persistence of enterococcal lineages in production environments. In contrast, clinically significant vancomycin-resistant E. faecium is primarily tracked in human invasive surveillance, where resistance levels vary substantially across countries and have remained a major public health concern in parts of Europe. Therefore, our One Health interpretation focuses on production-associated selection signatures and surveillance-relevant genomic payloads, while explicitly avoiding conflation with hospital-driven VRE epidemiology [48].

The extremely high vancomycin MICs observed in a small subset of isolates warrant targeted confirmation, as canonical glycopeptide resistance determinants (e.g., vanA/vanB operons) were not recovered by the applied gene-screening workflow. This discordance should be interpreted as a priority for focused re-analysis using dedicated glycopeptide-resistance callers and/or manual inspection of assemblies, alongside phenotypic re-testing, rather than as evidence for van-independent high-level resistance.

Finally, several limitations are intrinsic to the study design and should be stated as strengths of rigor rather than weaknesses. The WGS subset is built specifically for mechanistic/genomic profiling and should not be used to estimate population prevalence. Virulence-factor screening is database- and threshold-dependent, and VF catalog completeness varies across species and lineages. Moreover, gene presence does not equal expression, and host-specific pathogenicity cannot be inferred without phenotypic assays (e.g., biofilm/gelatinase/hemolysis) or infection models. These constraints, however, do not diminish the One Health value of the dataset; instead, they clarify the appropriate inference layer: our results map the distribution of virulence- and resistance-associated genetic potential among poultry-derived enterococci with genomic resolution, providing a rational basis for hypothesis generation, risk-stratified surveillance, and targeted functional follow-up in future work. A key limitation is the non-random, purposive selection of the WGS subset, which precludes prevalence inferences for ST frequencies and gene detection rates. Therefore, the One Health interpretation is framed as surveillance-oriented and hypothesis-generating, highlighting genomic signatures and potential risk patterns rather than population estimates.

5. Conclusions

In this study, we provide a species-stratified, genome-resolved overview of virulence-associated genetic potential and acquired antimicrobial resistance in poultry-derived Enterococcus isolates with paired phenotypic MIC data. The virulome displayed a clear species-dependent architecture: E. faecalis carried a broader and more coherent set of adhesion/biofilm- and regulation-linked determinants, whereas E. faecium showed a comparatively limited high-confidence virulence-gene repertoire in this dataset. In parallel, the acquired resistome profiles and phenotypic non-susceptibility patterns underscored that poultry-associated enterococci can harbor clinically relevant multidrug resistance determinants, reinforcing their relevance as One Health sentinels.

Taken together, these findings support a surveillance-oriented interpretation: poultry-derived enterococci may combine genetic features compatible with persistence/colonization with acquired resistance determinants, but the co-occurrence structure is primarily shaped by species and lineage background rather than by a universal coupling of virulence and resistance. Future work should integrate lineage-aware comparative genomics with targeted functional assays (e.g., biofilm/gelatinase activity) and mechanism-focused follow-up for genotype–phenotype discordances to refine risk stratification across the animal–food–environment–human interface.

Supplementary Materials

The following supporting information can be downloaded at: https://www.mdpi.com/article/10.3390/vetsci13020187/s1, Table S1. Acquired antimicrobial resistance gene repertoire of the WGS subset (n = 31) identified using the Comprehensive Antibiotic Resistance Database (CARD). The table provides isolate-level gene presence/absence together with the corresponding hit-level output retained for reproducibility. Table S2. Phenotypic antimicrobial susceptibility dataset for the study collection, including minimum inhibitory concentration (MIC) values and interpreted susceptibility categories (S/I/R) based on the applied breakpoints. For genotype–phenotype comparisons in this manuscript, analyses were restricted to isolates with whole-genome sequencing (WGS subset, n = 31). Table S3. Multilocus sequence typing (MLST) results for the WGS subset (n = 31), including sequence type (ST) assignments and species inference based on the MLST profiles. Table S4. Virulence-associated gene repertoire of the WGS subset (n = 31) screened against the Virulence Factors Database (VFDB). The table summarizes isolate-level gene presence/absence and retains the underlying hit metrics for transparency; virulence determinants are interpreted as genetic potential rather than clinical virulence. Figure S1. MLST sequence type distribution in the whole-genome sequenced Enterococcus faecalis (E. faecalis) and Enterococcus faecium (E. faecium) subset (n = 31), stratified by species; “ST not assigned” indicates isolates with incomplete MLST profiles.

Author Contributions

Conceptualization, Á.K. and Á.J.; methodology, Á.K. and K.B.; software, E.K.; validation, Á.J.; formal analysis, E.K. and K.B.; investigation, G.T.; resources, Á.K.; data curation, Á.K.; writing—original draft preparation, Á.K. and L.R.; writing—review and editing, Á.K.; visualization, Á.K.; supervision, Á.J.; project administration, G.T.; funding acquisition, Á.J. All authors have read and agreed to the published version of the manuscript.

Institutional Review Board Statement

Not applicable. This study used bacterial isolates from an existing strain repository and did not involve any additional sampling or animal procedures performed for research purposes.

Informed Consent Statement

Not applicable.

Data Availability Statement

The data presented in this study are openly available in NCBI under BioProject at https://www.ncbi.nlm.nih.gov/bioproject/PRJNA1396911, accession number PRJNA1396911 (accessed on 10 January 2026).

Conflicts of Interest

The authors declare no conflicts of interest.

Funding Statement

Project no. RRF-2.3.1-21-2022-00001 was implemented with support provided by the Recovery and Resilience Facility (RRF), financed under the National Recovery Fund budget estimate, RRF-2.3.1-21 funding scheme.