Environmental microbial community signatures associated with Listeria spp. detection in German meat processing facilities

ABSTRACT

Listeria monocytogenes can persist in niches of meat processing environments despite routine cleaning and disinfection. Its persistence may depend not only on stress tolerance but also on interactions with resident microbial communities, which may promote or inhibit survival. However, these ecological relationships remain poorly understood. We combined 16S rRNA V3/V4 amplicon sequencing, culture-based detection, and multilocus sequence typing (MLST) to characterize microbial communities in six German meat processing facilities over one year. We examined associations among community structure, sampling sites, and the occurrence of Listeria spp., including L. monocytogenes. Microbial communities were dominated by core genera typical of food-processing environments, particularly Pseudomonas spp. and Acinetobacter spp., but differed significantly among facilities (PERMANOVA, P = 0.001; pairwise R² = 0.023–0.079), indicating facility-specific communities. Culture-based analyses detected Listeria spp. in 51 of 370 environmental samples (13.8%), mainly from drains (44/51, 86.3%). L. monocytogenes was detected in five of six facilities, with 19 of 21 isolates originating from drains (90.5%). MLST of 74 typeable L. monocytogenes isolates revealed high diversity, comprising 21 sequence types across 15 clonal complexes, with lineage II predominating (86.5%). Overall microbial community composition was significantly associated with Listeria spp. and L. monocytogenes presence (PERMANOVA, P = 0.001; R² = 0.0137 and 0.0083). In drains, amplicon sequence variants (ASVs) assigned to the genera Acinetobacter, Rhizorhapis, and Vagococcus showed positive associations with Listeria spp.-positive samples. Together, these findings suggest that drains are key ecological niches for Listeria spp. and that associated taxa may indicate drain communities linked to Listeria spp. recovery.

IMPORTANCE

Listeria monocytogenes is a major foodborne pathogen that can persist in meat processing environments despite routine cleaning and disinfection. Resident microbial communities may influence its survival, but longitudinal studies linking those communities with culture-based Listeria spp. detection remain limited. Here, we characterized microbial communities in six German meat processing facilities over 1 year using 16S rRNA gene amplicon sequencing and culture-based Listeria spp. detection, and MLST of L. monocytogenes isolates. We identified facility-specific microbial communities, identified floor drains as key niches for Listeria spp., and observed repeated recovery of different L. monocytogenes sequence types across facilities. In drains, ASVs assigned to the genera Acinetobacter, Rhizorhapis, and Vagococcus were positively associated with culture-positive samples, identifying candidate taxa that may reflect microbial conditions associated with Listeria spp. recovery. These findings highlight the importance of considering not only whether Listeria spp. are detected but also the resident microbial communities that may support their fitness.

INTRODUCTION

Listeria monocytogenes is a foodborne pathogen of major public health concern due to its ability to persist in food-processing environments (FPEs). In humans, L. monocytogenes can cause listeriosis, a severe foodborne illness with high hospitalization and mortality rates and the highest fatality rate among the major zoonotic diseases reported in the EU in 2024 (1). In addition to its relevance for human health, L. monocytogenes is a major veterinary pathogen, underscoring the One Health implications of this organism (2). A key factor contributing to the persistence of L. monocytogenes in FPEs is its ability to grow at refrigeration temperatures and survive long-term in niches, particularly in biofilms (3). Other Listeria spp. are also frequently detected in food-processing environments and serve as index organisms for L. monocytogenes (4).

However, physiological resilience alone does not fully explain persistence. Increasing evidence indicates that ecological interactions within resident microbial communities play a critical role in microbial fitness. Biofilms formed by dominant colonizers, such as Pseudomonas spp., can enhance the resilience of L. monocytogenes against disinfectants (5, 6), while other community members may modulate survival through nutrient competition, inhibitory metabolites, or changes in biofilm structure and local microenvironmental conditions (7).

Stable, facility-specific microbial communities have been described across diverse FPEs, including breweries, fruit processing facilities, and meat processing facilities (8–10). These resident microbiota form persistent assemblages shaped by the environment, raw materials, processing steps, and cleaning regimens (11). Until recently, the microbial communities of meat processing environments were comparatively underexplored, and substantial knowledge gaps remained regarding their composition, stability, and ecological drivers (8, 12). Recent metagenomic and longitudinal studies have started to address these gaps, revealing that meat processing environments harbor diverse, facility-specific, and persistent microbial communities, including previously undescribed taxa and biofilm-forming strains (11). Understanding these microbial ecosystems is essential for identifying how individual community members may facilitate or inhibit the persistence of L. monocytogenes (3). This knowledge may reveal antagonistic interactions or microbiota-derived inhibitory compounds that could be exploited for preventive purposes and for the development of targeted hygiene strategies (8, 13).

In this study, we characterized the microbiota of six German meat processing facilities over a 1-year period and investigated the composition of the microbial communities and their associations with the occurrence of Listeria spp. and L. monocytogenes. We combined longitudinal 16S rRNA gene sequencing with culture-based detection of Listeria spp. and L. monocytogenes to determine whether specific microbial community signatures are associated with Listeria spp. occurrence and persistence in meat processing environments. Multilocus sequence typing (MLST) was used to subtype L. monocytogenes isolates and identify patterns of repeated recovery indicative of persistence. Beyond characterizing facility-specific microbial communities, we identified taxa whose abundance or prevalence was positively or negatively associated with culture-based Listeria spp. detection. These taxa provide an ecological context for Listeria spp. occurrence and represent candidates for future experimental interaction studies. Together, these analyses link microbiome composition with Listeria spp. occurrence and environmental reservoirs in meat processing environments.

MATERIALS AND METHODS

Sampling and sample processing

Samples were collected from six German meat processing facilities and covered all major production areas, including receiving, cold storage, raw material handling, and finished product processing. Per area, at least one floor drain and one food-contact surface (FCS) were sampled, resulting in 12 sites per facility (14 in one larger facility). The facilities processed beef (F5), pork (F2, F3, F4, F6), or both (F1). Facilities processing the same animal species produced heterogeneous product portfolios, including ground meat, sausages, patties, and ready-to-eat products, and product types could vary over time, including within the same sampling week or day. Therefore, facilities could not be consistently stratified by specific final product type, and comparisons were interpreted at the facility level rather than as controlled comparisons among facilities producing identical final products. Each facility was sampled five times at ~10-week intervals. Sampling was conducted on Mondays before production started to allow for the recovery of the microbiota during the weekend after routine cleaning and disinfection on Fridays. At each site, two adjacent but non-identical areas were sampled, yielding 740 samples (370 paired swabs): one for microbiota profiling (16S rRNA V3/V4 sequencing) and one for culture-based detection of Listeria spp., including L. monocytogenes, according to ISO 11290-1:2017, with species-level identification by MALDI-TOF MS. For both FCS and drains, ~225 cm² were sampled using sponge-sticks with neutralization buffer (Neogen Corporation, Lansing, MI, USA). If present, standing water was additionally collected from drains. All samples were cooled during transport to the lab and processed within 12 h.

For the samples intended for sequencing, sponges were aseptically transferred into sterile VWR Blender Bags (Filter Bag 129-0734; Avantor, Radnor, PA, USA) containing an integrated filter membrane, homogenized in 50 mL buffered peptone water (Merck KGaA, Darmstadt, Germany) for 30 s at 230 rpm (Stomacher 400 Circulator, Seward Ltd, Worthing, West Sussex, UK), and the filtered homogenate was split into two 25-mL aliquots. Both aliquots were centrifuged (16,100 × g, 35 min, room temperature) to obtain bacterial pellets. One pellet was used for 16S rRNA V3/V4 gene sequencing, while the other was resuspended in 1 mL Brain Heart Infusion broth supplemented with 15% glycerol and stored at −80 °C.

DNA extraction

For 16S rRNA V3/V4 gene sequencing, bacterial DNA was extracted from the pellet using the DNeasy Blood & Tissue Kit (Qiagen, Hilden, Germany) following the manufacturer’s protocol, including the pre-treatment step for Gram-positive bacteria. With each DNA extraction batch, one positive extraction control (mock microbial communities from ZymoBIOMICS Microbial Community Standard D6300; Zymo Research Europe GmbH, Freiburg, Germany) and one negative extraction control were included to assess extraction performance and monitor potential contamination. Negative extraction controls consisted of clean sponges that were processed in bags in the same way as the samples.

Samples were eluted in 50 µL of preheated Buffer ATE (65°C). DNA concentration was subsequently quantified using the Quantus Fluorometer with QuantiFluor dsDNA Dye (Promega, Madison, WI, USA). DNA was stored at −20°C until sequencing. Library preparation for V3/V4 16S rRNA gene sequencing was done using barcoded 341F (forward) and 806R (reverse) primers (14) by Novogene GmbH (Novogene, Munich, Germany). The PCRs were carried out with 15 μL of Phusion High-Fidelity PCR Master Mix, 0.2 μM of forward and reverse primers, and 10 ng template DNA. Thermal cycling consisted of initial denaturation at 98℃ for 1 min, followed by 30 cycles of denaturation at 98℃ for 10 s, annealing at 50℃ for 30 s, and elongation at 72℃ for 30 s, followed by a final 5-min hold at 72℃. Library preparation was done by Novogene GmbH (Novogene, Munich, Germany) using the NEBNext Ultra II DNA Library Prep Kit (New England Biolabs, Ipswich, MA, USA), and sequencing was performed on an Illumina NovaSeq 6000 (PE 250 bp reads, 50k reads/sample).

Sequencing data analyses

Sequencing data analyses were carried out in RStudio, R v. 4.5.0 (15) unless otherwise noted, using the DADA2 package v. 1.36.0 (16). Demultiplexed paired-end reads were analyzed as outlined in the DADA2 analysis.R script available on GitHub (see Data Availability). Briefly, sequences were inspected for quality and used to learn error rates. Amplicon sequence variants (ASVs) were inferred using DADA2. Forward and reverse reads were merged, sequences of non-target length were removed, and chimeras were identified and removed using the consensus method. Taxonomic assignment was performed using the SILVA v138.1 reference database (17). Negative extraction controls were used for contaminant identification using prevalence-based contaminant identification with decontam package v. 1.30.0, as outlined in the decontam.R script available on GitHub (18) in R version 4.5.1. Identified contaminant ASVs were removed prior to statistical analyses.

Statistical analysis of microbial community data

Statistical analyses were conducted as outlined in the following sections and in the statistical_analysis.R script provided on GitHub.

Beta diversity analysis

To account for the compositional nature of data, zero counts in the ASV table were imputed using the Count Zero Multiplicative (CZM) approach with zCompositions package v. 1.5.0.5 (19). Centered log-ratio (CLR) transformation was then applied to the filtered ASV table prior to principal component analysis (PCA) and plotting of PCA biplots to visualize clustering of samples. To statistically assess differences in beta diversity (i.e., microbiota composition), Aitchison distances were computed, and permutational multivariate analysis of variance (PERMANOVA) was conducted using the pairwiseAdonis package v. 0.4.1 (20). Models tested main effects and interactions among facility, sampling area, and Listeria spp. and Listeria monocytogenes presence, with 999 permutations. P-values were Bonferroni-adjusted.

Differential abundance analysis

MaAsLin3 v. 1.1.2 was used to identify ASVs associated with Listeria spp. or L. monocytogenes (21). Analysis was applied to all samples, as well as to a subset of samples from drains, due to more frequent recovery of Listeria spp. from drain samples. The analyses were conducted using an untransformed ASV table and total-sum scaling (TSS). Associations with a default false discovery rate-adjusted significance threshold of 0.1 were considered for interpretation.

Core microbiota analysis

Core microbiota were defined at the genus level using untransformed relative abundance data. For each facility, genera present in at least 90% of samples at a minimum relative abundance of 0.0001 were identified as core genera, and their mean relative abundance and standard deviation were calculated. Overall core genera were also identified, using the same criteria, for all samples collected in this study. Furthermore, the 10 most abundant genera per facility were plotted using ggplot2 v. 4.0.0 (22) to visualize the most abundant genera in each facility.

Alpha diversity analysis

Alpha diversity was assessed using the Shannon index calculated from the contaminant-filtered ASV count table after rarefaction to the minimum sequencing depth of 23,689 reads per sample. Rarefaction was performed using the rrarefy function in the vegan package v. 2.7.3 to account for differences in sequencing depth among samples, and Shannon diversity was calculated from the rarefied ASV count table using the diversity function in vegan. Rarefaction was used only for alpha diversity estimation to standardize sequencing depth before calculating within-sample diversity, whereas CLR transformation and Aitchison distances were used for beta diversity analyses because between-sample comparisons of microbiome data require approaches that account for their compositional nature. For longitudinal analyses, sampling sites were defined as repeatedly sampled locations identified by the sampling site code within each facility. The R code used for alpha diversity and longitudinal community-change analyses is provided in longitudinal_analysis.R in the GitHub repository listed under Data Availability.

Longitudinal community change

Temporal changes were quantified as the mean Aitchison distances in microbial community composition between samples collected from the same sampling unit at consecutive sampling events. Sampling units were defined by the sampling site code. Facility-stratified temporal trends in Shannon diversity and longitudinal community change were analyzed using linear mixed-effects models fitted with lme4 v. 2.0.1 and lmerTest v. 3.2.1. Models included sampling date as a numeric fixed effect and sampling site, or sampling unit for longitudinal community change, as a random intercept. Model outputs were extracted using broom.mixed v. 0.2.9.7, and P-values from facility-stratified models were adjusted using the false discovery rate method.

Sequencing and culture concordance

To assess the association between sequencing-based Listeria spp. signal and culture-based Listeria spp. detection, a logistic regression model was fitted using culture positivity as the binary response variable and Listeria species ASV abundance as the predictor.

Listeria spp. enrichment

Samples collected for culture-based detection of Listeria spp. and L. monocytogenes were processed according to ISO 11290-1:2017 to detect viable Listeria spp., as sequencing-based methods cannot differentiate viable from dead bacteria. Briefly, sample homogenates were subjected to primary enrichment in half-Fraser broth, followed by secondary enrichment in Fraser broth. After incubation, aliquots from the enrichment cultures were streaked onto Chromocult Listeria Agar according to Ottaviani and Agosti and GranuCult Oxford Agar, each supplemented with the respective Listeria-selective supplement according to the manufacturer’s instructions (all Merck KGaA, Darmstadt, Germany). Selective agar plates were evaluated after 48 h of incubation. For each positive sample, one colony showing morphology typical of Listeria spp. on the ALOA plate was subcultured on plate count plates for 24 h and subjected to MALDI-TOF MS identification to confirm the species. Colonies were prepared using a formic acid extraction protocol: bacterial material from a single colony was suspended in 300 µL deionized water, ethanol was added, and the suspension was centrifuged. The dried pellet was extracted with 70% formic acid and acetonitrile, and 1 µL of the supernatant was spotted onto a MALDI target plate, dried, overlaid with 1 µL HCCA matrix solution, and analyzed using the MALDI Biotyper Sirius system with the BDAL database version 2023 (Bruker Daltonics, Bremen, Germany). When colonies with different morphologies suggestive of Listeria spp. were observed from the same enriched sample, representative colonies of each morphology were subcultured separately and identified by MALDI-TOF MS. Thus, multiple Listeria species could be obtained from a single culture-positive environmental sample.

MLST of Listeria monocytogenes isolates

MLST was used to subtype L. monocytogenes isolates, assign sequence types and clonal complexes, and assess patterns of persistence within and across facilities. It was performed according to the standardized scheme (23, 24), targeting seven housekeeping genes of L. monocytogenes abcZ, bglA, cat, dapE, dat, ldh, and lhkA.

Isolates originated from environmental samples collected during the present study and from 55 additional isolates provided by the participating facilities as part of their routine environmental monitoring conducted between 2020 and 2025. Twelve of the latter were collected outside the sampling period of the present study (2024–2025).

Genomic DNA was extracted by heat lysis based on a previously described protocol (25), with minor modifications. For each isolate, a single colony was inoculated into 5 mL Brain Heart Infusion broth and incubated for 24 h. The entire overnight culture was centrifuged to obtain a cell pellet. Cell pellets were resuspended in 250 μL molecular-grade water, incubated at 99°C for 10 min, and centrifuged at 5,000 × g for 5 min. Then, 200 μL of the supernatant was transferred to a clean microcentrifuge tube and used as a DNA template in PCR amplification.

PCR amplification was performed using SensiFAST SYBR No-ROX Master Mix (Bioline, London, UK) on a Bio-Rad CFX96 real-time PCR system. Each 20 µL reaction contained 10 µL 2× SensiFAST SYBR No-ROX Master Mix, forward and reverse primers at a final concentration of 0.4 µM each, 7.4 µL molecular-grade water, and 1 µL template DNA diluted 1:1,000 for each sample before amplification. The same primer concentration was used for all MLST targets. Cycling conditions followed the Institut Pasteur MLST recommendations: 95°C for 5 min; 40 cycles of 95°C for 10 s, 45 or 52°C for 10 s, and 72°C for 20 s; followed by 72°C for 2 min. An annealing temperature of 45°C was used for bglA, cat, and ldh and 52°C for all other targets.

Sanger sequencing was performed by Eurofins Genomics (Ebersberg, Germany). The loci abcZ, cat, dat, and ldh were sequenced in a single direction using the forward primer only, while bglA, dapE, and lhkA were sequenced bidirectionally using both forward and reverse primers. Sequencing chromatograms were imported into Geneious Prime version 2025.0.3 for quality assessment. Chromatograms were inspected for signal intensity, base-calling accuracy, and the presence of mixed or ambiguous peaks. For bidirectionally sequenced loci, forward and reverse reads were assembled and used to generate consensus sequences. High-quality single-direction sequences and bidirectional consensus sequences were exported as FASTA files and uploaded to the Institut Pasteur BIGSdb-Lm database (26) for allele designation and sequence type (ST) assignment according to the canonical seven-locus MLST scheme.

Persistence was defined as the repeated isolation of Listeria monocytogenes isolates belonging to the same ST within the same facility at sampling times >6 months apart. This conservative time-based criterion was adopted because no universally accepted definition of Listeria spp. persistence in food-processing environments currently exists (27). Repeated recovery of the same ST at intervals < 6 months was described as recurrent detection. Because MLST has lower discriminatory power than cgMLST or whole-genome sequencing, this definition identifies persistence based on repeated recovery of the same ST within a facility over time but does not confirm isolate similarity at the whole-genome level.

RESULTS

Controls indicated reproducible sequencing performance and low contamination

Mock community samples showed consistent results across runs, with all expected standard taxa recovered (Fig. S1). The observed relative abundances did not fully match the theoretical composition due to lower extraction efficiency for Staphylococcus aureus, suggesting an extraction bias against this species and potentially other Gram-positive taxa. Negative controls produced zero reads in 11 out of 16 cases; rare low-read contaminants were identified in 5 samples (mean: 1,309.4 reads) and used as input for the decontam pipeline (18).

Distinct facility microbial communities shared core genera of Pseudomonas and Acinetobacter

Among the 370 samples submitted for 16S rRNA gene sequencing, 366 generated sufficient sequencing reads for downstream analysis, while 4 samples yielded no reads and were, therefore, excluded from subsequent microbiota analyses.

Microbial community composition differed significantly among facilities (Fig. 1). Pairwise PERMANOVA comparisons showed that all facilities differed significantly from each other (P = 0.001), but the proportion of variation explained by facility differed among pairwise comparisons. The largest effect sizes were observed for comparisons involving F6, particularly F1 vs F6 (R² = 0.079), F4 vs F6 (R² = 0.076), F5 vs F6 (R² = 0.062), and F6 vs F2 (R² = 0.062). In contrast, the smallest effect size was observed for F3 vs F2 (R² = 0.023).

Fig 1.

Pairwise PERMANOVA effect sizes (R²) for microbial community composition between meat processing facilities. Bars show the proportion of variance explained by facility (R²) for each pairwise comparison between the six meat processing facilities. All pairwise comparisons were significant (PERMANOVA, P = 0.001), indicating significant differences in microbial community composition among facilities.

Genus-level relative abundance profiles varied across facilities and sampling times (Fig. 2). Pseudomonas spp. and Acinetobacter spp. were repeatedly detected at high relative abundance in all six facilities, but their relative contributions varied across facilities. Pseudomonas spp. was prominently present in several facilities, whereas Acinetobacter spp. was particularly abundant in F3 and F5. Consistent with their broad distribution, Pseudomonas spp. and Acinetobacter spp. were the only genera identified as core genera across all samples, occurring in more than 90% of samples at a relative abundance of at least 0.0001. Their mean relative abundances were 0.17 ± 0.24 and 0.16 ± 0.22, respectively.

Fig 2.

Relative abundance of genera in six meat processing facilities over time. Stacked bar plots show the relative abundance of bacterial genera in individual samples collected from six meat processing facilities across five sampling times (T1–T5). Each bar represents one sample. For each facility, the 10 genera with the highest mean relative abundance across all samples from that facility are shown individually; all remaining genera are grouped as “Other.” The bar labeled “Mean” (far right) represents the mean relative abundance of these genera across all samples from the respective facility.

Core genera were defined as genera present in at least 90% of samples from a given facility at a minimum relative abundance of 0.0001. Facility-specific core genera differed among facilities (Fig. 4) and overlapped with several of the dominant genera shown in the facility-level mean profiles in Fig. 3. In F2, Pseudomonas spp. and Acinetobacter spp. were the only core genera, whereas the remaining facilities contained additional facility-specific core genera. These included Chryseobacterium, Janthinobacterium, Kocuria, and Rhodococcus species in F1; Microbacterium spp. in F3; Aeromonas spp. in F4 and F5; Chryseobacterium, Flavobacterium, and Janthinobacterium species in F5; and Enhydrobacter, Exiguobacterium, Kocuria, Microbacterium, Paracoccus, Psychrobacter, Rhodococcus, and Rothia species in F6. Several facility-specific core genera, including Rhodococcus, Psychrobacter, Flavobacterium, and Chryseobacterium species, were also among the abundant genera shown in Fig. 3.

Fig 3.

Ten genera with the highest relative abundance in each facility. This figure provides an expanded view of the rightmost “mean” bars shown in Fig. 2. For each facility, the mean relative abundance of the 10 genera with the highest mean relative abundance across all samples is displayed.

At the ASV level, fewer core taxa were identified than at the genus level, and core ASVs differed among facilities (Fig. 4). Using the same prevalence and relative abundance criteria as above, no core ASVs were identified in F2 or F3, indicating lower ASV-level core stability in these facilities. F4 and F5 each contained one core ASV (ASV42 and ASV16, respectively), both assigned to Pseudomonas spp., which was among the 10 most abundant genera in these facilities. In F1 and F6, several core ASVs were identified. In F1, core ASVs were assigned to Rhodococcus ASV1, Kocuria ASV7, Janthinobacterium ASV26, and Pseudomonas ASV64, whereas in F6, core ASVs were assigned to Rhodococcus ASV1, Rothia ASV11, Acinetobacter ASV14, and Enhydrobacter ASV25. These genera were also represented in the facility-level dominant genus profiles shown in Fig. 3.

Fig 4.

Facility-specific composition of core taxa (≥90% prevalence, ≥0.0001 relative abundance). Bars show the mean relative abundance of facility-specific core taxa across samples from each facility, with all remaining taxa grouped as “Other—non-core taxa.” The left panel shows core genera, whereas the right panel shows core ASVs aggregated to genus level for visualization. Genus colors were kept consistent across panels to facilitate comparison.

Temporal dynamics of microbial communities

Alpha diversity, as measured by the Shannon index, showed temporal variation across the study period (Fig. 5). Shannon diversity varied significantly across sampling events (LRT, q = 7.74 × 10⁻¹³), whereas production area was not significantly associated with Shannon diversity (q = 0.419). However, the interaction between production area and sampling event was significant (q = 0.025), indicating that temporal patterns in Shannon diversity differed among production areas. In addition, the categorical sampling-event model provided a significantly better fit than a model assuming a linear time effect (q = 2.00 × 10⁻¹⁰).

Fig 5.

Temporal dynamics of alpha diversity across production areas. Shannon diversity is shown for samples collected across the study period. Points represent individual samples and are colored by production area. Thick colored lines indicate area-level mean curves, and thin colored lines show the upper and lower limits of the 95% confidence intervals. Sampling event was modeled as a categorical predictor to assess temporal variation without assuming a linear trend.

As a measure for beta diversity, longitudinal community change, quantified as the Aitchison distance from the previous sampling event, was reported. It varied significantly across sampling events (LRT, q = 8.29 × 10⁻⁸) (Fig. 6). In contrast to Shannon diversity, production area was significantly associated with longitudinal community change after accounting for sampling event (q = 0.022), indicating that the magnitude of community turnover differed among production areas. The interaction between production area and sampling event was also significant (q = 0.0099), indicating that temporal patterns in community change differed among production areas. Consistent with the non-monotonic mean curves, the categorical sampling-event model provided a significantly better fit than a model assuming a linear time effect (q = 8.29 × 10⁻⁸).

Fig 6.

Longitudinal community change across production areas. Longitudinal community change was quantified as the Aitchison distance between samples collected from the same sampling unit at consecutive sampling events. Points represent individual sampling units and are colored by production area. Thick colored lines indicate area-level mean curves, and thin colored lines show the upper and lower limits of the 95% confidence intervals. Sampling event was modeled as a categorical predictor to assess temporal variation without assuming a linear trend.

Facility-stratified mixed-effects models were refitted, with the sampling event modeled as a categorical predictor, to assess temporal variation without assuming a linear trend. Shannon diversity varied significantly across sampling events in five of six facilities after FDR correction, namely, facilities F1, F2, F3, F5, and F6 (q = 1.0 × 10⁻⁵ to 0.0367) (Fig. S2), whereas no significant temporal variation was detected in facility 4 (q = 0.181).

Longitudinal community change, measured as the Aitchison distance from the previous sampling event, varied significantly over time in facilities F1, F2, F4, and F6 (q = 1.5 × 10⁻⁴ to 0.0262) (Fig. S3), while facilities F3 and F5 did not show significant temporal variation after FDR correction. These results indicate that temporal microbiome dynamics were facility-specific and not consistently monotonic across the study period.

Listeria spp. and L. monocytogenes were significantly more likely to be detected in drains than on food-contact surfaces

Of 370 environmental samples, 51 (13.8%) were positive for Listeria spp. Using a site-based environmental-monitoring zone classification, food-contact surfaces were classified as Zone 1 and drains as Zone 3. Accordingly, Listeria spp.-positive samples were recovered mainly from Zone 3 drain sites (44/51), whereas fewer positives were detected on Zone 1 food-contact surfaces (7/51). These positives were distributed across the predefined production areas as follows: receiving areas (n = 8), cold storage areas (n = 9), raw material handling areas (n = 26), and finished product processing areas (n = 8), with the highest number detected in raw material handling areas. Because multiple Listeria spp. were recovered from some positive samples, these 51 positive samples yielded 58 isolates representing 6 species (Table 1). L. monocytogenes was detected in five facilities, with the highest numbers in F2 and F3. Most L. monocytogenes isolates from the present survey were recovered from drains (19/21), whereas only two were recovered from FCS. Overall, Listeria spp.-positive samples were significantly more frequent in drains than on FCS (44/180 vs 7/190; Fisher’s exact test, P = 2.55 × 10⁻⁹; OR = 8.46). Co-occurrence of multiple Listeria spp. was observed in 7 out of 51 positive samples (13.7%), most commonly involving L. monocytogenes and L. innocua (5/7).

TABLE 1.

Distribution of Listeria species across the six processing facilities^a

Species	Facility						Total
	F1	F2	F3	F4	F5	F6
L. monocytogenes	1	6 (2)	6	0	4	4	21 (2)
L. innocua	0	4 (1)	8	1	1	7 (2)	21 (3)
L. welshimeri	4 (1)	2 (1)	3	1	0	0	10 (2)
L. grayi	0	0	0	2	1	0	3
L. seeligeri	0	0	0	0	0	2	2
L. newyorkensis	0	0	0	0	0	1	1
Total isolates	5 (1)	12 (4)	17 (0)	4 (0)	6 (0)	14 (2)	58 (7)

^a Numbers represent positive isolates per facility and species. Values in parentheses indicate food-contact-surface-positive samples. All other positive samples were obtained from drains.

Microbial community composition differed significantly between Listeria spp. positive and Listeria spp. negative samples, with specific ASVs associated with Listeria spp.-positive samples

Logistic regression revealed no significant association between Listeria ASV abundance and culture-based detection of Listeria spp. (estimate = −0.0012 ± 0.00086, P = 0.162), indicating limited concordance between sequencing-based Listeria spp. signal and culture positivity.

At the community level, the presence of Listeria spp. and L. monocytogenes was associated with minor but significant shifts in microbiota composition (PERMANOVA: R² = 0.0137 and 0.0083, respectively; both P = 0.001). Since Listeria spp.-positive samples were predominantly recovered from non-food-contact surfaces (i.e., drains), we conducted further differential abundance analyses on a subset containing only the drain samples (Fig. 7). In drain samples, three ASVs were significantly associated with culture-based Listeria spp. status at an FDR threshold of 0.1. All three were detected in the abundance model and were positively associated with Listeria spp.-positive samples: Acinetobacter ASV167 (β = 6.50, q = 0.064), Rhizorhapis ASV2147 (β = 4.19, q = 0.029), and Vagococcus ASV447 (β = 2.82, q = 0.029). No ASVs were significantly associated in the prevalence model, and no ASVs showed significant negative associations after FDR correction. In contrast, no ASVs were significantly associated with culture-based L. monocytogenes status in drain samples in either the abundance or prevalence model after FDR correction.

Fig 7.

ASVs associated with Listeria spp. in drain samples. Differential abundance of ASVs (circles) and log-odds of ASV presence (triangles) in Listeria spp.-positive samples are shown. Positive β coefficients indicate positive associations with Listeria spp.-positive drain samples, whereas negative β coefficients indicate negative associations. Associations were considered significant at an FDR-adjusted threshold of q < 0.1.

MLST revealed high genetic diversity and persistent sequence types

For MLST analysis, a total of 76 L. monocytogenes isolates were included, comprising 21 isolates recovered from environmental samples collected during the present study and 55 additional isolates provided by the participating facilities from their routine monitoring programs. They originated predominantly from food samples and FCS, whereas isolates recovered during our sampling campaign were mainly derived from drains. Persistence and recurrent detection were, therefore, assessed using this combined isolate collection rather than isolates from the 1-year sampling campaign alone. Of these, 74 yielded complete allelic profiles, whereas 2 could not be assigned to a sequence type (ST) because some alleles remained unassignable due to poor sequence quality, despite repeated resequencing attempts and DNA re-extraction using the DNeasy Blood & Tissue Kit (Qiagen, Hilden, Germany).

Among the 74 typeable isolates, most belonged to lineage II (86.5%), while lineage I accounted for 13.5% (Fig. 8). Among them, 15 clonal complexes (CCs) and 21 STs were identified: the most frequent STs were ST9 (n = 23), ST37 (n = 6), and ST451 (n = 5). ST diversity differed between facilities, with F4 showing the highest diversity and harboring 10 distinct sequence types. Several STs were repeatedly detected within the same facilities. Among the persistent STs (repeated isolation >6 months), ST204 was detected three times in F2, all from FCS. ST9 was detected 17 times in F3, predominantly from drains (12/17), but also from other non-food-contact surfaces, including the wheels of a smoking trolley (2/17) and the underside of a vat (1/17), as well as from FCS (2/17). In F4, ST87, ST121, and ST451 were detected three, three, and five times, respectively, and all corresponding isolates were recovered from FCS. ST9 was detected twice in F5, both times from drains (Fig. 9). In contrast, repeated recovery of ST3406, ST236, ST517, ST37, ST9, and ST223 over shorter intervals (<6 months) was classified as recurrent detection rather than persistence.

Fig 8.

Distribution of Listeria monocytogenes isolates assigned by MLST across lineages, clonal complexes, sequence types, and facilities. Hierarchical alluvial diagram showing the distribution of the 74 typeable L. monocytogenes isolates across genetic lineages, clonal complexes (CC), sequence types (ST), and the six processing facilities. Colors represent how often a given sequence type was observed within a facility.

Fig 9.

Detection time periods of persistent Listeria monocytogenes STs. Points indicate individual isolates, and horizontal lines indicate the period between first and last detection of each ST within the respective facility. Labels indicate the total detection period and number of isolates. Dashed vertical lines indicate the sampling period of the present study. Circles represent food-contact surface isolates, and triangles represent non-food-contact surface isolates.

DISCUSSION

Across all facilities, genus-level abundance profiles were characterized by the repeated dominance of Acinetobacter spp. and Pseudomonas spp., which were the only genera identified as core genera across the entire data set. Their broad distribution and high relative abundance suggest that these genera are particularly well adapted to environmental conditions present in meat processing facilities, including low temperatures, moisture, nutrient limitation, and repeated exposure to cleaning and disinfection. Pseudomonas spp. are frequently reported as dominant members of food-processing microbiota and include psychrotolerant, surface-associated, and biofilm-forming taxa that can persist under conditions typical of processing environments (9, 11, 28). Acinetobacter spp. have also been repeatedly detected in food-processing environments and have been associated with cold, moist, and nutrient-limited niches, supporting their role as important members of resident FPE microbiota (11, 29). Together, the prevalence of these genera suggests that meat processing environments select for psychrotolerant and biofilm-associated microorganisms (30).

In addition to these shared core genera, several other taxa contributed prominently to facility-specific community profiles, including Rhodococcus, Psychrobacter, Chryseobacterium, Flavobacterium, and Aeromonas species. Their distribution varied between facilities indicating that local conditions such as raw material input, moisture availability, surface properties, temperature, and sanitation efficiency shape the composition of resident microbial communities. Many of these genera have also been reported from cold, moist, or nutrient-limited food-processing environments and include taxa capable of surface association and biofilm formation (31, 32), for example, Flavobacterium spp. on solid surfaces (33) or Aeromonas spp. on glass and polystyrene (34). Thus, while Acinetobacter spp. and Pseudomonas spp. represented broadly distributed core members, the additional predominant genera likely reflect facility-specific ecological selection within individual processing environments.

Despite shared dominant taxa, community composition differed between facilities, indicating distinct resident microbial communities in individual facilities. This variation likely reflects differences in facility characteristics, such as size, hygiene practices, and processed animal species. Within-facility variation further indicates that the local microenvironment, influenced by surface properties, moisture, and cleaning efficiency, shapes microbial communities (8, 35).

Core microbiota analysis showed that microbial community stability depended on taxonomic resolution and facility context. At the genus level, Pseudomonas spp. and Acinetobacter spp. formed a shared core across all facilities, indicating common environmental selection pressures in meat processing environments. In contrast, no ASV was shared as a core taxon across all facilities, and no core ASVs were identified in F2 or F3, although both facilities contained core genera (Pseudomonas and Acinetobacter species). This suggests that similar ecological niches may be occupied by different ASVs across facilities or over time, resulting in genus-level stability but ASV-level turnover. This may also be reflective of the similar growth requirements between ASVs of the same genus. Similar genus-level stability together with finer-scale taxonomic variability has been reported in other food-processing environments (11, 36). Facilities lacking core ASVs, such as F2 and F3, may, therefore, represent more temporally dynamic microbial communities. This interpretation is consistent with the temporal analyses, which showed significant changes in Shannon diversity in F2 and F3 and significant longitudinal community change in F2. Such community turnover may reduce the likelihood that individual ASVs persistently meet core-microbiota criteria across sampling events. The temporal analyses further indicated that changes in Shannon diversity and longitudinal community composition did not follow a simple monotonic trend over the study period. Instead, the better fit of models treating sampling event as a categorical predictor suggests event-specific, non-linear temporal dynamics. These patterns may reflect episodic or facility-specific influences, such as variation in raw material inputs, moisture levels, sanitation efficacy, or microbial recolonization after cleaning, rather than a uniform directional trend over time.

These patterns may also provide ecological context for Listeria spp. occurrence. Facilities F2 and F3 had the highest numbers of L. monocytogenes detections in the present study, corresponding to detection prevalences of 8.6% in F2 (6/70 samples) and 10.0% in F3 (6/60 samples). Because most Listeria spp.-positive samples were recovered from drains, this pattern should be interpreted in the context of wet, difficult-to-sanitize niches. Temporally dynamic communities in wet or difficult-to-sanitize areas may indicate repeated microbial input, disturbance after sanitation, or changing nutrient and moisture conditions, all of which could create opportunities for Listeria spp. introduction, survival, or repeated recovery. Conversely, more stable communities, such as those observed in cold rooms, may reflect more constant environmental conditions, lower rates of microbial introduction and turnover, or less frequent disturbance. However, whether community stability inhibits or favors Listeria spp. cannot be inferred from community data alone, because stable communities may either resist invasion or provide persistent niches that support Listeria spp. survival.

Notably, production areas were distinguished more strongly by community composition than by alpha diversity, suggesting that functional zones differed primarily in the identity of their resident taxa rather than in overall within-sample diversity. This likely reflects area-specific selection pressures, including differences in moisture, temperature, organic residue availability, surface properties, and cleaning efficiency, which shape the composition and temporal stability of microbial communities in different production areas (8, 37). Together, these findings indicate that Listeria spp. occurrence should be interpreted in the context of facility-specific and temporally dynamic microbial communities rather than as the result of single taxa alone.

Culture-based analyses revealed a pronounced site-specific distribution of Listeria spp. The higher detection frequency in drains than on FCS highlights the importance of wet, difficult-to-sanitize niches, which can promote biofilm formation and persistence of Listeria spp. (3). Similar site-dependent patterns have been described in other FPEs (11). This observation is consistent with the view that hygienically challenging environmental niches contribute to Listeria spp. presence in FPEs (38). By including both FCS and drains, the study provides a differentiated view of contamination patterns within FPEs.

The lack of concordance between sequencing-based Listeria spp. signal and culture-based detection may reflect methodological differences between the two approaches. Culture-based detection after enrichment identifies viable Listeria spp., whereas 16S rRNA gene sequencing detects DNA and does not distinguish between viable and non-viable cells. In addition, Listeria spp. may have been present at low abundance near the sequencing detection limit, while enrichment-based culture can recover viable cells present at low levels. Finally, the paired swabs used for microbiota profiling and culture-based detection were collected from adjacent but non-identical areas, which may have further reduced direct concordance between the two approaches.

Although Listeria spp. and L. monocytogenes presence were significantly associated with overall microbial community composition, the corresponding PERMANOVA effect sizes were small. Thus, culture-based Listeria spp. status explained only a limited proportion of the total variation in community composition. The low explanatory power indicates subtle compositional differences between Listeria spp.-positive and -negative samples rather than strong compositional separation.

ASV-level associations in drain samples should be interpreted as ecological correlations rather than evidence of direct interactions. In addition, because V3/V4 16S rRNA gene sequencing provides limited resolution for species-level taxonomic assignment, ASVs assigned to the genera Acinetobacter, Rhizorhapis, and Vagococcus should not be interpreted as species-level identifications. Rather, they represent genus-assigned sequence variants, and these genera may include ecologically diverse members with potentially different interactions with Listeria spp. Their positive associations with Listeria spp.-positive drain samples may reflect shared preferences for wet, nutrient-rich, or biofilm-associated niches that also support Listeria spp. survival or repeated recovery. Genera such as Acinetobacter spp. (12, 39) and Flavobacterium spp. (8) have previously been reported in Listeria spp.-positive environments. It is also possible that Listeria spp. profit from the protected microenvironment that biofilms built by prolific biofilm producers like Acinetobacter spp. provide (31). These three ASVs may, therefore, represent candidate indicators of drain community states associated with Listeria spp. recovery. Their potential indicator value does not depend on being shared core ASVs across all facilities because informative ASVs may occur only in specific niches or vary in abundance with Listeria spp. status. However, they should not be interpreted as validated biomarkers or direct control targets. Their predictive value, specificity, and generalizability would require validation in independent data sets and targeted follow-up studies. Conversely, ASVs showing negative, non-significant trends may reflect habitat separation or environmental conditions less suitable for Listeria spp., such as drier surfaces, lower nutrient availability, or more effective sanitation, rather than direct inhibition. At the same time, a direct antagonistic effect is also plausible for some negatively associated taxa, for example, through bacteriocin-producing bacteria (LAB) such as Lactococcus spp., Streptococcus spp., Lactobacillus spp., and Enterococcus spp., which can suppress L. monocytogenes (40). Experimental co-culture and multi-species biofilm studies could determine whether these taxa directly affect Listeria spp. fitness, survival, or sanitizer tolerance.

Listeria spp. detection was performed according to ISO 11290, ensuring comparability with routine monitoring data. Nevertheless, enrichment-based detection is not equally sensitive for all Listeria spp. and may underestimate overall species diversity, particularly for less frequently recovered taxa such as L. seeligeri, L. grayi, and potentially L. ivanovii (41). L. seeligeri and L. grayi were detected in our study despite their potentially reduced recoverability. However, it is plausible that additional occurrences of these species were missed in other samples. Accordingly, our culture-based data likely provide a robust estimate of Listeria spp. occurrence but may not fully reflect the true species composition or relative abundance of Listeria spp. in sampled facilities.

MLST analysis revealed identical L. monocytogenes STs at product-associated sampling points, FCS, and floor drains within the same facilities. Recovery of isolates with identical STs from these sites is consistent with persistence, repeated introduction, or transfer within the same processing environment (42). Drain contamination may, therefore, provide useful information on local conditions and on the occurrence of L. monocytogenes lineages repeatedly recovered within a processing room (27, 43).

The predominance of lineage II may be explained by lineage-specific ecological differences. Lineage I is often associated with clinical cases and higher virulence potential, whereas lineage II is more frequently recovered from food and food-processing environments and includes clonal complexes repeatedly linked to environmental persistence (44). Accordingly, lineage II strains may be favored in FPEs because they are well adapted to refrigeration temperatures, nutrient limitation, surface-associated growth, and repeated cleaning and disinfection. This interpretation is consistent with the frequent detection of CC9, CC121, and CC204 in the present study, which have been repeatedly associated with meat processing environments and persistence (45, 46).

In the combined MLST data set, repeated recovery of specific STs within facilities met our definition of persistence. These STs belonged to CC9, CC11, CC121, and CC204, clonal complexes that include genotypes frequently reported from FPEs. While MLST does not provide sufficient resolution to infer persistence of identical strains, the repeated detection of STs within these clonal complexes supports the interpretation that these specific STs are well adapted to meat processing environments. This is consistent with previous studies reporting persistent or recurrent L. monocytogenes in food-processing environments, particularly among isolates belonging to CC9, CC121, and CC204 (42, 47), as well as studies from meat processing facilities where these clonal complexes have repeatedly been associated with persistence (45, 48). In addition, L. monocytogenes strains belonging to these recurrent clonal complexes have been associated with increased stress tolerance, survival after cleaning and disinfection, and the ability to colonize environmental niches and form biofilms (42, 48).

Altogether, this study provides a longitudinal view of microbial community structure and Listeria spp. occurrence in six German meat processing facilities by integrating microbial community profiling, culture-based detection, and MLST over 1 year. Our findings show that Listeria spp. recovery should be interpreted within the ecological context of facility-specific resident microbiota, wet and difficult-to-sanitize niches such as floor drains, temporal community dynamics, and co-occurring taxa associated with Listeria spp.-positive samples. This perspective moves environmental Listeria spp. surveillance beyond presence–absence detection, highlighting the importance of considering the microbial community background and niche characteristics of sampling sites. While routine monitoring should continue to prioritize high-risk sites such as FCS, microbiome-informed approaches may provide complementary information by helping to identify environmental niches with microbial community characteristics associated with L. monocytogenes survival or repeated recovery. Future studies integrating longitudinal microbial community profiling with routine environmental and end-product monitoring in commercial processing facilities could help determine whether microbiome-informed approaches provide added value for tracking microbial changes associated with Listeria spp. recovery over time. Repeated recovery of identical L. monocytogenes STs further supports the relevance of facility-associated reservoirs. In line with previous work showing that food-processing environments harbor facility-associated microbiomes linked to Listeria spp. occurrence (8) and diverse, facility-specific meat processing microbiomes (11), our results emphasize that Listeria spp. surveillance should be interpreted within the broader microbial ecology of each facility.

DATA AVAILABILITY

Sequencing data are available in the NCBI under BioProject PRJNA1261676. Metadata and sequencing data analysis code are available at https://github.com/kovaclab/environmental_microbiota_meat_processing_facilities.

SUPPLEMENTAL MATERIAL

The following material is available online at https://doi.org/10.1128/aem.01042-26.

File S1. aem.01042-26-s0001.csv.

ASV 16S analysis results, part 1.

File S2. aem.01042-26-s0002.csv.

ASV 16S analysis results, part 2.

File S3. aem.01042-26-s0003.xlsx.

Listeria spp. found during culture-based detection.

File S4. aem.01042-26-s0004.xlsx.

Metadata of the sponge samples taken during the study.

File S5. aem.01042-26-s0005.xlsx.

MLST results of L. monocytogenes isolates found during the study or provided from the facilities.

Supplemental figures. aem.01042-26-s0006.pdf.

Fig. S1 to S3.

Supplemental tables. aem.01042-26-s0007.docx.

Tables S1 to S8.

ASM does not own the copyrights to Supplemental Material that may be linked to, or accessed through, an article. The authors have granted ASM a non-exclusive, world-wide license to publish the Supplemental Material files. Please contact the corresponding author directly for reuse.