Microbial community structure of plant-based meat alternatives

The study addresses the growing consumption and market expansion of plant-based meat alternatives (PBMAs) amid environmental, public health, and sustainability concerns associated with conventional meat. Despite increased popularity, safety data—particularly microbiological risk profiles—are limited, and PBMAs not classified as ready-to-eat lack specific microbiological safety criteria. Potential hazards include chemical contaminants (pesticides, mycotoxins, heavy metals, mineral oil hydrocarbons) and microbiological risks, with recent evidence of mycotoxins in PBMAs suggesting current legal limits may be inadequate. Reducing food waste and improving shelf life are also priorities within sustainability goals. To fill knowledge gaps on microbial communities in PBMAs, the authors collected 32 products from Austrian supermarkets and investigated their microbial compositions using culture-dependent and -independent methods. They compared four product categories (pea- vs. soybean-based; minced vs. fibrous textures) and hypothesized that products within the same category would share more similar microbial communities than those between categories, due to similar protein processing.

The paper situates PBMAs in the broader context of environmental and health motivations for shifting to plant-based diets, noting continued growth in sales and product diversity. Prior literature has focused on market research and food technology, with limited studies on safety and microbiology. Documented hazards include pesticides, mycotoxins, toxic plant compounds, heavy metals, and mineral oil hydrocarbons; recent work reports notable mycotoxin levels in PBMAs and highlights gaps in regulatory coverage. Microbiological data are sparse, especially for non-ready-to-eat PBMAs lacking defined safety criteria. The review also references sustainability imperatives to reduce food loss and waste, emphasizing the role of microbial spoilage in losses and the need to understand product-specific microbiota for improved shelf life and packaging. It notes broader discussions about the potential health implications of reduced microbial exposure and loss of ancestral microbiota in Westernized lifestyles. Technologically, PBMA production commonly uses low- and high-moisture extrusion cooking (LMEC/HMEC), which effectively inactivates vegetative cells but can allow survival and potential reactivation of bacterial spores (e.g., Bacillus, Clostridia), suggesting post-extrusion contamination sources (spices, herbs, processing environment) may be influential.

Sample acquisition: Thirty-two PBMAs (vegan, pea- or soy-based; minced or fibrous texture) were purchased from four supermarket chains in Vienna, Austria (July 12–14, 2021), transported refrigerated, and stored at 4 °C. Selection excluded fermented products (e.g., tofu). Products varied in ingredients (4–27 per product; ~120 unique), shelf life (days to expiry or best-before), processing (pre-heating, freezing), and packaging (mostly modified atmosphere packaging, MAP). Sample preparation: Ten grams of each product were homogenized in 90 ml sterile PBS (120 s) in stomacher bags. Homogenates were centrifuged (300 × rcf, 2 min, RT) to remove coarse particles; supernatants were then centrifuged (3000 × rcf, 30 min, RT). Pellets were diluted 1:10 in PBS for immediate cultivation or stored at −20 °C (later cultivation) and −80 °C (DNA extraction). Cultivation: Two approaches were used. (i) Broad-range cultivation on non-selective and selective media (Columbia agar, VRBG, BHI + agar, PCA, Rose Bengal chloramphenicol agar, Baird Parker agar). Plates inoculated with 100 µl of 10−2 dilution were incubated at 37 °C under aerobic and semi-anaerobic conditions for 16–68 h, with further dilutions when overgrown. (ii) Targeted cultivation employed additional media (LB, nutrient agar, TSA, marine agar, corynebacterium agar, MRS, Pseudomonas agar F) at 25 °C for 48 h to isolate taxa detected by amplicon data but missed initially. Morphologically unique colonies were re-streaked for purity. Isolate identification and Sanger/WGS: DNA from pure cultures was extracted (Chelex-based lysis), followed by 16S rRNA gene PCR (27F/1492R) and, when needed, ITS2 PCR (ITS3/ITS4) for fungi. Amplicons were Sanger sequenced (one direction) and classified (RDP Naïve Bayesian Classifier via dada2). Selected isolates (potential pathogens and unclassified Enterobacteriaceae) underwent Oxford Nanopore MinION whole-genome sequencing (native barcoding). Long reads were filtered (Filtlong), assembled (Flye), polished (Racon ×4, Medaka), and analyzed with TORMES: taxonomy (Kraken2; 16S via Barrnap + RDP), MLST, annotation (Prodigal/Prokka), AMR (CARD via Abricate), virulence (VFDB via Abricate), and further classification (GTDB-Tk, rMLST). Bacillus paranthracis genomes were assessed with BTyper3. Good-quality LAB genomes were further analyzed (BlastKOALA, dbCAN3/CAZy, antiSMASH; candidate clusters confirmed by BLASTp). 16S rRNA gene amplicon sequencing: DNA was extracted from pellets (Qiagen DNeasy PowerFood Microbial Kit with modified lysis). The V3–V4 region was amplified (341F/805R), libraries prepared with Nextera XT, and sequenced on Illumina MiSeq (2×300 bp). Bioinformatics and statistics: Amplicon data were processed in QIIME2 (quality control, denoising with DADA2, paired merging, chimera removal, ASV inference). Taxonomy was assigned with a scikit-learn classifier trained on SILVA 138.1. Chloroplast/mitochondria ASVs were removed. In R, alpha diversity (Hill-Shannon, Hill-Simpson) was computed on 99.5% coverage-rarefied data (iNEXT). Group comparisons used Kruskal–Wallis with Bonferroni-corrected Dunn’s tests. For beta diversity, coverage-based rarefaction (phyloseq/metagMisc) was followed by Bray–Curtis, Jaccard, and Jensen–Shannon matrices; visualization used t-SNE (Rtsne; perplexity 5; up to 999 iterations). PERMANOVA (vegan: adonis; 999 permutations) assessed effects of main protein source and texture (dispersion checked by betadisper). LEfSe (phyloseqCompanion + Segata’s LEfSe; log10 LDA threshold 4.0) identified discriminatory taxa across groups.

- Product diversity: 32 PBMAs across pea/soy and minced/fibrous categories; most sold refrigerated and under MAP; some pre-heated or frozen once. - Cultivation outcomes: 470 colonies picked; 447 (95.1%) bacterial; 431 (91.7% overall) classified to genus level, spanning 38 genera from four phyla. Enterobacteriaceae isolates (n=16; 3.4%) could not be resolved below family by Sanger 16S. Fungal isolates (n=20; 4.3%) across 5 genera: Wickerhamomyces (7), Pichia (6), Yarrowia (3), Kurtzaniella (3), Geotrichum (1). Additional taxa were isolated in a second targeted cultivation (e.g., Brochothrix, Weissella, Psychrobacter), while several abundant amplicon taxa (Myroides, Pediococcus, Xanthomonas, Shewanella) were not cultured. - Prevalent cultured genera across samples: Bacillus (19/32 samples; 59.4%), Leuconostoc (18/32; 56.3%), Enterococcus (12/32; 37.5%), Latilactobacillus (10/32; 31.3%). Enterobacteriaceae were only found in pea-protein products from a single manufacturer. - 16S amplicon sequencing: 27 samples passed QC (883,427 total sequences; median 26,083/sample). ASVs spanned 25 phyla. Eighteen samples were dominated by Bacillota (10 with >90% relative abundance); nine were dominated by Pseudomonadota. Most common genera (detection frequency and relative abundance ranges): Leuconostoc (25 samples; 0.03–100%), Latilactobacillus (21; 0.02–86.38%), Pseudomonas (20; 0.36–35.25%), Serratia (19; 0.03–8.92%), Acinetobacter (17; 0.09–15.40%). Most abundant genus per sample: Leuconostoc (13 samples), Latilactobacillus (4), Shewanella (4). Several genera with >10% relative abundance in some samples were not isolated by culture. - Community profiles and diversity: t-SNE revealed three clusters characterized by Leuconostocaceae-, Latilactobacillus-, or Pseudomonadota-dominant profiles. Hill-Shannon and Hill-Simpson diversity differed significantly among these clusters (p<0.01). Across the four product groups (protein source × texture), Kruskal–Wallis showed a significant difference for Hill-Shannon (p=0.02) but not Hill-Simpson; only pea-fibrous vs. pea-minced differed in post-hoc Dunn’s test (p=0.02). - Drivers of composition: PERMANOVA indicated main protein source and texture significantly affected composition but explained only ~15.94–23.44% of variance. Including manufacturer (unbalanced design) increased explained variance to 53% (not used for final model). - LEfSe markers: Pea-fibrous was characterized by Leuconostoc and Enterococcus (and higher-level taxa); pea-minced by Pseudomonas; soy-fibrous by low-abundance Sphingomonadales and Burkholderiales; soy-minced by Enterobacterales and Flavobacteriales (Bacteroidota). - Genomic risks (WGS of isolates): E. coli genome harbored virulence factors (e.g., fimH, fyuA) and AMR genes (mdfA, ampH). Klebsiella (K. oxytoca, K. grimontii, K. pasteurii) carried class A beta-lactamase genes (e.g., blaoxy-2, blaoxy-6, blaxy-4). Other Enterobacteriaceae included Citrobacter braakii, Leclercia adecarboxylata, Lelliottia amnigena A. Bacillus paranthracis isolates carried the non-hemolytic enterotoxin (Nhe) complex and other virulence factors. Staphylococcus aureus harbored multiple virulence and AMR genes. LAB genomes (Leuconostoc, Latilactobacillus) encoded enzymes linked to spoilage-associated metabolites (diacetyl, acetoin, lactate, acetate, ethanol); one Leuconostoc mesenteroides carried a leucocin A/sakacin P family class II bacteriocin cluster. Leuconostoc mesenteroides showed broader carbohydrate utilization (26 glycoside hydrolases) and biosynthetic capabilities (multiple amino acids, menaquinone) than L. sakei (14 GHs).

The study provides foundational microbiological data for PBMAs, showing that lactic acid bacteria (Leuconostoc, Latilactobacillus) often dominate these products with generally low alpha diversity, while a subset is dominated by Pseudomonadota (e.g., Pseudomonas, Shewanella, Psychrobacter). The observed dominance patterns likely reflect processing histories: extrusion cooking effectively inactivates vegetative cells, whereas spores (e.g., Bacillus) can survive and later reactivate, and post-extrusion contamination from ingredients (spices, herbs) and the processing environment (e.g., biofilms) may shape communities. Despite the initial hypothesis, protein source and texture explained only a modest fraction of variance; manufacturer effects appear substantial, suggesting plant- and process-specific microbiomes. The presence of Enterobacteriaceae (restricted to products from one manufacturer) and WGS-detected virulence and AMR genes in E. coli, Klebsiella, Bacillus paranthracis, and Staphylococcus aureus highlight potential microbiological risks if recommended heating is not properly applied or kitchen hygiene lapses occur. Concurrently, LAB metabolic potential for producing spoilage-associated compounds implies they may significantly influence sensory deterioration and shelf life. These findings inform risk assessments, product formulation, processing hygiene (e.g., biofilm control), and consumer guidance (clearer cooking instructions), and underscore the need for targeted safety criteria for PBMAs not classified as ready-to-eat.

Combining culture-based methods, 16S rRNA amplicon sequencing, and isolate WGS, the study characterizes PBMA microbiomes from Austrian retail products. Most products were dominated by LAB (Leuconostoc mesenteroides, Leuconostoc citreum, Latilactobacillus sakei) with low alpha diversity, while others showed Pseudomonadota dominance. Genomic analyses revealed spoilage-associated metabolic capacities in LAB and potential pathogen/AMR gene carriage in some isolates (E. coli, Klebsiella spp., Staphylococcus aureus, Bacillus paranthracis). Protein source and texture contributed modestly to community variation, whereas manufacturer likely exerts a larger influence. These results suggest opportunities to optimize shelf life (targeting LAB-driven spoilage), improve processing hygiene, and refine consumer cooking guidance. Future work should include larger, balanced sampling across manufacturers, longitudinal shelf-life studies, direct assessment of raw materials and production environments (including biofilms), and development of PBMA-specific microbiological safety criteria.

- Unbalanced sampling across manufacturers limited formal modeling of producer effects; manufacturer inclusion increased PERMANOVA explained variance but was not used due to imbalance. - Lack of detailed production process data (e.g., exact extrusion parameters, ingredient sourcing, MAP gas composition) constrained attribution of contamination sources. - Only 27 of 32 products yielded amplicon datasets passing quality criteria, potentially limiting generalizability. - Products were heterogeneous in formulation, processing (pre-heated, frozen), and storage conditions, complicating direct comparisons. - Culture-based methods may miss viable but non-culturable taxa; conversely, amplicon data detected taxa not isolated in culture. - Cross-sectional design (single time point near purchase) limits inference on microbial succession over shelf life.