Microbiome profiling reveals manufacturer and batch differences in fermented grain beverages

A metabarcoding study of fermented plant- and grain-based beverages from seven manufacturers found that bacterial communities differed strongly by producer and production batch, while fungal communities showed substantial variation but less distinct clustering. The findings suggest that high-throughput sequencing can complement conventional microbiological testing by revealing community-wide patterns and unexpected taxa in fermented drinks.

The study examined commercially available, primarily acidic concentrates purchased in Klaipėda, Lithuania, from seven local manufacturers, designated M1 through M7. Two production time points representing separate batches were sampled for each product, with three independently processed biological replicates. The beverages included drinks made from grains such as buckwheat, oats and rice, along with products formulated with lactic acid bacteria, Bifidobacterium cultures, honey or juices. Detailed pH, ethanol and residual-sugar measurements were unavailable because manufacturers did not provide them and the researchers did not measure them.

Researchers extracted DNA from homogenized beverages and used amplicon sequencing on an Illumina NextSeq 2000 platform. The V3–V4 region of the bacterial 16S ribosomal RNA gene was targeted to characterize bacteria, while the fungal internal transcribed spacer 2 region was used for fungi. DADA2 processing, taxonomic assignment against SILVA 138.2 and UNITE 10.0, and contamination filtering produced 6,352,987 bacterial reads representing 1,332 amplicon sequence variants and 1,295,070 fungal reads representing 133 variants across 42 samples. Following rarefaction and removal of rare variants, the retained datasets contained 375 bacterial and 65 fungal variants.

Lactic acid bacteria were prominent across many beverages, including Lacticaseibacillus, Limosilactobacillus, Lactobacillus, Streptococcus and Sporolactobacillus. Bifidobacterium was detected intermittently, while Acetobacter and other genera also occurred. Community structure differed markedly among manufacturers: Lactobacillus represented approximately 41–85.5% of detected bacterial genera in selected M3, M4, M5 and M6 samples, whereas Clostridium accounted for 72.1% of M2’s second-batch community. M7 had more than 50% relative abundance of Weizmannia in both batches. Fungal profiles included the fermentative yeast Saccharomyces, secondary yeasts such as Pichia, and numerous environmental or crop-associated genera.

Production-batch changes were also observed within manufacturers. In M3, Lactobacillus declined from 58.6% of bacterial reads in the first batch to 17.5% in the second, while Acetobacter rose from 0.78% to 56.1%. In M5, Lactobacillus fell from 85.5% to 13%, accompanied by an increase in Limosilactobacillus. Fungal dominance also shifted among batches. For example, Fusarium represented 96.1% of the fungal community in M4’s first batch but 31.0% in its second, while environmental yeasts changed in relative prominence in several other products. By contrast, M4 retained broadly consistent bacterial dominance by lactic acid bacteria across its two batches.

Statistical comparisons supported these observed differences. The number of observed bacterial variants differed significantly among manufacturers in a Kruskal–Wallis test (p = 7.14 × 10^-6), whereas fungal richness did not show a significant difference (p = 0.396). PERMANOVA identified significant effects of manufacturer and batch, as well as their interaction, for both bacterial and fungal communities; all three effects had p-values of 0.001 in the bacterial analysis, while batch and interaction had p-values of 0.023 and 0.024 in the fungal analysis. Manufacturer, batch and their interaction together accounted for 98.02% of bacterial variation and 55.53% of fungal variation. However, significant differences in within-group dispersion indicate that heterogeneity may also have influenced the PERMANOVA results.

Functional annotation assigned bacterial profiles through FAPROTAX and fungal profiles through FUNGuild, followed by literature checks and manual classification. Most resolved bacterial variants were categorized as beneficial under the study’s criteria, and most manufacturers had bacteria classified as potentially harmful below 1% relative abundance. Exceptions included samples from both M6 batches and M3’s first batch. In M6’s second batch, 84.3% of classified bacterial taxa were categorized as potentially harmful or undesirable, including taxa associated with spoilage or off-flavor formation such as Clostridium tyrobutyricum and Anoxybacillus flavithermus. Detection of Clostridium perfringens was treated as a signal requiring follow-up, not confirmation of pathogenicity.

Fungal functional assignments were less clear because many taxa had mixed, unclear or plant-associated ecological roles. M4, M5 and M6 had higher proportions of genera classified as potentially harmful, including Fusarium and Issatchenkia. The high Fusarium signal in M4’s first batch is relevant because some Fusarium species can produce mycotoxins under suitable conditions and cereal substrates are associated with these compounds. Nevertheless, the assay detected DNA rather than viability or active toxin production, so the signal could reflect fungal material carried through from malted barley, wheat, rye or another raw ingredient.

The study has important methodological limits. Amplicon sequencing does not establish that detected organisms are viable, metabolically active or present as pathogenic strains, and conserved markers often limit identification below genus level. PCR bias, gene copy-number variation, compositional data and gaps in reference databases can affect abundance estimates and classification. Several low-fungal-read samples were removed, including all fungal samples from M7’s second batch, further limiting comparisons. The authors therefore proposed a tiered monitoring system: retain standard culture assays for baseline compliance, add periodic 16S and ITS profiling to identify community drift and out-of-panel taxa, and use targeted strain-specific quantitative PCR, viability assays and rapid LC–MS/MS mycotoxin screening for verification. Within that framework, the study positions metabarcoding as a complementary first-pass method rather than a substitute for culture-based safety testing or targeted chemical analysis.

Subject of Research: Metabarcoding assessment of bacterial and fungal communities in fermented grain beverages across seven manufacturers and two production batches

Article Title: High-throughput microbiome profiling of fermented grain drinks: insights across manufacturers and batches

Article References: Scriver, M., Zaiko, A., Gudaitytė, R., & Samuilovienė, A. (2026). High-throughput microbiome profiling of fermented grain drinks: insights across manufacturers and batches. Metabarcoding and Metagenomics, 10, Article e191385. https://doi.org/10.3897/mbmg.10.191385

Image Credits: AI Generated

DOI: 10.3897/mbmg.10.191385

Keywords: fermented beverages, metabarcoding, 16S rRNA, fungal ITS2, bacterial communities, batch consistency, food microbiology, microbial quality, probiotic bacteria, High-throughput, microbiome, profiling