Genomics and synthetic community experiments uncover the key metabolic roles of acetic acid bacteria in sourdough starter microbiomes

Abstract

While research on the sourdough microbiome has primarily focused on lactic acid bacteria (LAB) and yeast, recent studies have found that acetic acid bacteria (AAB) are also common members. However, the ecology, genomic diversity, and functional contributions of AAB in sourdough remain unknown. To address this gap, we sequenced 29 AAB genomes, including three that represent putatively novel species, from a collection of over 500 sourdough starters surveyed globally from community scientists. We found variations in metabolic traits related to carbohydrate utilization, nitrogen metabolism, and alcohol production, as well as in genes related to mobile elements and defense mechanisms. Sourdough AAB genomes did not cluster when compared to AAB isolated from other environments, although a subset of gene functions was enriched in sourdough isolates. The lack of a sourdough-specific genomic cluster may reflect the nomadic lifestyle of AAB. To assess the consequences of AAB on the emergent function of sourdough starter microbiomes, we constructed synthetic starter microbiomes, varying only the AAB strain included. All AAB strains increased the acidification of synthetic sourdough starters relative to yeast and LAB by 18.5% on average. Different strains of AAB had distinct effects on the profile of synthetic starter volatiles. Taken together, our results begin to define the ways in which AAB shape emergent properties of sourdough and suggest that differences in gene content resulting from intraspecies diversification can have community-wide consequences on emergent function.

Importance: This study is a comprehensive genomic and ecological survey of acetic acid bacteria (AAB) isolated from sourdough starters. By combining comparative genomics with manipulative experiments using synthetic microbiomes, we demonstrate that even strains with >97% average nucleotide identity can shift important microbiome functions, underscoring the importance of species and strain diversity in microbial systems. We also demonstrate the utility of sourdough starters as a model system to understand the consequences of genomic diversity at the strain and species level on multispecies communities. These results are also relevant to industrial and home-bakers as we uncover the importance of AAB in shaping properties of sourdough starters that have direct impacts on sensory notes and the quality of sourdough bread.

Keywords: acetic acid bacteria; comparative genomics; microbiome; sourdough starter; strain diversity; synthetic communities.

Fig 1

Overview of study design and geographic diversity of AAB. (A) Conceptual overview of study data and experiments. Leveraging 16S amplicon data from 500 sourdough starters, we isolated 21 AAB on selective media and obtained corresponding genomes. We obtained MAGs (N = 8) from metagenomes (sample N = 40) and also included 32 publicly available AAB genomes for a final genome set of 61 for our comparative genomics assessment. We also selected a subset of AAB isolates (N = 10) for synthetic sourdough experiments to measure the functional impact of AAB on starter microbiomes. (B) AAB were isolated from a set of 500 sourdough samples that were previously described using amplicon sequencing, collected from a global network of community scientists (14). Light blue-gray circles indicate the detection of AAB in 16S rRNA gene amplicon data, and bright blue circles denote the recovery of one or more AAB genomes from the sample. Orange circles denote the set of AAB genomes included from NCBI (National Center for Biotechnology Information), recovered from diverse environmental sources and geographic locations. Popout shows example AAB colonies on plate (s.b. 1 cm) and under the microscope at 100× (s.b. 10 µm).

Fig 2

Genomic diversity and abundance of AAB in sourdough starter microbiomes. (A) Summary of the mean % relative abundance of AAB by species taxonomic assignments across 500 sourdough starters. Some ASVs could only be assigned to the nearest cluster of species (Fig. S1) due to the low resolution of ASVs. Bar colors indicate the source of genomes obtained (isolate, MAG, both, or no genome obtained from sourdough). (B) Summary of the mean % relative abundance of AAB across starter samples. AAB were detected in 29.4% of the 500 starter samples at ≥1%. Breaks in bars represent the relative abundance of each AAB ASV detected in a sample. Dark blue bars denote samples where isolated taxonomy matched the expected identity from 16S amplicon sequencing of the community, and orange denotes samples where a distinct AAB taxon was recovered. (C) Representative genome tree of Acetobacter species including three putatively novel species (in bold) and other genera including Gluconobacter and Komagataeibacter, which were also detected in sourdough starter microbiomes. The number of genomes obtained for this study is listed in parentheses (via culturing, metagenome assembly, and download from NCBI). Species with isolates from sourdough or MAGs recovered from sourdough are highlighted in yellow. Triangles represent species that were included in our synthetic sourdough experiments, and filled-in triangles are species that also were included in our assessment of intraspecies strain diversity.

Fig 3

Genomic and metabolic diversity across AAB. (A) Pangenome and ANI cluster of 61 AAB genomes. Each ring represents a genome, colored by ANI cluster (Table S5). The core genome across AAB is highlighted, along with a subset of genes functionally enriched in sourdough genomes (for a full set, see Table S8). (B) Genome features of the 61 genomes included in the pangenome are reported by isolate source (including fermented beverage, fruit, fruit fly, sourdough, and other), ANI cluster, genome statistics (including % completion and % contamination, Table S4), and metabolic traits that were variable across AAB genomes, including carbohydrate-active enzymes (CAZy) families, acetate pathways, nitrogen metabolism, SCFA i.e. short chain fatty acids, and alcohol production). Genomes are labeled by species. Where multiple genomes per species are included, Refseq ID/original name is given.

Fig 4

Diversity within AAB species. (A) Strain pangenomes with ANI heat map (Table S5) and corresponding genome trees (SpeciesTree) of nine strains each of A. malorum/A, A. orientalis, and G. oxydans/potus. Pangenomes and trees are annotated by the isolation source. (B) Percentage of genes in the accessory genome within a subset of functional categories (mobilome, defense mechanisms, carbohydrate metabolism, energy metabolism, and translation) across the three species highlighting strain diversity. (C) Boxplot highlights GH13 alpha-amylase (starch cleavage) which was found to be enriched in sourdough starter vs other environments in multiple strains of the same species.

Fig 5

Sourdough starter acidification is determined by differences in genus, species, and strain-level variation in AAB. (A) Experimental design of synthetic communities. Ten isolates of AAB were selected as treatment groups and added to a background starter community of yeast (S. cerevisiae) and LAB (L. brevis). We also included a yeast + LAB only control and cereal-based fermentation medium (CBFM) blank where no microbes were added. All isolates were added in 5 µL at a total density of 20,000 colony-forming units (CFUs). (B) The total abundance of each member of SynCom129 measured via CFUs and plotted with log10 (top); acidification measured via pH from the liquid CBFM and reported at the end of 4 days of incubation and transfers (N = 5 replicate synthetic communities; dots represent individual observations and box plots summarize the distribution). (C) Plots of the relationships between AAB CFU count and emergent microbiome pH (top) and ANI vs pH distance (bottom) from SynCom129.

Fig 6

Sourdough VOCs are associated with microbial community type. (A) Dendrogram of VOC hierarchical clustering resulting in four sample clusters. (B) Superimposed NMDS plots of liquid sourdough communities and VOCs by community. The four main clusters have been colored. VOCs have been labeled when significantly different between clusters.