Strain dynamics of contaminating bacteria modulate the yield of ethanol biorefineries

Abstract

Bioethanol is a sustainable energy alternative and can contribute to global greenhouse-gas emission reductions by over 60%. Its industrial production faces various bottlenecks, including sub-optimal efficiency resulting from bacteria. Broad-spectrum removal of these contaminants results in negligible gains, suggesting that the process is shaped by ecological interactions within the microbial community. Here, we survey the microbiome across all process steps at two biorefineries, over three timepoints in a production season. Leveraging shotgun metagenomics and cultivation-based approaches, we identify beneficial bacteria and find improved outcome when yeast-to-bacteria ratios increase during fermentation. We provide a microbial gene catalogue which reveals bacteria-specific pathways associated with performance. We also show that Limosilactobacillus fermentum overgrowth lowers production, with one strain reducing yield by ~5% in laboratory fermentations, potentially due to its metabolite profile. Temperature is found to be a major driver for strain-level dynamics. Improved microbial management strategies could unlock environmental and economic gains in this US $ 60 billion industry enabling its wider adoption.

Fig. 1. Microbial dynamics during fermentation influence the performance of industrial bioethanol production.

A Outline of study sampling strategy, repeated 3 times in a single production season at 2 independent mills. Fresh media [1] are fed into fermentation vessels for 3 h. Samples were collected at 0–1.5 h [2], 1.5–3 h [3] and post-feeding [4]. Biomass is then centrifuged [5] and acid-washed [6] before re-entering the vessels. Vector images were obtained from Flaticon (www.flaticon.com), and figure as created using Adobe Illustrator. B Metagenomic profiling of microbial communities, grouped by relative batch performance (columns) per mill (row). Higher numbers of bacteria (blue) or eukaryotes (red) are not linked to better production performance. One sample from Mill A (starting broth, high-performing batch) did not contain sufficient DNA for sequencing. Error bars show variation across multiple samples, where applicable. For samples containing more than one datapoint, n = 3 biologically independent samples. Data are presented as mean values +/- SD. C Eukaryote-to-bacteria ratios across the production process. Each batch is connected by a line. In both mills, eukaryotic populations increased in the high-performing batches (orange, square) during fermentation steps (thick line) and decreased in lower-performing batches (light and dark brown, circle and triangle). Grey line denotes equal proportion of eukaryotes and bacteria (i.e. zero-fold difference). The y-axis indicates the fold change in the eukaryote to bacteria ratio. Higher fold values indicate a greater eukaryote abundance, and lower fold values a higher prokaryote abundance. D Genes associated with changes in fermentation performance. In both mills, the relative abundance of these 16 sets, involved in metabolism or membrane transport, differed between low- (brown) and high-performing (orange) batches (FDR < 0.1 for both mills). Increase in genes linked to bacteria-specific pathways, e.g. lipopolysaccharide biosynthesis and the phosphotransferase (PTS) system, is associated with better performance. Gene modules are ordered by decreasing relative abundance. KEGG Module IDs are provided in parentheses. For samples containing more than one datapoint, n = 3 biologically independent samples. The centre line denotes the median value (50th percentile), while the box contains the 25th–75th percentiles of dataset. The whiskers mark the 5th and 95th percentiles, and values beyond these upper and lower bounds are considered outliers, marked with dots. Source data are provided as a Source Data file.

Fig. 2. L. amylovorus and L. fermentum are the dominant bacteria in the industrial bioethanol microbiome and their interplay is linked to production performance.

A Taxonomic profiling of bacterial populations detected across process steps, grouped by relative batch performance (columns) per mill (row). L. amylovorus (light blue) and L. fermentum (dark blue) strains comprise the majority of bacteria across samples. Bacterial communities in the high-performing batch of Mill A and low-performing batch of Mill B undergo less change compared to others. For visual clarity, the 10 most abundant species are shown; beige colour represents other bacteria. Average values were used in process steps where multiple samples were collected. B Relative abundance of L. amylovorus and L. fermentum during fermentation steps. In both mills, high-performing batches contained more L. amylovorus strains. Across all batches and mills, L. fermentum increases and L. amylovorus decreases by the end of fermentation. C Bacterial species associated with changes in fermentation performance. High-performing batches showed increases in L. amylovorus and Weisella species (top row), while other lactic acid bacteria including L. fermentum, L.buchneri and L.plantarum decreased (bottom row) (FDR < 0.1 in both mills). Species are ordered by decreasing relative abundance. For samples containing more than one datapoint, n = 3 biologically independent samples. The centre line denotes the median value (50th percentile), while the box contains the 25th to 75th percentiles of dataset. The whiskers mark the 5th and 95th percentiles, and values beyond these upper and lower bounds are considered outliers, marked with dots. Source data are provided as a Source Data file.

Fig. 3. L. fermentum and other bacteria are associated with indicators of low fermentation performance.

A Correlation matrix of industrial fermentation parameters that showed strong associations throughout the production season (all sampling timepoints). Size and colour of points show correlation strength and direction, respectively. Low ethanol yield is associated with high acidity titres (dark red, Spearman’s ρ = −0.84), and increased bacterial cell count is linked to lower viability of yeast cells (orange, Spearman’s ρ = −0.72). Cross denotes correlations with FDR > 0.1. B Bacterial species associated with fermentation performance parameters. Species are ordered by decreasing relative abundance. C Increased L. fermentum is linked to higher acidity titres (Spearman’s ρ = 0.64, FDR < 0.05), and increased bacterial cell count (D; Spearman’s ρ = 0.63, FDR < 0.05). Source data are provided as a Source Data file.

Fig. 4. Specific L. fermentum strains reduce bioethanol production yield, possibly due to metabolic differences.

A Ethanol yield from pairwise fermentations of industrial yeast strain PE-2 with 3 L. fermentum industrial isolates (blue) and the 5 most abundant bacteria in the bioethanol production microbiome, compared to standalone fermentation (white). Ethanol yields were enhanced by L. amylovorus, P. claussenii and L. buchneri, and reduced only by L. fermentum strain (C) (*p < 0.05). For all experiments, n = 3 biologically independent samples. Data are presented as mean values +/- SD. Final ethanol yields were compared by multiple t-test (statistical significance analysis with alpha value of 0.05). B Cladogram of the 3 L. fermentum isolates and other published L. fermentum genomes, based on the alignment of 40 bacterial marker genes. L. fermentum strains broadly cluster into 4 groups; strain (C) belongs to a different clade than strains (A, B). Internal nodes are labelled with bootstrap values from 500 resamplings. Image generated using MEGAX, with the triangular root replaced for visual clarity. C Metabolite profiles measured from the supernatant of the 3 L. fermentum industrial isolates. Strains A and B showed similar production of acetate, lactate and ethanol. Strain C (dark blue, right) produced no ethanol and almost twice as much lactate as strains A and B. Source data are provided as a Source Data file.

Fig. 5. Temperature has different effects on L. amylovorus and L. fermentum growth rate during industrial fermentation.

A L. amylovorus-to-L. fermentum ratio at each stage of fermentation, summarised for both mills. Blue line denotes equal proportion of the two bacteria (i.e. zero-fold difference). L. fermentum populations can overtake L. amylovorus at the end of the bioprocess (right of blue line). For samples containing more than one datapoint, n = 3 biologically independent samples. The centre line denotes the median value (50th percentile), while the box contains the 25th–75th percentiles of dataset. The whiskers mark the 5th and 95th percentiles, and values beyond these upper and lower bounds are considered outliers, marked with dots. B Vat temperature at each stage of fermentation, summarised for both mills. Temperature decreases as the bioprocess progresses. For samples containing more than one datapoint, n = 3 biologically independent samples. The centre line denotes the median value (50th percentile), while the box contains the 25th – 75th percentiles of dataset. The whiskers mark the 5th and 95th percentiles, and values beyond these upper and lower bounds are considered outliers, marked with dots. C Growth rates of L. amylovorus and L. fermentum isolate strains A, B and C at 30°C and 37 °C. Higher fermentation temperatures hamper L. amylovorus growth (left, striped) and favoured L. fermentum strains (blue), with increases of up to 530%. Error bars denote variation across experimental replicates (*p < 0.05). For samples containing more than one datapoint, n = 3 biologically independent samples. Data are presented as mean values +/- SD. Final ethanol yields were compared by multiple t-test (statistical significance analysis with alpha value of 0.05). Source data are provided as a Source Data file.