Oats, peas, faba beans, and sunflower seeds: when the raw material changes, so does the aroma of fermented plant milk. A 2023 study combined four plant milks with 15 strains of lactic acid bacteria, looking for pairings that could reduce green-smelling compounds or increase aromas reminiscent of butter, fruit, and cheese.
Interestingly, although the materials differed substantially, the researchers also identified a candidate bacterium that increased buttery notes across all four. They also predicted aroma changes from instrumental composition data and compared those predictions with human assessments.
Finding bacteria suited to one material or several
The study aimed to improve plant-milk flavor through fermentation and identify bacteria that would work across different raw materials. Plant ingredients differ in their sugars, proteins, fats, and existing aroma compounds. A bacterium that works well in one material will not necessarily produce the same aroma in another.
The researchers mixed oats, defatted sunflower-seed material, peas, or faba beans with water, then homogenized and ultra-high-temperature-treated the samples. Rather than comparing commercial drinks as sold, they used experimental plant milks designed to make it easier to study combinations of ingredients and bacteria.
Each of 15 food-grade lactic acid bacterial strains was added separately, and fermentation proceeded without oxygen at 30°C for 24 hours. No sugar, animal milk, or other nutrients were added during fermentation. Each combination was tested in three biological replicates. Controls without added bacteria were also included to examine growth and changes in volatile compounds.
The same bacterium grows and produces aromas differently in different materials
Of the 15 strains, 14 grew in more than one plant milk. The remaining strain, Lactobacillus helveticus NCC1276, did not grow in any of the four, so subsequent aroma comparisons primarily involved samples fermented with the other 14 strains and unfermented samples.
Growth performance varied among strains. For example, Leuconostoc mesenteroides NCC2832 grew well in all four materials, whereas Lactiplantibacillus plantarum NCC2988 grew well in sunflower, pea, and faba milk but poorly in oat milk.
Gas chromatography–mass spectrometry identified 71 volatile compounds before fermentation and 27 additional compounds afterward. Common changes included reductions in aldehydes such as hexanal and increases in alcohols, acids, ketones, and esters. Hexanal is one compound associated with green, beany aromas.
Even the same bacterium produced different compounds depending on the material. L. johnsonii showed particularly notable alcohol production in sunflower and faba milk, but acid production in pea milk. The composition of compounds with woody, minty, floral, or citrus-like notes also changed in connection with terpenes present in sunflower and faba milk. The authors interpret the direction of aroma development as a product of the starting materials available and the bacterium’s metabolism.
Can the abundance of compounds alone predict aroma?
The researchers then tested a way to bring analytical data closer to the aroma people perceive. A compound’s contribution depends not only on how much is detected, but also on whether people can smell it at a very low concentration or only at a much higher one.
They calculated each compound’s proportion of the total analytical peak area and weighted it using odor-detection thresholds reported in the literature. They primarily used thresholds in air, converting values for compounds with only water-based thresholds available. They then grouped the compounds into eight aroma categories, including fruity/sweet, buttery/fatty, cheesy/sour, and green/grassy.
Of the 98 compounds detected, 69 had both aroma descriptions and threshold information available for these calculations. This was a relative predictive index, not the conventional odor activity value calculated from absolute concentrations. Thus, the change from 95% to 7% in the predicted green aroma of pea milk fermented with L. johnsonii describes a shift in the predicted aroma profile, not a measured reduction of that magnitude in the greenness people perceived.
Candidates for buttery and cheesy notes
The predictions identified Streptococcus thermophilus NCC1326 as a candidate for increasing buttery notes in all four plant milks. Production of compounds such as diacetyl and acetoin, associated with butter or cream, supported this direction.
Meanwhile, L. mesenteroides was highlighted as a candidate for cheesy notes, and L. johnsonii for terpene-related woody and minty notes. Increases in cheesy/sour notes were particularly apparent in pea and faba milk, and the authors discussed their potential for plant-based cheese alternatives.
For reducing green or beany notes, L. plantarum, L. mesenteroides, and L. johnsonii were promising. Rather than simply concluding that fermentation improves aroma, the study compared pairings to help select ingredients and bacteria for a desired aroma. The authors’ classification of desirable notes reflects product-development assumptions; the study did not measure preferences across the consumer population.
A small sensory panel also detected the predicted changes
To compare the predictions with human perception, the researchers used oat and pea milks fermented with six strains with differing characteristics. Before assessment, the samples were heated at 71°C for 2 minutes and cooled to room temperature. An untrained panel of 8–10 people rated increases or decreases in particular aroma notes relative to unfermented samples on a scale from −3 to +3.
Participants perceived increases in fruity/sweet and buttery/fatty notes in oat milk, and reductions in green/grassy notes and changes in cheesy/sour notes in pea milk. These ratings correlated with predictions from the composition data, suggesting that relatively accessible analytical peak information could help guide the selection of fermentation conditions.
However, sensory verification covered only two selected materials and six strains, not all four materials. The predictions could not fully account for interactions among compounds in mixtures, the effects of the material on aroma release, or compounds without available thresholds. The authors identify further analysis using more reference standards and methods that separate compounds for direct smelling as future work.
Summary
In fermented plant milks, the same bacterium showed different growth and changes in aroma compounds depending on the raw material. The study identified candidates such as S. thermophilus NCC1326 for adding buttery notes to all four materials, alongside candidates for cheesy or woody/minty notes and reductions in greenness.
Predictions that weighted relative compound abundance by odor thresholds also corresponded with small sensory assessments of oat and pea milks. This is a step toward designing aromas through ingredient–bacterium pairings. Future work needs to connect these findings with whole-product assessments, including liking, texture, color, and nutritional changes.
The study was funded by Nestec Ltd., which merged with Société des Produits Nestlé SA in 2019. Four authors were affiliated with Nestlé research or technology units at the time of the study. Open-access funding was supported by Projekt DEAL, and the authors declared no competing interests.
References
Tangyu M, Fritz M, Tan JP, et al. Flavour by design: food-grade lactic acid bacteria improve the volatile aroma spectrum of oat milk, sunflower seed milk, pea milk, and faba milk towards improved flavour and sensory perception. Microbial Cell Factories. 2023;22:133. doi:10.1186/s12934-023-02147-6
