In beverages made with pea protein, beany aromas, green notes, and bitterness can present challenges. Can fermentation improve them equally well with any microorganism?
A 2024 study progressively narrowed down candidates from 69 microbial strains. It reported that certain lactic acid bacteria and other strains reduced ratings of beany aroma and bitterness while increasing dairy-like aromas and umami. But it also found that the strains that changed the composition most dramatically did not necessarily produce good flavor.
- Examining taste compounds as well as aroma
- From 69 strains to 17, then six
- Some strains reduced bitterness and increased umami ratings
- Fewer green-smelling compounds, more dairy-like compounds
- Reduced bitterness was not simply a matter of bitter compounds disappearing
- The biggest compositional change is not necessarily the best
- What still needs checking before product development
- Summary
Examining taste compounds as well as aroma
The researchers aimed to select fermentation starters suitable for pea-protein beverages and examine the resulting changes in aroma and taste compounds. A starter is a microorganism added to initiate fermentation.
Rather than tracking only green-smelling volatiles, they also analyzed plant-derived compounds associated with bitterness and astringency, amino acids, and peptides formed by protein breakdown. The selection process combined human flavor perception with instrumentally measured compositional changes.
The beverage was prepared by dispersing a commercial pea-protein isolate in water and heat-treating it. Its protein concentration was 8.4%. Without optimizing the drink for each strain by adding nutrients or making similar adjustments, the researchers compared how active the microorganisms were in this beverage and the flavors they produced.
From 69 strains to 17, then six
The strains came from Nestlé’s culture collection. The initial 69 included a variety of microorganisms, including lactic acid bacteria, bifidobacteria, and Bacillus species.
- Stage 1: Fermentation lasted up to 72 hours, with measurements of microbial survival and activity, pH, and broad compositional changes. Five experienced assessors described the aromas and agreed on their characteristics through discussion.
- Stage 2: Seventeen strains were fermented for 48 hours. Taste and aroma intensities were compared with the unfermented beverage, and compositional analyses were repeated.
- Stage 3: Six strains were selected for detailed investigation after 48 hours of fermentation, including quantitative aroma analysis, amino acids and peptides, microbial counts, acidification, and changes in physical state.
The trained panel used for stages 2 and 3 comprised 14 people overall: seven men and seven women aged 23–35. They wore nose clips when assessing taste and removed them when assessing aroma, separating the two. Taste intensity was rated on a 0–5 scale and aroma on a 0–3 scale, using the unfermented product as a reference. This was not a consumer vote on which drink people liked best.
Some strains reduced bitterness and increased umami ratings
The six final candidates were two L. fermentum strains and one each of L. johnsonii, L. lactis, L. rhamnosus, and B. longum. Because strains of the same species can produce different results, the paper distinguishes them with NCC numbers.
L. johnsonii NCC533 and L. fermentum NCC660 attracted particular attention. Ratings shifted toward less bitterness and greater umami and saltiness. For example, in stage 3, the unfermented reference had bitterness and umami scores of 2.50 and 0.60, respectively, whereas the NCC533-fermented sample averaged 1.15 and 1.73.
For aroma, many candidates received lower beany ratings and higher dairy/cheesy ratings. NCC660 had a mean green-note score of about 0.21, compared with the unfermented reference of 0.80. However, not every candidate improved every aroma attribute: B. longum NCC2705 scored 0.85 for greenness, showing no reduction.
These are mean intensity ratings from trained assessors. Although the original paper describes some changes as significant, it does not provide the corresponding inferential statistical methods or p-values. Here, they are presented as the direction and magnitude of the observed changes.
Fewer green-smelling compounds, more dairy-like compounds
Chemical analysis showed lower levels of compounds including hexanal and benzaldehyde in samples fermented with the six finalists. NCC660 in particular greatly reduced several compounds, including hexanal, heptanal, and acetaldehyde.
For example, hexanal in the NCC660 sample fell from approximately 1,252 µg/kg before fermentation to approximately 64 µg/kg after 48 hours. This illustrates how changes in compounds associated with green or beany aromas corresponded with the direction of sensory ratings.
Meanwhile, NCC533 increased diacetyl and acetoin, which are associated with dairy-like aromas. NCC2378 also increased acetoin, but diacetyl remained almost unchanged. The aromas added by fermentation also depended on the strain.
The researchers also used odor activity values, or OAVs, calculated by dividing a compound’s concentration by its odor threshold, to identify potential aroma contributors. OAVs help select compounds for attention; they do not indicate how many times stronger the beverage’s overall aroma became.
Reduced bitterness was not simply a matter of bitter compounds disappearing
The taste analyses produced an intriguing result. In some samples fermented with lactic acid bacteria or other microorganisms that shifted flavor in a favorable direction, known plant-derived compounds associated with bitterness and astringency did not clearly decrease. Yet bitterness ratings still fell.
NCC533 and NCC660 showed increased analytical signals for free amino acids, nucleotide-related compounds, and small peptides. Potential contributors included glutamate, other umami-related compounds, and peptides associated with mouthfulness or kokumi.
The authors proposed that compounds produced by moderate protein breakdown may have enhanced umami and other sensations, making bitterness less noticeable. They also suggested that acidification-induced protein aggregation and changes in mouthfeel could contribute to reduced astringency.
At this stage, many taste-related compounds were compared using peak areas, rather than all being quantified as concentrations. The authors left confirmation of the mechanisms for future work, including tests that reconstruct the mixture to reproduce its taste.
The biggest compositional change is not necessarily the best
Some early candidates, including Bacillus species, substantially changed the metabolite profile, reduced fatty acids associated with bitterness, and produced many peptides. However, sensory assessments found increased bitterness or added rancid, fecal, and leathery notes, rather than desirable flavor.
The authors discuss how excessive protein breakdown may also produce unpleasant tastes. Even when analysis shows fewer potential bitter compounds, other bitter tastes or aromas may emerge. A central conclusion of the study is that chemical analysis needs to be complemented by actually tasting the samples.
What still needs checking before product development
Some samples coagulated during fermentation, while others remained liquid. A favorable flavor change does not necessarily make a sample easy to use as a beverage; acidity and texture also require adjustment.
Each selection stage involved one fermentation run, with the final candidates examined across all three stages. Technical replicates, in which the same sample is measured repeatedly, differ from independent production replicates. Reproducibility under different manufacturing conditions, consumer preferences, and quality during storage remain to be investigated.
Summary
Progressive screening of 69 strains identified certain lactic acid bacteria and other microorganisms that shifted ratings toward less beany aroma and bitterness and more dairy-like aroma and umami in pea-protein beverages. NCC533 and NCC660 were promising candidates for both taste and aroma.
The improvements involved not only reductions in green-smelling compounds but also increases in new aroma and taste-related compounds and changes in texture. Conversely, some strains caused large compositional changes that led to unpleasant tastes. The study shows why selecting fermentation strains requires moving back and forth between chemical analysis and human perception.
The research was supported by Société des Produits Nestlé and the CREATE program of Singapore’s National Research Foundation. Five coauthors were affiliated with Nestlé Research at the time. The authors declared no competing financial interests.
References
Spaccasassi A, et al. Screening of a Microbial Culture Collection: Empowering Selection of Starters for Enhanced Sensory Attributes of Pea-Protein-Based Beverages. Journal of Agricultural and Food Chemistry. 2024;72:15890–15905. Original article (University-hosted full text, Supporting information).
