The process the team studied differs fundamentally from conventional brine fermentation. Traditional Chinese pickled vegetables such as paocai and suancai are typically submerged in prepared brine, allowing salts, sugars, and microbial metabolites to circulate freely between the plant tissue and the surrounding liquid. Vacuum dry-salted cowpeas, by contrast, are mixed directly with salt at 8 percent of the fresh weight, sealed in food-grade plastic bags without any added liquid, and left to ferment in the dark at room temperature. Only then does liquid gradually exude from the salted tissues, creating a fermentation environment with restricted oxygen exchange and a slowly developing aqueous phase. Because these conditions could shape microbial succession and flavor chemistry in unique ways, the researchers sampled the fermenting cowpeas at five time points: days 0, 3, 7, 14, and 21.

The physicochemical data tell a story of rapid acidification. The pH of fresh cowpeas stood at 5.94, but it plunged to 3.68 by day 7 and bottomed out at 3.46 by day 14 before rising slightly to 3.90 at day 21. Titratable acidity climbed in parallel, from 1.92 grams per kilogram initially to a peak of 7.62 grams per kilogram at day 14. Reducing sugars, the fuel for fermentation, collapsed from 1.92 grams per 100 grams to a mere 0.10 grams per 100 grams over the three weeks, indicating that fermentable carbohydrates were almost entirely consumed. Meanwhile, amino acid nitrogen nearly tripled, rising from 0.87 to 2.18 grams per 100 grams by day 14, a sign that proteins were being broken down into amino acids and peptides that contribute savory taste and serve as precursors for aroma compounds.

Food safety emerged as a reassuring subplot in the data. Nitrite, a compound of concern in fermented vegetables because of its potential health risks, rose transiently during the early stage, peaking at 0.49 milligrams per kilogram on day 3, then declined steadily to 0.22 milligrams per kilogram by day 21. That final value sits far below the Chinese national safety limit of 20 milligrams per kilogram, suggesting that the acidification and microbial dynamics of vacuum dry-salted fermentation naturally suppress nitrite accumulation. The researchers attribute this decline to a combination of factors, including the inhibitory effect of low pH on nitrite-producing microorganisms and possible microbial reduction of nitrite itself.

Beneath these chemical shifts lay an equally dramatic ecological upheaval. Amplicon sequencing of the 16S rRNA gene and the fungal internal transcribed spacer region showed that the bacterial community underwent near-total replacement. At day 0, the phylum Pseudomonadota dominated, accounting for 91.67 percent of bacterial reads, a signature of the raw plant surface and its processing environment. By day 7, Bacillota had surged from undetectable levels to 93.75 percent and remained dominant through the end of fermentation. At the genus level, the fresh cowpea microbiota of Acinetobacter, Sphingomonas, and Methylobacterium gave way to a transitional community at day 3 in which Weissella, Lactiplantibacillus, Klebsiella, Kosakonia, Enterobacter, and Pediococcus coexisted, before Lactiplantibacillus took command from day 7 onward, ultimately reaching 66.67 percent relative abundance by day 21.

The fungal community followed its own arc of succession. Plant-associated genera such as Plectosphaerella and Verticillium, abundant at the start, were suppressed under the acidic, saline, and oxygen-poor conditions inside the sealed bags. In their place, yeast-like fungi flourished in stage-specific waves: Debaryomyces peaked at 20.7 percent on day 7, while Pichia and Millerozyma rose later, with Millerozyma reaching 8.8 percent by day 21. These genera are known flavor engineers. Debaryomyces is notably tolerant of salt and acid and can generate alcohols and esters from sugars and amino acids, while Pichia is recognized for its strong capacity to produce aroma-active volatiles. Statistical analysis confirmed that fermentation stage significantly structured both bacterial and fungal communities, with PERMANOVA yielding p-values of 0.001 for both.

The volatile chemistry captured by headspace solid-phase microextraction gas chromatography-mass spectrometry revealed 95 distinct compounds and a clear narrative of replacement. Fresh cowpeas were dominated by C6 aldehydes and alcohols born of plant lipid oxidation: hexanal at 655.09 micrograms per kilogram, 2-hexenal at 734.88 micrograms per kilogram, and (Z)-3-hexen-1-ol, all carriers of green, grassy, fresh-cut-vegetable notes. These compounds collapsed within days of fermentation onset. In their place, fermentation-associated volatiles accumulated. Phenylethyl alcohol, a rose-scented compound derived from aromatic amino acid metabolism, climbed steadily to 272.42 micrograms per kilogram by day 21, becoming the third most abundant volatile. Esters increased roughly eightfold, from 38.98 to 312.91 micrograms per kilogram, shifting from a single green-note ester to a diverse array including ethyl hexadecanoate and 2-phenylethyl acetate, which lend fatty, sweet, floral, and fruity nuances.

The single most striking transformation involved 4-ethylphenol, a heterocyclic-adjacent phenolic compound that surged from 22.58 micrograms per kilogram at day 3 to 619.13 micrograms per kilogram at day 7 and ultimately reached 641.29 micrograms per kilogram at day 21, accounting for roughly 29.3 percent of the total volatile content. This smoky, spicy, phenolic-smelling compound effectively defines the mature product’s aroma profile, though the authors caution that its sensory impact is concentration- and matrix-dependent, and excessive levels could tip into undesirable territory. By intersecting statistical screening with odor activity values, the team distilled 11 candidate volatile markers of fermentation stage, including hexanal and 2-hexenal for the fresh stage, lipid-derived aldehydes such as (E)-2-heptenal and 1-octen-3-one for the early phase, and phenylethyl alcohol and 4-ethylphenol for the late stage. Such markers could eventually underpin process-monitoring and quality-control tools for manufacturers.

Untargeted metabolomics added a third dimension, detecting 4,257 annotated features dominated by lipids, organoheterocyclic compounds, organic acids, and phenylpropanoids. The differential metabolites were overwhelmingly stage-specific rather than shared across comparisons, indicating that flavor development proceeds through successive metabolic waves rather than a fixed set of continuously changing compounds. Early fermentation was marked by lipid remodeling and the accumulation of oxidized fatty acid derivatives that mirrored the lipid-derived volatiles. Middle and late stages saw the enrichment of phenyllactic acid and hydroxyphenyllactic acid, signature products of lactic acid bacterial metabolism of phenylalanine and tyrosine, alongside accumulating peptides such as gamma-glutamylleucine that signal active proteolysis. Organoheterocyclic compounds tied to tryptophan and amino acid degradation rose as plant-derived glycosides and carbohydrate-related metabolites declined, a molecular reflection of the raw material being progressively consumed and transformed.

Correlation analysis wove these threads together into a network of 47 nodes and 82 edges linking microbes, volatiles, and non-volatile metabolites. Early-stage bacteria such as Acinetobacter and Methylobacterium co-varied perfectly with the green-note compound (Z)-3-hexen-1-ol, while Klebsiella tracked the lipid-derived aldehydes of the early phase. In the late stage, Lactiplantibacillus co-varied with phenylethyl alcohol and 4-ethylphenol, and the yeasts Pichia and Oligophagozyma showed parallel temporal associations with these same aroma compounds. The authors are careful to stress that these are co-occurrence patterns, not proof of causation, and that semi-quantified volatile data and a single fermentation batch limit generalizability. Still, the study delivers something the fermented vegetable industry has lacked for this product: a temporal framework connecting who the microbes are, what chemistry they drive, and when each flavor milestone occurs. Future work involving targeted quantification, sensory panels, and inoculation experiments will determine whether these candidate markers can be translated into practical tools for ensuring that every batch of vacuum dry-salted cowpeas delivers the same beloved sour, floral, and fermented character.

Subject of Research: Microbial succession and flavor compound evolution during vacuum dry-salted cowpea fermentation

Article Title: Flavor development in vacuum dry-salted cowpeas: changes in volatile compounds and non-volatile metabolites during fermentation

Article References: Xie, Q., Zhou, Z.-G., Sun, Y., Li, T.-Y., Zhang, K.-X., Ou, H.-Y., Man, R.-J., Li, C.-B., & Xin, M. (2026). Flavor development in vacuum dry-salted cowpeas: changes in volatile compounds and non-volatile metabolites during fermentation. Food Chemistry: X, Article 104605. https://doi.org/10.1016/j.fochx.2026.104605

Image Credits: AI Generated

DOI: Not provided

Keywords: cowpea fermentation, lactic acid bacteria, Lactiplantibacillus, volatile compounds, metabolomics, 4-ethylphenol, phenylethyl alcohol, food flavor chemistry, fermented vegetables, microbial succession, nitrite safety, vacuum dry-salting

The investigation was led by doctoral candidate Franziska Krause and principal investigator Martin Steinhaus, head of the Aroma Chemistry research group at the Leibniz Institute. Rather than relying on laboratory simulations alone, the team took their work into the field. Krause monitored fermentation processes on site at cocoa plantations in Nicaragua and Costa Rica, collecting samples at various stages of fermentation under real-world conditions. This field-based approach matters because cocoa fermentation is a living, microbial process that unfolds differently depending on climate, cocoa variety, and local technique. By anchoring their analytical work in genuine plantation practice, the researchers ensured that their conclusions would be relevant to the producers who actually grow and ferment the world’s cocoa.

Cocoa processing is traditionally divided into two broad stages. In tropical regions, farmers cultivate the cocoa tree, Theobroma cacao L., harvest the fruit, and extract, ferment, and dry the seeds that are commonly known as cocoa beans. The beans then travel, mostly to temperate regions, where they are transformed into chocolate and other cocoa products. The most important growing regions today lie in West Africa, Latin America, and Southeast Asia. Fermentation is the step where the raw seeds develop the precursor compounds that later give chocolate its characteristic aroma, but it is also the step where things can go wrong. Off-flavors that emerge during fermentation can compromise the quality of the final chocolate, and until recently, the odorants responsible for many of these defects were chemically unidentified.

Only in the last few years have researchers managed to pin down the specific odor-active compounds behind many cocoa off-flavors. Even so, how and under what conditions these compounds form remained poorly understood. To close that gap, Krause and Steinhaus compared samples of optimally fermented cocoa beans with samples that had been intentionally overfermented. Using modern odorant analysis techniques in the laboratory, they found that the overfermented samples contained elevated concentrations of short-chain carboxylic acids. These are not obscure molecules: they are the same class of compounds responsible for smells reminiscent of cheese, sweat, or vomit. Their presence at higher levels explains why overfermented cocoa can acquire rancid, dairy-like, and generally unpleasant aromatic notes that no chocolatier wants in their product.

The more surprising discovery concerned phenolic compounds. The overfermented samples contained significantly higher concentrations of 2-methoxyphenol and 4-ethylphenol, both of which smell smoky. The two molecules are not identical in character, however. 2-Methoxyphenol carries a rather sweet, smoky odor, while 4-ethylphenol is described as leathery and phenolic, with additional horse stable-like notes. Together, these compounds can give overfermented cocoa an aromatic profile that a trained sensory panel could easily mistake for smoke contamination. In other words, a batch of beans that never came anywhere near a smoky fire can still smell as though it did.

The comparison with genuine smoke contamination is what gives the new study its force. In an earlier investigation, the same research team had already identified and quantified the off-flavor compounds typical of wood smoke exposure in cocoa beans. That earlier work showed that when beans are dried over wood fires, smoke-derived phenols such as 2-methoxyphenol, 4-methylphenol, and 3-ethylphenol pass into the beans and penetrate deep into the cotyledons, the edible kernel of the seed. This is a crucial practical detail, because removing the shells afterwards does not significantly reduce the smoky off-flavor. The team also identified 2,6-dimethoxyphenol as a marker for smoke exposure, although that compound contributes little to the aroma itself. The new results now show that overfermented beans contain key smoke-like phenols at concentrations similar to those found in wood smoke-contaminated cocoa.

“Our results thus demonstrate for the first time, objectively and at the molecular level, that smoky off-flavors can arise not only from smoke contamination but also from overfermentation,” says food chemist Martin Steinhaus. The statement marks a conceptual shift for the industry. If smoky aromas can have two independent origins, then simply blaming open-fire drying is no longer sufficient, and quality-control strategies that focus only on drying practices may miss the real culprit. Krause adds an important nuance: “Another important finding of our study is that the quantitative spectra of the odor-active phenols differ between overfermented and wood-smoke-contaminated cocoa. These differences could help to identify the causes of smoky off-flavors more reliably in the future.” In essence, the relative pattern of phenolic compounds acts like a fingerprint, allowing analysts to distinguish whether smoke or fermentation is to blame.

Beyond diagnosis, the study offers something the industry can use immediately: a simple early-warning signal for overfermentation. At both study sites, in Nicaragua and Costa Rica, the researchers observed that the pH of the cocoa bean pulp in the fermentation mass increased significantly once the optimal fermentation duration had been reached. That means a measurement as basic as a test-strip pH reading could complement the standard checks of temperature, color, and odor that farmers and fermentary managers already perform. Detecting the onset of overfermentation early would allow producers to halt the process before cheesy, sweaty, or smoky off-flavors develop, protecting both quality and income. Notably, despite differences in climatic conditions, cocoa varieties, and fermentation methods between the two countries, the samples showed comparable results, suggesting that the pH signal may be robust across different production settings.

The work fits into a broader research program at the Leibniz Institute aimed at understanding the molecular background of cocoa quality. Previous studies by the group characterized the odorants causing smoky off-flavors in cocoa and clarified the molecular basis of a moldy-musty off-flavor, building a growing body of knowledge about the volatile compounds that make or break chocolate’s aroma. The current study adds an essential piece: a demonstration that a single sensory defect can have multiple, chemically distinguishable origins. Future research, the team says, will aim to clarify which microorganisms are involved in forming the identified compounds during fermentation, and at what concentrations these substances are actually perceived as off-flavors by consumers. Those answers could eventually translate into microbial management strategies for fermentaries.

For a commodity whose value depends so heavily on flavor, the implications are significant. Cocoa farmers, fermentary operators, chocolate manufacturers, and quality laboratories all stand to benefit from a more precise understanding of where off-flavors come from and how to catch them before they take hold. A cheap pH test strip, combined with the ability to read phenol fingerprints in suspect batches, could help ensure that the smoky notes in a chocolate bar come from intentional flavor design rather than an accident of processing. The study, published as “Off-Flavor Compounds in Fermented Cocoa: Impact of Overfermentation” in the Journal of Agricultural and Food Chemistry, was supported within the Industrial Collective Research program of the German Federal Ministry of Economic Affairs and Energy through the Research Association of the German Food Industry, with additional support for Krause through a doctoral scholarship from the Foundation of German Business. It is a reminder that in food science, as in chocolate making, the difference between delight and disappointment often comes down to chemistry measured in parts per billion.

Subject of Research: Molecular analysis of off-flavor compounds formed during overfermentation of cocoa beans

Article Title: Smoky off-flavors in cocoa aren’t caused solely by smoke

Article References: Smoky off-flavors in cocoa aren’t caused solely by smoke. (n.d.). Original publication

Image Credits: AI Generated

DOI: Not provided

Keywords: cocoa, chocolate, fermentation, overfermentation, off-flavors, food chemistry, aroma compounds, phenols, 2-methoxyphenol, 4-ethylphenol, smoke contamination, quality control