Bioconversion of ferulic acid and vanillylmandelic acid to 4-vinyl guaiacol and vanillin mediated by Oenococcus oeni: towards sustainable flavor production

Conventional extraction of vanilla pods is both insufficient and costly, while chemical synthesis fails to satisfy natural labeling requirements. Biotechnological approaches therefore offer a promising alternative for sustainable, high-value aromatic compound production. This study investigated the bioconversion of ferulic acid (FA), a hydroxycinnamic acid, and vanillylmandelic acid (VMA), a vanilloid, into vanillin and 4-vinylguaiacol (4VG) using three bacterial strains of Oenococcus oeni isolated from wine lees. Fermentation assays were performed with precursor supplementation at various concentrations and feeding strategies. Precursor consumption and aromatic compound production were monitored by Ultra-High-Performance Liquid Chromatography (UHPLC). For the highest-yielding fermentations, Downstream Processing (DSP) methods were applied to extract the target compounds. FA was metabolized into both 4VG and vanillin, while VMA was converted to vanillin via microbial catalysis. Maximum metabolic efficiency reached approximately 25.7% yield (referred to biocatalysis). These findings confirm the feasibility of biovanillin synthesis using O. oeni strains, supporting potential applications in natural flavor production and industrial scalability, and contributing to the development of more sustainable aromatic compound biotechnology.

Vanillin, the primary component of the vanilla bean’s aroma, is one of the most widely used flavoring agents in the world. Its distinctive sweet and creamy scent makes it indispensable in the fragrance, food, cosmetics, and pharmaceutical industries. The global demand for vanillin exceeds 20,000 metric tons annually, with the vast majority, over 85%, being produced synthetically from petrochemical sources or lignin (S. Kumar et al., 2025). However, growing consumer preference for natural and sustainable ingredients, coupled with environmental concerns, has driven interest in biovanillin over synthetic vanillin (Martău et al., 2021). This form is derived from renewable feedstocks such as ferulic acid (FA), eugenol, or agro-industrial waste, using microbial biocatalysis or enzymatic conversion (Converti et al., 2010; N. Kumar and Pruthi, 2014; Ma et al., 2022; Martău et al., 2021; Yan et al., 2016).

Vanillin is industrially synthesized from guaiacol through a two-step Rhône-Poulenc process. The first step involves the electrophilic substitution of glyoxylic acid onto guaiacol, producing vanillylmandelic acid (VMA) along with byproducts. The second step is the oxidative decarboxylation of VMA, which generates vanillin as the primary product alongside structurally related aldehydes such as o-vanillin derived from o-VMA (Taran et al., 1998). Due to the structural similarity of these compounds, the purification process is complex and requires advanced separation techniques. This synthesis pathway presents a trade-off between yield and purity, underscoring the importance of precise control during VMA formation to enhance the industrial feasibility of vanillin production. While the first step has been largely optimized to minimize byproducts, the second step depends on catalysts and oxidizing agents, which precludes the resulting vanillin from being classified as natural under U.S. regulatory standards. To qualify vanillin as natural, both guaiacol and glyoxylic acid must originate from natural sources, the synthesis of VMA must occur without synthetic chemical reagents, and the oxidation of VMA to vanillin must proceed without chemical catalysis (U.S. Food and Drug Administration (FDA), 2023).

Microbial production of vanillin offers a promising alternative to traditional methods, aligning with the principles of green chemistry and circular bioeconomy (D’Arrigo et al., 2024; S. Kumar et al., 2025). Engineered strains of Escherichia coli, Pseudomonas spp., and Amycolatopsis have demonstrated the ability to convert aromatic precursors into vanillin with increasing efficiency (Liu et al., 2023; Zhang et al., 2023).

Despite technical challenges such as low yields, product toxicity, and downstream purification, advances in metabolic engineering, synthetic biology, and bioprocess optimization are rapidly improving the feasibility of industrial scale biovanillin production (Ashengroph et al., 2011; Zamzuri and Abd‐Aziz, 2013).

The regulatory classification of vanillin plays a decisive role in its market value and industrial adoption. According to Article 3(2)(c) of Regulation (EC) No. 1334/2008 in both the EU and US markets, only vanillin produced through specific biotechnological processes from natural precursors can be labeled as ‘natural’, making microbial conversion routes commercially advantageous (European Parliament and Council of the European Union (2008).

FA and VMA are promising aromatic precursors for the microbial biosynthesis of vanillin, offering sustainable alternatives to petrochemical routes. FA, a ubiquitous phenolic acid in plant biomass and agro-industrial byproducts, is a well-established vanillin precursor due to its structural similarity and efficient microbial conversion (Ma et al., 2022). The canonical pathway involves a two-step process: FA is first converted to vanillic acid via feruloyl-CoA synthetase or FA decarboxylase, followed by reduction to vanillin via vanillin synthase or vanillate demethylase. In some systems, vanillic acid is further metabolized to protocatechuic acid, though this route is suboptimal for vanillin production (Kaur and Chakraborty, 2013). Notably, vanillin synthase (e.g., VpVAN) catalyzes the direct, one-step conversion of FA (and its glucoside) to vanillin (and vanillin glucoside) through C2 side-chain shortening via hydration and retro-aldol elimination (Gallage et al., 2014). VpVAN functions as a hydratase/lyase, enabling this transformation in vanilla pods and heterologous systems such as tobacco, barley, and rice cell cultures. However, as summarized in a recent publication of phenolic acid biotransformation (Lu et al., 2025) FA could undergo biotransformation through coordinated pathways including decarboxylation, reduction and hydrolysis and mediated by microbial enzymes with applications in flavor development and food fermentation (Kaur et al., 2023).

VMA, a metabolite of catecholamines and lignin-rich biomass, remains underexplored as microbial vanillin precursor. While its microbial conversion pathways are less characterized than those of FA, its vanillyl moiety suggests potential as a complementary substrate for biovanillin production. This study presents the first demonstration of VMA as a substrate for vanillin biosynthesis, highlighting its accessibility via chemical synthesis or natural isolation. Unlike FA, which primarily yields 4-vinylguaiacol (4VG) in Oenococcus oeni (O. oeni), VMA is converted exclusively to vanillin by this organism, potentially bypassing the need for multi-enzyme systems.

O. oeni, a lactic acid bacterium central to malolactic fermentation (MLF) in winemaking, possesses a versatile enzymatic repertoire—including glycosidases, esterases, and decarboxylases—that modulates aroma profiles by releasing bound volatiles (Pérez-Martín et al., 2015). Recent publication shows that O. oeni biofilms on oak surfaces enhance the transformation of aromatic compounds such as vanillin, eugenol, and furfural during wine aging (Bastard et al., 2016). Another publication has demonstrated that O. oeni exhibits strain-dependent glycosidase activities which contribute to aroma release during fermentation (Li et al., 2020). Strain-dependent activities further enable the release of terpenes, norisoprenoids, and C6 alcohols, underscoring its biotechnological potential beyond winemaking.

Several other microbial routes for vanillin production have been extensively explored, including the bioconversion of phenylpropanoid precursors such as eugenol, isoeugenol, and lignin using diverse bacterial, fungal, and engineered microbial systems (Paul et al., 2021). Existing approaches remain limited by low yields and overall process inefficiencies, highlighting the need for improved strategies.

This study investigates the microbial bioconversion of FA and VMA to vanillin by O. oeni under anaerobic fermentation conditions, aiming to contribute to the development of sustainable flavor production pathways that reduce reliance on fossil resources and address the increasing demand for natural ingredients. Particular emphasis is placed on evaluating VMA as a novel precursor for vanillin production, as it enables a functionally simpler biotransformation strategy compared to FA and results in higher vanillin concentrations. Overall, this work broadens the range of substrates available for microbial vanillin production and highlights the potential of strain-level biocatalyst selection for sustainable biotechnological applications.

Isolation and identification of O. oeni strains

A total of 30 bacterial isolates were obtained from five spontaneously fermented Xinomavro wines originating from different regions of Greece. One milliliter of each wine sample was aseptically transferred into 9 mL of sterile ¼ Ringer’s solution and serially diluted in the same diluent. Then, 0.1 mL of the appropriate dilution was spread onto de Man–Rogosa–Sharpe (MRS) agar (Biolife, Milan, Italy), adjusted to pH 5.7 and supplemented with 0.05% (w/v) cycloheximide (AppliChem GmbH, Darmstadt, Germany) to inhibit yeast growth, and incubated at 30 °C for ten days. The isolates were stored at −80 °C in MRS broth supplemented with 30% glycerol (AppliChem, Darmstadt, Germany). Before experimental use, each isolate was subcultured twice in MRS broth at 28 °C for five days.

Selection and genetic characterization of the bacterial isolates at the species level were performed through 16S rRNA gene sequencing according to Tzamourani et al. (2021). Briefly, PCR was performed in a 50 μL reaction mixture containing 500 ng of template DNA, 0.2 mM dNTPs, 1.5 mM MgCl 2, 0.2 μM of each primer (P1V1: 5′-GCG GCG TGC CTA ATA CAT GC-3′; P4V3: 5′-ATC TAC GCA TTT CAC CGC TAC-3′), and 1 U of Taq polymerase (Kapa Biosystems, USA). The amplification protocol included an initial denaturation at 94 °C for 2 min, followed by 30 cycles of denaturation at 94 °C for 30 s, annealing at 56 °C for 30 s, and extension at 72 °C for 90 s, with a final extension at 72 °C for 15 min. The LAB isolates identified as O. oeni were further differentiated at the strain level using multiplex PCR analysis following Claisse and Lonvaud-Funel (2012). The three newly isolated strains, MD00, MD02, and MD05, belonging to the collection of the Wine Microbiology Laboratory of UNIWA, were used in this study.

Growth conditions

Prior to use, bacterial strains were under anaerobic conditions (sealed conical vessels) cultured in MRS broth at 30 °C composed of: 20 g L -1 glucose, 4 g L -1 yeast extract, 10 g L -1 peptone, 10 g L -1 meat extract, 2 g L -1 K 2 HPO 4, 0.2 g L -1 MgSO 4 ×7H 2 O, 0.05 g L -1 MnSO 4, and 1mL L -1 Tween 80. The pH was adjusted to 4.5. The medium was sterilized by autoclave (121 °C for 15 minutes). Fermentations were initiated by liquid-to-liquid inoculation using actively growing cultures to achieve an initial cell concentration of approximately 10 7 cells mL -1. Cell density and growth was monitored by measurement of viable cells in improved Neubauer hemocytometer (ISOLAB Laborgeräte GmbH, Eschau, Germany), in combination with methylene blue staining (ratio 1:1) (Pérez-González et al., 2018).

Substrates

FA (purity > 98%) was purchased from Layn Europe S.r.l. (Savona, Italy) and VMA (purity = 80%) was manufactured at Vioryl S.A. (Athens, Greece). For VMA characterization calibration curve was used against commercial VMA standard (Sigma-Aldrich, ≥ 98% purity by HPLC) and screening took place for possible vanillin precursors with none of them being found. Precursors FA and VMA were added to the culture media and sterilized by autoclaving. Although FA is known to undergo thermal decomposition at elevated temperatures, it remains stable under standard sterilization conditions, while autoclaving was necessary because the precursors were insufficiently soluble in water at 20 °C to allow sterilization by membrane filtration (Brito et al., 2019). Two types of substrate feeding strategies were carried out. One-time feeding strategy at t 0, utilizing 4 mmol L -1 FA and 4 mmol L -1 VMA, with sampling performed on days 2 and 7 post-inoculation. To assess the impact of increased substrate loading on vanillin yields, additional fermentations were executed under continuous substrate supply. VMA was supplied at concentrations of 10.9 mmol L -1 and 16.36 mmol L -1 (only for strains MD02 and MD05), and the kinetic profile of the process was monitored through periodic sampling at 0, 2, 3, 6, 8, 10, 14, 16, 20, 22 and 24 days. For the first experiment, VMA was added at a concentration of 1.818 mmol L -1 on days 0, 2, 3, 6, 8 and 10, whereas for the second experiment, VMA was added at the same concentration (1.818 mmol L -1) on days 0, 2, 3, 6, 8, 10, 16, 20 and 22.

Fermentation conditions

Fermentations were conducted in 300 mL and 500 mL sealed conical flasks containing 250 mL (one-time feeding strategy) or 330 mL (continuous substrate supply) MRS broth, respectively. Flasks were inoculated with microbial strains and incubated at 30 °C with continuous stirring (120 RPM). Anaerobic conditions were maintained throughout the experiment. Samples were collected and centrifuged (8000 RPM, 10 min, room temperature). The resulting supernatant was sterile filtered through a 0.22 µm membrane.

Extraction and isolation of vanillin

At the end of fermentation, the culture medium was centrifuged at 10,000 rpm for 10 minutes at 2 °C to separate biomass. The liquid phase was subjected to liquid-liquid extraction with 250 mL diethyl ether (99.5% purity). Diethyl ether was selected due to its high efficiency for the extraction of non-polar aromatic compounds and was used under controlled laboratory conditions, ensuring complete removal prior to analysis in accordance with Directive 2009/32/EC residue requirements. The extraction was repeated twice to maximize vanillin recovery. Tween 80 formed a gel during extraction, which was disrupted by adding 30 g sodium chloride and 20 g sodium sulfate to dehydrate the organic phase. The organic phase was filtered to remove solids, and the filtrate was concentrated using rotary evaporation under vacuum.

Analytical methods

Standard curves were prepared for vanillin, FA, 4VG and VMA using authentic standards (Vioryl Chemical and Agricultural Industry, Research S.A.) to ensure accurate quantification. The formation and quantification of metabolites during the fermentation process were monitored using Ultra-High-Performance Liquid Chromatography (UHPLC). Calibration curves were constructed in triplicate for all standards of vanillin demonstrating excellent linearity (R² = 0.9992). Calibration was also performed in the matrix medium (MRS) to account for potential matrix effects. The method showed a limit of detection (LOD) of 0,0068 mg L -1 and a limit of quantification (LOQ) of 0,0022 mg L -1. Recovery experiments yielded an average recovery of 110%, indicating good method accuracy. Similar analytical performance was observed for the remaining compounds, with FA (LOD 0.0067 mg L -1, LOQ 0.0225 mg L -1, recovery 91.5%, R² = 0.9999), 4VG (LOD 0.45 mg L -1, LOQ 1.5 mg L -1, recovery 96.8%, R² = 0.9992) and VMA (LOD 15.04 mg L -1, LOQ 50.15 mg L -1, recovery 93.25%, R² = 0.9974).

In addition, to further confirm and characterize the chemical composition of the fermented extract, post-fermentation products were analyzed using Gas Chromatography–Mass Spectrometry (GC–MS).

Ultra-high-performance liquid chromatography

Analyses had been conducted with the use of an UHPLC-MS (LC-40DX3, LCMS-2020, Shimadzhu, Kyoto, Japan) equipped with UV detector and RP-UHPLC-C18 column (Purospher STAR, RP-18, 2μm). The mobile phases consisted of 0.1% (v/v) formic acid in water (Solvent A) and ACN (Solvent B). The elution gradient was programmed as follows: an initial hold at 10% solvent B for 3 minutes, followed by a linear increase from 10% to 30% B over 1 minute. This was maintained at 30% B for 13 minutes, after which the gradient was returned to 10% B and equilibrated from 18 to 25 minutes. The flow rate was maintained at 0.2 mL/min, and the injection volume was 5 µL. The UV detector operated at 254 nm. Mass spectrometric detection was conducted by operating in negative ion mode with an electrospray ionization (ESI) source. The source-related parameters were optimized as follows: a DL temperature of 250 °C, a nebulizer set at 1.5L min -1 and dry gas at 10L min -1. Data were acquired over a mass-to-charge (m/z) range of 100–1500. All mass spectra were averaged and background subtracted in the ESI negative mode. LabSolutions (Version 5.113 Shimadzu Corporation) software was used for data evaluation.

Gas chromatography-mass spectrometry

GC-MS analysis was performed using a Shimadzu GC/MS QP2010 ULTRA system with an HP-1ms column (20 m × 0.18 mm i.d., 0.18 μm film thickness; Agilent Technologies). Electron ionization was operated at 70 eV, with an injection temperature of 250 °C, an injection volume of 1 μL, and a split ratio of 1:30. The carrier gas flow was maintained at 0.46 mL/min under constant flow conditions. The oven temperature program began at 50 °C, ramped at 30 °C/min to 80 °C (held for 1 minute), then increased at 100 °C min -1 to 250 °C (held for 10 minutes). Compound identification was achieved by comparing retention times and mass spectra with entries in the Wiley/NIST05 libraries and an in-house database of authentic standards provided by Vioryl Chemical and Agricultural Industry, Research S.A., with additional confirmation using Kováts retention indices.

Bioconversion yield determination

The bioconversion yield of VMA to vanillin was determined based on reaction stoichiometry and experimentally measured concentrations. Assuming a 1:1 molar conversion, 1 mol of VMA theoretically yields 1 mol of vanillin. Concentrations initially expressed in mg/L were converted to mol/L using the respective molecular weights (198 g mol -1 for VMA and 152 g mol -1 for vanillin) to enable calculations on a molar basis. The theoretical vanillin production was estimated from the initial concentration of added VMA under the assumption of complete (100%) conversion, while the actual vanillin produced was obtained from the measured final concentration. Bioconversion yield was expressed as the percentage ratio of the produced vanillin to the theoretically expected amount. The calculation also accounted for the purity of the VMA substrate (80%). Accordingly, yield (%) was calculated as: (produced vanillin/theoretical vanillin) × 100. The bioconversion yield of VMA to vanillin for the inoculated strains was determined after correction for thermal decomposition, by subtracting the vanillin yield measured in the uninoculated control.

Statistical analysis

Data are presented as mean ± standard deviation (SD) from independent experiments, as described in each figure legend. Statistical comparisons between groups were performed using one-way analysis of variance (ANOVA), followed by Tukey–Kramer’s honestly significant difference (HSD) post hoc test to identify significant differences among group means. Differences were considered statistically significant at p< 0.05. All statistical analyses were performed using RStudio (version 4.2.3).

Bacteria isolation and genetic characterization

A total of 30 bacterial isolates were recovered from the final stage of alcoholic fermentation in five spontaneously fermented Xinomavro wines from different regions of Greece. All isolates were identified as O. oeni (Supplementary Table 1). Strain-level characterization by multiplex PCR revealed three distinct O. oeni profiles. Notably, despite differences in vintages, the same strain of O. oeni (MD00 in Xinomavro 1 & 2 and MD02 in Xinomavro 3 & 4) was detected in different wine samples.

Biotransformation of FA to produce 4VG and vanillin and biotransformation of VMA to produce vanillin utilizing two substrate feeding strategies

The conversion of FA into 4VG and vanillin follows a multi-step biochemical pathway that begins with a decarboxylation reaction. In this route, FA is initially decarboxylated to produce 4VG. To produce vanillin from the same substrate, probably a βoxidation occurs after hydration of the double bond. This intermediate may then undergo redox reactions, such as oxidation, to give rise to vanillin.

When VMA is used as the substrate a very limited amount of 4VG is observed, which confirms the difference between the two pathways. When comparing the yields between these two routes, distinct differences emerge in vanillin accumulation profiles. The FA pathway, though capable of converting into vanillin, typically results in a significant accumulation of the intermediate 4VG, with vanillin being formed in lower concentrations (Figures 1A, B).

In contrast, the use of VMA tends to produce vanillin in higher concentrations, about 9 times more than FA (Figure 1D). Notably, strain-specific performance—as observed in this study with strains MD00, MD02 and MD05—indicates that certain strains are more efficient at converting the substrate directly towards vanillin synthesis from VMA, thereby achieving higher final concentrations of vanillin. More specifically, in case of VMA, MD05 consumes more VMA (Figure 1C) produces more vanillin at the end of seven days of fermentation (Figure 1D).

This study evaluated two substrate feeding strategies for the biotransformation of VMA into vanillin: one-time substrate feeding and continuous substrate feeding, assessing their effects on vanillin production kinetics and microbial activity.

One-time feeding strategy

In this approach, VMA is introduced to the fermentation medium only once at the initiation of the process. This leads to a rapid initial conversion of VMA into vanillin due to substrate abundance and high microbial metabolic activity. However, substrate depletion occurs as fermentation progresses, resulting in a plateau in vanillin accumulation and a subsequent decline in production rate. This behavior reflects substrate-limited biotransformation kinetics, where the absence of additional VMA restricts further enzymatic conversion (Figure 1D).

Continuous feeding strategy

Alternatively, VMA is supplied incrementally and continuously over the entire fermentation period. This feeding modality maintains a near-constant substrate concentration, avoiding substrate exhaustion and sustaining the metabolic activity of the microbial cells. Continuous substrate replenishment ensures prolonged enzymatic activity of key biotransformation enzymes, leading to a steady, extended production phase of vanillin (Figure 1).

Role of thermal decomposition and uninoculated control

A significant consideration in the accurate quantification of biotransformation-derived vanillin is the chemical transformation of VMA during sterilization. Autoclaving and other sterilization procedures expose the fermentation medium to high temperatures and pressures, conditions under which VMA undergoes non-enzymatic thermal decomposition to vanillin. This abiotic reaction confounds vanillin quantification, as vanillin detected post-sterilization may not exclusively originate from microbial activity. To address this, an uninoculated control—nutrient medium supplemented with VMA but without inoculation of microbial cells—was subjected to the same sterilization and incubation conditions. This control enables quantification of vanillin formed strictly through thermal decomposition during sterilization and any possible abiotic reactions occurring during incubation. By establishing this baseline, the contribution of non-biological vanillin formation is accounted for and subtracted from total vanillin measurements in inoculated cultures, ensuring accurate assessment of microbial biotransformation efficiency.

Experimental observations

Figure 2 demonstrates the vanillin production rates under both feeding regimes. The microbial strain MD05 exhibited the highest vanillin synthesis rate, particularly evident on day 14 of fermentation. This indicates MD05’s enhanced enzymatic conversion capacity and effective utilization of the continuously supplied VMA.

Cell viability was monitored by colony-forming units per milliliter (cells mL -1). MD05 maintained comparatively stable and high cell counts throughout the fermentation, correlating with sustained metabolic activity. The temporal alignment of peak cell viability and maximum vanillin production on day 14 highlights the strong dependence of vanillin biosynthesis on microbial health and proliferation.

Integrating these datasets—vanillin yield corrected for thermal decomposition, production rate, and cell viability—validates that the continuous feeding strategy with strain MD05 yields superior and sustained vanillin production. This strain’s robustness under prolonged fermentation conditions and steady substrate supply underscores its potential for scalable and efficient biotransformation processes.

Instantaneous rate of bioconversion

Analysis in Figure 3 evaluates the vanillin production rates of O. oeni strains MD02 and MD05 under continuous VMA feeding at two concentrations: 10.9 mmol/L (2160 mg L −1) and 16.36 mmol/L (3240 mg L −1). The data presented in Figure 3, compare microbial biotransformation (red) to uninoculated controls (blue) over 14–24 days, providing insights into strain-specific performance and the impact of substrate concentration on vanillin biosynthesis.

At 10.9 mmol L - ¹ VMA (Panel A), strain MD02 exhibited a gradual increase in vanillin production, reaching a peak at approximately day 10, after which production levels plateaued. In contrast, the uninoculated control showed negligible vanillin formation, confirming the microbial origin of the compound. The observed plateau may indicate substrate limitation or the onset of product inhibition, suggesting that further process optimization is required to sustain production beyond this stage.

At 16.36 mmol L - ¹ VMA (Panel B), for the strain MD02 vanillin production was significantly higher relative to the lower substrate concentration, with peak levels observed between days 10 and 12, followed by a decline in vanillin concentration. As in Panel A, the uninoculated control remained negligible, further supporting the microbial dependence of vanillin formation. The post-peak decrease may be attributed to metabolic stress, substrate overloading, or product toxicity at elevated VMA concentrations. These findings highlight the importance of optimizing substrate concentration and improving strain tolerance to sustain efficient vanillin biosynthesis under high precursor conditions.

At 10.9 mmol L −1 VMA (Panel C) for the strain MD05 the vanillin production rate increased linearly throughout the 14-day period, with no plateau or decline. The control showed minimal vanillin formation, highlighting MD05’s robust and sustained biocatalytic activity at this substrate concentration. At 16.36 mmol L −1 VMA (Panel D) for the strain MD05 the production rate was markedly higher, with a steep, linear increase over 14 days. The control remained negligible, underscoring MD05’s exceptional efficiency in converting VMA to vanillin. The absence of a plateau or decline suggests superior enzymatic resilience and potential for high-yield, scalable bioprocesses. Strain MD05 consistently outperformed MD02 in vanillin production, particularly at the higher VMA concentration. The linear production trend for MD05 at 16.36 mmol L −1 indicates a strong capacity for sustained biotransformation, making it the preferred strain for industrial applications. Both strains demonstrated increased vanillin production at 16.36 mmol L −1 VMA, but MD05’s linear and sustained production highlights its adaptability to elevated substrate levels, minimizing risks of early inhibition or metabolic stress. Finally microbial versus abiotic contribution, is showed by the negligible vanillin production in uninoculated controls across all panels and confirms that vanillin formation is predominantly driven by microbial activity, with minimal contribution from thermal decomposition or abiotic reactions.

MD05 is the most promising candidate for industrial vanillin production due to its sustained, high-yield performance and tolerance to increased VMA concentrations.

Continuous VMA feeding at 16.36 mmol L −1 maximizes vanillin production, particularly with MD05. Further optimization of fermentation parameters (e.g., pH, temperature, nutrient supplementation) could enhance productivity and mitigate potential inhibition observed in MD02.

The linear production trend for MD05 at high VMA concentrations suggests significant potential for process scale-up. Future work should focus on elucidating the enzymatic or genetic factors underlying MD05’s superior performance and optimizing downstream processing to improve vanillin recovery and purity.

Vanillin extraction—downstream processing

Vanillin extraction from fermentation broths presents significant challenges due to its water solubility and susceptibility to loss during downstream processing. This study evaluated both solvent extraction efficiency and overall vanillin yield, enabling the assessment of results even in the absence of statistical validation with focus on practical recovery.

Post-fermentation, cultures were centrifuged to remove biomass, followed by liquid-liquid extraction using an organic solvent. The organic phase was concentrated under reduced pressure, while the aqueous phase underwent further extraction to recover residual vanillin. Total vanillin yield was calculated as the sum of vanillin recovered from both phases. Quantification was performed using UHPLC and GC-MS, with pre-extraction UHPLC measurements serving as a baseline for vanillin present in the broth. The difference between pre-extraction UHPLC and post-extraction GC-MS values quantify losses attributable to incomplete recovery, degradation, or adsorption during process (Table 1).

Uninoculated controls and reagent blanks were processed alongside samples to exclude instrumental artifacts and false positives, ensuring detected compounds originated solely from microbial biotransformation. GC-MS and cell viability assays confirmed the absence of toxic byproducts, validating the biosafety of the process.

UHPLC measurements consistently exceeded post-extraction yields, with losses ranging from 31% to 46% of initial values. These losses are attributed to incomplete phase partitioning, surface adsorption, volatilization, and potential degradation. Downstream recovery vanillin typically results in losses reaching 40% of initial values depending on the extraction strategy. Alternative downstream approaches such as absorption resins are suggested for improving vanillin recovery in future studies.

Strain MD05 demonstrated the highest recovery efficiency, with a smaller absolute difference (132 mg) relative to its UHPLC measurement (432 mg), suggesting superior extraction dynamics. The control exhibited the largest relative loss, likely due to lower vanillin titers and less efficient extraction in minimally converted media.

Figure 4 presents the total ion chromatogram (TIC) obtained by GC-MS analysis of the fermented extract (MD02) at the end of downstream processing. The chromatogram shows well-resolved peaks corresponding to the compounds identified in Table 2, including organic acids (e.g., lactic, butyric, isobutyric, and isovaleric acids), benzaldehyde, and vanillin. Vanillin is observed as a prominent peak, consistent with its accumulation during fermentation. No additional unexpected peaks associated with isomers or other toxic byproducts were observed under the applied analytical conditions.

As Conclusion, integrating direct analytical quantification (UHPLC/GC) extraction yields provides a robust framework for evaluating fermentative vanillin titers. While UHPLC/GC-MS values reflect total production, post-extraction yields offer a practical estimate of recoverable vanillin. GC-MS analysis also detected lactic acid, confirming expected bioconversion pathways (e.g., O. oeni malolactic fermentation). The absence of undesirable compounds further supports the process’s suitability for scalable applications.

Several microorganisms, including lactic acid bacteria (LAB) generally recognized as safe (GRAS), have demonstrated the ability to bioconvert ferulic acid into vanillin. Among LAB, species such as Lactobacillus hilgardii, Pediococcus damnosus, O. oeni, Lactobacillus brevis, and L. plantarum have been reported to possess vanillin-producing capabilities (Kaur and Chakraborty, 2013; Knockaert et al., 2012). For example, Kaur et al. (2013) reported bioconversion of ferulic acid (obtained from rice bran) into vanillin by a probiotic isolate, Pediococcus acidilactici M16. Beyond LAB, other bacterial species have also shown efficient vanillin production. Enterobacter hormaechei produced up to 5.2 g L -1 vanillin using rice bran oil as substrate (Mazhar et al., 2017), while Brevibacillus agri 13 demonstrated vanillin production at elevated temperatures (45 °C) (Wangrangsimagul et al., 2012). In addition, a vanillin-resistant strain, Pseudomonas putida BD16, was used for vanillin production from rice bran, yielding 1.06 g/L -1 vanillin (Chakraborty et al., 2017). Compared with these LAB species, O. oeni is uniquely adapted to the harsh wine environment, enabling sustained metabolic activity under conditions that inhibit many other bacteria. However, despite previous reports describing ferulic acid-derived aroma metabolism in LAB, evidence for the direct conversion of VMA to vanillin by O. oeni remains lacking. Therefore, the present study provides, to the best of our knowledge, the first phenotypic evidence of VMA bioconversion to vanillin by O. oeni.

The exclusive identification of O. oeni among isolates recovered at the end of alcoholic fermentation aligns with its established ecological dominance in wine. This species is uniquely adapted to post-fermentation conditions—low pH, high ethanol, and nutrient limitation—which enable it to outcompete other lactic acid bacteria and drive malolactic fermentation (Balmaseda et al., 2023; Lonvaud-Funel, 1999; Renouf et al., 2005).

At the strain level, the detection of identical O. oeni profiles across different wines and vintages suggests the persistence of resident populations within the winery environment. This persistence is likely facilitated by biofilm formation on winery surfaces, which enhances resistance to stressful conditions compared to planktonic cells (Bastard et al., 2016; Dimopoulou et al., 2018). The repeated isolation of strains MD00 and MD02 supports the hypothesis that only well-adapted genotypes persist in the cellar ecosystem, contributing to the microbial terroir of Xinomavro wines.

All three O. oeni strains efficiently converted ferulic acid (FA) to 4-vinylguaiacol (4VG), a transformation characteristic of phenolic acid decarboxylases (PADs). PADs are cofactor-independent enzymes that catalyze decarboxylation via a quinone methide intermediate, requiring a conjugated system between the β-carbon and the para-hydroxy group (Payer et al., 2019). The rapid and complete depletion of FA within 48 hours (Figure 1A) strongly indicates a PAD-mediated pathway, consistent with substrate specificity observed in Lactobacillus plantarum and Bacillus spp (Li et al., 2019; Payer et al., 2019).

Vanillin was detected at low but measurable concentrations (Figure 1B), reaching 4.3 mg/L from 776 mg/L ferulic acid under the tested conditions. Although the overall conversion efficiency remained limited, the production of vanillin under a defined ferulic acid bioconversion setup demonstrates the capability of the system to direct carbon flux toward aromatic aldehyde formation. In comparison, Taira et al. (2018) reported the occurrence of vanillin during Aspergillus luchuensis fermentation; however, their study did not quantify ferulic-acid-to-vanillin conversion yields, making direct productivity comparisons difficult. The gradual increase in vanillin concentration suggests ongoing, but less efficient, reduction activity, likely involving cofactor-dependent reductases (Eicher et al., 2024; Luziatelli et al., 2019).

In contrast, VMA was preferentially converted to vanillin. The structural differences between FA and VMA-specifically, the lack of a conjugated acrylic side chain and the presence of an α-hydroxyl group, preclude PAD involvement. The hypothetical conversion of VMA to vanillin suggests a cofactor-dependent decarboxylase/reductase system, possibly analogous to prFMN-dependent decarboxylases or metal-dependent enzymes (Payer et al., 2019). A hypothesis that could be tested through enzyme assays targeting decarboxylase activity, as well as genomic screening for candidate prFMN- or TPP-dependent enzymes in O. oeni.

The steady increase in vanillin production from VMA over the fermentation period supports the hypothesis of a distinct, non-PAD-mediated pathway. However, the lack of genetic or biochemical evidence in this study means the VMA-to-vanillin conversion remains phenotypically described. Decarboxylation is the enzymatic cleavage of a carboxylic acid moiety with concomitant CO 2 release. It is a fundamental biochemical transformation integral to both primary and secondary metabolism. While thermodynamically spontaneous, the uncatalyzed reaction proceeds at negligible rates, necessitating enzymatic intervention. Evolution has thus given rise to a diverse array of decarboxylases, categorized by their mechanistic strategies (heterolytic vs. homolytic C–CO 2 bond scission) and cofactor dependencies (e.g., thiamine pyrophosphate (TPP), pyridoxal 5′-phosphate (PLP), or prenylated flavin mononucleotide (prFMN) ((Nguyen et al., 2024; Payer et al., 2019) (Figure 5). An example is benzoylformate, a similar molecule to VMA. Benzoylformate Decarboxylase (BFD) is a TPP-dependent enzyme that catalyzes the decarboxylation of benzoylformate to benzaldehyde and CO 2. Ferulic Acid Decarboxylase (FDC) utilizes a prFMN cofactor to mediate the radical-based decarboxylation of FA, yielding 4VG. FDC’s substrate promiscuity extends to other hydroxycinnamic acids, underscoring its biocatalytic versatility. These enzymes exemplify the mechanistic plasticity of decarboxylases, which not only underpin core metabolic pathways but also drive the structural diversification of secondary metabolites. By expanding the chemical repertoire of natural products, decarboxylases influence the pharmacological properties of bioactive compounds, positioning them as prime targets for enzyme engineering and the development of novel biocatalysts for synthetic biology and drug discovery (Nguyen et al., 2024).

Strain-specific differences in vanillin yield (Figures 2D, 3) indicate variability in metabolic pathways or enzyme expression. MD05 consistently produced the highest vanillin concentrations, particularly in extended fermentations, suggesting enhanced enzymatic activity or greater tolerance to substrate and product accumulation or adaptive fitness traits (Zheng et al., 2024; Liu et al., 2023). Conversely, MD00 exhibited the lowest yield and was excluded from prolonged trials. These observations align with emerging evidence of genetic and phenotypic diversity among O. oeni isolates, influencing their aromatic compound metabolism. Such strain dependent metabolic performance is consistent with another study which reports variability in enzymatic activities among O. oeni strains contributed to aroma modulation (Li et al., 2020).

Cell viability remained stable throughout fermentation, though late-stage variability likely reflects metabolic stress from substrate depletion or product accumulation (Mehmood et al., 2022). Uninoculated controls confirmed the microbial dependency of vanillin formation (Figure 3, Supplementary Figure S1), ruling out abiotic contributions. Incremental precursor feeding sustained conversion activity, with MD05 demonstrating the highest productivity, followed by MD02. Vanillin yields and cell viability peaked on day 14, particularly in extended fermentations, highlighting the potential of prolonged, fed-batch processes for maximizing production.

This study demonstrates that O. oeni strains MD00, MD02, and MD05 employ distinct enzymatic strategies for the biotransformation of phenolic acids. FA is primarily converted to 4VG via potentially a cofactor-independent PAD pathway, with minor vanillin formation. In contrast, VMA is preferentially converted to vanillin, likely through a cofactor-dependent decarboxylase/reductase system, reflecting substrate-specific enzymatic flexibility.

The rapid conversion of FA and sustained vanillin production from VMA underscores the potential of these strains as natural flavor biocatalysts. MD05 exhibited superior biocatalytic efficiency and tolerance to product accumulation, while strain-specific differences in conversion efficiency and cell viability suggest opportunities for targeted strain selection and process optimization. Incremental precursor feeding effectively sustained activity over extended periods, validating a scalable approach for enhancing vanillin yields.

While FA conversion yielded relatively low vanillin concentrations, VMA proved a more effective precursor, emphasizing the importance of substrate specificity and metabolic flux. The observed variability among strains implies distinct enzymatic profiles or regulatory mechanisms, aligning with emerging evidence of genetic and phenotypic diversity among O. oeni isolates.

Future research should focus on elucidating the specific enzymatic pathways, particularly the putative cofactor-dependent system for VMA conversion and optimizing fermentation parameters for industrial applications. Comparative genomic analyses could further support strain selection by identifying variants associated with high-performing phenotypes, such as MD05. Scale-up studies and process integration will be critical for translating these findings into commercially viable natural flavor production technologies.