Enhancement of nutrition and flavor in the yellow suspension of a mutant strain of Auxenochlorella pyrenoidosa by fermentation with an isolated Lactiplantibacillus plantarum NH3

Abstract

Lactiplantibacillus plantarum HN3 was utilized to enhance the nutritional value and flavor of yellow suspension of Auxenochlorella pyrenoidosa mutant CX41 fermented for 10 h. High cell viability was observed, with no significant changes in hydrolyzed amino acid scores, although the free amino acid profile was altered. Total fatty acids and γ-aminobutyric acid increased by 54.02% and 35.00%, while lutein levels remained stable and in vitro antioxidant capacity was enhanced. Sensory evaluation indicated that reduced off-odors and increased floral, fruity, and herbaceous notes improved acceptance. The modified taste enhanced flavor complexity by decreasing sweetness and umami by 61.13% and 49.95%, eliminating bitterness, and increasing sourness by 69.38%. Metabolomic profiling revealed that reduced relative odor activity of 2-thiophenemethanethiol, isopentenyl mercaptan, 1-hexanol, 1-hexen-3-one, and δ-dodecalactone contributed to the improved sensory attributes. These findings demonstrate the promising application of L. plantarum HN3 fermentation in developing high nutritional and desirable flavor foods of A. pyrenoidosa.

Highlights

• •L. plantarum fermentation enhanced nutritional and sensory quality of A. pyrenoidosa.
• •Total fatty acids increased by 54% and GABA by 35% while lutein remaining stable.
• •Antioxidant capacity was improved significantly with maintained amino acid scores.
• •Sensory evaluation showed fewer off-odors and more floral, fruity, and grassy notes.
• •Taste shifted to less sweetness and umami, no bitterness, and increased sourness.

1. Introduction

The global economic development and rapid population growth have resulted in a significant increase in the demand for dietary protein (Y. Wang et al., 2022). The production of animal-derived proteins is closely associated with environmental degradation, climate change, and health risks arising from excessive consumption, raising concerns regarding their long-term sustainability (Sandner et al., 2020). Over the past decade, consumer interest in plant-based proteins for products like meat analogue, plant-derived beverages and egg substitutes has steadily increased, primarily driven by considerations of animal welfare and environmental protection (Zhang et al., 2022). Plant protein-based fermented milk has garnered considerable attention as a viable substitute for dairy milk, particularly among individuals with lactose intolerance, milk allergies, a hypercholesterolemia, and those following a vegan diet (Liu et al., 2025). At present, soy protein remains the dominant raw material in plant protein sources, with limited use of peanut and pea proteins (Ismail et al., 2020; McClements et al., 2021). However, many conventional sources of plant and microbial proteins suffer from intrinsic limitations such as an imbalanced essential amino acid profile and incomplete nutritional composition, which restrict their capacity to fully replace animal proteins in advanced food systems (Tallman et al., 2023).

The green microalga Auxenochlorella pyrenoidosa (formerly named as Chlorella pyrenoidosa) is recognized as a high-protein ingredient, containing 50–65% dry weight of protein, along with a rich array of amino acids, polysaccharides, pigments, polyunsaturated fatty acids (PUFAs), vitamins, and minerals (Shi, Li, et al., 2024). Its rapid growth and high cell density fermentation further underscore its potential as an alternative protein source for future food applications. However, its limited use in dietary supplements and health food products (such as tablets and capsules) has been attributed to its deep green colour and strong odor, which adversely affect consumer acceptance (Wang et al., 2025). In our previous study, we developed a novel yellow mutant, CX41, of A. pyrenoidosa through atmospheric and room temperature plasma (ARTP) mutagenesis and norflurazon-based screening (Shi, Li, et al., 2024). Under dark fermentation conditions, the CX41 strain exhibited a golden-yellow coloration and higher protein and lutein contents, resulting in improved sensory properties and nutritional quality, thereby overcoming significant barriers to its use as a consumer-friendly protein source. Currently, the food processing methods for A. pyrenoidosa remain relatively limited, which has prevented its market potential from being fully realized (Shi, Li, et al., 2024). Consequently, effectively harnessing the nutrients and bioactive compounds of the yellow CX41 mutant to develop innovative, high-protein foods with enhanced consumer acceptability has become a key challenge for the food industry.

Lactic acid bacteria (LAB) are a diverse group of gram-positive, non-spore-forming microorganisms widely utilized in fermented foods (Abdul Hakim et al., 2023). Recent genomic and functional studies on various Lactiplantibacillus plantarum strains have further elucidated their genotypic profiles, probiotic characteristics, safety attributes, and potential bioactive metabolites, reinforcing their promising applications in food fermentation (Aziz, Hangyu, et al., 2025; Aziz, Naveed, Shabbir, Sarwar, Khan, et al., 2024; Aziz, Naveed, Shabbir, Sarwar, Naseeb, et al., 2024; Aziz, Shabbir, et al., 2025). In recent years, the fermentation of LAB in plant-based substrates, such as soybean, oat, and rice proteins, as well as microalgae like Chlorella vulgaris and Arthrospira platensis, has been extensively studied due to its demonstrated ability to improve texture, enhance flavor, and increase nutritional value (Matos et al., 2025; Montemurro et al., 2021; Niccolai et al., 2020). During fermentation, LAB not only produce organic acids, including lactic acid and acetic acid, but also generate a variety of metabolites such as alcohols, aldehydes, and esters. Collectively, these compounds contribute to the development of complex flavor profiles, resulting in enhanced aroma and improved sensory attributes of fermented foods (Chen et al., 2025; Yang et al., 2024). Although LAB lack the ability to synthesize all essential amino acids, they compensate by secreting proteases and peptidases that hydrolyze plant proteins into amino acids and peptides (Yang et al., 2024). While plant-based raw materials typically contain little or no γ-aminobutyric acid (GABA), numerous studies have shown that LAB fermentation can significantly increase GABA levels through the conversion of L-glutamic acid (Icer et al., 2023). Furthermore, LAB fermentation creates an acidic environment that enhances vitamin stability and promotes vitamin synthesis in plant-based foods (Shi, Qin, et al., 2024). Their enzymatic reactions convert lipids, proteins, and carbohydrates into flavor precursors and aromatic compounds, resulting in fermented plant-based beverages characterized mainly by organic acids, amino acids, sugars, and volatile components (Yang et al., 2024). In addition, LAB fermentation enhances antioxidant capacity by increasing phenolics, vitamins, and other bioactive compounds (Filannino et al., 2014). The yellow, high-protein CX41 mutant of A. pyrenoidosa represents an ideal substrate for the development of fermented foods. However, research on LAB fermentation of this protein-rich microalga is still limited.

In this study, L. plantarum HN3, isolated from an outdoor raceway open pond containing A. pyrenoidosa, was utilized to ferment the yellow suspension of the A. pyrenoidosa mutant CX41. The variations in nutrients (proteins, amino acids, fatty acids, organic acids, and sugars) as well as functional compounds such as GABA and lutein, were systematically compared alongside viable LAB growth, total acidity, and antioxidant activity in the suspension before and after fermentation. Flavor profiles and variations were analyzed using an electronic tongue, an electronic nose, and metabolomics, elucidating the molecular basis for flavor enhancement induced by the fermentation of L. plantarum HN3.

2. Materials and methods

2.1. Strains, media and culture conditions

The yellow mutant CX41 of A. pyrenoidosa was bred and cultured in modified Basal medium (Table S1) for algal powder production at 30 °C, as reported in our previous study (Shi, Li, et al., 2024). L. plantarum HN3 was isolated from an outdoor raceway pond cultivating A. pyrenoidosa in Hainan, China. Isolation was performed by plating samples on MRS agar (Huankai, Guangzhou, China). The strain was identified by 16S rRNA gene sequencing performed by GenCefe Biotech Co., Ltd. (Jiangsu, China) (Table S2). Whole-genome sequencing was performed by Sangon Biotech Co., Ltd. (Shanghai, China), and the safety of the strain had been evaluated in previous work. It was preserved at −80 °C in 25% (v/v) glycerol and deposited in the Guangdong Microbial Culture Collection Center (GDMCC, Guangzhou, China) under accession number GDMCC 64816. A single colony of L. plantarum HN3 was inoculated into MRS broth and statically activated for two successive passages at 37 °C. The activated culture was then reinoculated into fresh MRS broth with an inoculum size of 1% (v/v) and statically incubated at 37 °C for 16 h. Following incubation, the culture was harvested by centrifugation at 10,000 ×g for 10 min at 4 °C. The cell pellet was washed twice with sterile physiological saline (0.85% NaCl), and resuspended in sterile saline to achieve to a final cell density of 1 × 10⁸ CFU/mL for use as the inoculum.

2.2. Preparation of LAB-fermentation suspension

Based on the optimal conditions determined in our previous studies, 60 g of yellow powder of CX41 mutant and 60 g of sucrose were accurately weighed and mixed with 880 g of ddH₂O. The mixture was subjected to high-pressure homogenization (UH-06; Union-biotech Co., Ltd. Shanghai, China) at 85 Mpa for 90 s at 20 °C. The resulting cell-disrupted algal slurry (referred to as the yellow suspension) was pasteurized at 90 °C for 10 min. After cooling to room temperature, an inoculum containing 1 × 10⁸ CFU/mL of L. plantarum HN3 was inoculated into the slurry at a 1% (v/v) ratio, followed by static fermentation at 37 °C for 10 h. Samples of the yellow suspension were collected before fermentation (0 h) and after 10 h of fermentation for subsequent analyses.

2.3. Sensory evaluation of the yellow suspension

Sensory evaluation of the yellow suspension samples was conducted with minor modifications in accordance with the Chinese Dairy Industry Standard RHB104–2020 for fermented milk (Li et al., 2025). The sensory panel comprised 10 trained panelists (five males and five females, aged 19–28 years). Evaluations were performed in a sensory laboratory maintained at a temperature of 23 ± 2 °C, adhering to the established criteria (Table S3). To minimize bias, the presentation order of the yellow suspension samples was randomized prior to evaluation. Statistical significance was assessed using a t-test, thereby ensuring the reliability and interpretability of the results. All participants voluntarily signed informed consent forms, confirming their agreement to participate in this sensory study and authorizing the use of their anonymized data for academic publication. The study was conducted in compliance with protocols to protect participants' rights and privacy throughout the execution process. Details regarding the ethical considerations and compliance of this study were provided in the “Ethical statement” section at the end of the manuscript.

2.4. Analytic methods

2.4.1. Cell density of L. plantarum HN3 during fermentation

Cell density of L. plantarum HN3 was determined using the plate counting method. Briefly, samples were serially diluted tenfold with sterile physiological saline (0.85% NaCl). Appropriate dilutions were spread onto MRS agar plates and incubated at 37 °C for 48 h. Colonies were counted on plates containing 30 to 300 colonies, and results were expressed as colony-forming units per milliliter (CFU/mL) (Lonvaud-Funel et al., 1991).

2.4.2. Biochemical composition of the yellow suspension

Crude protein concentration was determined using the Kjeldahl method (FOSS, Denmark) (Hayes, 2020). Free amino acids were analyzed by high-performance liquid chromatography (HPLC, Model 1260, Agilent Technologies Inc., USA) (Kowalska et al., 2022). Briefly, the yellow suspension collected at 0 h and 10 h was ultrasonicated for 30 min in a low-temperature water bath and thoroughly mixed. An aliquot of 1 mL of each sample was transferred into a 10 mL volumetric flask and brought to volume with 20% (v/v) ethanol containing 1 mmol/L HCl. Subsequently, 1 mL of the solution was filtered through a 0.22 μm membrane filter, transferred to an autosampler vial, and subjected to HPLC analysis. Chromatographic separation was performed using a HILIC-Z column (3.0 × 100 mm, 2.7 μm, Agilent Technologies Inc., USA). The mobile phase consisted of solvent A (75% acetonitrile in ddH₂O) and solvent B (0.1 mol/L sodium acetate). The flow rate was set at 0.3 mL/min, the column temperature was maintained at 35 °C, and the injection volume was 1 μL. The gradient elution program was as follows: 0 min, 10% A; 0–5 min, 10–50% A; 5–8 min, 50% A; 8–8.1 min, 10% A; 8.1–12 min, 10% A.

Hydrolyzed amino acids profile was analyzed using an automatic amino acid analyzer (L-8900, HITACHI, Japan) (Liu et al., 2024). Briefly, 20 mg of freeze-dried powder from the yellow suspensions was hydrolyzed with 5 mL of 6 mol/L HCl at 110 °C for 24 h. After cooling, the hydrolysate was filtered through a 0.22 μm membrane and diluted to 25 mL. For protein precipitation, 2 mL of the solution was mixed with 1 mL of 0.02 M HCl, filtered through a 0.22 μm membrane, and a 20 μL aliquot was analyzed using an amino acid analyzer. Tryptophan was determined separately following alkaline hydrolysis as follows. Briefly, 20 mg of sample was hydrolyzed with 1.5 mL of 4 M LiOH at 110 °C for 22–24 h. The hydrolysate was cooled, filtered, adjusted to 10 mL with 0.02 M HCl, passed through a 0.22 μm PES membrane, and analyzed using a 20 μL injection volume. The amino acid score (AAS) was calculated according to Eq. (1).

$$\mathrm{AAS}=\frac{\mathrm{aan}}{\mathrm{AAn}}\times 100$$ (1)

where, AAs represents the content of a specific essential amino acid (EAA) in the sample, and AAn represents the corresponding EAA content as outlined in the FAO/WHO reference pattern (Liu et al., 2024).

Total acidity was analyzed using the acid-base titration method (Tyl & Sadler, 2017). The contents of organic acids were determined by HPLC (Kim et al., 2025). The yellow suspensions collected at 0 h and 10 h were subjected to ultrasonication for 30 min in a low-temperature water bath and then thoroughly homogenized. An aliquot of 0.5 mL from each sample was diluted tenfold with a 0.1% phosphoric acid solution, and 1 mL of the diluted solution was filtered through a 0.22 μm membrane filter prior to analysis. Chromatographic separation was carried out using a C18 column (2.1 × 100 mm, 1.7 μm, Agilent Technologies Inc., USA). The mobile phase consisted of solvent A (ultrapure water containing 0.1% formic acid) and solvent B (100% methanol). The flow rate was set at 0.3 mL/min, the column temperature was maintained at 30 °C, and the injection volume was 1 μL. The gradient elution program was as follows: 0 min, 80% A; 0–1 min, 80% A; 1–3 min, 50% A; 3–10.5 min, 20% A; 10.5–10.6 min, 80% A; 10.6–13.5 min, 80% A.

The soluble sugars (glucose, fructose, sucrose, and rhamnose) were quantified in the complete fermentation system, consisting of a yellow CX41 mutant A. pyrenoidosa suspension supplemented with sucrose as an external carbon source and inoculated with L. plantarum. Samples were collected at 0 h (immediately after inoculation, serving as the baseline) and after 10 h of fermentation. Soluble sugars were determined as follows (Weiß & Alt, 2017): Aliquots (2 mL) of the yellow suspensions collected at 0 h and 10 h were subjected to ultrasonication for 5 min, and an appropriate volume was filtered through a 0.22 μm membrane filter prior to analysis. Chromatographic analysis was conducted using a Waters XBridge Amide column (4.6 × 150 mm, 5 μm, Agilent Technologies Inc., USA). The mobile phase consisted of solvent A (75% acetonitrile in water containing 0.2% triethylamine) and solvent B (80% acetonitrile in water containing 0.2% triethylamine). The gradient elution program was set as follows: 0 min, 10% A; 0–16 min, 70% A; 16–16.1 min, 10% A; 16.1–25 min, 10% A.

The fatty acid profile was determined using a modified method as previously described (Liu et al., 2024). Briefly, total fatty acids were extracted from the samples, converted to fatty acid methyl esters (FAMEs), and subsequently analyzed by gas chromatography mass spectrometry (GC–MS) equipped with a DB-23 capillary column (30 m × 250 μm, 0.25 μm, Agilent Technologies Inc., USA). Fatty acids were identified by comparison with an NIST mass spectral database and quantified using the internal standard curve with nonadecanoic acid (C19:0, 1 mg/mL) as the internal standard.

Lutein concentration in the yellow suspensions was quantified by HPLC (Liu et al., 2024). Extraction was performed by adding methanol-acetone (1,1, v/v) solution with 0.1% BHT to 1 mL of samples, followed by grinding with glass beads three times. The combined extracts were evaporated under a flow of nitrogen gas, reconstituted in 1 mL of methanol-tert-butyl methyl ether (1,1, v/v) containing 0.1% BHT, and filtered through a 0.22 μm membrane. Chromatographic separation was performed using methanol (solvent A) and tert-butyl methyl ether (solvent B) as the mobile phases. The gradient elution program was set as follows: 0 min, 90% A; 0–2 min, 80% A; 2–6 min, 60% A; 6–9 min, 50% A; 9–12 min, 10% A. The flow rate was maintained at 0.8 mL/min, the column temperature was set at 25 °C, and detection was carried out at a wavelength of 450 nm.

GABA concentration in the yellow suspensions was quantified by HPLC (Pencheva et al., 2022). Samples (5 mL) at 0 h and 10 h were ultrasonicated for 10 min, centrifuged at 8000 rpm for 5 min, and the extraction by ddH₂O was repeated twice. The combined supernatant was adjusted to 10 mL and stored at 4 °C. For derivatization, 1 mL of extract was mixed with triethylamine-acetonitrile (7,43, v/v) and 0.5 mL of 0.15% phenyl isothiocyanate in acetonitrile, reacted in the dark, and then extracted with n-hexane. The lower layer was filtered through a 0.22 μm membrane for HPLC analysis. The mobile phase consisted of solvent A (methanol) and solvent B (0.1 mol/L sodium acetate, pH 6.5, containing 0.5% triethylamine). The gradient program was as follows: 0 min, 40% A; 0–15 min, 60% A; 15–16 min, 100% A; 16–20 min, 100% A; 20–21 min, 40% A; 21–25 min, 40% A. The flow rate was 0.7 mL/min, and the detection was performed at 254 nm.

2.4.3. Antioxidant activity of the yellow suspension

The DPPH (2, 2-Diphenyl-1-picrylhydrazyl) radical scavenging activity of the yellow suspensions was assessed using a modified method (Jiang et al., 2021). Briefly, samples collected at 0 h and 10 h were centrifuged at 8000 ×g for 10 min at 4 °C, and the supernatants were subsequently collected. An aliquot of 1 mL of the supernatant was mixed with 1 mL of 0.2 mmol/L DPPH in ethanol and incubated in the dark for 30 min. The absorbance was measured at 517 nm (A_i), and the DPPH radical scavenging activity was calculated according to Eq. (2).

$$\text{DPPH radical scavenging activity}\left(\%\right)=\left(1-\frac{{\mathrm{A}}_{\mathrm{i}}-{\mathrm{A}}_{\mathrm{j}}}{{\mathrm{A}}_0}\right)\times 100\%$$ (2)

Herein, A_i represents the absorbance of the sample mixed with DPPH solution, A₀ represents the absorbance of DPPH solution mixed with ethanol, and A_j represents the absorbance of the sample mixed with ethanol.

The hydroxyl (•OH) radical scavenging activity was measured using a commercial kit (BC1325-100 T/96S, Solarbio, Beijing, China). The supernatants were sequentially mixed with reagents 1 to 4 and distilled water, incubated at 37 °C for 60 min. The absorbance was read at 536 nm. The control and blank reactions were prepared according to the kit's instructions. The scavenging rate was calculated according to Eq. 3.

$$\text{Hydroxyl radical scavenging activity}\left(\%\right)=\left(\frac{{\mathrm{A}}_{\mathrm{i}}-{\mathrm{A}}_{\mathrm{c}}}{{\mathrm{A}}_0-{\mathrm{A}}_{\mathrm{c}}}\right)\times 100\%$$ (3)

Herein, A₀ corresponds to the absorbance of the blank tube, A_c corresponds to the absorbance of the control tube, and A_i corresponds to the absorbance of the sample tube.

2.4.4. Flavor and taste attributes

An electronic nose (E-nose; PENS, AIRSENSE, Germany) and an electronic tongue (E-tongue; TS-5000Z, INSENT, Japan) were used to evaluate the flavor and taste attributes of the yellow suspensions. The E-nose analysis was performed with slight modifications to the method described (Qiu et al., 2021). Freeze-dried powder (5 g) of the yellow suspensions was sealed in 100 mL beakers and equilibrated for 2 h prior to analysis. The E-nose system consisted of an array of 10 metal oxide semiconductor sensors (Table S4). Headspace sampling was conducted at a flow rate of 400 mL/min with a sampling interval of 1 s and an analysis time of 80 s.

E-tongue analysis was conducted with minor modifications to the method reported (Woertz et al., 2011). The yellow suspensions collected at 0 h and 10 h were centrifuged at 3000 rpm for 10 min, filtered, and the supernatants were subjected to analysis. The E-tongue sensor array included sensors specific to six taste modalities: umami (AAE), astringency (AE1), bitterness (C00), sourness (CA0), saltiness (CT0), and sweetness (GL1). The system was calibrated using an artificial saliva solution consisting of 30 mM potassium chloride and 0.3 mM tartaric acid.

2.4.5. Volatile metabolites profiling

The yellow suspensions (1 mL) were cryogenically ground in liquid nitrogen, vortexed, and 0.2 mL was transferred into headspace vials with 20 μL of 3-hexanone (10 μg/mL) as internal standard and 0.2 g of NaCl. Volatile organic compounds (VOCs) were extracted by headspace solid-phase microextraction (HS-SPME, Agilent Technologies Inc., USA) at 40 °C for 5 min with a DVB/CWR/PDMS fiber (120 μm, Agilent Technologies Inc., USA) at 40 °C for 15 min. After sampling, VOCs were thermally desorbed from the fiber in the injection port of a GC apparatus (Model 8890, Agilent Technologies Inc., USA) at 250 °C for 5 min in splitless mode. Separation and analysis were performed by GC–MS (7000E, Agilent Technologies Inc., USA) equipped with a capillary column (DB-5MS, 30 m × 0.25 mm, 0.25 μm, Agilent Technologies Inc., USA). The injector was set at 250 °C with splitless injection; oven program: 40 °C for 3.5 min, 100 °C at 10 °C/min, 180 °C at 7 °C/min, 280 °C at 25 °C/min (hold 5 min). Mass spectra were acquired in electron impact (EI) mode at 70 eV, with the quadrupole, ion source, and transfer line at 150, 230, and 280 °C respectively, using SIM mode. Compounds were identified by matching retention times, fragmentation patterns, and m/z values with the metabolites database (MetWare Biotech. Co. Ltd., Wuhan, Hubei, China). The key odor compounds in the samples were evaluated using the relative odor activity value (ROAV) method, which quantifies the contribution of each volatile component to the overall aroma. ROAV was calculated using the following formula:

$${\text{ROAV}}_{\mathrm{i}}=\frac{{\mathrm{C}}_{\mathrm{i}}}{{\mathrm{C}}_{\text{stan}}}\times \frac{{\mathrm{T}}_{\text{stan}}}{{\mathrm{T}}_{\mathrm{i}}}\times 100\%$$ (4)

where, C_i is the relative content (%) of volatile component i; T_i is the odor threshold (μg/mL) of component i; C_stan is the relative content (%) of the component contributing most to the overall aroma; and T_stan is the corresponding odor threshold (μg/mL).

2.4.6. Non-volatile metabolites profiling

Non-volatile organic compounds (NVOCs) were extracted from 100 μL of the yellow suspensions at 0 h and 10 h using 1 mL of extraction solution (H₂O: acetonitrile: isopropanol, 1:1:1, v/v/v) in Eppendorf tubes. The mixtures were ultrasonicated for 30 min at 4 °C and then centrifuged at 12,000 rpm for 20 min at 4 °C. The supernatants were transferred to clean microtubes, vacuum-dried, and reconstituted in 200 μL of 30% acetonitrile (v/v) before being transferred to insert-equipped vials for analysis. To assess instrument stability and reproducibility, quality control (QC) samples were prepared by pooling 10 μL of each sample and analyzed alongside the other samples, with a QC sample measured after every 10 injections. NVOCs were analyzed using ultra-performance liquid chromatography (UPLC, Vanquish, ThermoFisher, USA) coupled with high-resolution mass spectrometry (HRMS, Orbitrap Exploris 120, Thermo Fisher, USA). Chromatographic separation was performed on a column (Waters BEH Amide, 2.1 × 50 mm, 1.7 μm, Waters, USA) at 40 °C with a flow rate of 0.3 mL/min and an injection volume of 2 μL. The mobile phases consisted of 0.1% acetic acid in water (A) and 0.1% acetic acid in acetonitrile (B), with the following gradient: 0–1 min, 0% B; 1–9 min, 0–95% B; 9–13 min, 95% B; 13–13.1 min, 95–0% B; 13.1–17 min, 0% B. Electrospray ionization (ESI) was performed in both positive and negative modes with a spray voltage of 3.0 kV / –2.8 kV, sheath gas at 40 arb, auxiliary gas at 10 arb, capillary temperature at 320 °C, and auxiliary gas heater at 350 °C. All analyses were conducted in triplicate. NVOC data were processed using ProteoWizard (v 3.0, ProteoWizard Software Foundation, Palo Alto, CA, USA) and analyzed for comprehensive identification with R packages based on mzXML and BiotreeDB software (v3.0, Shanghai Biotree Biomedical Technology Co., Ltd., China).

2.5. Statistical analysis

The t-test was used to assess the normality of sensory attributes, with significance set at p < 0.05. Normality and homoscedasticity were verified using the Shapiro-Wilk and Levene tests respectively, in IBM SPSS Statistics 27.0 (IBM Corp., Armonk, NY, USA). Principal component analysis (PCA) was performed on unit variance–scaled data to explore differences in VOCs and NVOCs. Orthogonal projections to latent structures discriminant analysis (OPLS-DA) was conducted on log₂-transformed, mean-centered data to identify differential metabolites between samples (Boccard & Rutledge, 2013). Model robustness was evaluated using 10-fold cross-validation and 200-permutation tests. Variable importance in projection (VIP) scores was used to rank the contributions of each variable, with VIP > 1.0 and fold change (FC) ≥ 2 or ≤ 0.5 considered as differential metabolites. P values were adjusted using the Benjamini-Hochberg procedure to control the false discovery rate (FDR), and variables with FDR < 0.05 and p < 0.05 were defined as significantly differential metabolites (SDM).

3. Results and discussion

3.1. Changes of basic properties in the yellow suspension

The changes in basic properties of the yellow suspension before and after fermentation by L. plantarum HN3 are shown in Fig. 1a-d. During fermentation within 10 h, the pH value decreased from 6.02 to 4.05 (Fig. 1a), which can be attributed to the continuous production of total organic acids including lactic acid by L. plantarum NH3. LAB produce lactic acid during fermentation to regenerate NAD⁺ from NADH, thereby allowing glycolysis to continue and generate ATP under anaerobic or low-oxygen conditions (Akpoghelie et al., 2025). The accumulation of lactic acid not only creates a favorable environment for LAB growth but also exerts a strong inhibitory effect on undesirable microorganisms, since the low pH disrupts cell membranes and inhibits the growth of pathogens like Listeria monocytogenes, Salmonella spp., and E. coli (Stabnikov et al., 2025), thereby ensuring the microbiological safety of the yellow suspension.

Fig. 1.

Comparative analysis of the yellow suspensions of A. pyrenoidosa mutant CX41 before and after fermentation for 10 h by L. plantarum HN3. (a) pH; (b) cell density of L. plantarum HN3; (c) total acidity; concentrations of (d) protein, (e) γ-aminobutyric acid (GABA), (f) lutein; scavenging rates of (g) 2,2-Diphenyl-1-picrylhydrazyl radical (DPPH) and (h) hydroxyl radical (OH). *, **, ***, and **** represent significant difference (p < 0.05) to highly significant difference (p < 0.01; p < 0.001; p < 0.0001), respectively; ns represents no significant difference (p > 0.05). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

After fermentation for 10 h, the cell density of L. plantarum NH3 in the yellow suspension increased significantly (p < 0.05), reaching 4.05 × 10⁸ CFU/mL (Fig. 1b), indicating that the cell-disrupted algal slurry serves as a highly suitable growth medium for L. plantarum NH3 by supplying sufficient nutrients. The CX41 mutant contains high protein over 45% dry weight, along with all essential amino acids, carbohydrates, fatty acids, minerals, pigments, carotenoids and vitamins (Liu et al., 2024). This finding aligns with previous reports indicating that Spirulina (Arthrospira) platensis powder is an ideal fermentation substrate for LAB. When Spirulina powder was mixed with sterilized water at a ratio of 1:20 (w/w) and fermented with Lacticaseibacillus paracasei No. 244 under anaerobic conditions at 30 °C for 24 h, the viable cell density reached as high as 8.55 log₁₀ CFU/g (Tolpeznikaite, Bartkevics, Skrastina, Pavlenko, Mockus, et al., 2023). Correspondingly, the total titratable acidity increased significantly (p < 0.05), from 0.74 g/L to 3.31 g/L after 10 h (Fig. 1c), whereas the protein concentration showed no significant change (Fig. 1d). This may be attributed to the fact that L. plantarum NH3 mainly perform limited proteolysis, hydrolyzing proteins into smaller peptides and free amino acids to support its metabolic activity and enhance digestibility, flavor, and functional properties, rather than extensively degrading proteins or utilizing them as a primary energy source (Wen et al., 2026).

3.2. Changes of functional components in the yellow suspension

GABA is a bioactive compound associated with various health benefits, including anxiety relief and sleep promotion (Hou et al., 2024). However, the GABA content in most natural foods is generally low. LAB fermentation has been widely recognized as an effective strategy for raising GABA levels in foods, with significant increases reported in products fermented by L. plantarum (Icer et al., 2023). As shown in Fig. 1e, the GABA concentration in the yellow suspension increased significantly from 40.00 mg/L (0 h) to 54.00 mg/L (10h) (p < 0.05). This increase can primarily be attributed to the high amino acid content in the CX41 mutant biomass, particularly its abundance of glutamic acid, which serves as a key precursor for GABA biosynthesis. In LAB, the conversion of glutamic acid (glutamate) to GABA is catalyzed by the glutamate decarboxylase (GAD) system, which functions as an acid resistance mechanism. During this process, intracellular protons are consumed, leading to an increase in cytosolic pH and enhanced bacterial survival under acidic conditions. The synthesized GABA is subsequently exported from the cell via a glutamate/GABA antiporter (Iorizzo et al., 2023).

Lutein is a bioactive carotenoid with multiple physiological functions, including antioxidant activity and visual protection, and is widely distributed in plants and microalgae (Mitra et al., 2021). As shown in Fig. 1f, no significant difference in lutein concentration (60.93 mg/L) was observed in the yellow suspension before and after fermentation by L. plantarum HN3. This result indicates that the lactic acid fermentation process exerts minimal impact on lutein, and its stability allows for the retention of its health-promoting properties in these fermented foods. In addition, the antioxidant activity of the yellow suspension was markedly enhanced by the fermentation of L. plantarum HN3. As shown in Fig. 1g & h, the radical scavenging activity of DPPH and hydroxyl in the fermented yellow suspension after 10 h were significantly increased by 12.29% and 44.91% respectively, compared to the levels at 0 h (p < 0.05). This enhancement may be attributed to increased levels of phenolic compounds, vitamins, and other bioactive substances in fermented plant-based products (Gholamhosseinpour et al., 2024). In principle, compared to unfermented products, the organic acids produced by L. plantarum fermentation create an environment that protects antioxidant compounds and enhances their stability (Yang et al., 2024). These findings are consistent with previous reports that the fermentation of Arthrospira platensis substrates using mixed cultures of LAB, including Lactobacillus acidophilus, Bifidobacterium bifidum, Lactobacillus casei, Bifidobacterium infantis, Bifidobacterium longum, and Lactococcus lactis, significantly enhances antioxidant capacity (Liu et al., 2011). These results indicate that L. plantarum HN3 fermentation does not compromise health-promoting components in the substrate, such as lutein, but instead confers additional functional compounds, thereby enhancing the overall functional value of the fermented yellow suspension.

3.3. Changes of biochemical composition in the yellow suspension

The changes of amino acid concentrations in the yellow suspension are shown in Fig. 2a and b. After 10 h of fermentation, the concentrations of both total hydrolyzed amino acids and total free amino acids decreased in the fermented yellow suspension. This finding aligns with previous studies indicating that after 24 h of Spirulina platensis fermentation by four LAB strains, namely L. plantarum B-531, L. rhamnosus B-442, L. helveticus B-4526, and L. casei B-1922, the contents of both total amino acids and total free amino acids were reduced compared to their concentrations before fermentation (Kurt et al., 2023). Specifically, after 10 h of fermentation, the concentrations of methionine and proline among the hydrolyzed amino acids decreased significantly by 69.00% and 14.10%, respectively, while no significant changes were observed for the other amino acids (Fig. 2a). The amino acid score at 10 h (85.21) showed no significant difference compared with that at 0 h (83.65). According to the FAO/WHO recommended nutritional criteria, the essential-to-total amino acid (E/T) ratios exceeds 0.40 (Liu et al., 2024), both E/T ratios at 0 h (0.462) and 10 h (0.456) in the yellow suspension met this criterion. This phenomenon may be attributed to the selective catabolism and utilization of amino acids by L. plantarum HN3. Methionine is often metabolized into volatile compounds or utilized through specific metabolic pathways, resulting in a substantial decrease in its overall pool. In contrast, proline can be oxidized via proline dehydrogenase, a reaction that contributes to NAD⁺ regeneration under anaerobic or acidic conditions, thereby supporting cellular redox balance during fermentation (Nicolescu et al., 2023).

Fig. 2.

Profiles of hydrolyzed amino acids (a), free amino acids (b), organic acids (c), soluble sugars (d), and fatty acids (e) in the yellow suspensions of A. pyrenoidosa mutant CX41 before and after fermentation for 10 h by L. plantarum HN3. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Free amino acids primarily influence flavor characteristics, with the concentration of arginine, glycine, and proline increasing significantly after fermentation by 43.09%, 48.98%, and 10.88%, respectively (Fig. 2b). Glycine and proline are classified as sweet-tasting amino acids, whereas arginine is categorized as a bitter-tasting amino acid (Hanagasaki & Asato, 2018). This increase may be attributed to protein hydrolysis induced by L. plantarum HN3 fermentation. LAB are unable to synthesize all the essential amino acids required for their growth and metabolism. However, they secrete proteases and peptidases that hydrolyze proteins into free amino acids and peptides, thus providing the necessary nutrients for their proliferation (Bai et al., 2025).

The fermentation of L. plantarum HN3 resulted in the production of various organic acids, as shown in Fig. 2c. After 10 h of fermentation, the concentrations of lactic acid, succinic acid, acetic acid, malic acid, and fumaric acid increased significantly. Notably, lactic acid reached 2355.83 mg/L, representing a 43.86-fold increase compared with that at 0 h. L. plantarum is categorized as a homofermentative LAB, primarily converting glucose and other hexoses into lactic acid through the glycolytic pathway, with a typical yield exceeding 85% (Chen et al., 2023). Acetic acid is predominantly produced via heterofermentative or mixed-acid pathways, which involve the cleavage of pyruvate to acetyl-CoA, followed by its conversion to acetate, accompanied by ATP generation. In contrast, succinic acid is generated as a minor byproduct through a reductive branch of the tricarboxylic acid cycle or citrate metabolism involving oxaloacetate, malate, and fumarate. Malic acid and fumaric acid primarily function as transient intermediates in these metabolic pathways and are occasionally consumed through malolactic fermentation or incompletely reduced to succinate rather than accumulating as major end products (Tsuge et al., 2016). Correspondingly, a decrease in soluble sugars was detected in the fermented yellow suspension at 10 h, as shown in Fig. 2d. The concentrations of rhamnose, glucose, fructose, and sucrose decreased significantly by 100.00%, 63.53%, 51.98%, and 11.61% respectively, indicating that glucose, fructose, and sucrose were efficiently utilized by L. plantarum HN3. Glucose and fructose, as monosaccharides, directly enter the glycolytic pathway, while sucrose, a disaccharide composed of glucose and fructose, is first hydrolyzed by enzymes such as sucrase or β-fructofuranosidase and subsequently transported into the cell via phosphotransferase systems (PTS) (Dai et al., 2023). In contrast, rhamnose utilization by LAB is strain variable and frequently negative in many strains. However, it has been reported that certain kimchi-derived isolates, such as L. plantarum ATG-K6 and ATG—K8, possess functional rhamnose utilization gene clusters and are therefore able to metabolize rhamnose efficiently (Cui et al., 2021).

The changes of fatty acid composition in the yellow suspension are shown in Fig. 2e. Compared with those at 0 h, the concentration of total fatty acids (TFA), total unsaturated fatty acids (TUSFA), and total saturated fatty acids (TSFA) increased significantly by 0.54-fold, 0.53-fold, and 0.53-fold after fermentation for 10 h, respectively (p < 0.05). The increase in total saturated fatty acids amounted to 495.41 mg/L, primarily due to marked increases in palmitic acid (C16:0) and stearic acid (C18:0), which rose by 381.42 mg/L and 113.99 mg/L, respectively. With respect to unsaturated fatty acids, the concentration of palmitoleic acid (C16:1) and hexadecadienoic acid (C16:2) increased by 39.88 mg/L and 91.86 mg/L, respectively. In addition, oleic acid (C18:1), linoleic acid (C18:2), and α-linolenic acid (C18:3) were observed to significant increase by 105.19 mg/L, 409.22 mg/L, and 206.33 mg/L, respectively. Overall, the TUSFA concentration increased by 0.53-fold, corresponding to an increase of 707.39 mg/L. These results are consistent with previous reports that fermentation by L. plantarum leads to increased levels of linoleic acid and α-linolenic acid in walnut milk (Fiorino et al., 2023). This increase is primarily attributed to the lipid biosynthesis activity of L. plantarum, which assimilate substrate-derived lipids or metabolic intermediates for the synthesis of membrane lipids required for cell growth and metabolic activity (Yang et al., 2024). Notably, linoleic acid and α-linolenic acid are essential fatty acids that cannot be synthesized endogenously and must be acquired through dietary sources (H. Wang et al., 2022). Therefore, L. plantarum HN3 significantly enhances these polyunsaturated fatty acids in the fermented yellow suspension, thereby substantially improving its nutritional value.

3.4. Flavor attributes of the yellow suspension

3.4.1. Evaluation by sensory panel and electronic nose analysis

As shown in Fig. 3a, the radar plot illustrates the sensory evaluation of yellow suspensions before and after fermentation (0 h and 10 h). Overall, fermentation led to noticeable improvements in aroma and taste scores. The fermented sample (10 h) exhibited a substantially higher aroma score compared with the unfermented sample, indicating a more pronounced and favorable fermented aroma. Similarly, taste scores increased markedly after fermentation, suggesting an improved flavor profile with enhanced palatability. In contrast, colour scores showed only a slight increase, indicating that fermentation had a limited effect on visual appearance. Appearance scores remained relatively stable, with only minor differences between the two samples. Collectively, these results demonstrate that fermentation primarily enhanced aroma and taste attributes, while exerting minimal influence on colour and appearance (Fig. 3b). PCA was performed on the response values of the E-nose sensors, with results depicted in Fig. 3c. The first two principal components (PC1 and PC2) accounted for over 77% of the total variance, indicating that they captured most of the variability and key features in the yellow suspensions. The distinct separation in the score plot further demonstrates clear differences in odor profiles at 0 h and 10 h, particularly evident in the responses of the W1W, W2W, W5S, W2S, and W1S sensors (Fig. 3d). The observed odor differences were mainly associated with sulfur-containing compounds, alcohols, aldehydes, ketones, nitrogen oxides, and a range of volatile compounds (Zhang et al., 2024).

Fig. 3.

The appearance of the yellow suspension of A. pyrenoidosa mutant CX41 before and after fermentation for 10 h by L. plantarum HN3 (a), radar chart of sensory evaluation by sensory panel (b); principal component analysis (PCA) of electronic nose detection results (c), and loading plot with sensor contribution rates (d).

3.4.2. Volatile metabolome profiling

In this study, a total of 726 VOCs were detected in the yellow suspension at 0 h, whereas 746 VOCs were identified after 10 h of fermentation with L. plantarum HN3. Compared to the sample at 0 h, the numbers of alcohols, amines, aromatic hydrocarbons, esters, ethers, heterocyclic compounds, ketones, and phenolic compounds increased after 10 h of fermentation. Specifically, the number of alcohols rose from 58 to 63, esters from 128 to 133, and heterocyclic compounds from 124 to 130, suggesting that L. plantarum HN3 may promote the formation of these compounds during fermentation (Fig. 4a). These results are consistent with previous reports that L. plantarum CCMA0743, when co-cultured to ferment a plant substrate composed of 75 g oats, 175 g sunflower seeds, and 75 g almonds for 24 h, led to increases in various volatile compounds, including acids, amides, alkenes, ketones, phenols, ethers, and others (Ferreira et al., 2022). In contrast, the numbers of sulfur-containing compounds and terpenes each decreased by one, whereas the numbers of amines, aromatic hydrocarbons, ethers, and phenolic compounds each increased by one (Fig. 4a). No changes were observed in the numbers of acids, aldehydes, halogenated hydrocarbons, hydrocarbons, ketones, nitrogen-containing compounds, or any compounds classified as others. In addition, significant differences were observed in the total peak areas of VOCs in the yellow suspensions at 0 h and 10 h of fermentation. Compared to the 0 h measurement, the total peak areas of alcohols, amines, aromatic hydrocarbons, hydrocarbons, and phenolic compounds increased at 10 h, whereas the peak areas of esters, ketones, nitrogen-containing compounds, sulfur-containing compounds, and terpenes decreased (Fig. 4b).

Fig. 4.

Number of volatile organic compounds (a), total peak area (b), OPLS-DA score plot (c), and heatmap cluster (d) analysis of volatiles in the yellow suspensions of A. pyrenoidosa mutant CX41 before and after fermentation for 10 h by L. plantarum HN3. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

To verify the differences in VOCs profile between both samples at 0 h and 10 h of fermentation, an OPLS-DA model was established (R²Y = 0.991, and Q² = 0.91) (Fig. 4c). Subsequently, 59 volatiles were identified as markers distinguishing the two yellow suspensions based on a VIP >1, FC ≥ 2 or ≤ 0.5 and p < 0.05. The 54 upregulated compounds mainly included terpenoids, aldehydes, alcohols, esters, ketones, phenols, heterocyclic compounds, ethers, aromatics, and amines (Fig. 4d). Among these, aldehydes, alcohols, and esters exhibited a significant increase after fermentation. For example, (E, Z)-2, 4-decadienal contributed fatty notes, while certain esters and alcohols unique to the fermented suspensions, including cis-3-hexenyl hexanoate and 3-hexanol, imparted floral and fruity aromas, thereby enhancing the sensory acceptability of the fermented yellow suspension (Feng et al., 2018; Liu et al., 2023). Conversely, 5 downregulated compounds were primarily sulfur-containing compounds, alcohols, and heterocyclic compounds (Fig. 4d). Notably, prenyl mercaptan, a compound associated with unpleasant odors (Dulsat-Serra et al., 2016), was absent from the fermented yellow suspension.

3.4.3. Key aroma compounds analysis

The ROAV was used to evaluate the contribution of volatile organic compounds to the aroma of yellow suspensions before and after fermentation, with the results summarized in Table 1. Before fermentation, the yellow suspension contained 24 key aroma compounds, including 2 roasted, 5 fruity, 1 sweet, 2 grassy, 1 floral, 2 fermented, 4 off-flavor, and 7 other compounds. After fermentation for 10 h, 26 key aroma compounds were detected, including 2 roasted, 6 fruity, 1 sweet, 3 herbaceous, 2 floral, 2 fermented, 3 off-flavor, and 7 other compounds, indicating slight increases in fruity, herbaceous, and floral notes alongside a reduction in off-flavors. Regarding key aroma compounds, fermentation preserved the original pleasant aromas of the yellow suspension, with 2-methyl-3-furanthiol and 5-methyl-2-furfurylthiol serving as the primary contributors to the roasted aroma (Zhang et al., 2024). Fruity notes were mainly attributed to (Z)-6-nonenal, while sweet notes were contributed by 5-ethyl-3-hydroxy-4-methyl-2(5H)-furanone (Piornos et al., 2020). The mushroom aroma was associated with 1-octen-3-one, resinous/waxy notes were associated with myrcene, and creamy notes were attributed to 3, −5-octadien-2-one. The herbaceous aroma was primarily due to α-muurolene (Wu et al., 2025). After fermentation for 10 h, the yellow suspension acquired two fatty aroma compounds, 2, 4-decadienal and (2E, 4Z)-2, 4-decadienal, whose ROAV values increased by 4.38-fold and 6.76-fold, respectively. Additionally, a new herbaceous aroma compound, 1-hexen-3-one, appeared with an ROAV of 1.91, uniquely contributing to the fermented sample (Bauer et al., 2022). Overall, these changes significantly enhanced the flavor contribution of the fermented yellow suspension, while the contribution of off-flavor compounds decreased. 2-Thiophenemethanethiol and isopentenyl mercaptan were identified as the primary off-flavor contributors; after fermentation, the ROAV of 2-thiophenemethanethiol decreased from 6.12 to 5.40, while isopentenyl mercaptan, which had a ROAV of 5.56 in the unfermented sample, was no longer detected (ROAV = 0) after fermentation. Regarding the modification of aroma compounds, L. plantarum HN3 also exerted a significant influence on the fermented suspension, with 1-hexanol contributing floral notes and δ-dodecalactone contributing fruity notes, thereby enriching the overall flavor profile of the fermented yellow suspension (Perpetuini et al., 2024).

Table 1.

Key aroma compounds in yellow suspensions of A. pyrenoidosa mutant CX41 before and after fermentation for 10 h by L. plantarum HN3.

Volatiles	OTV (μg/mL)	ROVA		Odor Descriptors
		0 h	10 h	Primary Classification	Secondary Classification
2-Methyl-3-furanthiol	0.0000048	100.00	100.00	Roasted	meat
(Z)-6-Nonenal	0.00002	30.10	26.68	Fruit	cucumber
2-Thiophenemethanethiol	0.00004	6.12	5.40	Off flavor	sulfury
Prenyl mercaptan	0.000002	5.56	0.00	Off flavor	pungent
5-Ethyl-3-hydroxy-4-methyl-2(5H)-furanone	0.0001	4.90	5.39	Sweet	honey
5-Methyl-2-furfurylthiol	0.00005	3.05	2.94	Roasted	coffee
α-Muurolene	0.0001	2.21	2.29	Herbaceous	woody
1-Octen-3-one	0.00004	1.86	2.71	Others	mushroom
3,5-Octadien-2-one	0.0005	1.35	1.08	Others	creamy
Myrcene	0.000015	1.11	1.09	Others	Spice
2,2,6-Trimethylcyclohexanone	0.0001	0.66	0.72	Off flavor	pungent
3-Octen-2-one	0.00003	0.65	0.63	Others	Nutty
2,4-Decadienal	0.0002	0.58	2.54	Others	fatty
(2E,4Z)-2,4-Decadienal	0.00004	0.55	3.72	Others	fatty
4-isocaranol	0.0005	0.50	0.45	Floral	floral
2-Ethyl-2H-furan-5-one	0.0097	0.49	0.47	Others	spice
Benzenemethanethiol	0.00008	0.40	0.38	Off flavor	oniony
Dodecanenitrile	0.00009	0.39	0.36	Fruit	orange
Safranal	0.003	0.38	0.32	Herbaceous	fresh
2-Ethoxy-3-methylpyrazine	0.0008	0.25	0.23	Fermented	earthy
3-Mercaptohexyl acetate	0.00002	0.19	0.18	Fruit	mango
3-Mercaptohexanol	0.00006	0.16	0.13	Fruit	mango
(6Z)-Nonen-1-ol	0.001	0.13	0.12	Fruit	cucumber
Geosmin	0.000015	0.12	0.16	Fermented	earthy
1-Hexen-3-one	0.00002	0.00	1.91	Herbaceous	Herb
1-Hexanol	0.0056	0.00	0.13	Floral	floral
δ-Dodecalactone	0.000523	0.00	0.11	Fruit	peachy

Note: Odor thresholds were reference to the database (MetWare Biotech. Co. Ltd., Wuhan, Hubei, China) and the compilation of odor thresholds of compounds (Yi et al., 2008).

3.5. Taste attributes of the yellow suspension

3.5.1. Electronic tongue analysis

In this study, an electronic tongue system was employed to evaluate the taste attributes of yellow suspensions before and after fermentation. As shown in Fig. 5a, pronounced differences were observed in sourness, bitterness, and umami between the two samples, with the greatest variation noted in sourness. Moderate differences were also detected in sweetness, while only minor changes were found in Aftertaste-B (bitterness aftertaste), Aftertaste-A (astringency aftertaste), and richness (umami aftertaste). In contrast, the intensities of saltiness and astringency in both samples were below the detection threshold when compared with simulated human saliva (30 mM KCl and 0.3 mM tartaric acid) (Woertz et al., 2011), indicating that neither samples exhibited a salty or astringent taste (Fig. 5b). Compared to the yellow suspension at 0 h, the fermented sample exhibited a pronounced increase in sourness, with the taste intensity increasing significantly (p < 0.001). In contrast, the intensity of sweetness decreased from 2.65 to 1.03, representing a 61.13% reduction, while the intensity of umami decreased from 11.07 to 5.43, corresponding to a 49.95% reduction (Fig. 5b). Notably, the bitterness intensity decreased from 5.40 to below the detection threshold, indicating that no bitter taste was perceived, potentially contributing to the improved palatability of the yellow suspensions (Fuke & Ueda, 1996). With respect to other taste attributes, the fermented yellow suspension showed slight increases in richness, astringent aftertaste, and bitter aftertaste compared to that before fermentation. However, these changes had only a minor influence on the overall taste profile. Specifically, the intensity of richness increased from 0.03 to 0.20, the aftertaste of bitterness remained close to the tasteless point with no significant difference before and after fermentation (p > 0.05), and the aftertaste of astringency increased slightly from 0.22 to 0.35. Overall, electronic tongue analysis demonstrated that L. plantarum HN3 significantly altered the taste of the fermented yellow suspension, leading to an overall improvement in flavor quality characterized by enhanced sourness, reduced sweetness and umami, and the complete elimination of bitterness, consistent with the sensory evaluation results and further confirming the effectiveness of L. plantarum HN3 fermentation in improving the taste of the fermented yellow suspension.

Fig. 5.

Radar chart of taste profile (a) and taste intensity (b) by electronic tongue analysis, relative contents of NVOCs (c), and OPLS-DA score plot (d) in the yellow suspensions of A. pyrenoidosa mutant CX41 before and after fermentation for 10 h by L. plantarum HN3. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

3.5.2. Non-volatile metabolome profiling

As shown in Fig. 5c, organic acid metabolites were markedly enhanced during fermentation. When comparing the sample at 0 h to the fermented yellow suspension, the relative abundance of organic acids and their derivatives exhibited a remarkable increase of 6.75-fold. In contrast, fatty acids and their derivatives as well as sugars and their derivatives decreased by 0.17-fold, suggesting that they were utilized or converted as carbon sources during fermentation. Nucleotides and their derivatives decreased by 0.33-fold, while amino acids and their derivatives decreasing by 0.44-fold, likely due to L. plantarum HN3 biosynthesis and amino acid metabolism, primarily in catabolic utilization for cell growth, energy production, and stress adaptation under anaerobic or acidic conditions during fermentation (Tolpeznikaite, Bartkevics, Skrastina, Pavlenko, Ruzauskas, et al., 2023). In addition, other classes of NVOCs decreased by 0.26-fold.

To reveal the differences in NVOCs before and after fermentation, an OPLS-DA model was established with R²Y = 0.96, and Q² = 0.949 (Fig. 5d). The volcano plot illustrates that a total of 411 NVOCs were identified as significantly differential NVOCs (SDNVOCs) based on the criteria of VIP > 1, FC ≥ 2 or ≤ 0.5, and p < 0.05 (Fig. 6a). Compared to the sample at 0 h, 150 SDNVOCs were significantly upregulated, while 261 SDNVOCs were significantly downregulated in the fermented sample (Fig. 6b). Specifically, amino acids and their derivatives, as well as saccharides, exhibited pronounced downregulation, with 102 and 13 metabolites decreasing respectively, while relatively few metabolites were upregulated, including 21 amino acid related metabolites and 1 saccharide related metabolite. In contrast, organic acids and their derivatives, along with fatty acids and their derivatives, were predominantly upregulated, with 31 and 23 metabolites increasing respectively, whereas fewer metabolites were downregulated, including 19 related to organic acids and 16 related to fatty acid. For nucleotides and their derivatives, 21 metabolites were upregulated and 8 were downregulated. These changes indicate that L. plantarum HN3 transitioned from an initial growth phase characterized by rapid utilization of sugars and proteins to a secondary metabolic phase involving extensive consumption of amino acids and carbohydrates, accompanied by the production and accumulation of organic acids and fatty acid derivatives. This stage represents a critical period for acidification, preservation, and flavor development in L. plantarum HN3 fermented suspension (Chen et al., 2025).

Fig. 6.

Volcano plot of NVOCs (a), number of significantly different NVOCs (b), and heatmap of significantly different taste-related metabolites (c) in the yellow suspensions of A. pyrenoidosa mutant CX41 after fermentation for 10 h by L. plantarum HN3. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Not all NVOCs contribute to the taste profile. To elucidate the metabolic basis underlying taste differences before and after fermentation, 411 SDNVOCs were further identified. As shown in Fig. 6c, in the fermented yellow suspension, the relative abundances of most umami-related compounds decreased significantly (p < 0.05), including inosine, guanosine, Val-Gly, Glu-Leu, Gln-Pro, Gln-Leu, Asp, Arg-Leu, Glu-Glu, Glu-Asp, Val-Val, and Leu-Val, with only Glu-Ala showing a significant increase due to fermentation. LAB can metabolize nucleotides, amino acids, and short peptides into organic acids, esters, and other flavor compounds, thus influencing the overall flavor profile and potentially enhancing nutritional value (Razola-Díaz et al., 2024). The relative abundances of fructose, glucose, serine, and alanine decreased, whereas lactate, phenyllactic acid, hydroxyphenyllactic acid, and indolelactic acid increased. The marked accumulation of these aromatic lactic acid derivatives not only imparted a fresh and bright sour taste but also enhanced the complexity and roundness of acidity, which is a key sensory attribute underlying the desirable flavor of many high-quality fermented foods (Lee et al., 2024). Regarding bitter compounds, poncirin increased, while 5′-methylthioadenosine, resveratrol, tyrosine, phenylalanine, methionine, and Arg-Pro decreased, and the high taste threshold of poncirin limited its contribution to overall bitterness (Mouly et al., 1998). Overall, fermentation by L. plantarum HN3 led to a substantial increase in lactate and its derivatives, which exerted synergistic effects on umami enhancement, bitterness suppression, and mouthfeel improvement, thereby contributing to a more rounded, full-bodied, and layered flavor profile in the fermented yellow suspension.

4. Conclusion

In this study, fermentation with L. plantarum HN3 significantly enhanced the nutritional, functional, and sensory attributes of the yellow suspension of Auxenochlorella pyrenoidosa mutant CX41. Notably, excellent probiotic viability was observed, alongside an increase in the composition of free flavor amino acids, specifically arginine, glycine, and proline, by 43.09%, 49.98%, and 10.88%, respectively. Additionally, concentrations of total fatty acids increased by 54.02%, γ-aminobutyric acid rose by 35.00%, while lutein levels remained stable. The in vitro antioxidant capacity was also improved. The fermentation process eliminated bitterness, reduced off-odors from sulfur-containing volatiles, and enhanced and floral and fruity aromas. Furthermore, intensified balanced sourness and improved flavor harmony contributed to a higher overall sensory acceptance. These findings highlight L. plantarum HN3 fermentation as a promising strategy for developing high-protein, microalgae-based functional foods like fermented milk and postbiotics.

CRediT authorship contribution statement

Yu Li: Writing – review & editing, Writing – original draft, Conceptualization. Bo Deng: Formal analysis, Data curation. Dong Wei: Writing – review & editing, Supervision, Conceptualization. Jucai Xu: Resources, Methodology.

Ethical statement

The sensory evaluation in this study was performed in accordance with the Standard Guide for Protection of Respondents and Informed Consent for Sensory Evaluation Studies (ASTM International, E3314–21). In this study, A. pyrenoidosa and L. plantarum are low-risk food-related species that have been approved or widely recognized for food and fermentation applications under relevant food safety regulations in China. In 2012, A. pyrenoidosa was approved as a novel food resource under the Food Safety Law of the People's Republic of China and the Administrative Measures for Novel Food Ingredients. The sensory evaluation involved healthy adult volunteers and was conducted in accordance with established ethical principles for low-risk sensory research involving commonly consumed food products. As the study did not involve clinical interventions, invasive procedures, vulnerable populations, or the collection of sensitive personal data, formal ethics committee approval was not required according to institutional and national guidelines. Participation in the sensory evaluation was entirely voluntary. All participants were informed of the study purpose, procedures, and their rights, including the right to withdraw at any stage without consequence, and written informed consent was obtained prior to participation. Participant privacy and confidentiality were strictly protected throughout the study.

Declaration of competing interest

The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: Dong Wei reports financial support was provided by Guangzhou Municipal Science and Technology Bureau. If there are other authors, they declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

Supplementary material

Data availability

Data will be made available on request.