The diatom strains P. tricornutum (CCAP 1052/6) and C. muelleri (MMDL50116) were cultivated in 1 L polycarbonate bottles sealed with porous filter membranes (pore size 0.2–0.3 μm). The cultures were maintained in f/2 medium (Guillard 1975) under a light intensity of 90 ± 5 μmol photons m⁻² s⁻¹, at 20°C, and with a 12-h light/12-h dark photoperiod. Multiple experiments with various GABA concentrations supplementation were conducted to determine the optimal cultivation conditions for the different microalgae species, as well as the appropriate concentration of exogenous GABA to be added. To ensure a scientifically sound comparison, we transitioned from external chemical concentration to a biologically normalized dosage (pmol cell⁻¹). When adding exogenous GABA, the content of the added exogenous GABA is calculated based on the cell concentration measured at the time of addition. This approach standardizes the metabolic pressure per unit of biomass, effectively eliminating the confounding effects of disparate cell densities at the onset of the exponential phase.

To ensure biological comparability, observations were synchronized based on equivalent physiological states specifically defined by the onset of the exponential phase instead of fixed chronological hours. Given the significant species-specific differences in growth rates and stress sensitivity, this approach accounts for their distinct developmental windows that absolute time points or uniform concentrations would fail to represent accurately. Based on batch culture data, cell densities of 5 × 10⁵ cells mL⁻¹ for P. tricornutum and 1 × 10⁵ cells mL⁻¹ for C. muelleri were established as the physiological thresholds for the exponential onset (Supplementary Fig. S1).

These specific thresholds were subsequently employed as the initial cell densities for the semi-continuous cultivation. To sustain this growth state, a semi-continuous cultivation strategy was employed with the medium refreshed every 72 h. After three preliminary cycles to ensure physiological stability, the cells were re-inoculated into fresh media at the aforementioned initial densities and supplemented with exogenous GABA at 0, 0.2, 0.5, and 1.0 pmol cell⁻¹ for P. tricornutum, and 0, 1.5, 6.2, and 16.2 pmol cell⁻¹ for C. muelleri. All treatments were conducted in triplicate. Samples harvested at the end of the experiment were subjected to comprehensive analyses, including the determination of biochemical compositions (carbohydrates, proteins, lipids, and fatty acids), assessment of photosynthetic activity, and quantification of antioxidant-related parameters.

For the batch cultivation experiment of C. muelleri, cells in the stable growth phase were inoculated at an initial density of 5 × 10⁴ cells mL⁻¹. On day 8, cultures were supplemented with varying concentrations of exogenous GABA: 0, 1.5, 6.2, and 16.2 pmol cell⁻¹. Each treatment was performed in triplicate, and the samples were harvested for carbohydrate, protein, and lipid determination on day 13.

Microalgal cell concentration and particle size were measured by particle cell counter (Z2TM; Beckman Coulter, Brea, CA, USA) in the middle of the light period. The specific growth rate (μ) was calculated as follows:

, in the formula, N_n and N_n-1 represent the microalgal cell concentration at t_n and t_n-1.

Lipid content was determined using the chloroform-methanol extraction method (Liu et al. 2014). On the final day of cultivation, 600 mL microalgal sample was collected, centrifuged at 5,000 rpm for 10 min, and the centrifuged sample was freeze-dried to produce powder, which was treated with 6 mL of chloroform-methanol solution (v : v = 2 : 1) and 1 mL of 0.88% NaCl solution. After mixing thoroughly at 2,000 rpm for 20 min, centrifuge at 8,960 rpm, 4°C for 10 min, and discard the upper aqueous phase. Add 6 mL of methanol-water solution (v : v = 1 : 1), mix thoroughly at 2,000 rpm for 20 min, and then centrifuge again (8,960 rpm, 4°C, 10 min). Use a Pasteur pipette to collect the lower organic phase and evaporate it under nitrogen until completely dry and reach a constant weight. The lipid content was calculated as follows:

, in the formula, w₁ represents the weight of the tube, mg; w₂ represents the total weight of the tube and the extracted lipids, mg; N represents the cell quantity.

Cells (at least 4 × 10⁷ per sample) were collected by centrifugation (6,000 rpm, 10 min, 4°C) to remove the culture medium. The cell pellets were washed three times with 1× phosphate-buffered saline (PBS) to thoroughly remove residual culture medium. Following the final wash, the cells were resuspended in 1× PBS and lysed via ultrasonication in an ice-water bath (10% power, ultrasound 2 s, interval 1 s, repeated 30 times). The resulting lysate was then centrifuged (12,000 rpm, 4°C, 10 min), and the supernatant was collected as the test sample solution. Soluble protein content in the cells was quantified using a BCA Protein Assay Kit (Beyotime, Shanghai, China) according to the manufacturer’s instructions. The absorbance was measured at 560 nm using a microplate reader (Infinite M200 Pro; Tecan, Männedorf, Switzerland), and the protein concentration of each sample was calculated based on a standard curve.

Carbohydrate content was determined by the phenol-sulfuric acid method. 50 mL of microalgal sample was filtered through a 0.8 μm polycarbonate membrane (Millipore, Watford, UK). The filters were then treated with 5 mL of 0.05 mol L⁻¹ H₂SO₄ solution, and incubated at 60°C for 30 min. After cooling, 2 mL of the extract was transferred to a glass tube, 0.05 mL of 80% phenol and 5 mL of sulfuric acid were added. The mixture was allowed to stand for 2 h to ensure color stabilization. Absorbance was measured at 485 nm using a microplate reader (Infinite M200 Pro; Tecan). Glucose served as the standard for generating a calibration curve, from which the carbohydrate concentrations of the samples were calculated.

Cells were collected at a minimum density of 2 × 10⁶ cells per sample. The cell suspension was centrifuged (6,000 rpm, 10 min, 4°C) to remove the culture medium. The pellet was washed three times with 1× PBS via resuspension and centrifugation under the same conditions to ensure the complete removal of residual medium components. After the final wash, cells were resuspended in 1× PBS. A portion of this suspension was reserved as a blank control and kept on ice under light-protected conditions until analysis. Another aliquot was incubated with 2′,7′-dichlorodihydrofluorescein diacetate (DCFH-DA) at a final concentration of 10 μM. Incubation was carried out in the dark at 20°C for 60 min, with gentle inversion every 10 min to ensure uniform probe exposure. Following incubation, the cells were harvested by centrifugation (6,000 rpm, 10 min, 4°C). To remove any extracellular, non-internalized DCFH-DA, the cell pellet was washed three times with 1× PBS using the same centrifugation parameters using the same centrifugation parameters. Finally, the cells were resuspended in 1× PBS. Both the probe-loaded sample and the blank control were transferred to a black-walled microplate. Fluorescence intensity was measured using a microplate reader with excitation at 488 nm and emission at 525 nm. The relative ROS level was calculated as follows:

, in this formula, F_treatment, F_control, and F_blank denote the fluorescence intensities of the GABA-supplemented samples, the untreated control, and the blank control (without probe), respectively.

Cells were collected at a minimum density of 4 × 10⁷ cells per sample. The cell suspension was centrifuged (6,000 rpm, 10 min, 4°C) to remove the culture medium. The pellet was washed three times by resuspension in 1× PBS followed by centrifugation (6,000 rpm, 10 min, 4°C) to thoroughly eliminate residual medium. After the final wash, the cell pellet was resuspended in 1× PBS. Cell lysis was performed on ice using an ultrasonic disruptor (10% power, ultrasound 2 s, interval 1 s, repeated 30 times). The lysate was subsequently centrifuged (12,000 rpm, 10 min, 4°C), and the resulting supernatant was collected as the sample extract for superoxide dismutase (SOD) activity assay. SOD activity was determined using a commercial SOD assay kit (NBT method; Beyotime) according to the manufacturer’s instructions. An adequate volume of the NBT/Enzyme working solution was prepared for each 160 μL reaction by thoroughly mixing 158 μL of SOD assay buffer, 1 μL of NBT solution, and 1 μL of enzyme solution. The Reaction Initiation working solution was prepared by diluting the Reaction Initiation solution (40×) 40-fold with SOD assay buffer (i.e., 1 μL of 40× solution plus 39 μL of buffer). Three types of blank control systems and the sample reaction system were assembled on ice. The blank controls were configured as follows: Blank 1 (20 μL SOD assay buffer, 160 μL NBT/Enzyme working solution, 20 μL Reaction Initiation working solution); Blank 2 (40 μL SOD assay buffer, 160 μL NBT/Enzyme working solution); Blank 3 (20 μL test sample, 20 μL SOD assay buffer, 160 μL NBT/Enzyme working solution). The sample reaction system consisted of 20 μL test sample, 160 μL NBT/Enzyme working solution, and 20 μL Reaction Initiation working solution. The absorbance of all systems was measured at 562 nm using a microplate reader (Infinite M200 Pro; Tecan). The SOD enzyme activity of the samples was calculated as follows:

, in the formula, m_protein represents the mass of intracellular soluble protein extracted in the sample. A_{Blank 1}, A_{Blank 2}, and A_{Blank 3} denote the absorbance values at 562 nm for Blank 1, Blank 2, and Blank 3, respectively. A sample represents the absorbance at 562 nm for the sample reaction system.

To evaluate real-time photosynthetic performance, the effective photosystem II quantum yield (Fv′/Fm′) was measured on day 3 of semi-continuous cultivation using a multi-color PAM fluorometer (MC-PAM; Heinz Walz GmbH, Walz, Germany). These measurements were conducted simultaneously with the collection of samples for ROS and SOD assays, ensuring a synchronized assessment of photosynthetic status and oxidative stress levels.

Add 2 mL of sulfuric acid-methanol solution (v : v = 2.0%) to the total lipid sample, purge with nitrogen for 20 s, seal the container, mix at 2,000 rpm for 10 min, and heat in a water bath at 80°C for 1 h. Then add 1 mL of saturated sodium chloride solution, 2 mL of hexane and 1 mg methyl heptadecanoate (C17:0, as an internal standard), mix at 2,000 rpm for 10 min, centrifuge at 3,000 ×g for 5 min, extract the supernatant, dry it with nitrogen gas and concentrate the sample with 1 mL of n-hexane, and transfer it through a 0.22 μm organic filter membrane into the injection vial. For external calibration, a 37-component fatty acid methyl ester mix (Supelco; sourced via Orileaf, Shanghai, China) was dissolved in n-hexane to prepare a series of multi-point standard curves through systematic dilution.

Fatty acid content was determined using a gas chromatograph-mass spectrometer Agilent 5977B (Agilent Technologies, Santa Clara, CA, USA). The injection volume was 1 μL, and analysis was performed using an Agilent CP8713 column (30 m × 0.25 mm × 0.25 μm) (Agilent Technologies). Pure helium was used as the carrier gas, with total flow rate of 7.4 mL min⁻¹, purge flow rate of 3 mL min⁻¹, and split ratio of 3 : 1. The temperature gradient was set as follows: the initial temperature was 110°C, which was increased at a rate of 10°C min⁻¹ to 170°C, held for 10 min, then increased at a rate of 3°C min⁻¹ to 205°C, held for 20 min, after that increased at a rate of 5°C min⁻¹ to 235°C, held for 2 min, finally increased at a rate of 10°C min⁻¹ to 250°C, and held for 3 min. The fatty acid content was calculated as follows:

, in the formula, FA_t represents the fatty acid concentration at peak time t, μg mL⁻¹; V represent the volume of the fatty acid sample, mL; N represent the cell quantity of microalgae.

The external standards were used to establish the linear relationship between peak area and concentration for each target analyte. The analytical performance was validated in terms of sensitivity and quantitation limits. The limit of detection (LOD) and limit of quantitation (LOQ) were established based on the characteristics of the linear regression and the sensitivity of the analytical system (Mocak et al. 1997), all reported fatty acid concentrations in biological samples were confirmed to be situated within the validated linear range with coefficients of determination (R²) typically exceeding 0.99 (all R² > 0.98) and accuracies ranging from 80% to 120%, ensuring the robustness and reliability of the quantitative results (Supplementary Table S1). All fatty acid concentrations reported in this study were well above the instrumental LOD and fell within the validated linear range starting from the LOQ, ensuring high data reliability.

The raw data and analysis scripts have been deposited at https://doi.org/10.6084/m9.figshare.31405839. All biochemical parameters (carbohydrates, proteins, lipids, and fatty acids) were normalized to both per-cell and volumetric bases. Specifically, per-cell content was calculated by dividing the total concentration by the synchronized cell count (X per cell = X_total/cell number) while volumetric productivity was expressed relative to the culture volume (volumetric X = X_total/V culture volume). Data were analyzed using Graphpad Prism 10.1.2, and the results are expressed as the mean values ± standard deviations. Two-way repeated-measures ANOVA was used to evaluate the impact of GABA concentrations on the growth curve over time, followed by post-hoc comparisons at each individual time point. In this analysis, GABA concentration and time were considered as factors, with time treated as a repeated measure.

A two-way ANOVA followed by post-hoc tests was performed to evaluate the impact of GABA concentrations on biochemical composition, where GABA concentration and biochemical composition (protein, lipid, and carbohydrate) were defined as the two main factors. For ROS, SOD, and Fv′/Fm′, a one-way ANOVA followed by post-hoc tests was conducted. Differences at p < 0.05 were considered significant between the treatments. Assessing differences in fatty acid profiles between groups were evaluated using R software version 4.3.3 (R Foundation for Statistical Computing, Vienna, Austria). For each fatty acid, an independent samples t-test was performed to obtain p-value (Raw_P). To solve the issue of multi-endpoint testing, the Benjamini-Hochberg procedure was applied to control the false discovery rate (FDR), yielding adjusted p-value (FDR_adjusted_P). Effect size was quantified to evaluate the magnitude of group differences. Eta squared (η2) was calculated to represent the proportion of total variance attributable to group membership. All analyses were conducted using the R packages lsr and effsize.

To determine the optimal GABA concentration for growth enhancement, we monitored the growth curves of P. tricornutum and C. muelleri following exogenous GABA supplementation. P. tricornutum was treated with 0.016–200 pmol cell⁻¹ GABA during both the inoculation and exponential phases, while C. muelleri received 0.09–20.47 pmol cell⁻¹ GABA in the late stage of exponential growth period and 1.5–16.2 pmol cell⁻¹ GABA during both the inoculation and exponential phases (Fig. 1). P. tricornutum exhibited high sensitivity to exogenous GABA, with concentrations exceeding 0.2 pmol cell⁻¹ significantly inhibiting both cell division and photosynthetic activity. In contrast, C. muelleri demonstrated greater tolerance, with higher GABA concentrations (1.5–16.2 pmol cell⁻¹) promoting cell division (Fig. 1).

Initial dose-response experiments revealed distinct sensitivity profiles between the two species (Fig. 1). P. tricornutum exhibited high sensitivity to exogenous GABA, with concentrations exceeding 0.2 pmol cell⁻¹ significantly inhibiting both cell division and photosynthetic activity. Conversely, C. muelleri demonstrated greater tolerance, with higher concentrations (1.5–16.2 pmol cell⁻¹) promoting cell growth. Guided by these observations, species-specific dosages (0.2, 0.5, and 1.0 pmol cell⁻¹ for P. tricornutum; 1.5, 6.2, and 16.2 pmol cell⁻¹ for C. muelleri) were selected for supplementation at the onset of the exponential phase during semi-continuous cultivation, aiming to further evaluate their metabolic effects during active growth. GABA supplementation (0.2–1.0 pmol cell⁻¹) at the onset of the exponential phase significantly suppressed the growth of P. tricornutum in semi-continuous cultures, with a concomitant and marked increase in cell size (Fig. 2A & B). The effects of GABA on C. muelleri were highly dependent on the timing of supplementation. While treatments initiated at the start of the exponential phase (1.5–16.2 pmol cell⁻¹) had no significant impact on growth or cell size (Fig. 2C & D), supplementation at the late exponential phase significantly promoted cell proliferation and led to a reduction in cell size (Fig. 2E & F).

GABA supplementation at the onset of the exponential phase in P. tricornutum under semi-continuous cultivation induced an increase in carbohydrate and lipid concentrations in a clear dose-dependent manner, reaching maximum levels at 1.0 pmol cell⁻¹. At this dosage, the carbohydrate and lipid content per cell were 3.30 and 2.88 times higher, respectively, than those of the control group (Fig. 3A). The cellular protein concentration reached a 1.71-fold increase over the control at 0.2 pmol cell⁻¹ GABA. Notably, further increasing the dosage to 0.5 and 1.0 pmol cell⁻¹ yielded only marginal further increases compared to the 0.2 pmol cell⁻¹ group (Fig. 3A). In terms of volumetric yield (per mL of culture), GABA supplementation in P. tricornutum led to marked reductions in the volumetric concentrations of carbohydrate, protein, and lipid (Supplementary Fig. S2).

GABA supplementation at the onset of the exponential phase in C. muelleri under semi-continuous cultivation did not significantly alter carbohydrate, lipid, and protein concentrations (Fig. 3B). However, GABA supplementation at the late stage of the exponential phase significantly influenced the carbohydrate, lipid, and protein concentrations per cell. GABA concentrations were inversely correlated with carbohydrate, lipid, and protein concentrations per cell, with the lowest cellular concentrations of carbohydrates and lipids observed at 16.2 pmol cell⁻¹ (Fig. 3C). Regarding the cellular protein concentration per cell, the 1.5 pmol cell⁻¹ GABA treatment yielded a marginally higher concentration compared to the other groups (Fig. 3C). No significant changes were observed in volumetric carbohydrate, lipid and protein concentrations with GABA supplement at the onset of exponential phase of C. muelleri under semi-continuous cultivation (Supplementary Fig. S2). GABA supplementation at the late exponential phase of C. muelleri showed a positive correlation with volumetric protein concentration, which peaked at 16.2 pmol cell⁻¹; however, volumetric carbohydrate and lipid concentrations remained unaffected (Supplementary Fig. S2).

The experimental results indicate that GABA supplementation in semi-continuous culture at the commencement of the exponential phase (0.2, 0.5, and 1.0 pmol cell⁻¹, respectively) significantly influenced the oxidative stress levels, antioxidant enzyme activity, and photosynthetic efficiency of P. tricornutum in a dose-dependent manner. The concentration of ROS increased sharply with higher GABA doses. Compared to the control group, the 1.0 pmol per cell dose induced 4.30-fold increase in relative ROS levels (p < 0.0001) (Fig. 4A). On the contrary, SOD activity showed a gradual upward trend with increase of GABA concentration. A significant increase was observed at 0.2 and 0.5 pmol per cell dose compared to the control group (p < 0.05) (Fig. 4B). Fv′/Fm′ also exhibited a significant decline as the GABA dose increased. The control group maintained the highest efficiency, while all GABA-treated groups showed a marked reduction (p < 0.001) (Fig. 4C). Unlike P. tricornutum, C. muelleri showed no marked response to vary concentrations of GABA (1.5, 6.2, and 16.2 pmol cell⁻¹, respectively) supplementation in semi-continuous cultures at the commencement of the exponential phase in terms of ROS levels, SOD activity, and Fv′/Fm′.

As shown in Fig. 5A and Supplementary Table S2, supplementation of varying GABA concentrations at the onset of the exponential growth phase significantly augmented the overall cellular fatty acid concentrations of P. tricornutum. Selected fatty acids, including α-linolenic acid (C18:3n3), eicosapentaenoic acid (EPA; C20:5n3), and docosahexaenoic acid (DHA; C22:6n3), increased significantly under GABA treatments in P. tricornutum, whereas no fatty acids were significantly altered in C. muelleri (Fig. 5A). In contrast, no fatty acids were altered by GABA supplementation in C. muelleri (Fig. 5B).