The diatom P. tricornutum was obtained from Shanghai Guangyu Biological Technology Co., Ltd. It was cultured in sterile natural seawater, supplemented with nutrients, trace metals, and vitamins following the f/2 recipe (Guillard & Ryther, 1962; Orlowska et al., 2017). The culture conditions for P. tricornutum included a temperature of 25 °C and a light-dark cycle of 12:12 hours.

The experiments aimed to analyze the effects of nutrient depletion (PO₄ ^3- and NO₃ ^-) and subsequent recovery on P. tricornutum. To acclimate the cells to the experimental conditions, daily dilutions were performed using sterile natural seawater with consistent nutrient concentrations. This adjustment process lasted 2-3 days before each experiment, ensuring the biomass matched that of the previous day. The cultures were then transferred to glass bottles and diluted to a volume of 15 L with sterile natural seawater, while maintaining an initial cell abundance of approximately 10⁵ cells per mL. The initial concentrations of nitrate, phosphate, and silicate for the phosphate-depletion and nitrate-depletion experiments were set at 50 µM, 2 µM, 50 µM, and 12 µM, 10 µM, 50 µM, respectively. Trace metals and vitamins were added based on the f/2 culture media. Throughout the nutrient-depletion experiments, non-limiting nutrients, trace metals, and sufficient vitamins were maintained to ensure only one nutrient became limiting. The strain was cultured under a light intensity of approximately 250 µmol photon m^-2 s^-1, with a 12:12 h light-dark cycle, at room temperature. The culture had a salinity of 28.5‰, a pH of 8.0, and was continuously stirred at 160 rpm. The first light exposure begins at time point zero. In the single nutrient depletion experiment, nutrient samples are collected every 12 hours, and cell abundance and transparent exopolymer particles samples are collected every 24 hours. Nutrient status was determined based on the growth rate of the cells, classified as: (1) nutrient-repletion (NR), representing log-phase growth; (2) nutrient-limitation (NL), with a growth rate of 0.3~0.1 d^-1; (3) nutrient-depletion (ND), where the growth rate approached zero. When the growth rate neared zero, the missing nutrient was added to the culture to restore replete conditions. This process is referred to as a nutrient spike. Sinking rate, protein, carbohydrate, total lipid, and absolute quantitative lipidomics analyses were conducted at three nutrient statuses and five time points: nutrient-repletion, nutrient-limitation, nutrient-depletion, and 2, 6, 12, 18, and 24 hours after adding the limiting nutrient.

Cell abundance was monitored using an automated cell counter from Shanghai RuiYu Biotech Co. Ltd. The growth rate (μ, d^-1) was calculated daily using the following Equation 1:

where C₀ and C₁ represent the cell concentrations at times t₀ and t₁, respectively.

The optimal photochemical efficiency of photosystem II (PSII) was measured daily using water pulse amplitude modulated (PAM) fluorometry from Heinz Walz GmbH (91090 Effeltrich, Germany). Prior to measurement, samples underwent a 20-minute dark treatment to obtain F_m (maximum fluorescence) and F₀ (minimum fluorescence). The maximum photosynthetic quenching capacity (F_v) was then calculated as the difference between F_m and F₀.

To analyze the nutrient content, water samples were filtered using 0.45 μm polycarbonate membranes. Thymol spectrophotometry (Osibanjo & Ajayi, 1980), Naphthylethylenediamine hydrochloride spectrophotometry (Tarafder & Rathore, 1988), molybdenum blue method (Holman, 1943), and silicon molybdenum blue method (Luke, 1953) were employed to measure the concentrations of nitrate, nitrite, phosphate, and silicate, respectively. The total nitrogen concentration was determined by summing the concentrations of nitrite and nitrate.

The sinking rate of the cells was determined using the SETCOL method outlined by Bienfang (1981). The SETCOL apparatus, as depicted in Figure 2 , consisted of three opaque plastic columns measuring 1.0 m in length and with a volume of 1111 mL. Each column was filled with a homogeneous water sample efficiently. Following this, the columns were left undisturbed at room temperature for 3 hours. Precipitated samples were collected from the bottom, middle, and upper compartments of the columns by sequential drainage. The phytoplankton biomass in each compartment was determined by measuring chlorophyll-a concentration (Chl-a), and the sinking rate was calculated using the following Equation 2:

where θ represents the sinking rate in meters per day (m d^-1), B s denotes the biomass settled into the bottom compartment, B t indicates the total biomass in the column, L signifies the column length in meters, and t represents the settling time in days.

For Chl-a analysis, the samples were filtered using 25 mm GF/F filters with a pore size of 0.65 μm under low vacuum pressure (< 0.04 MPa). Subsequently, the filtered samples were stored in the dark at -20 °C. After 24 hours of extraction with 90% acetone at -20°C, the Chl-a concentration was measured using a Turner-Designs Trilogy™ fluorometer (Caspers, 1970).

Concentrations of TEPs were measured using a modified version of a previously established method (Passow and Alldredge, 1995). Triplicate 10 mL samples were filtered through 0.4 μm pore size polycarbonate filters under a low constant vacuum (< 0.02 MPa). The filters were then stained with 0.5 mL of a 0.02% Alcian Blue solution for 2 seconds. To eliminate excess dye, the filters were rinsed twice with 2 mL of distilled water. Next, the filters were subjected to an extraction process using 6 mL of 80% sulfuric acid for 2 hours. The absorbance of TEPs was determined at 787 nm using a spectrophotometer, with Xanthan gum serving as the standard.

The cell morphology was analyzed using a scanning electron microscope (Model: TM-1000 Tabletop Microscope, company: Hitachi High-Technologies Corporation). To prepare the samples, 10 mL aliquots were filtered through 3.0 μm pore size isopore membrane filters and dried overnight. Subsequently, the filters were placed in a gold injector (Model: MSP-1S, company: Vacuum Device Inc) and coated with a thin layer of gold. Finally, the filters were subjected to scanning electron microscopy for photography.

Prior to protein, carbohydrate, and total lipid analysis, a pre-treatment process was applied to the algal sample. In brief, approximately 3 L of P. tricornutum cultures were passed through 0.4 μm pore size polycarbonate filters and washed thrice with distilled water. The filtered P. tricornutum cells were then dried at -70 °C under vacuum for 24 hours. Cell disruption was achieved through repeated freeze-thaw cycles. The cells were frozen below -20°C and subsequently thawed at approximately 4°C, repeating this process 3-4 times. This method utilized the formation of intracellular ice particles and the increased concentration of cytosolic salts to induce cell lysis and fragmentation.

Total lipids were extracted and quantified following the protocol described by Johnson and Wen (Johnson and Wen, 2009). Carbohydrate content was determined using the anthrone colorimetric technique as outlined in Laurentin and Edwards (Laurentin and Edwards, 2003). Protein quantitation was performed using the Lowry method (Waterborg, 2009).

Triplicate samples were obtained from the cultures during the phosphate depletion and spike experiment. P. tricornutum cultures (~50 mL) were filtered through 25 mm diameter 0.2 μm hydrophilic Durapore filters, rinsed thrice with distilled water, and immediately flash-frozen and stored at -80 °C. Lipids were extracted from the filters using the methyl tert-butyl ether (MTBE) method (Matyash et al., 2008). Briefly, the filters were thawed at 4°C and mixed with 200 μL water, 240 μL methanol, and 800 μL MTBE. The mixture was sonicated at low temperature (30 min/once, twice). After centrifugation at 14000 g for 15 min at 10°C, the upper layer was collected and dried under nitrogen.

To analyze the lipids, ultra-high performance liquid chromatography (UHPLC) mass spectrometry (MS) was performed using an UHPLC Nexera LC-30A (SHIMADZU, Japan) coupled to a Q-Exactive Plus high-resolution mass spectrometer (ThermoFisher Scientific, Waltham, MA, USA), following a previously reported method (Liu et al., 2022). Reverse-phase chromatography was employed for LC separation with a CSH C18 column (1.7 μm, 2.1 × 100 mm; Waters). The lipid extracts were re-dissolved in 200 μL of 90% isopropanol/acetonitrile, centrifuged at 14000 g for 15 min, and finally, 3 μL of the sample was injected.

The mobile phase consisted of acetonitrile–water (6:4, vol/vol) with 0.1% formic acid and 0.1 mM ammonium formate as solvent A, and acetonitrile–isopropanol (1:9, vol/vol) with 0.1% formic acid and 0.1 mM ammonium formate as solvent B. The initial mobile phase composition was 30% solvent B at a flow rate of 300 μL/min. It was maintained for 2 min, followed by a linear increase to 100% solvent B over 23 min, and then equilibrated at 5% solvent B for 10 min. Mass spectra were acquired using the Q-Exactive Plus in both positive and negative modes. The Electron Spray Ionization (ESI) parameters were optimized and preset for all measurements: source temperature at 300°C, the capillary temperature at 350°C, ion spray voltage at 3000 V, S-Lens RF Level at 50%, and the scan range set at 200–1800 m/z. For each full scan, 10 fragments (MS2scan, HCD) were collected. The resolution of MS1 was set at 70000 at m/z 200, and that of MS2 was set at 17500 at m/z 200.

To extract and identify the peaks of lipid molecules and internal standard lipid molecules, Lipidsearch (ThermoFisher Scientific, USA) was utilized. The main parameters were as follows: precursor tolerance: 5 ppm, product tolerance: 5 ppm, and product ion threshold: 5%. The raw data and the lipid category naming explanation for the absolute quantification lipidomics in the phosphate depletion and spike experiment are provided in Table S1 and Data Sheet 1 in the Supplementary Material, respectively.

A Pearson correlation and one-way ANOVA were conducted using IBM SPSS Statistics 22.0 to examine the correlation between sinking rate and detecting parameters (F_v/F_m, growth rate, protein, lipids, carbohydrate, and TEPs) in the nutrient depletion and spike experiments, as well as to determine the differences in each parameter under different nutrient regimes. Significance was determined by ANOVA probabilities< 0.05. The normal distribution of the data was checked first, and log10 transformation was applied if necessary. For data that could not be transformed into a normal distribution, the Spearman correlation and the nonparametric Kruskal-Wallis ANOVA were employed. Data visualization was performed using the R programming language.

During the initial 72 hours of the experiment, P. tricornutum exhibited log-phase growth as a result of phosphate repletion (PR). The growth rate ( Figure 3A ) and F_v/F_m ( Figure 5 ) initially increased but subsequently decreased. The mean values for the growth rate and F_v/F_m were 0.75 ± 0.34 d^-1 and 0.52 ± 0.04, respectively. P. tricornutum rarely settled during the PR phase, with a mean sinking rate of -0.06 ± 0.13 m d^-1 ( Figure 4A ). While the intracellular total lipid content remained stable, the protein and carbohydrate contents fluctuated ( Figure 6A ). The mean total lipid content was 0.16 ± 0.01 mg mg^-1. Throughout the PR phase, the concentration of TEPs consistently increased ( Figure 7 ).

After 72 hours of PR, the extracellular phosphate was depleted ( Figure 4A ), resulting in a significant decline in the growth rate and F_v/F_m to values of -0.08 ± 0.06 d^-1 and 0.37 ± 0.01, respectively. This indicated that P. tricornutum faced constraints due to limited phosphate availability and reduced photosynthetic efficiency. Additionally, the sinking speed significantly increased from -0.21 ± 0.08 to 0.58 ± 0.10 m d^-1 (P< 0.05). Under phosphate limitation (PL) conditions, there was minimal variation observed in carbohydrate and TEPs, while the protein concentration slightly decreased. Notably, the total lipid content significantly (P< 0.05) increased from 0.15 ± 0.01 to 0.25 ± 0.01 mg mg^-1.

In the phosphate depletion phase (PD), following 72 hours of PL, the growth rate exhibited fluctuating increases, rising from -0.08 ± 0.06 to -0.03 ± 0.01 d^-1 ( Figure 3A ). This fluctuation in growth rate could potentially be attributed to the re-release of restricted nutrients through cell fragmentation. Moreover, the sinking rate, total lipid content, and TEPs consistently increased, while the concentrations of protein and carbohydrate decreased.

Subsequent to the PD phase, a phosphate spike (PS) experiment was conducted. After the first 2 hours following the phosphate spike (R2h), both the growth rate and F_v/F_m significantly increased (P< 0.05). The growth rate rose from -0.03 ± 0.01 to 0.48 ± 0.08 d^-1, while the F_v/F_m increased from 0.34 ± 0.01 to 0.47 ± 0.01. However, after 6 hours following the PS (R6h), both the growth rate and F_v/F_m began to continuously decline. Furthermore, after 18 hours of phosphate recovery treatment (R18h), the growth rate decreased to -0.06 ± 0.04 d^-1, indicating a re-imposition of phosphate limitation on P. tricornutum. The sinking rate and total lipid content initially decreased, then increased during the PS phase. In contrast, the concentration of TEPs showed an initial increase followed by a subsequent decrease. After 12 hours of phosphate recovery (R12h), the sinking rate and total lipid content reached their lowest values of 0.01 ± 0.05 m d^-1 and 0.16 ± 0.02 mg mg^-1, respectively. However, with the re-imposition of phosphate limitation, the sinking rate and total lipid content increased again. The concentrations of protein and carbohydrates exhibited fluctuating changes during the phosphate recovery period.

Compared to the phosphate depletion experiment, the nitrate depletion experiment showed similar trends in cell abundance, growth rate, F_v/F_m, and TEPs variations. During the initial stage of the experiment (nitrate repletion, NR) from 0 to 72 hours, there was a rapid increase in cell abundance and TEPs, while the growth rate, settling rate, and F_v/F_m remained high. After 72 hours of nitrate sufficiency, the growth rate of P. tricornutum significantly decreased (P< 0.05) from 0.70 ± 0.07 to 0.12 ± 0.02 d^-1 ( Figure 3B ). Simultaneously, the F_v/F_m value declined from 0.46 ± 0.01 to 0.36 ± 0.02 ( Figure 5 ), indicating nitrate limitation (NL) in P. tricornutum. The sinking rate decreased to 0.16 ± 0.10 m d^-1, one-third of the rate under nitrate-replete conditions ( Figure 4B ). Under NL conditions, protein concentration ( Figure 6B ) and TEPs ( Figure 7 ) remained almost unchanged, while total lipid concentration slightly decreased. However, carbohydrate content significantly (P< 0.05) increased from 2.00 ± 0.14 to 7.04 ± 0.46 mg mg^-1.

After 72 hours of NL, the total nitrogen concentration and growth rate approached zero, marking the beginning of nitrogen depletion (ND). The sinking rate and intracellular content concentrations were within the range of measured values during NL. The TEPs concentration increased from 21.31 ± 0.05 × 10³ to 26.01 ± 10.99 × 10³ μg Xeq. L^-1.

In the first 6 hours after the nitrate spike (NS), from 216 to 222 hours, the growth rate and F_v/F_m had regained values typical of the nitrate-replete state. The sinking rate of P. tricornutum exhibited a slight fluctuating increase, and the TEPs concentration slightly increased during this time. Between 6 and 24 hours after the NS, P. tricornutum continued to show log-phase growth while maintaining relatively high values of F_v/F_m. However, the sinking rate decreased from 0.24 ± 0.05 to -0.05 ± 0.06 m d^-1. The intracellular contents and TEPs concentrations showed fluctuating changes during the NS period.

Figure 8A shows the correlations between sinking rate and physiological-biochemical parameters in phosphate depletion and spike experiments. The correlation plots reveal a significant positive relationship (P< 0.001) between sinking rate and lipid content (R = 0.91). Conversely, lipid content demonstrates a significant negative correlation (P< 0.05) with F_v/F_m (R = -0.77). Additionally, TEPs concentration exhibits a significant (P< 0.01) negative linear relationship (R = -0.89) with protein content.

In Figure 8B , the correlations among parameters in nitrate depletion and spike experiments are presented. The results demonstrate significant positive associations (P< 0.01) between F_v/F_m and growth rate (R = 0.85). Furthermore, TEPs show significant negative correlations (P< 0.05) with protein content (R = -0.60) and sinking rate (R = -0.71). The carbohydrate concentration exhibits significant negative correspondences (P< 0.05) with F_v/F_m (R = -0.81).

Figure 9 illustrates the overall cell population morphology in the four phosphate phases. In instances where phosphate was restricted and depleted, there was an enhancement in cell aggregation, resulting in a significant increase in the number of aggregates. However, within 24 hours after the phosphate spike, the number of cell aggregates decreased.

In the experiments involving phosphate depletion and spike, a significant correlation was observed between the settling rate and total lipid content. To further understand the buoyancy regulation mechanism of P. tricornutum, lipidomics analysis was conducted on the seven phases (PR, PL, PD, R2h, R6h, R12h, and R24h) of the phosphate depletion and spike experiment, and the findings are presented in Table 1 . The lipid composition of P. tricornutum mainly consists of glycerophospholipids, glycerolipids, serol lipids, sphingolipids, saccharolipids, prenol lipids, and fatty acyls. Among these, glycerophospholipids, glycerolipids, sphingolipids, and saccharolipids exhibited higher abundances. During the transition from PR to PL and PD, the content of glycerolipids and saccharolipids gradually increased, particularly in subclasses such as triglyceride (TG), digalactosyldiacylglycerol (DGDG), monogalactosyldiacylglycerol (MGDG), and sulfoquinovosyldiacylglycerol (SQDG). However, during the phosphate spike (PS) phase, the content of these lipids gradually decreased until 24 hours after PS, when their content began to rise again. Furthermore, subclasses such as phosphatidylinositol (PI) in glycerophospholipids and hexosylceramide (Hex1Cer) in sphingolipids exhibited similar trends of variation as these lipids.

Figure 10 illustrates the correlations between the sinking rate and lipid subclasses during the phosphate depletion and spike experiments. The sinking rate demonstrated significant positive correlations with TG (R = 0.89, P< 0.01), cholesterol ester (ChE, R = 0.88, P< 0.01), stigmasterol ester (StE, R = 0.89, P< 0.01), Hex1Cer (R = 0.93, P< 0.01), phytosphingosine (phSM, R = 0.82, P< 0.05), DGDG (R = 0.86, P< 0.05), MGDG (R = 0.81, P< 0.05), coenzyme Q (Co, R = 0.94, P< 0.01), and fatty acid (FA, R = 0.91, P< 0.01).