Clonal strains of Skeletonema costatum sensu stricto (Greville) Cleve and Chaetoceros lorenzianus Grunow, originally isolated from the Seto Inland Sea of Japan were used. The strains were maintained in f/2 medium (Guillard, 1975), with Na₂SeO₃ (final 1.0 × 10^-8 M) and hydroxymethyl aminomethane (final 2 mM) additions. The strains were incubated at 20°C under a photon flux density (PFD) of 150 µmol-photons m^-2 s^-1 (L:D=12:12 h) provided by a cool-white tube-type light-emitting diode (LED) light source.

Continuous cultures were used to obtain cells in which only one of the target nutrients (N or P) was depleted. In continuous culture, new medium is supplied continuously to a culture vessel, and the same amount of culture medium is removed (chemostat culture, Ukeles, 1973). The dilution rate in the culture vessel can be calculated as the daily outflow volume relative to the total volume of the culture vessel. For this purpose, 1 L custom-made cylindrical vessels with a side drainage port were fabricated ( Supplementary Figure 1 ). Teflon tubes (inner diameter = 2 mm) were inserted into a silicon plug that was placed in the vessel mouth, and one of these tubes was connected to a tubing pump (SMP-23AS, AS ONE, Japan) to supply medium. Another Teflon tube was used to obtain samples for the experiment. Air was supplied from the conical bottom of the vessel to mix the culture. A side drainage port was used to remove excess medium and pressurized air. For each diatom species, four continuous cultures (i.e., N- and P-limited conditions at two different dilution rates), were established; each diatom species was subjected to continuous culture at a high dilution rate (HDR: 0.44–0.54 day^-1) and a low dilution rate (LDR: 0.13–0.16 day^-1) under N- and P-limited conditions. Experiments were performed at 20°C under a PFD of 150 µmol-photons m^-2 s^-1 (L:D=12:12 h).

In the principle of our continuous culture experiment, a culture was maintained initially as batch culture until when the target nutrient was depleted (approx. for 2 weeks), then the continuous culture experiment was started by supplying new N- or P-limited f/2 medium. For the initial batch cultures, N-limited f/2 medium with a 1/10 NaNO₃ concentration (final 88.3 µM) or P-limited f/2 medium with a 1/10 NaH₂PO₄ concentration (final 3.6 µM) was initially placed in the culture vessels, which were then inoculated with S. costatum or C. lorenzianus to give initial cell densities of 1,000 cells mL^-1 and 500 cells mL^-1, respectively. This density ratio in the precultures was determined based on a previous study in nutrient uptake ratios and cell volumes of S. costatum and Chaetoceros gracile (Bienfang et al., 1982); nutrient uptake of C. gracile, whose cell volume was about three times larger than S. costatum, was 1.3–1.5 times faster than S. costatum. The mean cell volume of the C. lorenzianus culture strain in this study was about four times larger than the S. costatum culture strain (234.25 ± 74.99 µm³ (n = 32) and 946.56 ± 471.78 µm³ (n = 32), respectively), suggesting that nutrient uptake of C. lorenzianus was faster than C. gracile. The cultures were maintained as batch cultures for 2 weeks without supplying new medium. Upon entering the beginning of the stationary growth phase when the target nutrient was assumed to be totally depleted, the continuous culture experiments were started by supplying new N- or P-limited f/2 medium continuously at the two dilution rates mentioned above. An aliquot of the culture was retrieved from the Teflon tube using a syringe, and cell density, residual inorganic nutrients and photosynthetic parameters were measured daily. The cell densities were measured by counting under a microscope using a plankton counting chamber (MPC-200, Matsunami Glass, Japan). To avoid counting dead cells, live cell staining was conducted by adding neutral red solution (Sukisaki and Umino, 2013). Residual nitrogen (dissolved inorganic nitrogen (DIN: NO₃ + NO₂ + NH₄-N)) and phosphorus (dissolved inorganic phosphorus (DIP: PO₄-P)) were analyzed using a nutrient autoanalyzer (SWAAT, BLTEC, Japan). To prevent nutrient contamination, all instruments made of glass or plastic were soaked overnight in 1 N hydrochloric acid and then washed with MilliQ before use. Based on the constant cell densities and complete depletions of residual nutrients, cultures were assumed to have reached a steady state (balanced growth phase) at day 12, when the specimens were subjected to the experiments described below.

Before subjecting the continuous cultures to the prolonged high light exposure experiments, the photosynthetic parameters, Fv/Fm (i.e., maximum quantum yield of dark-adapted photosystem II) and quenching parameters were initially measured during rapid light curve measurements (Ralph and Gademann, 2005) using cells retrieved at day 12 using a pulse amplitude modulation (PAM) fluorometer (WATER-PAM, Heinz Walz, Germany). The measurements were performed in triplicate. Then, 1.2 mL of the culture that had been dark-adapted for more than 15 min was transferred to a cuvette and illuminated with eight increasing intensities of actinic lights (i.e., PFD = 158, 241, 356, 553, 825, 1179, 1648, 2743 µmol-photons m^-2 s^-1) for 30 s. The quenching parameters qP and NPQ were obtained at each PFD level using software (WinControl-3, Heinz Walz), and Fv/Fm, qP and NPQ (Stern-Volmer NPQ) were estimated using the following equations (Kitajima and Butler, 1975; Schreiber et al., 1986; Bilger and Björkman, 1990):

where, Fm is maximum fluorescence yield after dark acclimation; Fo is minimum fluorescence yield after dark acclimation; Fm’ is maximum fluorescence yield in light-acclimated state; F’ is variable basal fluorescence under actinic light; Fo’ is minimum fluorescence yield in the light-acclimated state. After measuring the parameters mentioned above, each 80 mL aliquot was separately dispensed into six 270-mL tissue culture flasks (Canted Neck, IWAKI, Japan) for eight different specimen materials (i.e., N- and P-depleted samples × two dilution rates × two species) ( Supplementary Figure 2 ). Among the six flasks, three were for observing the recovery responses after each nutrient was replenished, i.e., an NH₄Cl solution (final 100 µM) or a NaH₂PO₄ solution (final 10 µM) was added to the three flasks for the N- or P-depleted specimens, respectively. The other three flasks were used as controls. All of the flasks were incubated in the dark for 15 min to measure Fo and Fm. The samples were then transferred to a water bath and incubated at 20°C. Half of the bath was illuminated from the bottom with low light (LL: 50 µmol-photons m^-2 s^-1), and the remaining half was illuminated with high light (HL: 800 µmol-photons m^-2 s^-1). This high light intensity was determined based on field data obtained at the sea surface at noon on a sunny day (approx. 750 µmol-photons m^-2 s^-1 on average). To determine the initial values of qP and NPQ before HL exposure, the flasks were initially maintained under LL exposure for 60 min, followed by HL exposure for 60 min, and then under LL exposure for another 60 min to observe the recovery from HL-induced stress. To measure qP and NPQ, sample aliquots of approximately 5 mL were retrieved from the flask at 30 min and 60 min during the initial LL incubation step, and then at 15-min intervals during the HL and the final LL exposure steps. A 1.2 mL aliquot was subjected to PAM fluorometry analysis.

To simulate the summer season in our study area, S. costatum and C. lorenzianus were mixed in a P-limited continuous culture, and shifts in the relative abundance of the two species under two different light levels was performed in triplicate. The two diatom species were initially precultured separately in batch culture for 2 weeks at 25°C under a PFD of 150 µmol-photons m^-2 s^-1 (L:D=12:12 h); the initial cell densities were the same as in the above experiment, i.e., 1,000 cells mL^-1 for S. costatum and 500 cells mL^-1 for C. lorenzianus, respectively. After the 2 weeks, the cultures were subjected to semicontinuous culture for 1 week; 10% (volume) of the culture was replaced daily with new f/2 medium in order to minimize the likelihood of differences in nutrient stress, possibly due to differences in growth and/or nutrient consumption rates between species. Then, 150 mL of the precultured diatoms were mixed and transferred to a total of six continuous culture vessels. The vessels were filled to 1 L with P-limited f/2 medium, and new medium was supplied at a dilution rate of 0.21 day^-1. Then, three of the six vessels were illuminated with moderate light (ML: PFD = 150 µmol-photons m^-2 s^-1), and the other three were illuminated with high light (HL: PFD = 800 µmol-photons m^-2 s^-1). Residual inorganic nutrients, i.e., DIP and DIN were measured daily for the first 5 days and then once every two days thereafter. Cell densities and photosynthetic parameters were measured at day 1, 2, 4, 7, 9, 11, 13 and 15. Cell density measurements were performed as described in the “2.2 Continuous culture” section. To measure the photosynthetic parameters Fv/Fm and NPQ for each species, the cells from the cultures were placed on a glass slide with a cover slip and measurements were conducted by a microscopy-type PAM fluorometer (Micro-FluorCam FC-2000, Photon Systems Instruments, Czech Republic) as described in Higo et al. (2017). To obtain NPQ, light levels corresponding to ML or HL were supplied using the transmission light of the microscope for 1 min. Under this illumination, a fluorescence-induction curve was plotted with saturation pulses (five times) using the Quenching Analysis Wizard of the PAM software (FluorCam 7 ver. 1.2.5.24, Photon Systems Instruments). At least 15 colonies of each species were measured.

In this study, the prolonged high light exposure experiment and the competition experiment were conducted in triplicate. The average parameters were obtained from the prolonged high light exposure experiments, i.e., qP and NPQ, and the competition experiments i.e., cell density, Fv/Fm and NPQ. The above data were subjected to statistical analysis using Student’s t-test in R 4.2.2 (R Development Core Team, 2022). In this study, a p-value less than 0.01 was considered statistically significant.

To simulate the combination of P limitation and high light conditions, which appears to affect S. costatum negatively and has a limited effect on C. lorenzianus, an experiment consisting of a co-mixed culture of both species in a single continuous culture vessel was conducted in triplicate. As in the single-species continuous experiments above, the levels of the target nutrient, i.e., DIP, were totally depleted after day 5 or 3 under moderate light (ML: 150 µmol-photons m^-2 s^-1) ( Figure 4A ) or high light (HL: 800 µmol-photons m^-2 s^-1) ( Figure 4B ), respectively, while the DIN levels remained constant. Under ML conditions, both species showed gradual increases in cell density until day 4, when the cell density of S. costatum was higher than that of C. lorenzianus ( Figure 4C ). Similar trends were observed under HL conditions ( Figure 4D ). While C. lorenzianus maintained almost constant densities from day 4 to day 15 under both light conditions, the cell densities of S. costatum decreased 10-fold under ML conditions and decreased suddenly by 100-fold under HL conditions on day 7 from the values on day 4. Under ML conditions, S. costatum showed a gradual decrease in cell density and finally disappeared on day 15. Under HL conditions, S. costatum disappeared at day 9.

Because stationary-phase cultures were used prior to the start of the continuous culture experiments, the cellular Fv/Fm values on day 1 for both species were relatively low, with the values of S. costatum being higher than those of C. lorenzianus ( Figures 5A, B ). On day 2, under both light conditions, Fv/Fm values increased gradually as the cell density increased at the start of the continuous culture. While there was no significant difference between in the Fv/Fm values obtained for each species on day 2 under both ML (p = 0.09) and HL (p = 0.05) conditions, the cellular Fv/Fm values of C. lorenzianus were significantly higher than those of S. costatum on day 4 under both light conditions (p = 0.001 under ML conditions and p = 2.8 × 10^-14 under HL conditions, respectively). After day 7, Fo was too low to determine Fv/Fm and thus the values for S. costatum were treated as being below the detection limit. However, C. lorenzianus exhibited relatively higher Fv/Fm values on day 7 (0.57 ± 0.06 under ML conditions and 0.55 ± 0.09 under HL conditions, respectively), which then decreased slightly at day 15 (0.51 ± 0.60 under ML condition and 0.45 ± 0.60 under HL condition). NPQ values in S. costatum were higher under both light levels, and were higher than those of C. lorenzianus on day 1 ( Figures 5C, D ). On day 2, under ML conditions, NPQ values in C. lorenzianus increased while those in S. costatum remained similar to those on day 1 (0.49 ± 0.34 on day1 and 0.45 ± 0.34 day 2, respectively). Conversely, the NPQ values of both species increased significantly under HL conditions, although the increase was more obvious and significantly higher in S. costatum (1.18 ± 0.44 for S. costatum and 0.83 ± 0.60 for C. lorenzianus, p = 1.6 × 10^-5). At day 4, which was the turning point of the decrease in S. costatum, the NPQ values of both species decreased suddenly. After day 7, since the fluorescence levels of S. costatum were under the detection limits, NPQ could not be estimated. This was similar to the case in Fv/Fm values, but C. lorenzianus exhibited gradual increases of NPQ.

In the continuous culture experiments, similar cell densities were observed in S. costatum at either dilution rate under both N- and P-limited conditions on day 12 when the cells were in a balanced growth phase. However, in the C. lorenzianus cultures, the cell densities under LDR conditions were one order of magnitude lower than those under HDR conditions under both N- and P-limited conditions. Therefore, S. costatum appears to maintain constant cell densities irrespective of nutrient levels, while the cell densities of C. lorenzianus changed depending on nutrient availability. Nutrient uptake in phytoplankton increases as cell size increases (Marañón et al., 2013). Compared to C. lorenzianus, the smaller cells of S. costatum may have affected their lower nutrient uptake and helped to maintain higher cell densities; however, C. lorenzianus might possess another strategy to decrease cell density, which would enable this species to allocate more nutrients to each cell. This assumption is supported by the observation that C. lorenzianus maintained higher Fv/Fm values than S. costatum, even under the LDR conditions. Therefore, the following discussion on N-limited conditions should be interpreted after considering such differences between both species. In the case of P-limited conditions, the above assumed strategies in both species might be negligible because they had comparable Fv/Fm values under the same dilution rates.

From the results of the rapid light curve measurements, under N-limited conditions, both species exhibited relatively higher qP values (> 0.8) regardless of their N deficiency status (i.e., dilution rate) under ML exposure ( Figure 2A ). However, under HL exposure, the qP values in S. costatum decreased to 0.64 in the HDR treatment and more extensively to 0.56 in the LDR treatment, while NPQ reached almost 1 in both dilution rate treatments. In contrast, such a decline in qP and an increase in NPQ were not observed in C. lorenzianus, even under the same conditions. Li et al. (2021) reported that a higher NPQ was observed in the diatom Thalassiosira punctigera under a combination of high light and N-limited conditions. They proposed that the increase in NPQ was a passive response to a decrease in PSII performance caused by inhibition of the de novo synthesis of D1 protein and resulting imbalance of the repair-damage cycle of PSII under N deficiency. In another diatom, Phaeodactylum tricornutum, N deficiency has been reported to decrease the D1 protein (PsbA) (Levitan et al., 2015). Our results showed similar increases in NPQ values under N-limited and HL conditions in S. costatum, but no such increases were observed in C. lorenzianus. These findings imply that S. costatum was more susceptible than C. lorenzianus to the combination of N limitation and HL conditions and consequently, that NPQ should be induced as a photoprotective strategy. Alternatively, as mentioned previously, the lower densities of C. lorenzianus may have meant that more nitrogen could be allocated to each cell, obviating the need to induce NPQ. However, the high NPQ induction observed in S. costatum might not have been sufficient to mitigate against photoinhibition under N-deficient and HL conditions because qP decreased, as shown in Figure 2A .

Conversely, under prolonged light exposure ( Figure 3 ), C. lorenzianus induced NPQ to a similar level as that observed in S. costatum. NPQ induction is known to vary between diatom species and with light intensity and duration (Lavaud et al., 2004). Therefore, while the response of S. costatum occurred quickly, that of C. lorenzianus may have been slower. NPQ induction in N-limited C. lorenzianus was found only under prolonged HL exposure, implying that NPQ induction under continuous HL exposure may be required even in this species. By inducing NPQ, C. lorenzianus effectively maintained high qP levels under HL exposure ( Figure 3C ). Conversely, S. costatum appeared to be unable to maintain qP levels, irrespective of high NPQ induction ( Figure 3A ). These findings reinforce the above interpretation of the rapid light curve measurements in which the high NPQ induction in S. costatum may not have been sufficient for mitigating against photoinhibition under N-deficient and HL exposure. The reasons for the decrease in NPQ values observed in both species in the LDR treatments were unclear and remain speculative (see discussion below). Nevertheless, insufficient NPQ induction under LDR treatments caused a further decrease in qP values during HL exposure in S. costatum, but such a decrease was less apparent in C. lorenzianus.

Interestingly, decreases in NPQ values were also observed with N addition, but only in C. lorenzianus ( Figure 3C ). Such decreases in NPQ values, which were caused by N addition and not by N deficiency, were assumed to have occurred in response to NPQ relief due to a decrease in the need for the photoprotective function of C. lorenzianus. Indeed, the reintroduction of nitrogen has been shown to relieve NPQ in several microalgae due to the rapid recovery of photochemical functions (Liefer et al., 2018; Zhang et al., 2019). Such a quick recovery of the photochemical functions in C. lorenzianus would enable it to use higher light after reintroducing nitrogen, which further suggests the advantage that this species has over S. costatum under the combination of N limitation and HL conditions.

Our experiments showed that, under severely P-limited conditions, i.e., under LDR conditions, even ML exposure caused significant decreases in qP in both species ( Figure 2B ), indicating that even the ML intensity might be excessive and/or cause damage to PSII downstream. In rice plants, qP was also reduced under P-deficient conditions (Xu et al., 2007), implying that the proportion of the reduced state of quinone Q_A was elevated (Schreiber et al., 1986). Decreases in qP were more apparent under HL exposure, especially in S. costatum, suggesting that additional high light might lead to downstream deterioration and/or a marked reduction in Q_A. In addition, the residual DIP concentrations in S. costatum under N-limited condition were relatively lower than those in C. lorenzianus (shown in Figure 1A , LDR), suggesting S. costatum might have a higher demand for phosphorus than C. lorenzianus and this might lead to lower qP values in S. costatum than C. lorenzianus under the P-limited conditions. As with N limitation, both species induced NPQ passively under a combination of P limitation and HL exposure ( Figure 2D ), with the highest NPQ values observed in the LDR treatment. These findings suggest that P deficiency induced more severe photoinhibition than N deficiency, and that more photoprotection was necessary. Similar NPQ increases have been observed under P-limited conditions in various microalgae (Cui et al., 2017; Guo et al., 2018; Sun et al., 2019; Rocha et al., 2021). In these studies, NPQ activity was enhanced under P deficiency in order to dissipate excess light energy, protect against potential photooxidative damage and maintain the fluency of the energy flow (Cui et al., 2017). In Figure 3 , these NPQ-based photoprotection mechanisms were observed in both species under HL exposure, although they were more apparent in C. lorenzianus under prolonged HL exposure (the maximum NPQ under HL exposure in the LDR conditions was 0.79 for S. costatum and 1.52 for C. lorenzianus). As mentioned previously, C. lorenzianus changed cell allocation responses based on nutrient loads. Higher levels of NPQ induction in C. lorenzianus compared to S. costatum, even under P-limited conditions, might indicate a higher potential for NPQ induction in C. lorenzianus.

Such active NPQ induction in P-starved cells is unique and differs from the case of N deficiency, in which NPQ impairment was observed. It is known that NPQ induction in diatoms relies mainly on qE (Goss and Lepetit, 2015), a quenching mechanism that is controlled by the build-up of the trans-thylakoid proton gradient, the conversion of xanthophylls diadinoxanthin (Dd) into diatoxanthin (Dt), and the presence of Lhcx proteins. Dt needs to be bound to LHC proteins, most likely to Lhcx proteins, to enhance NPQ (Lepetit et al., 2013), so the production of Dt and Lhcx proteins could control NPQ. Elevated expression of specific Lhcx genes under N deficiency has been reported in the diatom Phaeodactylum tricornutum (Taddei et al., 2016). Although there have been no reports of elevated Lhcx expression under P-limited conditions in diatoms, Dt was reported to be generally high under P limitation in the coccolithophorid Emiliania huxleyi (Stolte et al., 2000). This may be due to a malfunction in ATP synthase, which would be unable to drain protons out of the lumen, resulting in an excessive accumulation of protons and an increase in the trans-thylakoid proton gradient (ΔpH), which would in turn activate pH-dependent diadinoxanthin de-epoxidase (Huang et al., 2019). While further investigations are needed, the increase and decrease in NPQ under P- and N-limited conditions could be affected by the size or status of the Dd/Dt pools and Lhcx expression, which may show species-specific variations.

Similar to the case of N limitation, relief of NPQ induction with P addition was observed only in C. lorenzianus ( Figure 3H ). The reintroduction of phosphate also improved qP in this species, which facilitated the quick recovery of this species (compared to S. costatum).

The above experiments revealed characteristic responses in both species. Especially under P deficiency, S. costatum was more susceptible to light stress and actively induced NPQ to protect itself against excessive light; however, as indicated by the decrease in qP, NPQ might be insufficient. In contrast, C. lorenzianus could maintain relatively higher qP levels without NPQ after exposure to HL for a short time. Indeed, even under prolonged exposure to HL, C. lorenzianus actively induced NPQ, which resulted in the maintenance of higher qP levels than in S. costatum. Based on recent reports that have described a decreasing trend in phosphorus in the Seto Inland Sea (Hayashi et al., 2000; Ohara et al., 2020), we employed continuous cocultures of S. costatum and C. lorenzianus under P-limited condition to simulate the typical summer environment where severe P deficiency and high light occur. The results showed a population shift from S. costatum to C. lorenzianus under both ML and HL conditions, which could support several of the observed phenomena of Chaetoceros succession from Skeletonema in the field, presumably due to ocean oligotrophication (Tada et al., 2014). Succession of Chaetoceros over Skeletonema under P-limited conditions was also reported by Roden and O'Mahony (1984). In their experiment, a field-collected phytoplankton assemblage containing S. costatum and Chaetoceros sp. was supplied with a P-limited medium (N:P = 34) in a semicontinuous culture. Their findings showed that S. costatum became dominant. under a higher dilution rate condition when P limitation was moderate; conversely, S. costatum was outcompeted by Chaetoceros sp. under a lower dilution rate condition. Given this background and the findings of this study, the rapid succession of C. lorenzianus under HL conditions in our experiment suggests the existence of a synergetic relationship between HL conditions and P limitation. In addition, the findings that C. lorenzianus seemed to be able to allocate more nutrients to cells and recover faster upon nutrient replenishment may also explain the observed succession dynamics in this species. These assumptions might be supported by the photosynthetic responses of C. lorenzianus and S. costatum, in which the former exhibited relatively higher Fv/Fm values than the latter at day 4, likely because of the better physiological status of C. lorenzianus compared to S. costatum. The higher NPQ in S. costatum at the beginning of culture under both ML and HL conditions in the single-species culture experiments was previously attributed to active photoprotection of the photosystem damaged by P deficiency. This defense mechanism, however, eventually deteriorated after day 4 and resulted in a rapid decline in the S. costatum population. Although NPQ also decreased in C. lorenzianus on day 4, unlike the decrease observed in S. costatum, this decrease may have been ascribed to relaxation of NPQ, as shown in Figure 3H , probably because C. lorenzianus could use all of the phosphorus due to the decrease in S. costatum. Recent severe P limitation and increased solar insolation in summer in the Seto Inland Sea is definitely disadvantageous for S. costatum. This could be one of the reasons for the recent replacement of Skeletonema spp. by Chaetoceros spp. in the Seto Inland Sea and might be applicable for other bays in western Japan where a shift from Skeletonema spp. to Chaetoceros spp. has also been observed (Tada et al., 2014; Satomichi et al., 2019).