Four P. lima complex strains, BS4F5 (morphotype 1), WZD145 (morphotype 2), SD4 (morphotype 3), and DS4G4 (morphotype 5), and one P. caipirignum strain, SE10 (morphotype 4), were cultured in natural seawater (salinity 30) enriched with f/2 medium and stored at 25°C in a thermostatic room under 100 μmol photons m^-2 s^-1 light intensity (light: dark=12h: 12h). After short-term UVR exposure, strains BS4F5 and DS4G4 were found to be sensitive to UVR, strain SE10 had moderate sensitivity, and strain SD4 was light tolerant ( Figure S.2 ). Strain WZD145 was obtained after the light exposure experiment. The results of these four morphotypes were significant and sufficient for proving diversity in their photo-physiological traits. Hence, the short-term UVR exposure was not repeated on the strain WZD145.

The five P. lima complex/P. caipirignum strains were treated under three different light conditions (PAR, PA, and PAB) for 18 days or 24 days. Cell division, photosynthetic status, and nutrient availability, were all dynamic during the batch culture. Together with strain-specific traits, these factors may have had impacts on the production of toxin. As a result, samples of cell density, nutrient concentration, photosynthetic response parameters, and toxin contents were taken every three days. This allowed for a more thorough description of the strain-specific traits and helped uncover the factors that drive toxin production with such adequate comparisons.

Chlorophyll a (Chl a) fluorescence, pigment and mycosporine-like amino acids (MAAs) content, which are types of UV-absorption compounds, were used to detect the photosynthetic responses. Maximum quantum efficiency (F_v/F_m ), rapid fluorescence transient (OJIP) curve, and non-photochemical quench (NPQ) were the parameters of Chl a fluorescence that, respectively, representing photosynthesis activity, electron transport between two photosystems, and photoprotection by energy dissipation. Additionally, measurements of the Chl a fluorescence’s excitation and emission spectra were made to reflect photosynthetic spectra. These chosen strains were then categorized according to their various photoacclimation characteristics.

With the exception of DST concentration per mL and per cell, the analysis of DST production focused more on the net toxin production rate (R _tox, pg cell^-1 d^-1), which eliminated the impact of cell division. To assess the effects of nutrients and growth on the production of DST, the nitrite and phosphate consumption rates (C _DIN and C _DIP, pmol cell^-1 d^-1) and growth rate (μ, d^-1) were calculated and correlated with R _tox. Finally, based on these variations in strain-specific traits, light treatments, growth rate, and nutrient consumption, the driving forces on R _tox were inferred.

Cells at the exponential growth phase were diluted 1:4 with an f/2 culture medium to a volume of 1.5 L in quartz tubes (Ø=10 cm, height 30 cm) in triplicate. The tubes were covered with different long-pass filters, including 395 nm filters (Ultraphan Digefra, Munich, Germany), 320 nm filters (Montagefolie, Folex, Dreieich, Germany) and 295 nm filters (Ultraphan, Digefra, Munich, Germany) and were exposed to artificial light sources (light: dark=12h: 12h), resulting in the PAR, PA (PAR+UVA), and PAB (PAR+UVA+UVB) treatment groups.

The PAR, UVA and UVB light sources were supplied by a light-emitting diode (Opple, China), UVA lamp (Philips, TL-D, German) and UVB lamp (Sankyo Denki, G15TBE, Japan), respectively. The light spectra were measured using a spectrometer (Hopoocolor, China) as previously described (Chen et al., 2021). The light intensities were measured using QSI2100 (Biospherical Instruments Inc, USA) and PMA2100 (Solar Light, USA) devices, which recorded the intensities of PAR, UVA and UVB as 300 μmol photons m^-2 s^-1, 4.8 W/m², and 0.05 W/m², respectively. The exposure to PAR, extra UVA and/or UVB produce different levels of photoinhibition in these strains. The light intensity settings for UVA and UVB permitted the survival of these strains. All strains were acclimated under each treatment for 24 days, except strain SD4, which was acclimated for 18 days. All strains were sampled every three days. Before sampling was conducted, cells at the bottom of each tube were resuspended using a magnetic stirrer controlled by a Teflon rotor.

Cell suspensions were fixed in final 1% Lugol’s solution, and cell counts (C_t ) were determined under a microscope. The specific growth rates (μ) of dinoflagellates were calculated using the following equation:

In Eq. 1, C₁ and C₂ represent the concentrations of cells at time 1 (t₁ ) and time 2 (t₂ ), respectively.

UVR-induced inhibition of growth was calculated using

In Eq. 2, C _P and C _X are the cell densities under the PAR and PAB treatments, respectively.

Rapid fluorescence transient (OJIP) curves were obtained for 15-min dark-adapted samples using a portable PAM fluorometer (Photon Systems Instruments, AP100, Czech Republic). The whole protocol lasted 2 s. The intensities of chlorophyll fluorescence at minimum (50 μs), J-step (2 ms), I-step (30 ms), and maximum were denoted as F₀, F_J, F_I, and F_M, respectively, and used for further calculations. V_J, the electron transport efficiency to plastoquinone A (Q_A), was calculated as (F_J-F₀)/(F_M-F₀). V_I, the electron transport efficiency from Q_A to plastoquinone B (Q_B), was calculated as (F_I-F₀)/(F_M-F₀) (Strasser et al., 2000). The OJIP curves were double-normalized as (F_t-F₀)/(F_M-F₀). Ft represented instantaneous chlorophyll fluorescence.

The Fluorescence profiles of the cell suspensions were captured with a fluorescence spectrophotometer (Hitachi, F-4600, Japan) at a scan rate of 120 nm min^-1. To determine the chlorophyll fluorescence when excited by the UVR waveband, excitation spectra were collected for samples at an emission Λ of 685 nm (excitation Λ range = 280-400 nm). A Λ650 nm long-pass filter was employed to prevent the entry of any excitation wavelengths shorter than Λ650 nm. The excitation spectra were calibrated using a Rhodamine B solution. Fluorescence emission spectra were collected from samples using an excitation Λ of 360 nm (emission Λ range = 400-700 nm).

Samples of 15 mL cell suspensions were frozen with liquid nitrogen and stored at -80°C before analysis. Thawed samples were concentrated on 25 mm GF/F filters (Whatman, UK) and extracted in 1.0 mL 90% acetone (HPLC grade, Sigma, USA) at 4°C for 12 h in the dark. The extracts were then filtered through 25 mm 0.2 μm Nylon filters to remove cell and filter debris. 100 μL of each sample was injected into an HPLC system (Agilent 1200, USA) with a ZORBAX SB-C₁₈ column (5 μm, 4.6×150 mm, Agilent, USA). Pigment separation was performed according to Barlow et al., 1993. Chlorophylls and carotenoids were detected using diode-array spectroscopy at 440 nm. Pigment peaks were identified and quantified by comparing retention times and contents with those of pure standards: chlorophyll a, peridinin, diadinoxanthin, diatoxanthin, chlorophyll c₂, and β-carotene (ChromaDex, USA).

Samples of 15 mL cell suspensions for the analyses of MAAs were concentrated onto 25 mm GF/F filters (Whatman, UK) and extracted in 2.0 mL of 20% methanol (HPLC grade, Sigma-Aldrich, USA) for 24 h at 37°C in the dark. The methanol extracts were frozen to dryness and reconstituted in 1 mL of 100% methanol. The extracts were evaporated to dryness at 45°C. The residues were re-dissolved in 200 μL of Milli-Q water and filtered through 0.2 μm pore syringe filters for HPLC analyses. 100 μL of each sample was injected into an HPLC system (Agilent 1200, USA) with a ZORBAX SB-C₁₈ column (5 μm, 4.6×150 mm, Agilent, USA). MAAs were separated using a solvent gradient system, as described by Taira et al. (2004). MAAs were determined using absorbances at 310 nm and 334 nm and scanned absorption between 280 to 400 nm. The MAA extracts were identified per the retention time and maximum absorption wavelength, being further verified using liquid chromatography coupled to tandem mass spectrometry using LC-ESI-Q-TOF-MS (LC-MS, TripleTOF 5600⁺, AB SCIEX, USA) (Lalegerie et al., 2019). Due to the lack of commercial standards for MAAs, the extinction coefficients ϵ (mol L^-1 cm^-1) at the wavelengths of maximum absorption of the identified MAAs as referred by Taira et al. (2004) were used, for shinorine (λ_max = 334 nm, 44,700), porphyra-334 (λ_max = 334 nm, 43,200), palythine (λ_max = 320 nm, 36,200), mycosporine-glycine (λ_max = 310 nm, 28,100), and palythene (λ_max = 360 nm, 50,000). The MAAs concentrations (C, mol mL^-1) were calculated using the method by Torsi et al. (1990).

In Eq. 3, F represents the flow rate (mL min^-1), S is the area of peak (AU×min), ρ (cm) is the cell thickness, and V (mL) is the injection volume.

The culture solutions after GF/F filtering were submitted to flow injection analysis (HACH, QC8500 S2, USA) to analyze the concentrations of nitrate and phosphate, which were measured using cadmium column reduction and phosphomolybdenum blue colorimetry. The nitrate and phosphate contents were calibrated with standard materials purchased from the National Marine Environmental Monitoring Centre (NMEMC, China). The nutrient consumption rates (C _nutrient) were calculated using Eq. (4):

N_t represents the nutrient concentration at the time point. Nitrogen and phosphate consumption rates were represented as C _DIN and C _DIP, respectively.

Samples of 30 mL cell suspensions for the analysis of DST were harvested using centrifugation at 1250 g for 10 min and reconstituted in 500 μL methanol. The methanol extracts were sonicated in an ice bath for 20 min and filtered through pore syringe filters (0.22 μm). DSP toxins were analyzed using liquid chromatography coupled to tandem mass spectrometry (LC-MS/MS, QTRAP 4500, AB SCIEX, USA) as described in Krock et al. (2008). OA and DTX-1 were quantified by external calibration using seven gradients of mixed standard solutions purchased from the National Research Council (NRC, Canada).

The cellular toxin content resulted from the concentration of toxin, T_t (amount of toxin mL^-1), divided by C_t (cells mL^-1). To eliminate the effect of cell division on toxin production, the net toxin production rate R _tox (toxin cell^-1 d^-1) was determined using Eq. 5 (Anderson et al., 1990):

where C ¯ is the average cell density, as per Eq. 6:

All statistical analyses were performed using SPSS, version 25.0 (IBM, USA), and triplicate measurements’ data are expressed as the average ± standard deviation (SD). Analysis of variance (two-way ANOVA) followed by Tukey’s post hoc HSD test was used to examine differences between treatments. The relationship between R _tox and C_DIP was analyzed using linear or exponential regression in the OriginPro 2021 software. Linear relationships were examined using Linear Fit with X Error, and Least-squares Linear regression was performed using the computation method of York. Non-linear curves were fitted using the Non-Linear Curve Fit tool, and exponential regression was performed with Orthogonal Distance Regression (Pro), which supports both x and y weights. The x and y errors were weighted by 1/σ_i ², where σ_i is the error value. The data for cell density, toxin content and nutrient concentrations from previous studies were obtained using the Digitize Image function of OriginPro 2021. The R _tox values were calculated according to Eq. 5. The relationships between OA and DTX-1 among the strains from the current and previous studies were examined independently using the Linear Fit tool of OriginPro 2021.

The cell density steadily increased from day 0 to 18 for all cases and then either slowly increased or did not increase for all morphotypes under PAR light conditions ( Figure 1 ). The growth rates from day 0-18 under PAR varied from 0.087 d^-1 (SD4) to 0.123 d^-1 (SE10, Figure S.3 ). UVA (PA compared to PAR) and UVR (PAB compared to PAR) comparisons showed that both had effects on growth. UVA addition significantly enhanced the final cell densities of strains BS4F4, WZD145 and strain SE10 by 26.6%, 35.1%, and 40.3%, respectively, on day 24 (one-way ANOVA, p< 0.05) but did not affect those of strains SD4 and DS4G4 ( Figure 1 ). The final cell densities of strains BS4F5 and DS4G4, sensitive to short-term exposure of UVR ( Figure S.2 ), were also decreased 26.5% and 76.8% by UVR, respectively, while other strains were not affected (one-way ANOVA, p > 0.05).

Regarding macronutrient levels during the culture, nitrogen was not depleted in the medium for any strains during the experiment, with the lowest concentration of 6.3 μmol mL^-1 in strain SE10 under PAB on day 24 ( Figure 1 ). Phosphate was depleted in the medium or was not consumed in strain SD4 under PA and PAB in the middle stage of culture (approximately on day 12). An exception was strain DS4G4 under PAB, which still had 5.9 μmol P mL^-1 on day 24 ( Figure 1 ). The time of phosphate depletion corresponded with the second decrease in F_v/F_m , reflecting photosynthetic activity. The first decrease of F_v/F_m occurred on day 3. Then, F_v/F_m started to recover in all cases except strain DS4G4 under PAR, which started on day 6, and it under PAB, which failed to recover ( Figure 1 ).

The five strains had similar responses regarding Chl a fluorescence and pigment ratios, so they were pooled to analyze the effects of the varying light treatments. The excitation spectra showed that Chl a fluorescence (685 nm) was mainly excited at the UVA waveband (320-400 nm, Figure S.4 ). Chl a, which yields fluorescence at 685 nm, was excited by light at 360 nm. Thus, UVA could power photosynthesis in P. lima complex and P. caipirignum ( Figure 2A , for specific variances among strains, see Figure S.5 ). After acclimation to UVA addition (PA), Chl a fluorescence at 685 nm was enhanced. In contrast, UVB addition (PAB) caused fluorescence at 685 nm to be reduced to the level of PAR condition ( Figure 2A ).

The OJIP curve, which indicates electron transport between PSII and PSI, had decreased J-P and I-P steps when UVA was added (both PA and PAB, Figure 2B ), meaning acceleration of electron transport. Throughout the culture period, the median value of V_J (O-J step) is decreased only in strains BS4F5 and SE10, whereas V_I (J-I step) is decreased in each morphotype under UVA addition except in strain DS4G4 ( Figure S.6 ). However, this decrease was only statistically significant for V_I of strain SE10 (two-way ANOVA, p< 0.05). V_J is not affected by light treatment when the strains are pooled together, whereas V_I is significantly decreased by UVA addition (one-way ANOVA, p< 0.001, Figure 2B ).

The UVA addition significantly reduced the ratios of the photosynthetic accessory pigment peridinin (Peri) to Chl a (fmol: fmol) when pooled together (one-way ANOVA, p< 0.001, Figure 2C ) and when independent (two-way ANOVA, all morphotypes except strain SD4, p< 0.05, Figure S.7 ). UVA also decreased the ratio of photoprotective pigments diatoxanthin (Dt) and diadinoxanthin (Dd) to Chl a. Dt/Chl a decreased significantly in strains WZD145 and SE10 (two-way ANOVA, p< 0.05, Figure S.7 ) and pooled together (one-way ANOVA, p< 0.001, Figure 2C ). Dd/Chl a did not affect by the UVA addition in the pooled ratio (one-way ANOVA, p > 0.05, Figure 2C ) or in independent strains (two-way ANOVA, p > 0.05, Figure S.7 ), except for strain BS4F5. β-caro/Chl a and Chl c₂/Chl a were not influenced in an independent strain by light environments (two-way ANOVA, p > 0.05, Figure S.7 ).

Five individual MAAs, mycosporine-glycine (My-G), palythine (Pali), palythene (Pale), shinorine (Shi) and porphyra-334 (Por), were detected in the five strains, except strain DS4G4 where Pali and Pale were not detected ( Table 1 ). Among these MAAs, porphyra-334 was the most abundant. The average total MAA content was significantly increased (2.7- to 6.6-fold) with the addition of UVA and UVB, with strain DS4G4 showing the highest increase. Strain DS4G4, however, had the lowest average MAA content within the PAB treatment (two-way ANOVA, p< 0.05), which was 3.5-fold lower than the highest MAA content (BS4F5).

For strains BS4F5 and DS4G4, both the de-epoxidation rate [Dt/(Dd+Dt)] and reaction-driven NPQ are noticeably lower than the other three sample groups (two-way ANOVA, p< 0.05, Figure 3A , indicated in red), while UVR inhibition of growth mostly exceeded 20% ( Figure 3B ). The increases in NPQ for strain DS4G4 under the PA and PAB treatments were not supported by the Dt/(Dd+Dt) data, which did not change with the UVA or UVR addition. Thus, NPQ in PA and PAB of strain DS4G4 is possibly induced by photodamage ( Figure 3A ). Principal component analysis suggested that responses to UVR among these strains could be divided into three groups: (i) intermediate MAA contents and high NPQ and Dt/(Dd+Dt) with low UVR inhibition on growth (WZD145, SD4 and SE10); (ii) the highest MAA content but low NPQ and Dt/(Dd+Dt) with intermediate UVR inhibition (BS4F5); and (iii) the lowest MAA content and low Dt/(Dd+Dt) with the highest UVR inhibition (DS4G4) ( Figure 3C ).

The maximum net toxin production rate (R _max) across the cultures evaluates the capacity of toxin production among strains. The R _max of DTX-1 ranged over three orders of magnitude, from 0.001 to 1 pg cell^-1 d^-1 in the five strains tested ( Figure 4A ). The strains are divided into three levels according to the magnitude of differentiation of the R _max of DTX-1.

The high level (HL, beyond 0.1 pg cell^-1 d^-1) of R _max of DTX-1 includes strains SD4 and DS4G4, with R _max for DTX-1 being 0.14-0.36 pg cell^-1 d^-1 and 0.18-0.23 pg cell^-1 d^-1 under light treatments, respectively. HL also contains the previously reported OMI29P, MIO12P, and AOF55P strains (Hashimoto et al., 2021). In particular, the strain MIO12P reached a magnitude above 1 pg cell^-1 d^-1, with a R _max of DTX-1 of 1.23 pg cell^-1 d^-1 on day 17 in the SWII&GJE1 medium.

The strain WZD145 has 0.078 ± 0.001 pg cell^-1 d^-1 R _max of DTX-1 on day 6 under PAR, with varying performances of R _tox on DTX-1 over time from the HL groups ( Figure 5B ). Although the rate is enhanced by UVA to 0.112 ± 0.001 and 0.111 ± 0.001 pg cell^-1 d^-1 on day 6-PA and -PAB, respectively, the R _max of DTX-1 in strain WZD145 could be categorized into the middle level (ML, 0.01-0.1 pg cell^-1 d^-1). ML also included the previously reported LM001 and LM003 strains (Moreira-González et al., 2019).

The low level (LL, 0.001-0.01 pg cell^-1 d^-1) of R _max on DTX-1 includes strains BS4F5 and SE10, with R _max for DTX-1 ranging from 0.0029-0.0065 pg cell^-1 d^-1 and 0.0015-0.0031 pg cell^-1 d^-1 under the light treatments, respectively.

The R _max of OA in the examined morphotypes displayed an extensive variance from 0.74 to 2.04 pg cell^-1 d^-1. The strain DS4G4 has the highest R _max for OA (1.76-2.04 pg cell^-1 d^-1), following strain BS4F5 (0.94-1.41 pg cell^-1 d^-1), SE10 (0.7-0.74 pg cell^-1 d^-1), and SD4 (0.51-0.81 pg cell^-1 d^-1), with strain WZD145 being the lowest (0.34-0.56 pg cell^-1 d^-1). Together with all previously reported strains, although different nutrient treatments and culture mediums influenced the R _max for OA, none of the R _max for OA exceeded 2.5 pg cell^-1 d^-1.

Positive linear correlations between R _tox of OA and DTX-1 were observed in all HL strains and the two ML strains of LM001 and LM003 ( Figure 4C and Table S.1 , R ² varied from 0.663 to 0.992). The fitting failed for ML strain WZD145 (R ² = 0.261) and two LL strains BS4F5 and SE10 (R ^2< 0.2). The data (pg: pg) between R _tox of OA and DTX-1 are variant in these strains, ranging from 1.15 ± 0.04 (MIO12P) to 24.52 ± 4.98 (LM001). The examined strains SD4 and DS4G4 had 7.93 ± 0.78 and 2.59 ± 0.25 production segments between OA and DTX-1, respectively. The data for fitting came from various light treatments or culture mediums over time. The linear correlation between the R _tox of OA and DTX-1 suggested that internal factors regulate the biosynthesis of the two compounds.

The OA and DTX-1 concentrations (ng mL^-1) generally increased over time in all strains ( Figure S. 8 ). The toxin levels would, however, occasionally drop. For example, OA and DTX-1 in strain DS4G4 on days 18 and 24 decreased. The OA concentration of strain SE10 initially increased 16.6-fold above 9 ng mL^-1 on day 6 and then abruptly decreased 22.6-fold below than 1 ng mL^-1 on day 9, remaining below 2.1 ng mL^-1. Except for DTX-1 of two LL strains BS4F5 and SE10 at some time points, light treatments did not change the trends in toxin concentration. However, the accumulation of toxic concentrations is influenced by light. For instance, the OA and DTX-1 concentrations of the most light-sensitive strain DS4G4 decreased considerably by adding UVB between days 12 to 24.

On the cellular toxin levels (pg cell^-1), no apparent or comparable trends were found ( Figure S. 9 ). After eliminating the effects of growth rate, the net toxin production rate trends (R _tox, pg cell^-1 d^-1) seemed to be more regular than those for cellular toxin content ( Figure 5 ). The combined effects of light treatments, traits of strains, toxin species, dynamic photosynthetic activity and nutritional availability during the culture process are involved in R _tox, resulting in it becoming a complex relationship to characterize.

The R _tox initially increased and then decreased, sometimes repeatedly, which shaped significant peaks or troughs. These inflection points might result from shifts in factors that direct R _tox. To simplify the analysis, the overall increase and decreasing process of R _tox was referred to as a period and was represented by a two-way arrow in Figure 5 . The time of the identified peak (indicated by a dotted line in Figure 5 ) during a period was listed in Table 2 . The different features in the examined strains, such as the light response and DTX-1 production ability, were also listed in Table 2 .

The first R _tox peaks in all sample groups varied between days 3 to 12 ( Table 2 ). All second peaks occur after day 12, except for the BS4F5-OA-PAR, whose second R _tox peak occurs on day 12 ( Table 2 ). Surprisingly, the first peaks of R _tox on OA in the HL (SD4 and DS4G4) and ML (WZD145) under the three light treatments occur on day 12 ( Figure 5A and Table 2 ). In contrast, two LL strains BS4F5 and SE10 showed earlier peaks occurring on days 3 or 6, depending on the light treatments. The peak value for R _tox was not the factor affecting the peak time because the peak values in LL were either higher or lower than those in HL ( Figures 4B , 5A ). The considerable variances of DTX-1 production capacity may cause the different R _tox peak times on OA of the LL strains (BS4F5 and SE10) than other strains ( Figure 4A ). Furthermore, the first peak times for R _tox on DTX-1 in ML and LL strains were also earlier than in the HL on day 12 ( Table 2 ). These findings show that strains with reducing DTX-1 production capability more easily reached their maximum DST production rates.

When comparing the peak times unchanged by light treatments in the HL strains (SD4 and DS4G4), the peak times of LL (BS4F5 and SE10) and ML (WZZD145) were more plastic and could be delayed or accelerated by adding either UVA or UVB, causing peak separation ( Table 2 and Figure 5 ). Normally, the delayed peaks in the LL strains were typically observed at the beginning, such as SE10-OA-PA &PAB on day 6 and SE10-DTX1-PAB on day 9, where R _tox was stagnant prior to the peak. UVA addition and, on occasion, UVB addition shortened the sustained increases under PAR and accelerated the peak times observed for LL and ML strains.

Stagnation of R _tox at the beginning was quite common, which was also observed in the BS4F5-OA-all light treatments and WZD145-DTX1-PA & PAB ( Figure 5 ). Additionally, at the beginning, the R _tox of some cases was negative, like WZD145-OA-PA & PAB and DS4G4-OA & DTX1-all light treatments. Whether R _tox is stagnant or negative at the beginning, which may arise for the same reason, they were represented by red right arrows in Table 2 . The resulting R _tox values from the light-tolerant strains including WZD145, SD4 and SE10 are not inhibited under PAR, whereas those from the light-sensitive strains BS4F5 and DS4G4 were all inhibited to almost nil or negative under the three light treatments. These trends demonstrate that inhibition of R _tox by light was relevant to the photoprotection ability.

Meanwhile, photosynthetic activity F_v/F_m also decreased at the beginning of the experiment ( Figure 1 ), indicating cells were experiencing a photoinhibition status. UVA or UVB addition increases the degree of photoinhibition. The strain DS4G4 was the most sensitive to UVR. On day 3, the R _tox of strain DS4G4 for DTX-1 yielded the following order: PAR, PA, and PAB ( Figure 5B ). The differences between the treatments were each significantly different (one-way ANOVA, p< 0.05, Figure 5B ). On day 9, the F_v/F_m of strain DS4G4 under PAR stopped decreasing ( Figure 1 ). At the same time, the R _tox on OA changed from negative to positive ( Figure 5A ). Therefore, the inhibition of DST production was caused by photoinhibition, and could be alleviated by photo-accumulation process.

In the LL (BS4F5 and SE10) and ML (WZD145) strains, the accelerated peaks as a result of UVR were observed in not only the first period but also the second period ( Figure 5 ). Especially in the first period, UVA significantly increases the maximum R _tox of OA in strain BS4F5 on day 3, and of DTX-1 in strains WZD145 (under PAR-day 6 versus PA and PAB-day 6) and SE10 (under PAR-day 6 versus PA-day 3 and PAB-day 9), respectively (two-way ANOVA, p< 0.05, Figures 4A and 5B ). In contrast, PAB results in two HL strains SD4 and DS4G4, having significantly lower peak levels of OA and DTX-1 on day 12 (one-way ANOVA, p< 0.05, Figure 5 ). But for OA and DTX-1 in strain DS4G4, the second peak on day 21 under PAB was the highest (one-way ANOVA, p< 0.05, Figure 5 ).

The advanced and increased peaks of R _tox induced by the addition of UVA in LL (BS4F5 and SE10) and ML (WZD145) strains resulted in their growth also being enhanced with the UVA addition ( Figure 1 and Table 2 ). UVA could enhance the photosynthesis levels ( Figure 2A ). Based on the earlier peaks under the UVA addition, the increasing processes of R _tox suggested higher photosynthetic yields for DST production.

However, the decreased R _tox remained unexplained. The earliest decreasing R _tox began on day 3, and the latest was on day 12. R _tox shows two decreasing patterns: i) R _tox decreases quickly and linearly, as in all OA cases, and DTX-1 in the HL strains (SD4 and DS4G4) and strain BS4F5, and ii) R _tox decreased slowly at first and then quickly for DTX-1 in the ML and LL strains, such as strains WZD145 and SE10 under PAR.

The troughs of R _tox were observed between days 9 to 18 in the middle of the experiment. The troughs appeared close in time to the depletion of phosphate and the second decrease of F_v/F_m , indicating limited nutrients available for photosynthesis. The phosphate consumption rate (C _DIP) was calculated to explore the relationship between toxin production and phosphate consumption.

Except for a notable increase in DS4G4-PAB on day 21 and drops and increases in BS4F5-PA and -PAB between days 3 to 6, C _DIP consistently decreased over time ( Figure S.10 ). Two different relationships between R _tox and C _DIP were observed, the inverse and positive relationships ( Figure S.11 ). R _tox was more strongly correlated with C _DIP than with nitrogen consumption (C _DIN, Figure S.12 ) or growth rates (μ, Figure S.13 ). The last two correlations seemed to be irregular. The pre-peak (increasing) R _tox, in the cases of sufficient points for fitting showing in Figure S.11 , had inversely linear correlations with C _DIP. In contrast, the post-peak (decreasing) R _tox in the cases of strains WZD145, SE10 and DS4G4 with sufficient points for fitting, had positively linear or exponential correlations with C _DIP (specifically presented in Figure 6 ). The functional parameters and statistical results of the linear and exponential correlations are listed in the Tables S.2 and S.3 . The coefficient of determinations, denoted by Pearson’s r and R square, the majority of which were above 0.98 ( Tables S.2 and S.3 ), indicate a strong correlation between R _tox and C _DIP.

Regardless of when the R _tox peak starts to decrease, periods, toxin species, or morphotypes with different DTX-1 production capacities, all downward R _tox decreased with C _DIP ( Figure 6 ). In the HL strain DS4G4 and ML strain WZD145, the peak time of R _tox on OA was not affected by the light treatments. Therefore, their R _tox on OA from different light treatments could be commonly fitted except DS4G4 under PAB due to strong inhibition on R _tox by adding UVB. The R _tox on OA in strain DS4G4 (maximum R _tox: 2 pg cell^-1 d^-1) was more saturated than in strain WZD145 (0.56 pg cell^-1 d^-1), corresponding to an exponential correlation with C _DIP in strain DS4G4 and a linear correlation with C _DIP in strain WZD145.

With reduced DTX-1 production capacities, R _tox for DTX-1 in ML strain WZD145 and LL strain SE10 under PAR, displaying the pattern II of decreasing, had saturated curves with C _DIP. The addition of UVA prompted the maximum R _tox and changed the outcome to linear lines. The linear correlations are most likely the initial part of the saturation curve. Furthermore, the uptrending R _tox of DS4G4-PAB on day 21, the only point of PAB significantly higher than the other light treatments, met the only increase in C _DIP and was located on the positive line between R _tox and C _DIP ( Figure S.10 , Figure 6 ). The relationship between C _DIP and R _tox is likely to be one of substrate-product.

Physiological studies of whole species inevitably require extrapolating from data at the strain level. However, conclusions drawn this way are sometimes insufficient or subject to high variation if the genetic distances between chosen strains are too close or the number of strains is insufficient (Schaeffer et al., 2007; Thessen et al., 2009). In particular, high genetic variability within a species (Burkholder and Glibert, 2006) is considered a sign of a cryptic species or functional diversity.

Though the five morphotypes of P. lima complex and P. caipirignum had similar photo-physiological responses to exposure to UVA, resulting in enhanced efficiency of photosynthesis ( Figure 2 ), they exhibited high diversity in photoprotection, especially strain DS4G4, which had extremely low NPQ and MAA content. NPQ, as well as the xanthophyll cycle indicated as Dt/(Dd+Dt), represents an efficient photoprotective mechanism that releases excess light energy as thermal energy and avoids damage to photosynthetic apparatuses. The MAAs absorb UVR with extremely high extinction coefficients and are essential to the survival of phytoplankton that experience UVR exposure. Thus, in outdoor cultures, strain DS4G4 could not survive under solar light (data not shown). Accordingly, it is reasonable to infer that strains from morphotype 5 would prefer deep water or conditions where light is obscured.

The collective environment of strain DS4G4 was not described by Zhang et al. (2015). According to the phylogenetic tree constructed using ITS and LSU rDNA sequences, strain LM001 shares the same clade as strain DS4G4 (Moreira-González et al., 2019). Strain LM001 was collected from a bloom of macroalgae Ulva lactuca in 2014. It was also found in the muddy tidal flats of the outer sector of the estuarine complex of Paranaguá Bay, Brazil, both of which exhibit obscured conditions (Moreira-González et al., 2019). While investigating vertical distributions in Japan (Nishimura et al., 2020), strains from P. lima complex subclade 1a and P. caipirignum subclades b and e, including light-tolerant strains WZD145 and SE10 in this study, were distributed only shallow (< 5m depth) water and in shallow to deep water (10-30 m depths), respectively, whereas P. lima complex subclades 3b and 3c, including light-sensitive strain BS4F5, were restricted to deep water. Overall, the performances of these morphotypes of the P. lima complex and P. caipirignum under UVR addition were all highly relevant to their adaptation to their respective field light conditions.

Benthic dinoflagellates exhibit certain benthic or epibenthic behaviours, such as relatively fixed life history modes, which exposes them to very different light environments due to physical shadowing by macroalgae (Nakahara et al., 1996; Villareal and Morton, 2002), sediments or sand (Reñé et al., 2021), tidal rhythms (Patil et al., 2017), and light attenuation in the water column. Variation in light conditions presents different selective pressures that influence photosynthetic optimization and, thus, long-term competitive adaptation. Therefore, the P. lima complex and P. caipirignum cells must continually adjust to variable intensities and wavelengths of light via photoprotection mechanisms as in light-tolerant strains or otherwise be restricted to narrowly defined light regimes as ‘ecotypes’, as in light-sensitive strains BS4F5 and DS4G4. From an ecological perspective, the divergence of multiple ecotypes with different photosynthetic physiologies allows them to better adapt to dynamic light environments and inhabit different depths of the water column and therefore enhance the competitive advantages for the population (Gallagher, 1982; Cohu et al., 2012; Suggett et al., 2015; Nishimura et al., 2020).

DTX-1’s production capacity was another factor that differed significantly between the P. lima complex and P. caipirignum strains, where the DTX-1 production capacity ranged across four orders of magnitude ( Figure 4A ). The R _tox of OA reached its maximum more easily when the DTX-1 production capacity was extremely low, such as in two LL strains (BS4F5 and SE10) ( Table 2 ). Otherwise, when the DTX-1 production capacity increased by one or two orders of magnitude, the R _tox between OA and DTX-1 experienced a linear correlation ( Figure 4C ). However, this linear relationship was not conserved, such as in ML strain WZD145, where R _tox of DTX-1 showed saturation (steady R _tox) and then lost linear relationship with the R _tox of OA ( Figures 4C and 5 ). This data infers that the relationship between OA and DTX-1 is competitive.

Nishimura et al. (2020) reported that three groups might be based on their OA and DTX-1 cellular content. In group I, DTX1 producers included P. lima complex subclades 1e and 1i, representing the MIO12P (HL) and AOF55P (HL) strains. In group II, low DTX-1 producers included P. lima complex subclades 1a and 1f, representing strains WZD145 (ML) and OMI29P (HL). In group III, no or very low DTX1 producers included P. lima complex subclades 1c, 1d, 1f, 1g, 1h, 1j, 3a, 3b, and 3c; Prorocentrum sp. Type 2; and P. caipirignum subclades b and e. The strains BS4F5 (LL) and SE10 (LL) were proposed as P. lima complex clade 3 and P. caipirignum subclade c by Nishimura et al. (2020), respectively. These two morphotypes could be regarded as group III based on their minimal DTX-1 content.

In groups I and II, the OA and DTX-1 content of the strains experienced linear relationships ( Figure 4C ). Such a relationship infers that the strains of these two groups follow a specific ratio between OA and DTX-1 to produce toxins most of the time, but not always (like WZD145). The following work, per Hashimoto et al. (2021), reported the toxin contents of three HL strains MIO12P, AOF55P, and OMI29P over time in the SWII & GJE1 culture medium. After the toxin contents were recalculated as R _tox, the strain OMI29P with the lowest R _max of DTX-1 among these three strains had the earliest OA peak on day 6. The AOF55P strain was next on day 9 and MIO12P strain on day 17, with the middle and highest R _max of DTX-1, respectively ( Figures 4A, B and S.13 ). Furthermore, in a previous study, toxin concentrations reported during the cell cycle suggested that OA and DTX-1 have separative transformation pathways (Pan et al., 1999). This outcome supports the hypothesis that production between OA and DTX-1 competes against each other.

According to the investigation of Nishimura et al. (2020), many more subclades of P. lima complex or P. caipirignum reduced the DTX-1 production capacity. In contrast, all reported strains of P. lima complex and P. caipirignum could reach relatively high R _tox of OA (more than 0.4 pg cell^-1 d^-1, Figure 4B ), indicating that OA might be more important than DTX-1 for DST producers in Prorocentrum. The fast accumulation of OA may be a selective advantage in reducing the DTX-1 production capacity. The function of DST for toxin producers is still unclear. Though OA and DTX-1 could inhibit other non-DST-producing microalgae (Windust et al., 1996), OA was not a primary growth-inhibitory compound in the medium of P. lima complex (Sugg and VanDolah, 1999).

An increase in light intensity inhibits or reduces DST content or production in both Prorocentrum and Dinophysis genera, even though the light intensity did not reach the saturation intensity for growth (Morton et al., 1992; Tong et al., 2011; López-Rosales et al., 2014; Praptiwi, 2014). A stagnant or even negative R _tox occurred from alteration of light conditions, associated with morphotype photoprotection to reduce light photodamage. Additionally, negative R _tox occurred not only during the change of light conditions but also when phosphate and nitrogen were limited ( Figures 5 and S.13D ), indicating that the same mechanism may cause these negative R _tox values.

According to the theory of photoinhibition, F_v/F_m occurs due to the photodamage and repair processes of photosystem II (PSII) (Takahashi & Murata, 2008). Photodamage is linearly related to light intensity, and each wavelength of light has a specific photodamage rate (Tyystjärvi & Aro, 1996). In contrast, environmental factors such as heat, osmotic stress, or nutrient limitation impair the repair process (Takahashi and Badger, 2011; Li et al., 2021). Photoinhibition occurs when F_v/F_m declines when the repair rate is lower than the PSII damage rate (Allakhverdiev & Murata, 2004). Photoinhibition is caused by UVR exposure and nutrient limitations for photodamage and repair processes. Accordingly, the inevitability of photodamage could be why the maximum R _tox decreased in D. acuminata (Tong et al., 2011) and the cellular toxin content reduced in Prorocentrum species, even when light levels were still low (Morton et al., 1992; López-Rosales et al., 2014; Praptiwi, 2014).

A follow-up experiment (Exp. F) was conducted on two morphotypes with different photosynthetic traits but similar R _tox trends (such as HL strains SD4 and DS4G4) under four light intensities varying between 12 to 400 μmol photons m^-2 s^-1. The initial R _tox values decreased exponentially from positive to negative with light intensities from 12 to 400 μmol photons m^-2 s^-1 (unpublished), demonstrating light’s inevitable inhibition effect on DST production under variable light conditions. DST production may be associated with the repair process of PSII. Thus, there is a contradiction, assuming DST production requires light (Pan et al., 1999; Tong et al., 2011) during the light phase (Pan et al., 1999; Jia et al., 2019) but is then inhibited by light.

Glycolate, as the starting unit of polyketide compounds, is only observed in dinoflagellates (Van Wagoner et al., 2014). For instance, it is seen in OA and DTX-1 (Needham et al., 1994) and the side chain of its diol esters DTX-4 and DTX-5a/b (Macpherson et al., 2003), and other polyketides such as amphidinol produced by Amphidinium carterae (Cutignano et al., 2017) and yessotoxin produced by Protoceratium reticulatum (Yamazaki et al., 2011). When oxygen concentrations increase, the ribulose-1,5-bisphosphate carboxylase/oxygenase (Rubisco) enzyme leads to the formation of both 3-phosphoglycera (PGA) and glycolate (Moroney et al., 2013). Glycolate then plays a role in the photorespiration pathway and exerts carbon and energy loss effects and photoprotection (Crawley et al., 2010). Glycolate in the P. lima complex participates in another donation pathway by converting pyruvate to hydroxypyruvate in the tricarboxylic acid (TCA) cycle (Needham et al., 1994). As it plays such a unique and essential role in the biosynthesis of DST and as a complex carbon source, whether it plays a crucial role in the regulation relationship between photoinhibition and DST production requires further study.

The increasing R _tox for OA and DTX-1 was observed in many strains of P. lima complex and P. caipirignum. This process is relevant to increasing photosynthetic yields for DST production because the increasing processes of R _tox in LL and ML strains are shorter than in HL, meaning that decreasing demand for DST production could reduce the increasing process. Accordingly, reducing the increasing process for DTX-1 in LL and ML strains under UVA addition was considered the result of increasing photosynthetic yield for the DST production. In the Exp. S (mentioned above), the increasing rate of R _tox was enhanced significantly by increasing light intensity and displayed an exponential correlation with light intensity (unpublished).

In Dinophysis species, mixotrophy helps provide additional carbon sources (Tong et al., 2015). DST production in Dinophysis species did not show a period of increase (Tong et al., 2011; Nielsen et al., 2013; Basti et al., 2018) before reaching a maximum OA R _tox of approximately 8 pg cell^-1 d^-1 (Nielsen et al., 2013). In contrast, the maximum OA R _tox in Prorocentrum species did not surpass 2.5 pg cell^-1 d^-1 ( Figure 4B ). Subsequently, the poor photosynthetic yield for DST production was correlated with carbon sources.

The carbon atoms for DST are sourced from intact or cleaved acetate units, except for glycolate, the starter unit. There is indirect evidence for enhancing DST production in mixotrophs in the P. lima complex. The addition of glycine, incorporated into DST production at C-37, C-38, C-45 and C-46 through conversion into acetate and glycolate, causes an increase in biomass. The increase is twice as much as the acetate addition in the P. lima complex (Needham et al., 1994). Further, the addition of amino acids that can be transformed into glycine or acetyl-CoA improved the maximum R _tox or yield of OA; in particular, lysine and serine resulted in approximately 1.5-fold and 1.6-fold improvements in the maximum R _tox and the total OA yield compared to the references ( Figures 4B and S.13E , Souto et al., 2001). If identified, a mixotroph with putative extra carbon addition would eliminate the acceleration process and improve the maximum R _tox values, providing benefits for gaining DST from Prorocentrum species, which has an extended culture period and relatively low DST production rate (Camacho-Muñoz et al., 2021).

The timing of the R _tox trough coincided with the deficiency of available phosphate ( Figures 1 , 6 ). Notably, all decreasing R _tox rates positively depended on C _DIP ( Figure S.11 ). The well-fitted linear or exponential correlations in strains WZD145, SE10 and DS4G4 reflected a supply and demand link from phosphate availability to DST production ( Figure 6 ). Furthermore, an unusually high R _tox in DS4G4-PAB on day 21 corresponded with an increased C _DIP and is located on the positive line with other decreasing R _tox values ( Figures 5 , 6 ). While further confirmation is needed, an inference about the high phosphate demand of DST production seems reasonable and could explain the complex trends in R _tox values.

Previous studies have found that deficiencies in available phosphate and nitrogen increase toxin levels (McLachlan et al., 1994; Varkitzi et al., 2010; Hou et al., 2016; Gu et al., 2019). For instance, the OA cellular levels increased by 1.9-fold and 2.3-fold under 1/50-N and 1/20-P of the f/2 concentration than under f/2 conditions, respectively (Vanucci et al., 2010). However, low nutrient concentrations also slowed down the growth rate. It is unclear whether macronutrient levels directly or indirectly affect DST production.

Previous studies also indicate that cellular DST levels in the P. lima complex at the stationary growth phase are higher than at the exponential growth phase (reviewed by McLachlan et al., 1994; Quilliam et al., 1996; Nascimento et al., 2005; Varkitzi et al., 2010; Lee et al., 2016). However, this finding has only been observed in specific strains of the P. lima complex, such as the CCAP 1136/11 and CCMP2579 strains. Investigations into DST production in strains isolated from different areas and representing different phylogenies of the P. lima complex show that the highest cellular toxin levels occur before the stationary growth phase in the LM001-3 strain isolated from Cuba and Brazil (Moreira-González et al., 2019). It also occurred before the SHG101, QD502, and XS336 strains, which were isolated from China (Wu et al., 2020), and the BS4F5, SE10, and DS4G4 strains in this study ( Figure S.9 ). Moreover, DST production in D. acuminata was found to mainly occur during the exponential growth phase (Tong et al., 2011), suggesting that limitation of growth is not required for DST production. Considering the constantly increasing process of R _tox in many strains of P. lima complex, the highest toxin content at the stationary phase might result from increasing R _tox values and decreasing growth rates.

D. acuminata does not utilize dissolved nitrate or phosphate, and its OA and DTX-1 production rates are directly associated with the availability of prey that has consumed these nutrients, which indirectly influences toxin production (Tong et al., 2015). The addition of organic substances obtained from sonicating ciliate Mesodinium rubrum (as prey) significantly enhanced the DST concentrations in D. acuminata (Nagai et al., 2011), indicating the mass nutrients required for DST production.

Nitrogen is not involved in DST production directly. Levels of recalculated R _tox were highest under the N-deficient treatment and lowest under stepwise nitrite addition (total 4.43 mM-N) and high nitrite treatment (1 mM-N) on day 20 ( Figure S.15 , McLachlan et al., 1994), demonstrating that DST production was prompted by nitrogen limitations and even nitrogen starvation conditions. The enhanced R _tox under restricted nitrogen levels might be due to changes in carbon flow. Under nitrogen limitations, the P. lima complex showed an increased expression of genes related to fatty acid biosynthesis and starch synthesis (Hou et al., 2018).

DST are polyketide compounds, and their biosynthesis is similar to the processes of fatty acids. Each acetate unit undergoing acetyl-CoA, ketoreduction (KR), dehydration (DH), and enoyl reduction (ER) prolongs the carbon chain (Van Wagoner et al., 2014). In contrast, some or all processes are omitted or skipped during PKS production. Though some carbons process carbon deletion, β-Alkylation, and pseudo-α-alkylation, most of the carbon atoms in OA come from the entire acetate unit. KR and ER processes need NADPH to reduce the keto group and double bond (Keatinge-Clay and Stroud, 2006). Finally, some other polyketides have been reported in the structure of ketoreductase with a cofactor NADP+ or bound with an NADPH (Korman et al., 2004; Keatinge-Clay and Stroud, 2006; Jiao et al., 2021). The regulation of NADPH in DST production could be the goal of further research based on the high correlation between R _tox and C _DIP Values.