The raphidophyte Heterosigma akashiwo (strain code UOL_HA1) was isolated from a water sample taken from the Hartenbos Estuary along the warm-temperate south coast of South Africa in 2019 (i.e., less than one year until use in experiments), and the dinoflagellate Heterocapsa rotundata was obtained from the Roscoff Culture Collection, France (strain code RCC3042). Both algae were cultivated in f/2 medium (Guillard, 1975) prepared from 0.2 μm filtered and sterilized seawater with a salinity of 30, taken from the Jade Bay in Wilhelmshaven, Germany. By dilution with f/2 medium, the cultures were kept in exponential growth phase and were cultivated under controlled conditions in a climate chamber at 18°C, 70 μmol photons m^-2 s^-1 and a 12:12 h light:dark period. Eggs of the copepod Acartia tonsa were obtained from the AWI Biological Institute Helgoland (BAH), Germany, and stored at 6°C in total darkness until use. The rotifer Brachionus plicatilis was obtained from the aquatic retailer Interaquaristik, Germany, and was grown under the same culture conditions as described for the phytoplankton species. Every two weeks, the Brachionus culture was transferred into new medium and was fed with the green algae Tetraselmis sp. (also derived from the Roscoff Culture Collection, France).

A competition experiment was conducted to investigate the growth and competitive ability of H. akashiwo in the presence of the dinoflagellate H. rotundata under variable estuarine conditions by establishing two levels of temperature (16 and 22°C) in combination with two levels of salinity (15 and 30) in a full factorial design. Temperature and salinity levels were selected based on the findings of Lemley et al. (2018b) in the Sundays Estuary, South Africa, and thus represent typical ranges for the naturally occurring variability of temperature and salinity in South African estuaries. Both algae were inoculated at equal biovolume proportions to account for the size difference of the two species. For that purpose, the dimensions of 30 cells of each species were measured using a Zeiss Axiovert 10 inverted microscope and the individual cell biovolume of H. akashiwo and H. rotundata was calculated according to Hillebrand et al. (1999). Both species were inoculated at a total biovolume of 2.43 × 10⁶ μm³, corresponding to an initial cell density of 5 × 10³ cells ml^-1 for H. akashiwo and 21 × 10³ cells ml^-1 for H. rotundata (cell volume H. akashiwo: 487 μm³, cell volume H. rotundata: 114 μm³). Both species were set up in monoculture and polyculture, using an additive design, i.e., inoculating each species with the same biovolume in monoculture and polyculture, resulting in twice the total biovolume in the polycultures compared to monocultures. This additive design was chosen in consideration of the subsequent grazing assays, where the consumers were incubated with the same algal monocultures and polycultures. This assured equal amounts of the harmful algae (i.e., equal concentrations of potentially harmful substances) in all treatments. Each treatment was replicated by four, resulting in a total of 48 experimental units (3 species combinations x 2 salinity levels x 2 temperature levels x 4 replicates), containing 200 ml culture in a sterile 250 ml Erlenmeyer flask. To establish the lower salinity level of 15, f/2 medium was diluted with fully desalinated water enriched with f/2 nutrient concentrations and the salinity was checked with a Multi 3630 IDS SET KS2 sensor. The experimental units were incubated, using the empty containers of an indoor mesocosm facility (Planktotrons, (Gall et al., 2017) at their respective experimental temperature (16°C and 22°C) and a light intensity of 100 μmol photons m^-2 s^-1 with a light:dark period of 12:12 hours. To ensure equal illumination among the experimental units and avoid cell aggregation the flasks were exposed to slow mechanical rotation and were additionally homogenized by hand every 12 hours. Sampling was conducted every 48 hours and comprised subsamples of 500 μl from each experimental unit. Samples were fixed in a 48-well microplate (83.3923 Sarstedt AG & Co KG ©) with formalin at a final concentration of 2% for quantification conducted with a Leica DMIL inverted bright field microscope at a magnification of 100x. Every fourth day, an additional 10 ml of each experimental unit were sampled for subsequent dissolved macronutrient analysis. Furthermore, at the end of the experiment, 10 ml samples were taken from each experimental unit for particulate macronutrient analysis. The experiment was terminated after 28 days, when the last treatment had reached the stationary growth phase.

A grazing assay was conducted to evaluate whether and how the potential harmful effects of H. akashiwo on copepod nauplii are altered by different temperature and salinity regimes, as well as by the presence of a non-harmful competitor. Prior to the main experiment, a pilot study was performed in which the effect of a H. akashiwo culture (4 x 10⁴ cells ml^-1) on A. tonsa nauplii (1 individual ml^-1) survival was compared to a starvation control (sterile filtered seawater without algal food) in a five-day incubation. This was done to test whether the harmful effects of H. akashiwo on copepod nauplii exceed a pure starvation effect, i.e., whether the potential mucus production of the harmful algae mainly hinders filter feeding in the consumer, leading to starvation, or whether the harmful effects are stronger, indicating the production of allelochemicals or other harmful substances.

For the subsequent full factorial grazing assay, the same experimental salinity and temperature levels and the same algae species combinations were used as in the competition experiment (Section 3.2), but with the addition of A. tonsa nauplii at an initial density of 1 individual ml^-1. To provide sufficient prey for the nauplii, H. akashiwo was inoculated with an initial cell density of 1.5 × 10⁴ cells ml^-1, corresponding to 6.4 × 10⁴ cells ml^-1 H. rotundata, resulting in equal biovolume proportions of each phytoplankton species in monoculture and polyculture. Phytoplankton were inoculated again using the additive design, resulting in twice the total algal biovolume in polyculture to ensure that the grazers were exposed to the same biovolume of H. akashiwo and thus the same amount of potentially harmful substances in all treatments (see above). Due to space constraints, no additional non-grazer control treatments containing only the phytoplankton species were set up; instead, the competition experiment, which was conducted only a few weeks before, was used as non-grazer control.

After hatching, nauplii stock cultures were prepared according to Meunier et al. (2016). Experimental units were set up with growth medium and phytoplankton/consumer stock cultures in a total volume of 200 ml contained in 250 ml Erlenmeyer flasks. The preparation of the experimental units in terms of salinity was identical to the procedure described for the competition experiment. After incubation in the Planktotron containers, phytoplankton sampling was conducted every 24 hours with the identical procedure as described for the competition experiment. For evaluation of zooplankton mortality, additional 10 ml subsamples were taken daily from each experimental unit and transferred into 6-well microplates for quantification with an inverted light microscope. The mortality was then calculated by dividing the number of dead individuals in the live samples by the total number of nauplii counted in each sample after fixation with formalin at 2% final concentration. An individual was considered dead after no physical movement for at least 60 seconds was observed. The experiment was terminated after five days, when the first treatment showed a mortality of 100%. At the end of the experiment, the developmental stages of the nauplii in different treatments were evaluated by micrographs of three individuals within each replicate taken with a Jenoptik Gryphax Kapella camera connected to a Zeiss Axiovert 10 microscope. At the end of the experiment, additional 10 ml subsamples were taken from each experimental unit for dissolved and particulate macronutrient analysis, respectively. For the particulate macronutrient subsamples, grazers were removed by filtration through a 20 µm nylon mesh.

A second grazing assay was conducted with the rotifer Brachionus plicatilis to test for potential differences in grazer sensitivity to the harmful algae. Since the results of the prior competition experiment (Section 3.2) indicated that salinity effects depended on temperature, the rotifer bioassay was reduced to the two salinity levels (15 and 30) at a fixed experimental temperature of 16°C. This reduced experimental setup allowed for the addition of non-grazer control treatments containing only the phytoplankton species in monoculture and polyculture. All treatments were replicated three times.

Prior to the start of the experiment, the rotifers were starved for 24 hours by careful rinsing with sterile seawater on a 50 μm nylon mesh to remove any prey items, followed by resuspension in 500 ml beakers. After the starvation period, these cultures were used as stocks for the experiment. Next, the experimental units, with a total volume of 150 ml each, were inoculated with the rotifers at a density of 1 individual ml^-1. The preparation of the experimental units in terms of salinity, as well as the initial cell densities of the two phytoplankton species, was identical to the copepod nauplii bioassay (Section 3.3). After incubation in the Planktotron containers under the same controlled conditions at 16°C, grazer and phytoplankton sampling for microscopy were conducted every third day (including day 0) with identical procedures as previously described in Section 2.2 and 2.3, respectively. The experiment was terminated after six days.

Samples for dissolved inorganic nutrients (nitrogen, phosphorus, silicate) were prefiltered through 0.2 µm syringe filters (Sarstedt Filtropour S) and stored in PE bottles at -20°C until analysis using a SAN++ Continuous Flow Analyzer (CFA, Skalar Analytical B.V., Netherlands). For particulate intracellular carbon (C), nitrogen (N), and phosphorus (P) analyses, two 10 ml subsamples were filtered on pre-combusted and acid washed Whatman GF/F filters for CN and P analyses, respectively. The CN elemental composition was measured using a CHN analyzer (Thermo, Flash EA 1112). Particulate phosphate was measured as orthophosphate by molybdate reaction after sulfuric acid digestion according to Grasshoff et al. (2009).

To test for optimum growth conditions of the two phytoplankton species, as well as for the impact of the competitor in the competition experiment (Section 3.2), treatment effects on phytoplankton biovolume were tested for each species separately, using linear mixed models (LMMs) with temperature, salinity, presence of the competitor, and time as fixed factors, as well as the index number of the experimental unit representing the random factor. The LMMs were conducted using the “lmer” function from the “lmerTest” package (Kuznetsova et al., 2017) that includes the Satterthwaite approximation for obtaining p-values. To meet the model assumptions of homoscedasticity, algal biovolumes were square root transformed prior to the analysis. For testing dominance shifts over time in the polycultures, log ratios of the two phytoplankton species biovolumes were calculated for each sampling point which were then analyzed separately for each factorial treatment combination against time using Kruskal-Wallis tests. Subsequently, non-parametric pairwise Wilcoxon rank sum post hoc tests were used to compare the log ratios of every sampling day to the log ratio of day 0 at which both species were inoculated at equal biovolume proportions. Moreover, non-linear models were fitted for each experimental unit individually by means of the “nls” base R function using sampling data from the start until the day of the maximum achieved biovolume to obtain the maximum growth rates and carrying capacities of the two phytoplankton species in each treatment. Only significant model fits were used to extract parameters for maximum growth rates and carrying capacities. These were then tested separately for both species using two-way ANOVAs with salinity and temperature as independent variables. To test for differences in phytoplankton particulate molar C:N ratios at the end of the competition experiment and nauplii grazing assay, a three-way ANOVA was performed with species composition, temperature, and salinity as independent variables. Model assumptions of homogeneous variances and normally distributed residuals were tested with Fligner-Killeen tests and Shapiro-Wilk tests, respectively. Due to technical problems with the particulate phosphorous analysis, P-concentrations could not be considered here and no C:P or N:P ratios could be calculated.

In the nauplii and rotifer grazing assays, effects on mortality were analyzed using LMMs with temperature, salinity, prey composition, and time as fixed factors, as well as the index of the experimental unit as random factor for the nauplii grazing assay. The same factors, except for temperature (i.e., experiment was conducted at only one temperature level) were used for rotifers. To calculate grazing rates, phytoplankton growth rates in the absence of consumers (gross growth rate = μ) and with consumers (net growth rate = r) were calculated between day 0 and day 5 by subtracting ln-transformed starting biovolume from ln-transformed final biovolume and dividing this by the respective number of experimental days (five days for A. tonsa nauplii, six days for B. plicatilis). The net population grazing rate (g) was then calculated as μ − r. For the prey mixture treatments that contained both phytoplankton species, biovolumes were pooled prior to the grazing rate calculations. As previously described, there were no additional non-grazer controls for the copepod grazing assay so that the competition experiment was used as such. Consequently, the resulting grazing rates require cautious interpretation as the two experiments were set up with different initial biovolumes. However, no density-dependent effects on phytoplankton growth could be expected in any of the experiments during the first five days of the experiments (duration of the nauplii grazing assay). Moreover, the experiments were conducted in the same mesocosm containers with only a short time interval in between and represented identical conditions, allowing to compare phytoplankton growth rates in the different setups. To test for differences in relative population grazing rates of the nauplii in different temperature and salinity treatments, generalized least squares (GLS) models were conducted, to compensate for groups with non-homogeneous variances by using weights. Effects on rotifer relative population grazing rates were tested using three-way ANOVAs with temperature, salinity, and prey composition as independent variables. Moreover, to test for selective grazing effects in the mixed prey cultures, separate grazing rates were calculated for both phytoplankton species which were then tested in a three-way ANOVA with temperature, salinity, and phytoplankton species identity for the nauplii bioassay, and a two-way ANOVA with salinity and phytoplankton species identity for the rotifer bioassay. For comparing the sensitivity to harmful effects of A. tonsa and B. plicatilis, another LMM was performed on grazer mortality as dependent variable and salinity, prey composition, grazer identity, and time as independent variables. However, from the nauplii grazing assay, only data from the lower temperature treatments were considered here, since the rotifer grazing assay was conducted only at 16°C. Differences in population grazing rates between the two grazer species were tested using a three-way ANOVA with the same set of independent variables (except for time).

Post-hoc analyses were conducted with estimated marginal means pairwise comparisons using the “emmeans” function from the “emmeans” package (Lenth, 2021) for GLS models and Tukey post-hoc tests for LMMs and ANOVAs, respectively. All statistical analyses and graphs of this study were performed using the R statistical environment 4.0.2. (R Core Team, 2020).

H. akashiwo and H. rotundata biovolumes were both significantly and interactively affected by salinity, temperature, time, and the presence of the competitor, except for H. rotundata biovolumes that did not show a significant effect by its competitor ( Table 1 ).

In monoculture, H. akashiwo biovolume was significantly higher at the higher experimental temperature and lower salinity level ( Figures 1A, D ). Low salinity treatments resulted in higher carrying capacities for H. akashiwo ( Table 2 ). The positive temperature effect was reflected in higher maximum growth rates for H. akashiwo at 22°C compared to 16°C (p < 0.001), being highest (0.53 day^-1, Table S1 ) in the low salinity – high temperature treatment, differing from all other treatment combinations (p < 0.004). Depending on the respective treatment, H. akashiwo reached its maximum biovolume in monoculture between day 18 and 24, with the 16°C treatments generally achieving a time-delayed, but notably higher carrying capacity than at 22°C ( Figures 1A, D ).

H. rotundata biovolume was promoted by the higher salinity level of 30 in combination with the lower experimental temperature of 16°C compared to the other treatment combinations (p < 0.001, Figures 1B, E ), which resulted in a maximum growth rate of 0.52 day^-1 in this treatment and the highest carrying capacity (p < 0.001). By contrast, under low salinity – high temperature conditions, H. rotundata biovolume production was strongly inhibited, reflected by a lower maximum growth rate compared to the other treatments ( Table S1 ), and a maximum biovolume accounting for only 29% of the maximum biovolume achieved under the optimum high-salinity low-temperature conditions. Compared to H. akashiwo, the dinoflagellate generally exhibited lower total carrying capacities that were, however, achieved between day 6 and 14 depending on the treatment, and therefore several days earlier than for H. akashiwo in monoculture.

Temperature and salinity effects on H. akashiwo biovolume were generally stronger in polycultures compared to monocultures ( Figures 1C, F ). Within the high temperature treatments, H. akashiwo showed slightly higher maximum growth rates at the lower salinity level and a suppressed carrying capacity at the higher salinity level accounting for only 78% of the carrying capacity achieved in the corresponding monoculture treatment. At higher temperature, H. akashiwo dominated over H. rotundata for most of the experimental duration ( Figures 1F , S1 ) indicated by significant differences of the log ratio from day 0 ( Table S2 ). More precisely, within the high temperature treatments, H. akashiwo dominated over H. rotundata at the lower salinity level from day 2 onwards, whereas at the higher salinity level H. rotundata exhibited a temporary dominance from day 4 to 6, after which H. akashiwo dominated again from day 12 until the end of the experiment. Within the low temperature treatments, H. akashiwo dominated the low salinity treatment from day 12 onwards but was significantly inhibited by its competitor at high salinity until day 18, after which it dominated the community again. As such, competitive interactions strongly depended on salinity, especially within the low temperature treatments ( Figure 1C ), which was also reflected by the significant interactive effects of temperature, salinity, and competitor presence for H. akashiwo biovolume, maximum growth rate, and carrying capacity ( Tables 1 and 2 ). At low temperature – low salinity, H. akashiwo performed similar in the presence of H. rotundata as under the corresponding monoculture conditions. At the higher salinity level, H. akashiwo showed strongly suppressed growth rates and reached only 20% of the maximum biovolume of its corresponding monoculture.

Dissolved nutrient concentrations indicated phosphorous depletion in the H. akashiwo monocultures as well as in the polycultures starting from day 12 onwards ( Figure S2 ). Dissolved nitrogen was depleted several days later at day 20 ( Figure S3 ). However, C:N ratios across the different treatment combinations were relatively low and ranged from 5.1 to 9.1, indicating that algal cells were not N limited. C:N ratios were significantly affected by species composition, with H. akashiwo generally having higher C:N ratios than H. rotundata (p < 0.001), yet showing interactive effects with temperature and salinity ( Figure S4 ; Table S3 ).

In the preliminary pilot study, the mortality of the A. tonsa nauplii was higher when incubated with H. akashiwo compared to the filtered seawater control (p = 0.032). In the filtered seawater treatment, nauplii mortality showed a steady linear increase over time ( Figure S5 ), resulting in a mean mortality of 72% after five days of incubation. In contrast, mean nauplii mortality increased from 3 to 83% between day 1 and day 2 in the H. akashiwo treatment.

In the full factorial nauplii grazing assay conducted under the same temperature and salinity levels used in the competition experiment, nauplii mortality was interactively affected by salinity, temperature, prey composition, and time ( Table 3 ; Figure 2 ). Nauplii mortality at the end of the experiment was higher when incubated with H. akashiwo monocultures compared to H. rotundata monocultures or the polyculture treatments containing both phytoplankton species (p < 0.001, Table S4 ). In the H. akashiwo monocultures, nauplii mortality was higher at 22°C than at 16°C (p < 0.001), starting from day 2 onwards. Contrastingly, in the 16°C treatments, mortality was still 0% after two days of incubation and started to increase thereafter. Across both temperatures, nauplii mortality increased earlier at the lower salinity level compared to high salinity but converged in both salinities over the course of the experiment, which is reflected in the significant interaction of salinity and time ( Table 3 ). The highest overall mortality was observed in H. akashiwo monocultures under low salinity – high temperature conditions, i.e., the optimal growth conditions of H. akashiwo, with the first experimental unit reaching a mortality of 100% on day 3. Treatments containing the non-harmful prey H. rotundata, either in monoculture or in polyculture with H. akashiwo, were similar to each other and showed low nauplii mortality over time with no notable differences among each other ( Table S4 ). Regarding the developmental stages of the nauplii at the end of the experiment, nauplii exposed to H. akashiwo monocultures did not develop beyond the N1 state in any treatment combination of temperature and salinity, while nauplii in the H. rotundata monocultures and the polycultures developed to stages N4 and N3, respectively, at 16°C and even reached copepodite stages C2 – C4 in the 22°C treatments ( Figure S6 ).

Population grazing rates of the A. tonsa nauplii were significantly and interactively affected by prey composition and temperature, while the salinity effect was not significant, but depended on prey composition and temperature (twofold interactions of salinity with both factors, Table 3 ). Average nauplii grazing rates were lowest in H. akashiwo monoculture treatments irrespective of temperature and salinity levels ( Figure 3 ). In the H. rotundata monoculture as well as in the polyculture treatments, however, nauplii grazing rates strongly differed with temperature and salinity. While grazing rates were generally very low at 16°C and did not exceed a mean of 0.1 day^-1, grazing rates increased with temperature, especially at low salinity, resulting in significant differences of this treatment compared to most other treatment combinations ( Table S5 ). H. rotundata monoculture treatments experienced the highest grazing impact (0.39 day^-1). Considering separate grazing rates for both phytoplankton species in the polyculture treatments, grazing rates on H. rotundata were higher than on H. akashiwo in the higher experimental temperature (p < 0.001, Figure 4 ; Tables 4 , S6 ), regardless of salinity.

Dissolved nutrient concentrations indicated no macronutrient depletion of phosphorous or nitrogen during the experiment ( Figures S7, S8 ). C:N ratios of the phytoplankton ranged between 6.7 and 8.2 ( Figure S9 ).

The second grazing assay was conducted with the rotifer B. plicatilis at only one temperature (16°C), but both salinity levels which were also used in the nauplii grazing assay. Here, rotifer mortality was significantly affected by prey composition and time; both factors depending on salinity (two-fold interactive effects of both factors with salinity, Figure 5 ; Table 5 ).

As in the nauplii grazing assay, grazer mortality at the last sampling day was higher in H. akashiwo monoculture treatments compared to the H. rotundata monoculture or polyculture treatments (p < 0.004, Table S7 ). Within the H. akashiwo treatments, the highest mortality was found at the high salinity level (p = 0.023) with a mean mortality of 65% at the end of the experiment, which was notably higher than the mortality at the low salinity level (20%). Rotifer mortality in treatments containing the non-harmful prey H. rotundata in monoculture or in polyculture with H. akashiwo did not differ from each other and did not exceed a mean mortality of 10%.

The relative population grazing rates were very low across all treatments and included negative values, indicating that for some treatments phytoplankton growth rates were even higher in the presence of grazers compared to the non-grazed controls ( Figure 6 ). However, the relative population grazing rate of B. plicatilis differed with salinity as well as interactively with salinity and prey composition ( Table 5 ). This was reflected in similar grazing rates across different prey compositions at the low salinity level, but significant differences at the higher salinity level ( Table S8 ). However, despite these significant differences, the overall grazing impact of the rotifers and therefore also the effect size was very low.

In polycultures, grazing rates on H. rotundata were higher than for H. akashiwo at the lower salinity level (p = 0.012, Tables 4 , S9 ). Yet, differences between the species-specific relative grazing rates were also relatively small (< 0.1) and are likely an artefact of the enhanced growth of H. akashiwo in this treatment since B. plicatilis is known to have a non-selective feeding behavior ( Figure S10 ).

Comparing the mortality of B. plicatilis and A. tonsa nauplii in the 16°C treatments, grazer mortality was significantly and interactively affected by prey composition and time (significant main factors and interaction), while salinity effects were only important in relation to time ( Table S7 ). Grazer mortality did not differ notably between the two consumer species, indicating similar sensitivities of both grazers in response to the harmful algae. However, salinity affected grazer mortality in H. akashiwo monocultures in contrasting ways, with B. plicatilis exhibiting higher mortality at high salinity and nauplii showing higher mortality at the low salinity level ( Figures 2 , 5 ; Table S11 ).

Population grazing rates of the two grazers differed from each other and were interactively affected by salinity and prey composition (significant main effects and two-way interactions, Table S10 ). The grazing rates of the two consumers did not exceed an average of 0.1 day^-1 in the lower temperature level. The observed interactive effect of consumer identity and prey composition indicated that the presence of non-harmful prey led to higher grazing rates for the A. tonsa nauplii. In contrast, the presence of non-harmful prey did not result in increased grazing for B. plicatilis based on its overall minor grazing impact across all factorial combinations ( Figures 3 , 6 ; Table S12 ).