Intraspecific trait variation and trade-offs within and across populations of a toxic dinoflagellate.

Intraspecific trait diversity can promote the success of a species, as complementarity of functional traits within populations may enhance its competitive success and facilitates resilience to changing environmental conditions. Here, we experimentally determined the variation and relationships between traits in 15 strains of the toxic dinoflagellate Alexandrium ostenfeldii derived from two populations. Measured traits included growth rate, cell size, elemental composition, nitrogen uptake kinetics, toxin production and allelochemical potency. Our results demonstrate substantial variation in all analysed traits both within and across populations, particularly in nitrogen affinity, which was even comparable to interspecific variation across phytoplankton species. We found distinct trade-offs between maximum nitrogen uptake rate and affinity, and between defensive and competitive traits. Furthermore, we identified differences in trait variation between the genetically similar populations. The observed high trait variation may facilitate development and resilience of harmful algal blooms under dynamic environmental conditions.


INTRODUCTION
Ecosystem functioning is often linked to biodiversity, with enhanced productivity and stability in systems with higher species richness (Cardinale et al. 2006;Ptacnik et al. 2008;Cardinale 2011;Schindler et al. 2015). By increasing the number of species in the community, complementarity of functional traits within the system may increase, resulting in higher resource use efficiencies, enhanced resistance to consumers and resilience to environmental changes (MacArthur 1955;Hooper et al. 2005;Bolnick et al. 2011;Cardinale 2011). Similarly, a high intraspecific trait diversity, as well as variation within traits may facilitate population growth by enhanced complementarity (Violle et al. 2012). Moreover, high standing genetic and trait variation within populations allows species to adapt more quickly to changes in the environment (Barrett & Schluter 2008). This variation is ultimately responsible for the diversification and evolutionary adaptation of species, which can even happen on ecological time scales (Carroll et al. 2007;Pelletier et al. 2009). Especially with respect to climate change, intraspecific trait variation could modulate species' responses to changing environments and buffer biodiversity losses (Reusch et al. 2005;Albert et al. 2010;Sj€ oqvist & Kremp 2016).
Phytoplankton populations can possess a very high genetic diversity with substantial intraspecific trait variation (Maranda et al. 1985;Medlin et al. 2000;Burkholder & Glibert 2006;Alpermann et al. 2010;Godhe et al. 2016). Specifically, harmful algal bloom (HAB) populations have been shown to exhibit a high variation in the expression of traits Hakanen et al. 2014;John et al. 2015). HAB species are known to possess specialized functional traits, including toxin production, as well as N 2 fixation and mixotrophy (Fogg 1969;Hallegraeff 1993;Burkholder et al. 2008), which are expected to contribute to the success of HAB populations (Paerl & Huisman 2009;Litchman et al. 2010;Carey et al. 2012). Among common HAB-forming phytoplankton groups are dinoflagellates, a ubiquitous group of protists that particularly contributes to primary production in coastal waters (Falkowski & Knoll 2007). The toxic dinoflagellate genus Alexandrium is among the most common HAB species, imposing an eminent threat to human health and ecosystem functioning by causing shellfish poisoning and fish kills Durbin et al. 2002;Anderson et al. 2012). Like most HAB species, Alexandrium populations possess a high functional trait diversity (Ogata et al. 1987;Cembella et al. 2002;Masseret et al. 2009;Tillmann et al. 2009;Alpermann et al. 2010). Indeed, strains can substantially differ in the production of paralytic shellfish poisoning (PSP) toxins, with some strains even lacking the ability to produce these toxins (John et al. 2014;Murray et al. 2015). Furthermore, some Alexandrium species and strains can produce two additional types of neurotoxins; gymnodimines (GYM) and spirolides (SPX; Cembella et al. 2000;Kremp et al. 2014; Van de Waal et al. 2015), as well as allelopathic compounds that are excreted from the cell and can lyse competing phytoplankton species and small protozoan grazers (Tillmann & John 2002;Tillmann et al. 2007).
The high functional trait variation within Alexandrium populations can partly be attributed to their life histories. In addition to vegetative growth, Alexandrium reproduces sexually and forms resting cysts, allowing recombination of genetic material . Cysts may hatch continuously throughout the vegetative growth season and thereby supply novel genotypes to the water column Angles et al. 2012;Kremp et al. 2015). This mechanism may contribute to a high genotypic diversity, and support a high variation in traits within a population ). High trait variation is further enhanced by shifting selection pressures that can occur during toxic algal blooms (Driscoll et al. 2015). For instance, before the onset of a bloom, interactions occur mainly interspecifically, and competition with other phytoplankton species together with a high grazing pressure may select for defensive traits, such as toxin and allelochemical production. During the development of a bloom, however, when population densities increase and nutrients become depleted (Brandenburg et al. 2017), intraspecific interactions become more important, and selection may shift towards higher resource use efficiencies. Such low nutrient conditions may support the success of non-toxic or non-allelopathic strains, as production of secondary metabolites involves biochemical and energetic costs (Pohnert et al. 2007). Consequently, populations may exhibit intraspecific trade-offs, for instance between defensive and competitive traits. Trade-offs between functional traits stand at the base of eco-evolutionary dynamics, and are key drivers underlying phenotypic diversity and variation, as they constrain trait evolution and thereby prevent selection for a 'super-clone'. Trade-offs furthermore facilitate coexistence of species and strains, and are important for maintaining diversity (Tilman 1982;Grover 1997;Kneitel & Chase 2004;Blows & Hoffmann 2005;Van de Waal et al. 2011). The variation in selection pressures during a bloom can concede and maintain a high genetic and functional diversity within a population, and result in genetic differentiation over very short time scales (Richlen et al. 2012). However, empirical data on intraspecific trade-offs in HAB populations is still lacking.
Here, we assessed the variation in eight traits in 15 isolates from two geographically distinct toxic Alexandrium ostenfeldii populations. A. ostenfeldii is a globally widespread dinoflagellate species that can produce several types of neurotoxins (PSP, GYM and SPX), as well as allelochemicals (Kremp et al. 2009;Van de Waal et al. 2015). It generally occurs in background levels, but may also cause dense harmful blooms in brackish waters including embayments, creeks and coastal marine ecosystems (Kremp et al. 2009). One of the studied populations is relatively large and already established between 8000 and 3000 BP in the Baltic Sea (Tahvanainen et al. 2012). The other population is much smaller and only established recently in a small creek in the Netherlands (Burson et al. 2014; Van de Waal et al. 2015). For all isolates, we first determined general traits such as growth rate, cell size and stoichiometry; nitrogen (N) acquisition traits including the maximum N uptake rate and affinity (since N is the limiting nutrient during summer blooms at both locations); and defensive traits including production of toxins (i.e. PSP toxins, GYM and SPX), as well as allelopathic potency, and afterwards tested for putative trade-offs. We aim to assess how different environments affect diversification of traits, and how physiological constraints lead to intraspecific trade-offs. Specifically, we expect a higher trait variation in the longer established Baltic Sea population as this population had more time to diversify, and we predict selection of defensive traits (i.e. toxin quota and allelochemical potency) in the recently established Dutch population as this would support its colonisation. Moreover, we anticipate distinctive trade-offs between traits that scale with size, such as growth rate and cell volume, as well as between defensive and competitive traits. Lastly, we expect a positive relationship between relative phosphorus (P) content and growth rate following the growth rate hypothesis, which predicts that cells require more P in order to grow faster (Elser et al. 1996(Elser et al. , 2003Sterner & Elser 2002).

Culturing
Alexandrium ostenfeldii strains from the Netherlands were isolated in summer blooms from the Ouwerkerkse Kreek Clonal isolates were genotyped using six microsatellite markers (Nagai et al. 2014), which confirmed that strains represented different haplotypes (Table S1). We randomly selected 15 A. ostenfeldii strains (five from the Netherlands in 2015, five from the Netherlands in 2016, and five from the Baltic Sea), which we cultured in ½Kmedium (after Keller et al. 1987) with 0.2 lm sterilized North Sea water adjusted to a salinity of 10. Cultures were grown at a temperature of 18°C and an incident light intensity of 85 lmol photons m À2 s À1 , with a light:dark cycle of 16:8 h. First, the cultures were acclimated to experimental conditions for 2 weeks (approximately five generations), after which they were transferred to 500 mL Erlenmeyer flasks using four biological replicates. Cultures were grown as batch starting at a cell density of 400 cells mL À1 . Each replicate was grown for three to five generations, and cultures were harvested for the assessment of traits while in exponential growth phase at densities of approximately 5.0 9 10 3 cells mL À1 .

Growth rate and cell size
To estimate cell densities, 5 mL samples were taken every second or third day and fixed with neutral Lugol's iodine solution (Lugol) to a final concentration of 1%. Samples were stored in the dark at 4°C until analysis. A. ostenfeldii cells were counted on an inverted microscope (DMI 4000B; Leica Microsystems CMS GmbH, Mannheim, Germany). Maximum growth rates (l) were calculated for each replicate by fitting an exponential function through cell counts over time (excluding potential lag phases; see also Fig. S1 and Table S2), following: where N t refers to the cell concentrations at time t, and N 0 to the cell concentrations at the start of each experiment (or exponential growth phase).
To establish cell size data, Lugol fixed samples from the last day were analysed using a FlowCam (Fluid Imaging Technologies, Yarmouth, ME, USA), where surface area was measured and used to calculate cell volume based on a sphere. For each replicate, at least 100 cells were measured.

Elemental composition
Cellular carbon (C), nitrogen (N) and phosphorus (P) were determined at the end of the experiment in 20 mL of culture material collected on a prewashed (100 mL distilled water) glass microfiber filter (Whatman GF/F, Maidstone, UK). Filters were dried overnight at 60°C and stored in the dark until further analyses. For C and N analyses, a subsample (14%) of each filter was folded into a tin cup and analysed on a FLASH 2000 organic elemental analyser (Brechbueler Incorporated, Interscience B.V., Breda, the Netherlands). Cellular P was analysed by combusting the remainder of the filter (86%) for 30 min at 550°C in Pyrex glass tubes, followed by a digestion step with 5 mL persulfate (2.5%) for 30 min at 120°C. This digested solution was measured for PO 4 3À on a QuAAtro39 AutoAnalyzer (SEAL Analytical Ltd., Southampton, UK).

Allelopathic potency
The allelopathic potency of the A. ostenfeldii cultures was determined after Tillmann et al. (2009), using a short-term bioassay with the cryptophyte Rhodomonas salina as target species. Lytic effects were quantified by adding A. ostenfeldii filtrate to an R. salina culture at concentrations reflecting different cell densities, and subsequently assess the A. ostenfeldii cell densities required to reduce the R. salina abundance by 50%.
To collect the filtrate, 200 mL of A. ostenfeldii culture from the end of the experiment was centrifuged at 168 g for 10 min (Thermo Scientific SL16, Thermo Fisher Scientific, Langenselbold, Germany). Hereafter, the cell pellet was used for the N uptake bioassays (see below) and the filtrate was filtered over a sterile 0.2 lm cellulose acetate filter and stored at À20°C until further analyses.
Cell counts from the end of the experiment were used to estimate the cell densities of A. ostenfeldii associated with the filtrate. From the filtrate, eight dilutions were made corresponding to a range of 0.3-8 9 10 3 cells mL À1 in vials of 4 mL. Medium without filtrate was used during each run as a control.
For the bioassays, 0.1 mL of R. salina culture was added to each vial containing A. ostenfeldii filtrate reaching a final cell density of 10 9 10 3 cells mL À1 . The samples were incubated for 24 h under the same culture conditions as the experiment.
At the end of the incubations, R. salina cells were counted using an automated cell counter (Beckman Coulter Multisizer 3, Indianopolis, IN, USA).
In order to estimate the EC 50 for each strain (i.e. the A. ostenfeldii cell density leading to a 50% decline in R salina cell densities), the following function was fitted through R. salina counts as a function of log-transformed A. ostenfeldii cell concentrations associated with the filtrate (Fig. S2): where N final is the final target cell density of R. salina using A. ostenfeldii filtrates, N control is the final target cell density of R. salina in control samples, x is the log-transformed A. ostenfeldii cell density, and EC 50 and h are fitted parameters, where h is the Hill coefficient.

Nitrogen uptake kinetics
Determination of the N uptake kinetics of the different A. ostenfeldii strains was done using 15 N bioassays, after Moschonas (2015). Centrifuged cell pellets of A. ostenfeldii were re-suspended in medium lacking N and incubated under experimental culture conditions for 24 h.
Hereafter, samples for cell counts were taken and the A. ostenfeldii culture was divided and transferred to 7 9 50 mL tubes. Afterwards, 15 N labelled NO 3 À (Sigma Aldrich, Steinheim am Albuch, Germany, 10 atom % 15 N) was added to yield the following concentrations: 0.1, 0.3, 0.6, 1, 3, 10 lmol L À1 . For every A. ostenfeldii strain, a control without addition of 15 N labelled NO 3 À was taken in order to estimate background 15 N values. The samples were incubated for 1 h under the same culture conditions, and afterwards filtered through glass microfiber filters (GF/F; Whatman, Maidstone, UK). Filters were dried overnight at 60°C and stored in the dark until further analyses.
A subsample (14%) of every filter was folded in a tin cup and analysed on an elemental analyser (Flash 2000, Thermoscientific, Langenselbold, Germany) coupled to an isotope ratio mass spectrometer (IRMS, Thermoscientific), and d 15 N values were calculated as the difference in 15 N abundance between the sample and a reference gas (with 0.366 atom% 15 N). The d 15 N values were subsequently used to calculate N uptake rates using the equation from Dugdale & Wilkerson (1986): where q is the absolute N uptake rate (in lmol N L À1 h À1 ), 15 N(t 0 ) source is the isotopic ratio of the enriched dissolved N pool (in atom%), 15 N(t 0 ) sink is the natural abundance of 15 N (in atom%), 15 N(t) sink is the abundance of 15 N in the sample after the incubation (in atom%), t is the incubation time (in hours) and PON, the particulate organic nitrogen concentration (in lmol N L À1 ).
The absolute N uptake rates calculated for each concentration of NO 3 À can then be used to estimate the N uptake kinetics for each A. ostenfeldii strain. To this end, a Michaelis-Menten function was fitted through all data points within the bioassays for each replicate of each strain: where q (in lmol N L À1 h À1 ) is the N uptake rate, V max the maximum uptake rate (in lmol N L À1 h À1 ), K 1/2 the half saturation constant (in lmol N L À1 ) and NO 3 À the nitrate concentration. Subsequently, V max was divided by the cell densities at the end of each experiment and units were converted to pmol N cell À1 day À1 .

Toxin sampling and analyses
Samples for toxin analyses (PSP toxins and cyclic imines) were taken by filtration of 20 mL of A. ostenfeldii culture over glass microfiber filters (GF/F, Whatman, Maidstone, UK), which were stored at À20°C until further analysis. PSP toxins were determined by ion pair liquid chromatography coupled to post-column derivatization and fluorescence detection, as described in Krock et al. (2007) and Van de Waal et al. (2015). The cyclic imine toxin measurements (SPX and GYM) were performed on an Agilent 1100 LC liquid chromatograph (Waldbronn, Germany) coupled to a 4000 Q Trap triple-quadrupole mass spectrometer (AB-Sciex, Darmstadt, Germany) with a Turbo V ion source. Toxins were quantified by external calibration curves of SPX-1 and GYM A with standard solutions ranging from 10 to 1000 pg lL À1 , each. Other SPXs and GYMs for which no standards are available were calibrated against the SPX-1 and GYM A calibration curve, respectively, and expressed as SPX-1 or GYM A equivalents. For further details, also see Van de Waal et al. (2015).

Statistical analyses
Statistical analyses were performed in R version 3.2.2 (R Core Team 2015). For the measured traits across all strains, coefficients of variation (CV) were calculated as the ratio of the standard deviation to the mean. Significant differences in the expression of traits between the three groups of A. ostenfeldii strains were determined using a one-way ANOVA, including all replicates, followed by post hoc comparison of the means using Tukey's HSD. Trade-offs and relationships were recognized using a linear regression between the log-transformed trait values, including all replicates. Differentiation of traits within the distinct populations was assessed with the R package 'vegan' (Oksanen et al. 2017), where the multivariate homogeneity of group dispersions was analysed. Significant differences between the groups were determined using a permutation test.

RESULTS
We observed substantial variation of the analysed traits among the 15 studied A. ostenfeldii strains within and among their corresponding populations. Maximum growth rates ranged from 0.12 to 0.39 d À1 in different strains (Figs 1 and 2a), and cell volumes from 7.5 to 32.4 9 10 3 lm 3 (Fig. 2d), with CVs of 0.32 and 0.38, respectively. PSP toxin quotas varied from 1.1 to 81.7 pg cell À1 (CV 0.73) and cyclic imine toxin quotas from 18.8 to 443 pg cell À1 (CV 0.66; Fig. 2g-h; Fig. S3). Allelopathic potencies also varied strongly between the strains, with estimated EC 50 values from 0.9 to 33.5 9 10 3 A. ostenfeldii cells mL À1 (CV 1.09; Fig. 2i). For N uptake kinetics, K 1/2 values ranged from 0.3 to 11.9 lmol N L À1 (CV 0.8), and V max values from 3.6 to 5.4 pmol N cell À1 day À1 (CV 0.74; Fig. 2ef). Moreover, we also observed variation in the elemental stoichiometry of the different strains, ranging from 5.4 to 7.2 for cellular C:N ratios (CV 0.08; data not shown), from 50.9 to 154.6 for cellular C:P ratios (CV 0.21) and from 7.6 to 18.5 for cellular N:P ratios (CV 0.23; Fig. 2b-c).
Significant differences in the expression of traits between the Dutch and Baltic Sea A. ostenfeldii populations were found for growth rate (P < 0.001), cell size (P < 0.001), C:P ratio (P < 0.001), N:P ratio (P < 0.001), PSP toxin quotas (P < 0.001), cyclic imine toxin quotas (P < 0.001) and allelopathic potencies (P < 0.01; Fig. 2). Differences were less pronounced for N uptake kinetics, with significant differences only between the Baltic Sea population and the 2016 Dutch population for K 1/2 (P < 0.01; Fig. 2). Strains from the Dutch population isolated in different years were generally similar in their trait expression. Significant differences were only found for PSP toxin quotas (P < 0.01), cyclic imine toxin quotas (P < 0.05) and K 1/2 (P < 0.05; Fig. 2).
Using non-metric multidimensional scaling in an ordination plot we could define two distinct clusters that clearly separated the two geographically distinct A. ostenfeldii populations (Fig. 3a). There was more variation in the expression of traits within the Baltic Sea population as compared to the Dutch population from both years together ( Fig. 3b; P < 0.001).
The analysed selection of traits showed various distinct relationships. For instance, we observed strongest positive correlations between cell size and toxin quota and between PSP toxins and cyclic imines, while we observed negative  correlations between growth rate and cell size as well as between growth rate and toxin quota, which may represent putative trade-offs (Table 1). We note that these trade-offs were mainly driven by clustering of the two distinct A. ostenfeldii populations (Fig. S4). Interestingly, we observed a very distinct trade-off between V max and K 1/2 , both within and across populations (Fig. 4).

DISCUSSION
Our results reveal large intraspecific variation in trait expression within one phytoplankton species at two different levels, namely across and within two distinct populations. In fact, even within the recently established Dutch population, we found substantial variation in trait expression, which may point towards rapid diversification. We also observed distinct differences in the analysed traits between the two populations. Overall, the Baltic Sea strains grew faster, were smaller, contained relatively more P and produced less toxins and allelopathic compounds than the Dutch strains (Fig. 2). These particular traits are closely linked, and exhibit fundamental relationships and trade-offs.
The observed intraspecific trade-off between growth rate and cell size is in line with earlier findings on interspecific trade-offs, where smaller phytoplankton cells typically grew faster than larger cells (Banse 1976;Finkel et al. 2010). Furthermore, in order to grow faster, the growth rate hypothesis predicts that organisms should contain relatively more P, as more P-rich rRNA is required (Elser et al. 1996(Elser et al. , 2003Sterner & Elser 2002). The applicability of the growth rate hypothesis for phytoplankton, however, has been debated . Phytoplankton may accumulate P under non-limiting growth conditions, as well as at limitations other than P (Elrifi & Turpin 1985;Hillebrand et al. 2013). Thus, the relationship between growth rate and the relative cellular P contents will strongly depend on the growth-controlling factors. Our results show intrinsic differences in growth rate between strains cultured under the same non-limiting conditions, with , showing variation in (a) maximum growth rates, (b) C:P ratios, (c) N:P ratios, (d) cell sizes, (e) V max , (f) K 1/2 , (g) PSP toxin quotas, (h) cyclic imine toxin quotas and (i) EC 50 , isolated from the Netherlands (2015, light grey; 2016, dark grey) and the Baltic Sea (white). Significance differences between the populations are indicated by the asterisks (***P < 0.001, **P < 0.01, *P < 0.05).

Letter
Intraspecific trait variation of a toxic alga 1565 a close association with the relative P contents (i.e. C:P ratio, Table 1). This was further confirmed by a positive correlation between growth rate and cellular P contents corrected to cell volume (R 2 = 0.16, P < 0.01). Our results thus revealed an ideal test for the growth rate hypothesis, showing that the Baltic Sea A. ostenfeldii cells contained relatively more P, and could indeed grow faster than Dutch isolates. Faster growth and smaller cells were also associated with lower toxin quota in A. ostenfeldii (Table 1). Interestingly, the negative relationships between growth rate and PSP toxins (R 2 = 0.21, P < 0.001) as well as cyclic imines (R 2 = 0.52, P < 0.001) remained significant after correcting for cell size. This may represent a trade-off between toxin synthesis and growth, as the production of these secondary metabolites has certain biochemical and energetic costs (Pohnert et al. 2007). Cellular toxin content also positively correlated with C:P and N:P ratios (Table 1), suggesting that a relatively lower P content is associated with higher toxin quota. This relationship, however, might as well be caused by differences in cell size, where larger cells contain relatively less P and more toxins.
Most of the analysed traits were relatively conserved within A. ostenfeldii. Specifically, the observed intraspecific variation in growth rate, cell size, V max for nitrogen and elemental composition across the tested A. ostenfeldii strains was lower as compared to the interspecific variation across phytoplankton species (Fig. S5). This is presumably caused by the vast range in cell size across species, that affects growth rate, nitrogen acquisition and elemental composition Edwards et al. 2015a, b;Garcia et al. 2018). The intraspecific variation in K 1/2 values for nitrogen, however, was comparable to the earlier reported interspecific variation across marine phytoplankton species (Fig. 5;Edwards et al. 2015a). These ranged from high K 1/2 values (i.e. low affinities) observed for larger phytoplankton groups such as dinoflagellates, down to low values (i.e. high affinities) observed for small phytoplankton groups such as coccolithophores. This emphasises that trait variation occurring in a single phytoplankton species can be substantial, and may have important implications for its ecological functioning. Indeed, K 1/2 largely determines the competitive ability for a resource (Tilman 1982;Litchman et al. 2007;Edwards et al. 2011). Consequently, low K 1/2 values may support strains under low nutrient conditions, while high K 1/2 values associated with high nutrient uptake rates (Fig. 4;Litchman et al. 2007;Edwards et al. 2011) may increase the competitive ability of strains under nutrient replete conditions.
Both of the studied A. ostenfeldii populations are phylogenetically closely related, belonging to the same ribotype, but live in different habitats Van de Waal et al. 2015). The Baltic Sea isolates showed more pronounced trait variation than the Dutch isolates (Fig. 3). Besides habitat constraints, this might be explained by the presumably different histories of the populations in their respective habitats. Colonisation of the Baltic Sea by marine organisms is estimated to have occurred after the postglacial opening of the North Sea, approximately 8000-3000 BP. The population genetic structure of A. ostenfeldii in the Baltic Sea, together with long plankton records, supports this early colonisation scenario (Bj€ orck 1995; Tahvanainen et al. 2012). In contrast, the A. ostenfeldii population in the Netherlands was first reported in 2012 (Burson et al. 2014). The Dutch population has presumably colonised the small creek very recently, and may have only started with a small inoculum through the bottleneck effect selecting for functional traits supporting the establishment of an invasive species in a new environment (Carson 1959;Nei et al. 1975;Olden et al. 2004). For instance, the Dutch population exhibited a stronger expression of defensive traits such as the production of toxins and allelopathic compounds, which might be more important in novel environments with high interspecific competition. The recent establishment of the Dutch population is supported by genetic analysis of the A. ostenfeldii isolates using microsatellites. The applied markers could not differentiate between all the different Dutch genotypes, suggesting that the tested strains were genetically very similar while showing a high variation in phenotypic trait expression, which is consistent with rapid evolution. The same microsatellite markers clearly differentiated the Baltic Sea isolates and thus indicate larger genetic differences (Table S1). The observed difference in intraspecific variation between the Baltic Sea and Dutch population might thus represent an example of established and founder populations, respectively (Carson 1959;Dlugosch & Parker 2008). The differences in trait expression between both populations may not only derive from different evolutionary histories, but also from distinct environmental conditions. Both populations occur in relatively calm and shallow water bodies, with high nutrient inputs from the surrounding land. Yet, the environmental conditions in the F€ ogl€ o area of the Baltic Sea are likely more stable than in the creek where the Dutch A. ostenfeldii population resides, which is strongly influenced by nutrient and salinity fluctuations caused by rainfall (Brandenburg et al. 2017). In contrast to the Dutch creek, nutrient conditions quickly become limiting at the F€ ogl€ o bloom site in the Baltic Sea (Hakanen et al. 2012). The cell density of competing phytoplankton species or abundance of grazers during the blooms in the Dutch creek appears higher than during blooms in the F€ ogl€ o area (Fig. S6). The more stable conditions in the F€ ogl€ o area together with a relatively weaker competition with other algal species, lower grazing pressure and stronger nutrient limitation may have selected for smaller and faster growing, and thereby less toxic genotypes. In contrast, the more intensive competition, the higher grazing pressure and the higher availability of nutrients in the Dutch creek may have selected for more toxic and allelopathic, and thereby larger but slower growing genotypes. In addition, a higher genetic diversity and trait variation within populations can be attained through gene flow (Slatkin 1987;Sexton et al. 2014). In the northern Baltic Sea bloom areas, multiple differentiated local A. ostenfeldii populations have established in close proximity due to a highly fragmented coast line (Tahvanainen et al. 2012). Gene flow between these different populations allows for genetic exchange, which is in a strong contrast to the isolated bottleneck population from the Dutch creek. This may explain the higher degree of differentiation in the Baltic Sea isolates. We note that the described traits were derived from cultures grown under a constant salinity and temperature, while both factors can vary substantially during a bloom and may affect the expression of traits.
Overall, we found considerable intraspecific trait variation within and across populations of A. ostenfeldii, which was for one trait even comparable to the interspecific trait variation across marine phytoplankton species. Such large trait variation can be achieved and maintained through shifting selection pressures during toxic algal bloom periods, which can possibly contribute to their prevalence and success (Driscoll et al. 2015). Varying ecological pressures may thus result in high functional trait variation within populations, and lead to intraspecific trade-offs between, for instance, defensive and competitive traits. Moreover, our results suggest the importance of evolutionary history that may have led to the higher trait variation in the Baltic Sea population. We here report a high functional trait variation within A. ostenfeldii populations that allows for rapid diversification, as was observed for the Dutch population. This intraspecific trait variation may facilitate bloom development under dynamic environmental conditions and promote the success of HABs under global environmental change.