We isolated Ostreococcus sp. from pico-phytoplankton community samples obtained during a RV ALKOR cruise (AL505) in 2018 (see Figure 1A and Table 1 for sampling dates and locations) using a Niskin bottle at 5 m. Community samples were immediately passed through a 35 μm sieve to remove grazers and large debris, and then further size fractioned via gentle filtration through a 2 μm membrane filter (kept filtrate) and a 0.2 μm filter (kept filter and rinsed gently). From these samples, we successfully isolated eight new strains of Ostreococcus sp. (see Table 1 for details); five from the Kiel area and three from the Bornholm Basin. We determined the species identity via 18S rRNA sequencing (Cai et al., 2010). The isolation by dilution method where samples were diluted to below 1 cell per specific volume and grown in wellplates for ca. 15–30 days, proved best for Ostreococcus of the Baltic Sea (unpubl. results). Stock cultures were kept at their respective isolation salinities in f/2 media (Guillard, 1975) that was prepared using artificial seawater. In a common garden approach, all stocks were kept at 18°C after isolation. 18°C corresponds to a mean temperature over the two regions where the strains were isolated from. All cultures were kept in gentle rotation at a 12:12 day and night cycle with light intensities of ca. 100 μE. Based on earlier studies and the characteristics of phytoplankton populations (up to million cells/mL) we assume that the isolated strains are genetically distinct (Boyd et al., 2013; Hattich et al., 2017; Godhe and Rynearson, 2017). Furthermore, SNP analysis of Mediterranean O. tauri strains isolated by the same procedure have shown that strains are indeed genetically distinct (Gwenael et al., personal communication).

The eight new strains were cultured to determine their growth rate and metabolism response to 15 and 22°C (spanning late spring and late summer temperatures). The experiment had to be carried out in three subsequent batches due to the limited number of metabolic measurements possible at the same time (see Figure 1B and Table 1). The batches were all set up the same way: each strain was replicated three times and each replicate inoculated with 3000 cells/mL in 40 mL f/2 media (Guillard, 1975) of the respective salinity of isolation (Table 1). All replicated cultures were exposed to the two treatment temperatures with a 12:12 day and night cycle at 100 μE light intensity for 18 days ensuring growth through a whole microbial growth cycle (Figure 1C). Starting at day three of microbial growth we measured cell numbers daily via flow cytometry (BD Accuri C6 flow cytometer) and starting on day 4–5, we measured photosynthetic metabolic activity daily via optical O₂ measurements (Figure 1D) (see below for explanation of measurement).

After the main experiment, we set up a second smaller experiment to determine the potential of growth on organic carbon sources using six strains of the eight isolated Ostreococcus strains (four from Kiel and two from Bornholm, respectively) (Figure 1E). Each strain was inoculated at 3000 cells/mL in 200 mL of f/2 media (Guillard, 1975) of the respective salinity of isolation and exposed to both 15 and 22°C with a 12:12 day and night cycle at 100 μE light intensity. In addition to growth and carbon use efficiency measures, we here investigated the potential of each strain to grow on 31 different organic carbon sources using ecoplates (Biolog EcoPlate^TM) at three time points during the microbial growth curve (determined via the preceding growth experiments).

We determined cell counts to calculate growth rates in the experimental cultures and increase of cell numbers in the Biolog EcoPlates. Cells were counted at a rate of 65 μl/min on a flow cytometer (Accuri BD C6 Plus) using gating of cells characterized by the height of forward scatter and red fluorescence signal (Supplementary Figure 1). The forward scatter was set to a threshold of ca. 200 that allowed us to determine debris and potentially high amounts of bacteria in the experimental cultures. Ostreococcus is large enough to be well distinguished from bacterial cells or cell debris (Supplementary Figure 2) which came in handy when measuring the Ostreococcus increase on the Biolog EcoPlates with organic carbon sources.

On the daily cell count measurements, we fitted a growth curve containing a lag phase, exponential phase and carrying capacity. To analyze the shape of the growth curves, non-linear curve fitting of a gompertz growth model (Buchanan et al., 1997) was carried out using the “nlsLoop” function in the R package, “nlsLOOP” (version 1.2-1). Parameter estimation was achieved by running 1,000 different random combinations of starting parameters for cell count at carrying capacity, duration of lag phase, and maximum growth rate picked from a uniform distribution. The script then retained the parameter set that returned the lowest Akaike information criterion (AICc) score, yielding μmax and day at μmax. In addition, we calculated growth rates at early and late exponential phase (3 days before and after day at μmax, respectively) using the following formula [ln(N_t−N_t–1)] with N being the number of cells/ml. The day at μmax was important for subsequent analysis of growth on organic carbon sources (see Supplementary Table 1).

Net photosynthesis and respiration rates were measured on PreSens^® SDR Sensor Dish optodes. We measured oxygen production for 15 min in the light, and respiration for 15 min in the dark under the light and temperature conditions set in the incubator (i.e., all experimental units at their assay temperatures). In non-axenic cultures, bacterial respiration is also measured both in light and in the dark. However, over the course of the microbial growth curve we find that bacterial counts (as part of the debris smaller than ca. 0.5 μm) did not increase or change such that we can assume that measures compared over one batch cycle are based on changes in autotrophic phytoplankton cells. All characterizations were carried out at the same time of day (9–11 a.m.). From the measurements we calculated the rate of oxygen evolution within a 5–10 min constant time window as changes from light to dark can have consequences on the absolute values measured on the PreSens SDR Sensor Dish. Due to oxygen drifts within the sample vials we always ran a control with MQ in each measurement and subtracted this drift from the rates calculated. From the rate of oxygen evolution in the dark (R) and light (NP) we then calculated gross photosynthesis as follows: [R]+NP = GPP.

Carbon use efficiency and was then calculated as follows: C⁢U⁢E=1-([R]G⁢P⁢P)

Here we were interested in, if and how much a culture could grow on a number of different carbon sources that we provided via the ecoplates (Supplementary Figure 3). Based on Exp. 1–3 we identified three time points (corresponding to three different days during the microbial growth cycle) at which we then measured growth on organic carbon. These time points were “early exponential phase” 3 days prior to day at μmax, “mid exponential phase” at μmax and “late exponential phase” 3 days later than day at μmax. The actual days of inoculation into ecoplates varied between the different strains of Ostreococcus and treatment temperatures (see Supplementary Table 2 for details). Each ecoplate contained 31 different organic carbon sources in triplicates and three controls with water (see Rutgers et al., 2016 for list of sources and groups thereof and Supplementary Figure 2). Each of the 96 wells was inoculated with 200 μl of culture (except one control of water with only MQ) and then left to grow for 24 h at the same experimental condition as the respective original culture. After 24 h we fixed the samples with sorbitol (10 μL of a 1% sorbitol solution per sample) for 24 h at 4°C in the dark and then froze them for later analysis via flow cytometry. After thawing the ecoplates overnight, we counted the cells in each well and calculated the relative change in cell numbers on organic carbon compared to the control (on water). Since our cultures were not axenic, we cannot exclude, that present bacteria readily use the organic compounds provided in the ecoplate. Therefore, we calculated the relative difference of cell numbers on an organic source specifically identifying phytoplankton cells in the cytometry fingerprints. We provide an example for control sample (no culture), culture on water (control), and on an amino acid in Supplementary Figure 2. In addition, we calculated an overall value of relative change in cell numbers on organic carbon the following way: First, we calculated the mean of cells/mL for the control wells; second, we calculated the relative change in cells/mL for each of the 93 wells that contained a carbon source compared to the control (i.e., “water,” no other organic carbon source). This relative change could be between −1 and 1: anything between −1 and 0 meant that less cells than in the control were present after 24 h whereas anything between 0 and 1 meant that more cells in the control were present after 24 h. Third, we counted the number of wells where the relative change in cell numbers was positive. And last, the overall potential use of an organic carbon source was then calculated as the mean of the relative change of cells normalized by the number of wells on which the relative increase was positive.

We found that the transfer of culture from the 150 mL culture flasks to 200 μL on the ecoplates presented a stressful condition in itself. This was seen as rapid decrease of cell numbers even in the control samples on water only. However, one well contained only water, and thus there was no change in culturing condition other than being transferred into a smaller volume and we could still determine the effect of the presence of an organic carbon compound on the Ostreococcus strains.

All data were analyzed in the R programming environment (version 4.0.3.) using the packages “nlme,” “ggpot2,” “lme4,” “emmeans,” “vegan,” “reshape2,” and “multcomp.”

The growth rates (μmax, growth rate at early and late exponential phase) were analyzed with a linear mixed effects model (lme within the nlme package). Growth was first analyzed via a global model that included sampling location (Kiel area or Bornholm Basin), assay temperature (15 and 22°C) and time as fixed factors in full interaction. The “experiment” (Exp 1–3 for μmax and net primary production) was computed as a nested random effect within region. The global model was reduced to the single and interacting factors containing the lowest AICc score with a minimum difference of 2 (Supplementary Table 3.1). The best model for growth only included the single factors and no interaction (Supplementary Table 3.2).

CUE, gross photosynthesis and respiration were analyzed with a generalized linear squares (gls) model (within the nlme package) that accounted for autocorrelation over time. We first constructed a global model that included the sampling location and assay temperature as fixed factors in full interaction—using the lowest AICc score with a minimum difference of 2 (Supplementary Tables 4.1, 5.1, 6.1, respectively) then yielded the model that best explained each response variable (Supplementary Tables 4.2, 5.2, 6.2, respectively.

The changes in cell numbers on the organic carbon compounds was analyzed in two ways: First we analyzed the differences in the groups of carbon compounds via a linear mixed effects model containing the effect of timing at 22°C (Supplementary Tables 7.1, 7.2) and temperature (Supplementary Tables 7.3, 7.4) in mid exponential phase as well as the effects of region and “carbon group” (Supplementary Table 8). The lme model was reduced to the single and interacting factors containing the lowest AICc score with a minimum difference of 2. Second, we confirmed the statistical analysis on the relative growth on carbon and changes in cell numbers via a PCA analysis and subsequent permanova that tested again for the effect of timing at 22°C (see above) and temperature in mid exponential phase (Supplementary Tables 9, 10). The difference to the first analysis is, that it includes the differences between all 31 carbon sources and how the complete use of all the sources differed between time-points, temperature and regions.

Following the expected shape of a microbial growth curve, the highest growth rates were reached in the middle of the microbial growth curve and lower at the early and late stage of the microbial growth curve both in the Kiel and Bornholm strains (Figure 2 and Supplementary Table 5; Global model (TI) Effect of “Timing” [F_{(2, 155)} = 619.787, p < 0.0001]. In the Kiel strains, the growth rates were higher in the early and mid-phases of exponential growth compared to the Bornholm region [Figure 2 and Supplementary Table 3.2; Effect of “Region” F_{(1, 155)} = 28.458, p < 0.0001]. All strains of Ostreococcus increased their growth rate from 15 to 22°C [Figure 2 and Supplementary Table 3.2; Effect of “Temperature” F_{(1, 155)} = 16.724, p < 0.0001] at early and late stage of exponential growth. The increase of growth rate between 15 and 22°C is higher in the Kiel strains (e.g., min and max growth mid exponential phase 0.52–1.06) compared to the Bornholm (e.g., min and max growth mid exponential phase 0.43–0.71) strains indicated in the maximum and minimum growth rates of all strains within a region. The difference between regions indicated an effect of the absolute difference in experienced past environment including the variability therein; in other words, their evolutionary history.

Both gross photosynthesis (GP) and respiration (R) decrease rapidly over the course of a batch cycle (correlation of time in statistical models) (Figure 3). The statistical analysis did not show a significant effect of either temperature or region on either GP or R (Figure 3 and Supplementary Tables 4, 5). The high variability over the course of the microbial growth cycle could be responsible. Since carbon use efficiency (CUE) puts GP and R in relation to each other per sample we have—apart from GP—another measure of primary productivity that allows us to interpret the capacity to grow on the available inorganic resource.

The CUE in both Kiel and Bornholm strains increased with temperature at mid exponential phase [Figure 3 and Supplementary Table 6.2; Effect of “Temperature” F₍₁₎ = 9.934, p = 0.002] and varied throughout the microbial growth curve (all models included correlation of time). Specifically, CUE increases in the first 6–8 days of experiment and then stays relatively constant. In addition, CUE was generally higher in the Kiel strains compared to the Bornholm strains [Figure 3 and Supplementary Table 6.2; Effect of “Region” F₍₁₎ = 11.680, p = 0.001]. 20% of the data were also explained by an inclusion of the interaction of temperature and region that explains a higher CUE especially at 22°C (Supplementary Table 6.1).

Due to experimental limitations, we were only able to measure use of organic carbon sources on one representative replicate of four strains from the Kiel area and two strains from the Bornholm area, respectively. Thus, a comparison between the strains was statistically not possible. Averaged over all the organic sources we did find a higher use of organic carbon sources by isolates from Kiel compared to Bornholm at mid exponential phase in both assay temperatures [Figure 4 and Supplementary Table 7.4; Effect of “Region” F_{(1, 4)} = 11.416, p = 0.028] as well as early and late exponential phase at 22°C [Figure 4 and Supplementary Table 7.2; Effect of “Region” F_{(1, 4)} = 6.531, p = 0.063]. In addition to the overall difference between the regions we also found effects of the timing of measurement and temperature: on the one hand, the use of organic carbon in both Kiel and Bornholm strains decreased from early to mid and late exponential phase at 22°C [Figure 4 and Supplementary Table 7.2; Effect of “Timing” F_{(2, 10)} = 12.398, p = 0.002]. On the other hand, at mid exponential phase, the use of organic carbon sources was higher at 15°C compared to 22°C in both areas [Figure 4 and Supplementary Table 7.4; Effect of “Temperature” F_{(1, 5)} = 5.730, p = 0.062].

In addition to the differences in use when all carbon sources were considered together, we found that at different sampling times and temperatures the carbon sources were used in different quantities (Figure 5). At 22°C we found that in the early exponential phase on all carbon groups except for the polymers (Cyclodextrin and Glycogen), cell numbers compared to the water control were higher in the strains from Kiel, whereas in the Bornholm strains this was mainly observed on carbohydrates and carboxylic acids [Figure 5A, Supplementary Table 8.2; Effect of “Region^∗Carbon Group” F_{(5, 1525)} = 3.798, p = 0.002]. At mid and late exponential phase, there were fewer sources on which cell numbers were higher than in the control in the Kiel strains and there was no further increase in cell numbers on carbon sources in the Bornholm strains [Figure 5A and Supplementary Table 8.2; Effect of “Timing” F_{(2, 1525)} = 67.528, p < 0.0001]. In the Kiel strains, there was still potential use of amines and carbohydrates at mid exponential and phenolic compounds in the late exponential phase [Figure 5A and Supplementary Table 8.2; Effect of “Timing^∗Region” F_{(2, 1525)} = 9.290, p < 0.0001). In addition, at 15°C compared to 22°C there were more carbon groups potentially used in both Kiel and Bornholm strains [Figure 5B and Supplementary Table 8.4; Effect of “Temperature” F_{(1, 1058)} = 37.202, p < 0.0001], however, differently so between the regions [Figure 5B and Supplementary Table 8.4; Effect of “Temperature^∗Region” F₍₁, ₁₀₅₈₎ = 9.334, p = 0.002]. In the Kiel strains resources from all groups except for the polymers lead to an increase in cell numbers and in Bornholm strains this was only the case on amines and carboxylic acids [Figure 5B and Supplementary Table 8.4; Effect of “Region^∗Carbon Group” F_{(5, 1058)} = 5.007, p < 0.0001].