Four strains of E. huxleyi (National Center for Marine Algae: CCMP379, CCMP374, and Plymouth Algal Culture Collection: DHB607, DHB624) were used in these experiments. All phytoplankton cultures were grown in 0.2-μm sterile-filtered autoclaved natural seawater (FSW), enriched with f/2 nutrients minus silica (Guillard and Hargraves, 1993). All cultures were maintained on a 12:12 h light:dark cycle at 18°C, salinity of ~30, and a light intensity of 85–100 μmol photon m⁻² s⁻¹ for the phytoplankton cultures, and 20–30 μmol photon m⁻² s⁻¹ for O. marina. The grazer, O. marina (LB1974; UTEX Culture Collection) was cultured in FSW only. Isochrysis galbana (CCMP1323) was used to rear the predator prior to grazing experiments. Although the cultures were maintained using rigorous aseptic technique, the cultures were not received as axenic, nor were they axenic at the time of the experiments performed in this study. Phytoplankton cultures were transferred every 7–10 days to maintain exponential growth. Prior to the grazing experiments, stocks of the predator were maintained on I. galbana (50,000 cells ml⁻¹ final concentration), fed twice weekly. Unless otherwise specified, cell concentrations of both predator and prey cultures were determined by microscope counts using samples fixed with 1% Lugol's solution.

This experiment was as performed as in the companion study by Harvey et al. (2015). Briefly, the functional grazing response of O. marina to four strains of E. huxleyi was measured by calculating the change in prey abundance over 48 h, at five initial prey concentrations (5000–80,000 cells ml⁻¹; 70–1080 ng C ml⁻¹) for each strain examined. We chose four strains of E. huxleyi that displayed phenotypic variation in calcification states and intracellular DMSP concentration. Initial grazer concentration was ~200 cells mL⁻¹. Seawater-, grazer-, and prey-only controls were also monitored. Prior to the experiment set up, O. marina cultures were starved for 24 h, and the concentration of I. galbana was less than 1000 cells mL⁻¹ at the beginning of the experiments. Organisms were first combined in 2 L cultures at the appropriate concentrations; mixed gently, and then triplicate aliquots were gently poured into 250 mL clear polycarbonate bottles. Sample bottles were placed randomly on a plankton wheel with a rotation of 1 rpm, and maintained on a 12:12 h light:dark cycle at 18°C, and light intensity of 20–30 μmol photon m⁻² s⁻¹ for 48 h. The 2 L bottle was sampled for initial measurements and the polycarbonate bottles after 48 h. After 48 h, aliquots were taken for cell abundances to calculate grazing coefficients, ingestion rates, and growth rates of predators (O. marina) and prey species (E. huxleyi; Frost, 1972; Heinbokel, 1978).

To calculate E. huxleyi abundance, 200 μL aliquots were taken from each replicate and pipetted into a 96-well plate and run on a flow cytometer (Guava, Millipore, Billerica MA). Prior to the experiment, the optimal flow cytometer settings for E. huxleyi were determined based on chlorophyll a (692 nm) fluorescence and side scatter. To enumerate O. marina, 5 mL aliquots were preserved using 1% Lugol's solution and enumerated using a Sedgwick Rafter counting chamber and a ZEISS light microscope.

From each replicate at the end of the experiment, 80 mL of culture was decanted into duplicate 50 mL falcon tubes (40 mL of culture in each). Samples were also taken at t = 0 h for the following treatments: prey only, predator only, and three replicates of predator + prey. Cells were removed by gentle centrifugation (3000 rpm, 10 min) and the supernatant was filtered through GF/F filters (GE Healthcare Bio-Sciences, Pittsburgh, PA) and the filtrate was collected and frozen at −20°C until extraction. Dissolved organic constituents in the filtrate were extracted using Waters Oasis HLB columns (200 mg/6 mL) as follows: cartridges were conditioned with 3 mL ethyl acetate followed by 3 mL methanol; then equilibrated with 3 mL of MilliQ water. Samples (80 mL of filtrate per replicate) were loaded onto the column (at a flow rate of ~1 mL min⁻¹), followed by a 4 mL wash of MilliQ water to remove salts. We eluted organic compounds with 2 mL of methanol followed by 2 mL ethyl acetate, repeated once each, resulting ~8 mL final extract volume. Extracts were then dried in vacuo, using a Savant speed vac concentrator at 50°C.

For metabolic footprinting (Hollywood et al., 2006), extracts were re-dissolved in 95:5 water/methanol and were either normalized to the volume of culture extracted, or the number of E. huxleyi cells. Because we attempted to assess the responses E. huxleyi populations to grazing, extracts of strains CCMP374, DHB607, and DHB624, were brought to equivalent concentrations based on the number of E. huxleyi cells present at the end of the 48 h grazing experiment, (e.g., Poulson-Ellestad et al., 2014; Mausz and Pohnert, 2015), in order to observe per cell amounts of compounds in the environment. For the purposes of this study, we were concerned with only pronounced responses of the E. huxleyi exometabolome to the presence of the grazer O. marina. We therefore combined replicates in silico across E. huxleyi cell concentrations in order to search for chemical features, or metabolites, that varied in concentration due to grazing. Because strain CCMP379 was heavily grazed, samples were brought up to volumetric equivalents, instead of being normalized to cell numbers. Extracts of “grazer only” controls were reconstituted at equivalent O. marina cell concentrations. Seawater only controls were also filtered, extracted, dried, and re-dissolved at the same concentration factor as the most concentrated grazed E. huxleyi samples, in a conservative effort to identify background chemical features present in the filtered seawater medium.

Samples were then profiled as in Whalen et al. (2015) with an untargeted metabolomics approach, using high performance liquid chromatography coupled to time of flight mass spectrometry (Agilent Technologies 6230 Time of Flight MS with a Dual Agilent Jet Stream Electrospray Ionization; HPLC/MS-ToF). Our reversed-phase HPLC solvent scheme was as follows: hold at 5% methanol for 5 min, ramp to 40% methanol over 5 min, and hold at 40% for 2 min. Then ramp to 95% methanol over 5 min, hold at 95% methanol for 3 min, drop to 5% methanol, and hold for 3 min. All solvents were acidified with formic acid, 0.1% (v/v) final concentration. The first 2 min of each run was diverted into waste to avoid contaminating the mass spectrometer with salt, with a 5 min column equilibration time between samples. We used a Kinetex C18 column (2.6 μm; 2.1 × 150 mm; Phenomenex), with a flow rate of 0.2 mL min⁻¹ and injected 3 μL of sample. The column temperature was held at 35°C.

Extracts were run in positive ionization mode, scanning from 20 to 3000 m/z, with mass correction ions of 922.0098 and 121.0509 m/z that were injected at the ESI source. Gas and sheath gas temperatures were 350°C, drying gas flow rate was 8 L min⁻¹ and sheath gas flow rate was 10 L min⁻¹. Nebulizer pressure was 40 psi, with capillary voltage of 6500 V and a nozzle voltage of 1000 V. Fragmenter and skimmer voltages were 135 and 65 V, respectively. To minimize the impacts of batch variability since each grazing experiment was performed separately, one strain at a time; we analyzed strains in independent runs and only statistically compared profiles within each strain of E. huxleyi.

Spectral processing was performed in MZmine 2.11 (Katajamaa and Oresic, 2005) and included peak picking (minimum spectral count of 5000), chromatogram building, retention time normalization, and spectral alignment. All MZmine settings were as reported in Whalen et al. (2015) and Macintyre et al. (2014). This resulted in a peak list containing unique m/z-retention time pairs, referred to as chemical features, as well as peak areas from each sample. After generating a peak list, the list was filtered to only include features that were (1) present in two of three replicates in at least one treatment group. (2) present at abundances >10 times that of the seawater blanks, and (3) present at concentrations greater than twice that of the “grazer only” controls. In such cases where polyethylene glycol (PEG) was present in our extracts (particularly in strains DHB607 and DHB624), PEG ions were easily detected and manually removed due to their retention times between 15 and 18 min, and characteristic in-source fragmentation pattern with neutral losses of 44 a.m.u. Molecular formula generation was performed on selected molecular ions using the molecular formulae generator in Mass Hunter Workstation version B0.4.00. Molecular formulae were then searched against the Metlin database for metabolite annotation (Smith et al., 2005).

Before multivariate statistical analyses were performed, data were centered and variance stabilized by autoscaling (van Den Berg et al., 2006) in the PLS Toolbox in Matlab version 8.2.0.7. Multivariate analyses including principal component analysis (PCA) and orthagonalized partial least squares-discriminant analysis (OPLS-DA, cross validated with venetian blinds with six splits) were performed in PLS Toolbox (Poulson-Ellestad et al., 2014 and citations within). Loadings of chemical features from PCA were examined to determine their influence on the differences between exometabolomes (Robertson et al., 2007). Univariate unpaired t-tests comparing relative peak areas with a false discovery rate (FDR) correction (< 0.05; Benjamini and Hochberg, 1995) were performed in Microsoft Excel, while unpaired t-tests on principal component scores were performed in Graph Pad Prism v. 6.05. We also performed 2-way ANOVAs with Sidak's multiple comparisons test in order to determine the influence of E. huxleyi cell concentrations on chemical footprints. Pearson correlation coefficients were calculated, in Graph Pad Prism, in order to determine correlations between metabolite peak areas and predator activities; specifically grazing coefficients, predator ingestion rates, and predator growth rates.

To aid in the annotation of a particular chemical feature (m/z 144.0479) we also analyzed authentic standard of the metabolite 4-methyl-5-thiazoleethanol (Sigma-Aldrich Cat. 190675-25G), diluted to 8 μM in 95:5 water/methanol, with the same chromatographic and spectral methods described above.

We pooled all replicates across E. huxleyi cell concentrations as a single treatment in order to facilitate our search for metabolites that varied in concentration due to grazing (Figure 1), but see Harvey et al. (2015) for important discussions regarding the response of O. marina as a function of E. huxleyi concentration. We observed different grazing coefficients as well as predator growth rates (μ; Supplementary Figure 1), when O. marina was fed four different E. huxleyi strains. When comparing grazing on noncalcified E. huxleyi, O. marina grazed less on strain CCMP374 than CCMP379, with grazing at rates of 0.84 (± 0.16) d⁻¹ and 1.6 (±0.33) d⁻¹, respectively, (ANOVA with Tukey post-test; p < 0.0001, n = 10–13), when grazing coefficients were averaged across all concentrations of E. huxleyi tested (Figure 1). O. marina also grazed on the two calcified strains of DHB607 and DHB624 with significantly different rates of 0.27 (±0.20) d⁻¹ and 0.80 (±0.13) d⁻¹, respectively, indicating that O. marina does not graze all strains of E. huxleyi equally (p < 0.0001; n = 13; Figure 1). Strains DHB624 and CCMP374 were grazed at nearly identical rates (p = 0.97, Figure 1).

From strain CCMP374, which was heavily grazed, we limited our analysis to 24 chemical features (i.e., distinct m/z –retention time pairs), that remained after applying our selection criteria. Therefore, although we initially detected 553 chemical features we focused our efforts on less than 5% of these features.

In order to explore the overall response of the CCMP374 exometabolome to grazing, we analyzed the peak areas of the 24 metabolites from strain CCMP374 using PCA, and found a striking and significant separation between exometabolomes from grazed cultures and prey only control cultures along the first principal component (unpaired t-test, p < 0.0001; Figure 2). Additional analyses revealed that there was a significant influence of grazing, but not of E. huxleyi cell concentration on the PC1 scores (2-way ANOVA with Sidak's multiple comparison test; effect of grazing p < 0.0001, cell concentration p = 0.6145; Supplementary Table 1, Supplementary Figure 2). There was a significant impact of the interaction between grazing and E. huxleyi cell concentration (grazing x cell concentration p = 0.0128; Supplementary Table 1). The PCA model indicated that there were nine metabolites with increased concentrations as a result of grazing (features with loadings >0.2 on principal component 1; Supplementary Table 2). Additionally, there was a group of five metabolites that were found to be more prevalent in prey only samples, (features with loadings < -0.15, Supplementary Table 2), indicating that these chemical features may reflect normal activity of CCMP374 populations. Of all metabolites with concentrations significantly altered by the presence of grazers in the PCA model, three were not found in the grazer only treatment, indicating that these are most likely produced by E. huxleyi cells in response to O. marina.

We detected several metabolites that could be candidates for biomarkers of protist grazing on CCMP374. Twelve metabolites had significantly varying concentrations as a result of grazing: ten metabolites had positive fold changes from 2.55 to 6.36 and two metabolites with −1.82 and −2.54-fold reductions in concentration (t-test with FDR < 0.05 correction; Figure 3; Supplementary Table 2). Only four of these metabolites were not detected in the exometabolomes of grazer only cultures, three of which (m/z 251.1500, 329.1104, and 400.1467; potential molecular formulae C₁₂H₁₈N₄O₂, C₁₃H₂₀N₄O₂S, and C₁₅H₂₁N₅O₈, respectively) were not detected in E. huxleyi only cultures, indicating that these metabolites are E. huxleyi stress response compounds or O. marina-specific digestion metabolites (Supplementary Table 2). Additionally, one metabolite (m/z 399.2340, retention time 19.1 min with potential molecular formulae C₁₇H₃₀N₆O₅ or C₁₆H₃₄N₂O₉) that was also absent in extracts of O. marina cultures, experienced a −2.54-fold reduction in relative concentration in grazed cultures, suggesting this chemical feature is an indicator of non-stressed E. huxleyi strain CCMP374.

From the highly grazed strain CCMP379, we detected 29 features that were adequate (as discussed above) for further analyses. For this strain, we normalized extract concentrations to volume extracted vs. final E. huxleyi cell concentration, as done with the other strains, due to minimal E. huxleyi cells remaining at the end of the experiment. PCA on these 29 chemical features revealed a significant separation (p = 0.0052; Figure 2) between grazed and control exometabolomes on the second principal component of a three-component model (88.3% of variance captured). When assessing the relative impact of cell concentrations vs. grazing on exometabolomes, we found significant impacts of grazing (p = 0.0003), E. huxleyi concentration (p = 0.0022) and the interaction of the two factors (p = 0.030; Supplementary Table 3). A Sidak's multiple comparisons test revealed that there was a significant difference between control and grazed samples at only the highest concentration of E. huxleyi tested (p = 0.0003, Supplementary Figure 2). There were 11 metabolites with loadings >0.15 on the second principal component, (Supplementary Table 4), suggesting their concentrations increased as a result of grazing; whereas the relative concentrations of 12 metabolites appeared to be reduced by grazing (i.e., enhanced concentrations in controls relative to grazed samples), with loadings < -0.15 on principal component 2 (Supplementary Table 4). However, the concentrations of only two features were significantly altered by grazing (Figure 3). Concentrations of one of these features, m/z 297.1015 (C₁₂H₁₆N₄O₃S, retention time 3.9 min) significantly increased 1.89-fold whereas the concentrations of the other feature, m/z 309.2049 (C₁₈H₂₈O₄, retention time 20.0 min), significantly decreased by −5.76-fold as a result of grazing (Figure 3). Both features were also present in grazer only extracts, indicating that they are possibly O. marina metabolites.

Our analysis revealed that there were still several candidate metabolites specific to E. huxleyi CCMP379, despite detecting only a few compounds that varied in concentration due to grazing. Eleven features were not found in extracts of the T0 and/or T48 h grazer only cultures, indicating that they are either E. huxleyi specific, or are only produced by O. marina during grazing (Supplementary Table 4). Of these 11, eight had negative loadings in the PCA model and negative fold changes, suggesting that together this suite of compounds may be characteristic of healthy E. huxleyi CCMP379. Additionally, three of these 11 features were found to have increased concentrations in response to grazing — all had positive loadings on PC2, but fold change was not calculated for one (m/z 227.1064, C₁₅H₁₄O₂) because it was not present in CCMP379 only cultures (Supplementary Table 4).

From the DHB624, the concentrations of 50 chemical features were analyzed. The relative concentrations of 47 metabolites were significantly altered in response to grazing (t-test with FDR < 0.05; Figure 3; Supplementary Table 5). Additionally 21 of these 50 features were not present in extracts of grazer only cultures at the final sampling point (t = 48 h), while five of these were also not present in grazer only cultures at t = 0 h. Of these five, four had significantly greater concentrations in grazed samples, with fold changes ranging from 5.32 to 9.69. These features included m/z 283.1163 (C₁₁H₁₀N₁₀, retention time 14.2 min), 322.1852 (C₁₁H₁₉N₁₁O, retention time 14.8 min), 366.2118 (C₁₃H₂₃N₁₁O₂, retention time 15.5 min); and 396.2582 (C₁₅H₂₉N₁₁O₂, retention time 18.6 min). The concentrations of one feature, m/z 297.1316 (C₁₆H₁₆N₄O₂, retention time 18.8 min) were significantly reduced by -0.33-fold in grazed samples (Supplementary Table 5).

When we analyzed these 50 metabolites with PCA, we determined that there were significant separations of chemical profiles from grazed and control cultures on the first and third principal components (p = 0.0017 and p = 0.0402, respectively; Figure 2). This was a three-component model accounting for 92.3% of total variance in the dataset. When considering the influence of E. huxleyi cell concentration on these scores from PC1, we found significant effects of E. huxleyi concentration (p < 0.0001), grazing (p < 0.0001), and the interaction of both factors (p < 0.0001, Supplementary Table 6). Results of the multiple comparisons test showed that there were significant differences between grazed and prey only controls at the three lowest concentrations of E. huxleyi tested (p < 0.0001 for all; Supplementary Figure 2). There were 14 features that had small, negative loadings on principal component one (loadings ≤ −0.05), and had relative concentrations that were significantly greater in grazed samples (Supplementary Table 5). In contrast, PCA revealed 32 metabolites with reduced concentrations in grazed samples compared to control samples. Based on the separation obtained on the third principal component, there was one feature with significantly greater concentrations in the exometabolomes of grazed cultures (m/z 549.7926; retention time 17.4 min; fold change 3.17) and that was also present in extracts of strain DHB607, however this metabolite did not appear discriminatory on the first principal component (Supplementary Table 5).

In order to narrow down which features may have predictive or discriminatory power for distinguishing between grazed DHB624 and control exometabolomes, we analyzed these data with an orthagonalized PLS-DA model, with grazed and control treatments as our two classes. This model was partially successful in classifying grazed and prey only exometabolomes, with a 23% cross validation error rate. The variance captured on the first latent variable was 34.3%. From this model, 15 features had selectivity ratios ≥1.0, indicating that they may have predictive power in distinguishing between grazed and control extracts (Rajalahti et al., 2009). The relative concentrations of 14 features were greater in extracts of grazed cultures (Supplementary Table 5). Only one feature decreased in concentrations in response to grazing (fold change −6.43). We found that the same four features that were never detected in extracts of grazer only cultures, but had concentrations significantly increased due to grazing, had selectivity ratios >1.0 (Supplementary Table 5), and are most likely E. huxleyi response compounds or exuded waste metabolites of O. marina.

The presence of a grazer did not alter the exometabolomes of strain DHB607, which is highly calcified and poorly grazed by O. marina (Figure 1). None of the 29 metabolites we analyzed had concentrations that were significantly altered in response to grazing or the presence of the grazer (t-test with FDR < 0.05; Figure 3, Supplementary Table 7). In addition, only two features were not present in the predator only extracts (Supplementary Table 7). A five PC model (87.0% of cumulative variance explained), demonstrated that there was a small, but significant separation between grazed and control exometabolomes along the second principal component, which accounts for 23.4% of total variance (p = 0.045; Figure 2). The 2-way ANOVA revealed significant effects of both grazing (p = 0.0002) and E. huxleyi concentration (p < 0.0001), with a significant impact of interaction between the two factors (p = 0.023; Supplementary Table 8). Sidak's multiple comparisons test revealed that the effect of grazing is driven by the significant differences of the chemical profiles of grazed and controls cultures at the second lowest E. huxleyi concentration tested (p = 0.0006; Supplementary Figure 2). Based on this model, we found 11 features with increased relative concentrations due to grazing, (loadings >0.18; fold changes in relative concentration ranged from 0.29 to 2.03). We also found eight features with negative fold changes and negative loadings (< −0.07) on the second principal component indicating that this might be a suite of compounds that are characteristic of non-stressed E. huxleyi DHB607. However, it is important to note that the individual relative concentrations of these features were not significantly altered due to grazing, despite being relevant in the PC model. In addition, OPLS-DA was unable to sufficiently classify grazed and prey only exometabolomes, with 38% cross validation error, and no features from this model had selectivity ratios >1.0.

One striking result was that there was no metabolite, which remained after our stringent feature filters, common to all E. huxleyi strains. However, a small number of chemical features were shared by more than one strain (Figure 4). The strains that shared the largest number of features were the non-calcified strains CCMP379 and CCMP374, which shared six features, while both strains shared four of these features with strain DHB607 (Figure 4). Concentrations of five of these shared features significantly increased in response to grazing in strains CCMP374 (while only one increased in concentration in CCMP379). However, all six (including the four metabolites also detected in extracts of DHB607) were additionally found in O. marina only cultures, as well as E. huxleyi only cultures (Supplementary Tables 2, 4, 7), suggesting they are common to marine eukaryotic cells.

When we considered the effects of grazer-specific rates on the concentrations of shared metabolites, we found several that correlated with grazer activities. For instance, concentrations of one feature, m/z 144.0479, (retention time 5 min) were positively correlated to O. marina grazing coefficients on CCMP374 (Pearson correlation coefficient r = 0.6355, p = 0.048), CCMP379 (r = 0.6860, p = 0.0096), and DHB607 (r = 0.9387, p < 0.0001). Concentrations of this metabolite also significantly increased in response to grazing in strains CCMP374 and CCMP379 when pooled across all treatments. After determining the molecular formula of this feature as C₆H₉NOS (molecular ion [M+H]+), we then annotated this metabolite as 4-methyl-5-thiazoleethanol, based on both retention time and molecular weight matches (mass accuracy 0.0–1.38 delta ppm for all strains) to the authentic standard (Supplementary Figure 3). Although concentrations of this metabolite correlated with grazing coefficients for three of our four E. huxleyi strains, it displayed variable expression patterns when considering other O. marina activities. For example, in strain CCMP374 concentration of this metabolite did not correlate with O. marina ingestion (r = 0.4893, p = 0.1512) nor growth rates (r = 0.6160, p = 0.058), yet was positively correlated with these O. marina rates in CCMP379 (r = 0.8728–0.9342, p < 0.0001 for both). However, concentrations negatively correlated with O. marina ingestion rates on strain DHB607 (r = −0.7046, p = 0.0072), indicating that concentrations of this metabolite are not altered uniformly across all strains in response to grazing pressure.

Strain specific responses to variable grazing pressures were also observed when considering the other shared metabolites between these three strains. Concentrations of three metabolites, m/z 181.1089 (C₈H₁₂N₄O, retention time 2 min), m/z 265.1295 (C₁₂H₁₆N₄O₃, retention time 4 min), and m/z 297.1015 (C₁₂H₁₆N₄O₃S, retention time 4 min), positively correlated with O. marina grazing coefficients in E. huxleyi strains CCMP374 (r = 0.7409−0.9502, p < 0.02 for all) and CCMP379 (r = 0.5817–0.6759, p < 0.02 for all), and were either positively or not correlated to O. marina ingestion and growth rates. Concentrations of one feature, m/z 265.1294 (retention time 5 min) was positively correlated with grazing, ingestion, and growth rates (r = 0.6320–0.9074, p < 0.02 for all rates) in CCMP379 while being negatively correlated with these rates in extracts of CCMP374 (r = −0.8263 − −0.9398, p < 0.003 for all). The calcified strains DHB607 and DHB624 shared one common feature m/z 549.7961; retention time 17 min; Figure 4), concentrations of which were significantly greater in extracts of grazed DHB624 cultures, but were unaffected in strain DHB607. Due to low ion abundances and isobaric interferences in this region of the spectra, we were unable to determine a plausible chemical formula for this feature. Concentrations of this feature also positively correlated with O. marina grazing rates on both DHB607 and DHB624 (r = 0.6769, p = 0.011; r = 0.7019, p = 0.0075, respectively) suggesting that this metabolite may have been produced by grazing O. marina.

We also detected chemical features that appeared to be E. huxleyi specific, although they had concentrations that were not impacted by grazing. One feature (m/z 357.2734, C₁₅H₃₂N₈O₂, retention time 18 min), common to both CCMP379 and CCMP374, was not found in the exometabolomes of O. marina cultures indicating that this feature is specific to E. huxleyi. However, concentrations of this metabolite did not significantly differ (p = 0.047, p = 0.046 t-test with FDR correction) between grazed and control cultures, despite a sometimes large, positive fold change in response to grazing (fold change 0.59 and 9.55). Additionally, peak areas of this metabolite did not correlate with grazing coefficients, ingestion rates, or predator growth rates (p > 0.20 for all correlations) for either strain. There was one metabolite shared between CCMP379 and DHB624, however retention time of this feature experienced a 2 min shift between strains. Concentrations of this metabolite also did not differ between grazed and control exometabolomes; however this feature was not found to be produced by O. marina monocultures, qualifying it as a candidate for an E. huxleyi specific metabolite.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.