Identification of Effector Metabolites Using Exometabolite Profiling of Diverse Microalgae

ABSTRACT Dissolved exometabolites mediate algal interactions in aquatic ecosystems, but microalgal exometabolomes remain understudied. We conducted an untargeted metabolomic analysis of nonpolar exometabolites exuded from four phylogenetically and ecologically diverse eukaryotic microalgal strains grown in the laboratory, freshwater Chlamydomonas reinhardtii, brackish Desmodesmus sp., marine Phaeodactylum tricornutum, and marine Microchloropsis salina, to identify released metabolites based on relative enrichment in the exometabolomes compared to cell pellet metabolomes. Exudates from the different taxa were distinct, but we did not observe clear phylogenetic patterns. We used feature-based molecular networking to explore the identities of these metabolites, revealing several distinct di- and tripeptides secreted by each of the algae, lumichrome, a compound that is known to be involved in plant growth and bacterial quorum sensing, and novel prostaglandin-like compounds. We further investigated the impacts of exogenous additions of eight compounds selected based on exometabolome enrichment on algal growth. Of these compounds, five (lumichrome, 5′-S-methyl-5′-thioadenosine, 17-phenyl trinor prostaglandin A2, dodecanedioic acid, and aleuritic acid) impacted growth in at least one of the algal cultures. Two of these compounds (dodecanedioic acid and aleuritic acid) produced contrasting results, increasing growth in some algae and decreasing growth in others. Together, our results reveal new groups of microalgal exometabolites, some of which could alter algal growth when provided exogenously, suggesting potential roles in allelopathy and algal interactions. IMPORTANCE Microalgae are responsible for nearly half of primary production on earth and play an important role in global biogeochemical cycling as well as in a range of industrial applications. Algal exometabolites are important mediators of algal-algal and algal-bacterial interactions that ultimately affect algal growth and physiology. In this study, we characterize exometabolomes across marine and freshwater algae to gain insights into the diverse metabolites they release into their environments (“exudates”). We observe that while phylogeny can play a role in exometabolome content, environmental conditions or habitat origin (freshwater versus marine) are also important. We also find that several of these compounds can influence algal growth (as measured by chlorophyll production) when provided exogenously, highlighting the importance of characterization of these novel compounds and their role in microalgal ecophysiology.

stability of algal systems, both natural and engineered, are influenced by interactions of algae with other microorganisms and with each other, and exometabolites released into the water are important mediators of these interactions (4).
The specific compositions of algal exometabolite pools are important for modulating algal growth, and the interactions of algae and their associated communities through nutrient exchange, signaling, growth promotion and inhibition, and defense. A range of different plant hormones (indole-3-acetic acid, gibberellic acid, kinetin, 1-triacontanol, and abscisic acid) have all been shown to influence the growth of the model green alga Chlamydomonas reinhardtii (5). One strain of the coccolithophore Emiliana huxleyi has been shown to produce the plant hormone indole-3-acetic acid, while the growth of another strain was impacted by the same metabolite (6). Some algae have also been shown to inhibit the growth of competitor algae through the release of allelochemicals (7). Algal exudates have been hypothesized to play similar roles in the phycosphere, the microenvironment around algal cells, to those of root exudates in the rhizosphere (8). In plant systems, specific metabolites in root exudates have been shown to impact the growth of different microorganisms in the rhizosphere, favoring some microbial taxa over others (9). Similarly, combinations of metabolites known to be produced by algae have been shown to modulate microbial community composition depending on the combination of metabolites (10). In another study, rosmarinic acid and azelaic acid were shown to be produced by the diatom Asterionellopsis glacialis in response to bacterial presence and were also found to impact bacterial growth (11). Dimethylsulfoniopropionate produced by E. huxleyi feeds the bacterium Phaeobacter inhibens, but when the alga begins to produce p-coumarate, the bacterium responds by shifting its role from a mutualist, promoting algal growth, to a parasite, through the release of exometabolites (12).
Most studies of algal metabolomics have focused on the composition of intracellular metabolomes. For instance, of the organisms used in our study, the intracellular metabolomes of Chlamydomonas reinhardtii and Phaeodactylum tricornutum have been characterized for their responses to different growth conditions (13,14). In contrast, microalgal exometabolomes, particularly for marine microalgae, have not been well characterized, partly due to the technical challenges of low metabolite concentrations and high salt content in the samples (15). Some recent studies have begun to analyze algal exometabolomes: Shibl et al. investigated the exometabolome of the diatom Asterionellopsis glacialis, and its response to the addition of a microbial community (11), while Ferrer-González et al. characterized the exometabolome of the diatom Thalassiosira pseudonona in comparison to cocultures of the diatom with three different bacterial strains (15). Another study used an untargeted metabolomics analysis to compare the exometabolomes of cyanobacteria (three Prochlorococcus strains and two Synechococcus strains) to diatoms (two Thalassosira species and one Phaeodactylum) (16). Metabolite exudation and uptake have also been studied in the cyanobacterium Synechococcus sp. strain PCC 7002 (17). To better understand microalgal interactions, there is a need to characterize exometabolites from a broader range of microalgae.
In this study, we characterized the nonpolar exometabolomes of four phylogenetically and ecologically diverse eukaryotic microalgal strains: the freshwater green alga Chlamydomonas reinhardtii, the brackish green alga Desmodesmus sp. strain C046, the marine diatom Phaeodactylum tricornutum, and the marine eustigmatophycean alga Microchloropsis salina (also referred to as Nannochloropsis salina). We chose these strains to cover a phylogenetically diverse group of microalgae with different environmental origins. C. reinhardtii and Desmodesmus are the most closely related of these algae, both belonging to the plant kingdom in the phylum Chlorophyta and the class Chlorophyceae. P. tricornutum and M. salina are both members of the kingdom Chromista but belong to different phyla. Additionally, C. reinhardtii and P. tricornutum are both model organisms that have been well studied and whose genomes have been sequenced (18,19). Desmodesmus is also of particular interest because it can grow in both freshwater and saltwater, allowing us to examine metabolomic shifts under these two conditions. In addition to profiling the exometabolomes, the goal of our analysis was to determine whether specific exometabolites detected from these strains impacted algal growth and might therefore be effector metabolites involved in interactions within or between algal strains.

RESULTS
Algal growth. Algal chlorophyll fluorescence, an indicator of algal growth, increased with incubation time for all algal strains through the eighth day after inoculation, when samples were collected for metabolomic analysis (Fig. 1). Chlorophyll A fluorescence began to decrease by day 12 for P. tricornutum and by day 18 for all algal cultures except for C. reinhardtii, indicating that all cultures were in mid to late exponential growth when samples were collected. Ash-free dry weight measurements were made for the sampling times before and after day 8 (day 4 and day 12). Average biomass (ash-free dry weight) differed between the highest and lowest biomass alga by 1.6-fold on day 4 and 1.7-fold on day 12 (see Fig. S1a in the supplemental material). Increasing biomass between days 4 and 12 supports that at day 8 the algae were all in mid to late exponential growth.
The total organic carbon concentration of algal spent medium was measured for samples from day 8. Organic carbon concentrations ranged from 4 6 2 ppm (330 6 170 mM) for Desmodesmus (SW) to 13 6 2 ppm (1,080 6 170 mM) for C. reinhardtii (mean 6 standard deviation for five replicates) (Fig. S1b). Based on these measurements, the total organic carbon loading from 30 ml spent medium onto solid-phase extraction columns ranged from 0.1 mg to 0.4 mg, well below the manufacturer listed maximum loading capacity of the columns (5 mg to 50 mg). Spent medium from algae grown in freshwater (C. reinhardtii and Desmodesmus [FW]) had higher concentrations of organic carbon than spent medium from algae grown in saltwater (P. tricornutum, M. salina, and Desmodesmus [SW]). Within each growth condition, the relative concentrations of organic carbon in spent medium followed the patterns of chlorophyll A fluorescence (C. reinhardtii . Metabolite profiles of diverse microalgae. To minimize matrix effects resulting from differences in salt concentrations, we focused on nonpolar metabolite analysis using both C 18 -based solid-phase extraction and C 18 -based liquid chromatography. It is important to note that this extraction method recovers only a fraction of the dissolved organic carbon present in the sample and is biased toward recovery of nonpolar metabolites. To assess matrix effects on metabolite detection, we analyzed detection of 13 C-and 15 N-labeled internal standards. Signal suppression of internal standards was generally low and comparable across sample groups, with the exception of the saltwater exometabolite samples (Fig. S2). For saltwater exometabolite samples, the median log 2 fold change compared to the value for extraction controls ranged from 21.4 to 22.0 for the different sample groups. For all other sample groups, median log 2 fold change ranged from 20.15 to 20.68.
Initial untargeted analysis of liquid chromatography-tandem mass spectrometry (LC-MS/MS) data for all cell pellet, spent medium, and control samples detected 14,916 features, of which 4,259 features were determined to be significantly different from medium blank controls (adjusted P value of ,0.05 and fold change compared to control value of .2) for at least one sample group, where a sample group consisted of all five replicates for a single organism and sample type (cell pellet or spent medium) combination (see Data Set S1 in the supplemental material).
Within the subset of significant features, overall exometabolome composition differed between the different algal cultures (Fig. 2). Principal-component analyses of the exometabolite profiles separated samples from the different algae, including separating metabolite profiles for Desmodesmus grown in freshwater from Desmodesmus grown in saltwater (Fig. 2). The first two principal components accounted for a total of 67.4% of the variance in exometabolite profiles. Saltwater cultures separated along the first principal-component axis, and freshwater cultures separated largely along the second principal-component axis.
Hierarchical clustering of LC-MS/MS (Fig. 3, top dendrogram) revealed distinct sets of predominantly cell pellet features (gray) and predominantly extracellular features (purple), as well as distinct sets of features associated with the different algal strains. Across all 4,259 features and sample groups (sample type and organism), the median log 2 fold change in one sample group compared to the sample group with maximum signal was 25.6, corresponding to a 48.4-fold change. Hierarchical clustering of samples (Fig. 3, side dendrograms) indicated that Desmodesmus metabolite profiles from freshwater and saltwater were distinct but were more similar to each other than to metabolite profiles from other algae. Further, although Desmodesmus and C. reinhardtii are more closely related to each other than to the other algae, their metabolite profiles did not cluster together in the hierarchical clustering analysis. To further assess whether the metabolite profiles correlated with phylogeny, we conducted a Mantel test comparing phylogenetic distances between algae to Euclidean distances between normalized exometabolite profiles. That test did not detect any significant correlation between phylogeny and exometabolome (r = 0.07 and P value = 0.29 when the saltwater metabolite profiles were used for Desmodesmus and r = 0.17 and P value = 0.17 when the freshwater metabolite profiles were used).
To obtain putative annotations for some features and to assess potential relationships between features through feature-based molecular networking, we analyzed the peak detection results with the Global Natural Products Social Molecular Networking (GNPS) tool (20,21). This analysis resulted in preliminary spectral matches for 263 features based on tandem mass spectrometry (MS2) mass spectra, of which 137 (52%) had signal intensities at least twofold greater than controls for at least one sample group (see Table S1 in the supplemental material). This represents 3.2% of all 4,259 features that were significantly greater than medium controls. For features of interest discussed below, we manually curated the best spectral matches (from up to 100 matches per feature) based on the GNPS cosine score, delta m/z, number of shared peaks, and adducts. Final curated annotations are listed in Table S1. Only the features with the highest quality matches (GNPS cosine score of .0.7, delta m/z of ,10 ppm, number of shared peaks . 5, common adducts: [M1H] 1 or [M1Na] 1 for positive mode or [M-H] 2 or [M1Cl] 2 for negative mode) were considered putative annotations due to a lack of retention time information for authentic standards, with a metabolomics standards initiative (MSI) level 2 identification (22,23). Two metabolites were subsequently confirmed using authentic standards as described below.
Characterization of algal exudates. To differentiate metabolites resulting from lysis versus exudation, we identified features significantly enriched in the exometabolome in paired samples of extracellular and cell pellet metabolites. Enrichment was assessed based on the ratio of peak heights between paired exometabolome and cell pellet samples. Because of inherent differences between sample types, this ratio should be interpreted as a qualitative estimate of enrichment. Of the 138 features with spectral matches from GNPS, 41 (29%) were significantly enriched in the exometabolome of at least one alga (Fig. 4). Enrichment ratios for these features ranged from 2.1 to infinite (feature not detected in cell pellet), with a median enrichment ratio of 6.3.
Although most exometabolome-enriched features were enriched in the spent medium of only one alga, two features (positive features 4889 and 7957, identified as lumichrome and 59-S-methyl-59-thioadenosine, respectively) were enriched in four of the five algal systems tested. The identities of these features were further confirmed by matching retention times to standards analyzed in our laboratory, resulting in MSI level 11 identifications. The levels of lumichrome and 59-S-methyl-59-thioadenosine differed between algal cultures, with P. tricornutum producing the highest levels of both metabolites among the exometabolomes (Fig. S3). Additionally, four features were enriched in three of the algal systems, and three were enriched in two of the algal systems (Fig. 4).
Three features potentially related to prostaglandins were enriched in the exometabolomes of some algae, with at least one of these features enriched for all algal strains tested except Desmodesmus. Across the full data set, seven features had best spectral matches to FIG 4 Exometabolome-enriched metabolites with putative annotations. Circle sizes are proportional to the base 2 logarithm of the average relative enrichment ratio in the exometabolome compared to the cell-associated metabolome for five biological replicates. MSI level 11 identified compounds are indicated with an asterisk. Spectral matches in bold type indicate high-quality putative annotations (MSI level 2). Spectral matches in plain (roman) type indicate quality matches to unusual adducts, suggesting related molecules with different chemical formulae, with potential difference indicated in parentheses (MSI level 3). Spectral matches in italic type indicate matches with high m/z error, suggesting different, but potentially related compounds. Spectral match names in purple were selected for exogenous metabolite addition experiments.
prostaglandins, all with either spectral matches to unusual adducts or had m/z errors greater than 10 ppm, indicating that these were not exact matches but instead represented potentially related compounds. To further understand this group of features, we evaluated these features in the context of the GNPS feature-based molecular networking analysis (20), which resulted in five subnetworks containing features with spectral matches to prostaglandins (Fig. S4). The features in these subnetworks were generally associated with individual algal strains, with different features being dominated by different strains. In addition to prostaglandin-like features, others had high spectral similarity to steroids and other lipids. Moreover, many of the edges in the molecular network (linking two features with similar spectra) corresponded to m/z differences indicating the specific loss or gain of small numbers of carbons and hydrogens (C 2 H 2 , C 4 H 4 , etc.), suggesting that this may represent a related group of lipids with small variations in chemical formulae.
Several features putatively annotated (MSI level 2) as dipeptides and tripeptides were also enriched in the exometabolomes of the different algae, with each alga having at least one of these enriched in its exometabolome. These features (positive features 23650, 12382, 12591, and 16719 and negative feature 6526) were putatively annotated as Ala-Phe-Arg, Ile-Arg, Phe-Arg, Val-Arg, and Tyr-Trp, respectively. Across the full data set, di-and tripeptide features were highly associated with the exometabolomes compared to cell pellet metabolomes, with the exception of Arg-Trp which was most abundant in the cell pellet metabolome of C. reinhardtii (Fig. S5).
Response of Desmodesmus to medium conditions. Desmodesmus sp. strain C406 is unique among the algae tested here in its ability to grow in both the saltwater and freshwater medium. Between the two growth media, both the extracellular and cell pellet metabolite profiles differed significantly. Because high salt concentrations can interfere with detection of compounds by LC-MS/MS, making some metabolites appear to be lower or not detectable from high salt samples due to matrix effects (24,25), we focused on metabolites that were detected at significantly higher levels in the saltwater samples. Three features with spectral matches to database compounds were significantly enriched in the exometabolome of Desmodesmus grown in saltwater compared freshwater (Fig. S7). These enriched features included the putatively annotated dipeptides Tyr-Trp and Arg-Trp, as well as negative feature 1013, which had high spectral similarity to the [M-H-H 2 O] 2 adduct of dodecanedioic acid. Several of the metabolites resulted in decreased growth in Desmodesmus (FW) but did not have significant effects on Desmodesmus (SW). Notably, in the exogenous metabolite addition experiment above, Desmodesmus responded positively (44% increase in chlorophyll A) to dodecanedioic acid when grown in saltwater but negatively when grown in freshwater (14% decrease in chlorophyll A) (Fig. 5).

DISCUSSION
Our study builds on previous work to expand our understanding of microalgal exometabolome diversity, composition, and impacts on algal interactions. We set out to characterize exometabolomes from a phylogenetically and ecologically diverse set of microalgae, including two green algae (one freshwater, one freshwater and marine), a marine diatom, and a marine eustigmatophycean alga to discover exometabolites that may mediate algal interactions and ecophysiology. Given our interest in comparing algal responses to different salinities, we focused our analysis on nonpolar metabolites and used solid-phase extraction to minimize the influence of salinity on metabolite analysis. We find the following. (i) Exometabolite profiles are distinct between these eukaryotic microalgal strains, but we did not detect any correlation of exometabolomes with phylogeny among this limited number of strains. (ii) Based on the distinct responses to exogenous addition and salt and freshwater culturing, exometabolite profiles may instead primarily reflect algal ecophysiology. (iii) Lumichrome, dipeptides, and prostaglandin-like compounds were enriched in the algal exometabolomes, none of which are well characterized in algal systems. (iv) Several compounds acted as effector metabolites, having significant impacts on algal growth. The exometabolites profiled may play roles in growth promotion, defense, and interactions as has been shown in other systems.
In our comparison across phylogenetically diverse algal strains, we found distinct metabolite profiles associated with each strain. Although we did observe matrix effects on the detection of internal controls (see Fig. S2 in the supplemental material), the fold changes for features between different cultures (median log 2 fold change of 25.6) were much larger than the observed matrix effects (median log 2 fold change of 20.15 to 22.0 for different sample groups) and can therefore be interpreted as real differences between metabolite levels in different cultures. One previous study has compared the exometabolomes of two cyanobacterial groups (three Prochlorococcus strains and two Synechococcus strain) with diatoms (two Thalassosira species and one Phaeodactylum) (16). That study found that the differences in the exometabolome profiles reflected the phylogenetic differences (prokaryote versus eukaryote) between these groups. Our results suggest that this phylogenetic signal is not consistently present within eukaryotic microalgae. C. reinhardtii and Desmodesmus, both members of the phylum Chlorophyta and class Chlorophyceae, are more closely related than the other strains used in our study (a diatom and an eustigmatophycean alga). However, their metabolite profiles did not cluster together in our analysis (Fig. 3, side dendrograms). Further, Mantel tests did not detect significant correlation between algal phylogenetic distance and exometabolome composition distance. A study of both macro-and microalgae found that environmental origin influenced intracellular metabolome composition more than phylogenetic relatedness (26). The results of our study suggest that exometabolite profiles are also not phylogenetically determined. However, because we studied only four algal strains, more work is needed to fully investigate the impact of phylogeny on microalgal exometabolomes.
The distinct profiles and responses of Desmodesmus under freshwater and saltwater conditions as well as the distinct responses of the microalgae to the exogenously added metabolites suggest that algal ecophysiology may be an important determinant of exudation. Given the signal suppression at high versus low salt concentrations that was observed in the analysis of the internal standards (even using solid-phase extraction) (Fig. S2), we focus on metabolites that increased with high salt. We find that Desmodesmus drastically shifts its exometabolome in response to salt-and freshwater culturing and has contrasting growth responses to addition of three of the eight tested metabolites under the different conditions, highlighting the responsiveness of the exometabolome to shifting environmental conditions. There have been some studies examining intracellular responses to different salinities by a variety of microalga, including Scenedesmus sp., Amphora subtropica, Dunaliella sp., Desmodesmus armatus, Mesotaenium sp., and Tetraedron sp. (27)(28)(29)(30). Our work supports the finding that salinity impacts algal metabolite composition and extends it to the extracellular environment, suggesting that these differences will influence the surrounding community as well.
We assessed exometabolome enrichment as the ratio of exometabolome to cell pellet peak heights in paired samples. This ratio normalizes exometabolite peak heights to cell pellets and provides an estimate of exometabolome enrichment to enable identification of exometabolites of interest. However, this ratio is an imperfect approximation of enrichment. One limitation of this approach is that exometabolites accumulate over the course of the 8-day experiment, while endometabolites accumulate over the generation time of 1.3 to 2.4 days (estimated from algal growth data [ Fig. 1]), with biomass concentrations increasing over time. Thus, the time scales of accumulation differ by a factor of approximately four. Because exometabolites are accumulating on the longer time scale, this may lead to an overestimate of enrichment based on this ratio.
Additionally, matrix effects, which can lower detection of metabolites in high salt samples (24,25), differ between exometabolite and cell pellet samples, particularly for the saltwater cultures (Fig. S2). Saltwater exometabolite samples showed more signal suppression of internal standards than corresponding cell pellet samples. Thus, matrix effects also impact this ratio, biasing it toward an underestimate of enrichment, in the opposite direction from that expected for the potential time scale bias. Because of these uncertainties, it is important to note that this ratio is a qualitative measure of enrichment used to identify metabolites present at significant levels outside the cell. To minimize false-positive results in this analysis, we required enrichment ratios of at least two, in addition to statistical significance, to identify exometabolome-enriched metabolites in this analysis.
Several of the exometabolome-enriched metabolites may act as effector metabolites, with potentially important roles in algal growth and interactions with bacteria. Lumichrome (MSI identification level 11), which was enriched in the exometabolomes of all cultures except M. salina, has been found to promote plant development and growth and has also been shown to impact bacterial quorum sensing (31,32). Chlamydomonas is known to produce lumichrome (31), but lumichrome production has not been reported previously in P. tricornutum or Desmodesmus. Most previous studies of algae and lumichrome or riboflavin (vitamin B2), of which lumichrome is a derivative, have focused on possible lumichrome/riboflavin production by algal growth-promoting bacteria, as opposed to algal production (33)(34)(35). In addition to biological production, lumichrome can be produced through photochemical degradation of riboflavin (36). Since biomass concentrations were similar across the algal strains, it is unlikely that differences in light penetration could account for the observed differences in lumichrome accumulation. However, observed lumichrome levels likely represent a combination of direct biological production of lumichrome and photodegradation of biologically produced riboflavin. Because riboflavin was not detected in the exometabolomes, lumichrome and riboflavin levels could not be compared. Two recent studies found that exogenous lumichrome had positive growth impacts on the green algae Chlorella sorokiniana and Auxenochlorella protothecoides (33,34), while another found that riboflavin increased chlorophyll production in Chlorella vulgaris (35). Supporting these studies, we found that P. tricornutum, which produced the highest level of lumichrome (Fig. S3), exhibited increased growth in response to exogenous lumichrome addition (Fig. 5), indicating it may be particularly important for this model diatom. To our knowledge, this is the first study showing impacts of lumichrome on a diatom or any saltwater microalga, expanding the range of systems where this compound may have important impacts. Further, given that lumichrome was enriched in four of the five exometabolomes and its role in bacterial quorum sensing, it could also play a role in modulating the phycosphere bacterial community.
Prostaglandin-like compounds were another group which were enriched in the exometabolomes of several of the algal strains (C. reinhardtii, P. tricornutum, and M. salina). Our data suggest that these features may represent related but novel compounds because these spectral matches were to unusual adducts or had a m/z error greater than 10 ppm. Prostaglandins are hormone-like signaling molecules found in a range of animals and plants (37)(38)(39). Originally discovered and studied in animals, prostaglandins have only recently been detected in microalgae, first being reported in the diatom Skeletonema marinoi (38). Although the roles of these compounds in microalgae have not yet been elucidated, these molecules are known to be involved in intracellular communication in other organisms. Prostaglandins have also been found to have diverse roles in marine macroalgae, including as defense molecules against pathogens in red macroalgae and as part of the oxidative response in brown macroalgae (37). A recent study in P. tricornutum detected the production of the isoprostanoids, which are structurally related to prostaglandins, under oxidative stress (40). That study found that the addition of 5 to 50 mM concentrations of several different isoprostanoids to P. tricornutum cultures decreased cell counts and increased lipid accumulation, while not affecting photosynthetic efficiency. In comparison, we found that the addition of 10 mM 17-phenyl trinor prostaglandin A2 increased chlorophyll A production by P. tricornutum by 20% and decreased chlorophyll production of Desmodesmus grown in freshwater but not saltwater (Fig. 5). Interestingly, negative feature 13470, which showed spectral similarity to 17-phenyl trinor prostaglandin A2, was enriched in the exometabolome of P. tricornutum, but not that of Desmodesmus (Fig. 4). Together, these results suggest that prostaglandins and related compounds have important roles affecting microalgal growth.
Another interesting group of detected compounds included the di-and tripeptides, of which at least one was enriched in each of the five algal system exometabolomes profiled, indicating that this group of exometabolites is widespread across diverse microalgae. We identified one tripeptide and several dipeptides enriched in the alga exometabolomes, with specific dipeptides highly associated with either P. tricornutum or C. reinhardtii (Fig. S5). Dipeptides are known to be involved in interkingdom interactions as well as bacterial quorum sensing. For example, dipeptides have been implicated in coral-alga symbiosis (41,42), as well as the interactions between land plants and arbuscular mychorhizal fungi (43). Cyclic dipeptides are important compounds for quorum sensing in Gram-positive bacteria and are also produced by other organisms to interfere with that quorum sensing (44). Linear dipeptides, like those detected in our study, have also been found to affect bacterial quorum sensing systems (45). The specificity of these dipeptides to particular algal taxa suggests that these metabolites may not be the result of general proteolysis but instead that these particular peptides may have important roles for these algae. Because these metabolites are enriched in the exometabolome, they could play a role in interkingdom interactions, as has been seen in these other systems.
The contrasting growth responses to exogenous addition of specific effector metabolites suggest there is delicate and distinct homeostasis with respect to these compounds and their extracellular and intracellular concentrations. Some metabolites are excreted to reduce toxicity but can also have growth promoting effects. Indole-3-acetic acid, a plant hormone also produced by some algae, stimulates growth of Desmodesmus at low concentrations but inhibits growth at concentrations above 200 mM (46). Contrasting growth responses of different algae to the same metabolite could also indicate allelopathy, in which one alga releases metabolites to inhibit competitors. In our study, negative feature 3830, with spectral similarity to the [M-H-H 2 O] 2 adduct of aleuritic acid, was enriched in the exometabolomes of C. reinhardtii, M. salina, and P. tricornutum, but not in the exometabolome of Desmodesmus (Fig. 4). In the growth impact experiment, aleuritic acid inhibited the growth of Desmodesmus but not the other algae (Fig. 5). Interestingly, although features with spectral similarity to the [M-H-H 2 O] 2 and [M1H-3H 2 O] 1 adducts of dodecanedioic acid were enriched in the exometabolome of P. tricornutum (Fig. 4), the addition of 0.01 mM dodecanedioic acid significantly inhibited P. tricornutum growth (Fig. 5). However, this inhibition was concentration dependent, with the lowest concentration tested (0.0001 mM) slightly increasing growth (Fig. S6). This result suggests that overaccumulation of some exometabolites could result in self inhibition.
There are some limitations to this study which are important to consider. First, our analysis has focused on nonpolar C 18 extractable metabolites, which enabled analysis of changes in response to salt concentration and analysis of compounds relevant to signaling. However, excluding polar metabolites provides a limited view of overall exudates. Because the extraction method is biased toward and against certain groups of metabolites and because the presence of other metabolites can affect extraction, extraction efficiencies can vary depending on the metabolite composition of the sample. We did not measure extraction efficiencies in our study, but a previous study of estuarian water estimated a dissolved organic matter extraction efficiency of 23% to 27% for the solid-phase extraction columns used in our study (47), while another study found that extraction efficiencies for a different C 18 extraction column ranged from 35% to 45% between deep sea and estuarian samples (48). These represent much smaller differences (log 2 fold change = 0.2 to 0.4) than were observed between sample groups in our study (median log 2 fold change = 25.6 across features). Extraction efficiencies of individual compounds or compound classes of interest may be more relevant to consider for these comparisons than overall organic matter extraction efficiency. Johnson et al. found that extraction efficiencies varied across a suite of isotopically labeled spiked-in metabolite standards between marine cultures (49). The fold changes in extraction efficiencies between cultures observed in that study were small (median log 2 fold change of 0.5) compared to those observed in our data. Unfortunately, determination of extraction efficiencies for individual metabolites of interest, which were unknown at the outset of our study, was not feasible for this exploratory untargeted study. Differences in extractions are also an important consideration when relating these results to previous algal metabolomics studies. For instance, other studies of algal exometabolites have used different solid-phase extraction protocols (e.g., to target small polar molecules) (11,15) or derivatize samples prior to extraction (15). Known differences in extraction protocols should be considered when comparing our study, which focuses on nonpolar C 18 extractable metabolites, to other studies with different focuses. A second limitation of this study is that, given our focus on diverse nonpolar metabolites, relatively few were identified. Therefore, an important future direction will be the extension of this work to the analysis of polar metabolites which include many well-characterized molecules such as sugars, amino acids, and small organic acids. Further, given the large phylogenetic distances and limited number of taxa profiled, there are some limitations to what can be concluded about phylogenetic correlations and the specificity of exometabolites to phyla. However, because of the wide diversity of the algal systems studied, the similarities provide evidence for exometabolites that are commonly exuded by microalgae.
This work adds to our understanding of microalgal exometabolomes. We found that variation in eukaryotic microalgal exometabolome composition is not primarily phylogenetically driven and is affected by growth conditions. We also identified impacts of specific effector metabolites, many of which have not been identified or characterized in algae, on algal growth. Further work is now needed to understand the roles of these metabolites in more complex algal-bacterial systems and to extend this work to include polar metabolites.

MATERIALS AND METHODS
Algal strains and growth conditions. This study used four microalgal strains, as detailed in Table 1. Two of the algal strains (C. reinhardtii and Desmodesmus) were grown in freshwater medium, and three (P. tricornutum, M. salina, and Desmodesmus) were grown in saltwater medium.
Saltwater growth medium was prepared using salts for enriched saltwater artificial water (ESAW) medium (50,51). Saltwater medium contained 21.194 PO 4 . To make the freshwater and saltwater conditions more comparable, the same F/2 levels of trace metals and vitamins listed above for saltwater medium were also added to the freshwater medium.
Algae were grown in glass 125-ml Erlenmeyer flasks that had previously been baked in a muffle furnace at 550°C for 2 h to remove trace organics that could contaminate metabolomics analyses. Each flask contained 50 ml of the appropriate medium and was capped with a foam stopper and covered with aluminum foil to avoid contamination during the experiment. Flasks were incubated with shaking at 90 rpm in a light incubator, with a 12-h day/night light cycle. Daytime illumination was 3,500 lx, and the incubator temperature was maintained at 22°C. Algal inocula were prepared by first growing algal stock cultures under experimental conditions (flasks, media, incubation) for 1 week to acclimate cultures to those conditions. After 7 days, experimental flasks were inoculated by transferring 2 ml of inoculum culture to flasks with the appropriate medium. Twenty-five biological replicate cultures were prepared for each algal and control condition to allow destructive sampling of five replicates at each sampling time point (0, 4, 8, 12, and 18 days).
Sample collection. Chlorophyll A measurement samples were collected by transferring 1 ml of culture to a clear 5-ml polystyrene tube. Microscopy samples were collected by transferring a 1-ml sample from the flask to a 1.5-ml microcentrifuge tube, adding 100 ml of formaldehyde to fix cells, and storing the sample at 4°C until performing microscopy to determine the axenicity of the culture.
After removal of samples for chlorophyll measurements and microscopy, the remainder of the culture was destructively harvested to collect cell pellets and spent medium. Spent medium and cell pellets were separated by centrifugation. Half of the culture was transferred to a 50-ml Falcon tube and centrifuged at 5,000 Â g for 8 min at 4°C. The supernatant was carefully transferred to a clean Falcon tube, the remainder of the culture was added to the tube containing the pellet, and the centrifugation procedure was repeated. After separation by centrifugation, the supernatant was filtered through a 0.45-mm-poresize syringe filter to remove residual algal cells. Filtered spent medium and cell pellets were frozen on dry ice and stored at 280°C until extraction.
To disrupt cells for metabolite extraction, cell pellets were resuspended in 1-ml sterile water, frozen at 280°C, and lyophilized. Lyophilized pellets were disrupted using a sterilized steel ball and vortexing three times for 5 s. Disrupted pellets were then resuspended in 50-ml sterile water and filtered and frozen in the same manner as spent medium. Thus, the spent medium and cell pellet samples corresponded to the same original sample volume.
Chlorophyll A fluorescence, biomass, and organic carbon measurements. Chlorophyll A fluorescence of samples from flask grow cultures was measured using a Trilogy fluorometer (Turner Designs, San Jose, CA, USA) with the Chlorophyll A In-Vivo Module.
Biomass was measured as the ash-free dry weight of the cell pellets collected on days 4 and 8. Cell pellets were transferred to aluminum dishes and dried at 95°C overnight. The dry weight of the pellet and aluminum dish was measured. Pellets were then ashed at 500°C for 4 h, and the weight of the remaining ash and aluminum dish was measured. Ash-free dry weight was calculated as the difference between the two measured weights.
Samples were prepared for total organic carbon (TOC) measurement by diluting 1.4-ml sample (spent medium or cell pellet extract) with 5.6-ml ultrapure water for a 1:5 dilution. TOC was measured on a Shimadzu TOC analyzer.
Metabolite profiling. Samples for LC-MS/MS metabolomics analysis were extracted by solid-phase extraction using Waters Sep-Pak C 18 3-ml cartridges. Because chlorophyll and biomass content were similar across samples, samples were normalized to the same volume (30 ml). Cartridges were first preconditioned with 3 ml of methanol and then rinsed with 6 ml of ultrapure water. Metabolites were extracted by passing 30 ml of sample through the cartridge. Columns with extracted metabolites were rinsed with 6 ml of ultrapure water. Metabolites were eluted with 1 ml of methanol into a 1.5-ml polypropylene tube and then dried in a vacuum centrifuge with a refrigerated vapor trap.
Dried metabolite extracts were prepared for analysis as previously described (9,53,54). Briefly, samples were resuspended in 150 ml of methanol with 13 C-and 15 N-labeled matrix control internal standards (see Table S2 in the supplemental material), vortexed for 5 to 10 s, and sonicated in a water bath sonicator with ice for 10 min. Resuspension in 150 ml of methanol was performed to obtain more concentrated extracts for analysis, and internal standards were used to evaluate matrix effects, which can impact metabolite detection (55). Samples were centrifuged at 10,000 Â g for 5 min at 4°C to pellet insoluble salts and proteins. Supernatant was transferred to a 0.2-mm-pore-size centrifuge filter device and centrifuged at 5,000 Â g for 5 min at 4°C. Filtrate was transferred to an autosampler vial for LC-MS/MS analysis.
LC-MS/MS was performed as previously described (56), and the LC gradient and MS conditions are provided in Table S2. Briefly, chromatographic separation of metabolites in samples were performed using reverse-phase liquid chromatography on a Zorbax EclipsePlusC18 RRHD 1.8 mm, 2.1 Â 50 mm column (Agilent Technologies, Santa Clara, CA, USA) on an Agilent 1290 series ultra high-performance LC system. Mass spectrometry analyses were performed using a Thermo Fisher Scientific Q Exactive Hybrid Quadrupole-Orbitrap mass spectrometer (Thermo Fisher Scientific, Waltham, MA, USA). LCMS parameters used for analysis are defined in Table S2.
Untargeted analysis of LC-MS/MS data was performed to identify metabolites as previously described (56). Briefly, LC-MS/MS features, corresponding to specific retention time and m/z combinations, were detected using the MZMine software v. 2.53 (57,58). Detected features were analyzed using the Global Natural Products Social Molecular Networking tool (GNPS) to construct feature-based molecular networks (20,21). This analysis creates networks of potentially related features, using spectral similarity as a proxy for structural similarity. Preliminary feature annotations were determined based on the highest scoring MS2 match to the GNPS database. For features identified as statistically significant in subsequent analyses, putative annotations were made by hand curation of the top 100 matches to the GNPS database, based on GNPS cosine score, delta m/z, number of shared peaks, and common adducts.
Impacts of select metabolites on algae. Impacts of selected metabolites on algal chlorophyll A production were tested in 48-well plates. Each metabolite was tested individually for each of the algal strains at concentrations of 0.01, 0.001, and 0.0001 mM. The highest concentration of 0.01 mM was selected based on a previous study of the impacts of phytohormones on C. reinhardtii growth (5). Each well contained 750 ml of media with or without added metabolite and was inoculated with 15 ml of 1-week-old algal inoculum. Each metabolite was tested for each alga in triplicate in randomly positioned wells. Plates were incubated with a 12-h light/dark cycle for 12 days with shaking at 90 rpm. Chlorophyll A fluorescence of plate growth assays was measured in a Cytation5 plate reader (BioTek, Winooski, VT, USA) with excitation and emission wavelengths of 440 nm and 680 nm, respectively.
Statistical analyses. LC/MS-MS features significantly different from background were identified by comparing feature peak heights for test samples to those of medium blanks with two-tailed t tests and using the Bonferroni correction to account for multiple comparisons. Features were considered significant if at least one sample group had an adjusted P value of ,0.05 and at least twofold above the signal in control samples.
Features significantly enriched in the exometabolome were identified by first calculating the ratio of the peak heights for paired exometabolome and pellet metabolome samples. A one-sample t test was used to compare mean of the base two logarithm of these ratios to zero, using the Benjamini-Hochberg adjustment to correct for multiple comparisons. A feature was considered significantly exometabolome enriched for a particular alga if the adjusted P value was less than 0.05 and average ratio of extracellular to cell pellet peak height was at least 2.
Mantel tests were conducted using the "mantel" function in the "vegan" package v. 2.5-7 in R v. 4.0.5. Phylogenetic distances were calculated based on multiple sequence alignment of 18S rRNA genes to the global SILVA alignment for rRNA genes (59). Metabolome distance matrices were determined based on Euclidean distances of normalized metabolite profiles. Separate Mantel tests were performed using either the freshwater or saltwater exometabolomes for Desmodesmus in the metabolome distance matrix.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. DATA SET S1, XLSX file, 10.4 MB.