Particulate Metabolites and Transcripts Reflect Diel Oscillations of Microbial Activity in the Surface Ocean

Our sampling targeted an anticyclonic eddy and followed two drifters to facilitate Lagrangian sampling of surface ocean water with minimal mixing or forcing other than the day-night cycle. Samples were collected for two multiday sampling periods in summer 2015 (period one: July 26, 6:00 to July 30, 6:00; period two: July 31, 18:00 to August 3, 18:00). The eddy was characterized by warm, nutrient-deplete surface waters typical of the persistently oligotrophic NPSG (19, 38) (Table 1). Photosynthetic picoeukaryotes and the cyanobacteria Prochlorococcus and Crocosphaera contributed substantially to phytoplankton biomass (21) (Fig. 1). POC, which includes bulk community biomass, and picophytoplankton-specific biomass oscillated with significant 24-h periodicity (Fig. 1). Cell abundances and total biomass of Prochlorococcus and Crocosphaera populations increased between the first and second sampling periods (Table 1). Wind speed also increased between the first and second sampling periods, resulting in an increase in the mixed layer depth from 21 ± 5 to 36 ± 6 m. Additionally, we observed a decrease in the number of significantly diel metabolite oscillations during the second sampling interval, from 55 to 9 (see Table S1 in the supplemental information). Here, we focus our analysis on data collected during the first sampling period where metabolite oscillations were more pronounced.

A total of 79 targeted metabolites were detected across samples (Table S1; Data set S1). Total particulate metabolite concentration (as molar concentration and as a percentage of POC or PN) increased during the day and decreased at night (Fig. 2). The most abundant compounds were osmolytes, including glycine betaine (GBT), homarine, 2,3-dihydroxypropane-1-sulfonate (DHPS), and dimethylsulfoniopropionate (DMSP), nucleobases (particularly guanine), and amino acids related to nitrogen metabolism, such as glutamic acid and glutamine (Tables S1 and S2; Fig. 2). At dusk, quantified metabolites totaled 1.7% ± 0.2% of POC and 3.1% ± 0.6% of PN (Fig. 2), with free nucleobases and amino acids comprising most of the metabolite derived nitrogen (2.5% ± 1.1% of PN at dusk, Table S2).

Multivariate analyses were used to determine if time of day influenced the community metabolome. NMDS analysis indicated that samples collected at different times had different overall metabolomes (ANOSIM, R = 0.19, P = 0.001). Samples collected near sunrise (6:00) had metabolomes that were more similar to one another than those of samples collected at other times of day, and the metabolomes collected at 6:00 were most dissimilar to those from samples collected near sunset (Fig. S1; Table S3).

To determine whether oscillations in particulate metabolite concentrations were driven by changes in biomass or by changing cell physiology, we calculated metabolite concentrations as moles of particulate metabolite relative to water volume filtered, resulting in values of molar concentration (nmol liter⁻¹), and as molar concentration relative to POC concentration, resulting in values of nmol per μmol POC. Bulk and individual metabolite concentrations oscillated both in molar concentration and relative to POC (Fig. 2; Fig. 3A and B) (Rhythmicity Analysis Incorporating Nonparametric methods [RAIN] false-discovery rate [FDR]-corrected P < 0.05). The molar concentration (nmol liter⁻¹) of 55 metabolites (70%) had significant 24-h oscillations, with 26 reaching a maximum in concentration within 2 h of 18:00 and 20 reaching their peak concentration within 2 h of 14:00 (Fig. 3A and B; Table S1). Relative to POC (nmol μmol POC⁻¹), 37 compounds (47%) showed diel oscillations (Table S1), and the mean time of peak concentration shifted to earlier in the afternoon (Fig. 3A). POC reflects total community biomass and detritus, so to avoid assumptions of metabolite source, we present molar concentrations throughout except when metabolite source can be constrained to a specific phytoplankton type, in which case we present metabolite concentration relative to the cell number or biomass of the source organism.

Metabolites with significant oscillations (Fig. 3C) had daily fold changes ranging from 2 to 12.8, all of which exceeded the 1.2- and 1.8-fold changes of POC and the sum of FCM-derived phytoplankton biomass, respectively (Fig. 4). These compounds function in cell turgor, anabolism, energy storage, and redox balance (5, 11, 13, 39–50). The disaccharides trehalose and sucrose displayed the most robust oscillations (P value < 1 × 10⁻¹³) (Fig. 4 and 5). Trehalose and sucrose are known osmolytes, compounds that accumulate as compatible solutes (47), and nearly all other identified osmolytes (9/10) showed diel oscillations (Fig. 4 and Fig. 6A; Table S1). Glutamic acid is the only known osmolyte that did not have a significant oscillation in molar concentration (Table S1).

Metabolites with significant diel cycles involved in cell growth include the three methionine-cycle compounds that were detected, S-adenosyl methionine (SAM), S-adenosyl homocysteine (SAH), and methionine. 5′-Methylthioadenosine (MTA) is produced from SAM during polyamine synthesis and had a temporal pattern that closely matched that of SAM (Fig. 3C and Fig. 6A), such that SAM/MTA remained relatively constant. The cofactor pantothenate (vitamin B₅) was one of the few compounds that peaked in the morning (Fig. 3C). Cofactors riboflavin and niacin (vitamins B₂ and B₃) oscillated with maxima near dusk. Reduced glutathione oscillated with an afternoon peak (Fig. 3C).

To investigate the relationships between gene expression and metabolite concentration we used the Kyoto Encyclopedia of Genes and Genomes (KEGG) database to connect metabolites with transcripts annotated as encoding proteins that directly produce or degrade each metabolite. KEGG was chosen because it currently contains the greatest number of metabolites compared to that in other databases that link genes, enzymes, and metabolites (51). Using KEGG, all but four of the diel metabolites were linked to at least one annotated prokaryotic or eukaryotic transcript (Fig. S2). Glucosylglycerol, ergosterol, and isethionic acid are in the KEGG database, but no transcripts were annotated in our data set as directly producing or degrading them, while homarine is not included in the KEGG database.

Transcripts provide insight into the number and identity of organisms and pathways that may be responsible for the metabolite’s synthesis and degradation, despite potential database biases and sequencing depth. The orders containing Crocosphaera, Prochlorococcus, Pelagibacter ubique, and other unclassified alphaproteobacteria comprised ∼50% of all prokaryotic transcripts that could be linked to metabolites (Data set S2). Dinoflagellates (Dinophyceae), nondiatom stramenopiles (Stramenopiles), haptophytes (Haptophyceae), nonmetazoa opisthokonts (Opisthokonta), and diatoms (Bacillariophyta) comprised ∼70% of eukaryotic transcripts linked with metabolites (Data set S2).

AMP, SAM, and SAH stand out as the diel metabolites with the greatest number of associated diel transcriptional patterns, with 181, 124, and 113 transcripts, respectively (Fig. 6B; Fig. S2). Most diel transcripts associated with SAM and SAH encoded methyltransferases that use SAM and produce SAH (Fig. 6; Data set S2). In most other cases, there were few diel transcripts associated with a metabolite (e.g., only 6 diel transcripts were associated with trehalose) (Fig. 6B; Fig. S2).

As a first step to investigate the temporal relationship between gene expression and metabolite concentration, we estimated the lag-time between peaks in metabolites and in transcripts that exhibited significant diel periodicity in abundance. This analysis showed a broad distribution in the lag-times between metabolites and transcripts (Fig. S3), with no significant relationship between the peak time of prokaryotic or eukaryotic transcripts and their associated metabolites (Pearson correlation P value > 0.05).

We observed trehalose-related transcripts from eukaryotic phytoplankton and from Crocosphaera (Fig. 6). Using published Ostreococcus cellular trehalose concentrations (6) and picoeukaryote cell counts, we estimated that picoeukaryote contribution to trehalose was 0.2 to 3.0 pmol liter⁻¹, a small fraction of environmental trehalose (274 pmol liter⁻¹ at 1,800 on average; Table S2). The abundance of Crocosphaera (Table 1; Fig. 1) and diel oscillations in the Crocosphaera transcript for trehalose 6-phosphate synthase/phosphatase (Fig. 6; Data set S2) suggest Crocosphaera as the main contributor of trehalose during this field study. To further explore this hypothesis, we grew Crocosphaera watsonii WH8501 under a 12:12 light:dark cycle and measured 0.8 and 0.07 fmol trehalose cell⁻¹ at the end of the light and dark periods, respectively (Fig. 5). Given the Crocosphaera abundance during our sampling and assuming similar intracellular concentration, this accounts for 1.8 to 670 pmol liter⁻¹ particulate trehalose, comparable to total particulate trehalose during our sampling (2.8 to 627 pmol liter⁻¹ across both sampling periods; Fig. 5).

Multiple taxa expressed transcripts related to production and degradation of sucrose, including Prochlorococcus (Fig. 6). To assess the potential contribution of Prochlorococcus to environmental sucrose concentrations, we measured the cellular sucrose quota in a culture of Prochlorococcus MIT1314 harvested midday during exponential growth. Using the cellular quota of sucrose in these cultures (range in biological triplicates: 1.4 to 2.1 amol cell⁻¹) and the abundance of Prochlorococcus during the environmental sampling, it is possible that all the observed sucrose could have been in Prochlorococcus during this study (Fig. 5).

The diel oscillations in POC and FCM-resolvable phytoplankton biomass reflect the alternation of carbon fixation, anabolism, and growth during daylight hours and respiration, catabolism, and mortality during the night (Fig. 1) (17, 19, 27). The community metabolome reflected these patterns, with an overall increase in concentration throughout the day while cells are growing (Fig. 2). Macromolecular measurements show that phytoplankton use sugars and lipids overnight (22, 26, 28, 32, 46) and synthesize protein in the early morning in order to optimize photosynthesis during the day (31). The observations of diel oscillations in primary metabolites here highlight the timing and extent of this preparation in a natural environmental community.

Multivariate similarity of samples collected at 6:00 a.m. indicates a consistent morning phenotype with low concentrations of metabolites (see Fig. S1A and B in the supplemental material), reflecting nighttime use of energy stores and recovery from daytime oxidative stress (46). Nearly half of the diel metabolites (26/55) had peak molar concentrations near dusk (Fig. 3), corresponding with a peak in carbon biomass. However, for most (46/55) diel metabolites, the daily enrichment of a metabolite exceeded that of POC or total FCM-resolvable phytoplankton biomass, which had daily fold changes of 1.2 and 1.8, respectively (Fig. 4). This suggests that these metabolites likely had oscillations in intracellular concentration, as previously observed for many primary metabolites in non-marine cyanobacteria (45).

Primary metabolites are particularly powerful indicators of biochemical activity on the community scale. SAM, SAH, and AMP are compounds involved in biosynthesis and growth that had diel oscillations with daytime increases (Fig. 3 and 6). Transcripts associated with these molecules displayed diel oscillations in abundance, with peaks at different times of day, across myriad pathways and microbial taxa (Fig. 6; Fig. S2). Despite this diversity in use, the sum of community activity was reflected in diel oscillations of metabolite concentrations, which were synchronized with daytime biomass accumulation. Further evidence of this daytime community-scale anabolism is the diel oscillation of pantothenate (vitamin B₅), a component of coenzyme A as well as acyl carrier protein. Pantothenate peaked in the morning (Fig. 3), suggesting that the community was poised to assemble these cofactors for daytime biosynthesis.

SAM is a ubiquitous methyl donor used by all living cells. During methylation, SAM is converted to SAH, which is then regenerated back to SAM via methionine. In addition to its role in methylation, SAM is essential for polyamine synthesis and is the most common riboswitch effector in prokaryotes (53). SAM riboswitches have been observed in native Station ALOHA bacterioplankton populations (54). SAH had an afternoon peak time, such that the SAM/SAH ratio was at a minimum during the day (Fig. 6). This ratio reflects methylation potential (55), suggesting that the demand for methylation outstripped the supply of SAM in the light despite rising intracellular concentrations of SAM during the day, concentrations which likely aid in maintaining high reaction rates. Over the dark period, SAM/SAH ratios recovered, suggesting that catabolic processes dominated and the need for SAM was diminished. Many cells require cobalamin (vitamin B₁₂) to catalyze the reactions that regenerate methionine, and SAH is elevated relative to SAM during cobalamin stress as cells struggle to complete the cycle (11). Thus, it is possible that the lower SAM/SAH ratio additionally reflects a daytime increase in cobalamin demand among auxotrophic community members.

All living cells produce reactive oxygen species, and redox homeostasis is a critical part of maintaining cellular function. Reactive oxygen species produced during photosynthesis accumulate over the day and present a continuing challenge for cells at night (40). Strategies for managing oxidative stress range from the acute reactions that detoxify reactive oxygen species to mitigation and avoidance strategies, such as accumulating glycogen to assimilate excess reducing power in high-light conditions and avoid dangerous levels of membrane redox potential (46). Reduced glutathione is the active form of a key component of the reactive oxygen species detoxification system and peaked during the afternoon (Fig. 3B), as has been observed in cultures and field studies (41). The daytime peak possibly reflects production to compensate for increased oxidative stress during the day and a subsequent decrease in production and oxidation of the residual pool overnight.

Riboflavin and niacin (vitamins B₂ and B₃) are precursors to cofactors flavin mononucleotide/flavin adenine dinucleotide (FMN/FAD) and NAD/NADP, respectively. These cofactors are involved in electron transport chains for photosynthesis and respiration and are therefore key components of redox processes within cells. The daytime accumulation of riboflavin and niacin (Fig. 3; Table S1) underscore that the community-wide processes of growth metabolism and redox maintenance occur in step with the diel cycle.

Metabolites with osmolyte properties are among the most abundant compounds within marine microbial cells (5, 13, 44, 47, 48, 56) and exhibited diel oscillations in concentration (Fig. 5 and 6). One exception to this observation was glutamic acid, which plays other critical roles as a component of protein and in regulating nitrogen assimilation in addition to its osmotic properties (47). In the absence of fluctuations in salinity or temperature, oscillations in osmolyte concentrations occurred in excess of or out of sync with biomass oscillations and point to alternative roles for this compound group such as previously hypothesized roles in short-term energy and nutrient storage (47) (Fig. 4; Table S1). Intracellular accumulation of metabolites occurred predominantly during the day when electron flow through the photosystems and the Calvin cycle exceeds that required to maintain maximum division rates. The resulting need to dissipate reductant is typically channeled into the production of carbohydrates like glycogen (22, 39, 46), into exopolymeric substances (57, 58), or into storage lipids (26, 32). Cyanobacteria, for example, manage excess energy during the day by both storing glycogen and producing small molecules that can either be stored or excreted (39, 40, 46, 49, 50). These energy stores are used to fuel cellular respiration and other activities at night, such as protein synthesis and preparing cells for photosynthesis (26, 31, 32, 39, 46). Unlike starch and storage lipids, osmolytes do not necessarily need to go through hydrolysis, β-oxidation, or glycolysis prior to entering the tricarboxylic acid (TCA) cycle and could be used as readily available substrates for energy production and as biosynthetic intermediates while macromolecular pools are being mobilized by the cell (49).

Trehalose was the most prominent diurnally oscillating compound with diel oscillations in concentration in seawater and relative to POC (Fig. 4 and 5; Table S1). Trehalose is an osmolyte produced by the unicellular diazotroph Crocosphaera (42, 44), some heterotrophic bacteria, and some phytoplanktonic picoeukaryotes, including Ostreococcus (6). Transcriptomic evidence motivated us to measure trehalose in cultures of Crocosphaera, which was more concentrated at the end of the day than at the beginning of the day, similar to what we saw in the environment. Assuming trehalose in the environment is produced primarily by Crocosphaera, our results strongly suggest that intracellular trehalose concentrations have diel oscillations in the field (Fig. 5).

Crocosphaera temporally separate photosynthesis and nitrogen fixation to protect nitrogenase from oxygen (59–61), as reflected in their gene expression for photosynthesis during the night and early morning and nitrogen fixation at dusk (21). To draw down cellular oxygen and fuel nitrogen fixation, Crocosphaera need a nighttime energy source (62, 63). Crocosphaera has at least one gene encoding a protein homologous to glycoside hydrolases, family 15 (64), which contains enzymes that hydrolyze a variety of glycosidic bonds, including trehalose. Thus, it is possible that Crocosphaera uses trehalose as a fuel for generating the electrons and ATP required for nitrogen fixation. Using the stoichiometry of these reactions (62, 65), we estimated that trehalose catabolism could have fueled 9 to 28% of the nighttime nitrogen fixation during this expedition (calculation in supplemental material calculation on trehalose fueling nitrogen fixation) (21). As much as 60% of total dark respiration by Crocosphaera is used to draw down cellular oxygen rather than to directly fuel nitrogen fixation (62), and, if we adjust our calculation accordingly, trehalose can produce 3.6 to 11% of the required respiratory substrates needed for Crocosphaera to effectively fix nitrogen at the rates measured (21). In addition to providing energy, trehalose could be providing the carbon needed to generate TCA cycle intermediates when biosynthesis of other molecules, such as amino acids, is consuming those intermediates.

The flux of carbon through trehalose may be an indicator of the accumulation and degradation of a larger glycogen pool that accumulates during the day and is used at night (66). Shi et al. (67) suggest that Crocosphaera cells are depleted of storage compounds at night, since prolonged dark does not result in increased nitrogen fixation. If this hypothesis is correct, the total amount of nitrogen fixation possible is limited by the amount of energy stored in substrates such as trehalose and glycogen during daytime, and the ability to accumulate and use these compounds could have impacts on the nitrogen budget of the microbial community.

Another disaccharide osmolyte, sucrose, displayed an oscillation with a maximum daily concentration at 22:00 when calculated relative to seawater and to POC (Table S1). Sucrose is the major compatible solute in high-light Prochlorococcus (56), and the observed environmental variation may reflect the in situ accumulation and use of glycogen by Prochlorococcus. Though other organisms also expressed sucrose-related genes (Fig. 6), Prochlorococcus was the numerically dominant sucrose-producing organism detected in these populations (Table 1). Prochlorococcus has strong diel gene expression for anabolic and catabolic processes (23) and is known to accumulate polysaccharides during the day, particularly under nitrogen limitation (68). If we assume that cellular quotas of sucrose in Prochlorococcus grown in culture are similar to those in the environment, Prochlorococcus alone could explain the sucrose concentrations seen in the environment (Fig. 5). Sucrose had a diel oscillation when calculated relative to Prochlorococcus cell counts and biomass (Fig. 5). These potential intracellular oscillations lead us to hypothesize that Prochlorococcus uses sucrose for energy storage and not only as a compatible solute, as has been observed in nonmarine cyanobacteria (45, 49).

Homarine and DMSP are known eukaryotic osmolytes (5, 47, 48, 69). The amplitude and timing of the diel oscillations in these two compounds differ from those observed in phytoplankton picoeukaryote biomass (Fig. 4), again suggesting that these compatible solutes play multiple roles within the microbes that use them as osmolytes. This diversity of functions is well established for DMSP, which influences grazing behaviors and can function as an antioxidant (5, 43). DMSP is also a source of carbon and reduced sulfur in the microbial community, with uptake and assimilation both tied to light availability (70, 71). In our analysis, the only annotated transcript related to DMSP encodes a SAR11 DMSP demethylase required for DMSP degradation (72) (Fig. 6). A dearth of data on the roles of homarine in marine microbes and a lack of genetic information about homarine synthesis and degradation limit our ability to infer the sources and sinks for this abundant compound. The high concentration and diel dynamics of homarine call for further investigation.

Both isethionic acid and DHPS are associated with fast-growing eukaryotes that need to mobilize cellular machinery to transport materials into the mitochondria for respiration (13, 73), and recent work has suggested that DHPS has potential osmotic capabilities (13). These two metabolites had large diel oscillations, implicating them as temporary stores of energy or intermediates that can be mobilized quickly. Our data implicate SAR11 and Rhodobacteraceae as likely DHPS degraders at Station ALOHA (Fig. 6), although genes for the production of DHPS are not in the KEGG database and thus were not identified by our analyses. If production and degradation of these compounds are separated along phylogenetic lines (36), then these compounds are likely excreted into the dissolved phase by eukaryotes and subsequently available for use by bacteria, as suggested in Durham et al. (13). This may explain the midday maximal expression of an hpsN-like Rhodobacteraceae DHPS degradation gene (Fig. 6).

Glycine betaine is an osmolyte found within many marine microorganisms and can play multiple cellular roles, including modulating buoyancy, acting as a methyl donor, and providing a nitrogen source (5, 47, 74). Glycine betaine had a significant diel oscillation with an evening peak and similar fold change to that of the total FCM-quantifiable biomass (Fig. 4). Chlorophyta, Dinophyceae, and diatoms expressed glycine betaine synthesis genes with diel periodicity while many more groups expressed related genes that lacked diel periodicity (Fig. 6). It is possible that osmotic balance is maintained in certain phytoplankton by the relatively stable intracellular concentration of glycine betaine while other osmolytes are more dynamic pools with higher daily fold change.

Although our data suggest that over diel cycles many metabolites are respired directly by the phytoplankton that produce them, other metabolites quantified here are known to fuel heterotrophic bacterial growth in marine ecosystems (71, 75–78). DMSP, for example, can support up to 9.5% of the bacterial carbon demand at Station ALOHA (70). Additionally, glycine betaine can support heterotrophic bacterial growth, and both natural marine populations and isolated bacteria are known to have high-affinity glycine betaine transporters (79–81). The oscillations of particulate metabolite concentrations observed in this study call for further investigation into the hypothesis that these compounds are important substrates for community interactions and resources for the microbial loop. For compounds that exhibited diel oscillations, the difference between the daily maximum and minimum values provides a daily net production and degradation rate. We estimated a total net turnover rate of over 32 nmol C liter⁻¹ d⁻¹ from our targeted metabolites, with several metabolites exhibiting individual turnover rates of over 1 nmol C liter⁻¹ d⁻¹, including arachidonic acid, trehalose, homarine, sucrose, glycine betaine, glucosylglycerol, and DHPS (Table S1). These are conservative estimates since the instantaneous flux may be much higher than the daily net change and we did not measure excretion of metabolites into the dissolved pool. For example, dissolved DMSP has a turnover time of 4.5 h at Station ALOHA (70) and has been shown to be produced at night and during the day (82). Both of these observations about DMSP would substantially increase the baseline estimate of DMSP production made here, which does not account for rapid turnover and only includes a daytime increase in intracellular concentration.

While the fate of the metabolites measured here remains unclear, conservative estimates of carbon and nitrogen flux through these small pools was large, comprising around 2% of the ¹⁴C based estimates of primary productivity during this study (26). These compounds are potentially used for cellular requirements by the organisms synthesizing them, as discussed above, or released into the labile dissolved pool. When they enter the dissolved pool through excretion or cell lysis, these compounds are important components of the labile dissolved organic matter pool (77) and play a role in organism interactions (83, 84).

The light-dark cycle plays a dominant role in structuring marine microbial activity. Previous work has shown diel oscillations of community processes, such as daily accumulation and depletion of POC (27), and diel oscillations of transcriptional activity, which have provided new information on temporal dynamics and raise hypotheses about the activity of individual taxa (23, 24). Measurements of in situ metabolites in native planktonic microbial populations reported here support the hypotheses that diverse microbial taxa in the NPSG are synchronized to daily oscillations of light energy and photosynthesis, with metabolites accumulated during the day and depleted at night. The diel oscillations of ubiquitously used primary metabolites are a direct manifestation of photoautotrophic organisms dominating the community and driving anabolic processes during the day and catabolic processes at night. The combination of transcript abundances, metabolite concentrations, and taxa-specific biomass in the field and in culture allows us to postulate that Crocosphaera uses trehalose as a short-term energy source to drive nighttime nitrogen fixation. Trehalose and the other osmolytes we measured are highly abundant in cells and, in addition to playing multiple roles within producers, likely fuel respiration in heterotrophic bacteria. Studies in model organisms suggest mechanisms for why some common metabolites, such as niacin and riboflavin, have diel oscillations, but the hypotheses presented here need to be validated with future studies. Though metabolic flux is often regulated at other points along metabolic pathways rather than at the individual transcript levels, measuring gene expression remains one of the most useful ways we have of probing a community’s metabolic state. However, metabolite concentrations cannot be predicted from transcripts in a single organism in pure culture, let alone in a complex natural community. Pairing quantitative measurements of particulate metabolites with transcriptomes is a key step toward understanding how regularly oscillating gene expression in microbial communities is reflected in the net community processes we observe and further elucidates the currencies of the microbial community.

Samples were collected on the R/V Kilo Moana in the NPSG (near 24.5° N, 156.5° W) every 4 h for two sampling periods in summer 2015 (period one: July 26, 6:00 to July 30, 6:00; period two: July 31, 18:00 to August 3, 18:00). To limit variability unrelated to solar forcing, we conducted Lagrangian sampling following two drifters in an anticyclonic eddy (21). Samples for particulate metabolites and transcripts were collected from 15 m water depth using Niskin bottles attached to a conductivity, temperature, depth array (CTD). Ancillary measurements for nutrients and heterotrophic bacterial abundance (reported in Wilson et al. [21]) were collected and analyzed with standard Hawaii Ocean Time-series protocols (http://hahana.soest.hawaii.edu/index.html).

POC concentrations were derived from particulate beam attenuation at 660 nm measured via a hyperspectral absorbance and attenuation meter (ac-s, Wetlabs, as published in White et al. [27]). Particle attenuation at 660 nm (c_p 660, m⁻¹) was calibrated against discrete POC samples taken near dawn and dusk (n = 30; r of a type II regression = 0.78). Discrete POC and PN samples were collected by filtration of the ship’s underway flow through seawater onto combusted GF/F filters. Analysis is further described in the section on particulate carbon and particulate nitrogen in Text S1.

Continuous underway flow cytometry (SeaFlow) (85) was used to count Prochlorococcus, Synechococcus, picoeukaryotes (eukaryotic phytoplankton 2 to 4 μm in size), and Crocosphaera. These data were supplemented with discrete flow cytometry sample analysis as in Wilson et al. (21). Cell diameters of individual cells were estimated from light scatter by the application of Mie theory (86) to a simplified optical model and converted to carbon quotas assuming spherical particles, as described in Ribalet et al. (87). Carbon biomass was estimated by multiplying cell abundance by carbon quotas.

Metabolite samples were collected in triplicate at each time point by filtering 3.5 liters of seawater onto a 47 mm by 0.2 μm polytetrafluoroethylene (PTFE) filter (Omnipore) using a peristaltic pump, polycarbonate filter holder, and Masterflex PharMed BPT tubing (Cole-Parmer). Filters were frozen in liquid nitrogen immediately after filtration and stored at −80°C. Metabolite extractions employed a modified Bligh-Dyer method (4, 13, 88), resulting in aqueous and organic soluble metabolites with heavy stable isotope-labeled extraction and injection internal standards added to both fractions (Text S1, “metabolite sample extraction”). Unused filters extracted alongside the samples served as methodological extraction blanks.

Metabolomics data were collected by paired liquid chromatography mass spectrometry (LC-MS) with a Waters Acquity I-Class ultraperformance liquid chromatography (UPLC) system and a Waters Xevo TQ-S triple quadrupole with electrospray ionization in selected reaction monitoring mode with polarity switching, targeting over 200 compounds (4). Two separate analysis were performed for each sample, one using hydrophilic liquid interaction chromatography and the other using reversed phase chromatography. The software Skyline was used to integrate LC-MS peaks (89), and resulting peak areas were normalized to the peak area of internal standards using best-matched internal standard normalization to reduce variability introduced through the extraction and analysis process (4). Detection limits for the metabolites using these methods are published in Boysen et al. (4). A subset of these data are presented in Durham et al. and Muratore et al. (13, 29).

Metabolites with isotopologue internal standards were quantified in all samples. Trehalose, sucrose, and 2,3-dihydroxypropane-1-sulfonate (DHPS) were quantified with standard additions. For all other metabolites (Table S1), concentration (pmol liter⁻¹) was calculated from injections of known concentrations of authentic standards in both water and a pooled sample from filtered euphotic zone seawater from the same cruise to correct for ion suppression. Dimethylsulfoniopropionate (DMSP) loss is known to occur during methanol-based extractions, so concentrations are considered a minimum estimate (90). Details are in Text S1 in the section on the quantification of select metabolites. The amount of each metabolite in each sample is presented in units of metabolite concentration where the amount is normalized to liters of water filtered (nmol metabolite/liter) and percent POC or percent PN where the amount is normalized to POC or PN (percent POC and percent PN calculated as mol C or N in metabolite/mol C or N in POC or PN ×100).

Whole community transcript data are referred to here as prokaryotic transcript data, as they were enriched in bacterial and archaeal RNA. These metatranscriptome samples were collected on 0.2-μm filters simultaneously with the metabolomic data reported here, and the sample collection and processing have been described in Wilson et al. (21) and Aylward et al. (20). Briefly, the metatranscriptome sequence reads were quality trimmed, end-joined, mapped, and quantified with molecular standards. Metatranscriptome sequence reads were aligned to the ALOHA gene catalog (91) using LAST version 959 (92). Sequence reads were quantified using transcript count normalization, leveraging molecular standards as described in Gifford et al. (93). Sequence reads were summed if assigned to the same taxonomic order and Kyoto Encyclopedia of Genes and Genomes (KEGG) orthologue (94).

Poly A+ selected transcript data (referred to here as eukaryotic transcript data) are from the metatranscriptomes presented in Durham et al. (13). These samples were collected on 0.2-μm filters concurrently with the metabolomic samples and include only the first sampling period. Quality-controlled short reads were assembled using Trinity de novo transcriptome assembler version 2.3.2 (95) on the Pittsburgh Supercomputing Center’s Bridges Large Memory system. Parameters include using in-silico normalization, a minimum k-mer coverage of 2, and a minimum contig length of 300. The raw assemblies were quality controlled with Transrate version 1.0.3 (96). To eliminate redundancy and duplication, the assemblies were merged and clustered at the 99% amino acid identity threshold level with linclust in the MMseqs2 package (97). Using DIAMOND version 0.9.18 (98), translated eukaryotic contigs were aligned to a reference sequence database of marine organisms that includes peptide sequences from hundreds of marine eukaryotic transcriptomes (MarineRefII reference database, http://roseobase.org/data/, with additions discussed in Text S1 in “eukaryotic metatranscriptome reference database”). Taxonomy was assigned with DIAMOND by using the top 10% of hits with E value scores below 10⁻⁵ to estimate the lowest common ancestor of each contig. We assigned putative function using hmmsearch (from HMMER 3.1b2 [99], minimum bitscore 30) to find the best-scoring KEGG gene family from KOfam and linked the specific KO term associated with the KOfam to the contig (version 2019-03-20) (100). Contig abundances were quantified by pseudoalignment of the paired reads to the assemblies with kallisto (101) and normalized to the total number of assigned reads for a taxonomic group. Sequence reads assigned to the same taxonomic group and KEGG ortholog were summed.

Metabolites and transcripts were associated with one another using the KEGG database as a scaffold to match metabolites with transcripts coding for enzymes that directly use or produce those metabolites. The R package KEGGREST (102) was used to access the KEGG database followed by manual curation of these matches.

Diel periodicity was evaluated for all signals using Rhythmicity Analysis Incorporating Nonparametric methods (RAIN) (20, 21, 103). Metabolites and transcripts were considered significantly periodic if they had a false-discovery rate (FDR) (104) corrected P value of <0.05. For each significantly oscillating signal, the time of peak abundance was estimated by fitting a periodic function (Text S1, “detecting periodicity and estimating time of peak concentration”), recognizing that the precision of these peak times is limited by sampling resolution. Diel periodicity in metabolites was identified for the two different sampling periods independently and jointly.

Cultures of phytoplankton were grown in combusted borosilicate tubes in diurnal incubators with a 12:12 light:dark cycle. Crocosphaera watsonii strain WH8501 was grown at 27°C with 50 μmol photons m⁻² s⁻¹ in YBC-II artificial seawater medium (105) supplemented with 0.9 mM nitrate; exponentially growing cells were collected just before the lights turned on and just after the lights turned off. Cells were enumerated via a Beckman Z2 Coulter counter. Prochlorococcus MIT1314 (HLII clade [106]) were grown at 20°C with 20 μmol photons m⁻² s⁻¹ in Pro99 media (107) prepared with Turk Island salt solution and supplemented with 6 mM sterile sodium bicarbonate and 1 mM N-Tris(hydroxymethyl)methyl-3-aminopropanesulfonic acid (108). Prochlorococcus cells were collected 6 h into the light period during exponential phase and enumerated using the flow cytometer BD Influx cell sorter. Axenicity of Prochlorococcus cultures was verified regularly with SYBR-staining and flow cytometry (FCM) and plating on bacterial 1/2 yeast tryptone sea salts (YTSS) agar. Samples for metabolomics were collected by gentle filtration onto 0.2-μm hydrophilic polyvinylidene fluoride (Durapore) filters using combusted borosilicate filter towers.

Information for the KM1513/HOE Legacy II cruise can be found online at http://hahana.soest.hawaii.edu/hoelegacy/hoelegacy.html. Raw sequence data for the diel eukaryotic metatranscriptomes are available in the NCBI Sequence Read Archive under BioProject ID PRJNA492142. Raw sequence data for the prokaryotic metatranscriptomes are available in the NCBI Sequence Read Archive under BioProject ID PRJNA358725. Raw and processed metabolomics data are available in Metabolomics Workbench under Project ID PR000926.