Seasonal and daily patterns in known dissolved metabolites in the northwestern Sargasso Sea

Organic carbon in seawater plays a significant role in the global carbon cycle. The concentration and composition of dissolved organic carbon reflect the activity of the biological community and chemical reactions that occur in seawater. From 2016 to 2019, we repeatedly sampled the oligotrophic northwest Sargasso Sea in the vicinity of the Bermuda Atlantic Time‐series Study site (BATS) to quantitatively follow known compounds within the pool of dissolved organic matter in the upper 1000 m of the water column. Most metabolites showed surface enrichment, and 83% of the metabolites had significantly lower concentrations with increasing depth. Dissolved metabolite concentrations most notably revealed temporal variability. Fourteen metabolites displayed seasonality that was repeated in each of the 4 yr sampled. Concentrations of vitamins, including pantothenic acid (vitamin B5) and riboflavin (vitamin B2), increased annually during winter periods when mixed layer depths were deepest. During diel sampling, light‐sensitive riboflavin decreased significantly during daylight hours. The temporal variability in metabolites at BATS was less than the spatial variability in metabolites from a previous sample set collected over a broad latitudinal range in the western Atlantic Ocean. The metabolites examined in this study are all components of central carbon metabolism. By examining these metabolites at finer resolution and in a time‐series, we begin to provide insights into the chemical compounds that may be exchanged by microorganisms in marine systems, data which are fundamental to understanding the chemical response of marine systems to future changes in climate.

amount of carbon present in dissolved form is 200 times the amount found in particulate form (Hansell et al. 2009).Over the last 30 yr, scientists have made advances in quantifying where DOC is produced, consumed, and stored in the global ocean (Carlson and Hansell 2015).Microorganisms are the main consumers of DOC, and multiple researchers have examined the connections between the diversity of microorganisms in a marine system and their ability to consume dissolved organic matter (DOM) (Bercovici et al. 2021;Liu et al. 2020a;Stephens et al. 2020).These and other interactions between biology and chemistry play a key role in defining the factors that control organic carbon distributions in marine systems, both now and in the future ocean.
Time-series research in the northwest Sargasso Sea extends back to 1954 at Hydrostation S and began in 1988 at the Bermuda Atlantic Time-series Study site (BATS).Repeated sampling at these sites has revealed long-term (multiyear) changes in properties such as temperature, dissolved inorganic carbon, and dissolved oxygen (Bates and Johnson 2020), and changes in the water masses in the vicinity of BATS (Stevens et al. 2020).These studies have further identified seasonally varying physical and biogeochemical processes (Michaels et al. 1994;Bates et al. 1996;Steinberg et al. 2001) that drive changes in DOC concentrations (Carlson et al. 1994;Hansell and Carlson 2001), nutrient cycling and carbon export (Lomas et al. 2013), carbon isotopes (Gruber et al. 1998), oxygen levels (Fawcett et al. 2018;Billheimer et al. 2021), cellular carbon quotas (Casey et al. 2013), and carbon export by phytoplankton (De Martini et al. 2018;Lomas et al. 2022).Seasonal variability at BATS is also evident in the biological community, including viruses (Parsons et al. 2012), autotrophic microorganisms (DuRand et al. 2001;Lomas et al. 2010), heterotrophic microorganisms (Morris et al. 2005;Treusch et al. 2009), and mesozooplankton (Blanco-Bercial 2020).Logically, the temporal variability observed in these chemical and biological parameters should also be expressed in the distribution, quantity, and composition of organic compounds found in the northwest Sargasso Sea.However, quantitative measurements of the individual compounds that comprise DOM and how they change over time at BATS are lacking.
Investigations into individual organic compounds provide insight into biogeochemical processes in marine systems.Yet, the analytical methods needed to track individual organic compounds in marine systems are relatively recent advances.Preseparation techniques such as liquid chromatography, have enabled both targeted investigations of biologically relevant known organic compounds (Durham et al. 2015;Widner et al. 2021) and untargeted investigations that characterize previously unknown dissolved organic compounds in marine systems (Longnecker and Kujawinski 2017;Petras et al. 2021).Our application of these tools to characterize the small, known organic compounds in the marine environment stems partially from our prior research with cultured microorganisms.For example, we have learned that Synechococcus releases select metabolites as waste products (Fiore et al. 2015), while the heterotrophic Ruegeria pomeroyi changes its metabolite release as a function of growth substrate (Johnson et al. 2016).In the ocean, studies have shown that particulate metabolites can vary on diel timescales (Boysen et al. 2021) and sinking particles contain compositionally distinct organic compounds compared to suspended particles (Johnson et al. 2020).However, previous work in the Atlantic Ocean found that the composition of metabolites in suspended particulate organic material is distinct from the dissolved metabolites (Johnson et al. 2023).Therefore, patterns in particulate organic matter do not directly correlate with DOM, and we need to explicitly probe the temporal variability of dissolved metabolites.
As a component of BIOS-SCOPE, a multiyear, transdisciplinary program to study microbial processes, structure, and function in the Sargasso Sea, we used targeted metabolomics to track and quantify a set of compounds central to cellular carbon metabolism.Our first aim was to determine the presence and concentration of dissolved metabolites over multiple time scales (diel to seasonal) at BATS.These compounds were measured in the dissolved organic fraction collected within the upper 1000 m at regular temporal intervals (monthly to seasonal) between 2016 and 2019.Each July, multiday BIOS-SCOPE process cruises enabled higher frequency (6-h) sampling to investigate diel patterns in the summer.We used these data to connect observed patterns in dissolved metabolites with hydrographic trends and biological activity.

Hydrographic data
From July 2016 through July 2019, the upper ocean was sampled at three locations in the Sargasso Sea: BATS (31 40 0 N, 64 10 0 W), Hydrostation S (32 10 0 N, 64 30 0 W), and east of BATS (AE1916, 32 10 0 N, 64 13 0 W).On a subset of BATS cruises, water samples were collected from a minimum of six vertical levels spanning the surface down to 1000 m.During select times (July each year, September 2016, April 2017, and May 2019), the station was occupied for multiple days (process cruises) allowing for higher sampling frequency (Supporting Information Table S1).Because local water mass structure responds to winter mixing, thermal stratification, mesoscale eddies and varying light penetration, we assigned sample depths to a vertical zonation of the water column (Curry et al. unpublished).We have samples from four seasons (mixed, spring, stratified, and fall) with season boundaries set by the relative positions of the chlorophyll maximum (CM) layer and the mixed layer depth (MLD), which was defined by density threshold criteria as the depth where sigma-theta (σ 0 ) exceeds the surface density by 0.125 kg m À3 .We have samples from nine of the vertical zones (VZs) defined by Curry et al., and we grouped our samples into the photic zone (VZ 0 , VZ 1 , and VZ 2 ), the subphotic region (VZ 3 ), and the deep ocean (VZ 4 through VZ 10 ).The details on the bounds used to define the VZs are provided in Supporting Information Table S2.
Water samples were collected using 12-L Ocean Test Equipment bottles mounted on a rosette equipped with a conductivitytemperature-depth, fluorometer, and an oxygen sensor.Water samples were processed to obtain concentrations of particulate organic carbon, dissolved/total organic carbon (TOC), nutrients (nitrate + nitrite, phosphate), bacterial abundance using epifluorescence microscopy, and bacterial production via 3 H-leucine incorporation (process cruises only) using established methods described in Knap et al. (1996), Halewood et al. (2022), andLiu et al. (2022).Both DOC and TOC samples were collected during this project and the two are statistically indistinguishable in oligotrophic waters at the resolution of the high temperature combustion method (Halewood et al. 2022); in this article, we refer to DOC or TOC concentrations as "bulk organic carbon" concentrations.

Metabolite extractions
Water (4-10 L) was collected directly from the sampling bottles into polytetrafluoroethylene (PTFE) or polycarbonate containers.Water was then filtered using a peristaltic pump through a 0.2 μm filter (Omnipore, EMD Millipore) held in a Teflon filter holder.The filtrate was acidified with 12 M HCl to $ pH 2-3.While at sea, the dissolved organic molecules were extracted from the filtrate using solid phase extraction (SPE) with Agilent Bond Elut PPL cartridges (1 g, 6 mL; Dittmar et al. 2008;Longnecker 2015).Briefly, the cartridges were preconditioned with methanol and the filtrate was pulled through the cartridges via PTFE tubing using a vacuum pump.The cartridges were rinsed with $ 24 mL of 0.01 M HCl and then allowed to dry by pulling air over the cartridge for 5 min.The samples were eluted with methanol into a glass test tube.The extracts were stored for up to 1 week at À20 C before they were evaporated to near dryness using a Vacufuge (Eppendorf).Immediately prior to analysis the extracts were reconstituted in 250 μL of 95 : 5 (v/v) water/acetonitrile or 100% water with isotopically labeled compounds that serve as injection standards (Kido Soule et al. 2015).Our prior work has shown that filtration-induced leakage of metabolites from cells will have minimal impact on the measurement of dissolved metabolites in oligotrophic environments (Johnson et al. 2023).

Selection of metabolites
The project focuses on metabolites that play a central role in cellular carbon metabolism.To characterize the number of metabolic reactions known to involve each compound, we used the framework linking metabolites and genomic information established by KEGG (Kanehisa et al. 2023).We used Biopython (Cock et al. 2009) to access KEGG through a REST-style application programming interface.With two exceptions (n-acetyltaurine and chitotriose), all the compounds examined in this project are present in KEGG with corresponding compound numbers (Supporting Information Table S3).We used the compound numbers to query KEGG to identify the chemical reactions associated with each compound.In KEGG, these reactions are assembled into pathways and the query results also provide the number of pathways for each compound.For compounds that we cannot analytically separate (the isomer pairs sarcosine/alanine and threonine/homoserine), we queried both KEGG numbers and separately present the results for each query.The code used to organize the data for these queries and export the results is available at GitHub (https://github.com/KujawinskiLaboratory/metaboliteMath).

Targeted mass spectrometry
Samples were analyzed using ultra high performance liquid chromatography (Accela Open Autosampler and Accela 1250 Pump, Thermo Scientific) coupled to a heated electrospray ionization source (H-ESI) and a triple quadrupole mass spectrometer (TSQ Vantage, Thermo Scientific) operated under selected reaction monitoring mode (Kido Soule et al. 2015).Separation was performed at 40 C on a reversed phase column (waters aquity HSS T3, 2.1 x 100 mm, 1.8 μm) equipped with a Vanguard (waters) precolumn.Mobile phase A was 0.1% formic acid in water and mobile phase B was 0.1% formic acid in acetonitrile.The flow rate was maintained at 0.5 mL min À1 .The gradient began at 1% B for 1 min, increased to 15% from 1 to 3 min, then increased to 50% B from 3 to 6 min, and then increased to 95% B from 6 to 9 min.The mobile phase was maintained at 95% B until 10 min and then decreased to 1% B from 10 to 10.2 min and held at 1% B for the remainder (12 min total run time).Samples were run in both positive and negative ion modes using a 5 μL injection for each.We used two precursor-to-product ion transitions to identify each metabolite, one for quantification and a second to confirm metabolite identity; Kido Soule et al. (2015) provide details on the sourcing for each metabolite.The complete list of metabolites analyzed in this project is provided in Supporting Information Table S3.
Raw data files were converted to mzML format using msConvert (Chambers et al. 2012), and El-MAVEN (Agrawal et al. 2019) was used to identify and integrate peaks for all samples, standards, and pooled samples.Metabolite concentrations were calculated using a standard curve of at least five points.The standard curve was made by adding the known compounds to a pooled sample generated by collecting aliquots from each sample in a batch; each batch of samples contained up to 100 samples.For metabolites with an extraction efficiency > 1%, we corrected the measured concentrations using the extraction efficiency data available from Johnson et al. (2017) and Swarr et al. (unpublished data) for metabolites added since 2017.Concentrations of metabolites with extraction efficiencies of < 1%, see Supporting Information Table S3, should be viewed with caution.Quality control checks for peak shape and instrument response were implemented as described in Kido Soule et al. (2015).

Pre-extraction derivatization to enhance capture of polar dissolved metabolites
In May 2019, we had an opportunity to use a new method developed by Widner et al. (2021) that derivatizes metabolites in filtered seawater prior to SPE to improve their extraction efficiency and quantification.We processed a subset of samples by this method, replicate samples from 8 depths (cast 6) and 12 depths (cast 8).Briefly, sodium hydroxide (8 M) was added to filtered seawater, followed by the addition of 5% v/v benzoyl chloride in acetone.After shaking, samples were neutralized with concentrated phosphoric acid and stored frozen until processing on land through SPE to remove salts.The samples were then analyzed by liquid chromatography coupled to an ultrahigh resolution mass spectrometer (Fusion Lumos tribrid mass spectrometer, Thermo Scientific) to quantify known (derivatized) metabolites.The data files were processed using Skyline (Henderson et al. 2018;Pino et al. 2020).Here, we focus on the results for malic acid, a compound that is not wellcaptured with PPL-based SPE resins.

Statistics and data availability
Statistical analyses and plotting were done using MATLAB version 2019b.We used a non-parametric Kruskal-Wallis test to examine seasonal differences in metabolites and differences in metabolites during daylight (sunrise to sunset) compared to nighttime samples.Gridding of data prior to plotting was done with a MATLAB implementation of data interpolating variational analysis (DIVA for MATLAB, Troupin et al. 2012).Targeted metabolomics data for this project are available at MetaboLights (http://www.ebi.ac.uk/metabolights/) as study accession number MTBLS2356.Environmental data are available from the biological and chemical oceanography data management office (BCO-DMO) (http://lod.bco-dmo.org/id/dataset/861266, and http://lod.bco-dmo.org/id/dataset/3782)and from the BATS FTP data site (http://bats.bios.edu/bats-data).

Hydrography and environmental data
Over each annual cycle at the study site, the upper ocean exhibited large changes in stratification and depth of the surface mixed layer, with enhanced thermal stratification and shallow MLDs (< 20 m) characterizing the warm, summer months, a progressive deepening in the fall, and maximum MLDs (100-300 m) in late winter/early spring.The onset of warming occurred in April in each year of the project with the MLD shoaling to < 20 m by June (Fig. 1).During the July periods, when repeated sampling was possible, the depth of the CM ranged from 80 to 100 m (Supporting Information Fig. S1).
We partitioned the upper 1000 m into vertical layers to facilitate statistical comparisons of metabolite concentrations across seasonal and interannual boundaries.From a total of 372 samples, 75% corresponded to the summer (stratified) period when we occupied the station for multiple days during dedicated process cruises.Approximately 15% of the samples were collected during the winter (mixed) period, and the remaining 10% were associated with the relatively brief spring and fall transition periods (Supporting Information Table S2).Sample collection was evenly distributed throughout the upper 1000 m of the water column, with 25% of samples collected in the surface mixed layer (VZ 0 ), which ranged between 20 and 270 m over the 4-yr period.
Querying the KEGG database revealed that the metabolites examined in this project were found in at least one pathway within KEGG, with the highest number of pathways observed for glutamic acid that appears in 52 different pathways (Supporting Information Table S3).The range of chemical reactions is even broader, with some compounds found in one reaction extending to nicotinamide adenine dinucleotide that appears in over 1000 reactions at KEGG.Thus, aside from the two compounds not listed within the KEGG, the compounds examined in the Sargasso Sea are broadly represented within reactions and biochemical pathways at KEGG.Out of the 95 metabolites available in the analytical methods for this project, 41 compounds passed the quality checks and were present in all 4 yr of the dataset.The majority of dissolved metabolites not measured have a low extraction efficiency in seawater (Supporting Information Table S3) and therefore their absence was not unexpected.Of the metabolites we detected, 34 (83%) exhibited higher concentrations near the surface and decreased in concentrations with increasing depth in the water column (Supporting Information Fig. S2).Thus, correlations between the concentration of each metabolite and other environmental variables generally revealed significant negative correlations with depth and nutrients, and positive correlations with temperature, bacterial production, and the concentration of bulk organic carbon (measured as TOC or DOC) (Supporting Information Fig. S3).By contrast, some metabolites including syringic acid and cyanocobalamin were present in all 4 yr of the project, were highly variable, and had no significant correlations with measured environmental parameters.In the sections that follow, we focus our discussion on metabolites with repeatable seasonal and diurnal patterns.We also compare the temporal variability of metabolite concentrations at BATS to the geographical variability of metabolite concentrations sampled along a latitudinal transect between 38 S and 55 N in the western Atlantic Ocean (Johnson et al. 2023).

Seasonality of metabolites: Mixed vs. stratified season
We had sufficient samples from two seasons, mixed and stratified, to allow for rigorous statistical comparisons of dissolved metabolite concentrations in the photic zone, the subphotic zone, and the deep ocean (Supporting Information Table S2).Differences in metabolite concentrations were most pronounced when comparing the photic zone between the winter (mixed) and summer (stratified) seasons (p-values < 0.05, Kruskal-Wallis test) (Fig. 2).Within the photic zone a total of 14 metabolites showed statistically significant differences in concentration across seasons.For example, the nucleosides adenosine and xanthosine were elevated in the winter, as were S-(5 0 -adenosyl)-L-homocysteine, the vitamins pantothenic acid (B 5 ) and riboflavin (B 2 ), and desthiobiotin, the vitamin precursor to biotin (vitamin B 7 ).In contrast to desthiobiotin, two other vitamin precursors were lower in the winter: 4-methyl-5-thiazoleethanol (HET), the metabolic precursor to thiamine (vitamin B 1 ), and 4-aminobenzoic acid, the metabolic precursor to folic acid (vitamin B 9 ).Yet while thiamine and folic acid were both measured at BATS during this project, neither showed a seasonal difference.A heterogeneous set of compounds, including amino acids and the nucleoside inosine were also higher in the summer.

Seasonality of specific metabolites
Pantothenic acid, taurocholic acid, and tryptophan presented the strongest examples of a repeatable, seasonal pattern over the 4 yr when samples were collected.Each dissolved metabolite showed higher concentrations in the upper 300 m of the water column.

Pantothenic acid
Concentrations ranged from below detection (0.7 pM) to 34.5 pM, with the lowest concentrations at the onset of the stratified period (Fig. 3a).In the winter (mixed) period, pantothenic acid accumulated over the upper 300 m, although its concentrations in the winter of 2017/2018 were notably low compared to the two other winter periods.Pantothenic acid concentrations integrated within the mixed layer showed a recurring annual pattern of decreased stocks during stratified periods when the MLD had shoaled to < 20 m (May, July, and September), whereas stocks increased when the MLD extended deeper than 150 m (January, February, March, and April) (Fig. 3b).

Taurocholic acid
Mean concentrations of taurocholic acid reached a relative maximum of 2.2 AE 0.8 pM in VZ 0 during thermal stratification in July (Supporting Information Fig. S4).Below 100 m, concentrations of taurocholic acid were lower than 1 pM; in the deep ocean a weak seasonal pattern with increased concentrations during the summer (stratified) periods was observed (Fig. 2).

Tryptophan
Regular seasonal variability was also observed for tryptophan with the greatest concentrations observed in July; however, unlike taurocholic acid the temporal variability was most pronounced deeper than the mixed layer depth (MLD) and between 40 and 120 m where the mean concentration was 2.6 AE 3.3 pM (Fig. 4).

Temporal and vertical variability of metabolites during summer stratification
Metabolite samples were collected every 6 h over 3 days during the July process cruises in each year of the study.

Riboflavin
Concentrations of riboflavin demonstrated consistent diel variability in each year of the study, with lowest concentrations in shallowest depths during the mid-day when sampling was coincident with the highest PAR values (Fig. 5).The difference in riboflavin concentrations in the daytime (based on sunrise and sunset times at BATS) vs. nighttime samples was statistically significant (p-value < 0.01).The exception was in July 2019 when riboflavin reached maximum concentrations deeper in the water column (between 80 and 100 m) with values mostly below detection in the surface samples (Supporting Information Fig. S5).Thus, riboflavin presented significant differences on both a diel cycle and a seasonal cycle, as noted in the previous section.

Malic acid
Generally higher concentrations of malic acid ($ 500 pM) were observed in the surface ocean and decreased with depth (Fig. 6).However, the attenuation patterns over depth were not consistent from year to year.For example, in 2016 the measured concentration for malic acid was greatest at 200 m.One caveat to consider for this metabolite is that the extraction efficiency of malic acid with PPL cartridges is low (0.7%, Johnson et al. 2017) and therefore, we did not correct the measured concentrations given the potential error associated with the correction at low extraction efficiencies.The actual concentrations are likely much higher than those shown in Fig. 6a.During the July 2019 cruise, olites where the summer season was higher, while orange shows the metabolites where the winter season was higher.In white are the cases where there was no significant difference between the two seasons.The water column was grouped into three VZ regions for this comparison: photic zone (VZ 0 , VZ 1 , and VZ 2 ), subphotic (VZ 3 ), and deep (VZ 4 through VZ 10 ).Comparisons marked with ** have p-values < 0.001, all other comparisons marked in orange or blue have 0.001 ≤ p-values < 0.05.
we used a pre-extraction chemical derivatization method described in Widner et al. (2021) to validate our observations of malic acid.This method derivatizes the -OH group on malic acid using benzoyl chloride, thereby enhancing its extraction with the PPL resin.Using this method, we quantified higher concentrations of malic acid approaching 2000 pM at the surface and decreasing with depth to a minimum at 600 m.The increased sampling resolution made possible with the benzoyl chloride derivatization method also revealed a deep secondary mesopelagic maximum ranging between 600 and 1700 pM at a depth range of 200-600 m (Fig. 6b).
Additional examples of metabolites that showed generally higher concentrations in the surface and decreased throughout the euphotic zone include 4-hydroxybenzoic acid and 5 0methylthioadenosine (MTA).However, there was significant interannual variability in the shape and magnitude of the depth profile for these metabolites (Supporting Information Fig. S5).For example, 4-hydroxybenzoic acid was higher throughout the upper water column, except in 2018 where it was below detection in all samples collected at the surface.For MTA, both the magnitude and location of the highest concentrations varied by year.In 2017, samples from depths < 10 m had the highest concentrations of MTA, while the samples collected in 2018 and 2019 near the deep CM presented elevated concentrations (Supporting Information Fig. S5).

Comparing temporal vs. spatial variability in metabolites
Data from the current study reveal temporal variability in metabolite concentrations over a 4-yr period in one geographic location.To compare this temporal variability to geographic variability, we calculated the percent relative standard deviation (RSD, standard deviation divided by the mean, multiplied by 100) for 11 metabolites found both in the vicinity of BATS and in a study conducted in the western Atlantic Ocean between 38 S and 55 N latitude in which samples were collected in the austral fall and boreal late summer to early fall of 2013 (Johnson et al. 2023).For the latitudinal study, we restricted the analysis to samples collected from the upper  1000 m of the water column to allow an explicit comparison between the two datasets.The metabolites in this comparison are all observed in picomolar concentrations (Supporting Information Fig. S2), and by calculating the RSD we can compare variability in space and time without considering differences in mean values.
When RSD values from the BATS site are plotted against those from the latitudinal transect, most of the metabolite abundances showed patterns of higher geographic variability compared to temporal variability (Fig. 7; symbols below the 1 : 1 line).Only the amino acid phenylalanine showed higher temporal variability compared to the latitudinal transects, and three metabolites (4-aminobenzoic acid, caffeine, and tryptophan) had the same degree of variability in time and space.As a comparison, the RSD of bulk organic carbon at the BATS site was 15.6% compared to 22.3% in the samples collected along the latitudinal transect in the western Atlantic Ocean.

Discussion
The inventory of marine DOC represents the largest reservoir of exchangeable reduced carbon in the ocean.The biological and physical dynamics that control the sources and sinks of marine DOC over space and time play a critical role in the global carbon cycle.Our understanding of the changes in bulk DOC has evolved over the years (Carlson and Hansell 2015), and about half of marine primary production cycles through the labile pool of DOC (Moran et al. 2022a).Recent advances in analytical methods (Steen et al. 2020) now allow us to track individual molecules traded between microorganisms within the marine carbon cycle (Moran et al. 2022b).This is a corollary to the advances made through GEOTRACES' investigations into inorganic trace elements and isotopes that have been used as markers of a range of oceanographic processes (Anderson 2020).With this project, we focus on compounds that are involved in central carbon metabolism, a small subset of the 19,000 compounds present in KEGG.Of the 41 compounds that met our quality control metrics, there was not a single factor to describe the dynamics of all compounds.Therefore, we group our discussion of the dissolved metabolites into those with annually repeating patterns, those that vary on seasonal and diel cycles, and end with a comparison of temporal and spatial variability of metabolites.

An annually repeatable pattern: Pantothenic acid
The water-soluble vitamin pantothenic acid (vitamin B 5 ) is the clearest example of a compound with an annual pattern that repeats regularly in each of the 4 yr at BATS.Pantothenic acid was present in three-quarters of the samples collected in the upper 300 m, reflecting the origin of its name from the Greek "pantos" meaning from everywhere.Its concentration, however, was not constant throughout the year.We measured the lowest concentrations in the late spring/early summer periods when thermal stratification was greatest and the highest concentrations during periods of vigorous mixing in winter and early spring as the MLD increased.We also observed slightly lower values in the 2018 mixing period which was characterized by shallower maximal MLDs compared to other years.Pantothenic acid was discovered in the 1930s as a growth factor for yeast (Williams et al. 1933), and subsequent work has shown that pantothenic acid is incorporated intracellularly into CoA, a cofactor used in common metabolic pathways including lipid synthesis, processing of fatty acids, and the tricarboxylic acid cycle (Novelli 1953;Leonardi et al. 2007).Microbial production of pantothenic acid can exceed demand for the compound (Jackowski and Rock 1981), which may lead to its release from cells.Yet, details on which microorganisms produce dissolved pantothenic acid and why it is released extracellularly in marine systems are sparse.In laboratory cultures, three strains of the abundant marine phytoplankton Prochlorococcus have been shown to release pantothenic acid to the surrounding media (Kujawinski et al. 2023).At BATS, the distribution of Prochlorococcus is seasonally variable (Olson et al. 1990) with maximum concentrations usually between 60 and 80 m (DuRand et al. 2001), overlapping with the depth range where the maximal concentration of pantothenic acid occurred, suggesting that Prochlorococcus could be a source of pantothenic acid.Furthermore, in heterotrophic marine organisms, genetic information reveals that the abundant SAR86 group lacks a putative metabolic pathway to produce pantothenic acid (Dupont et al. 2012), which indicates this group of bacteria likely require external sources of pantothenic acid.Liu et al. (2022) proposed that some groups of bacterioplankton may shift their metabolisms as the MLD shoals following deep convective mixing, resulting in enhanced scavenging of pantothenic acid in the surface 200 m at BATS.Furthermore, because the short-term scavenging of pantothenic acid observed in April (Liu et al. 2022) is also an annually repeating pattern, this metabolite may be a good model for tracking long-term changes in the sources and sinks of metabolites in the northwest Sargasso Sea.

Seasonal shifts in the balance of dissolved metabolites
The seasonal pattern of bulk DOC dynamics and the role of deep convective mixing in the biological carbon pump have been well established at BATS (Carlson et al., 1994;Hansell and Carlson 2001).The DOC concentrations increase in the euphotic zone as the water column stratifies in late spring or early summer and remains elevated until the mixed layer extends deeper than the euphotic zone in the winter or early spring.During winter convective mixing to depths between 200 and 300 m, DOC is homogeneously redistributed throughout the mixed layer and as a result a portion of the seasonally accumulated DOC is exported from the euphotic zone into the mesopelagic zone (Carlson et al. 1994;Hansell and Carlson 2001).DOC accumulation within the euphotic zone results from a relative imbalance between DOC production processes and heterotrophic bacterial consumption (Carlson et al. 1996).Factors that can affect bacterial consumption of DOC and control the accumulation of DOC include potential inorganic limitation of heterotrophic bacterioplankton production (Cotner et al. 1997;Thingstad et al. 1997), the production of recalcitrant organic compounds (Aluwihare and Repeta 1999) that resist rapid microbial remineralization, and the composition and varying metabolic potential of the resident microbial communities within the euphotic zone (Carlson et al. 2004).At BATS, this imbalance results in the accumulation of dissolved combined neutral sugars (Goldberg et al. 2009) and total combined amino acids (Liu et al. 2022) during the summer stratified period.In the winter, seasonal mixing redistributes DOC that accumulated in the surface to deeper in the water column; organic carbon at deeper depths is then converted to inorganic carbon via microbial respiration as the water column re-stratifies (Hansell and Carlson 2001).
Based on the metabolites examined in this project, compounds with higher concentrations in the mixed layer in the ) in dissolved metabolite concentrations from 2016 to 2019 in the Sargasso Sea (y-axis) compared with data from a latitudinal transect in the western Atlantic Ocean (x-axis, Johnson et al. 2023).The black line is the one-to-one line where the variability in time (Sargasso Sea) matches the variability in space (western Atlantic Ocean transect).The arrows are used to connect metabolite names with data points when crowding would prevent a label from appearing adjacent to its datapoint.The gray diamond is the RSD of bulk concentrations of TOC/DOC in each dataset.
winter were predominantly vitamins, in contrast to amino acids, select nucleic acid and vitamin precursors, and other metabolites that were most pronounced in the summer.Our in situ dissolved metabolite concentration data reveal changes in the standing stock of a metabolite, or the balance between changes in metabolite production and consumption.An increase in the amount of a metabolite measured in the water column can indicate an increase in its production or a decrease in its consumption, or a change in the physical transport of metabolites in and out of the ecosystem.With these caveats in mind, we next consider the metabolites that had relatively higher concentrations in the winter (mixed) period compared to the summer (stratified) period.
Excess S-(5 0 -adenosyl)-L-homocysteine (SAH) in the water column observed during the winter period could be an indication of greater release by cells seeking to avoid the negative effects of higher intracellular levels of SAH.SAH is produced when S-adenosyl-L-methionine (SAM) donates methyl groups to molecules such as DNA, RNA, or proteins (Parveen and Cornell 2011); however, the cellular accumulation of SAH inhibits this methylation.Halophilic cyanobacteria can generate SAH during the synthesis of compatible solutes used to endure increased salt concentrations (Sibley and Yopp 1987), and SAH can be converted to adenosine and homocysteine by prokaryotic cells (Shimizu et al. 1984).Thus, the higher levels of dissolved SAH and adenosine observed during the winter at BATS could have been released from cells.Another possible explanation arises with SAR11, an abundant cosmopolitan marine heterotrophic Alphaproteobacteria, where the transcription of genes producing SAH are enhanced under sulfur-limited growth (Smith et al. 2016).This is of particular relevance given SAR11's inability to use oxidized forms of sulfur such as sulfate and its growth requirement for exogenous reduced sulfur (Tripp et al. 2008).At BATS, the lowest levels of the organic sulfur compounds dimethylsulfide (DMS) and dimethylsulfoniopropionate (DMSP) occur during the winter (Levine et al. 2016).Hence, if SAR11 has an inadequate supply of reduced sulfur in the winter at BATS, it could increase its transcription of genes producing SAH and thereby release increased amounts of SAH into the water column, which would match our observations.
The biochemical reactions described above whereby SAH is produced during the methylation of molecules also produce MTA (Parveen and Cornell 2011).We have measured MTA in the dissolved phase in both the north Atlantic (Johnson et al. 2023), and in laboratory cultures with R. pomeroyi (Johnson et al. 2016) and Prochlorococcus MIT9313 (Kujawinski et al. 2023).Intracellularly, MTA is significantly less abundant in Thalassiosira pseudonana when cobalamin concentrations are low (Heal et al. 2014), which indicates a link between internal MTA levels and conditions experienced by phytoplankton.At the BATS site, dissolved MTA concentrations reach maxima at different depths in the water column, in different years, a pattern that was not correlated to any single environmental factor.Furthermore, dissolved MTA levels did not show significant seasonal differences indicating that further investigation is required to determine the environmental conditions that control the distribution of MTA in the water column.
Two vitamins (pantothenic acid and riboflavin) and one vitamin precursor (desthiobiotin) showed higher concentrations in the mixed layer during the winter mixed period.The production and consumption of vitamins occurs in both phytoplankton and heterotrophic bacteria (Warren et al. 2002;Rodionov et al. 2003;Koch et al. 2012;Sañudo-Wilhelmy et al. 2014), and vitamins are exchanged within microbial communities (Joglar et al. 2021;Wienhausen et al. 2022a;Zoccarato et al. 2022).Vitamin precursors may enable cellular growth in the absence of the vitamin itself, as has been observed in some, but not all, bacterial cells tested for the ability to use desthiobiotin in the absence of biotin (Wienhausen et al. 2022b).In addition, many species of eukaryotic phytoplankton require external sources of vitamins (Croft et al. 2005(Croft et al. , 2006)), and eukaryotic picophytoplankton are prevalent in the winter period at BATS (Treusch et al. 2012) and presumably consume vitamins.However, the winter mixed season is the time when we see relatively higher vitamin concentrations, which contradicts the pattern that would be expected if consumption of vitamins by eukaryotic phytoplankton were the dominant controlling factor.Furthermore, while light and increased temperature can cause degradation of vitamins (Carlucci et al. 1969;Gold et al. 1966), only riboflavin revealed significantly lower concentrations in the day compared to night, suggesting that the accumulation of the other vitamins are not directly controlled by lower light levels in the winter.
The compounds with elevated levels during the summer are select amino acids such as leucine and isoleucine, nucleic acid precursors, and several miscellaneous compounds.Extracellular enzyme activity, especially peptidase activity that cleaves proteins into free amino acids, has been shown to increase with warmer summer temperatures leading to excessive amino acid production relative to consumption (Piontek et al. 2014); this excess production would explain our observations of higher leucine and isoleucine concentrations in the summer.Theoretical calculations show that leucine and isoleucine are among the amino acids that are energetically costly to produce (McClelland and Montoya 2002;Yamaguchi et al. 2017), which may indicate that cells can only produce excess leucine and isoleucine during the summer.Furthermore, inosine and 4-aminobenzoic acid are among the known compounds excreted by copepods (Maas et al. 2020).Zooplankton are known to vary seasonally at BATS (Blanco-Bercial 2020), which could result in a temporally-variable source of these compounds to the mixed layer at BATS.D-ribose 5-phosphate, malic acid, and taurocholic acid are also relatively higher in the summer compared to the winter, as are the vitamin precursors HET and 4-aminobenzoic acid, although the reasons remain unclear.

Diel shifts in riboflavin (vitamin B 2 ) during the summer
The subtropical North Atlantic Ocean experiences higher incoming solar radiation than the temperate and polar latitudes.Riboflavin, which showed significant decreases in concentration during the daylight hours in the surface water masses, was the only measured organic compound responding to daily changes in light.Riboflavin is a water-soluble vitamin that is a precursor for flavin mononucleotide or flavin adenine dinucleotide, cofactors involved in electron transfer, and has been widely studied since it was first isolated in the 1880s (Eggersdorfer et al. 2012).Riboflavin is subject to direct photodegradation due to its highly-conjugated, aromatic structure, and the history of research on the sensitivity of riboflavin to light dates to the 1930s when the first oxidation products of riboflavin were isolated (Warburg and Christian 1932).The loss of riboflavin occurs at wavelengths of light between 350 and 520 nm, with the highest levels of damage in the narrower window between 415 and 455 nm (Ahmad et al. 2006).The interactions between light and riboflavin also vary as a function of pH (Ahmad et al. 2004), ionic strength (Ahmad et al. 2016), and temperature (Sattar et al. 1977).In addition to its role in metabolism, riboflavin can also act as a quorum-sensing molecule (Rajamani et al. 2008).Our dissolved riboflavin concentrations, from below detection to 0.6 pM, are at the low end of the dynamic range of existing data, although coastal seawater from the North Sea had even lower concentrations at < 0.04 pM (Bruns et al. 2022).Along a north-south transect in the western Atlantic Ocean, Johnson et al. (2023) measured values averaging 0.3 pM (range 0-12 pM) with the highest concentrations found between 50 and 150 m, with the exception of the northernmost station where values approached 12 pM at 1 m.The values from the Atlantic Ocean are comparable to dissolved riboflavin concentrations off the coast of California (Sañudo-Wilhelmy et al. 2012).In contrast, in a coastal inlet in the northwest coast of the United States, Heal et al. (2014) found riboflavin had the highest concentrations of the B vitamins, with values ranging from 40 to 120 pM.In laboratory cultures, riboflavin is released by Synechococcus (Fiore et al. 2015) and Thalassiosira (Kujawinski et al. 2017;Longnecker and Kujawinski 2017), but is not released by Prochlorococcus (Kujawinski et al. 2023).Coral reefs (Weber et al. 2022) and sponges (Fiore et al. 2017) are also sources of riboflavin to the marine environment.
The daytime decrease in riboflavin concentrations in the surface water samples could be due to the response of riboflavin to light, consumption by the in situ microbial community, or a combination of both processes.In the Pacific Ocean, intracellular concentrations of riboflavin reached maximal levels at the end of the day (Boysen et al. 2021).Furthermore, while riboflavin is known to react to light, there are certainly other, unknown, compounds in the euphotic zone that exhibit this behavior.To our knowledge, there are no organisms known to be auxotrophic for riboflavin nor has a daily cycle in riboflavin release been examined.The daily decrease in riboflavin in the surface ocean has implications for the microorganisms who rely on production of vitamins by other organisms because excess dissolved riboflavin is only present at night.

Additional compounds of interest
Dissolved malic acid concentrations were higher in the summer in the upper euphotic zone, in VZ 0 .Malic acid is an intermediate in the citric acid cycle and is produced in the first step in the Hatch-Slack, or C 4 carbon fixation, pathway.The C 4 carbon fixation pathway is less common, and on land is dominated by grasses (Sage 2016).In marine ecosystems there is equivocal evidence for a C 4 photosynthetic pathway in diatoms (Mackey et al. 2015), and diatoms with the C 4 photosynthetic pathway may use it to dissipate excess light energy rather than to fix carbon dioxide (Haimovich-Dayan et al. 2013).The Tara Oceans dataset revealed low transcript levels for the enzymes of the C 4 pathway (Pierella Karlusich et al. 2021), supporting the possibility that the higher levels of malic acid in the surface ocean are due to exudation by eukaryotic phytoplankton because they lack an active C 4 carbon fixation pathway.Malic acid is also one of multiple organic compounds measured within marine aerosols collected in the western Arctic Ocean (Kawamura et al. 2012) and western Pacific Ocean (Kawamura and Sakaguchi 1999).Thus there may be atmospheric sources of malic acid to the surface ocean.
The pattern of 4-hydroxybenzoic acid in the water column is an example of a metabolite with irregular vertical and temporal variability during our multiyear project.4-Hydroxybenzoic acid was completely absent from the shallowest samples in 2018, yet was present deeper in the water column at similar concentrations for all years.Furthermore, the return of 4-hydroxybenzoic acid to the surface waters in 2019 indicates a transient decoupling between the production and removal processes in the upper water column.In laboratory cultures, cyanobacteria release 4-hydroxybenzoic acid (Fiore et al. 2015;Kujawinski et al. 2023) or use it as a carbon source (Mou et al. 2007), which complicates defining sources and sinks within the water column.In shallow reef habitats, marine sponges remove 4-hydroxybenzoic acid from the water column, likely due to the actions of the microbial community residing within the sponges (Fiore et al. 2017).The presence of 4-hydroxybenzoic acid in the water column will have unknown effects on the microbial community as it can both stimulate and inhibit prokaryotic activity in a manner that varies for different microbial species (Czerpak et al. 2001;Kamaya et al. 2006).Our previous work (Liu et al. 2020b) demonstrates that we can analytically separate the isomers 3-hydroxybenzoic acid (m-hydroxybenzoic acid) and 2-hydroxybenzoic acid (o-hydroxybenzoic acid), from 4-hydroxybenzoic acid ( p-hydroxybenzoic acid); however, these isomers were not analyzed during this project.This is relevant because the effects of 2-and 3-hydroxybenzoic acid differ from 4-hydroxybenzoic acid on the microbial community, as 2-hydroxybenzoic acid strongly simulates growth, 3-hydroxybenzoic acid inhibits growth, and 4-hydroxybenzoic acid weakly stimulates growth (Czerpak et al. 2001).Thus, the biological community responds differently to each isomer, underscoring the care that must be taken with respect to measuring structural isomers in marine metabolomics.

Scale of variability in metabolites
By comparing data from this multiyear sampling with our previous data from the western Atlantic Ocean (Johnson et al. 2023), we find that the variability in dissolved metabolite abundance across space is generally greater than the variability over time.The greater spatial variability in metabolites is consistent with the expected variability in biological communities over large geographic distances.Furthermore, the variability in the concentrations of individual dissolved metabolites is considerably higher than the variability in bulk organic carbon concentrations, emphasizing that change in bulk DOM can mask changes in the components that make up the pool of DOM.The relatively higher variability in metabolites along a latitudinal transect indicates that a single time-series station cannot serve as a model for the global ocean.However, the repeated sampling at the BATS site enables us to investigate long-term patterns in metabolites as the ocean encounters future changes in climate.Since the 1980s, the BATS site has become warmer and more acidic (Bates and Johnson 2020).While we cannot yet observe long-term changes in our dataset, in future work we can track metabolites over time to view how specific organic compounds change within the context of a changing climate and the subsequent impact on the marine microbial food web.

Conclusions
The dissolved metabolites measured in the northwestern Sargasso Sea during this multiyear study are central carbon metabolites and examining their variability in time provides insight into the processes that underlie chemical variability in a marine ecosystem.These time-series data reveal recurring temporal patterns of known dissolved organic compounds on diel, seasonal, and annual timescales.Understanding how the temporal variability of dissolved metabolites is linked to the sources and sinks of other biological and biogeochemical variables is the next challenge to determine how marine metabolites will respond to future changes in the marine environment.

Fig. 1 .
Fig. 1.Water temperatures in the upper 1000 m of the water column from July 2016 through July 2019, the depth and time range of samples that were collected during this project.The gray dots show the locations where discrete samples were collected.The colored boxes at the top indicate the season in which the samples were collected.The thin contour lines depict chlorophyll fluorescence measured in relative fluorescence units (RFU).The thick black line is the MLD.

Fig. 2 .
Fig. 2. Summary of differences between the winter (mixed) season and the summer (stratified) season in metabolites.Boxes in blue are dissolved metab-

Fig. 3 .
Fig.3.Concentrations of dissolved pantothenic acid in the water column from July 2016 through July 2019, (a) shows pantothenic acid (in pM) in the upper 300 m of the water column.The gray dots represent discrete samples and the black line is the MLD over the sampling period, (b) presents the amount of pantothenic acid integrated to the MLD (in units of μM m À2 ).Values were grouped by month and year before averaging in order to present data from each year over a 12-month annual cycle.The gray line is the averaged MLD.

Fig. 4 .
Fig. 4. Dissolved tryptophan from July 2016 through July 2019 in the upper 300 m of the water column.The black line is the MLD.Colors represent the concentration of tryptophan in pM.

Fig. 5 .
Fig. 5. Concentration of dissolved riboflavin concentrations (pM) in the surface water samples from VZ 0 , sampled over 24 h periods during the July cruises in all years of the project.The PAR data show the average light levels by hour of the day from the 2017 cruise.The x-axis shows local time.

Fig. 6 .
Fig. 6.Two methods for sample preparation were used to measure dissolved malic acid during this project.(a) Shows data from SPE using a PPL resin that was used for all 4 yr of samples collected, and (b) shows malic acid captured with benzoyl chloride derivatization used prior to PPL extraction in the July 2019 cruise.The data in (a) are the discrete samples (filled dots) and interpolated profiles generated using DIVA gridding (open circles).The concentration of malic acid with the SPE method in (a) has not been corrected for the extraction efficiency of malic acid.
Fig.7.The percent RSD (standard deviation divided by the mean, * 100%) in dissolved metabolite concentrations from 2016 to 2019 in the Sargasso Sea (y-axis) compared with data from a latitudinal transect in the western Atlantic Ocean (x-axis,Johnson et al. 2023).The black line is the one-to-one line where the variability in time (Sargasso Sea) matches the variability in space (western Atlantic Ocean transect).The arrows are used to connect metabolite names with data points when crowding would prevent a label from appearing adjacent to its datapoint.The gray diamond is the RSD of bulk concentrations of TOC/DOC in each dataset.