Production and transformation of dissolved neutral sugars and amino acids by bacteria in seawater

. Dissolved organic matter (DOM) in the ocean consists of a heterogeneous mixture of molecules, most of which are of unknown origin. Neutral sugars and amino acids are among the few recognizable biomolecules in DOM, and the molecular composition of these biomolecules is shaped primarily by biological production and degradation processes. This study provides insight into the bioavailability of biomolecules as well as the chemical composition of DOM produced by bacteria. The molecular compositions of combined neutral sugars and amino acids were investigated in DOM produced by bacteria and in DOM remaining after 32 days of bacterial degradation. Results from bioassay incubations with natural seawater (sampled from water masses originating from the surface waters of the Arctic Ocean and the North Atlantic Ocean) and artiﬁcial seawater indicate that the molecular compositions following bacterial degradation are not strongly inﬂuenced by the initial substrate or bacterial community. The molecular composition of neutral sugars released by bacteria was characterized by a high glucose content (47 mol %) and


Introduction
Approximately 700 petagrams (10 15 g) of carbon in the ocean are in the form of dissolved organic matter (DOM) and consist of a broad range of different chemical compounds spanning a continuum of sizes and reactivities (Hansell, 2013;Siegenthaler and Sarimento, 1993). Although all organic matter originates from organisms, only about 6.6 % of surface DOM and 2 % of deep ocean DOM is identified as specific biomolecules such as neutral sugars and amino acids (Benner, 2002). Despite their low concentrations, the rapid turnover of simple biomolecules suggests that they play an important role in the cycling of carbon and nitrogen in the ocean (Rich et al., 1997(Rich et al., , 1996Skoog et al., 1999). However, several studies have also indicated that some biomolecules can resist bacterial degradation over year-long timescales Ogawa et al., 2001). Heterotrophic bacteria are well known sources of semi-labile and refractory DOM (Kaiser and Benner, 2008;McCarthy et al., 1998;Ogawa et al., 2001), and ∼ 23 % of the entire DOC pool is estimated to derive from bacteria via the microbial carbon pump (microbial transformation of bioavailable DOM to refractory DOM) (Benner and Herndl, 2011;Jiao et al., 2010). However, still only little is known about microbial production of specific semi-labile and refractory biomolecules. This is in part due to the low concentration of individual Published by Copernicus Publications on behalf of the European Geosciences Union.
biomolecules in seawater and the consequent analytical challenge involved.
The most common biomolecules in the oceanic DOM pool are carbohydrates (Benner et al., 1992), with neutral sugars constituting up to half of the total carbohydrate pool (Biersmith and Benner, 1998). Seven different neutral sugars are commonly detected (fucose, rhamnose, arabinose, galactose, glucose, mannose and xylose), and the amount of each neutral sugar relative to the total amount of neutral sugars (the molecular composition, mol %) differs between DOM released by different groups of organisms (Lazareva and Romankevich, 2012). The molecular composition can therefore be used as a tracer of different organisms or processes; e.g., freshly released neutral sugars from different algal species exhibit great variation in mol %, but generally have high contributions of glucose and galactose, each sometimes reaching values above 50 mol % (Amon and Benner, 2003;Biersmith and Benner, 1998;Hama and Yanagi, 2001). Surface ocean DOM is also rich in glucose and galactose, with many studies reporting ∼ 20 mol % for each of the two compounds (Goldberg et al., 2009;Kaiser and Benner, 2009;Skoog and Benner, 1997). With depth, glucose and fucose increase in relative abundance, accounting for about 20-40 % and 16-19 % in the deep ocean, respectively McCarthy et al., 1996). The molecular composition of neutral sugars directly released by heterotrophic bacteria during growth in the ocean, however, remains unknown. The molecular composition found in the ocean reflects both the neutral sugars released by different organisms and the neutral sugars persisting in organic matter after long-term bacterial degradation, and the sources and sinks are difficult to distinguish. Knowledge of the composition of neutral sugars produced and transformed by heterotrophic bacteria is key to understanding the origin and fate of these specific DOM components in the ocean. D-enantiomers of amino acids are useful as bacterial biomarkers, since bacteria are the predominant source of Damino acids in seawater (Kaiser and Benner, 2008). Four different pairs of L-and D-amino acids are important in DOM (aspartic acid, glutamic acid, serine and alanine) and the amount of D relative to L (the D/L ratio) has been used as an indicator of amino acid bioavailability (Jørgensen et al., 1999). Freshly released DOM from phytoplankton has a low D/L ratio, but the ratio increases during bacterial degradation of DOM (Amon et al., 2001) due to direct release of D-amino acids by bacteria during growth (Kawasaki and Benner, 2006); viral lysis of cells and the subsequent release of cell wall D-amino acids (Middelboe and Jørgensen, 2006); and a presumably higher bioavailability of L-amino acids (Amon et al., 2001;Hopkins et al., 1994;Pérez et al., 2003). The presence of grazers such as flagellates can also increase the bacterial uptake of D-amino acids, possibly due to a higher release from bacterivory and subsequent microbial uptake (Pérez et al., 2003). The D/L ratio does not follow a certain pattern with depth in the ocean. Both a decrease, an increase and no change has been observed with depth in different studies (Jørgensen et al., 1999;Kaiser and Benner, 2008;McCarthy et al., 1998;Pérez et al., 2003), making it difficult to distinguish between different sources and sinks and the balance between these. Knowledge of the D/L ratio in freshly produced DOM and in DOM remaining after long-term microbial degradation is necessary to understand the origin and fate of amino acids in the ocean.
In the present study, we investigated the concentration, composition and bioavailability of neutral sugars and amino acids in two seawater samples collected between Greenland and Iceland, representing cold seawater originating from the Arctic Ocean and warm seawater originating from the Atlantic Ocean, respectively. During 32-day bioassay incubations of seawater samples and parallel glucose-enriched artificial seawater samples, the bacterial production and decomposition of individual neutral sugars and amino acids were investigated. The aim was to compare the molecular composition of biomolecules in DOM produced by bacteria and in DOM remaining after long-term bacterial degradation, and to use these molecular signatures of bacterial activity to further understand the origin of neutral sugars and amino acids in the ocean.

Sampling site and incubation experiments
Seawater for the 32-day incubations was collected from two different locations in the Denmark Strait: at 10 m depth in the warm (9.8 • C) northwards traveling North Icelandic Irminger Current (henceforth referred to as the Atlantic sample) and at 80 m depth in the cold (−1.6 • C) southwards traveling East Greenland Current (henceforth referred to as the Arctic sample, Fig. 1). The Arctic Ocean receives a large amount of freshwater, supplying 25-36 Tg (10 12 g) in terrigenous dissolved organic carbon (DOC) per year (Raymond et al., 2007), and up to 41 % of this DOC is exported through the Fram Strait (Opsahl et al., 1999). The East Greenland Current transports cold Arctic water from the Fram Strait south along the Greenland shelf, where it mixes with water from the North Atlantic Ocean (Stein, 1988). The seawater collected for bioassay incubations included terrigenous DOM from the Arctic Ocean as well as marine DOM from the Atlantic Ocean.
Six different treatments, all consisting of 90 % 0.2 µm filtered seawater and 10 % GF/C filtered inoculum, were incubated for 32 days in the dark at 18 • C. Three treatments of Arctic and three treatments of Atlantic samples were incubated: a natural seawater sample (NSW), an artificial seawater sample with 60 µM glucose C as the only carbon source (ASW glu ) and a natural seawater sample spiked with 60 µM glucose C (NSW glu ). Hence, the carbon pool consisted of either natural DOM, glucose or both. Only two different bacterial inoculums were used: one from the Arctic and one from the Atlantic sampling site. All samples were amended with inorganic N and P to have 10 µM KNO 3 and 2 µM Na 2 HPO 4 at the beginning of the incubations. N and P were added in excess to reduce the risk of nutrient limitation during the incubations.
Onboard the ship, the majority of the seawater was filtered through a 0.2 µm cartridge filter (Millipore Opticap) to obtain the DOM fraction, and a small bacterial inoculum was filtered through a GF/C filter (Whatman, ∼ 1.2 µm). Both filters were cleaned with ample amounts of seawater before the filtrates were collected. The filtrates were acclimated to 18 • C in the dark for approximately 24 h, and the incubations were initiated by the addition of bacterial inoculum to the 0.2 µm filtered water in acid-washed amber glass bottles. Each of the six treatments consisted of nine amber glass bottles (200 mL) of seawater, and three bottles were sacrificed during each subsampling. The artificial seawater for ASW glu samples was prepared as described by Kester et al. (1967), and adjusted to the salinities of the NSW samples (35.1 and 33.5 for Atlantic and Arctic seawater, respectively) by further addition of MilliQ water. Subsamples for bacterial abundance (all heterotrophic prokaryotes), DOC concentration, dissolved combined neutral sugars and amino acids were taken on days 0, 6 and 32. Bacterial abundance and DOC were measured from all triplicate bottles, while neutral sugars and amino acids were measured from a single bottle. All samples from day 0 were taken approximately 12 h after the addition of inoculum, glucose and nutrients.

Sample analysis
Subsamples for determination of bacterial abundance were fixed with glutaraldehyde (1.3 % final concentration) and stored frozen (−80 • C) until measurement. Bacteria were counted by flow cytometry on a BD FACS Canto II flow cytometer using the nucleic acid stain SYBR Green to stain the fixed cells (Marie et al., 1997). Subsamples for measurement of DOC concentration were 0.2 µm filtered (Acrodisc) and collected in acid-cleaned high-density polyethylene (HDPE) bottles. The samples were acidified with 2 M HCl to a pH of 2 and stored cold (5 • C) until analysis on a Shimadzu TOC-V CPH analyzer. The instrument was calibrated using a standard series made from acetoanilide, and performance was evaluated using deep-sea water reference material made available by the Hansell CRM program. The measured concentration of DOC in the deep-sea reference (41-43 µM) corresponded well to values from the .
Neutral sugars and amino acids were measured from 0.2 µm filtered subsamples collected in acid-cleaned HDPE bottles that were stored frozen (−20 • C) until analysis. The concentration of free and combined hydrolyzable neutral sugars (fucose, rhamnose, arabinose, galactose, glucose, mannose and xylose) was measured on a Dionex 500 ion chromatography system with pulsed amperometric detection (PAD) as described by Skoog and Benner (1997) and Kaiser and Benner (2009). The relative deviation from the mean of replicate hydrolysis (excluding replicate experiments) was 29-40 % for concentrations less than 20 nM, and 10-33 % for concentrations greater than 20 nM of individual neutral sugars. For total hydrolyzable neutral sugars the relative deviation from the mean was 15 %. Free and combined hydrolyzable amino acids were analyzed according to Kaiser and Benner (2005) on an Agilent 1260 UPLC using a fluorescence detector and corrected for hydrolysis induced racemization (Kaiser and Benner 2005). The concentration of the following amino acids was measured: L-and D-asparagine (asparagine and aspartic acid), L-and D-glutamine (glutamine and glutamic acid), L-and D-serine, L-histidine, Lthreonine, glycine, L-arginine, β-alanine, L-and D-alanine, γ -aminobutyric acid, L-tyrosine, L-valine, L-methionine, Lphenylalanine, L-isoleucine, L-leucine and L-lysine. The relative deviation from the mean of duplicate hydrolysis (excluding replicate experiments) was 4-23 % for concentrations less than 10 nM, and 3-8 % for concentrations greater than 10 nM of individual amino acids. For total hydrolyzable amino acids, the relative deviation from the mean was 8 %.

Terminology and calculations
The sum of all neutral sugars and the sum of all amino acids measured are termed total hydrolyzable neutral sugars (THNS) and total hydrolyzable amino acids (THAA), respectively. The contributions of THNS and THAA to total DOM (the yields) were calculated as the ratio of carbon bound in either THNS or THAA to the total concentration of DOC. For THAA, the calculations excluded the non-protein amino acids β-alanine and γ -aminobutyric acid. The mol % of a specific compound was calculated as the molar ratio of the compound concentration to the total concentration of compounds, THNS or THAA.
To investigate the neutral sugars produced by bacteria after utilization of the added glucose, the concentration and composition of neutral sugars on day 6 in glucose-enriched treatments were compared to the corresponding values of the NSW samples. For the NSW glu samples, it was necessary to take into account that a fraction of neutral sugars was bound in natural oceanic DOM. Assuming that bacterial transformation of natural oceanic DOM was similar in the NSW and NSW glu samples, a subtraction of the concentration of individual neutral sugars (NS) in the NSW samples from the corresponding concentrations in the NSW glu samples allowed an estimation of neutral sugars produced by bacteria in the NSW glu samples: For ASW glu samples, only the neutral sugars bound in the 10 % inoculum had to be subtracted, assuming similar concentrations and compositions in the inoculum and the NSW samples:

Bacterial abundance and dissolved organic carbon
The abundance of bacteria was low at the beginning of the experiments, with fewer than 0.4 × 10 6 cells mL −1 in Atlantic samples and fewer than 0.2 × 10 6 cells mL −1 in Arctic samples ( Fig. 2a, b and c). During the first 6 days of the experiment, the abundance increased considerably in all treatments, reaching 1.3 × 10 6 and 0.4 × 10 6 cells mL −1 in Atlantic and Arctic NSW samples and between 3.7 × 10 6 and 5.2 × 10 6 cells mL −1 in glucose-enriched treatments. From day 6 to day 32, the bacterial abundance stayed constant or decreased slightly in all samples. The initial DOC concentrations in Atlantic and Arctic samples were approximately 60, 115 and 70 µM in NSW, NSW glu and ASW glu samples, respectively (Fig. 2d, e and f). DOC concentrations decreased in all treatments during the first 6 days and stayed approximately constant from day 6 to day 32. The net DOC consumption in the Atlantic samples was 4 ± 3.6, 52 ± 3.6 and 58 ± 1.4 µM in NSW, NSW glu and ASW glu samples, Biogeosciences, 11, 5349-5363, 2014 www.biogeosciences.net/11/5349/2014/ respectively. The corresponding values for the Arctic samples were 2 ± 1.5, 52 ± 0 and 56 ± 2.1 µM DOC. The DOC consumption after 32 days corresponded to 7 and 1 % of total DOC in the Atlantic and Arctic NSW samples, respectively.

Neutral sugars
The initial concentrations of THNS ranged from 210 to 339 nM in the NSW samples, and from 7041 to 7850 nM in the glucose-enriched samples (Fig. 3a, b and c). THNS decreased in all treatments over time, particularly during the first 6 days of the incubations. After 32 days, the net consumption of THNS in the Atlantic samples was 130, 7585 and 6992 nM in NSW, NSW glu , and ASW glu , respectively.
The corresponding values for Arctic samples were 84, 6990 and 6943 nM. Initially, THNS constituted 3.5 and 2.4 % of total DOC in the Atlantic and Arctic NSW samples, respectively (Fig. 3d). The yields decreased to 2.5 and 1.7 % on day 6, and continued to decrease at a slower pace during the remaining of the incubations. In the glucose-enriched treatments, the yields were approximately 40 and 63 % in the NSW glu and ASW glu samples, respectively, decreasing to ∼ 3 and 9 % after 6 days ( Fig. 3e and f). The molecular composition (mol %) of neutral sugars changed over time, but varied little between treatments (Table 1). Galactose and glucose were the most abundant neutral sugars in all samples, together comprising between 38 and 74 mol %. The glucose mol % increased from day 0 to day 6 in all samples (disregarding glucose treatments, day 0), followed by a decrease from day 6 to day 32. The galactose mol % followed the opposite pattern: decreasing during the first 6 days, followed by an increase from day 6 to day 32. At the end of the experiments, galactose was the most abundant neutral sugar, comprising between 21 and 47 %. At the same time, glucose only comprised 12 to 32 %. Due to the high concentration of glucose on day 0 in the glucose-enriched treatments, these samples were diluted 100 times prior to analysis. Consequently, no other neutral sugars were determined on day 0. The molecular composition of neutral sugars produced by bacteria in the glucose-enriched treatments during the first 6 days was calculated as described in the "Methods" section. It was characterized by a high glucose mol % (47 ± 12) and a relatively low mol % of the remaining neutral sugars, ranging from 3 ± 2 to 14 ± 9 (Table 1). After 32 days, the molecular composition in all samples was characterized by a high galactose mol % (33 ± 11), a lower glucose mol % (22 ± 8), and an even lower mol % of the remaining neutral sugars, ranging from 7 ± 5 to 11 ± 3.

Amino acids
The initial concentrations of THAA ranged from 170 to 244 nM in all treatments (Fig. 4a, b and c). During the incubations, the concentration of THAA decreased in the Arctic and Atlantic NSW samples. The glucose-enriched sam-ples showed an increased concentration of THAA on day 6, followed by a decrease. The only exception was the Arctic ASW glu sample, which like the NSW samples showed a decreasing THAA concentration throughout the incubation. The yield of THAA in NSW samples decreased from 1.4 to 1.1 % and from 1 to 0.8 % for the Atlantic and Arctic treatments, respectively (Fig. 4d). The yield in NSW glu samples increased from 0.7 and 0.6 % on day 0 to 1.5 and 1.8 % on day 6, and ended at 1.5 and 0.8 % on day 32 in the Atlantic and Arctic samples, respectively (Fig. 4e). The yield in ASW glu samples increased from 1.1 and 1.0 % on day 0 to 7.3 and 2.1 % on day 6, and decreased to 2.3 and 0.7 % on day 32 in the Atlantic and Arctic samples, respectively (Fig. 4f).
The molecular composition (mol %) of amino acids varied little from sample to sample, and also did not show any notable changes over time (Table 2). Glycine was the most abundant amino acid, comprising between 26 and 40 mol %. The four amino acids from which both the L-and the Denantiomers have been measured (asparagine, glutamine, serine and alanine) were the second-most abundant amino acids, comprising between 6 and 18 mol %. The concentration of D-amino acids only changed slightly, while L-amino acids generally decreased in concentration during the time course of the experiments, leading to an increased D/L ratio with time (Table 2).

Concentration and bioavailability
In the NSW samples, the initial DOC concentrations (61-63 µM) were similar to previous measurements from the region (e.g., Benner et al., 2005). After 32 days, 7 and 1 % of the DOC were consumed in the Atlantic and Arctic NSW samples, respectively. These values are comparable to estimates by Amon and Benner (2003) who estimated around 10 % of Atlantic and 2 % of Arctic DOC to be labile. The initial concentrations of THNS in NSW samples (210 and 339 nM) were within the range of concentrations (60-409 nM) found in ultrafiltered surface samples (< 0.1 µm, < 100 m) from the region (Amon and Benner, 2003), and GF/F filtered surface samples from the Sargasso Sea which ranged from ∼ 180 to 450 nM (Goldberg et al., 2009). A considerable fraction of the neutral sugars in the NSW samples were labile, as indicated by the preferential removal of neutral sugars (Fig. 3d). This trend has also been observed in previous studies (Amon and Benner, 2003;Amon et al., 2001). The initial concentrations of THAA in NSW samples (189 and 244 nM) corresponded well with values from surface waters of the Sargasso Sea (∼ 150-200 nM, Lee and Bada 1977;Benner 2008, 2009) and the Arctic Ocean (∼ 150-500 nM, Dittmar et al., 2001;Shen et al., 2012). The preferential removal of amino acids in NSW  Bacterially produced neutral sugars 3 ± 2 8 ± 4 7 ± 8 14 ± 7 47 ± 12 8 ± 3 14 ± 9 Composition after 32 days 8 ± 5 11 ± 3 10 ± 3 33 ± 11 22 ± 8 8 ± 5 7 ± 5 samples indicated that a fraction of amino acids was labile (Fig. 4d). Our study supports the general notion that neutral sugar and amino acid yields can be used as biochemical indicators of DOC bioavailability (Amon et al., 2001), since the highest DOC consumption was observed in the Atlantic NSW sample, which was also associated with the highest yields of THNS and THAA.
In the glucose-enriched samples, 49-56 µM DOC was consumed during the experiment. The DOC samples on day 0 were collected 12 h after the addition of glucose, and the actual DOC consumption was therefore likely up to 12-17 µM higher (estimated from the missing free glucose). Taking this into account, the DOC consumptions agreed well with the amount of added glucose and, possibly, a small contribution from the original DOM pool. The DOC consumption estimates were calculated from the missing free glucose: glucose was added to a final concentration of 10 000 nM before addition of the inoculum. However, the concentrations measured at the beginning of the experiment ranged from 7041 to 7850 nM. Sample hydrolysis possibly also altered some of the added glucose, resulting in lower measured concentrations (Skoog and Benner, 1997). The DOC consumption in the NSW glu samples (52 µM) was lower than in the ASW glu samples (56-58 µM), despite a higher DOC concentration and an equal amount of glucose. Although this discrepancy is small, it may represent a minor contribution of www.biogeosciences.net/11/5349/2014/ Biogeosciences, 11, 5349-5363, 2014 Table 2. Molecular composition (mol %) of amino acids and D / L ratios in the six different treatments during the incubations. The amino acids His, Thr, Tyr, Val, Met, Phe, Ile and Lys all had mol % below 5, and are therefore not included in this table. The amino acids provided in the table accounted for 91-98 mol %. The molecular composition and D / L ratios after 32 days are given as means and standard deviations calculated from all six treatments. The only exception is the D / L ratio of Asx, which did not include the Arctic ASW glu sample. labile DOC associated with the inorganic salts used to prepare the ASW glu samples, or an inhibition of degradation of natural DOM in the presence of a simple labile substrate (Gontikaki et al., 2013). However, further studies are warranted to resolve the importance of the latter process. The significant drop in neutral sugar yields clearly reflects the labile nature of added free glucose. The amino acid yields increased at the beginning of the incubations and decreased from day 6 to day 32 -a trend also observed in previous studies (Kawasaki and Benner, 2006;Ogawa et al., 2001). The initial increase in amino acid yields is due to the significant increase in bacterial abundance and the subsequent release of amino acids.

Bacterial production of neutral sugars during the first 6 days
Bacterial production and the subsequent release of neutral sugars was calculated from the glucose-enriched treatments as described in the "Methods" section. The amount of neutral sugars produced during the first 6 days of the incubations (104-208 nM) was within the range of concentrations of neutral sugars observed in the ocean (20-800 nM; Benner, 2002). The molecular composition was characterized by a high glucose mol % (47 ± 12) and a relatively low mol % of the remaining neutral sugars (3-14 ± 2-9), and is strikingly similar to the composition of bacterial DOM found in a study by Ogawa et al. (2001) after 7 days of incubation (Fig. 5).
Calculations of the neutral sugars produced in the NSW glu samples on day 6 are associated with uncertainty, since bacterial degradation of the natural DOM is unknown. However, if only the ASW glu samples are used when calculating the bacterially derived neutral sugars on day 6, the results are almost identical: a high glucose mol % of 50 and low mol % of the remaining neutral sugars, ranging from 4 to 11. Similar patterns have also been observed in algal-derived DOM (Hama and Yanagi, 2001;Lazareva and Romankevich, 2012), e.g., in DOM from a fresh diatom culture (Biersmith and Benner, 1998) and from sea ice algae (Amon and Benner, 2003). However, DOM released by other algal cultures exhibit different molecular compositions of neutral sugars with galactose or xylose being most abundant (Biersmith and Benner, 1998). In the ocean, glucose is generally the most abundant neutral sugar; however, values above 30 mol % are only observed occasionally (Table 3). Amon and Benner (2003) suggested that degradation processes rather than production processes determine the neutral sugar composition in the ocean, based on the similarities observed in samples of different origin (terrestrial and marine) and from different locations (oceanic regions and water masses). This can possibly explain the difference in the molecular composition of DOM in the ocean and the bacterially produced DOM after 6 days in the present study. Only small differences in neutral sugar composition between the Atlantic and Arctic samples were observed after 6 days of incubation, despite the different bacterial inocula. Moreover, there was a striking similarity in the bacteria-derived neutral sugar composition observed in the present study and in the study by Ogawa et al. (2001). Together, these results suggest that bacterially produced neutral sugars, independent of bacterial community structure, are important in shaping the molecular composition of neutral sugars in seawater.  Kaiser and Benner (2009)

Bacterial degradation of neutral sugars
After 32 days, a clear degradation signature had emerged, with glucose being less important and galactose being more important (Fig. 6). This trend has also been seen in studies of marine sediments (Oakes et al., 2010). The majority of the remaining DOM after 32 days was assumed to be semi-labile or refractory, and the striking similarity between treatments indicates that the molecular composition of semi-labile or refractory neutral sugars (i.e., material persisting longer that 32 days) attains a fairly specific molecular composition, irrespective of initial DOM composition. Furthermore, the similarity between treatments suggests that bacterial degradation processes shape the composition of semi-labile or refractory neutral sugars. After 32 days, about 85 to 87 % of the neutral sugars present in the ASW glu samples are of bacterial origin (i.e., produced by bacteria during glucose assimilation), while only 13 to 15 % originate from the DOM added with the inoculum (excluding glucose), assuming that neutral sug-ars in the inoculum follow the same degradation pattern as in the NSW samples, and that no refractory neutral sugars were added with the inorganic salts while preparing the artificial seawater. Hence, the molecular composition observed in the ASW glu treatments is primarily the result of bacterially produced and bacterially altered molecules containing neutral sugars. In the NSW samples, however, the molecular composition after 32 days primarily reflects the natural background level of refractory and semi-labile neutral sugars present in the seawater when collected. Since this background signature is approaching the molecular composition of bacterially produced neutral sugars in ASW glu samples remaining after 32 days, we hypothesize that semi-labile and refractory neutral sugars primarily originate from bacterial processing of DOM and bacterial remains. Mesocosm studies have shown that 91-94 % of dissolved neutral sugars accumulated during an algal bloom were degraded within 15-20 days (Kragh and Søndergaard, 2009;Meon and Kirchman, 2001), and here we Biogeosciences, 11, 5349-5363, 2014 www.biogeosciences.net/11/5349/2014/ find that only 32-49 % of the neutral sugars produced by bacteria in the ASW glu samples were degraded within a period of 26 days, from day 6 to day 32. Neutral sugar containing molecules produced by bacteria appear to be less bioavailable than those produced by algae. The molecular composition of neutral sugars in the deep ocean is significantly different from that observed at the end of our incubation experiments, although both reflect a high degree of bacterial degradation and transformation. Our results indicate that bacterially degraded neutral sugars have a high galactose mol % (Table 1, Fig. 6), but measurements in the deep ocean reveal a high glucose mol % (Table 3). This discrepancy has been seen in other studies as well (e.g., Amon et al., 2001), and can be due to fundamental environmental differences between the ocean interior and bioassay incubations and the different timescales of carbon cycling. However, the difference can possibly also be due to differences in diagenetic state of the neutral sugars in the deep ocean and in the surface waters sampled for this study. In addition, it is possible that bacterial degradation of particulate organic material (POM) plays a major role in shaping the molecular composition in the deep ocean. POM supports a major fraction of the respiratory carbon demand below the photic layer (Arístegui et al., 2002) and POM is known to have a high glucose content (Hernes et al., 1996;Panagiotopoulos and Sempéré, 2007). In our incubations, only the 10 % inoculum contained POM, and it is likely that our results mainly reflect bacterial degradation of DOM, either DOM produced by bacteria from glucose or DOM initially present in the samples, while deep ocean observations mainly reflect bacterial degradation of POM. Finally, the high glucose mol % in the deep ocean possibly reflects bacterially produced neutral sugars with a high glucose content (Fig. 5), or other neutral sugar sources also having a high glucose mol %, e.g., submarine vent microbes (Skoog et al., 2007). In the Pacific Ocean, the glucose mol % is significantly higher than in the Atlantic Ocean , supporting the hypothesis that the bacterially produced neutral sugars become more important with time in the deep ocean. However, further studies are needed to fully understand bacterial degradation of neutral sugars bound in DOM and POM and the connection between in situ measurements and incubation studies.

Bacterial degradation of amino acids
The molecular composition of amino acids did not vary considerably between samples, nor did it change over the time course of the experiments ( Table 2). The most abundant amino acid was glycine with a mol % between 26 and 40 followed by aspartic acid, glutamic acid, serine and alanine, which all typically ranged from 10 to 15 mol %. The same trends are seen in the ocean, except that the mol % of glycine is somewhat lower (15-32) and the mol % of γ -aminobutyric acid increases with depth (Table 4) Means and standard deviations were calculated from all samples on day 32 (n = 6). Asx did not include the artificial Arctic sample, which had a considerably higher D/L ratio than the rest of the samples (indicated with a star).
is generally associated with highly degraded organic matter (Dauwe et al., 1999), however, the high mol % observed in our incubations are rarely seen in the ocean (Table 4). The γ -aminobutyric amino acid is known to increase with depth in the ocean, and has therefore been used as an indicator of organic matter diagenesis (Dauwe and Middelburg, 1998;Davis et al., 2009). No consistent increase in γ -aminobutyric acid is seen during our incubations, suggesting that the increase observed in the ocean could result from long-term degradation. This hypothesis is consistent with results from a study by Davis et al. (2009), who found an increase in γaminobutyric acid of only 0-7 mol % during 20-33 days of incubation experiments. The four measured amino acids with L-and Denantiomers exhibited a clear degradation pattern (Fig. 7). Despite very different D/L ratios at the beginning of the experiment, all treatments ended up having almost identical ratios after only 32 days ( Table 2). The D/L ratios of amino acids in NSW samples generally remained constant or increased slightly during the incubations. In the glucoseenriched samples, however, the ratio increased significantly, and after 32 days, all treatments had almost similar D/L ratios. This trend was especially clear for aspartic acid (except for the Arctic ASW glu sample indicated in the figure) and serine, where the variability after 32 days was minimal. As exceptions, the Atlantic and Arctic samples seemed to have slightly different D/L ratios of glutamic acid after 32 days, with Atlantic samples being below 0.27 and Arctic samples above 0.35. Hence, the amount of D-amino acids produced relative to L-amino acids utilized was higher in www.biogeosciences.net/11/5349/2014/ Biogeosciences, 11, 5349-5363, 2014  (1996) a Most sampling sites were shallower than 200 m. b Samples collected above 300 m were GF/F filtered and samples collected below 300 m were unfiltered.
Arctic samples than in Atlantic samples. Previous studies have also reported increased D/L ratios with time in degradation experiments (Amon et al., 2001;Jørgensen et al., 1999) and with depth in the ocean (Dittmar et al., 2001;Kaiser and Benner, 2008). The common endpoints in the D/L ratios observed here have not previously been observed, and suggest that a degradation signature exists. It appears that DOM sources (Arctic DOM, Atlantic DOM and glucose) and the bacterial community (Arctic versus Atlantic) have a minor influence on D/L amino acid ratios, except for glutamic acid, where the bacterial community structure to some extent might influence the ratio. The fact that D/L ratios in the NSW samples equal bacterially produced D/L ratios in ASW glu samples after 32 days indicates that bacteria are the dominant source of amino acids in semi-labile and refractory DOM. The D/L ratios observed at the end of the incubation experiments were considerably higher than D/L ratios in the deep ocean (Kaiser and Benner, 2008;McCarthy et al., 1998;Pérez et al., 2003). In our experiments as well as in the ocean, the highest D/L ratio is that of aspartic acid, followed by the D/L ratio of alanine and the (almost equal) D/L ratios of glutamic acid and serine. However, the D/L ratios observed after 32 days in our incubations were about 0.1 to 0.3 units higher than in the deep ocean (Fig. 7, McCarthy et al. 1998). The timescale of the incubation experiments was much shorter than the timescale of the ocean circulation, and it is possi-ble that deep ocean amino acids exhibit a different D/L ratio due to a higher degree of diagenesis. The observed difference could also reflect the fact that the incubations included a larger fraction of DOM of bacterial origin and therefore had a higher D-amino acid content than DOM and POM in the ocean. The main sources of D-amino acids are bacterial cell wall membrane components, including peptidoglycans, lipopolysaccharides and lipopeptides (Kaiser and Benner, 2008). The ASW glu samples demonstrated that during bacterial utilization of glucose and subsequent degradation of bacterial remnants, DOM with high D/L amino acid ratios is produced. The DOM and POM available for bacterial degradation in the ocean, however, is probably derived from many different sources (Kaiser and Benner, 2008).

Implications
Bacteria play an important role in shaping the concentration and composition of neutral sugars and amino acids in the ocean. Our results show that bacteria are capable of changing the composition of biomolecules bound in DOM significantly within short timescales, and that biomolecule yields can be used as indicators of DOC bioavailability. Bacterial transformation of labile DOM to refractory DOM via the microbial carbon pump has been suggested as an important production pathway of refractory DOM (Jiao et al., 2010;Ogawa et al., 2001). Results from the present incubations indicate that the microbial carbon pump also applies Biogeosciences, 11, 5349-5363, 2014 www.biogeosciences.net/11/5349/2014/ for neutral sugars and amino acids: the molecular composition of biomolecules produced by bacteria and biomolecules remaining after degradation of bacterial remnants in ASW glu samples was strikingly similar to the composition of semilabile or refractory biomolecules remaining in NSW samples after 32 days, suggesting that bacterially produced biomolecules can persist for long periods in the ocean. The present study provides preliminary indications of microbial production of refractory biomolecules via the microbial carbon pump, but further studies are needed to test this hypothesis and better understand its quantitative importance.