Size-fractionated Dissolved Primary Production and Carbohydrate Composition of the Coccolithophore Emiliania Huxleyi

Extracellular release (ER) by phytoplankton is the major source of fresh dissolved organic carbon (DOC) in marine ecosystems and accompanies primary production during all growth phases. Little is known, so far, on size and composition of released molecules, and to which extent ER occurs passively, by leakage, or actively, by exuda-tion. Here, we report on ER by the widespread and bloom-forming coccolithophore Emiliania huxleyi grown under steady-state conditions in phosphorus-controlled chemostats (N : P = 29, growth rate of µ = 0.2 d −1) at present-day and high-CO 2 concentrations. 14 C incubations were performed to determine primary production (PP), comprised of particulate (PO 14 C) and dissolved organic carbon (DO 14 C). Concentration and composition of particulate combined carbohydrates (pCCHO) and high-molecular-weight (> 1 kDa, HMW) dissolved combined carbohydrates (dCCHO) were determined by ion chromatography. Information on size distribution of ER products was obtained by investigating distinct size classes (< 0.4 µm (DO 14 C), < 0.45 µm (HMW-dCCHO), < 1000, < 100 and < 10 kDa) of DO 14 C and HMW-dCCHO. Our results revealed relatively low ER during steady-state growth, corresponding to ∼ 4.5% of primary production, and similar ER rates for all size classes. Acidic sugars had a significant share on freshly produced pCCHO as well as on HMW-dCCHO. While pCCHO and the smallest size fraction (< 10 kDa) of HMW-dCCHO exhibited a similar sugar composition, dominated by high percentage of glucose (74–80 mol %), the composition of HMW-dCCHO size classes > 10 kDa was significantly different, with a higher mol % of arabinose. The mol % of acidic sugars increased and that of glucose decreased with increasing size of HMW-dCCHO. We conclude that larger polysaccha-rides follow different production and release pathways than smaller molecules, potentially serving distinct ecological and biogeochemical functions.


Introduction
The global ocean inventory of dissolved organic carbon (DOC) is estimated to be in the range of 662-700 Gt (Hansell and Carlson, 1998;Ogawa and Tanoue, 2003).A common classification of marine DOC relies on its reactivity and discriminates between labile (LDOC), semi-labile (SLDOC), semi-refractory (SRDOC), refractory (RDOC) and ultrarefractory (URDOC) DOC with lifetimes of hours to days, weeks to months, month to years, centuries or even millennia (Kirchman, 1993;Carlson and Ducklow, 1995;Anderson and Williams, 1999;Hansell, 2013).Only a small fraction of marine DOC is considered reactive: LDOC (< 0.2 Gt) and SLDOC (6 ± 2 Gt) (Hansell, 2013).In general, these compounds are freshly produced by plankton and represent the major nutritional resource for heterotrophic microorganisms (Cherrier et al., 1996;Amon and Benner, 1996;Amon et al., 2001;Benner, 2002;Azam and Malfatti 2007;Davis et al., 2009).Especially during the summer season, SLDOC can accumulate in temperate waters, becoming available for deep convective mixing, contributing to the biological carbon pump (Hopkinson and Vallino, 2005;Hansell et al., 2009).Microbial assimilation of DOC, as well as the formation of gel particles, such as transparent exopolymer particles (TEPs), leads to a repartitioning of DOC into the particulate organic carbon (POC) pool (Alldredge et al., 1993;Chin et al., 1998;Engel et al., 2004), the sinking of which represents Published by Copernicus Publications on behalf of the European Geosciences Union.
C. Borchard and A. Engel: Size-fractionated dissolved primary production and carbohydrate composition another pathway for carbon export and storage in the ocean.In addition, microbial processing of fresh DOC may result in formation of recalcitrant compounds with longer residence time, also increasing the carbon dioxide (CO 2 ) storage potential in the ocean (Jiao and Zheng, 2011).Thus, deeper insights into the origin and quality of DOC in the ocean can greatly abet our ability to quantify carbon and nutrient cycling in the ocean.
ER is a normal function of healthy algae cells during all stages of growth (Fogg, 1966;Mague et al., 1980;Bjørnsen, 1988;Borchard and Engel, 2012;Lopez-Sandoval et al., 2011) and can comprise up to 80 % of primary production (Sharp, 1977;Mague, 1980;Fogg, 1983;Bjørnsen, 1988).Two conceptual models have been proposed for phytoplankton ER: (i) the passive diffusion model that describes the leakage of smaller molecules from inside the cell to its surrounding environment (Fogg, 1983;Bjørnsen, 1988), and (ii) the overflow model that assumes an energy consuming exudation of HMW compounds (Fogg, 1983;Nagata, 2000;Schartau et al., 2007).According to the passive diffusion model, DOC crosses the cell membrane independently from primary production (PP) during day and night, and ER correlates to phytoplankton biomass and cell size.A higher relative contribution of ER to total PP would therefore be expected in communities dominated by small cells due to their higher surface-to-volume ratio (Bjørnsen, 1988;Kiørboe and Hansen, 1993;Marañón et al., 1996).
Central aspects of the overflow model are a dependence of ER on PP rates, the absence of ER at night and a high share of HMW substances (Williams, 1990, and references therein;Nagata, 2000).Fogg (1966) proposed that photosynthesis and build-up of organic carbon is primarily regulated by irradiance, while cell growth is controlled by the availability of inorganic nutrients.The discharge of photosynthesates not utilized for cell growth was suggested to be more energy efficient than intracellular storage (Wangersky, 1978;Wood and van Valen, 1990).In accordance with the overflow model, data from coastal, marine and estuarine systems revealed a linear relationship between PP and ER, and factors influencing PP were suggested to also affect ER (Baines and Pace, 1991).Such effects were shown for light (Zlotnik and Dubinsky, 1989) and later also suggested for CO 2 (Engel, 2002) and temperature (Moran et al., 2006).Under nutrient limitation, however, substantial ER was observed when PP was reduced, leading to higher percentage of extracellular release (PER) (Myklestad et al., 1989;Goldman et al., 1992;Obernosterer and Herndl, 1995;Halewood et al., 2012).Under such conditions, decoupled from PP and biomass, ER becomes difficult to estimate, both in terms of quantity and quality.Moreover, phytoplankton cells display a large physiological plasticity for nutrient requirements, i.e., the nutrient cell quota, which varies with environmental conditions or among different taxonomic groups (Geider and LaRoche, 2002).
Despite their role in marine carbon cycling, processes involved in the production, consumption and remineralization of extracellular organic matter are little understood and have largely been neglected in biogeochemical models (Flynn et al., 2008;Repeta and Aluwihare, 2006;Hansell et al., 2009;Hansell, 2013).So far, it is not known whether extracellular products are mainly released by leakage or by exudation processes, or how much leakage and exudation products differ.We also do not know whether and how the physiological status of the cell influences the composition of extracellular products, and whether or not such differences in chemical signatures subsequently affect their microbial cycling, remineralization rate or affinity to form gel particles.
In order to improve our understanding on ER, we conducted a chemostat experiment with E. huxleyi under fully controlled nutrient supply and growth rate.Emiliania huxleyi is a bloom-forming cosmopolitan coccolithophore species, and known to produce a methylated, acidic polysaccharide that plays a central role in coccolith formation and agglutination (Fichtinger-Schepmann, 1979;De Jong, 1979).ER by E. huxleyi cells was reported earlier (Aluwihare and Repeta, 1999;Biddanda and Benner, 1997;Borchard and Engel, 2012) and carbohydrates were shown to provide a substantial fraction of freshly produced HMW-DOC (35-94 %) (Aluwihare and Repeta, 1999;Biddanda and Benner, 1997).
This study was part of a larger experiment investigating carbon and nutrient cycling under different pCO 2 conditions at steady-state growth in E. huxleyi.No effect of the CO 2 treatment was observed for elemental stoichiometry of cells as well as for TEP production (Engel et al., 2014).This study focuses on primary production of POC and DOC by E. huxleyi, the carbohydrate composition of cells and for the first time on different size fractions of released compounds.
With our study we wanted (i) to determine ER of DOC and carbohydrates by combining rate measurements for particulate and dissolved primary production with analyses of carbohydrate concentration, and (ii) to characterize monomeric carbohydrate composition in different size classes of DOC in order to elucidate mechanisms of ER.We chose the continuous culture approach, because here cells can be grown under nutrient limitation at steady-state biomass.Thus, in a chemostat the increase in extracellular organic matter can primarily be attributed to growing phytoplankton cells and not to cell lysis and decay, processes that co-occur with ER when batch cultures or natural populations become nutrient depleted.

Experimental setup
A calcifying strain of E. huxleyi (PML B92/11) was grown as a continuous culture in two chemostats (∼ 9.2 L each) at a constant dilution rate of D = 0.2 d −1 .A more detailed description of the chemostat principle and the experimental setup are given by Borchard et al. (2011), respectively.Temperature was set to 14.0 ± 0.1 • C. Irradiance was provided at a 16 h / 8 h light / dark cycle with a photon flux density of 19 µmol photons m −2 s −1 (TL-D Delux Pro, Philips; QSL 100, Biospherical Instruments Inc.).Nutrient medium was prepared from sterile-filtered (Sartobran P, 0.2 µm capsule, Sartorius) aged natural seawater (NSW) with a salinity of 33, total alkalinity (TA) of 225 µmol kg −1 seawater and a pH of 8.24.The seawater was enriched with nutrients according to the f/2 recipe of Guillard and Ryther (Guillard and Ryther 1962) with final concentrations of 43 µmol L −1 NO − 3 and 1.5 µmol L −1 PO 3− 4 .The nutrient medium was treated for 3 h with UV irradiation (Microfloat 1/0, a.c.k.aqua concept GmbH) for sterilization before the addition of sterilefiltered (0.2 µm, Minisart, Sartorius) f/2 vitamins.Axenic conditions, however, could not be maintained in the 9.2 L chemostats over the long period of time.
Equilibration of the medium with CO 2 was obtained by constant aeration with 380 and 75 µatm CO 2 , respectively.To minimize effects of calcification by E. huxleyi on carbonate chemistry in the incubators, TA in the reservoir tank was increased by addition of bicarbonate (LaRoche et al., 2010), resulting in 2460 µmol kg −1 seawater.E. huxleyi cells were pre-cultured for 30 days at prescribed CO 2 concentrations and temperature conditions in f/2 media in order to avoid short-term stress effects on cell physiology.Each chemostat incubator was then inoculated to a final density of ∼ 5000 cells mL −1 .Cultures were grown in batch mode for 5 days until the constant medium supply was applied at a dilution rate (D) of D = 0.2 d −1 .Cells were kept in suspension by gentle mixing at 50 rpm.Here, we report data derived from samplings during steady-state growth on experimental day 30, 34, 38, 42 and 44 for 14 C rate measurements and on day 38, 42 and 44 for carbohydrate analyses and size fractionations of those and 14 C exudation.All samples were taken 3 h after lights-on to avoid biases due to physiological variations during the day-night cycle.

Cell density and chemical analysis
Cell density was determined daily as the mean of three consecutive measurements of 50 µL by an electronic particle counter (Coulter Multisizer III, Beckman Coulter) equipped with a 10 µm aperture.In order to dilute the samples to 1 : 100, 0.2 µm of pre-filtered (Minisart, 2000; Sartorius) NSW with a salinity of 33 was used.After microscopic inspection, particles with an equivalent spherical diameter in a range of 3.2 to 8 µm were identified as E. huxleyi cells.
Primary production and exudation were measured by applying the 14 C incubation method according to Steemann Nielsen (Steemann Nielsen, 1952) and Gargas (Gargas, 1975).Triplicate samples (75 mL each) were taken from each chemostat, transferred into cell culture flasks (25 cm 2 , Corning ® ) and spiked with approximately 5 µCi NaHCO − 3 (Hartmann Analytics, specific activity 40-60 mCi mmol −1 ).Each triplicate set was incubated for about 4 h at original experimental light and temperature settings, but without aeration.Simultaneously, dark uptake was measured in triplicate from 75 mL samples incubated in the dark.Added activity in the samples was determined by removing a 10 µL aliquot from three dark bottles prior to incubation and transferred to 6 mL liquid scintillation vials in which 200µl of 2N NaOH was placed.Four milliliters of liquid scintillation cocktail (Ultima Gold AB) was added before counting.Incubations were stopped by gentle filtration on 0.4 µm polycarbonate filters (Nucleopore) at low vacuum (< 150 mbar) to avoid cell breakage.The filters (PO 14 C) were covered with 25 µl of 1 M HCl in order to remove inorganic 14 C.After a few seconds they were rinsed with 10 mL of filtered seawater.Filters were transferred to 6 mL scintillation vials, 4 mL of liquid scintillation cocktail (Ultima Gold AB) was added, and samples were stored overnight before being counted in a Packard Tri Carb Liquid Scintillation Counter.Carbon incorporation rates were calculated in accordance to Borchard and Engel (2012).

HMW-dCCHO DO 14 C
Total 1 kDa< HMW-dCCHO < 0.45 µm DO 14 C < 0.40 µm Very large 1000 kDa < HMW-dCCHO < 0.45 µm 1000 kDa < DO 14 C < 0.40 µm Large 100 kDa < HMW-dCCHO < 1000 kDa 100 kDa < DO 14 C < 1000 kDa medium 10 kDa < HMW-dCCHO < 100 kDa 10 kDa < DO 14 C < 100 kDa Small 1 kDa < HMW-dCCHO < 10 kDa DO 14 C < 10 kDa For the determination of released dissolved organic carbon (DO 14 C), 4 mL of the filtrate was transferred into 20 mL scintillation vials and acidified to pH < 2 by the addition of 10 µL of 1 M HCl and left open under the fume hood for 24 h.For size fractionation of DO 14 C, triplicate sets of 10 mL sample were transferred into Macrosep ® centrifugal devices with a membrane cut-off of < 1000, < 100 and < 10 kDa, respectively.After centrifugation (Heraeus, Megafuge ® 1.0 R) for 15 min at 4000 rpm, 4 mL of sample was transferred into 20 mL liquid scintillation vials.In the following, samples were treated as the whole DO 14 C samples and after the outgassing of inorganic 14 C, 15 mL of liquid scintillation cocktail was added.Counting and calculations were performed following Borchard and Engel (2012).
Primary production (PP) was derived from the sum of PO 14 C and DO 14 C.The PER was calculated as (DO 14 C / PP) × 100.
For size fractionation of HMW-dCCHO, 10 mL of sample was transferred into Macrosep ® centrifugal devices with a molecular weight cut-off (MWCO) of 1000, 100 and 10 kDa, respectively.After centrifugation (Heraeus, Megafuge ® 1.0 R) for 15 min at 4000 rpm, samples were transferred into combusted (8 h at 500 • C) glass vials and stored at −20 • C. Before usage, Macrosep ® devices were rinsed twice via centrifugation with ultrapure water to avoid any contamination with carbohydrate compounds in the membrane.Concentrations of CCHO in these blanks were tested to be below the detection limit and therefore did not affect analyses.
Size fractions of DO 14 C and HMW-dCCHO obtained using Macrosep ® centrifugal devices were subtracted from each other in order to present data for each size class.Definitions for size classes are given in Table 1.
For total alkalinity (TA), 25 mL of each sample was measured by titrating with 0.05 M HCl until the buffering capacity of the water samples was consumed and all bases of interest were protonated to zero-level species.Analysis was done with an automatic titrator (TitroLine ® alpha plus, SI Analytics) equipped with a sample changer (TW alpha plus, SI Analytics) and a piston burette (Titronic ® 110 plus, SI Analytics).The pH was monitored by a two-point-calibrated (buffer solution pH 4.006 and pH 6.865; Applichem, standardized according to DIN 19266) electrode (Schott ® Instruments Io-Line).The concentration of TA in µmol kg −1 seawater was calculated from linear regression of the absolute numbers of protons in solution and the total volume (sample plus HCl) in the range of pH 4 and 3. Determination of the seawater carbonate chemistry was conducted using the program CO2SYS (Lewis and Wallace 1998), with pH (calibrated using reference materials provided by A. Dickson) and TA being the input parameters.

Data treatment
All samplings were performed during the steady-state period of the experiment, when the growth rate (µ) was equal to the dilution rate (D).The samplings over time thus represent replicates of the same physiological state and values of the respective parameters are given as average ± standard deviation.Since CO 2 induced no differences between the presentday and high-CO 2 chemostat, they were used as replicate treatments and values are given as mean values with 1 standard deviation if not stated otherwise.
In order to relate daily rates (µmol L −1 d −1 ) directly to concentrations (µmol L −1 ), data were converted into each other by applying a growth rate of 0.2 d −1 .For cellnormalized carbon values, concentrations and rates were divided by the cell number.
Differences in carbohydrate composition for the different size fractions were tested by means of analysis of covariance (two-way ANOVA).Differences as a response to CO 2 conditions were tested by means of a t test.Statistical significance was accepted for p < 0.05.All calculations were performed using the software package Sigma Plot 10.01 (SysStat).

Growth, nutrients and carbonate chemistry
Growth and biogeochemical composition of Emiliania huxleyi as well as carbonate and nutrient chemistry during this chemostat experiment are described in more detail in Engel et al. (2014).Briefly, on day 28 of the experiment, the steady state was reached, with the dilution rate (D) being equal to the growth rate (µ) of E. huxleyi.Cell abundances and basic parameters such as particulate organic carbon (POC), nitrogen (PON), phosphorus (POP) and chlorophyll a (Chl a) remained constant until the end of the experiment, proving the constant physiological state of E. huxleyi ( Engel et al., 2014).
During the steady-state period, cell densities were similar in the present-day and high-CO 2 treatment and averaged 5.2 × 10 8 ± 18.6 and 5.1 × 10 8 ± 19.7 % cells L −1 , respectively.High variations resulted exclusively from intensive sampling between days 42 and 44.Until day 42 variations did not exceed 11.6 % and biomass production was accepted as balanced growth as a result of controlled nutrient supply.During steady state (days 30-44), both NO − 3 and PO 3− 4 concentrations were below the detection limit in both treatments.P limitation was likely more severe than N-limitation, given a nutrient supply N : P ratio of ∼ 29 and indicated also by PON : POP ratios clearly > 16 (Engel et al., 2014).pCO 2 was calculated from pH and TA and yielded significantly different values between treatments of 337 ± 94 (present day) and 623 ± 139 (high CO 2 ) µatm.Time-averaged values given here differ slightly from those given by Engel et al. (2014), as that study used data from replicate chemostats per CO 2 treatment, while only one chemostat per treatment was sampled for the purpose of this study.

Primary production and exudation
As determined for cell densities, PO 14 C and DO 14 C production rates derived from replicate sampling during steady-state growth varied < 11 %, confirming the physiological steady state of E. huxleyi grown in the chemostats.PO 14 C production of 173 ± 17 and 168 ± 16 µmol C L −1 d −1 and DO 14 C production of 8.0 ± 0.7 and 8.2 ± 1.1 µmol C L −1 d −1 were determined for the present-day and high-CO 2 µatm treatment, respectively (Fig. 1).Production rates of PO 14 C and DO 14 C were not significantly different between the CO 2 treatments (Mann-Whitney rank-sum tests and t tests, n = 5, p > 0.69) and were thus averaged for both treatments: 171 ± 16 (PO 14 C) and 8.1 ± 0.9 µmol C L −1 d −1 (DO 14 C).
Cell-normalized production of PO 14 C and DO 14 C during the steady-state period was on average 0.33 ± 0.04 and 0.015 ± 0.002 pmol C cell −1 d −1 , respectively for both treatments.Similar PO 14 C and DO 14 C production rates in both chemostats are reflected in comparable PER of 4.42 ± 0.22 (present day) and 4.70 ± 0.92 % (high CO 2 ); also, for the size classes of DO 14 C, no CO 2 effect was determined (Fig. 1).
Averaged for both treatments, size-fractionated (see Table 1 for definition) DO 14 C production ranged between 1.27 ± 0.53 (medium) and 2.74 ± 0.88 µmol C L −1 d −1 (very large).Relative contribution of different DO 14 C size classes to total DO 14 C was 33.6 ± 9.31 (very large), 24.6 ± 7.90 (large), 15.9 ± 7.15 (medium) and 25.8 ± 3.55 % (small).Thus, total DO 14 C was comprised of comparable shares of DO 14 C in these size classes with slightly higher proportions in the very large fraction.
Table 2. Size class resolved production rates of high molecular weight (> 1 kDa) carbohydrates (HMW-CCHO) and of fresh organic carbon ( 14 C) during the chemostat experiment, as well as contribution of carbon contained in HMW-CCHO to primary production ( 14 C) in particulate matter and in different size fractions of dissolved organic carbon .Values represent averages ± standard deviation of replicate samplings and both treatments, n =6.

HMW -CCHO
14 C HMW-CCHO : ) for fresh HMW-dCCHO, and hence very similar between the two CO 2 treatments (Fig. 2).Averaged for both treatments, 87 ± 3 % of tCCHO was present in the particulate fraction (pCCHO).E. huxleyi produced pCCHO on the order of 104 ± 27 µmol C L −1 (0.20 ± 0.02 pmol C cell −1 ), equivalent to 20.7 ± 5.3 µmol C L −1 d −1 at a growth rate of 0.2 d −1 , representing about 12.5 % of the daily produced PO 14 C (Table 2).Freshly produced HMW-dCCHO was 15 µmol C L −1 (0.043 ± 0.004 pmol C cell −1 ), equivalent to about 40 % of freshly produced DO 14 C. Fresh carbohydrate concentrations in various size classes (see Table 1 for definition) also revealed a strong similarity between the present-day and the high-CO 2 treatment (Fig. 2, t tests, n = 6, p > 0.269) and are therefore given as average values in the following.

Carbohydrate composition of exudates
Sugar monomers of three different types comprised the combined carbohydrates (CCHO) determined during the present experiment: neutral sugars (Fuc, Rha, Ara, Gal, Glc and coeluting Man / Xyl), amino sugars (GalN and GlcN) and uronic acids (Gal-URA and Glc-URA).Various amounts of these monomers were detected in HMW-dCCHO of the initial NSW used for the present experiment (Fig. 3b (p > 0.462).However, relative to the other size fractions Man / Xyl was slightly enriched in the small size fraction, while a smaller proportion of Fuc was detected in this fraction.No significant differences in monomeric composition of CCHO produced by Emiliania huxleyi were determined between the present-day and high-CO 2 treatment (p > 0.881).
Therefore, average values are given for replicate sampling during steady-state growth and both treatments in the following (Fig. 3 and Table 3).
Man / Xyl, Ara and Glc-URA ranged between 3.2 ± 1.7 and 4.2 ± 1.0 mol %, while Fuc and the amino sugars GalN and GlcN contributed only a minor fraction (< 0.5 mol %) to pCCHO (Table 3).Hence, composition of pCCHO was substantially different from the composition determined for freshly produced HMW-dCCHO (p < 0.002), except for the proportions of Gal and Glc-URA.This difference is mainly attributed to a smaller proportion of Glc in the dissolved fraction along with a more than 10-fold higher share of Ara and also higher proportions of Man / Xyl and Gal-URA (Table 3).
Carbohydrate composition of the investigated dCCHO size fractions was also significantly different (p < 0.002).Ara was dominant in very large, large and medium HMW-dCCHO, but not in the small fraction, in which its contribution was significantly smaller than in all other size classes (p < 0.01).Most interestingly, the proportion of Glc increased with decreasing size class, while the proportion of Gal-URA clearly decreased (Fig. 4).In the small fraction, Glc contribution was 80 ± 12 mol % and significantly higher than in all other size classes of HMW-dCCHO (p < 0.002) (Table 3).The contribution of Glc to very large dCCHO was negligible (< 0.5 mol %).In contrast, Gal-URA contributed 18 mol % to the very large fraction, but only 1 mol % to the small fraction.Proportions of Gal also decreased with the smaller HMW-dCCHO size classes, albeit not as clearly as for Gal-URA.Gal ranged from 6 (very large) to < 0.5 mol % (small).Contributions of Rha and Man / Xyl varied among size classes mol % of Fuc and both amino sugars as well as GalN and GlcN were negligible.

Particulate and dissolved primary production
Nutrient limitation and low growth rate did not hamper organic carbon production of Emiliania huxleyi during the present study.Cell normalized production of PO 14 C was on average ∼ 0.33 pmol C cell −1 d −1 and well within the range of published values (0.12-0.64 pmol C cell −1 d −1 ; Biddanda and Benner, 1997;Borchard and Engel, 2012).The partitioning of organic carbon between dissolved and particulate pool was shown earlier to be highly influenced by environmental conditions such as light, temperature and nutrient supply (Myklestad and Haug, 1972;Zlotnik and Dubinsky, 1989;Staats et al., 2000;Wetz and Wheeler, 2007).Nutrient depletion, however, seems to be the major factor leading to excess DOC excretion from algae cells to the surrounding environment and was reported from a variety of field and lab experiments (Fogg, 1983;Wood and Van-Valen, 1990;Smith and Underwood, 2000;Lopez Sandoval, 2010, 2011).Extracellular release (ER) in the range of 0-80 % has been reported over the past decades, and only after a long-lasting debate primarily concerning methodological constraints (Sharp, 1977;Mague, 1980;Fogg, 1983;Bjørnsen, 1988) has it become accepted today that ER is a normal function of healthy algae cells occurring during all stages of growth.In exponentially growing cells in culture, ER typically ranges between 2 and 10 %, while in natural marine environments ER is generally higher by 10-20 % (see Nagata, 2000, and references therein).A relatively constant PER of 20 % was reported for field samples over different ecosystems covering oligotrophic and eutrophic regions (Marañón et al., 2005).Increased PER (up to 37 %), however, were observed for nutrient-limited algae during the transition period of exponential to stationary growth and during senescence of natural phytoplankton communities (Lopez-Sandoval, 2010, 2011;Engel et al., 2013).In chemostats, despite the strict control of nutrient supply and growth rate, cells still grow exponentially.A decoupling of carbon to nutrient metabolism in continuous cultures can occur due to a change in growth rate (e.g., change the inflow of nutrient media) and results in changes in the partitioning between dissolved and particulate carbon pools, as shown with the same E. huxleyi strain (B 92/11) by Borchard and Engel (2012).In their study, down-regulation of the growth rate from µ = 0.3 d −1 to µ = 0.1 d −1 induced a slight increase in DO 14 C production, while the PO 14 C production was significantly minimized, resulting in higher PER.Cells then adapted to the steady state and high PER remained constant.During the present study, growth of E. huxleyi was also balanced to the nutrient supply but cells were not exposed to any stress due to nutritional changes.Thus, production of DO 14 C was not explicitly stimulated by changing experimental conditions, and, albeit constantly P limited, the cell-normalized DO 14 C production of ∼ 0.015 pmol C cell −1 d −1 represented an ER of ∼ 4.5 %, well within the abovementioned range for non-stressed algae.Full acclimation to environmental conditions during steady-state growth may also explain the absence of a CO 2 effect on primary production and exudation during this study, and shows that E. huxleyi is in principle capable of acclimating to different CO 2 concentrations.Engel et al. (2014) suggested that exudation may be more sensitive to changes in pCO 2 during transient growth phase, such as towards the end of phytoplankton blooms, when cells become nutrient limited.Indeed, significant responses of ER to changes in pCO 2 have mainly been reported for phytoplankton blooms (Engel et al., 2013), batch and semi-continuous cultures (Thornton, 2009;Barcelos and Ramos, 2014), or when growing conditions changed during chemostat studies (Borchard and Engel 2012).

Combined carbohydrate production
HMW-dCCHO freshly produced by C. huxleyi during steady-state growth represented about 40 % of freshly produced DO 14 C (Table 2).This is a lower estimate because low-molecular-weight DOC (< 1 kDa, LMW) would be detected with the 14 C-incubation method (Steemann Nielsen, 1952) during the determination of DO 14 C, but would escape the analysis of HMW-dCCHO due to the molecular cut-off > 1 kDa during desalinization of seawater samples (Engel and Händel, 2011).In the surface ocean, HMW compounds of dissolved organic matter (DOM) were found to be more abundant (30-35 %) compared to deeper waters (20-25 %) and it was concluded that HMW-DOM inherits a higher reactivity and shorter lifetimes, while LMW-DOM is rather refractory (Amon and Benner, 1996;Ogawa and Tanoue, 2003).Major reaction processes of HMW compounds are heterotrophic degradation (Amon and Benner, 1996;Guo et al., 2002;Aluwihare and Repeta, 1999) and gel particle formation (Mari and Burd, 1998;Leppard, 1995;Passow, 2000;Passow, 2002, and references therein).Thus, the HMW-DOM pool is directly linked to processes significant for organic carbon dynamics, nutrient cycling and oxygen consumption in the ocean.Assembly and coagulation of polymeric precursors has been proposed as a mechanism leading to the formation of marine gel particles, such as TEPs.Specifically, divalent cation bridging of acidic sugars, such as uronic acids, is assumed to be involved in bonding between polysaccharide chains.The release of larger polysaccharides with relatively high mol % Gal-URA as observed for E. huxleyi in this study may be an important first step for high TEP concentrations observed previously (Engel et al., 2004;Harlay et al., 2009).However, absolute rates of ER were relatively low and apparently insufficient to induce TEP formation during this study.Engel et al. (2014) suggested that responses to variations in environmental factors, specifically to changes in nutrient supply, are responsible for excess carbon accumulation inside the cell and for exudation of carbohydrates.Sampling during this study was conducted during the period of steady-state growth.This may explain the observed relatively low rates of ER, including potential TEP precursors.

Monomeric composition of CCHO
NSW used in the present study to prepare the nutrient media was collected from the North Sea and kept under dark and cool conditions for several months before usage.HMW-dCCHO monosaccharide composition of NSW was dominated by Glc (24 mol %) and Man / Xyl (24 mol %).Also, high mole percentage (∼ 10) of Fuc, Gal, Gal-URA and Glc-URA were determined, while other monomers were of minor importance (Fig. 3b, left panel).The composition of the aged NSW used here differs from what was obtained from the Northwest Atlantic, the Sargasso Sea and the Gulf of Mexico (Aluwihare et al., 1997, and references therein), especially concerning comparably low proportions of Rha and Gal (Fig. 3b, left panel).Differences in carbohydrate composition of the seawater can be explained by seasonal or geographical divergences as well as by storage time of NSW.Monomeric composition of HMW-dCCHO released by E. huxleyi during the present experiment was substantially different from the initial NSW composition (Fig. 3b) and the compositional shift was primarily induced by a profound relative increase in Ara.The HMW-dCCHO and pCCHO derived from E. huxleyi during this experiment contained a similar composition as determined earlier for cellular and extracellular carbohydrates derived from this species (De Jong et al., 1979;Fichtinger Schepman et al., 1979;Nanninga et al., 1996;Bilan and Usov, 2001).Cellular pCCHO of E. huxleyi clearly differed from not only NSW but also HMW-dCCHO (Fig. 3b, right panel).This is in accordance with previous studies showing differences between intracellular and extracellular CCHO compositions for various algae (Mague, 1980;Aluwihare, 1999Aluwihare, , 2002)).
Neutral sugars generally dominated the HMW-dCCHO composition with ∼ 83 mol %.These results are consistent with findings by Aluwihare (1999), who report on HMW exudates from E. huxleyi being mainly composed of neutral polysaccharides with Ara as the dominant monomer (30 mol %).However, the fraction of Ara observed during this study is considerably higher than reported for ultrafiltered DOM (> 1 kDa) by Biersmith and Benner (1998), who also investigated non-axenic E. huxleyi as a batch culture, and for HMW-dCCHO sampled during a field study in the Bay of Biscay where coccolithophores and presumably E. huxleyi was the dominating phytoplankton organism (Engel et al., 2012); both studies reported Ara of ∼ 3 % mol.Apart from well-documented species-specific differences in CCHO composition (Aluwihare and Repeta, 1999;Myklestad, 1974;Myklestad et al., 1989), variations in the composition of algal extracellular carbohydrates may be related to physiological and ecological functions.Although freshly produced DOC is generally a primary substrate for heterotrophic uptake, E. huxleyi exudates were shown to exhibit recalcitrant features (Nanninga et al., 1996).Degradation experiments with the diatom Thalassiosira weissflogii revealed a special role of Ara in carbohydrate accessibility, as it escaped bacterial degradation over a period of 2 weeks (Aluwihare and Repeta, 1999).Bacterial cell numbers during the present experiment were relatively high, between 2 and 3 × 10 6 mL −1 , contributing ∼ 2 % to particulate organic carbon (POC) and ∼ 3 % to DOC (Engel et al., 2014).Assuming a bacterial growth efficiency of 60 % (upper limit; Del Giorgio and Cole, 1998), the bacterial carbon demand could have been about 2 % of POC and 5 % of DOC.Relative to the freshly produced DO 14 C derived from rate measurements, however, a share of up to 20 % may have been channeled into heterotrophic turnover.This means that PER would be underestimated by 20 % at most.The HMW-CCHO was thus to some extent subject to bacterial reworking, and the high proportions of Ara may be a result of the selective removal of other monomers.In accordance with the findings of Aluwihare (1999), concentration of Ara in dCCHO remained unchanged during a degradation experiment with the same E. huxleyi strain investigated here, while dCCHO was reduced by ∼ 60 % (Piontek et al., 2010;J. Piontek, personal communication, 2014).However, we would expect extensive microbial degradation of larger dCCHO to lead to an increase in Ara mol % in the small size fraction.However, this was not observed.
Alternatively, high-mol % Ara and low-mol % Glc may indeed be a characteristic of larger carbohydrate molecules released by E. huxleyi that are recalcitrant to microbial decomposition.Assuming these components are bad substrates for microbial utilization, their controlled exudation, if physiologically necessary, may be ecologically advantageous for algal cells that are competing with bacteria for nutrients such as phosphorus.This corroborates earlier findings of DOM produced at P depletion being more resistant to bacterial degradation (Obernosterer and Herndl, 1995;Puddu, 2003).On the other hand, bacteria recycle organic phosphorus, and a certain degree of bacterial activity will be advantageous for regenerated productivity of algal cells.So far, little is known on how nutrient limitation affects the composition of algal release products.We suggest that nutrient availability may be one factor responsible for variability in carbohydrate composition observed during various studies (Giroldo et al., 2005;Goldberg et al., 2010;Engel et al. 2013).
Assuming a certain degree of microbial modification, another explanation for the difference of CCHO composition between culture studies and those observed in natural seawater may be the highly specific linkage between algal release and bacterial community response, proposed in a series of recent studies (Teeling et al., 2012;Taylor et al., 2014;Kabisch et al., 2014).These studies showed that the release of algal polysaccharides can induce a succession of bacterial communities inhabiting different abilities for enzyme expression related to specific carbohydrate degradation.Because the majority of marine bacteria cannot be kept in culture, bacteria present in this chemostat study, and likely in all culture experiments, represent only a small fraction of the natural diversity.The bacteria present in this study may have left a different fingerprint on polysaccharide composition than natural communities.Short-term incubation studies with natural bacterial communities may be required to better understand the microbial fingerprint on DOM, specifically polysaccharide degradation.A better understanding of the microbial fingerprint on DOM could also allow for tracing microbial degra-dation activities in specific environments, such as the ocean's anoxic zones, or the extreme oligotrophic seas.

Size fractionation of CCHO and DOCconsiderations on extracellular release
Quantitatively, each DO 14 C size fraction contributed similar amounts to total DO 14 C with slightly higher proportions in the very large fraction (Fig. 1 and Table 2).Release rates of HMW-dCCHO were similar for the different size fractions but highest in the very large fraction (Fig. 2, Table 2).On a total basis, ∼ 40 % of produced DO 14 C was characterized as freshly produced HMW-dCCHO (Table 2).Contribution of dCCHO to fresh DOC was lowest in the small size fraction (30 %) and highest in the very large fraction (60 %) (Table 2).Monomeric composition of different size classes of dCCHO enriched by E. huxleyi exudates was profoundly different from those of the aged NSW used as culture media (Fig. 3).In aged NSW, monomers were more evenly distributed among size fractions (Fig. 3b, left panel).In comparison, differences in monomeric composition of size classes in E. huxleyi exudates were largely due to changes in Ara, Glc, and Gal-URA.Most remarkably, Ara the dominant monomer in all larger dCCHO size classes, was of minor importance in the small dCCHO size fraction and lowest in the particulate fraction (Fig. 3, right panel).This is in accordance with the findings of Biersmith and Benner (1998), who also observed lower mol % of Ara for particulate components of an E. huxleyi culture as well as for the cell lysate.In contrast to Ara, mol % of Glc in our study was highest in the particulate and small fraction, relatively small in the medium to large fraction, and negligible in the very large fraction.This also agrees well to earlier findings; Skoog et al. (2008) observed larger mol % of Glc in LMW-CCHO than in HMW-CCHO, while reporting less mol % of Ara in LMW-CCHO than in HMW-CCHO.Thus, differences in size fractions of combined sugar molecules may be one factor responsible for differences in CCHO composition of DOC between study sites.In general, carbohydrate composition in the smallest size class was similar to cellular pCCHO composition, while larger molecules were more distinct (Fig. 3, right panel).The 14 C method (Steemann Nielsen, 1952) applied here to measure primary production and ER of organic carbon does not allow for distinguishing whether DOC is released from the cell passively (i.e., by leakage) or actively (i.e., by exudation).Leakage is hypothesized to be directly related to biomass and cell size, suggesting a constant value of passive PER.The composition of the small size class, and particularly the high share of Glc, resembled the cellular carbohydrate composition (Fig. 3b, right panel).This finding suggests a non-selective, i.e., passive, release of carbohydrates in the smallest size class determined here.Storage glucans in algae are comprised exclusively of Glc in D formation and have a molecular weight of 5-10 kDa.D-Glc was reported as a major component of coccolith polysaccharide (CP) of E. huxleyi (Fichtinger Schepman, 1979).For chloroplasts in higher plants, porins are described that allow transmembrane passage of hydrophilic molecules like sugars and amino acids up to a molecular weight of 10 kDa without the use of energy (Flügge and Benz, 1984;Mohr and Schopfer, 1992).The existence of porins in cell membranes of algae is likely but not explicitly reported.If DO 14 C > 1 and < 10 kDa and associated carbohydrates leak from the cell in accordance to the passive diffusion model, this extracellular release is presumably linear correlated to biomass ("property tax" -Sharp, 1977).For molecules > 10 kDa, however, different mechanisms for the extracellular release are to be expected, since larger molecules cannot pass the membrane by diffusion, and CCHO composition clearly differs from intracellular CCHO (Fig. 3b, right panel).If active release, i.e., exudation, follows the overflow model, biomass growth and dissolved primary production might be strongly decoupled ("income tax" - Sharp, 1977).Moreover, exudation requires a series of physiological processes involved in the synthesis, transport and trans-membrane release of exudates.Hence, exudates likely vary in composition.Data obtained during the present study indicate that components > 10 kDa rich in Ara and Gal-URA and poor in Glc are transported actively through the cell membrane.

Conclusions
Carbohydrates of high molecular weight (> 1 kDa) as a product of primary production are released from nutrient-limited E. huxleyi during steady-state growth.Compositional differences between size fractions of combined carbohydrate suggest that dCCHO > 10 kDa is released by active exudation across the cell membrane, whereas lower-molecular-weight carbohydrates (< 10 kDa) can pass the membrane passively by leakage.The underlying mechanism of the release, however, needs to be further elucidated.To unravel whether the presence of Ara is indeed an indicator of less degradable exudates, as suggested by this study, or whether Ara degradation requires activities of specific bacterial assemblages, further exploration is needed, i.e., by using axenic phytoplankton cultures combined with the addition of natural bacterioplankton communities.At present our understanding of how microbial processes shape the molecular composition of DOM, specifically of carbohydrates, is still in its infancy.This study suggests that dCCHO composition and size may be valuable indicators of processes related to autotrophy such as primary production and exudation but that they may also keep the fingerprint of heterotrophic degradation.A better understanding of compositional changes in dCCHO, as a major fraction of semi-labile DOC, may therefore help to unravel carbon cycling and ecosystem dynamics in the ocean.

Figure 1 .
Figure 1.Dissolved (DO 14 C, left) and particulate (PO 14 C, right) primary production [µmol C L −1 d −1 ] of Emiliania huxleyi at present-day (filled bars) and high-CO 2 (open bars) conditions.Daily rates are additionally given for each DO 14 C size fraction.Each bar corresponds to the average ( ± standard deviation) of replicate samplings (sampling 1-5, n = 5) performed during the steadystate period of the experiment.

Figure 3 .
Figure 3. Concentration [µmol C L −1 ] (a) and composition [mol % CCHO] (b) of high-molecular-weight (> 1 kDa) dissolved combined carbohydrates (HMW-dCCHO).Data are shown for natural seawater used to prepare the experimental culture media (left panels) and composition in natural seawater enriched with freshly produced HMW-dCCHO derived from E. huxleyi (rights panels) grown in chemostats.Due to the strong similarity between the present-day and high-CO 2 treatment, both were treated as replicates.Stacked bars show the average of replicate samplings (samplings 3-5, n = 6) performed during the steady-state period of the experiment.Data for HMW-dCCHO for natural seawater and E. huxleyi taken from Aluwihare (1999) for comparison.Here, only neutral carbohydrates are included, since amino and acidic HMW-dCCHO were not analyzed.

Figure 4 .
Figure 4. Proportions of glucose (Glc) and galacturonic acid (Gal-URA) in high-molecular-weight (HMW > 1 kDa) dissolved combined carbohydrates (dCCHO) of different molecular weight size classes as defined in Table1.Due to the strong similarity between the present-day and high-CO 2 culture, both were treated as replicates.Bars show the average (± standard deviation) of replicate samplings (sampling 3-5, n = 6) performed during the steady-state period of the experiment.

Table 3 .
Freshly produced combined carbohydrates (CCHO) in various size fractions.Average values (bold) and standard deviations (nonbold) in mol % CCHO are given for replicate samplings and both treatments, n = 6.Fuc, GalN and GlcN were always < 0.5 mol % and are not included.