Effects of CO2 perturbation on phosphorus pool sizes

Introduction Conclusions References


Introduction
Increasing emissions of anthropogenic CO 2 into the atmosphere and subsequent acidification of the ocean can potentially affect the diversity of organisms and the functioning of marine ecosystems (Eisler, 2011). The rise of the atmospheric CO 2 content was accelerated from 3.4±0.2 PgC yr −1 in the 1980s to 4.0±0.2 PgC yr −1 in the 2000s lead-5 ing to CO 2 elevation in ocean surface waters the same rate (IPCC, 2013). Atmospheric CO 2 is predicted to rise to 750-> 1000 ppm in 2100 (IPCC, 2001) corresponding with a decrease in pH by 0.3-0.5 units (Caldeira and Wickett, 2005) from the present pH of 8.1. Although this process is of global significance and all parts of the oceans are at risk, there will be regional differences in the degree of acidification (Borges et al., 10 2005). Thus, to determine the CO 2 -related changes in the oceans, multiple studies in different regions are required. Semi-enclosed coastal regions, such as the Baltic Sea, can be more sensitive to CO 2 elevation than open ocean waters due to high freshwater inputs resulting in a reduced buffer capacity (Orr, 2011).
In the Baltic Sea, several studies of CO 2 effects are done on the organismic level 15 of fish (Frommel et al., 2013), zooplankton (Pansch et al., 2012;Vehmaa et al., 2012), macrophytes (Pajusalu et al., 2013), benthic organisms (Hiebenthal et al., 2013;Stemmer et al., 2013), and filamentous cyanobacteria (Czerny et al., 2009;Eichner et al., 2014;Wannicke et al., 2012). The impacts of elevated CO 2 at the ecosystem level, however, have thus far been limited to the Kiel Bay in the western Baltic Sea (Engel In general, there is little knowledge on how the P cycle is affected by ocean acidification and how related changes in P availability influence the response of organisms to CO 2 elevation. In CO 2 manipulation experiments, particulate phosphorus dynamics were studied to determine effects on C : P stoichiometry of phytoplankton (Riebesell and Tortell, 2011) and PO 4 concentration dynamics to estimate its utilization . CO 2 effects on phosphorus pool sizes and PO 4 uptake have so far only been studied by Tanaka et al. (2008) in the Raunefjorden, Norway and by Unger et al. (2013) and Endres et al. (2013) in laboratory experiments with cultures of Nodularia spumigena. Thus, there is a gap of knowledge on how the phosphorus cycle may be affected under future CO 2 conditions. We therefore studied the impact of elevated 15 CO 2 on phosphorus pool sizes, the DOP composition, and PO 4 uptake of northern Baltic Sea plankton community. These measurements provide important information on potential changes in P cycling under future conditions and thus will contribute to a better understanding of potential impacts of increased CO 2 levels in brackish water ecosystems. 20 2 Material and methods

Experimental design and CO 2 manipulation
The study was conducted in the northwestern Gulf of Finland, in the proximity of the Tvärminne Zoological Station (TZS) (Fig. 1), between 17 June and 4 August 2012, using the KOSMOS mesocosm system . Nine mesocosms (M1-Introduction Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | ter depth of ∼ 30 m. Only six of them were included throughout the whole study period since leakages in the remaining three rendered them unusable. Equipment and deployment procedures are described in detail by Paul et al. (2015b). Briefly, polyurethane enclosure bags of 2 m in diameter and 18.5 m in length were mounted in floating frames and lowered in such a way that ∼ 17 m of each bag were immersed in the water column 5 and ∼ 1.5 m remained above the water surface. Large organisms were excluded from the mesocosms by a 3 mm mesh installed at the top and bottom of the bags before closure. The mesocosms were deployed 10 days prior to CO 2 manipulation to rinse the bags and for full water exchange. Sediment traps were mounted on the lower ends to close them water tight, while the upper ends were raised above the water surface to prevent water entry during wave action. The mesocosms were covered with a dome shaped roof to prevent nutrient input by bird and potentially significant fresh water input by rain. Salinity gradients were removed by bubbling the mesocosms with compressed air for 3.5 min, so that 5 days before the start of the experiment (day −5) the water body was fully homogeneous. 15 CO 2 treatment started on day 0 and was repeated on subsequent 4 days by pumping various quantities of 50 µm-filtered and CO 2 -enriched fjord water into seven of the mesocosms as described by Riebesell et al. (2013). The intended CO 2 and pH gradients were reached after the last treatment on day 4. Details are described in Paul et al. (2015b). For the two untreated (control) mesocosms, only filtered fjord water was 20 added to adjust the water volume to that of the treated mesocosms. To compensate for outgassing, the CO 2 manipulation was similarly repeated in the upper 7 m layer of the mesocosms on day 16.

Sampling
Daily sample collection started 3 days before the first CO 2 injection (day −3). Parallel 25 samples were taken from the surrounding fjord. Sampling over the entire 17 m depth was carried out using an integrating water sampler (IWS HYDROBIOS-KIEL) that was Introduction lowered slowly on a cable by hand. The sampling frequency differed depending on the parameter to be observed as shown in the overview by Paul et al. (2015b). Phosphorus pool parameters and uptake rates were determined every second day, except for dissolved organic phosphorus (DOP) components, which were measured every 4 days. Termination of the measurements varied due to logistical constrains.

5
Thus, total phosphorus (TP) and DOP were sampled only until day 29 whereas other parameters were sampled until day 43.
The collected water was filled in HCl-cleaned polyethylene canisters that had been pre-rinsed with sample water. All containers were stored in the dark. Back on land, subsamples were processed immediately for each P-analysis. The other analyses were 10 carried out within a few hours of sample collection and sample storage in a climate room at in situ temperature.

Temperature, salinity, and carbonate chemistry
Measurements in the fjord and in each mesocosm were conducted using a CTD60M 15 memory probe (Sea and sun technology, Trappenkamp, Germany) lowered from the surface to a depth of 17 m at about 0.3 m s −1 in the early afternoon (1: 30-2:30 p.m.).
For these parameters, the depth-integrated mean values are presented here. The carbonate system is described in detail in Paul et al. (2015b). The pH was determined using the spectrophotometric method (Dickson et al., 2007) using a Cary 100 tion of 450 nm and an emission of 670 nm to determine Chl a concentrations (Jeffrey and Welschmeyer, 1997).
A segmented continuous-flow analyzer coupled with a liquid-waveguide capillary flow-cell (LWCC) of 2 m length was used to determine phosphate (PO 4 ) and the sum of nitrite and nitrate (NO 2 + NO 3 ) at nanomolar precision (Patey et al., 2008). The PO 4 determination was based on the molybdenum blue method of Murphy and Riley (1962), and NO 2 +NO 3 on the method of Morris and Riley (1963). PO 4 concentrations from the same subsample were also measured manually using a 5 cm cuvette (Grasshoff et al., 1983). In most of the samplings PO 4 data obtained from both methods did not differ significantly (paired t test: p = 0.0262, t = 1.127, n = 109).

Dissolved organic phosphorus (DOP)
For the determination of DOP, duplicate 40 mL subsamples were filtered through precombusted (6 h, 450 • C) glass fiber filters (Whatman GF/F) and stored in 50 mL vials (Falcon) at 20 • C until further processing. The thawed samples were oxidized in a microwave (MARSX press, CEM, Matthews, USA) after the addition of potassium peroxy- 25 disulfate in an alkaline medium (Grasshoff et al., 1983). The P concentration, measured as PO 4 in a 10 cm cuvette, represents the total dissolved phosphorus (DP) concentration. DOP was calculated as the difference between the DP concentrations in the filtered and digested samples and the corresponding PO 4 concentration analyzed as described above.

5
For all analyzed components, subsamples were pre-filtered through pre-combusted (6 h, 450 • C) filters (Whatman GF/F) followed by filtration through 0.2 µm cellulose acetate filters. Subsamples were prepared for storage according to the specific method used for each compound. After the analyses, the phosphorus content of measured DOP compounds was summed and the amount subtracted from the total DOP concen-10 tration. The difference is defined as the uncharacterized DOP.

Dissolved ATP
The method of Bjorkman and Karl (2001), adapted to Baltic Sea conditions , was used to determine dissolved adenosine triphosphate (dATP). A Mg(OH) 2 precipitate, including the co-precipitated nucleotides, was obtained by treat- 15 ing 200 mL of the filtrate with 2 mL of 1 M NaOH (0.5 % v/v). The precipitate was allowed to settle overnight and then centrifuged at 1000 g for 15 min. The supernatant was discarded and the precipitate was transferred into 50 mL Falcon tubes, centrifuged again (1.5 h, 1680 g). The resulting pellet was dissolved by drop-wise addition of 5 M HCl. The samples were frozen at −20 • C until further processing. The pH of the 20 thawed samples was adjusted to 7.2 by the addition of TRIS buffer (pH 7.4, 20 mM). The final volume was recorded. The dATP concentrations were measured in triplicate using the firefly bioluminescence assay and a Sirius luminometer (Berthold Detection Systems Pforzheim, Germany), as described by Unger et al. (2013). Standard concentrations were prepared as described above, using aged Baltic Sea water and six BGD 12,2015 Effects of CO 2 perturbation on phosphorus pool sizes M. Nausch et al. Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | A2383) ranging from 1 to 20 nmol L −1 . The detection limit of the bioluminescence assay was 2.5 pmol mL −1 . The fluorescence slope of the standard concentrations was used to calculate dATP concentrations, correcting for the final sample volume. The P-content of the ATP (ATP-P) was calculated by assuming that 1 mol of ATP is equivalent to 3 mol P.

Dissolved phospholipids
The phosphate content of the dissolved phospholipids (PL-P) was analyzed using a modified method of Ingall (2001, 2004). Briefly, 400 mL subsamples of the filtrate were stored at −20 • C until further processing. The samples were then thawed in a water bath at 30 • C and extracted twice with 100 mL of chloroform. The 10 chloroform phase was collected, concentrated to 5 mL in a rotary evaporator (Heidolph Instruments, Schwabach, Germany), and then transferred into microwave tubes. The chloroform was completely evaporated by incubating the tubes in a 60 • C water bath overnight. After the addition of 20 mL of deionized water (Milli-Q, Millipore), the samples were digested with potassium peroxydisulfate in alkaline medium and microwaved 15 as described for the DOP analysis. Six standard concentrations of phospholipids, ranging from 0 to 125 µg L −1 , were prepared by adding the respective amounts of a stock solution containing 5 mg of l-phosphatidyl-dl-glycerol sodium salt (PG, Sigma Aldrich, P8318) mL −1 to the aged seawater. The detection limit was 0.8 nmol L −1 . The blanks contained only chloroform and were processed as for the samples.

Dissolved DNA and RNA
Dissolved DNA and RNA (dDNA and dRNA) concentrations were determined according to Karl and Bailiff (1989) and as described by Unger et al. (2013). For each sample, 200 mL of the filtrate was gently mixed with the same volume of ethylene-diaminetetracetic acid (EDTA, 0.1 M, pH 9.3, Merck, 1.08454) and 4 mL of cetyltrimethylammonium bromide (CTAB, Sigma-Aldrich, H5882) and stored frozen at −20 • C for BGD 12,2015 Effects of CO 2 perturbation on phosphorus pool sizes M. Nausch et al. at least 24 h. After thawing the samples, the precipitate was collected onto combusted (450 • C, 6 h) glass fiber filters (25 mm, GF/F Whatman), placed into annealed vials, and stored frozen at −80 • C until further analysis.
Coupled standards (DNA + RNA) containing 1-10 µg DNA (Sigma Aldrich, D3779) L −1 and 20-120 µg RNA (Sigma Aldrich, R1753) L −1 were prepared in aged seawater as described above. A reagent blank served as the reference and aged seawater as the background control. The P-contents of the DNA and RNA were calculated by multiplying the measured values by a factor of 2.06 nmol P per µg dDNA and 2.55 nmol P per µg dRNA. The latter values were determined by the microwave digestion of standard substrates.

Particulate organic phosphorus, carbon, and nitrogen
Particulate organic phosphorus (POP) was analyzed using two methods in parallel. In 15 the "aqueous method", 40 mL of unfiltered subsamples were frozen at −20 • C and analyzed as described for DOP. The measured PO 4 concentration represents total phosphorus (TP). PP is the difference between the total PO 4 concentration in the unfiltered digested sample and the sum of DOP + PO 4 . In the "filter-method", 500 mL sub precombusted GF/F-filters that were then placed into Schott bottles containing 40 mL of 20 deionised water. PP was digested to PO 4 by the addition of oxidizing decomposition reagent (Oxisolv ® , Merck) followed by heating in a pressure cooker for 30 min. The PO 4 concentrations of the cooled samples were determined spectrophotometrically according to Grasshoff et al. (1983). Paired t test revealed significant differences between both methods; however, both means of 0.19 ± 0.03 µmol L −1 for the filter method and 25 of 0.16 ± 0.04 µmol L −1 for the aqueous method differed near the detection limit of the methods. Thus, solely the mean values obtained from both measurements are used in the following. a blank were prepared. The blank was obtained by the addition of formaldehyde (1 % final concentration) 10 min before radiotracer addition, in order to poison the samples. At defined time intervals within the incubation, 5 mL subsamples were taken from each of the parallel samples and filtered onto polycarbonate filters pre-soaked with a cold 20 mM PO 4 solution to prevent non-specific [ 33 P]PO 4 binding. The filters were rinsed 15 with 5 times 1 mL of particle-free bay water and placed in 6 mL scintillation vials. Scintillation liquid (4 mL IrgaSafe; Perkin Elmer) was added and the contents of the vials were mixed using a vortex mixer. After allowing the samples to stand for at least 2 h, the radioactivity on the filters was counted in a Perkin Elmer scintillation counter. Filters of 0.2 and 3 µm pore sizes (Whatman and Millipore, respectively) were used to determine 20 uptake by the whole plankton community and the size fraction > 3 µm, respectively. Picoplankton uptake was calculated as the difference between the activity on the 0.2 µm and 3 µm filters.
[γ 33 P]ATP (specific activity of 111 TBq mmol −1 , Hartmann Analytic GmbH) was added to triplicate 10 mL samples and a blank, each in a 20 mL vial, at a concen- the samples, which were then filtered and processed as described for the PO 4 uptake measurements.

Bacterial production (BPP)
Rates of bacterial protein production (BPP) were determined by incorporation of 14 [C]leucine ( 14 C-Leu, Simon and Azam, 1989) according to Grossart et al. (2006). Trip-5 licates and a formalin-killed control were incubated with 14 C-Leu (213 mCi mmol −1 ; Hartmann Analytic GmbH, Germany) at a final concentration of 165 nmol L −1 , which ensured saturation of uptake systems of both free and particle-associated bacteria. Incubation was performed in the dark at in situ temperature (between 7.8 • C and 15.8 • C) for 1.5 h. After fixation with 2 % formalin, samples were filtered onto 5.0 µm (attached 10 bacteria) nitrocellulose filters (Sartorius, Germany) and extracted with ice-cold 5 % trichloroacetic acid (TCA) for 5 min. Thereafter, filters were rinsed twice with ice-cold 5 % TCA, once with ethanol (96 % v/v), and dissolved with ethylacetate for measurement by liquid scintillation counting. Afterwards the collected filtrate was filtered on 0.2 µm (free-living bacteria) nitrocellulose filters (Sartorius, Germany) and processed 15 in the same way as the 5.0 µm filters. Standard deviation of triplicate measurements was usually < 15 %. The sum of both fractions (free-living bacteria and attached bacteria) is referred to total BPP. The amount of incorporated 14 C-Leu was converted into BPP by using an intracellular isotope dilution factor of 2. A conversion factor of 0.86 was used to convert the protein produced into carbon (Simon and Azam, 1989). 20

Statistical analyses
The Grubbs test, done online (graphpad.com/quickcalcs/Grubbs1.cfm) was applied to identify outliers in all data sets. The outliers were removed from further statistical analyses.
BGD 12,2015 Effects of CO 2 perturbation on phosphorus pool sizes M. Nausch et al. Spearman Rank correlations were carried out to describe the relationship between the development of the parameters over time in the mesocosms and in the fjord using Statistica 6 software.
Short-term CO 2 effects on POP concentrations at days 0-2 and 23-43 between the CO 2 treatments were verified with an ANCOVA analysis using the SPSS software. The 5 "days" were treated as a covariate interacting with the treatments. Paired t test was applied to check the differences in PO 4 concentrations between the treatments.

Development in the mesocosms
3.1.1 CO 2 , pH, temperature and salinity 10 The different mesocosms were characterized based on their averaged f CO 2 and pH values from day 1 until day 43 ( Fig. 2a and

Phytoplankton biomass
Chlorophyll a (Chl a) reached maximum concentrations of 2.06-2.48 µg L −1 at day 5 ( Fig. 4). Average concentrations of 1.94 ± 0.23 µg L −1 in phase I exceeded those in phases II and III when Chl a decreased to a mean of 1.08 ± 0.16 µg L −1 . The increase in Chl a in the high CO 2 mesocosms by 24 % in phase III was statistically significant 5 (Paul et al., 2015b), and differences of 0.27 µg L −1 were only marginal.
We observed a significant relationship between Chl a and PO 4 in the untreated and intermediate treated mesocosms that was diminished with increasing f CO 2 indicated by lower p values. The significance got lost in the highest f CO 2 mesocosms (Table 2). 10 Total phosphorus (TP) concentrations in the mesocosms ranged between 0.49 and 0.68 µmol L −1 (Fig. 5a) without significant differences between the different CO 2 treatments. Shortly after the bags were closed, the decline in TP concentrations began and continued until the beginning of phase II. On average, TP concentrations decreased from 0.63 ± 0.02 µmol L −1 on day −3 to 0.51 ± 0.01 µmol L −1 on day 21. Thereafter, the 15 mean TP remained constant at 0.54±0.03 µmol L −1 until the end of the measurements.

Phosphorus Pools
Thus, the loss of phosphorus (116 nmol L −1 ) from the 17 m layer during the 29 day measurement period was calculated to be 4.0 nmol L −1 day −1 . The decline in TP can be explained by loss through sedimentation of POP (Paul et al., 2015b). Particulate organic phosphorus (POP) concentrations varied from 0.10 to 20 0.23 µmol L −1 in all CO 2 treatments (Figs. 5b and 6). We expected that the decrease in TP was reflected in POP. However, parallel changes occurred only periodically. POP concentrations increased during the first 5 days after the bags were closed. This increase was stimulated by the CO 2 treatments from day 0 to day 2 (ANCOVA: p = 0.004, F = 20.811) (Fig. 7a) (Table 5). Figures 3 and 6b show that the increase in Chl a was delayed by 2-3 days compared to the increase in POP during the first growth event. A correlation between POP and Chl a was detected only for the untreated mesocosms (Table 2). Dissolved organic phosphorus (DOP) concentrations in the mesocosms ranged be-10 tween 0.18 and 0.36 µmol L −1 constituting 32-71 % of the TP pool (Fig. 5). DOP did not change significantly in response to the CO 2 perturbations, and were similar to the concentrations in fjord water. Concentrations ≥ 0.3 µmol L −1 were measured on days 6 and 7 (phase I) and on day 23 (phase II); the high DOP value in the intermediate CO 2 treatment at day 19 was an outlier (Grubbs test) (Fig. 5c).

15
In phase I, DOP initially increased in parallel with Chl a and BPP but reached its maximum 1-2 days later, after which it decreased only marginally until the end of this phase, independent of changes in BPP and Chl a (Fig. 8c,d). In phase II, the peak conformed to that of BPP. DOP correlated with temperature only in the high f CO 2 mesocosms (Table 2). In addition, the composition of DOP did not change with increasing CO 2 . The 20 sum of RNA (∼ 47 %) plus the unidentified fraction constituted 98-99 % of the DOP pool whereas the other measured compounds delivered only 1-2 % (Table 3).
Phosphate (PO 4 ) concentrations ranged between 0.06 and 0.21 µmol L −1 , with variations occurring only in the nanomolar range. The mean contribution of PO 4 to TP was 25 ± 6 %, which was the lowest among all P fractions (Fig. 6). From the start of 25 the measurements to day 13, PO 4 declined by 0.06 µmol L −1 (or 3.5 nmol L −1 day −1 ) from initial values of 0.16 ± 0.01 µmol L −1 (Fig. 5d). Subsequently, concentrations increased again, by an average of 2.6 nmol L −1 day −1 , until the end of the experiment.
There were no significant differences between CO 2 treatments until day 23, when high CO 2 concentrations led to slightly lower PO 4 concentrations (Fig. 5d). Afterwards, PO 4 concentrations in the high f CO 2 mesocosms were significantly lower than those in the untreated mesocosms (t = 6.51, p = 0.0003). This observation is in accordance with the dynamics of POP and Chl a concentrations, which were significantly elevated in the high CO 2 treatments. Thus, the transformation of PO 4 to POP via stimulated biomass 5 formation may have been promoted under high CO 2 conditions in phase III.
Since PO 4 was never fully exhausted, phosphorus limitation of phyto-and bacterioplankton can be excluded. This interpretation is supported by the POC : POP ratios, which varied between 84.4 and 161.1 in all treatments (Paul et al., 2015b). 10 PO 4 turnover times of 1.5-8.4 days (mean 4.0 ± 1.2 days, n = 112) in all mesocosms indicated no dependency on the CO 2 treatment (Fig. 9a). Gross PO 4 uptake rates were in the range of 0.6-3.9 nmol L Table 4). The rates were highest on days 4 and 9 (phase I) and decreased thereafter until day 15,15 followed by an increase to a mean maximum rate of 2.3 ± 0.5 nmol L −1 h −1 (n = 6) at day 27. The size fraction < 3 µm was responsible for 59.1 to 98.4 % of the total PO 4 uptake (mean 86.5 ± 7.6 %) whereas the size fraction > 3 µm accounted for only 1.6-40.9 % (mean 13.5±7.4 %). Thus, PO 4 was taken up mainly by picoplankton. However, only the uptake rate by the size fraction > 3 µm was positively related to Chl a and in-20 versely related to the P content of the biomass (Table 2). Thus the PO 4 uptake was obviously stimulated when the phytoplankton biomass increased and at simultaneous decrease of the cellular P. The relationship between PO 4 uptake by this fraction and Chl a became evident only in the CO 2 -amended conditions indicating that the interaction between P uptake, cellular P-content and growth of phytoplankton was stimulated 25 under elevated CO 2 conditions. ATP turnover times of 0.2 to 3.6 days (mean 0.94 ± 0.74 days, n = 90) were much shorter than the PO 4 uptake rates and did not vary between the treatments. Turnover BGD 12,2015 Effects of CO 2 perturbation on phosphorus pool sizes M. Nausch et al. times were longest on day 9 and shortest on day 4 (Fig. 9c). Between 0.05 and 0.36 nmol ATP L −1 h −1 (mean 0.14 ± 0.08 nmol L −1 h −1 , n = 36) were degraded, corresponding to a P supply of 0.14 and 1.08 nmol L −1 h −1 (mean 0.44 ± 0.25 nmol L −1 h −1 , n = 36). Thus, phosphorus additionally supplied from ATP accounted for ∼ 25 % of that provided by PO 4 . The picoplankton size fraction (< 3 µm) was responsible for 90-99 % 5 of ATP uptake, with only a marginal portion (1.6-9.5 %) attributable to the phytoplankton fraction > 3 µm (Table 4).

In situ CO 2 and pH conditions
Large variations in f CO 2 and pH occurred in fjord water during the period of investigation (Table 1). The relationship of f CO 2 with temperature and salinity indicated that the CO 2 conditions were influenced predominantly by changes in the water masses, specifically by upwelling which affected both the relationship of f CO 2 with PO 4 and probably the correlation of f CO 2 with Chl a and PC (Table 2). f CO 2 ranged from 207 µatm (Fig. 2a) at days 12-16 when temperatures were highest to 800 µatm at day 33 when 15 deep water input occurred which was indicated by low pH (7.75).

Phytoplankton biomass
Chl a concentrations in the fjord were between 1.12 and 5.46,µg L −1 (mean 2.29 ± 1.11 µg L −1 ; n = 38), with distinct phases similar to those of temperature and salinity. However, the Chl a maximum occurred at the beginning of phase II, which was 1-2 20 days after the maximum temperature. Shortly thereafter, Chl a decreased to its lowest level before it increased again, albeit only marginally to 1.93 µg L −1 during phase III (Fig. 4). Chl a concentrations correlated positively with temperature and pH (

Phosphorus-Pools
TP concentrations from day −3 until day 29 ranged between 0.54 and 0.70 µmol L −1 (mean 0.61 ± 0.04 µmol L −1 ; n = 19) (Figs. 5a and 6a). The progression of TP differed from that of the hydrographic parameters or the Chl a concentrations. With a general decreasing tendency, TP undulated with a frequency of about 10 days in the period of 5 phases 0 to the first half of phase I and of 6 days in the second half of phase I to II. For the period under investigation, the TP fractions had the following characteristics: POP concentrations varied from 0.13 to 0.30 µmol L −1 (mean 0.20 ± 0.04 µmol L −1 ; n = 29), thus accounting for 23.4-51.8 % (mean 34.7 ± 7.9 %; n = 19) of the TP pool. The development of POP over time did not follow that of TP (Fig. 6b). POP concen-10 trations were highest between days 8 and 19, when the accumulation of POP in the biomass was reflected in declining C : P ratios from 180 to 107 and thereafter remained at the low ratio until the end of the measurements. The POP increase in phase III occurred in parallel to Chl a and to the PO 4 decrease (Table 5). Thus PO 4 was transformed into POP via biomass production. The calculated P content of phytoplankton DOP substantially contributed (26-45 %) to the TP pool (Fig. 6). Concentrations ranged between 0.19 and 0.29 µmol L −1 (mean 0.24 ± 0.03 µmol L −1 ; n = 17), with high concentrations occurring in parallel to those of TP in phases I and II (Fig. 5c). The very low DOP value of 0.11 µmol L −1 , on day 29, was an outlier (Grubbs test) and was ex-20 cluded from the calculation. For the whole study period, DOP concentrations correlated positively with both POP and PO 4 turnover times and inversely with PO 4 concentrations ( Table 5). A similar behavior between DOP and Chl a was restricted to phases 0 and I, whereas the relationship was inverse in phase II (Fig. 8b)  was not feasible because DOP and BPP were not always sampled on the same day and only very few data pairs were available. Phosphorus, derived from the sum of ATP, PL, RNA, and DNA, constituted 42.8-72.0 % (mean 59.7 ± 10.7 %; n = 7) of the DOP pool (Table 3) The changes in all of these components over time were not related to changes in the total DOP pool (Fig. 10). PO 4 concentrations ranged between 0.06 and 0.41 µmol L −1 (mean 0.21 ± 0.09 µmol L −1 , n = 21), thus comprising 24.3 ± 11.2 % (n = 21) of the TP pool (Fig. 6). With a few exceptions, PO 4 concentrations declined from the beginning of the study period until the end of phase I and increased during phase II and the beginning of 15 phase III. These changes were caused by upwelling of PO 4 enriched deep water of higher salinity and lower temperatures. The subsequent decline in PO 4 between days 33 and 40 was caused by the stimulation of phytoplankton production, as indicated by the increase in Chl a concentration (Fig. 4). For the whole experimental period, the Spearman rank correlation showed an inverse relationship between PO 4 and particu-20 late organic matter such as Chl a, PC, and PN (Table 5).

Uptake of PO 4 and ATP
Applying [ 33 P]PO 4 , PO 4 turnover times in the fjord were in the range of 30-379 h (1.2-15 days) (mean 139 ± 98 h, n = 18) (Fig. 9a), corresponding to uptake rates of 0.73-3.37nmol L −1 h −1 (mean 1.64 ± 0.82 nmol L −1 h −1 , n = 18) (Table 4). These rates were influenced by multiple factors, including temperature, phytoplankton biomass, and DP, POP, and PO 4 concentrations as deduced from data pairs, the total PO 4 uptake rate correlated with total BPP and with BPP in the fraction < 5 µm (r = 0.886; p = 0.0188; n = 6 for each relationship). Within the experimental period, the turnover times shortened on days 15-17 (Fig. 9a), when temperature and Chl a (Figs. 3 and 4) reached a maximum and PO 4 concentrations were lowest (Figs. 5d). Although the shortest turnover times were ex-5 pected to be coupled with the highest uptake rates, the latter were estimated 2 days later, between days 17 and 19. The day-to-day variations conformed to the small changes in temperature and PO 4 concentrations. Uptake was dominated by the size fraction < 3 µm in most of the measurements (Table 4), which accounted for 17.4-92.3 % (mean 72.2 ± 20.6 %) of the total uptake rate. The mean contribution of the size fractions > 3 µm was 27.8 ± 20.6 %. Assuming that autotrophic organisms were largely responsible for the uptake by this fraction, the specific PO 4 uptake rates of phytoplankton, calculated from the size fraction > 3 µm and the Chl a concentration, were 0.02 and 0.46 nmol (µg Chl a) −1 h −1 .

Discussion
An increase in CO 2 in marine waters and the associated acidification may potentially have multiple effects on organisms and biogeochemical element cycling (Gattuso and Hansson, 2011). However, reported findings indicate wide ranging responses, probably depending on the investigated species and growth conditions. For example, CO 2 5 stimulation as well as lack of stimulation were found for primary production and carbon fixation (Beardall et al., 2009;Boettjer et al., 2014), DOC release (Engel et al., 2014;MacGilchrist et al., 2014) and phytoplankton growth (Riebesell and Tortell, 2011). Thus, the responses of organisms and ecosystems to enhanced CO 2 concentrations are complex and still poorly understood. The present study is the first to determine the 10 effects of increased CO 2 levels on the phosphorus cycle in a brackish water ecosystem.

Response of P-pools and P-uptake to enhanced CO 2 in the mesocosms
The Finish side of the Gulf of Finland is one of the most important upwelling regions in the Baltic Sea. During our investigation in 2012, surface temperatures, obtained from the NOAA satellite (Siegel and Gerth, 2013) showed that upwelling persisted during 15 the whole study period but with varying intensity. The intensity of upwelling shaped the pattern of temperature in the fjord and in the mesocosms varying from 7.8 to 15.9 • C. Such variations in temperature influence the phosphorus transformation and interleave with CO 2 effects. While nutrients were added in previous mesocosms experiments (Riebesell et al.,20 2008; , no amendments were undertaken in this study in order to be close to natural conditions. Initial PO 4 concentrations of only 0.17 ± 0.01 µmol L −1 were measured, however, PO 4 was never exhausted (Figs. 5 and 6). Cellular C : P and N : P ratios were close to the Redfield ratio. Therefore, phosphorus limitation unlikely occurred in this experiment. Simultaneous low nitrate and ammonium concentrations (Paul et al., 2015b) formed nutrient conditions that benefit the growth of diazotrophic cyanobacteria. However, any cyanobacteria bloom failed to appear, despite the low- of Chl a in phase I and rising to ∼ 85 % in phase III (Paul et al., 2015b). However, a positive correlation of f CO 2 with the Chl a size fraction > 20 µm was estimated. The abundance of diatoms that could be a part of this fraction increased from ∼ day 23 to day 30 and might have an influence on this relationship. Against this background, the CO 2 perturbation did not cause significant changes 10 in phosphorus pool sizes, DOP composition, and P-uptake rates from PO 4 and ATP when the whole study period was considered. However, small but nevertheless significant, short-term effects on PO 4 and POP pool sizes were observed in phases I and III (Fig. 7). CO 2 elevation stimulated the formation of POP until day 3 (Fig. 5b) when chlorophytes, cyanobacteria, prasinophytes and the pico-cyanobacteria started 15 to grow (Paul et al., 2015b). The effects of CO 2 addition on PO 4 and POP pool sizes were evident from day 23 onwards (Figs. 5 and 7). PO 4 concentrations were slightly, but significantly lower in the high CO 2 treatment than in the untreated mesocosms, accompanied by significantly elevated POP concentrations indicating that the transformation of PO 4 into POP was 20 likely stimulated under high CO 2 conditions. Since Chl a was elevated as well at similar POP : Chl a ratios, the PO 4 taken up was used for new biomass formation. However, the elevated transformation of PO 4 into POP was not detected in the PO 4 uptake rates which can be seen as gross uptake rates. Thus, it is likely that not the gross uptake but rather the net uptake was modified, e.g. via reduction in P-release from biomass under 25 CO 2 elevation.
It is hard to assess the short-term effects that we have found in phase I. Uptake and release are assumed to be continuous processes and can alter the P pool sizes on timescales shorter than one day. Thus, variations and differences in the treatments BGD 12,2015 Effects of CO 2 perturbation on phosphorus pool sizes M. Nausch et al. can be overseen at daily sampling. Unger et al. (2013) demonstrated that an accelerated PO 4 uptake by the cyanobacterium Nodularia spumigena under elevated CO 2 incubations could only be observed during the first hours. Thereafter, the differences were balanced and the same level of radiotracer labeling was reached in all treatments. An acceleration in formation of particulate P concentrations under CO 2 elevation with-5 out any changes of PO 4 turnover times was also observed by Tanaka et al. (2008). They observed an increase of the POP amount and an earlier appearance of the POP maximum under CO 2 elevation. Correlations calculated by using the Spearman rank test between P pools or uptake rates and other parameters for each mesocosm are presented in Table 2. The 10 relationships between POP and TP with Chl a disappeared at elevated f CO 2 , whereas correlations developed between POP and PC as well as between the PO 4 uptake by phytoplankton > 3 µm and the POP : Chl a ratio (Table 2). These shifts could be caused by changes in the phytoplankton composition deduced from CO 2 effects on the pigment composition (Paul et al., 2015b). 15 Independent of the CO 2 treatment, TP decreased by 2.6 nmol L −1 day −1 in all mesocosms over the course of the experiment, in agreement with the measured sedimentation rates (Paul et al., 2015b). The strongest decrease (∼ 3.2 nmol L −1 day −1 ) occurred during phase I. Of the total TP removal during this phase (48 nmol L −1 ), 84 % (∼ 40.5 nmol L −1 ) could be explained by the decrease in POP and 16 % (∼ 8 nmol L −1 ) 20 by changes in the dissolved pool. However, the PO 4 decline (∼ 34.5 nmol L −1 ) was stronger than that of the total dissolved P pool since DOP increased in parallel by ∼ 26.5 nmol L −1 . Thus, about 77 % of the PO 4 reduction was retrieved as DOP and remained in the dissolved P-pool being the main pathway of PO 4 transformation.

25
Measurements of P-pool sizes and P uptake in the fjord provided new information about the phosphorus dynamics in a Baltic Sea upwelling system and in times when diazotrophic cyanobacteria did not dominate the phytoplankton community. conditions were mainly determined by upwelled waters, which were depleted in dissolved inorganic nitrogen and enriched in PO 4 , as reported for other upwelling areas of the Baltic Sea (Lass et al., 2010). Thus, ammonium and NO 2 + NO 3 concentrations in the surface water were only in the nanomolar range (Paul et al., 2015b). PO 4 increased in parallel with the increase in salinity and decrease in temperature, indicating 5 their coupling with upwelling (Table 5). Maximum PO 4 concentrations of 0.33 µmol L −1 and 0.42 µmol L −1 (Figs. 5 and 6) were observed at the end of the upwelling events in phases 0 and II, respectively. The correlation with Chl a and POP indicated that PO 4 was utilized during plankton growth in the subsequent relaxation phases I and III. However, PO 4 was not fully depleted (Fig. 6d) which can be attributed to "low" Pdemand of organisms as deduced from the relatively low PO 4 uptake rates in fjord water and in the mesocosms. As in the mesocosms, the phytoplankton community was unlikely P-limited indicated by PC : POP ratios of 86-189 (mean 125, n = 23) (Paul et al., 2015b). The close correlation of POP with Chl a indicated a large contribution of phytoplankton to particulate P. However, its P content deduced from POP : Chl a ratios 15 of 0.05-0.15 µmol P (µg Chl a) −1 was somewhat lower than those observed during an upwelling event along the east coast of Gotland, where the ratios were between 0.1 and 0.2 µmol P (µg Chl a) −1 (Nausch et al., 2009). POP concentrations of 0.13-0.3 µmol L −1 were in the range typically observed in the Baltic Proper (Nausch et al., 2009;Nausch et al., 2012). However, POP concentrations 20 in the Gulf of Finland may reach higher values, as was the case in the summer of 2008, when the observed POP concentration was 0.35 ± 0.07 µmol L −1 (Nausch and Nausch, 2011). DOP exhibits vertical gradients with maximum concentrations in the euphotic surface layer (Nausch and Nausch, 2011) and lower than 0.1 µmol L −1 at depths below 25 m. 25 Thus, the observed DOP dynamics in surface water can be assumed to be the result of release, consumption and mineralization by organisms. The DOP increase in phase I coincided with increases in Chl a and initially BPP (Fig. 8), while the development of DOP and BPP showed opposing trends in the second part of phase I and thereafter.
BGD 12,2015 Effects of CO 2 perturbation on phosphorus pool sizes M. Nausch et al. Thus, the increased DOP concentrations in phase I were due to release by phytoplankton supplemented by bacterial release exceeding the consumption or degradation. During phase II, phytoplankton biomass was low and DOP release should thus be minor. Since the small mesozooplankton increased in the fjord similar to those reported for the mesocosms in phases II and III (Paul et al., 2015b) DOP could be released during 5 grazing combined with the observed temporal offset of BPP and DOP maxima. The DOP concentration of 0.27 ± 0.02 µmol L −1 during our study was similar to that detected in the Gulf of Finland in the summer of 2008 (Nausch and Nausch, 2011). On average, more than half (59.1 %) of the DOP consisted of the measured compounds ATP, PL, DNA, and RNA; the other sources remained uncharacterized. ATP levels in 10 the Storfjärden were approximately ten times higher than in the surface water of the subtropical Pacific (Bjorkman and Karl, 2001) but were similar to those measured during a spring bloom in the Antarctic (Nawrocki and Karl, 1989).
The contribution of PL to the DOP pool was in the same range as reported by Suzumura (2005) whereas the contribution of DNA was relatively small. DNA concentrations 15 were much lower than either those measured in the northern Baltic Sea (Riemann et al., 2009) or those reported by Karl and Bailiff (1989) for the Pacific Ocean. RNA, however, was the dominant DOP component, contributing about half of the DOP pool in this study. In studies of Karl and Bailiff (1989) in various marine systems RNA concentrations were 6-10 times higher than DNA concentrations. However, they have measured 20 such high RNA concentrations as detected in our study only in a pond.
The uptake of phosphorus from radioactively labeled ATP is used to monitor DOP utilization (Karl and Björkman, 2002). However, ATP is a component of the labile P fraction and is thus preferred over other substrates (Siuda and Chrost, 2001). The mean ATP turnover times of 23 ± 14 h were similar to those measured in the Gotland 25 Basin in May and June 2001 (Nausch et al., 2004), when temperatures were below 12 • C. During mesocosm experiments at the Tvärminne station in July 2003 (Lovdal et al., 2007), ATP turnover times at temperatures > 18 • C were between 2 and 6 h. In our study, 0.04-0.51 nmol ATP-P L −1 h −1 were taken up mainly (85-98 %) by pico- for only 3-20 % of the total ATP uptake. In our study, ATP uptake rates correlated with the contribution of the fraction < 2 µm to total Chl a whereas no such correlation could be established for BPP. These observations might be an indication that autotrophic picoplankton dominates the ATP uptake. PO 4 turnover times varied between 1 and 5 days. Longer turnover times (10- 10 15 days) occurred only in phase 0 while the shortest turnover time (1 day) was at the end of phase I, when temperatures and phytoplankton biomass were highest and PO 4 concentrations lowest. Nevertheless, a 1 day turnover time indicated no Plimitation, as under P-limited conditions reported turnover times are < 1 h (Nausch et al., 2004). The PO 4 uptake rates of 0.9-2.8 nmol L −1 h −1 and the specific uptake  (Nausch et al., 2009;2012). As in the summer of 2001, two-thirds of the PO 4 uptake was realized by the size fraction < 3 µm, although with progressive PO 4 depletion this percentage may rise to ∼ 90 % (Nausch et al., 2004). This result suggests that picoplankton has an advantage in the 20 competition for phosphorus at low concentrations. The close correlation of total PO 4 uptake and BPP < 5 µm (Table 2) suggests that free-living heterotrophic bacteria were the main consumers of PO 4 . This relationship was probably determined by the 3 to 5 µm size fraction, since there was no apparent correlation between BPP < 5 µm and PO 4 uptake by the size fraction < 3 µm (Table 2). 25 The phosphorus demand of heterotrophic bacteria is influenced by carbon and nitrogen availabilities. The shorter turnover times of ATP, DNA, and PO 4 following nitrogen and carbon amendments in the mesocosm experiments of Lovdal et al. (2007) in July 2003 suggested that the bacterioplankton community at Tvärminne station is C-and BGD 12,2015 Effects of CO 2 perturbation on phosphorus pool sizes M. Nausch et al. N-limited. This may also have been the case during our study. Thus, the relatively long PO 4 turnover times might have been caused not only by low temperatures but also by the reduction in bacterioplankton activities due to C and N limitations and could be the reason that PO 4 was not depleted completely in the mesocosms and in fjord water.

5
Surface water in Storfjärden showed highly variable f CO 2 conditions and reached levels up to 800 µatm, which is similar to that expected in ca. 100 years from now. Deduced from the high frequency of upwelling events there, organisms are confronted with elevated f CO 2 more or less regularly and are used to high f CO 2 variability. This could explain the minimal response of the phytoplankton community. A general impact of f CO 2 on P pools and P uptake rates could not be identified for the overall period of investigation. However, temporary responses to f CO 2 elevation were observed for the transformation of PO 4 into POP. Although statistically significant, it is difficult to assess if the differences between the treatments are of ecological relevance. Potentially, such short-term variations are possible in the phosphorus dynamics since the trans- 15 formation can take place on hourly scales and transformations are in the nanomolar concentration range. There are also indications that relationships of P pool sizes or uptake with Chl a and PC can change as f CO 2 increases. This would have an effect on biogeochemical cycles. This study also provides information on the phosphorus cycle in an upwelling-driven ecosystem of the Baltic Sea. P pool sizes were in the range 20 characteristic for spring and cooler summers when low temperatures inhibit cyanobacteria bloom formation. The transformation of PO 4 into DOP may be the major pathway of phosphorus cycling under hydrographical and phytoplankton growth conditions as occurred in our experiment.
BGD 12,2015 Effects of CO 2 perturbation on phosphorus pool sizes M. Nausch et al. Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | measurements. We appreciate the assistance of Jehane Ouriqua in the nutrient analysis and that of many other participants who carried out the samplings. We also appreciate the collegial atmosphere during the work and thank everyone who contributed to it. We would also like to acknowledge the staff of the Tvärminne Zoological Station for their hospitality and support, for allowing us to use the experimental facilities, and for providing CTD data for the summers of 5 2008-2011. Finally, we thank Jana Woelk for analysing the phosphorus samples in the IOW. This study was funded by the BMBF project BIOACID II (FKZ 03F06550). 12,2015 Effects of CO 2 perturbation on phosphorus pool sizes M. Nausch et al.  Hiebenthal, C., Philipp, E. E. R., Eisenhauer, A., and Wahl, M.: Effects of seawater pCO 2 and temperature on shell growth, shell stability, condition and cellular stress of Western Baltic Sea Mytilus edulis (L.) and Arctica islandica (L.), Mar. Biol., 160, 2073-2087. IPCC, 2001 Figure 2. (a) f CO 2 values in the mesocosms and in the fjord throughout the experiment. Small black dots show the f CO 2 in the ambient fjord water. Treatment of the mesocosms with CO 2 saturated fjord water at the beginning of the experiment (days 0-4) created different f CO 2 levels in the mesocosms: blue symbols represents the untreated mesocosms, grey the intermediate, and red the high CO 2 treated mesocosms. The treatment was repeated at day 16. (b) Corresponding pH ranges in the mesocosms during the four phases. Despite decreasing trend over time, a gradient between the mesocosms was kept over the whole period.

BGD BGD
BGD 12,2015 Effects of CO 2 perturbation on phosphorus pool sizes M. Nausch et al.