Effects of ocean acidification on primary production in a coastal North Sea phytoplankton community

We studied the effect of ocean acidification (OA) on a coastal North Sea plankton community in a long-term mesocosm CO2-enrichment experiment (BIOACID II long-term mesocosm study). From March to July 2013, 10 mesocosms of 19 m length with a volume of 47.5 to 55.9 m3 were deployed in the Gullmar Fjord, Sweden. CO2 concentrations were enriched in five mesocosms to reach average CO2 partial pressures (pCO2) of 760 μatm. The remaining five mesocosms were used as control at ambient pCO2 of 380 μatm. Our paper is part of a PLOS collection on this long-term mesocosm experiment. Here, we here tested the effect of OA on total primary production (PPT) by performing 14C-based bottle incubations for 24 h. Furthermore, photoacclimation was assessed by conducting 14C-based photosynthesis-irradiance response (P/I) curves. Changes in chlorophyll a concentrations over time were reflected in the development of PPT, and showed higher phytoplankton biomass build-up under OA. We observed two subsequent phytoplankton blooms in all mesocosms, with peaks in PPT around day 33 and day 56. OA had no significant effect on PPT, except for a marginal increase during the second phytoplankton bloom when inorganic nutrients were already depleted. Maximum light use efficiencies and light saturation indices calculated from the P/I curves changed simultaneously in all mesocosms, and suggest that OA did not alter phytoplankton photoacclimation. Despite large variability in time-integrated productivity estimates among replicates, our overall results indicate that coastal phytoplankton communities can be affected by OA at certain times of the seasonal succession with potential consequences for ecosystem functioning.


Introduction
Atmospheric CO 2 partial pressure (pCO 2 ) is currently rising at an unprecedented rate due to anthropogenic activities. This leads to enhanced CO 2 uptake by the oceans and a decrease in a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 ocean surface water pH, referred to as ocean acidification (OA) [1,2]. From 1765 until 1994, pH values were calculated to have already decreased by 0.08 units. Present-day CO 2 concentrations of around 400 μatm are predicted to more than double by the year 2100, which will result in a further acidification of the ocean [3]. After the Polar Oceans, the North Atlantic is expected to show strongest changes in response to rising pCO 2 [3,4]. As a major sink of anthropogenic CO 2 , the North Atlantic Ocean basin stores almost a quarter of the global oceanic anthropogenic CO 2 , although covering only 15% of the global ocean area [5]. The projected changes in ocean carbonate chemistry may thus not only have strong effects on the marine biota, but also on the oceanic carbon cycling.
Phytoplankton take up inorganic carbon (C i ) in the photic zone and fix it into organic compounds, thereby providing a carbon and energy source for higher trophic levels. The key enzyme of carbon fixation, the CO 2 -binding enzyme Ribulose 1,5-bisphosphate Carboxylase/ Oxygenase (RubisCO), exhibits a generally low affinity for its substrate CO 2 [6][7][8]. To avoid C i limitation, many phytoplankton species operate carbon concentrating mechanisms (CCMs) [9,10]. The efficiency in CO 2 fixation depends on both the type of RubisCO as well as the mode of CCMs so that the response of phytoplankton to OA cannot be generalized across taxa [11][12][13]. Various studies have provided mechanistic insights into the CO 2 -dependent regulation of CCMs and thus CO 2 fixation over a range of phytoplankton species (e.g. [14][15][16]). Besides species-specific differences, also strains of the same species may respond differently (e.g. [17][18][19]), which further complicates predictions on OA-driven changes in primary production.
To test these effects directly, numerous studies have exposed natural phytoplankton communities to high pCO 2 , either in bottle incubations or mesocosms, often finding higher rates of CO 2 fixation under OA [20]. In these experiments, which lasted only a couple of days up to a month, the effects were yet relatively small. Here, we investigated the impact of OA on primary production by a natural phytoplankton community over an entire winter-to-summer succession. Experiments were performed in large scale mesocosms, deployed in the Gullmar Fjord located in Southwest Sweden at the Skagerrak coast in 2013 [21]. Depending on the wind direction and tides, the fjord consists of high saline bottom water from the North Atlantic, a low salinity thin surface layer fed with water from the river Ö rekil, and in between a layer fed by the Baltic current. Monitoring data from over 100 years have shown that the phytoplankton spring community in the Gullmar Fjord is typically dominated by diatoms, whereas summer blooms often comprise dinoflagellates [22,23]. We assessed primary production of the phytoplankton community from the mesocosms as well as the fjord by applying 14 C incubations over 24 h [24]. We furthermore assessed the light dependency of CO 2 fixation by performing photosynthesis-irradiance response curves in short incubations (80 min.).

Material and methods
Primary production experiments did not involve endangered or protected species The KOSMOS 2013 mesocosm experiment was performed in the Gullmar Fjord (Kristineberg, Sweden) from March until July 2013 as part of the project BIOACID (Biological Impacts of Ocean ACIDification) phase II. Ten mesocosms were deployed near Kristineberg, with permission from the Sven Lovén Centre for Marine Infrastructure. The mesocosms were cylindrical polyurethane bags with a 2 m diameter mounted in a floatation frame [25]. The bags reached a depth of 17 m and were closed at the bottom with a 2 m long conical sediment trap [26]. Two days prior to the experiment (i.e. t-2), a water body was enclosed inside the mesocosms by lifting the upper end about one meter above the surface.
All mesocosms had a salinity of about 29, and nitrate, phosphate and silicate concentrations of about 7, 0.8, and 10 μmol L -1 , respectively. CO 2 enrichment was conducted on t-1 and t0, for which sterile-filtered and CO 2 -saturated seawater from the Gullmar Fjord was added to  five mesocosms (M2, M4, M6, M7, M8). The remaining five mesocosms (M1, M3, M5, M9,  M10) were treated as controls and received no CO 2 -enriched seawater. Average pCO 2 (based on dissolved inorganic carbon (DIC) and spectrophotometric pH T measurements) in the 'low' and 'high' CO 2 treatments were about 380 and 760 μatm, respectively. The systems were open and allowed a gas exchange at the sea surface. To account for CO 2 losses to the atmosphere by outgassing and for CO 2 consumption by primary production, CO 2 was added on a regular basis to the 'high' CO 2 treatments. As a consequence, CO 2 concentrations remained above the control treatment at all times (for more details see [21]). Sampling of seawater from each mesocosm was done with a depth-integrated water sampler (Hydro-Bios). After initial sampling on t0 and t1, samples were taken every other day until t109 (i.e. t3, t5, t7 etc.). For further information on the design and set-up of the experiment, as well as the CO 2 perturbation and sampling techniques, we refer to [21].

Sampling for primary production
For our measurements, integrated water samples from 0-17 m depth were taken in a four day interval (i.e. t1, t5, t9, etc.) from each of the ten mesocosms, and an additional sample was taken from the fjord. Sampling usually took place between 9 and 12 a.m. and aliquots from well mixed water samples were filled in gas-tight and headspace-free bottles (Schott) of 250 mL (for the 24 h incubations) and 500 mL (for the photosynthesis-irradiance response (P/I) curves). Samples were brought directly to the laboratory, where they were gently filtered over a 500 μm mesh-size filter to remove larger zooplankton from the samples, and were kept at the in situ water temperature until incubations started. Over the course of the entire experiment, the temperature in the fjord increased from 1.5˚C at t1 towards 15.5˚C at t109, and we adjusted the incubation temperatures accordingly (Fig 1A). Only at the beginning of the experiment, when productivity and biomass was still low, we could not fully match the temperature from the fjord as our incubator was not able to maintain temperatures below 4˚C. Light was provided by daylight tubes (OSRAM) from the side in a 16:8 h light-dark cycle. To account for the increase in light intensities over the course of the experiment in the mesocosms, the light intensity was stepwise increased in the incubator (Fig 1B). Using a spherical micro quantum sensor (Walz), we increased the photon flux density (PFD) every 16 days (i.e. after 4 sampling days) by about 20 μmol photons m -2 s -1 , starting with around 100 μmol photons m -2 s -1 at t1 and ending with 240 μmol photons m -2 s -1 at t109.

Primary production measurements
Primary productivity was measured according to Steeman Nielsen [24]. Despite limitations [27], this approach has remained the method of choice, especially for field work, as it allows assessing rates even at times of low productivity. One has to keep in mind, however, that measured rates have different meanings depending on the incubation time [27]. In our 80 min. incubations for 14 C-based photosynthesis-irradiance response (P/I) curves, we obtained rates of gross primary production because there is only little loss of incorporated 14 C via respiration and exudation over such short timescales. In our 24 h incubations for the 14 C-based primary production measurements, respiration lowers the 14 C incorporation and thus net rates of primary production are obtained. To account for fixed 14 C ending up in the dissolved phase, which can be a significant proportion under nutrient deplete conditions, we included values of the filtrate in our PP T estimates. 14

C-based primary production measurements
For the 24 h incubations, 40 mL sample volumes were spiked with 20 μL of 14 C-labeled sodium bicarbonate (NaH 14 CO 3 ; from a 1 mCi mL -1 = 37 MBq mL -1 stock solution; Perki-nElmer). Two incubation vials for each mesocosm, and the fjord water, were prepared accordingly (i.e. 22 vials in total, of which 11 were used for the light and the remaining 11 for the dark incubations). Determination of total 14 C-spike addition was done from an extra 40 mL 14 C-spiked water sample. For this purpose, 1 mL was directly transferred into a 20 mL scintillation vial (PerkinElmer) containing 10 mL scintillation cocktail (Ultima Gold AB; PerkinElmer) and counted in a liquid scintillation analyzer (Beckman LS6500). Blank determination was done by transferring 1 mL from the extra 14 C-spiked water sample into 6 mL of 6 M HCl, which degassed for 48 h and was then counted after adding 10 mL scintillation cocktail. All incubations were placed on an orbital shaker in a temperature-controlled incubator.
Incubations were stopped after 24 h by vacuum filtration onto GF/F filters (Whatman). To estimate the amount of C i fixation into particulate organic carbon (POC), filters were rinsed twice with 20 mL of sterile filtered seawater (0.2 μm), and subsequently placed in scintillation vials containing 300 μL of 3 M HCl to remove 14 C-labeled DIC. To estimate the amount of C i fixation ending up in the pool of dissolved organic carbon (DOC), 6 mL of filtrate was transferred into a scintillation vial, acidified with 1 mL 6 M HCl, and placed under a fume hood for 48 hours for degassing DI 14 C. Prior to measurements, 10 mL of scintillation cocktail was added to each vial and filter, thoroughly mixed, and counted in a liquid scintillation analyzer. Primary production (PP) was calculated according to: where DPM represents the decays per minute and t represents time. Correction for non-specific 14 C fixation in the dark was done by subtracting dark incubations from light incubations. Dark 14 C fixation accounted for about 1 to 6% of the light incubations during times of high and low productivity, respectively. Based on the phytoplankton community composition [21], some primary producers were smaller than the pore size of our filters (i.e. <0.7 μm). We therefore reported total primary production (PP T ; μmol C L -1 h -1 ) from the 24 h incubations as the sum of CO 2 fixation into POC and DOC.

C-based photosynthesis-irradiance response curves
For the photosynthesis-irradiance response (P/I) curves, 300 mL sample volume from each mesocosm was spiked with 100 μCi of NaH 14 CO 3 -(PerkinElmer) and subdivided into seven 40 mL glass vials. From the remaining 14 C-spiked seawater, 200 μL aliquots were transferred into a 10 mL scintillation cocktail to determine total spike addition for each P/I curve. While one vial was incubated in the dark, the six remaining vials were exposed to increased light intensities ranging from about 10 to 700 μmol photons m -2 s -1 in a custom-made photosynthetron. Light was supplied from below and the PFD was assessed prior to each experimental day. The photosynthetron was placed in the same incubator as the 24 h incubations. Additional temperature control was achieved via a water bath connected to the sample holder. After an incubation time of 80 min. at the respective light conditions, samples were filtered on GF/F filters (Whatman). Analysis of PO 14 C was determined following the same procedure as for the 24 h incubations and data was fitted according to: where P max is the light-saturated rate of photosynthesis, α is the light-limited (i.e. initial) slope of the P/I curve representing the maximum light-use efficiency, I is the irradiance, and I k is the light saturation index. Rates of PP P/I were normalized to chlorophyll a (Chl a) concentrations in the samples from the particular day and mesocosm [21].

Statistics
Differences in PP T , Chl a, P max , I k and α between the CO 2 treatments were tested over time by a two-way repeated measures Analysis of Variance (rmANOVA), and the association between PP T and Chl a was tested by Pearson product-moment correlations. Variables were log+1 or square root transformed if this improved normality or homogeneity of variances, as tested by the Shapiro-Wilk test or Levene's test, respectively. All statistics were performed with Sigmaplot 12.5 (Systat).

Total primary production
For the first three weeks of the experiment, estimates on PP T were lower in the mesocosms than in the fjord (Fig 2A). All mesocosms showed comparable development in PP T , with an initial period of low productivity (phase I, t1-t16), a first spring bloom of highest productivity around t33 (phase II, t17-t40), followed by a second bloom of highest productivity around t57 (phase III, t41-77), and a subsequent period of low productivity until the end of the experiment (phase IV, t78-t109; Fig 2A, Table 1). Dynamics in primary production in the mesocosms differed from that in the fjord. For example, PP T was higher in the fjord during phase I, while PP T was higher in the mesocosms during phase II. Also, a small increase in PP T present in the fjord at the start of phase IV was lacking in the mesocosms (Fig 2A). High pCO 2 yielded higher mean estimates on PP T during both blooms, although differences during both blooms were not significant ( Table 2). Highest PP T was observed during the first bloom at t33, with up to 16.1 ± 6.7 μmol C L -1 d -1 at high pCO 2 and 12.4 ± 7.0 μmol C L -1 d -1 at low pCO 2 . During the second bloom, PP T amounted to highest values of 11.3 ± 5.0 μmol C L -1 d -1 at t53 for high pCO 2 , and 6.0 ± 1.9 μmol C L -1 d -1 at day t57 for low pCO 2 At the peak of the second bloom, PP T appeared to be higher at high pCO 2 , though this difference was marginally significant and dependent on time (Table 2; rmANOVA, Time x CO 2 treatment, P = 0.098). During both blooms phases, Chl a remained unaltered in response to OA (Table 2), though at times showed higher concentrations at high pCO 2 [21]. Furthermore, Chl a was strongly correlated to PP T (σ = 0.87, P < 0.0001).
When cumulated over the experimental period of 109 days, the PP T data yielded a total of 92 ± 29.21 and 110 ± 25.79 μmol C L -1 at low and high pCO 2 , respectively. In the fjord, cumulative PP T yielded 95 μmol C L -1 (Fig 3A), being more comparable to PP T in the mesocosms at low pCO 2 . The difference in cumulative PP T between low and high pCO 2 was about 20% and closely matched the observed difference in Chl a concentration of about 15%. Consequently, no differences in the yields were observed when normalizing cumulated PP T to Chl a (as to account for changes in phytoplankton biomass). In both treatments, we got a cumulative value of around 600 μg C (μg Chl a) -1 until the end of the experiment (Fig 3B). Chl a-normalized cumulative PP T in the fjord was higher than in the mesocosms and amounted to a total of about 700 μg C (μg Chl a) -1 (Fig 3B).
Photoacclimation P/I curves provided information on the photoacclimation of the phytoplankton communities in the mesocosms and the fjord. P max was on average 3.17 ± 0.54 and 3.38 ± 0.26 μg C (μg Chl a) -1 h -1 at low and high pCO 2 , respectively. There was no apparent CO 2 effect on P max during both blooms (Table 2), which furthermore strongly varied between mesocosms and sampling days (Fig 4A). I k , indicating the light intensity at which phytoplankton shifts from light limitation to light saturation, changed over the course of the experiment (Fig 4B). More specifically, in the period prior to the first bloom (phase I), I k remained around 100 μmol photons m -2 s -1 and increased towards the end of the first bloom phase reaching mean values of approximately 160 and 250 μmol photons m -2 s -1 at high and low pCO 2 , respectively. In the course of the second bloom, I k decreased resulting in lowest values of 50 μmol photons m -2 s -1 around t61 (Fig  4), after which it increased again to values of around 150 μmol photons m -2 s -1 (Fig 4B). Besides these general changes over the season, we did not observe a significant CO 2 effect on I k values during both blooms ( Table 2). The maximum light-use efficiency also changed in the course of  the phytoplankton succession. Highest α values coincided with the phytoplankton blooms during phases II and III and were observed around t30 and t56 in all mesocosms (Fig 4C). Similar to the other parameters, there was no significant CO 2 effect on α values during both blooms ( Table 2).

Discussion
We did not observe a sustained effect of OA on primary production during the investigated winter-to-summer plankton succession. When focusing on the peak of the second spring bloom in phase III, however, PP T showed a marginally significant increase under high pCO 2 ( Table 2). During this distinct phase, the availability of inorganic nutrients was low and primary production was fueled by in situ remineralization [21]. Integrated over the entire experimental period, OA yielded about 20% more CO 2 fixation. Such enhanced primary production is in line with the higher Chl a concentration under these conditions. At the onset of the experiment, concentrations of major nutrients in the mesocosms were higher than in the fjord [21]. The lower concentrations in the fjord were the result of higher primary production compared to the mesocosms right after closure of the mesocosms  Table 2. Output of the repeated measures ANOVA for phase II, phase III, peak of bloom 1 and peak of bloom 2, with degrees of freedom (df), the Fvalue and the P-value. Significant outcomes are indicated with P < 0.001 (***), P < 0.01 (**), P < 0.05 (*) and P < 0.1 (Á). (beginning at t-2, Fig 2). Although initial conditions in the mesocosms were largely comparable to the situation in the fjord, perturbations induced during the set-up of the mesocosms [21], e.g. the water column mixing (t0) or the establishment of CO 2 treatments (t-1 and t0), may have contributed to the delay in primary production.

Parameter
In the mesocosms, PP T as well as Chl a concentrations remained relatively low during phase I and started to increase more pronounced around t20, leading to a first phytoplankton bloom with highest PP T around t33 (Fig 2). During this phase II, major nutrients such as inorganic phosphate and nitrogen were depleted to very low values (for more details, see Bach et al. [21]). This nutrient depletion, particularly for nitrogen, together with grazing presumably caused the collapse of the phytoplankton bloom and the decrease in PP T as well as Chl a concentrations (Fig 2). At the same time at the Kristineberg field station (~3 km distance to mesocosm deployment site), a sudden drop in the in situ light intensity was detected (Fig 1B), coinciding with the peak of the first bloom (Fig 2A). In fact, the average water column light intensities (0-19 m depth) during midday for all ten mesocosms were reduced to about 35 μmol photons m −2 s −1 for several days (Fig 4B), and dropped below the I k values (about 115 μmol photons m −2 s −1 ). Such lower light levels may possibly have, at least temporally, limited photosynthesis and thereby affected the response of phytoplankton to low nutrient levels. While there were dynamic changes in P max , I k , and α over the course of the phytoplankton succession (Fig 4), there was no effect of OA on photoacclimation.
Dissolved phosphate and inorganic nitrogen concentrations remained low during phase III, while PP T and Chl a concentrations increased again, causing the second bloom (Fig 2). An earlier study in the Gullmar Fjord also showed a relatively high primary production during summer months, despite low nutrient concentrations [28,29]. According to this long-term Gullmar Fjord time-series study, nutrients were not only derived from recycled production, but also from local precipitation, run-off, and input from the Kattegat [29]. As the mesocosms were isolated from the surrounding water, nutrient input for primary production should have derived from regeneration only. In fact, dissolved organic nitrogen and NH 4 + concentrations in the mesocosms remained low, indicating a rapid cycling of nutrients in the food web [21]. Interestingly, it is /under these conditions of recycled production and low concentrations of inorganic nutrients that we observed the strongest response in PP T towards OA (Fig 2A). Under nutrient-limited conditions, effects of elevated pCO 2 on phytoplankton productivity, standing stock and community composition were often found to be stronger [30][31][32] and comparable findings were also reported with respect to iron limitation [33]. Since nitrogen, phosphorus and iron predominantly limit phytoplankton growth in the global surface oceans [34,35], more studies investigating the combined effects of elevated pCO 2 and resource limitation are required to provide a mechanistic understanding on the impacts of OA on future primary production. Effects of ocean acidification on primary production in a coastal North Sea phytoplankton community Effects of ocean acidification on primary production in a coastal North Sea phytoplankton community Even though we did not find a consistent CO 2 response over the entire winter-to-summer plankton succession, the stimulation in primary production under elevated pCO 2 at the peak of the second bloom was also observed in earlier studies looking at mixed natural assemblages as well as monoclonal laboratory cultures. During a mesocosm study in Bergen (Norway), for instance, DIC uptake increased under a comparable OA scenario by about 40% [30,36]. Moreover, a mesocosm study in Kongsfjorden (Svalbard, Norway) showed an OA-induced increase in primary production of 10 to 60% over the experimental period [37]. Such increases in primary production may derive from physiological changes in predominant species and/or shifts in community composition both leading to higher phytoplankton biomass buildup. At a higher taxonomic level, the phytoplankton community remained largely unaltered and was dominated by diatoms [21]. Under nutrient-replete as well as nutrient-limiting conditions, elevated pCO 2 resulted in an increased abundance of picoeukaryotes [21]. Specific changes within phytoplankton groups will be discussed elsewhere in this special issue (see S1 Table in [21]). With regard to the dominating role of diatoms in our experiment, several studies found this group to enhance their C i fixation rates in response to elevated pCO 2 , which was often attributed to the down-regulation in the CCM activities under these conditions (e.g. [38][39][40]). Such enhanced OA-driven efficiencies in C i fixation may, at least partially, have contributed to the higher phytoplankton biomass during the second bloom in our experiment.
Our results indicate an OA-dependent increase in primary production during certain times of the spring-to-summer phytoplankton succession, particularly under NO 3 limitation (phase III) being accompanied by a significant increase in picoeukaryotes during this period [21]. With respect to higher trophic levels, OA showed differential growth effects on several predominant mesozooplankton species, though as a whole, the community remained rather unaltered under OA (Alguero et al. in prep.). OA led, however, to an increase in the survival rate of herring larvae (being planted in the mesocosms on t63), which could be linked to higher prey abundances (Swaat et al. in prep.). Hence, the observed changes in primary production under OA have a high potential to restructure phytoplankton communities in the future coastal North Sea with likely consequences for higher trophic levels.