Effects of Increasing Seawater Carbon Dioxide Concentrations on Chain Formation of the Diatom Asterionellopsis glacialis

Diatoms can occur as single cells or as chain-forming aggregates. These two strategies affect buoyancy, predator evasion, light absorption and nutrient uptake. Adjacent cells in chains establish connections through various processes that determine strength and flexibility of the bonds, and at distinct cellular locations defining colony structure. Chain length has been found to vary with temperature and nutrient availability as well as being positively correlated with growth rate. However, the potential effect of enhanced carbon dioxide (CO2) concentrations and consequent changes in seawater carbonate chemistry on chain formation is virtually unknown. Here we report on experiments with semi-continuous cultures of the freshly isolated diatom Asterionellopsis glacialis grown under increasing CO2 levels ranging from 320 to 3400 µatm. We show that the number of cells comprising a chain, and therefore chain length, increases with rising CO2 concentrations. We also demonstrate that while cell division rate changes with CO2 concentrations, carbon, nitrogen and phosphorus cellular quotas vary proportionally, evident by unchanged organic matter ratios. Finally, beyond the optimum CO2 concentration for growth, carbon allocation changes from cellular storage to increased exudation of dissolved organic carbon. The observed structural adjustment in colony size could enable growth at high CO2 levels, since longer, spiral-shaped chains are likely to create microclimates with higher pH during the light period. Moreover increased chain length of Asterionellopsis glacialis may influence buoyancy and, consequently, affect competitive fitness as well as sinking rates. This would potentially impact the delicate balance between the microbial loop and export of organic matter, with consequences for atmospheric carbon dioxide.


Introduction
Amongst the most recent (180 Ma [1]) planktonic unicellular autotrophs of Earth's Oceans, diatoms exhibit diverse morphologies and ecological strategies. For instance, diatoms can occur either as single cells or as colonies, influencing buoyancy, predator evasion, light absorption and nutrient uptake. The processes by which adjacent cells in chains establish connections determine strength and flexibility of the bonds, and their distinct cellular locations define colony structure. In a constantly changing environment cells, whether solitary or in colonies, need to be able to regulate their gene expression, physiology and signalling. Colonial species such as the diatom Skeletonema costatum (S. costatum), have been shown to alter chain formation, namely by increasing chain length with temperature (from 6 to 17uC) and nutrient availability [2]. Finaly, chain length has been found to be positively correlated with growth rates [2] and to follow the inverse trend in senescent populations [3][4].
Adjacent cells in a chain of S. costatum attach by external silica tubes at the margin of the valves ( [4], http://www.protistcentral. org/index.php/Taxa/get/taxa_id/2843). However, this is not a unique strategy and other species establish cell-cell connections by means of mucus, bands or even septa fusion [5]. In the case of the cosmopolitan Asterionellopsis glacialis (A. glacialis), cells attach at the valve apices by exuded polysaccharides which form mucilage pads [1]. Changes in seawater chemistry could influence the binding strength or secretion of polysaccharides, especially when charged, and potentially affect chain formation. This in turn may influence buoyancy, predator or pathogen evasion, light absorption and nutrient uptake ( [6], summarized in Beardall et al. [7]). Furthermore, cells in a chain such as in spirals of A. glacialis may develop a microenvironment with lower CO 2 concentrations/ higher pH in the centre of the colony during daylight where photosynthetic removal of CO 2 can lead to diffusional limitation and localized depletion. However, virtually nothing is known about the effects of varying environmental conditions (e.g. pH) on chain length of A. glacialis.
Atmospheric CO 2 has been increasing since the industrial era, reaching values (currently ,400 matm) above those observed in the last 800 000 years (from ,180 to ,280 matm). In a business as usual scenario [8] CO 2 is projected to continue to increase, reaching about 750 matm by the year 2100. As CO 2 increases in the atmosphere, it also enters the ocean by air-sea gas exchange, increasing its average concentration and shifting the carbonate chemistry to a more acidic environment (termed ocean acidification).
Recent studies have revealed that changes in carbonate chemistry as expected in the future ocean [8] can affect marine phytoplankton in various ways (e.g. [9][10]). Until now, studies with diatoms mostly focussed on carbon acquisition [11,12] or found higher growth, carbon fixation rates and/or increased efficiency of energy conversion to photosynthesis [13][14][15][16][17] under high CO 2 concentrations. Considering that these silica shielded planktonic primary producers are thought to account for up to 45% of net primary productivity in the ocean [18], a further increase in carbon fixation could act as negative feedback for atmospheric CO 2 . However, the response of diatoms to increasing CO 2 is still poorly understood and the importance of organization strategies has been mostly overlooked so far [19]. Here we investigate whether changing seawater carbonate chemistry affects the physiology (cell division and organic matter production rates and element stoichiometry) and colony/chain formation of the cosmopolitan diatom A. glacialis. Additionally, we provide reasoning that the observed response of A. glacialis is driven by distinct parameters of the carbonate system (carbonation versus pH) depending on the CO 2 concentration to which the cells were being exposed.

Experimental Setup
Freshly isolated monospecific cultures of the cosmopolitan Asterionellopsis glacialis (strain isolated offshore the Azores (CCMMG_1, October 2011)) were grown semi-continuously under varying CO 2 levels (between approximately 320 and 3400 matm, pH total scale of ,8.15 to 7.24, for more detail see Table 1) for a minimum of 20 generations before the start of the experiments. No specific permissions were required for these location (38u379N27u159W)/activities at the time of collection. The field studies did not involve endangered or protected species. All cultures were grown in 0.2 mm sterile filtered North Atlantic water (salinity of 36) enriched with approximately 4 mmol l 21 phosphate and 64 mmol l 21 of nitrate and silicate (the increase of total alkalinity upon addition of Na 2 SiO 3 was compensated for by HCl addition), and with trace metals and vitamins following f/8 [20] at 20uC, a photon flux density of 220 mmol m 22 s 21 (supplied from OSRAM L 18W/840, Lumilux, coolwhite) and a 14/10 h light/dark cycle.
The media was acclimated to the temperature of the experiment before inoculation. Cell densities varied on average between 140 and 16000 cell ml 21 , therefore minimizing changes in seawater carbonate chemistry (average DIC drawdown of 3.9%). All cultures were vertically rotated (10 times gently) daily one hour after the beginning of the light phase to avoid aggregation, sedimentation and self-shading during the light phase.

Carbonate system
The carbonate system was manipulated by adding calculated amounts of NaHCO 3 and HCl in a closed system following Schulz et al. [21]. Alkalinity was measured by potentiometric titration following Dickson et al. [22], using a Metrohm Titrino Plus 848 equipped with a 869 Compact Sample Changer, and calibrated with certified reference material supplied by A. Dickson. The pH was measured using a glass electrode (WTW, pH 340i) and calibrated with a TRIS seawater buffer, supplied by A. Dickson. Carbonate chemistry was calculated from measured temperature, salinity, silicate and phosphate concentrations, and pH and TA using CO2sys [23], with the equilibrium constants determined by Mehrbach et al.

Nutrients
Samples for the determination of nutrients at the start and end of incubations were filtered through a polyethersulfone (PES) 0.2 mm syringe filter and stored at 220uC until being analysed. Concentrations of nitrate, silicate and phosphate were measured following Hansen and Koroleff [26], by means of a spectrophotometer (Cary 50 Probe, Varian).

Cellular element quotas and dissolved organic carbon exudation
Samples for cellular particulate organic carbon (POC), nitrogen (PON) and phosphorus (POP) were gently filtered (200 mbar) through pre-combusted GF/F filters (6 h, 450uC) and stored at 220uC until analyses. POC and PON samples were then dried (4 h, 60uC), packed in tin boats and analysed in a gas chromatograph (EURO EA Elemental Analyser, EUROVEC-TOR equipped with a thermal conductivity detector and an element analyzer) following Sharp [27]. POP filters were oxidized by potassium peroxydisulphate to dissolved inorganic phosphorus and measured colorimetrically by means of a spectrophotometer (UV-1202, UV-VIS Spectrophotometer, SHIMADZU) following Hansen and Koroleff [28]. Daily production rates were calculated by multiplying cellular quotas (POC, PON, POP per cell abundances) with the corresponding cell division rates m (see below).
Dissolved organic carbon was estimated as the difference between calculated (from TA and pH) inorganic carbon consumption (DDIC) and net build-up of organic matter (DPOC).

Cell numbers and growth rates
Cell abundances (on average ,800 cells per sample were counted) and the number of cells in a chain were determined from samples fixed with Lugol (2% final concentrations) by means of an inverted microscope (Leica DMIL) at 2006 magnification. Cell division rate (m) was calculated as: where Ce and Ci refer to end and initial respectively of concentrations of cells, POP, POC or PON, and Dt to the duration of the incubation period in days. The equation used for fitting cell division rates (m tl , tendency line) based on cell numbers and on all parameters (cell concentrations, POP, POC and PON) followed a modified Michaelis Menten kinetic [29], allowing for optimum curve characteristics, as: in which a (cell 3.35, all parameters 3.46) and b (cell 89.6, all parameters 93.21) are random fitting parameters, c (cell 0.0006739, all parameters 0.0006819) describes the CO 2 sensitivity and pCO 2 (matm) refers to the CO 2 level.

Statistical analysis
Statistical significance of the data was tested for by Anova (significance determined as 99%, p,0.01), using the program R.

Results
When CO 2 was increased from approximately 320 to 3400 matm, the relative number of chains composed of 1 to 6 cells decreased (p,0.01) while longer chains with 7 to 18 cells increased (7 to 12 cells p,0.01 and 13 to 18 cells p = 0.05, i.e. significant at a 95% confidence level) (Fig. 1, Fig. 2). Data was fitted linearly.
Cell division rates based on cell numbers and organic matter (POC, PON and POP) followed a modified modified Michaelis Menten curve (R 2 all data = 0.69), not varying significantly from 320 to 600 matm, but decreasing with rising CO 2 between ,600 to 3400 matm (Fig. 3). In fact, cell division rates based on cell numbers decreased on average 2.3 fold (p,0.01) with higher CO 2 between the interval considered (,600 to 3400 matm). For CO 2 levels ranging from ,600 to 1470 matm, the decrease was associated with an increase by approximately 1.5 fold of the  cellular quotas of carbon (C), nitrogen (N), and phosphorus (P) (Fig. 4). This trend was not maintained at CO 2 levels higher than 1470 matm, at which C, N and P quotas decreased. A similar trend, but now following a modified Michaelis Menten kinetic, to cellular contents was observed for N, C and P production rates (Fig. 5).
Despite the observed trends in cellular quotas and production rates no significant correlation was obtained for organic element ratios (C to N, C to P and N to P), showing a proportional storage at all CO 2 levels tested (Fig. 6). Finally, associated with the decrease in cell division rate, there was a ,2 fold increase (p,0.01) of exudation of carbon in the form of dissolved organic carbon (DOC) as depicted in the linear increasing difference between calculated (from TA and pH, except for one value since TA was not precise) inorganic carbon consumption (DDIC) and net build-up of organic matter (DPOC) from 1260 to 3400 matm (range of CO 2 levels corresponding to positive values of exudation). This trend could be pinpointed to increased cellular exudation (Fig. 7) and was maintained also as cellular exudation rates (data not shown).

Discussion
Growth response to enhanced CO 2 /decreased pH Despite the importance of diatoms in the marine carbon and silica cycles only a few studies have considered the effects of varying CO 2 concentrations on their physiology. Indeed, diatoms are thought to be less sensitive to increasing CO 2 than other phytoplankton groups such as coccolithophores. A number of studies have analysed the influence of CO 2 levels on diatom carbon concentration mechanisms (e.g. [12,30,31,32,33]). Nevertheless, there are only a few studies directly addressing the potential effects of enhanced CO 2 levels on diatoms, with the majority showing null to little effects (Thalassiosira weissflogii under 36 to 1800 ppmv [12]; Thalassiosira pseudonana under 380 and 760 ppmv [34] and Phaeodactylum tricornutum from ,20 to 800 ppmv [17]) or positive effects (e.g. enhanced growth rate, carbon fixation and/or increased efficiency of energy conversion to photosynthesis; as in the case of experiments done with Phaeodactylum tricornutum (about 380 and 1000 ppmv CO 2 , [15]), S. costatum (350 and 1000 ppmv CO 2 , [14]), Thalassiosira pseudonana (,390 to 750 ppmv CO 2 , [16])), Asterionellopsis glacialis, Thalassiosira punctigera and Coscinodiscus wailesii (from ,20 to 800 ppmv [17]) under increasing CO 2 . Moreover, studies with natural diatom-rich phytoplankton communities have shown dominance of larger diatoms under enhanced CO 2 concentrations in the Ross Sea [35] and Southern coast of Korea [36]. The positive response of diatoms is thought a consequence of an associated down-regulation of carbon concentrating mechanisms (CCMs) with increasing CO 2 concentration (e.g. [37][38]), since the energy saved by a down-regulated CCM operation could be reallocated to carbon fixation and growth. Indeed both solitary (Phaeodactylum tricornutum and Thalassiosira pseudonana) and colony forming species (S. costatum and Thalassiosira weissflogii) have been shown to continue to increase cell division rates at CO 2 levels higher than 600 matm. In the present study the CO 2 threshold isn't conclusive. However, the tendency line estimated suggest that A. glacialis cell division rates increased until ,600 matm CO 2 , potentially driven by the excess energy saved from the CCM, decreasing at CO 2 values higher than 600 matm probably related to low pH values. The apparent slightly higher sensitivity of A. glacialis to enhanced CO 2 concentrations, for the CO 2 treatments considered, might be species-specific, but an effect of the colony structure (shaped as star, zigzag and spiral) and the consequent proximity of sister cells cannot be excluded. Similar to other colonial phytoplantkon species, such as the cyanobacteria Anabaena sp. and Nodularia spumigena [39], the centre of A. glacialis colonies might have relatively high pH/low CO 2 concentrations during the day in comparison to the bulk media, decreasing diffusive CO 2 supply. Hence, the initial positive effect of increased CO 2 availability, here more visible in photosynthesis (carbon  production rate) than growth, was counterbalanced by the effects of the pH decrease. Increased chain length of A. glacialis may have influenced CO 2 supply, but more importantly at more extreme conditions of pH exposure, may have maintained localised external pH closer to optimum. Finally, the modifications in colony length of A. glacialis might come as a compensation for a higher pH optimum (more alkaline) of this species, at the expense of energy and cell division rate. This may oppose the response of other species such as Thalassiosira weissflogii or S. costatum which form more linear colonies. In Proboscia alata cells formed spirals under the combined effect of low CO 2 concentrations (below present concentrations and the range considered in this study) and high light [19]. However, in this case, the modification in morphology might be related to a strategy to reduce excess light penetration under low CO 2 supply, thereby reducing reactive oxygen species production and keeping growth rate constant.

Influence of carbonate chemistry speciation on chain length
Lower cell division rates found in this study were associated with longer chains of A. glacialis. In contrast, the growth rate of S. costatum has been found to be positively correlated with chain length both in cultures and enclosed natural communities [2]. Discrepancy in the correlation between colony growth and metabolic rates has been previously reported [7]. The increased chain length and proximity of the cells due to the observed colony structure under high CO 2 concentrations might be a strategy to increase pH in the centre of the colonies during the light phase or may simply be a consequence of the nature of the bonds established between adjacent cells in a chain. These bonds vary from septa fusion [1] to attachment at the valve apices by exudation of polysaccharides [5] depending on the diatom species. Adjacent cells of S. costatum establish low flexibility bonds by connection of external tubes, which may break with increased turbulence. Under a low turbulence environment, as cell division rate rises, the number of cells in a given chain and time should increase independent of the type of cell-cell connections. A. glacialis cells bind by mucilage polysaccharide pads with high C:N and C:P [7]. However, no detectable increase in polysaccharide production was observed either by using Calcofluor White staining (data not shown) nor by a change in the C:P and C:N ratios under increasing CO 2 concentrations/decreasing pH values. Therefore, longer chains under decreased cell division rates and slight turbulence due to mixing in this study might be connected to stronger bonds between polysaccharides at lower pH conditions. Interestingly, the elemental ratios found in this study were lower than Redfield and distinct from those found for the same species (different strain and growth conditions) by Burkhardt et al. [40], but within the range found in previous studies for cells (e.g. similar C:N to [16]) under nutrient replete conditions (C:P 27-135 and N:P 5-19, for a revision see [41]).

Cell uptake/exudation balance
Under nutrient-replete conditions, lower cell division rates would be expected to be accompanied by increased cellular element quotas as observed here until 1470 matm CO 2 . However, N, P and C quotas decreased with increasing CO 2 from 1470 to 3400 matm in spite of the decreasing trend in cell division rates. This is potentially due to increased exudation of dissolved organic compounds or variable nutrient uptake. Nutrient drawdown data is not conclusive (data not shown), but it is evident that P and Si uptake were higher at 3400 matm CO 2 concentrations than 600 matm while nitrate showed no trend. Hence, the decrease of cellular quotas could not be explained by lower nutrient uptake. Similarly, in Thalassiosira weissflogii there wasn't a significant difference in Si uptake with increasing CO 2 concentrations from ,370 to 750 matm, changing the rates of dissolution, efflux and incorporation into the frustule from ,100 to 750 matm [42]. Here, the difference between P, Si and nitrate drawdown may reflect a number of factors related to cell signalling (e.g. unsaturated aldehydes, see [43,44]), energetics and membrane permeability.
Enhanced exudation of organic matter, namely carbohydrates, has been previously observed as a response to stressors such as increasing CO 2 concentrations in coccolithophores [45] and nutrient limitation at the end of phytoplankton blooms [46]. Exudation as a response to low pH values might indeed explain the observed trend in carbon cellular quotas as depicted by uptake rates (DIC) that are greater than the accumulation rate of organic  carbon. This is further supported by the POC production rate decrease with the increase of dissolved organic exudation rates at higher CO 2 levels.

Summary and conclusions
The present study shows that cell division rates of A. glacialis did not change significantly from 320 to 750 matm of pCO 2 , but started to decrease towards higher CO 2 levels. This decrease was accompanied by an increase in cellular element quotas and organic matter production rates until 1470 matm, and by increased DOC exudation at CO 2 levels higher than that, with no changes in stoichiometric element ratios. Moreover, the relative number of cells per chain (chain length) increased at elevated CO 2 , potentially limiting nutrient diffusion under deplete conditions. Longer chains and modified chain morphology could influence buoyancy and sinking rates as in the case of other species [3,4,47]. If A. glacialis follows the response of S. costatum [4] the increased buoyancy with chain length could in turn positively affect growth in the natural environment since cells closer to the surface of the ocean will be exposed to an increased average light intensity. Hence, the chain formation strategy (i.e. longer chains) displayed by A. glacialis might be advantageous under future scenarios of elevated CO 2 where increased light supply might further increase photosynthesis. Depending on the sensitivity of co-occurring species, these changes could affect the plankton community composition. Finally, the increased exudation of dissolved organic carbon might increase aggregation and potential for sinking of particles. Figure 7. Cellular net exudation with increasing CO 2 (pCO 2 ). Exudation calculated here as the difference between calculated inorganic carbon consumption (DDIC) and particulate organic carbon build-up (DPOC) in relation to CO 2 levels (pCO 2 ) per cell (p,0.01). Solid line was obtained by linearly fitting the data. doi:10.1371/journal.pone.0090749.g007