Diel variability in the elemental composition of the marine cyanobacterium Synechococcus Contribution to the Themed Section: Scaling from individual lankton to marine ecosystems

The ratio of elements such as carbon:nitrogen:phosphorus (C:N:P) in phytoplankton is known to vary substantially within single isolates and across environmental gradients. In addition, C:N:P is known to vary throughout the day due to diel patterns in nutrient acquisition and storage. It has been hypothesized that small phytoplankton such as marine cyanobacteria have relatively invariable elemental ratios during a 24 h period, whereas larger phytoplankton have a greater capacity to store elements and thus a wider diel range of C:N:P. To test this hypothesis, we examined diel variability in cellular C:N:P, using a chemostat culturing system, for one of the most abundant marine cyanobacteria, Synechococcus (WH8102) during two 24 h periods. The cellular C quota nearly doubled during the 14 h light period and was subsequently reduced during the dark period. The cellular N quota also varied considerably, whereas the P quota remained relatively stable. These daily changes in elemental quotas led to highly variable C : N cell and C : P cell . Furthermore, the magnitude of variability in cellular elemental stoichiometry of Synechococcus was positively related to the growth rate. We constructed a model to test the extent to which variation in C : N cell and C : P cell is related to reserve carbon accumulation and depletion over each light–dark cycle. Results imply that, in addition to growth-related respiratory losses, Synechococcus also purges excess C during the dark period in order to maintain a nutritive balance within cells. Our data suggest that diel variation in C : N cell and C : P cell of Synechococcus is of the same order of magnitude as stoichiometric variation within plankton communities between major ocean environments.


I N T RO D U C T I O N
In the 80 years since its inception, the identification of a nearly constant, atomic C:N:P ratio in the marine environment has remained a central foundation of marine biogeochemical processes (Redfield, 1958). Despite substantial plasticity in cellular C:N:P in both field and laboratory observations (Karl et al., 2001;Bertilsson et al., 2003;Heldal et al., 2003), fixed elemental stoichiometry in marine phytoplankton has been consistently used to infer global primary production (Martin et al., 1987), carbon export (Krishnamurthy et al., 2009) and nitrogen fixation and loss rates (Deutsch et al., 2001(Deutsch et al., , 2007. More recent observations and models indicate that the elemental stoichiometry of marine particles varies between vast N-and P-limited ocean basins and regions (Martiny et al., 2013a,b;DeVries and Deutsch, 2014;Teng et al., 2014). The mechanisms that drive variation in elemental stoichiometry have been attributed to nutrient-limited growth (Rhee, 1978;McCarthy and Goldman, 1979;Sterner and Elser, 2002), light (Laws and Bannister, 1980;Urabe et al., 2002), temperature (Toseland et al., 2013) and taxonomy (Geider and La Roche, 2002;Ho et al., 2003;Weber and Deutsch, 2010;Quigg et al., 2011;Martiny et al., 2013a). Thus, environmental controls on basic growth physiology and community composition of marine phytoplankton contribute to spatial and temporal patterns in elemental stoichiometry of marine biota (Flombaum et al., 2013, Martiny et al., 2013a. In addition to broad spatial and temporal patterns, elemental stoichiometry of phytoplankton is known to vary in short daily rhythms within cells (Geider and La Roche, 2002;Granum et al., 2002;Talmy et al., 2014). For example, the marine cyanobacterium Crocosphaera watsonii separates inorganic C and N 2 fixation during the day and night (Berman-Frank et al., 2007), leading to daily patterns in cellular C:N ratios (Mohr et al., 2010;Saito et al., 2011;Gradoville et al., 2014). In chemostat cultures of the diatom Skeletonema costatum, cellular carbon cycling during light and dark periods was shown to confer a daily pattern in C:N (Anning et al., 2000). In contrast, small phytoplankton such as Prochlorococcus and Synechococcus have been suggested to have a smaller capacity to store C and N and therefore a relatively invariable C:N ratio (Talmy et al., 2014) when compared with large phytoplankton such as diatoms. This stronger flexibility in large cells may lead to a competitive advantage in terms of energy storage and growth in environments in which the light intensity is highly variable, such as in deeply mixed layers, or in which the light period is short, as it can be in high latitude waters (Van Oijen et al., 2003).
Specific cell components can create diel variability in elemental stoichiometry of phytoplankton (Geider and La Roche, 2002). For example, some studies report diel variability in nucleic acid concentrations and pigment fluorescence of Prochlorococcus and Synechococcus (Vaulot et al., 1995;Jacquet et al., 2001). Other data indicate that amino acid uptake has a diel pattern (Mary et al., 2008), which is likely synchronous with cellular protein concentrations (Matallana-Surget et al., 2014). Cycling of cellular pools containing a high C:P ratio (e.g. proteins and carbohydrates) relative to those with a low C:P ratio (e.g. nucleic acids and polyphosphates) could also create diel variability in cellular elemental stoichiometry.
In order to determine how cellular elemental stoichiometry varies over a diel period in small phytoplankton, we first asked how does cellular C, N and P content in steady-state chemostat cultures of Synechococcus WH8102 change over two 24 h periods. Secondly, we asked how differences in the growth rate contribute to diel variability in elemental stoichiometry. Thirdly, using a mathematical model, we identify putative physiological mechanisms that contribute to this variability. Synechococcus contributes to a large fraction of net ocean primary production (Flombaum et al., 2013) and thus plays a major role in cycling of biogeochemical elements. Fine-scale resolution of elemental stoichiometry in dominant lineages of marine plankton is needed to more accurately model marine biogeochemical processes (Prézelin, 1992).

M E T H O D
Chemostat cultures of Synechococcus (WH8102) were grown in 8 L-polycarbonate bottles at 248C. Ambient light (195 mmol quanta m 22 s 21 ) was supplied using cool white fluorescent lamps on a 14:10 light:dark cycle (Mouginot et al., 2015). We selected the temperature and light conditions on the basis of optimal growth conditions reported by Mackey et al. (2013) and Moore et al. (1995). All parts of the chemostat system were autoclaved before adding 60 mL inoculum culture of Synechococcus over an open flame to minimize biological contamination. The culture medium (artificial seawater) was autoclaved and cooled to room temperature before adding 0.2 mm-filtered bicarbonate, carbonate, nitrate, phosphate and trace metals in concentrations described by Garcia et al. (in review) to limit biomass accumulation by nitrate (added as NaNO 3 ), where measured nutrient concentrations in the medium were 15.9 mM NO 3 2 and 9.2 mM PO 4 32 (added as K 2 HPO 4 ; N:P input ¼ 1.7). Phosphate and nitrate concentrations were measured with the colorimetric assay, described in the Bermuda Atlantic Time-series Study protocols (Michaels et al., 1997a, b) using a ThermoVis 2000 spectrophotometer at 885 and 543 nm, respectively. We selected these specific sources of N and P because of their prominence throughout the world's oceans. All parts used to culture Synechococcus were autoclaved, and transfer of media and cultures to the chemostat system was performed in a laminar flow hood over an open flame. The culturing system was sealed with 0.2 mm-filtered air pumped into the chamber and a 0.2 mm filter attached to an air outlet. Dilution rates (i.e. growth rates) were controlled by modulating the culture volume (2.3-5.25 L) rather than the flow rate.
We collected samples for elemental ratio estimates and cell measurements (1 mL) from steady-state cultures every 3 h during two 24 h periods. The second 24 h period was measured 7 days after the first period. Culture samples for the analysis of particulate organic carbon and nitrogen (POC and PON; 150 mL) and particulate organic phosphorus (POP; 50 mL) were passed through pre-combusted (4508C) GF/F Whatman glass fiber filters under low pressure (5 psi). Samples for the analysis of POC and PON were dried at 60 -808C ( 48 h), pelletized and analyzed using an EA CN Elemental Analyzer (Thermo-Scientific). Samples were rinsed with 0.17 M NaSO 4 , and samples for POP analysis were then dried at 60-808C with 2 mL of 0.017 M MgSO 4 and combusted at 4508C for 2 h before adding 5 mL 0.2 M HCl and baking at 80 -908C. The resulting orthophosphate concentrations were measured using the Bermuda Atlantic Time-series Study protocols (Michaels et al., 1997a). We analyzed the culture cell density with a flow cytometer (Accuri C6) by identifying particles with forward scatter (FSCH) and Chl a fluorescence. We also measured the fluorescence of phycoerythrin with the flow cytometer. A script was then used in MATLAB (The MathWorks, Inc., Natick, MA, USA) to fit a least-squares sine function to cell quota data for each growth rate treatment ( Fig. 1).
Although we do not have direct measurements of heterotrophs from our chemostats, other estimates suggest that heterotrophic contaminants represent a small portion of the total biomass of our Synechococcus cultures. In chemostat cultures grown under nearly identical conditions, the number of fluorescent cells was always .90% of all particles visible with light microscopy at 40Â. In addition, we have good agreement between estimates of stoichiometry in the data reported here and elsewhere, where we monitored chemostats with fluorescence microscopy (Garcia et al., in review).
To explore the influence of reserve carbon accumulation and depletion on stoichiometric variation over a diurnal cycle, we used a model that connects organism growth rate with carbon fixation and protein synthesis. A full model description is provided in the Supplementary data. The model structure and main assumptions are similar to previously published model forms (Geider et al., 1998;Flynn, 2008;Talmy et al., 2014). We built on previous models by exploring various assumptions regarding the day-night dependence of the carbon loss rate and considered the implications of our findings for understanding the ecology of Synechococcus. For all simulations, the nutrient concentration in the input reservoir (S in ) and the dilution rate (D) exactly matched the experimental conditions. With this design, we tested the influence of two contrasting assumptions on the model dynamics: (i) growth-rate-independent carbon losses, which may include maintenance respiration Gray bands indicate dark periods during two 24 h periods. Break in x-axis denotes 7-day period between first and second 24 h sampling period. Sine curves were best fit to the data using a least-squares script. and excretion, are equal to 0.01 day 21 and thus have a negligible influence on diurnal changes in elemental stoichiometry and (ii) a preferential purging of carbon by Synechococcus (4.0 day 21 ) in a manner that is independent of growth rate. Furthermore, carbon losses are assumed to be negligible in the day (0.01 day 21 ), but elevated at night (Supplementary data, Table SII).

R E S U LT S
To examine diel patterns in cellular elemental stoichiometry, we grew Synechococcus (WH8102) in chemostat cultures, limiting biomass accumulation with a low supply of nitrate (N:P input ¼ 1.7). We analyzed cellular elemental quotas during two 24 h periods and observed a strong daily pattern in the cellular carbon and nitrogen quotas (Q C and Q N ) (Fig. 1). This included substantial (up to 93%) increases during the light period and decreases during the dark period. The percent change in Q C and Q N during the day scaled positively with increasing growth rate (Fig. 1). In the fast-growing culture (0.74 day 21 ), Q C increased by 93% (increasing from 113 to 218 fg C cell 21 ), whereas Q N only increased by 50% (20-31 fg N cell 21 ; Table I and Fig. 1). In the slowest growing culture, however, Q C only increased by 44% (from 90 to 130 fg C cell 21 ), whereas Q N increased by 7% (16-17 fg N cell 21 ; Table I and Fig. 1). Changes in the cellular phosphorus quota (Q P ) were similar, but percent increases during the day were generally smaller, increasing by 27% (3.12-3.96 fg P cell 21 ) in the fast-growing culture (0.74 day 21 ) and 13% in the slow-growing culture (0.36 day 21 ; Table I and Fig. 1). Cell size was monitored independently with flow cytometry as FSCH (a proxy for cell size based on cell diameter) and revealed a pattern similar to that of Q C ; the cell size was highly correlated with Q C (R ¼ 0.93, P , 0.001; Fig. 2). Thus, the range of variation in Q C and cell size over the diel cycle scaled positively with growth, and the largest differences between cultures were observed at the conclusion of the light period. Q N and Q P , however, exhibited peaks  roughly midway through the light period and troughs roughly midway through the dark period ( Fig. 1B and C), particularly in the two fast-growing cultures. The sinusoidal spline interpolation through the Q N and Q P of these cultures confirmed that the maxima occurred 8 h prior to the conclusion of the light period ( Fig. 1B and C). Diel variation in Q C , Q N and Q P led to large oscillations in C:P cell and C:N cell (Table I and Fig. 3), which scaled positively with growth. Although diel ratios of C:P cell and C:N cell increased by 66% and 69% in the fast-growing culture (0.74 day 21 ), respectively, diel ratios of N:P cell only increased by 20% (Table II). In the slow-growing culture (0.36 day 21 ), percent changes in C:P cell , C:N cell and N:P cell were smaller: 43%, 41% and 11%, respectively. Daily maximum values of C:N cell and C:P cell occurred at the end of the 14 h light period, leading to the hypothesis that daily carbon acquisition and losses were the main drivers of diel variation in cellular elemental stoichiometry of Synechococcus.
Pigment fluorescence data also exhibited diel variability with maxima 8 h before the conclusion of the light period ( Fig. 4A and B). Because the majority of the main pigment phycoerythrin is composed of nitrogen-rich proteins, we compared diel variability in Q N to variability in fluorescence of pigments ( Fig. 4C and D). This apparent relation was analyzed by simple linear regression, and Q N was highly correlated with pigment fluorescence (R ¼ 0.92, P , 0.001) with a coefficient of determination (r 2 ) of 0.85 for both phycoerythrin and Chl a. To explore the functional cause of diurnal variation in Synechococcus elemental stoichiometry, a cellular model was parameterized to mimic the chemostat culture conditions [Supplementary data, Table SI, Equation (S1); Talmy et al., 2014]. Our model was able to recapitulate the bimodal pattern that was observed for C:N cell and C:P cell over the course of two 24 h cycles (Fig. 3). Furthermore, our assumption of constant N:P cell during diel cycles yielded suitable fits to observations (Fig. 3B and E). Because cells were limited by N throughout the experiment, the intracellular concentration of reserve, inorganic nitrogen (N R ) and the N concentration in the chamber (S) were both low by comparison to all other modeled biomass variables (C R , N F ,C F , Table S2; data not shown). Thus, the stoichiometric variation observed in the model arises due to changes in the functional and structural apparatus and the carbon reserve.
We present results from our model using two contrasting assumptions regarding the carbon loss rate. First, when the main carbon loss is assumed to be a function of growth rate, the parameter R, which constrains growth-rate-independent carbon losses, was set to an arbitrarily low number for the duration of the experiment. With this assumption, the amplitude of C:N model and C:P model depends strongly on the dilution rate within the chamber (Fig. 3A and C). At the lowest dilution rates, the model overestimates C:N cell and C:P cell at dawn, which suggests that there is an additional night time loss of carbon, which is not accounted for in this configuration. In contrast, when we assumed a high night time loss of carbon (Supplementary data,  (S9)], the model produces a bimodal pattern in C:N cell and C:P cell with amplitude similar to the observational data, even for different dilution rates (Fig. 3D and F). Under this assumption, the C:N model and C:P model at dawn are much closer to the observed values ( Fig. 3D and  F). Thus, our model and data suggest that the carbon loss rate at night is high and could be due to respiration or carbon purging.

D I S C U S S I O N
The plasticity in cellular quotas between light and dark periods resulted in pronounced deviation from Redfield stoichiometry. Consistent with previous work showing the accumulation of glycogen throughout light periods and subsequent use to maintain essential chemical processes throughout dark periods (Smith, 1983), Q C varied substantially during each 24 h period and was synchronized with the 14:10 light -dark cycle. In addition, Q C was positively related to the growth rate and significantly correlated with cell size, as documented previously (Garcia et al., in review). Because of lower fluctuations in Q N and Q P (Fig. 1B), N:P cell was not highly variable (Fig. 3E) and Q C was responsible for the majority of the diel variation in C:N cell and C:P cell in Synechococcus (Fig. 3D and F).
Due to their small cell size, Synechococcus is expected to have relatively limited diel variation in elemental stoichiometry (Talmy et al., 2014). When averaged over the diurnal cycle, Synechococcus C:N cell and C:P cell do indeed vary over a relatively limited range [e.g. C:N cell between approximately 6 and 8 (by moles) and C:P cell between approximately 100 and 140] in comparison to that observed in other, nutrient-limited eukaryotes (e.g. C:N cell between approximately 6 and 20 and C:P cell between approximately 100 and 600; Caperon and Meyer, 1972;Laws and Wong, 1978;Laws and Bannister, 1980;Elrifi and Turpin, 1985;Geider et al., 1998). Surprisingly, however, Synechococcus does have high diurnal variation in C:N cell and C:P cell (Fig. 3), similar to variable elemental stoichiometry in N 2 -fixing cyanobacteria, which temporally separate C and N 2 fixation between light and dark periods (Mohr et al., 2010;Dron et al., 2012). This relationship among Q C , cell size and growth rate has consequences not only for field measurements of stoichiometry, but also for growth rates, as diel variability in cell size has been used to estimate growth rates of Synechococcus (Hunter-Cevera et al., 2014).
As suggested by previous modeling and empirical studies (Pirt, 1982;Vincent, 1992;Clark et al., 2002;Flynn et al., 2002), we suspect that the temporal decoupling of cellular N and P from C (Q C peaks are separated from Q N and Q P peaks by several hours; Fig. 1) could be due in part to the utilization of accumulated carbon reserves (Fig. 1A) to drive N and P uptake during the dark hours, particularly when cells are growing fast. Although some data suggest that nitrate uptake in Synechococcus is restricted to the light period (Paerl et al., 2012), the growth rate differential could modulate light and dark uptake of nitrate. For example, although slow-growing cells may rely on the energetically favorable day-time N acquisition, fast-growing cells may need to acquire N during the day and during the night to keep up with the high N demand. Certainly, the growth rate differential is responsible for modulating utilization of different N sources in other organisms within cyanobacteria that fix N 2 (Garcia and Hutchins, 2014), a trade-off that is also thought to be thermodynamically controlled. Just as other organisms within cyanobacteria (such as Crocosphaera) rely on carbon reserves to supply energy needed to assimilate N 2 during the dark period, we recognize the possibility that Synechococcus acquires nitrate during the dark period when the N demand for growth is high. The pre-dawn rise in Q N and Q P could be associated with dark N and P acquisition, and our model parameterizes the associated energy costs as respiration of reserve carbon (Supplementary data, Equation  S3). This perspective presumes that stoichiometric variation within phytoplankton could depend on the thermodynamics of nutrient assimilation. For example, ammonium assimilation is presumed to use less energy relative to nitrate assimilation, possibly influencing stored carbon utilization rates. Thus, the temporal decoupling of Q C , Q N and Q P could result from combined effects of timing in cell division and nutrient acquisition and may depend on nutrient source.
Previous models of elemental stoichiometry have assumed that the carbon loss rate is dependent primarily on the growth rate, with other carbon losses negligible (Geider et al., 1998;Pahlow, 2005;Bonachela et al., 2012;Talmy et al., 2014). Using this assumption, the model was able to reproduce the general bimodal pattern in C:N cell and C:P cell that emerges over our sampling periods ( Fig. 3A and C), but fails to adequately reproduce the amplitude of fluctuation in different dilution rates. When it was assumed that dark-period carbon losses are high (4.0 day 21 ), however, the model was able to reproduce the full amplitude of diurnal changes in C:N cell and C:P cell ( Fig. 3D and F). The improved model fits suggest that Synechococcus purges carbon at night, through either respiration or excretion.
We suggest three possible mechanisms describing how Synechococcus might benefit from purging excess carbon. The first is that carbon accumulated during the light period is used as energy for respiratory costs associated with repair of core functional apparatus that may have been damaged during the photoperiod (Raven, 1989;Long et al., 1994;Vasilikiotis and Melis, 1994;Talmy et al., 2013;Talmy et al., 2014). The second possibility is related to cell size. In the oligotrophic gyres, there is strong evolutionary pressure to optimize the surface-area-to-volume ratio through small cell size . Excess carbon in the form of carbohydrate and lipids may take up valuable space within the cell. Synechococcus may have evolved a night-time carbon excretion mechanism to enable it to shrink to the smallest possible size, even when growth in a high-light environment necessitates the accumulation of excess carbohydrates. Finally, the mechanism could be important for ocean ecosystem functioning, as excretion of energy-rich carbon polymers is likely to play a role in algal-heterotrophic interactions (Grossart and Simon, 2007). Excretion of carbon polymers could stimulate bacterial growth and increase nutrient remineralization rates-a potentially crucial adaptation to chronic nutrient limitation.
In addition to cell quotas and elemental ratios, we also observed oscillation in pigment fluorescence ( Fig. 3A and  B). Although Q C and cell size are directly linked to light availability (Vaulot et al., 1995;Sato et al., 2015) and co-oscillate in our data (Figs 1 and 2), the variation in pigment fluorescence is more tightly correlated with variation in Q N (Fig. 4C and D). Pigments change over a diel cycle in natural phytoplankton populations (Sester et al., 1982;Sato et al., 2015), and phycoerythrin is known to be a nitrogen storage compound in Synechococcus (Wyman et al., 1985;Yeh et al., 1986).
Diel patterns in elemental stoichiometry may also be linked to patterns in gene expression of different proteins (Stöckel et al., 2011;Mella-Flores et al., 2012;Diamond et al., 2015). However, the large variation in C:N cell and C:P cell data indicates that Q C is dominant whereas Q N and Q P are minor drivers. This leaves few candidate molecules for driving diel patterns in elemental ratios, and direct measurements of glycogen have been shown to cycle throughout the day in other cyanobacteria (Saito et al., 2011). The relatively invariable N:P cell suggests that nucleic acids may be responsible for the minor variation in Q N and Q P as proteins do not typically contain a large quantity of P. Cellular nucleic acids are known to increase in response to growth (Churchward et al., 1982;Sterner and Elser, 2002;Garcia et al., in review) and cycle with a daily rhythm (Vaulot et al., 1995). Direct measurements of specific compounds are needed, however, especially as there are no data describing diel patterns in phytoplankton polyphosphates.

C O N C L U S I O N
Our data show pronounced diel oscillations in the elemental quotas of Synechococcus with variability largely depending on the growth rate (Figs 1 and 3) and carbon accumulation driving the majority of variable C:P cell and C:N cell ratios. The variability in C:N cell is larger than expected based on models of variability among phytoplankton groups, but still smaller than the variability observed in large eukaryotic phytoplankton. Our model suggests that Synechococcus purges excess carbon during the dark period of a 24 h cycle. The variability observed over a single day is comparable to that over large-scale environmental regimes and should be further investigated in the field for its implications on modeling biogeochemical processes. Differences in diel stoichiometric variability between groups of large and small phytoplankton may contribute to competitive niches within large biogeographical domains.