Combined Effects of Ocean Acidification and Light or Nitrogen Availabilities on 13C Fractionation in Marine Dinoflagellates

Along with increasing oceanic CO2 concentrations, enhanced stratification constrains phytoplankton to shallower upper mixed layers with altered light regimes and nutrient concentrations. Here, we investigate the effects of elevated pCO2 in combination with light or nitrogen-limitation on 13C fractionation (εp) in four dinoflagellate species. We cultured Gonyaulax spinifera and Protoceratium reticulatum in dilute batches under low-light (‘LL’) and high-light (‘HL’) conditions, and grew Alexandrium fundyense and Scrippsiella trochoidea in nitrogen-limited continuous cultures (‘LN’) and nitrogen-replete batches (‘HN’). The observed CO2-dependency of εp remained unaffected by the availability of light for both G. spinifera and P. reticulatum, though at HL εp was consistently lower by about 2.7‰ over the tested CO2 range for P. reticulatum. This may reflect increased uptake of (13C-enriched) bicarbonate fueled by increased ATP production under HL conditions. The observed CO2-dependency of εp disappeared under LN conditions in both A. fundyense and S. trochoidea. The generally higher εp under LN may be associated with lower organic carbon production rates and/or higher ATP:NADPH ratios. CO2-dependent εp under non-limiting conditions has been observed in several dinoflagellate species, showing potential for a new CO2-proxy. Our results however demonstrate that light- and nitrogen-limitation also affect εp, thereby illustrating the need to carefully consider prevailing environmental conditions.


Introduction
Anthropogenic activities have caused the partial pressure of CO 2 (pCO 2 ) in the atmosphere and oceans to increase at an unprecedented rate [1]. This will shift marine carbon speciation towards increasing CO 2 and bicarbonate (HCO 3 -) concentrations, and decreasing carbonate ion (CO 3 2-) concentration and pH [2]. Along with these changes in carbonate chemistry, global temperatures are expected to rise by 2 to 6°C within this century [1], which likely leads to enhanced (thermal) stratification for most oceanic regions [3]. Enhanced stratification can cause primary production to decrease, as observed in low-latitude oceans [4], where the mixed layer depth is already relatively shallow and upwelling of nutrient-rich deeper water masses is suppressed. Alternatively, enhanced stratification may increase primary production in regions with deep mixed layer depths, such as in high latitude oceans. At such locations, phytoplankton may be light-limited due to the deep convective turnover [5]. Irrespective of the net effect on primary production, shoaling of the thermocline causes phytoplankton to be more often restricted to the upper layers of the water column, characterized by high irradiance and low nutrient concentrations [6]. Such changes in light intensity and nutrient concentration may affect marine phytoplankton, including dinoflagellates.
Dinoflagellates are unicellular eukaryotes and can reach high densities under favorable environmental conditions, which may lead to harmful algal blooms with adverse effects not only for the aquatic food web, but also for human health (e.g. [7; 8]). Strategies that add to their success include toxin production, allelopathy, mixotrophy and cyst formation [9; 10; 11; 12]. While studies have investigated how dinoflagellates are influenced by changes in pH and/or pCO 2 [13; 14; 15; 16; 17], less is known about the combined effects of CO 2 and light availabilities (as daylength, see [18]) or CO 2 and nitrogen-limitation [19]. Like all phytoplankton, dinoflagellates fix CO 2 with the carboxylation enzyme Ribulose-1,5-bisphosphate Carboxylase/ Oxygenase (RubisCO), which discriminates between carbon isotopes, favoring 12 C over 13 C (e.g. [20; 21; 22]). The inorganic carbon (C i ) species taken up by phytoplankton differ in their isotopic composition, with CO 2 being 13 C-depleted compared to HCO 3 -. Under elevated CO 2 concentrations, dinoflagellates may take up relatively more CO 2 , resulting in higher 13 C fractionation (ε p ) [23; 14]. Similarly, high CO 2 efflux:total C i uptake (i.e. leakage) prevents the accumulation of 13 C within the intracellular carbon pool, thereby increasing ε p [23; 14]. Indeed, ε p values in different phytoplankton groups, including dinoflagellates, were shown to increase with elevated pCO 2 [18; 24; 14; 25; 17]. Organic dinoflagellate cysts are ubiquitously preserved in marine sediments (e.g. [26]). The CO 2 dependency of their isotopic composition may be reflected in their cysts, thus potentially providing a proxy for past CO 2 concentrations. However, the CO 2 dependency in ε p may be affected by other environmental conditions, such as the availability of light and nutrients (e.g. [27; 28; 29; 30]). Here, we investigate the combined effects of elevated pCO 2 and low-light conditions or nitrogen-limitation on particulate organic carbon (POC) production (μ c ), Chlorophyll-a (Chl-a):POC ratios and ε p in four marine dinoflagellate species. We grew Gonyaulax spinifera and Protoceratium reticulatum under low-light conditions (LL) and Alexandrium fundyense and Scrippsiella trochoidea under nitrogen-limiting conditions (LN) and compared these responses to results from an earlier study, where the same species were grown under high-light and nitrogen-replete conditions (HL and HN).
As growth medium, filtered North Sea seawater (cellulose acetate membrane, 0.2 μm pore size, Sartorius, Göttingen, Germany) with a salinity of 34 and enriched with 100 μmol L -1 nitrate and 6.25 μmol L -1 phosphate was used. FeCl 3 (1.9 μmol L -1 ), H 2 SeO 3 (10 nmol L -1 ) and NiCl 2 (6.3 nmol L -1 ) were added according to K medium [33], and metals and vitamins were added according to f/2 medium [34]. Bottles were placed on a roller table in order to avoid sedimentation. Daylight tubes (Lumilux HO 54W/965, Osram, München, Germany) provided incident light intensities of 250 ± 25 μmol photons m -2 s -1 at a 16:8 h light:dark cycle. In order to determine the carbonate chemistry, pH was measured every other day using a WTW 3110 pH meter equipped with a SenTix 41 Plus pH electrode (Wissenschaftlich-Technische Werkstätten GmbH, Weilheim, Germany), which was calibrated prior to each measurement to the National Bureau of Standards (NBS) scale. The precision of pH measurements during the experiments was ±0.02 units. Cells were acclimated to the pCO 2 treatments for at least 7 generations (i.e. >21 days) prior to each experiment.
For the low-light treatments, the same conditions as the nutrient-replete dilute batch conditions were applied, except that incident light intensities were reduced to 55 ± 5 μmol photons m -2 s -1 . In these incubations, CO 2 concentrations ranged between ca. 16 and 50 μmol L -1 , according to pCO 2 values of 380, 800 and 1200 μatm. Nitrogen-limited conditions were achieved in gently mixed continuous cultures [35]. Cultures were grown as chemostats with fixed dilution rates representing~33% of maximum growth for each species, with 0.15 ± 0.01 d -1 for A. fundyense and 0.2 ± 0.01 d -1 for S. trochoidea, yielding nitrate concentrations below 0.8 μmol L -1 for both species. In these incubations, CO 2 concentrations ranged between ca. 8 and 40 μmol L -1 , according to pCO 2 values of 220, 800 and 1000 μatm (A. fundyense), and 280, 590 and 770 μatm (S. trochoidea). Steady state was reached after 22-43 days of acclimation, and samples were taken during this phase over 4 consecutive sampling points with time intervals of 2-3 days. For more details on the setup of the continuous culture experiment we refer to Eberlein et al. [19].

Sampling and Analyses
For total alkalinity (TA) analysis, 50 mL culture suspension was filtered over cellulose acetate syringe filters (0.45 μm pore size, Thermo Scientific, Waltham, Massachusetts, USA) and stored in gas tight borosilicate bottles at 3°C. Samples were then analyzed in duplicates using an automated TitroLine burette system (SI Analytics, Mainz, Germany) with a precision of ±13 μmol L -1 . Certified Reference Materials (CRMs) supplied by A. Dickson (Scripps Institution of Oceanography, USA) were used to correct for inaccuracies of TA measurements. TA was measured at the beginning and the end of each experiment, and during steady-state conditions in the continuous cultures. Minor changes in TA over the course of the experiments combined with the pH measurements every other day allowed for a complete resolution of the carbonate chemistry. The carbonate chemistry was assessed with the program CO2sys [36] using TA and pH (following recommendations of Hoppe et al. [37]) as well as temperature, salinity and phosphate concentration. We used the dissociation constants of carbonic acid and sulfuric acid of Mehrbach et al. [38], refitted by Dickson and Millero [39] and Dickson [40], respectively.
Duplicate samples of 20 mL culture suspension were fixed with neutral Lugol's solution (2% final concentration) and counted every day or every other day with an inverted light microscope (Axiovert 40C, Zeiss, Germany). Growth rates during the exponential phase of growth were assessed separately for each biological treatment by fitting an exponential function through the cell numbers over time according to: with N referring to cell number per mL at time t in days, N 0 to the cell number per mL at the start of the experiment, and μ referring to the specific growth rate (d -1 ). At the end of the experiment, when cells where still in exponential growth, we took samples to analyze Chl-a, POC and its isotopic composition (δ 13 C POC ). For the analysis of Chl-a, duplicate samples of 200 mL of culture suspension were filtered over cellulose acetate filters (Whatman, Maidstone, UK). Filters were rapidly frozen in liquid nitrogen and stored at -80°C. Chl-a was extracted using 90% acetone with subsequent sonification for 0.5 min. Fluorescence was assessed using a TD-700 Fluorometer (Turner Designs, Sunnyvale, CA), and Chl-a concentrations were calculated according to Knap et al. [41]. To measure POC and PON quota and δ 13 C POC , 300-400 mL of culture suspension was filtered over pre-combusted GF/F filters (6 h, 500°C). Filters were stored in pre-combusted glass Petri dishes and 200 μL of HCl (0.2 mol L -1 ) was added to remove any inorganic carbon before they were dried overnight and stored at -25°C. POC quota and δ 13 C POC of dilute batch experiments were then measured in duplicate with an Automated Nitrogen Carbon Analyser mass spectrometer (ANCA-SL 20-20, SerCon Ltd., Crewe, UK), with a precision of ± 0.5 μg C and 0.3‰, respectively. POC and PON quota and δ 13 C POC of the continuous cultures were measured with a Delta S (Thermo) isotopic ratio mass spectrometer connected to an elemental analyzer CE1108 via an open split interface (Finnigan Conflow II). δ 13 C POC is reported relative to the Vienna PeeDee Belemnite standard (VPDB). μ c was calculated by multiplying μ with POC quota.
For isotopic measurements of the dissolved inorganic carbon (δ 13 C DIC ), 4 mL of culture suspension was sterile filtered over 0.2 μm cellulose acetate filters (Thermo Scientific, Waltham, Massachusetts, USA) and stored at 3°C. 0.7 mL of the filtrate was then transferred to 8 mL vials, which contained three drops of 102% H 3 PO 4 solution, and headspaces filled with helium. After equilibration, the isotopic composition in the headspace was measured using a Gas-Bench-II coupled to a Thermo Delta-V advantage isotope ratio mass spectrometer, with a precision of ±0.1‰. ε p was calculated relative to the isotopic composition of dissolved CO 2 in the water (δ 13 C CO2 ) with an equation modified after Freeman and Hayes [42]: In order to calculate the isotopic composition of CO 2 (δ 13 C CO2 ) from δ 13 C DIC , we calculated the isotopic composition of HCO 3 -(δ 13 C HCO3-) based on δ 13 C DIC according to a mass balance relation following Zeebe and Wolf-Gladrow [43] and the temperature-dependent fractionation factors between CO 2 and HCO 3 and CO 3 2-and HCO 3 -, as determined by Mook et al. [44] and Zhang et al. [45], respectively. For further details on the determination of carbon isotope fractionation we refer to Van de Waal et al. [25].

Statistical analysis
Shapiro-Wilk tests confirmed normality of the data. Linear regressions were used to determine the relations between the tested variables and CO 2 . Significant differences between CO 2 treatments were confirmed by one-way ANOVA followed by post hoc comparison of the means using the Tukey HSD (α = 0.05). A covariance analysis (ANCOVA) was used to determine homogeneity of slopes. When slopes were significantly different, i.e. when there were interactive effects of CO 2 with light or nitrogen, the Johnson-Neyman technique (J-N; Johnson and Neyman [46]) was applied to identify the range of CO 2 over which the investigated parameter was different. To improve the homogeneity of variances, as tested by Levene's test, we used log 10 transformed data for analysis of POC quota, μ c and Chl-a:POC ratios of G. spinifera, and for analysis of Chl-a:POC and ε p of S. trochoidea.

Results
Elevated pCO 2 and light availability In G. spinifera, μ c did not change with CO 2 availability under LL, but increased under HL, which was mainly driven by increased POC quota in the highest CO 2 treatment (Fig 1A; linear regression; R 2 = 0.60; P = 0.003) (see also [17]). Moreover, μ c was lower under LL (ANCOVA; P<0.001; 95% CI [-0.635; -1.026]), which was due to decreased POC quota in all CO 2 treatments, and due to lowered μ in all but the highest CO 2 treatment ( , respectively). Moreover, CO 2 and light availability showed interactive effects on the Chl-a:POC ratios (ANCOVA; F 1,20 = 9.453; P = 0.007 and F 1,19 = 9.149; P = 0.008, respectively). In other words, the effect of CO 2 depended on the light availability, with P. reticulatum showing a significant increase in Chl-a:POC ratios with CO 2 availability under LL only (linear regression; R 2 = 0.45; P = 0.05). Similarly, under LL Chl-a:POC ratios were significantly higher in the higher pCO 2 treatments of G. spinifera (ANOVA; P<0.05).

Production rates, quotas and stoichiometry
Our results show differential effects of elevated pCO 2 in combination with light availability on growth, POC quota, μ c and Chl-a:POC ratios in G. spinifera and P. reticulatum (Fig 1; Table 1). In G. spinifera, μ c increased with CO 2 under HL, but there was no sensitivity towards elevated pCO 2 under LL (Fig 1A). Low-light furthermore caused lowered POC quota and μ c , while μ remained unaffected (Fig 1A and 1B; Table 1). At the same time, Chl-a contents and Chl-a: POC ratios increased under LL (Fig 1C and 1D; Table 1). Such higher ratios are needed to Table 1. Overview of the growth parameters in the HL and LL treatments. Growth rate (μ, d -1 ), POC quota (pg C cell -1 ), Chl-a content (pg cell -1 ) and 13 C fractionation ε p (‰) of G. spinifera and P. reticulatum grown under high-light and low-light conditions. Values represent the mean of triplicate incubations (n = 3 ±SD). Superscript letters indicate significant differences between pCO 2 treatments (P<0.05). Superscript symbols refer to earlier published data in Hoins Table 2. POC production and Chl-a:POC ratios have previously been published in [15] and [18], and ε p in the HN treatments in Hoins et al. 2015. doi:10.1371/journal.pone.0154370.g002 Combined Effects of OA, Light and Nitrogen on Four Dinoflagellate Species capture more light, which is a general response of phytoplankton to light-limitation. For P. reticulatum, the low light conditions did not yield changes in POC quota, μ and μ c (Fig 1A and 1B; Table 1). This suggests a high flexibility of P. reticulatum to deal with low-light conditions. Cells did synthesize more Chl-a, thereby showing elevated Chl-a:POC ratios, which were apparently sufficient to compensate for the low-light conditions. Both species showed increasing Chl-a:POC ratios with increasing CO 2 availability when grown under low-light. This suggests that CO 2 influences the ability of cells to synthesize Chl-a, and therefore their ability to cope with low-light conditions. We observed generally minor effects of elevated pCO 2 under LN, while nitrogen-limitation alone exerted a much stronger control (Fig 2, Table 2). Specifically, μ c was lower in both A. fundyense and S. trochoidea, although POC quota in S. trochoidea grown under LN was significantly higher. Decreased μ c was mainly a result of low μ, i.e. the imposed dilution rate which was set at about 33% of the μ of the respective species obtained from experiments under replete conditions. Nitrogen-limitation was confirmed by the higher POC:PON ratios in S. trochoidea in all tested pCO 2 treatments, while POC:PON ratios of A. fundyense grown under LN were only higher in the lowest pCO 2 treatment (Table 2) [19]. The Chl-a:POC ratios in LN were comparable to HN in S. trochoidea, while in A. fundyense these Chl-a:POC ratios showed a CO 2 -dependent increase under LN, and only differed between HN and LN under low CO 2 concentrations. Thus, although nitrogen was limiting μ c and caused an increase in POC:PON ratios, this did not strongly affect the Chl-a:POC ratios.
Irrespective of the light intensity or nitrogen concentration, CO 2 effects on growth rates, POC quotas and POC production in our study were either absent or relatively minor, suggesting the presence of effective carbon concentrating mechanisms (CCMs). Dinoflagellates possess RubisCO type II with lowest CO 2 affinities compared to all other eukaryotic algae [47; 48], Table 2. Overview of the growth parameters in the HN and LN treatments. Growth rate (μ, d -1 ), POC quota (pg C cell -1 ), Chl-a (pg cell -1 ), POC:PON ratios (molar) and ε p (‰) of A. fundyense and S. trochoidea grown under nitrogen-replete conditions and nitrogen-limitation. Values represent the mean of duplicate incubations (n = 2 ±SD). Superscript letters indicate significant differences between pCO 2 treatments (ANOVA; P<0.05; only applied when n>2). Superscript symbols refer to earlier published data in Hoins et al. 2015  which make effective CCMs a prerequisite to maintain growth under low CO 2 concentrations. Indeed, earlier work has shown that A. fundyense and S. trochoidea are able to actively take up HCO 3 -, thus increasing their intracellular C i pool [16]. Additionally, high extracellular activities of carbonic anhydrase, the enzyme accelerating the otherwise slow interconversion between CO 2 and HCO 3 -, have been found in S. trochoidea [16]. Consequently, at least the investigated dinoflagellate species do not seem to be CO 2 -limited in any of tested CO 2 concentrations, irrespective of the light or nutrient supply, explaining why μ, POC quotas and μ c did not respond to elevated CO 2 concentrations.
In the cyanobacterium Trichodesmium and the coccolithophore Emiliania huxleyi, limitation by light has been shown to cause enhanced sensitivity towards elevated pCO 2 [49; 50]. The CO 2 -dependent stimulation of μ c was most pronounced under light-limitation, which was explained by larger CO 2 -dependent benefits due to the CCM down-regulation and thus energy reallocation under light-limitation. In the tested dinoflagellate species, however, μ c remained largely unaltered over the applied CO 2 range (Figs 1A, 1B, 2A and 2B). Yet, we observed a CO 2 -dependent increase in Chl-a:POC quota in G. spinifera and P. reticulatum grown under low-light. Thus, with elevated pCO 2 more energy is acquired via photosynthesis, while the same level of μ c is maintained. It is further conceivable that their CCMs are down-regulated with elevated pCO 2 , lowering the energetic costs for carbon acquisition. The likely higher availability of energy with elevated pCO 2 under low-light conditions, however, seems not to be allocated to μ c (Figs 1C, 1D and 2C). This suggests either a lower overall efficiency to convert energy to biomass under these conditions, or a shunting of energy to alternative processes not accounted for in our study. Similarly to the Chl-a:POC ratios in G. spinifera and P. reticulatum under low-light conditions, Chl-a:POC ratios in A. fundyense grown under nitrogen-limitation also increased at elevated CO 2 concentrations. When grown under nitrogen-limitation, excess energy from a down-regulation of CCMs may be shunted to nitrogen acquisition. Indeed, POC:PON ratios decreased under elevated pCO 2 for both A. fundyense and S. trochoidea (Table 2) (see also [19]). Such lower POC:PON ratios (i.e. relatively more nitrogen) may favor synthesis of nitrogen-rich biomolecules such as Chl-a. Overall, elevated pCO 2 seems to have only minor effects on growth and μ c in the tested dinoflagellates, and yet it apparently causes intracellular shifts in energy and resource allocation under light-or nitrogen-limited conditions.

C fractionation
The 13 C fractionation of phytoplankton is influenced by the interplay between 1) CO 2 supply, 2) inorganic carbon demand (i.e. μ c ), and 3) active uptake of CO 2 and HCO 3 -(i.e. CCMs). If CO 2 supply in the growth medium increases, ε p increases because more of the 13 C-depleted CO 2 may be taken up in comparison to the 13 C-enriched HCO 3 -. In contrast, ε p may decrease with increasing μ c as CO 2 is fixed at a higher rate than total carbon is taken up, and the ability of RubisCO to express its full preference for 12 CO 2 is reduced. CCMs can influence ε p in various ways, e.g. as they determine the relative uptake of CO 2 and HCO 3 as well as leakage (Fig   3). Under HL and HN conditions, ε p shows a clear increase with increasing CO 2 concentrations in all four tested dinoflagellate species ([17]; Figs 1 and 2). Under LL, similar CO 2 dependencies were observed, although ε p shifted to higher values in P. reticulatum. Under LN, ε p was not CO 2 sensitive, and remained relatively high also at lower CO 2 concentrations for both A. fundyense and S. trochoidea. Light-or nutrient-limitation cause changes in the availability of energy (ATP) and reductants (NADPH) that in turn may affect μ c and CCM activity, eventually influencing ε p (Fig 3). Under low-light conditions, for instance, less photons arrive at the photosystems, thereby  (Fig 3B). The lowered electron and proton fluxes then result in lower amounts of ATP and NADPH. ATP is required to operate the energetically costly CCMs, while both ATP and NADPH are required for CO 2 reduction in the Calvin Cycle to produce biomass, and for reducing nitrate (NO 3 -) to ammonium (NH 4 + ) to eventually produce particulate organic nitrogen (PON). Thus, differences in the availability of light but also nitrogen alter the availability of ATP and NADPH, which may be one reason for the differences in ε p responses between types of incubations (e.g. [27; 51; 52; 30]). High-light intensities may provide the cells with more energy than required for CO 2 fixation, which will enhance the active uptake of C i that in turn serves as an energy sink for excess light [53]. Depending on how much C i is taken up in relation to the amount of CO 2 that is fixed, a high C i uptake may be accompanied by a high leakage [54]. A high C i uptake by G. spinifera at both LL and HL, in concert with high leakage, would explain its relatively high ε p . In contrast to our expectations, however, ε p in P. reticulatum was substantially lower under HL conditions. In this species, an increasing contribution of energetically costly HCO 3 uptake under HL may support the dissipation of excess energy, avoiding damage to photosystem II. If this active HCO 3 uptake does not lead to higher leakage, it could in fact explain the lower ε p under HL. Comparable to light, also nitrogen availability may alter ε p as it indirectly changes cellular energy budgets (Fig 3). As mentioned, NADPH is used to reduce CO 2 to organic carbon, and NO 3 to NH 4 + . As a consequence, less NADPH is needed when μ c is low and/or when NO 3 is limiting. Under these conditions, cyclic electron flow "around" photosystem I may be up-regulated, thereby circumventing NADPH production while maintaining ATP generation (Fig 3C). Such a putative excess of ATP over NADPH, in turn, may be used for active inorganic carbon uptake. As ε p in both A. fundyense and S. trochoidea was higher under nitrogen-limitation, CO 2 and not HCO 3 may have been taken up actively. Alternatively, increasing overall C i uptake despite low μ c may have increased leakage and thus ε p . Nonetheless, even the highest ε p of~14‰ in our study was low compared to earlier studies investigating the effect of nitrogenlimitation on ε p in other algal species [27; 51; 28; 55]. This is in line with the generally high uptake of HCO 3 observed in earlier studies on CCMs in dinoflagellates [14; 16]. Moreover, maximum 13 C fractionation of RubisCO in the tested dinoflagellate species may be lower than the typical 24-30‰, as was also found for a RubisCO isolated from E. huxleyi (i.e. 11‰; [56]).

Proxy development
The CO 2 -dependency of ε p in dinoflagellates can potentially serve in the development of a proxy for past pCO 2 in the atmosphere [17]. As indicated before, however, additional experiments focusing on environmental variables other than pCO 2 , physiological underpinning of the recorded response, quantification of fractionation between dinoflagellate cells and cysts, as well as field calibration studies are required to establish a reliable proxy [17]. Here, we investigated the possible role of environmental variables other than pCO 2 , including light and nitrogen availability.
Electrons are transported by plastohydroquinone (PQ), thereby pumping more protons into the lumen. The cytochrome b 6 f complex oxidizes PQ molecules, thereby producing electrons, which are then transported to Photosystem I (PSI) where they reduce NADP + to NADPH. Protons are transported to F-ATPase to synthesize ATP. (B) Under light-limitation, the overall decreased amount of energy arriving at PSII causes a decrease in water oxidation, thereby producing less electrons and protons, and thus also less ATP and NADPH. (C) Under nitrogen-limitation, less NADPH is needed for NO 3 reduction, thus the excess electrons are transported back to PSII by cyclic energy flow. Protons are still pumped by F-ATPase, thereby increasing the amount of ATP synthesized. The results show that under low-light conditions, the general response of ε p towards elevated pCO 2 remains largely unaltered in G. spinifera and P. reticulatum, i.e. slopes remained largely similar. In contrast, ε p becomes insensitive to changes in CO 2 under nitrogen-limitation in A. fundyense and S. trochoidea. Elevated pCO 2 in the past was presumable accompanied by water column stratification, thereby not only affecting the water depth at which dinoflagellates fixed carbon, but also the potential upwelling of nutrient-rich deeper water masses. Consequently, it is crucial to take into account the light conditions and nutrient concentrations during the dinoflagellate lifetime.
Application of an eventual proxy based on dinoflagellate ε p would likely be most valuable at study sites where nitrate concentrations are non-limiting and stable through time. For such settings, the equilibrium between dissolved (recorded in dinoflagellates) and atmospheric (the proxy target) pCO 2 is typically sub-optimal. This results in an interesting paradox since study sites are required for which CO 2 is equilibrated between the ocean and atmosphere, and also bear sufficient nutrients to force a CO 2 response in ε p . Moreover, intense blooms of dinoflagellates may deplete seawater not only in CO 2 [57; 58], but also in nutrients, leading to a potential bias in ε p .
Thus, although ε p shows largely consistent CO 2 dependencies across four tested dinoflagellate species under optimal growth conditions [16], other environmental factors, notably nitrogen limitation, complicate and possibly negate the suitability of dinoflagellate ε p as a proxy for past pCO 2 .