CO₂-dependent carbon isotope fractionation in dinoflagellates relates to their inorganic carbon fluxes

Abstract

Carbon isotope fractionation (ε_p) between the inorganic carbon source and organic matter has been proposed to be a function of pCO₂. To understand the CO₂-dependency of ε_p and species-specific differences therein, inorganic carbon fluxes in the four dinoflagellate species Alexandrium fundyense, Scrippsiella trochoidea, Gonyaulax spinifera and Protoceratium reticulatum have been measured by means of membrane-inlet mass spectrometry. In-vivo assays were carried out at different CO₂ concentrations, representing a range of pCO₂ from 180 to 1200 μatm. The relative bicarbonate contribution (i.e. the ratio of bicarbonate uptake to total inorganic carbon uptake) and leakage (i.e. the ratio of CO₂ efflux to total inorganic carbon uptake) varied from 0.2 to 0.5 and 0.4 to 0.7, respectively, and differed significantly between species. These ratios were fed into a single-compartment model, and ε_p values were calculated and compared to carbon isotope fractionation measured under the same conditions. For all investigated species, modeled and measured ε_p values were comparable (A. fundyense, S. trochoidea, P. reticulatum) and/or showed similar trends with pCO₂ (A. fundyense, G. spinifera, P. reticulatum). Offsets are attributed to biases in inorganic flux measurements, an overestimated fractionation factor for the CO₂-fixing enzyme RubisCO, or the fact that intracellular inorganic carbon fluxes were not taken into account in the model. This study demonstrates that CO₂-dependency in ε_p can largely be explained by the inorganic carbon fluxes of the individual dinoflagellates.

Highlights

• •To understand ¹³C fractionation in dinoflagellates, inorganic carbon fluxes were measured by MIMS and used for modeling.
• •Changes in cellular carbon fluxes, i.e. HCO₃⁻ contribution and CO₂ leakage were CO₂-dependent.
• •These CO₂-dependencies could largely explain the CO₂-dependent fractionation patterns observed in the four tested species.

1. Introduction

During photosynthetic carbon fixation, the lighter carbon isotope ¹²C is preferred over the heavier carbon isotope ¹³C, thereby causing carbon isotope fractionation (ε_p) between the inorganic carbon (C_i) source and the organic carbon. Values for ε_p of marine phytoplankton have been shown to be CO₂-sensitive (e.g. Degens et al., 1968), and thus were discussed to serve as a proxy for past CO₂ concentrations (Jasper and Hayes, 1990, Pagani, 2014, Van de Waal et al., 2013, Hoins et al., 2015). Large species-specific differences in ε_p have been described, which are yet poorly understood (e.g. Hinga et al., 1994, Burkhardt et al., 1999). Moreover, irrespective of the phytoplankton species investigated, most of these studies have solely described the relationship between ε_p and CO₂, and only few have investigated the underlying physiological processes. Such mechanistic understanding is, however, needed to identify the reasons of the CO₂-dependency of ε_p.

Carbon isotope fractionation of phytoplankton is primarily driven by the enzyme ribulose-1,5-bisphosphate Carboxylase/Oxygenase (RubisCO), which is responsible for the fixation of CO₂ into organic compounds. The intrinsic fractionation associated with RubisCO (ε_f) has been estimated to range between ~ 22 and 30‰ (e.g. Roeske and O'Leary, 1984, Guy et al., 1993, Scott et al., 2007), even though a recent study obtained values as low as 11‰ for the RubisCO of the coccolithophore Emiliania huxleyi (Boller et al., 2011). While RubisCO principally sets the upper limit of fractionation, other processes strongly determine the degree to which RubisCO can express its fractionation (Sharkey and Berry, 1985, Burkhardt et al., 1999, Rost et al., 2002). First, there is leakage, i.e. the amount of CO₂ diffusing out of the cell in relation to C_i uptake. With higher leakage, the intracellular C_i pool is ‘refreshed’, thereby preventing accumulation of ¹³C and allowing RubisCO to approach its upper fractionation values. Second, the relative contribution of bicarbonate (HCO₃⁻) to total C_i uptake plays a role, as HCO₃⁻ is enriched in ¹³C by ~ 10‰ relative to CO₂ (Mook et al., 1974). An increasing HCO₃⁻ contribution thus lowers ε_p. The enzyme carbonic anhydrase (CA), which accelerates the otherwise slow interconversion between CO₂ and HCO₃⁻, can also influence ε_p under certain conditions, e.g. by influencing leakage as well as the relative HCO₃⁻ contribution. All these processes play a role in the CO₂-concentrating mechanisms (CCMs) of phytoplankton. Assessing the mode of CCMs may therefore help to understand the reasons for CO₂-dependent changes in ε_p and species-specific differences therein.

Dinoflagellates are cosmopolitan unicellular algae that occur in many different environments, including eutrophic coastal regions and oligotrophic open oceans. In this study, we investigated whether the CO₂-dependency of ε_p, which was found in the dinoflagellate species Alexandrium fundyense, Gonyaulax spinifera, Protoceratium reticulatum and Scrippsiella trochoidea (Burkhardt et al., 1999, Hoins et al., 2015), can be explained by changes in their C_i fluxes. Characteristics of CCMs in the tested species, including their CA activities and C_i fluxes, were measured by means of membrane-inlet mass spectrometry (MIMS). Results were fed into a single-compartment model that considers cellular leakage, the relative HCO₃⁻ contribution as well as the carbon isotope fractionation of RubisCO (Sharkey and Berry, 1985, Burkhardt et al., 1999). The calculated carbon fractionation (ε_p-mod) was then compared to the measured carbon fractionation (ε_p-meas).

2. Material and methods

2.1. Incubations

Cultures of the dinoflagellate species A. fundyense (formerly Alexandrium tamarense strain Alex5; John et al., 2014), S. trochoidea (strain GeoB267; culture collection of the University of Bremen), G. spinifera (strain CCMP 409) and P. reticulatum (strain CCMP 1889) were grown in 0.2 μm filtered North Sea water (salinity 34), which was enriched with 100 μmol L^− 1 nitrate and 6.25 μmol L^− 1 phosphate. Metals and vitamins were added according to f/2 medium (Guillard and Ryther, 1962), except for FeCl₃ (1.9 μmol L^− 1), H₂SeO₃ (10 nmol L^− 1) and NiCl₂ (6.3 nmol L^− 1) that were added according to K medium (Keller et al., 1987). Each of the strains was grown in 2.4 L air-tight borosilicate bottles at 15 °C and 250 ± 25 μmol photons m^− 2 s^− 1 at a 16:8 h light:dark cycle. Bottles were placed on roller tables in order to avoid sedimentation.

Dissolved CO₂ concentrations ranged from ~ 5–50 μmol L^− 1 and were reached by pre-aerating culture medium with air containing 180, 380, 800 and 1200 μatm pCO₂. The carbonate chemistry was calculated based on pH and total alkalinity (TA), using the program CO2sys (Pierrot et al., 2006). pH values were measured using a WTW 3110 pH meter equipped with a SenTix 41 Plus pH electrode (WTW, Weilheim, Germany), which was calibrated prior to measurements to the National Bureau of Standards (NBS) scale. An automated TitroLine burette system (SI Analytics, Mainz, Germany) was used to determine TA. Dissolved inorganic carbon (DIC) was determined colorimetrically using a QuAAtro autoanalyser (Seal Analytical, Mequon, USA). For more details on the carbonate chemistry in the acclimations, please refer to Eberlein et al. (2014) for A. fundyense and S. trochoidea and to Hoins et al. (2015) for G. spinifera and P. reticulatum.

To determine ε_p values, the isotopic composition of the organic material was measured using an Automated Nitrogen Carbon Analyser mass spectrometer (ANCA-SL 20–20, SerCon Ltd., Crewe, UK), and the isotopic composition of the DIC in growth medium was measured using a GasBench-II coupled to a Thermo Delta-V advantage isotope ratio mass spectrometer (see Hoins et al., 2015 for details on isotope analysis). Prior to assays, cells were acclimated to the different CO₂ concentrations for at least 7 generations (i.e. > 21 days). To prevent changes in the carbonate chemistry, i.e. keeping drawdown of DIC < 3%, incubations were terminated at low cell densities (< 400 cells mL^− 1).

2.2. MIMS assays

A custom-made membrane-inlet mass spectrometer (MIMS; Isoprime, GV Instruments, Manchester, UK; see Rost et al., 2007 for details) was used to determine CA activities and C_i fluxes of A. fundyense and S. trochoidea acclimated to four different pCO₂ (i.e. 180, 380, 800 and 1200 μatm; Eberlein et al., 2014), and of G. spinifera and P. reticulatum acclimated to a low and high pCO₂ (i.e. 180 and 800 μatm). Assays were performed in an 8 mL temperature-controlled cuvette, equipped with a stirrer. Assay tests over ~ 1 h confirmed that conditions during the assay do not cause physiological stress (i.e. no decline in O₂ production rates), and subsequent microscopic inspection did not reveal any visual effects on cell morphologies. Prior to the measurements, acclimated cells were concentrated using a 10 μm membrane filter (Millipore, Billerica, MA) by gentle vacuum filtration (< 200 mbar) and stepwise transferred into C_i-free medium buffered with a 4-(2-hydroxylethyl)-1-piperazine-ethanesulfonic acid (50 mmol^− 1 HEPES) solution at 15 ± 0.3 °C and a pH of 8.0 ± 0.1. Chlorophyll a (Chl-a) concentrations were determined fluorometrically by using a TD-700 Fluorometer (Turner Designs, Sunnyvale, CA, USA) and ranged between 0.15 and 1.70 μg mL^− 1 during the assays.

To quantify activities of extracellular CA (eCA), the ¹⁸O depletion rate of doubly labeled ¹³C¹⁸O₂ in seawater was determined by measuring the transient changes in ¹³C¹⁸O¹⁸O (m/z = 49), ¹³C¹⁸O¹⁶O (m/z = 47) and ¹³C¹⁶O¹⁶O (m/z = 45) in the dark, following the approach of Silvermann (1982). If cells possess eCA, exchange rates of ¹⁸O are accelerated relative to the spontaneous rate. To monitor the spontaneous rate, NaH¹³C¹⁸O₃ label was injected to the cuvette, waiting until the m/z = 49 signal reached a steady-state decline. This rate was then compared to the steady-state decline after cells were added. Following Badger and Price (1989), eCA activity is expressed as percentage decrease in ¹⁸O-atom fraction upon the addition of cells, normalized to Chl-a. Consequently, 100 units (U) correspond to a doubling in the rate of interconversion between CO₂ and HCO₃⁻ per μg Chl-a.

Photosynthetic O₂ and C_i fluxes were determined following Badger et al. (1994). Making use of the chemical disequilibrium, this approach estimates CO₂ and HCO₃⁻ fluxes during steady-state photosynthesis. It is based on the simultaneous measurements of O₂ and CO₂ concentrations during consecutive light and dark intervals with increasing amounts of DIC. Oxygen fluxes in the dark and light are converted into C_i fluxes by applying a respiratory quotient of 1.0 and a photosynthetic quotient of 1.1 (Burkhardt et al., 2001, Rost et al., 2003). The light intensity in the cuvette was adjusted to the acclimation conditions (i.e. 250 ± 25 μmol photons m^− 2 s^− 1). Net CO₂ uptake was calculated from the steady-state decline in CO₂ concentration at the end of the light period, corrected for the interconversion between CO₂ and HCO₃⁻. The uptake of HCO₃⁻ was calculated by subtracting net CO₂ uptake from net C_i uptake, and the CO₂ efflux from the cells was estimated from the initial slope after turning off the light. Rate constants k₁ and k₂ were determined based on temperature, salinity and pH (Zeebe and Wolf-Gladrow, 2001, Schulz et al., 2006), yielding mean values of 0.9241 (± 0.0506) min^− 1 and 0.0085 (± 0.0008) min^− 1, respectively. To eliminate any eCA activity, a prerequisite to apply the rate constants, we added dextran-bound sulfonamide (DBS; 50 μmol L^− 1) to the cuvette. For more details on the calculations, please refer to Badger et al. (1994) and Schulz et al. (2007).

2.3. Single-compartment model

To calculate ε_p-mod, results for the relative HCO₃⁻ contribution and leakage were fed into a single-compartment model after Sharkey and Berry (1985) and Burkhardt et al. (1999):

$${\mathrm{\epsilon }}_{\mathrm{p}-mod}={\mathrm{R}}_{\mathrm{HCO}3}\times {\mathrm{\epsilon }}_{\mathrm{s}}+{\mathrm{L}}_{\mathrm{CO}2}\times {\mathrm{\epsilon }}_{\mathrm{f}}$$ (1)

where R_HCO3 represents the ratio of HCO₃⁻ to total C_i uptake, ε_s the equilibrium fractionation between CO₂ and HCO₃⁻ (− 10‰; Mook et al., 1974), L_CO2 the ratio of CO₂ efflux relative to total C_i uptake, and ε_f the kinetic fractionation associated with the CO₂ fixation of RubisCO, which was here assumed to be 28‰ after Raven and Johnston (1991).

2.4. Statistical analysis

Shapiro–Wilk tests confirmed normality of the data. Significant differences between CO₂ treatments were confirmed by a one-way ANOVA followed by post hoc comparison of the means using the Tukey HSD (α = 0.05; Table 1).

Table 1.

Experimental conditions in dilute batch culture incubations (see also Eberlein et al., 2014, Hoins et al., 2015): average CO₂ concentrations (μmol L^− 1), total alkalinity (TA; μmol L^− 1), dissolved inorganic carbon (DIC; μmol L^− 1) and pH (NBS scale). HCO₃⁻ contribution, leakage, modeled carbon isotope fractionation (ε_p-mod) and measured carbon isotope fractionation (ε_p-meas) was derived under the same conditions.

pCO2 μatm	CO2 μmol L− 1	TA μmol L− 1	DIC μmol L− 1	pH NBS	HCO3− contribution	Leakage	εp-mod ‰	εp-meas ‰
A. fundyense
180	5.9 ± 0.9a	2434 ± 3	1992 ± 10a	8.50 ± 0.06a	0.22 ± 0.03	0.44 ± 0.01a	10.1 ± 0.2a	9.0 ± 0.3a
380	11.5 ± 2.1b	2439 ± 1	2117 ± 8b	8.27 ± 0.07b	0.24 ± 0.04	0.46 ± 0.02a	10.6 ± 0.5a	10.2 ± 0.5b
800	25.9 ± 5.8c	2434 ± 2	2245 ± 8c	7.97 ± 0.10c	0.24 ± 0.04	0.53 ± 0.02b	12.6 ± 0.6b	12.7 ± 0.4c
1200	36.5 ± 9.3d	2418 ± 1	2283 ± 5d	7.83 ± 0.12d	0.23 ± 0.08	0.63 ± 0.05c	15.3 ± 0.8c	12.1 ± 0.2c
S. trochoidea
180	6.6 ± 0.2a	2386 ± 1	1872 ± 2a	8.45 ± 0.01a	0.53 ± 0.03a,b	0.56 ± 0.06	10.4 ± 1.5	6.0 ± 0.5a,b
380	13.1 ± 0.5b	2388 ± 2	2096 ± 3b	8.21 ± 0.02b	0.55 ± 0.04a	0.53 ± 0.06	9.4 ± 1.5	5.0 ± 0.1a
800	28.8 ± 2.0c	2385 ± 1	2223 ± 3c	7.91 ± 0.03c	0.48 ± 0.03b,c	0.54 ± 0.01	10.3 ± 0.5	7.1 ± 0.7b
1200	41.5 ± 3.6d	2386 ± 4	2268 ± 9d	7.77 ± 0.04d	0.46 ± 0.04c	0.48 ± 0.04	8.8 ± 1.1	11.8 ± 0.7c
G. spinifera
180	6.0 ± 1.1a	2447 ± 5	1962 ± 15a	8.50 ± 0.05a	0.19 ± 0.11	0.61 ± 0.01	15.6 ± 0.9a	7.8 ± 1.0a
380	11.7 ± 2.5b	2461 ± 12	2083 ± 1b	8.27 ± 0.07b	–	–	–	9.4 ± 0.4a
800	27.9 ± 7.4c	2475 ± 13	2224 ± 9c	7.96 ± 0.10c	0.19 ± 0.11	0.71 ± 0.01	18.6 ± 1.7b	11.7 ± 0.7b
1200	42.4 ± 7.9d	2459 ± 4	2293 ± 5d	7.78 ± 0.06d	–	–	–	8.1 ± 0.5a
P. reticulatum
180	7.1 ± 0.5a	2460 ± 8	2002 ± 2a	8.43 ± 0.04a	0.44 ± 0.13	0.50 ± 0.06	9.58 ± 2.0	8.4 ± 1.8
380	13.9 ± 0.8b	2455 ± 2	2121 ± 4b	8.21 ± 0.02b	–	–	–	8.4 ± 0.7
800	31.0 ± 4.7c	2461 ± 12	2249 ± 23c	7.88 ± 0.08c	0.49 ± 0.19	0.48 ± 0.09	9.2 ± 1.9	8.6 ± 2.0
1200	45.2 ± 6.9d	2473 ± 19	2288 ± 16d	7.75 ± 0.05d	–	–	–	9.9 ± 0.8

Values represent the mean of triplicate incubations (n = 3; ± SD). Superscript letters indicate significant differences between pCO₂ treatments (P < 0.05).

3. Results

3.1. CA activity

In A. fundyense and P. reticulatum, eCA activities were low with maximum activities of 156 U (μg Chl-a)^− 1 and 44 U (μg Chl-a)^− 1, respectively. In S. trochoidea and G. spinifera, eCA activities were comparably high with up to 1600 U (μg Chl-a)^− 1 and 1100 U (μg Chl-a)^− 1, respectively. In neither of the species, eCA activities were responding to changes in pCO₂. Please note that for G. spinifera and P. reticulatum no statistics could be applied due to the lack of replication.

3.2. HCO₃⁻ contribution and leakage

Relative HCO₃⁻ contribution was around 0.2 in A. fundyense and G. spinifera (Figs. 1A and 3A; Table 1), whereas S. trochoidea and P. reticulatum showed higher values of ~ 0.5 (Figs. 2A and 4A; Table 1). In other words, in A. fundyense and G. spinifera approximately 80% of the C_i taken up is in the form of CO₂, whereas in S. trochoidea and P. reticulatum this was 50%. There was a significant decrease in HCO₃⁻ contribution with increasing CO₂ concentration in S. trochoidea, while no such CO₂-dependency was observed in any of the other tested species. Leakage differed significantly between the tested species, with values of up to 0.7 at 800 μatm pCO₂ in G. spinifera (Fig. 3A; Table 1) and lowest average values of ~ 0.5 in S. trochoidea and P. reticulatum, respectively (Figs. 2A and 4A; Table 1). Only in A. fundyense, leakage was significantly CO₂-dependent and increased from 0.44 to 0.63 (Fig. 1A; Table 1). For details on the kinetics of O₂, CO₂ and HCO₃⁻ fluxes in A. fundyense and S. trochoidea, please refer to Eberlein et al. (2014).

Fig. 1.

Relative HCO₃⁻ contribution, leakage and ε_p-mod and ε_p-meas in A. fundyense. Each data point represents the mean ± standard deviation (n = 3).

Fig. 3.

A) Relative HCO₃⁻ contribution, leakage and B) ε_p-mod and ε_p-meas in G. spinifera. Each data point represents the mean ± standard deviation (n = 3).

Fig. 2.

A) Relative HCO₃⁻ contribution, leakage and B) ε_p-mod and ε_p-meas in S. trochoidea. Each data point represents the mean ± standard deviation (n = 3).

Fig. 4.

A) Relative HCO₃⁻ contribution, leakage and B) ε_p-mod and ε_p-meas in P. reticulatum. Each data point represents the mean ± standard deviation (n = 3).

3.3. C_i flux based ε_p calculations

Estimates for ε_p-mod are in the same range as ε_p-meas in A. fundyense and P. reticulatum (Figs. 1B and 4B; Table 1), while the model overestimated the fractionation by up to 5‰ and 8‰ in S. trochoidea and G. spinifera, respectively (Figs. 2B and 3B; Table 1). Except for S. trochoidea, ε_p-mod generally matches trends observed in ε_p-meas. In A. fundyense, for instance, ε_p-mod increased significantly from 10.1 to 15.3‰, thereby closely matching ε_p-meas values (Fig. 1B; Table 1). Also in G. spinifera, ε_p-mod matches trends observed in ε_p-meas, if the highest pCO₂ treatment of ε_p-meas is excluded. In this treatment, carbon isotope fractionation dropped significantly, most likely due to 2.5-fold increased cellular carbon contents (see discussion in Hoins et al., 2015). In S. trochoidea, neither the relative HCO₃⁻ contribution nor leakage showed a CO₂-dependency; hence ε_p-mod did not match the increase in ε_p-meas with increasing CO₂ concentration (Fig. 2B; Table 1).

4. Discussion

4.1. CA activity plays a minor role in C_i fluxes

By expressing CA, many marine algae species accelerate the otherwise slow interconversion between CO₂ and HCO₃⁻, thereby possibly facilitating the C_i uptake and internal C_i fluxes. In line with previous studies on dinoflagellates (Rost et al., 2006, Ratti et al., 2007), A. fundyense and P. reticulatum exhibit rather low eCA activities, even under low CO₂ concentrations. In view of this, eCA is not expected to significantly influence C_i fluxes or ε_p in these species. In S. trochoidea and G. spinifera, however, eCA activities were high at all tested CO₂ concentrations, comparable to values observed for temperate diatoms (Trimborn et al., 2009). Hence, the inhibition of eCA by the inhibitor DBS during the MIMS assay might have biased the C_i fluxes, i.e. underestimated CO₂ uptake (Rost et al., 2003), thereby potentially causing an underestimation of ε_p. As these were the species for which the model overestimated ε_p values, however, it can be concluded that eCA (and its inhibition by DBS during the C_i flux measurements) did not influence the C_i fluxes much.

4.2. Species-specific differences in C_i fluxes

The HCO₃⁻ contribution differed considerably between the tested species. While A. fundyense and G. spinifera showed a strong preference for CO₂, S. trochoidea and P. reticulatum used CO₂ and HCO₃⁻ in equal proportions. The latter contradicts with the findings of an endpoint pH-drift experiment, suggesting that P. reticulatum is not able to efficiently use HCO₃⁻ (Ratti et al., 2007). Testing other dinoflagellates with a modified pH-drift method, including the genus Protoceratium, demonstrated that the high pH itself can affect growth and thus interpretations about the used C_i source based on pH-drift experiments must be considered with caution (Hansen et al., 2007). From an energetic point of view, high CO₂ usage would be of advantage as CO₂ can be taken up passively by diffusion, while HCO₃⁻ is charged and thus has to be taken up by active uptake. And yet, the tested species covered a large part of their C_i demand by HCO₃⁻, as observed in S. trochoidea and P. reticulatum (Figs. 2A and 4A).

Similarly high HCO₃⁻ contributions were observed for other dinoflagellate species (Rost et al., 2006) and cyanobacteria (Price et al., 2008, Kranz et al., 2011). This preference for HCO₃⁻ has been associated with the very low CO₂-affinity of RubisCO type IB, which is expressed in cyanobacteria, and RubisCO type II expressed in dinoflagellates (Badger et al., 1998, Whitney and Andrews, 1998). To compensate for this low affinity, high amounts of C_i have to be accumulated, which in seawater can more easily be accomplished with the abundant HCO₃⁻ ion rather than with CO₂. In addition, HCO₃⁻ is about 1000-fold less permeable to lipid membranes than CO₂, making it the preferred C_i form to be accumulated within the cell (Price et al., 2008). In the case where species covered the majority of their C_i demand by CO₂, as observed in A. fundyense and G. spinifera (Figs. 1A and 3A), one could therefore speculate about chloroplast-based C_i accumulation rather than pumping of HCO₃⁻ at the cell wall (Badger et al., 1998). The observed differences in the preferred C_i source and likely consequences for internal C_i fluxes may also affect the overall leakage of cells.

Also leakage differed considerably among the species. A. fundyense showed a relatively low leakage of 0.44 at 180 μatm pCO₂, which increased to 0.63 at 1200 μatm pCO₂. Leakage in G. spinifera varied between 0.60 at 180 μatm pCO₂ and 0.70 at 800 μatm pCO₂. Leakage estimates in S. trochoidea and P. reticulatum were lower with ~ 0.50 and remained constant over the applied range of pCO₂. These differences may be caused by different membrane permeabilities, which again potentially relate to the preferred C_i source. In fact, A. fundyense and G. spinifera both preferred CO₂ over HCO₃⁻ and likewise showed the highest degrees of leakage, thereby suggesting highly permeable membranes with respect to CO₂. In these species, also CO₂-related changes in the membrane permeability are indicated as they show significantly increased leakage under higher pCO₂ (see also Eberlein et al., 2014 for A. fundyense, formerly A. tamarense).

4.3. Patterns in carbon isotope fractionation can be explained by C_i fluxes

Using results for HCO₃⁻ contribution and leakage obtained in this study, carbon isotope fractionation was calculated and compared to previous measurements (Figs. 1B–4B; see also Hoins et al., 2015). Generally, there is a good agreement as ε_p-mod and ε_p-meas values were in the same range (A. fundyense, S. trochoidea, P. reticulatum) and/or followed the same trend (A. fundyense, G. spinifera, P. reticulatum; Figs. 1B-4B). Despite the overall agreement between flux-based estimates and directly measured carbon isotope fractionation, ε_p-mod was overestimated in S. trochoidea and G. spinifera. Such offsets could principally be attributed to biases in the C_i flux measurements, i.e. uncertainties in the estimation of HCO₃⁻ contribution and/or leakage. It has been argued, however, that the MIMS approach tends to overestimate the HCO₃⁻ contribution (due to the constant pH of 8.0 during the assay, see Burkhardt et al., 2001), and rather underestimates cellular leakage (due to fact that CO₂ fixation does not cease instantly upon darkening, see Badger et al., 1994). Hence, by correcting for these potential biases, i.e. assuming slightly lower HCO₃⁻ contribution and higher leakage values, we would actually overestimate the fractionation even more for S. trochoidea and G. spinifera.

An alternative explanation for the overestimation by the model may be attributed to the fractionation factor of RubisCO, which we assumed to be 28‰ (Raven and Johnston, 1991). Recent studies have found lower values, even as low as 11‰ as in the case of the coccolithophore Emiliania huxleyi (Boller et al., 2011). Even though there are no indications for such low fractionation values in the highly conserved type II RubisCO, a lower fractionation would bring modeled and measured ε_p values closer in S. trochoidea and G. spinifera. In A. fundyense and P. reticulatum, however, it would lead to underestimated ε_p-mod. Hence, we would refrain from assuming much lower fractionation values for RubisCO type II in dinoflagellates in our calculations. Lastly, the fact that internal C_i fluxes were not taken into account might have also contributed to the offsets between ε_p-mod and ε_p-meas. Models that incorporate internal C_i cycling have, however, caused even higher ε_p-mod, as these processes work against the ¹³C accumulation within the chloroplasts (Cassar et al., 2006, Schulz et al., 2007) or the carboxysome (Eichner et al., 2015). Therefore, although the values and trends in carbon isotope fractionation are relatively well understood based on our physiological experiments, differences between theory and measurements are at present not fully resolved.

5. Conclusions

Our study demonstrates that carbon isotope fractionation in dinoflagellates can, to a large degree, be explained by considering their C_i fluxes. Relative HCO₃⁻ contribution and/or leakage were CO₂-dependent in A. fundyense, S. trochoidea and G. spinifera, which in turn can explain the CO₂-dependency of their ε_p observed in previous studies (Hoins et al., 2015). To further advance our understanding of the ε_p patterns in dinoflagellates, C_i fluxes measurements should be performed at in situ pH (Kottmeier et al., 2014, Kottmeier et al., 2016) and ideally differentiate between ¹³C and ¹²C fluxes (McNevin et al., 2006).