Monoclonal cultures of Coccolithus pelagicus (RCC4092) and an Arctic strain of Emiliania huxleyi (RCC3533) were obtained in June 2012 through single cell isolations from surface water samples collected in the Greenland Sea (67.83∘ N, 16.42∘ W and 66.79∘ N, 25.14∘ W, respectively) during the 2012 UK Ocean Acidification Arctic cruise (JR271). These cultures have been deposited into the Roscoff Culture Collection (RCC). North Atlantic Ocean strains of Coccolithus braarudii (RCC1198) and E. huxleyi (RCC1228) were obtained from the RCC.

Cultures were grown in sterile-filtered (0.2 µm) modified K/20 medium (modified from Keller et al., 1987; following Gerecht et al., 2014); aged natural seawater was enriched with 28.8 µM nitrate and 1.8 µM phosphate. Experiments on parallel cultures of either the Arctic strains (C. pelagicus and E. huxleyi RCC3533) or the Atlantic strains (C. braarudii and E. huxleyi RCC1228) were carried out over a range of temperature and light conditions, under a 12 h light–12 h dark cycle.

To reflect a realistic in situ environment (Poulton et al., 2010; Ryan-Keogh et al., 2013), different experimental conditions were used for the Arctic and Atlantic cultures. The Arctic strain experiments were carried out at 6, 9 and 12 ∘C, with a daily photon flux ranging from 1.30 to 8.21 mol photons m-2 d-1 (30–190 µmol photons m-2 s-1) between experiments, while the Atlantic strain experiments were carried out at 12, 14, 16 and 19 ∘C, with a daily photon flux ranging from 1.94 to 10.54 mol photons m-2 d-1 (45–244 µmol photons m-2 s-1). Cells were acclimated to experimental conditions for approximately 10 generations and grown in dilute batch cultures in duplicate. Cultures were grown in ventilated flasks and to low cell densities to avoid biological effects on the carbonate system (150 000–470 000 cells mL-1, 4500–8700 cells mL-1 and 5300–16 000 cells mL-1 for E. huxleyi, C. braarudii and C. pelagicus, respectively) and sampled during the mid-exponential phase to avoid nutrient limitation (Langer et al., 2009; Hoffman et al., 2014).

For determination of cell density, samples were taken daily or every other day and counted immediately in triplicate using either a Sedgwick rafter cell for C. braarudii and C. pelagicus (Langer et al., 2006) or a Coulter Multisizer™3 (Beckman Coulter) for E. huxleyi (Langer et al., 2009). Cell density was plotted against time, and growth rates (μ) were calculated by exponential regression (Langer et al., 2006).

Biometric measurements of coccolithophores were made on samples collected on cellulose nitrate (0.8 µm) and polycarbonate (0.8 µm) filters, and prepared following Poulton et al. (2010) and Daniels et al. (2012), respectively. Light microscopy was used for all biometric measurements of Coccolithus (Gibbs et al., 2013), while a combination of light microscopy and scanning electron microscopy (SEM) was used to study E. huxleyi. Measurements of coccolith size and the number of coccoliths per coccosphere were used to estimate cellular calcite content following the relationship of Young and Ziveri (2000). Cellular particulate organic carbon (POC) was estimated from measured internal cell diameters and cell biovolume following Menden-Deuer and Lessard (2000). Samples for determination of cellular chlorophyll a (Chl a) were collected on Fisherbrand MF300 filters (effective pore size 0.7 µm), extracted in 8 mL of 90 % acetone (HPLC grade, Sigma) for 24 h and analysed on a Turner Designs Trilogy Fluorometer calibrated using a solid standard and a chlorophyll a extract. All experimental data included in the paper are available from the data repository PANGAEA (Publishing Network for Geoscientific & Environmental Data) via Sheward et al. (2014).

Samples for coccolithophore abundance were collected from three RRS Discovery cruises spanning the Irminger and Iceland basins of the North Atlantic during the period of April to August 2010. Two cruises (D350, D354) were part of the (UK) Irminger Basin Iron Study (IBIS), while the third cruise (D351) occupied the Extended Ellett Line. In all three cruises, surface water samples (0.2–1 L) were filtered through cellulose nitrate (0.8 µm) and polycarbonate (0.45 or 0.8 µm) filters, oven dried (30–40 ∘C, 6–12 h) and stored in Millipore PetriSlides. The filters were examined using a Leo 1450VP scanning electron microscope, with coccolithophores identified following Young et al. (2003), and enumerated from 225 fields of view (Daniels et al., 2012). The detection limit was estimated to be 0.2–1.1 cells mL-1. All field data included in the paper are available from the British Oceanographic Data Centre (BODC) via Daniels et al. (2014).

Through manipulation of experimental conditions (temperature and irradiance), a wide range of growth rates was achieved, ranging from 0.16 to 0.85 d-1 (Fig. 2). Emiliania huxleyi RCC1228 (0.50–0.85 d-1) grew significantly faster (Student's t test, t=6.8, df=10, p<0.001) than C. braarudii (0.32–0.58 d-1). For the Arctic strains, the growth rate of E. huxleyi (0.16–0.58 d-1) was significantly different (Student's t test, t=3.5, df=6, p<0.02) to that of C. pelagicus (0.18–0.49 d-1), growing faster in all but the experiment with the slowest growth rates (Fig. 2).

Although E. huxleyi always grew faster than C. braarudii and was generally faster than C. pelagicus, the differences in growth rates were smaller than previously reported, with E. huxleyi growing on average only 12 % (-11 to 26 %) faster than C. pelagicus, and 28 % (12–49 %) faster than C. braarudii. In contrast, Buitenhuis et al. (2008) observed that, when grown in conditions comparable to ours (12–15 ∘C, 14/10 L/D, 4.20 mol photons m-2 d-1), the growth rate of C. braarudii was 42–51 % that of E. huxleyi, although the strain of E. huxleyi used by Buitenhuis et al. (2008) was a non-calcifying mutant of a type that has been observed to have higher growth rates (Paasche, 2002).

While our maximum growth rate of E. huxleyi (0.85 d-1) was lower than in some recent studies (e.g. 0.98–1.64 d-1; Langer et al., 2009), they are well within the range of reported growth rates (0.4–1.9 d-1; Paasche, 2002). Strain-specific variability is likely to partly contribute to this large range in growth rates (e.g. Langer et al., 2009). However, it is also likely that our lower maximum growth rates are due to the effect of the day length used in our study (12 L/12 D), as day lengths shorter than 16 h have been observed to reduce phytoplankton growth rates (Paasche, 1967). Although our E. huxleyi growth rates were lower than those obtained in 16 h day length studies (e.g. Langer et al., 2009; Hoppe et al., 2011), they were similar to another 12 h day length study (0.6–1 d-1; Iglesias-Rodriguez et al., 2008). This is also the case for C. braarudii and C. pelagicus; the maximum growth rate of C. braarudii (0.58 d-1) was below that observed in 16 h day length studies (0.73–0.82 d-1; Langer et al., 2006; Gibbs et al., 2013), but above both 12 h (0.42–0.5 d-1; Taylor et al., 2007; Gerecht et al., 2014) and 14 h (0.4 d-1; Buitenhuis et al., 2008) day length experiments. Although there are few studies of C. pelagicus, our maximum growth rate (0.49 d-1) was greater than the 12 h day length study (0.36 d-1) by Gerecht et al. (2014) but lower than a 16 h day length experiment (0.58 d-1) by Gibbs et al. (2013). Given these differences between experiments, and no literature consensus on recommended day length (Probert and Houdan, 2004), we are therefore confident that our growth rates are representative of these coccolithophore species.

Both temperature and irradiance had a measurable effect on growth rates (Table 2, Supplement Fig. S1). Temperature was the primary driver of growth rates for both E. huxleyi (r2=0.84, p<0.001, n=18) and Coccolithus (r2=0.62, p<0.001, n=18), while irradiance had a secondary, but significant, effect on both E. huxleyi (r2=0.33, p<0.02, n=18) and Coccolithus (r2=0.23, p=0.04, n=18). The growth rate of C. braarudii declined between 16 and 19 ∘C, suggesting that 19 ∘C was above the optimum temperature for C. braarudii. No such decline was observed in the temperature range experienced by C. pelagicus (6–12 ∘C).

In general, a decrease in absolute growth rates was coupled with a smaller difference in the relative growth rates of E. huxleyi and Coccolithus (Fig. 2). As the variability in growth rate was primarily driven by temperature, this suggests that growth rates of Coccolithus and E. huxleyi may be most comparable in cold waters (<10 ∘C), while the growth rate of E. huxleyi will become increasingly greater relative to Coccolithus in temperate waters. As a cold-water species (Winter et al., 1994), with a biogeography spanning the Arctic and sub-polar Northern Hemisphere (McIntyre and Bé, 1967; McIntyre et al., 1970), C. pelagicus could therefore potentially dominate calcite production in this region. As C. braarudii is a more temperate species, seemingly present only in coastal waters of the North Atlantic (Cachao and Moita, 2000; Daniels et al., 2012) and upwelling pockets (Giraudeau et al., 1993; Cubillos et al., 2012), we expect the difference in growth rate between C. braarudii and E. huxleyi to be greater in areas where they are both present. However, as a heavily calcified species, where the coccosphere calcite of one cell is equivalent to ∼ 78 cells of E. huxleyi (Table 1), C. braarudii still has the potential to dominate calcite production in these regions.

The potential for C. pelagicus and C. braarudii to dominate calcite production in their respective environments is dependent on both their relative growth rates and cellular calcite inventories, as well as the relative abundance of these species compared to other coccolithophores. In the context of our study, we consider daily contributions to calcite production, as this is the minimal time length over which we can realistically expect relative abundances to be least variable. Also, much of the work measuring calcite production by natural field communities is based on daily integrals (e.g. Poulton et al., 2010; Poulton et al., 2013).

We examine the potential relative daily calcite production by modelling a simplified community comprised of just E. huxleyi and either C. pelagicus or C. braarudii. Assuming steady state in terms of the cellular quota across a day, calcite production for a given species is the product of its growth rate (μ), cellular calcite (C) and abundance (N) (Leynaert et al., 2001; Poulton et al., 2010). Therefore, we can calculate the percentage of calcite production by a specific species (%CPsp), such as Coccolithus,within a mixed community, using the following equation: %CPsp=μspCspNsp∑i=1nμiCiNi×100. The model was parameterised using a range of relative growth rates that spans the range measured in our culture experiments (Fig. 2, Table 2), but which has also been extended down to 10 % to investigate the effect of Coccolithus having a much lower relative growth rate. The relative abundance of Coccolithus and E. huxleyi in our simple model community is represented as the ratio of E. huxleyi to Coccolithus and was varied from 0 to 80. Cellular calcite values for each species were experimentally determined (Table 1). The percentage calcite production by Coccolithus is inversely related to its relative growth rate, cellular calcite and abundance, and linearly related to the ratio of E. huxleyi to Coccolithus (demonstrated in Fig. 3). As the ratio of E. huxleyi to Coccolithus increases, or the relative growth rate of Coccolithus decreases, a decrease in the percentage calcite production by Coccolithus is observed (Fig. 3).

Coccolithus braarudii is the major source (> 50 %) of calcite production in 56 % of the model, and 64 % of the model when considering only the range of relative growth rates of C. braarudii observed in this study (51–88 %, Fig. 3a). At its average relative growth rate (72 %), C. braarudii will dominate (> 50 %) calcite production if the ratio of E. huxleyi to C. braarudii is less than 53:1, whilst with the same growth rates C. braarudii calcifies at a rate equivalent to 74 cells of E. huxleyi. However, if C. braarudii is only able to grow at a relative growth rate of 10 % that of E. huxleyi, its calcite production is reduced to only 7 times that of an E. huxleyi cell. Therefore, unless C. braarudii both is in a very low relative abundance and has a very low relative growth rate, we would expect C. braarudii to be a major source of calcite compared to E. huxleyi.

Coccolithus pelagicus has a lower cellular calcite content than C. braarudii (16.6 and 38.7 pmol C cell-1, respectively; Table 1) and thus only dominates 29 % of its total model, and 44 % of the model when constrained to observed relative growth rates (74–110 %). When growing at its average observed relative growth rate (88 %), C. pelagicus dominates calcite production when the ratio of E. huxleyi to C. pelagicus is less than 34:1 (Fig. 3b). Equivalent growth rates require a ratio less than 39:1 for C. pelagicus to dominate cellular calcite production, whilst a growth rate of only 10 % that of E. huxleyi results in calcite production from C. pelagicus being only 3.5 times that of an E. huxleyi cell. Although a greater relative abundance of C. pelagicus is required to dominate calcite production compared to C. braarudii, we still find that it would also be a large source of calcite unless both relative growth rates and abundances are low.

Although we have modelled the effect of growth rate and relative abundance on the role of Coccolithus as a calcite producer, the relative calcite production of the two species in these models is highly dependent on the cellular calcite quotas attributed to both E. huxleyi and Coccolithus (Table 1), as calcite production is the product of growth rate and cellular calcite. Estimates of the cellular calcite content of E. huxleyi vary significantly between studies (Balch et al., 1996; Paasche, 2002; Langer et al., 2009; Poulton et al., 2010), which is likely due to both ecophysiological and methodological differences (Young and Ziveri, 2000; Poulton et al., 2010, 2013; Hoffman et al., 2014). Our estimates of E. huxleyi cellular calcite (0.43–0.52 pmol C cell-1) are similar to recent estimates based on similar biometric measurements (Hoffman et al., 2014) and are within the range of literature values (0.22–1.1 pmol C cell-1; Fritz and Balch, 1996; Paasche, 2002; Hoppe et al., 2011). Our value for C. braarudii cellular calcite is greater than previously measured (28 pmol C cell-1, Langer et al., 2006; 17 pmol C cell-1, Gerecht et al., 2014), while the value for C. pelagicus cellular calcite is lower (26 pmol C cell-1; Gerecht et al., 2014).

To address the impact of variability in cellular calcite on calcite production, we have varied the parameters of our model by concurrently increasing the calcite content of E. huxleyi and decreasing that of Coccolithus, by 1 standard deviation each (Table 1), or vice versa (Fig. 3c–f). In doing this, we capture most of the reported range of E. huxleyi calcite as it is the equivalent of varying E. huxleyi RCC3533 calcite by 0.23–0.75 pmol C cell-1 and RCC1228 by 0.33–0.79 pmol C cell-1, while the value for Coccolithus is held constant.

Reducing the calcite content of C. pelagicus (12.7 pmol C cell-1) and C. braarudii (32.5 pmol C cell-1) and increasing that of E. huxleyi (0.57–0.66 pmol C cell-1) reduces the dominance of Coccolithus in the model (Fig. 3c–d). Thus C. braarudii dominates only 37 % of the total model (Fig. 3c), 43 % of the model when constrained to observed relative growth rates, and calcifies at a rate equivalent to 49 cells of E. huxleyi when growth rates are the same. With the same reductions in cellular calcite content, C. pelagicus is the major calcite producer in only 17 % of the total model (Fig. 3d), 26 % of the model when constrained to observed relative growth rates, and with the same growth rate it will dominate calcite production if the ratio of E. huxleyi to C. pelagicus is less than 22:1.

An increase in the calcite content of C. pelagicus (20.5 pmol C cell-1) and C. braarudii (44.9 pmol C cell-1), coupled with a decrease in that of E. huxleyi (0.29–0.38 pmol C cell-1), results unsurprisingly in an increased dominance of both C. braarudii (Fig. 3e) and C. pelagicus (Fig. 3f). Coccolithus braarudii dominates 75 % of the total model and 93 % of the observation-constrained model, while C. pelagicus dominates 53 % of the total model and 81 % of the observation-constrained model.

Cellular calcite clearly has a significant influence on our calculation of percentage calcite production and therefore needs to be constrained more tightly, particularly in the case of Coccolithus. However, we still observe notable levels of calcite production deriving from Coccolithus rather than E. huxleyi in the models using even the lowest values of cellular calcite for Coccolithus.

The model scenarios clearly highlight the importance of relative cellular calcite quotas, relative growth rates and relative abundances when determining the relative role of E. huxleyi and Coccolithus in calcite production. While cellular calcite and growth rates will affect relative calcite production at a cellular level, it is the relative abundance of E. huxleyi and Coccolithus within a population that will determine the proportion of calcite production that derives from Coccolithus. Using data from field communities, we can examine whether populations exist where C. pelagicus has the potential to be a significant calcite producer.

Coccolithophore abundances were determined from samples collected on three cruises in the Irminger and Iceland basins of the North Atlantic, a region in which both E. huxleyi and C. pelagicus are present (McIntyre and Bé, 1967). A physicochemical description of the region is available in Ryan-Keogh et al. (2013), which indicates nutrient replete conditions for the phytoplankton community in spring and nutrient depleted (iron and/or nitrate) conditions in summer. Although other species of coccolithophore were present, we have extracted only the abundances of E. huxleyi and C. pelagicus, so that the data are comparable to our model scenarios in Sect. 3.2. Of the 37 samples analysed, E. huxleyi and C. pelagicus were observed in 29 samples, with E. huxleyi present in a further 6 samples in which C. pelagicus was absent (Fig. 4). When present, concentrations of E. huxleyi ranged from 2 to 980 cells mL-1, while C. pelagicus ranged from 0.1 to 74 cells mL-1. The relative abundance of E. huxleyi to C. pelagicus (0.7–85) was generally comparable to our modelled range, with a relatively low median average of 12.7. However, in two samples (Supplement Table S1), the relative abundance was much higher (155–212), such that C. pelagicus was unlikely to be a significant calcite producer in these samples.

Assuming the original model scenario of measured cellular calcite (Table 1, Fig. 3a and b) and the average relative growth rate for C. pelagicus of 88 %, the minimum relative abundance of E. huxleyi to C. pelagicus required for E. huxleyi to dominate calcite production (34:1) was exceeded in only 5 out of 29 samples. Taking into account those samples in which C. pelagicus was absent, C. pelagicus is a greater calcite producer than E. huxleyi in 69 % of the samples. If equivalent growth rates are assumed, then C. pelagicus remains the major calcite producer in 69 % of the samples.

Under the more conservative model scenario (Fig. 3d), with a relative growth rate of 88 %, C. pelagicus remains the major calcite producer in 57 % of the samples, which is reduced to 51 % if the lowest measured relative growth rate (74 %) is used. If C. pelagicus has a higher nutrient requirement and lower nutrient affinity than E. huxleyi, then in low-nutrient conditions we would expect a lower relative growth rate. As we do not know the relative nutrient affinities, we have used an extreme in our original model where C. pelagicus has a relative growth rate of 10 %. Under this scenario, C. pelagicus is the major calcite producer in 14 % of the samples, although it would still form a significant component of the total calcite production (7–49 %) in other samples when present.

Using experimentally determined relative growth rates and cellular calcite quotas, in conjunction with relative abundances from field populations, we have shown that C. pelagicus is likely to be a major source of calcite in the sub-polar North Atlantic. Data on relative abundances of E. huxleyi and C. braarudii in field communities were not available for an equivalent comparison study.

While the difference in growth rates between E. huxleyi and Coccolithus is comparatively small, the difference in cell volume of C. pelagicus (∼ 1100 µm3) and C. braarudii (∼ 2100 µm3) compared to E. huxleyi (∼ 50 µm3) is relatively large. These differences are reflected in their cellular Chl a and cellular calcite : POC ratio (Table 1), with the species having similar ratios of carbon : Chl a (25–36 g g-1) across the experimental conditions. Larger cells have a lower surface-area-to-volume ratio, which reduces the diffusive nutrient uptake per unit volume of the cell (Lewis, 1976; Finkel et al., 2009), and thus maximal growth rates generally increase with decreasing cell size (Sarthou et al., 2005). Hence, although we expect E. huxleyi maximal (optimal) growth rates to be higher than Coccolithus, the relatively small difference in growth rate (Fig. 2) compared to cell volume (Table 1) implies that Coccolithus must have efficient (competitive) nutrient uptake pathways, or that these experimental conditions are less optimal for E. huxleyi than Coccolithus.

It is also worth considering the implications of relative differences in cell size and surface area to volume for nutrient requirements to support growth. From our estimates of cellular POC (Table 1) and assuming Redfield stoichiometry (Redfield, 1958), we can also estimate that the cellular particulate organic nitrogen (PON) and particulate organic phosphorus (POP) content of E. huxleyi, C. pelagicus and C. braarudii is, respectively, 0.10, 2.0 and 3.6 pmol N cell-1 and 0.006, 0.12 and 0.22 pmol P cell-1. Our estimates of cellular quotas for E. huxleyi are similar to Langer et al. (2013), who measured cellular quotas of 0.69 pmol C cell-1, 0.12 pmol N cell-1, and 0.003 pmol P cell-1. Cellular quotas of both C. pelagicus and C. braarudii have recently been measured by Gerecht et al. (2014). While the cellular PON (1.9 pmol N cell-1) and POP (0.19 pmol P cell-1) of C. pelagicus were generally similar to our study, the value for cellular POC was slightly larger (20 pmol C cell-1), suggesting a lower nutrient requirement per unit POC. However, Gerecht et al. (2014) report C. braarudii cellular quotas of POC (13 pmol C cell-1) and PON (1.5 pmol N cell-1) that are much lower than their values for C. pelagicus. This is unexpected, as it is generally accepted that C. braarudii is a larger species of coccolithophore than C. pelagicus (Geisen et al., 2004) and we would therefore expect a higher POC content for C. braarudii than C. pelagicus (Table 1) if POC scales with cell size. Clearly further cellular measurements of POC, PON and POP for different coccolithophore species are needed to fully examine cellular nutrient requirements.

For culture media with a given nitrate concentration of 10 µmol N L-1, the maximum cumulative cell concentration that could be supported using our estimated cellular PON would therefore be ∼ 1 × 105, ∼ 5000 and ∼ 2800 cells mL-1, respectively, for E. huxleyi, C. pelagicus and C. braarudii. This corresponds to cumulative calcite concentrations, using cellular calcite quotas from Table 1, of ∼ 50, ∼ 80 and ∼ 110 µmol C L-1. Therefore despite lower cell densities, for a given nutrient concentration, a population of C. pelagicus and C. braarudii would be a greater source of calcite than E. huxleyi.

Emiliania huxleyi regularly forms seasonal blooms in excess of 1000 cells mL-1, particularly in the high latitudes of the Northern and Southern hemispheres (Tyrrell and Merico, 2004; Poulton et al., 2013). For a bloom with a magnitude of 1000 cells mL-1, this would require a nitrate concentration of only ∼ 0.1 µmol N L-1. Comparatively, although rare, C. pelagicus has also been reported in concentrations exceeding 1000 cells mL-1 in the high-latitude North Atlantic (Milliman, 1980), requiring a much larger nitrate concentration of 2 µmol N L-1. The seasonal drawdown of nitrate in the North Atlantic is estimated be ∼ 10 µmol N L-1 (Sanders et al., 2005; Ryan-Keogh et al., 2013), and thus a C. pelagicus bloom of 1000 cells mL-1 represents the utilisation of a significant amount of the available nutrients. For a bloom of this magnitude to occur, we would expect C. pelagicus to be a significant proportion of the total phytoplankton community with a relatively low mortality rate, as nutrient drawdown will be related to gross production by the total phytoplankton community. Reduced mortality has also been discussed as a possible factor in the formation and persistence of E. huxleyi blooms in the southeast Bering Sea (Olson and Strom, 2002).

The function of coccoliths is not well understood but may have a significant role in reducing mortality by providing a certain level of protection from zooplankton grazing (Young, 1994; Tyrrell and Young, 2009). If this is the case, then we would speculate that C. pelagicus has a relatively lower mortality than E. huxleyi due to both its larger cell size and its much larger and heavier coccosphere. A lower mortality may explain how C. pelagicus is able to form high-density populations, while the large nutrient requirement would restrict C. pelagicus blooms to populations where it heavily dominates the plankton community, and this may explain the scarcity of reported C. pelagicus blooms.