Temperature affects the morphology and calcification of Emiliania huxleyi strains

The global warming debate has sparked an unprecedented interest in temperature effects on coccolithophores. The calcification response to temperature changes reported in the literature, however, is ambiguous. The two main sources of this ambiguity are putatively differences in experimental setup and strain specificity. In this study we therefore compare three strains isolated in the North Pacific under identical experimental conditions. Three strains of Emiliania huxleyi type A were grown under non-limiting nutrient and light conditions, at 10, 15, 20 and 25 C. All three strains displayed similar growth rate versus temperature relationships, with an optimum at 20–25 C. Elemental production (particulate inorganic carbon (PIC), particulate organic carbon (POC), total particulate nitrogen (TPN)), coccolith mass, coccolith size, and width of the tube element cycle were positively correlated with temperature over the sub-optimum to optimum temperature range. The correlation between PIC production and coccolith mass/size supports the notion that coccolith mass can be used as a proxy for PIC production in sediment samples. Increasing PIC production was significantly positively correlated with the percentage of incomplete coccoliths in one strain only. Generally, coccoliths were heavier when PIC production was higher. This shows that incompleteness of coccoliths is not due to time shortage at high PIC production. Sub-optimal growth temperatures lead to an increase in the percentage of malformed coccoliths in a strain-specific fashion. Since in total only six strains have been tested thus far, it is presently difficult to say whether sub-optimal temperature is an important factor causing malformations in the field. The most important parameter in biogeochemical terms, the PIC : POC ratio, shows a minimum at optimum growth temperature in all investigated strains. This clarifies the ambiguous picture featuring in the literature, i.e. discrepancies between PIC : POC–temperature relationships reported in different studies using different strains and different experimental setups. In summary, global warming might cause a decline in coccolithophore’s PIC contribution to the rain ratio, as well as improved fitness in some genotypes due to fewer coccolith malformations.

The ratio of particulate inorganic carbon (PIC) and particulate organic carbon (POC) influences surface wateratmosphere gas exchange as well as the composition of matter exported from surface waters to the deep ocean (Ridgwell and Zeebe, 2005;Findlay et al., 2011).The response of PIC and POC production and their ratio in the prolific species E. huxleyi to temperature is a necessary first step towards an understanding of its possible impact on global biogeochemical cycles.
The relationship of PIC production/PIC : POC and temperature in E. huxleyi is not clear.De Bodt et al. (2010) found that PIC production was higher at lower temperatures in a strain grown at 13 and 18 • C, while Sett et al. (2014) found the opposite in another strain grown at 10, 15 and 20 • C. De Bodt et al. (2010) found higher PIC : POC ratios at lower temperatures for a strain of E. huxleyi and Gerecht et al. (2014) found a similar relationship for a strain of the species Coccolithus pelagicus.Sett et al. (2014), however, found a different relationship for the PIC : POC ratio in another strain of E. huxleyi, which is not supported by the experiment of Langer et al. (2007) on the same strain.Feng et al. (2008) did not find differences in the PIC : POC ratio in another strain grown at 20 and 24 • C.These discrepancies between studies might stem from different experimental setups and a lack of knowledge of the optimum growth temperature or indeed strain-specific differences (Hoppe et al., 2011).Therefore, it is necessary to test more than one strain for its temperature response under otherwise identical conditions.This we have done in the present study.
Apart from biogeochemical considerations, global warming might also be of interest in terms of the ecological success of coccolithophores, because different groups of organisms might be differently affected by warming and therefore ecological succession patterns, grazing pressure etc. might change.The latter was proposed to depend on coccolith morphology more than it does on PIC production (Langer et al., 2011).The effect of temperature on coccolith morphogenesis is evident in field observations (Bollmann, 1997;Ziveri et al., 2004) and is best assessed with respect to the optimum growth temperature in laboratory experiments.While the effect of supra-optimal temperature is unequivocally detrimental (Watabe and Wilbur, 1966;Langer et al., 2010), it is not clear whether there is an effect of sub-optimal temperature at all (Watabe and Wilbur, 1966;Langer et al., 2010;De Bodt et al., 2010).A temperature increase in the sub-optimal range is probably what most coccolithophore clones will experience in the course of global warming (this study Buitenhuis et al., 2008;Langer et al., 2009;Heinle, 2014), and therefore this temperature range is particularly interesting.In the present study we focus on coccolith morphology under suboptimal temperature, doubling the amount of data currently available, and thereby clarifying whether sub-optimal temperatures can cause malformations.We selected three strains of E. huxleyi from a single area, the Japanese coast in the North Pacific Ocean, in order to assess the plasticity within strains originating from a particular environmental setting.

Pre-culture and batch culture experiments
Clonal cultures of Emiliania huxleyi were obtained from the Roscoff Culture Collection.We selected three strains of E. huxleyi, two from the Japanese coast in the North Pacific Ocean (RCC1710 -synonym of NG1 and RCC1252 -synonym of AC678 and MT0610E) and a third strain from the same region but of unknown exact origin and strain name, named here IAN01.Strain RCC1710 was collected off Nagasaki at Tsushima Strait (Japan) and RCC1252 at Tsugaru Strait (Japan); both places are strongly influenced by the Tsushima warm current.Additional information about the strain RCC1252 can be found at http: //roscoff-culture-collection.org/.
The culture media was sterile-filtered North Sea water (filtered through 0.2 µm pore size sterile Sartobran 300 filter cartridges, Sartorius, Germany) supplemented with nutrients (nitrate and phosphate), metals and vitamins according to Guillard and Ryther (1962).Cell densities were determined using a Multisizer 3 Coulter Counter (Beckman Coulter for particle characterization).To prevent significant changes in seawater carbonate chemistry, maximum cell densities were limited to ≈ 1×10 5 cells mL −1 (e.g.Oviedo et al., 2014).We used a 16/8 light/dark cycle, and an irradiance of ≈ 300 µmol photons s −1 m −2 .The three strains were grown for at least 20 generations.
The dilute batch culture experiments were conducted in triplicate, for the strains RCC1710 and RCC1252 at 10, 15, 20 and 25 • C of temperature, and for IAN01 at 15, 20 and 25 • C. The strains were grown in 2 L of sea water within transparent sterilized 2.3 L glass bottles.Cell density at inoculation was 500 to 1000 cells mL −1 , and at harvest it was a maximum of 1 × 10 5 cells mL −1 .Harvesting was done 9 h after the onset of the light period.
Growth rate was calculated from exponential regression according to µ where c 1 and c 0 are the final cell concentration and the initial cell concentration, respectively, and t is the duration of incubation in days.Averages of triplicates and SD were used in tables and figures (Table 1 and Fig. 1a).

Carbonate chemistry
The seawater carbonate system was monitored because temperature and coccolithophore production alter the system.We  employed the dilute batch method (Langer et al., 2013) to minimize production effects.
During the harvesting, samples for total alkalinity (TA) measurements were sterile-filtered (0.2 µm pore size) and stored for less than 2 months prior to measurement in 25 mL borosilicate flasks at 4 • C. TA was calculated from linear Gran plots (Gran, 1952) after potentiometric titration (in duplicate) (Bradshaw et al., 1981;Brewer et al., 1986).
Samples for dissolved inorganic carbon (DIC) were sterile-filtered (0.2 µm pore size) with gentle pressure using cellulose-acetate syringe filters and stored bubble-free for less than 2 months prior to measurement at 4 • C in 5 mL borosilicate flasks.DIC was measured, in triplicate, using a Shimadzu TOC 5050A.
The carbonate system was calculated from temperature, salinity (32 ‰), TA and DIC, using the program CO2SYS (Lewis and Wallace, 1998), applying the equilibrium constants from Mehrbach et al. (1973), refitted by Dickson and Millero (1987).For an overview of carbonate chemistry final conditions in all treatments, see Table 2.

Particulate organic and inorganic carbon, particulate nitrogen and calcite
Duplicate samples for the determination of total particulate carbon (TPC) and total particulate nitrogen (TPN) were filtered onto pre-combusted (500 Particulate inorganic carbon (PIC) was calculated measuring calcium content of samples with 3.6×10 6 E. huxleyi cells filtered onto 47 mm polycarbonate (PC) filters (0.8 µm pore size).PC filters were immersed overnight in an acid solution of 1 % HNO 3 to dissolve calcite.Calcium was determined by analysing an aliquot of the samples using an inductively coupled plasma mass spectrometer (ICP-MS, Agilent model 7500ce).Cellular PIC was calculated from the molecular mass of calcite, using the following equations: where PIC cell −1 is the cellular PIC (in pg), PIC s is the PIC sampled contained in the filter (in pg), c is the cell concentration (in cells L −1 ), V s is the volume sampled (in L), [Ca 2+ ] s is the calcium content in the sample (in pg), 12.0107 corresponds to the relative atomic mass of carbon, and 40.078 corresponds to the relative atomic mass of calcium.Particulate organic carbon (POC) was calculated as the difference between TPC and PIC.PIC, POC and TPN production (P PIC , P POC , P TPN ) were estimated as the product of cellular PIC, POC or TPN, and growth rate.Calcite (CaCO 3 ) per cell (concomitant of PIC) can also be estimated, substituting in Eq. ( 2) the calcium carbonate molecular mass (100.0869) in place of the relative atomic mass of carbon.The ratio between PIC and POC (PIC : POC) and the ratio between POC and TPN (POC : TPN) were also calculated.

Coccolith morphology -by scanning electron microscopy
Thirty millilitres of culture was filtered onto polycarbonate filters (0.8 µm pore size) and dried at 60 • C for 24 h.A small portion (∼ 0.7 cm 2 ) of each filter was mounted on an aluminium stub and coated with gold (EMITECH K550X sputter coater).Images were captured along random transects using a ZEISS-EVO MA10 scanning electron microscope (SEM).
Emiliania huxleyi SEM images were used to measure and categorize 300 coccoliths per sample (e.g.Langer et al., 2009); the coccoliths were on coccospheres.The tube width (width of the tube elements cycle) of each coccolith (Fig. 2c) was the average of the tube width measured on the two semiminor axes (along the coccolith width) on the distal view of the coccolith.Tube width measurements were manually taken using the program Gimp-2.8.Examples of the tube width variations in the three different strains are shown in Fig. 2. The 300 coccoliths were classified as normal, malformed or incomplete (e.g.Langer et al., 2011), as described in Table 3, with examples in Figs. 3 and 4.

Coccolith length and mass -by polarized light microscopy
Between 10 and 30 mL of culture was filtered with ∼ 200 mbar onto cellulose nitrate filters (0.2 µm pore size) and dried at 60 • C for 24 h.A radial piece of filter was embedded and made transparent in immersion oil on microscope slides (e.g.Ziveri et al., 1995).
Images were taken at a magnification of 1000× with a Leica DM6000B cross-polarized light microscope (LM) equipped with a SPOT Insight camera (e.g.Bach et al., 2012;Horigome et al., 2014).Between 50 and 200 image frames from each sample were taken along radial transects and analysed using SYRACO software (Dollfus and Beaufort, 1999;Beaufort and Dollfus, 2004).A minimum of 300 coccolith images were automatically identified by the software and measured in pixels.The software also automatically measures the grey level for each pixel by a birefringence method based on the coccolith brightness when viewed in crosspolarized light (Beaufort, 2005).Coccolith length and mass were subsequently calculated from the pixels and from the measured grey level, respectively, following Horigome et al. (2014) and Beaufort (2005).Therefore, coccolith length was converted from pixels to micrometres, where 832 pixels correspond to 125 µm, and coccolith mass was converted from grey level units to picograms, where 2275.14 grey level units were equivalent to 1 pg of calcite.

Statistics
For the three E. huxleyi strains together, ANOVA (two-factor with replication) was used to test whether a response variable (i.e.growth rate, element variables, morphological variables and mass) presented significant (p < 0.05) differences between the temperature treatments, to test whether the effect was strain-independent or strain-specific (p < 0.05), and to test whether there were significant differences in the interaction between treatment and strain (p < 0.05) and therefore whether the different strains respond similarly or not regardless of whether they were presenting differences between them.If the temperature effect was strain-specific, further ANOVA was used for pairs of strains.
If a response variable presented significant differences between the temperature treatments, and the variable also presented a significant strain-independent response to temperature, or at least the same response on two of the strains, the variable for the similar strains was analysed with simple and multiple linear regressions, including CO 2 partial pressure (pCO 2 ), CO 2− 3 concentration and pH, in order to find the useful coefficients (t statistics, p < 0.05) of the significant equation (F test, p < 0.05) that would estimate the assessed variable value, e.g. the single or combined variables significantly estimating growth rate.

Population growth
The three strains of E. huxleyi presented a stable growth rate (per day) that changed with temperature (Fig. 1a, Table 1), with significant differences between the temperature treatments (F = 244.11,p = 0.000).The strains RCC1710 and RCC1252 presented similar growth rates, not statistically different from one another (F = 0.372, p = 0.550).From 15 to 25 • C, the IAN01 growth rate was significantly different from the other two E. huxleyi strains (F = 4.53, p = 0.025), but there was no significant difference in the interaction between treatment and strain (F = 0.71, p = 0.597), so the three strains behaved significantly similarly.The optimum temperature for the three strains was 25 • C. When RCC1710 and RCC1252 were analysed together, changes in growth rate only depended significantly on temperature (linear regression: R 2 = 0.91, F = 229.58,p = 0.000); the carbonate system variables (Table 2) did not much increase the coefficient of determination (maximum to an R 2 = 0.92) and none of them were significantly useful in predicting growth rate when used together with temperature (t statistics: p > 0.05).According to Eq. ( 1), on the three strains, a minimum of one duplication per day was obtained from 15 to 27.5 • C.

Element measurements, ratios and production
Cellular PIC (and its concomitant calcite), POC and TPN (pg cell −1 ) did not show a consistent trend related to temperature when comparing the three strains of E. huxleyi (Fig. 1b,  e, h; Table 1).When cellular PIC and TPN response to temperature (from 15 to 25 • C) were statistically analysed (ANOVA), significant differences were found between treatments (F = 113.42,p = 0.000 and F = 36.52,p = 0.000,

Normal
Regular coccolith in shape, with well-formed distal shield elements aligned forming a symmetric rim.Considered normal when zero or only two malformations were present.

Malformed
Irregular coccolith in shape or size of individual elements and a general reduction in the degree of radial symmetry shown; teratological malformation (Young and Westbroek, 1991).Considered malformed when three or more malformations were present in the coccolith.

Incomplete
Coccolith with variations in its degree of completion according to its normal growing order, with no malformations.Primary calcification variation (Young, 1994).
There was no consistent explanatory variable for cellular PIC, POC, and TPN when analysing the three strains independently.
In the three strains, production of PIC (and its concomitant calcite), POC and TPN (pg cell −1 day −1 ) showed a positive relationship with temperature (Fig. 1c, f, i; Table 1).Highest PIC and POC production was in general reached at 25 • C, except for RCC1710, which reached it at 20 • C. From the statistical analysis, PIC and POC production response to temperature, when comparing the three strains of E. huxleyi together, was significantly different between the temperature treatments (F = 8.36, p = 0.003) and the response was strain-independent (F = 0.89, p = 0.428).Highest TPN production was in general reached at 20 • C, except for RCC1252, which reached it at 25 • C. The latest was supported statistically, as TPN production response, with significant differences between temperature treatments (F = 499.96,p = 0.000), was strain-specific (F = 65.92,p = 0.000) when comparing the three strains of E. huxleyi together, and yet still the strains RCC1710 and IAN01 presented a similar interaction between treatment and strain (F = 3.52, p = 0.062); thus, the two strains had a similar behaviour in the TPN production response despite the different values between the strains (F = 19.02,p = 0.000).
Changes in PIC production on the three strains of E. huxleyi mostly depended on temperature (linear regression: R 2 = 0.89, F = 217.36,p = 0.000); pCO 2 with [CO 2− 3 ], when used together with temperature, only slightly increased the coefficient of determination (R 2 = 0.93).Changes in POC production on the three strains of E. huxleyi only depended significantly on temperature (linear regression: R 2 = 0.85, F = 157.71,p = 0.000).
The PIC : POC ratio decreased from 10 to 20 • C in the three strains of E. huxleyi (Fig. 1d).POC was higher than PIC only in the strains RCC1710 and IAN01 at 20 • C. From the statistical analyses, the only significant similitude obtained was in the interaction between treatment and strain for RCC1252 and IAN01 (F = 2.12, p = 0.163), which means that the PIC : POC ratio behaves similarly towards temperature in these two strains.
The POC : TPN ratio (Fig. 1h) relationship with temperature was strain-specific (F = 9.59, p = 0.001).The differences between the temperature treatments were significant (F = 16.95,p = 0.000).There were no significant differences between the strains RCC1710 and RCC1252 (F = 2.71, p = 0.119), in which the lowest POC : TPN ratio was found at 10 • C; however, there were significant differences in the interaction between treatment and strain (F = 3.52, p = 0.039), as observed in the different temperatures at which maximum POC : TPN ratios were found for each strain (20 and 25 • C, respectively).The strain IAN01 showed a much different relationship with temperature, with a minimum POC : TPN ratio found at 20 • C.

Coccolith morphology and mass
Although there was great variation between replicates, mean tube width of coccoliths (Fig. 5a, Table 4) presented a positive trend with temperature, independent of the strain of E. huxleyi (F = 1.73, p = 0.204).Changes in tube width on the three strains of E. huxleyi only depended on temperature (linear regression: R 2 = 0.47, F = 28.09,p = 0.000); pCO 2 and [CO 2− 3 ] did not much increase the coefficient of determination (R 2 = 0.51) and none of them were significantly useful in predicting tube width when used together with temperature (t statistics: p > 0.05).
Coccolith length (Fig. 5b, Table 4) showed a positive trend with temperature, especially on strains RCC1252 and IAN01.The positive trend in strain RCC1710 was not so clear; however, minimum length was also found at 10 • C and maximum length also at 25   between them (F = 2.12, p = 0.171); temperature gave the highest coefficient of determination (R 2 = 0.62, F = 24.03,p = 0.000) and was the only useful coefficient in estimating coccolith length when making any combination with pCO 2 , [CO 2− 3 ] or pH.The strain RCC1710 was analysed independently of the other two strains: temperature presented a low and non-significant coefficient of determination (R 2 = 0.28, F = 3.55, p = 0.092); instead, pH presented the highest coefficient of determination (R 2 = 0.65, F = 16.87,p = 0.002).
The positive relationship of the mean tube width with temperature reflects the increased coccolith calcite quota at higher temperature.Coccolith mass and coccolith size are positively correlated.Why coccolith mass or size should increase with temperature cannot be decisively answered based on our data.
Regardless of the strain, coccolith calcite mass (Fig. 5c, Table 4) showed a positive trend with temperature; significant differences were found between treatments (F = 35.59,p = 0.000) and no significant differences were found in the interaction between treatment and strain (F = 2.53, p = 0.08).The strains RCC1252 and IAN01 were analysed together as they did not show significant differences between them (F = 0.65, p = 0.425).Temperature presented the highest coefficient of determination for RCC1252 and IAN01 (R 2 = 0.75, F = 45.93,p = 0.000) and also for RCC1710 (R 2 = 0.87, F = 58.58,p = 0.000), and adding other coefficients was not significantly useful in estimating coccolith mass.On average, coccolith mass increased with temperature ∼ 2.2× from 10 to 25 • C, ∼ 1.5× from 15 to 25 • C, and ∼ 1.2× from 20 to 25 • C; on average, coccolith mass increased 1.28× (or 0.45 pg) each 5 • C.
The percentage of malformed coccoliths per sample (Fig. 6a, Table 4) did not show a consistent trend with temperature when comparing the three strains of E. huxleyi (F = 113.21,p = 0.000).Only one strain (RCC1252) presented significant differences between the temperature treatments, with higher percentage at the lowest experimented temperature.
Only in strain RCC1710, the percentage of incomplete coccoliths presented a significant increase with temperature (Fig. 6b, Table 4).Higher percentages of incomplete coccol- iths in strain RCC1710 were found at 25 • C. ANOVA results showed that, between the three strains, there were no significant differences between only the strains RCC1252 and IAN01 (F = 0.06, p = 0.810) and their interaction between treatment and strain (F = 2.33, p = 0.139), though in this case (analysed from 15 to 25 • C) there were also no significant differences between the temperature treatments (F = 3.78, p = 0.053).Significant strain-independent and strain-specific responses of E. huxleyi to temperature, found in the three strains of this study, are summarized in Table 5.

Growth rate, elemental production and incomplete coccoliths
All three E. huxleyi strains investigated here displayed similar growth rate versus temperature relationships, with an optimum at 20-25 • C (Fig. 1a).This is a typical range for many E. huxleyi strains (e.g.Watabe and Wilbur, 1966;Van Rijssel and Gieskes, 2002;Sorrosa et al., 2005;De Bodt et al., 2010;Langer et al., 2009).We expect that strains isolated, for example, in the Arctic will have a lower temperature optimum, though.Also not untypical, elemental production (PIC, POC, TPN) increased with temperature over the sub-optimum to optimum temperature range (Fig. 1; Langer et al., 2007;Sett et al., 2014).It is intuitive that, approaching optimum, higher temperature increases elemental production, because biochemical rates are temperature-dependent.It is also intuitive that the percentage of incomplete coccoliths should increase with higher P PIC , as indeed observed in RCC1710 (Fig. 6b).The idea underlying this intuition is that less time is taken to produce one coccolith and that the production process is stopped before the coccolith is fully formed.A comparison of RCC1710 and RCC1252 shows how wrong this idea is (Table 6).The percentage of incomplete coccoliths increases in the former only.While it is true that coccolith production time in RCC1710 decreases from 31 min at 10 • C to 22 min at 25 • C, this decrease is even more pronounced in RCC1252 (from 88 to 23 min).Hence, RCC1252 should show a steeper increase in incompleteness than RCC1710.This is not the case.Please note that the increase in incompleteness in RCC1252 (Fig. 6b) is not significant, because the increase is well below 10 % and the error bars overlap (see also Langer et al., 2013, for a discussion of this criterion).Another piece of evidence which does not fit the "premature release of coccoliths because of time shortage" idea is that both RCC1710 and RCC1252 manage to produce heavier coccoliths in a shorter time at higher temperature (Tables 4  and 6).We do not know why the stop signal for coccolith growth is affected by temperature in RCC1710.Nothing is known about the biochemical underpinning of that stop signal, so it is unfortunately impossible to speculate about the mechanism of a temperature effect.It was, however, argued that the processes involved in the stop signal are different from those producing teratological malformations (Young and Westbroek, 1991;Langer et al., 2010Langer et al., , 2011)).This is supported by our data, because there is no correlation between incompleteness and malformations (Fig. 6).We will discuss malformations in Sect.4.3.Interestingly coccolith mass is positively correlated with temperature (and P PIC ) in all strains tested here.The positive correlation of coccolith mass and P PIC was also observed by Bach et al. (2012) in a carbonate chemistry manipulation experiment and is the basis of using coccolith mass as a proxy for P PIC (Beaufort et al., 2011).This is an interesting option, because in field samples coccolith mass might be a promising indicator of P PIC .There are only few proxies available to reconstruct past coccolithophore P PIC , the traditional one being the calcite Sr / Ca ratio, established at the turn of the millennium (Stoll and Schrag, 2000).Analysing Sr / Ca, how-Table 5. Significant strain-independent and strain-specific responses of E. huxleyi to temperature found in the three strains of this study.-Cellular PIC, POC and TPN (pg per cell).-POC : TPN ratio.However, in the two strains tested at 10 • C (RCC1710 and RCC1252), the POC : TPN ratio was lowest at 10 Strain ever, requires either a sizable sample or comparatively sophisticated secondary ion mass spectrometry (SIMS) measurements (Stoll et al., 2007;Prentice et al., 2014).Recently, coccosphere diameter and coccolith quota were introduced as growth rate proxies (Gibbs et al., 2013).However, complete coccospheres are the exception rather than the rule in sediment samples, so it is important to have a proxy based on individual coccoliths.Hence, coccolith mass and size (which are correlated; Fig. 5, Table 4) are an option which it is worthwhile exploring in the future.

Emiliania huxleyi PIC : POC response
As detailed in the introduction there is considerable variability in the PIC : POC response of E. huxleyi to temperature changes.This variability cannot be traced back to strainspecific features, but it might partly reflect the fact that different temperature ranges were investigated, mostly without the knowledge of the optimum temperature.Other experimental conditions, such as light intensity and nutrient concentrations, varied and might have also played a role (Hoppe et al., 2011).In this study we ran three strains under identical conditions and, for the first time, are presented with a coherent picture.All three strains display a bell-shaped curve with lowest PIC : POC close to the optimum growth temper-ature (Fig. 1d).Although our data on the right-hand side of the PIC : POC minimum are not conclusive for RCC1252, the bell-shaped curve is discernible in the latter strain.This finding seems to fit data on other E. huxleyi strains (De Bodt et al., 2010;Sett et al., 2014) and on C. pelagicus (Gerecht et al., 2014).This comparison is, however, not straightforward since two of the studies (De Bodt et al., 2010;Gerecht et al., 2014) employed two temperatures, one of the studies employed three temperatures (Sett et al., 2014), only without determining the optimum temperature in all three studies.Be that as it may, based on our data, we might conclude that E. huxleyi tends to show the lowest PIC : POC close to its optimum growth temperature.In the context of global warming, that would mean that, in the future, E. huxleyi and possibly coccolithophore PIC : POC will tend to decrease because most strains live at sub-optimal temperatures in the field (Buitenhuis et al., 2008;Langer et al., 2009;Heinle, 2014).This trend might be pronounced because global warming is accompanied by lower surface water nutrient levels and ocean acidification (Cermeño et al., 2008;Doney et al., 2009).All these changes apparently cause a decrease in E. huxleyi's PIC : POC (our data; Hoppe et al., 2011;Oviedo et al., 2014).A marked decline in coccolithophore PIC : POC will have implications for long-term carbon burial and might even affect surface water carbonate chemistry on short timescales, i.e. 1 year (Barker et al., 2003;Ridgwell and Zeebe, 2005;Cermeño et al., 2008).

Coccolith malformations
The coccolith shaping machinery is, besides the ion transport machinery, an essential part of coccolith formation (for an overview see Holtz et al., 2013).The latter commences with heterogeneous nucleation on an organic template, the so-called base plate.The nucleation determines crystal axis orientation.Crystal growth proceeds in principle inorganically, with the notable exception that crystal shape is strongly modified by means of a dynamic mould, which essentially consists in the coccolith vesicle shaped by cytoskeleton elements and polysaccharides inside the coccolith vesicle.Malformations can be due to an abnormal base plate which would affect crystal axis orientation, aberrations in the composition or structure of the polysaccharides, and disturbance of cytoskeleton functionality.The last of these would most likely also cause a decline in growth rate, which is why this mechanism was disregarded in the case of carbonate-chemistryinduced malformations (Langer et al., 2011).By the same reasoning, temperature-induced malformations might be due to cytoskeleton disturbance, because temperature does also alter growth rate (Fig. 1a).However, it is not straightforward to see why lower than optimum temperature should disturb cytoskeleton functionality (see also Langer et al., 2010).At any rate, coccolith malformations are most likely detrimental to fitness, because malformed coccoliths result in fragile coccospheres, which are regarded as instrumental in coccolithophore fitness (Dixon, 1900;Young, 1994;Langer and Bode, 2011;Langer et al., 2011).One of the many hypotheses concerning function of calcification is that the coccosphere confers mechanical protection (Dixon, 1900;Young, 1994).After more than a century of research, it still remains the most plausible hypothesis.Coccolith malformations, i.e. disturbances of the coccolith shaping machinery, occur in both field and culture samples, but usually more so in the latter (Langer et al., 2006(Langer et al., , 2013)).The causes of malformations are only partly known.In cultured samples, artificial conditions (not present in the field) such as cell densities of 10 6 cells mL −1 , cells sitting on the bottom of the culture flask, stagnant water, and confinement in a culture flask play a role in inducing the surplus of malformations compared to field samples (Langer et al., 2013;Ziveri et al., 2014).However, in the field malformations do occur, and sometimes in considerable percentages (Giraudeau et al., 1993;Ziveri et al., 2014).The environmental conditions leading to elevated levels of malformations have long since been disputed.Besides nutrient limitation (Honjo, 1976), temperature and carbonate chemistry are conspicuous candidates.Although the range of temperatures used here exceeds 2100 projections (IPCC, 2013), we used it not only on physiological grounds but also for ecological reasons.Over the course of the year, coccolithophores in the North Pa-cific experience the whole range of temperatures used here (http://disc.sci.gsfc.nasa.gov/giovanni/,maps in the Supplement).In a seminal experimental study it was shown that moving away from the optimal growth temperature increases malformations in E. huxleyi (Watabe and Wilbur, 1966).This result was confirmed for higher than optimum temperature in another strain (Langer et al., 2010) but could not be confirmed for sub-optimal temperature in two strains (De Bodt et al., 2010;Langer et al., 2010).The sub-optimal temperature range is of particular interest because most clones live at sub-optimal temperatures in the field.Here we investigated sub-optimum to optimum temperatures in three further strains.While RCC1710 showed no change in the percentage of malformations and IAN01 featured a shallow gradual increase from 25 to 15 • C, RCC1252 was insensitive over the latter range but displayed a steep increase in malformations at 10 • C (Fig. 6).Based on our own and the literature data, we conclude that the sub-optimal temperature effect on morphogenesis is strain-specific.The fact that the base level of malformations in cultured coccolithophores differs between species and strains (and also varies with time) has been recognized for many years and is now well documented (e.g.Langer and Benner, 2009;Langer et al., 2011Langer et al., , 2013)).Also, the response of the morphogenetic machinery to environmental factors is strain-specific (Langer et al., 2011).We currently do not have enough accessory information to formulate a hypothesis why exactly one strain differs from another.The fact that they do indeed differ, however, probably reflects the high genetic diversity in E. huxleyi.
Can we see a pattern in this strain specificity?It is intriguing that E. huxleyi clones fall into two distinct groups characterized by their temperature preference: the warm-water and the cool-water group (Hagino et al., 2011).Of the strains analysed for morphology, the following belong to the warmwater group: BT-6 (Watabe and Wilbur, 1966), RCC1710, RCC1252, and possibly RCC1238 (Langer et al., 2010).The latter was unfortunately not included in the study by Hagino et al. (2011).Since these strains display different responses to temperature, their being part of the warm-water group does unfortunately not help finding common features of sensitive strains.However, only a few strains have been studied so far, and it might be worthwhile testing a statistical number from the warm-water and the cool-water group.

Conclusions
1. Temperature, PIC production, coccolith mass, and coccolith size are positively correlated.Since the positive correlation between coccolith mass and PIC production was observed in response to seawater carbonate chemistry changes as well (Bach et al., 2012), it can be hypothesized that coccolith mass might be a good proxy for PIC production independent of the environmental parameter causing the change in PIC production.
2. Sub-optimal growth temperature was identified as one of the potential causes of coccolith malformations in the field.Since the effect of sub-optimal temperature on coccolith morphogenesis is strain-specific, a statistically relevant number of strains have to be tested in order to clarify whether this effect is indeed ecologically relevant.
3. We consistently showed for the first time that E. huxleyi features a PIC : POC minimum under optimum growth temperature.Taken together with literature data, this finding suggests that global environmental change will lead to a marked decrease in PIC : POC of E. huxleyi and possibly coccolithophores as a group.
The Supplement related to this article is available online at doi:10.5194/bg-13-2913-2016-supplement.

Figure 1 .
Figure 1.Results at different temperatures.Growth rate (a) (extra temperatures from pre-experiments are included and shown as empty symbols); cellular PIC and its concomitant calcite (b), POC (e) and TPN (h) content; PIC (c), POC (e) and TPN (i) production (linear trend lines and r squared values are shown); and PIC : POC ratio (d) and POC : TPN ratio (g).Standard deviations of the triplicate experiment results are shown.Three different strains of E. huxleyi were used.

Figure 2 .
Figure 2. Examples of tube width variations observed in E. huxleyi RCC1710 (a-c), RCC1252 (d-f), and IAN01 (g-i) coccoliths.Tube width (c) was measured along the two semi-minor axes (along the coccolith width) of each coccolith and averaged.Scale bar equal to 1 µm.

Figure 5 .
Figure 5. Changes in coccolith morphometry (a, b) and mass (c), at different temperatures.Standard deviations of the triplicate experiment results are shown.Three different strains of E. huxleyi were used.

Figure 6 .
Figure 6.Percentage of malformed (a) and incomplete (b) coccoliths in three E. huxleyi strains grown at different temperatures.Standard deviations of the triplicate experiment results are shown.

Table 1 .
Growth rate and cellular PIC, POC, and TPN content and production of the three strains of E. huxleyi at different temperatures.Standard deviation of the triplicates in parentheses.Measured growth rates for extra temperatures from the pre-experiments are included, but PIC, POC and TPN were not measured for these temperatures.

Table 3 .
Morphological categorization of coccoliths (from SEM images) of E. huxleyi used in this study.

Table 4 .
Coccoliths morphology and mass.Standard deviation of the triplicates is shown in parentheses.

Table 6 .
Coccolith production time.Standard deviation of the triplicates is shown in parentheses.Lith: coccolith; d: day; h: hour; min: minutes.