2.1 Preparation of the culture media

We opted to conduct our cultures in artificial seawater, as diluting natural seawater would have led to covariations between salinity and δ ¹⁸O_sw. We thus maintained all ion ratios constant to better mimic natural environmental changes (Kirst, 1990). In total, we prepared two batches of 50 L each at salinity S=39 (hereafter, Sal39) by adding salts to de-ionised water following the ESAW recipe (Keller et al., 1987), and subsequently prepared the lower salinity batches with by re-diluting volumetrically the Sal39 batch with de-ionised water. We checked the salinity of each of the Sal29 to Sal39 batch by measuring the conductimetry of the media at room temperature. We used a TetraCon^® 325 conductimetry cell calibrated with 0.01 M KCl solution with a value of 1413 µS cm⁻¹ at 25 ^∘C. The correlation line linking conductimetry and salinity exhibited a root-square coefficient (r²) of 0.99 indicating the good yield and accuracy of the targeted salinities. Therefore, in the following account, we will express the salinities of our batches by the target values, as they depart from less than 0.5 ‰ with respect to the actual measurements.

The various media were subsequently amended in nitrate, phosphate, silica and trace metals, and vitamins as indicated in the f∕2 recipe to ensure that macro and micronutrient availability did not change with salinity. Medium pH was adjusted to 8.2 by addition of 1 N NaOH solution. The last step of the preparation consisted of steri-filtration of the artificial medium using 0.22 µm Millipore^® Stericup^® devices and the transfer of the medium into sterile culture vessels for further algal inoculation.

2.2 Strains selected and acclimatation phase

We chose the following species and strains to conduct our experiments: Emiliania huxleyi morphotype A (strain RCC1256), Emiliania huxleyi morphotype B (strain RCC1212), Gephyrocapsa oceanica (strain RCC1314) and Gephyrocapsa ericsonii (strain RCC4032) (see Fig. 1 for SEM images of their coccospheres). The cultures were conducted at the biogeochemistry laboratory operating within the Institut des Sciences de la Terre de Paris. Firstly, the cells were transferred from natural seawater originating from the English Channel to our Sal35 batch and left for a month (i.e. more than 20 generations). We sub-cultured the algae every week in fresh medium. Subsequently, the four strains acclimated in ESAW medium at Sal35 were transferred in each salinity, and the acclimatation phase to the new salinity was made for another month with the same sub-culturing frequency. The transfer of RCC1314 G. oceanica and RCC4032 G. ericsonii into the two extreme conditions of Sal29 and Sal39, respectively, failed despite several attempts.

Figure 1Scanning electron micrographs of the four strains of coccolithophores (Haptophyta) cultured in the present study. Key to strain ID: RCC1314 is Gephyrocapsa oceanica (a), RCC4032 is Gephyrocapsa ericsonii (b); RCC1212 is Emiliania huxleyi Morphotype B (c); and RCC1256 is Emiliania huxleyi Morphotype A (bottom right). Note the higher magnification for the smallest cells studied, namely RCC4032 (d).

Download

2.3 Implementation of the batch cultures

After the acclimation phase, the proper culture began with an initial concentration of cells around 200 cells mL⁻¹ of medium in 150 cm² polystyrene Thermo Scientific Nunc^® culture flasks. Each strain/salinity experiment was conducted in duplicate. The medium was refreshed every 2 days following a semi-continuous batch strategy (see Laroche et al., 2010). The cultures grew in a Panasonic^® MLR 352 illuminated growth chamber. Temperature was kept constant at 15 ^∘C, light irradiance was ∼150 µmol photons m⁻² s⁻¹ and the photoperiod lasted 14 h per day. These flasks were gently stirred twice a day, and prior to pipetting out 1 mL sterilely for cell numeration. Cell numeration was achieved using a Beckman^® Coulter Counter^® Multisizer IV fitted with a 70 µm aperture tube. The measurements were made 3 h after the start of the illuminating period in the growth chamber. Calculation of the specific growth rates (μ) was made following Eq. (1):

where μ is the specific growth rate, cell conc d is the cell density in cells mL⁻¹ at harvest, and cell conc d−n is the cell density n day before this measurement. n is the number of days elapsed between these two measurements. The unit of μ is day⁻¹.

The culture experiments were stopped when the densities reached 8000∕10 000 cells per mL. A final cell numeration was undertaken including the measurements of the size of the coccospheres (provided as a diameter-equivalent quantity by the Multisizer IV). The same aliquots were acidified by addition of 200 µL of 0.2 M HCl in the cuvette, left for 5 min and the measurements were repeated to provide the diameter of the naked cells. Removal of the coccospheres that originally surrounded the cells were checked under an inverted microscope Zeiss Axio Vert.A1^® at 630 times magnification. Standardisation of size measurements on the Multisizer IV was done via a calibration procedure that used certified latex beads L10 with a 10.13 µm nominal diameter (batch 9747089F).

Harvest consisted in the suspensions containing the cells and free coccoliths to be poured onto polycarbonate membranes (0.8 µm nominal aperture) and gathered by vacuum-filtration. The culture residues were rinsed with deionised water to prevent precipitation of salt during air-drying in a desiccating cabinet.

2.4 Determination of isotopic ratios

Aliquots of 0.22 µm filtered water samples were analysed for their δ¹⁸O_sw using a PICARRO CRDS (cavity ring-down spectrometer; model L2130-I Isotopic H2O) at the LOCEAN-IPSL lab in Paris. The δ¹³C of the dissolved inorganic carbon (DIC) was measured from CO₂ (g) released from water after injection of phosphoric acid (H₃PO₄ 85 %), and determined using a dual inlet-isotopic ratio mass spectrometer (SIRA9-VG) in the same laboratory. These procedures and the standardisation are described in Racapé et al. (2010) and Benetti et al. (2017). The media were measured with δ¹⁸O${}_{\mathrm{sw}}=-\mathrm{6.55}\mathrm{‰}\pm \mathrm{0.1}$ ‰ VSMOW, and δ¹³C of the DIC with values of $-\mathrm{12.40}\mathrm{‰}\pm \mathrm{0.3}$ ‰ VPDB. The isotopic values were indistinguishable between batches of distinct salinities (n=12).

For coccolith calcite, approximately 100 µg of dry culture residue was weighted out and transferred into borosilicate vials. Calcite reacted with purified phosphoric acid at 70 ^∘C in a Kiel IV preparation device attached to a Thermo Scientific Delta V Advantage isotope ratio mass spectrometer at ISTeP, Sorbonne University. Stable isotope values were calibrated relative to the Vienna Pee Dee Belemnite (‰ VPDB) via the NBS-19 international standard. Reproducibility of measurements is ±0.1 ‰ for the δ¹⁸O ratios and ±0.05 ‰ for the δ¹³C ratios.

3.1 Specific growth rates

All strains exhibit significant changes in their specific growth rates in response to varying salinities (Fig. 2). Overall, the relationship between salinity and μ can be described by second-order polynomial curves (Supplement Table S1), with the exception of G. ericsonii, for which the measurements do not show a relation between μ and salinity (partly due to the Sal31 measurements). Gephyrocapsa oceanica, the largest cell studied here, grew faster at the lowest salinities Sal31 (0.52 day⁻¹), but did not grow at Sal29. The μ for this strain progressively decreased down to 0.17 day⁻¹ with increasing salinity. The relatively small cell of G. ericsonii shows the fastest division rate achieving more than one division per day for Sal33 and Sal35. The lowest salinities Sal29 and 31 exhibit μ values around 0.65 day⁻¹. They were significantly lower at the highest salinity Sal37, while no growth occurred at Sal39 for this strain.

The strain of Emiliania huxleyi producing coccoliths belonging to the Morphotype A (RCC1256) exhibits a bell curve for its salinity∕μ relationship (r²=0.91; p value < 0.0005) with a growth optimum at Sal33 (Fig. 2). On both sides of the maximum of ca. 0.7 day⁻¹ (for reference, representing a doubling population of cells every day on average), the growth rates are significantly lower, with a symmetrical response. Thus, growth rates are roughly halved at Sal29 and Sal39 compared to the optimum. The other strain of Emiliania huxleyi, RCC1212 Morphotype B, which is also much smaller in size, shows a different salinity∕μ relationship, as growth rates did not change significantly from around 0.7 day⁻¹ with our different salinity treatments, except for Sal29 that showed growth rates of only ∼0.3 day⁻¹.

Figure 2Effect of ambient salinity on the specific growth rates of coccolithophore algae. The data show a prominent modulation of growth rates, which appears to be strongly strain specific, with optimum growth achieved at salinities around 33, except for G. oceanica. All fits (red lines) correspond to a second-order polynomial law (not significant for RCC4032).

Download

3.2 Cell sizes (diameters)

The strain RCC1314 of Gephyrocapsa oceanica exhibits an increase in cell diameter from ∼6 to ∼7 µm from Sal31 to Sal39 conditions (Fig. 3), which corresponds to an increase of the cellular volume by 60 % (116 to 188 µm³). This change can also be described by a second-order polynomial fit (r²=0.86; p value < 0.0005). The size of Gephyrocapsa ericsonii is also highly dependent on salinity (r²=0.95; p value < 0.05), with a progressive increase of the diameter and volume with increasing salinity, as observed for G. oceanica. We note, however, that the sizes for Sal35 and Sal37 are similar. For the small G. ericsonii cell, the relative variation in cell size is modest, less than 20 % compared to its close-relative species G. oceanica.

Even though cell sizes of RCC1212 Emiliania huxleyi grown at different salinities range between 5 and 5.6 µm (65 and 95 µm³ in volume), there is no clear trend in size or volume with ambient salinity (Fig. 3). The morphotype A of the same species shows a response in the cellular volume with salinity, as these two parameters are statistically linked by a polynomial fit (r²=0.81; p value < 0.0005) with a minimum value at the intermediate salinity (Sal31–27 – 47 µm³ on average), whereas Sal29 and Sal39 are characterised by higher cell volumes (87 µm³ for Sal29).

Figure 3Effect of ambient salinity on the cell diameters of coccolithophore algae. As for growth rate, variations in the size of the cells (lightly decalcified coccospheres – see Methods) are species and strain dependent. All fits (red lines) correspond to a second-order polynomial law.

Download

The volume of coccolithophore cells is an important physiological parameter reflecting the metabolic rates of the cells (Aloisi, 2015). Such a relationship (the higher the growth rate, the smaller the cell size) is expressed for G. oceanica (RCC1314) and the Morphotype A (RCC1256) of E. huxleyi (r²=0.82; p value < 0.005 and 0.62; p value < 0.005, respectively) (Fig. 4), but is not for the two other strains, which do not exhibit responses describing a “bell curve” (second-order polynomial fit) with p values between 0.4 and 0.6 (Fig. 3).

3.3 Isotopic ratios of coccolith calcite

3.3.1 Oxygen isotopes

A practical means to express the isotopic results from calcite minerals in biogeochemistry and palaeoceanography is to use the “δ – δ” notation, by which a pseudo-fractionation factor is expressed by the offset between δ¹⁸O_calcite and δ¹⁸O_sw. Cultured coccoliths show a wide spread of δ¹⁸O_calcite – δ¹⁸O_sw values between species, from 0.5 ‰ for G. oceanica to 2 ‰ for E. huxleyi Morphotype A, but the variations for a given strain are much more limited, typically within a ∼0.35 ‰ range (Fig. 5). Overall, there is no statistically significant relationship between salinity and the oxygen isotope composition of the coccoliths (r² < 0.6; p value around 0.01). It seems that coccoliths produced by G. ericsonii exhibit increased δ¹⁸O values with increasing ambient salinity, but this correlation is not strong, given the reproducibility of the replicates with respect to the range of measured values on the Sal29–37 spectrum.

Figure 4Covariations between growth rate and cell size of coccolithophores grown under various salinities. The conversion of the cell diameters into volumes have been achieved considering that the cells have are spherical. RCC1314 G. oceanica and RCC1256 E. huxleyi A show correlated changes in these two ecological-relevant parameters, however with distinct influence of salinity.

Download

Figure 5Effect of ambient salinity on the oxygen isotope composition of coccolith calcite. The variations in the isotopic composition of the coccolith are relatively minor, less than 0.35 ‰ for a considered strain with our salinity treatments. The spread of isotopic data is slightly larger for RCC1212 E. huxleyi (0.5 ‰), but there is no salinity dependence on growth rates observed for this strain. For reference, the Kim and O'Neil's calcite composition is around 0 ‰ VPDB. At 15 ^∘C, the vital effect can thus be quantified by the δ¹⁸O_calcite – δ¹⁸O_sw offset. These positive values indicate that all the investigated species belong to the “isotopic heavy group”. All fits (red lines) correspond to a second-order polynomial law (see Table S1).

Download

In our study, the δ¹⁸O_sw did not significantly change with salinity, and temperature was kept constant at 15 ^∘C, implying that changes in the isotopic composition of calcite correspond to a modulation of the expression of the vital effects by the coccolithophores. It is possible to compute the δ¹⁸O composition of an inorganic calcite following work by Kim and O'Neil (1997). This inorganic reference has been recently found to depart from equilibrium conditions (Watkins et al., 2013), but all previous literature on coccolith biogeochemistry, including the assignment of coccoliths into isotopically light and heavy groups (Dudley et al., 1986; Hermoso, 2014; Ziveri et al., 2003) are based on the Kim and O'Neil's study. As our aim is to compare the isotopic composition of coccolith calcite grown by microalgae exposed to various salinities, without change in temperature and δ¹⁸O_sw, the use of this reference will not lead to an artefact for the biogeochemical signatures of the coccoliths. We calculated the oxygen isotope composition of such an inorganic calcite as −6.59 ‰ VPDB. The average magnitudes of the oxygen isotope vital effect are thus +0.7 ‰ for RCC1314; +1.6 ‰ for RCC4032; +1.6 ‰ for RCC1212; +1.9 ‰ for RCC1256 – results that are in line with previously published coefficients of the vital effect, except for RCC4032 G. ericsonii for which no prior value exists. For RCC1314 Gephyrocapsa oceanica, the magnitude of the vital effect is slightly lower, by 0.5 ‰, than previous reports with cultures of the same strain grown in natural seawater (Hermoso et al., 2016a). This difference may be related to lower growth rates measured in our study at Sal 33–35 (0.45 day⁻¹ in an artificial medium vs. 0.85 day⁻¹ in natural seawater). In all cases, the important feature of this dataset is that we found no influence of salinity on the oxygen isotope vital effects.

3.3.2 Carbon isotopes

The spread of δ¹³C_calcite – δ¹³C_DIC offsets for a given strain is within 1 ‰ disregarding a non-replicated measurement for RCC1256 E. huxleyi at Sal37 (Fig. 6). Both Gephyrocapsa cells (RCC1314 and 4032) show comparable trends for the δ¹³C of their coccoliths with salinity, although the absolute values differ significantly (the former being 2 ‰ isotopically depleted compared to the latter). A negative trend is observed with increasing salinity, disregarding their lowest salinity data points. Such a trend is also evident for RCC1212 E. huxleyi, whereas for the other strain of E. huxleyi, there is not apparent link between δ¹³C values and salinity. Interestingly, the magnitude of these decreases in δ¹³C values seems to be common to all species, and can thus be quantified to −1 ‰. For G. ericsonii, δ¹³C and μ are strongly correlated by a polynomial function (r²=0.91; p value < 0.05). This power fit is the mathematical consequence of the logarithmic expression of growth rate. Such a correlation is also seen, albeit statistically weaker, in RCC1212 E. huxleyi (r²=0.49), but not for the two other strains.

For the carbon isotopes, the composition of the inorganic calcite is given by Romanek et al. (1992) and is equal to δ¹³C_DIC+1, hence −11.40 ‰ VPDB in our case study. Thus, G. oceanica appears to be close (±0.5 ‰) to “equilibrium” conditions, whilst the other strains shift towards positive vital effects (up to +3.3 ‰ for RCC1212 E. huxleyi). These observations are also in good agreement with published literature (Hermoso, 2015; Hermoso et al., 2014; Holtz et al., 2015; McClelland et al., 2017; Rickaby et al., 2010). There seems to be, in contrary to the oxygen isotope system, an influence of salinity on coccolith δ¹³C values, at least for three of the four strains being examined here.

Figure 6Effect of ambient salinity on the carbon isotope composition of coccolith calcite. There is an influence of salinity, or a parameter itself influenced by salinity (growth rate) on the stable carbon isotope composition of the coccoliths. The is a general decrease in δ¹³C values with salinity, particularly well expressed on the Sal33–39 spectrum. RCC1256 E. huxleyi shows an appreciable spread in the data without a clear correlation between salinity and δ¹³C values.

Download

Figure 7Compilation of previously published data documenting the variations of the specific growth rates of coccolithophores with salinity, and data of the present study shown on Fig. 2. The source for previous studies is inset. It is apparent that growth rates diminish with increasing salinity with a large variability in the response according to the considered strain, especially for E. huxleyi. By contrast, data for G. oceanica seem to be in good agreement between studies and strains (red curve), disregarding the fact that no growth was achieved by RCC1314 at Sal29 in our study. Overall, this graph highlights the euryhaline nature of open ocean microalgae that typifies the ecology of the coccolithophores thriving in the oceans and maintained in the laboratory under salinities nearing 33.

Download

4.1 The effect of salinity change on the physiology of the coccolithophores

The seminal culture survey by Brand (1984) revealed the euryhaline nature of the coccolithophores Emiliania huxleyi and Gephyrocapsa oceanica, which have been successfully grown at salinities between 25 and 45, with some strains of E. huxleyi exhibiting limited growth under salinity as low as 15 %. Our data and other previous studies indicate that changing ambient salinity of the coccolithophores induces large changes in their growth rate (Fig. 7). In particular, it has been found that E. huxleyi and G. oceanica achieved better growth at the lowest salinities, with a lower limit of 25 for the former species (Fisher and Honjo, 1989; Saruwatari et al., 2016; Schouten et al., 2006). Except for Sal29, a salinity at which no growth was achieved by G. oceanica in our batches, our data for this species, and for G. ericsonii, are consistent with this overarching observation that the cells exhibit diminished growth rate with increasing salinity (Figs. 2 and 7). However, the salinity effect on growth rate for the two strains of E. huxleyi (RCC1212 and RCC1256) being examined here markedly differs from observations made by Fisher and Honjo (1989) and Schouten et al. (2006) (strains MHC1 & G4 and PML B92/1, respectively), but are in line with those made on strain PLY B92/11 by Fielding et al. (2009). An optimum in division rates at salinity around 33 (the mean oceanic value and the salinity at which the strains have been maintained in culture for many years) is evident in our data for E. huxleyi Morphotypes A and B. The reason(s) behind this discrepancy are not straightforward to discuss beyond methodology, and may pertain to physiological differences and distinct adaptabilities to changing environment of the multiple ecotypes of E. huxleyi (e.g. Rickaby et al., 2016).

Our experiments only deal with short-term adaptability of the coccolithophores exposed to relatively abrupt changes in salinity, necessitating rapid adaptation to the osmotic stress imposed on the cells in our treatments. The physiological implications regarding turgor stressor are complex, involving osmotic adjustment and the understanding of the precise mechanisms at play is beyond the scope of this paper. These aspects are not described in any great detail in the existing literature. We see contrasting responses in cellular sizes forced by ambient salinity among the strains grown in this study. Gephyrocapsa oceanica and G. ericsonii exhibit relatively smaller cell sizes at the lowest salinity (Fig. 4), while the opposite response is observed for the Morphotype A of E. huxleyi. Meanwhile, there seems to be no response for the Morphotype B of E. huxleyi. The presence of active transporters (or membrane pumps) may vary in type, number and effective rates between various coccolithophore species. As a consequence, these possibly distinct strain-specific features will affect the transport and diffusion of water across the membrane, thus influencing the size of the cell and leading to the observed differences between G. oceanica, G. ericsonii, and the two strains of E. huxleyi.

The preparation of our culture media did not only change the ionic concentrations of the dominant Na⁺ and Cl⁻ ions, but also induced coeval compositional changes in elements relevant to photosynthetic and calcifying unicellular organisms. Indeed, from the Sal39 down to Sal29 conditions, [Ca²⁺] and [DIC] decreased in the same ratio as salinity. Thus, a secondary effect from simultaneous changes in Ca²⁺ and DIC concentrations is therefore possible. For the lowest salinities, diluted Ca²⁺ and DIC may become limiting for coccolithophore growth and intracellular calcification (calcium has also a role in cell signalling), superimposing a response to salinity per se. If the concentration and availability of DIC substantially decreased with decreasing salinity (constant 39:29 ratio), the availability of aqueous CO₂ in equilibrium with the lab air first and with the headspace in the culture flask during the batch culture remained relatively unchanged. It is well established that coccolithophore algae predominantly source the carbon resource from their external environment in the form of aqueous CO_2aq (McClelland et al., 2017). Thus, the ambient concentration in CO₂ is the primary parameter dictating cellular growth with implications for the expression of the isotopic vital effects (Bolton and Stoll, 2013; Hermoso, 2014). We calculated the variations in [CO_2aq] using CO2CALC software (Robbins et al., 2010) and found indeed only relatively modest changes (on the order of 1 %) from Sal39 to Sal29. Salinity has little effect on the CO_2air∕CO_2aq equilibrium according to Henry's law, conversely with temperature. Furthermore, the linear changes in the concentrations of DIC, CO_2aq and calcium on the entire salinity spectrum are not compatible with the bell curves observed in the measurements of coccolithophore physiology that exhibit optima around Sal33–35. This question was also dealt with in work by Paasche et al. (1996) who conducted calcium and bicarbonate enrichment in low salinity waters, and found no alleviation of depressed growth rates. It is therefore unlikely that lowered growth rates at salinities lower than the optimum values are caused by decreases in the carbon and calcium resources. Therefore, we can confirm that the modulations of growth observed in our study are salinity-induced changes. Which ionic species (such as Na⁺, Cl⁻, ${\mathrm{SO}}_{\mathrm{4}}^{\mathrm{2}-}$) exerts the dominant control of salinity on the coccolithophores at the cellular level remains an open question.

4.2 The effect of salinity change on the isotopic composition of coccolith biominerals