Effects of high CO 2 levels on the ecophysiology of the diatom Thalassiosira weissﬂogii differ depending on the iron nutritional status

m [Si]) of the diatom Thalassiosira weissﬂogii under high iron ([Fe ′ ¼ (cid:3) pmol l 2 1 ) and low iron ([Fe ′ ] ¼ (cid:3) pmol l 2 1 ) investi-gated. Cell volume decreased with increasing p CO 2 , whereas intracellular C, N, and P concentrations increased with increasing p CO 2 only under high p CO 2 .ItreﬂectshigherproductionofnetC,N,andPwithnocorrespondingchange in net Si production under high p CO 2 and high iron conditions. In contrast, signiﬁcant linear relationships between measured parameters and p CO 2 wererarelydetectedunderlowironconditions.WeconcludethattheincreasingCO 2 levelscouldaffectonthebiogeochemicalcyclingofbioelements selectively under the iron-replete conditions in the coastal ecosystems.


Introduction
Ongoing increases in CO 2 in seawater resulting from anthropogenic activities affect carbonate chemistry and result in a decrease in pH and an increase in the partial pressure of CO 2 (pCO 2 ; i.e. ocean acidification; Doney et al., 2009). The effect of changing carbonate chemistry on phytoplankton has attracted much research attention over the last decade. Because phytoplankton is a crucial component of the marine ecosystem, understanding the effects of ocean acidification on phytoplankton processes is important to predict the future health of the marine ecosystem (Doney et al., 2009;IGBP, IOC, and SCOR, 2013).
The bioavailability of trace metals is affected by changing carbonate chemistry, which in turn affects the relationship between trace metals and pH-sensitive inorganic/organic coordination chemistry (Millero et al., 2009). Shi et al. (2010) reported that the conditional stability constant of iron -organic ligand complexes increased as pH decreased when the ligand was an acidic binding group. Previous CO 2 perturbation studies have indicated that iron availability can decrease in low pH -high CO 2 conditions (Shi et al., 2010;, but the results may largely depend on the sensitivity of each phytoplankton community to changes in iron availability Yoshimura et al., 2013Yoshimura et al., , 2014. In addition, the chemical nature of iron-binding ligands in natural seawater is largely uncertain. Such uncertainty implies that simultaneous changes in carbonate chemistry and iron availability in natural seawater make it difficult to discriminate the effect of each factor on phytoplankton ecophysiology. To properly interpret the results of ocean acidification studies in the natural environment, it is necessary to understand the individual effects of changes in carbonate chemistry and iron availability on phytoplankton ecophysiology. Iron has important roles in the metabolism of phytoplankton such as photosynthesis, nitrogen (N) assimilations, and respiration (Raven et al., 1999). One of the most bioavailable form of iron for eukaryotic phytoplankton is dissolved inorganic iron (Fe ′ ) species mainly composed of Fe(OH) 2 + , Fe(OH) 4 2 , and Fe(OH) 3 0 in seawater (Morel et al., 2008). In contrast, iron uptake from iron-organic ligand complex is often a very slow process for eukaryotes, indicating that such complex shows low iron bioavailability. Therefore, to investigate the effect of iron availability on phytoplankton processes, Fe ′ rather than total iron in seawater must be considered. The binding affinities of iron with ethylene diamine tetraacetic acid (EDTA) and with bacterial hydroxamate siderophore desferrioxamine B are sensitive to changes in seawater pH (Shi et al., 2010). For example, 100 mmol l 21 EDTA bound with 100 nmol l 21 ferric iron under 208C and 100 mmol photon m 22 s 21 , 12:12 h light:dark conditions was used to demonstrate that a decrease in pH from 8.0 to 7.8 potentially decreased the concentration of Fe ′ ([Fe ′ ]) from 90 to 26 pmol l 21 (Sunda and Huntsman, 2003). Recently,  demonstrated that a combination of iron, EDTA, and CO 2 concentrationcontrolled air can be used to manipulate [Fe ′ ] and carbonate chemistry separately and simultaneously. However, most previous studies did not discriminate the effects of changes in iron availability from changes in carbonate chemistry.
Dissolved iron (,0.2 mm) concentrations in coastal waters (. 1 nmol l 21 ) are generally higher and more spatio-temporally dynamic than that of oceanic waters (, 0.5 nmol l 21 ; Kuma et al., 2000;Nishioka et al., 2013). The dynamic nature of dissolved iron concentrations in coastal environment reflects the iron sources, and physical and chemical properties of seawater. For example, riverine inputs and vertical mixing involving sediments could increase iron concentrations, whereas intrusion of oceanic waters and the low solubility of Fe ′ in seawater could decrease the concentrations (Kuma et al., 2000). Furthermore, one of the most important factors determining dissolved iron concentrations is the concentration and binding affinity (Log K Cond FeL,Fe ′ ) of organic ligands (Kuma et al., 2000;Bundy et al., 2015). Therefore, iron bioavailability, namely the [Fe ′ ], could dynamically change depending on the organic and inorganic chemistry of seawater.
Iron requirement is generally higher in coastal phytoplankton species than in oceanic species, probably reflecting the environmental conditions of their habitats (e.g. Sunda and Huntsman, 1995). As a result, the effects of changes in iron availability with changes in CO 2 conditions could differ depending on the nature of the iron requirement of each phytoplankton species. One of the notable effects of iron availability on phytoplankton is change in the elemental composition of diatoms (e.g. Takeda, 1998). Changes in CO 2 concentrations can also alter the cellular elemental composition of diatoms (e.g. Sun et al., 2011;, indicating that changes in either iron availability or CO 2 concentration could affect marine biogeochemical cycling of bioelements. However, other than the study by , no studies have demonstrated the individual effects of iron availability and CO 2 conditions on the elemental composition of phytoplankton. Because the response to the changes in iron availability and CO 2 variations differ among species (Burkhardt et al., 1999;Doney et al., 2009;Bucciarelli et al., 2010), more studies on the elemental composition of phytoplankton are needed to better understand the biogeochemistry of the ocean.
To date, prediction of future environments has relied on in silico simulations. One of the major problems in estimating biogeochemical cycling of bioelements in modelling studies is the behaviour of the elemental composition of phytoplankton, which is usually fixed as the canonical Redfield ratio (Passow and Carlson, 2012). In forecasts on global and long-time-scales, Redfield ratio-based simulation is assumed to be representative of the nutrient budget in the ocean. However, models simulating changes on regional, and short-time-scales should consider the flexible elemental composition of phytoplankton to represent the ecological stoichiometry in dynamic environments (Smith et al., 2011;Passow and Carlson, 2012). Quigg et al. (2003) reported that individual phytoplankton taxa have unique elemental stoichiometry. The regional characteristics of elemental composition of marine particulate matter therefore reflect the elemental stoichiometry of dominant phytoplankton taxa (Weber and Deutsch, 2010). Similarly, changes in the elemental composition of phytoplankton communities in response to environmental changes differ among reports, probably as a result of differences in the community composition Eggers et al., 2014;Richier et al., 2014;Yoshimura et al., 2014). Therefore, more information is needed on species-specific differences in the effects of environmental disturbances on the ecophysiology of phytoplankton.
Here, we examine the individual effects of carbonate chemistry and iron availability on the elemental composition (carbon [C], [N], phosphorus [P], and silicon [Si]) of estuarine and coastal diatom Thalassiosira weissflogii. In addition to the dynamics of iron, coastal and estuarine systems are also more dynamic in terms of carbonate chemistry than oceanic systems (Wootton et al., 2008;Dore et al., 2009). This study can provide useful information about first-order biogeochemical processes of nutrients in neritic ecosystems.

Experimental set-up
Seawater used for the experiment (salinity 34.2) was collected from a coastal region in the North Pacific, near Onjuku, Chiba, Japan (35818 ′ N, 140838 ′ E). About 50 l of seawater was filtered through a 0.22-mm cartridge filter (Advantech Co. Ltd, Tokyo, Japan). Stock solutions of macronutrients were passed through a Chelex 100 resin (Bio-Rad, CA, USA) to eliminate trace metals (Price et al., 1988(Price et al., /1989, and then added to the filtered seawater to make final concentrations of 100 mmol l 21 NO 3 2 , 6 mmol l 21 PO 4 2 , and 150 mmol l 21 Si(OH) 4 . The seawater with added nutrients was aged for more than 1 month in an acid-washed 50-l polypropylene carboy, to precipitate dissolved iron excess to its solubility (Sugie et al., 2010;. The aged seawater was passed through a 0.1-mm cartridge filter (Merck Millipore, MA, USA) to sterilize it and to eliminate particulate iron before use. The background iron concentration of the filtered seawater was 0.4 nmol l 21 , as measured by flow-injection with chemiluminescence detection (Obata et al., 1993). The culture medium used in this study was modified Aquil medium (Price et al., 1988(Price et al., /1989, which was composed of nutrient-enhanced 0.1 mm filtered seawater and Aquil metals chelated with 100 mmol l 21 of EDTA. Only iron Effects of pH and Fe on diatom concentrations were modified during the culture experiment. An algal strain of T. weissflogii (Grunow) G. Fryxell and Hasle [currently accepted as Conticribra weissflogii (Grunow) K. Stachura-Suchoples and D.M. Williams; CCMP1336] was maintained in modified Aquil medium at 208C under cool fluorescent light at 150 mmol photons m 22 s 21 (measured using QSL radiometer; Biospherical Instrument Inc., CA, USA), and a 12:12 h light:dark cycle. All equipment used in the culture experiment was acid-washed (soaked for at least 24 h in either 1 or 4 mol l 21 HCl solution) followed by thorough rinsing with Milli-Q water (.18.0 MV cm 21 , Merck KGaA, Darmstadt, Germany). Preparation and sampling were conducted in a class 1000 clean room and at a class 100 clean bench, respectively, to avoid inadvertent trace metal contamination.

Culture design
Iron, other trace metals, and EDTA stock solutions were mixed and left for 1 h in 4-l polycarbonate culture bottles before pouring 3-l of nutrient-enhanced 0.1-mm filtered seawater. Next, CO 2 concentration (xCO 2 )-controlled air (Nissan Tanaka Corp., Saitama, Japan) was injected to manipulate the carbonate chemistry of the culture medium. The xCO 2 of the injected air was set at 386, 614, or 795 ppm, corresponding to the present and two possible future CO 2 conditions, respectively, and these high CO 2 conditions could also be found in the present coastal region during winter (Wootton et al., 2008). The injected air was humidified by passing through Milli-Q water and it was also passed through a 0.2 mm inline filter to avoid contaminations from the gas cylinder or lines. The flow rate of xCO 2 -controlled air was set at 40 ml min 21 for 4 d before the start of the experiment to ensure steady state of carbonate chemistry and to equilibrate [iron -EDTA], [EDTA], and [Fe ′ ] conditions. The [Fe ′ ] was set at 30 or 250 pmol l 21 for low iron (LFe) and high iron (HFe) treatments, respectively. The values for [Fe ′ ] were determined based on a presurvey of the relationship between specific growth rates and [Fe ′ ] ( Figure 1). Non-linear fitting of the growth rate and [Fe ′ ] data to the Monod equation, resulted in a maximum specific growth rate (m max ) ¼ 1.24 + 0.04 (d 21 ) and a half saturation constant for growth (k m ) ¼ 28.1 + 3.2 (pmol l 21 ) (R 2 ¼ 0.98, p , 0.001). The [Fe ′ ] of 30 and 250 pmol l 21 were selected to produce 50 and .90% of the m max of T. weissflogii, respectively. In coastal seawater, given the dissolved iron and iron-binding ligand concentrations of 1 -150 and 10 -500 nmol l 21 , respectively, with (Log K Cond FeL,Fe ′ ) of ligands ranging 8.7 -12 (Kuma et al., 2000;Laglera and van den Berg, 2009;Hassler et al., 2011;Bundy et al., 2015), the [Fe ′ ] could range from 0.1 to 297 pmol l 21 . Given these iron conditions and the dynamics of carbonate chemistry in the coastal region, this study could partially simulates the present coastal environment as well as future high CO 2 conditions. Six treatments based on different xCO 2 and [Fe ′ ] were established as follows: LFe386, LFe614, LFe795, HFe386, HFe614, and HFe795. When estimating [Fe ′ ] according to the method of Sunda and Huntsman (2003), the background concentration of iron was included in the added iron concentrations. Defined [Fe ′ ] were obtained using total iron and EDTA concentrations, pH, light intensity and duration, and temperature. The pH values were calculated from measurements of total alkalinity (TA) and dissolved inorganic C (DIC) concentrations (Table 1). Cultures were set in duplicate bottles for each treatment.
The dilute batch culture technique was used in this study (LaRoche et al., 2011); the diatom cells were harvested ,20% of the carrying capacity of the culture medium to avoid large changes in carbonate chemistry and iron availability. To acclimate and to reduce intracellularly stored iron of the T. weissflogii cells to experimental conditions, cells in stock culture (exponential growth phase) were transferred to each xCO 2 with 1/100 Aquil trace metals medium and grown in the exponential growth phase corresponding to 10 cell divisions. The xCO 2 -acclimated cells were then transferred to the experimental media to give an initial cell density of 100 and 50 cells ml 21 in the LFe-and HFe-treated bottles, respectively. Cultures were continued for 10 and 6 d in the LFe and HFe treatments, respectively, representing 6-7 cell divisions under the experimental conditions. Geider et al. (1993) reported that iron-limited diatom cell could acclimate under highiron conditions within 24 h, and thus, T. weissflogii should acclimate each iron conditions during the experiment. After inoculation of the diatoms into the experimental media, the flow rate of xCO 2controlled air was reduced to 10 ml min 21 .

Measurements
Growth was monitored daily at a fixed time (mid of the light phase) using a Multisizer 4 Coulter Counter (Beckman Coulter Inc., CA, USA) to calculate specific growth rates and to measure cell volume and surface area (SA). Specific growth rates (m, d 21 ) were determined from the slope of a plot of the natural log density of biovolume (i.e. abundance × cell volume) against time during the exponential growth phase. At the end of the experiment, cells were harvested on precombusted GF/F filters for particulate organic C (POC), particulate N (PN), and particulate P (PP) analysis. For biogenic Si (BSi) analysis, samples were collected on a 0.8-mm polycarbonate filter. The methods for POC, PN, PP, and BSi measurements have been described elsewhere . In brief, for POC and PN analysis, filters were freeze-dried, and the concentrations were measured using a CHN analyser. Filters for PP analysis were combusted in high temperature followed by acid hydrolysis and measured using a spectrophotometer as described by Soló rzano and Sharp (1980). For BSi analysis, the filter was digested in 0.5% Na 2 CO 3 solution (Paasche, 1980) and measured with a QuAAtro-2 continuous flow analyser. For chlorophyll a Figure 1. Change in the specific growth rate of T. weissflogii against the change in dissolved inorganic iron (Fe ′ ) concentration. The hatched area represents where solid iron hydroxides are precipitating (e.g. Stumm and Morgan, 1996), and therefore [Fe ′ ] may not be correct. The dotted hyperbolic line was obtained by non-linear fitting of the data on the relationship between the specific growth rate of T. weissflogii and [Fe ′ ] to the Monod equation.

682
K. Sugie and T. Yoshimura (Chl a) measurements, seawater samples were filtered on GF/F filters, soaked in N,N-dimethylformamide (Suzuki and Ishimaru, 1990) and quantified by fluorometry (Welschmeyer, 1994 The DIC and TA samples were collected and measured at days 0, 2, and 6 for HFe conditions and days 0, 3, and 10 for LFe conditions. Both parameters were measured using a potentiometric Gran plot method with 0.1 mol l 21 HCl (Wako Co. Ltd, Osaka, Japan) and a TA analyser (Kimoto Electric Co. Ltd, Osaka, Japan), according to the method of Edmond (1970). Titration data below pH 4 were eliminated because EDTA began absorbing protons below pH 4. The stability of the DIC and TA analyses was checked using DIC reference material (KANSO Co. Ltd, Osaka, Japan) and the analytical errors in this study were ,0.1% for both DIC ( 1.1 mmol kg 21 ) and TA ( 1.4 mmol kg 21 ). The DIC and TA values of the reference material were traceable to the certified reference materials supplied by Professor Andrew Dickson, University of California, San Diego, USA. Carbonate chemistry was calculated using the CO2SYS program and DIC and TA data (Lewis and Wallance, 1998). Samples for macronutrient analysis were collected at the end of the experiment and measured using a QuAAtro-2 continuous flow analyser.

Statistics
Linear regression was used to test the relationship between measured parameters and pCO 2 . When significant linearity was found (p , 0.05), the regression line was described in the figure. When the hypothesis of linearity was rejected, differences among the different xCO 2 treatments were tested with Tukey's HSD ANOVA test. To test the effect of the different [Fe ′ ] of each of three xCO 2 treatments, two-tailed, paired-sample t-tests were performed. All statistics were carried out using PASW statistical software (version 17.0 SPSS Inc., Chicago, IL, USA). Significant differences are reported at the 95% confidence level.

Results
The pH and pCO 2 were clearly different among the xCO 2 tested throughout the experiment with a slight increase in pH and a decrease in pCO 2 during incubations ( Figure 2, Table 1). The [Fe ′ ] was not statistically different among the three xCO 2 treatments under both LFe and HFe conditions (Figure 3, Table 1). The specific growth rate differed significantly between the HFe614 and HFe795 treatments, but there was no significant difference in specific growth rate between the HFe386 and HFe614 treatment or between the HFe386 and HFe795 treatment (Figure 4a). Under LFe conditions, the specific growth rate was slightly increased by the change in carbonate chemistry, but the difference was not statistically significant (p ¼ 0.13, ANOVA; Figure 4a, Table 2). The cell volume decreased linearly with increasing pCO 2 under HFe conditions (Table 2), whereas no linearity was found under LFe conditions (Figure 4b). Under LFe conditions, there was a significant decrease in the cell volume for the LFe614 treatment compared with the LFe386 and LFe795 treatments (Figure 4b).
A  Table 2). Only In cell [P] decreased significantly with increasing pCO 2 under LFe conditions (Supplementary Figure S1c, Table 2). The [Si]/SA did not differ significantly because of a difference in carbonate chemistry under both LFe and HFe conditions (Supplementary Figure S1d). The NCP, NSiP, and NChlaP did not differ significantly among the three xCO 2 treatment in both iron conditions (Supplementary Figure S2a, d, and e). The NNP and NPP under HFe conditions increased significantly with increasing pCO 2 (Table 2), but those trends were not seen under LFe conditions (Supplementary Figure S2b and c). The C:N, C:P, and N:P ratios did not show any significant trends or differences among the three xCO 2 conditions under HFe and LFe conditions (Figure 6a -c). The Si:N, Si:C, and Si:P ratios under HFe conditions decreased significantly with increasing pCO 2 , but the ratios did not differ significantly under LFe conditions (Figure 6d-f, Table 2). Different Fe ′ conditions produced significant differences in most of the parameters measured in this study, except for In cell [P], C:N, and Si:P ratios (Table 3). Lower iron availability decreases the cell quota of C and N synchronously. The different iron conditions affected the Si cell quota and Si/SA, but the degree of the change was small compared with the change for C and N, resulting in increasing Si:C and Si:N ratios with increasing iron availability (Table 3). The Chl a cell quota, In cell [Chl a], and NChlaP decreased dramatically with a decrease in iron availability (Supplementary Figure S2, Table 3).

Discussion
This study shows that the cell size and elemental composition of the diatom T. weissflogii changed with both separate and simultaneous changes in iron availability and carbonate chemistry. The proper control of bioavailable [Fe ′ ] in seawater under different pCO 2 conditions enables the investigation of individual effects of iron availability and carbonate chemistry on the cellular elemental dynamics of phytoplankton species . The present study demonstrates that the ecophysiological properties of the coastal and estuary diatom T. weissflogii changed in response to the change in CO 2 levels, mainly under iron-replete condition.  Figure S1), suggesting the increasing concentration of intracellular constituents. However, the cell quota of C, P, and Chl a did not change (Table 2), because cell size decreased with increasing pCO 2 under HFe conditions. A decrease in cell size in response to high pCO 2 conditions has been reported for micro-and nano-sized phytoplankton such as the diatom Pseudo-nitzschia pseudodelicatissima  and the coccolithophore Emiliania huxleyi (Lohbeck et al., 2012). Previous studies have demonstrated that the increase in temperature also decrease cell size of marine phytoplankton 2.5 -4% 8C 21 (Montagnes and Franklin, 2001;Atkinson et al., 2003). According to the IPCC AR5 (IPCC, 2013), temperature increase in the range 1 -38C (RCP2.6) or 3 -108C (RCP8.5) has been predicted with atmospheric CO 2 concentrations ranging from 421 to 936 ppm by 2100. Such temperature rises could decrease phytoplankton cell size by 2.5 to .40%. Our study estimated the decrease in cell volume to be 1.4 × 10 22 % per pCO 2 increases ( Table 2), suggesting that the increase in CO 2 levels in 2100 from the present conditions ( 400 matm) could decrease cell volume by 0.3 -7.5% under RCP2.6 and RCP8.6 scenarios, respectively. These estimations suggest that the effect of temperature on cell volume is stronger

684
K. Sugie and T. Yoshimura than pCO 2 , but pCO 2 increase could contribute substantially (12 -19% under RCP2.6 and RCP8.6 scenarios, respectively) to the decrease in cell volume of T. weissflogii in the forecasted future conditions, given the decrease in cell size emerge due to the additive effects of hightemperature and pCO 2 under iron-replete conditions. Seebah et al. (2014) reported that the aggregate size and sinking velocity of T. weissflogii decreased under high pCO 2 , which can increase the remineralization of organic matters in the water column. The reduction in cell size found in the present study may partly explain the sinking behaviour of T. weissflogii reported by Seebah et al. (2014). In addition to the slower sinking rate of T. weissflogii under high CO 2 conditions (Seebah et al., 2014), the relatively stable cell quota of C, N, P, and Chl a suggest that increasing pCO 2 could have negative impact on the export flux of these bioelements in the high CO 2 conditions. In contrast, the [Si]/SA and NSiP were not affected significantly by changes in pCO 2 (Table 2).  reported that the [Si]/SA and NSiP of P. pseudodelicatissima decreased with increasing pCO 2 under various [Fe ′ ] conditions, suggesting that the effect of increasing CO 2 levels is species-specific or differs between centric and pennate diatoms. Milligan et al. (2004) examined the Si metabolism of T. weissflogii under iron-replete conditions and found that the rates of intracellular Si efflux and frustule dissolution increased in 750 ppm CO 2 compared with 100 ppm CO 2 . During that experiment, the Si cell quota did not differ significantly between 380 and 750 ppm CO 2 because of the high Si uptake rate under the high CO 2 conditions. The small but insignificant decrease in [Si]/SA with increasing pCO 2 under LFe conditions may result from the pH-dependent change in Si metabolism as previously reported (Milligan et al., 2004;Hervé et al., 2012). Muggli et al. (1996) found that phytoplankton cells suffered from energy limitation under iron-limited conditions. Because Si uptake and mineralization are associated with energy from respiration (Martin-Jézéquel et al., 2000), the [Si]/SA may be slightly affected by changing pCO 2 under conditions of low iron availability.

Effects of different CO 2 levels on elemental compositions
The present study confirms that the cellular C:N, C:P, and N:P ratios do not change with changes in pCO 2 under iron-replete or iron-limited conditions (Figure 6). Previous studies also observed   Tables 2 and 3. Labels above the symbols represent the results of Tukey's group test (p , 0.05), which did not show a linear trend.
Effects of pH and Fe on diatom that these ratios of T. weissflogii were not affected by changes in pCO 2 under both macronutrient-replete and -depleted conditions (Burkhardt et al., 1999;Clark et al., 2014). In the studies by Burkhardt et al. (1999) and Clark et al. (2014), carbonate chemistry was manipulated by adding acid or base, whereas in the present study, it was manipulated by injecting xCO 2 -controlled air. The method of carbonate chemistry manipulation is reported to be one of the most important factors affecting biological responses to ocean acidification (Shi et al., 2009;Hoppe et al., 2011). The major difference between these manipulation methods is DIC concentrations: the bubbling of high CO 2 air increases DIC, whereas the acid/base addition did not change DIC (Burkhardt et al., 2001;Shi et al., 2009). Taken together, the present and previous findings indicate that increasing CO 2 levels have no effect on the C:N, C:P, and N:P ratios of T. weissflogii, irrespective of the different manipulation methods.
However, some previous studies have reported that cellular C:P and N:P ratios of phytoplankton are affected by increases in pCO 2 under iron-replete conditions. Many centric and pennate diatom species showed an increase in C:P ratio in response to increases in pCO 2 (Burkhardt et al., 1999;King et al., 2011;Sun et al., 2011;). An increase in the C:P ratio with increasing pCO 2 has also been found in Dinophyceae, Raphidophyceae, and Prymnesiophyceae Fu et al., 2008). Some diatom species such as Asteionellopsis glacialis and Phaeodactylum tricornutum did not show changes in the C:P ratio with changing pCO 2 (Burkhardt et al., 1999), similar to the finding for T. weissflogii in the present study. Sugie and Yoshimura (2013) demonstrated using P. pseudodelicatissima that a steeper decline in In cell [P] and NPP compared with those of C and N with increasing pCO 2 is a key mechanism underlying increasing C:P and N:P ratios in high CO 2 conditions. They suggested that increases in C:P and N:P ratios with increasing CO 2 availability were derived from decreases in the expenditure of P-rich cellular constituents such as ATP, which is used to maintain/ operate carbon concentration mechanisms (CCMs; Hopkinson et al., 2011). A notable characteristic of C assimilation in T. weissflogii is the presence of both C 3 and C 4 photosynthetic pathways, where the latter functions as a biochemical CCM. In contrast, other diatoms are mainly C 3 photosynthesizers (Roberts et al., 2007). The initial products of DIC uptake are glycerate-P and hexose-P in C 3 diatoms, and malate in C 4 diatoms of T. weissflogii (Roberts et al., 2007). If C 4 pathway-mediated CCMs decrease with increasing CO 2 availability, the contribution of sugar-P compounds such as glycerate-P mediated from C 3 photosynthesis can increase with increasing CO 2 availability. Such adaptive photosynthetic strategies result in an increase in NPP (Supplementary Figure S2), which could offset increases in C:P and N:P ratios of T. weissflogii with increasing pCO 2 .
In contrast to the relatively stable C:N, C:P, and N:P ratios in the present study, we found that the ratios of Si to C, N, and P changed in response to changes in carbonate chemistry ( Figure 6). This discrepancy reflected different trends in the effect of pCO 2 on the rates of NEP between Si and C, N, and P (Supplementary Figure S2). Under HFe conditions, the decrease in Si:N and Si:P with increasing pCO 2 was derived from the significant increases in NNP and NPP with increasing pCO 2 . The decreasing trend of Si:C with increasing pCO 2 resulted from the small but insignificant (p ¼ 0.16, ANOVA) Table 2. Statistical results of the relationships between partial pressure of CO 2 (pCO 2 , matm) and specific growth rate (m, d 21 ), cell volume (CV, mm 3 ), intracellular organic carbon (In cell [C], mol l 21 ), intracellular nitrogen (In cell [N], mol l 21 ), intracellular phosphorus (In cell [P], mol l 21 ), silicon concentration per unit surface area ([Si]/SA, mmol m 22 ), intracellular chlorophyll a (In cell [Chl-a], mg l 21 ), Net C, N, P, Si, and chlorophyll a productions (NCP, NNP, NPP, NSiP, and NChl-aP; pmol cell 21 d 21 for C, N, and Si, and fmol cell 21 d 21 for P, pg cell 21 d 21 for chlorophyll a), and elemental compositions. Decreasing Si relative to C, N, and P in response to increasing pCO 2 has been previously reported for diatoms (Pseudo-nitzschia multiseries, Sun et al., 2011;P. pseudodelicatissima, Sugie and Yoshimura, 2013). It has also been reported that cell cycledependent uptake of Si and C, N, and P were uncoupled (Claquin et al., 2002). The present results further support such uncoupling and indicate that ratios of Si to other nutrients in diatoms can change in response to changes in environmental conditions including carbonate chemistry. Under LFe conditions, elemental compositions did not differ significantly among the three pCO 2 conditions because of constant  Figure S2), which are clearly different from the results obtained in the HFe conditions ( Figure 6). This discrepancy may be caused by unexplored physiological mechanisms. We assume that the processes responsible for the increase in NEP with increasing pCO 2 as was found in the HFe conditions require additional energy for nutrient uptake. Under the energy-limited LFe conditions (Muggli et al., 1996), the CO 2 -related change in NEP was regulated by the iron-limitation. Hoppe et al. (2013) also observed that the effects of high CO 2 conditions on natural phytoplankton communities in the Southern Ocean were minimal under iron-depleted (siderophore desferrioxamine B added) conditions compared with iron-replete conditions. However, the rates of change in NEP of the diatom P. pseudodelicatissima were affected by changes in pCO 2 under conditions of low iron availability . The notable difference between T. weissflogii and P. pseudodelicatissima is k m value that T. weissflogii and P. pseudodelicatissima showed k m ¼ 28 and 1.5 -5.0 pmol [Fe ′ ] l 21 , respectively. These results  Tables 2 and 3. 688 K. Sugie and T. Yoshimura indicate that responses to the interaction between changes in pCO 2 and changes in iron availability vary among species depending on the affinity for uptake of [Fe ′ ]. Previous studies have reported that the elemental composition of diatoms is not always related linearly to iron availability (Price, 2005;Bucciarrelli et al., 2010). More experiments manipulating various Fe ′ and pCO 2 conditions using more phytoplankton species are required to better understand the effect of different CO 2 conditions on biogeochemical cycling of bioelements.

Effect of different iron availability on ecophysiology of T. weissflogii
Decreasing iron availability affected most of the measured parameters of T. weissflogii, indicating that iron availability is more important to determine ecophysiological processes of diatoms than variations in carbonate chemistry. Yoshimura et al. (2014) also suggested that changes in pCO 2 do not play a major role in the dynamics of particulate and dissolved organic matter produced by subarctic plankton communities during the exponential growth phase. Price (2005) reported that the C:N:Si:P ratio of T. weissflogii changed from 97:14:4.7:1 to 70:10:5.9:1 under iron-replete and iron-limited conditions, respectively. In the present study, we obtained similar results under HFe (96:17:9.7:1) and LFe (80:15:8.8:1) conditions in the 386 ppm xCO 2 treatments. By calculating Q E ratios (RQ E ) between LFe386 and HFe386 (RQ E ¼ Q E-LFe /Q E-HFe , where Q E-LFe and Q E-HFe represent cell quota of each element in LFe386 and HFe386 treatment, respectively), we obtained 0.76, 0.78, 0.90, and 0.82 for RQ C , RQ N , RQ P , and RQ Si , respectively (Supplementary Figure S3). This estimate suggests that Q C , Q N , and Q Si are highly variable compared with Q P in response to the different iron conditions. In addition, the changes in carbonate chemistry modulated the elemental composition of T. weissflogii possibly via uncoupling of the cell cycle-dependent uptake of each bioelement (Claquin et al., 2002). Specifically, the difference in the Si:N ratio between HFe and LFe conditions increased with increasing pCO 2 (Figure 6d). Such unexplored flexible nutrient stoichiometry may be a consequence of optimality of nutrient uptake or assimilations (Milligan et al., 2004;Smith et al., 2011), the mechanisms of which should be incorporated into the model simulations to better predict the biogeochemical cycling of nutrients.

Relevance of this study for ocean acidification study
Results of our short-term experiment contribute to better understanding of first-order biogeochemical processes of nutrients in dynamic coastal environment, but the progress of ocean acidification in nature is a slower process compared with our short-term experiment. Extrapolating our results to the future acidified environment require attention because long-term adaptive responses often differ from short-term phenotypic plasticity (Lohbeck et al., 2012;Collins et al., 2014). Crawfurd et al. (2011) and Tatters et al. (2013) reported that there was no clear difference in the effect of high CO 2 conditions on the ecophysiology of diatoms, when comparing the results of short-term plastic response with acclimation/adaptation responses for .100 cell divisions. On the other hand, Torstensson et al. (2015) reported that the specific growth rate decreased in the high CO 2 conditions only after the acclimation of 60 cell divisions. Because phenotypic plasticity may be important for the subsequent adaptation or evolution of organisms (Collins et al., 2014), plastic response found in this study could provide relevant information for the future acidified ocean ecosystem. However, there is very limited information available to date concerning adaptation of diatoms, and no reports are available on the adaptive response of cell size and cellular elemental content of diatoms. We need more information on the long-term effect of CO 2 on phytoplankton ecophysiology.

Conclusions
Our study demonstrated that some ecophysiological properties of coastal diatom T. weissflogii changed in response to changing CO 2 levels under iron-replete conditions. In coastal environment, iron concentration is generally higher than the oceanic region (Kuma et al., 2000;Nishioka et al., 2013;Bundy et al., 2015), suggesting the change in the ecophysiological properties of T. weissflogii could occur with increasing CO 2 levels in the modern coastal seas (Wootton et al., 2008). These changes may affect the food availability of higher trophic levels, sinking behaviour of T. weissflogii, and nutrient availability for other phytoplankton in the coastal regions. However, in addition to the differences in the response of individual phytoplankton species to changes in CO 2 levels (e.g. Doney et al., 2009;Collins et al., 2014), previous studies of the impacts of increasing CO 2 levels on the elemental composition of natural plankton communities have produced conflicting results, including no effect (Martin and Tortell, 2006;Feng et al., 2010;Tortell et al., 2010), increased particulate Si:N and Si:C ratios , and inconsistent changes in the production of dissolved and particulate organic matters (Yoshimura et al., 2010(Yoshimura et al., , 2014. To interpret these results, it is necessary to improve knowledge of the effects of carbonate chemistry and iron availability on phytoplankton processes using both natural phytoplankton communities and laboratory-based unialgal cultures. Table 3. Results of two-tailed, paired-sample t-test (n ¼ 12) to test the effect of iron availability on the measured parameters, comparing the LFe (Fe ′ ¼ 26 pmol l 21 ) and the HFe treatments (Fe ′ ¼ 240 pmol l 21 ). Effects of pH and Fe on diatom