Low Fe Availability for Photosynthesis of Sea-Ice Algae: Ex situ Incubation of the Ice Diatom Fragilariopsis cylindrus in Low-Fe Sea Ice Using an Ice Tank

Sea-ice algae play a crucial role in the ecology and biogeochemistry of sea-ice zones. They not only comprise the base of sea-ice ecosystems, but also seed populations of extensive ice-edge blooms during ice melt. Ice algae must rapidly acclimate to dynamic light environments, from the low light under sea ice to high light within open waters. Recently, iron (Fe) deficiency has been reported for diatoms in eastern Antarctic pack ice. Low Fe availability reduces photosynthetic plasticity, leading to reduced ice-algal primary production. We developed a low-Fe ice tank to manipulate Fe availability in sea ice. Over 20 days in the ice tank, the Antarctic ice diatom Fragilariopsis cylindrus was incubated in artificial low-Fe sea ice ([total Fe] = 20 nM) in high light (HL) and low light (LL) conditions. Melted ice was also exposed to intense light to simulate light conditions typical for melting ice in situ. When diatoms were frozen in, the maximum photochemical quantum efficiency of photosystem II (PSII), Fv/Fm, was suppressed by freezing stress. However, the diatoms maintained photosynthetic capability throughout the ice periods with a stable Fv/Fm value and increased photoprotection through non-photochemical quenching (NPQ) via photoprotective xanthophyll cycling (XC) and increased photoprotective carotenoid levels compared to pre-freeze-up. Photoprotection was more pronounced in the HL treatment due to greater light stress. However, the functional absorption cross section of PSII, σPSII, in F. cylindrus consistently increased after freezing, especially in the LL treatment (σPSII > 10 nm2 PSII–1). Our study is the first to report such a large σPSII in ice diatoms at low Fe conditions. When the melted sea ice was exposed to high light, Fv/Fm was suppressed. NPQ and XC were slightly upregulated, but not to values normally observed when Fe is not limiting, which indicates reduced photosynthetic flexibility to adapt to environmental changes during ice melt under low Fe conditions. Although ice algae can optimize their photosynthesis to sea-ice environments, chronic Fe starvation led to less flexibility of photoacclimation, particularly in low light conditions. This may have detrimental consequences for ice algal production and trophic interactions in sea-ice ecosystems if the recent reduction in sea-ice extent continues.


INTRODUCTION
Sea ice is one of the largest biomes on Earth and a significant driver of the biogeochemistry of polar oceans (Arrigo, 2017;van Leeuwe et al., 2018). Sea ice harbors diverse and productive microbial communities that occur mainly at the bottom of sea ice (Meiners et al., 2012van Leeuwe et al., 2018). Ice algae, sympagic microalgae proliferating near the bottom of sea ice, are a major primary producer, estimated to contribute 9-25% of the annual primary production (PP) in perennial ice zones (e.g., Legendre et al., 1992;Arrigo et al., 1997;Arrigo, 2017). In addition, they also contribute more than 50% of the total PP in ice-covered zones (Satoh et al., 1989;McMinn et al., 2010;Fernández-Méndez et al., 2015). When sea ice melts, algal cells are released from the sea ice and can seed extensive ice-edge microalgal blooms (Smith and Nelson, 1986;Syvertsen, 1991). Ice algae therefore are a significant player in the ecology and biogeochemistry of sea ice and marginal ice zones. However, sea ice is a challenging environment for photosynthesis due to low temperatures, high salinities, low light availability, and reduced access to nutrients. In these conditions, light and nutrient availability effectively control photosynthesis in ice algae, although salinity stress can also be important. Pankowski and McMinn (2008) first reported iron (Fe) limitation in Antarctic pack ice in spring/summer. They suggested that it was due to the low Fe supply from the Fe-deficient surface waters of the Southern Ocean, where off the continental shelf, the average concentration of dissolved Fe (DFe), which is considered the most bio-available form (Lannuzel et al., 2014), is 0.31 ± 0.45 nM (Tagliabue et al., 2012). It is well-known that Fe is a crucial element for photosynthesis in microalgae, as it is required for pigments and cytochrome synthesis, nitrate reduction, and the detoxification of reactive oxygen species (ROS) (Geider and LaRoche, 1994;Sunda and Huntsman, 1995;Behrenfeld and Milligan, 2013;Twining and Baines, 2013). Ice algae are able to acclimate to the dark environment frequently encountered in sea ice by increased synthesizing of chlorophylls and carotenoids to maximize the efficiency of light utilization (e.g., Falkowski and Owens, 1980;Morel and Bricaud, 1981;Bricaud et al., 1995). However, pigment synthesis requires Fe; the low availability of Fe can thus be exacerbated due to the antagonistic Fe-light colimitation (i.e., more Fe required in a darker environment; Sunda and Huntsman, 1997;Maldonado et al., 1999;Moore et al., 2007). The reduced cellular chlorophyll a and accessory carotenoids due to Fe-limited conditions (e.g., Greene et al., 1992;Behrenfeld and Milligan, 2013) would lead to serious photo-damage by ROS because Fe-limited cells cannot cope with the excess absorbed energy that overruns the reduced electron transport capacity due to decreased Fe-containing electron donors (e.g., cytochromes and Photosystem I; van Oijen et al., 2004;Petrou et al., 2014;Roncel et al., 2016). This would pose a significant challenge when Fe-limited ice algal cells are exposed to high light during release from the bottom of the sea ice into open waters during ice melt (i.e., the setting of ice-edge blooms). It is thus ecologically and biogeochemically important to investigate how Fe availability and light combine to control the photosynthetic performance of ice algae to further understand their role in sea-ice and marginal ice-zone ecosystems.
Sea ice can act as a reservoir of Fe, especially close to lithogenic coastal iron sources in the case of fast ice (e.g., de Jong et al., 2013), but also in pack ice far from the coast (Lannuzel et al., 2016b). Pack ice might accumulate biogenic Fe by directly incorporating it during growth in autumn and winter (Lannuzel et al., 2016a) or by transferring DFe from seawater into particulate Fe (PFe) (e.g., van der Merwe et al., 2011). According to the limited data available, pack-ice DFe concentrations range from about 0.2-36.8 nM (average 5.9 ± 6.2 nM from 40 cores; Lannuzel et al., 2007Lannuzel et al., , 2008Lannuzel et al., , 2016avan der Merwe et al., 2011;de Jong et al., 2015;Janssens et al., 2016), up to about two orders of magnitude greater than the underlying seawater (Tagliabue et al., 2012). When sea ice melts in spring and summer, 70% of the DFe can be lost within 10 days (Lannuzel et al., 2008), but PFe, which is about an order of magnitude greater in average concentration, needs to be physically released from both the ice, and gellike exopolymeric substances (EPS) lining the brine network (Krembs et al., 2002;van der Merwe et al., 2011;Lannuzel et al., 2013Lannuzel et al., , 2016b. These gel-like EPS are produced in situ by seaice organisms in response to the evolving sea-ice environment and could in turn control the solubility of Fe transferred back to the dissolved fraction (Thomas and Dieckmann, 2002;Lannuzel et al., 2015Lannuzel et al., , 2016bGenovese et al., 2018), possibly maintaining in-ice DFe within the concentration range described.
With regard to climate change, it has been reported that ocean acidification (OA), associated with the increasing anthropogenic CO 2 input, modifies Fe availability (Millero et al., 2009). OA can suppress Fe availability by stabilizing Fe-ligand complexation (Sunda and Huntsman, 2003;Shi et al., 2010) and also extending the oxidation rate of Fe(II), which is believed to be the most prolific bio-available source of Fe (Millero et al., 1987(Millero et al., , 2009Breitbarth et al., 2009). Although Boyd et al. (2015) estimated that climate change enhances Fe availability in the Southern Ocean, sea ice has greater fluctuations in pH than in pelagic waters (Matson et al., 2014;McMinn et al., 2014). In addition, future predictions of sea-ice algal biomass and photosynthesis in the changing polar oceans are one of the major missing pieces in sea-ice biogeochemical models (Vancoppenolle et al., 2013;Constable et al., 2014;Steiner et al., 2016;van Leeuwe et al., 2018). Recently, Tedesco et al. (2019) has demonstrated future changes in algal productivity in the Arctic combining a biogeochemical model for sympagic algae with sea-ice drivers from an ensemble of 18 CMIP5 climate models. However, little is known about temporal changes in sea-ice algal biomass and productivity in the changing Southern Ocean.
Methodological limitations to the investigation of ice-algal photosynthesis, however, still remain-almost all studies so far have worked on ice algal photosynthesis in the laboratory, during which ice algae are resuspended in a water medium after being melted out from ice samples. This melt process hinders obtaining realistic photophysiological information because water media cannot reproduce the temperatures in sea ice (<−1.8 • C). Kennedy et al. (2012) and Kameyama et al. (2020) successfully incubated ice algae in artificial sea ice produced in laboratory ice tanks. Ice tank techniques enable ex situ incubation of ice algae within ice and manipulation of the ice environment by maintaining a stable ice thickness. Yoshida et al. (2020) also incubated the model polar diatom Fragilariopsis cylindrus under Fe-replete condition ([total dissolvable Fe] = 400 nM) in artificial ice using an ice tank. Our previous study suggested that ice diatoms, with high Fe availability, flexibly acclimated their physiology to the dynamic ice environments described above. However, photosynthetic responses of ice algae to low Fe availability have not been well-understood yet. Here, we use a newly developed low-Fe ice tank, made of titanium, to control Fe concentration in the ice. Using this novel low-Fe ice tank, we attempted to investigate how low Fe availability in sea ice affects the photosynthetic plasticity of ice algae exposed to dynamic fluctuations in light availability, from extremely low light when sea ice is present, to intense light exposure during the ice melts. Here, the polar diatom F. cylindrus, a model ice diatom with a fully sequenced genome, was used (Mock et al., 2017), enabling the application of molecular techniques to understand the underlying mechanisms driving photophysiology. Photophysiological responses of F. cylindrus were monitored with variable chlorophyll (chl) a fluorescence and their pigment composition throughout the ice tank experiments.

Ice Tank Incubation and Preparation of Low-Fe Medium
The polar pennate diatom Fragilariopsis cylindrus, isolated from Antarctic pack ice in 2015 at Davis Station, East Antarctica (Kennedy et al., 2019), was incubated in a purpose-designed low-Fe ice tank (Island Research, Tasmania; see details in Yoshida et al., 2020). The low-Fe ice tank, which was constructed of titanium to minimize Fe contamination, was placed into a freezer (-20 • C). Ice thickness and temperature gradient were controlled by interactions between a basal heater and the adjustable ambient freezer temperature. This enabled an ice thickness of approximately 5.5 cm to be maintained during the experiment. Incubations were conducted in Aquil media (Price et al., 1989;Pankowski and McMinn, 2009;Yoshida et al., 2020) buffered with ethylenediaminetetraacetic acid (EDTA) (final concentration 20 µM) at a salinity of 35 and at 150 and 30 µmol photons m −2 s −1 (PAR; white LED, PANEL-300-18W; LED Lighting Products, Sydney, NSW, Australia) as high light (HL) and low (LL) treatments, respectively. The diatom was pre-incubated in the same low-Fe Aquil media at 2.5 • C and 150 µmol photons m −2 s −1 before inoculation of the ice tank. DFe concentrations were set at 20 nM, which is within the concentration range of 0.2-36.8 nM that has been found in pack ice (Lannuzel et al., 2007(Lannuzel et al., , 2008(Lannuzel et al., , 2016avan der Merwe et al., 2011;de Jong et al., 2015;Janssens et al., 2016), where the concentrations of total inorganic forms of Fe (Fe ) were 4.0 pM calculated using the software Visual MINTEQ, ver. 3 1 . Pankowski and McMinn (2009) confirmed, with Fe-related protein analysis, that low Fe stress was evident at 15.5 pM of Fe . Two independent ice tank runs were conducted for each light treatment. Results of the low-Fe ice tank runs were compared with those from Fe-replete ice tank experiments ([DFe] = 400 nM) of Yoshida et al. (2020).
The low-Fe experiment commenced with the incubation of F. cylindrus in seawater at 2.5 • C after 3-day acclimation to the ice tank environment. A seawater sample was obtained to assess the original physiological state of the algae (day-05, hereafter). A freezing cycle was then started by setting the ice tank to −1.8 • C to initiate ice formation. When ice had formed on day 2, the under-ice water was partially replaced with ultrapure water to adjust the salinity to 35; because the brine rejection from sea ice had increased the salinity to ∼38. After a 2-day acclimation to the new salinity, ice samples were obtained every 5 days for 20 days (i.e., days 0, 5, 10, 15, and 20; Figure 1). Ice samples were randomly recovered with a trace metal-free hand drill (2 cm in diameter) to minimize the heterogeneity among the samples (Yoshida et al., 2020). After all ice sampling at day 20, the remaining ice in the tank was completely melted at 2.5 • C to assess the stress of ice melt and high light exposure at 800 µmol photons m −2 s −1 using the white LED panel (hereafter Melt and Light samples, respectively) (Figure 1). The ice tank incubation procedures are shown in Yoshida et al. (2020). Because it was difficult to obtain the concentrations of macro-nutrient samples from the thin artificial ice (5.5 cm), macro-nutrient concentrations of under-ice seawater were measured. It was assumed that the under-ice seawater was the only nutrient source for ice algae in the ice tank. An unstable brine salinity profile and brine volume fractions in excess of percolation threshold (shown in Yoshida et al., 2020) indicated that brines should have been exchanging with the under-ice seawater (Golden et al., 1998). At the beginning and end of the ice tank runs (days −05 and Melt, respectively), concentrations of nitrate and phosphate in under-ice seawater were determined using a QuickChemR 8000 Automated Ion Analyzer (LaChat Instruments) (Britton et al., 2019). Silicate concentrations of under-ice seawater were determined following Parsons et al. (1984) with the alternate reductant, L-ascorbic acid (Sigma), using a spectrophotometer (Model S-22, BOECO). Exponential growth rate (r; d −1 ) of F. cylindrus in the ice was calculated (Wood et al., 2005): where N Day20 and N Day00 were algal cell abundance (cells mL −1 ) at day 20 and day 00, respectively.

Total Dissolvable Fe Concentration
Concentrations of total dissolvable Fe (TDFe) in the under-ice seawater were monitored to check whether contamination by external Fe (e.g., sampling-induced aerosol deposition on the ice) had occurred. Fe concentrations in the ice were difficult to determine because the ice was too thin to obtain a brine sample. In addition, the Fe source for the ice algae in the artificial sea ice would have been restricted to from the underice water. Fe concentrations in the water samples were thus determined by the Ferrozine colorimetric method in a laminar flow hood (Stookey, 1970;Farid et al., 2018). Fe samples were placed into a 60 mL low-density polyethylene (LDPE), previously acid-cleaned, bottle (Thermo Fisher Scientific) following the GEOTRACES protocol (Cutter et al., 2017). Fe samples were acidified to pH < 2 with Suprapur HCl (Merck) at least 2 months before the measurements (Farid et al., 2018). A 50 mL ferrozine cocktail was prepared with 10 mM Ferrozine (Sigma), 1.44 M hydroxylamine hydrochloride (Trace metal analysis grade, Wako) and ultrapure water acidified with ultrapure hydrochloric acid (HCl) (Ultrapure grade, Wako). The 0.4 mL of ferrozine cocktail was added to a 20 mL sample. The sample was heated at 70 • C for 15 min to accelerate the reduction of Fe(III) to Fe(II) and for detaching Fe(II) from the Fe-EDTA complex. After cooling down the sample, 0.4 mL of an ammonium acetate (Fe analysis grade, Wako) buffer solution, prepared with an ammonia solution (Trace metal grade, Wako), was added to the sample. The buffered sample was incubated at room temperature for 24 h to fully recover Fe in the sample (Farid et al., 2018). Fe concentrations were determined with a spectrophotometer measuring the absorbance of Fe(II)-ferrozine complexes at a wavelength of 562 nm (UV-2450, Shimadzu) with a 50 mm path length glass tube. An Fe standard curve was drawn with a series of known Fe concentrations by diluting an Fe standard solution (Wako) (r> 0.999, n = 7). The detection limit of the measurement was 10 nM, which was consistent with previous studies (Farid et al., 2018). The quantification limit, defined as 3 SD, was 15 nM.

Ice Structure
Prior to the low-Fe incubation experiments, an ice section was collected to assess the ice structure. Artificial sea ice was produced using the same seawater medium, and a 5 cm × 10 cm × 5.5 cm ice sample was collected from the ice tank using a metal saw. The ice sample was thinned to a 0.5 cm thick vertical section using a band saw (BARNES Junior, BARNCO, Australia) in a −20 • C temperature-controlled laboratory. The ice section was planned smooth with a handheld microtome blade before being photographed between cross-polarizing filters. Ice structural measurements were performed in accordance with Langway (1958) and Wongpan et al. (2018).

Fast Repetition Rate (FRR) Fluorometry
A bench-top type Fast Repetition Rate fluorometer (FRRf; FastOcean Act2Run Systems, Chelsea Technologies) was used to monitor the photophysiology of F. cylindrus during the ice tank experiments. Variable chlorophyll a fluorescence data were processed with Act2Run software (Chelsea Technologies). Melted ice samples were buffered with filtered seawater (FSW) (ice: FSW = 1: 1) at 2 • C and kept in the dark for 30 min. Variable chl a fluorescence of the melted ice samples was measured with a single turnover protocol. Briefly, one hundred 2-µs flashlets at a wavelength 450 nm excited reaction centers of PSII (RCII) with 2 µs intervals, and twenty 1-µs flashlets were used for relaxation. Eighteen light steps were applied to generate a rapid light curve (RLC) from 0 to 800 µmol photons m −2 s −1 . Each step took ∼15 s; one RLC was thus completed in 5 min. To obtain chl a fluorescence parameters, six induction and relaxation curves were averaged and fitted to the model of Kolber et al. (1998 ;  Table 1). Absolute electron transport rate through the reaction centers of PSII (RCII) (ETR RCII ) was calculated as follows: following Suggett et al. (2011) and Schuback et al. (2016), wherein ETR RCII is absolute electron transport rate through RCII, E is actinic light intensity, σ PSII is functional absorption cross section of PSII, the fluorescence ratio was effective quantum yield of PSII photochemistry at E, RCII is the quantum yield of RCII assumed as 1, and the following number is a conversion factor to mol quanta mol RCII −1 s −1 , respectively. The resultant ETR RCII -E relationship was fitted to the model of Platt et al. (1980) to obtain photosynthesis-irradiance parameters ( Table 1). We adopted NPQ NSV , non-photochemical quenching based on the normalized Stern-Volmer (S-V) coefficient, to assess the heat dissipation from algal cells following McKew et al. (2013): where F o is minimum fluorescence yield after a relaxation sequence following Oxborough and Baker (1997 ; Table 1). NPQ NSV were measured at the incubation light intensities (i.e., HL: 150 µmol photons m −2 s −1 ; LL: 30 µmol photons m −2 s −1 ) throughout the ice tank experiments.

Pigment Composition
Algal chlorophylls and carotenoids were quantified with Ultra-High Performance Liquid Chromatography (UHPLC) . Fast and directly melted ice samples and seawater samples were filtered onto a 25 mm GF/F filter (Whatman) with gentle vacuum (<0.013 MPa) passing through a 25 mm polypropylene in-line filter holder (Swinnex, Merck). Fast direct melting has little effect on pigment analysis (Rintala et al., 2014;Yoshida et al., 2020). The filter was flash frozen in liquid nitrogen and stored in a deep freezer (−80 • C). Prior to the UHPLC pigment quantification, the thawed filter was blotted dry with filter paper and bead-beat in N, N-dimethylformamide (DMF) to extract pigments . Extracted pigments suspended in DMF were injected into an UHPLC. As an index of photoprotection, ratio of diadinoxanthin (DD) and diatoxanthin (DT) was calculated as DT/(DD + DT), xanthophyll deepoxidation state (DES) (Katayama and Taguchi, 2013;Katayama et al., 2017;Yan et al., 2019). Biomass of algae was assessed as total chl a (Tchl a: sum of chl a, chlorophyllide a, chl a-allomer, and chl a-epimer) concentrations, whereas contributions of chlorophyllide (chlide) a to Tchl a were also calculated as an index of breakdown of chl a. As indices of photoprotective potentials of Gene Expression of Photosynthesis-Related Genes, psbA and rbcL Gene expression measurement shows the potential mechanisms of responses to drastic changes such as freezing and melting of sea ice. The two photosynthesis-related genes, psbA and rbcL, were targeted for gene expression measurements. The psbA and rbcL genes encode the D1 protein in PSII reaction center and the large subunit of ribulose-1,5-bisphosphate carboxylase/oxygenase (RuBisCO), respectively. To stabilize the RNA, RNAlater (Sigma) was immediately added to the ice core after collection. Melted ice and seawater samples were filtered onto two 25 mm, 2 µm polycarbonate Isopore membrane filters (Millipore) with gentle vacuum (<0.013 MPa) passing through a 25 mm polypropylene in-line filter holder (Swinnex, Merck). RNA samples were suspended in 600 µL RLT buffer (Qiagen) in a cryotube, to which 10 µL of β-mercaptoethanol (Sigma-Aldrich) was added. Both RNA and DNA samples were placed in a cryotube, flash frozen in liquid nitrogen and stored in a deep freezer (−80 • C). DNA was extracted following Endo et al. (2013), and RNA was extracted following Endo et al. (2015). The extracted RNA was reverse transcribed to complementary DNA (cDNA) with the PrimeScriptTM RT Master Mix (RR036, Takara) reagent according to the manufacturer's specification. Quantitative PCR (qPCR) and quantitative reverse-transcribed PCR (qRT-PCR) were performed to determine the copy number of DNA and cDNA, respectively. Gene expression of the psbA and rbcL genes were calculated as a ratio of cDNA/DNA. The primer sets and PCR conditions are shown in Supplementary Table S1.

Statistical Analysis
Statistical analyses were conducted using the SigmaPlot software program ver. 11.0 (SysStat Software, Inc., San Jose, CA, United States). One-way ANOVA with Tukey's test was performed on obtained data. Variations in a parameter were tested at the individual sampling time point in each light treatment. A difference in a parameter was regarded as significant if p< 0.05. Additionally, two-way ANOVA with Tukey's tests was also performed at individual sampling time points between light treatments to identify variations in the parameters if no interaction between light availabilities and sampling time points was observed. The results of two-way ANOVA tests were also found in Supplementary Table S2. Shapiro-Wilk's and Levene's tests were performed to confirm the normality and equal variance of the data after normalization with the R statistical software using the package MASS. No statistical analysis was conducted for the results of the Fe-replete ice tank experiments, which were shown for comparison with this low-Fe study. Statistical results of the Fe-replete ice tank runs can be found in Yoshida et al. (2020).

Ice Physics and Ice Algal Growth
The ice thickness was stable at 5.5 cm with little basal ice melting or sublimation of ice from the ice-air interface. During the experiments, the temperature within the ice matrix increased from the ice-air interface (−22.5 • C) to the ice-water interface (−2.2 • C), and seawater temperature beneath was maintained at ∼ −1.8 • C throughout the incubations, as also found in Yoshida et al. (2020). The artificial sea ice growth had commenced with the settling out of individual frazil crystals at the surface of the cooled water column. This layer of fine crystals accumulated to a thickness of 1 cm before downward-growing larger crystals were quickly geometrically selected which favored downward growth. Indeed, the latter resembled columnar ice growth and contained brine channels and pockets (Figure 2). There were also some incorporated granular crystals, which indicate turbulence, in the under-ice seawater, possibly due to convection driven by the basal heater (Figure 2). Convection driven by the basal heater could have minimized the thickness of the boundary layer at the icewater interface and kept nutrients mixed in the water-column under the ice. Macro-nutrients were not depleted throughout any of the incubations ( Table 2). During our incubations, a trace amount of Fe was inevitably introduced, but this contamination was considered to be low and negligible ( Table 2). TDFe concentrations ranged from 11 to 43 nM, which were equivalent to 2.2 and 8.7 pM of Fe , respectively. Initial Fe concentrations FIGURE 2 | An ice section (0.5 cm thick) of artificial sea ice from the low-Fe ice tank. An ice block (5 cm × 10 cm × 5.5 cm) was sectioned to 0.5 cm with a band saw at −20 • C. After planning with a microtome blade, the thick section was photographed between cross-polarized filters.

Non-photochemical Quenching (NPQ NSV )
Non-photochemical quenching (NPQ NSV ) was relatively high at the initial stage on day −05 (Figures 5A,B). Freezing enhanced NPQ NSV in the HL treatment (q = 5.14, p = 0.016), whereas no significant change was observed in the LL treatment (q = 2.49, p> 0.05). However, NPQ NSV was also significantly upregulated in the early stage of the frozen period at day 05 in the LL treatment (q = 5.55, p = 0.008). In addition, NPQ NSV was gradually upregulated and reached plateaus in both light treatments (Figures 5A,B). After the ice melt, NPQ NSV decreased to similar levels observed on day −05 in both light treatments, while light exposure to the melted samples led to different responses between the HL and LL treatments. Light exposure to the melted samples from the HL treatment little affected NPQ NSV (q = 3.16, p> 0.05), whereas the NPQ NSV in the LL treatment was upregulated upon light exposure (q = 6.25, p = 0.002).

Photosynthesis-Irradiance (ETR RCII -E) Curve
The initial slopes of ETR RCII -E curves; α, regarded as light utilization efficiency, did not exhibit any change as the F. cylindrus cells were integrated into the freezing ice matrix (Figures 6A,B). During the frozen periods, α values gradually and constantly increased under both HL and LL conditions. The ice melt event decreased α values to the initial levels. The light exposure did not change α (Figures 6A,B) (HL: q = 0.47, p> 0.05; LL: q = 0.77, p> 0.05). Maximum electron transport rates (ETR max ) showed significant decreases in both treatments during the frozen period compared with the initial ETR max values   ( Figures 7A,B) (HL: q<6.28, p< 0.01; LL: q<10.05, p< 0.001). However, the ice algal cells in the HL sustained the initial ETR max values at the beginning of the frozen period at day 00 (q = 2.36, p> 0.05). Once the ice had melted, ETR max values recovered to their pre-freezing levels (Figures 7A,B); however, light exposure had minimal effects on ETR max (Figures 7A,B Figures 8A,B). Melting increased the E k values to the initial levels before freezing on day −05, whereas the light exposure did not change the E k values in either treatment (Figures 8A,B) (HL: q = 2.50, p> 0.05; LL: q = 1.05, p > 0.05).

Pigment Composition
Initial Tchl a concentrations were comparable regardless of light availability (Welch's test, t = 0.19, p> 0.05, Figures 9A,B); however, the LL treatment showed higher Tchl a biomass during the frozen period (q = 6.24, p< 0.001). The contribution of chlide a to Tchl a gradually increased during the course of the incubations (Figures 9A,B). When the cells were exposed to high light after the ice had melted, there were substantial increases in the chlide a to Tchl a ratio in both light treatments (HL: q = 7.25, p = 0.002; LL: q = 5.20, p = 0.033, Figures 9A,B).
The DD-DT pool size in the HL treatment was significantly higher than that in the LL treatment throughout the incubations (q = 3.02, p = 0.041; Figures 10A,B). The DD-DT pool size at the end of the frozen period at day 20 was significantly higher than the initial size in the HL treatment (q = 5.05, p = 0.041, Figure 10A), whereas no significant change in the pool size was observed in the LL treatment (p> 0.05, Figure 10B). Values of DES increased slightly later than the algal cells that had been frozen into the ice and remained stable in both light treatments (Figures 10A,B). The DES levels decreased when algal cells were melted out in both light treatments (HL: q = 55.89, p< 0.001; LL: q = 14.83, p< 0.001). Light exposure did not change the DES levels in the HL treatment (q = 4.56, p> 0.05, Figure 10A); however, the DES in the LL treatment significantly decreased after the light exposure (q = 5.91, p = 0.013, Figure 10B).
The HL treatment showed a gradual and sharp increase in PPC/PSC during the frozen period (Figure 11A), whereas no conspicuous change was observed in the LL treatment throughout the incubation (p> 0.05, Figure 11B). When the ice melted, there was a significant decrease in the PPC/PSC in the HL treatment (q = 5.73, p = 0.016), down to the level prior to the start of the frozen period ( Figure 11A). The light exposure little affected the PPC/PSC ratios in both light treatments (HL: q = 0.21, p< 0.05, LL: p< 0.05, Figures 11A,B).
The PPC/PSC level of the HL treatment was significantly higher than that of the LL treatment throughout the incubations (q = 5.39, p < 0.001).

Gene Expression of Photosynthesis-Related Genes rbcL
Gene expressions of the rbcL gene were highly upregulated when the algal cells were frozen into the ice in both light treatments (HL: q = 9.23, p< 0.001; LL: q = 5.14, p = 0.036, Figures 12A,B). During the frozen periods, however, gene expression levels dropped sharply (HL: q = 10.39, p< 0.001; LL: q = 7.10, p = 0.003), to a lower level than that of the values from the seawater prior to the initiation of freezing (Figures 12A,B). The low level was sustained throughout the frozen period. Ice melt also did not affect the transcriptional activity of the rbcL gene. In addition, the rbcL gene expression did not show any conspicuous changes in either treatment after light exposure (HL: q = 0.44, p> 0.05; LL: q = 0.24, p> 0.05, Figures 12A,B). Interestingly, the HL treatment showed a significantly higher rbcL gene expression than the LL treatment throughout the incubations (q = 3.93, p = 0.009).
psbA Unlike the rbcL gene, gene expression of the psbA gene behaved differently with differences in light availability (Figures 13A,B). Under HL conditions, transcriptional levels of the psbA gene dynamically changed; the highest value was observed while algal cells were frozen within the ice matrix ( Figure 13A). This peak value was significantly higher than the low expression levels at the end of the frozen period (q = 7.17, p = 0.002, Figure 13A), which was similar to the rbcL variation in the HL treatment ( Figure 12A). The LL treatment, on the other hand, showed a constant psbA gene expression with no significant change (p> 0.05) ( Figure 13B). Both ice melt and light exposure had minimal effects on the transcriptional activity of the psbA gene in both light treatments (HL: q> 1.15, p> 0.05; LL: p> 0.05, Figures 13A,B).

Ice Physics and Chemistry
Our experiments resulted in the first successful incubation of the diatom Fragilariopsis cylindrus in artificial sea ice at a low Fe concentration using a purpose-designed low Fe ice tank (Figure 1). Our prolonged ice tank runs demonstrated the photophysical responses and competency of the ice algae, F. cylindrus, to low Fe availability in the Southern Ocean. The thick section of artificial ice contained elongated brine channels with some brine pockets in the ice matrix, which consisted of frazil ice granules at the top and columnar ice below the upper 1 cm (Figure 2). This structure suggests that the ice tank realistically reproduced a natural sea-ice environment (e.g., Petrich and Eicken, 2017). As discussed in our previous work using the same ice tank (Yoshida et al., 2020), the sharp decrease in temperature (top: −22.5 • C; bottom: −2.2 • C in 5.5 cm) was consistent with the winter-time pack ice environment (Petrich and Eicken, 2017). Macro-nutrients were sufficient to maintain growth throughout the experiments and to ensure that Fe was the limiting nutrient. The consistently low Fe concentration confirmed that the ice tank runs had been conducted at low Fe levels throughout the experiments ( Table 2). Although low Fe availability suppressed PSII photochemical efficiency (discussed below in 4.2), the observed growth rates (HL: 0.022 ± 0.006; LL: 0.019 ± 0.009 d −1 ) were identical to those in Fe-replete ice tanks runs (0.020 d −1 ; Yoshida et al., 2020). This suggests acclimation to the low Fe condition. Pankowski and McMinn (2008) reported that moderate Fe limitation occurred at ∼5 pM Fe for F. cylindrus, which is a similar level to that of this study (HL: 4.6-6.5 pM Fe ; LL: 4.1-8.7 pM Fe ; Table 2). It is thus suggested that our ice tank incubation experiments were appropriate to assess how freezing and melting stress under chronic Fe starvation affects the photosynthetic physiology of ice algal F. cylindrus.

Freezing Event
Freezing stress lowered the photochemical efficiency and stemmed the electron transport. At the beginning of the incubation, F v /F m values were reasonably low (∼0.3) (Figures 3A,B), which was ca. 30 % lower than the previous Fereplete ice tank incubation experiments (F v /F m = ∼0.45; Yoshida et al., 2020; Figures 3C,D). Kolber et al. (1994) and Suzuki et al. (2002) demonstrated that Fe availability affects photochemical reactions at RCII, since PSII has 2 Fe atoms and a high turnover rate (Greene et al., 1992;Geider and LaRoche, 1994;Govindjee et al., 2010). These observations demonstrate that we successfully incubated ice algae under low-Fe conditions for the first time.
The freezing event gradually decreased both F v /F m and σ PSII values until day 05 (Figures 3A,B, 4A,B). In principle, σ PSII is a function of the optical absorption cross section of PSII (σ opt PSII ) and maximal quantum yield of PSII photochemistry ( max PSII , or F v /F m ) (e.g., Huot and Babin, 2011): The concomitant decrease thus suggests that the reduced photochemical efficiency of PSII (i.e., F v /F m ) lowered the σ PSII when F. cylindrus cells were frozen into the ice. Also, NPQ NSV gradually increased, while F v /F m conversely decreased with the freezing stress (Figures 3A,B, 5A,B), suggesting that the photosynthetic apparatus of F. cylindrus was well-protected from freezing stress even under the Fe-starved condition. Although σ PSII decreased (Figures 4A,B), the light utilization index, α, did not respond to the freezing event (Figures 6A,B). In general, α is a product of σ PSII and the concentration of functional RCII (n PSII ): The stable α values suggest that F. cylindrus could increase n PSII in response to low-light acclimation (n PSII strategy, hereafter) with freezing stress. However, these were unexpected physiological responses, because it is Fe-costly to synthesize PSII (e.g., Geider and LaRoche, 1994;Twining and Baines, 2013). Interestingly, ETR max maintained higher values (Figures 7A,B),   which were similar to those of the Fe-replete ice tank experiments (Yoshida et al., 2020; Figures 7C,D). Schuback et al. (2015) also reported relatively high ETR max values in Fe-limited waters. They hypothesized the existence and upregulation of alternative pathways for electrons to dissipate the excess energy, rather than xanthophyll cycling. These included pseudo-cyclic and cyclic electron flow (Prášil et al., 1996;Feikema et al., 2006;Cardol et al., 2011), which would have reduced the high transmembrane pH under low Fe conditions. The resultant E k value eventually showed the dark-acclimated state (Figures 8A,B). The contribution of chlide a to Tchl a was stable when cells had frozen into the ice, indicating a smaller effect of the melting process on pigment breakdown and the production of chlide a (Figures 9A,B). The size of DD-DT pool was also stable, and the xanthophyll cycle had not been activated at day 00 (Figures 10A,B). This suggests that other non-photochemical quenching pathways were responsible for the NPQ. However, the xanthophyll cycle activated at day 05 with the upregulation of NPQ NSV , suggesting that diatoms had gradually acclimated to the ice environment by employing their photoprotective capability (Figures 5A,B, 10A,B). Gene expression of the rbcL gene were upregulated in both light treatments (Figures 12A,B), emphasizing the cold acclimation strategy. The carboxylation rate of RuBisCO critically slows down at low temperatures (Young et al., 2015). The rbcL gene upregulation could enhance the synthesis rate of RuBisCO to complement the low enzymatic activity by increasing the cellular enzyme concentration (Devos et al., 1998;Lyon and Mock, 2014;Young et al., 2015;Yoshida et al., 2018). A contrasting response was found for the psbA gene, i.e., upregulation in the HL treatment while constant in the LL treatment (Figures 13A,B). The upregulation of the psbA in the HL treatment could be a strategy to overcome overexcitation associated with HL and possibly the freezing stress causing electron clogging. This upregulation also suggests that the turnover rate of D1 proteins was accelerated due to a higher rate of PSII photodamage.

Frozen Period
Sea ice provides a stable platform for photosynthesis of ice algae, which might optimize their photophysiology in spite of the low Fe availability. Regardless of light intensity, F v /F m values during the frozen period were comparable to the previous study under Fe-replete conditions (Yoshida et al., 2020; Figures 3C,D). This suggests that both Fe and light availability had little effect on PSII reaction center activity and growth of F. cylindrus in the sea-ice environment. Pankowski and McMinn (2008) reported similar growth rates of F. cylindrus both in Fe-replete and Fe-starved culturing media. Although F v /F m values were stable in our experiments, σ PSII showed gradual increases during the course of the incubation (Figures 3A,B, 4A,B). Interestingly, the other index of light absorption efficiency, α, increased synchronously with σ PSII (Figures 4A,B, 6A,B), suggesting that the increase   in α was due to increasing photosynthetic antennae size (i.e., increase in σ PSII ; "σ PSII strategy, " hereafter), but not increasing the number of photosynthetic units (n PSII strategy). By conducting culture incubation experiments using representative Southern Ocean species, Strzepek et al. (2012Strzepek et al. ( , 2019, demonstrated that Southern Ocean phytoplankton in Fe-limited waters employed the σ PSII strategy (σ PSII > 10 nm 2 PSII −1 ; Strzepek et al., 2019, similar in value to those of this study; Figures 4A,B). They proposed that the σ PSII strategy is used to overcome Fe-light co-limitation in the Southern Ocean; this study is the first to demonstrate that ice algae in the Antarctic also appear to employ the same strategy. On the contrary, the Fe-replete ice tank experiments did not show any increase in both σ PSII and α (Figures 4C,D, 6C,D), indicating that the σ PSII strategy could be a unique photophysiological response of polar diatoms to low-Fe and low-light conditions. Also, F. cylindrus had changed their low-light acclimation strategy; they increased n PSII while being frozen in (n PSII strategy; discussed above), whereas they enhanced σ PSII over the duration of the frozen periods (σ PSII strategy). This physiological switching in F. cylindrus could be significant for acclimation to ice environments. Interestingly, NPQ NSV seemed to synchronize with DES during the frozen periods, whereas NPQ NSV and DES did not covary during the freezing event at day 00 (Figures 5A,B, 10A,B), which was another evidence of the physiological switching. This response was feasible because DD and DT do not possess Fe, although there are some effects of Fe availability on the genetic translations of the pigments, such as operon regions (e.g., Laudenbach et al., 1988;Geider and LaRoche, 1994;Lommer et al., 2012;Georg et al., 2017). The DD-DT xanthophyll cycle was thus largely responsible for photoprotection in diatoms under the low-Fe condition, as well as high-Fe environments (Yan et al., 2019;Yoshida et al., 2020). The pool size of DD plus DT in the HL treatment was larger than that of the LL counterpart (Figures 10A,B), again suggesting more photoprotective capability under the HL condition. The significantly higher PPC/PSC ratio in the HL treatment was also supportive evidence of the large range of photoprotective capability (Figures 11A,B). NPQ NSV capability (maximum values of NPQ NSV ) was, however, comparable between the HL and LL treatments (Figures 5A,B). This was possibly due to the effectiveness of the xanthophyll cycle rather than other factors, e.g., the cyclic electron flow via ferredoxin or flavodoxin (Prášil et al., 1996;Feikema et al., 2006;Cardol et al., 2011). The DD-DT xanthophyll cycle prevents further build-up of the transmembrane pH because further electron sinks might not have been active. Indeed, Yoshida et al. (2020) reported that DES was not exclusively responsible for NPQ NSV ; other electron pathways play supportive roles for photoprotection.
It is also of interest that both Schuback et al. (2015) and this study reported higher ETR max values even under Fe starvation compared with Fe-replete conditions (Yoshida et al., 2020). The higher ETR max values might also be due to charge recombination at PSII (i.e., reflux of electrons back to RCII after the excitation event; electron recoupling to RCII; Vass, 2011), leading to overestimation of electron transportation at PSII (Schuback et al., 2016). The freezing stress could lead to high excitation pressure to the electron transport chain. Also, if the xanthophyll cycle was exclusively responsible for photoprotection, as discussed above, electron recoupling could be one of the possible pathways to address the enigmatic high ETR max values under Fe starvation. Transcriptional activity of the rbcL gene was sharply downregulated in the ice and maintained at a low level (Figures 12A,B), suggesting that algal cells acclimated to their ice environment and optimized the balance between photochemical reactions and energy allocation to the dark reaction processes, such as carbon fixation by RuBisCO. It was of interest that rbcL gene expression in the HL treatment was significantly higher than the LL treatment (Figures 12A,B), suggesting that more RuBiSCO might be required as an electron sink with the higher excitation pressure in the HL treatment. The constant gene expression of the psbA gene in the LL treatment seems reasonable, as the constant freezing stress in the ice could have damaged the fragile D1 protein in PSII by over-reduction ( Figure 13A). Downregulation in the transcriptional activity of the psbA gene in the HL treatment, however, was enigmatic because light stress was more evident at the higher light intensity. The relationship between the psbA gene and the transcription of the D1 protein is not directly inferred by the considerable post-transcriptional regulation (e.g., mRNA splicing and RNA editing; Kettunen et al., 1997). The significant upregulation of the psbA gene when frozen in (discussed above), associated with the sudden freezing stress, might thus provide excess mRNA to maintain synthesis of the D1 proteins.

Light Exposure After the Ice Melt Event
Even under low Fe availability, ice algae coped with light stress when they were released from sea ice. The high ETR max, values, which were higher than or similar to the Fe-replete values, also support the photosynthetic plasticity of F. cylindrus to the chronic low-Fe stress even after ice melt and light exposure (Figure 7; Yoshida et al., 2020). The σ PSII strategy, which enlarges photosynthetic antenna size, was quite effective for photosynthetic acclimation to overcome low-Fe and low-light availability in sea ice (Strzepek et al., 2012(Strzepek et al., , 2019. This low-Fe acclimation, at low temperatures, resulted in the similar growth rate to that of the Fe-replete ice tank runs. Light exposure significantly decreased F v /F m values due to the excess excitation pressure in the reaction centers of PSII, i.e., photoinhibition (Figures 3A,B; Melis, 1999). This photoinhibition was also evident even in the Fe-replete ice tank incubation experiments (Yoshida et al., 2020 ; Figures 3C,D). NPQ NSV also responded to light exposure to dissipate excess energy (Figures 5A,B); however, photoprotective capability was not evident compared with Fe-replete condition (Figures 5C,D). The NPQ NSV values in this low-Fe study were higher than those of the Fereplete ice tank runs after the ice melt event (Figure 5). This upregulation of NPQ could be related to the σ PSII strategy with the quite large photosynthetic antenna size. A larger antenna has a longer residence time of excitons within the PSII, which results in more energetic loss as NPQ (Raven, 1990;Wientjes et al., 2013;Strzepek et al., 2019). After the light exposure, the NPQ NSV enhancement was less efficient (HL: 20.9% increase; LL: 57.8% increase) than in those of the Fereplete conditions (HL: 98.5%; LL: 89.5% increase) reported in our previous ice-tank work (Yoshida et al., 2020; Figure 5). This indicates that the low Fe conditions suppressed NPQ ability. Moreover, the xanthophyll activity, shown as DES, was downregulated in the high light stress in the LL treatment, but not in the HL treatment (Figures 10A,B). This different response to light exposure reflects their different light history during the frozen period; lower photoprotective ability under lower light availability. In addition, the small change and lower DES after light provide a reasonable explanation of the lower NPQ NSV enhancement. The xanthophyll pool and PPC/PSC ratio were, however, relatively stable, because pigment synthesis is slower than photoprotective processes (i.e., xanthophyll cycles and D1 protein repair) (Kuczynska et al., 2015).
The significant accumulation of chlide a after light exposure reflected the high level of absorbed energy that subsequently broke down intact chl a, possibly due to the presence of ROS (Figures 9A,B). The stable contribution of chlide a again implied that the production of chlide a was not due to ice melt events. Large concentrations of chlide a, however, have been observed at the surface of the Southern Ocean during ice retreat in spring (Bidigare et al., 1986;Wright et al., 2010). Exposure to high light after ice algae are released from ice is more likely the reason for the high chlide a contributions in situ. Interestingly, repair of damaged PSII, as indicated from gene expression of the psbA gene, was not activated after light exposure (Figures 13A,B). In addition, the slower repair of damaged PSII at low temperature (Kropuenske et al., 2009;van de Poll et al., 2011;Lacour et al., 2018) could exacerbate the photoinhibition. The diatom F. cylindrus displayed high photosynthetic flexibility and successfully acclimated to the low-Fe availability even after chronic Fe starvation in a simulated sea-ice environment. However, the chronic Fe starvation led to reduced photoprotective capability, which could affect the survival of ice algae when the cells are released to high-light pelagic waters. This might lead to changes in ice-edge bloom species and the extent of ice-edge blooms unless an ample amount of Fe is supplied to the cells during ice melt.

CONCLUSION
For the first time, we successfully incubated the polar diatom F. cylindrus under low Fe conditions in artificial sea ice using the ice tank. The artificial sea ice was similar to natural winter seaice characteristics including vertically elongated ice crystals and brine channels as well as brine pockets underneath a thin layer of fine granular frazil ice. Ice algae are capable of optimizing their photosynthesis for sea-ice environments during the freezing, ice melt and light exposure events even under chronic Fe starvation in sea ice. However, chronic Fe starvation led to reduced photoprotective capabilities. This may have detrimental consequences for ice algal production if the recent reduction in sea ice extent continues, although regional variabilities in the sea ice extent are evident around Antarctica (Stammerjohn and Maksym, 2017;Parkinson, 2019). Our study has revealed that Fe availability modifies the photophysiological plasticity of ice algae. This also indicates that increased Fe availability in surface water would enhance phytoplankton growth and overwhelm ice algae after chronic Fe starvation in sea ice. A model simulation suggests a continued increase in Fe availability in the Southern Ocean (Boyd et al., 2015). Furthermore, glacial ice melt enhances the input of Fe to the surface water (Statham et al., 2008;Herraiz-Borreguero et al., 2016). Consequently, an Fe-related change in the production of ice algae and switching to phytoplankton dominance would change the diversity and trophic dynamics in the Southern Ocean (Constable et al., 2014). Our study emphasizes the importance of quantifying and modeling Fe availability in pack ice to predict changes in primary production and ecosystem functioning in low ice concentration zones around the Antarctic continent, including the Marginal Ice Zone (Boyd et al., 2016;Lannuzel et al., 2016b).

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available at: http://hdl.handle.net/2115/80350.

AUTHOR CONTRIBUTIONS
KY primarily wrote the manuscript, collected and interpreted the data. AM and KS supervised KY and conceptualized the study. KY, AS, and AM designed the ice tank experiments. KY, AS, KK, and AM developed low-Fe ice tank techniques. KY, MC, and PH conducted the ice structure measurement. MC and PH interpreted results of the ice structure measurement and wrote the manuscript. AS, MC, PH, AM, and KS critically revised the manuscript. All authors contributed to the article and approved the submitted version.

FUNDING
This study was supported by the Australian Antarctic Programme (AAS4319). The Fragilariopsis cylindrus collection was supported under the Australian Research Council's Special Research Initiative for Antarctic Gateway Partnership (Project ID SR140300001). The structural analysis was supported by AAS4506. This study was also partly supported by the JSPS Grants-in-Aid for Scientific Research (JP18H03352 and JP17H00775).