Ice and Dust Samples.

Antarctic dust of Last Glacial Maximum age (LGM) was obtained from the European Project for Ice Coring In Antarctica (EPICA) Dome C ice core from Dome C on the East Antarctica Plateau (EDC; 75°06’S 123°21’E, 3,233 m asl). This ice core provides a climate and dust archive stretching back 800,000 years [1–3,21,25]. For this study, we chose ice samples to cover the highest-dust portion of the LGM (21–26 kyr BP; [21]), with dust concentrations of 0.6–1 mg kg^-1 [21], providing sufficient Fe to observe any possible response in culture. Despite these relatively-large dust concentrations compared to other intervals [21], it was still necessary to combine dust from several kg of ice to provide enough dust for incubations. Ice was supplied by EPICA as 8 bags (55 cm length) of ‘A cuts’ (approx. ¼ of the core), with ages of 22.1–22.3 and 24.7–25.8 kyr BP [40]. We calculated that the total mass of dust provided by the ice was 2.84 mg, based on the high-resolution dust record of Lambert et al. [21]. Sub-samples of ice from these bags and ice of similar age were sampled and analysed for total and soluble Fe concentrations in our recent study on Fe solubility [24], allowing us to calculate the soluble Fe that would be expected to be released in seawater. There is large variability in both total and soluble Fe at high-resolution in the ice core, which means it is difficult to calculate a representative average soluble Fe per 55 cm bag accurately (e.g. the Fe which dissolves at pH 5 ranges from 1.5–28 ng g^-1 for these EDC bags (mean 9±9 ng g^-1; 1SD)) [24]. However, based on measured Fe values and the weight of ice used from each bag, we estimate that the total mass of dust available from these EDC bags would provide ~0.2 mg of total Fe and ~0.02 mg of soluble Fe available to be used for culture experiments [24].

Ice Sample Processing and Laboratory Procedures.

EDC ice samples were cut with a band saw and then decontaminated within a class 100 flow hood in a cold room (-20°C) at the British Antarctic Survey according to published methods [24]. All plastic used was rigorously acid cleaned in the Department of Earth Sciences, University of Cambridge, and strict handling procedures were used to minimize contamination [24]. Briefly, each piece of ice was handled with acid-cleaned polyethylene gloves and scraped sequentially with ceramic blades to remove the outer edges of ice, which are contaminated with trace metals by collection, drilling and handling [24,41]. Ice samples were then taken for sublimation [24], which allows extraction of the dust without melting of the ice. This is important because release of Fe from dust is strongly pH dependent, and Fe solubility and speciation can be affected by exposure of dust to water [24,42]. Thus dry-extraction, together with the fact that the major deposition mode of dust deposited at Dome C during the LGM was dry [25], means that the dust used in this study should be representative of natural aerosol dust that would have been deposited to the nearby Southern Ocean at the LGM.

Decontaminated ice samples were sublimated using the larger ice mass configuration described by Conway et al. [24] (see S1 Fig). Sublimation of ice was achieved by lowering the pressure below the triple-point of water (0.01°C, 6.1173 mbar). Briefly, decontaminated ice samples were placed within open 1 L wide-mouthed Nalgene low-density polyethylene (LDPE) bottles. The 1 L bottles were placed within closed 4 L high-density polyethylene bottles (with holes drilled in the lids). The 4 L bottles were placed within a Schott glass desiccator and the desiccator closed. All transfer of ice to the desiccator was carried out under class 100 filtered air at -20°C. The sealed desiccator was carried out to the warm lab and attached to a vacuum pump and freezer trap [24] (see S1 Fig). The pump was turned on, the pressure was reduced to 0.8–1.6 mbar and the ice was allowed to sublime, following procedures described previously. Once all the ice had gone, and the pressure had returned to the background of the chamber (0.27 mbar), the desiccator was closed, the pressure released and the desiccator returned to the flow bench in the cold room for opening and addition of more ice. Sublimation proceeded at a rate of ~ 10 g hr^-1, meaning that it required several weeks to extract enough dust for incubations, with fresh ice added to the same bottle every ~2 days. Testing established that Fe contamination by sublimation was very low (<40 pg per ~500 g 2 day session [42]), which is insignificant compared to the micrograms of seawater-soluble Fe that are present from dust within 500 g of ice [24].

Laboratory Procedures.

All incubation processing and sampling of incubation materials was carried out under class 100 laminar airflow in the Department of Chemistry at the University of Otago, using strict clean procedures to prevent trace metal or biological contamination. All plastic equipment and pipette tips in contact with incubations was rigorously acid-cleaned using standard methods [24,43,44], which included rinsing thoroughly with 1% v/v quartz-distilled hydrochloric acid, ultrapure water (UPW; 18.2 MΩ) and clean sub-Antarctic seawater (see below) prior to use.

Seawater.

Natural sub-Antarctic seawater for incubation experiments was collected onboard the R/V Polaris II on the 14th July 2009 from a station at the distal end of the Munida Transect, close to New Zealand (45.8°S 171.0°E; [45]; for which permission was not required). Water was collected using trace-metal clean towfish and sampling lines [46], filtered with 0.45 μm Pall Aquaprep 600 capsules into acid-cleaned polycarbonate carboys (in the dark) onboard ship and frozen at -20°C to preserve organic ligands. Prior to use in incubations, the seawater was thawed and filtered with Sartobran 150 filter cartridges (0.2 μm with 0.45 μm prefilter), before being used in incubation experiments in July-August 2009. The dissolved Fe concentration in the seawater (low-Fe seawater) was 0.223 ± 0.023 nmol L^-1, measured by CL-FIA (see later) and the organic ligand concentration was approx. 0.5 nmol L^-1, measured by competitive ligand stripping voltammetry [47]. The pH of the seawater was 8.05, measured using photometric methods [48].

Diatom Species.

For this study, we chose monospecific cultures of two medium-large diatom species Eucampia antarctica (50–100 μm) and Proboscia inermis (150–200 μm) which are both important members of the modern Antarctic diatom assemblage [11,49–52]. E. antarctica has a well-documented bloom response to Fe and was also a widely distributed species at the LGM [11,50,53,54]. In fact, E. antarctica abundance in sediment cores has been found to be several fold greater during glacial intervals, with E. antarctica comprising up to 40% of the diatom assemblage [11]. The record of P. inermis is less well documented, as is the species’ response to Fe addition; however, both a growth and lack of growth response to Fe have been described previously [39,55]. We chose these species in order to test the effect of Fe from atmospheric dust on diatoms which are known to have different Fe requirements in order to gain a more realistic picture of how the whole diatom assemblage at the LGM would respond to dust addition. The cultures used in this study were isolated from the Southern Ocean (57°51’S, 139°50’E) in 2001 [39] and had been grown at 3°C in Aquil media at the University of Otago. Each culture was grown into Fe limitation for a number of generations in sterile 0.22 μm filtered low-Fe seawater, with only major nutrients (nitrate, phosphate and silicate) added to the culture. Fe limitation in cultures prior to incubation experiments was confirmed by measurement of the quantum use photosynthetic efficiency of photosystem II (Fv/Fm) at values less than 0.3 [56,57].

Incubation Setup.

Diatom culture incubations were designed to match natural conditions as closely as possible. We used natural sub-Antarctic seawater without EDTA to complex dissolved metals and used realistic constant light (30 μmol photons m^-2 s^-1 from cool fluorescent tubes) and temperature (3°C) conditions. In addition to incubations with added EDC dust, control incubations (with only macronutrient addition to represent Fe-limited conditions) and incubations with addition of dissolved Fe standard (FeCl₃; to assess the bioavailability of dissolved Fe) were carried out under identical conditions.

Incubation media were prepared in six acid-cleaned 8 L polycarbonate canisters, three for each diatom species (control, +dust and +FeCl₃ conditions). Each bottle was filled with 6 L of low-Fe seawater, and 650 μL of each of silicate, nitrate and phosphate solutions added to make final concentrations of ~40 μmol L^-1 silicate and nitrate and ~3 μmol L^-1 phosphate. The nutrient solutions were either cleaned prior to addition with Chelex ion-exchange resin to remove dissolved Fe (nitrate, phosphate) or were low-metal standard solutions (silicate). Following mixing, the canisters were amended with either 0.5 L of a seawater-dust suspension (+dust), 0.5 L of low-Fe seawater and an aliquot of FeCl₃ solution (+FeCl₃) or simply 0.5 L of low-Fe seawater (control). The seawater-dust suspension was prepared by addition of 1 L of low-Fe seawater to the 2.84 mg of extracted EDC dust (present in three 1 L LDPE bottles), with extensive shaking/sonication in the bottles. The seawater-suspension was then combined and subdivided so that 500 mL of the mixture was added to each of the two +dust canisters, corresponding to a calculated average dust addition of 0.22 mg L^-1 dust. The measured dissolved Fe concentration by CL-FIA (see later) immediately after dust addition for the two +dust canisters was 14 and 9 nmol L^-1. Addition of FeCl₃ to the +FeCl₃ canisters was designed to match these values, by addition of aliquots of a 179 μmol L^-1 FeCl₃ solution (prepared by 100x dilution of a 17.9 mmol L^-1 FeCl₃ 2% HNO₃ solution with UPW) to each canister.

A 3.25 mL inoculum of E. antarctica or P. inermis culture was pipetted into each canister, and the contents gently mixed. P. inermis was added to the 9 nmol L^-1 dissolved Fe +dust or +FeCl₃ canisters, and E. antarctica to the 14 nmol L^-1. Following mixing, the canisters were subdivided into five 1 L transparent polycarbonate bottles, providing 5 replicate bottles for each incubation, and the remaining volume in the canister taken for measurement of a number of initial parameters (see below). The 1 L incubation bottles were sealed, double bagged within transparent ziplock bags and placed in thermostat controlled circulating water baths. Bottles were gently mixed and re-arranged within the incubator once per day to ensure constant light. Incubations were designed to last for 18 days but were stopped at 12 (E. antarctica) and 17 (P. inermis) days to prevent nutrient limitation, based on visual observation of extensive diatom biomass.

Sub-sampling for Growth Parameters.

Cell number, chlorophyll a, dissolved Fe concentration (dFe), macronutrient concentrations, and the quantum use efficiency of photosystem II (Fv/Fm) were measured throughout the experiment at regular intervals. For each treatment, these parameters were measured in subsamples from 3 of the 5 incubation bottles. Subsamples for cell number and Fv/Fm were taken approx. every 2 days, and for the other parameters approx. every 4 days. The initial canisters were sampled for each of these parameters and particulate organic carbon (POC) at the beginning of the experiments, and all 5 incubation bottles were subsampled for these parameters, together with cell surface area/volume and POC concentration at the completion of each incubation.

Growth Parameters and Analysis.

Photosynthetic quantum-use efficiency of photosystem II (Fv/Fm) provides a sensitive indicator as to whether a culture is Fe-stressed, and responds rapidly to relief of Fe-limitation [58]. This parameter has been used in a range of recent studies to demonstrate the effect of ash-fertilisation on phytoplankton [59,60,32] and Fe-limitation in the open ocean [61], with values ≤0.3 considered to be indicative of Fe-limitation [59,61]. Fv/Fm in this study was measured using a PhytoPAM (Walz, Germany) following dark adaptation of samples at low temperature, after Kolbowski and Schreiber [62]. Unfortunately, Fv/Fm data is only available from later stages of the experiments due to analytical issues with the PhytoPAM. Diatom cell number, length and width were determined using light microscopy in small volumes of solution (5–20 mL) after fixation with acid Lugol’s solution [63]. Cell surface areas and volumes were then calculated following either Hillerbrand et al. [64] for P. inermis, or using the ideal shape calculations described in S1 Text for E. antarctica. In the latter case we chose to use this calculation to ensure inclusion of the large horns of Eucampia sp., which visibly contain cytoplasm material. The cultures were not axenic but were visibly checked by microscope both before and during incubations, and no bacterial growth or any other phytoplankton species were present at any point.

Subsamples for determination of chlorophyll a and macronutrient concentration were filtered using GF/F filtration, and both the filter and filtrate were frozen at -20°C for later analysis for chlorophyll a and macronutrients respectively. Macronutrient concentrations (nitrate, phosphate and silicate) were determined using standard photometric methods [65]. Prior to analysis of chlorophyll a, the frozen filters were placed in 10 mL of 90% acetone in polypropylene vials, vortexed and then re-frozen at -20°C for 18–24 hours. Samples were allowed to melt, centrifuged at -5°C for 3 minutes (4100 rpm), kept in the dark and the supernatant pipetted into a glass vial for chlorophyll a analysis by a Turner fluorometer using standard procedures [66]. Subsamples for POC were filtered through pre-combusted GF/F filters and the filter frozen at -85°C. Filters were then dried at 65°C and stored dry until analysis with standard methods [67]. A blank value for POC was obtained from unused pre-combusted GF/F filters, and this was subtracted from all measurements.

Subsamples for measurement of dissolved Fe were collected with acid-cleaned polypropylene syringes, and filtered through 0.2 μm Millex GP PES Express Sterile 3 mm filtration units, which were rinsed with quartz-distilled HNO₃ and UPW prior to use. Samples were collected into 125 mL LDPE bottles, acidified to pH ~1.6 with quartz-distilled HCl and were stored for several months before analysis. Dissolved Fe concentrations were determined in acidified seawater using chemi-luminescence flow injection analysis (CL-FIA) with luminol (5-amino-2,3-dihydro-1,4-phthalazine-dione) modified by Steve Rusak (pers. comm.) from previous studies [68,69]. Samples were pre-concentrated onto a Toyopearl AF-chelate resin column, analysed with a Waterville Analytical flow injection analyser and calibrated using standard additions, prepared daily. The limit of detection was defined as three times the standard deviation of the mean of a series of replicate analysis on a given sample. The method was checked by measurement of the NASS-5 reference material; we measured Fe as 4.197 ± 0.238 nmol L^-1, within 95% confidence of the certified value (3.707 ± 0.627 nmol L^-1).

Analysis of Variance.

Analysis of variance (ANOVA) and least significant difference (LSD) were calculated using GenStat (Release Version 15.1) on each growth parameter for the three treatments for each species, testing whether the populations were significantly different from each other in each case.

Initial dissolved Fe.

Dissolved Fe concentration (dFe) was measured in each canister before and after the addition of diatom cultures (Fig 1). In the control canisters, the mean [dFe] was 1.73 and 1.92 nmol L^-1 prior to diatom addition, and 2.1 and 4.4 nmol L^-1 for E. antarctica and P. inermis respectively after addition of culture (Fig 1). These initial control dFe concentrations are higher than would be expected in the open HNLC Southern Ocean (e.g. 0.05–0.10 nmol L^-1; see introduction), but occurred despite our best efforts and deployment of trace metal clean procedures throughout. Typically culture studies use Fe-binding ligands such as EDTA or DFB to control free Fe and Fe-limit phytoplankton [37–39]. In this study, however, we chose not to use an Fe-binding ligand because the ligand might have enhanced dissolution of Fe from dust, or changed the speciation of Fe in solution and would have prevented us from measuring dFe. This decision meant that it was much more difficult to maintain cultures at open-ocean concentrations, given the necessary addition of nutrients and cultures. The increase from 0.2 to ~2 nmol L^-1 could be the result of Fe contamination by handling or more-likely from the pre-cleaned nutrient solutions, whilst addition of E. antarctica culture contributed ~0.2 nmol L^-1 and P. inermis contributed ~2 nmol L^-1, despite these cultures being in an Fe-limited state prior to addition, based on Fv/Fm measurements.

However, despite the higher than expected initial dFe in the controls, dFe in the canisters following Fe/dust (and culture) addition was significantly higher, with 15 and 13 nmol L^-1 for the +dust, and 6 and 8 for the +FeCl₃ for E. antarctica and P. inermis respectively. The dFe concentrations added by dust addition were thus 9 and 11 nmol L^-1 for P. inermis and E. antarctica, respectively. This compares reasonably well with the predicted increase in Fe addition based on [24]: the 0.02 mg Fe dust added should result in an average increase in dFe of ~32 nmol L^-1 (range of 6 to 109 nmol L^-1 based on soluble Fe range from [24]). Whilst the +FeCl₃ treatments were designed to match the measured dFe in solution following just dust addition, the measured dFe in the +FeCl₃ canisters were lower than these values. This was likely due to precipitation of the concentrated Fe(III) standard, an unavoidable consequence of diluting a concentrated acidic Fe(III) solution with higher pH ultrapure water and then seawater, where Fe(III) solubility is much lower and so ~50% of dissolved Fe precipitates rapidly (see Suppl. Disc. of [24], and [70]). This meant that the +FeCl₃ incubations began with lower dFe than the +dust treatments (Fig 1). Nevertheless, the initial dFe concentrations in the +FeCl₃ incubations were higher than the control incubations and still provided a useful comparison of Fe released from dust with Fe provided from a dissolved Fe³⁺ addition (Fig 1).

Dissolved Fe and Fv/Fm throughout the incubations.

Despite high initial values in the control incubations, dFe concentrations for each species declined relatively quickly (Fig 1), reaching 0.51±0.30 and 0.91±0.58 nmol L^-1 by day 4 in the control incubations for E. antarctica and P. inermis respectively. For E. antarctica, the control then remained at ~0.5 nmol L^-1 until the end of the incubations, whilst for P. inermis, there was more variability, although dFe concentrations remained below 2 nmol L^-1 (Fig 1). A similar pattern was observed in the +FeCl₃ treatments, with dFe values of 0.3–0.8 nmol L^-1 for E. antarctica and 0.3–1.1 nmol L^-1 P. inermis from day 4 onwards (Fig 1). This reduction in measured dissolved Fe in both control and +FeCl₃ treatments may indicate relatively rapid precipitation of dissolved Fe³⁺ species from solution, to near the organic ligand binding capacity of the seawater (0.5 nmol L^-1), as well as uptake by diatom growth. Such a finding is consistent with experiments with seawater at ambient pH, which showed significant loss of naturally occurring or added dissolved Fe over the first 72–75 hours [71,70].

In contrast to the +FeCl₃ and controls, the dFe concentrations in the +dust treatments remained high for the first 8 days of the experiments for both cultures (Fig 1). dFe concentrations decreased from 14.9 to 9 nmol L^-1in the first 8 days in E. antarctica and from 13.9 to 9.5 nmol L^-1 in P. inermis. Thereafter, dFe concentrations were similarly low in all treatments in both cultures (between 0.5 and 1.68 nmol L^-1; Fig 1). Thus, whilst measured dFe did decline over the course of the experiment in the +dust treatments for both diatom species, dFe remained much higher than the controls over most of the sampling period. This may indicate that there was continuous inorganic or organic-mediated release of Fe from dust particles, or that the dissolved Fe released was in a form that was more easily stabilized or solubilized in seawater, such as Fe(II) species or Fe colloids which persist for longer [71]. The eventual decline to dFe values similar to control may then represent the uptake of Fe into diatom cells, or the conversion of soluble Fe species to those such as Fe hydroxides which may be more readily precipitated/adsorbed from solution [72,73].

Photosynthetic efficiencies (Fv/Fm) of <0.30 are generally taken to indicate stress or Fe-limitation of the diatom species [61,74]. By day 6 for E. antarctica and day 8 for P. inermis, the control incubations for both species showed significantly lower Fv/Fm values compared to the Fe addition treatments (Figs 2a and 3a). This difference was especially pronounced for E. antarctica (Fig 2a), with Fv/Fm values of just ~0.2 from days 6–12, and to a lesser extent for P. inermis (Fig 3a) with values of 0.2–0.4 from day 8–17. This observation, paired with the rapid reduction in dFe in the control incubations (Fig 1) to dFe concentrations where E. antarctica has been shown to be unable to grow (0.5 nmol L^-1; [39]), suggests that the control incubations were Fe-limited by day 6–8, and at least for E. antarctica, perhaps from much earlier in the experiment (Fig 2a). By contrast, Fv/Fm remained high (>0.4) for P. inermis in both the +dust and +FeCl₃ treatments over the course of the whole incubation period (Fig 3a), suggesting that neither culture was experiencing Fe limitation or stress. The same was true for E. antarctica in the +dust treatment over the course of the incubations, while for the +FeCl₃ treatment Fv/Fm values declined to Fe-limited conditions similar to the control after day 10. Thus, the evidence suggests that, despite having initial dFe concentrations higher than open-ocean conditions, both diatom species experienced Fe-limitation or stress in the control incubations from early on the incubation experiments, a state that was alleviated in the +dust and at least initially in the +FeCl₃ treatments.

Other Trace Metals.

Although we did not measure the release of trace metals in the dust, it is possible they may have also affected growth, either acting as a toxin (e.g. Cd, Cu) or alleviating limitation (e.g. Mn, Zn) [75]. Although we can likely rule out toxicity from the other metals because Cd, Co, Cu, Zn and other metals are only found at low concentration (an order of magnitude less than Fe and Mn) from dust in the EDC core [76,77], and trace metal toxicity (e.g. Cu) is suppressed at relatively high levels of Mn³⁺ or Fe³⁺ [78], it is possible that Mn supplied from atmospheric dust could also contribute to a positive diatom growth response as recently hypothesized for volcanic ash [60].

Eucampia Antarctica.

In the control treatment of E. antarctica, cell numbers increased from 27 ± 3 cells mL^-1 to 786 ± 76 mL^-1 in 12 days and chlorophyll a concentrations increased from 0.31 to 3.9 ± 0.42 μg L^-1 (Fig 2, Table 1). Also, nutrient drawdown was relatively low with 10.1, 6.19, and 0.88 μmol of silicate, nitrate, and phosphate taken up respectively (Fig 2, Table 1). By contrast, there was a strong growth response to dust addition in the +dust E. antarctica treatments, with Fv/Fm values typical of Fe replete conditions (0.4–0.5), a large increase in cell number from 27 ± 3 cells mL^-1 to 5620 ± 408 mL^-1 in 12 days and a chlorophyll a increase from 0.31 to 51.3 ± 2.15 μg L^-1, coupled with rapid nutrient uptake of nitrate, phosphate and silicate of 31.1, 35.6, and 1.95 μmol, respectively (Fig 2, Table 1). Cell numbers and nutrient uptake were all significantly higher in the +dust treatments than the controls at the 99.9% confidence level (Table 1). Similarly, much more particulate organic carbon (POC) was produced in the +dust treatments compared to the controls (1.96 ± 0.17 compared to 0.28 ± 0.05 μg mL^-1) (Figs 2 and 4; Table 1). These differences in chlorophyll a and POC concentration, which correspond to an increase of 13x and 6x, respectively in the +dust compared to the controls, are significantly different at the 99.9% confidence level.

A similar growth response as in the +dust treatments was observed for E. antarctica in the +FeCl₃ treatments (Figs 1 and 3; Table 1). Here, we observed similarly high Fv/Fm values, a rapid nutrient drawdown of 27.2, 34.9 and 1.91 μmol silicate, nitrate and phosphate, respectively. Cell numbers increased from 27 ± 3 cells mL^-1 to 3880 ± 290 mL^-1 in 12 days, and POC concentrations from 0.01 to 1.68 ± 0.07 μg ml^-1. This is a significant difference compared to the control for all mentioned parameters at the 99.9% confidence level (Table 1). While the nutrient uptake was very similar between the +dust and the +FeCl₃ treatments, Fv/Fm, cell numbers and chlorophyll a concentrations showed some significant differences. We observed a greater final chlorophyll a concentration in the +FeCl₃ treatments compared to the +dust (87.1 μg L^-1 vs 51.3) although this was coupled with lower final Fv/Fm, cell counts and POC than in the +dust (ANOVA suggests significant at the 99.9% confidence level; Figs 1 and 3; Table 1).

In contrast to the +dust treatments, the strong growth in the +FeCl₃ treatments did not persist for the full 12 days, with no increase in cell number after day 10 (Fig 2). Fv/Fm values declined from day 6 onwards, reaching values which are indicative of Fe-limitation (<0.3) by day 10 (Fig 2), likely reflecting the decline in dFe in the +FeCl₃ treatments to those of the controls after day 4. Despite this decline in both Fv/Fm and dFe, E. antarctica cell counts in the +dust and +FeCl₃ treatments were equivalent until day 10 (Fig 2). After day 10, however, the cell numbers in the +dust treatment continued to increase (with Fv/Fm >0.3) while in the +FeCl₃ treatments there was no increase in cell number (Fig 2). Overall, the lower final cell counts in the +FeCl₃ compared to the +dust treatments (3880 compared to 5620) corresponds to 1 less doubling of cells in the +FeCl₃ treatment and is significantly different at the 99.9% confidence level (Table 1). Since Fe is directly required for production of the photosystems, chlorophyll a production is a good indicator that sufficient ‘bioavailable’ Fe is present. In this context it is interesting that significantly more chlorophyll a production was observed in the +FeCl₃ treatment which had not only lower final cell numbers, but also lower dFe concentrations throughout the incubations (Fig 1a). This observation suggests that not all the Fe released from the dust in the operationally defined dFe parameter was bioavailable to E. antarctica, consistent with previous work on crystalline desert dust that documented that only a fraction of Fe released from dust may be bioavailable [79].

The low Fv/Fm values after day 10 in the +FeCl₃ treatments, despite the fact that major nutrients were still present (Fig 2), suggests that diatoms in the +FeCl₃ treatments began to experience Fe limitation by day 10, resulting in no further growth. It is, however, interesting to note that the decline in Fv/Fm lagged the onset of low dFe by several days, perhaps suggestive of ‘luxury uptake’ of Fe by diatoms during periods of higher Fe concentrations [80]. These high internal Fe concentrations can be passed on as cells divided and sustain growth for several days beyond the onset of Fe limitation. It is also interesting to note that the Fe-limitation inferred for the +FeCl₃ and control treatments in the later stages of the incubations, based on Fv/Fm, corresponds to dFe concentrations of ~0.5 nmol L^-1. This observation may suggest that the Fe requirements for E. antarctica are relatively high, or that not all the dFe was in a form available for utilization by the diatoms.

Proboscia inermis.

For P. inermis, although low Fv/Fm values (~0.3) indicated Fe limitation in the control incubations by day 8, compared to higher values in the +dust and +FeCl₃ treatments throughout the experiment (>0.4), there was no obvious response to dust or Fe addition in terms of cell number (Fig 3; Table 1); in fact, there were less cells in both the +FeCl₃ and +dust treatments (388 ± 55 and 244 ± 62 cells mL^-1) at the end of the experiment compared to the controls (654 ± 104 cells mL^-1). Until day 14, there was almost no difference in the cell numbers between the control and the + dust treatment while cell numbers were lowest in the +FeCl₃ treatment throughout (Fig 3). The similar growth of P. inermis both with and without dust or Fe is comparable to that observed previously by de Baar and Boyd [55], who documented no obvious growth response to +2 nmol L^-1 dFe treatments for this species, although this is different to experiments by Strzepek et al. who found that P. inermis cells placed under Fe-limitation using the siderophore DFB to complex dissolved Fe (a ratio of 4:400 or 4:200 Fe:DFB) showed growth rates reduced by ~50%, or using seawater with 0.49 nmol L^-1 showing a 67% reduction in growth rates [39,81]. Thus the lack of obvious growth response in our experiment in terms of cell number could indicate that P. inermis has a lower Fe requirement than the dissolved Fe present in any of our treatments, which might not be surprising given the relatively high dFe concentrations maintained in P. inermis cultures compared to E. antarctica. However, this would be in apparent disagreement with Fv/Fm measurements which suggest that the P. inermis control was ‘Fe-limited’, whilst the +dust and +FeCl₃ treatments were not.

This apparent discrepancy could be resolved by investigation of other parameters, such as POC and chlorophyll a production. Indeed, in contrast to cell number results, final chlorophyll a concentration was higher in the +dust P. inermis treatments compared to the other P. inermis treatments. Both the +dust and the +FeCl₃ treatments showed significantly higher chlorophyll a concentrations compared to the control but were not significantly different from each other until day 8. Thereafter, chlorophyll a concentrations were higher in the +dust treatment (Fig 3; Table 1). The mean POC concentration in the +dust treatments was also slightly higher (Table 1; Fig 4), but this is only significant at the 97.2% confidence level. Taken together, these results suggest that P. inermis exhibits a different response to Fe addition than E. antarctica, possibly as a result of luxury Fe uptake, as discussed in more detail below, or possibly because E. antarctica has a higher requirement for Fe than P. inermis, consistent with previous work which showed E. antarctica to be unable to grow at Fe:DFB ratios >4:50 whilst P. inermis could grow at ratios up to 4:200 [39]. However, we note that the growth of P. inermis in that study could also be because P. inermis is better able to acquire Fe from DFB by Fe-reduction [39] than E. antarctica.

Notably, in contrast to E. antarctica, P. inermis diatoms in the +FeCl₃ treatment produced overall less chlorophyll a than those with the +dust, perhaps suggesting that Fe from dust was more ‘bioavailable’ to P. inermis than E. antarctica (or that the dissolved Fe precipitated more rapidly in P. inermis treatments). However, in considering the implications of this result for bioavailability of Fe from dust for P. inermis, we must also consider the overall number of diatoms cells which were significantly fewer in the +FeCl₃ treatment (see below).

Chlorophyll a or particulate organic carbon per cell.

Cellular carbon and chlorophyll a concentrations and nutrient uptake for the final day of the incubations are shown in Table 2, Fig 4 and cell volume is shown in S1 Data. For E. antarctica, the concentrations of chlorophyll a per cell in the +dust treatment were about twice as high as those in the control and more than four times higher than the control in the +FeCl₃ treatment (Fig 4c). For P. inermis, chlorophyll a concentrations per cell in both +dust and +FeCl₃ treatments were statistically equivalent and more than two times higher compared to the control (Fig 4c, Table 2). This shows that both diatom species increased chlorophyll a production dramatically in response to either FeCl₃ or dust addition, something that was not obvious from total chlorophyll a production alone in the case of P. inermis. This increase in chlorophyll a production can also be further observed in photomicrographs showing an increased number of chloroplasts per cell in the +FeCl₃ and +dust treatments for both species compared to the controls (Figs 5 and 6).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Photomicrographs of Eucampia antarctica cells, showing higher number and density of chloroplasts in amended treatments.

Panels are labeled control, +dust or +FeCl₃ to denote the incubation conditions.

https://doi.org/10.1371/journal.pone.0158553.g005

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Photomicrographs of Proboscia inermis cells, showing higher number and density of chloroplasts in amended treatments.

Panels are labeled control, +dust or +FeCl₃ to denote the incubation conditions.

https://doi.org/10.1371/journal.pone.0158553.g006

The POC production by P. inermis was only slightly lower in the control compared to the other two treatments, but the difference was not significant (Fig 4a, Table 1). However, because of the lower cell numbers in the +FeCl₃ and +dust treatments, the POC concentration per cell was significantly higher in the +dust and +FeCl₃ treatments by more than a factor of two (Table 2, Fig 4b), and highest in the +FeCl₃ treatments. This shows that POC fixation increased for P. inermis in response to Fe addition. Conversely, in E. antarctica overall POC concentrations at the end of the experiment were six to seven times higher in response to dust or FeCl₃ addition, respectively compared to the control (Table 1, Fig 4a). The equivalent POC concentrations per cell, however, showed no significant difference between the three treatments. These observations highlight the differing response of the two species to Fe addition, with E. antarctica showing a strong positive growth response, producing more cells and more chlorophyll a but not increasing relative carbon uptake per cell. In contrast, P. inermis has a more subtle response, fixing more carbon and chlorophyll a per cell but perhaps not increasing cell number. However, lower Fe concentration experiments might be required to test this definitively. Furthermore, the composition of the cells of both species does suggest a stronger growth response to FeCl₃ than dust addition, supporting the previous assertion that not all Fe released from dust was bioavailable.

Nutrient uptake ratios.

Both species demonstrated a statistically different change in their nutrient uptake ratios in response to FeCl₃ or dust addition, compared to the controls (Table 2). For E. antarctica, the Si:N ratio dropped from 1.61 in the control to 0.87 for +dust and 0.78 for +FeCl₃ (with these identical at the 99.9% confidence level). Similarly, P. inermis cells changed their Si:N uptake ratio from 1.52 in the control to 0.56 and 0.47 in the +dust and +FeCl₃ treatments (again statistically identical). However, the reason for these nutrient uptake changes differs between the two species. For E. antarctica the silicate/cell decreased by ~2x for the +FeCl₃ or +dust treatments relative to cells in the control incubations, but nitrate/cell showed no significant difference between the three treatments (Table 2). By contrast, for P. inermis, the silicate/cell was the same for all three treatments, but the +FeCl₃ and +dust incubations showed marked increases in nitrate/cell (3-4x) relative to cells in the control incubations (Table 2).

A similar pattern is observed for Si:POC or Si:chlorophyll a, with both species dramatically reducing ratios in response to +Fe or +dust but again for opposing reasons (Table 2); E. antarctica reduced silicate/cell relative to POC/cell or chlorophyll a/cell in response to Fe or dust, while in contrast P. inermis increased POC/cell and chlorophyll a/cell for the same uptake of silicate/cell. This means that although both species increased their carbon fixation relative to silicate in response to both FeCl₃ or dust, E. antarctica produced lighter less-silicate rich cells while P. inermis fixed more carbon and nitrate for the same amount of silicate. So while these changes in macronutrient uptake ratios are consistent with observation in both open ocean or under laboratory conditions, where diatoms grown under Fe replete conditions uptake relatively less Si compared to C and N, resulting in relatively less silicified diatom frustules [82–87], it is interesting that the reason for this varies. However, it is perhaps in line with the response of the two species, with E. antarctica producing more cells in response to Fe or dust, while P. inermis instead changed cell composition. Despite this difference in response, the fact that a change in Si:N ratio was observed for P. inermis in this experiment thus also lends support for the assertion that the control incubations were at least partially Fe-limited.

E. antarctica also significantly changed Si:P (14–16) and N:P (18) ratios in response to both Fe and dust, compared to the controls (10 and 6 for Si:P and N:P respectively; Table 2). Negligible uptake of phosphate in the P. inermis cultures meant that it was not possible to tell how phosphate uptake varied with Fe addition in these experiments.