Iron sources alter the response of Southern Ocean phytoplankton to ocean acidification

Article

Reference

Iron sources alter the response of Southern Ocean phytoplankton to ocean acidification

TRIMBORN, Scarlett, et al.

Abstract

The rise in anthropogenic CO₂and the associated ocean acidification (OA) will change trace metal solubility and speciation, potentially altering Southern Ocean (SO) phytoplankton productivity and species composition. As iron (Fe) sources are important determinants of Fe bioavailability, we assessed the effect of Fe-laden dust versus inorganic Fe (FeCl₃) enrichment under ambient and high pCO₂levels (390 and 900 μatm) in a naturally Fe-limited SO phytoplankton community. Despite similar Fe chemical speciation and net particulate organic carbon (POC) production rates, CO2-dependent species shifts were controlled by Fe sources. Final phytoplankton communities of both control and dust treatments were dominated by the same species, with an OA-dependent shift from the diatom Pseudo nitzschia prolongatoides towards the prymnesiophyte Phaeocystis antarctica. Addition of FeCl

3 resulted in high abundances of Nitzschia lecointei and Chaetoceros neogracilis under ambient and high pCO₂, respectively. These findings reveal that both the characterization of the phytoplankton [...]

TRIMBORN, Scarlett, et al. Iron sources alter the response of Southern Ocean phytoplankton to ocean acidification. Marine Ecology Progress Series, 2017, vol. 578, p. 35-50

DOI : 10.3354/meps12250

Available at:

http://archive-ouverte.unige.ch/unige:95492

Disclaimer: layout of this document may differ from the published version.

1 / 1

1

Research article

I

S

A

T

R

S

O

P

O

A

Running page head: CO₂ and Fe source alter phytoplankton growth

Scarlett Trimborn^1,2*, Tina Brenneis¹, Clara J. M. Hoppe¹, Luis M. Laglera³, Louiza Norman⁴, Juan Santos-Echeandía⁵,Christian Völkner¹, Dieter Wolf-Gladrow¹, Christel S.

Hassler⁶

1Alfred Wegener Institute, Helmholtz Centre for Polar and Marine Research, Bremerhaven, 27570, Germany

2University of Bremen, Leobener Straße NW2, 28359 Bremen, Germany

3FI-TRACE,University of Balearic Islands, Palma, 07122, Spain

4University of Cambridge, Cambridge, CB23EA, United Kingdom

5Spanish Institute of Oceanography (IEO), San Pedro del Pinatar, 30740, Spain

6Department F.-A. Forel, University of Geneva, Geneva, 1211, Switzerland

*Corresponding author: [email protected], phone: +49 471 4831 1038

Keywords: climate change, ocean acidification, phytoplankton, iron, dust, Southern Ocean, community composition, diatoms, Phaeocystis

2 Abstract

1

The rise in anthropogenic CO2 and the associated ocean acidification (OA) will change trace 2

metal solubility and speciation, potentially altering Southern Ocean (SO) phytoplankton 3

productivity and species composition. As iron (Fe) sources are important determinants of Fe 4

bioavailability, we assessed the effect of Fe-laden dust versus inorganic Fe (FeCl3) 5

enrichment under ambient and high pCO₂ levels (390 and 900 µatm) in a naturally Fe-limited 6

SO phytoplankton community. Despite similar Fe chemical speciation and net particulate 7

organic carbon (POC) production rates, CO₂-dependent species shifts were controlled by Fe 8

sources. Final phytoplankton communities of both control and dust treatments were 9

dominated by the same species, with an OA-dependent shift from the diatom Pseudo- 10

nitzschia prolongatoides towards the prymnesiophyte Phaeocystis antarctica. Addition of 11

FeCl3 resulted in high abundances ofNitzschia lecointei and Chaetoceros neogracilis under 12

ambient and high pCO₂, respectively. These findings reveal that both the characterization of 13

the phytoplankton community at the species level and the use of natural Fe sources are 14

essential for a realistic projection of the biological carbon pump in the Fe-limited pelagic SO 15

under OA. As dust deposition represents a more realistic scenario for the Fe-limited pelagic 16

SO under OA, unaffected net POC production and dominance of P. antarctica can potentially 17

weaken the export of carbon and silica in the future.

18

3 Introduction

19

At present, due to anthropogenic emissions atmospheric carbon dioxide (CO2) concentrations 20

are increasing at an unprecedented rate (Hoegh-Guldberg & Bruno 2010) and projected to 21

reach between 720 and 1000 μatm by the end of this century (RCP6.0 scenario, IPCC 2014).

22

The dissolution of CO2 in seawater alters its chemistry by increasing the dissolved CO2

23

concentration and lowering pH (called ‘ocean acidification’, OA). Since the beginning of the 24

industrial revolution, the ocean has absorbed about a third of the CO2 emissions. Among the 25

world oceans, the Southern Ocean (SO) sequesters a disproportionally large share of 26

anthropogenic CO2, accounting for about 40% of the global oceanic uptake of anthropogenic 27

CO2 (Sabine et al. 2004, Landschützer et al. 2015). In this region, the biological sequestration 28

potential is, however, constrained by iron (Fe) input(Martin et al. 1991, Boyd et al. 2007, 29

Smetacek et al. 2012). In fact, in-situ Fe-fertilization of SO surface waters relieved Fe 30

limitation and triggered growth of predominantly diatoms (Boyd et al. 2007, Smetacek &

31

Navqui 2008), accompanied with significant CO2 drawdown and in some cases sinking of 32

organic matter (Blain et al. 2007, Smetacek et al. 2012). These studies, however, lack an 33

assessment of OA impacts. Recent CO₂-Fe-bottle experiments with natural phytoplankton 34

assemblages have demonstrated profound impacts on primary productivity and species shifts 35

within the diatom assemblage with potential implications for carbon export (Tortell et al.

36

2008, Feng et al. 2010, Hoppe et al. 2013).

37

To date, SO Fe enrichment experiments were mostly performed using dissolved 38

inorganic forms (Fe(III): FeCl₃ and Fe(II): FeSO₄) (Boyd et al. 2007, Smetacek et al. 2012, 39

Feng et al. 2010, Hoppe et al. 2013), which are considered highly bioavailable to 40

phytoplankton(Shaked et al. 2005, Morel et al. 2008). It has been shown that most of the Fe 41

in the ocean is bound to organic ligands (Boye et al. 2001, Boyd & Ellwood 2010) with 42

consequences for its bioavailability(Hutchins et al. 1999). In pelagic SO waters, upwelling is 43

the major Fe input, resulting in enrichment of organically bound Fe(Boyd & Ellwood 2010).

44

4

Atmospheric dust deposition is another Fe source (Moore & Braucher 2008, Boyd & Ellwood 45

2010), which is known to release soluble Fe and to form colloidal Fe (Fishwick et al. 2014).

46

Indeed, Fe associated with dust has been found to be poorly bioavailable to the two SO 47

diatoms Actinocyclus sp. and Thalassiosira(Visser et al. 2003). In line with this, the two SO 48

diatoms Eucampia antarctica and Proboscia inermis were found to respond more strongly to 49

FeCl₃ than to dust enrichment whereas Fe from dust was more bioavailable to P. inermis than 50

to E. antarctica (Conway et al. 2016). Due to the observed species-species responses to dust 51

and FeCl₃ in the latter study, there is the need to assess their influence on natural 52

phytoplankton assemblages of the SO. Fe-fertilization via dust deposition has been suggested 53

to play an important role for SO biogeochemistry during glacial times, resulting in elevated 54

primary production and carbon export, thereby decreasing atmospheric CO₂ concentrations 55

(Sigman & Boyle 2000). In the future, some important dust producing areas such as central 56

Australia are predicted to become dryer (Durack et al. 2012), resulting in a change in 57

atmospheric Fe-laden dust inputs to the SO, particularly in the context of OA.

58

OA will impact Fe chemistrywith enhanced Fe(III) solubility in ligand-free seawater 59

even though this effect may be rather small under OA (Millero et al. 2009, de Baar &

60

Gerringa 2009), it could on the other hand change Fe(III)-ligand binding affinity, potentially 61

causing a decline in the bioavailability of Fe(III) to marine phytoplankton as previously 62

reported (Shi et al. 2010, Sugie et al. 2013). A comprehensive understanding of how 63

concurrent changes in pCO2 and dust input will affect primary productivity and carbon export 64

efficiency of SO Fe-limited open waters is still lacking. The aim of the present study was 65

therefore to assess the effect of Fe-laden dust versus inorganic Fe (FeCl3) enrichment under 66

present-day and future pCO2 levels (390 and 900 µatm) in a phytoplankton community of the 67

Atlantic sector of the SO.

68

5 Material and methods

69

Experimental set-up 70

As Fe sources are important determinants of Fe bioavailability, the influence of pCO₂ and 71

different Fe sources was investigated in a naturally Fe-limited open ocean phytoplankton 72

community (initially dFe = 0.23 ± 0.06 nmol Fe L^-1). The phytoplankton population was 73

sampled south of the Polar Front in the Atlantic sector of the Southern Ocean (53°0.8’S, 74

10°1.5’E) on 21 January 2012 during the RV Polarstern expedition ANTXXVIII/3. Using a 75

Teflon membrane pump, naturally Fe-depleted Antarctic seawater was collected from 24 m 76

depth and ducted directly into a laminar flow hood inside of a trace-metal clean (TMC) van.

77

The collection of the seawater and the containing phytoplankton by the membrane pump was 78

done under a pressure of ~2 bar and led to no physical damage of the collected phytoplankton 79

as observed from onboard light microscopy observations. All sampling and handling of the 80

incubations was conducted in the TMC van using TMC techniques to avoid any 81

contamination. Tubing, bubbling systems, reservoir carboys, incubation bottles and other 82

equipment were acid-cleaned prior to the cruise using TMC techniques: After a 2-day 83

Citranox detergent bath and subsequent rinsing steps with Milli-Q (MQ, Millipore), 84

equipment was kept in acid (5N HCl for polyethylene and 1N HCl for polycarbonate 85

materials) for 7 days, followed by 7 rinses with MQ. Equipment was kept triple-bagged 86

during storage and experiments. 4L polycarbonate bottles for incubation were stored under 87

acidified conditions (addition of 500 μL 10N suprapure quartz distilled HCl, Carl Roth, in 500 88

mL MQ) and rinsed twice with seawater prior to the start of the experiment. To remove any 89

large grazers, we also filtered seawater containing the natural phytoplankton community for 90

the incubation experiments through an acid-cleaned 200 μm mesh. In addition, 200 L 91

seawater were filtered through TMC filter cartridges (0.2 µm, AcroPak 1500, PALL) and 92

collected for later use as dilution seawater.

93

6

All 18 4L polycarbonate incubation bottles were placed into a cold room at 3 ±1°C and 94

exposed to a constant daylight irradiance of 30 ±5 μmol photons m^-2 s^-1 (Philips Master TL-D 95

18W daylight lamps, adjusted by neutral density screens). This irradiance level was chosen on 96

the basis of in-situ irradiance measurements at 30 m depth to mimic natural light conditions as 97

close as possible and to prevent light limitation. Triplicate incubation bottles were 98

continuously bubbled through sterile 0.2 µm air-filters (Midisart 2000, Sartorius stedim) with 99

humified air of CO2 partial pressures (pCO2) of 390 or 900 μatm (e.g. ambient and high pCO2

100

treatments). CO₂ gas mixtures were generated using a gas flow controller (CGM 2000 MCZ 101

Umwelttechnik), in which CO2-free air (<1 ppmv CO2, Dominick Hunter) was mixed with 102

pure CO2 (Air Liquide Germany). The CO2 concentration in the mixed gas was regularly 103

monitored with a non-dispersive infrared analyzer system (LI6252, LI-COR Biosciences) 104

calibrated with CO2-free air and purchased gas mixtures of 150 ±10 and 1000 ±20 ppmv CO2

105

(Air Liquide Germany). The influence of the Fe availability on the phytoplankton community 106

was investigated by growing the incubations under natural total dissolved Fe (dFe) 107

concentrations of 0.23 ± 0.06 nmol L^-1 (e.g. control treatment) or under Fe enrichment of 0.5 108

nmol L^-1 through either the addition of FeCl₃ (ICP-MS standard, TraceCERT, Fluka; e.g.

109

FeCl3 treatment) or 0.25 mg L^-1 Australian dust (e.g. dust treatment) originating from the 110

Buronga region, New South Wales. The mineral dust used in this experiment was collected 111

during a dust storm on 26th September 2009, using a High Volume Air Sampler situated on 112

the roof (4th floor) of the Environmental Sciences building at Griffith University, Nathan 113

Campus, Brisbane, QLD, Australia. Prior addition into the respective experimental bottle, the 114

dust was dissolved in 10 mL Milli-Q water and subsequently shaken for 2 min to homogenize 115

the solution. Under these conditions, most of the Fe associated with the dust material used 116

was indeed colloidal (0.02 to 0.2 μm) or particulate, releasing only 1% of Fe into the 117

dissolved phase (C. Hassler, unpublished data). As the concentration of in-situ organic ligands 118

exceeded the FeCl₃ enrichment (Table 2), it is expected that the added Fe was buffered by 119

7

naturally present ligands rather than forming inorganic colloids. Each experimental treatment 120

was run in triplicate. To check that the pCO2 and iron manipulations were successful, in 121

addition to the incubations bottles Fe and carbonate chemistry was determined from abiotic 122

control bottles, which contained only filtered seawater (0.2 μm) and that were exposed to the 123

same experimental treatments as the incubation bottles (pCO2 and Fe availability). Such 124

abiotic control bottles confirmed the successful manipulation by addition of FeCl₃ or dust, 125

yielding similar high dFe concentrations irrespective of the Fe source added (Table 1).

126

Initial nitrate, phosphate, and silicic acid concentrations in the collected seawater were 127

24.6, 1.6, and 32.1 µmol L^-1, respectively. No additional macronutrients were added to the 128

incubation bottles. To monitor nutrient drawdown as an indirect indicator of phytoplankton 129

growth, macronutrient concentrations were determined colorimetrically onboard on a regular 130

basis (1-2 days) with a Technicon TRAACS 800 Auto-analyzer following procedures 131

improved after Grasshoff et al. (1999). When nitrate concentrations fell below 14 µmol L^-1, 132

all incubations were sampled apart from 200 mL, which were topped up with 4L of collected 133

filtered seawater to prevent significant changes in seawater chemistry. While control 134

treatments were diluted once, dust and FeCl₃ treatments were diluted twice and freshly 135

amended with Fe (FeCl3 or dust, Fig. 1). The growth phase prior and after the first dilution 136

denotes the first and the second experimental phase, respectively (Fig. 1). In total, incubation 137

experiments lasted between 25 and 34 days depending on experimental treatment.

138 139

Seawater carbonate chemistry 140

For the determination of the seawater carbonate chemistry, samples for total alkalinity (TA), 141

dissolved inorganic carbon (DIC) and pH were collected. TA samples were taken from the 142

filtrate (Whatman GFF filter, ~0.6 µm), fixed with 0.03% HgCl₂ and stored in 100-mL 143

borosilicate flasks at 4 °C until further analysis. TA was estimated from duplicate 144

potentiometric titration (Brewer et al. 1986) at the Alfred Wegener Institute (Germany) using 145

8

a TitroLine alpha plus (Schott Instruments) and calculated from linear Gran Plots (Gran 146

1952). DIC samples were gently sterile-filtered (Sartorius stedim, 0.2 µm), fixed with 0.03%

147

HgCl₂ and stored in 5-mL borosilicate flasks free of air bubbles at 4°C until they were 148

measured with a QuAAtro Autoanalyzer (Seal Analytical) at the home laboratory. Seawater 149

pH was measured on board using a pH/ion meter (model 713, Metrohm) that was calibrated 150

(three-point calibration) using National Institute of Standards and Technology-certified buffer 151

systems. Seawater carbonate chemistry (including pCO2) was calculated from TA and DIC, 152

silicic acid (31 µmol kg^-1), phosphate (1.6 µmol kg^-1), temperature (3 °C), and salinity (34) 153

using CO2SYS (Pierrot et al. 2006). Equilibrium constants of Mehrbach et al. (1973) refitted 154

by Dickson et al. (1987) were chosen. The calculated pH(total scale) and pCO2 values are given 155

in Table 1.

156 157

Determination of total dissolved Fe concentrations and Fe chemical speciation 158

Total dissolved Fe (dFe) concentrations were determined in the initial seawater (SW), at the 159

end of the first and second experimental phase (including the second dilution for dust and 160

FeCl₃ treatments) while Fe chemical speciation was estimated only in the initial SW and in 161

SW samples taken at the end of the first experimental phase. Samples for dFe and Fe 162

chemical speciation were filtered through HCl-cleaned polycarbonate filters (0.2 µm poresize, 163

47 mm, Nucleopore). While Fe speciation samples were double bagged, frozen and stored at - 164

20 °C until further analysis, total dFe samples were determined onboard by voltammetry 165

following the protocol described by Laglera et al. (2013). The conditional chemical speciation 166

of Fe was determined using the competitive ligand exchange adsorptive cathodic stripping 167

voltammetry and the ligand 2-(2-thiazolylazo)-p-cresol (TAC, 10 µmol L^-1, LOT 30549, Alfa 168

Aesar) according to Croot & Johansson (2000). All samples were gently defrosted in the 169

refrigerator, dispensed in 10 mL polypropylene tubes and allowed to reach room temperature 170

for 4h. Then 4-(2-Hydroxyethyl)-1-piperazinepropanesulfonic acid (EPPS, 5 mmol L^-1) and 171

9

increasing inorganic Fe (FeCl₃, ICP-MS Standard, Fluka) were added to titrate the natural 172

ligands (L) present in the SW at a fixed pH of 8.1. Following 2h equilibration, TAC was 173

added and samples were left to equilibrate overnight prior to analysis. Measurements were 174

done with a bioanalytical system (BASi) consisting of an EC epsilon potentiostat and a 175

controlled growth mercury electrode (CGME). The working electrode medium mercury drop 176

(size 8) was used with the static mercury drop electrode (SMDE) instrument setting, together 177

with an Ag/AgCl reference electrode and a platinum wire counter electrode. An α = 327 was 178

determined with a titration using diethylene triamine pentaacetic acid (DTPA). To determine 179

Fe conditional chemical speciation, both the concentration of ligands (sum of ligands within 180

our detection window, ∑L) and the conditional stability constants (KFe’∑L) were calculated 181

according to the non-linear fit method of Gerringa et al. (1995) and the linearization method 182

van den Berg (1982). In order to also inform on the potential bioavailable Fe forms, the 183

concentration of Fe exchangeable after addition of 10 µmol L^-1 of the ligand TAC to form 184

Fe(TAC)2 complexes (Fe Labile) and the inorganic Fe concentrations are given as well as the 185

side-coefficient reactions for Fe’ and Fe³⁺. 186

187

Chlorophyll a fluorescence 188

Chlorophyll a fluorescence was determined on a regular basis (1-3 days) using a fluorescence 189

induction relaxation system (FIRe, Satlantic, Halifax, Canada). Samples were 1 h dark- 190

acclimated prior to measurements to ensure that all photosystem II (PSII) reaction centers 191

were fully oxidized. The duration of the dark acclimation was chosen after testing different 192

time intervals (data not shown). Samples were then exposed to a strong short saturating flash 193

(80 µs Single Turnover Flash, STF), which was applied in order to cumulatively saturate 194

PSIIs. A period (60 µs) of 40 weak modulated light pulses followed to record the relaxation 195

kinetics of fluorescence yield. Afterwards, a longer saturating pulse (20 ms Multiple Turnover 196

Flash, MTF) was applied in order to saturate PSII as well as the PQ pool. From this 197

10

measurement, the minimum (F_o) of the STF and maximum (F_m) fluorescence of the MTF was 198

determined. Using these two parameters, the maximum quantum yield of PSII (Fv/Fm) was 199

calculated according to the equation (F_m - F_o)/ F_m. Blank corrections at each gain setting were 200

performed with 0.22-µm-filtered seawater. The photosynthetic parameters Fo and Fm were 201

fitted using the FIRePro software provided by Satlantic inc. (v. 1.20., Halifax, Canada). All 202

measurements were conducted at 3 °C.

203 204

Phytoplankton community characterization and biomass estimates 205

To determine the taxonomic phytoplankton compositions, aliquots of 200 mL unfiltered 206

seawater were preserved with both hexamine-buffered formalin solution (2% final 207

concentration) and lugol (1% final concentration) at the start, at the end of the first (prior to 208

the first dilution), during (second dilution) and at the end of the second phase (after the first 209

dilution) of the experiments. Preserved samples were stored at 4 °C in the dark until further 210

analysis by inverted light microscopy (Axiovert 200, Zeiss). After transfer of 10 mL of 211

sample into Hydrobios sedimentation chambers, allowing settling of the cells for at least 24h, 212

the dominant phytoplankton species Pseudo-nitzschia prolongatoides, Nitzschia lecointei, 213

Chaetoceros neogracilis, Fragilariopsis curta, and Phaeocystis antarctica at the end of the 214

first and second experimental phase were enumerated according to the method of Utermöhl 215

(1958) following the recommendations of Edler (1979). Please note that other phytoplankton 216

species were present and counted, but represented less than 5% of the total phytoplankton 217

community and therefore were not taken into account. Each aliquot was examined until at 218

least 400 cells had been counted in stripes. The dominating phytoplankton species were 219

identified using scanning electron microscopy (Philips XL30) according to taxonomic 220

literature (Tomas & Hasle 1997). Net growth rates (µ) of the dominant phytoplankton species 221

as well as of the whole phytoplankton community were calculated as 222

μ = (ln Nt2 – ln N_t1)/Δt (1) 223

11

where N_t1 and N_t2 denote the cell abundances at the respective sampling days t₁ and t₂, and Δt 224

is the corresponding incubation time in days. According to the microscopic determination and 225

counting microzooplankton grazer abundance (<200 µm) remained unaltered in any treatment 226

at any sampling time over the incubation experiment. For analysis of particulate organic 227

carbon (POC), seawater was filtered onto precombusted GF/F-filters (15h, 500 °C) at the end 228

of the first and second experimental phase. Filters were stored at -20 °C and dried for >12h at 229

60 °C prior to sample preparation. Analysis was performed using an Automated Nitrogen 230

Carbon Analyser mass spectrometer system (ANCA-SL 20-20, SerCon Ltd.). POC content 231

was corrected for blank measurements and normalized to filtered volume. Taking into account 232

the corresponding incubation time in days, net daily POC production rates were calculated.

233

Samples for the determination of biogenic silica (BSi) were filtered through a cellulose 234

acetate filter (Sartorius, 0.4 µm) and stored at -20 °C. The filters were then digested in 0.2 N 235

NaOH, at 95 °C for 60 minutes, neutralized with 1 M HCl according to Brzezinski and Nelson 236

(1995) and analyzed colorimetrically for silicate using standard spectrophotometric 237

techniques (Koroleff, 1983). BSi content was normalized to filtered volume and POC content.

238 239

Fe uptake 240

At the end of the first and second experimental phase, Fe uptake capacities were estimated by 241

addition of 1 nM ⁵⁵Fe (Perkin Elmer, 33.84 mCi mg^-1 as ⁵⁵FeCl₃ in 0.5 N HCl) to the 242

unfiltered seawater sample after 2-4 h dark-acclimation. Generally, 2 mL were taken from all 243

samples to determine the initial amount of ⁵⁵Fe. Subsequently, cells were exposed for at least 244

24h to 30 µmol m^-2 s^-1 continuous light. At the end of the incubation time, the sample was 245

filtered onto GF/F-filters and rinsed 5 times with oxalate solution that was gravity-filtered for 246

approx. 2 min between each rinsing step, then the filter was rinsed 3 times with natural 247

seawater and filtered (Hassler et al. 2011). Finally, each filter was collected in a scintillation 248

vial, amended with 10 mL scintillation cocktail (Ultima Gold, Perkin Elmer) and mixed 249

12

thoroughly (Vortex). Counts per minute were then estimated for each sample on the shipboard 250

scintillation counter (Tri-Carb 2900TR). Counts per minute were then converted into 251

disintegrations per minute taking into account the radioactive decay and custom quench 252

curves. ⁵⁵Fe uptake was then calculated taking into account the initial ⁵⁵Fe concentration and 253

the total dissolved iron concentration (background and added). ⁵⁵Fe uptake rates were 254

normalized to POC.

255 256

Statistics 257

Interactive effects of the two pCO2 (390 and 900 µatm) and Fe treatments (control, dust, 258

FeCl3) on experimental parameters were statistically analyzed using two-way ANOVA with 259

Bonferroni’s post tests. Statistical analyses were performed using the program GraphPad 260

Prism (Version 5.00 for Windows, Graph Pad Software, San Diego California, USA). All 261

significance testing was done at the p < 0.05 level. The dissimilarity analysis of 262

phytoplankton community composition for the different treatments was performed according 263

to Zuur et al. (2007). A dissimilarity index (DI) of 1.00 denotes 100% dissimilarity.

264

13 Results

265

Seawater chemistry 266

Initial seawater carbonate chemistry represented present-day levels (pH: 8.00, pCO₂: 443 267

µatm, DIC: 2176 µmol kg^-1, TA: 2293 µmol kg^-1, Table 1). After sampling, all Fe treatments 268

(control, dust, and FeCl3) were continuously bubbled with a pCO2 of 390 or 900 μatm, 269

resulting in a stable average pH of 7.99 ± 0.03 and 7.72 ± 0.02, respectively, over the course 270

of the experiment (Table 1). Initial nitrate, phosphate, and silicic acid concentrations in the 271

collected seawater were 24.6, 1.6, and 32.1 µmol L^-1, respectively. Over the course of the 272

experiment, concentrations of nitrate never fell below 13.2 µmol L^-1 (Fig. 1d-f, while silicic 273

acid and phosphate concentrations were always above 11.4 and 0.54 µmol L^-1, respectively 274

(data not shown). Total dissolved Fe (dFe) concentration in the initial seawater was 0.23 ± 275

0.06 nmol L^-1 (Table 1). The enrichment with dust and FeCl3 significantly increased total dFe 276

concentrations in abiotic control bottles (p < 0.0001). With increasing pCO₂, total dFe 277

concentrations of the respective abiotic control bottles (control, dust, and FeCl3) remained 278

unaltered. Over the course of the experiment, total dFe remained unaltered in control 279

treatments, but was drawdown in dust and FeCl₃ treatments (Table 1).

280 281

Chlorophyll a fluorescence 282

The maximum quantum yield of photosystem II (PSII; F_v/F_m) of the initial phytoplankton 283

community was 0.33 ± 0.06 (Fig. 1). On day 3, Fv/Fm values went up to 0.48 ± 0.04 in all 284

treatments. In the following, F_v/F_m decreased ≤ 0.2 in all control treatments while F_v/F_m of the 285

dust- and FeCl3-treatments did not fall below 0.2 until the end of the first experimental phase.

286

After dilution with the initially collected filtered seawater, Fv/Fm values increased to ~0.3 in 287

control treatments and to ~0.4 in dust and FeCl₃ treatments. Over the duration of the second 288

experimental phase, Fv/Fm remained ≤ 0.2 in all control treatments whereas Fv/Fm values did 289

14

not fall below 0.2 in dust- and FeCl₃-treatments. Generally, dust addition resulted in similar 290

trends in Fv/Fm as FeCl3 enrichment. Over the course of the experiment, Fv/Fm was 291

significantly influenced by Fe availability (p < 0.0001), but not by pCO₂ (p > 0.05).

292 293

Fe chemical speciation 294

Fe chemical speciation was determined using cathodic stripping voltammetry in the initial 295

seawater (SW) and at the end of the first experimental phase (Table 2). At the end of the first 296

experimental phase, among all treatments dFe had returned to its initial value, being on 297

average 0.22 ± 0.05 nmol L^-1. There was no significant trend by CO2 and/or Fe on 298

concentrations of labile Fe (Fe Labile) and the sum of inorganic Fe forms (Fe’), being on 299

average 0.05 ± 0.04 nmol L^-1 and 0.16 ± 0.13 pmol L^-1, respectively. Variation within the 300

same treatment (e.g., Control 390, Table 2) for concentrations of the sum of all ligands (ΣCL) 301

and their conditional stability constants with respect to Fe′ (Log KFe′L) prevented any 302

statistical differences among treatments. ΣCL calculated according to Van den Berg (1982) 303

and Gerringa et al. (1995) were on average 0.78 ± 0.37 nmol L^-1 and 0.77 ± 0.35 nmol L^-1, 304

respectively. Log KFe′L calculated according to Van den Berg (1982) and Gerringa et al.

305

(1995) were also similar. Using ΣCL and Log KFe′L, the side coefficient of dissolved Fe 306

complexes ligands (log α_Fe3+L; being α_Fe3+L = K times C_L) was calculated as indicator of the 307

overall Fe binding to organic ligands. In response to changes in Fe availability and pCO₂, log 308

αFe3+L showed a strong Fe complexation in all treatments, ranging between 12.64 and 13.65.

309 310

Phytoplankton community characterization and biomass estimates 311

As the diatoms Pseudo-nitzschia prolongatoides, Nitzschia lecointei, Chaetoceros 312

neogracilis, and Fragilariopsis curta as well as the prymnesiophyte Phaeocystis antarctica 313

contributed to more than 5% to the community at the end of the first and second experimental 314

phases (Fig. 2), their relative contributions are shown in Fig. 3 and Table 3. At the end of the 315

15

first phase, diatoms dominated all treatments, accounting for 62 ± 3% up to 95 ± 1% (Fig. 2, 316

3, Table 3). Under ambient pCO2, P. prolongatoides dominated all communities (52 ± 3% - 317

72 ± 3%) while P. antarctica became the most abundant species in all high pCO₂ treatments 318

(35 ± 6% - 39 ± 3%). Hence, there was a significant OA-dependent species shift within the 319

phytoplankton community (Control treatments: Dissimilation index, DI = 0.89, Dust 320

treatments: DI = 0.91, FeCl₃-treatments: DI= 0.77). At the end of the second phase, P.

321

prolongatoides still dominated the control and dust treatments under ambient pCO2 (85 ± 6%

322

and 62 ± 10%, respectively), but not the FeCl₃ treatments with similar carbonate chemistry. In 323

the latter, N. lecointei was the most prevalent species (45 ± 5%, Figs. 2, 3, Table 3). In 324

response to OA, P. antarctica still dominated the control and dust treatments (57 ± 12% and 325

47± 8%, respectively) while C. neogracilis became the most abundant species in FeCl₃ 326

treatments (46 ± 4%, Figs. 2, 3, Table 3). At both pCO2 levels, the phytoplankton community 327

composition of control treatments was similar to the one of the dust treatments (390 328

treatments: DI = 0.13, 900 treatments: DI = 0.06), but significantly differed from the FeCl3

329

treatments (390 treatments: DI = 0.71, 900 treatments: DI = 0.39).

330

At the end of the first phase, net growth rates (µ) of P. antarctica were not altered by Fe 331

availability, but significantly increased with increasing pCO2 in all treatments (p = 0.0014) 332

(Fig. 4a). In comparison, net growth rates of the diatom community were controlled by the 333

availability of both pCO₂ (p < 0.0001) and Fe (p = 0.0325) (Fig. 4b). With increasing pCO₂, 334

net growth rates significantly declined in all Fe treatments, with the control treatment 335

exhibiting the strongest decline (posthoc: p < 0.001). At the end of the experiments, net 336

growth rates of P. antarctica were significantly influenced by Fe availability (p < 0.0001) as 337

well as changes in pCO2 (p = 0.0072) (Fig. 4c). Under ambient pCO2, µ increased by 42%

338

(posthoc: p < 0.05) and 50% (posthoc: p < 0.01) after addition of dust and FeCl₃, respectively.

339

Within high pCO2 treatments, only the addition of FeCl3 stimulated µ by 19% (posthoc: p <

340

0.05). Net growth rates of the diatom community showed a significant effect only by Fe 341

16

availability (p < 0.0001), with a significant increase by 59% and 75% under ambient (posthoc:

342

p < 0.001) and high (posthoc: p < 0.0001) pCO2, respectively, after enrichment with FeCl3, 343

but not with dust (Fig. 4d).

344

At the end of the first phase, net daily POC production was significantly affected by Fe 345

availability (p < 0.0005) and changes in pCO2 (p < 0.0007) (Table 4). At both pCO2 levels, 346

net daily POC production significantly increased in response to addition of FeCl₃ (posthoc p <

347

0.05) and dust, but in the latter only under high pCO2 (posthoc p < 0.05). In response to OA, 348

net POC production significantly declined in control and FeCl₃ treatments (posthoc p < 0.05, 349

but not in response to dust enrichment). Final net POC production did neither change in 350

response to pCO2 nor to Fe availability.

351

The BSi: POC ratio of the initial phytoplankton community was 0.69 ± 0.02 mol mol^-1 352

(Table 4). At the end of the first phase, BSi: POC ratios did not change in response to changes 353

in Fe availability and pCO₂. A significant OA-dependent reduction by 50% in BSi: POC 354

ratios of final phytoplankton communities were observed in both control and dust treatments 355

(p < 0.05).

356 357

Fe uptake 358

The Fe uptake: POC ratio of the initial phytoplankton community was 26.51 ± 3.97 µmol 359

mol^-1 (Table 4). At the end of the first phase, Fe uptake: POC ratios remained unaffected by 360

changes in Fe availability and pCO2 (Table 4). At the end of the experiments, Fe uptake: POC 361

ratios were strongly controlled by Fe availability (p = 0.001) and pCO₂ (p < 0.0001) as well as 362

their interactive effects (p = 0.0006) (Table 4). Following addition of dust or FeCl3, Fe 363

uptake: POC ratios were not altered under ambient, but under high pCO2, being reduced by 364

29% (posthoc: p < 0.01) and by 52%, respectively (posthoc: p < 0.0001). With increasing 365

pCO2, Fe uptake: POC ratios of control treatments and dust treatments were significantly 366

17

enhanced by 127% (posthoc: p < 0.0001) and 73% (posthoc: p < 0.01), respectively, but 367

remained unaltered in FeCl3 treatments.

368

18 Discussion

369

Due to the importance of the SO in sequestering anthropogenic CO2 (Sabine et al. 2004, 370

Landschützer et al. 2015), understanding the effect of natural Fe sources under different CO₂ 371

scenarios on SO primary productivity and phytoplankton species composition can help to 372

elucidate their combined effects on the biological carbon pump at present-day and in the 373

future. To assess their potential impacts, CO₂-Fe perturbation bottle experiments with natural 374

phytoplankton assemblages can serve as valuable tools as they account for species 375

interactions and reduce the complexity by targeting the investigated environmental factors.

376

These studies are, however, also biased by possible bottle effects as they remove the 377

phytoplankton community from its natural into an artificial environment (i.e. lack of sinking 378

and grazers larger than 200 µm; Venrick et al. 1977, Calvo-Diaz et al. 2011), complicating 379

thereby projections to the real world. Yet, to date multiple stressor experiments represent the 380

only tool to simulate future potential climate change scenarios and their impact on future SO 381

phytoplankton, hence these experiments can facilitate the interpretation of cause-effect 382

relationships and have the potential to identify phytoplankton species in the field that could be 383

sensitive/tolerant toward the tested climate change scenarios. Here we present results from a 384

bottle incubation experiment with a phytoplankton community from SO pelagic waters 385

elucidating the impact of different Fe sources (dust versus FeCl₃) and pCO₂ levels (390 versus 386

900 µatm) on Fe chemistry, particulate organic carbon (POC) production and phytoplankton 387

species composition.

388

The initial phytoplankton community, sampled south of the Polar Front in the Atlantic 389

sector of the SO, was composed of numerous diatom species and the single-celled 390

prymnesiophyte Phaeocystis antarctica. Biogenic silica (BSi) to particulate organic carbon 391

(POC) was high (0.69 ± 0.02 mol mol^-1, Table 5), indicating a dominance of diatomsunder 392

Fe-deficient (Trull et al. 2015; BSi: POC ~0.6 mol mol^-1) compared toFe-replete(BSi: POC 393

~0.15 mol mol^-1) conditions (Hutchins & Bruland 1998, Takeda 1998, Hoffmann et al. 2007, 394

19

Assmy et al. 2013). Both the low in-situ total dissolved Fe (dFe) concentration (0.23 ± 0.06 395

nmol L^-1, Table 1) and the reduced maximum quantum yield of photosystem II (PSII; Fv/Fm of 396

0.33 ± 0.06, Fig. 1) were similar to values typically observed for Fe-limited waters of the SO 397

(Hopkinson et al. 2007, Klunder et al. 2011, de Jong et al. 2012, Trimborn et al. 2015), 398

suggesting an Fe-limited phytoplankton assemblage at the start of the experiment. Fe 399

fertilization was successfully achieved through addition of 0.5 nmol L^-1 totaldissolved Fe 400

(dFe) of dust or FeCl3, yielding similar initial total dFe concentrations (Table 1).

401

At the beginning of the first experimental phase, F_v/F_m values increased up to 0.48 ± 402

0.04 in all treatments (Fig. 1a-c). This rise in Fv/Fm was not related to total dFe concentrations 403

(Table 1), but was most likely a response to the constant 30 µmol m^-2 s^-1 light supply(Feng et 404

al. 2010). After acclimation to this irradiance, F_v/F_m decreased to ≤ 0.2 in all control 405

treatments, confirming severe Fe-limitation, while Fv/Fm of the dust- and FeCl3-treatments 406

never fell below 0.2 until the end of the experiments (Fig. 1a-c). In line with this, nitrate 407

concentrations were drawdown to 10 µmol L^-1 already after 10 days or at latest after 13 days 408

in FeCl3- and dust-enriched treatments whereas it took 15 days or longer for control 409

treatments (Fig. 1d-f), confirming therefore that Fe-limitation was relieved and phytoplankton 410

growth was stimulated through addition of either FeCl3 or dust. In fact, enrichment by dust 411

showed similar trends in F_v/F_m relative to FeCl₃ amendment irrespective of the pCO₂, 412

suggesting that dust can successfully relieve Fe-limitation, as previously observed (Mélancon 413

et al. 2016, Conway et al. 2016). Accordingly, total dFe concentrations increased initially 414

after Fe enrichment (dust and FeCl₃) whereas no significant differences in total dFe were 415

observed at the end of the first phase (Table 1). These findings further suggest that this dust 416

was rapidly, rather than continuously, releasing Fe (Baker & Croot 2008, Shi et al. 2011, 417

Fishwick et al. 2016). Furthermore, similar total dFe concentrations were measured within the 418

respective Fe treatments under both pCO2 levels (Table 1), even for the dust treatment, 419

supporting previous observations that the effect of OA on Fe(III)' solubility is rather small in 420

20

seawater over the pH range of 7.5-9 (Kuma et al. 1996, Liu & Millero 2002, Fishwick et al.

421

2014). Hence, our results suggest that Fe solubility associated with dust was not significantly 422

enhanced under OA as previously observed (Fishwick et al. 2014, Mélancon et al. 2016).

423

Previous studies, however, reported changes in Fe(III) complexation, resulting in a decline in 424

the bioavailability of Fe(III) to marine phytoplankton (Shi et al. 2010, Sugie et al. 2013).

425

According to our results, (derived from pH 8.1 titrations) no changes in ligand concentrations 426

(ΣCL) and Fe(III)-ligand conditional binding affinity were detected in response to OA (Table 427

2). Considering, however, that variations even within the same experimental treatment for 428

ΣCL were high (e.g., Control 390, Table 2), further studies are required to verify this.

429

Even though Fe additions (dust or FeCl3) imposed strong changes in Fv/Fm (p < 0.0001, 430

Fig. 1a-c), only a CO₂-dependent change in the phytoplankton community structure was 431

observed (Figs. 2, 3, Table 3). At the end of the first phase, at both pCO2 levels net growth 432

rates were negligible for the single-celled P. antarctica (Fig. 4a), but positive for the diatom 433

community (Fig. 4b). Such rapid growth of diatoms is commonly referred to as boom-and- 434

bust strategy (Smetacek et al. 2004). The difference in the phytoplankton community 435

composition between ambient and high pCO₂ mainly resulted from the absence of P.

436

prolongatoides under all OA treatments (Fig. 2, 3, Table 3), as shown by the unaltered 437

relative abundances of Nitzschia lecointei, Chaetoceros neogracilis, Fragilariopsis curta and 438

Phaeocystis antarctica under these conditions (Supplement Fig. S1). Such CO₂ sensitivity of 439

the genus Pseudo-nitzschia has been previously observed in bottle incubation experiments 440

with natural phytoplankton assemblages from the Ross Sea and the Weddell Sea (Tortell et al.

441

2010; Hoppe et al. 2013). Similarly, growth of Pseudo-nitzschia spp. remained unaffected by 442

OA in bottle incubation experiments with natural phytoplankton assemblages from the Bering 443

Sea (Sugie et al. 2013) as well as in laboratory experiments with monocultures of the 444

Antarctic Pseudo-nitzschia subcurvata (Trimborn et al. 2013) or the temperate Pseudo- 445

nitzschia pseudodelicatissima (Sugie & Yoshimura 2013). An OA-dependent stimulation in 446

21

growth was reported yet only for temperate species such as Pseudo-nitzschia multiseries (Sun 447

et al. 2011) and Pseudo-nitzschia fraudulenta: (Tatters et al. 2012), pointing out the 448

vulnerability of Antarctic Pseudo-nitzschia species to OA in particular.

449

During the second experimental phase, the similarity of the community composition for 450

dust and FeCl3 treatments prior and after the second dilution (Table 3) suggests that a steady 451

state situation among species was reached. During this phase, both CO₂ and Fe sources were 452

key modulators of the phytoplankton community composition. Under ambient pCO2, FeCl3- 453

enrichment induced a shift within the diatom community from P. prolongatoides to N.

454

lecointei (Fig. 2, 3, Table 3). This floristic shift in response to FeCl3 addition was further 455

modulated by pCO2, resulting in high abundances of C. neogracilis under OA (Fig. 2, 3, 456

Table 3). At both pCO₂ levels, diatoms became dominant following FeCl₃ addition (Table 3).

457

Such diatom-specific response to inorganic Fe enrichment has been previously reported 458

(Tsuda et al. 2005, Boyd et al. 2007, Feng et al. 2010, Smetacek et al. 2012, Assmy et al.

459

2013, Hoppe et al. 2013). In both control and dust treatments, high pCO2 caused the single- 460

celled P. antarctica to reach accumulation rates that were as high as those observed for the 461

overall diatom community (Fig. 4c, d), resulting in it being the most abundant species in these 462

treatments (Table 3).

463

Surprisingly, phytoplankton community composition of control and dust treatments 464

were similar (390 treatments: DI = 0.13, 900 treatments: DI = 0.06), with P. prolonatoides 465

and P. antarctica dominating both treatments under ambient and high pCO2, respectively 466

(Fig. 2, 3, Table 3). This finding suggests also relatively low Fe bioavailability for Fe 467

associated with dust. When comparing the phytoplankton community composition of FeCl3

468

and dust treatments, our results reveal that the dominating phytoplankton species markedly 469

differed between Fe sources (Figs. 2, 3, Table 3), under ambient pCO₂ with P. prolongatoides 470

and N. lecointei (DI = 0.48) and under high pCO2 with P. antarctica and C. neogracilis (DI = 471

0.30) dominating the dust and FeCl₃ treatments, respectively. Hence, the difference in 472

22

phytoplankton community composition between dust and FeCl₃ treatments suggests that in 473

fact FeCl3 fails to mimic dust enrichments. The reason for this may potentially result from 474

species-specific abilities to access different Fe pools (e.g. Hutchins et al. 1999, Maldonado &

475

Price 2001) and/or Fe bioavailability (e.g.Visser et al. 2003, Shaked et al. 2005, Morel et al.

476

2008, Conway et al. 2016). In response to OA, Fe uptake: POC ratios of control treatments 477

and dust treatments were further significantly enhanced by 127% (posthoc: p < 0.0001) and 478

73% (posthoc: p < 0.01), but remained unchanged in FeCl3 treatments (Table 5), indicating 479

higher Fe uptake rates by the P. antarctica-dominated relative to the diatom-dominated 480

assemblage (Hassler et al. 2011, Trimborn et al. 2015). We, therefore, suggest that low Fe 481

bioavailability in conjunction with high pCO2 may favor single-celled P. antarctica over 482

diatoms.

483

Despite similar net daily POC production rates among all treatments at the end of the 484

experiment (Table 5), potential changes in phytoplankton community structure can bear 485

important implications for the marine carbon cycle. Indeed, the strength of the biological 486

pump depends on the functional types of phytoplankton present, which act as differentially 487

efficient vectors for vertical carbon export. Phaeocystis antarctica, in particular as a singled- 488

cell form, is considered to be insignificant for vertical transport of biogenic matter 489

(Schoemann et al. 2005; Reigstad & Wassmann 2007). Diatoms, the other dominating 490

functional group observed here, can significantly affect carbon export depending on the 491

degree of silicification of their frustules(Assmy et al. 2013). In line with previous findings on 492

enhanced diatom frustule dissolution (Milligan et al. 2004) and reduced cellular BSi quotas 493

(Sun et al. 2011, Hoppe et al. 2015, Sugie & Yoshimura 2016) in diatoms under OA, we 494

observed a significant OA-dependent reduction by 50% in BSi:POC ratios of both control and 495

dust treatments (p < 0.05; Table 5). The phytoplankton community shift towards P. antarctica 496

and weakly silicified diatoms could decrease the strength of the biological pump under OA 497

23

across the Fe-limited pelagic SO, an important region accounting for ~90% of the total annual 498

primary production south of 50°(Arrigo et al. 2008).

499

Even though multiple stressors bottle incubation experiments are biased by possible 500

bottle effects (artificial environment, lack of large grazer, sinking), currently they are the only 501

tool to assess future potential climate change scenarios and their impact on future SO 502

phytoplankton. Our results clearly support previous observations that under future increased 503

CO2 conditions phytoplankton community structure and potential carbon export of SO HNLC 504

waters can be altered (Tortell et al. 2008,Feng et al. 2010, Hoppe et al. 2013). In particular, 505

we reveal that FeCl3, commonly used in Fe-enrichment experiments, versus more 506

environmentally relevant Fe-laden dust enrichments, resulted in diverging effects on the 507

dominating phytoplankton species, demonstrating that FeCl₃ has to be used with caution to 508

reliably assess future phytoplankton community composition and export in an acidified 509

pelagic SO. Although it is difficult to predict how Southern Ocean phytoplankton species will 510

respond to the future climatic scenarios, it is obvious from this study that both the 511

characterization of the phytoplankton community at the species level and the use of natural Fe 512

sources are essential for a realistic projection of the biological carbon pump in the Fe-limited 513

pelagic SO under OA. The results of this study furthermore highlight the need to better 514

constrain the impact of OA on Fe bioavailability to SO phytoplankton species, as different Fe 515

sources differently affect species competition with important implications for future 516

biological CO2 sequestration by the SO. Moreover, multifactorial perturbation experiments 517

need to be designed that include factors, such as grazing as well as the aggregation capacity of 518

the phytoplankton community to encompass their potential for biological carbon export, to 519

fully understand the ecological responses of phytoplankton assemblages to OA in a future SO.

520

24 Acknowledgements

521

We thank S. Ossebar for macronutrient analysis and J. Hölscher for BSi and DIC 522

measurements. Thanks also to H. de Baar and M. Rijkenberg who supported us in setting up 523

the voltammetry at the AWI. We also thank F. Hinz for scanning electron microscopy pictures 524

for identification of the species and U. Schüssler for having placed a trace metal clean 525

container at our disposal. We gratefully thank M. Ellwood for providing the membrane pump 526

for trace metal clean seawater sampling. Also, we would like to thank the three reviewers for 527

their detailed and helpful comments on our manuscript. Finally, we would like to thank the 528

captain and crew of R.V. Polarstern during ANTXXVIII/3. S.T. was funded by the Deutsche 529

Forschungsgemeinschaft (DFG) in the framework of the priority programme ‘Antarctic 530

Research with comparative investigations in Arctic ice areas’, project TR 899/2. S.T., T.B.

531

and C.V. were funded by the Helmholtz Impulse Fond (HGF Young Investigators Group 532

EcoTrace). C.S.H. was funded by a Swiss National Science Foundation Professor Fellowship 533

(PP00P2_138955) and a UTS Chancellor Post-doctoral Fellowship. Furthermore, C.H. and 534

L.N. were funded by the Australian Research Council (Discovery Project DP1092892).

535

L.M.L. and J.S.-E. participation was funded by MINECO of Spain (CGL2010-11846-E). This 536

work was further supported by research grants from the European Research Council (ERC) 537

under the European Community’s Seventh Framework Programme (to C.J.M.H., FP7 ⁄ 2007- 538

2013, ERC grant agreement no. 205150).

539