Ocean acidification with (de)eutrophication will alter future phytoplankton growth and succession

Human activity causes ocean acidification (OA) though the dissolution of anthropogenically generated CO2 into seawater, and eutrophication through the addition of inorganic nutrients. Eutrophication increases the phytoplankton biomass that can be supported during a bloom, and the resultant uptake of dissolved inorganic carbon during photosynthesis increases water-column pH (bloom-induced basification). This increased pH can adversely affect plankton growth. With OA, basification commences at a lower pH. Using experimental analyses of the growth of three contrasting phytoplankton under different pH scenarios, coupled with mathematical models describing growth and death as functions of pH and nutrient status, we show how different conditions of pH modify the scope for competitive interactions between phytoplankton species. We then use the models previously configured against experimental data to explore how the commencement of bloom-induced basification at lower pH with OA, and operating against a background of changing patterns in nutrient loads, may modify phytoplankton growth and competition. We conclude that OA and changed nutrient supply into shelf seas with eutrophication or de-eutrophication (the latter owing to pollution control) has clear scope to alter phytoplankton succession, thus affecting future trophic dynamics and impacting both biogeochemical cycling and fisheries.

Axenic stock cultures were used to start experiments, and precautions taken to minimise the risk of any subsequent contamination in experimental flasks. No significant contamination (as judged by microscopy) was noted during experiments. Two-point calibrated pH probes (pH 7.01 and 10.01; HI 1131b, Hanna Instruments, Woonsocket, RI, USA) were surface sterilized with 2 mol·L -1 HCl for 15 min and rinsed with sterile pure water before being inserted into experimental vessels under aseptic conditions. Experimental vessels were sealed gas-tight with sterilized silicon bungs through which passed the pH probe lead, a sampling tube, two further tubes which facilitated pH correction, and a tube for the entry of N 2 gas to replace the liquid volume removed in sampling.
A bespoke software based pH monitoring and control unit was constructed at Plymouth Marine Laboratory (U.K.) for use in these experiments. Using this, cultures were independently monitored and, as required, the pH controlled. The unit was operated in 2 modes; a "fixed" mode where pH was maintained at a set value throughout the experiment, and a "drift" mode in which pH was monitored as it varied from a pre-determined starting point due to phytoplankton growth (increasing with net Cfixation, and decreasing with net respiration overnight). The values at which pH was either fixed, or from which it drifted, were pH 8.2 representing extant conditions and pH 7.6 ("acidic" treatments) representing an acidified ocean. In addition, a series of experiments commencing at an elevated pH (8.8; "basic" treatments) were also run.
The pH values were measured using a bench top pH meter (HI 931400, Hanna Instruments, Woonsocket, RI, USA). Both temperature, measured in tap-water filled vessels placed adjacent to experimental flasks to avoid the potentially toxic effects of stainless steel probes (HI 7669/2W, Hanna Instruments, Woonsocket, RI, USA), and culture pH were measured at a time step of 70 seconds. For cultures at fixed pH, a tolerance of 0.05 pH units was assigned and 5 consecutive measurements beyond this tolerance resulted in pH correction; metering pumps added either 1 mL of HCl or NaOH to individual cultures. During the growth cycle in these closed system cultures, dissolved inorganic carbon (DIC) concentration and buffering capacity decreased. For cultures grown at fixed pH the concentration of acid/base added to cultures was decreased as culture age increased in order to minimize the amplitude of pH change during the automated correction procedure. An initial HCl/NaOH concentration of 1.2/0.25 µmol·L -1 was used for the acclimation stage and first experimental day, decreasing to 0.12/0.125 µmol·L -1 at day 2 and 0.05/0.0625 µmol·L -1 at day 4 respectively. A log of temperature, pH and acid/base addition for each experimental vessel was recorded by the pH control unit. Changes in total alkalinity (TA) were calculated from changes in nutrient (including DIC) and pH by reference to methods used previously [22]; see model description, below.
Inoculated experimental containers were placed on magnetic stirrers within a constant temperature room, connected to the pH control unit, and kept for 14 to 16 hours for temperature and light acclimatization. During this stage pH was maintained at a constant level of 7.6, 8.2 or 8.8 as required. After acclimatization, pH control was switched to "drifting" or "fixed" modes to initiate experiments (t0).
Sample collection -Samples were collected from vessels on a daily basis for days 0-6, and then at days 8 and 10. Samples were drawn from a PTFE tube that was permanently submerged in culture media.
The first 10 mL of drawn media was discarded, and the subsequent 250-450 mL retained. N 2 gas was introduced to the atmosphere above cultures to replace the volume removed by sampling. Sample volume was stored in the dark at 4 °C and sub-sampled for a range of procedures which were completed within 4 hours.
Cell numbers and biovolume -Cell numbers and biovolume of sub-samples were measured with a FlowCam (single species experiments with Rhodomonas sp and T. weissflogii; Fluid Imaging Technologies, Yarmouth, ME, USA) equipped with 90 µm field of view flow cell and a x10 objective, and/or with a Coulter Counter (E. huxleyi and triple species experiments; Multiplier 4, Beckman Coulter GmBH, Krefeld, Germany). For the FlowCam, image data processing software "Visual Spreadsheet" was used (version 2.1.20 beta, FlowCam and software: Fluid Imaging Technologies, Yarmouth, ME, USA). Analyses of particles of the size range of 5 to 30 µm were carried out with a picture rate of 6 s -1 in auto-image mode. Images in which algal cells were absent were excluded following visual examination of images. Typically, 500 particles were counted although in low density cultures measurements were stopped after 4 minutes to avoid sedimentation effects. Biovolume was computed using the ABD FlowCam algorithm, which equates to a volume estimated from equivalent spherical diameter. Data represent the median of at least 3 replicate measurements.
Analysis of inorganic and organic nutrients -Under clean conditions (within a filtered air stream in a laminar flow cabinet), sub-samples of culture were gravity filtered through pre-combusted (550 °C, 20 min) glass fibre filters with a nominal pore size of 1 µm (Type A/E, 13 mm, Pall Corporation, Ann Arbor, MI, USA). Filters, which retained cells, were stored frozen for subsequent elemental and compositional analyses.
A segmented flow auto-analyzer (model AA3, Seal Analytical Ltd., Fareham, UK), consisting of a random access XY-2 auto-sampler, a high precision pump, chemistry module trays and a dual beam high resolution digital colorimeter operating within the range 340-900 nm, was used for inorganic analyses. Nitrate and nitrite were determined by the cadmium reduction method and sudan-1 synthesis, measured at 550 nm. The determination of silicate was based on the reduction of silicomolybdate in acidic solution to molybdenum blue by ascorbic acid and the coloured complex was measured at 820 nm. The Berthelot reaction was used for the determination of ammonium with indophenol measured at 660 nm. The analysis of inorganic phosphorous [42] produced a blue phosphor-molybdenum complex measured at 880 nm. The determination of DIC was based [43] on inverse chemistry with the use of buffered phenolphthalein indicator and was measured at 550 nm.
Analysis of particulate C, N and P -Particulate C and N content were analyzed against isoleucine standards using an elemental analyzer and mass spectrometer (ANCA PDZ-Europa and 20-20 IRMS PDZ-Europa, SerCon Ltd Crewe, UK) combined with Callisto 530 software (SerCon Ltd Crewe, U.K.).
Both samples and standards were wrapped in tin (Sn) foil. Helium gas served as system carrier and auto-sampler purge. The combustion column, which was packed with quartz wool/chromium oxide/silver wool, operated at 1000 °C, with sample combustion taking place in the presence of an O 2 pulse to improve combustion efficiency. Combustion gases were reduced in a reduction column composed of silica wool/copper wire at 600 °C. Sample derived CO 2 and N 2 gases were separated by a gas chromatography column at 70 °C and analyzed by mass spectrometry.
Particulate P content was derived by alkaline persulphate oxidation followed by analysis of the solubilized phosphate. Algal samples retained on glass fibre filters were heated for 20 minutes at 121 °C (autoclave) in sealed glass ampoules containing 2 mL of a solution containing potassium peroxodisulphate (50 g·L -1 ), boric acid (30 g·L -1 ) and sodium hydroxide (15 g·L -1 ). After shaking, 1.5 mL of the supernatant was pipetted into micro-centrifuge tubes, glass fibre remains were removed by centrifugation (20 000 x G for 5 minutes) and 1 mL of supernatant was stored frozen (-20 °C) until analysis. Inorganic P content was measured in samples diluted using purified water (ELGA Purelab ultra, VWS UK Ltd., High Wycombe, U.K.) and analyzed as phosphate in the segmented flow autoanalyzer (as described for nutrient phosphate, above).

Commentary on experiment design
The experimental design deployed was used in an attempt to achieve a mass balance for carbon across all forms of carbon in the system (see [21]). To this end, the systems were sealed to prevent ingress of CO 2 , and pH was controlled by acid/base additions rather than through changes in pCO 2 . It is noteworthy that in natural systems the equilibration of the atmosphere with the surface mixed layer (which may comprise several 10s of meters depth), is a slow process, of the order of months.
Preventing gas exchange in the experiments over a period of 10d does not thus induce an unnatural scenario.
The method used for control of pH (addition of acid/base) introduces the potential for significant changes in alkalinity, and is not the favoured approach for studying ocean acidification [44,45]. However, for the processes explored in this work attempting to rigidly maintain a constant pH (by any means) is artificial because the events in question centre around the fact that seawater pH is not constant but changes with biological activity.
The primary experimental data used for this work, for the tuning (calibration) of the models, comes from the "drift" systems, which were not subjected to any additions of acid or base other than that required for the initial setting of pH ("extant", "acidic" or "basic"). The initial chemistry of the "drift" flasks thus differed only slightly with respect to total alkalinity (TA). The fact that the highest rates and extent of growth were seen in fixed pH systems, which underwent very significant changes in TA over the duration of the experiments (figure S1), indicates that TA has no damaging influence on cell physiology, and certainly not one that is discernible against a background of changing DIC and pH in the experimental systems deployed. This subject is discussed further in [21].

Simulations
Modelling was undertaken using the same platform (Powersim Constructor v2.51, Isdalstø, Norway) and source model code as deployed previously [22]. This included a variable stoichiometric acclimative model of phytoplankton physiology, coupled to a carbonate chemistry module. The only additional code added were equations linking growth and death rates to pH and nutrient status; these are provided in Tables S1 . The carbonate  The model was tuned against experimental data for cell-C and cell N:C quotas derived for each individual organism (see Experimental) using the evolutionary tuning method supported by Powersim Solver v2 (Powersim, Isdalstø, Norway). This provided, for each organism, a single set of constants controlling physiology. Including TA as a parameter during tuning of the models had no effect on the outcome of this modelling analysis.
Dynamic sensitivity analyses -Sensitivity analyses were undertaken using the Powersim Solver v2 platform, running a Latin-hypercube routine with an initial pH set for "extant" (pH 8.2), "acidic" (pH 7.6) or "basic" (pH 8.8) together with a set standard deviation. Simulations of open water pH scenarios -Simulations were run using the phytoplankton models that had been configured against the experimental data, but now within scenarios of a water column with gas exchange at the surface, and exchange of water at the interface of the mixed layer. The basis of these simulations in terms of abiotic factors is the same as used previously [22]. The major differences, other than the lack of simulated predator activity, was that the future pCO 2 was set at 1000ppm (prediction for end of this century [47,48]), rather than 750ppm.  Tables S1 Relationship between growth rate, death, pH and nutrient status

DESCRIPTION OF FUNCTIONS
The growth function describes a bell-shaped curve with optimum, minimum and maximum pH values. The function returns a value between 0 (no growth potential) and 1 (maximum potential). As the nutrient status decreases, so the breadth of pH supportive of growth also decreases (narrows). Thus under optimal conditions of zero nutrient status, growth proceeds at a maximum rate and also over the widest pH range. As the organism becomes stressed, so the ability to withstand pH extremes lessens; the width of the bell narrows and its optimum may change. The growth rate then becomes a function of the incident irradiance and of the nutrient status (both described via the usual functions; see (22) and references therein), and of pH.
The death function describes a default death rate that under optimal nutritional status is relatively flat and low over the pH range. As nutrient stress develops so the death rate even under optimal pH for growth increases, and the rate increases even more so at pH values either side of the optimum value. The function thus resembles a 'U' shape that becomes higher and steeper with increasing nutrient stress.

PARAMETER AND EQUATIONS
The equations and definitions are defined here specifically for the diatom Thalassiosira (_diat) description. The tabulated constants provide values for the alternative values for the description of the flagellate Rhodomonas (replace _diat with _flag), and microalgae Emiliania (replace _diat with _alg).
Equations are given as ASCII text to facilitate their use in models or spreadsheets, and also as traditional formal equations. The equations contain Boolean logic terms, which take the value of 1 if true and 0 otherwise; these may be replaced by IF-THEN-ELSE constructs as required.       Figure S1. As for figure 1, but also showing changes in pH (with selected experimental data as symbols) and total alkalinity (TA). Conditions of pH were extant fixed at pH 8.2 (EF), extant drifting from pH 8.2 (ED), acidic fixed at pH 7.6 (AF), acidic drifting from pH 7.6 (AD), basic fixed at pH 8.8 Rhodomonas Figure S5. Changes in carbonate chemistry for the diatom Thalassiosira during growth under conditions of pH that were acidic fixed at pH 7.6 (AF), acidic drifting from pH 7.6 (AD), extant fixed at pH 8.2 (EF), extant drifting from pH 8.2 (ED), basic fixed at pH 8.8 (BF) or basic drifting from pH 8.8 (BD). H 2 CO 3 and HCO 3 are substrates for photosynthesis (see [49]). These values were computed by the model (see model description for comment on carbonate chemistry computation). See also figure S1.

Extant Fixed
Time ( Thalassiosira Figure S6. Experimental data (symbols) for the prymnesiophyte Emiliania huxleyi, cryptophyte Rhodomonas sp. and diatom Thalassiosira weisflogii grown together under conditions of pH that were acidic fixed at pH 7.6 (AF), or acidic drifting from pH 7.6 (AD), extant fixed at pH 8.2 (EF) or extant drifting from pH 8.2 (ED). Lines indicate model output; model configurations were those generated from fits to data in figures 1, S1, & S2. Differences between data and model output for Emiliania huxleyi likely indicate some form of allelopathic interaction between species that were unaccounted for in the model. Also, shown is simulated growth in mixed cultures for conditions of basic fixed at pH 8.8) and basic drift from pH 8.8; experiments were not conducted under these conditions. Note the value for Emiliania in the acidic fixed system at d10 is obscured by the symbol for the cryptophyte.

Extant Fixed
Time ( Figure S7. Dynamic sensitivity analysis of growth of mixed species communities commencing with fixed (equal) proportions of start biomass, but with variable "acidic" pH start points, centred at pH 7.6 with standard deviation of pH 0.2 (see model description). Each plot shows, for the indicated phytoplankton type and in a regime with either fixed or drifting pH, the average (red), +/-standard deviation (blue), with the yellow field showing the 95% limits. The relative extent of the 95% confidence limits is the key issue in these plots. Biomass has units of mgC m -3 . Systems were ultimately N-limiting. Figure S8. Dynamic sensitivity analysis of growth of mixed species communities commencing with fixed (equal) proportions of start biomass, but with variable "extant" pH start points. The initial pH was centred at pH 8.2 with standard deviation of pH 0.1 (see model description). Each plot shows, for the indicated phytoplankton type and in a regime with either fixed or drifting pH, the average (red), +/standard deviation (blue), with the yellow field showing the 95% limits. The relative extent of the 95% confidence limits is the key issue in these plots. Biomass has units of mgC m -3 . Systems were ultimately N-limiting. Figure S9. Dynamic sensitivity analysis of growth of mixed species communities commencing with fixed (equal) proportions of start biomass, but with variable "basic" pH start points, centred at pH 8.8 with standard deviation of pH 0.05 (see model description). Each plot shows, for the indicated phytoplankton type and in a regime with either fixed or drifting pH, the average (red), +/-standard deviation (blue), with the yellow field showing the 95% limits. The relative extent of the 95% confidence limits is the key issue in these plots. Biomass has units of mgC m -3 . Systems were ultimately N-limiting.