Interacting Effects of Light and Iron Availability on the Coupling of Photosynthetic Electron Transport and CO2-Assimilation in Marine Phytoplankton

Iron availability directly affects photosynthesis and limits phytoplankton growth over vast oceanic regions. For this reason, the availability of iron is a crucial variable to consider in the development of active chlorophyll a fluorescence based estimates of phytoplankton primary productivity. These bio-optical approaches require a conversion factor to derive ecologically-relevant rates of CO2-assimilation from estimates of electron transport in photosystem II. The required conversion factor varies significantly across phytoplankton taxa and environmental conditions, but little information is available on its response to iron limitation. In this study, we examine the role of iron limitation, and the interacting effects of iron and light availability, on the coupling of photosynthetic electron transport and CO2-assimilation in marine phytoplankton. Our results show that excess irradiance causes increased decoupling of carbon fixation and electron transport, particularly under iron limiting conditions. We observed that reaction center II specific rates of electron transport (ETRRCII, mol e- mol RCII-1 s-1) increased under iron limitation, and we propose a simple conceptual model for this observation. We also observed a strong correlation between the derived conversion factor and the expression of non-photochemical quenching. Utilizing a dataset from in situ phytoplankton assemblages across a coastal – oceanic transect in the Northeast subarctic Pacific, this relationship was used to predict ETRRCII: CO2-assimilation conversion factors and carbon-based primary productivity from FRRF data, without the need for any additional measurements.


Introduction
The photosynthetic assimilation of inorganic CO 2 into organic carbon by marine phytoplankton accounts for almost half of total global primary productivity [1], and variations in phytoplankton primary productivity can profoundly affect ecosystem dynamics and global climate (e.g. [2][3][4][5]). However, despite its recognized importance, it remains challenging to accurately quantify marine primary production at the temporal and spatial resolution needed to relate its variability back to external environmental conditions. In vast oceanic regions, the availability of iron (Fe) limits marine phytoplankton primary productivity [6][7][8]. This element plays a fundamental role in the photosynthetic electron transport chain (ETC) and therefore the conversion of light energy to organic carbon products [9][10][11].
Approaches currently used to measure phytoplankton primary production quantify rates at different points of the photosynthetic process (evolution of O 2 , assimilation of CO 2 , electron transport in photosystem II). These various rates can be decoupled in response to changes in environmental conditions or phytoplankton taxonomy [12]. For this reason, it is likely that iron limitation will affect the conversion factors between these various productivity metrics. Phytoplankton CO 2 -assimilation can be measured directly using the radioisotope tracer 14 C [13,14]. This technique has been widely applied in biological oceanography over the past 60 years, despite a number of well-known limitations (e.g. low spatial and temporal resolution, high cost and labour intensity, bottle artifacts due to exclusion of grazers and contamination, requirement for radio-isotopes, ambiguity of whether net or gross production is measured [14][15][16][17][18]). In recent years, bio-optical approaches have emerged as an attractive alternative to overcome these limitations. Chlorophyll a fluorescence (ChlF) yields, measured by Pump and Probe, FRR, or PAM fluorometry, can be used to estimate rates of linear electron transport (i.e. rates of charge separation) in photosystem II (ETR RCII ) [19][20][21][22][23], thus providing a measure of gross photosynthesis. Being non-intrusive, instantaneous and relatively inexpensive, these approaches can be used to examine phytoplankton photophysiology at unmatched spatial and temporal resolution, and improve the coverage of productivity estimates over vast oceanic domains.
Despite significant potential, active ChlF approaches are currently not widely applied to monitor rates of phytoplankton primary productivity. This is due, in part, to uncertainty in the conversion of ETR RCII to ecologically relevant rates of CO 2 -assimilation [12,24]. Numerous studies conducted over the past decades have collectively shown that the conversion factor linking ETR RCII to CO 2 -assimilation in phytoplankton is not constant, but changes in response to taxonomy and environmental conditions [12,20,. On the physiological level, ETR R-CII and CO 2 -assimilation can be uncoupled by a number of energy-allocation processes that evolved to maximize photosynthetic efficiency while preventing photodamage. Marine phytoplankton evolved an exceptional photosynthetic plasticity to achieve this balance under low nutrient and fluctuating light conditions. A number of recent studies have examined this finetuning of electron transport and energy allocation within the phytoplankton photosynthetic apparatus, providing mechanistic insight into the processes decoupling CO 2 -assimilation and photosynthetic electron transport (e.g. [54][55][56][57][58][59][60][61][62][63]).
In this study, we examine the interacting effects of iron levels and instantaneous light availability on the coupling of ETR RCII and CO 2 -assimilation in marine phytoplankton. We derived rates of ETR RCII normalized to PSII reaction center content (mol emol RCII -1 s -1 ), resulting in a conversion factor consisting of two parameters: the amount of chlorophyll a (chl a) functionally connected to each RCII (1/n PSII , mol chla -1 mol RCII -1 ), and the electron requirement for carbon fixation (F e:C , mol emol C). Working with natural phytoplankton assemblages in the Northeast subarctic Pacific, and mono-specific phytoplankton cultures in the laboratory, we conducted simultaneous measurements of FRRF-derived ETR RCII and 14 C-based CO 2 -assimilation over a range of irradiances (PvsE curves) under high and low iron conditions. Our results demonstrate significant and interactive effects of irradiance and iron availability on the coupling of ETR RCII and CO 2 -assimilation, with an increase in the conversion factor F e:C /n PSII under excess light and low iron conditions. From a photophysiological point of view, increased decoupling appeared to be caused by the effects of increased excitation pressure on the photosynthetic ETC, resulting in a strong correlation between the derived conversion factor and the expression of non-photochemical quenching (NPQ) in the antennae of PSII. This correlation can, in turn, be used to derive rates of carbon-based productivity from FRRF data, without the need for any additional measurements.

Methods
In this study, we utilized three separate datasets. First, we examined the coupling of ETR RCII and CO 2 -assimilation in a mixed phytoplankton assemblage during a 6 day ship-board iron addition experiment in iron-limited waters of the subarctic Pacific (Fig 1). Secondly, we conducted experiments with two mono-specific phytoplankton cultures grown under controlled light and iron conditions in the laboratory. These experiments were conducted to examine the physiological effects of iron and light on the conversion factor F e:C /n PSII , in the absence of potentially confounding taxonomic shifts. Finally, we applied the results obtained from the iron addition experiment to derive a conversion factor predicting rates of CO 2 -assimilation along a coastal to open ocean transect in the NE subarctic Pacific (Line-P, https://www. waterproperties.ca/linep/) (Fig 1). All fieldwork for this project was conducted under the authorization and permits of Fisheries and Oceans Canada.

Iron addition experiment
All fieldwork was conducted on board the CCGS John P. Tully in August-September 2013. A 6 day iron addition experiment was initiated at P20 (49°34 N, 138°40 W) (Fig 1), located in iron-limited HNLC waters. Water was collected before dawn from 7 m depth using a trace metal clean pumping system and an on-deck class 100 laminar flow hood (cf. [64]). In order to eliminate macro-zooplankton, the water was pre-filtered through acid washed 200 μm Nitex mesh. Six trace metal-cleaned 10 L cubitainers were rinsed and filled in random order. Triplicate iron-addition treatments were amended with 1 nM Fe (ammonium iron (II) sulfate hexahydrate ((NH 4 ) 2 Fe(SO 4 ) 2 Á6H 2 O), dissolved in 0.05M HCl), while triplicate controls were left unamended. Cubitainers were kept in on-deck incubators continuously supplied with seawater pumped from 5 m depth. Light intensity was adjusted to~50% of full sunlight with neutral density screening and irradiance was continuously logged using a LI-1000 radiation sensor (LI-COR, USA), located 2 m above the incubator. This level of light reduction was chosen to avoid exposing the phytoplankton to irradiances higher than in situ values. On days 1, 3 and 5 at exactly 2 hours after local sunrise, 500 mL of water were sub-sampled from each cubitainer using trace metal clean techniques. Sub-samples were analyzed for total chlorophyll a concentration ([chl a]), photophysiological parameters and rate measurements as outlined below. On the last day of the experiment, additional samples were collected for pigment analysis by high pressure liquid chromatography (HPLC), and the determination of absorption spectra using the quantitative filter technique (QFT) [65].

Laboratory culturing
The oceanic centric diatom Thalassiosira oceanica (CCMP isolate 1003, Sargasso Sea) and the oceanic prymnesiophyte Chrysochromulina polylepis (NEPCC isolate 242, NE subarctic Pacific) were grown under iron-replete and iron-limiting conditions. We chose these two species as representative eukaryotic open ocean species, common in the region of our field study [66,67]. Iron-limited growth conditions were chosen to achieve an approx. 50% reduction in growth rate. Both species were cultured in 28 mL acid-cleaned polycarbonate tubes using the artificial seawater medium AQUIL [68], prepared as described by Maldonado et al. [69]. All cultures were kept at 19°C in continuous, sub-saturating light (ca. 40 μmol quanta m -2 s -1 ). Growth was monitored by daily measurements of in vivo chl a fluorescence using a Turner 10-AU Fluorometer, and cultures were kept in exponential growth phase using semi-continuous batch culturing [70]. Cultures were considered acclimated when growth rates during ca. 40 cell divisions (five successive transfers), varied by <15% [70]. Acclimated, exponentially growing cells were used to inoculate triplicate 200 mL cultures. These 200 mL cultures were subsampled several times for FRRF measurements (see below), which demonstrated that cells maintained steady-state photophysiology throughout the sampling phase. During early to midexponential phase, each replicate culture was sampled for duplicate ETR RCII -PvsE curves, duplicate 14

Station sampling
In addition to the iron addition experiment, seawater samples were collected at five hydrographic stations (P4, P12, P16, P20, and P26) spanning a coastal to open ocean transect in the NE subarctic Pacific (Fig 1). Collection of water column hydrographic profiles was performed with a CTD (SeaBird Electronics, model 911 plus) equipped with a dissolved oxygen sensor (SBE 43), fluorometer (Seapoint), and an underwater photosynthetically active radiation (PAR) sensor (Biospherical QSP-400). At each of the stations, water was collected from Niskin bottles at three depths exactly two hours after local sunrise and processed immediately for rate measurements, photophysiological parameters, and [chl a] as described below.  For the 3 sets of experiments outlined above, samples for [chl a] were filtered onto pre-combusted 25 mm glass fiber filters (GF/F) using low vacuum pressure (<5 mm Hg) and analyzed following the method of Welschmeyer [71]. In the field, triplicate 100-300 mL samples were filtered and stored at -20°C until analysis within three weeks of collection. In the laboratory, triplicate culture samples (10 mL, 20 mL and 30 mL) were collected and analyzed immediately. Each sample was analyzed in duplicate.

Carbon assimilation
For both laboratory and field work, rates of carbon assimilation were measured as small volume PvsE curves in a custom built photosynthetron [72]. In the field, 300 mL of water were spiked with 150 μCi NaH 14 CO 3 (final concentration 0.5 μCi mL -1 , 52.5 mCi mL -1 specific activity) (Perkin-Elmer) immediately after sampling. Spiked samples were mixed gently but thoroughly, aliquoted into 20 ml glass scintillation vials and placed into the photosynthetron. Temperature was kept within 1°C of in situ temperature by circulating water from a water-bath through an aluminum cooling jacket (the offset from in situ temperature was larger for station samples because samples from different depth had to be incubated simultaneously). Light was provided by high power light emitting diodes (LEDs) located under each scintillation vial. Each PvsE curve consisted of 11 light levels spanning intensities from 3 to 600 μmol quanta m -2 s -1 . Actual light intensities were measured before and after each experiment using a 4π quantum sensor (QSL-2100, Biospherical Instruments) immersed in water inside a scintillation vial. Incubations lasted for 3-4 hours and were ended by gentle filtration onto 25 mm GF/F filters. Filters were pre-combusted to reduce nominal pore size to approx. 0.4 μm. For each curve, three time-zero samples were taken by filtering 20 mL immediately after spiking. The total 14 C activity added was determined from three 1 mL aliquots of the spiked sample added to 1 mL 1 M NaOH. All work was done under low light and filters were stored in scintillation vials at -20°C until processing within 1 month of the experiment. During laboratory processing, 500 μL of 3 M HCl was added to each filter and vials were left to degas for >24 hours to eliminate any inorganic 14 C remaining in the samples. Ten mL of scintillation cocktail (Scintisafe plus, Fisher) were added to each vial, and vials were then vortexed and left to stand in the dark for >12 hours before analysis on a liquid scintillation counter (Beckman). Disintegrations per minute (DPM) were derived from scintillation counts using a quench curve prepared from commercial 14 C standards (Perkin-Elmer). DPM were converted to units of carbon biomass following Knap et al. [73]. The 14 C protocol used for laboratory cultures was the same as outlined above with the following exceptions. We spiked 80 mL of exponentially growing culture with 40 μCi NaH 14 CO 3 and 3 mL aliquots were incubated in the photosynthetron for 30 minutes. Duplicate curves were measured for each sample. The incubation was terminated by adding 1 mL of 1 M HCl to each vial and samples were dried completely, omitting the filtration step. After drying, salts were re-suspended in 1 mL MilliQ water. For both laboratory and field measurements 14 C-PvsE curves were fit following Webb et al. [74], as described below.

Chl a fluorescence parameters and ETR RCII
A bench-top FRRF instrument (Soliense Inc.) was used for all active ChlF measurements. In the field, opaque bottles were used for sub-sampling from the rosette or iron addition experiment, and light in the laboratory was kept low at all times to allow oxidation of the ETC and relaxation of NPQ. For all measurements, background fluorescence blanks were prepared by very gently filtering a small amount of sample through a pre-combusted GF/F. Single turnover (ST) flash protocols consisted of 100 flashlets with 1.0 μs length and 2.5 μs interval (46200 μmol quanta m -2 s -1 peak power intensity, resulting in a ST flash length of 250 μs, pro-viding~5-10 quanta per RCII). The excitation power was selected at the beginning of the cruise to saturate the observed fluorescence transients within the first half of the ST excitation protocol. Our experience indicates that this approach offers the best signal-to-noise ratio in the recovered parameters, while accommodating significant variations in the photosynthetic properties of the local phytoplankton populations along the cruise track, without re-adjusting of the excitation protocol. Excitation power was provided by an array of eight LEDs at four wavelengths centered on 445 nm, 470 nm, 505 nm, and 530 nm (equal intensity at each wavelength; see S1 Fig for more information on the spectral distribution). We measured steady state light curves (SSLC), where each sample was exposed to 10 actinic 'background' irradiances ranging from 0 to 1000 μmol quanta m -2 s -1 , also provided at four wavelengths (S1 Fig). The relatively long duration of the SSLCs in this study could create some potential for the settling of cells which could influence the ChlF yield. However, our sampling region is known to be dominated by small cells [66], which should have a slow settling rate. Equally, the laboratory isolates used during this study stay in suspension for many hours, and it is thus unlikely that rapid settling of cells drastically altered our results.
All ChlF yields and parameters described below were derived by an iterative non-linear fitting procedure, applying the four parameter biophysical model of Kolber et al. [21] to a mean of 20 consecutive ST flashes using custom software (Z. Kolber). This software accounts for a formation of fluorescence quenching, most likely due to formation of a P680 triplet, which reduces the maximum fluorescence yield attainable during the ST flash by 3-6%. Throughout the SSLC, ST flashes were applied continuously (at 1 s interval), while the length of each light step was optimized to allow all derived parameters to reach steady state (ca. 5 min). ChlF yields and parameters corresponding to each light level were obtained from the mean of the last three acquisitions at each light level. In this way, we derived the fluorescence yields F o and F m (in dark regulated state) as well as F 0 and F m 0 (in the light regulated state for each light level of the SSLC  [76]) and varying stress levels experienced by natural phytoplankton assemblages. Five fluorescence signals, F o , F m , F 0 , F m 0 and F o 0 were used to calculate ChlF parameters, following Roháček [77]. In the dark-regulated state, we derived the commonly used F v /F m ratio as [78]. For each light level of the SSLC protocol we have calculated the following ChlF parameters: (1) The photochemical quenching of variable fluorescence, , which quantifies the fraction of functional RCII in the open state (i.e. primary quinone acceptor Q A in the oxidized state) [79]. (2) The maximum quantum yield of PSII photochemistry, which can be used to quantify the extent to which photochemistry in PSII is limited by competition with thermal decay of excitation energy [75]. (3) The overall quantum efficiency of photochemical energy conversion in PSII at a given light intensity (note that numerous definitions for this parameter exist in the literature), ). Furthermore, the functional absorption cross section of PSII, σ PSII (Å 2 RCII -1 ), was derived from the rate of closure of RCII in the dark-regulated and at each light-regulated state [20,21]. The connectivity parameter, ρ, was also calculated, but not used in our analysis. Rates of charge separation (i.e. ETR RCII ) in functional RCII (mol emol RCII -1 s -1 ) were estimated as the product of incident irradiance (E), the fraction of irradiance absorbed by PSII (σ PSII ) and the efficiency with which charge separation occurs in RCII. We calculated ETR RCII as where E (μmol quanta m -2 s -1 ) is the actinic irradiance at each light level, σ PSII 0 (Å 2 RCII -1 ) is the functional absorption cross section at E and F q 0 /F v 0 is the photochemical capacity of PSII at E. The number 6.022 x 10 −3 converts μmol quanta to quanta and Å 2 to m 2 . Because of potential systematic errors in the calculation of F o 0 , we also calculated ETR RCII as which does not require the knowledge of F o 0 . Both calculations are equivalent, assuming that non-photochemical quenching processes affecting ChlF can be adequately accounted for in either the absorption term (Eq 1) and the efficiency term (Eq 2). While Eq 2 does not require F o 0 (which was not measured directly) or σ PSII 0 (which is difficult to derive at high irradiances), it does rely on parameters measured in the fully dark-regulated state, which can be difficult to achieve in field assemblages. For all ETR RCII calculated during our iron addition experiment (n = 345) the difference between values calculated in both ways ranged from 0.5 to 21% with a mean coefficient of variance of 5.5%. Both approaches thus provided similar results in the analysis of our data, and the differences observed were not systematically related to the treatment (high vs low Fe). Non-photochemical quenching (NPQ) at each light level was estimated as the normalized Stern-Volmer quenching coefficient, defined as NPQ NSV . Quantification of NPQ using NPQ NSV instead of the more commonly used Stern-Volmer coefficient of quenching, defined as NPQ SV = (F m -F m 0 )/F m [80], is appropriate for our data-set, as it resolves differences between NPQ present in the dark-regulated state.

PvsE curves
Measurements of CO 2 -assimilation and ETR RCII were plotted against irradiance, and the exponential model of Webb et al. [74] was fit to the data using a non-linear least squares regression procedure in Matlab. For the CO 2 -assimilation data, an intercept parameter was added to force the regression through the origin and provide a good fit in the linear part of the PvsE curve [28,81]. For both rates of productivity, we derived the light saturated maximum rate P max and the light utilization efficiency α. When photoinhibition was observed at high irradiances, the data-points were excluded from the fitting procedure.

Derivation of conversion factor
Because we derived ETR RCII in units of mol emol RCII -1 s -1 and CO 2 -assimilation in units of mol C mol chl a -1 s -1 , the conversion factor between the two rates accounts for changes in chl a functionally associated with each RCII (1/n PSII , mol chl a mol RCII -1 ) and the number of charge separations in RCII needed per CO 2 -assimilated into organic carbon products (F e:C , mol emol C -1 ).
In this approach, we attribute the observed decoupling between ETR RCII and CO 2 -assimilation to changes in both, 1/n PSII and F e:C . We recognize that combining F e:C and 1/n PSII into one conversion factor obscures the mechanistic underlying of the observed decoupling. Nevertheless, as we will show, our approach has the potential to provide FRRF-derived estimates of phytoplankton primary productivity in carbon units without the need for many of the auxiliary measurements and inherent assumptions used in previous studies.
The value of 1/n PSII is known to change significantly as a function of taxonomy [22], light [22,82], macro-nutrients [83], and iron availability [84][85][86][87][88][89][90]. Therefore we could not assume a constant value for 1/n PSII , as has been done in most previous studies [24]. Although 1/n PSII can be directly measured from oxygen flash yield experiments (e.g. [91][92][93]), the approach is labourintensive and not practical for routine field sampling. A new approach to derive [RCII] directly from FRRF measurements has been developed [94,95], but not implemented in our study because the inherent assumption that the ratio of rate constants of photochemistry and fluorescence (k p /k f ) is confined to a narrow range, does not hold under varying levels of iron limitation [62,87,94].
Having established a relationship between light intensity and rates of CO 2 -assimilation and ETR RCII for each sample, we were able to model the light dependency of the conversion factor F e:C /n PSII . This approach allowed us to observe how the coupling of ETR RCII and CO 2 -assimilation is modulated by incident irradiance, and how, in turn, iron limitation influences the light-dependent response. Additionally, we used α and P max of each rate to derive the conversion factor under sub-saturating and saturating light conditions, respectively.

Results
Effect of iron addition on phytoplankton community composition, photophysiology, ETR RCII and CO 2 -assimilation in the NE subarctic Pacific Phytoplankton assemblages at station P20 in the NE subarctic Pacific (Fig 1) responded strongly to iron addition in a ship-board incubation experiment (Fig 2). Six days after iron addition, [chl a] increased by an order of magnitude, whereas the control (i.e. no iron addition) showed only a small increase in [chl a]. This result confirms that the initial phytoplankton assemblage was iron-limited (Fig 2A), and that we were able to carry out the manipulation experiment without significant contamination of the control bottles. The slight increase in [chl a] in the control treatments is likely attributable to a decrease in grazing pressure and to changes in the light environment (i.e. lower and less fluctuating light). Iron addition also significantly affected phytoplankton photophysiology, as demonstrated by rapid changes in the parameters σ PSII and F v /F m derived in the dark-regulated state (Fig 2B and 2C). F v /F m initially increased in both treatments, but then remained low in the control while continuing to increase in the iron addition treatment (Fig 2B). While the functional absorption cross-section of PSII, σ PSII (Å 2 RCII -1 ), remained high and relatively constant in the iron-limited control, it declined rapidly after iron addition, and remained~25% lower than that of the initial phytoplankton assemblage (Fig 2C). The observed changes in F v /F m and σ PSII may have resulted from both, photophysiological responses and from changes in species composition. CHEMTAX analysis of pigments sampled on day 6 of the experiment showed that the addition of iron changed the taxonomic composition of the phytoplankton assemblage (S2 Fig). Most prominently, the abundance of chlorophytes decreased from 7% to 1%, prymnesiophytes decreased from 55% to 22%, pelagophytes increased from 17% to 39%, and diatoms increased from 1% to 16% in iron amended bottles. A similar response has been observed in previous iron addition experiments conducted in this region [96].
We measured PvsE curves of short-term CO 2 -assimilation and ETR RCII five times during the iron addition experiment (Fig 3). Both rates show the expected light dependency, and were affected by iron addition. However, the response to iron addition differed for CO 2 -assimilation and ETR RCII . Chlorophyll a-normalized CO 2 -assimilation showed a small, though not statistically significant, increase after iron addition (Fig 3A-3E). The observed increase in the chl anormalized rate was small, because cellular chl a content increased in parallel with CO 2 -assimilation (under all nutrient limitations, cellular chl a in phytoplankton is drastically reduced, a condition referred to as chlorosis, e.g. [97]). The strong effect of iron addition on CO 2 -assimilation can be seen more clearly when rates are normalized to volume. Indeed, volume-normalized CO 2 -assimilation rates increased more than 8-fold after iron addition in this experiment (S3 Fig). In contrast to rates of CO 2 -assimilation, ETR RCII decreased significantly after iron addition, when compared to the iron-limited control treatment (Fig 3F-3J).
The response of CO 2 -assimilation and ETR RCII to iron addition is further visualized in Fig 4, which shows changes in light-limited slopes (α) and light saturated rates (P max ), as well as the derived conversion factor F e:C /n PSII for α and P max , throughout the experiment. Values for α and P max were derived from the 14   Iron Limitation Decouples Phytoplankton ETR_RCII from 14C-Uptake statistically significant change in values of α could be determined for either chl a-normalized CO 2 -assimilation, ETR RCII or F e:C /n PSII (p-value > 0.05). Similarly, the P max for chl a-normalized CO 2 -assimilation remained relatively constant in the control, and did not show a statistically significant increase after iron addition (p-value > 0.05) (Fig 4D). In contrast, there was a significant (p-value < 0.05) decrease in P max for ETR RCII following iron-addition, as compared to the control treatments, which exhibited a small increase in this variable over the course of Response of rates of CO 2 -assimilation (mol C mol chl a -1 s -1 ) and ETR RCII (mol emol RCII -1 s -1 ) during the iron addition experiment. Both rates were measured as a function of irradiance, and PvsE curves were fit with the exponential model of Webb et al. [74]. Shown are mean values from three biological replicates where error bars represent standard error of mean and are sometimes smaller than symbols. Iron Limitation Decouples Phytoplankton ETR_RCII from 14C-Uptake the experiment (Fig 4E). The observed changes in the P max for CO 2 -assimilation and ETR RCII , resulted in a decrease in F e:C /n PSII in the iron addition treatment compared to the relatively constant value observed in the iron-limited control (Fig 4F). This difference was statistically significant for the last 2 days of the experiment (p-value < 0.05). When compared to the initial value on day 0 of the incubation, the conversion factor F e:C /n PSII for P max decreased by 66% after iron addition, and by 16% in the iron-limited control (Fig 4F). These results indicate that Iron Limitation Decouples Phytoplankton ETR_RCII from 14C-Uptake the iron-dependent changes in F e:C /n PSII are most readily apparent under high irradiance conditions where photosynthesis is light-saturated.
To better explain the iron-dependent decrease in ETR RCII and F e:C /n PSII observed in our data, we examined changes in additional FRRF-derived ChlF parameters, measured on day 3 after iron addition. We choose day 3 for the in-depth analysis of our data, but trends observed on this day were representative of those observed throughout the experiment. The parameter F q 0 /F v 0 represents the efficiency of charge separation in functional RCII (Fig 5A). It is an estimate of the fraction of open RCII (i.e. Q A oxidized) at any given light level, and therefore always equals one at zero irradiance. On day 3 after iron addition, we observed higher F q 0 /F v 0 for the iron-limited control at all irradiance levels (Fig 5A), indicating a greater fraction of open reaction centers. The parameter F v 0 /F m 0 , the efficiency of excitation energy capture by the  [19], can be used to quantify the extent to which photochemistry in RCII is limited by thermal energy dissipation in the antenna [75]. This parameter was significantly reduced in the iron-limited control relative to the iron addition treatment (Fig 5B), indicating that the efficiency of excitation energy transfer in the light-harvesting antenna was comprised. The overall efficiency of charge separation per quantum absorbed in PSII (F q 0 /F m 0 ) is the product of F q 0 /F v 0 and F v 0 /F m 0 [19,77]. On day 3, at all light levels, F q 0 /F m 0 was higher in the iron addition treatment than in the iron-limited control (Fig 5C). We used our PvsE measurements of CO 2 -assimilation and ETR RCII to examine the lightdependent response of the conversion factor F e:C /n PSII . Our results (Fig 6) show that F e:C /n PSII increased with increasing irradiance, regardless of iron treatment and day of the experiment (Fig 6A-6E). However, this light-dependent increase was much more pronounced in the ironlimited control treatment. It is important to note that the magnitude and light-dependency of F e:C /n PSII in the iron-limited control treatment changed over the course of the experiment relative to the initial sample (Fig 6A). This shift in F e:C /n PSII in the absence of iron addition likely reflects changes in light quality and quantity in the incubation bottles relative to the ambient water column.
Also shown in Fig 6 is the light and iron dependency of NPQ NSV , estimated as F o 0 /F v 0 . This parameter showed a light and iron-dependent response that was remarkably similar to F e:C / n PSII , with values increasing with increasing light, regardless of treatment and day of the experiment, and decreasing in response to iron addition (Fig 6F-6J). The NPQ NSV values measured in our initial sample ( Fig 6F) were higher than those measured in either control or iron addition treatments during the following days. We attribute this effect to a more stable light environment in the incubation bottles, relative to in situ irradiance levels.
Given the similar light and iron-dependent responses of F e:C /n PSII and NPQ NSV , we sought to examine the relationship between these two variables. In order to do so, however, it was necessary to derive NPQ NSV and F e:C /n PSII values at a standard set of light levels, matching those of the FRRF derived ETR RCII -PvsE curves. For each sample, ETR RCII -PvsE curves consisted of 14 light levels spanning from 0 to 1000 µmol quanta m -2 s -1 . These light levels did not exactly match those used for the CO 2 -assimilation experiments. We thus used the PvsE curve fits of our 14 C data to derive the CO 2 -assimilation values at light levels matching those of the ETR R-CII -PvsE curves. In this way, we were able to compile a dataset of 298 paired values for NPQ NSV and F e:C /n PSII , derived from 27 sets of ETR RCII and 14 C PvsE curves during the iron addition experiment. Plotting these F e:C /n PSII values against the corresponding NPQ NSV reveals a strong and statistically significant correlation (R 2 = 0.70, p-value < 0.0001, for quadratic fit) (Fig 7).

Effects of iron limitation on photophysiology and rates of ETR RCII and CO 2 -assimilation in mono-specific phytoplankton cultures
Using methods analogous to those applied to mixed phytoplankton assemblages in the NE subarctic Pacific; we measured PvsE curves of CO 2 -assimilation and ETR RCII in mono-specific laboratory cultures of two open ocean phytoplankton species. The results, summarized in Table 1, show similar trends as observed in our field data. Steady-state growth rates (μ, d -1 ) in the low iron cultures were 68% and 49% of iron-replete growth rates in T. oceanica and C. polylepis, respectively (Table 1). For both species, F v /F m in iron-limited cultures was reduced (by 32% and 20% in T. oceanica and C. polylepis, respectively). In iron-limited T. oceanica, σ PSII increased by 15%, while it increased by 5% in C. polylepis. The iron dependent changes in μ, F v /F m and σ PSII was statistically significant in both species (one tailed p-value < 0.0001 and < 0.01 for T. oceanica and C. polylepis, respectively). Chlorophyll a-normalized CO 2assimilation at P max remained relatively constant in both species (p-value > 0.05). In contrast, we observed a 90% increase in ETR RCII at P max in T. oceanica under iron-limited growth conditions. C. polylepis also exhibited an increase in ETR RCII at P max under iron-limited conditions, but this increase was not statistically significant (p-value > 0.05). Regardless of species-specific differences, both species showed the same trend of increased F e:C /n PSII and NPQ NSV under iron limitation (Table 1), which is consistent with our field observations. Furthermore, the Units of in Φ e:C /n PSII are (mol e-mol C) / (mol chl a mol RCII -1 ). The curves were derived by dividing corresponding values of ETR RCII and CO 2 -assimilation from the PvsE curves presented in Fig 3. NPQ was estimated as the normalized Stern-Volmer quenching coefficient NPQ NSV = F o 0 /F v 0 and is unitless [65]. Error bars are the standard error from three biological replicates and often smaller than symbols.   Iron Limitation Decouples Phytoplankton ETR_RCII from 14C-Uptake species-specific differences observed in our laboratory experiments are consistent with changes in phytoplankton assemblage composition observed in our iron addition experiment, where the abundance of diatoms (lower F e:C /n PSII ) was increased in the iron addition treatment and the abundance of prymnesiophytes (higher F e:C /n PSII ) was decreased (S2 Fig).
Thalassiosira oceanica and Chrysochromulina polylepis were grown in steady state ironreplete and iron-limited conditions. The mean growth rate μ, derived from successive measurements in semi-continuous batch cultures, is given in d -1 . The error is the SD of 3 biological replicates, and number of consecutive batch transfers (ca. 4 cell divisions per transfer) used to calculate growth rates are given in brackets. F v /F m and σ PSII are values from cultures in the dark regulated state (10 min of 5 μmol quanta m -2 s -1 at 730 nm), measured on the day of CO 2 -assimilation experiments. The error is SD of 3 biological replicates. Changes in these parameters are statistically significant for T. oceanica (p-value < 0.0001) and C. polylepis (p-value < 0.01). P max for CO 2 -assimilation (mol C mol chla -1 s -1 ) and ETR RCII (mol e-mol RCII -1 s -1 ) were derived from PvsE curves as described in the methods section. The error is the 95% confidence interval of the P max derived from the fit to data from 6 whole curve measurements (duplicate curves each from 3 biological replicates). The conversion factor F e:C /n PSII for P max was derived as the quotient of P max for ETR RCII and P max for CO 2 -assimilation. The error is the propagated error from numerator and denominator. NPQ NSV was estimated as

Discussion
Our results provide new insight into the effects of iron and light availability on the coupling between CO 2 -assimilation and photosynthetic electron transport in natural phytoplankton assemblages and mono-specific laboratory cultures. We show that both of these environmental variables significantly influence F e:C /n PSII , which has important implications for the use of FRRF measurements to infer rates of CO 2 -assimilation in oceanic waters. Below, we first discuss the observed increase in F e:C /n PSII under excess light and low iron conditions in the context of previously reported values. We then discuss the effects of iron and light on phytoplankton photophysiology, and suggest a simple conceptual explanation for the observed increase in ETR RCII under iron limitation. We hypothesize, that iron and light-dependent changes in F e:C /n PSII are driven by the need to dissipate excess excitation energy, caused by either excess light, or the effects of iron limitation on the ETC. In this context, we discuss the correlation between F e:C /n PSII and NPQ NSV , and examine the potential significance of this finding in the context of marine primary productivity studies.

Magnitude of the observed conversion factor
The conversion factor F e:C /n PSII , derived from our measurements of ETR RCII and CO 2 -assimilation, varied significantly in response to light and iron availability. In our field experiment, the addition of iron caused the value of F e:C /n PSII at light saturation (P max ) to decrease by 66% within 6 days ( Fig 4F). Furthermore, short-term changes in light availability had a major effect on the value of F e:C /n PSII , and this effect was enhanced under iron limitation. A recent metaanalysis of variability in experimentally determined F e:C from 14 field studies found values ranging from 1.15 to 54.2 with a mean of 10.9 ± 6.91 mol emol C -1 [24]. This analysis comprised a wide range of oceanic regions, but did not include observations from the NE subarctic Pacific or other HNLC regions. Due to our experimental approach, we are unable to derive absolute values for F e:C . However, if we assume 1/n PSII to be 500 mol chla mol RCII -1 [20], as has been done in most previous studies [24], F e:C values on day 3 of the iron-addition experiment range from 13 to 39 mol emol C -1 . Using a constant value of 1/n PSII for both treatments is unlikely to be realistic. Even though iron-limited phytoplankton possess less chl a per cell, 1/n PSII , the ratio of chl a to RCII, has frequently been observed to increase under low iron conditions [84,85,88,90,98]. If we thus assume 700 mol chl a mol RCII -1 for the iron-limited control treatment and 500 mol chl a mol RCII -1 for the iron addition treatment [85], F e:C ranges from 13 to 28 mol emol C -1 . These F e:C values represent the range observed across different irradiance levels in our PvsE experiments. At the time of sampling, cells in the on board incubator were exposed to~40 µmol quanta m -2 s -1 . Assuming 700 and 500 mol chl a mol RCII -1 for the iron-limited and iron-replete treatments, respectively, we derive F e:C values of~18 and 15 mol emol C -1 . Values of F e:C estimated from our data are thus within the range reported in previous field studies [24], with no estimate falling below the theoretical minimum of 4 mol emol C -1 .
Ideally, measurements of ETR RCII and CO 2 -assimilation should be performed simultaneously on the same sample, eliminating differences in incubation time and spectral quality of the light sources used. As discussed in detail in the supplementary material, the differences in spectral distribution of the light sources used for FRRF and 14 C measurements could have led to an underestimation of absolute values of F e:C /n PSII (S1 Fig). However, these differences cannot explain the large iron dependent changes we observed in F e:C /n PSII , since the absorption spectra of iron-limited and iron-enriched treatments did not differ drastically (S1 Fig). Furthermore, differences in incubation times could have influence the absolute magnitude of the derived conversion factor. Incubation times used for the PvsE curves were ca. 5 min for FRRF measurements (applied incrementally to the same sample), vs. 3-4 hours in the field and 30 min in the laboratory for 14 C-assimilation experiments (light levels applied simultaneously to different samples). As has been shown by Halsey et al. [16,17,99] and Pei and Laws [18], the use of fixed incubation times for cells growing at different growth rates could lead to an overestimation of our conversion factor F e:C /n PSII in the iron-limited relative to iron-replete samples. Additionally, the longer incubation time in CO 2 -assimilation experiments might have exacerbated cumulative processes such as photodamage under excess irradiance. To address this issue, we did not utilize the part of the PvsE curves showing photo-inhibition. However, we cannot rule out any differential cumulative effects of photoinhibition on ETR RCII and 14 Cassimilation at P max . This could potentially decrease CO 2 -assimlation at P max relative to ETR R-CII at P max and lead to overestimation of our F e:C /n PSII values at P max . Notwithstanding these potential sources of uncertainty in the absolute value of F e:C /n PSII , the good agreement between our estimated F e:C (assuming~500-700 mol chl a mol RCII -1 ) and those of previous studies suggests that our observations are robust. More importantly, potential offsets in the absolute values of F e:C /n PSII do not diminish the significance of the relative, iron and light-dependent changes we observed in this parameter (discussed below).
Interacting effects of iron and light on the conversion factor Φ e:C /n PSII Our data show strong and interacting effects of iron and light availability on the conversion factor F e:C /n PSII in phytoplankton field assemblages and mono-specific laboratory cultures (Fig 4C, 4f and 6, Table 1). It has been shown that the magnitude of both 1/n PSII and F e:C vary significantly between phytoplankton taxa (e.g. [22,93]). Changes in F e:C /n PSII in field experiments was thus likely influenced by both, physiological changes and taxonomic shifts. These two sources of variability are, to a large extent, intrinsically linked, since changes in phytoplankton community composition (S2 Fig) reflect the selection of better adapted species under any particular set of environmental conditions (i.e. iron limitation). In the following, we discuss the observed changes in F e:C /n PSII from a predominantly photophysiological point of view, since our laboratory results specifically demonstrate such physiological effects.
Numerous metabolic processes, acting between ETR RCII and CO 2 -assimilation can act to increase F e:C , and therefore the conversion factor F e:C /n PSII (e.g. [61,59,100]). In addition to its role in reducing CO 2 to organic carbon products, reductant (NADPH) formed at the end of the ETC can also be used for nitrate and sulphate reduction [101], photorespiration [102], or respiration via the malate shunt [103]. These alternative pathways decouple ETR RCII from CO 2assimilation, increasing the value of F e:C . Similarly, before the formation of NADPH, pseudocyclic electron flow can reduce O 2 and create a water-water cycle of electron transport, also increasing F e:C (e.g. [104]). Pseudo-cyclic electron transport pathways can divert electrons from the ETC before (short water-water cycling, e.g. [105]) or after PSI (Mehler-reaction, e.g. [106]). Cyclic electron transport (CET) around PSII [107,108] and charge recombination in PSII [109,110], act more closely to the initial charge separation in RCII, and can also cause an increase in F e:C .
We suggest that the higher F e:C /n PSII observed in response to iron limitation and shortterm increases in incident irradiance during the PvsE experiments (Fig 6A-6E) results predominantly from increases in the alternative electron flow pathways prior to reductant formation. These pathways, which are diagramed conceptually in Fig 8, can act as 'safety valves' to keep the primary quinone acceptor Q A oxidized when excitation pressure on the ETC is high, thereby decreasing the potential of damage to RCII [111,54,55,61,56,104,112,60].
Iron limitation directly affects the photosynthetic ETC and thereby modulates the lightdependent changes in the conversion factor F e:C /n PSII (Fig 6A-6E). Importantly, iron limitation has been shown to alter the stoichiometry of ETC components (i.e. expression of iron-rich PSI and cytochrome b 6 f complexes is down-regulated to a higher extent than PSII) (e.g. [62,84,[116][117][118]). Low levels of electron acceptors downstream of PSII ultimately restrict the flow of electrons away from PSII during light exposure. This exacerbates the need for short (i.e. acting before PSI) alternative electron flow pathways to dissipate excess excitation energy and prevent over-reduction of RCII (Fig 8). A number of recent studies have suggested that re-routing electrons to a midstream plastoquinol oxidase (PTOX) to bypass the electron flow bottleneck of PSI is a common strategy in open ocean phytoplankton [113,55,54,[56][57][58]62]. Importantly, up-regulation of pseudo-cyclic electron flow under iron limitation not only protects RCIIs from photodamage, but also helps to maintain a high ΔpH across the thylakoid membrane, providing energy for cell maintenance and growth [62,119]. Cyclic electron flow around PSII [107,108,114,61,120] and increases in charge recombination at PSII [109,110,115] are two additional mechanisms that can act to prevent over-reduction and damage of RCII when excitation pressure is high and the electron flow bottleneck is prior to PSI. Unlike PTOXmediated water-water cycling, these processes do not contribute to an increase in ΔpH across the thylakoid membrane. They would, however, contribute to a high ETR RCII and therefore F e:C /n PSII (Fig 8) .
While ambient light intensity has a well-documented effect on values of 1/n PSII , these changes act on timescales longer than those of short-term PvsE experiments, and are thus unlikely to have caused the light-dependent changes we observed in F e:C /n PSII (Fig 6). On longer time-scales, however, iron limitation causes a reduction of chl a per cell (chlorosis), and an increase in chl a per functional RCII (1/n PSII ) [87,90]. This well documented response, which has been attributed to preferential down-regulation of RCII [87], and up-regulation of ironstress-induced light harvesting complexes (isiLHCs) [62,90], would act to further increase F e:C /n PSII under iron limitation, regardless of light intensity (Fig 6A-6E). However, rates of linear electron transport (LET) and CO 2 -assimilation saturate at a light intensity determined by the physiological state of the phytoplankton, resulting in a typical PvsE curve. Under optimal growth conditions, it is the resupply of NADP -(predominantly from CO 2 -assimilation) which limits LET, while under short-term exposure to excess light and under iron limitation, the 'bottleneck' of LET will be located before PSI. Whenever exitonic influx exceeds the chemical outflux at the level of RCII, excess excitation pressure needs to be safely dissipated to prevent photodamage. (B) Under optimal growth conditions and sub-saturating light, all absorbed photons are used for charge separation in RCII, and the majority of electrons will be used for LET and CO 2 -assimilation, resulting in minimum Φ e:C . (C) Conditions of high excitation pressure can be caused by short-term exposure to high light, but also by iron limitation, which In summary, we suggest that it is the effect of high excitation pressure, which causes a decoupling of ETR RCII and CO 2 -assimilation. This high excitation pressure may be a result of short-term exposure to excess irradiance as well as the effect of iron limitation on the ETC. This purely photophysiological interpretation can be extended to observations made in mixed phytoplankton communities. Here, fluctuating light and low iron conditions will select for species with the best ability to control high excitation pressure by adjusting the flow of excitation energy into, and the flow of electrons out of PSII.

Iron limitation increases ETR RCII
To our knowledge, this is the first study which shows that ETR RCII increases under iron limitation. This observation may seem counter-intuitive, and it is important to emphasize that our results do not imply an overall increase in photosynthetic electron transport under low iron conditions. Rather, our observations point to an increase in the rate of charge separation at each individual RCII, independent of the reduced total cellular concentration of these RCII. We show that the overall efficiency of PSII photochemistry in the light-regulated state, F q 0 /F m 0 (= Ф PSII 0 ), is reduced under iron limitation (Fig 5C), as expected. However, deconvolution of this parameter into its constituents F q 0 /F v 0 ( Fig 5A) and F v 0 /F m 0 ( Fig 5B) shows that F q 0 /F v 0 , representing the fraction of open RCII (Q A oxidized) at each given light level, increased under iron limitation. We hypothesize that this is likely achieved by increased alternative electron transport pathways acting to keep RCIIs open (Q A oxidized) and bypassing the electron flow bottleneck at PSI, when excitation pressure is high (Fig 8). In contrast to F q 0 /F v 0 , the parameter F v 0 /F m 0 is much lower when iron is limiting (Fig 5B), indicating that the excitation energy transfer in the antennae is compromised. Based on our experimental observations, we suggest a simple mechanistic explanation for the observed increase in ETR RCII under iron limitation. Cellular iron demand can be significantly reduced by economizing on iron-rich components of the photosynthetic apparatus and 'funneling' more electrons down fewer RCIIs (i.e., increasing ETR RCII ). In line with this explanation is the observation that values of σ PSII are high under iron limitation, and rapidly decrease after iron addition (Fig 2) [121,122,84,85,[123][124][125][126]87]. Strzepek et al. [127] suggested that increased σ PSII compensates for fewer iron-rich photosynthetic reaction centers in Southern Ocean phytoplankton species. Similarly, Ryan-Keogh et al. [128] noted that increasing the absorption cross section of RCs by the expression of isiLHCs allows cells to reduce the cellular iron requirement while maintaining the same light absorption capacity.
In conclusion, our results and interpretation support a scenario where photosynthetic electron flow has been fine-tuned to maximize energy conversion as well as photo-protection under conditions where ETC component abundance and stoichiometry are compromised by the availability of iron.

Link to NPQ NSV
Above, we discussed how mechanisms acting down-stream of the initial charge separation in RCII are likely to be enhanced under conditions of high excitation pressure, resulting in high comprises the functioning of the ETC and has been shown to create a 'bottle neck' for LET before PSI. Under these conditions, PTOX-mediated pseudocyclic electron flow (e.g. [54][55][56][57][58]62,105,113]), cyclic electron transport around PSII (e.g. [107,108,114]), and charge recombination in RCII (e.g. [109,110,115]), have been suggested to safely dissipate excess excitation energy after RCII (but before PSI). Up-regulation of these alternative electron flow pathways could explain the high ETR RCII (and Φ e:C /n PSII ) observed in our iron-limited samples. Excess excitation energy can also be dissipated in the light harvesting antenna, before charge separation in RCII. Collectively, a number of different molecular processes dissipating excess excitation energy in the PSII antenna can be quantified as NPQ NSV . doi:10.1371/journal.pone.0133235.g008 Iron Limitation Decouples Phytoplankton ETR_RCII from 14C-Uptake ETR RCII and F e:C /n PSII . High excitation pressure can also be dissipated in the pigment antenna, before reaching RCII [104]. Fig 8 shows schematically the 'safety mechanisms' used for the dissipation of excess energy at both sides of RCII. Because processes dissipating excess excitation pressure in the antenna also quench ChlF yields measured by FRRF, they have collectively been called non-photochemical quenching (NPQ). NPQ, which is present in all oxygenic photosynthetic organisms, encompasses a wide variety of mechanisms acting to dissipate absorbed light energy as heat before it reaches RCII [129][130][131][132][133][134]. Following the approach of McKew et al. [80], we estimated NPQ from FRRF measurements as so-called normalized Stern-Volmer quenching (NPQ NSV ). We observed a strong correlation between the conversion factor F e:C /n PSII and the expression of NPQ NSV (Fig 7). We note that F e:C /n PSII and NPQ NSV are not entirely independent parameters, and therefore the strong correlation observed in Fig 7 is in part a result of their co-dependence on the ChlF parameter F v 0 (which we used in the derivation of both NPQ NSV and F e:C /n PSII ). At this point, the relationship between Ф e:C /n PSII and NPQ NSV shown in Fig 7 is empirical rather than mechanistic. However, while there are a number of processes which will influence Ф e:C /n PSII and NPQ NSV differentially, there are many processes related to the amount of excitation pressure experienced by the ETC that would influence both in a consistent manner. Numerous studies have shown that Ф e:C increases if light is saturating, i.e. when excitation pressure is high (e.g. [33,36,40]). Clearly, excess light would also increase the expression of NPQ NSV . Indeed, very recent work has pointed to a mechanistic link between alternative electron sinks involving PTOX and the expression of NPQ NSV [105].
A possible approach towards improved prediction of CO 2 -assimilation from FRRF data While it remains to be seen how strong the correlation between F e:C /n PSII and NPQ NSV (Fig 7) may be for other datasets, our results provide a potential basis for improved estimates of CO 2assimilation from FRRF measurements alone. A number of factors make this approach more desirable than the use of static, regional conversion factors. First, the magnitude of F e:C /n PSII in phytoplankton assemblages will be determined by a multitude of interacting environmental variables. The use of NPQ NSV as an integrated physiological measure of environmental effects on electron transport processes will therefore help to constrain the relationship between F e:C / n PSII and various environmental stressors. Secondly, as our data show, the magnitude of F e:C / n PSII can vary significantly within the same sample in response to short-term variations in incident light. Such small scale changes would be lost using a static (regional) conversion factor, but are captured with our NPQ NSV -based approach, as every single ETR RCII estimate is paired with a corresponding NPQ NSV estimate. Finally, a non-static conversion factor is crucial if the goal is to monitor the effects of environmental change on marine primary productivity, since physiological responses to environmental change will likely affect the conversion factor itself before productivity changes are observed.
As a test of the validity of our approach, we used the F e:C /n PSII vs. NPQ NSV correlation determined from our iron addition experiment (Fig 7) to predict the CO 2 -assimilation rates from FRRF-derived ETR RCII and NPQ NSV measured along the Line-P transect. In this case, in situ phytoplankton assemblages were collected from within and below the mixed layer, and rate measurements were conducted immediately after collection, without any experimental manipulation (see methods). As shown in Fig 9, we obtained a strong correlation between the predicted and measured CO 2 -assimilation rates (Spearman' s r = 0.90, n = 95 and two-tailed pvalue < 0.0001 on non log-transformed data). Our approach consistently underestimates values from the deepest sampling depth, which can likely be attributed to the lack of spectral correction of our data. The RMSE for the values predicted using our approach and measured values is 48.4 mol C mol chl a -1 hr -1 . This error represents~10% of the total range of values observed along the transect during this study, suggesting that rates of productivity can be predicted with reasonable accuracy. In comparison with our approach, computation of CO 2assimilationfrom FRRF data assuming a constant 1/n PSII value of 500 mol chl a mol RCII -1 and 4 mol emol C -1 , significantly under-predicts observed CO 2 -assimilation rates (RMSE = 837.3 mol C mol chl a -1 hr -1 ). Even if we use a constant conversion factor derived from the average of the F e:C /n PSII measured during our iron addition experiment, the model error remains larger compared to that derived using our variable, NPQ NSV -based conversion factor (Fig 7). Our data therefore show significant potential in the application of a variable, NPQ NSV -derived conversion factor and associated quantification of carbon uptake rates from FRRF data.

Conclusion
Deriving rates of phytoplankton CO 2 -assimilation from bio-optical approaches like FRRF has the potential to provide estimates of primary production at unprecedented spatial and temporal resolution. High resolution measurements, covering large oceanic regions, are essential for Rates of CO 2 -assimilation (mol C mol chl a -1 hr -1 ) derived from FRRF measurements plotted against rates measured by 14 C-assimilation experiments. Samples were taken at one to three depths at five stations along Line-P in the NE subarctic Pacific (see Fig 1). FRRF based PvsE curves were used to derive ETR RCII and NPQ NSV at 8 light levels for each sample, and Φ e:C /n PSII values were then derived from the relationship presented in Fig 7. Φ e:C / n PSII and ETR RCII for each light level were used to calculate CO 2 -assimilation rates. Error bars for predicted CO 2 -assimilation rates represent the propagated error from the ChlF yields of the last three ST acquisitions of each light level during the FRRF PvsE curve used to derive NPQ NSV and ETR RCII . Error bars for measured CO 2 -assimilation rates represent the mean coefficient of variance derived from all duplicate measurements (n = 46). The correlation between all predicted and measured data points (n = 95) was statistically significant (Spearman's r = 0.90, two-tailed p-value < 0.0001). All statistics are for non logtransformed data. doi:10.1371/journal.pone.0133235.g009 Iron Limitation Decouples Phytoplankton ETR_RCII from 14C-Uptake the monitoring and modelling of marine food webs and global biogeochemical cycles. Further, such measurements are indispensable for the development and validation of algorithms estimating global marine primary productivity from remote sensing.
Crucial to this approach is a sound characterization of the conversion factor between FRRFderived ETR RCII and primary productivity in carbon units. Our data demonstrate that the conversion factor varies significantly in response to iron and light availability in phytoplankton field assemblages and mono-specific laboratory cultures. We interpret the observed variability in the conversion factor F e:C /n PSII as a manifestation of the extreme photophysiological flexibility which evolved in phytoplankton to maximize growth under dynamic light and nutrient regimes [135,136]. We hypothesize that, to a large extent, changes in F e:C /n PSII represent a suite of coordinated photophysiological adjustments acting to balance light absorption with CO 2 -assimilation under given environmental conditions. These will be manifested on the physiological as well as on the taxonomic level. On the taxonomic level, a low nutrient and / or fluctuating light environment will select for species with the best ability to control high excitation pressure by adjusting the flow of excitation energy into, and the flow of electrons out of PSII (manifested in changes of NPQ NSV , 1/n PSII and F e:C ). Future studies will be needed to evaluate the relationship between NPQ NSV and F e:C /n PSII in a number of oceanic regions in order to evaluate the potential for improved CO 2 -assimilation estimates from FRRF data.
Supporting Information S1 Fig. Spectral distribution of light sources used for FRRF and Photosynthetron assays, and absorption spectra of phytoplankton assemblages on day 6 of the iron-addition experiment. (a) The FRRF instrument used during this study contains LEDs with peak output at four wavelengths (445 nm, 470 nm, 503 nm, 530 nm). In our FRRF instrument, excitation as well as actinic background irradiance is applied from the same LEDs. (b) Spectral distribution of the LEDs used in the photosynthetron used for 14 C-uptake experiments. (c) Spectral overlap of the two light sources. The overlap is good in the region of maximal light absorption by photosynthetic pigment (ca. 450 nm). However, in direct comparison with the photosynthetron, the FRRF instrument provides a higher proportion of photons in the region > 480 nm. This could have led to an underestimation of ETR RCII values relative to CO 2 -assimilation values measured in the photosynthetron, resulting in an under-estimate of F e:C /n PSII . In addition to knowledge of spectral differences in the light sources used (a-c), spectral correction of our data would require light absorption spectra of the phytoplankton assemblages examined. Relative absorption spectra of the phytoplankton communities on day 6 after iron-addition (measured using the quantitative filter technique [65]) are shown in (d-f). Spectra from 3 biological replicates of the control (d) and two biological replicates of the iron addition treatment (e) were averaged, and these spectra are shown together in panel (f). The results show relatively small changes in the relative light absorption between the two treatments, and it is unlikely that these changes would have significantly influenced the large iron and light-dependent effects in F e:C /n PSII . Because we did not measure absorption spectra for all sampling points of the iron addition experiment and stations along the transect, we were unable to spectrally correct our data. Furthermore, because we are not deriving absolute values for F e:C /n PSII , we did not apply a constant correction factor (estimated from e.g. the data shown in a-f). The taxonomic composition of phytoplankton assemblages (% of total chl a) was derived from HPLC analysis of accessory photosynthetic pigment. Average values are shown from three biological replicates for the iron-limited control and the iron addition treatment on day 6 of the experiment. One to1.5 L of water were filtered on 25 mm GF/F and stored at -80°C until analysis. Pigments were extracted and quantified as described by Taylor et al. [137]. Pigment ratios were then used to estimate phytoplankton assemblage composition using CHEMTAX as described by Taylor et al. [137]. The initial pigment ratio matrix used for our data was taken from Lee et al. [138], table 5, which is specific to North Pacific phytoplankton isolates. (TIFF) S3 Fig. Response of volume normalized rates of CO 2 -assimilation (mol C m -3 hr -1 ) during the iron addition experiment. The rates were measured as a function of irradiance, and PvsE curves were fit with the exponential model of Webb et al. [74]. Shown are mean values from three biological replicates where error bars represent standard error of mean and are sometimes smaller than symbols. Results shown in this figure confirm a strong stimulatory effect of iron additions on primary productivity in the experimental bottles. (TIFF)