Improving the accuracy of single turnover active fluorometry (STAF) for the estimation of phytoplankton primary productivity (PhytoPP)

Photosystem II (PSII) photochemistry is the ultimate source of reducing power for phytoplankton primary productivity (PhytoPP). Single turnover active chlorophyll fluorometry (STAF) provides a non-intrusive method that has the potential to measure PhytoPP on much wider spatiotemporal scales than is possible with more direct methods such as 14C fixation and O2 evolved through water oxidation. Application of a STAF-derived absorption coefficient for PSII light-harvesting (aLHII) provides a method for estimating PSII photochemical flux on a unit volume basis (JVPII). Within this study, we assess potential errors in the calculation of JVPII arising from sources other than photochemically active PSII complexes (baseline fluorescence) and the package effect. Although our data show that such errors can be significant, we identify fluorescence-based correction procedures that can be used to minimize their impact. For baseline fluorescence, the correction incorporates an assumed consensus PSII photochemical efficiency for dark-adapted material. The error generated by the package effect can be minimized through the ratio of variable fluorescence measured within narrow wavebands centered at 730 nm, where the re-absorption of PSII fluorescence emission is minimal, and at 680 nm, where re-absorption of PSII fluorescence emission is maximal. We conclude that, with incorporation of these corrective steps, STAF can provide a reliable estimate of JVPII and, if used in conjunction with simultaneous satellite measurements of ocean color, could take us significantly closer to achieving the objective of obtaining reliable autonomous estimates of PhytoPP.


1
Introduction 31 Phytoplankton contribute approximately half the photosynthesis on the planet (Field, 1998), thus 32 forming the base of marine food webs. Reliable assessment of Phytoplankton Primary Productivity 33 (PhytoPP) is crucial to an understanding of the global carbon and oxygen cycles and oceanic 34 ecosystem function. Consequently, PhytoPP has been recognized as an Essential Ocean Variable 35 (EOV) within the Global Ocean Observing System (GOOS). PhytoPP is a dynamic biological 36 process that responds to variability in multiple environmental drivers including light, temperature and 37 nutrients across spatial scales from meters to ocean basins, and time scales from minutes to tens of 38 years. This poses significant challenges for measuring and monitoring PhytoPP. 39 Historically, the most frequently employed method for assessing PhytoPP has been the fixation of 40 14 C within closed systems over several hours of incubation (Marra, 2002;Milligan et al. 2015). 41 Despite the widespread use of the 14 C method, which has led to measurements of PhytoPP by the 14 C 42 method providing the database against which remote sensing estimates of primary production are 43 calibrated (Bouman et al. 2018), there is considerable uncertainty in what exactly the 14 C method 44 measures and the accuracy of bottle-incubation based methods for obtaining PhytoPP in oligotrophic 45 ocean waters (Quay et al. 2010). 46 According to Marra (2002), the 14 C technique measures somewhere between net and gross carbon 47 fixation, depending on the length of the incubation. In this context, net carbon fixation is defined as 48 gross carbon fixation minus carbon respiratory losses and light-dependent losses due to 49 photorespiration and light-enhanced mitochondrial respiration (Milligan et al. 2015). Although it may 50 seem intuitive that short incubations should provide a good estimate of gross carbon fixation (and 51 closely match PhytoPP), several authors have reported that short-term 14 C fixation does not reliably 52 measure net or gross production (e.g. Halsey et al. 2013; Milligan et al. 2015). It should also be noted 53 that short-term, in the context of 14 C fixation, is several hours incubation. This clearly imposes major 54 limitations on the spatiotemporal scales at which PhytoPP can be assessed using this method. 55 Gross photosynthesis by phytoplankton is defined here as the rate at which reducing power is 56 generated by photosystem II (PSII) through the conversion of absorbed light energy (PSII 57 photochemistry). Within this study, gross photosynthesis is quantified by measuring the rate at which 58 O2 is evolved through water oxidation by PSII photochemistry (Ferron et al. 2016) and is termed 59 PhytoGO. Although measurement of O2 evolution provides some advantages over 14 C fixation, in 60 that both gross and net primary production can be obtained, the spatiotemporal limitations are 61 similar. 62 It is now widely accepted that active fluorometry can provide a non-intrusive method for measuring 63 PSII photochemistry on much wider spatiotemporal scales than either 14 C fixation or O2 evolution. 64 Within oceanic systems, where optically thin conditions are the norm, the most appropriate form of 65 active fluorometry is the single turnover method (Kolber and Falkowski 1993 under actinic light. It follows that JVPII can, in principle, provide a proxy for PhytoPP (Oxborough et 84 al. 2012). 85 An important caveat to using JVPII as a proxy for PhytoPP is that there are a number of processes 86 operating within phytoplankton that can uncouple PhytoPP from PhytoGO and PhytoGO from PSII 87 photochemistry (Geider and  4. The proportion of these complexes in the open state has routinely been estimated through the qP 142 parameter (Kolber et al. 1998) which is mathematically equivalent to the photochemical factor 143 (Fq'/Fv') defined by Baker and Oxborough (2004). This requires determination of Fo', using the 144 equation provided by Oxborough and Baker (1997) or through direct measurement after 1 -2 s dark-145 adaptation following a STAF measurement under actinic light (Kolber et al. 1998

Phytoplankton cultures (N-limited experiments) 206
Semi-continuous phytoplankton cultures were maintained and adapted to nutrient-replete conditions. 207 All cultures were grown in f/2 medium with silicates omitted where appropriate (Guillard, 1975). 208 The experimental work covered a period of several months. The initial work was conducted at the 209 University of Essex and incorporated six phytoplankton species (Table 1) The N-limited cultures were sub-cultured from the nutrient-replete cultures. High light-grown 222 cultures were used for the six species interrogated at the University of Essex. In all cases, the growth 223 photon irradiance of the original culture was maintained after sub-culturing. All N-limited cultures 224 were grown into the stationary growth phase using N-limiting f/2 medium before experimental 225 measurements were made. 226

Phytoplankton cultures (package effect experiments) 227
All package effect experiments were conducted at CTG Ltd. Cultures were maintained as 30 mL 228 aliquots within filter-capped tissue culture flasks (Fisher Scientific, UK: 12034917). A growth 229 temperature of 20 °C was maintained by placing the flasks within a water bath (Grant SUB Aqua Pro 230 2 L, USA). Low light illumination (photon irradiance of 30 µmol photons m -2 s -1 ) was provided from 231 white LED arrays (Optoelectronic Manufacturing Corporation Ltd. 1ft T5 Daylight, UK). The L:D 232 cycle was set at 12 h:12 h. 233

Setup for OLCs and flash O2 measurements 234
All OLCs and flash O2 measurements were made using an Oxygraph Plus system (Hansatech 235 Instruments Ltd, Norfolk, UK). The sample volume was always 1.5 mL and a sample temperature of 236 20 °C was maintained using a circulating water bath connected to the water jacket of the DW1 237 electrode chamber. The sample was mixed continuously using a magnetic flea (as supplied with the 238 Oxygraph Plus system). Illumination was provided from an Act2 laboratory system (CTG Ltd, as 239 before). The source comprised three blue Act2 LED units incorporated within an Act2 Oxygraph 240 head. Automated control of continuous illumination during OLCs or the delivery of saturating pulses 241 during flash O2 measurements was provided by an Act2 controller and the supplied Act2Run 242 software package. 243

Dilution of samples between flash O2 and STAF measurements 244
The N-limited and dual waveband experiments included determination of sample-specific Ka values. 245 In all cases, the required dark STAF measurements of Fo and PII were made after the flash O2 246 measurements. In all cases, filtered medium was used to dilute the sample between Oxygraph and 247 STAF measurements. 248

Chlorophyll a extraction 249
In all cases, the concentrated sample used for flash O2 or OLCs was normalized to the parallel dilute 250 STAF sample used to generate Fo and PII or FLC data through direct measurement of chlorophyll a 251 concentration from both samples. 252 Chlorophyll was quantified by pipetting 0.5 mL of each sample into 4.5 mL of 90% acetone and 253 extracting overnight in a freezer at -20 °C (Welschmeyer, 1994). Samples were re-suspended and 254 centrifuged at approximately 12,000 x g for 10 minutes and left in the dark (~ 30 minutes) to 255 equilibrate to ambient temperature. Raw fluorescence from a 2 mL aliquot was measured using a 256 Trilogy laboratory fluorometer (Turner, UK). The chlorophyll a concentration was then calculated 257 from a standard curve. 258

Setup for dark STAF measurements and FLCs (N-limited experiments) 259
All STAF measurements for the N-limited experiments were made using a FastOcean sensor in 260 combination with an Act2 laboratory add-on (CTG Ltd, as before). The Act2 FLC head was 261 populated with blue LEDs. A water bath was used as a source for the FLC head water jacket, 262 maintaining the sample temperature at 20 °C. 263

Flash O2 measurements for determining sample-specific Ka values 264
The density of photochemically active PSII complexes within each sample was determined using the The concentration of photochemically active PSII centers is proportional to the product of gross O2 269 evolution rates (E0) and the reciprocal of flash frequency (Hz). The basic theoretical assumptions are 270 that all photochemically active PSII centers undergo stable charge separation once during each flash, 271 that all photochemically active PSII centers re-open before the next flash and that four stable charge 272 separation events are required for each O2 released. In reality, small errors are introduced because 273 some centers do not undergo stable charge separation with each flash (misses) while some centers 274 will undergo more than one stable charge separation event with each flash (multiple hits). 275 The following checks were applied with all samples: • The proportion of PSII centers closed during each flash was verified by comparison with 277 sequences of 120 µs flashes on a 24 ms pitch at a photon irradiance of 13,800 µmol photons 278 m -2 s -1 279 • The default flash pitch of 24 ms was compared against 16 ms and 36 ms to assess the 280 accumulation of closed PSII centers, with 120 µs flashes of 22,000 µmol photons m -2 s -1 281 being applied in all three cases 282 • Sequences of 180 and 240 µs flashes on a 24 ms pitch at a photon irradiance of 22,000 µmol 283 photons m -2 s -1 were applied to assess multiple hits 284 In all cases, a flash duration of 120 µs duration at a photon irradiance of 22,000 µmol photons m -2 s -1 285 on a 24 ms pitch provided more than 96% saturation, with no evidence of a significant level of 286 multiple hits or the accumulation of closed PSII centers. 287

Parallel OLC and FLC measurements (N-limited experiments) 288
A series of parallel replicate OLC/FLC measurements were made on all nutrient-replete cultures, as 289 well as for the N-limited T. weissflogii culture ( sequences on a 100 ms pitch. The auto-LED and auto-PMT functions incorporated within the 296 Act2Run software were always active. 297 The reported gross O2 evolution rates (E0) were taken as the sum of measured net O2 evolution (Pn) 298 and Rd (Equation 9). 299 E 0 = P n + R d Equation 9 300

OLC and FLC curve fits (N-limited experiments) 301
OLCs and FLCs are variants of the widely used P-E (photosynthesisphoton irradiance) curve. For 302 OLCs, the metric for photosynthesis is the rate at which O2 is evolved through water oxidation by 303 PSII. For FLCs, the metric for photosynthesis is the relative rate of PSII photochemistry, which is 304 assessed as the product of PII and E. In the absence of baseline fluorescence (when Fb = 0), the 305 parameter Fq'/Fm' can be used to provide an estimate of PII. It follows that FLC curves can be 306 generated by plotting E against the product of baseline-corrected Fq'/Fm' (Fq'/Fmc') and E. 307 There are three basic parameters derived from all P-E curve fits: , Ek and Pm. The value of  308 provides the initial slope of the relationship between E and P. Ek is an inflection point along the P-E 309 curve which is often described as the light saturation parameter (Platt and Gallegos, 1980). Pm is the 310 maximum rate of photosynthesis. 311 The FLC curve fits within this study were generated by the Act2Run software (CTG Ltd, as before). 312 The curve fitting routine within Act2Run is a two-step process which takes advantage of the fact that 313 the signal to noise within FLC data is highest during the initial part of the FLC curve. Kromkamp, 2012). The overall fit is an iterative process that minimizes the sum of squares of the 316 difference between observed and fit values. During the Alpha fit, a significant weighting on the initial 317 points (low actinic E values) is generated by multiplying each square of the difference by (Fq'/Fmc')2. 318 This approach normally generates a good fit up to Ek, but overshoots beyond this point. 319 Consequently, the Pm values generated by the Alpha phase are generally too high. 320 In the second step (the Beta phase) Equation 11 is used to improve the value of Pm. This step includes 322 a second exponential which is only applied to data points at E values above the Ek value generated by 323 the Alpha phase. The sum of squares of the difference between observed and fit values is not 324 weighted during the Beta phase. This approach forces PII at Ek to be 63.2% of . 325

Terminology 386
A structured approach has been taken in derivation of the parameters used within this manuscript. As 387 baseline fluorescence is central to this study, new fluorescence terms to describe baseline-corrected 388 values of existing fluorescence terms have been introduced. Otherwise, the parameters are structured 389 around root terms that are widely used within the fluorescence community. 390 Table T1 provides terms used to describe the fluorescence signal at any point. Table T2 provides  391 commonly used parameters derived from the terms in Table T1. Tables T3, T4 and T5 show the  392  derivation of terms used for the yields, rate constants, absorption cross sections and absorption  393 coefficients applied to PSII energy conversion processes. The remaining terms used are covered 394 within Table T6. 395 The root terms and subscripts provided in Tables T3 and T4 Table 1). In all cases, the N-limited sample-specific Ka values are 409 significantly lower than for the nutrient-replete samples they were sub-cultured from (Figure 2A Where Fmc is the Fb-corrected value of the measured Fm (see Terminology). When using this 422 equation, Fm and Fv are measured from the sample and Fv/Fmc is an assumed baseline corrected value 423 of Fv/Fm for the photochemically active PSII complexes within the sample (see Figure 1).  and B, respectively). The slope for the sample-specific Ka data (D), at 0.778, is significantly lower 473 than the ideal of 1.0. This lower slope may be at least partly due to differences in the curve fits 474 applied to OLC and FLC data (see Methods). 475

The stability of Fb under actinic light 476
Clearly, the consensus Fv/Fmc (0.518) in Equation 14 generated a good match between Ka values for 477 all but one of the nutrient-replete and N-limited cultures in Figure 2A.  Figure 5A shows the maximum PhytoGO values (PhytoGOm), measured as O2-evolution (x-axis) or 494 calculated using the sample-specific Ka value from the nutrient-replete T. weissflogii of 15,868 m -1 . 495 For these values, Fb was set to zero for both the nutrient-replete cultures and the N-limited cultures. 496 Clearly, while there is good agreement between the measured and calculated values of PhytoGOm 497 from the nutrient-replete cultures, most of the calculated PhytoGOm values from the N-limited 498 cultures are much higher than the measured values. 499 For Figure 5B, Equation 14 was used to generate a consensus Fv/Fmc specific to the N-limited 500 cultures. This consensus value was reached by minimizing the sum-of-squares for the regression line 501 through the N-limited data by allowing Fb to vary. The mean consensus Fv/Fmc from this fit (0.502) is 502 within 3% of the consensus value derived from the dark-adapted data presented in Figure 2. In 503 contrast, the average NPQ-dependent decrease from dark-adapted Fm to the light-adapted Fm' 504 measured at Pm was always more than 30% (data not shown). Consequently, these data do not imply 505 significant quenching of Fb between the dark-adapted state and Pm. 506

Dual waveband STAF measurements to correct for the package effect 507
We hypothesized that the variance of sample-specific Ka values within Figure 2A could be at least 508 partly due to variable package effect. As prevoiously noted within Materials and methods, three 509 FastBallast units (B730, B680 and B682) were used to measure fluorescence centered at 730 nm and 510 680 nm (both 10 nm FWHM) and 682 nm (30 nm FWHM), respectively. 511 To test the viability of a STAF-based approach to quantifying the package effect, we generated ratios 512 of the Fv measured by B730 as a proportion of the Fv measured by B680 (Fv 730/680 ) or B683 (Fv 730/683 ). 513 Within Figure 6, these values are plotted against sample-specific values of KR ( Figure 6A and D, 514 respectively). The Fv ratios from Figure 6A and D were used to generate Fv-derived values of KR 515 ( Figure 6B and E, respectively). 516 Calculated K R = as appropriate. The calculated KR values within C and F were generated by combining the + BB3 520 Fv 730/680 and Fv 730/683 data with the Slope and Intercept from A and D, as appropriate. 521 One feature that is immediately clear from these data is the much tighter grouping of points along the 522 regression lines for the Fv 730/680 data (A to C) than the Fv 730/683 data (D to F). This indicates that the 30 523 nm FWHM of the 682 nm bandpass filter is too broad to adequately isolate the fluorescence 524 generated close the 680 nm absorption peak and, consequently, that the 10 nm FWHM 680 nm 525 bandpass filter is the better choice for these measurements. 526 All 11 species used within the package effect tests were grown under nutrient-replete conditions and 527 exhibited Fv/Fm values that were above the consensus value of 0.518 generated from the first part of 528 this study. The addition of BB3 to each sample within the package effect tests was to simulate the 529 lower Fv/Fm values that are frequently observed under conditions of stress. The expectation was that 530 fluorescence from the added BB3 would increase Fb but have minimal impact Fv and, as a 531 consequence, that the slope of the relationship between calculated and measured KR values would not 532 be significantly affected by a BB3-dependent increase in Fb. The absence of significant changes in 533 slope between B and C and E and F are consistent with this expectation. PhytoGO and PhytoPP on much wider spatiotemporal scales than O2 evolution or 14 C fixation, 537 respectively, through determination of JVPII. This study was undertaken to assess the extent to which 538 baseline fluorescence and the package effect could introduce errors into the calculation of JVPII 539 (Equation 5). 540 With regard to baseline fluorescence (Fb), the underlying question was whether sub-maximal dark-541 adapted value of Fv/Fm could be attributed to Fb or downregulation of PSII photochemistry by dark-542 persistent Stern-Volmer quenching or some combination of the two. The data presented within Figure  543 2A provides strong evidence that, for the examples presented within this study, Fb is by far the 544 dominant contributor to sub-maximal Fv/Fm values. Although this interpretation may not hold for all 545 phytoplankton species and environmental conditions, this study provides a straightforward, practical 546 approach to addressing the question of how universally valid an Fb correction to low sub-maximal 547 dark-adapted Fv/Fm values might be. 548 We conclude that no correction for baseline fluorescence should be applied when the dark-adapted 549 Fv/Fm is above a certain consensus value. In situations where the dark-adapted Fv/Fm is below this 550 consensus value, Equation 14 should be used to calculate a value for Fb. From the data presented 551 here, a consensus value (Fv/Fmc) of between 0.50 and 0.52 seems an appropriate default value. 552 Clearly, the value of Fb generated by Equation 14 is dependent on a STAF measurement made on a 553 dark-adapted sample. The data presented in Figure 5 indicate that, for this specific example at least, 554 the dark-adapted Fb could be applied at the other end of the FLC scale, to correct the value of Pm. 555 With regard to the package effect, the wide range of Ka values within Figure 2A is entirely consistent 556 with a significant proportion of the fluorescence emitted from functional PSII complexes being 557 reabsorbed through this process. This interpretation is clearly supported by data presented in Figure  558 3, where use of the sample-specific Ka value in place of Ka FO provides a much stronger match 559 between the FLC and OLC data. The dual waveband data presented in Figure 6 provide strong 560 evidence that the package effect-induced error could be decreased significantly through incorporation 561 of a Fv 730/680 -derived correction factor applied to a default instrument-type specific Ka value such as 562 Ka FO . From a practical point of view, routine implementation of this correction step will require either 563 two detectors with different filters or a single detector with switchable filtering. On balance, the latter 564 option is likely to prove more cost-effective and easier to calibrate. 565 Overall, the conclusions reached can be summarized by Equation 16 566 where Ka TS is the instrument type-specific Ka value and RPE is a dimensionless sample-specific 568 correction factor. All other terms are as before or are defined in Terminology. 569 For the species and conditions examined in this paper, the data presented provide strong evidence 570 that baseline correction and package effect correction can increase the accuracy of estimates of 571 PhytoGO from STAF. We anticipate that development and deployment of STAF instrumentation that 572 will allow Equation 16 to be applied will take us significantly closer to achieving the objective of 573 obtaining reliable autonomous estimates of PhytoGO. Such measurements, if used in conjunction 574 with simultaneous satellite measurements of ocean color, will likely lead to improved estimates of 575 local, regional or global pelagic PhytoPP. 576

Acknowledgements 577
The authors wish to thank Tania Cresswell-Maynard, James Fox and Philipp Siegel (University of 578 Essex, UK) for supplying the phytoplankton cultures. We would also like to thank Hoi Ga Chan 579 (University of Princeton, US) and Mark Moore and Anna Hickman (Southampton University, UK) 580 for helpful discussions. RJG acknowledges support from NERC grant NE/P002374/1. 581

Author contributions statement 582
KO conceived of the study and developed the software required to conduct the experiments. The 583 initial baseline experiments were designed by KO, RG and TB. All of the baseline experiments were 584 conducted by TB who also processed the primary data. The dual waveband experiments for 585 assessment of the package effect were conceived of by KO and conducted by TB. Package effect data 586 were processed by TB and jointly analyzed by KO and TB. Figure 1 was produced by KO, all 587 remaining Figures were produced by TB. The initial draft of the main text was produced by KO. 588 Iterations of the manuscript were implemented by KO, TB and RG. The submitted version of the 589 manuscript is approved for publication by KO, TB and RG. 590

Conflict of interest statement 591
The authors (KO, TB and RG) declare that the research was conducted in the absence of any 592 commercial or financial relationships that could be construed as a potential conflict of interest. 593

Contribution to the field statement 594
Phytoplankton photosynthesis is responsible for approximately half of the carbon fixed on the planet. 595 As a process, photosynthesis is responsive to variability in multiple environmental drivers including 596 light, temperature and nutrients across spatial scales from meters to ocean basins, and time scales 597 from minutes to tens of years. This poses significant challenges for measurement and monitoring. 598 While direct measurement of the carbon fixed by photosynthesis can only be applied on very limited 599 spatial and temporal scales, active chlorophyll fluorescence has enormous potential for the accurate 600 measurement of phytoplankton photochemistry, which provides the reducing power for carbon 601 fixation, on much wider spatiotemporal scales and with much lower operational costs. This study 602 identifies practical measures that can be taken to improve the accuracy of such measurements. We 603 are confident that these measures will have minimal impact on the frequency at which phytoplankton 604 photochemistry is assessed and that they will be suitable for application on autonomous measurement 605 platforms. 606 Webb WL, Newton M, Starr D (1974