Use of active fluorescence to estimate phytoplankton photosynthesis in situ

Transcription

1 Limnol. Oceunogr.. 38(8), 1993, , by the American Society of Limnology and Oceanography, Inc. Use of active fluorescence to estimate phytoplankton photosynthesis in situ Zbigniew Kolber and Paul G. Falkowski Oceanographic and Atmospheric Sciences Division, Brookhaven National Laboratory, Upton, New York Abstract We describe the theory and practice of estimating photosynthetic rates from light-stimulated changes in the quantum yield of chlorophyll fluorescence. By means of a pump-and-probe fluorescence technique, where weak probe flashes are used to measure the change in the quantum yield of fluorescence induced by the strong pump flash, it is possible to derive the absolute absorption cross sections for photosystem 2, the quantum yield for photochemistry, and the maximum rate of photosynthetic electron transport at light saturation. In conjunction with a semiempirical biophysical model of photosynthesis, these parameters can bc used to calculate the instantaneous rate of gross photosynthesis in situ under ambient irradiance. A profiling pump-and-probe fluorometer was constructed and interfaced with a CTD, and vertical profiles of variable fluorescence were obtained on four cruises in the northwest Atlantic Ocean. The derived photosynthetic rates were compared with concurrent estimates of production based on radiocarbon uptake. The correlation coefficient between the two estimates of primary production, normalized to Chl a, was 0.86; linear regression analysis yielded a slope of There is a 3-4-fold range in the maximum change in the quantum yields of photochemistry and absorption cross-sections in natural phytoplankton communities. Uncertainties in the pump-and-probe-derived estimates of photosynthesis arc primarily due to temporal mismatches between instantaneous and time-integrated measures of production and in biological variability in the ratio of the number of PS2 reaction centers to total Chl a. Almost all measurements of phytoplankton photosynthesis in situ are based on the timedependent incorporation of radiocarbon into particulate matter or on changes in concentration of dissolved oxygen in the bulk fluid. Among other things, uncertainty in these measurements arises from artifacts associated with isolating natural phytoplankton assemblages in bottles (e.g. Eppley 1980), differentiating between net and gross photosynthesis (Bender et al. 1987), light shock, and, in the open ocean, trace metal toxicity (Carpenter and Lively 1980; Fitzwater et al. 1982). In situ, fluorescence-based measurements of phytoplankton photosynthesis are attractive because they potentially overcome all of these problems. Fluorescence measurements can be made rapidly, conveniently, and continuously without an incubation, thereby eliminating bottle effects. The measurements reflect gross photosynthetic rates and, because they are practically instantaneous, light shock and trace metal toxicity are Acknowledgments This research was supported by NASA grant UPN 16 l and DOE contract DE-AC02-76CHOOO 16. We thank Marcel Babin, Richard Geider, Bernard Genty, Dale Kiefer, Stephen Long, Miguel Olaizola, Tom Owens, Andre Morel, Kevin Wyman, and Crcighton Wirick for comments and discussions negligible (Falkowski and Kiefer 1985; Topliss and Platt 1986; Kiefer et al. 1989). However, interpretation and use of fluorescence signals to derive photosynthetic rates are not straightforward (Falkowski and Kiefer 1985; Kiefer and Reynolds 1992; Owens 199 1). Here we examine the theory and practice of estimating photosynthetic rates with an active fluorescence technique. Active fluorescence methods use an artificial light source (as opposed to natural solar radiation) to stimulate chlorophyll (Chl) fluorescence. The specific technique we describe is based on pump-and-probe fluorometry (Mauzerall 1972; Falkowski et al. 1986; Kolber et al. 1990), in which a controlled change in the quantum yield of fluorescence, measured with a weak probe flash, is induced by an actinic (pump) flash. Modifications of the basic technique and instrumentation potentially allow for extended in situ observations from moorings (e.g. Falkowski et al. 1988a) and remote sensing of phytoplankton photosynthesis from LIDAR systems aboard ships or fixed-wing aircraft. The relationship between fluorescence and photosynthesis is based on biophysical models (e.g. Butler and Kitajima 1975; Weis and Berry 1987; Falkowski et al ; Genty et al. 1989; Kiefer and Reynolds 1992). We have devel-

2 Fluorescence-photosynthesis relationship 1647 oped a model that quantitatively predicts photosynthetic rates from changes in the quantum yields of fluorescence. The model formally describes some of the relevant photosynthetic processes, including light absorption, primary photochemistry, oxygen evolution, and nonphotochemical fluorescence quenching. In this paper our goals are to examine the theoretical biophysical bases for fluorescence-photosynthesis relationships, develop a mathematical model relating photosynthesis to fluorescence over the entire range of photon flux densities experienced by phytoplankton in the ocean, derive model parameters from experimental data, validate the model with independent in situ data, and examine the application of the pump-and-probe methodology in estimating the instantaneous rates of phytoplankton photosynthesis in the ocean. A biophysical model of photosynthesis and fluorescence-photosynthetic O2 evolution can be described by a simplified model as shown in Fig. 1. There are two photoreactions in all oxygenic photoautotrophic organisms; these are designated photosystem 1 (PSl) and photosystem 2 (PS2). In PS2, water is photochemically oxidized, forming O2 while simultaneously generating electrons and protons. In PS 1, light energy is used to further transfer the electrons to terminal acceptors, such as CO, or nitrate. We define photosynthesis as the photochcmical oxidation of water leading to the formation of molecular oxygen and the generation of reductant, i.e. gross photosynthesis. In the steady state, electron flow between water and the terminal acceptors is coupled, and the molar ratio of O2 evolved to CO, assimilated is the photosynthetic quotient. The reaction center in PS2 consists of a primary electron donor, a special Chl a molecule (designated PhsO), and a primary electron acceptor (a quinone designated Qa). In the dark, P680 is reduced while Qa is oxidized, and the reaction center is said to be open. Upon exposure to light, photons may be absorbed by the photosynthetic pigments and the excitation energy transferred to the reaction center, leading to charge separation. During charge separation, PhgO is oxidized to PhgO+, and Q, is reduced to Q,-. Under these conditions, the reaction center cannot utilize a subsequent excitation for photochemical charge separation and is said to be closed. Only after P& is re- reduced and Qa- re-oxidized can the energy of another absorbed photon be used to successfully promote the next photochemical electron transfer. The re-reduction of PhgO+ is very rapid (< 1 ps) and is not rate limiting to the reopening of a PS2 reaction center. Qn- is oxidized by a secondary acceptor, Q,, (a plastoquinone), which, upon reduction, physically dissociates from the reaction center protein complex to become part of the plastoquinone (PQ) pool. The vacated binding site is then occupied by another PQ, allowing subsequent electron transfers from Qa. Typically, the concentration of PQ exceeds that of Q, by a factor of 5-30 (Sukenik et al. 1987; Greene et al. 1992); hence, the PQ pool operates as a capacitor capable of temporarily storing electrons between PS2 and PS 1. Each Qb molecule must accept two electrons (and two protons) to form plastoquinol, PQH2, whereas Q, can accept only one electron at a time (Crofts and Wraight 1983). More importantly, however, the time constant for the oxidation of PQH2, TP, is controlled on the acceptor side of PSI by the dark reactions of photosynthesis (2-15 ms), whereas the time constant for Qa- oxidation, TQ, is about an order of magnitude faster (-0.6 ms) (Crofts and Wraight 1983; Sukenik et al. 1987). At low photon fluxes, the rate of photon absorption by PS2 reaction centers is less than the rate of re-oxidation of PQH2, and electrons are efficiently transferred from Qa- to PQ (Falkowski et al. 1986; Kolber et al. 1988). As irradiance increases, however, the PQ pool becomes progressively reduced, and re-oxidation of Qa- slows due to the paucity of oxidized PQ at any given instant. Under such conditions, the fraction of closed reaction centers increases, leading to a reduction in the quantum efficiency for photosynthesis. At physiological temperatures, fluorescence emanates almost entirely from PS2 (Barber et al. 1989). The theoretical basis for relating fluorescence to photosynthesis postulates that excitation energy delivered to PS2 can either be converted to chemical energy via photochemical charge separation, re-emitted as fluorescence, or dissipated nonradiatively in the process of thermal deactivation (Fig. 1). If the reaction center is open, the probability that excitation energy will be re-emitted as fluorescence is (Qa was originally called Q

3 1648 Kolber and Falkowski 1 Fluorescence Fn.-A\ 'Q \ \ F&2 cycle I A. PS2 reaction center open B. PS2 reaction center closed Fig. 1. Model of photosynthetic O2 evolution. Incident irradiance (E) absorbed by a light-harvesting chlorophyll complex (LHC2) with absorption cross-section u PsZ can migrate by resonance through excited chlorophyll molecules (Chl*), and randomly encounter a reaction center (RC2). If the reaction center is open (A), P,,, may become oxidized to P cm+ 9 and the first stable acceptor, Q., will bc reduced to Q.. Under these conditions, fluorescence is minimal (F,). If, at the instant of the photon absorption, Qa is reduced [i.e. the reaction center is closed (B)], the absorbed energy can be reradiated as fluorescence, increasing the fluorescence yield to F,,. The fluorescence signal Fobserved at irradiance E is an average of F. and F, weighted by the fraction of open and closed reaction centers and is calculated as qr. P,,, + is rereduced through Z by electrons derived from the photochemically driven oxidation of water (S,-S,), leading to the evolution of 0,. Q, is oxidized by a secondary electron acceptor, Q,,, with a time constant of TP. Upon accepting two electrons and two protons from two successive reduction reactions of Q,, Qb dissociates to become a member of the PQ pool. PQH,, is oxidized by the cytochrome b,/f complex and the electrons are subsequently passed through PSl to a terminal acceptor. At light saturation, the maximum overall rate at which electrons can pass from water to the terminal electron acceptors is l/r,,. At very high irradiance levels, when the probability of finding an oxidized PQ is very low, - 15% of the Q; pool may become reoxidized through a cycle around PS2, reducing P,,, + without oxidizing water. This cycle effectively leads to a slight uncoupling bctwecn the change in the quantum yield of fluorescence and that of oxygen evolution (see Fig. 2). for quencher of fluorescence; Duysens and vided). Under ambient light, the fluorescence Sweers 1963). If the reaction center is closed, yield, F, increases as Q, becomes reduced, the probability that the excitation energy will reaching a maximum, F,, when all reaction be re-emitted as fluorescence increases to centers are closed (see Van Kooten and Snel (Schatz et al. 1988). In the dark, when 1990). Variable fluorescence, F,, can be de- Q, is completely oxidized, fluorescence is at a fined as F, = (F,, - FO). Thus, in its simplest minimal level, FO (a list of notation is pro- form, the relative change in the quantum yield

4 Fluorescence-photosynthesis relationship 1649 Fo, Fm f qn hs2 PSU02 v PO2R PfRC2 PC J R a* a E, yo2 Notation In vivo fluorescence yield induced by a weak probe flash in the dark (initial), and following a saturating flash (maximal), measured in dark-adapted state where nonphotochemical quenching is at a minimum (i.e. qn = 0); F, < F, In vivo fluorescence yield induced by a weak probe flash in the dark, under ambient light, and following a saturating hash, all measured in light-adapted state (i.c. qn L 0); Ffo < F < F,,, Variable fluorescence (F, = F,, - Fo; I;, = F,,, - F,); a difference between the max and min fluorescence yield measured in a dark-adapted and light-adapted state Maximum change in the quantum yield of fluorescence (A&, = F,IF,,,; A#, = F,IF,) measured in light-adapted and dark-adapted state, respectively, dimensionless Photochemical quenching, a measure of the fraction of open reaction centers under ambient photon llucnce rates [=(F,, - F )/(F, - F,,)], dimensionless number between 0 and 1 Nonphotochemical quenching, a measure of the contribution of thermal energy dissipation to the quantum yield of in vivo fluorescence [=(F,,, - F,)/(F,, - F,)], dimensionless number between 0 and 1 Fraction of potentially functional PS2 reaction centers in the dark, sometimes called maximum photosynthetic energy conversion efficiency (=A&,/O.65), dimensionless number between 0 and approaching 1 Minimum time required to transfer an electron from water to the terminal electron acceptor in the steady state at light saturation (= 1 IE,gPS2), time, usually ms Minimum average time required for oxidation of Qn-, time, usually -600 ps.. Functional absorption cross-section of PS2, m2 quanta- Quantum yield of photosynthesis, mol 0, (mol quanta)-- Quantum yield of electron transfer with a reaction center (assumed to be unity), electrons quanta - Actual quantum yield of electron transport, accounting for losses due to thermal dissipation and electron cycling around PS2, dimensionless number between 0 and approaching 1 Ratio of PS2 reaction centers to Chl a (=4n,,,), mol electrons (mol Chl a) Photosynthetic unit (see text), mol Chl (mol 0,) Photosynthetic electron flow calculated from fluorescence yields, mol electrons Chl- * time- Steady state oxygen evolution normalized to Chl a, mol0, (mol or g Chl a)- time- Steady state photosynthesis normalized to PS2 reaction centers, electrons (RC2) time Rate of photosynthetic C fixation normalized to Chl, mol C Chl - time- Flash intensity, quanta m-2 Ratio of 0, evolved : CO, fixed, mol mol Spectrally avcragcd in vivo absorption cross-section normalized to Chl a, rn (mg Chl a) Initial slope of P vs. E curve (=a*&,, = u,,~~~~)j, 0, Chl m quanta Intercept corresponding to Po2 Ia, PEinst m z s- Yield of 0, produced by a brief saturating flash, 0, flash- of fluorescence reflects the level of reduction of Q, (Duysens and Sweers 1963; Butler and Kitajima 1975), and the relationship between photochemistry and fluorescence is inverse and controlled by a single process, namely the redox state of Q,. In reality, however, fluorescence rarely follows a simple inverse relationship with photochemistry. Operationally, fluorescence yields are controlled by two basic types of quenching phenomena. These are called photochemical quenching, qf, which is associated with the redox state of Qa, and nonphotochemical quenching, qn (Schreiber et al. 1986; Weis and Berry 1987; Owens 199 l), which may arise from a variety of sources, including transmembrane electrochemical potentials and ph gradients within chloroplasts, the conversion of fluorescence to heat by carotenoids in the xanthophyll cycle (Demmig-Adams 1990), cyclic

5 1650 Kolber and Falkowski electron flow around PS2 (Falkowski et al. 1988b), and the photodegradation of PS2 reaction centers (Horton and Hague 1988; Owens 199 1). Empirical models relating photosynthesis to fluorescence usually attempt to partition fluorescence quenching between q, and qn (Weis and Berry 1987; Genty et al. 1989; Falkowski et al ). We now examine the formal relationship between photosynthesis and fluorescence in the context of a method for deriving the former from the latter. Biophysical description of a photosynthesisirradiance curve-photosynthesis can be expressed as a function of irradiance (E) in a general form: Po2B(E) = fn(ah, Pn, j, E). (1) PozR(E) is the rate of gross photosynthetic oxygen evolution per unit Chl a [mol O2 evolved (g Chl a)- time- 1, a is the initial slope of the photosynthesis-irradiance curve [m2 (g Chl a)- mol O2 (mol photons)- ] (Jassby and Platt 1976), P, is the light-saturated rate, and E is the incident photosynthetically active irradiance (mol photon m-2 time-l). Equation 1 can be rewritten in terms of the optical absorption cross-section (a*) and the quantum yield of photosynthesis ($,,): Po,~(E) = a*$,,(e)e. (2) a* describes the efficiency of light absorption by the ensemble of photosynthetic pigments normalized to Chl a (Dubinsky 1992), is the quantum yield for oxygen evolution at irradiance E. Equation 2 can be further recast in terms of the functional absorption crosssection and number of PS2 reaction centers: Po,~(E) = ~ps2@rc~pwlf nps2e. (3) ups2 is the functional absorption cross-section of Rc is the quantum yield of photochemistry within PS2 (i.e. the value of %,tc is taken to be unity and constant; 1 electron is transferred from PbgO to Q, per quanta absorbed and delivered to the reaction center), nps2 is the ratio of PS2 reaction centers to Chl a, and fis the fraction of PS2 reaction centers that are capable of evolving oxygen. 4dE) can be described by the product of two components which can be measured by fluorescence. These are the photochemical quenching coef- ficient, qp (a parameter which is a measure of the fraction of open PS2 reaction centers at any given instant at irradiance E) and the quantum yield of electron transport through PS2, $, [mol O2 evolved (mol electron) - 1. Thus, Eq. 3 can be further modified to POEM = ~ps2~rcsi~e)~c(e~ nps2e (4) where qp and 4, are irradiance-dependent. Equation 4 forms the basis of the relationship between photosynthesis and fluorescence; its solution requires knowledge of one independent variable, E, and five parameters, cps2, q,,, (6,, J: and nibs2 (Table 1). We now examine the experimental basis for deriving these parameters by active fluorescence. The pump-and-probe technique-a pumpand-probe fluorometer generates a three-flash sequence consisting of a weak probe flash, an actinic pump flash of variable intensity, and a second weak probe flash. The intensity of probe flashes is kept low enough that the excitation energy does not alter the measured fluorescence signals. The pump flash energy can be varied from near zero to a supersaturating level, allowing transient reduction of Q, in a controlled fashion. By adjusting the intensity of the pump flash or the time delay between the pump flash and second probe flash, the fluorometer can measure gps2 (photochemical quenching) and qr, (the kinetics of electron flow from PS2 to PSl) and can provide an assessment of number of functional PS2 reaction centers (Falkowski et al. 1986; Kolber et al. 1988). Measurement of bply2 -The functional absorption cross-section of PS2, cps2, is the product of the light-harvesting capability of the light-harvesting pigments and the efficiency of excitation transfer to the reaction center (Ley and Mauzerall 1982; Mauzerall and Greenbaum 1989; Dubinsky 1992). cps2 can be measured by the pump-and-probe method by gradually increasing the intensity of the pump flash and following the flash-intensity saturation curve of variable fluorescence. This curve can bc fit by a cumulative one-hit Poisson function (Ley and Mauzerall 1982; Falkowski et al. 1986), where (F - F,)I(F, - FO) = 1 - cxp( Y~>~~J). (5) FO is the fluorescence yield preceding the pump

6 Fluorescence-photosynthesis relationship 1651 Table 1. Example of source data and calculated photosynthetic rates with pump-and-probe fluorescence parameters applied to Eq. 4. Cp, is calculated assuming T,, = 4.3 ms; nras2 = ff,,,,(a&,,/o.64); iips2 is assumed to be (1 electron per 500 Chl a molecules); P,,,.ps2 = E (~~~~4~45~ (from PS2 to PSl); P,:c.ps2 = P,,c ps2 (3,600/4) [rate of carbon fixation in CO1 molecules per PS2 RC, calculated assuming that four electrons in PS2 are required to fix one molecule to carbon (i.e. R = l)]; Pl = P,,,. s2m,.im, I, u (Chl a-specific rates of carbon fixation calculated from fluorescence signals, MC. and J4.1,1 (, are molecular masses of C and Chl Chl a I,,,, (clcctrons I, t I > 2 (ml E* (A* quanta ) 4,. 9, s (pg lilcx ) (clcctrons Chl ) R( s ) (C RC 11 ) P, t p, I, , , , , , , * heinst m 2 s-l. tmgc(mgch1) h I.,,I \,I flash of energy, J, and F is the fluorescence yield immediately (l-l 00 hs) following the pump flash (note, FO I F I F,). Simultaneous measurements of bps2 from the flash intensity saturation curve of O2 and the change in the quantum yield of fluorescence are virtually identical (Falkowski et al. 1988c), supporting the assertion that the photochemical process responsible for variable fluorescence is identical to the O,-evolving process and that the quantum yield for photochemical charge separation, arc, is unity. Estimates of qp -The photochemical quenching coefficient, q,, can be defined as the probability that a reaction center will bc capable of charge separation at a given instant. Given a large population of reaction centers, this probability is equal to the fraction of open reaction centers, A (Genty et al. 1989; Owens 199 1). qrj can be calculated from the change in fluorescence yields of probe flashes preceding or following a saturating pump flash at any ambient irradiance level; thus, qp = (F,, - F )l(f, - F,). (6) F and F, are the fluorescence yields induced by a weak probe flash preceding and immediately following the pump flash, respectively, measured under ambient light, and FIO is the fluorescence yield induced by a weak probe flash in a situation when all PS2 reaction centers are open (i.e. following l-2 s of dark adaptation). The rationale for this measurement is as follows: after a short period of dark adaptation (l-2 s), all reaction centers become open and the variable fluorescence, Ftv (=F, - Fto), is maximal. Under an ambient continuous light, however, photons from the continuous light source (e.g. the sun) may have closed some reaction centers prior to the first probe flash. In such a case, FtO will rise to a level F, and the remaining change of fluorescence yield as induced by a saturating flash (i.e. F, - F ) will be smaller. The ratio (F,, - F ) : (F, - F,) will describe the fraction of reaction centers still capable of photochemical charge separation. If nonphotochemical quenching proportionally affects both FtO and F, (see below), Eq. 6 will be valid independent of the level of qn* What is the experimental evidence that qp really is a measure of the fraction of open PS2 reaction centers? With a bare platinum oxygen rate electrode, it is possible to measure the oxygen evolved from a continuous background light; simultaneously, with a pump-and-probe technique, it is possible to measure both qp and the 0, evolved from the same (saturating) pump flash that was used for the fluorescence measurements. The O2 produced by a flash must come from the photochemical oxidation of water in PS2 and can be produced only if PS2 reaction centers are open (Falkowski et al. 1988b). The O2 produced by a saturating flash ( YmaX) is a quantitative measure of functional, open PS2 reaction centers. Under very low irradiances (i.e. < 1 PEinst m-2 s-l), Y,,, is less than qp. The loss in photochemical efficiency occurs at very low irradiance levels because the rate of photon absorption is so low that electrons can leak out of the oxygen-evolving complex, leading to an overall loss of 02-

7 1652 Kolbcr and Falkowski b 1.0 -:: 0.8 E s 0.6 -i Om4 E 0.2 k a a 0.6 o l 1 o Background Irradiance (peinst m -2 s -1 ;" at nm Y/Ymax Fig. 2. [a.] The effect of continuous background irradiance on I; (0) and F,, (V), q,, (0), 0, flash yields, Y/Y,,,,, (III), and steady state oxygen evolution PIP,,,, (0) in Chaetoceros gracilis. Experimental details are described by Falkowski et al. (1988b). The changes in the yields preceding (V) and 70 ps after (V) a saturating plump flash are shown as a function of continuous background irradiance, As continuous background irradiance is increased from darkness to a very low level of light there is a slight decrease in q,, that is not accompanied by measurable oxygen evolution; this is attributed to establishment of electrochemical and ph gradients across the thylakoid membranes (Horton and Hague 1988). As light intensity is increased, F and F, reach a baseline level and q,, stabilizes. This region of the F vs. E: curve corresponds to (Y in the P vs. E curve. Further increases in background light lead to large increases in F and a sharp decrease in q,,, reflecting the closure of reaction centers. The inflection corresponds to EL. As background light is increased beyond this level, both F and F,, decrease and q,, reaches an asymptotic value approaching -0.18, while Y/Y,,,,, approaches zero. The effect of the background light on the evolution of oxygen in the steady state is shown as P/P,,,,. [b.] The relationship between the change in the q,, and Y/Y,,,,. The different symbols arc representative of different experiments. At very high irradiance levels q,, slightly overestimates the fraction of open PS2 reaction centers. flash yields (Ley and Mauzerall 1982). As the continuous irradiance level is increased above this threshold, q, and Y,,,, follow each other very closely, and the change in the quantum yield of fluorescence (as determined by qp) quantitatively follows the fraction of oxygen evolving PS2 reaction centers over most of the range of irradiance levels (Falkowski et al. 1988c). Estimates of 6, -At low irradiances (but above the threshold for the leakage of electrons from the oxygen-evolving complex) when all reaction centers are open, the photochemical quenching coefficient (qij) is 1. Four successive electron transfer reactions are required to generate 1 02, and thus the maximum quantum yield of electron transport for 0, evolution in PS2, &, is 0.25 O2 quanta- I. Under these conditions, energy losses occur primarily during the process of excitation energy transfer from the light-harvesting antenna pigments to the reaction center. The saturation level of photosynthesis is determined by the maximum rate at which electrons can be transferred from water to the terminal acceptors. If we designate TP as the turnover time for whole-chain electron flow in steady state photosynthesis, C#J, can be expressed as (b, = , = 0.25/(EcPS2qp7,>) if Eaps2qr, > l/~~, (W i.e. as long as the rate of primary photochemistry does not exceed the maximal rate of electron transfer from PS2 to PSl (which is controlled by the dark photosynthetic reactions) (= l/~~), four electrons are used to evolve one molecule of oxygen. At rates of charge separation exceeding l/~~, a portion of the electrons transferred to Q, will be wasted due to limitations imposed by processes downstream of PS2, resulting in a decrease of 4,. Estimation of rp -To calculate &, we must determine TP,. Although it is technically difficult to measure 7p directly from fluorescence signals, it can be calculated from the change in q,, with irradiance (Fig. 2). The rationale for this is as follows: as irradiance increases and the PQ pool becomes reduced, the rate of Q,- oxidation decreases, and ho approaches 7/>. The

8 Fluorescence-photosynthesis relationship 1653 change in the time constant for oxidation of Qa- is observed as a sudden decline of qp at an irradiance corresponding to E,, on a P vs. E curve (Fig. 2). From knowledge of El, and flfs2, we can calculate rp: TP = Wk+s2). (81 Note that this calculation requires that we know gps2 in absolute terms of m2 quanta- I; i.e. it is not a relative absorption cross-section (Mauzerall and Greenbaum 1989). Estimates of nps2 and the concentration of functional PS2 reaction centers- At present, one of the largest uncertainties in the calculation of photosynthetic electron flow per unit Chl a is related to the measurement of nps2. nzps2 is not easily measured by fluorescence techniques alone. Laboratory data indicate that the size of the photosynthetic unit (PSUo,) is relatively constrained, averaging -2,000 molecules Chl a per molecule O2 (Mauzerall and Greenbaum 1989). As each PSUo, contains the equivalent of four PS2 reaction centers, this corresponds to an average nps2 value of 1 PS2 per 500 Chl a molecules. This average value is used in all subsequent calculations where we derive Po2 from fluorescence measurements. We recognize, however, that the value of nps2 can diverge from the typical value of l/500. Functional changes in the fraction of PS2 (i.e. those capable of O2 evolution) are more important than variability of nps2. This fraction, J varies by a factor of -3 in the oceans (Kolber et al. 1990). In laboratory cultures of phytoplankton grown under optimal conditions, the maximum change of variable fluorescence, A&, = (F, - F,)/F,, is 0.65, This value is not unity because of inefficiencies in energy transfer and charge recombination within the reaction center. Nonetheless, the maximum value of 0.65 is remarkably constant in all species of algae examined to date. We assume that when A&, is 0.65, close to 100% of the PS2 reaction centers are functional. When the growth conditions deviate from the optimal, the observed A&, as well as the measured PB,,, decline (Kolber et al. 1988). The decline in A4, occurs because some PS2 reaction centers become photosynthetically incompetent due either to inefficient excitation transfer from the pigment bed to the reaction center, impairment of primary photochemistry, or disruption of electron transport between Qil and PQ. In all these cases, the incompetent reaction centers will exhibit the maximal fluorescence yield, F,,, irrespective of irradiance conditions. The measured A&, in this situation will be Ah = -E, - [ffo + (1 - ff miwn, = f(fm - F,,)/F,m = 0.65 f (9) where f is the fraction of the functional reaction centers. Therefore, A&,/O.65 is a measure of J and consequently F nps2 = nps2(&)m/0.65). (10) The natural variability in A&, (or J) appears to be a major factor affecting the chlorophyllspecific rates of photosynthesis in the ocean. Operationally, the measurement of A& is quite simple with a pump-and-probe fluorometer and is determined simultaneously with that Of%2. In principle, A$, can also be measured with any basic fluorometer by measuring the change in the quantum yields of fluorescence with and without the herbicide DCMU (Falkowski and Kiefer 1985; Owens 199 1; Geider et al. 1993). Nonphotochemical quenching--in practice, fluorescence usually does not follow a simple, inverse relationship with photochemistry. If it did, at higher ambient irradiance levels we would expect to see the quantum yield of fluorescence increase as PS2 reaction centers became closed (e.g. figure 15.9 of Rabinowich and Govindjee 1969). More often, however, the quantum yield of fluorescence shows a complex relationship with irradiance, increasing initially as more and more PS2 reaction centers close and then decreasing at high irradiance (Falkowski and Kiefer 1985; Owens 199 1). Since the high-light-induced fluorescence quenching does not appear to be due to the photochemical reduction of Q,, it is called nonphotochemical quenching, qn (Schreiber et al. 1986). The partitioning of fluorescence between qp and q, is based on the premise that photochemical charge separation ceases immediately upon exposure to the dark. Thus, the fluorescence yield measured in the light has contributions from both q,, and qn, but if the

9 1654 Kolber and Falkowski Time (m;n) Fig. 3. Kinetics of the changes in qn and q,, in the dark, immediately following exposure to natural solar irradiance. At least two processes can be extracted from the analysis of the kinetics of the decay in qiv (0). The faster of the two components (half-time on the order of min) has little or no cffcct on q,, (0), while the slower (halftime on the order of l-3 h) is correlated with changes in FJF,. These data were obtained from a natural phytoplankton sample from the Middle Atlantic Bight taken from the upper 5 m in midday (-PAR 1,000 PEinst m- * s ) and followed continuous use of a flow cell of a benchtop pump-and-probe fluorometer. sample is placed in the dark, any residual fluorescence quenching must be due to non-photochemical quenching processes. qn can be calculated as qhf = UL - F,)/(F, - FO). (11) F,,, is the fluorescence yield immediately following the saturating flash in the dark-adapted state and F, is measured in the light under the steady state conditions. If qn were negligible, the difference (F, - F,) would be zero. qn is sometimes considered a safety valve protecting PS2 reaction centers from damage due to multiple excitation events at high photon flux densities (Demmig-Adams 1990), and it is significant only after exposure to supersaturating irradiances. We observe at least two components in qn relaxation in natural phytoplankton communities (Kolber et al. 1990; Falkowski 1992). The first decays with a half-time on the order of 5-l 5 min and has no effect on A&, (Fig. 3). This component is not blocked by the inhibitor of chloroplast-directed protein synthesis, chloramphenicol, and appears to be correlated with the xanthophyll cycle (Mortain-Bertrand and Falkowski 1989; Falkowski 1992; Olaizola 1993). The second decays with half-times on the order of 2-4 h, can be blocked by chloramphenicol, and is highly correlated with changes in A$,. The relative proportion of the two components is somewhat variable; however, the fast component dominates (i.e. is >50% of qn). We interpret the faster of the two components as corresponding to energy dissipation reactions in the antenna pigments and the slower component as corresponding to damage to PS2 reaction centers. From a practical viewpoint, qn can affect two important parameters in the fluorescencephotosynthesis relationship, namely gias2 and nps2. Our measurements of the changes in gps2 under a variety of background irradiance levels suggest that the effect of q, on the cross-section is relatively small ( crps2 change per unit of q, change) (Falkowski et al. 1986). The effect of the slower component on nps2 is taken into account by measurement of A&,,. Integrating model parameters- We have shown how the photosynthetic parameters in Eq. 8 and 9 are related to fluorescence-based measurements. We can determine a fluorescence-based photosynthetic rate, P,H, with units of electrons Chl- l time : qf(e) = Eaps2qr~enps2(ASm/0.65). (12) Equation 12 is valid over the entire range of irradiance levels experienced by phytoplankton in the sea. To calculate the rate of photosynthesis in terms of carbon fixed per unit Chl a, PcB, Eq. 12 must be modified, requiring some assumptions concerning electron utilization for CO, fixation. While four electrons are required. to reduce a single molecule of CO2 to the level of carbohydrate, not all electrons derived from water are used to reduce CO,; some are used to reduce N03- and (to a lesser extent) SO4 2-. The fraction of photochemitally produced electrons that actually are allocated to carbon fixation depends primarily on the C: N ratio of the cells, the nitrogen source, and photorespiratory losses (Myers 1980; Falkowski et al. 1985; Laws 199 1). The ratio of 0, evolved : CO2 fixed, R, can be assumed or measured. The regression of fluorescence-based estimates of phytoplankton photosynthesis with carbon fixation is an estimate of this ratio. We calculate PC as I <, (13) Calculation of CO2 fixation rates From fluo-

10 Fluorescence-photosynthesis relationship 1655 rescence measurements requires prior estimates of R. Additionally, from an operational viewpoint, carbon- or oxygen-based measurements of photosynthesis inherently include respiratory costs, while the fluorescence-based estimate of photosynthesis, using Eq. 13, does not. The difference, (c//r) - PCR, approximates respiration, and an average value can be derived by regression analysis. Materials and methods, Measurement of variable fluorescence in situ-to measure the fluorescence signals in situ, we constructed a submersible pump-andprobe fluorometer capable of vertically profiling the water column (Falkowski and Kolber 1990). The instrument can be interfaced to a CTD or used alone (Fig. 4). The submersible fluorometer produces a sequence of probepump-probe flashes at a repetition rate of 1 Hz. The actinic (pump) flash, with an output of - loi photons cm-2 between 400 and 500 nm, was preceded 500 ms by the first probe flash and then followed 70,US by the second probe flash; the intensity of both probe flashes was - 5 x 1012 photons cmv2. The flashes were generated by xenon flash tubes (EG&G 405B); the half-width of the probe and pump flashes was 0.7 and 2.5 ps, respectively. Blue light from both flashes was isolated with Corning 4-96 and 5-60 filters, with an additional Corion LS-550F cut-off interference filter for the probe flash. The flashes illuminated an upward facing Plexiglas sample chamber 1 cm in diameter and 2 cm deep. The fluorescence signal was isolated with two Corion LG670 cutoff filters and a 685 nm bandpass filter from MicroCoatings, Inc. The emission signal was detected with a 1.26-cm end-window photomultiplier (Hamamatsu R 1463). The instrument was lowered at - 5 m min- to accommodate the 1 -Hz sampling rate while providing high resolution profiles. The sample chamber was open at one end. During deployment, seawater continuously, passively flushed through the upward-facing sample chamber; consequently, the sample was exposed to ambient downwelling irradiance throughout the cast. The fluorescence signals were sent in real time through a conducting cable to either a CTD deck receiver or directly to a desk-top computer. The two signals sent by the submersible fluorometer were the fluorescence Mounting bracket :hicld Fig. 4. Schematic cut-away section showing the major components of the pump-and-probe profiling fluorometer. Pump-and-probe flashes are focused from two flash units on the upward-facing cuvette, and the actively stimulated fluorescence from the probe flashes is detcctcd with a photomultiplier tube. The acquired data are relayed as analog signals to a CTD deck unit and displayed on board in realtime. yields preceding and succeeding a saturating pump flash. The absolute voltage of either signal can be used to semiquantitatively profile the vertical distribution of chlorophyll in the water column and to estimate the concentration of PS2 reaction centers, nps2 (see below). The submersible instrument was mounted on a General Oceanic 12-bottle CTD rosette in a manner similar to mounting a 12-liter Niskin bottle. The top of the instrument (containing the sample chamber) clears the upper portion of the rosette frame, thereby reducing shadow effects. A 2a Lambda LiCor 190 quantum sensor was mounted next to the instrument, and simultaneous measurements of

11 1656 Kolber and Falkowski downwelling PAR ( nm), temperature, conductivity, and pressure were made during each fluorescence-profiling cast. The absorption cross-sections of PS2 and the dark-adapted fluorescence signals were measured with a custom-built deck-based fluorometer (Kolber et al. 1990) using water samples taken from discrete depths during the profile with Niskin or Go-Flo bottles mounted on the CTD rosette. Unlike the submersible instrument, the deck-based fluorometer permits continuous control of pump flash intensity and the time delay between pump and probe flashes. Additionally, the deck fluorometer was used to monitor curves of fluorescence yield recovery in the dark, and from these curves we estimated the level of nonphotochemical quenching. The absolute intensity of the pump flash was measured with a Hamamatsu S BK photodiode coupled with a charge-sensitive preamplifier. The photodiode signal was converted into radiometric units (quanta m-2) by using continuous light of spectral quality similar to the pump flash (continuous xenon lamp filtered by identical excitation filters) to intercalibrate the photodiode current with an NBS-traceable, calibrated Lambda LiCor 190 quantum sensor. For deck-based measurements, 500-l,000 ml of sample from 6-8 discrete depths were dispensed into precleaned polycarbonate flasks and dark adapted for 5 min in a circulating water bath at in situ temperatures. The sample was pumped continuously through silicone tubing with a diaphragm bellows pump through a 250~~1 quartz fluorescence flow-through cuvette. In this configuration, samples were exposed to only one pump flash during measurement, thereby eliminating unwanted long-term actinic effects of the pump flash. With the deck fluorometer, measurements of gps2 (acquiring data for 24 pump flash intensities and signal averaging 10 flashes) required - 5 min. The surface samples were processed first, thereby minimizing the effect of qn recovery on measured ups2 and nps2. The results of deck measurements (a,,, and nps2) were interpolated over the euphotic zone with a second-order spline curve-fitting procedure. Invariably, this procedure neglects the effects of the fast-relaxing portion of qn and will result in overestimation of the in situ gps2. Intercalibration with incorporation of 14C0, - Data were obtained from four cruises in the northwest Atlantic Ocean in March 1988 and March, June, and October At each station, primary productivity was measured on discrete samples by following the incorporation of NaH14C0,. Samples were collected simultaneously with the fluorescence profiles, withdrawn from Niskin or Go-Flo bottles as soon as possible upon retrieval of the rosette, and prefiltered through 200~pm-mesh Nitex netting to remove larger zooplankton. Duplicate samples were dispensed into acid-cleaned 250-ml milk dilution bottles, covered with screens to attenuate light to 60, 35, 25, 15, 10, 5, and 1% of the surface irradiance, and incubated in a deckboard Plexiglas incubator with continuously flowing seawater to maintain ambient temperatures. Irradiance was recorded with a deck integrator interfaced with a Lambda LiCor 185 quantum sensor. The samples were incubated for 4 h with 5 PCi of NaH14C03 and filtered on Millipore HA filters under low vacuum. The filters were placed over fuming HCl for 60 s, dried, and radioactivity was assayed with a liquid scintillation counter with an external standard to correct for quenching. Dark bottles were run for each depth and the counts (usually < 1 O/o) subtracted from the corresponding sample. Samples for chlorophyll were filtered on Whatman GF/F glass-fiber filters and extracted by homogenization with a motor-driven Teflon pestle in a glass mortar with 90% acetone. The in vitro fluorescence was analyzed with a Turner Designs model 10 fluorometer before and after acidification et al. 1965). The fluorometer with pure Chl a. (Holm-Hansen was calibrated Results and discussion In situ measurements with the pump-and-probe fluorometer -Three representative vertical profiles of fluorescence signatures acquired with the in situ fluorometer in an unstratified euphotic zone are presented in Fig. 5. Night profiles of F and F,,, are parallel and conform to the vertical structure of phytoplankton chlorophyll. Under these conditions, qp e 0 because the only ambient light is night-sky light and decklights from the ship.

12 Fluorescence-photosynthesis relationship 1657 Chlorophyll (pg liter- ) Chlorophyll (pg liter- ) Chlorophyll (kg liter- ) Irrad. (PEinst m -2 s -1 ) -2 Irrad. (PEinst m s -1 ) x 100 x o $ Productivity Productivity qp [mg C (mg Chl)- h-l] [mg C (mg Chl)- h-l] q, x 0.25 qp x 0.25 Fig. 5. Representative vertical profiles of F, F,,,, q,>, Py (a- the fluorescence-based estimates of primary production with Eq. 4), PC j (V- 14C-based estimates of primary production), downwelling irradiance ( nm) (E), and Chl a (0). The profiles were taken from a site at N, W in the western North Atlantic in March Profiles (a) at night, when qlj is 1: 1, photosynthesis is nil and F' and F',,, are parallel and correspond closely to the distribution of Chl; (b) in the early morning of an overcast day, q,, is depressed slightly in the upper portion of the water column; and (c) under high irradiance lcvcls (note the sharp inflection of q,, at - 10 m, corresponding to the Ex value in situ). A similar profile recorded early in the morning during an overcast day is also presented in Fig. 5. As the irradiance level increases toward the surface, F changes relative to F,. Since the fast-relaxing component of qn does not affect q, (Fig. 3), the relative changes in F and F, must have resulted from F increase relative to F, (a decrease in qp): The vertical profile of F, still reasonably follows the distribution of chlorophyll throughout the water column, although near the surface F, declines relative to Chl a, indicating a small contribution of qn. During a bright, sunny day the effect of the absorbed radiation on qp is much more pronounced. qp shows at least two distinctive phases. In the upper portion of the water col- umn, it decreases toward the surface due to the increase in irradiance. This behavior corresponds to the decreased probability of finding an open PS2 reaction center at higher photon flux densities. At some irradiance level, qp reaches steady state. In the northwest Atlantic, the inflection depth usually corresponds to irradiances of between 100 and -250 PEinst m-2 s-l. The steady state level of qp at lower photon flux densities indicates that the probability of finding an open PS2 reaction center is relatively constant, corresponding to the initial slope of the photosynthesis-irradiance curve, ar. Consequently, the inflection depth at which qp starts to decrease corresponds to Ek. For data as in Fig. 5b (no Ek point), determination of TP is not required [i.e. 4, = 0.25

13 1658 Kolber and Falkowski l 3 a r $ 0.6- IL : a, IY Wavelength Fig. 6. Spectral quality of the actinic flash used in the pump-and-probe fluorometer. over the entire water column (Eq. 7a)]. On bright days, F can deviate significantly from the distribution of extracted chlorophyll in the upper portion of the euphotic zone. The deviation is caused by qn. Natural variability in ~~~~~~ -Pump-andprobe-based calculations of phytoplankton photosynthesis require knowledge of the rate of light absorption in PS2. Correct measurements of gps2 depend critically on radiometric calibration of the pump flash intensity. Furthermore, ups2 is a function of wavelength and will reflect the actual effective absorption cross section only if the spectrum of the pump flash is comparable to that of the in situ spectral distribution of irradiance. Because spectral irradiance continuously varies with depth, water type, and time of day (Morel 1978), it is impossible to select filters for the pump flash that absolutely mimic the spectral light quality found in situ. We selected a broadband excitation spectrum in a range nm (Fig. 6). This spectrum is similar to the blue-green window of the sun irradiance penetrating the water column in case 1 waters; it is somewhat blueshifted relative to deep case 2 waters (Morel and Prieur 1977). Our measurements of glbs2 deep in the water column, especially in the coastal ocean where spectral irradiance is narrower and red-shifted, might be overestimated. Although not ideal, our choice of filters reasonably mimics spectral distributions under a wide variety of in situ conditions. Measurements of gljs2 with the pump-and-probe technique reveal that the absolute cross-sections are highly variable in natural phytoplankton (nm) communities. Our measurements yield values ranging from 250 to 1,000A2, averaging - 500&. In the upper portion of the euphotic zone where spectral irradiance is more broadband, our estimates of gps2 are probably low. However, photosynthesis is likely to be light saturated in this region of the water column and no longer controlled by light absorption but by nps2, 43,, and rp. Natural variability in r,-in laboratory cultures of phytoplankton, 7p can increase from - 1 to 10 ms as cells adapt to lower growth irradiance levels (Myers and Graham 197 1; Falkowski et al ; Sukenik et al. 1987). The change appears to be due to the inability of the dark carbon fixation pathways to keep pace with the rate of supply of reductant from the electron transport chain at light saturation (Sukenik et al. 1987). From knowledge of Ek and o-ps2, we used Eq. 9 to calculate 7/J for natural phytoplankton communities in situ. Our results indicate that TP, is less variable than measurements in culture, ranging from to 6 ms and averaging ms (mean * SD). Estimating nljs2 - The conversion of pho- tosynthetic electron flow per unit time per reaction center to a photosynthetic rate per unit chlorophyll requires measurement of nps2. If we assume n,3s2 = electron/chl a, f nps2 is calculated from Eq. 8. The validity of this approach depends critically on the assumption that nps2 is constant. Laboratory data suggest that under physiological conditions this number may vary with growth irradiance (Fal- kowski et al ; Dubinsky et al. 1986) and nitrogen limitation (Herzig and Falkowski 1989). These changes, however, are relatively small compared with changes in J; A&,, or gps2. The second assumption (nps2 = electron/chl a) can be tested by regression analysis of the fluorescence-based estimates of carbon fixation rates per PS2 reaction center, 4Rc 2 *. PfRc2 = E~s2hPc, (14) and the simultaneously acquired 14C estimates of photosynthetic rates normalized to the fraction of functional Chl a, Per = PI 4C.f3A&, If we assume that the photosynthetic quotient is 1, the slope of the regression line (=O.OO 183, or 1 electron per 545 Chl a molecules) is the lower end estimate of nps2. Although the as-