Photosynthesis and photoprotection in top leaves respond faster to irradiance fluctuations than bottom leaves in a tomato canopy Authors: Bingjie Shao, Yuqi Zhang, Elena Vincenzi, Sarah Berman, Silvere Vialet-Chabrand, Leo F M Marcelis, Tao Li, Elias Kaiser, Tracy Lawson Published: 2024 ABSTRACT AbstractAccounting for the dynamic responses of photosynthesis and photoprotection to naturally fluctuating irradiance can improve predictions of plant performance in the field, but the variation of these dynamics within crop canopies is poorly understood. We conducted a detailed study of dynamic and steady-state photosynthesis, photoprotection, leaf pigmentation, and stomatal anatomy in four leaf layers (100, 150, 200, and 250 cm from the floor) of a fully grown tomato (Solanum lycopersicum cv. Foundation) canopy in a greenhouse. We found that leaves at the top of the canopy exhibited higher photosynthetic capacity and slightly faster photosynthetic induction compared with lower-canopy leaves, accompanied by higher stomatal conductance and a faster activation of carboxylation and linear electron transport capacities. In upper-canopy leaves, non-photochemical quenching showed faster induction and relaxation after increases and decreases in irradiance, allowing for more effective photoprotection in these leaves. Despite these observed differences in transient responses between leaf layers, steady-state rather than dynamic photosynthesis traits were more influential for predicting photosynthesis under fluctuating irradiance. Also, a model analysis revealed that time-averaged photosynthesis under fluctuating irradiance could be accurately predicted by one set of Rubisco activation/deactivation parameters across all four leaf layers, thereby greatly simplifying future modelling efforts of whole-canopy photosynthesis.Keywords: Canopy photosynthesis, dynamic photosynthesis, non-photochemical quenching, tomato, photosynthetic induction, stomatal conductance Introduction Crops often grow in monoculture and canopies are formed by groups of plants. Plant canopies are a collection of leaves (Anten et al., 2016) that are characterized by spatial heterogeneity in irradiance, temperature, and humidity (Pearcy et al., 1994; Slattery et al., 2018; Tanaka et al., 2019; Durand et al., 2021; Zhuang et al., 2021; Long et al., 2022). These factors affect leaf properties (Campany et al., 2016; Liu et al., 2019), but knowledge on how dynamic responses of net photosynthetic rate (A) to changes in irradiance are affected by leaf position in the canopy is sparse. A better understanding of canopy A, including its dynamic responses to changes in the environment, may improve predictions of terrestrial carbon fluxes. Such knowledge may help with crop or climate management, utilization of lighting in controlled-environment agriculture, and improvements in productivity. Leaves at different positions in the canopy have structural, anatomical, and metabolic differences, all of which affect A (Niinemets and Sack, 2006; Niinemets, 2007). Irradiance is typically lowest at the bottom of the canopy, and throughout the canopy can change by up to 50-fold (Niinemets et al., 2015; Anten et al., 2016). These conditions typically result in upper-canopy leaves being thicker and smaller (Van Wittenberghe et al., 2012; Murchie et al., 2022) than their older and less strongly illuminated counterparts. Also, the red: far-red ratio (R/FR) drops markedly throughout the canopy (Huber et al., 2021), as leaves absorb red much more strongly than they do far-red. Finally, leaf age increases from the top to the bottom of the canopy, at least in crops with indeterminate growth, such as tomato cultivars used in greenhouse production. Leaves acclimate to each of these factors by changing traits such as specific leaf area (SLA), nitrogen concentration ([N]), chlorophyll concentration ([Chl]), stomatal size, and stomatal density (Sharkey et al., 1996; Hikosaka, 2005, 2016; Legner et al., 2014; Scartazza et al., 2016). In nature and in greenhouses, solar irradiance often fluctuates, causing A to change rapidly between repeated shade-to-sun and sun-to-shade transitions (Kaiser et al., 2018; Durand and Robson, 2023; van Westreenen et al., 2023). Photosynthesis in leaves that are initially adapted to low irradiance responds with a gradual increase upon sudden exposure to a high irradiance (photosynthetic induction), which is mainly controlled by the slow activation of Calvin–Benson–Bassham (CBB) cycle enzymes and stomatal opening (Sassenrath-Cole and Pearcy, 1992; Kaiser et al., 2015, 2016, 2018; Sakoda et al., 2021; Liu et al., 2022). Upon a sun-to-shade transition, A drops within seconds while deactivation of CBB cycle enzymes and stomatal closure proceed at timescales of a few to several tens of minutes (Pearcy et al., 1996). Depending on the initial activation state of the leaf, it can take A between seconds and tens of minutes to respond to shade-to-sun transitions (Pearcy et al., 1996), resulting in a calculated loss of 10–40% of diurnal A (Taylor and Long, 2017; Wang et al., 2020; Long et al., 2022). Under irradiance fluctuations, a faster rate of photosynthetic induction may thus enhance sunfleck-use efficiency (Durand et al., 2022). While a large body of literature exists on the effects of leaf position in a canopy on steady-state A (e.g. Evans, 1989; Niinemets, 2007; Li et al., 2014; Niinemets et al., 2015), much less is known on dynamic traits, and published results differ (Tausz et al., 2005; Urban et al., 2007, 2008; Durand et al., 2022). Furthermore, almost all research concerning leaf position and dynamic A has been conducted on woody species, whereas knowledge on herbaceous crops is sparse (Acevedo-Siaca et al., 2021a). Dynamic A is under diffusional and biochemical limitations (Kaiser et al., 2017). Stomata, small pores on the leaf surface that dynamically balance the influx of CO2 against the efflux of water vapor, are a key limitation to photosynthetic induction (Kaiser et al., 2019; Long et al., 2022; N. Zhang et al., 2022). For example, a higher stomatal conductance (gs) at the start of photosynthetic induction often leads to faster photosynthetic induction (Kaiser et al., 2020). Within a plant canopy, gs tends to be higher in top leaves (Gay and Hurd, 1975; Urban et al., 2007; Eensalu et al., 2008; Igarashi et al., 2015; Durand et al., 2022), which may thus exhibit a faster rate of photosynthetic induction than bottom leaves. The late phase of photosynthetic induction and stomatal opening were previously shown to slow down with leaf age (Faralli et al., 2019; Y. Zhang et al., 2022). Transient biochemical limitations during photosynthetic induction involve ribulose-1,5-bisphosphate (RuBP) regeneration activation and ribulose-1,5-bisphosphate carboxylase/oxygenase (Rubisco) activation (Mott and Woodrow, 2000), the latter of which is under control of the chaperone Rubisco activase (Carmo-Silva and Salvucci, 2013). The rate of change in these limitations can be calculated from transient changes in maximum carboxylation rate (Vcmax) and linear electron transport rate under high irradiance (J1500; Soleh et al., 2016; Salter et al., 2019); both traits often scale positively with [N]. From the top to the bottom of a canopy, leaf [N] at first increases due to leaf development and then declines as a function of the irradiance gradient (Hikosaka et al., 2016), resulting in highest [N] in fully developed, sun-exposed leaves. Several recent studies have found positive correlations between [N] and the rate of photosynthetic induction (Liu et al., 2021; Sun et al., 2022; Xiong et al., 2022), and negative correlations between [N] and the rate of deactivation of photosynthesis after a switch from high to low irradiance (‘loss of photosynthetic induction’; Sun et al., 2016). Together, this suggests that leaves with high [N] (i.e. leaves at the top of the canopy) are able to utilize fluctuating irradiance more effectively than low [N] leaves. Under high irradiance, leaves can incur photoinhibition. Non-photochemical quenching (NPQ) dissipates excess energy as heat, thereby protecting leaves from photoinhibition (Kasajima et al., 2011; Murchie and Niyogi, 2011; Ruban, 2016; Hubbart et al., 2018). NPQ is a compound term that includes rapid energy-dependent quenching (qE), as well as the slower processes state transitions (qT), zeaxanthin-dependent quenching (qZ), and photoinhibition (qI; Iwai et al., 2010; Nilkens et al., 2010; Minagawa and Tokutsu, 2015; Long et al., 2022). These processes can be estimated by fitting a double exponential function to the time course of NPQ relaxation after a sun-to-shade transition (Long et al., 2022). The rate of NPQ relaxation has been linked to productivity in the field (Kromdijk et al., 2016). However, to what extent NPQ induction and relaxation differ in various canopy layers is under-researched (Tausz et al., 2005; Acevedo-Siaca et al., 2021a). Here, we asked whether the dynamics of photosynthesis and photoprotection under irradiance fluctuations differ between leaves at different heights of a canopy, which physiological mechanisms may explain the differences observed, and what the relative effects of steady-state versus dynamic photosynthesis traits were on photosynthesis under fluctuating irradiance. To this end, we conducted a detailed investigation of key properties of photosynthesis and photoprotection in a fully grown tomato canopy. We hypothesized that (i) leaves in the upper canopy induce A more quickly, (ii) this response is due to higher initial gs and faster activation of Rubisco, and (iii) NPQ induction and relaxation are faster in upper-canopy leaves. Materials and methods Plant material and growth conditions Tomato seeds (Solanum lycopersicum cv. Foundation; BASF Nunhems, the Netherlands) were sown in stonewool plugs and initially grown in a greenhouse in Nunhem, the Netherlands. Plugs were soaked in a nutrient solution and covered with a thin layer of vermiculite. Around 2 weeks later, uniform seedlings were transplanted into rockwool blocks. Once they were ~50 cm tall (~4–5 weeks after sowing), plants were brought to Wageningen University, Wageningen, the Netherlands (52°N, 5°E). Plants were planted in two adjacent Venlo-type glasshouse compartments on 27 September 2021 [defined as day 1 of experiment (DOE)], and grown until 15 February 2022. Each compartment had an area of 12 × 12 m2. Four gutters were arranged from east to west per compartment, and the distance between adjacent gutters was 1.5 m. Plants on the gutter were planted every 25 cm and alternatively trained to two high wires located 30 cm to the right and left of the gutter. Shoot density was 2.66 plants m−2. Plants were grown on rockwool slabs (Grodan, Roermond, the Netherlands) with drip irrigation according to commercial practice. The pH of the nutrient solution was 6.0 and mean electrical conductivity was 3.0 dS m−1; these values were checked weekly. Plants were lowered weekly, to maintain a height of ~300 cm from the uppermost (first) leaf to the floor, and leaves below the second-lowest truss (50–70 cm from the floor) were removed, as is common practice in greenhouse tomato cultivation. Fruits were harvested when they reached a red-ripe stage. Supplementary lighting was applied by GPL Philips top lighting 1.2DRW-LB 400V modules (Signify, Eindhoven, the Netherlands). Average photosynthetic photon flux density (PPFD) from supplemental light was 226 μmol m−2 s−1 at ~320 cm from the floor. Lamps emitted 86% red, 6% green, and 8% blue light, with very small amounts of green and far-red light (Supplementary Fig. S1). During the measurement period, lamps were on between 00.00 h and 16.00 h. When outside radiation surpassed 500 W m−2, lamps were turned off, and were turned on again at 400 W m−2. A greenhouse climate computer (Hogendoorn, Vlaardingen, the Netherlands) was used to control the climate, with identical setpoints in both compartments. During the entire experiment, average day/night temperature was 21.7/19.2 °C, air relative humidity was 73%, and CO2 concentration was 550 μmol mol−1. Measurements All photosynthesis, light environment, and chlorophyll content measurements were conducted between DOE 65 and DOE 141. Nitrogen content and stomatal observations were conducted afterwards. The first leaf in this study was defined as the first leaf that was longer than 5 cm. Measurements were conducted when the first leaf had reached a height of 300 cm from the floor. Light intensity and spectrum Solar radiation was continuously measured by a pyranometer (Apogee, USA) that was located above the greenhouse. To calculate solar PPFD inside the greenhouse, the following conversions were made: (i) the fraction of PPFD in the total solar radiation was assumed to be 47% (Britton and Dodd, 1976), (ii) the conversion factor from energy flux to PPFD was assumed to be 4.57 μmol J−1 (McCree, 1970), and (iii) the light transmissivity of both compartments was determined to be 51%, and was calculated by comparing PPFD above the LEDs and PPFD estimated from the pyranometer data. Four sensors just above the LED modules (LI-190; LI-COR Biosciences, Lincoln, NE, USA) logged PPFD every minute. To determine the light extinction coefficient (k) of the canopy, two measurement areas (0.75 × 0.75 m) were chosen per compartment. The areas were selected directly underneath and between plants, thus accurately representing the light conditions within the greenhouse (Supplementary Fig. S2). Each designated area encompassed two plants. In each area, two equidistant points along the gutter were measured using a line quantum sensor (LI-191R, LI-COR). The line quantum sensor was positioned parallel to the ground and at a 45° angle from the center of gutter to the midpoint of the pathway, to measure a representative area of the canopy. PPFD was measured at gutter height (approx. 40 cm above the floor), and then every 30 cm until reaching the plant’s apex (approx. 310 cm above the floor). Leaf area (LA) of two plants in the PPFD measurement areas was estimated non-destructively, by measuring leaf length (L) and width (W) using a soft tape, together with leaf number and height. Relationships between LA, L, and W were obtained by taking leaves from three border plants before photosynthesis measurements (see below), and measuring LA, using the LI-3100 leaf area meter (LI-COR; Higashide and Heuvelink, 2009): where α is a proportionality factor, and was determined to be 0.2953. PPFD and leaf area measurements were repeated four times (DOE 65, 126, 137, and 139) after the canopy reached full closure and remained at a height of around 300 cm from the end of November 2021 (DOE 65) through mid-February 2022 (DOE 141). The PPFD profile within the canopy was calculated (Monsi and Saeki, 2005): where I(L) is PPFD (μmol m−2 s−1) at canopy depth L, I0 is the incident PPFD at the top of the canopy, k is the extinction coefficient, and LAI(L) is the cumulative leaf area index (LAI; m2 leaf area per m2 ground area) from the top of the canopy until canopy depth L. Using this information, the daily light integral (DLI; mol photons m−2 d−1) was calculated per canopy layer. The height profile of the light spectrum was measured by a spectrometer (LI-180; LI-COR) every 50 cm in the range 50–300 cm from the floor. Light spectrum measurements were repeated three times (DOE 93, 103, and 140), at four spots per compartment. These spots were randomly selected to be equidistant along the gutter and positioned 30 cm from the center of the gutter. Photosynthetic gas exchange and chlorophyll fluorescence The canopy was divided into four leaf layers at distances of 100, 150, 200, and 250 cm from the greenhouse floor. Leaves from each layer were selected randomly. Measurements were conducted in two adjacent greenhouse compartments, with each compartment accounting for half of the measurements taken. All measurements were conducted using the LI-6800 photosynthesis system (LI-COR) equipped with the leaf chamber fluorometer (LI-COR part no. 6800-01A; enclosed leaf area: 2 cm2), and at a leaf temperature of ~25 °C, leaf to air vapor pressure deficit of ~1 kPa, a flow rate of 500 μmol s−1, and [CO2] of 400 μmol mol−1 (unless specified differently). Irradiance was provided by a mixture of red (90%) and blue (10%) LEDs in the fluorometer, with peak intensities at 635 and 465 nm, respectively. Chlorophyll fluorescence (CF) was measured using a rectangular saturating flash of 16 000 μmol m−2 s−1 intensity that was applied for 500 ms, and at a data acquisition rate of 100 Hz. Measurements were conducted on six to eight biological replicates per layer, from DOE 68 to DOE 141, between 08.00 h and 17.00 h, and were fully randomized between leaf layers and type of measurement. Steady-state light and CO2 response curves Leaves were adapted to 2000 μmol m−2 s−1, until A was stable (~40 min). Then, irradiance was lowered in the following steps: 1800, 1500, 1200, 1000, 800, 600, 400, 300, 200, 150, 100, 50, 10, 0 μmol m−2 s−1. At each irradiance, A was logged 10 times at intervals of 5 s, after reaching a steady state (1–2 min per step), and average A per step was used in subsequent data analysis. At each irradiance, maximum (Fmʹ) and operating CF (Fʹ) were determined. Then, leaves were adapted to 1500 μmol m−2 s−1 and 400 μmol mol−1 [CO2]. After A reached a steady state, [CO2] was changed in steps of 300, 200, 150, 100, 50, 400, 600, 800, 1000, 1200, 1500, 1800, 2000 μmol mol−1, and data were logged as described above, after matching the IRGAs at each [CO2]. Dynamic photosynthesis Several protocols with low (50 μmol m−2 s−1) and high (1500 μmol m−2 s−1) irradiance were executed, to obtain a full characterization of dynamic A in different layers of the canopy. To obtain rate constants of A and gs changes in a dynamic irradiance environment, gas exchange and CF were measured after stepwise increases and decreases in irradiance. Leaves were first dark-adapted for 20 min by wrapping them in aluminum foil, and were then inserted into the darkened cuvette for another 10 min. Minimal (Fo) and maximal (Fm) CF were recorded to determine maximum quantum efficiency of photosystem II (PSII) photochemistry [Fv/Fm=(Fm−Fo)/Fm]. Irradiance was then increased to 50 μmol m−2 s−1, until A and gs reached stability (~30 min). Gas exchange and CF were recorded for 10 min, then irradiance was increased to high irradiance (1500 μmol m−2 s−1) for 60 min, after which it was dropped back to low irradiance (50 μmol m−2 s−1) for another 60 min. Gas exchange was logged every 1–2 s, and Fs and Fmʹ were additionally recorded every 1 min in the first 10 min after a change in irradiance, and every 2 min thereafter. Photosystem II electron transport efficiency (ΦPSII) was calculated as: ΦPSII=(Fmʹ−Fs)/Fmʹ, and non-photochemical quenching (NPQ) was calculated as: NPQ=Fm/(Fmʹ−1). To determine changes in Vcmax and J1500 during photosynthetic induction, A and Ci were measured during photosynthetic induction after low-to-high irradiance step changes at different [CO2], namely 50, 100, 150, 200, 300, 400, 600, 800, 1000, 1200 μmol mol−1 [CO2] (Taylor and Long, 2017; Salter et al., 2019). Leaves were adapted to low irradiance until A was stable (~20 min), then irradiance was increased to 1500 μmol m−2 s−1 in a stepwise change. Gas exchange was logged every 2 s for 10 min under low irradiance, and for 20 min under high irradiance. To determine the effect of low-irradiance durations before high irradiance on rate constants of photosynthetic induction (i.e. loss of photosynthetic induction), leaves were exposed to different durations of low irradiance, after reaching full photosynthetic induction at high irradiance. Leaves were first adapted to 1500 μmol m−2 s−1 until A and gs were stable, after which irradiance was dropped to 50 μmol m−2 s−1 for 1, 3, 5, 10, 15, 30 min (as well as 60 min, see above). Then, irradiance was returned to 1500 μmol m−2 s−1, and gas exchange was logged every 2 s. Periods of low irradiance were randomized to prevent artefacts from pathway dependency effects of previous irradiance exposure. To observe the response of A and gs under an artificially fluctuating irradiance regime (lightfleck sequence), leaves were first adapted to 50 μmol m−2 s−1 irradiance, until A and gs were stable. Then, leaves were subjected to 2 min of low irradiance, followed by 2 min of high irradiance, and this cycle was repeated six times. Leaf anatomy and pigmentation To determine specific leaf area (SLA), 16 leaf samples per leaf layer were collected. Each sample contained four to six leaflets, which had previously been used for photosynthesis measurements (10 samples were collected on DOE 102, and another six samples were collected on DOE 142). Fresh weight and leaf area were measured. Then, samples were dried at 70 °C for 3 d, and dry weight was recorded. SLA was calculated as leaf area divided by leaf dry weight. To determine leaf nitrogen content ([N]), six samples were taken from each layer, each containing four to six leaflets. Samples were dried at 70 °C for 3 d, and were ground in a mortar. Of the resulting powder, 0.15 g per sample was used for analysis and a C/N analyser (CHN828, LECO, USA) was used to determine [N]. To determine leaf pigment contents, discs of 1 cm Ø were taken from the leaves used for photosynthesis measurements, stored in liquid N, and transferred to glass bottles containing 1.5 ml N,N-dimethylformamide. Leaf disks were stored at −20 °C in darkness for 2 weeks for pigment extraction. The absorbance of the extract was measured at 480, 647, 664, and 750 nm using a SpectraMax iD3 Multi-mode Microplate Reader (Molecular Devices, San Jose, CA, USA). Chlorophyll a, chlorophyll b, and total carotenoid contents were calculated after Wellburn (1994). To analyse stomatal anatomy, imprints of six leaves per layer were made on both sides of the leaf, using the silicone rubber impression technique (Savvides et al., 2012). Impressions were placed on microscope slides and analysed using an optical microscope (Leica DM2000 LED, Germany). Five fields of view were randomly selected per sampling area, where stomatal and epidermal cells numbers were counted. Twenty-five stomata were randomly selected per area, on which stomatal length and width were measured (Ando analysis biology 3.8.6, China), following Savvides et al. (2012). Calculations Steady-state A A non-rectangular hyperbolic function was fitted to the light response curve of A (Equation 6 in (Lobo et al., 2013). Parameters, including maximum A (Asat), mitochondrial respiration rate (Rd) and apparent quantum yield (α), were derived using the Excel solver. From A–Ci curves, maximum carboxylation rate (Vcmax), electron transport rate at high irradiance (J1500), and maximum rate of triose phosphate use (TPU) were derived by using the Sharkey Excel tool (Sharkey et al., 2007; Sharkey, 2016). For purposes of A–Ci curve fitting, Rd was assumed to be 50% of the value determined from light response curves (Sharkey, 2016), and mesophyll conductance at 400 μmol mol−1 atmospheric [CO2] was calculated using the variable J method (Harley et al., 1992). Photosynthetic induction Transient A and gs responses after step increases and decreases in irradiance were averaged over five data points to reduce measurement noise, using a moving average filter. Photosynthetic induction state was calculated after Zipperlen and Press (1997): where A50 is steady-state A under 50 μmol m−2 s−1 before the increase in irradiance (averaged over ~120 s), A1500 is final steady-state A under 1500 μmol m−2 s−1, and A is transient A at time t during photosynthetic induction. Photosynthetic induction state 60 s after the irradiance increase, IS60, as well as the times required to reach 50% and 90% of full photosynthetic induction, t50-A and t90-A, were then calculated. gs50 and gs1500 were calculated analogously to A50 and A1500. Intrinsic water use efficiency (WUEi) was calculated as WUEi=A/gs. Stomatal kinetics Two empirical models were used to quantify the temporal response of gs to a stepwise change in irradiance (Vialet-Chabrand et al., 2017). According to the shape of the gs response to a step increase in irradiance, a sigmoidal model was used: where λ represents the initial time lag (time after the increase in irradiance during which no change in gs is observed), and ki represents the time constant for an increase of gs. ki does not directly represent a time to reach a percentage of gs, as this also depends on λ. Therefore the time constant, τi, which represents the time to reach 63% of the total gs increase, was calculated as: After a step decrease in irradiance, the following exponential model was determined to be the best fit for stomatal closure in tomato leaves (Y. Zhang et al., 2022): where the time constant τd represents the time to reach 63% of the total decrease in gs. NPQ kinetics Separate models were fit to the induction (Equation 7) and dark relaxation (Equation 8) of NPQ to obtain their respective rate constants (b) (Ferguson et al., 2023): where in both equations, a represents the amplitude of the response, c is an offset to account for a non-zero intercept, and t is time. Different NPQ components were further quantified by fitting a double exponential function (Long et al., 2022) to NPQ relaxation: where Aq1 corresponds to amplitude of qE, Aq2 corresponds to amplitude of qZ and qT, Aq3 corresponds to amplitude of qI, τq1 represents half-life of qE relaxation, and τq2 represents half-life of qZ and qT relaxation. Dynamic A and gs model To simulate A during lightflecks, A was modelled using steady-state equations (FvCB model; Farquhar et al., 1980), which were included in differential equations as targets to describe time evolution similarly to Vialet‐Chabrand et al. (2021). Only the differences with the original FvCB model are described here. Vcmax was dependent on irradiance, and its induction by Rubisco activase followed an exponential response (Taylor et al., 2022): with VcT the Vcmax reached in a steady state, τc_i/d the time constants for an increase or a decrease in Vcmax. Only two irradiances (low and high) were used during the experiment, and VcT was set to a low value under low irradiance and increased under periods of high irradiance. Under high irradiance, the model was considered to be RuBP-saturated and the following equation was used with the dynamic estimation of Vcmax: with Γ∗ the photorespiratory CO2 compensation point, Cc the chloroplastic [CO2], Km the apparent Michaelis–Menten constant for CO2 of rubisco and Rd the mitochondrial respiration. Under low irradiance, the model was considered to be RuBP-limited, and the following equation was used: with J the electron transport rate. As only one low irradiance was used in the protocol, only one value was estimated for J. In this equation, Γ ∗ was set to a higher value post-illumination to account for part of the post-illumination CO2 burst (Pearcy et al., 1997; Kirschbaum et al., 1998). With a change in irradiance every 2 min, stomata had little time to react and gs was modelled in three parts: with g0 the initial gs, gday a long-term trend as a function of time (t) and gtw an hydropassive response due to light intensity: with τgi/d the time constant for an increase or a decrease in gs and gT the steady-state target. To calculate Ci, the following equation was used: with Ca the atmospheric CO2 concentration, and gt the total conductance to CO2, which was calculated as: with gbw the boundary layer conductance to water vapor as set in the gas exchange chamber. Parameter optimization The model was written in Julia (v1.10.4) and parameters were optimized using Bayesian Inference (Turing.jl). The value of gb was set to 3 mol m−2 s−1, gm to 0.3 mol m−2 s−1, Γ ∗ to 40 µmol mol−1 (Kaiser et al., 2020) and Rd to the value estimated for each layer during the photosynthesis light response curve. All chains converged (Rhat between 0.95 and 1.05; ess>100) without divergent transitions. Stomatal and biochemical limitations Transient stomatal limitation (LS, %) and biochemical limitation (LB, %) during photosynthetic induction were calculated according to Kaiser et al. (2017) and Urban et al. (2007), where A*Ca and A*Ci were A corrected for changes in transient Ca and Ci: V cmax and J1500 induction Using data of photosynthetic induction at several [CO2] (50–1200 μmol mol−1), A–Ci curves were generated every 2 s during photosynthetic induction. The Farquhar et al. (1980) model was fitted to these curves using the ‘plantecophys’ package in R (Duursma, 2015; Salter et al., 2019). To determine how fast Vcmax and J1500 were generated, the times to reach 50% of full Vcmax and J1500 induction (t50-Vcmax, t50-J1500) were calculated analogously to t50-A, respectively. Averages under fluctuating irradiance To quantify transient changes of key parameters, the averages of A, gs, and ΦPSII in the first 5 min after a shade-to-sun (Aavg_i, gsavg_i, ΦPSIIavg_i) and a sun-to-shade transition (Aavg_d, gsavg_d, ΦPSIIavg_d) were calculated. Furthermore, to calculate the effects of rapid and slow NPQ relaxation on A, A50 was subtracted from A after the sun-to-shade transition. Then, average A was calculated 30–150 s after the sun-to-shade transition (time window during which rapid NPQ relaxation may affect A; Kromdijk et al., 2016), and 58–60 min after the sun-to-shade transition (long-lasting effects of slow NPQ relaxation or photoinhibition on A; A50dcorrfast, A50dcorrslow; Supplementary Fig. S3). Statistical analysis One-way ANOVA was performed to test for differences between leaf layers. The normality of residuals was tested by a Shapiro–Wilk test at P=0.05. If the requirement for normal distribution was not fulfilled, Kruskal–Wallis one-way ANOVA was performed. If the data were normally distributed and the variance was homogeneous, the least significant difference (LSD) test was used to test for differences between any two different layers at a significance level of P=0.05. If data were normally distributed but the variance was not homogeneous, the Tamhane test was used to test for differences between any two different layers at P=0.05. All statistical analyses were performed in IBM SPSS Statistics 24.0 (IBM Corp., Armonk, NY, USA), except for the correlation matrix, which was generated using the ‘corrplot’ package in R (https://www.r-project.org/). Results Light environment inside a fully grown tomato canopy We conducted detailed measurements of photosynthetic properties and climate variables in four leaf layers in a fully grown tomato canopy in the greenhouse; layer 1 was located 250 cm above the greenhouse floor, and layers 2–4 at 200, 150, and 100 cm, respectively. As expected, irradiance showed a strong vertical gradient, declining from ~300 μmol m−2 s−1 just above the canopy to only ~50 μmol m−2 s−1 below it, showing the strongest decreases between layers 1 and 3 (Supplementary Fig. S4A). R/FR also declined from layer 1 to 3, and stabilized between layers 3 and 4 (Supplementary Fig. S4B). Irradiance declined exponentially as a function of LAI, resulting in an extinction coefficient of 0.67 (Supplementary Fig. S4C). Consequently, throughout the course of the experiment, layer 1 received the highest DLI (~12 mol m−2 d−1), almost twice as much as layer 2 (~7 mol m−2 d−1), and approximately three times as much as layers 3 and 4 (which received ~4 and 3.6 mol m−2 d−1, respectively; Supplementary Fig. S5). Steady-state photosynthesis, leaf morphology, and pigmentation As expected, there was a steep decline in steady-state photosynthetic capacity, with leaves in layer 1 showing the largest and those in layer 4 showing the smallest capacity (Fig. 1). Light response curves of A, gs, and ΦPSII clearly scaled with leaf position in the canopy, as did A–Ci curves (Fig. 1A–C; Supplementary Fig. S6A–C), resulting in clear downward gradients in almost all derived parameters, i.e. Asat, Rd, Vcmax, J1500, and TPU (Fig. 1D–I). The quantum efficiency of CO2 fixation (α) also showed a downward gradient (Fig. 1F), suggesting that leaves in layer 1 were the most efficient in using incoming light to fix CO2. A showed a decrease at high Ci in layer 1 (Fig. 1C), paralleled by a strong decrease of ΦPSII (Supplementary Fig. S6B), indicating a ‘reverse sensitivity’ of A to CO2 in layer 1. TPU was not different between layers 1 and 2, but was significantly lower in layers 3 and 4 (Fig. 1I). Dark-adapted Fv/Fm was significantly reduced in layer 4, with no differences between the other layers (Supplementary Fig. S7G). Steady-state photosynthetic response curves and derived parameters in four layers of a fully grown tomato canopy. Layer 1 is the uppermost (top) layer. (A, B) Net photosynthetic rate (A) and photosystem II electron transport efficiency (ΦPSII) as a function of PPFD; (C) A as a function of leaf internal CO2 partial pressure (Ci); (D) light-saturated A (Asat); (E) mitochondrial respiration rate (Rd); (F) apparent quantum yield of CO2 fixation (α); (G) maximum carboxylation rate (Vcmax); (H) electron transport rate at 1500 μmol m−2 s−1 (J1500); (I) maximum rate of triose phosphate use (TPU). Mean values ±standard error of the mean (SEM) from six to seven biological replicates are shown. Box plots indicate 0% (maximum), 25%, 50% (median), 75%, and 100% (minimum), points show single data, and grey dashed lines indicate mean values. One-way ANOVA was performed between four layers; different letters indicate statistically significant differences (P<0.05). SLA was smaller in the upper two layers (Fig. 2F), indicating thicker leaves at the top of the canopy. Remarkably, there was a tendency for [N] to increase from the top to the bottom of the canopy, but this was not significant on a leaf area (P=0.145; Fig. 2A) and almost significant on a mass basis (P=0.053, Supplementary Fig. S7A). Concentrations of chlorophyll a ([Chla]) and chlorophyll b (Chlb]) were lowest in layer 1 and highest in layers 2 and 3, while their ratio declined linearly from layer 1 to 4 (Fig. 2C; Supplementary Fig. S7B, C). Leaves in layer 1 showed a higher stomatal density and smaller stomatal size than all other leaves (Fig. 2D, E), while stomatal index did not differ between layers (Supplementary Fig. S7E). Leaf morphology and pigmentation in four layers of a fully grown tomato canopy. Leaf nitrogen content ([N]; A); total chlorophyll content (B); chlorophyll a:b ratio (Chla/b; C); stomatal density (SD; D); stomatal size (SS; E); and specific leaf area (SLA; F) in four layers of a fully grown tomato canopy. Box plots indicate 0% (maximum), 25%, 50% (median), 75%, and 100% (minimum); points show single data; and grey dashed lines indicate mean values. One-way ANOVA was performed between four layers; different letters indicate statistically significant differences (P<0.05). Transient photosynthetic and photoprotective responses to shade-to-sun and sun-to-shade transitions Upon sudden exposure of shade-adapted leaves to an increase in irradiance, A in layer 1 exhibited a significantly faster response compared with the other layers (Fig. 3A, B), resulting in the smallest t50-A, as well as a slight tendency for a smaller t90-A in layer 1 (P=0.57; Fig. 3K, L). The induction state 60 s after an irradiance increase, IS60, was significantly higher in layer 1 than in layer 3, while values in layers 2 and 4 were intermediate (Fig. 3Q). Dynamic increases in A were paralleled by those in ΦPSII (Fig. 3F). Steady-state A at high irradiance (A1500) scaled with layer (Supplementary Fig. S6E), as did gs, transpiration rate, and ΦPSII (Fig. 3C, F; Supplementary Fig. S8A). Faster and stronger responses caused average A, gs, and ΦPSII during the first 5 min of photosynthetic induction to be larger in layer 1 and smallest in layer 4 (Fig. 4A, C, E). Dynamic photosynthesis, stomatal behavior, and NPQ components in four layers of a fully grown tomato canopy. (A) Time course of leaf net photosynthesis rate (A); (B) photosynthetic induction; (C) stomatal conductance (gs); (D) intrinsic water use efficiency (WUEi); (E) non-photochemical quenching (NPQ); (F) photosystem II electron transport efficiency (ΦPSII); (G) stomatal limitation (LS); (H) biochemical limitation (LB); (I, J) induction of maximum carboxylation rate (Vcmax) and electron transport rate (J1500); (K, L) times to reach 50% and 90% of full photosynthetic induction (t50-A and t90-A); (M, N) time constants of stomatal opening (τi) and closure (τd); (O, P) times to reach 50% of maximum Rubisco carboxylation induction rate (t50-Vcmax) and electron transport induction rate (t50- J1500); (Q) induction state 60 s after irradiance increase (IS60); (R, S, T) components of NPQ relaxation, energy-dependent quenching (qE), zeaxanthin-dependent quenching (qZ), state transition-dependent quenching (qT), and photoinhibition (qI). Mean values ±SEM are shown (n=8). Grey dashed lines in box plots indicate mean values; box plots indicate 0% (maximum), 25%, 50% (median), 75%, and 100% (minimum); points show single data. One-way ANOVA was performed between four layers; different letters indicate statistically significant differences (P<0.05). Averaged rates of photosynthesis, and of processes underlying it, during the first 5 min after a shade-to-sun (left panel) and a sun-to-shade (right panel) transition. (A, C, E) Average A, gs, and ΦPSII in the first 5 min after increasing irradiance; (B, D, F) average A, gs, and ΦPSII in the first 5 min after decreasing irradiance. (G, H) Rate constants of NPQ induction and relaxation. Grey dashed lines in box plots indicate mean values; box plots indicate 0% (maximum), 25%, 50% (median), 75%, and 100% (minimum); points show single data. One-way ANOVA was performed between four layers; different letters indicate statistically significant differences (P<0.05). During photosynthetic induction, gs was highest in layer 1 and lowest in layer 4 (Fig. 3C). Larger gs in layers 1 and 2 was at the expense of WUEi, which was highest in layers 3 and 4, under both low and high irradiance (Fig. 3H). Also, leaf temperature at high irradiance was consistently higher in layers 3 and 4, by approx. 0.5 °C, compared with layers 1 and 2 (Supplementary Fig. S8B). Contrary to large differences in overall gs, the rate of stomatal opening was relatively similar between layers, both in terms of reaching 50% and 90% of full opening, and in terms of the time constants of stomatal opening and closure (Fig. 3M, N; Supplementary Fig. S6H, I). Upon a sun-to-shade transition, stomatal closure was slowest in layer 3 (Fig. 3N); nevertheless, average gs in the first 5 min after the sun-to-shade transition scaled positively with leaf layer (Fig. 4D). Leaves in layer 1 reached larger values of dynamic Vcmax and J1500 during 20 min high irradiance exposure (Fig. 3I, J), similar to steady-state A (Fig. 1G, H). Also, Vcmax induced fastest in layer 1 and slowest in layer 3 (Fig. 3O), and there was a non-significant tendency for J1500 to induce faster in layers 1 and 2 (Fig. 3P). NPQ showed a rapid increase during photosynthetic induction, especially in layers 1 and 2 (Figs 3E, 4G). However, after high irradiance exposure for 1 h, NPQ had reached similar levels in all leaves (Fig. 3E). After a drop in irradiance, NPQ in layer 1 relaxed the fastest, and in layers 3 and 4 the slowest (Figs 3E, 4H), and these differences were associated with qZ, qT, and qI (for which layer 1 showed the lowest values; Fig. 3S, T), but not qE (Fig. 3R). Fast NPQ relaxation in upper leaf layers coincided with larger transient ΦPSII in the minutes following the sun-to-shade transition (Figs 3F, 4F), but A directly after the sun-to-shade transition did not seem to be affected by NPQ relaxation (Supplementary Fig. S3A). After 60 min of low irradiance, NPQ in layer 1 had the lowest value (~0.2) and seemed to have reached a steady state, whereas NPQ in layers 2–4 was still declining slowly and scaled with canopy layer, such that NPQ in layer 4 was ca. five times higher than in layer 1 (Fig. 3E). However, these NPQ differences had no significant effects on A at the same moment (Supplementary Fig. S3B). Photosynthetic induction after various durations of shade exposure The duration of shade exposure had strong effects on photosynthetic induction rate after re-exposure to high irradiance (Fig. 5). Especially leaves adapted to 60 min low irradiance took longer to reach 50% and 90% of full photosynthetic induction, compared with those that had been shaded 30 min or less. Shade exposure for 30 min or 60 min tended to result in slower induction in layers 2–4 compared with layer 1, but these differences were subtle. Shorter-term shade exposure (1–15 min) did not produce differences between layers. Photosynthetic induction following various shade exposure durations in four layers of a fully grown tomato canopy. (A, B) Times required to reach 50% (A) and 90% (B) of full photosynthetic induction (t50-A and t90-A) after re-exposure to high irradiance. Mean values ±SEM from six biological replicates are shown. Measured and modelled dynamic gas exchange under a sequence of lightflecks Gas exchange of shade-adapted leaves that were exposed to a lightfleck sequence was strongly affected by leaf position in the canopy. During phases of high irradiance, leaves in layer 1 quickly reached ~15 μmol m−2 s−1, those in layer 2 reached ~10 μmol m−2 s−1, and the last two layers only reached ~5 μmol m−2 s−1 (Fig. 6A). During phases of low irradiance, leaves in layer 1 exhibited the lowest A (likely due to high Rd; Fig. 1D). Also, gs and Ci were highest in layer 1, and lowest in layer 3 and 4 (Fig. 6B, C), suggesting diffusional limitations during lightflecks in bottom leaves. During the entire lightfleck sequence, A in layer 1 showed an increase of 29% from the first high irradiance phase to the final high irradiance phase. Lower layers showed even stronger increases in A: 58% (layer 2), 107% (layer 3), and 104% (layer 4; Fig. 6A). At the same time, gs in canopy layer 1 barely responded to lightflecks, increasing by only 0.42% from the first to the final high irradiance phase, whereas gs noticeably increased in layers 2 (43%), 3 (70%), and 4 (121%). Dynamic photosynthetic gas exchange under lightfleck sequence in four layers of a fully grown tomato canopy. (A, B, C) Time courses of net photosynthesis rate (A; A), stomatal conductance (gs; B), and leaf internal CO2 partial pressure (Ci; C). Dark broken lines in (A) show modelled dynamic A. Each curve represents the mean (±SEM) of six biological replicates. Grey dashed lines in box plots indicate mean values; box plots indicate 0% (maximum), 25%, 50% (median), 75%, and 100% (minimum); points show single data. To gain more insight into the limitations underlying the dynamic response of A to lightflecks, we generated a new model based on Vialet-Chabrand et al. (2021). The model was fit to data from each replicate separately, to obtain a realistic parameter distribution. At each leaf layer, modelled A closely resembled actual data (Fig. 6A), with R2=0.94-99, and a root mean squared error of 0.46–0.98 µmol m−2 s−1, between observed and modelled data (Supplementary Fig. S9). Among the model parameters (Fig. 7; Supplementary Fig. S10) describing the dynamic response of A to lightflecks, the time constants for Rubisco activation and deactivation (τVi, τVd) were most strongly affected by leaf layer: τVi was ~3 min in layers 1 and 2, and ~7 min in layers 3 and 4 (Fig. 7C), suggesting more than twice as fast Rubisco activation in the upper half of the canopy. Also, τVd was by far the smallest in layer 1 (~5 min; Fig. 7D), and substantially increased in layer 2 (~30 min), 3, and 4 (both ~50 min), suggesting much faster Rubisco deactivation in the uppermost leaf layer. Parameter values related to Rubisco of the dynamic gas exchange model under lightflecks, estimated per leaf layer. (A, B) Minimum (Vcmin) and maximum carboxylation rates (Vcmax) at low and high irradiance, respectively; (D, E) time constants of Rubisco activation (τVi) and deactivation (τVd), respectively. Error bars represent mean ±SE. One-way ANOVA was performed between four layers; different letters indicate statistically significant differences (P<0.05). We then explored to what extent lightfleck A would change if all leaf layers possessed Rubisco activation/deactivation kinetics of layer 1 (Supplementary Fig. S11) or of layer 4 (Supplementary Fig. S12), or if A in all cases responded instantaneously to increases and decreases in irradiance (Supplementary Fig. S13). We found that time-averaged A was hardly affected by either layer 1 or 4 Rubisco kinetics (Fig. 8), as there was no significant difference at any of the leaf layers between average A as predicted by the original model (‘modelled’), average A as predicted using layer 1 Rubisco kinetics (‘layer 1 kinetics’), or average A as predicted using layer 4 Rubisco kinetics (‘layer 4 kinetics’). Also, time-averaged A would increase substantially if A would respond instantaneously to irradiance, although in layer 3 this was not statistically significant (Fig. 8). The simulated increase in A under this scenario was largest in layer 1 (~45%), and ~20% in layers 2 and 4 (Fig. 8). Results of model analysis of dynamic A under lightflecks. Boxplots show time-averaged A during lightflecks (see Fig. 6A for time courses). Each panel displays (i) observed A, (ii) modelled A using the original model, (iii) modelled A assuming the same Rubisco activation and deactivation kinetics as in layer 1, and (iv) modelled A assuming its rapid (instantaneous) response to increases and decreases in irradiance. Box plots indicate 0% (maximum), 25%, 50% (median), 75%, and 100% (minimum); points show single data. One-way ANOVA was performed between four layers; different letters indicate statistically significant differences (P<0.05). Correlations between traits Traits describing dynamic responses of A (t50-A, t90-A) and gs (τi, τd) generally showed weak, non-significant correlations with each other, with steady-state A traits, leaf anatomical traits and pigment concentrations, and with DLI and R/FR (Fig. 9). Large exceptions to this observation were t50-Vcmax and t50-J1500, which both showed strong negative correlations with rate constants of NPQ induction and relaxation, all steady-state A traits, as well as the Chl a/b ratio, DLI, and R/FR. Rate constants of NPQ induction and relaxation showed strong positive correlations with steady-state A and gs traits, Chl a/b, DLI, and R/FR. Time constants of stomatal movement (τi, τd) did not correlate with either stomatal density or size. Strong positive correlations were observed among the various steady-state A traits (Asat, Rd, α, Vcmax, J1500, TPU), which additionally showed strong positive correlations with stomatal density, Chl a/b, DLI, and R/FR, and strong negative correlations with SLA. Total chlorophyll, carotenoid, and nitrogen concentrations as well as stomatal size showed only weak correlations with other traits. Correlation matrix including major physiological traits, as well as major drivers of environmental acclimation (DLI and R/FR). Significant linear correlations are shown as: *P<0.05, **P<0.01, ***P<0.001. Negative correlations are shown in red, positive correlations in blue, with darker colors signifying larger Pearson correlation indices. Abbreviations: α, apparent quantum yield; Asat, light-saturated A; [Carsx+c], carotenoid concentration; [Chla+b], total chlorophyll concentration; Chla/b, chlorophyll a:b ratio; DLI, daily light integral; gs50 and gs1500, steady-state gs at low and high irradiance; J1500, electron transport rate at 1500 μmol m−2 s−1; [N], leaf nitrogen content (g/m2); NPQ induction and relaxation, rate constants of NPQ induction and relaxation; qE, energy-dependent quenching; qZ and qT, zeaxanthin- and state transition-dependent quenching; qI, photoinhibition; R/FR, red:far-red ratio; Rd, mitochondrial respiration rate; SA, stomatal size; SD, stomatal density; SLA, specific leaf area; τi and τd, time constants of stomatal opening and closure; t50-A and t90-A, times to reach 50% and 90% of full photosynthetic induction; t50-Vcmax and t50-J1500, times to reach 50% of Vcmax and J1500 induction; TPU, triose phosphate use; Vcmax, maximum carboxylation rate. Discussion Crop canopies are a complex ensemble of leaves that differ in age and light history (both irradiance and spectrum), which affect leaf photosynthetic acclimation. While the vertical gradients in steady-state photosynthesis (A), stomatal conductance (gs), and NPQ are well characterized, their dynamic responses to rapid fluctuations in irradiance are not, even though they are crucial for predicting how field crops use natural fluctuations in irradiance. Therefore, we studied how dynamic A, gs, and NPQ differed at various heights in a fully grown tomato canopy, and used a model analysis to quantify the effects of Rubisco activation/deactivation kinetics on A under fluctuating irradiance. We found that leaves in the upper half of the canopy demonstrated an enhanced capacity to quickly up- and down-regulate A and NPQ under dynamic irradiance. Nevertheless, one set of parameters describing transient responses was sufficient to model A under fluctuating irradiance in all leaf layers. Top leaves show the fastest responses to changes in irradiance The question whether leaves at the top of the canopy respond differently to irradiance fluctuations than those at the bottom has repeatedly received attention in the last three decades (especially in woody species), but with mixed results. For instance, some researchers (Tausz et al., 2005; Bai et al., 2008) found that bottom leaves had slower photosynthetic induction (larger t90-A), while others (Durand et al., 2022) found the opposite (smaller t50-A and t90-A). Some studies also found mixed results: Urban et al. (2008) found that in four out of five species, t90-A was significantly smaller in bottom leaves, while Poorter and Oberbauer (1993) as well as Zipperlen and Press (1997) found significantly smaller t90-A in bottom leaves of one of two studied species each; other species in these three studies showed no significant effects of top or bottom leaf on t90-A. Some studies found no significant differences in the rate of photosynthetic induction between top and bottom leaves (Rijkers et al., 2000; Urban et al., 2008; Acevedo-Siaca et al., 2021b). In addition, several studies found that initial gs was reduced in bottom leaves (Urban et al., 2007, 2008; Bai et al., 2008; Durand et al., 2022), although this does not seem universal (Campany et al., 2016; Acevedo-Siaca et al., 2021a). Among the few studies that assessed NPQ kinetics between top and bottom leaves, Tausz et al. (2005) documented faster NPQ induction in bottom leaves, whereas Acevedo-Siaca (2021a) did not find significant differences. Our results suggest that leaves at the top of a fully grown tomato canopy were remarkably ‘ready for the sun’ in many aspects: apart from having the largest photosynthetic capacity and quantum efficiency (Fig. 1), they showed faster responses of A during photosynthetic induction (Fig. 3B, K, Q), largest gs (Figs 3C, 6C; Supplementary Fig. S6A, C), fastest up-regulation of Vcmax during photosynthetic induction (Fig. 3O) and Vcmax down-regulation during low irradiance phases between lightflecks (Fig. 7E) as well as the fastest up- and down-regulation of NPQ under irradiance transients (Figs 3E, 4G, 4H). Nevertheless, our model analysis suggested that leaves in layer 1 would profit most from even faster (near-instantaneous) responses of A to lightflecks (Fig. 8), likely due to their larger photosynthetic capacity that they would be able to exploit to a larger extent (than leaves in layers 2–4) if A responded instantaneously (Supplementary Fig. S13). Fast up-regulation of photochemistry and NPQ during photosynthetic induction (Fig. 3E) also likely protected upper-canopy leaves from accumulating photoinhibition under high irradiance, as the value of qI after a return to shade was lowest in these leaves (Fig. 3T). Many factors—leaf age, leaf [N], and light intensity and spectrum—may affect these observed differences in dynamic A. While in our study these factors were entangled (Fig. 2A; Supplementary Figs S4, S5), precluding conclusions as to the involvement of any specific factor, literature can provide some answers. The late phase of photosynthetic induction (i.e. t90–95-A) was slower in older tomato (Y. Zhang et al., 2022) and wheat leaves (Faralli et al., 2019), while photosynthetic induction was generally increased with higher leaf [N] across a number of species (Liu et al., 2021; Sun et al., 2022; Xiong et al., 2022). Growth light intensity overall had very mixed effects, in that it did not affect the rate of photosynthetic induction in Arabidopsis (Matthews et al., 2018), and caused faster, slower, or no differences in induction in several tree species (Han et al., 1999; Kang et al., 2022). Finally, growth under different R/FR had no discernible effects on kinetics of photosynthesis or NPQ (Ernstsen et al., 1999; Lazzarin et al., 2024). Combining these results from literature, and adding that [N] in our canopy did not show a significant decline with leaf rank (Fig. 2A), one may reason that the observed differences in dynamic A were mainly caused by differences in leaf age. Nevertheless, studies disentangling leaf age from leaf light acclimation (such as when plants are grown horizontally instead of vertically; Trouwborst et al., 2011) may shed further light on this matter. Transient limitations to dynamic photosynthesis Differences in the rate of photosynthetic induction between leaf layers partially coincided with large differences in steady-state gs (Fig. 3C; Supplementary Fig. S6), which strongly influenced the extent of transient stomatal limitation (Fig. 3G) and WUEi (Fig. 3H), but not the rate of photosynthetic induction (Fig. 9). Also, the rate of stomatal opening was relatively similar between layers (Fig. 3M), and showed no significant correlations with t50-A or t90-A (Fig. 9). These results suggest that the intracanopy variation in stomatal traits did not affect photosynthetic induction. While differences in average biochemical limitation between layers were subtle (Fig. 3D), Vcmax showed faster increases during photosynthetic induction in the upper two leaf layers (Figs 3O, 7D), and increases in J1500 showed a similar, non-significant tendency (Fig. 3P). Similar to the increases in A (Fig. 3A, K), the increases in Vcmax and J1500 tended to be faster on a relative (i.e. t50; Fig. 3O, P) as well as an absolute basis (Fig. 3I, J). This suggests that irradiance-regulated enzymes in the CBB cycle were present at greater numbers, and were able to up-regulate their activities more quickly, in upper-canopy leaves. The range of t50 values found for Vcmax and J1500 induction (1–5 min; Figs 3O, P, 7D) is well in line with those found for several wheat cultivars (Salter et al., 2019) and horticultural genotypes, including tomato (N. Zhang et al., 2022). We observed remarkable differences in the rate of NPQ relaxation between leaf layers (Figs 3E, 4H), which have been shown to affect photosynthetic efficiency and growth in the field (Zhu et al., 2004; Kromdijk et al., 2016; De Souza et al., 2022; Long et al., 2022). In our data, NPQ relaxation coincided with larger average ΦPSII (Fig. 4F), but not with increased average A (Fig. 4B), after the sun-to-shade transition. However, because of large differences in Rd (Fig. 1D), initial A at low irradiance (A50) was lower in upper-canopy leaves (Supplementary Fig. S6D). When correcting transient A after the sun-to-shade transition for this initial difference in A50, there was a tendency for transient A in layers 1 and 2 to be larger than in layers 3 and 4 (Supplementary Fig. S3A), but this was not significant (P=0.408). Thus, faster NPQ relaxation in upper-canopy leaves may have contributed to enhanced photosynthetic efficiency under fluctuating irradiance in these leaves, but our data did not show a significant effect. Lack of correlation between dynamic and steady-state photosynthesis traits We observed - more than threefold variation in DLI from the top to the bottom of the canopy, from 12 mol m−2 d−1 in layer 1 to 3.6 mol m−2 d−1 in layer 4 (Supplementary Fig. S5). At the same time, R/FR roughly halved from the top to the bottom of the canopy, changing from ~12.5 at layer 1 to ~7.5 at layer 4 (Supplementary Fig. S4B). Surprisingly, however, neither DLI nor R/FR showed strong correlations with the traits characterizing A and gs kinetics (t50-A, t90-A, τi, τd; Fig. 9). Additionally, we observed a striking lack of correlation between the traits quantifying dynamic (e.g. t50-A, t90-A) and steady-state A (Fig. 9). While this has not been studied before within canopies, many recent studies on genotypic variation have documented a similar lack of correlation between photosynthetic capacity and rate of photosynthetic induction within and between species (McAusland et al., 2016; Soleh et al., 2016; Acevedo-Siaca et al., 2020, 2021a, b; De Souza et al., 2020; N. Zhang et al., 2022). Our data thus fit this pattern, reinforcing the notion that the mechanisms that govern photosynthetic capacity and those governing rapid responses to irradiance fluctuations are different (Soleh et al., 2016). It also seems that in terms of environmental acclimation, t50-A is driven by a factor, or a number of factors, other than DLI and R/FR. There is often a strong relationship between photosynthetic capacity and [N] (Evans, 1989). However, to our surprise, [N] tended to increase from the top to the bottom of the canopy (Fig. 2A; Supplementary Fig. S7A), even though photosynthetic capacity showed the characteristic vertical decline (Fig. 1) that is often observed in a canopy (Niinemets, 2007). An explanation for this phenomenon could be that N is stored in lower leaves without directly contributing to photosynthetic capacity, as discussed in Townsend et al. (2018). Another explanation could be that LAI was never larger than ~3 in our study due to regular removal of leaves low in the canopy (Supplementary Fig. S4C), whereas in field crop and forest canopies this value can reach 6–7 (Dufrene and Breda, 1995; Gitelson et al., 2003), and the stronger self-shading inside such canopies could provide a larger driver for reductions in [N] in bottom leaves. One set of parameters was sufficient to capture canopy dynamic A under lightflecks While the modelling of leaf photosynthetic gas exchange to irradiance fluctuations is well-advanced (Pearcy et al., 1997; Kirschbaum et al., 1998; McAusland et al., 2016; Vialet-Chabrand et al., 2016, 2017; Morales et al., 2018a, b; Joubert et al., 2023), the modelling of such responses at the canopy scale is in its infancy (Wang et al., 2020). While such models are important to predict whole-canopy CO2 assimilation under field conditions, developing them raises some questions: to what extent do leaves at different locations in the canopy differ in their dynamic response to irradiance fluctuations? If there are significant differences, does each leaf rank require its own set of parameters characterizing dynamic A? In other words, how much measurement and computational effort is required to capture these responses accurately at the whole-canopy level? Our work can provide some tentative answers to these questions. Firstly, mathematical modelling of lightfleck A responses suggests that using one set of parameters is sufficient to simulate A at any leaf layer (Fig. 8). Specifically, when Rubisco activation and deactivation kinetics of either leaf layer 1 or leaf layer 4 were used to predict dynamic A responses to lightflecks in all leaf layers, time-averaged A under this scenario did not significantly differ from their measured or originally modelled responses to lightflecks (Fig. 8). Secondly, observed statistically significant differences in t50-A, IS60 and kinetics of stomatal movement were minor (Fig. 3), and no major differences in the rate of loss of photosynthetic induction were found (Fig. 5). Thirdly, when correlating A under fluctuating irradiance (first 5 min of photosynthetic induction, and average lightfleck A responses) with t50-A or with steady-state A at high irradiance (A1500; Supplementary Fig. S14), we found that time-averaged dynamic A correlated very strongly with steady-state A (P<0.01; R2=0.98–0.99), but less strongly with t50-A (P=0.011–0.13; R2=0.75–98). Altogether, these results suggest that (i) steady-state A traits have larger effects on time-integrated A under fluctuating irradiance, and (ii) one set of dynamic A parameters, for example estimated on leaf layer 1 only, sufficiently captures the dynamic A behavior of all leaves in the canopy. However, it should be noted that the contribution of each of these factors will strongly depend on characteristics such as sunfleck frequency and amplitude (Morales et al., 2018a; Durand and Robson, 2023). Work on characterizing key properties of sunflecks in nature and in protected horticulture is still in its infancy (Durand et al., 2021; van Westreenen et al., 2023). Much is still to be learned regarding what constitutes a ‘typical’ sunfleck in a given canopy, and how this affects whole-canopy A. Conclusions It is well known that leaves at the top of a canopy usually have larger photosynthetic and photoprotective capacity than leaves at the bottom of the canopy. Our research additionally suggests that in a fully grown tomato canopy, top leaves show faster responses of photosynthesis to increases and decreases in irradiance. Surprisingly, however, these faster responses do not seem to allow them to use sunflecks more efficiently than leaves lower down in the canopy, suggesting that larger carbon gain under fluctuating irradiance in top leaves is mostly achieved by a larger steady-state photosynthetic capacity rather than faster dynamic responses. Additionally, we found that non-photochemical quenching, a central photoprotective mechanism, was both up- and down-regulated more rapidly in top than in bottom leaves, thereby reducing the extent of photoinhibition during high irradiance exposure, and transiently increasing the quantum yield of photosystem II electron transport after a sun-to-shade transition. Our results contribute to the modelling of whole-canopy A under fluctuating irradiance (Wang et al., 2020), as they provide leaf layer-specific time constants for changes in A, gs, and NPQ. Crucially, while we did find differences in time constants across leaf layers, our modelling analysis suggests that one set of dynamic parameters is likely sufficient to model dynamic A in all leaves, thereby simplifying future modelling efforts of whole-canopy A under fluctuating irradiance. Supplementary data The following supplementary data are available at JXB online. Fig. S1. Spectral distributions of supplementary light. Fig. S2. Areas in the canopy for measuring light interception. Fig. S3. A in low irradiance after photosynthetic induction, corrected for differences in A before the start of photosynthetic induction. Fig. S4. Light climate in a fully grown tomato canopy. Fig. S5. Daily light integral during the experiment per leaf layer. Fig. S6. Gas exchange and chlorophyll fluorescence traits. Fig. S7. Leaf pigmentation and stomatal traits. Fig. S8. Time courses of transpiration rate (E) and leaf temperature (Tleaf) under transient irradiance changes. Fig. S9. Observed versus modelled A during lightflecks. Fig. S10. Parameter values for the dynamic gas exchange model. Fig. S11. Results of model analysis with leaf layer 1 Rubisco kinetics. Fig. S12. Results of model analysis with leaf layer 4 Rubisco kinetics. Fig. S13. Results of model analysis with instantaneous responses to lightflecks. Fig. S14. Relationships between traits of steady-state and dynamic A, and time-integrated A under fluctuating irradiance. Acknowledgements We thank Joke Oosterkamp for help with the optical microscope, and Tijmen Kerstens for help with determining greenhouse light transmissivity. Glossary Abbreviations: net photosynthetic rate maximum apparent quantum efficiency of CO2 fixation Calvin–Benson–Bassham chlorophyll fluorescence concentrations of chlorophyll a, b, their sum, and their ratio leaf internal CO2 partial pressure daily light integral day of experiment maximum quantum yield for photosystem II stomatal conductance photosynthetic induction state 60 s after irradiance increase electron transport rate at near-saturating irradiance light extinction coefficient leaf area index nitrogen concentration non-photochemical quenching photosynthetic photon flux density photosystem II electron transport efficiency energy-dependent quenching photoinhibition state transition-dependent quenching zeaxanthin-dependent quenching red:far-red ratio mitochondrial respiration rate ribulose-1,5-bisphosphate specific leaf area time required to reach 50% and 90% of full photosynthetic induction time required to reach 50% of full induction of Vcmax and J1500 time constants of stomatal opening and closure, respectively triose phosphate use maximum carboxylation rate intrinsic water use efficiency Contributor Information Bingjie Shao, Institute of Environment and Sustainable Development in Agriculture, Chinese Academy of Agricultural Sciences, Beijing, China; Horticulture and Product Physiology, Department of Plant Sciences, Wageningen University, Wageningen, The Netherlands. Yuqi Zhang, Institute of Environment and Sustainable Development in Agriculture, Chinese Academy of Agricultural Sciences, Beijing, China. Elena Vincenzi, Horticulture and Product Physiology, Department of Plant Sciences, Wageningen University, Wageningen, The Netherlands. Sarah Berman, Horticulture and Product Physiology, Department of Plant Sciences, Wageningen University, Wageningen, The Netherlands. Silvere Vialet-Chabrand, Horticulture and Product Physiology, Department of Plant Sciences, Wageningen University, Wageningen, The Netherlands. Leo F M Marcelis, Horticulture and Product Physiology, Department of Plant Sciences, Wageningen University, Wageningen, The Netherlands. Tao Li, Institute of Environment and Sustainable Development in Agriculture, Chinese Academy of Agricultural Sciences, Beijing, China. Elias Kaiser, Horticulture and Product Physiology, Department of Plant Sciences, Wageningen University, Wageningen, The Netherlands. Tracy Lawson, University of Essex, UK. Author contributions EK, TL, LFM, and BS: conceptualization; EK, TL, SVC, SB, and BS: methodology; BS, EV, and SB, investigation; BS, SVC, and YZ, formal analysis; BS: writing—original draft preparation; EK, TL, LFM, SVC, EV, and YZ: writing—review and editing; EK, TL, and LFM: supervision; EK, TL, YZ, and LFM: funding acquisition. Conflict of interest The authors declare no conflict of interest. Funding The authors gratefully acknowledge financial support from the National Natural Science Foundation of China (no. 32172654; no. 32102465), the Central Public-interest Scientific Institution Basal Research Fund (no. BSRF202201; no. BSRF202310) and Young Elite Scientists Sponsorship Program by CAST (2022QNRC001). Data availability All data supporting the findings of this study are available within the paper and within its supplementary data published online. References Associated Data Supplementary Materials Data Availability Statement All data supporting the findings of this study are available within the paper and within its supplementary data published online.