The experiment was conducted from 28 January to 12 June 2020, in a compartment (8 × 8 m) of a Venlo-type glasshouse located in Wageningen, the Netherlands (52°N, 6°E). Four growth tables were situated in the compartment. All genotypes were grown in the same compartment to avoid artifacts caused by different growth conditions. In total, 19 genotypes of six horticultural crop species were used, including two flower crops, chrysanthemum (Chrysanthemum morifolium) and rose (Rosa hybrida); two fruit vegetables, cucumber (Cucumis sativus L.) and tomato (Solanum lycopersicum L.); and two leafy vegetables, basil (Ocimum basilicum) and lettuce (Lactuca sativa L.; Table 1). Cultivars for each crop were chosen based on their commercially relevant traits as horticultural merchandise: cultivars of the flower crops differed in flower color and number, those of fruit vegetables differed in fruit size, and those of leafy vegetables differed in leaf color and texture. For basil, cucumber, lettuce, and tomato, seeds were sown in rockwool plugs (diameter: 2 cm). Following germination, the seedlings were transferred to rockwool cubes (10 × 10 cm). Chrysanthemum plants were grown in plastic pots (diameter: 14 cm) filled with potting soil. Rose plants were grown in rockwool cubes (7 × 7 cm).

For basil, cucumber, lettuce, and tomato, the seeds were sown weekly. For chrysanthemum and rose, the plants were cut back weekly at the third or fourth node, counting from the base, to allow for the formation of a new axillary bud. Two weeks after sowing seeds or cutting back plants, two plants per genotype were placed on a growth table in a grid of four rows (distance between rows: 50 cm; distance between plants within the row: 30 cm). Plant positions were randomized two times per week to minimize any effects of a heterogeneous climate in the greenhouse compartment on plant growth. Plants were placed on the growth table for 2–3 weeks (i.e., 4–5 weeks after sowing seeds or cutting back plants), after which the measurements were conducted. This protocol was repeated weekly until data of 7–9 replicates per genotype had been collected.

A mixture of high-pressure sodium lamps (600 W, Philips, Eindhoven, Netherlands) and white light-emitting diodes (LEDs) (GreenPower LED toplighting module, Signify, Eindhoven, Netherlands) were used between 02:00 and 18:00 (a photoperiod of 16 h). Lamps were switched on during the photoperiod whenever global radiation (GR) outside the greenhouse dropped below 150 W m^–2 and were switched off when GR > 250 W m^–2. Photosynthetically active radiation (PAR) from both lamp types combined was, on average, 226 ± 16 μmol m^–2 s^–1 at the canopy level (mean ± S.D.; Supplementary Figure 2). A shading screen (HARMONY 4215 O FR, Ludvig Svensson, Hellevoetsluis, Netherlands) was closed when GR > 600 W m^–2 and was opened when GR < 500 W m^–2. Day and night temperatures were set to 20 and 19°C, respectively. Relative humidity was set to 60%. Climate settings were designed to provide reasonably optimal growth conditions for all genotypes, in discussion with greenhouse cultivation experts at the Wageningen University. Average values of daily PAR (from both solar light and supplemental lamps), air temperature, relative humidity, and [CO₂] inside the greenhouse during the experiment were 241 ± 48 μmol m^–2 s^–1, 21.5 ± 1.4°C, 63 ± 6%, and 445 ± 11 ppm, respectively (mean ± SD; Supplementary Figure 3). Plants were irrigated four times per day between 7:00 and 19:00 with a customized nutrient solution suitable for all six greenhouse crops (pH: 6.3; EC: 2.2 mS cm^–1; Supplementary Table 1).

Net photosynthesis rate (A) and stomatal conductance to water vapor (g_s) were measured on the youngest fully expanded leaf, using a gas exchange system (LI-6800, Li-Cor Bioscience, Lincoln, NE, United States) equipped with a 6 cm² leaf chamber fluorometer. No correction for the leaf area was needed for any of the gas exchange measurements as the leaves always fully filled the leaf chamber. All measurements were performed at an air temperature of 23°C, relative humidity of 65%, and a flow rate of air through the system of 500 μmol s^–1. Irradiance was provided by a mixture of red (90%) and blue (10%) LEDs in the fluorometer. Before any gas exchange measurement, single plants were preconditioned to a low irradiance (ca. 50 μmol m^–2 s^–1) for 40–60 min in the greenhouse compartment, using a custom-built shading construction. The shading construction was covered by opaque plastic films and with LEDs installed at the top, which produced an irradiance at around 50 μmol m^–2 s^–1 with 90% red and 10% blue colors. Preadaptation under the shading construction for 40–60 min was applied to ensure that during subsequent gas exchange measurements, the leaves were sufficiently shade-adapted to produce comparable data.

Photosynthetic induction was measured under a range of [CO₂]: 50, 100, 250, 400, 600, 800, and 1,000 ppm. At 400 ppm of CO₂, the leaf was first exposed to a low irradiance of 50 μmol m^–2 s^–1 for 30 min in the gas exchange chamber, after which the irradiance was increased in a single step to a high level (1,000 μmol m^–2 s^–1) for an additional 30 min. A low irradiance rather than darkness was used for the initial light conditions, as in natural environments; shade-adapted leaves are often suddenly exposed to high light (due to cloud movements or wind), whereas the exposure of an entirely dark-adapted leaf to a high irradiance is unlikely in nature and greenhouses. Gas exchange data was logged every 2 s. If a steady-state A value was not reached after 30 min under high irradiance, measurements continued until A reached a steady state. For other [CO₂], the leaf was clamped into the cuvette at 50 μmol m^–2 s^–1 for 5 min, after which the irradiance was increased to 1,000 μmol m^–2 s^–1 for 15 min. All measurements ([CO₂] × genotype) were randomized. Every week, a group of new plants (one per genotype) was chosen and measurements at different [CO₂] were randomized among these plants during the day, to avoid potential diurnal effects being entangled with treatment effects. Once photosynthetic induction was measured on a given plant, the particular plant was not used for another measurement for at least 40 min, to avoid interference from previous conditions. All gas exchange measurements were done between 8:00 and 16:00 h.

Samples to measure the stomatal size and density were taken on the same leaf used for gas exchange measurements. In addition, leaf light absorptance and chlorophyll and carotenoid contents were measured.

Using a nonlinear regression with the GAUSS method in PROC NLIN of SAS (SAS Institute Inc., Cary, NC, United States), parameters of the dynamic g_s model and V_cmax kinetics during A induction were estimated. Statistical analyses were conducted using R¹. First, normality was tested using the Shapiro–Wilk test, and homogeneity was tested using Levene’s test to determine whether residuals showed equal variances. For traits that did not show equal variance, log transformation of data was applied. Differences between genotypes were detected using one-way ANOVA (p < 0.05), by taking into account, the different weeks of sowing/cutting as a block effect. When a significant difference was detected, a post hoc test was conducted for pairwise comparisons between genotypes, using Fisher’s Protected Least Significant Difference (LSD) test (p < 0.05).

The kinetics of A induction varied substantially among the 19 horticultural genotypes tested (Figures 1A–C). Variation in the key traits of photosynthesis dynamics tended to be larger between different crop species than between cultivars of the same species (Figure 2). CV of T₂₀, T₅₀, and T₉₀ was 55, 61, and 42%, respectively, while CV of average A during the first 300 s of induction (A_avg,300) was 22% (Table 2). The rate of A induction in rose was the fastest as demonstrated by small values for T₅₀ and T₉₀ in all three rose cultivars (Figures 1B, 2A,B). Chrysanthemum and lettuce, on the other hand, tended to have the highest T₉₀ values (Figure 2B) and thus showed relatively slow induction. Chrysanthemum also had high T₅₀ values, except for cv. Anastasia (CHA) had a very low T₅₀, whereas lettuce showed a relatively smaller T₅₀ (Figure 2A). Tomato and cucumber had intermediate T₅₀ and T₉₀ (Figures 2A,B). Most crops showed a relatively small variation between cultivars, except for basil, which showed relatively large variations in T₅₀ and T₉₀ for its three cultivars (Figures 2A,B).

The kinetics of the g_s response to increases of irradiance also varied substantially among genotypes (Figures 1D–F). The value of g_s in rose barely responded to an irradiance increase; hence, g_s,i and g_s,f of rose cultivars were nearly identical (Figure 1E and Supplementary Figure 4). Therefore, the parameters representing the temporal response of g_s (k, λ, and Sl_max) were not estimated for rose. Values for the CV of k, Sl_max, and λ (among the remaining 16 genotypes) were, respectively, 23, 68, and 62% (Table 2). Both tomato and cucumber tended to have fast g_s increases, as well as exhibit stomatal oscillations (Figure 1F). Lettuce had medium Sl_max, followed by chrysanthemum and basil, which had a relatively smaller Sl_max (Figure 2C). The CV of average g_s and water use efficiency in the first 5 min of A induction (g_s,avg,300 and iWUE_avg,300) were 22 and 21%, respectively, which were smaller than CV for most dynamic g_s parameters (λ and Sl_max; Table 2).

Additionally, steady-state A and g_s varied strongly among genotypes (Figures 1A–C). A_i and g_s,i had a CV of ∼20% each (Table 2). For A_i, basil had the lowest value and tomato the highest, while for g_s,i, chrysanthemum showed the lowest value and rose showed the highest (Table 2). Steady-state A and g_s at high irradiance (A_f and g_f) showed CV of 30 and 46%, respectively, with tomato showing the highest and rose the lowest A_f and g_s,f (Table 2).

Based on dynamic A vs. C_i curves, the kinetics of V_cmax and J, as well as the stomatal and nonstomatal limitations to photosynthesis during A induction, were quantified (Figure 3 and Supplementary Figure 5). Both V_cmax and J induction kinetics varied between crops and cultivars of the same crop (Figures 3A,B and Supplementary Figures 5A,B). After 15 min in high irradiance, V_cmax and J of rose were the smallest, while tomato and chrysanthemum showed higher values for final V_cmax and J (Figures 3A,B). Interestingly, tomato and lettuce showed transient drops in V_cmax and J induction during the first 3 min after exposure to high irradiance (Figures 3A,B). Variations in time constants for V_cmax induction were smaller than those describing A induction. Both τ_fast and τ_slow had CV values of 22% (Table 2). Generally, τ_fast varied between 0.5 and 1 min, with chrysanthemum showing the largest τ_fast (Figure 4A). The values of τ_slow of basil, chrysanthemum, cucumber, and lettuce were generally around 5 min, and rose had the smallest τ_slow (3.1–3.6 min; Figure 4B). Surprisingly, large variations of τ_slow between cultivars were found in basil and tomato; basil cv. Eleonora (BEL) showed significantly smaller τ_slow than the other two basil cultivars, while tomato cv. Merlice (TM) showed a significantly larger τ_slow than the other two tomato cultivars (Figure 4B).

Within the first 15 min of exposure to high irradiance, transient non-stomatal and stomatal limitations of A induction showed substantial genotypic variation, with a greater variation in the level of transient stomatal limitation than in non-stomatal limitation (Figure 3C and Supplementary Figures 5C,D). Chrysanthemum showed the largest transient stomatal limitation among all crops, which went up to 70% during the first 1–2 min of induction and remained high (up to 40%) after 15 min in high irradiance (Figure 3C). Rose hardly exhibited any transient stomatal limitation during photosynthetic induction (Figure 3C), which can be explained by its non-responsive g_s to an irradiance increase (Figure 1E). Tomato showed a fast decrease in transient stomatal limitation (from ∼50% to ∼10% in 15 min) after an irradiance increase (Figure 3C). Transient non-stomatal limitation decreased sharply in the first 4–5 min after an irradiance increase (Supplementary Figure 5C). Most genotypes showed a transient non-stomatal limitation at around 10% after 15 min in high irradiance, except for cucumber, which still had ∼20% nonstomatal limitation after 15 min of high irradiance (Supplementary Figure 5C). Some crops (basil, chrysanthemum, and tomato) also showed relatively large variations between cultivars for transient stomatal and nonstomatal limitations (Supplementary Figures 5C,D).

Stomatal density and size showed large CV, especially adaxially, and this was partly due to the fact that the rose had no stomata at the adaxial side (Table 2). The CV of stomatal density and size at the leaf abaxial side were, respectively, 78 and 57% (Table 2). Generally, large variation in the stomatal density and size occurred between crop species, while the variation between cultivars was relatively small (Figures 5A,B). Chrysanthemum had the largest, and cucumber had the smallest stomata (Figure 5A). Both chrysanthemum and lettuce had low stomatal density, while cucumber had the highest stomatal density (Figure 5B). These large variations in the stomatal density and size resulted in large variation in theoretical maximum stomatal conductance (g_s,max): the CV of g_s,max was 50%, with cucumber showing the highest g_s,max (up to ∼5 mol m^–2 s^–1), followed by tomato (up to ∼4 mol m^–2 s^–1), and lettuce, having the lowest g_s,max (∼1 mol m^–2 s^–1; Figure 5C and Table 2).

Using values of g_s,max and observed g_s during photosynthetic induction (Figures 1D–F), absolute pore area opening was calculated. Kinetics of absolute pore area opening during A induction varied substantially between crops (Figures 6A–C). In cucumber leaves, individual pore area was found to be increased from ∼1 to ∼3 μm² after 30 min in high irradiance, resulting in an increase of g_s from about 0.1 to 0.3 mol m^–2 s^–1 (Figure 6C and Supplementary Figure 4). The pore area of tomato was calculated to increase more strongly, from ∼4 to ∼16 μm², leading to a g_s increase from about 0.1 to 0.4 mol m^–2 s^–1 (Figure 6C and Supplementary Figure 4). In contrast, the pore area of chrysanthemum and lettuce required a larger extent of opening to achieve a comparable g_s increase with cucumber and tomato from ∼18 to ∼36 μm² in chrysanthemum and from ∼10 μm² to ∼29 μm² in lettuce (Figures 6A–C).

Surprisingly, when calculating the percentage of pore area opening relative to the maximum pore area, variation between crops was much smaller than for other traits (Figures 6D–F). During photosynthetic induction, all genotypes opened their stomata by less than 10% of the theoretical maximum pore area (calculated from pore length; Figures 6D–F). For example, the absolute pore area after 30 min in high irradiance reached ∼40 μm² in chrysanthemum (which was the largest among all crops), which only accounted for 3–5% of the maximum pore area (Figures 6B,E). The pore area of cucumber only reached ∼3 μm² after 30 min in high irradiance, which was also ∼3% of the maximum pore area (Figures 6C,F).

Leaf pigment concentrations showed relatively less variation among genotypes compared with most stomatal traits, with a CV of 25–30% for chlorophyll and carotenoid contents (Table 2). Since leaf color differed between cultivars (e.g., purple leaves in BR and brownish leaves in LGI), pigment types varied between crop species and cultivars (Supplementary Figures 6A,B). The chlorophyll a:b ratio showed little genotypic variation (CV: 7%), and an average of 2.7 across genotypes (Table 2). Leaf light absorptance was even more conserved, with a CV of 5%, the lowest value among all traits (Table 2). Small but significant differences in leaf light absorptance occurred between crop species, whereas variations between cultivars were not found, except for basil and lettuce which had cultivars (BR and LGI) with different leaf colors (Supplementary Figure 6C).

Generally, steady-state gas exchange traits correlated well with one another (e.g., g_s,f vs. A_f, V_mf vs. A_f), as did dynamic traits (e.g., k vs. T₉₀; Figure 7). Some steady-state traits also correlated well with dynamic traits (e.g., Sl_max vs. g_s,f; Figure 7). Importantly, we identified key traits that showed strong correlations with indicators of the rate of photosynthetic induction (i.e., T₂₀, T_50, or T₉₀); these key traits were relevant to either stomata and their rate of movement (g_s,i and k) or Rubisco activation (f, τ_slow, and V_mf) (Figure 8). Furthermore, these traits represented either dynamic (f, τ_slow, or k) or steady-state traits (g_s,i or V_mf), suggesting that both types of the trait were relevant for the rate of photosynthetic induction. The value g_s,i correlated negatively with T₂₀ and T₅₀ (Figures 8A,C). T₂₀ was also correlated with f, and T₅₀ was correlated with τ_slow (Figures 8B,D). Both k and V_mf were positively correlated with T₉₀, and k and T₉₀ showed an especially strong linear correlation (Figures 8E,F). Given the strong correlations between photosynthetic induction traits (T₂₀, T₅₀, and T₉₀) and stomatal parameters (g_s,i or k), we further tested whether stomatal conductance-related parameters were correlated with traits characterizing stomatal anatomy (stomatal size and density). The stomatal size was not correlated with either g_s,i, k, or Sl_max, but was negatively correlated with stomatal density across species (except for chrysanthemum, which had large stomates; Supplementary Figures 7A–D). Interestingly, there was a very strong linear correlation between stomatal size on the abaxial side with that on the adaxial side of the leaf (Supplementary Figure 7E), with stomatal size on the adaxial leaf surface being ∼93% of the size on the abaxial leaf surface for all species except for rose.

Increasing the rate of photosynthesis is expected to increase crop yields (Ort et al., 2015; Simkin et al., 2019). Although the harvested product for horticultural crops can be very different from staple food crops, e.g., fresh flowers, fruits, and flavor additives, biomass production (thus photosynthesis) remains the basis for high yield and good product quality. For example, flower number was positively correlated with plant dry weight in chrysanthemum (Carvalho and Heuvelink, 2003), and extra assimilates contributed by the canopy improved the stem quality of cut-rose (Zhang et al., 2020). Therefore, increasing photosynthesis is important for optimizing horticultural crop production, especially where growth in most crops can be assumed to be a source rather than sink limited for most of the production season (Marcelis, 1994; Li et al., 2015).

Natural genetic variation is an important resource for breeding. The genotypic variation of photosynthesis has been examined widely regarding its steady-state traits (Flood et al., 2011). However, steady-state photosynthesis does not provide an accurate representation of operating photosynthesis under fluctuating light (which often happens in the field and greenhouses, e.g., Supplementary Figure 1), given that time constants of induction/relaxation of photosynthesis reduce the time-integrated rate of carbon fixation (Kromdijk et al., 2016; Morales et al., 2018). Speeding up photosynthetic induction has been suggested as an important breeding target (Tanaka et al., 2019; Qu et al., 2020). Large genotypic variation of photosynthetic induction has been found in field crops (e.g., rice, wheat, soybean, and cassava) and woody species in forestry systems, proving that breeding for improving the dynamic crop photosynthesis is feasible (Valladares et al., 1997; Soleh et al., 2017; Salter et al., 2019; Acevedo-Siaca et al., 2020; De Souza et al., 2020). Here, we show that a large genotypic variation in photosynthetic induction also exists between major horticultural crops, and generally this variation for dynamic traits is larger than the variation for steady-state traits. Also, variation between crops in photosynthetic induction was generally larger than the variation between cultivars of the same crop.

Variation of photosynthetic induction in these 19 horticultural genotypes was quantified under near-optimal conditions, i.e., climate control management in the greenhouse was done similarly as in commercial greenhouse production. This is similar to other studies that aimed to quantify genotypic variation in crops, such as rice and cassava (De Souza et al., 2020; Acevedo-Siaca et al., 2021). However, abiotic stresses often occur during crop growth, not only in the field but also in low-tech greenhouses. When testing a genotypic variation of steady-state photosynthetic traits under both well-watered and drought conditions, drought accounted for a larger proportion of total variation compared with the genotypic variation (Gu et al., 2012). Genotypic variation of dynamic photosynthetic traits could potentially be coupled with variations induced by environmental fluctuations other than irradiance. For example, genotypic variation in intrinsic water-use efficiency found in our and other studies (Acevedo-Siaca et al., 2021) could lead to different crop performance between genotypes when drought occurs. Additionally, a recent study suggested that taking into account photosynthetic induction effects led to a reduction of 2–7% in the estimation of daily carbon gain (Murakami and Jishi, 2021), which is much smaller than the estimation error predicted by earlier studies (Naumburg and Ellsworth, 2002; Taylor and Long, 2017) and in real measurements (Adachi et al., 2019). The patterns of irradiance fluctuations appear to be very important in determining the discrepancy between simulating daily carbon gain with and without the effects of photosynthetic induction (Murakami and Jishi, 2021). However, only few studies quantified irradiance fluctuations in greenhouses at the relevant time scales (van Westreenen et al., 2020), hampering such estimations for the greenhouse production context. Moreover, previous irradiances potentially affect photosynthetic induction responses to the upcoming irradiance (Jackson et al., 1991; Kaiser et al., 2017). Further studies are needed to quantify the genotypic variation of dynamic photosynthesis under stress conditions and to evaluate their importance in different irradiance fluctuation patterns under greenhouse conditions with considering photosynthetic induction rates across different irradiances.

Photosynthetic induction is mainly regulated by three transient limitations: RuBP regeneration, Rubisco activation, and stomatal opening (Pearcy, 1953). We found large (CV up to 68%, Table 2) genotypic variation in the kinetics of stomatal responses to an irradiance increase, compared to the genotypic variation in the other two limitations. A large variation in the stomatal opening time was also found across 15 vascular plants including fern, gymnosperm, and angiosperm species (Deans et al., 2019), indicating that strong genotypic variation of stomatal response kinetics exists in many species. In our study, Sl_max was 0–0.3 μmol m^–2 s^–2, and values for k varied between 8 and 16 min (Table 2). These values of Slmax and k are within the range of those found for other species that had partially grown and evolved outdoors (McAusland et al., 2016), suggesting that the specific indoor growth conditions horticultural crops experienced do not influence the rapidity of stomatal opening. Faster stomatal opening tends to speed up photosynthetic induction (Shimadzu et al., 2019; Yamori et al., 2020), and our results showed a strong linear correlation between k and T₉₀ (Figure 8E), indicating that genotypes that require less time to open their stomata reach full photosynthetic induction faster. The strong correlation between the time constants of stomatal opening and the time to approach full photosynthetic induction also suggests that stomatal effects are typically the major ones left in the later phase of photosynthetic induction. Moreover, a higher initial g_s before an irradiance increase led to a faster speed of photosynthetic induction (Figures 8A,C), which is confirmatory of many previous studies (Soleh et al., 2017; Kaiser et al., 2020; Sakoda et al., 2020). These results highlight the importance of stomatal traits to explain the variations in photosynthetic induction, not only between genotypes of the same crop but also among different crops.

The speed of the stomatal response to environmental changes is generally considered to be related to the stomatal size (Hetherington and Woodward, 2003; Raven, 2014). A negative correlation exists between the stomatal size and the speed of g_s increase upon an irradiance increase, which has been found in many species (Drake et al., 2013; Kardiman and Ræbild, 2018). In addition, the relationship between average pore aperture and g_s is nonlinear (Kaiser and Kappen, 2000, 2001), which means that similar stomatal opening responses could result in different g_s kinetics, depending on the anatomical features of the stomatal complex in a given species. This nonlinear change in scale can also result in different time constants (e.g., time to reach 50% of the total variation) for the kinetic of pore aperture compared to g_s. We found that the pore area of cucumber and tomato (which had relatively small stomates) tended to reach a plateau earlier after an irradiance increase than that in chrysanthemum (which had relatively large stomates) (Figures 5A, 6B,C). The time needed to reach 50% of the final pore area in high irradiance was found to be higher in chrysanthemum than in cucumber and tomato (Supplementary Figure 8), suggesting that horticultural species with larger stomates require more time to open their stomata. However, this does not necessarily lead to a close correlation between parameters of g_s kinetics (k and Sl_max) and stomatal size (Supplementary Figures 7B,C). This could be due to the fact that stomatal density also determines g_s, and the range of genotypic variation in these traits may also be too small to identify the correlations. Both tomato and cucumber showed large absolute changes in g_s for low- and high irradiance adapted leaves (Supplementary Figure 4). However, changes in absolute pore area for low- and high-irradiance adapted leaves in cucumber and tomato were rather small, compared to other crops (Figure 6C). This could result from relatively high stomatal density in cucumber and tomato (Figure 5B), magnifying small changes in an individual pore area. In the hypothetical situation of a cucumber leaf having a low stomatal density, such as that of chrysanthemum, stomata in this cucumber leaf would need to open their individual pore area up to ∼40 μm² to achieve the observed increase in g_s (Supplementary Figure 9A). In contrast, the pore area of chrysanthemum substantially increased after exposure to high irradiance, but due to a low stomatal density, this did not lead to a large increase in g_s (Figures 5B, 6B and Supplementary Figure 4). When using the hypothetical situation of a chrysanthemum leaf possessing the stomatal density of a cucumber leaf, stomata in the chrysanthemum leaf only needed to open to a very small extent (∼3 μm²) to achieve the observed g_s increase (Supplementary Figure 9B). These results suggest that species having small but many stomates are more efficient in adjusting g_s to changes in irradiance, as it only requires small changes in individual pores to achieve large changes in g_s.

Interestingly, the actual pore area opening generally accounted for less than 10% of the theoretical maximum pore area in all genotypes (Figures 6D–F), resulting in an average ratio between g_s,f and g_s,max (determined by anatomical traits) of 0.1 across genotypes (Supplementary Figure 10). This average g_s,f/g_s,max ratio among horticultural crops is generally lower than what has been found in previous studies (McElwain et al., 2016; Murray et al., 2020). In a modeling study, Dow et al. (2014) predicted an optimal ratio between operating g_s and anatomical g_s,max of 0.2; their study suggested that at 20% operating capacity, guard cells could increase the pore size efficiently when favorable conditions persisted, but could also close the pore just as quickly under stress (Dow et al., 2014). While experimental studies on the g_s-g_s,max relationship across species are scarce, some have described a relatively constant ratio of 0.25 between operating g_s and g_s,max in shrub and tree species (McElwain et al., 2016; Murray et al., 2020). Our results suggest that for horticultural crops, operating g_s at 10% of its maximum capacity may be already sufficient for guard cells to function efficiently.

The initial, fast phase of photosynthetic induction involves the availability of RuBP and other Calvin cycle intermediates and is assumed to last 1–2 min (Pearcy, 1953; Sassenrath-Cole and Pearcy, 1992). The time constant for the fast phase of V_cmax induction (τ_fast; Figure 4A) may indicate the speed of completing the initial phase of photosynthetic induction, but this assumption needs to be verified using Calvin cycle metabolomics studies. Here, τ_fast varied between 0.5 and 1.1 min, which is generally larger than what has been found across wheat cultivars (0.3–0.5 min; Salter et al., 2019). This may suggest higher activities and/or amounts of fructose-1,6-bisphosphatase (FBPase), sedoheptulose-1,7-bisphosphatase (SBPase), and phosphoribulokinase (PRK) in field agronomic crops than in horticultural crops, given that the activation of RuBP regeneration is mainly limited by these three enzymes (reviewed by Kaiser et al., 2018). Nevertheless, our results confirm that the time needed to complete the initial phase of photosynthetic induction is generally rapid, and this is especially true when the leaf was adapted to low irradiance instead of darkness before switching to a high irradiance (Kaiser et al., 2017), as was the case in this study.

The following, slow phase of photosynthetic induction involves light-dependent activation of Rubisco by Rubisco activase, and this phase seems to show more variation between species: time constants of 4–5 min were reported for Alocasia macrorrhiza and Spinacia oleracea, and 2–4 min for wheat (Pearcy, 1953; Salter et al., 2019). For the horticultural genotypes examined here, we found slightly larger time constants (τ_slow) of 3–7 min (Figure 4B). The rate of Rubisco activation has been found to be an important determinant of photosynthetic induction in many species (e.g., wheat and soybean) (Soleh et al., 2017; Salter et al., 2019). However, we found a relatively smaller variation in Rubisco activation rate compared to variations found in many other traits. A CV of 22% was found for τ_slow, which was less than the CV of photosynthetic induction (e.g., 61% for T₅₀) and traits related to stomatal opening (e.g., 68% for Sl_max; Table 2). This corresponds with the findings of Deans et al. (2019) who found that the biochemical activation response time (5–25 min) was much more conserved between species (including angiosperms, ferns, and gymnosperms) than the time required for stomatal opening (10–150 min). It is worth noting that although the dynamic A vs. C_i approach has been used in many studies to quantify V_cmax kinetics during photosynthetic induction (Soleh et al., 2016; Taylor and Long, 2017; Salter et al., 2019; De Souza et al., 2020), the original FvCB model describes steady-state photosynthesis. By applying the FvCB model on dynamic A vs. C_i, it was assumed that the slow A induction changes are mainly caused by Rubisco activation. Although the role of Rubisco activation during A induction has been verified experimentally (Taylor et al., 2022), other processes, such as changes in mesophyll conductance could also play a role during A induction (Liu et al., 2021; Sakoda et al., 2021). However, mesophyll conductance changes have been suggested to be far more rapid than the observed V_cmax kinetics presented here, and the relative importance of mesophyll conductance for A induction is still under debate (De Souza et al., 2020; Liu et al., 2021; Sakoda et al., 2021). We conclude that the variation in Rubisco activation kinetics among the six horticultural crops may not be the primary cause for the large variation found in photosynthetic induction.

In some species (chrysanthemum, lettuce, and tomato), photosynthetic induction in the first 1–2 min exhibited a transient drop when photosynthetic induction was measured under high CO₂ (>600 ppm; Supplementary Figure 11). This is likely caused by a limited amount of inorganic phosphate (Pi) in the metabolite pool of the Calvin cycle, due to insufficient and slow activation of sucrose-phosphate synthase (SPS) during the initial phase of the light increase (Stitt and Quick, 1989; Huber and Huber, 1992). Supposedly, during the first 1–2 min of the irradiance increase, the amount of free Pi is sufficient to support photosynthesis independently of any end-product synthesis. However, once Pi is exhausted, photosynthesis is inhibited until the conversion of triose-phosphates to sucrose in the cytosol releases enough Pi, which can then be translocated back into the chloroplast (Stitt and Quick, 1989). The activation of SPS is regulated by irradiance in some species (e.g., barley and maize) but not in others (e.g., soybean, tobacco, and cucumber) (Huber et al., 1989), leading to species variations in the level of Pi limitation. This may explain why in our results, the transient drop of photosynthesis in high CO₂ was seen in some species only (Supplementary Figure 11).

We showed that in major horticultural crops, transient limitations to photosynthetic induction appeared to be species-dependent, but the general trend was that there was a large genotypic variation in the level of transient stomatal limitation, whereas the extent of transient non-stomatal limitation during photosynthetic induction was relatively conserved (Supplementary Figures 5C,D). Previous studies showed that in rice, the primary transient limitation was biochemical, whereas, in cassava, primary limitations were caused by stomata (Yamori et al., 2012; De Souza et al., 2020). For horticultural species, photosynthesis transients of some crops (e.g., cucumber) tended to be limited by biochemistry and stomata to a comparative extent, whereas those in other crops (e.g., lettuce and chrysanthemum) tended to be more strongly limited by stomata (Supplementary Figures 5C,D). Stomatal size may partially regulate the level of stomatal limitation during photosynthesis induction. Species (e.g., rice) with smaller stomata have been found to show a low level of stomatal limitation (Acevedo-Siaca et al., 2020). In our study, chrysanthemum, which had the largest stomata among the tested greenhouse crops, showed the highest level of transient stomatal limitation (Figures 3C, 5A and Supplementary Figure 5D). This is possibly due to the fact that larger stomata need more time to open until a new steady state has reached (Drake et al., 2013; also refer to Figures 6A–C and Supplementary Figure 8), resulting in a higher level of transient stomatal limitation during photosynthetic induction.

Species with small stomata displayed high stomatal density, which in the case of incomplete stomatal closure may lead to high transpiration and increased water demand (e.g., during the night). Reduced stomatal density improves drought tolerance in species, such as rice and barley (Hughes et al., 2017; Caine et al., 2019). We found that the two cut-flowers have relatively low total stomatal density (including both leaf surfaces), which possibly favors vase life by increasing water conservation, such as in other cut-flowers (e.g., Antirrhinum majus L., Schroeder and Stimart, 2005). Altogether, these results suggest that manipulating stomatal traits rather than biochemical traits is more relevant for horticultural crop breeding.

Additionally, we found a highly conserved ratio between stomatal size at the abaxial and adaxial leaf surface, as well as between the stomatal densities on both leaf sides in all crops, except for rose (Supplementary Figures 7E,F). Stomatal size and density at the adaxial leaf surface were respectively 93 and 71% of the size and density at the abaxial leaf surface. A linear correlation between the stomatal densities of both leaf sides has previously been found in rice and tomato, with more stomata on the abaxial leaf surface (Fanourakis et al., 2015; Zhang et al., 2019). The distribution of stomatal density between the two leaf sides is relevant for total g_s partitioning between leaf sides (Ticha, 1982). A more uniform g_s partitioning favors CO₂ diffusion inside the leaf, and therefore gas exchange (Parkhurst and Mott, 1990; Muir et al., 2014), but may come at the expense of stress resilience in the field. Milla et al. (2013) found that wild species showed a larger difference in stomatal density between leaf sides, while domestication tended to reduce the difference of stomatal density between leaf sides, by lowering the stomatal density at the abaxial side. Interestingly, wild but not domesticated tomato genotypes showed even stomatal distribution between leaf sides (Koenig et al., 2013; Fanourakis et al., 2015). Given the potential effects of g_s partitioning between leaf sides on gas exchange, further studies are needed to explore whether or not a uniform g_s partitioning favors photosynthetic induction and the underlying mechanisms that regulate the distribution of stomatal density between leaf sides for breeding.