In vivo detection of spectral reflectance changes associated with regulated heat dissipation mechanisms complements fluorescence quantum efficiency in early stress diagnosis

Summary

• Early stress detection of crops requires a thorough understanding of the signals showing the very first symptoms of the alterations in the photosynthetic light reactions. Detection of the activation of the regulated heat dissipation mechanism is crucial to complement passively induced fluorescence to resolve ambuiguities in energy partitioning.

• Using leaf spectroscopy, we evaluated the capability of pigment spectral unmixing to calculate the fluorescence quantum efficiency (FQE) and simultaneously retrieve fast absorption changes in a drought and nitrogen deficiency experiment with tomato. In addition, active fluorescence measurements and pigment analyses of xanthophylls, carotenes and chlorophylls were conducted.

• We observed notable responses in noninvasive proximal sensing‐retrieved FQE values under stress, but as expected, these alone were not enough to identify the constraints in photosynthetic efficiency. Reflectance‐based detection of the 535‐nm peak absorption change was able to complement FQE and indicate the activation of regulated heat dissipation for both stress treatments under growing light conditions. However, further complexity in the light harvesting energy regulation needs to be accounted for when considering additional light stress.

• Our results underscore the potential of complementary in vivo quantitative spectroscopy‐based products in the early and nondestructive stress diagnosis of plants, marking the path for further applications.

Introduction

With the escalating global population projected to exceed 9 billion by 2050 (Dhankher & Foyer, 2018), it is imperative to develop forward‐thinking approaches to early warning signals of plant performance in agricultural systems. In particular, the availability of nutrients and water, two of the most critical resources (Tyner & Webb, 1946; Boisvenue & Running, 2006), requires efficient use and an early detection of their deficiencies.

Dealing not only with the applicability to large monitoring scales but also with the complexity of various potential triggers, the early stress detection of crops could greatly benefit from the understanding of the signals that show the very early alterations in the photosynthetic light reactions (Peñuelas & Filella, 1998; Porcar‐Castell et al., 2021). The quick acclimations of plants to excessive energy under the given environmental conditions are steered by the photon‐trapping photosynthetic antennae, which are flexible in terms of adaptation of their stoichiometry, structure and connectivity (Vetoshkina et al., 2023). Both in the short term and in the long term, the antenna can respond to assure an efficient distribution of the absorbed energy and a downregulation of the energy channelled towards photosynthesis (Bag, 2021; Bassi & Dall'Osto, 2021; Ruban & Wilson, 2021). The most important short‐term reversible photoprotective process is the so‐called nonphotochemical energy quenching (NPQ) of Chl fluorescence, that is the de‐excitation of singlet excited states (1Chl*) reducing the excitation pressure on the reaction centra (Lambrev et al., 2012). The fast NPQ mechanism is often referred to as the ‘feedback de‐excitation’ or the ‘energy‐dependent’ quenching component (qE) of NPQ since it is primarily induced by the high pH gradient across the thylakoid membrane and responds to fluctuations in light intensity in question of seconds to minutes (Ruban et al., 2012). But the mechanistic processes behind qE remain difficult to disentangle as several interrelated processes occur on a short timescale (from picoseconds to minutes) in the thylakoid membrane involving both chemical and structural changes in the antenna (Betterle et al., 2009; Zaks et al., 2012). The key role in chemical changes is undoubtedly the xanthophyll (Xan) cycle (Jahns & Holzwarth, 2012; Ruban, 2016). The violaxanthin de‐epoxidase (VDE) enzyme catalyses the conversion of violaxanthin (Vio) to antheraxanthin (Ant) and zeaxanthin (Zea) under high light while zeaxanthin epoxidase (ZE) catalyses the reverse reaction under low light, giving rise to the so‐called VAZ cycle. Similarly, the accumulation of Lutein (Lut) formed out of lutein‐epoxy (Lx), known as the LxL cycle is induced in the case of several species (García‐Plazaola et al., 2007; García‐Plazaola et al., 2012). The explicit role of Xans in the intramolecular deactivation of the excitation energy (dynamic quenching) and the molecular mechanism of the regulated heat dissipation mechanism remains, however, discussed. A role as direct quencher of singlet excited Chl has been raised for Ant (Gilmore & Yamamoto, 1993, 2001; Gilmore et al., 1994), Zea (Niyogi et al., 1998; Holt et al., 2005) and Lut (Li et al., 2009; Leuenberger et al., 2017), but has also been questioned by others (Dreuw & Wormit, 2008; Gray et al., 2022). To induce the Xan cycle, it is generally considered that lumenal VDE intrinsically requires ascorbate and a low pH (Gilmore & Yamamoto, 1992; Pfundel et al., 1994). However, experiments have shown that xanthophyll de‐epoxidation, paralleled by a loss of photochemical efficiency, can take place in the total absence of light for intact vascular plants, questioning the dependency of a light‐driven intrathylakoidal pH gradient (Fernández‐Marín et al., 2009, 2021). Other abiotic stresses, such as heat applied under darkness, also have been demonstrated to trigger the accumulation of Zea (Streb et al., 2003). The pH gradient has been further considered as the necessary trigger for conformational changes in the PsbS protein that lead to the development of the regulated thermal energy dissipation (Li et al., 2000; Jahns et al., 2009). Hence, despite that the molecular mechanism of the fast activation of the regulated heat dissipation remains unsolved, most of the current evidence points to the involvement of different xanthophyll‐bound conformational states in the antenna (Johnson et al., 2009; Ilioaia et al., 2011; Liguori et al., 2015, 2017; Balevičius et al., 2017; Li et al., 2021).

Spectroscopy‐based methods, either in vitro by ultrafast transient absorption spectroscopy or in vivo by leaf spectroscopy, have probed various absorption difference spectra with the intention of disentangling the quencher and potential energy transfer or quenching mechanisms. Observing and describing the natural dynamic effects in whole leaves has the advantage of upscaling the mechanisms at larger monitoring scale, but also in vivo the origin of the observed absorbance changes remains debated. Light‐induced absorbance changes in vivo c. 535 nm have been used to monitor the reversible qE activation (Ruban et al., 1993; Bilger & Björkman, 1994), while also an electrochromic shift of carotenoids c. 515 nm driven by the electric field generated by the electrochemical proton gradient (Kramer et al., 1999; Bailleul et al., 2010; Klughammer et al., 2013) is described as mechanism in the same spectral range. Hence, the underlying cause‐effects of several absorbance increases c. 515–535 nm remain discussed, mainly because the changes envelop several effects at various timescales. For example, from the kinetic trends of the quick 500–600 nm absorbance changes, different underlying causes have been suggested. Ultrafast changes in the millisecond to second time range were on the one hand shown to evidence several phases linked to the different speeds in PSI and PSII charge separations, followed by the activity of cytochrome b ₆ f (Joliot & Delosme, 1974). On the second to minute scale, on the other hand, kinetical behaviour of the absorbance changes were modelled by three distinct Gaussians, coinciding with the chemical Ant and Zea formation (30 s–1 min), and followed by an additional amplified absorbance effect (1–3 min) of the Ant‐related contribution c. 535 nm (Van Wittenberghe et al., 2021a). It is the latter effect, built up with a delay in respect to the earlier Chl fluorescence quenching and Ant formation, which is suggested by the authors as the major impact on the total absorbance increase in vivo, in the order of a few %. This amplified absorbance effect may hence also be seen or derived from the reflectance signal under steady‐state conditions, decoupled from earlier xanthophyll formation or fluorescence quenching.

Pulse amplitude‐modulated (PAM) Chla fluorescence (ChlF) measurements in combination with saturating pulses have been the standard way to monitor the relaxed (‘unquenched’) and actual antenna states in vivo. However, nonphotochemical quenching of fluorescence derived from the information of saturating pulses (F‐based NPQ, hereafter called NPQ_F) contains other mechanisms besides the fast NPQ component (qE). Slower quenching effects related to the accumulation of Zea (qZ) and other processes such as chloroplast motion (qM) (Brugnoli & Björkman, 1992) and antenna‐regulated state transitions (qT) (Krause & Jahns, 2004) may affect also the NPQ_F parameter, but not necessarily increase the regulated thermal dissipation of excess excitation energy. This adds further complexity to the disentangling of energy redistribution within the antenna (e.g. through chemical conversions), absorbed light avoidance by the chloroplasts and antenna‐regulated heat dissipation, only based on the NPQ_F parameter.

Hence, it is evident that fluorescence measurements should be complemented with additional information on the antenna‐driven energy dissipation mechanisms and absorption mechanisms, also in the case of passive ChlF emission. Retrieval of passively emitted ChlF from vegetation, using certain absorption bands in the surface solar irradiance spectrum or using filtered leaf surface radiance, is commonly referred to as sun‐induced fluorescence (SIF) when the sun is used as the light source (Meroni et al., 2009; Cendrero‐Mateo et al., 2015; Mohammed et al., 2019). Fundamentally, the ChlF or SIF signal, here defined as the total photon emission in the 650–850 nm region, cannot be used as a direct indicator of the energy partitioning yet (Porcar‐Castell et al., 2014). First, the normalization of surface‐emitted fluorescence energy is required, that is through the calculation of the fluorescence quantum efficiency (FQE), defined as the ratio of emitted fluorescent photons and the flux of photons absorbed by Chla, or APAR Chla (Van Wittenberghe et al., 2021b). Theoretically, a quantum efficiency can range from 0 to 1, but in practice, leaf‐determined FQE is of a very small order due to the strong reabsorption by the high density of Chls in the photosynthetic tissues (Agati et al., 1993; Pedrós et al., 2008; Frankenberg & Berry, 2018). Further, the missing information on regulated heat dissipation needs to be provided, as photosynthetic quantum yield and fluorescence quantum yield are principally not always linearly related, and certainly not under stress (Rosema et al., 1998; Porcar‐Castell et al., 2014; van der Tol et al., 2014; Moreno, 2021). For this purpose, the 500–600‐nm absorbance changes have been used in remote sensing studies through the photochemical reflectance index (PRI) (Gamon et al., 1990; Gamon & Surfus, 1999). The changes in PRI can be strongly related to qE in the short term (Ripullone et al., 2011) but are also affected by long‐term changes in the pigment pools due to spectral overlap with the Chls (Wong & Gamon, 2015; Moncholi‐Estornell et al., 2022). With the goal to retrieve the regulated dissipation by means of direct changes in spectral absorption properties, independently of spectral overlap, a 535‐nm feature was recently included in a spectral unmixing method for pigment absorptions (Van Wittenberghe et al., 2024). However, it remains to be tested if such an absorption unmixing protocol can: detect the activation of the fast regulated heat dissipation; and, tackle the ambiguity in the fluorescence‐photosynthesis relationship, currently an issue in remote sensing applications. To achieve this, we applied a spectral endmember unmixing approach to obtain simultaneously FQE and the 535‐nm absorption feature under drought and nitrogen deficiency, in a time‐course experiment with tomato (Solanum lycopersicum L.). FQE is expected to first decrease and then increase depending on the stress phase, while a 535‐nm absorption peak component is expected to increase for both treatments. Our further goal was to investigate the diagnosis of the early stress and recovery through the observed FQE and 535‐nm absorption peak changes. Our hypothesis is that, if 535‐nm absorption, indicative for the fast component of regulated heat dissipation, can be detected under early stress, this information is able to complement FQE in the remotely sensed diagnosis of photosynthetic downregulation.

Materials and Methods

Experimental design

A total of 156 tomato (S. lycopersicum L. cv Moneymaker) plants were grown in pots with 80% expanded clay (Arlita™, Madrid, Spain) – 20% coconut fibre substrate inside a culture chamber with a temperature regime of 25°C : 18°C (day : night) and a photoperiod of 16 h : 8 h (PAR 300 μmol m⁻² s⁻¹, LED lights L28 NS12, Valoya), irrigated every 2–3 d with half‐strength Hoagland n°2 solution. After 38 d from seed sowing, the treatments started (Day 0, d0). Apart from a control treatment, maintaining the initial irrigation regime and nitrogen supply (8 mM), a drought stress and nitrogen deficiency stress treatments were initiated at this point for 84 plants each. For the drought treatment, irrigation was stopped, while N was applied at 0.5 mM for the N deficiency treatment (N deficit). The stress experiment lasted 7 d. At the end of Day 7 (d7), all plants were returned to the initial irrigation and fertilization regime (recovery phase, until d14).

During the stress phase, point measurements and sampling took place on Days 0, 2 (d2), 4 (d4) and 7. At each time point, 12 plants per treatment were set aside with three labelled leaves per plant. Active fluorescence and gas exchange were initially measured, followed by passive fluorescence, reflectance and transmittance measurements on the same leaves according to the protocols further described later. After that, samples for pigment analysis were taken for three plants per each treatment. Finally, all leaf material (12 plants per treatment) was further processed and collected for molecular analysis. In total, 36 plants were processed each measuring day.

After 7 d, all plants were irrigated with the control solution (8 mM N), starting the recovery phase. After 7 d of recovery treatment, all measurement and sampling protocols were repeated at d14.

To assess the responses of the plants to both nitrogen and water shortage, biomass parameters were measured during the experiment. Fresh and dry weight (FW and DW) of the aboveground organs (stems and leaves) were determined at each sampling date. In addition, the total content of elemental N (%DW) was determined in the first fully developed leaf (4^th leaf from the apex) with a CHN elemental analyzer at the CEBAS Laboratory of Ionomics (CEBAS‐CISIC, Murcia, Spain), using four different biological replicates.

Gas exchange and active fluorescence measurements

Instantaneous values of net CO₂ assimilation rate (A _net, μmol m⁻² s⁻¹) were determined using a LI‐6400 portable photosysthesis system (Li‐Cor Biosciences, Lincoln, OR, USA) equipped with a 6400‐40 leaf chamber fluorometer with a light source of independently controlled LEDs (three blue, one far‐red and two red). The conditions in the leaf chamber were controlled at an air flow of 500 μmol s⁻¹, a 420 ppm CO₂ concentration and 60–70% relative humidity. Leaf temperature was maintained at 25°C, and leaves were carefully positioned filling the entire sample cuvette (2 cm²). Two consecutive light‐driven measurements were taken for one leaf (3^rd or 4^th leaf from the apex) of each plant (12 plants per treatment). First, the measurements were performed at a total PAR of 300 μmol m⁻² s⁻¹ (‘PAR300’), that is the illumination at which plants were grown. Next, the same protocol was repeated at a total PAR of 1000 μmol m⁻² s⁻¹ (‘PAR1000’), adding additional light stress.

Dark‐adapted fluorescence and gas exchange measurements were taken before the start of the photoperiod (PAR = 300 μmol m⁻² s⁻¹), and light‐adapted measurements commenced 4 h after the start of the photoperiod and were completed within c. 4 h. The same leaf from each plant was analysed for dark‐ and light‐adapted conditions. Measurements were taken when A _net and other gas exchange measurements reached steady‐state conditions.

Measured active fluorescence parameters included minimal fluorescence in the dark (F ₀), steady‐state fluorescence (F _s), and dark‐ and light‐adapted maximal fluorescence (F _m, F _m′). Maximum fluorescence yields were achieved by exposing the leaf to a 0.8 s saturating flash (8000 μmol m⁻² s⁻¹). Next, the maximum quantum yield of photosystem II photochemistry (F _v/F _m = (F _m − F ₀)/F _m), actual PSII operating efficiency (ΦPSII = (F _m′ − F _s)/F _m′), that is proxy for the energy directed to PSII over the energy absorbed, PQ, nonphotochemical quenching (NPQ_F = (F _m – F _m′)/F _m′) and parameters related to the ‘reversible’ and ‘sustained’ NPQ and qL (Porcar‐Castell, 2011) were estimated based on the measured parameters.

In addition to these steady‐state measurements, light‐response curves (LCs) were performed on d1, d3 (stress) and d8 (early recovery) on three different plants for each treatment, by varying the incident PAR levels stepwise between 1800, 1200, 800, 500, 300, 200, 100, 50 and 0 μmol m⁻² s⁻¹. Leaves were acclimated for 1–4 min at each PAR level until stable gas exchange parameters were observed. The data collected through the LC scheme were further processed as F _s vs ΦPSII trends. To explore the different phases in the F _s‐ΦPSII LCs across treatments and along time, the first derivative δF _s/δΦPSII of the curves was calculated. This was done after applying a polynomial fitting function (5^th degree) to get a proper fit of the original distribution for the PAR‐interval sampled F _s‐ΦPSII data. Based on this derivative, the different slope phases were defined.

Leaf spectroscopy measurements: fluorescence, reflectance and absorbance

Following the active fluorescence measurements, passive fluorescence, reflectance and transmittance measurements were performed on the labelled leaves (Fig. 1a). Upward and downward scattered leaf radiance (L _up, L _dw) were simultaneously measured, respectively, by an upward and downward nadir‐pointing fibre optic inserted in the upper and lower opening of a FluoWat leaf clip (Alonso et al., 2007; Van Wittenberghe et al., 2013, 2019). Each fibre optic measuring under a 25° field of view was connected to a high‐sensitivity VIS–NIR spectroradiometer (QEPRO; Ocean Insight Inc., Orlando, FL, USA), and both spectroradiometers were controlled by the OceanView software (Ocean Insight). A 400–800 nm broadband incoming illumination spectrum was provided by a high‐voltage single LED (High Cri LED 10 W 17 V 3050–5900 K; Yuji International Co. Ltd, Beijing, China). Similarly to the active fluorescence protocol, all measurements and derived parameters (reflectance ‘R’, transmittance ‘T’, absorbance ‘A’) were performed at a PAR of c. 300 and 1000 μmol m⁻² s⁻¹. For this, we kept the LED input current and voltage constant and used a polarized filter to lower the irradiance level by a factor of 3.5. The intensity of the LED was kept constant by an external power source (0.50 A, 15 V). Steady‐state fluorescence diffusively emitted from either leaf sides, that is F _up and F _dw, was directly measured by inserting a 650‐nm high‐pass filter at the light opening of the clip (Fig. 1a). Passively emitted fluorescence was in this way obtained in the 650–800 nm range, allowing the calculation of the total fluorescence flux (F _tot = F _up + F _dw, μmol m⁻² s⁻¹). For more details on the set‐up, the reader is kindly referred to Van Wittenberghe et al. (2019). The different measuring steps followed and the processing protocol are further described in Supporting Information Table S1.

Fig. 1.

Leaf spectroscopy set‐up and spectral fitting analysis. (a) Leaf clip and VIS–NIR spectroradiometer set‐up for the measurements of upward and downward radiance (L _up, L _dw), reflectance (R), transmittance (T) and upward, downward and total fluorescence (F _up, F _dw, F _tot) at photosynthetically active radiation (PAR) conditions PAR300 and PAR1000. (b) Pigment spectral unmixing algorithm after Van Wittenberghe et al. (2024) applied on the logarithm of the R signal to obtain the estimated effective absorbances (${\widehat{A}}_{\mathrm{eff}}$) for Chla, Chlb, β‐carotene (β‐Car), anthocyanins (Anc) and the absorption changes indicative for regulated heat dissipation (${\widehat{A}}_{\mathrm{eff}}$ 535‐nm) in the fitting range 500–780 nm. (c) Full processing scheme with ${\widehat{A}}_{\mathrm{eff}}$ 535‐nm, absorbed photosynthically active radiation (APAR) and fluorescence quantum efficiency (FQE) as essential parameter outcomes.

Spectral retrieval of FQE and 535‐nm absorbance peak

To remotely obtain the fluorescence quantum efficiency and the absorption related to the regulated heat dissipation, we used a pigment spectral unmixing methodology, recently described in Van Wittenberghe et al. (2024). In short, estimated effective absorbance (${\widehat{A}}_{\mathrm{eff}}$) of individual pigments, that is Chla and Chlb, beta‐Carotene (β‐car) and regulated dissipation by means of changes in spectral absorption properties in the 500–600 nm absorption region (modelled as a peak c. 535 nm) are retrieved based on a non‐negative least squares (NNLS) linear fitting strategy using absorption coefficients of the pigments (Fig. 1b). The absorption coefficients used for Chla, Chlb and β‐Car were based on the photoacoustic spectra measured by Nagel et al. (1989), while the fast absorption changes rising from additional regulated heat dissipation behaviour was presented by a Gaussian‐based shape in the 500–600 nm range, modelling the quick amplified absorbance changes observed during the qE process and chemical conversion of the VAZ cycle according to Van Wittenberghe et al. (2021a). The 535‐nm peak feature was proposed as a true electronic change upon formation of Antheraxanthin, being strongly red‐shifted by an electrochromic shift due to the Stark effect compared with in vitro measurements. Formation of Zea was also observed during the peak development (but without amplified Gaussian effect), wherefore some contribution related to qZ in addition to qE cannot be entirely excluded. The spectral fitting relies on the minimization of the squared difference between the measured absorbance spectrum and the modelled spectrum by adjustment of the abundancy weights given to each pigment absorption, not allowing overfitting of the final result. Next, FQE was calculated as the ratio between the total energy quanta emitted as fluorescence (F _tot) and the total energy quanta absorbed by Chla (Figs 1c, S1). The latter was obtained by the multiplication of the effective absorbance by Chla (result from the NNLS fitting) and the PAR irradiance measurement described in the leaf spectroscopy protocol (Van Wittenberghe et al., 2024).

Photosynthetic pigment analysis

Plants were acclimated to PAR 300 μmol m⁻² s⁻¹ for 10 min before sampling by punching three leaf discs (5 mm Ø) on one leaf from three plants per treatment. The punched leaf discs were immediately frozen in liquid N₂ and stored at −80°C. Frozen leaf discs were extracted and filtered according García‐Plazaola & Becerril (1999). Extracts were further analysed in an HPLC (high‐performance liquid chromatography) system equipped with a reverse‐phased Waters Spherisorb ODS1 column, and pigments were quantified using a PDA detector (Waters model 996) at 445 nm. Chla and Chlb, β‐car, Vio, Ant, Zea, neoxanthin (Neo), Lut and Lx were identified and quantified by comparison with pure standards. Xanthophyll pools, normalized by the Chla pool to evaluate the plasticity of the antenna stoichiometry to stress, were statistically compared between treatments and vs the control measurements of the same day.

Xanthophyll‐related enzyme expression

Gene expression of the enzymes involved in xanthophyll cycle pigments' interconversions was analysed by reverse transcription quantitative polymerase chain reaction. Total RNA was obtained using the Plant/Fungi Total RNA Purification kit (Norgen Biotek, Thorold, ON, Canada), and further retrotranscribed to complementary DNA, using the PrimeScript™ kit (TaKaRa, Kusatsu, Japan), following both kit's protocols. Next, gene expression of beta‐carotene hydroxylase (CrtR‐b2), violaxanthin de‐epoxidase (VDE) and zeaxanthin epoxidase (ZEP) were analysed (StepOnePlus Real‐Time PCR; Applied Biosystems, Foster City, MA, USA). Four biological replicates were analysed for each treatment and day, using the mean value of three technical replicates for each of them. The relative expression of each gene was calculated using as reference gene the ribosomal protein SolRPS18 gene, and the LinRegPCR software (Ruijter et al., 2009) was employed for the analysis (Table S2).

Statistical analysis

The data were analysed using Ibm Spss Statistics (v.26.0). One‐way analysis of variance (ANOVA) test was used to compare differences between different experimental days, followed by Duncan's multiple range test for post hoc comparisons. Comparisons between two groups (generally control vs stress, or PAR300 vs PAR1100) were carried out using Student's t‐test. Normality and homoscedasticity of data was assessed using a Shapiro–Wilk's and a Levene's test, respectively. In all cases, P < 0.05 values were considered statistically significant.

Results

Active fluorescence‐based measurements

Under the growth chamber illumination condition (‘PAR300’), statistical differences in the active fluorescence measurements between both treatments vs the controlled plants were found for all parameters from d2 (F _s, NPQ_F, ΦPSII) or d4 (A _net) onward (Figs 2a–d, S2, S3). At d2, F _s significantly decreased and NPQ_F significantly increased (P < 0.05) for both stress treatments, with respect to the control, affecting also ΦPSII. While NPQ_F progressively increased for both stress treatments along d2–d4–d7, F _s measurements showed no statistical differences with the control on d4. In the case of the drought treatment, F _s further increased until d7, while this was not the case for the N deficit treatment. By contrast, NPQ_F levels of both stress treatments remained elevated and significantly higher (P < 0.05) than the control during the whole stress phase (Fig. 2c). Along the experiment, both treatments showed a decrease in maximum or actual F _v/F _m, mainly due to decreases in F _m (dark and PAR300) and/or increases in F ₀ (dark and PAR1000) (Fig. S2). These trends, in combination with the calculation of ‘reversible’ and ‘sustained’ NPQ components and qL parameters obtained under the Lake model (Fig. S4), showed a relatively stronger impact of sustained NPQ during the dark‐adapted and growing light conditions (PAR300), while photo‐inhibition related effects (monitored by the increase in F ₀) were apparently stronger when exposed to PAR1000 at the end of the stress period (Fig. S2c). Control plants that remained watered with 8 mM N showed decreased F _s values at d7 compared with d0, and slightly elevated NPQ_F levels, but without any strong effect on ΦPSII, A _net nor dark F _v/F _m. Under additional light stress, all treatments further downregulated photosynthesis compared with the control plants (Figs 2e–h, S2h,i). By formula definition, the ‘sustained’ NPQ component showed independency of light treatment (Fig. S4b,f), while the ‘reversible’ component (NPQ_r, similar to NPQ_F) showed additional quenching mechanisms under high light, especially for the control plants (Fig. S4e). This indicates that control plants retained more adaptability under higher light conditions. Significantly higher sustained NPQ was found for both stress treatments along the experiment, with highest values for the N‐deficit plants. This was further reflected in the significant decrease in functional reaction centra for both treatments (Fig. S4d,h). The fraction of open reaction centra (qLr) decreased strongly between PAR300 and PAR1000, with lowest values at d7 for the drought treatment (Fig. S4c,g). After 1 wk of recovery (d14), the treated plants appeared to have fully recovered as there was no longer any statistical difference between the control and stress treatments for F _s, NPQ_F and A _net. Significant stress appearing from d2 until d7, and recovery at d14, was confirmed also by the leaf nitrogen content, indicating only significant differences during the stress phase of both treatments (Fig. S5). Nitrogen‐deficit plants, however, showed a lower (P < 0.05) plant DW after recovery (Fig. S5).

Fig. 2.

Measurements of active fluorescence (F _s) and derived parameters (ΦPSII, NPQ_F, A _net) for each treatment and measured day using the PAR300 (left panels a, b, c, d) and PAR1000 (right panels e, f, g, h) protocols (n = 12, mean ± SD). Statistical differences between each measured parameter of the stressed plants (Drought and Nitrogen deficit) vs those of the Control for the same day are indicated with asterisks (*, P < 0.05; **, P < 0.001; Student's t‐test) and statistical differences along the days for each treatment, are indicated by different lowercase letters (one‐way ANOVA test for independent samples followed by Duncan's test, P < 0.05).

Trends in F _s or fluorescence yield behaviour under stress – starting with an initial decrease in F _s followed by an increase with progressive stress – were analysed through the light curves (d1, d3 and d8), shown as F _s vs ΦPSII plots (Fig. 3a–f), in combination with the F _s and ΦPSII data obtained by the PAR300 and PAR1000 protocols (d0, d4 and d7). Despite the mismatch in measurement days, the F _s‐ΦPSII data sets showed reasonable to good agreement between the LCs and point measurements of the Li‐Cor protocols. The LCs measured during the early stress phase (d1 and d3) were similar within each treatment but slightly different between the treatments. In the panels (g)–(i), the first derivatives of the F _s‐ΦPSII LCs are shown, marking the F _s‐ΦPSII slope changes, and in the panels (j)–(l), the evolution of NPQ_F is shown. The activation and saturation of the NPQ_F parameter is seen as the mirror of a sigmoidal curve and illustrates the dominant control of fast qE onto the F _s‐ΦPSII slope changes. Overall, there is a negative relationship between PSII operating efficiency (ΦPSII) and fluorescence yield (F _s). This means that excessive energy is dominantly released as fluorescence, but when NPQ_F is strongly activated (see NPQ strong activation phase ‘2’, Fig. 3j–l), the quenching effect is superimposed onto the original trend, inducing a short phase with a positive slope in the F _s‐ΦPSII curve (Fig. 3a–f). This positive slope segment was less pronounced during the stress phase (d1, d3) of the drought treatment, resulting in a higher F _s value at PAR300 (Fig. 3b). It shows that within the first days of water deprivation, the actual NPQ capacity was insufficient, resulting in an increase in F _s. By contrast, the control F _s values remained located within the early NPQ_F activation phase (Fig. 3a). Finally, the NPQ_F parameter saturated for all plants along the LC protocol (Fig. 3j–l), with F _s values at PAR1000 of all treatments showing NPQ_F saturation (Fig. 3d–f). Interestingly, the LCs taken on d8 (early recovery) showed a lowering of the F _s‐ΦPSII curves for all treatments, indicating an overall decrease in fluorescence yield as the first visible response to recovery.

Fig. 3.

Mean F _s‐ΦPSII light curves (LC) along the PAR interval protocol as an unsmoothed line for each treatment at d1, d3 and d8 (n = 3, mean ± SD) (a–f), plotted together with the steady‐state F _s‐ΦPSII measurements (n = 12, mean ± SD) at d0, d4 and d7, measured according to the PAR300 (a–c) and PAR1000 (d–f) protocols. (g–i) First derivative of F _s against ΦPSII (δF _s/δPhiPSII) after a 5^th‐degree polynomial fit of the LCs. (j–l) NPQ_F induction curve as a function of ΦPSII obtained by the LC protocol illustrating the NPQ_F early activation (‘1’), the strong NPQ_F activation (‘2’) and the NPQ_F saturation phase (‘3’).

FQE and F shape dynamics obtained from leaf spectroscopy

Apparent FQE calculated from leaf spectroscopy measurements (Figs 1, S1; Table S1) showed agreement with F _s, demonstrating overall higher values at 1000 μmol m⁻² s⁻¹ compared with those obtained at 300 μmol m⁻² s⁻¹ (Fig. 4). The linear correlation between the active‐based F _s and the FQE was significant (at P < 0.05) for the control (R ² = 0.50) and N‐deficit (R ² = 0.63) treatments, while it was not the case for the drought treatment (R ² = 0.19).

Fig. 4.

Scatter plot and linear correlations between mean active fluorescence F _s (n = 12, mean ± SD) and mean proximal sensing‐based fluorescence quantum efficiency FQE (n = 12, mean ± SD) measured at the same leaf, shown by treatment including both light treatments (PAR300 and PAR1000).

In Fig. 5, the active F _s measurements were used as a blueprint for the interpretation of the FQE measurements. FQE values are shown in combination with the corresponding ΦPSII values (measured at the same leaf, see the Materials and Methods section) and plotted on top the closest corresponding LC. Similar to the active F _s, there were few differences in FQE between the treatments during d2–d4 (Fig. 5a–c). On d7, the FQE values for the drought treatment showed a significant increase (P < 0.05) compared with the previous days, and compared with the control and N‐deficit plants (P < 0.05) (Fig. 5c,d). At the end of the recovery phase (d14), all plants returned having similar FQE values as previous prestress values of d2 or d4 (Fig. 5e,f). This indicates that the FQE can confirm the plant's recovery once the stressor has ceased. Along the course of the stress experiment, the FQE values also demonstrated the fluctuating, and therefore ambiguous, relationship with ΦPSII, with a distinct behaviour between drought and N‐deficit plants on d7. Both treatments showed a clear downregulation of photosynthesis (ΦPSII = 0.3) (Fig. 5c), but the mean FQE for drought on d7 resulted similar as the mean FQE of the (untreated) plants at d0 (ΦPSII = 0.5) (Fig. 5a). Hence, only based on FQE, no difference was found between early drought stress on d7 and healthy plants on d0. Clearly, this illustrates that the FQE value alone is not sufficient to identify early plant stress, and information equivalent to NPQ_F (Fig. 3j–l) is needed.

Fig. 5.

Active fluorescence‐based F _s‐ΦPSII light curves (LC, left y‐axis) of, respectively, d1, d3 and d8, for all treatments (Control, N deficit, Drought) shown as an unsmoothed line (n = 3, mean ± SD) at PAR300 (left columns) and PAR1000 (right columns), together with proximal sensing‐based fluorescence quantum efficiency fluorescence quantum efficiency (FQE) (n = 12, mean ± SD, right y‐axis) in combination with the ΦPSII measurements (n = 12, mean ± SD) taken from the same leaf samples on d0 and d2 (a, b), d4 and d7 (c, d), and d14 (e, f). The arrows in (c) indicate the opposite change in FQE from d4 to d7 for N‐deficit (green arrow) and Drought (magenta arrow) treatments.

Regarding the F spectral shape evolution (Fig. 6), results of d7 were of most interest as it was the only day when treatments showed distinctive F shape behaviour relative to the control plants. Both drought and N‐deficit plants showed an increase and widening of the red fluorescence peak and a small decrease in the far‐red peak (Fig. 6c), indicating a shift in the absorbance behaviour. This also resulted in a statistically higher (P < 0.001) red : far‐red fluorescence peak ratio (F689 : F740) for both treatments compared with the control plants on d7 (Table S3).

Fig. 6.

Mean upward emitted fluorescence (F _up) (n = 12) measured at PAR300 and normalized for the total emission for each treatment (Control, Drought, N deficit) at d2 (a), d4 (b), d7 (c) and d14 (d).

Effective absorbance fitting of the photosynthetic and photoprotective pigment pools

For each treatment and measuring day, the NNLS linear fitting resulted in estimated abundancy weights and spectrally resolved effective absorbance (${\widehat{A}}_{\mathrm{eff}}$) obtained for Chla, Chlb, β‐car, Anc and the regulated heat dissipation by means of associated changes in spectral absorption properties at the 535‐nm peak. The highest peak values of all pigment ${\widehat{A}}_{\mathrm{eff}}$ values are shown for the PAR300 (Table 1) and PAR1000 (Table S4) conditions, while Fig. 7 illustrates the effective absorbance fittings obtained at d7. Almost no differences were found in the pigment abundancy weights for the main antenna pigments Chla, Chlb and β‐car between days, nor between treatments. Only in the case of Chlb, a statistical difference (P < 0.05) was found between the control plants and the N‐deficit treatment (d7) and drought treatment (d14), respectively. HPLC analyses of all pigments showed significant decreases under both stress treatments (Table S5), while Chla : Chlb ratios did not show this, and Chl/Car ratios only in a few cases (Table S6). Abundancy weights obtained for the anthocyanins (photoprotective pigments in the vacuoles), and the 535‐nm absorption peak showed more statistical differences than the ones for Chl, either along the measurement days, or in comparison with control (Table 1). Anthocyanins were fitted in all cases except one sample of drought plants on d7. Interestingly, the 535‐nm absorption feature was not fitted in the case of control plants throughout the whole experiment. On d2 and d4 (early stress), the 535‐nm feature was fitted in few cases for the drought (d2: 5/10; d4: 1/12) and N‐deficit plants (d2: 2/10; d4: 3/12), despite overall lower surface‐based Xan pool concentrations (Table S5). On d7, the 535‐nm feature was fitted in the majority of the samples for both the drought (7/12) and N‐deficit (9/12) treatments (Fig. 7; Table 1). After the recovery phase (d14), only one case of drought stress remained with a 535‐nm feature fitting (Table 1).

Table 1.

Effective absorbance at the maximum peak (${\widehat{A}}_{\mathrm{eff}}$ peak) in the 500–780 nm range for the spectral fitting of Chla, Chlb, β‐carotene (β‐car), anthocyanins (Anc) and the fitted 535‐nm absorbance changes at PAR300.

${\widehat{A}}_{\mathrm{eff}}$ Peak (pigment)	Day	Control	Drought	N deficit
Chla	D0	0.807 (6/6)a
	D2	0.850 (10/10)a	0.846 (10/10)a	0.835 (10/10)a
	D4	0.862 (12/12)a	0.827 (12/12)a	0.840 (12/12)a
	D7	0.860 (12/12)a	0.814 (12/12)a	0.851 (12/12)a
	D14	0.870 (12/12)a	0.863 (12/12)a	0.837 (12/12)a
Chlb	D0	0.241 (6/6)a
	D2	0.210 (10/10)a	0.221 (10/10)a	0.215 (10/10)a
	D4	0.226 (12/12)a	0.217 (12/12)a	0.210 (12/12)a
	D7	0.236 (12/12)a	0.209 (12/12)a	0.206 (12/12)*,a
	D14	0.246 (12/12)a	0.208 (12/12)*,a	0.211 (12/12)a
β‐Car	D0	0.177 (6/6)a
	D2	0.269 (10/10)a	0.258 (10/10)a	0.339 (10/10)a
	D4	0.252 (12/12)a	0.247 (12/12)a	0.284 (12/12)a
	D7	0.236 (12/12)a	0.316 (12/12)a	0.326 (12/12)a
	D14	0.227 (12/12)a	0.303 (12/12)a	0.285 (12/12)a
‘535‐nm peak’	D0	0 (0/6)a
	D2	0 (0/10)a	0.0328 (5/10)*,a	0.0175 (2/10)*,a
	D4	0 (0/12)a	0.0126 (1/12)a	0.0392 (3/12)*,a
	D7	0 (0/12)a	0.0186 (7/12)*,a	0.0287 (9/12)**,a
	D14	0 (0/12)a	0.0189 (1/12)a	0 (0/12)a
Anc	D0	0.247 (6/6)a
	D2	0.248 (10/10)a	0.186 (10/10)a	0.231 (10/10)a
	D4	0.263 (12/12)a,b	0.266 (12/12)b	0.209 (12/12)a,b
	D7	0.283 (12/12)a,b	0.182 (11/12)**,a	0.157 (12/12)**,b
	D14	0.305 (12/12)b	0.287 (12/12)b	0.271 (12/12)*,a

The mean values are shown for every condition and measurement day (D0–D14), omitting the nonfitted cases (when the abundancy weight was zero), indicating in brackets how many cases were fitted out of the total analysed. Statistical differences between stressed plants (Drought and N deficit) vs Control for the same day are indicated with asterisks (*, P < 0.05; **, P < 0.001; Student's t‐test) and statistical differences along the days for the each treatment, are indicated by different lowercase letters (one‐way ANOVA test for independent samples followed by Duncan's test, P < 0.05).

Fig. 7.

Effective absorbance (${\widehat{A}}_{\mathrm{eff}}$) in the fitting range (500–780 nm) for Chla, Chlb, β‐carotene (β‐car), anthocyanins (Anc) and the absorption changes indicative for regulated heat dissipation (${\widehat{A}}_{\mathrm{eff}}$ 535‐nm) estimated through the spectral fitting method on day 7 (d7) of the stress experiment for each treatment (Control, Drought and N deficit) at PAR300.

Dynamic xanthophyll pools and their effect on NPQ parameters and F _v/F _m

Total xanthophyll pool (Vio + Ant + Zea + Lut + Lx + Neo) normalized by the Chla pool showed statistical increases (P < 0.05 or P < 0.001) for N‐deficit plants compared with the control plants on d4 and d7 (Fig. 8a). A gradual increase in the mean Zea/Chla and Lut/Chla was observed for the N‐deficit and drought plants along the stress days d2–d4–d7, resulting in significantly higher relative pools (P < 0.05 or P < 0.001) compared with the control plants on d4 and d7 (Fig. 8b,d). Interestingly, the control plants initially showed a slight (nonsignificant) increase in Ant/Chla from d0 to d4, while subsequently a slightly higher Zea/Chla pool was formed on d7, which significantly reduced the Ant/Chla pool compared with d4 (Fig. 8c,d). On d14, the Lut and Zea pools downregulated again as a result of the recovery, except for the Zea pool of the N‐deficit plants.

Fig. 8.

Total xanthophyll (Xan tot, a) and individual lutein (Lut, b), antheraxanthin (Ant, c) and zeaxanthin (Zea, d) pools normalized by the Chla pool, sampled at PAR300 (n = 3, mean ± SD). Statistical differences between treatments (Drought and Nitrogen deficit) vs Control for the same day indicated are indicated by asterisks (*, P < 0.05; **, P < 0.001; Student's t‐test) and differences along the days are indicated by lowercase letters for each treatment (one‐way ANOVA test for independent samples followed by Duncan's test, P < 0.05).

Analysis of expression of the genes encoding the xanthophyll‐related enzymes is shown in Fig. S6. CrtR‐b2, the enzyme involved in the conversion from α‐ and β‐carotene to the corresponding Xans (Zea and Lut), is significantly higher (P < 0.05) in drought plants compared with nitrogen‐deficit and control plants on d4 and d7, while it goes down again on d14 (Fig. S6). VDE, the enzyme involved in the conversion from Vio to Zea (or Lx to Lut), is only significantly downregulated in N‐deficit plants compared with the other groups on d7. ZEP, which converts Zea to Vio (and Lut to Lx), expression is only different for drought plants on d7, being significantly higher than N‐deficit and control plants (Fig. S6).

Significant correlations (P < 0.05 or P < 0.01) were found between the total xanthophyll pool and the maximum or actual F _v/F _m (Fig. S7), indicating a strong link between the photoprotective pool (sampled at PAR300) and the quantum yield of photochemistry at each measuring day. At PAR300, the total xanthophyll pool also correlated significantly with NPQ_F (Fig. 9a). Among the xanthophylls, stronger correlations for the larger Lut (R ² = 0.61) and Zea (R ² = 0.50) pools with the NPQ_F value were found (Fig. 9b,d), while Ant did not show any correlation, mainly due to the higher concentrations in the control plants (Fig. 9c). Differentiating between ‘reversible’ and ‘sustained’ NPQ components showed in addition that both components were significantly Xan pool‐related at PAR300 (P < 0.01), whereas at PAR1000, the additional reversible NPQ could not be related to the Xan pool data derived at PAR300 (Fig. S7).

Fig. 9.

Linear regression with parameters (R ², P‐value) between the total xanthophyll (Xan tot, a) or individual lutein (Lut, b), antheraxanthin (Ant, c) and zeaxanthin (Zea, d) pools normalized by the Chla pool (n = 3, mean ± SD), and the NPQ_F parameter at PAR300 (n = 12, mean ± SD), pooling all the treatments (Control, Drought and N deficit) together.

535‐nm absorption fitting vs pigments and NPQ parameters

Overall, the NNLS fitting resulted in an average fitting error of around maximal 10%, with slightly higher errors for the control samples compared with the treatment's samples (Fig. 10a). Slightly higher errors in the PAR1000 data set fittings compared with the PAR300 data set fittings were found due to a lower fitting performance in the green region of the spectrum (Fig. 10b). At PAR300, the fitted 535‐nm absorbance peak showed a significant (P < 0.05) correlation with the available Xan pool (Lut + Ant + Zea) relevant during the activation of NPQ_F (Fig. 10c). We further tried to relate the 535‐nm absorbance fitting result at both illumination conditions to the active‐based NPQ parameters (Fig. 10d–f) taking in mind that, due to differences in F _v/F _m linked to the differences in accumulated Xan pools (Fig. S7), comparisons should be made with care. Correlations between the 535‐nm absorbance peak fit and the NPQ_F were stronger at PAR300 than at PAR1000 and could also be described as being due to a better fit with the ‘sustained’ NPQ component for both PAR conditions. This component was shown indeed more dependent on the Xan pool accumulation for both PAR conditions (Fig. S7). Only few significant ${\widehat{A}}_{\mathrm{eff}}$ 535‐nm peak fits were found for the drought and N‐deficit treatments in the PAR1000 data set, while none at all were made for the control plants, despite a substantial increase in ‘reversible’ NPQ, and hence NPQ_F.

Fig. 10.

Evaluation of the spectral fitting method and meaning of the effective absorbance fitting related to the fast regulated heat disspation (${\widehat{A}}_{\mathrm{eff}}$ peak 535 nm). (a) Mean fitting error of the total effective absorbance estimation along the fitting range 500–780 nm for each treatment (Control, Drought and N deficit) and each illumination condition (PAR300, PAR1000) and (b) difference in spectral fitting error between PAR300 and PAR100 fittings along the fitting range (ΔError_300–1000). (c) Linear regression (R ², P‐value) between the fitted ${\widehat{A}}_{\mathrm{eff}}$ peak 535‐nm (mean ± SD) at day 7 at PAR300, excluding the zero‐weighted fits in case of insignificant fittings (for n, see Table 1), and the sum of the lutein (Lut), antheraxanthin (Ant) and zeaxanthin (Zea) pools normalized by the Chla pool. Panels (d), (e) and (f) show respectively the linear regression (R ², P‐value) between the fitted ${\widehat{A}}_{\mathrm{eff}}$ peak 535‐nm (mean ± SD) and the corresponding mean NPQ_F, NPQ_s and NPQ_r parameters (n = 12, mean ± SD) for both PAR300 and PAR1000 data sets.

Discussion

FQE as insufficient parameter in stress detection

Several reviews and perspective papers on remotely estimated SIF have emphasized the remaining challenges to unleash the true potential of passively emitted fluorescence measurements in the support of better estimations of the global carbon cycle (Porcar‐Castell et al., 2014, 2021; Wohlfahrt et al., 2018). To deal with the omnipresent scaling issue, the so‐called ‘downscaling’ approaches have been often used as part of the challenge to establish a better link with carbon sequestration (Badgley et al., 2017; Dechant et al., 2020). Another strategy for imaging spectroscopy data starts from a bottom‐up approach, using the full spectral information based on unique signals (Van Wittenberghe et al., 2024). Understanding and disentangling the spectral pigment and antenna behaviour at the leaf level, and in particular the dynamics observed when antenna states are changed, is therefore the starting point for further applications at larger spatial scales. The calculation of FQE, based on the energy absorbed by Chla, is one of the first steps to account for the variability in absorbed energy. FQE relates to F _s (Fig. 4) as seen from other studies (Cendrero‐Mateo et al., 2015; Loayza et al., 2023); wherefore, it is expected that FQE also relates to ΦPSII in a nonlinear way along the course of NPQ_F activation and saturation (Fig. 5). The changes in the FQE‐ΦPSII trend, that is the evolution from a positive (NPQ_F strong activation) towards a negative (NPQ_F saturation) slope, occur during a critical transition for plants. It occurs when available NPQ_F mechanisms (e.g. pools) are activated but gradually reach saturation (Fig. 3j–l), resulting in the most significant downregulation of photosynthesis. Under these conditions, FQE values vary in a limited range (0.01–0.014) while ΦPSII values significantly vary between 0.2 and 0.6. Remote sensing studies have also revealed that SIF explains limited variability in gross primary productivity (GPP) changes during sudden stress events, such as heat waves (Wohlfahrt et al., 2018). In the field, an increase in passive fluorescence yield has been observed under extreme events, such as heat wave stress, showing a clear ΦF‐ΦP trend break under very high NPQ_F values (Martini et al., 2022). It is clear, therefore, that neither SIF (measured under constant APAR) nor FQE alone is sufficient to diagnose the critical impairment of photosynthesis.

Detection of the antenna‐regulated heat dissipation by changes in spectral absorption

The activation of the regulated heat dissipation by means of changes in nondestructive spectral absorption properties has been commonly sought through the PRI (Gamon et al., 1997), which closely follows the photochemical yield when Chl content or APAR remains constant (Gamon et al., 1992; Atherton et al., 2016). The proposed spectral unmixing approach aims to disentangle the specific 535‐nm absorbance peak change, which is suggested to find its origin quickly after the formation of the de‐epoxidised Xans (Van Wittenberghe et al., 2021a). Electrochromic absorption shifts have been suggested to act in the same spectral range in the same (or faster) timescale (Kramer et al., 1999; Viola et al., 2019). While electrochromism can play a role in the antenna behaviour, we consider it rather difficult to resolve the different effects (in this case electrochromic and electronic) through a single‐peak feature fitting method. In vivo occurring electrochromic effects, expected to demonstrate an observable shift, are therefore considered less representable by an overall feature at a fixed wavelength. Here, we found that the 535‐nm absorption was able to show the activation of controlled heat dissipation mechanism in the majority of the N‐deficit leaf samples, and in most of the drought‐stressed plants, on d7, distinguishing them from the control plants (Fig. 7). This occurred when the pools of photoprotective xanthophylls were at highest levels for both treatments (Fig. 8). Higher relative Lut and VAZ pools (Esteban et al., 2015) but also decreasing Chla : Chlb ratios (Hu et al., 2023) have been seen earlier under drought stress. In accordance with previous results (Van Wittenberghe et al., 2024), the fitted values for the 535‐nm peak were overall small (0.01–0.03) (Table 1). The early stages of NPQ_F activation in the control plants, apparently only controlled by the Lut pool (Fig. 9b–d), were clearly not well‐captured by the current 535‐nm absorption feature (Fig. 10a). By contrast, N‐deficit and drought treatments, showing increased Zea pools on top of increased Lut pools (Fig. 9b–d), showed more successful fittings. N‐deficit leaves, which showed overall higher Zea/Chla pools than drought plants on d4 and d7, also interestingly showed a better 535‐nm feature fit on these days (Table 1).

The observed issue with fitting a single fixed 535‐nm feature, which seemed to work rather for additional increased Zea than for increased Lut pools, is that the features are known to be dynamic in peak position due to xanthophyll composition as seen from mutants' studies (Li et al., 2009; Ilioaia et al., 2011), but also when measuring under different natural conditions for a single species (Van Wittenberghe et al., 2019; Sukhova and Sukhov, 2020). This shortcoming should be resolved by elucidating individual Xans from the spectral fitting method, which are known to have different relative quenching abilities (Short et al., 2023). Other confounding dynamic absorption effects that may have played a role in the fitting result are the potential conformational changes within the antenna. Transitions between conformational states in the antenna are expected to alter the entire spectrum due to changes in the pigment electronic states and are observed as distinct shifts in their absorption and emission energies (Krüger et al., 2011; Chmeliov et al., 2016). Absorbance shifts expected for these conformational changes should be in future measured and described for the in vivo absorbance behaviour, with the goal to decouple the distinct behaviour from the proposed 535‐nm feature.

Interestingly, as a symptom of NPQ_F saturation, indicating a strongly quenched state, drought plants could be distinguished from N‐deficit plants by an additional increase in FQE (Fig. 5c). Therefore, the combination of 535‐nm absorption and FQE retrieval was able to differentiate between different stages of the early stress progression, specifically in the case of NPQ_F activation (case of N‐deficiency) vs NPQ_F saturation (case of drought). This result is encouraging as it may help to diagnose the origin and the severity of the stress remotely.

Discrepancy between 535‐nm absorbance fitting and F‐based NPQ

Support for the fitting of the 535‐nm absorption as an indication of the activation of regulated heat dissipation in the antenna is shown by the increased presence of the de‐epoxidised xanthophylls, which are considered relevant in the prior control of NPQ_F, that is Lut, Ant and Zea (Fig. 10a). This is in agreement with previous studies that have also found a close correlation between NPQ_F and xanthophyll pools across different species and conditions, including studies that have used VAZ‐cycle mutants (Johnson et al., 2008, 2010; Nilkens et al., 2010; Förster et al., 2011; Ware et al., 2016; Leuenberger et al., 2017; Bernardo et al., 2022). Validating the fitting of 535‐nm absorption with the ‘classical’ NPQ_F parameter is, however, not straightforward for several reasons. On the one hand, NPQ_F may contain several apparent nonphotochemical quenching effects, related to the energy‐dependent (qE), zeaxanthin‐cycle (qZ), photoinhibition (qI), state‐transitions (qT) (Kalaji et al., 2014) and chloroplast movement (qM) processes (Cazzaniga et al., 2013). While qE may be the dominant process, our results indicating a correlation with Zea accumulation (Fig. 10c) also suggest that qZ‐related quenching may play an additional role in the 535‐nm peak feature.

By contrast, our results indicated that other quenching mechanisms, activated under saturating light conditions (PAR1000), were not represented by the current method. N‐deficit and drought plants showed increases in F _s and FQE, with a relatively small increase in NPQ_F (Figs 3d–f, 5b,d,f). At the same time, fitting results showed lower detection of ${\widehat{A}}_{\mathrm{eff}}$ 535‐nm for PAR1000 conditions than the PAR300 results (Fig. 9b; Table S4). In the case of control plants, the main ‘reversible’ NPQ component remained undetected by the peak at 535 nm (Fig. 10). We suspect that the reason for this discrepancy lies in the fitting methodology, which may currently lack the necessary complexity to account for all the expected variability in antenna‐related absorbance adjustments. Additionally, a divergence between the 535‐nm fitting approach and the NPQ_F parameter under extremely saturating light conditions may be explained by other components of NPQ_F such as state transitions, which reduce the PSII relative absorption cross section, altering the fluorescence signal (Horton & Hague, 1988). The state transitions simply increase the NPQ_F parameter because PSI has a lower fluorescence quantum yield than PSII (Krause & Weis, 1991). Other photoprotection may be caused by chloroplast light avoidance movements, contributing to the apparent decrease in fluorescence lifetime, but not necessarily to the decrease in fluorescence quantum yield (Cazzaniga et al., 2013). Probably, the effect of state transitions or other adjustments in the membrane cannot entirely be ruled out here, as we could observe on d7 a significant increase in the red vs far‐red fluorescence peak ratio between control and treatment plants (Fig. 6; Table S3), suggesting changes in the (red‐edge) absorption behaviour. Between PAR300 and PAR1000 measurements, the fluorescence peak ratio also increased in all the cases, but this was only significant in one case (Table S3).

Another discrepancy was noticed during recovery, when both treatments showed a reversible response in both FQE and 535‐nm absorption at d14. By contrast, the Zea pool of N‐deficit plants remained elevated after recovery (Fig. 8d), indicated by a significantly higher sustained NPQ (Fig. S4b,f) and lower whole‐plant biomass at recovery (Fig. S5b). Hence, even when the 535‐nm absorption fitting has ceased or remains undetected (Table 1), a sustained stress‐generated Zea pool was still present and delayed full biomass recovery. These observed sustained NPQ effects remaining after relaxation of the regulated heat dissipation should be considered together with the effects of other plant acclimations such as Chla : Chlb and total Car : total Chl ratios to improve absorbance‐based fitting strategies in order to estimate actual photosynthesis remotely.

Conclusions

The findings of this work show that early stress responses in the light reactions are manifested by subtle changes in the absorbance and fluorescence signals, which can be detected by nondestructive leaf spectroscopy and signal processing. As expected, FQE is found to be insufficient as a single parameter for diagnosing early stress due to its fluctuating relationship with photosynthesis during the activation and saturation of the quenching mechanisms. The unmixing of the 535‐nm absorbance peak changes from the leaf reflectance succeeded to monitor the early activation of regulated heat dissipation. Based on this complementary information, a distinction could be made between nitrogen deficiency and drought stress, indicating different levels of plant stress response. Further improvements in the fitting strategy, disentangling more underlying mechanisms through unique spectral signatures, may potentially help to further resolve the regulated heat dissipation dynamics, which will be of great interest to harmonize early stress detection in photosynthesis over different scales.

Competing interests

None declared.

Author contributions

SP‐D was involved in methodology, data curation, investigation, formal analysis and writing – original draft. MPC‐M was involved in methodology, investigation, data curation and writing – review & editing. AM‐E was involved in methodology and investigation. AR‐F, MIA, BR‐M, BF‐M, JIG‐P, RVM and CR were involved in investigation. JM was involved in writing – review and editing. SGN and IG‐R were involved in resources, investigation and writing – review and editing. SVW was involved in conceptualization, methodology, investigation, supervision, writing – original draft and funding acquisition.

Supporting information

Fig. S1 Conceptual presentation of the fluorescence quantum efficiency retrieval from the pigment spectral unmixing algorithm.

Fig. S2 Additional active fluorescence‐measured parameters (F ₀, F _m and F _v/F _m) per each treatment, light condition and measuring day.

Fig. S3 Additional active fluorescence‐measured photochemical quenching, YNPQ and YNO per each treatment, light condition and measuring day.

Fig. S4 Sustained and reversible nonphotochemical energy quenching and qL measured from active fluorescence per each treatment, light condition and measuring day.

Fig. S5 Plant physiological stress indicators during the experiment given by leaf nitrogen content, plant fresh and dry biomass and relative water content.

Fig. S6 Analysis of expression of the genes encoding the xanthophyll‐related enzymes done by reverse transcription quantitative polymerase chain reaction.

Fig. S7 Linear regression between the total xanthophyll pool and F _v/F _m, NPQ_F, NPQ_r and NPQ_s at the different light conditions.

Table S1 Steps of the leaf spectroscopy protocol.

Table S2 Primers used to analyse the expression of genes involved in xanthophyll biosynthesis.

Table S3 Statistical differences of the upward emitted fluorescence red : far‐red ratio (calculated as F689 : F740).

Table S4 Effective absorbance at the peak (${\widehat{\mathit{A}}}_{\mathrm{eff}}$ peak) estimated from the spectral fitting of each pigment at PAR1000.

Table S5 Pigment pools per leaf area for each treatment along the experiment.

Table S6 Chla : Chlb and Car : Chl ratios between treatments and control plants.

Please note: Wiley is not responsible for the content or functionality of any Supporting Information supplied by the authors. Any queries (other than missing material) should be directed to the New Phytologist Central Office.

Data availability

Raw and processed data are available open access through the Zenodo repository (doi: 10.5281/zenodo.12800064).