Quantifying spatial heterogeneity of chlorophyll fluorescence

Preparing to load PDF file. please wait...

0 of 0
Quantifying spatial heterogeneity of chlorophyll fluorescence

Transcript Of Quantifying spatial heterogeneity of chlorophyll fluorescence

Quantifying spatial heterogeneity of chlorophyll fluorescence during plant growth and in response to water stress
Bresson et al.
Bresson et al. Plant Methods (2015) 11:23 DOI 10.1186/s13007-015-0067-5

Bresson et al. Plant Methods (2015) 11:23 DOI 10.1186/s13007-015-0067-5



Open Access

Quantifying spatial heterogeneity of chlorophyll fluorescence during plant growth and in response to water stress
Justine Bresson1,2,3†, François Vasseur4†, Myriam Dauzat1, Garance Koch1, Christine Granier1 and Denis Vile1*†

Background: Effects of abiotic and biotic stresses on plant photosynthetic performance lead to fitness and yield decrease. The maximum quantum efficiency of photosystem II (Fv/Fm) is a parameter of chlorophyll fluorescence (ChlF) classically used to track changes in photosynthetic performance. Despite recent technical and methodological advances in ChlF imaging, the spatio-temporal heterogeneity of Fv/Fm still awaits for standardized and accurate quantification.
Results: We developed a method to quantify the dynamics of spatial heterogeneity of photosynthetic efficiency through the distribution-based analysis of Fv/Fm values. The method was applied to Arabidopsis thaliana grown under well-watered and severe water deficit (survival rate of 40%). First, whole-plant Fv/Fm shifted from unimodal to bimodal distributions during plant development despite a constant mean Fv/Fm under well-watered conditions. The establishment of a bimodal distribution of Fv/Fm reflects the occurrence of two types of leaf regions with contrasted photosynthetic efficiency. The distance between the two modes (called S) quantified the whole-plant photosynthetic heterogeneity. The weighted contribution of the most efficient/healthiest leaf regions to whole-plant performance (called Wmax) quantified the spatial efficiency of a photosynthetically heterogeneous plant. Plant survival to water deficit was associated to high S values, as well as with strong and fast recovery of Wmax following soil rewatering. Hence, during stress surviving plants had higher, but more efficient photosynthetic heterogeneity compared to perishing plants. Importantly, S allowed the discrimination between surviving and perishing plants four days earlier than the mean Fv/Fm. A sensitivity analysis from simulated dynamics of Fv/Fm showed that parameters indicative of plant tolerance and/or stress intensity caused identifiable changes in S and Wmax. Finally, an independent comparison of six Arabidopsis accessions grown under well-watered conditions indicated that S and Wmax are related to the genetic variability of growth.
Conclusions: The distribution-based analysis of ChlF provides an efficient tool for quantifying photosynthetic heterogeneity and performance. S and Wmax are good indicators to estimate plant survival under water stress. Our results suggest that the dynamics of photosynthetic heterogeneity are key components of plant growth and tolerance to stress.
Keywords: Arabidopsis thaliana, Chlorophyll fluorescence imaging, Heterogeneity of Fv/Fm values, Modelling, Photosynthetic performance, Pixels distribution, Plant growth, Plant survival, Sensitivity analysis

* Correspondence: [email protected] †Equal contributors 1Laboratoire d’Ecophysiologie des Plantes sous Stress Environnementaux (LEPSE), INRA, Montpellier SupAgro, UMR759, F-34060 Montpellier, France Full list of author information is available at the end of the article
© 2015 Bresson et al.; licensee BioMed Central. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.

Bresson et al. Plant Methods (2015) 11:23

Page 2 of 14

Background High-throughput phenotyping is increasingly used for dissecting the genetic and eco-physiological determinisms of plant performance and stress tolerance. Over the last decade, efficient automated imaging systems have been developed for the acquisition of visible, bioluminescence, fluorescence and multi-spectral images. A rising difficulty is now to extract valuable, i.e., biologically meaningful, preferably quantitative, information from the large collection of images generated by these systems [1].
Chlorophyll fluorescence (ChlF) imaging has become one of the most powerful and popular tools to track changes in the photosynthetic capacities of plants in response to abiotic and biotic factors [2-4]. Pulse-amplitude modulated ChlF techniques provide non-invasive assessment of the photosystem II (PSII) efficiency to supply electrons to the photosynthetic machinery. Light energy absorbed by chlorophyll molecules can undergo one of three competing fates: (i) driving photosynthesis (photochemistry); (ii) being dissipated as heat; or (iii) being re-emitted as ChlF. These three processes take place in a competitive manner, and under stress conditions, the photochemistry declines whereas heat dissipation and ChlF emission increase (for recent reviews, see [5,6]). ChlF is estimated by the quantification of the light re-emitted (in the red wavebands) after the application of a saturating flash (usually for a few seconds) to the photosynthetic organs [5]. The saturating flash induces the transport of electrons through PSII centres, driving the reduction of QA, the primary stable electron acceptor of PSII. Once reduced, QA cannot accept new electrons before electrons are transferred to the next acceptor (the reaction centre is considered to be ‘closed’), and the excess of energy is dissipated through heat and fluorescence.
Amongst the different ChlF parameters, the Fv/Fm ratio is a useful and rapid parameter that reflects the maximum quantum efficiency of the PSII photochemistry [7]. In dark-adapted leaves (in which all PSII reaction centres are in the ‘open’ state; QA fully oxidized), a measuring beam is applied to elicit the minimal value of ChlF, F0 (i.e., basal fluorescence). F0 represents the energy dissipation via light-harvesting antenna pigments when excitation energy is not being transferred to the PSII reaction centres. After reaching F0, the application of a brief saturating pulse induces a maximum value of ChlF, Fm (PSII reaction centres get ‘closed’ because of electron accumulation; QA fully reduced). The difference between F0 and Fm is the variable fluorescence, Fv and Fv/Fm is given by (Fm-F0)/Fm (for more details, see [5]). Low Fv/Fm indicate substantial photoinhibition or down-regulation of PSII that occurs when plants experience stress. It has been shown that Fv/Fm is a robust indicator of plant health. Healthy photosynthetic tissues of most plant species exhibit a mean Fv/Fm at ca. 0.83, while lower values are

indicative of an impaired physiological status [8,9]. Rapid modifications of Fv/Fm are for instance reported in response to many environmental factors, such as water stress [8,10], temperature [11-13], wounding [14], photoinhibition [11,15], biotic interactions such as pathogenic as well as beneficial bacteria [16-19].
Soil water availability is one of the most important environmental factors for plant growth and development. The impact of water deficit on the photosynthetic performance of plants depends on the severity and duration of the stress. In the short-term, decrease in water supply usually induces stomata closure to maintain a favourable leaf water status, what in turn leads to a reduction of internal CO2 concentration [20]. Hence, stomata closure under water stress promotes an imbalance between the photochemical activity of PSII and the electron requirement for carbon fixation, leading to over-excitations and subsequent photoinhibitory damages to PSII reaction centres [21]. As a consequence, substantial decline in Fv/Fm in response to moderate water deficit is observed in various plant species (see references in [2]), and was closely related to decreased relative leaf water content [8]. With increasing stress severity or duration, carbon starvation and hydraulic failure, which strongly alter Fv/Fm at the whole-plant level, lead to partial (or total) senescence or leaf abscission [22]. Even though exacerbated leaf senescence can be lethal, sacrificing a few leaves might be a good strategy to ensure survival under severe resource limitation [23]. Growth recovery following severe water stress is then associated with the (partial) reestablishment of the photosynthetic capacities of the senescing leaves, and/or with the development of new leaves with optimal photosynthetic performance [24].
ChlF imaging has revealed that photosynthetic performance is extremely heterogeneous at the leaf surface, as well as between leaves, when plants experience environmental stresses. For examples heterogeneity in ChlF is reported in response to changing CO2 concentration [25], light stimuli [26], ozone-induced perturbations [27], low growth temperature [28], chilling [29], pathogen attack [16], drought [10,30] or treatment with abscisic acid [31]. Spatio-temporal heterogeneity across photosynthetic areas has been assessed by visual inspection of leaves [24,26,30], by measurements at spatially different small areas on the leaf surface [10,29,30], or by visual inspection of the shape of Fv/Fm distributions across leaves [26,28,29,32]. ChlF imaging of leaves of Arabidopsis grown under water stress for instance reveals a progressive decline of Fv/Fm beginning at the leaf’s tip [10]. However we still lack an automatic and standardized method for the quantification of the spatial heterogeneity of Fv/Fm values, which is crucial to compare photosynthetic performance depending on the developmental stage, the genotype, or the environmental conditions.

Bresson et al. Plant Methods (2015) 11:23

Page 3 of 14

Here, we analysed the distribution of Fv/Fm to estimate the spatial heterogeneity of photosynthetic efficiency (S) and the weighted contribution of the most efficient/ healthiest leaf regions to whole-plant photosynthetic performance (Wmax). We first showed that the changes in S and Wmax were related to the survival of the Arabidopsis Col-0 accession to a severe water deficit (SWD). A sensitivity analysis of S and Wmax to simulated dynamics of Fv/Fm distributions showed to what extent S and Wmax can vary depending on plant tolerance and/or stress intensity. Finally, we found that a significant part of the variation in biomass accumulation in six contrasted Arabidopsis accessions is explained by the variation of Wmax in the course of plant development.
Analytical framework: severe water deficit strongly affects
plant growth, photosynthetic efficiency and induces plant
mortality Arabidopsis Col-0 plants were grown in the PHENOPSIS automaton [33] (Figure 1A). Plants were subjected to SWD by withholding irrigation from the four-leaves stage (L4; stage 1.04, [34]; Figure 1B) in order to progressively reach a very low soil relative water content (RWCsoil) of 6% g H2O g−1 dry soil (corresponding to water potential ca. of −9.52 MPa; see Additional file 1: Figure S1). Thereafter, irrigation was resumed to progressively reach the well-watered (WW) soil condition (35% g H2O g−1 dry soil; 0.07 Mpa, Additional file 1: Figure S1) maintained until the flowering of surviving plants (Figure 1B). These two soil conditions allowed the investigation of Fv/Fm heterogeneity with highly contrasted physiological status and thus, with a wide range of leaf damages and senescence. Plant growth and Fv/Fm were daily measured from early developmental stages (i.e., emergence of the two first leaves, stage 1.02, [34]) to the emergence of the flowering stem (i.e., bolting, stage 5.01, [34]; Figure 1C), with a high-throughput ChlF imaging system (Imaging-PAM M-Series, Maxi-version, Heinz Walz GmbH, Germany) implemented on the automaton (Figures 1A, C). We developed an ImageJ (1.47v, Rasband, Bethesda, Maryland, USA) macro “PHENOPSIS-Fluo” to semi-automatically extract the whole-rosette Fv/Fm mean, the distribution of Fv/Fm values across the rosette and the projected total leaf area from ChlF images.
Under SWD, 40% of the plants survived, resumed growth and flowered whereas the remaining plants failed to recover, perished and decomposition of tissues started (Figures 1D, E). Whole-rosette mean Fv/Fm followed the variation of RWCsoil and was therefore dramatically affected by the SWD (Figures 1B and 2A). Whole-rosette mean Fv/Fm of stressed plants remained stable at 0.812 ± 0.041 (n = 4–30) during the 14 days after water withholding, similar to plants grown under WW conditions

(0.813 ± 0.019; n = 4–31; Figure 2A). Then, whole-rosette mean Fv/Fm of stressed plants decreased dramatically (Figure 2A). This was mainly due to the decrease of Fv/ Fm in the oldest leaves of the rosette, notably with a gradient from the tip to the base of the leaves (see 3-D representations in Figure 2B and Additional file 2: Figure S2). Just before rewatering, SWD resulted in a significant 38% and 43% decrease of mean Fv/Fm in surviving and perishing plants, respectively (Figure 2A). Upon rewatering, mean Fv/Fm continued to decline steadily for three further days. Afterwards, surviving plants progressively recovered Fv/Fm values up to 88% of their initial values after 6 days following rewatering (Figure 2A). This was mainly achieved by shedding of almost all senescing leaves (Figure 2B). In contrast, mean Fv/Fm of perishing plants continued to decrease to reach undetectable threshold of photosynthetic activity (i.e., plants were completely senescing or decomposing; Figure 2B and Additional file 2: Figure S2). A clear separation of mean Fv/Fm between surviving and perishing plants was visible four days after rewatering (Figure 2A).
Computing and quantifying the heterogeneity of plant
photosynthetic efficiency during growth and under
severe water deficit During SWD, Fv/Fm values at the plant surface became heterogeneous, as illustrated by the changes in the mean and distribution of Fv/Fm values (Figures 2A, B). We notably observed the establishment of multimodal distributions during SWD, reflecting the spatial variability of Fv/Fm in the rosette (Figure 2B). To explore the heterogeneity of Fv/Fm values during time course, we applied the Hartigan’s non-parametric significance test for unimodality [35-37]. As expected, the proportion of stressed plants showing multimodal distributions increased strongly after stress exposure (Figure 2C). Under WW conditions, the proportion of plants that displayed significant multimodal distributions also increased from < 10% to > 90% between 1 to 10 days after L4 stage. Stressed plants even displayed a slightly lower proportion of multimodal distributions compared to plants grown under WW conditions (Figure 2C).
After distinguishing the plants that exhibited significant multimodal distributions, we used the REBMIX algorithm for finite mixture models [38] to characterize each mode i of the mixture of distributions of Fv/Fm values (i.e., mean μi, standard deviation σi and weight ρi) for each individual rosette. All distributions displaying multimodality were accurately represented by bimodal mixtures of normal distributions where the distributions are composed of two clusters of Fv/Fm values grouping in two modes. The higher mode (maximum; μmax, σmax and ρmax; with the highest Fv/Fm values) represented the photosynthetically most efficient/healthiest parts of the rosette. The

Bresson et al. Plant Methods (2015) 11:23

Page 4 of 14

Figure 1 High-throughput analysis of Arabidopsis growth and chlorophyll fluorescence in the PHENOPSIS automaton. (A) A. thaliana
plants are grown in controlled environmental conditions in the PHENOPSIS platform equipped with a chlorophyll fluorescence imaging system.
(B) Dynamics of soil relative water content in two watering scenarios including constant well-watered conditions (WW) and water withdrawing from the four-leaves stage (L4; beginning of stress) followed by rewatering after 1 day at 6% g H2O−1 dry soil (SWD). Data are means (± SE) of 13 and 48 plants under WW and SWD, respectively. (C) Plant growth (top) and whole-rosette Fv/Fm (bottom) during plant development and under SWD. Fv/Fm values are represented by false colour scale ranging from black (pixel values 0) through red, yellow, green, blue to purple (ending at 1). (D) Visible images of surviving and perishing plants (left) and survival percentage of plants under WW and SWD conditions (right). Asterisks indicate significant differences following Chi2 test between plants grown in WW conditions (n = 13) and plants under SWD (n = 19 and 29 for surviving
and perishing plants, respectively; ***: P < 0.001). (E) Total projected leaf area of plants under WW conditions and SWD (surviving and perishing plants) as a function of days after L4 stage until bolting. Data are means (± SE) of 13–29 plants.

second mode (minimum μmin, σmin and ρmin; with the lowest Fv/Fm values) represented the least efficient or senescing parts of the rosette (Figure 3A). In case of
unimodal distribution, the mode was considered as the single maximum mode.
For plants grown under WW conditions, each param-
eter was roughly constant during plant development (Figures 3B-G). In stressed plants, while μmax essentially followed the same variation of whole-rosette mean Fv/Fm (Figures 2A and 3C), μmin decreased to reach very low

values (μmin = 0.24 ± 0.13 and 0.37 ± 0.17 for perishing and surviving plants, respectively; Figure 3B). Standard deviation σmax progressively increased during SWD establishment. However, while σmax of surviving plants recovered values similar to WW plants after rewatering, σmax continued to increase in perishing plants (Figure 3E). By contrast, standard deviation σmin increased more in surviving than in perishing plants, but recovered their initial value
13 days after rewatering (Figure 3D). In addition, the weight, i.e. the proportion, of the minimum mode ρmin

Bresson et al. Plant Methods (2015) 11:23

Page 5 of 14

Figure 2 Effects of severe water deficit on plant photosynthetic efficiency. (A) Dynamics of whole-rosette mean Fv/Fm of well-watered (WW) plants and stressed (surviving and perishing) plants under severe water deficit (SWD) as a function of days after four-leaves stage (L4; beginning of SWD) until bolting. Data are means (± SE) of 13–29 plants. (B) 3-D representations of vegetative rosettes under WW and SWD conditions in Fv/Fm false colour (from black pixel values (0) through red, yellow, green, blue to purple (ending at 1)) and their corresponding Fv/Fm distributions during time courses. Asterisks indicate p-value < 0.01 (Hartigan’s dip test) meaning significant departure from unimodality of Fv/Fm values. Arrows indicate rewatering step. (C) Dynamics of the proportion of non-unimodal (i.e., multimodal) plants under WW and SWD after L4 stage until bolting following the Hartigan’s dip test.

increased to a greater extent in perishing plants (and the weight of the maximum mode ρmax decreased likewise) compared to surviving plants (Figures 3F, G).
A quantification of the disparity between the two modes of a bimodal distribution, i.e. the heterogeneity of the values, is given by the ‘bimodal separation’ S = (μmax μmin) / 2(σmax + σmin) [39]. S is roughly the distance between the two peaks, and S > 1 when the two modes do not overlap. Here, the Fv/Fm heterogeneity across the plant increased regardless of soil water conditions during time course (Figure 4). However, S increased more in plants that survived the SWD than in others plants, whereas perishing plants had the same heterogeneity than those grown in WW conditions. A clear difference between S values of surviving and perishing plants was visible just before rewatering (Figure 4), i.e. four days earlier than mean Fv/Fm.
Quantifying the effect of photosynthetic heterogeneity on whole-plant performance: description Under SWD, S accurately represented the photosynthetic heterogeneity and allowed deciphering surviving and

perishing plants. However, it failed to quantify the effect of photosynthetic heterogeneity on plant performance and stress tolerance, as shown by the overlap of S values between WW and perishing plants (Figure 4). This is because the deviation of both modes to the photosynthetic optimum is as important as the disparity between the two modes.
It was shown from energy conversion modelling of PSII that theoretical optimum of Fv/Fm is about 0.87 in unstressed dark-adapted leaves [40,41]. However, a healthy plant displays a typical maximal mean Fv/Fm = 0.83 [8,9] and shows considerable variation around the mean. The theoretical optimum would be reached if a plant exhibits a unimodal distribution of mean 0.87 and variance 0. Hence, the photosynthetic deviation of each mode i to the theoretical optimum can be estimated as the bimodal separation Si such as Si = (0.87 - μi) / 2 σi (i.e., Smax and Smin; Figure 5A). High Si represents low photosynthetic performance of the mode i. Then, the weighted deviation to the optimum, which measured the size-corrected performance of a given mode, was calculated as Smax × ρmax and Smin × ρmin, for the maximum and the minimum

Bresson et al. Plant Methods (2015) 11:23

Page 6 of 14

Figure 3 Dynamics of the parameters describing the bimodal distributions. (A) Example of a bimodal distribution composed of two
clusters of pixels grouping in two modes. The higher mode (max; with the highest Fv/Fm values) represents the healthiest parts of the rosette whereas the second mode (min; with the lowest Fv/Fm values) represents damaged/senescing parts of the rosette. Each mode i of the mixture distribution of Fv/Fm values is characterized by mean μi, standard deviation σi and weight ρi. (B-G) Dynamics of μmax and μmin of Fv/Fm values, σmax and σmin, and, ρmax and ρmin in well-watered (WW) plants and under severe water stress (SWD; surviving and perishing plants) as a function of days after four-leaves stage (beginning of SWD) until bolting. Data are means (± SE) of 13–29 plants.

Bresson et al. Plant Methods (2015) 11:23

Page 7 of 14



SWD x surviving rewatering

SWD x perishing


Bimodal separation S


stop of watering






Days after four-leaves stage

Figure 4 Dynamics of spatial heterogeneity of whole-plant photosynthetic efficiency during development and severe water deficit. Bimodal separation (S) of Fv/Fm values of well-watered (WW) plants and stressed plants (SWD; surviving and perishing) as a function of days after four-leaves stage (beginning of SWD) until bolting. S = (μmax - μmin) / 2 (σmax + σmin) measures the distance between the modes and is superior to 1 essentially if the two modes do not overlap. Data are means (± SE) of 13–29 plants.

modes, respectively (Figure 5A). To estimate the spatial efficiency of a photosynthetically heterogeneous plant to convert light energy into chemical energy (Wmax), we calculated the proportion of Smax × ρmax (i.e., the weighted deviation to the optimum of the most efficient leaf regions) in the distribution of Fv/Fm values, as Wmax = (Smax × ρmax - Smin × ρmin) / Smax × ρmax (Figure 5A). By definition, for a unimodal distribution Wmax = 0 because there is no spatial heterogeneity (Smax × ρmax = Smin × ρmin). Basically, increase or decrease in Wmax indicates that the contribution of the most efficient/healthiest regions to the whole-plant photosynthetic performance is more or less important, respectively, than the contribution of the least efficient or senescing regions (note that Wmax has a maximum value of 1). For a heterogeneous surface (i.e., not in the first stages of plant development which display Wmax = 0 because of unimodal distributions), Wmax = 0 is assumed to be the compensation point, where the healthiest leaf regions compensate the negative effect of the less efficient leaf regions. Negative values of Wmax appear when the contribution of senescing leaf regions is prevailing.
Quantifying the effect of photosynthetic heterogeneity on whole-plant performance: applications In plants grown in WW conditions, Wmax increased progressively during development from 0 to ca. 0.85 (Figure 5B). This reflects the increase in the heterogeneity of whole-plant photosynthetic performance (i.e., a switch from unimodality to bimodality) with a very low

Figure 5 Dynamics of the spatial efficiency of a photosynthetically heterogeneous plant (Wmax). (A) Illustration of the mixture parameters in the case of a bimodal distribution. Wmax is calculated as the proportional difference in the weighted bimodal separation of each mode (Smax and Smin) to the theoretical optimum of photosynthetic performance (0.87, with standard deviation = 0), such as: Wmax = (Smax × ρmax - Smin × ρmin) / Smax × ρmax. Wmax estimates the relative contribution of the most efficient/healthiest leaf regions to the whole-plant photosynthetic performance. (B) Dynamics of Wmax of plants under well-watered (WW) and severe water deficit (SWD; surviving and perishing) conditions as a function of days after four-leaves stage (beginning of SWD) until bolting. Data are means (± SE) of 13–29 plants.
and negligible effect of the minimum mode compared to the maximum mode. In stressed plants, the increase of Wmax was delayed and reduced (Figure 5B). In surviving plants, Wmax started to decrease at 15 days after L4 stage, and recovered shortly (2 days) after rewatering. At bolting, surviving plants exhibited a Wmax of ca. 0.65, i.e. 23% less than WW plants at the same developmental stage (Figure 5B). By contrast, in perishing plants, Wmax started

Bresson et al. Plant Methods (2015) 11:23

Page 8 of 14

to decrease after 14 days following the L4 stage and became negative ten days later.
We used simple mathematical functions to model the dynamics of the parameters of the bimodal distributions in various stressing conditions, and simulate the associated variations of S and Wmax (see Additional file 3). First, this simulation exercise was sufficient to reproduce what has been observed in this paper in plants grown under SWD that did not survive the stress. The parameters of these functions were then varied to simulate different scenarios of photosynthetic heterogeneity generated by different stress intensities. Our sensitivity analysis of Wmax showed that it becomes as negative as (1) the rate of decrease in means and (2) the increase in proportion of damaged leaf regions, are high (i.e., low stress tolerance, and/or diffuse stress effect, high stress intensity). Conversely, its decrease is delayed when the rates of decrease in means and rates of changes are low (i.e., high stress tolerance, stress effects with high patchiness, and/or low stress intensity; see Additional file 3).
To explore further the possible applications of Wmax, we performed the same analysis on two other datasets. First, we used an independent dataset (not generated with the PHENOPSIS platform) to explore the genetic variability in photosynthetic performance in six accessions of Arabidopsis from contrasted geographic locations. The plants displayed little variation during plant development in mean Fv/Fm values (Figure 6A). However, we observed an increase in photosynthetic heterogeneity S and Wmax during plant development (see Additional file 4: Figure S3). We calculated the increase in Wmax during development as the slope of the relationship between Wmax and plant age. Interestingly, we found that 72% of the variability in plant dry mass at 48 days after stratification (DAS) was explained by the variation of Wmax between 17 and 48 DAS (P < 0.05; R = 0.85; Figure 6B).
Second, we investigated the effect of soil inoculation with Phyllobacterium brassicacearum STM196 strain, a plant growth-promoting rhizobacteria (PGPR) that improves plant tolerance to moderate water deficit [42] and also increases plant survival under SWD [19]. Bresson et al., 2014 [19] showed that STM196-inoculated surviving plants also exhibit a higher growth recovery after rewatering, leading to a higher plant biomass than non-inoculated plants [19]. Here, we showed that STM196-inoculation induced a faster and higher increase in Wmax than noninoculated plants from 2 days after rewatering (Additional file 5: Figure S4). This suggests that the positive effects of STM196 on growth recovery, biomass production and plant survival may be related to its effects on whole-plant photosynthetic heterogeneity.

Figure 6 Variation of Fv/Fm and relationship between Wmax and growth in six accessions of A. thaliana. (A) Dynamics of wholerosette mean Fv/Fm as a function of days after stratification (DAS). Pots (n = 4) were manually watered three times per week to maintain good (non-stressing) soil moisture. (B) Relationship between the slope of Wmax in the course of plant development and plant dry mass at 48 DAS. The accessions were collected from six different geographic origins (ICE107: South Italia; Sha: Kazakhstan; ICE111: South Italia; ICE50: Spain; Yeg-1: Caucasus; ICE228: South Tyrol). R: Pearson’s product–moment correlation coefficient.
Discussion Analysing the effects of environmental conditions on plant growth, survival and yield requires massive, rapid and non-invasive tools to track changes in plant performance. Non-invasive ChlF imaging has been developed to give insights into plant photosynthetic capacities and explore the ability of plants to tolerate various environmental stresses (e.g., [8,16,43]). Most often the mean values of various indices of ChlF, including the widely used Fv/Fm, of an organ or a plant is used to characterize the response to a stressor (e.g., [8,11]). However, a ChlF image is composed of a panel of pixels in a given range (Fv/Fm = [0; 1]). Hence, using mean values does not give a clear clue of the disparity of values that corresponds to contrasted physiology. Heterogeneity in the photosynthetic capacities of plants has been observed but rarely quantified in responses to a wide variety of external stimuli (e.g., [10,16,25-32]). For instance, the establishment

Bresson et al. Plant Methods (2015) 11:23

Page 9 of 14

of Fv/Fm heterogeneity in response to stress has been described by sampling small areas on the leaf surface [10,29,30], visual inspection of the shape of distributions [25-29,32], or by Fv/Fm clustering [16]. However, this is prone to large variations depending on the species, experimenter and stress. ChlF heterogeneity is often admitted but its standardized, objective and reproducible quantification is still lacking. For instance, previous methods used threshold-based analysis to quantify the area and progression of senescence or damages [8,16]. Here, we proposed a method to quantify (1) the heterogeneity of Fv/Fm values at whole-plant level (S) and (2) the spatial efficiency of a photosynthetically heterogeneous plant (Wmax). Although we applied our method to measurements of Fv/Fm in the Arabidopsis rosette under a severe water deficit scenario, we argue that the approach can be used with other ChlF parameters (e.g., ΦII, NPQ) as well as in response to other stressing conditions that induce variations of the physiological status.
Photosynthetic heterogeneity is intrinsic to the
development of plants Our results showed that the distribution of Fv/Fm values shifted from unimodal to bimodal distributions both under WW and SWD conditions, and this despite a constant mean Fv/Fm in WW plants. This result indicates that heterogeneity in photosynthetic efficiency (i.e., the increase in the proportion of bimodal Fv/Fm distributions) does not appear only under stress but is intrinsic to the development of plants. Importantly, S and Wmax in WW plants also significantly increased during development. It therefore indicates that, even in the absence of visible senescence, (1) there were leaf regions exhibiting lower Fv/Fm, (2) low-efficiency leaf regions increased during development and, (3) the contribution of these latter was minor on whole-plant photosynthetic performance under WW conditions. There might be different sources of photosynthetic heterogeneity. First, at the whole-plant level, photosynthetic heterogeneity in plants might be caused by age-induced leaf senescence, i.e. by visible and non-visible cell death and nutrient remobilization, notably on the edges of the oldest leaves. In addition, the increase in the size of leaf veins with increasing leaf size can also induce a decrease in the mean Fv/Fm, as well as an increase in Fv/Fm heterogeneity. Second, at the sub-cellular level, some of PSII centres are inactive to linear electron transport. Functional PSII heterogeneity is for instance expected since 70-80% of PSII are located in the stacked grana region and the remaining PSII are located in the stroma-exposed region of the thylakoid membrane [44-47].
We also showed that the variation in photosynthetic heterogeneity might be a key trait related to plant growth, as suggested by the significant correlation between the

increase in Wmax during development and biomass in six contrasted Arabidopsis accessions and despite no distinct differences in the mean Fv/Fm between genotypes. The analysis of the distributions of Fv/Fm values, as proposed with S and Wmax, allows the quantification of the whole-plant heterogeneity and may be more informative than the whole-plant mean value to investigate changes during plant development and genetic variation in plant performance.
The indicators of photosynthetic heterogeneity (S and Wmax) are linked to plant tolerance to severe water deficit Our analysis revealed that SWD affected the establishment of the intrinsic heterogeneity in plants during development. The heterogeneity of Fv/Fm values (quantified by S) across the rosette increased differently depending on the state of the plants. Importantly, S was a more sensitive indicator of the plant physiological status than the mean Fv/Fm. Indeed the mean Fv/Fm was stable during the first 14 days in stressed plants, while a strong photosynthetic heterogeneity was already present (Figures 2A and 4). S allows the discrimination between surviving and perishing plants earlier, ca. four days, than the whole-rosette mean Fv/Fm. The lag time before recovery was also shorter in S values than the mean Fv/Fm.
Surprisingly, surviving plants displayed a higher increase of S than the others plants during stress establishment, and perishing plants exhibited S dynamics similar to plants grown under WW conditions. This did not reflect the lower absolute values of Fv/Fm in perishing plants. The higher photosynthetic heterogeneity in surviving plants can be explained by the establishment of a gradient of Fv/Fm values from the tip to the base in the oldest leaves, often observed under water stress [10] and with high Fv/Fm values in the youngest leaves (as suggested in this study, see Figure 2). The analysis of the different parameters of bimodal distributions shows that SWD did not induce a global decrease of Fv/Fm, but plants rather maintained leaf regions with near-optimum Fv/Fm and sacrificed other leaf regions. Moreover, plant survival to SWD was associated to a large variability in Fv/Fm of the most damaged/senescing leaves; but to a low variability in the healthiest leaves or leaf regions (Figure 3). After rewatering, we showed that surviving plants recovered optimal mean Fv/Fm values with decreasing S, by loss of senescing leaves and/or by development of new leaves with optimal Fv/Fm. This is in accordance with the survival strategy of plants aiming at recycling and reallocating resources from the oldest or senescing leaves to active growing organs [23]. On the contrary, perishing plants displayed a decrease of Fv/Fm values in their oldest but also youngest leaves, resulting in a lower and constant value of bimodal separation S across the rosette. Perishing
PlantsWmaxSwdHeterogeneityPhotosynthetic Heterogeneity