A. thaliana ecotypes Landsberg erecta (Ler), Wassilewskija (Ws) and a gpt2- mutant (Ws WT background) were selected based on their differing abilities to undergo dynamic acclimation (Athanasiou et al., 2010) The seeds were vernalized at 4°C in a water suspension for 48 hours, prior to transfer into 6.5 cm diameter pots containing Levington M3 compost. One week after germination, seedlings were transplanted into individual pots containing 25 g of Levington M3 compost. Plants were cultivated in a Fytoscope 3000 (Photon System Instruments, PSI, Czech Republic) growth chamber, which uses a combination of red and blue LEDs (1:1 ratio throughout on a photon flux basis) plus far-red LEDs (set to a constant 10 µmol m ^-2 s^-1 throughout the day). The cabinet was set to a 12 hour photoperiod, with a 20°C day temperature, 16°C night temperature and 50% relative humidity; these conditions remained constant throughout the experiment.

Plants were split into two groups and were subject to two different light treatments: Constant light (CL) and fluctuating light (FL). The CL plants were grown under 266 ± 10 µmol m ^-2 s ^-1 for the duration of the experiment, i.e. up to 37 days. This constant pattern included an initial ramp up and final ramp down stage to represent sunrise and sunset, respectively. The FL group was subject to constant light for 28 days and then transferred into a fluctuating light for the remainder of experiment (i.e. 9 days). 9 days was selected to ensure full dynamic acclimation (Retkute et al., 2015). During growth, plants were kept well-watered. The experiment was repeated four times for Ws and gpt2- and three times for Ler.

Due to the short growth span of Arabidopsis, response to the FL treatment represents a combination of developmental and dynamic acclimation, where growth for the first 28 days was under a constant light.

The fluctuating light pattern was designed as a re-occurring 3 h 20 min light motif, which was repeated 3 times throughout the day, combined with an initial ramp up and final ramp down phase ( Figure 1 ). Each light step was a minimum of 20 min long to discount changes as a result of induction. Care was taken to accommodate different steps in light intensity, decreasing from 400 µmol m ^-2 s ^-1 to 100 µmol m ^-2 s ^-1 or increasing from 50 µmol m ^-2 s ^-1 to 400 µmol m ^-2 s ^-1. Overall, the fluctuating light pattern had the same daily integrated photon dose as the constant light pattern.

Both destructive and non-destructive measurements were made on plants. Analysis of rosette area was performed on all plants starting at 21 days after sowing (DAS). Plants were briefly removed from the Fytoscope 3000 growth cabinet into the adjoining room every other working day and were photographed using a RGB camera (Canon EOS 650D SLR, Canon Europa N.V., The Netherlands) using ambient lighting and a scale. Images were analyzed using Image J for rosette area (Schneider et al., 2012).

We assumed that increases in the size of plants followed the exponential growth as a function of time:

where d is a day from sowing, and a ₁ is relative growth rate. Curve fitting for rosette area was carried out using Mathematica (Wolfram, UK).

Following gas exchange measurements (see below), chlorophyll assays were carried out. A size 4 leaf borer was used to take 2 leaf discs per plant from leaves in the 3rd whorl, which were placed immediately in cold, 80% acetone and kept dark. Leaf samples were ground in 80% acetone and made up to 5 ml before being centrifuged for 5 min at 3000 r.p.m, 1600 g. The chlorophyll content and a:b ratios in the supernatant was determined according to (Porra et al., 1989) using absorption with a spectrophotometer at 646.6, 663.6 and 750 nm.

Whole plant light response curves (LRCs) were taken using the LI-COR 6400XT (Li-COR, Nebraska, USA) using the whole plant chamber attachment (6400-17) and RGB Light source (6400-18A for LRCS) at the end of the experiment (36 DAS+). The small size of some of the leaves precluded the use of the standard LI-COR 6400XT chamber, and measurement of the whole plant allows the response of both developmental and dynamic acclimation to be monitored. For all gas exchange measurements, plants were not dark-adapted prior to measurements. The block temperature was maintained at 20°C using a flow rate of 600 l min^-1 . For LRCs, light was provided by a combination of in-built red, blue and green LEDs, set to ‘white’ light. Illumination occurred over a series of 12 PAR values between 0 and 1500 µmol m ^-2 s ^-1, with a minimum of two minutes at each light level. At least 6 replicates were taken per experimental repeat for both CL and FL plants.

During the FL treatment period, changes in photosynthesis were measured using the LI-COR 6400XT with the whole plant chamber attachment and sun and sky lid. An ‘autologging’ program was created that took measurements every 15 seconds throughout the FL. The LI-COR was placed inside the Fytoscope chamber, with the chamber providing the light pattern to the individual plant being measured through the sun and sky lid. CO ₂ was maintained at 400 p.p.m. throughout. Due to the repeating light signature (3 hours 20 minutes long; see Materials and Methods, Plant Growth), 3 replicates were taken per day on days 1, 3, 5 and 8- post treatment (corresponding to 28, 30, 32 and 35 DAS). Autologging was carried out for two full repeat experiments of the WTs Ler and Ws (i.e. 6 replicates per post treatment day) and one full experiment for the gpt2- mutant. The data was normalized according to the average photosynthesis during the last 10 time points at the end of the light pattern and then averaged.

Curve fitting for LRCs was carried out using the Mathematica (Wolfram, UK). The net photosynthetic rate, or assimilation, A, as a function of irradiance, L, can be described using the non-rectangular hyperbola [41]:

The nonrectangular hyperbola is defined by four parameters: the quantum use efficiency (QY), ϕ; convexity, θ; maximum photosynthetic capacity, P_max and; the rate of dark respiration, R_d . We assumed that the rate of dark respiration is proportional to the maximum photosynthetic capacity, according to the relationship R_d = αP_max (Givnish and Vermeij, 1976; Niinemets and Tenhunen, 1997; Retkute et al., 2015). Fitting was performed using the Mathematica command ‘FindFit’ with a minimum constraint on dark respiration at 0.05 and convexity at 0.8.

A model incorporating a ‘fading memory’ of the recent light pattern in the form of a time-weighted average for the light was introduced in (Retkute et al., 2015):

This describes the ability for plants to respond more strongly to recent changes in light history. This effectively accounts for photosynthetic induction state, which is very hard to quantify in situ as it varies according to the light history of the leaf.

This fading memory was incorporated into the light response functions when calculating instantaneous photosynthetic rate, P, at a time t. The model was adapted so the time-weighted average was only applied during the transition from low to highlight (to represent induction) but not from high to low light, during which photosynthesis can almost immediately respond.

We estimated distributions of τ, P_max , α, θ and ϕ for each accession and each day. Parameters were fitted using adaptive multiple importance sampling (Retkute et al., 2021) with likelihood formulated assuming normally distributed errors between photosynthetic rate measured in the experiments and simulated using Eq.4. Parameter prior distributions were assumed to be uniform within following ranges: τ ϵ [0,60], P_max ϵ [0,30], α ϵ [0.03,1], ϕ ϵ [0,1] and θ ϵ [0.6,0.99].

Analysis of variance (ANOVA) was carried out using GenStat for Windows, 17th Edition (VSN International Ltd.). An unbalanced design was used to account for differences in the number of replicates each round (i.e. due to plant mortality). The data was checked to see if it met the assumption of constant variance and normal distribution of residuals. For all statistical analyses, data from each of the lines (Ws, Ler and gpt2-) were treated independently because of their differing responses to a change in light. Rosette area was analyzed at 28 DAS and 37 DAS. The former was carried out to ensure that plant growth was same in the CL treatment relative to FL treatment during the period of growth under which they were subject to the same light pattern (grey horizontal line; Figure 1 ) whilst the latter was to determine whether the fluctuating light pattern influenced growth and final rosette area. Chlorophyll a:b ratio, total chlorophyll content and length of memory fading window were analyzed using ANOVAs.

Rosette area of each accession under CL and FL during the course of the experiment is given in Figure 2 . Whilst this does not account for overlapping leaves or leaf thickness, it can be used as an approximation of growth rate. Two treatment comparisons were performed on each accession in order to see if there were any differences in growth of the plants: first at 28 DAS to ensure consistent growth prior to FL treatment and; secondly at the end of the experiment (37 DAS). At 28 DAS there was no significant difference in rosette area between the treatments in any line, which suggests that up to that point the plants grew similarly. Similarly, at 37 DAS there was no significant difference in the rosette area for any line indicating that 9 days of FL did not significantly alter growth. We estimated the relative [visible] leaf area expansion to be between 5.5 and 6.5 cm ² per day (fitted curves in Figure 2 ).

LRCs indicate a significant increase in P_max for both the wild type accessions ( Figure 3 : Ws p=0.023; Ler p<0.001), indicative of acclimation to high light. In comparison, the P_max for the gpt2- mutant significantly decreased (p=0.013). Direct comparisons between Ws and gpt2- under CL showed no significant difference in P_max . However, under FL, P_max in the mutant was significantly lower, indicating the importance of gpt2- to FL.

Similar to that seen for P_max , fitted values indicate that for both the WT accessions, QY was significantly higher in plants under FL compared to those under CL (p<0.001 for Ws and p=0.006 for Ler); a further indicator of high light acclimation. There was no significant difference in QY for gpt2- under CL versus FL. There was a significant decrease in LCP in the FL plants for Ws (p=0.012) but not Ler or gpt2-. A decrease in LCP is an indicator of acclimation to low light. Ler exhibits an increase in R_d in fluctuating light, however, no difference in R_d was found in the other two genotypes.

There was no significant difference in chlorophyll a:b in Ws, and for both Ler and gpt2- the chlorophyll a:b ratios were significantly lower in the FL plants compared to the CL plants (p=0.049 and p=0.004, respectively) ( Figure 4 ). Moreover, for both Ws and gpt2- the total Chl content was significantly lower in plants under FL compared to those under CL (p<0.001 and p=0.002, respectively). For Ler, no significant change was observed between total Chl amount.

The light motif was split into 8 stages according to the irradiance level (Colored bars; Figure 4 ). For stages 1-4 and 6 (corresponding to the light intensity of 400, 100, 200, 50 and 100 µmol m ^-2 s ^-1, respectively), the average normalized photosynthesis value of the last 50 time points during the step (i.e. at steady state) was calculated. For stage 5, the time taken to reach a normalized photosynthesis value of 0.7, 0.8 and 0.9 was calculated as a proxy for rate of change. There was no significant difference in days post treatment for any of the lines at stages 1-3 and 6. For gpt2- and Ws, there was no significant difference during stage 4 (i.e. the step at 50 µmol m ^-2 s ^-1) or 5 (i.e. the step from 50 to 400 µmol m ^-2 s ^-1). However, for Ler there was a significant difference (Stage 4, p= 0.017; Stage 5, p=0.044 and p=0.036 for a normalized photosynthesis of 0.8 and 0.9, respectively). This indicates that in the days following a change in the light environment, Ler was able to respond more quickly to a change in irradiance compared to the other genotypes, thus indicating the importance of the light history.

The time-weighted average (Equation (3) acts as a ‘fading memory’ of the recent light pattern and uses an exponentially decaying weight. If τ = 0 then a plant will able to instantaneously respond to a change in irradiance, whereas if τ< 0 the time-weighted average light pattern will relax over the timescale τ. Previous data from Arabidopsis indicates that τ ≈ 0 hours (Retkute et al., 2015). This value of τ (0.3h) represented a maximum leaf ‘memory’ of around 18 minutes that exponentially declines according to time spent in the light.

For this study, a model, given by Eq.(4), was fit to values of photosynthetic rate during stage 5 of the light motif, i.e. after switching irradiance from 50 µmol m ^-2 s ^-1 to 400 µmol m ^-2 s ^-1 ( Figure 5 ; Supplementary Figure S1 ). This time period was selected because it showed the strongest response to change in irradiance levels. There were statistically significant differences between accessions (p<2e^-16) and between days (p<2e^-16). Ler had the highest values of the fading memory window, corresponding to the slowest photosynthetic induction (16.9 - 20.8 minutes; Figure 5A ). Ler plants showed tendency for decrease in τ with more days spend under FL regime, corresponding to an increase in P_max during the latter days of the experiment ( Figure 5B ). The length of photosynthetic response for Ws was estimated to be between 13.8 and 16.5 minutes. The fastest response to increase in light intensity was for gpt2-. However, the response time increased with number of days spend under FL for both Ws and gpt2- ( Figure 5A ), yet there was still an increase in P_max during the course of the experiment for gpt2- but not the WT Ws. Overall, a good correspondence between the fitted model and experimental measurements was found for all accessions and days ( Supplemental Figure S1 ).

One of the commonly cited functions of acclimation is maintenance of photosynthetic efficiency under the new light regime. For example the lowering of light compensation point, R_d and antenna size under low light. However naturally fluctuations create a dilemma in which both low light and high light acclimation states would be beneficial during the same photoperiod. How do plants deal with this problem? We argue that these features, often seen as fixed to high light or fixed to low light, are not necessarily in conflict ( Figure 6 ). Low light acclimated leaves can support high photosynthetic rates as long as photoprotective mechanisms are engaged and other stress factors such as high leaf temperature are not present. Up-regulation of electron transport components can induce a higher P_max during dynamic acclimation without change in Calvin cycle components (Murchie et al., 2005; Vialet-Chabrand et al., 2017).

In consensus with the literature, both WT plants exhibited an increase in P_max under fluctuation light ( Figure 3 ) (Murchie and Horton 1997; Yin and Johnson, 2000; Walters, 2005; Li et al., 2009; Athanasiou et al., 2010; Dyson et al., 2015; Retkute et al., 2015; Vialet-Chabrand et al., 2017). Increasing P_max under high light is usually mirrored by a step-wise increase in Chl a:b ratio, due to loss of light harvesting complex (LHCII) (Anderson, 1980; Anderson, 1986; Murchie and Horton, 1997; Bailey et al., 2001; Scheibe et al., 2005). Vialet-Chabrand et al. (2017) found that in the accession Col-0, there was an increase in LHCa1 in non-fluctuating conditions, indicating that fluctuating light was inducing a preferential high light response despite the occurrence of low light periods in the regime. Within this study, a lower Chl a:b ratio was observed for Ler in the FL treatment compared to CL treatment ( Figure 7 ), this combined with the significant increase in P_max ( Figure 3 ) plus a change in photosynthesis at 50 µmol m ^-2 s ^-1 in the days following a change in the light treatment ( Figure 4 ) suggests features of acclimation to both the high- and low- light are present (Anderson, 1980; Anderson, 1986; Bailey et al., 2001; Walters et al., 2004) ( Figure 6 ). Whilst Chl a:b and leaf Chl content were typical of a low light response, P_max was independent of this. With the exception of Ws, the Chl a:b responses indicate acclimation of the antenna to low light, not high light upon transfer to FL. Therefore, chlorophyll traits operated independently to the lack of P_max acclimation conferred by the GPT2 gene. The reduction in chlorophyll content under FL could therefore have been achieved by a reduction in chloroplast number, size or cell size whilst the stromal fraction for Rubisco content increased to confer the higher P_max . The origin of the higher P_max would therefore need to be determined.

Contrary to previous experiments, there was no significant difference in R_d for any accession (Yin and Johnson, 2000; Dyson et al., 2015), although this may be due to the difficulty in measuring dark respiration within the whole plant chamber. R_d is an essential component of light acclimation and normally rises and falls in line with P_max . A high R_d would be disadvantageous under low light periods.

Bailey et al. (2001) used constant light to identify three distinct phases of acclimation in Ler; a low light response, a high light response and a less pronounced response at intermediate light intensities, which they linked to changes in the content and composition of the thylakoid components as well as both photosystems. This is consistent with the suggestion that the regulation of P_max and Chl a:b is largely independent (Bailey et al., 2001). Whilst acclimation to high light was also exhibited by Ws, the corresponding change in chlorophyll was not seen. However, a decrease in Chl a:b ratio under FL vs CL was also observed in the gpt2- mutants suggesting that the mutant was still able to acclimate to lower light intensities, but not higher intensities. It is unclear why the Ws background would show a different response to gpt2- when the P_max shows the typical response to HL. These findings suggest that the effect of fluctuating light on acclimation is to invoke multiple pathways that allow an acclimation response to both high and low irradiances.

Athanasiou et al. (2010) grew plants outdoors in unheated green houses, and found that under naturally fluctuating light, WT Ws had a higher fitness relative to gpt2- mutants and WT Col. The same study also showed the inability for Col to dynamically acclimate to an increase in irradiance under controlled conditions (see Supplementary Figure 1 in Athanasiou et al. (2010)). The fact that Col expresses GP T2 but does not acclimate (Niewiadomski et al., 2005) supports the hypothesis that alternative pathways, not involving GPT2, are necessary for dynamic acclimation under fluctuating light. The role of GPT2 is to translocate G6P into the chloroplast to enable starch synthesis (Kunz et al., 2010). This results in an increased chloroplastic phosphate pool, causing changes in gene expression enabling acclimation to high light. In contrast, gpt2- mutants have a higher photosynthesis rate at lower light levels than the parental WT Ws plants (Dyson et al., 2015). However, this improved photosynthetic rate may be a result of developmental acclimation, not dynamic. Nevertheless, combined with the significantly lower Chl a:b ratio found within this study, this suggests that acclimation components can operate independently.

So far we have taken a simplistic approach, focusing on antenna size as given by Chl a:b and features of the light response curve notably P_max which may be limited by Rubisco, stomatal conductance and electron transport rate depending on conditions. We may consider any component of the chloroplast to be part of acclimation and further work is needed to determine how these operate in relation to each other. For example, changes in Rubisco concentration and activity, along with molecular changes such as cytochrome-b/f activity and LHC and photosystem stoichiometry (Murchie and Horton, 1997; Walters et al., 1999; Yano and Terashima, 2001; Walters et al., 2004; Suorsa et al., 2012). We may determine some general trends, i.e., levels of Cytochrome-b/f, ATPase, plastoquinone and Rubisco will be needed to achieve high P_max , the question is how much. Vialet-Chabrand et al. (2017) showed that election transport was more important than Rubisco in Arabidopsis Col-0. There may be much variation in nature: there is evidence that in developmentally acclimated plants, some species do not change their chlorophyll a:b ratios (Murchie and Horton, 1997; Zivcak et al., 2014). Furthermore, some changes in chlorophyll a:b ratio have been attributed to genes involved in LHCII distribution, which is a known induction n response (Allen and Forsberg, 2001; Depge et al., 2003; Bellafiore et al., 2005; Vainonen et al., 2005; Suorsa et al., 2012; Mekala et al., 2015; Retkute et al., 2015).

In previous work we formulated a mathematical framework of dynamic acclimation that defined the optimal adjustments to net photosynthesis under fluctuating light conditions (Retkute et al., 2015). Applied within this study we describe two key aspects of the acclimation process: first the rate of acclimation itself and second ‘entrainment’ of P_max by the fluctuating light. As described previously, transient high light events induce a ‘fading memory’ which influences both the induction state and the likelihood of inducing acclimation ( Figure 5 ). We show a new feature of acclimation in fluctuating light which is the ‘fading memory’ or τ which varies according to genotype. This indicates that there is genetic variation for both speed of response to fluctuating light but also in the sensitivity to which the plants sense and measure light transients.

Previous research on variation among Arabidopsis accessions found variation between Ler-0 and Ws-2 in the PSI/PSII ratio, the lateral mobility of the thylakoid membrane and in chlorophyll protein complexes (Kaiser et al., 2020; Wójtowicz and Gieczewska, 2021). In particular, the authors proposed that increased value of the nonphotochemical quenching qN or NPQ reported for Ler-0 under control conditions might limit the capacity for the photosynthetic apparatus to adapt to changing light intensities. NPQ is known to be sensitive to changes in the energy status of the chloroplasts (energy-dependent quenching) and thus presents the most sensitive parameter for the early detection of such changes. Whilst Kaiser et al. (2020) did not assess Ws, they found that certain photosynthetic traits were correlated with the ecological niche to which the accession originated. For example, they found that a reduction in PSII operating efficiency (ϕPSII) under fluctuating light correlates with latitude: with those originating further north exhibiting the lowest ϕPSII. Therefore our conclusion for variation in terms of the timing and rate of onset of acclimation is consistent with the known variation among Arabidopsis accessions.

Fluctuating light experiments have been performed on plants under natural conditions, these found that the ability to acclimate provided a fitness advantage by optimizing photosynthetic efficiency for a new environment (Athanasiou et al., 2010; Suorsa et al., 2012). However, in these experiments, plants were subject to fluctuations in temperature and humidity as well as light thus entangling photosynthetic acclimation to irradiance from that of temperature or humidity is difficult to achieve. Nevertheless, more realistic representations of the natural environment will be critical for determining the adaptive significance of acclimation and determining the limits placed on plants. However, quantifying the physiological response of plants under environmentally relevant conditions is extremely difficult (Walters, 2005).

Photosynthesis in nature responds largely to fluctuating light in addition to the fixed longer term square waves commonly used for studies in photoacclimation (Poorter et al., 2016; Vialet-Chabrand et al., 2017). Confounding this are species- and genotype-specific differences in plant structure as well as physiological capacity will influence the overall impact of growth conditions on performance (Murchie and Horton, 1997; Athanasiou et al., 2010; Burgess et al., 2017; Burgess et al., 2021). Acclimation is a complex process potentially involving most photosynthetic components in the chloroplast and experimental data indicates that the past light history of a leaf is critical in determining the optimal P_max under a given light level (e.g. Figure 5 ; Retkute et al. (2015)). Whilst this can be controlled or determined relatively easily within small plants with simple structures, such as Arabidopsis, knowledge of the past light history is difficult to obtain for larger plants, or crop plants, like rice (Oryza sativa) and wheat (Triticum aestivum) (Murchie et al., 2002; Murchie et al., 2008; Townsend et al., 2018; Burgess et al., 2021). The complex canopy structure of these plants combined with environmental factors such as weather conditions and wind, cloud or solar movement mean that a given section of leaf within the same plant will be subject to light changes that vary in frequency and longevity (Burgess et al., 2015; Burgess et al., 2017; Burgess et al., 2021). Knowledge of the underlying mechanisms of this process, what fitness advantages acclimation provides and how it could be manipulated will therefore be critical in targeting crops for improved productivity and yield.