Shade avoidance responses become more aggressive in warm environments

2.1 Plant material

We used seedlings of A. thaliana of the wild-type (WT) Columbia background in all the experiments. WT plants and the phyA-211, phyB-9, phyA-211 phyB-9, cry1-304, cry2-1, and cry1-304 cry2-1 mutants (for references, see Sellaro, Pacín, & Casal, 2012) were included in all the experiments, except those aimed to model the action of either phyA or cry1 and cry2, which included only the relevant mutants. For all the experiments, 10–15 seeds per genotype were sown on 8 ml of 0.8% agar in clear plastic boxes and kept in the dark at 4°C for 4 days. Then, they were transferred to white light conditions for 3 days in a growth room at 22°C with a photoperiod of 10 hr under 100 μmol·m⁻²·s⁻¹ (400–700 nm) provided by a mixture of fluorescent and halogen lamps (red/far-red ratio = 1.1). At the time of initiation of the photoperiod of the fourth day, the seedlings were transferred to the corresponding experimental light conditions.

2.2 Hypocotyl length increment

For field experiments and all indoor experiments under constant conditions, pictures of the seedlings were taken at the beginning and at the end of the photoperiod of the fourth day. For experiments with fluctuating temperatures, pictures were taken every 2.5 hr during the photoperiod of the fourth day. For experiments with fluctuating light, pictures were taken at the beginning, at the middle, and at the end of the photoperiod of the fourth day. Hypocotyls were measured using image processing software (Sellaro, Hoecker, Yanovsky, Chory, & Casal, 2009), and length increments were divided by the duration of the period to obtain growth rates. Data obtained in experiments under controlled conditions are shown in Table S1, and those obtained in the field are shown in Table S2. When using growth data to either adjust or test the growth model, the hourly growth rates were converted to length increments during a period of 9 hr (by multiplying by 9) for consistency with the original model (Legris et al., 2016).

2.3 Light treatments under controlled conditions

To model the contribution of phyA (see Figure 1a,b), we used five different mixtures of red and far-red light provided by 150-W incandescent R95 lamps (Philips) in combination with a water filter, a yellow, orange, and red acetate filter set (LEE filters 101, 105, and 106, respectively), either alone or plus a green acetate filter (LEE filters 138, 121, or 089), or six blue acrylic filters (Paolini 2031, Buenos Aires, Argentina). Scans of the light fields were obtained with a spectroradiometer (USB400, Ocean Optics, USA). The red/far-red ratios were 1.09, 0.44, 0.20, 0.07, and 0.001, and the irradiances ranged from 63 to 834 μmol·m⁻²·s⁻¹ (spectra are shown in Figure S1a). Spectral data were used in combination with photoconversion cross-section data (Mancinelli, 1994) to obtain the rates of photoconversion from Pr to Pfr (k1) and from Pfr to Pr (k2). To model the contribution of cry1 and cry2 (see Figure 1c,d), blue light was provided by light-emitting diodes (the spectrum is shown in Figure S1b).

To test the model under stable conditions of light and temperature, simulated sunlight was provided by a combination of fluorescent and halogen lamps (100 μmol·m⁻²·s⁻¹, between 400 and 700 nm), and increasing degrees of shade were simulated by the combination with different green acetate filters (LEE filters 138, 121, or 089 for mild, medium, and deep shade, respectively). The red/far-red ratios were 0.76, 0.41, and 0.12, and the irradiances between 400 and 700 nm ranged from 18 to 70 μmol·m⁻²·s⁻¹ (spectra are shown in Figure S1e). Different growth chambers were used for the different temperature conditions (14°C, 22°C, and 28°C).

For experiments with temperature fluctuations (see below Figure 2), we used a growth chamber (Percival E-308; Percival Scientific Inc., Perry, IA, USA) with fluorescent light tubes (30 μmol·m⁻²·s⁻¹ between 400 and 700 nm, the spectrum is shown in Figure S1c) and simulated real temperature patterns obtained from climate stations at three different locations in Argentina (25 de Mayo, Ingeniero Jacobacci, and Puerto Pirámides) for a date when photoperiod was 10 hr. The three cases show similar medium temperature but different thermal amplitudes.

For experiments with light fluctuations (see Figure 3), we used a light-emitting diode panel with three light channels: red, far red, and blue (spectra are shown in Figure S1d). For the control, each one of the three channels was kept constant at 9 μmol·m⁻²·s⁻¹. For the blue light treatment, the red and far-red channels were kept constant at 9 μmol·m⁻²·s⁻¹, and blue light intensity was increased 20% every hour (starting with 3.6 μmol·m⁻²·s⁻¹) during the first half of the photoperiod (when it reached maximum intensity, 18 μmol·m⁻²·s⁻¹) and then decreased symmetrically, that is, 20% every hour during the second half of the photoperiod. The same pattern and intensity values were used for red plus far-red treatment, where the blue channel was kept constant and the red and far-red channels were shifted during the photoperiod. For the blue plus red plus far-red treatment, all the channels were shifted during the photoperiod following the same pattern.

2.4 Field experiments

On the fourth day, light-grown seedlings of all the genotypes were transferred either to unfiltered sunlight or under one of two different canopy shade conditions (mild and deep shade) in the experimental field of Facultad de Agronomía, University of Buenos Aires (latitude: 34.5914, longitude: 58.4798). Ten different experiments were performed at different dates, starting in June (winter) and ending in February (summer) to obtain a range of different light and temperature combinations. Temperature inside the growth boxes that contained the seedlings was recorded hourly by using a datalogger (DS1922L-F5# Thermochroni Button, Maxim Integrated, California, USA). Photosynthetically active radiation, blue, red, and far-red light, were recorded hourly by using a 4-channel light sensor (SKR 1850, Skye Instruments, Llandrindod Wells, UK).

2.5 Weather data

To investigate the impact of diurnal temperature fluctuations on hypocotyl growth rate, we searched a local weather database (Sistema de Información y Gestión Meteorológica INTA, Argentina, http://siga2.inta.gov.ar/). We identified locations and dates where the photoperiod was 10 hr. Then, we selected tree cases where mean temperature was similar (close to 10°C) and thermal amplitude was contrasting. One of the locations (25 de Mayo) showed considerable thermal amplitude; the second (Ingeniero Jacobacci) showed small thermal amplitude; and the third (Puerto Pirámides) showed thermal inversion caused by the important oceanic influence.

To apply the growth rate model at a global scale of current weather, maximum temperature, minimum temperature, and solar radiation data for 100 different locations were obtained from the Climate Change, Agriculture, and Food Security-Climate data portal (http://www.ccafs-climate.org). The processing of data sets was conducted by using GIS software (QGIS 3, http://www.qgis.org/). The data correspond to dates when the photoperiod is close to 10 hr (this excludes latitudes where days are longer throughout the year).

For future projections (year 2080) of weather conditions, the climate model mpi-echam5 (http://ccafs-climate.org/data_spatial_downscaling/) and the emission scenario A1B were used. The latter scenario predicts an integrated world characterised by rapid economic growth, a population that reaches 9 billion by 2050 and then declines gradually, and the rapid development of alternative energy sources that facilitate increased economic growth while limiting and eventually reducing carbon emissions. The starting point (current weather conditions) was defined by the conditions obtained from the Climate Change, Agriculture, and Food Security-Climate data portal, as described above.

2.6 Bioluminiscence assay

For bioluminescence assays, we exposed 3-day-old light-grown seedling of the Columbia WT carrying the PIF4:PIF4-LUC transgene to a photoperiod with irradiance or temperature manually adjusted hourly to obtain a peak at midday, compared with a control under constant light and temperature. Seeds carrying the bioluminescent PIF4 reporter were kindly provided by Salomé Prat (Centro Nacional de Biotecnología, Madrid, Spain) and will be described in further detail elsewhere. In the morning, light increased linearly from 41 μmol·m⁻²·s⁻¹ during the first hour to 230 μmol·m⁻²·s⁻¹ during the fifth hour. Similarly, temperature increased linearly from 13°C during the first hour to 27°C during the fifth hour. In both cases, the opposite trend was applied during the afternoon. The control was exposed to 135 μmol·m⁻²·s⁻¹ and 20°C throughout the 10-hr photoperiod. Twenty-four hours before starting luciferase readings, 20 μL of 0.5-mM d-luciferin were added to each well. Luciferase activity was detected with a Centro LB 960 (Berthold, Bad Wildbad, Germany) luminometer.

2.7 Structure of the growth rate model

The final structure of the growth rate model is as follows (Table S3):

where G represents the length increment during 9 hr and G ₀ the maximum attainable length increment. The denominator includes terms related to conditions that reduce growth rate from its maximal plus one; that is, in the absence of inhibitory conditions, growth rate is maximal (G = G ₀/1). The first three inhibitory terms are as described previously (Legris et al., 2016) and, respectively, incorporate the effect of phyB (which depends on light and temperature conditions), the effect of low temperature, and the effect of the interaction between phyB and low temperature. We use these three terms throughout the paper. The original model contained a fourth inhibitory term to incorporate the inhibitory effect caused by other photoreceptors. Here, we replace this term for two more specific terms, one for phyA and the other for cry1 and cry2. These terms were parameterized by analysing the impact of phyA, cry1, and cry2 mutants under different light inputs. We then run the model with five inhibitory terms against growth rate data including four light/shade conditions, seven genotypes, and three temperatures. There, we observed that the model failed to predict the maximum growth rate (G ₀) and the effect of low temperature (a) in the phyA phyB double mutant (that had not been used in previous steps of the model). Furthermore, we observed biases for the effect of cry1 plus cry2 (e) at 28°C in all genotypes and at all temperatures (but with different intensities depending on whether temperature was either ≥22°C or <22°C) in phyA phyB. Therefore, we incorporated specific parameters for these cases (Table S3). The five inhibitory terms (some with different parameters for specific genotypes/conditions) were used in all subsequent steps. Although the model improved with the definition of parameters for different temperature ranges, this should not be taken as evidence for the occurrence of thresholds because only three temperatures were used in the analysis. The last term represents the impact of thermal amplitude and was parameterized in experiment with WT and photoreceptor mutant genotypes under conditions where the plants were grown under similar average temperatures but different thermal amplitudes. This analysis yielded different effects of thermal amplitude (f) depending on the genotype. We finally tested the model against field growth rate data.

3.1 Basic growth rate model under stable light and temperature

The model developed by Legris et al. (2016) predicts the average hypocotyl growth of a 3-day-old light-grown seedling during the photoperiod as a function of phyB activity (Table S3a), temperature (Table S3b), the interaction between phyB and temperature (Table S3c), and white light irradiance. The last term is used to incorporate the activity of the additional photosensory receptors (phyA, cry1, and cry2), which are not included in the term corresponding to phyB. To make the model more accurate under natural radiation conditions, where changes in spectral composition can alter the relative activation of different photosensory receptors, we modelled the specific contributions of phyA, cry1, and cry2 under controlled conditions. As in previous models, growth rate (G) is represented as a function of maximum growth rate (G ₀) divided by the sum of all the terms that reduce growth below the maximum (phyB, low temperature, interaction between phyB and low temperature, phyA, cry1, and cry2) plus 1 (Table S3), that is, G = G ₀ if there are no inhibitory terms.

To parameterize the term corresponding to phyA, we measured G both in WT and the phyA mutant 3-day-old light-grown seedlings exposed to a photoperiod of different mixtures of red plus far-red light, each one of them at a range of irradiances. Figure 1a shows the G ₀/G ratio for one of these red plus far-red light mixtures plotted against the sum of the rates of photoconversion from Pr to Pfr (k1) and from Pfr to Pr (k2). G ₀/G increases above 1 when there is inhibition (i.e., G < G ₀). For a given spectral composition, k1 + k2 increases with irradiance. The phyA mutant retained a significant degree of inhibition (equivalent to the phyB-mediated inhibition predicted by Legris et al., 2016), and the difference in slope between the WT and phyA provides an estimate of the inhibitory effect of phyA. Figure 1b shows the difference in slope between the WT and phyA observed in plots similar to that shown in Figure 1a but with different red/far-red ratios. The effect of spectral composition is captured by the k1/(k1 + k2) ratio (we prefer the latter to the red/far-red ratio itself because the photoconversion rates are calculated with information provided by the whole spectrum [Mancinelli, 1994]). The inhibitory effect is maximal for almost pure far-red light (the lowest k1/[k1 + k2]) and steeply decreases with increasing proportions of red light to reach a plateau. This pattern is consistent with the well-known maximal effect of phyA under far-red light (Quail et al., 1995), the contribution of phyA even under pure red light when the irradiance is high enough (Franklin et al., 2007), and the lack of effects of red/far-red ratio in the upper range of ratios (Sellaro et al., 2010). Then, the contribution of phyA was modelled by incorporating to the denominator a term where the activity of phyA depends on irradiance (k1 + k2) in a manner that, in turn, depends on spectral composition and more specifically red/far-red ratio (k1/[k1 + k2]; Table S3d).

To parameterize the terms corresponding to cry1 and cry2, we measured G both in WT and cry1 cry2 mutant 3-day-old light-grown seedlings exposed to a photoperiod of different irradiances of blue light. Figure 1c shows the G ₀/G ratio plotted against blue light irradiance (400–500 nm), and the difference in slope indicates the combined contribution of cry1 and cry2. To discriminate the role of each one of these receptors, we compared the double mutant with the cry1 and cry2 single mutants under a range of simulated sunlight and shade conditions and calculated the average ratio between G for each mutant and G for cry1 cry2. Figure 1d indicates strong genetic redundancy between cry1 and cry2 because the cry2 mutation on its own causes a reduction of G (reflected mathematically as “negative inhibition”) but significantly released the growth potential when combined with the cry1 mutation. The contribution of cry1 and cry2 was modelled by incorporating a term where blue light irradiance is multiplied by the slope of response when both cry1 and cry2 are present (Figure 1d) multiplied by the specific contribution (Figure 1e, Table S3e).

To test the performance of the model, seedlings of the WT and of the phyA, phyB, phyA phyB, cry1, cry2, and cry1 cry2 were exposed to different simulated sunlight and shade conditions and different temperatures. Figure 1e shows the relationship between observed and predicted G, which did not significantly depart from the 1:1 line and showed adequate accuracy.

3.2 The impact of daily temperature fluctuations

Although the model was able to capture the impact of temperature, the above experiments were conducted under conditions where this parameter remained constant through the photoperiod, which is not typical of natural settings. To investigate the impact of temperature fluctuations, we first searched for natural patterns characterised by divergent daily trends and not very large differences in average temperature, in places with a photoperiod of 10 hr (i.e., similar to the photoperiod used in our indoor experiments). We selected three locations in Argentina (Figure 2a), and we exposed 3-day-old light-grown seedling of the WT and photosensory receptor mutants to a stable photoperiod in a growth chamber manually set to reproduce the three temperature patterns and a constant temperature control. We then measured G at different times of the photoperiod (Figure 2b). The analysis of the residuals unaccounted by the growth rate model indicated that the accuracy of fit was improved by a factor reducing growth rate with increasing thermal amplitude in seedlings of the WT and of the phyA and cry2 mutants (Table S4). This effect was not significant for the phyB, cry1, and cry1 cry2 mutants and was actually negative in the phyA phyB double mutant (Table S4). Thermal amplitude was calculated as the difference between maximum and minimum temperature with a negative signal if temperature decreases rather than increase towards midday, as in the case of Puerto Piramide (Figure 2a). Our interpretation of this effect is that in the WT (as well as phyA and cry2), during the photoperiod maximum, hypocotyl growth rate occurs in the morning, and then growth decreases gradually (Nozue et al., 2007; Sellaro et al., 2012). Therefore, if a given average temperature results from lower morning temperatures and higher values at midday, there is a coincidence between lower temperature and the time of maximal growth potential, which lowers average G. The phyB, cry1, cry1 cry2, and phyA phyB mutants have higher rates not in the morning but later in the photoperiod and, therefore, do not become affected by temperature fluctuations in the same way. The model incorporating this additional term (Table S3f) accounted for the G values observed under fluctuating temperatures with reasonable accuracy (Figure 1c).

3.3 The impact of daily light fluctuations

Following the same argument used for the analysis of fluctuating temperatures, we exposed 3-day-old light-grown seedling of the WT and photosensory receptor mutants to a photoperiod with irradiance manually adjusted hourly to obtain a peak at midday. We modified the red plus far-red light or the blue light component leaving the other constant or both components, and we also included a control without temperature fluctuations (Figure 3a). We then measured G during the first and second halves of the photoperiod (Figure 3b). The model (which is based on average irradiance values) was able to predict G under these conditions with reasonable accuracy, and no need for a correction associated to light fluctuations was apparent (Figure 3c).

3.4 The impact of temperature or light fluctuations on PIF4 dynamics

The above experiments indicate that, compared with a stable condition with the same average temperature and irradiance, fluctuations in temperature peaking at midday tend to reduce growth, whereas fluctuations in irradiance do not have large effects. Because one of the most important players in the control of hypocotyl growth by light and temperature is PIF4 (Franklin et al., 2011; Hornitschek et al., 2012; Huq & Quail, 2002; Niwa, Yamashino, & Mizuno, 2009), we investigated whether light and/or temperature fluctuations affect PIF4 dynamics. Figure 4 shows that compared with the stable conditions, increasing thermal amplitude significantly decreased PIF4 levels at midday, whereas fluctuating light only had effects at the end of the photoperiod. This observation offers a molecular correlate to the observed pattern of the physiological output.

3.5 Model performance in the field

To test the model developed under controlled conditions, we exposed 3-day-old light-grown seedling of the WT and photosensory receptor mutants to a photoperiod (10 hr) of natural radiation either under full sunlight or under two levels of natural shade provided by grass canopies of different stature. The experiment was repeated on 10 different dates between the beginning of winter and mid-summer to obtain a range of temperatures and irradiances. Light and temperature conditions inside the boxes where the plants were grown were monitored every hour. The daily average values were used as inputs to the model to estimate G (Table S3h). Figure 5 shows the relationship between observed and predicted data. The relationship did not departure significantly from the 1:1 ratio. For some of the mutants, the model underestimated field growth rates under conditions that combine shade and warm temperature (Figure S2). Conversely, the WT showed no obvious bias, and, therefore, the model can predict G of WT seedlings under field conditions with reasonable accuracy. The analysis of adjusted R² values indicates that all the terms contributed to account for the field data (Table S6). The weakest contribution was that of the term reflecting the interaction between low temperature and high phyB activity, likely because the latter combination was not as frequent in our field conditions as under controlled conditions.

3.6 Impact of weather conditions on shade avoidance

To investigate the impact of weather conditions on shade avoidance, we first downloaded light and temperature data from 100 locations distributed over the Earth at the time of the year when the photoperiod is 10 hr (Figure 6a, Table S5). This photoperiod occurs only within certain latitudes, and very high latitudes are not included due to their extremely low temperatures (which are not suitable for Arabidopsis growth). Light and temperature data were used to estimate G under sunlight conditions and corrected by the impact of the canopy to estimate G under mild and strong shade conditions. Figure 6b shows estimated G values under sunlight and shade conditions, plotted against the mean irradiance, mean temperature, or thermal amplitude corresponding to each location (G values are the same in the three plots). The environmental values correspond to sunlight conditions (even though shade conditions were used to estimate G) because the aim is to explore the impact of weather. Of the three environmental variables, only mean temperature affected the magnitude of the shade avoidance response (the interaction between mean temperature and light/shade condition is significant at p < 10⁻²⁶). The magnitude of the difference between G under sunlight compared with shade conditions increased under warmer temperatures.

Figure 6c shows a similar plot, but, in this case, G corresponds to the values measured in our field experiments (Table S2, rather than model estimates) and their corresponding environmental data. Compared with the values observed in nature, the range is shifted towards higher temperatures because in our experiments, the seedlings were exposed to natural radiation for 10 hr even in the warm season, when natural photoperiod is above 10 hr. Despite this, the results show the same pattern providing independent support to the conclusion based on the application of the model: Of the three environmental variables, only mean temperature affected the magnitude of the shade avoidance response (the interaction between mean temperature and light/shade condition is significant at p < 10⁻¹⁶).

Finally, in Figure 6d, G was again estimated by using the model and the variables corresponding to the 100 world locations, but, in this case, only the variable plotted in abscissas was varied, whereas for the other two, the average for the 100 locations was used as input. This procedure eliminates the natural correlations among temperature, irradiance, and thermal amplitude. The results of this exercise show that irradiance and thermal amplitude theoretically do have impact on shade avoidance, which becomes more intense under better irradiated conditions with weaker thermal amplitude. The comparison of Figure 6a,c indicates that the effect of irradiance and thermal amplitude are not obvious in natural settings because they become diluted and/or counteracted by variations in the other variables.

3.7 Predicted impact of climate change on the magnitude of shade avoidance responses

By using current weather data in combination with a model that estimates climate trends, we obtained predicted whether conditions for 2080 at a global scale. These data were used as input for the growth rate model (Table S3h) to estimate current and future growth rate under sunlight and two different degrees of shade. In the plot of future versus current growth rate, most of the points are above the 1:1 line (Figure 7). This pattern is consistent with the fact that the elevation of average temperature with global warming increases hypocotyl growth. The extent of change (future minus current estimated growth) is significantly higher for the seedlings grown under deep shade, compared with mild shade or sunlight (Figure 7, inset). This indicates that shade avoidance responses are predicted to increase in a scenario of global warming. It must be emphasised, however, that this exercise is valid only to investigate the impact of future combinations of temperature, irradiance, and thermal amplitude, whereas a prediction of the actual growth rates would require the incorporation of other variables to the model (e.g., water availability).