Drought responses in Coffea arabica as affected by genotype and phenophase. II – photosynthesis at leaf and plant scales

Summary

The aim of this work was to compare gas exchanges from leaf to whole plant scales, in two Ethiopian accessions (‘E083’ and ‘E027’), and two bred cultivars (Iapar 59 and Catuaí 99) of Arabica coffee (Coffea arabica L.) cultivated under irrigated and rainfed conditions. Variations in gas exchanges were evaluated over four phenophases (leaf expansion – BE1 and BE2, and berry harvesting – BH1 and BH2), covering the first two production years in the coffee life cycle. We addressed the following questions: Are gas exchanges modified by water availability at leaf and/or plant scales? Do bred cultivars and wild accessions differ in their physiological responses to water availability and phenophases? Photosynthesis (A), stomatal conductance (g_s), and transpiration (E) were measured on the recently fully expanded leaves at the upper canopy stratum. The functional-structural plant modelling (FSPM) was used to integrate A at whole plant photosynthesis (A”_p), based on 3D virtual trees constructed under VPlants modelling platform. Despite high A values of ‘E083’ overall phenophases, a strong decline in A”_p under rainfed condition was observed due to lower plant leaf area as compared to irrigated condition. Catuaí 99 and ‘E083’ were more sensitive to drought than Iapar 59 and ‘E027’, considering photosynthesis at leaf and plant scales. At the last BH2 phenophase, A, g_s, E, and carboxylation efficiency were similar between irrigated and rainfed conditions for all genotypes, suggesting some acclimation of leaf gas exchange to the environment. However, A”_p benefited by water management in all phenophases as plant leaf area increased. These findings revealed the need to develop methodologies for structural and functional analyses at plant scale, an important step towards the realistic responses of plants and orchards to the surrounding environment.

Introduction

Climate change has increased the frequency, duration, and intensity of drought events, especially in arid and semi-arid regions (Hosseinizadeh et al., 2015), while frequency and intensity of rainfalls and storms have also increased (Chiang et al., 2021), especially in tropics (Makarieva et al., 2023). In such complex scenario, plant responses to varying water availability are not easy to predict. Process-based models have been used for this purpose, but they require experimental data to estimate the resistance of plants and ecosystems under low water availability in large scales (Zhou et al., 2019). The scaling of functional responses at smaller scales, such as organs up to plants and even orchards, can be obtained with functional-structural plant modelling, also known as FSPM (Vos et al., 2010). Here we are interested in Coffea arabica L. species, in which some geometrical and FSPM were done considering multiscale analyses, from metamer, axis, canopy stratum to plant scales (Rakocevic et al., 2021a), berry distribution over plagiotropic ranks (Rakocevic et al., 2018a), and chemical bean quality over the canopy profile (Rakocevic et al., 2023a). FSPM helped to simulate the increased C-assimilation at whole plant scale and architectural modifications indicating the rejuvenation of the plant structure in C. arabica under elevated air CO₂ concentration (Rakocevic et al., 2018b), and to compare C-assimilation at whole plant scale between Coffea. canephora genotypes (Rakocevic et al., 2023b).

Single-leaf net CO₂ assimilation rate (A) of C. arabica saturates when the irradiance is between 600 and 700 μmol photons m^–2 s^–1 (Rodrigues et al., 2016; Rakocevic et al., 2021b), although it is closely dependent on N–nutrition levels (Ramalho et al., 2000). At daily scale, direct irradiance on leaves and canopies fluctuates due to changes in solar angle and cloud cover, as well as due to wind-induced leaf movements. Leaves acclimated to high irradiance respond more rapidly to sudden changes in light environment and are usually more tolerant to co-occurring heat and water stresses than shaded leaves (Niinemets, 2007). When irradiance increases, A of shadeadapted leaves does not immediately increase to a new steady-state level, but increases progressively due to stomatal opening, rather than to ribulose–1,5–bisphosphate carboxylase/oxygenase (RuBisCO) activation (Zhang et al., 2022). Coffee plants displayed a relatively low A, mainly due to limitations to CO₂ diffusion from the atmosphere to the chloroplasts, which is caused by low stomatal (g _s) and mesophyll (g _m) conductances (DaMatta et al., 2019). In fact, g _s is the major factor limiting coffee photosynthesis and such influence is amplified under low water availability (Martins et al., 2014; Peloso et al., 2017). Water use efficiency (WUE) has become a major target for improving the resistance and productivity of C3 and C4 crops in a changing environment (Leakey et al., 2019). Under water deficit, increased WUE is based on the balance between A and transpiration (E), with the latter being more sensitive to low soil water availability than the former. Improvements in WUE can be achieved, namely, by targeting key aspects of plant architecture and morphology. For instance, reduced height, more productive tillers, and an expanded root system increased WUE in rice plants (Lo et al., 2017).

As a perennial species, coffee trees are usually exposed to water deficit during the cold and dry winter occurring in subtropical climates, but cold itself can promote important leaf dehydration even under well-watered conditions (Ramalho et al., 2018). The complete reproductive cycle of coffee trees takes two years, being composed of vegetative (first year) and reproductive (second year) phenophases. In total, six phenophases are defined for this species (Camargo and Camargo, 2001). After the first harvest, there are two parallel phenophases (flower bud induction and vegetative growth) in branches of a given plant, or in neighbouring metamers of the same plagiotropic axis (Rakocevic et al., 2021a). Under rainfed conditions, single-leaf A is more sensitive to high light during berry maturation than during berry and leaf expansion (Rakocevic et al., 2021b). Near harvest, g _s increases and crop water stress index diminishes under irrigated conditions (Abreu et al., 2023). On the other hand, decreases in A are more common at the end of berry maturation than during berry and leaf expansion phenophases under rainfed conditions (Rakocevic et al., 2021b). Among various coffee species, growth of C. arabica is highly sensitive to water deficit in growth parameter responses (Vu et al., 2018) and such sensitivity is clearly genotype-dependent. Some genotypes display important intrinsic resistance to drought (Chekol et al., 2023), namely at physiological and biochemical levels (Ramalho et al. 2018; Semedo et al., 2021; Chekol et al., 2024).

Coffee breeding is primarily focused on increased yield, but also on selection of accessions able to cope with drought and supra-optimal temperatures (Cheserek and Gichimu, 2012). Coffee is unusual among major crop trees, with a relatively short period of domestication and reduced genetic distance from wild to cultivated genotypes, except in the cases of interspecies hybrids (Leroy et al., 2006). Over the domestication, whether the ability of wild coffee genotypes to respond to low water availability has been reduced in commercial genotypes or not is still an open question. In this study, we aimed to evaluate gas exchange in wild and bred C. arabica trees grown under irrigated (IR) or rainfed (NI) conditions. We also examined variations in photosynthesis over four phenophases (leaf and berry expansion in year 1 [BE1] and year 2 [BE2], and berry harvesting in year 1 [BH1] and year 2 [BH2]), covering the first two production years of field-grown coffee trees. Using FSPM, leaf gas exchange was integrated at the whole plant scale, and two Ethiopian wild accessions (‘E083’ and ‘E027’) and two bred cultivars (Iapar 59 and Catuaí 99) were compared along the two first productive years. Finally, we quantified A at the leaf and whole plant scales, comparing wild and bred genotypes in both water regimes.

Material and Methods

Plant material and experimental design

Young coffee plants of six months raised from seeds were planted in the experimental field in Londrina (23°18′37″S, 51°09′46″W, 620 m a.s.l.), Paraná, South Brazil. Coffee rows were oriented East–West, with 2.5 m distance between them and 0.5 m between plants in the row (density of 8,000 plants ha^–1). Two largely used coffee cultivars, Iapar 59 and Catuaí 99, and about a hundred planted Ethiopian wild accessions were randomly distributed in the experimental field. Among the accessions, two were chosen for our experiment, ‘E083’ and ‘E027’, because of their outstanding architectural characteristics. The ‘E083’ had elongated leaves, a lot of flowering, and quick vegetative space occupation, while the ‘E027’ had large leaf blades, branched structure, and reduced flower number. The period of experiment comprised the two first productive years, i.e., two (year 1) and three (year 2) years after planting, when plants were 2.5- and 3.5-year-old, respectively. Four dry periods were registered during the experimental period, three in autumn/winter seasons as expected, and one unusual in the hot summer season of year 2 (details in Rakocevic et al., 2024).

The soil was a dusky-red dystrophic latosol (Nunes et al., 2021) and the climate of the region is subtropical, Köppen–Geiger climate type Cfa, with average annual precipitation of about 1,585 mm, ranging from 55 mm in the driest month (August) to 245 mm in wettest one (January). The limiting factors for C. arabica survival in such Cfa climate are dry periods and minimum air temperatures causing strong frost in winter (Meireles et al., 2009). Photosynthetic photon flux density (PPFD, μmol photons m^–2 s^–1) was measured with LI–190R light sensor (LICOR, Lincoln NE, USA) at 2 m above the soil. A datalogger model 21X (Campbell Scientific, Logan UT, USA) was used for data recording. PPFD data were collected every 60 s and integrated as mean values for each 15 min interval. The mean values for two-month period in each phenophase (n = 40–60) were calculated (Supplementary Material Fig. S1) and used in plant photosynthesis estimations (see section Plant architectural measurements and photosynthesis estimations).

Water regimes

Coffee plants were cultivated either under rainfed (NI) or irrigation (IR). Drip irrigation was triggered based on soil water balance method and aiming to restore the full soil water field capacity (details in Rakocevic et al., 2024). The irrigation flow rate of each dripper was 3.5 L h^–1, and drippers were distributed at 0.5 m of linear distance, with hoses near coffee trunks. NPK (20:5:15) fertilisation was done at 1,000 kg ha⁻¹ year⁻¹, split into four equal applications at flowering: two at berry and leaf expansion and one at the beginning of berry maturation, while weeds were controlled mechanically.

The experimental design was completely randomised, with one plant as an experimental unit, and the number of replications was four. Low number of repetitions is common when coding plants (Rakocevic et al., 2011; 2023b). Plant coding at each of four phenophases took about one month and required four people to complete (Rakocevic et al., 2024). Only in ‘year 2’ was the ‘E083’ accession represented by three plants in rainfed conditions, because one died after the dry winter of ‘year 1’. All other plants, including those two cultivars and ‘E027’ accession, were killed by the strong frost that occurred after the harvest of the ‘year 2’, imposing the end of the experiment. Only four ‘E083’ plants under irrigated conditions survived the frost (Rakocevic and Matsunaga, 2018).

Single-leaf gas exchange measurements

Single-leaf net CO₂ assimilation rate (A, µmol m^–2 s^–1), stomatal conductance (g _s, mol m^–2 s^–1), transpiration (E, mmol m^–2 s^–1), and intercellular CO₂ concentration, (C _i, μL L^–1 or Pa) were measured using a portable gas exchange analyzer (LC–Pro, ADC, Hoddesdon, UK). Fully recently expanded leaves belonging to the third to fifth pair of leaves from the terminal apex of the branch were always considered. One leaf was measured on each plant, positioned at the 2^nd-order plagiotropic branches oriented to the north. Measurements were taken ‘in situ’, between 9h00 and 11h30, in each of the four phenophases: (1) expansion of leaf area and berries, in January–February of the ‘year 1’ (BE1); (2) berry harvesting of the first production, in June–July of the ‘year 1’ (BH1); (3) expansion of leaf area and berries of second production, in January–February of the ‘year 2’ (BE2); and (4) end of second production berry harvest, in June–July of the ‘year 2’ (BH2). During the measurements, the mean air temperature and PPFD were 33 ± 2°C and 1060 ± 125 µmol photons m^–2 s^–1 at BE1/BE2, while 28.2 ± 2° C and 710 ± 120 µmol photons m^–2 s^–1 at BH1/BH2. The intrinsic water use efficiency (iWUE, µmol mol^–1) was calculated as A/g _s, and the instantaneous carboxylation efficiency (CE, µmol m^–2 s^–1 Pa^–1) as A/C _i.

Plant architectural measurements and photosynthesis estimations

Topological and geometric traits were coded as described in Rakocevic et al. (2024). Briefly, 3D reconstructions and visualisations were performed as proposed by Matsunaga et al. (2016), using VPlants platform and modules AmostraCafe3D, VirtualCafe3D, and Cafe3D. These reconstructed plants were visualised in 3D using PlantGLViewer (Pradal et al., 2009) and exported to VegeSTAR, a software that allows the calculation of leaf area, PPFD interception, and A, based on light distribution among the elements of plant canopy or plants (Adam et al., 2006).

Photosynthesis calculation under VegeSTAR requires information about azimuth and height of the Sun, global radiation, diffuse radiation, air temperature, and CO₂. Physical and meteorological parameters were measured during the four phenophases (from BE1 to BH2), e.g., irradiance (Fig. S1) and temperature (Rakocevic et al., 2024) or calculated in VegeSTAR (Sun azimuth and height). For the whole plant photosynthesis modelling, the maximum carboxylation rate of RuBisCO (V _cmax) and the maximum rate of electron transport (J _max) varied in the range of 25–70 µmol m^–2 s^–1 and 63–95 µmol m^–2 s^–1, respectively, depending on light conditions (Araujo et al., 2008) and genotype (Rakocevic et al., 2023b). Dark respiration values (R _d) were between 0.2 and 1.9 µmol m^–2 s^–1, based on Rakocevic et al. (2018b and 2021b).

For leaves reconstructed with sixteen triangles (Rakocevic et al., 2024), the single-leaf photosynthesis (A’, µmol m^–2 s^–1) was simulated. These simulated values were compared to the measured ones, considering leaf position in each coffee tree of all genotypes growing under two water regimes and at four phenophases. The validation was performed by the following procedure: (i) R _d, V _cmax, and J _max were included in the 3D reconstruction files (‘vgx’) of each plant and in each phenophase, then water availability and light conditions were varied; (ii) the sequences of physical and meteorological parameters, necessary for VegeSTAR run, were built for each of the four phenophases, considering 15 min intervals between 9h00 and 11h30 (morning period) when A measurements were taken. Then, outputs of VegeSTAR simulations (estimated A’ values) were compared with the measured A values; (iii) instantaneous mean plant photosynthesis (A”_p) was estimated as the integration of single-leaf photosynthesis, considering the plant leaf area in m² (Supplementary Material Fig. S2), using the sequences of physical and meteorological parameters at 15-min intervals for the photoperiod of 11 hours for BH1 and BH2, and 13 hours for BE1 and BE2. Mean A”_p was then integrated at daily scale to calculate net CO₂ assimilated per leaf surface and day in each phenophase (A”_m).

Statistical analyses

The R software (R Core Team, 2022) libraries and functions were used for statistical analyses. ANOVA was applied to analyse leaf gas exchange and plant photosynthesis. Two-way ANOVA considered a mixed linear model (lme function from ‘nlme’ package) and maximum likelihood to test the significance of differences in leaf gas exchange due to the two fixed factors: (i) water regime (IR and NI); and (ii) genotypes (‘E083’, ‘E027’, Iapar 59, and Catuaí 99). If no significant interaction was found, the model reduction was applied (and fitted by using lme function, considering both factors as fixed). When interaction between factors was significant, only P–value of interaction was shown in the figures. One-way ANOVA was additionally applied to compare phenophases for each genotype. For comparing average values estimated by ANOVA, we used the Tukey Honestly Significant Difference and ‘lsmeans’ and ‘multcompView’ packages. The accuracy of modelling was estimated through linear regressions between measured (A) and estimated (A’) photosynthesis of each genotype, considering RMSE, R², and bias (package ‘qpcR’). The Pearson method was applied to correlate traits related to leaf gas exchange and architecture of each plant, using the ‘corrplot’ package. We used the following architectural traits: leaf area per plant; branch area (length x diameter) per plant; leaf number per plant; individual leaf area (ILA); leaf angle in relation to soil (0–90°); 2^nd order branch angle in relation to soil (0–90°); number of metamers per orthotropic axis; height of orthotropic axis; and number of 2^nd order plagiotropic branches, as reported in Rakocevic et al. (2024). A significance level of 0.05 was used for all analyses.

Results

Single-leaf gas exchange as affected by water regime, genotypes, and phenophases

Globally, single-leaf photosynthesis (A) did not differ among phenophases (P = 0.258). Also, all genotypes exhibited decreases in A under rainfed (NI) as compared to irrigated (IR) conditions in the first three phenophases (Figure 1A–C). At BH2, only Catuaí 99 had decreases in A due to low water availability (Figure 1D). At BE1 and BE2, genotypes did not differ in relation to A, within each water regime (Figure 1A, C). On the other hand, ‘E083’ plants showed higher A than ‘E027’ or Catuaí 99 at BH1 (Figure 1B), and higher than Iapar 59 at BH2 (Figure 1D), regardless of the water regime. Under NI conditions, the wild accessions showed higher A than bred cultivars at BH2 (Figure 1D).

Figure 1. Single-leaf net CO₂ assimilation rate (A) of four Coffea arabica genotypes grown under irrigated (IR) or rainfed (NI) conditions, measured during leaf/berry expansion of ‘year 1’ (BE1, in A) and ‘year 2’ (BE2, in C), and harvest of ‘year 1’ (BH1, in B) and ‘year 2’ (BH2, in D). Estimated mean ± SE and P–values (bold when significant) are shown (n = 3–4). Different lower case letters indicate significant differences among four genotypes within each water regime, while different upper case letters indicate significant differences between the water regimes for each genotype, always in a given phenophase. Same red upper case letters at upper right corners indicate no global differences among phenophases (P _phase = 0.2581).

Stomatal conductance (g _s) at BE was globally higher than at BH phenophase (Figure 2). Notably, IR treatment allowed greater g _s in all genotypes at BE1 and BE2 (Figure 2A, C), as well as in Catuaí 99 at BH1 (Figure 2B). Interestingly, ‘E027’ plants under NI conditions showed greater g _s than their irrigated counterparts at BH1, although the latter ones maintained greater A values. At BH1, no differences in g _s were observed among the four genotypes under IR conditions, while leaves of ‘E027’ showed higher g _s than ‘E083’ and Catuaí 99 under NI conditions (Figure 2B). It is also noteworthy that despite some differences among genotypes along the experiment, no significant differences in g _s were depicted at the last phenophase BH2, regardless of water regime or genotypes (Figure 2D).

Figure 2. Stomatal conductance (g _s) of four Coffea arabica genotypes grown under irrigated (IR) or rainfed (NI) conditions, measured during leaf/berry expansion of ‘year 1’ (BE1, in A) and ‘year 2’ (BE2, in C), and harvest of ‘year 1’ (BH1, in B) and ‘year 2’ (BH2, in D). Estimated mean ± SE and P–values (bold when significant) are shown (n = 3–4). Different lower case letters indicate significant differences among four genotypes within each water regime, while different upper case letters indicate significant differences between the water regimes for each genotype, always in a given phenophase. Red upper case letters at upper right corners indicate global differences among phenophases (P _phase < 0.0001).

Leaf transpiration (E) was most intensive at BE1, while the lowest values were noticed at BH1 and BH2 (Supplementary Material Fig. S3). As evidence that plants were under water deficit, the IR treatment increased E for all genotypes at BE1 and BE2 (Fig. S3A,C) as compared to NI condition, while no effect of water availability was noticed during the two harvest periods (Fig. S3B, D). At BH1, ‘E083’ and Iapar 59 showed higher E than ‘E027’, while Catuaí 99 presented intermediate values (Fig. S3B). The E was insensitive to water regime and did not vary among genotypes at BH2 (Fig. S3D), as found for g _s (Figure 2D).

The lowest iWUE occurred at BE1, closely associated with the usually greater g _s in this phenophase, whereas the highest iWUE occurred at BH1, followed by BE2 and BH2 phenophases (Figure 3). Interestingly, iWUE did not vary due to water regime or among phenophases, but strong differences were found among genotypes. At BE1, the bred cultivars showed higher iWUE than the wild accessions (Figure 3A). However, the wild accession ‘E083’ was more efficient than Catuaí 99 at BH1 (Figure 3B), than both bred cultivars at BE2 (Figure 3C), and than Iapar 59 at BH2 (Figure 3D).

Figure 3. Intrinsic water use efficiency (iWUE) of four Coffea arabica genotypes grown under irrigated (IR) or rainfed (NI) conditions, measured during leaf/berry expansion of ‘year 1’ (BE1, in A) and ‘year 2’ (BE2, in C), and harvest of ‘year 1’ (BH1, in B) and ‘year 2’ (BH2, in D). Estimated mean ± SE and P–values (bold when significant) are shown (n = 3–4). Different lower case letters indicate significant differences among four genotypes within each water regime, while different upper case letters indicate significant differences between the water regimes for each genotype, always in a given phenophase. Red upper case letters at upper right corners indicate the differences among phenophases (P _phase = 0.0001).

The instantaneous carboxylation efficiency (CE) did not vary (P = 0.1695) over the four phenophases (Figure 4). On the other hand, the IR treatment impacted positively on CE of all genotypes (Figure 4A–C), except at BH2 (Figure 4D). Moreover, no significant differences were observed in CE among genotypes within each water regime at BE2 (Figure 4C). In contrast, significant effects of both genotype and water regime on CE were found at BE1 and BH1 (year 1), with ‘E083’ always showing the highest values in comparison to the other genotypes, regardless the water regime (Figure 4A, B). At BH2, the lowest CE values were measured in Iapar 59, regardless of the water regime (Figure 4D).

Figure 4. Carboxylation efficiency (CE) of four Coffea arabica genotypes grown under irrigated (IR) or rainfed (NI) conditions, measured during leaf/berry expansion of ‘year 1’ (BE1, in A) and ‘year 2’ (BE2, in C), and harvest of ‘year 1’ (BH1, in B) and ‘year 2’ (BH2, in D). Estimated mean ± SE and P–values (bold when significant) are shown (n = 3–4). Different lower case letters indicate significant differences among four genotypes within each water regime, while different upper case letters indicate significant differences between the water regimes for each genotype, always in a given phenophase. Same red upper case letters at upper right corners indicate no global differences among phenophases (P _phase = 0.1695).

Photosynthesis of reconstructed leaves and plants

The estimated single-leaf photosynthesis (A’) was slightly underestimated when compared to the measured values (A) of each genotype (Supplementary Material Fig. S4). On average, such underestimation varied between 6.8% (‘E083’) and 11.1% (Iapar 59), based on linear regressions. The bias varied between –0.39 (Catuaí 99) and –0.797 (‘E083’) µmol m^–2 s^–1 for A’ up to 10.6 µmol m^–2 s^–1, with RMSE ranging from 0.039 (‘E083’) to 0.503 (Iapar 59).

The instantaneous photosynthesis integrated to plant leaf area (A”_p) gradually increased over the time, usually more clearly after BH1 (Figure 5), and followed the increases in plant leaf area (Fig. S2). A”_p was always significantly impacted by water regime, regardless of genotype and phenophase (Figure 5). At BE1 and BH1 (year 1), the highest difference in A”_p between the IR and NI plants was noticed in ‘E083’, roughly five times, while this difference was about two times for the other three genotypes (Figure 5A,B). At BE1, the highest A”_p values were estimated in Iapar 59 under both water regimes and in ‘E083’ under IR condition (Figure 5A). At BH1, the highest A”_p was again found in Iapar 59 plants under IR condition, while ‘E083’ had the lowest A”_p among the four genotypes under NI condition (Figure 5B). Among the studied factors, only the water regime impacted on A”_p at BE2 (Figure 5C). At the last phenophase (BH2), the highest A”_p was estimated in Iapar 59 and the lowest in ‘E027’ when plants were irrigated, while ‘E083’ and Catuaí 99 showed lower A”_p than Iapar 59 and ‘E027’ under NI conditions (Figure 5D).

Figure 5. Instantaneous plant photosynthesis (A”_p) of four Coffea arabica genotypes grown under irrigated (IR) or rainfed (NI) conditions, measured during leaf/berry expansion of ‘year 1’ (BE1, in A) and ‘year 2’ (BE2, in C), and harvest of ‘year 1’ (BH1, in B) and ‘year 2’ (BH2, in D). Estimated mean ± SE and P–values (bold when significant) are shown (n = 3–4). Different lower case letters indicate significant differences among four genotypes within each water regime, while different upper case letters indicate significant differences between the water regimes for each genotype, always in a given phenophase. Red upper case letters at upper right corners indicate global differences among phenophases (P _phase < 0.0001).

The integrated diurnal net CO₂ assimilation per leaf surface area (A”_m) in C. arabica genotypes was higher in younger plants (BE1) and gradually decreased until BH2 (Figure 6). The IR treatment always had a positive effect on A”_m, regardless of the phenophases. From BE1 to BE2, ‘E083’ had greater A”_m values than the other genotypes, irrespective of water regime (Figure 6A–C). At BE1, the lowest A”_m values were found in Iapar 59 (Figure 6A). Within each water regime, the four genotypes showed similar A”_m values only at BH2 (Figure 6D), as noticed for the total leaf area per plant (Fig. S2).

Figure 6. Diurnal net CO₂ assimilation per leaf surface (A”_m) of four Coffea arabica genotypes grown under irrigated (IR) or rainfed (NI) conditions, measured during leaf/berry expansion of ‘year 1’ (BE1, in A) and ‘year 2’ (BE2, in C), and harvest of ‘year 1’ (BH1, in B) and ‘year 2’ (BH2, in D). Estimated mean ± SE and P–values (bold when significant) are shown (n = 3–4). Different lower case letters indicate significant differences among four genotypes within each water regime, while different upper case letters indicate significant differences between the water regimes for each genotype, always in a given phenophase. Red upper case letters at upper right corners indicate global differences among phenophases (P _phase < 0.0001).

Correlations among photosynthetic and architectural traits

The correlations among architectural and physiological traits were analysed in each water regime (Figure 7), giving special focus on A”_p and A”_m. Overall, A”_p was always negatively correlated with A”_m. A”_p was positively and A”_m negatively correlated with architectural traits at plant scale, such as total leaf and branch area, total leaf number per plant, individual leaf area, number of metamers at orthotropic axes, height of orthotropic axis, and number of plagiotropic branches of 2^nd order. Interestingly, the average leaf angle was not correlated with A”_p or A”_m, while branching angle (2^nd order plagiotropic elevation in relation to horizontal plane) was positively correlated with A”_p under IR condition (Figure 7A). Considering the underlying processes driving photosynthesis, A”_p was significantly correlated only with g _s, while A”_m was positively correlated with A, E, g _s , and CE, always in IR plants (Figure 7A). E was negatively correlated with leaf and branch area, leaf number and ILA, and with orthotropic axis traits, such as height and number of metamers of orthotropic axis, and number of 2^nd-order plagiotropic branches (Figure 7), i.e., single-leaf E decreases when increasing plant/canopy structure (Fig. S3). Similar correlations were also observed for g _s but only under IR condition (Figure 7A). As expected, iWUE was positively correlated with A and negatively with g _s, irrespective of the water regime. However, iWUE was positively correlated with orthotropic axis traits only for irrigated plants (Figure 7A). Higher number of significative correlations among the structural and functional traits in coffee plants was established under IR than under NI conditions (Figure 7).

Figure 7. Graphical presentation of Pearson’s correlation coefficients (values corresponding to ellipse size and colour intensities) and P–values (significant when < 0.05, not crossed ellipses, n = 3–4) for correlations among instantaneous plant photosynthesis (A”_p, AP), integrated diurnal photosynthesis per surface (A”_m, APMD), architectural traits (leaf area – LA, branch area – BA, leaf number per plant – LN, individual leaf area – ILA, leaf angle – LANG, 2^nd order branch angle – BANG, number of metamers per orthotropic axis – NMO, height of orthotropic axis – HO, number of 2^nd order plagiotropic branches –NP2) and instantaneous single-leaf gas exchanges (photosynthetic rate – A; transpiration rate – E; stomatal conductance – g _s; intrinsic water use efficiency – iWUE; carboxylation efficiency – CE) in Coffea arabica plants grown under irrigated (A) or rainfed (B) conditions.

Discussion

Photosynthetic responses to water availability: Upscaling from leaf to canopy

Coupling single-leaf photosynthesis (A) with plant architecture revealed that A”_p strongly increased under IR as compared to NI conditions (Figure 5). The A”_p was positively correlated with architectural traits (Figure 7), suggesting that plant structure was closely associated with the total plant CO₂ assimilation. Although this idea has been mentioned earlier as a coffee response to drought (DaMatta, 2004), testing such hypothesis was not an easy task from a methodological perspective. In fact, we believe this is the first paper to show how the structure (leaf area distribution including leaf size and position) and function (photosynthetic carbon assimilation) are interlinked in coffee plant responses to drought, using the FSPM approach.

With plant structure increasing from BE1 to BH2 (Fig. S2), the average plant photosynthesis per leaf surface (A”_m) decreased (Figure 6). This pattern is likely associated with reduced irradiance at leaf scale due to an increased self-shading in the inner part of plant canopy, and at the basal canopy strata (Rodrigues et al., 2016; Rakocevic et al., 2024). Over the four phenophases, the integration of instantaneous photosynthesis from leaf to plant scale (A”_p) revealed an increasing trend, which again was driven by increasing plant leaf structure. When plants exhibit a large canopy, as coffee plants grown in monoculture under high light (Rakocevic et al., 2024), there is a need for allocating more assimilates to root and sapwood formation, so that the increased water demand by the aboveground organs is met (Oliveira et al., 2023). In such a way, the investment in branch structure can be promoted in some genotypes, as was in ‘E083’, more than in others (Rakocevic et al., 2024). Previously, we found evidence of C-allocation to root system of young Arabic coffee plants under dry conditions (Rakocevic et al., 2022), which is a possible phenomenon in our experiment. In fact, NI plants presented lower leaf area than IR plants (Fig. S2), a consequence of reduced CO₂ uptake and/or changes in biomass partitioning in field-grown coffee trees.

In VegeSTAR, the A”_p and A”_m were estimated based on light distribution within the plant/canopy, i.e., with the increased leaf area over phenophases, the average A”_m decreases while average A”_p increases, thus leading to a negative correlation between these two traits (Figure 7). Despite high A values of ‘E083’ overall phenophases compared to other genotypes (Figure 1), there was a strong reduction in A”_p under NI as compared with IR conditions. IR plants had 4 to 5-fold higher CO₂ uptake at BE1 and BH1 (Figure 5), which was due to differences in leaf area between IR and NI plants (Fig. S2). At leaf scale, water deficit did not reduce A at BH2 in all genotypes (Figure 1D), which was not true at plant scale (Figure 5D). Here again the architecture (plant leaf area) impacted the plant scale photosynthesis. As leaf area increased over time, the overlapping of leaves reduced light use efficiency at plant scale by reducing light availability to deep plant strata (Slattery et al., 2017).

The integration of photosynthesis at the plant scale is dependent on light distribution over the plant canopy, and our data revealed that the more the complex canopy structure – which happens along the plant phenological phases – the more light competition occurs among leaves and a disconnection between leaf and plant scales would be expected when evaluating the coffee responses to environmental changes. Therefore, there is a need to develop and adopt methodologies that permit the structural and functional analyses at plant scale and then a glance at realistic responses of plants and orchards.

Physiological responses of bred and wild genotypes to water availability over phenophases

Coffee would need more water and photoassimilates during the leaf and berry expansion than during the fruit maturation, which is explained by changes in source-sink with coffee ageing and fruit development (Unigarro Muñoz et al., 2021). As a consequence, one would expect higher g _s and E values and higher coffee sensitivity to drought at BE1/BE2, when both leaves and berries are sinks, than at BH1/BH2 (Figures 2 and S3). Focusing on genotype stability in terms of CO₂ assimilation along the experimental period, the reduced sensitivity in A over time could suggest the wild accessions and Iapar 59 as drought-tolerant genotypes, while Catuaí 99 would be the most sensitive genotype after two years under field conditions (Figure 1D). A possible explanation for such variation in drought sensitivity relies on the origin of studied genotypes. Iapar 59 originated from the cross between the cultivar Villa Sarchi CIFC 971/10 and the hybrid of Timor CIFC 832/2, representing C. canephora introgression obtained by spontaneous specific crossbreeding with C. arabica (Anthony et al., 2002), while Catuaí 99 is a high-productive C. arabica cultivar with high cup quality (Pérez–Molina et al., 2021) and compacted architecture (Guerreiro–Filho et al., 2018). Here, we would like to highlight that bred cultivars are basically designed for high berry production, the main sink for assimilates.

Although there are differences in their responses to drought, bred cultivars have many characteristics of wild genotypes adapted to the shade environment of the Ethiopian forests, including partial closure of the stomata when transpiration is increased due to large leaf-to-air water vapour deficit (>1.6 kPa), an adaptive mechanism to minimise water loss under high light (Carr, 2001). This physiological strategy helps to explain differences in g _s, E, and iWUE among phenophases (Figures 2 and S3). Interestingly, variations in iWUE were genotype-dependent but were not related to water regimes. Regardless of the water regime, the wild ‘E083’ had the highest iWUE during the year 2, i.e., BE2 and BH2 (Figure 3C, D). Overall, such response was induced by higher A and/or reduced g _s in wild ‘E083’ when compared to bred cultivars. Generally, high g _s allows greater A and E, which is beneficial when environmental conditions are not limiting (Diao et al., 2024). As an example, domesticated tomato plants show higher g _s and E than wild plants, and there is no difference in A between wild and domesticated plants under drought (Lupo and Moshelion, 2024). The combination of low g _s and a similar A would lead wild coffee accessions in the ‘year 2’ to have higher iWUE than bred ones (Figure 3C, D). The absence of iWUE differences between water availability conditions was common to the four studied coffee genotypes and this could be envisaged as an acclimation to the environmental conditions in such a long-term experiment.

Coffee photosynthetic responses over phenophases

Measurements of single-leaf gas exchange taken at the last phenophase (BH2) seem to point to an acclimation of coffee leaves to varying water availability after four dry periods during the experimental biennial period (climatic details in Rakocevic et al., 2024). If such acclimation is true – as evidenced by an absence of photosynthetic response to irrigation – acclimatory responses were more effective in wild accessions and Iapar 59 than in Catuaí 99 (Figure 1D). As an alternative explanation, one would argue that irrigation was not effective in increasing photosynthesis in wild accessions and Iapar 59, which we are not able to address because the lack of data on plant water status. Even so, it is clear that our treatment for increasing water availability (irrigation) modulated coffee responses, as noticed for Catuaí 99, a drought-sensitive genotype. Such sensitivity to low water availability found in Catuaí 99 could be a consequence of low acclimation potential, a topic for further studies.

Acclimation to repeated episodes of drought depends on an orchestrated reprogramming of plant metabolism, involving key processes such as photosynthesis, respiration, photorespiration, and the antioxidant system (Menezes–Silva et al., 2017). Some long-term experiments report improved leaf gas exchange for woody plants under drought (Zhou et al., 2019). RuBisCO content has been found to increase in sunflower leaves under prolonged drought, conferring drought tolerance to such genotypes (Panković et al., 1999). Such kind of photosynthetic acclimation would justify similar CE between IR and NI conditions (Figure 4D) and suggest some biochemical acclimation at leaf scale in all studied coffee genotypes.

The integration of physiology and structure did not clearly indicate any acclimation of photosynthesis at plant scale, as revealed when differences between IR and NI plants were reduced over time. Both A″_p and A″_m were impacted by water deficit over the phenophases (Figures 5 and 6). While A″_p increased from BE1 to BH2 due to increased total leaf area of coffee plants, A″_m reduced because of greater self-shading promoted by increasing leaf area (Fig. S2). One interesting finding was that A″_p under NI conditions reached the highest values at BH2, with such response being more evident in ‘E027’ and Iapar 59 (Figure 5). Among the four studied genotypes, the highest bean yield was recorded in ‘E083’ and Iapar 59 under irrigation in both berry harvest phenophases, while ‘E027’ showed the lowest bean yield regardless of the water regime (Rakocevic et al., 2023a). On the other hand, bean yield of bred Iapar 59 and Catuaí 99 was less affected under rainfed conditions, while the bean yield of wild accession was strongly affected by the lack of irrigation, especially at the second berry harvest phenophase (Rakocevic et al., 2023a). The physiological basis of differential sensitivity to changes in water availability among coffee genotypes should be revealed in further studies addressing not only photosynthesis but also plant water status and water transport in plants, giving special attention to Iapar 59.

Conclusions

From the beginning of the experiment, the greatest differences in estimated plant photosynthesis between irrigated and rainfed conditions were found in ‘E083’ accession, while this accession showed only slight changes in leaf photosynthesis between the two water regimes along the four studied phenophases. This indicates that the differences in plant photosynthesis of ‘E083’ were due to plant leaf area, which varied between irrigated and rainfed conditions. The wild ‘E027’ had similar trend in leaf photosynthesis as ‘E083’ accession under the two water regimes over time, but more stable leaf area. On the other hand, the bred cultivar Catuaí 99 maintained its photosynthetic sensitivity to drought during the two-year study. Considering the maintenance of CO₂ assimilation in a changing environment throughout the four phenophases, Iapar 59 stands out as the most stable genotype and its ability to deal with varying water availability should be studied. Overall, bred cultivars had variable ability when dealing with low water availability, suggesting that drought resistance/sensitivity was likely affected by breeding. Modelled photosynthesis at the plant scale using FSPM was increased due to irrigation in all phenophases. Our data suggest acclimation of photosynthesis at leaf scale, which was not found at plant scale due to a strong interplay between function and structure in coffee trees. At plant scale, irrigation always benefited coffee photosynthesis, with whole plant photosynthesis increasing over the experimental period due to leaf area development and integrated diurnal photosynthesis per leaf area decreasing due to self-shading within coffee canopy.