Introduction

Among the species of the genus Theobroma, cacao (Theobroma cacao L.) is the most commercially important. The species is native to the Amazon basin and has great economic, social and environmental importance1,2. After fermented, dried, roasted and grounded, its beans are the main raw material of interest for the food and pharmaceutical industries3,4,5 and especially for the production of chocolate6,7. South America accounts for 12.5% of the world’s cocoa bean production, and its producing countries earn US$ 2.4 billion/year from its exports. Brazil is the second-largest producer in South America and the seventh-largest producer in the world (269,731 t)8. However, there has been a reduction in yield, mainly attributed to the occurrence of witch’s broom disease9 and to climate change, particularly due to abiotic and biotic stresses10,11. According to the latest report of the Intergovernmental Panel on Climate Change (IPCC), the increase in extreme weather events related to climate change may pose a risk to food security in Brazil due to frequent and extreme droughts12. These effects negatively impact cocoa yield13,14. According to the World Cocoa Foundation15, climate change could cause severe damage to cocoa cultivation in the next decades. This fact is mainly associated with water stress and prolonged periods of drought in the areas of cultivation16,17.

Water deficit in cacao was pointed out by Alvim18 as the most serious limitation that it can face compared to any other risk, including pests and diseases. Alvim18 assured in the 1970s that cocoa yield is considerably more controlled by rainfall than by any other environmental condition. Water scarcity and climate change predicted for the coming decades reinforce the importance of searching for and selecting drought-resistant genotypes with high production capacity. Knowledge in this area is still incipient, given the gravity of the situation: (i) water stress is the main limiting factor for cacao yield19; (ii) future scenarios indicate that hotter and drier climates will reduce cacao yield20; (iii) there will be a need to expand cultivated areas (which is hardly possible) or select genotypes tolerant to unfavorable climatic conditions21. It should be emphasized that, with the growth of cities and increased water demand for human consumption, environmental issues such as licensing and granting of water for cacao cultivation in Brazil should guide discussions on the future of irrigated farming22.

From the technological point of view, current research results indicate that cacao can be grown with irrigation in non-traditional areas, such as semi-arid regions23,24. Even in traditional growing regions, where water deficit did not exist, complementary irrigation for water supply is becoming necessary, as dry spells are getting more frequent. The southern region of Bahia State, for instance, recorded an extreme dry spell with five months (September/2015 to January/2016) of drought. This dry spell, associated with high temperatures, caused a sharp decrease in cocoa production that year and the following year. Deaths of 13.9% of cacao trees in the region were also recorded, caused by drought or fires22.

Cacao originates from the Amazon rainforest and, as such, tolerates a dry season of up to three months25. Cacao’s root system concentrates 83% and 86% of thin and thick roots up to a depth of 40 cm, facilitating the cycling of nutrients, but making the cultivation sensitive to long dry spells16. Water stress reduces plant development, interfering in several biochemical and physiological processes26,27. When cacao is under stress, there is a reduction in leaf area, leading to lower net assimilation rate, stomatal conductance, yield28 and transpiration29, affecting gas exchange and photosynthesis30. Nutrient uptake from the soil and nutrient transportation within the plant are also impaired31,32.

Fortunately, the severity of water stress is genotype-dependent, making genetic improvement a viable and sustainable alternative to mitigate its effects by selecting resistant genotypes33. Nutritional and physiological traits have been used to select drought-tolerant genotypes. Different selection strategies have been used to evaluate and identify superior genotypes, such as tandem selection, independent levels and selection indices, in addition to those based on principal components34,35,36,37. Yan and Rajcan37 proposed the Genotype-by-Trait (GT) biplot approach to evaluate data of multiple traits, aiming to rank and select the best ones based on joint performance. This approach has already been used in crops such as soybean37, common bean38, and sunflower39. Its use in cocoa has not yet been reported.

The present study aimed to promote the simultaneous selection of genotypes for nutritional and physiological traits under water-stressed conditions using a GT biplot approach. We expect to bring positive impacts on cacao breeding and agriculture by demonstrating how a robust approach such as GT biplot can aid in the selection of drought-tolerant genotypes. The success of this approach will not only contribute to the sustainability of cocoa production but also provide farmers with valuable tools to address specific challenges associated with cacao cultivation.

Methods

Genotypes, experiment and design

Eighteen genotypes were evaluated at the “Filogônio Peixoto” Experimental Station (ESFIP) (Table S1), in Linhares (latitude 19° 24′ 52″ S, longitude 40°03′ 54″ W and altitude of 19 m above sea level), Espírito Santo, Brazil. The seeds of these genotypes were pre-germinated, sown in 288 cm\(^3\) tubes and irrigated daily by an automated system. After six months, the seedlings were transplanted into polyethylene bags with capacity of 18 L and taken to the greenhouse. The substrate was prepared in a 3:2:1 ratio, resulting from the mixture of soil:sand:cattle manure, and fertilized with 5 kg of single superphosphate and 1 kg of potassium chloride per m\(^3\). The trial was set up in a greenhouse in a randomized complete block design, in a 2 \(\times\) 18 factorial arrangement (with and without irrigation, and 18 genotypes), with three replicates and five plants per plot, at 12 months of age. In the irrigated treatments, the individuals were kept on a substrate close to field capacity. In the treatments without irrigation, the water supply was gradually suspended. After an acclimatization period, genotypes with more uniform characteristics were divided into two groups, subjected to different treatments related to two distinct irrigation treatments. One group was kept as a control, maintaining constant humidity. Meanwhile, the second group was subjected to a water deficit for a sufficient period to reach and/or fall below − 1.5 MPa in \(\Psi _w\) values, indicating a level of “severe water stress” for cacao plants40.

It is worth pointing out that the 18 genotypes evaluated were previously selected from the yield monitoring of 39 genotypes under field conditions. Of these 39 genotypes, 11 were selected for their good agronomic traits and higher fruit production per plant, or higher bean yield, quantified during the period of intense drought that occurred between 2014 and 2016 (three years) at the ESFIP. The remaining 7 genotypes were selected at five cacao farms in the municipality of Linhares, adopting the same yield monitoring also carried out during the same three years.

Evaluation of gas exchange

Gas exchange measurements were made 28 days after suspension of irrigation, on the first fully developed leaf from the apex, between 8:00 and 10:00 a.m., using a portable infrared gas analyzer (IRGA), model LI 6400 (LI-COR, USA), under irradiance of 800 μmol of photons s\(^{-1}\). The CO2 concentration was 400 μmol m−2 s−1 (both conditions), and the chamber temperature was 25 \(^\circ\)C. The following physiological variables were determined, in irrigated and non-irrigated environments: net CO2 assimilation rate (A, μmol CO2 m\(^{-2}\) s\(^{-1}\)), transpiration (E, mmol H2O \(\text{m}^{-2} \text{s}^{-1}\)), stomatal conductance (gs, mol H2O \(\text{m}^{-2} \text{s}^{-1}\)), intercellular and atmospheric CO2 concentration (Ci/Ca, μmol CO2 mol air\(^{-1}\)), intrinsic water use efficiency (A/gs), instantaneous water use efficiency (A/E) and instantaneous carboxylation efficiency (A/Ci).

Evaluation of macro and micronutrient contents

To determine the contents of macronutrients (N, P, K, Ca, Mg and S) and micronutrients (Fe, Zn, Cu, Mn and B), the same leaves evaluated for gas exchange were collected, also 28 days after suspension of irrigation. Chemical analyses for these determinations were performed according to Malavolta et al.41.

Statistical analysis

ANOVAs and Scott–Knott test

Preliminarily, the experimental data were analyzed for homogeneity of variances and normality. Subsequently, the ANOVAs were processed. Once the significance of each effect (Genotype, Irrigation treatment (IT) and Genotype \(\times\) IT interaction) for each variable was verified, the means were grouped using the Scott–Knott test42 which allowed multiple group comparisons. To avoid multicollinearity in the data set, we excluded variables that originated the derivatives (A, E, and gs) in the multivariate analysis.

The coefficient of determination \(R^2\) was determined by:

$$\begin{aligned} R^{2} = 1 - \frac{\text {SS}_\text {res}}{\text {SS }_\text {tot}} \end{aligned}$$
(1)

where \(SS_\text {res}\) is the sum of squares of residuals and \(SS_\text {tot}\) is the total sum of squares (proportional to the variance of the nutritional and physiological data).

Genotype-by-Trait biplot (GT biplot)

The variables were standardized according to the expression below:

$$\begin{aligned} P_{ij}=\frac{T_{ij}-\bar{T_j}\ }{S_j} \end{aligned}$$
(2)

where \(P_{ij}\) is the standardized value of the genotype i for the trait j; \(T_{ij}\) is the average value of the genotype i for the trait j; \(\bar{T_j}\)is the average value of the trait j in all genotypes; and S\(_j\) is the standard deviation of the trait.

The GT biplot approach was based on the first two principal components (PC) resulting from the singular value decomposition (SVD) of the matrix of standardized variables:

$$\begin{aligned} P_{ij}=\left( d\lambda _1^\alpha \xi _{i1}\right) \left( \frac{\lambda _1^{1-\alpha }\tau _{1j}}{d}\right) +\left( d\lambda _2^\alpha \xi _{i2}\right) \left( \frac{\lambda _2^{1-\alpha }\tau _{2j}}{d}\right) +\varepsilon _{ij} \end{aligned}$$
(3)

where \(\xi _{i1}\) and \(\xi _{i2}\) are the eigenvalues for PC1 and PC2, respectively, for the genotype i; \(\tau _{1j}\) and \(\tau _{2j}\) are the eigenvalues for PC1 and PC2, respectively, for the trait j, and \(\varepsilon _{ij}\) is the adjusted residual of PC1 and PC2 for the genotype i in the trait j; \(\lambda _1\) and \(\lambda _2\) are the eigenvalues for PC1 and PC2, respectively, and \(\alpha\) is the partition factor of the singular value. If \(\alpha\) = 1, the biplot is said to be genotype-focused and is suitable for comparing genotypes. When \(\alpha = 0\), the biplot is said to be trait-focused and is suitable for viewing correlations between traits. The GT biplot was constructed by plotting \((d\lambda _1^{\ \alpha }\xi _{i1})\) against \((d\lambda _2^{\ \alpha }\xi _{i2})\), for genotypes, and plotting \({[(\lambda }_1^{1-\alpha }\tau _{1j})/d]\) against \({[(\lambda }_2^{1-\alpha }\tau _{2j}/d)]\), for trait in the same plot. The analysis described was performed using the traits’ mean values between irrigated and non-irrigated treatments. The heatmap for Pearson’s correlation was created using the ComplexHeatmap package43. The GT biplot approach was processed using the metan package44, both in R software45.

Ethical statement

The plant species used here is a cultivated plant (T. cocoa L.), and the genotypes employed were provided by the “Filogônio Peixoto” Experimental Station (ESFIP), located in the state of Espírito Santo, Brazil. We confirm that we have complied with all the necessary regulations for this type of research.

Results

Effects of genotype, irrigation, and their interaction on nutritional and physiological traits of cocoa

The effect of genotypes (G) was significant (p < 0.05) for the nutritional variables N, P, Mg, S, Zn, Cu and Mn and physiological variables A, gs, E, Ci/Ca, A/gs, A/E and A/Ci. The irrigation treatment (IT) effect was significant for the nutritional variables P, K, Ca, Mg, S, Fe, Mn and B and the physiological variables A, gs, E and Ci/Ca. The G \(\times\) IT interaction was significant for the physiological variables A, gs and E (Table 1).

Table 1 ANOVA for nitrogen (N), phosphorus (P), potassium (K), calcium (Ca), magnesium (Mg), sulfur (S), iron (Fe), zinc (Zn), copper (Cu), manganese (Mn), boron (B), net CO2 assimilation rate (A), stomatal conductance (gs), transpiration (E), internal carbon concentration (Ci/Ca), intrinsic water use efficiency (A/gs), instantaneous water use efficiency (A/E), and instantaneous carboxylation efficiency (A/Ci) in 18 cocoa genotypes evaluated in irrigated and non-irrigated environments.
Full size table

Traits’difference between cacao genotypes under different irrigation treatments

The genotypes had different responses for the physiological traits evaluated, allowing the selection of superior individuals. CCN 51, CP 196, CP 223 and CP 236 showed predictable behavior in both environments evaluated, indicating tolerance to water stress (Table 2).

Table 2 Scott–Knott test for grouping of means (p < 0.05) relating cocoa genotypes in irrigated (IR) and non-irrigated (NIR) irrigation treatments for the physiological variables A, gs, and E.
Full size table

The 18 genotypes were divided into two groups related to the contents of the nutrients N, P, Mg, S, Mn, Cu and Zn (Table 3). These contents ranged from 24.2 to 27.5 (N), 2.1 to 2.9 (P), 4.0 to 5.6 (Mg), 2.3 to 2.9 (S), 459.6 to 986.0 (Mn), 3.3 to 5.8 (Cu) and 37.6 to 56.3 (Zn) under the evaluated conditions. Overall, the genotypes ESFIP 04, CP 223, CP 43 and PH 16 had higher nutrient contents under the evaluated conditions. Furthermore, the nutrient content was generally higher in the irrigated environment than in the water-stressed environment (Table 4).

Table 3 Scott–Knott test for grouping of means comparing the contents of the nutrients N, P, K, Mg, S, Fe, Zn, Cu, Mn, and B in 18 cocoa genotypes. 1See codes in Table 1. Means followed by the same letter do not differ significantly from each other, at 5% probability level.
Full size table
Table 4 Scott–Knott test for grouping of means (p< 0.05) comparing the contents of the nutrients P, K, Mg, S, Fe, Zn, Cu, and B in irrigated and non-irrigated environments. 1See codes in Table 1. Means followed by the same letter in the column do not differ significantly from each other, at 5% probability level.
Full size table
Figure 1
figure 1

Pearson’s correlation coefficients between nutritional and physiological traits evaluated in 18 cocoa genotypes, in irrigated and non-irrigated environments. See codes in Table 1 for physiological and nutritional traits.

Full size image

Most correlations between traits were low, except for physiological traits. The highest correlation coefficients were observed between Ci/Ca and A/gs (r=-0.93), A/gs and A/E (r=0.92), A/E and A/Ci (r=0.71), A/gs and A/Ci (r=0.63), Fe and B (r=0.48), and Zn and Mg (r=0.43) (Fig. 1).

Selection of cacao genotypes based on the nutritional and physiological traits

Three plots were generated using the GT biplot method to understand the relationship between genotypes and nutritional and physiological traits (Fig. 2). The“Which-Won-Where/What” biplot was used to identify the superior genotypes for a set of traits in the irrigated environment. At the vertices of the polygon lies the superior genotype in its respective sector (Fig. 2A). The polygon was divided into five sectors, and the superior genotypes for each sector are TSH 1188 in Sector 01, CCN 51 in Sector 02, CP 197 in Sector 03, CP 196 in Sector 04, and BN 34 in Sector 05.

Figure 2
figure 2

GT Biplot approach in the evaluation of the performance of 18 cocoa genotypes in irrigated environment: (A) “Which-Won-Where/What”, (B) ranking of genotypes, (C) discriminant versus representative. PC 1 = Principal component 1, PC 2 = Principal component 2. 1See codes in Tables 1 and 2 for physiological and nutritional traits.

Full size image

The “ideal”genotypes are those whose projections of the ATC (Average tester coordination) on the horizontal axis correspond to the largest vectors, that is, those genotypes that have combined high performance and good stability for the different traits. CP 41, CP 43, CP 197 and CP 176 were the “ideal”genotypes for most of the traits (Fig. 2B). Nevertheless, CCN 51 had a superior performance in A/Ci, A/gs and A/E. Traits with longer vectors and with smaller angles in the ATC in the biplot C are considered more discriminant and representative, respectively (Fig. 2C). In general, the traits were able to discriminate the genotypes well, especially A/Ci, A/gs, A/E, Ci/Ca, K and Zn. S, A/gs and A/E were the most representative traits.

Figure 3
figure 3

GT Biplot approach in the evaluation of the performance of 18 cocoa genotypes in non-irrigated environment: (A) “Which-Won-Where/What”, (B) ranking of genotypes, C) discriminant versus representative. PC 1 = Principal component 1, PC 2 = Principal component 2. 1See codes in Tables 1 and 2 for physiological and nutritional traits.

Full size image

In the non-irrigated environment, the genotypes with the best performance were SJ-02 (sector 1), PH 16 (sector 2), ESFIP 04 (sector 3) and CCN 51 (sector 4) (Fig. 3A). The last-mentioned genotype showed similar behavior to that observed in the irrigated environment, mainly with high performance in the respective sector for the traits A/Ci, A/gs and A/E. The genotypes CP 43, CP 196, CP 197 and CP 176 were superior in the non-irrigated environment (Fig. 3B). CP 196 and CP 197 showed predictable performance in both environments. The traits Ci/Ca, Cu, Mg, K, P, A/gs and A/E were the most discriminant, while S, Ci/Ca, A/gs, and A/E were the most representative (Fig. 3C).

Discussion

Cacao has wide genetic diversity for agronomic46 and biochemical47 traits. The difference in the concentration of nutrients in cacao is due to differences in the genotypic constitution, requirement of nutrients and efficiency in their use. Furthermore, cacao has several adaptive mechanisms acting in favor of the survival and development of the plant under water stress, which are also genotypic-dependent48. It is expected that a drought-tolerant genotype reunites several physiological resources that enable it to thrive in yield and biomass production, even in limited environments. Following this principle and leveraging the genetic differences in physiological behavior and nutrient uptake under water-stressed conditions, we selected genotypes that had more evidence of being drought-tolerant.

When cacao faces water stress, significant changes occur in parameters such as leaf area, thickness, and leaf number49 The initial stage of seed development is dramatically impacted48. These alterations result in constraints on sugar transport and lipid availability in the seeds. As a common adaptive response to this challenging scenario, cacao tends to reduce biomass allocation to the roots. This is manifested in the frequent occurrence of premature leaf abscission, an effective strategy to minimize water loss through transpiration. Additionally, there is an observed development of thicker leaves, accompanied by a reduction in the number of stomata on the leaves50,51. These adaptations highlight the complex physiological response of cacao to water stress, underscoring specific mechanisms that contribute to its survival in under conditions. Nitrogen, in the form of nitrate in the vacuole, confers greater tolerance to water stress by contributing to the maintenance of cell turgor. Phosphorus and potassium, when accumulated in the plant, improve osmotic adjustment. Changes in phosphorus concentration may lead to higher efficiency in water use and stomatal conductance3. The absorption of magnesium, sulphur and manganese is also influenced by water stress. Deficiency of these macronutrients and micronutrients can directly affect photosynthesis52. This complex interaction between soil nutrients and the physiological responses of cacao to water stress allows us to comprehend the mechanisms of adaptation and tolerance of plants under water stress conditions.

In this study, the CP 43 genotype had better performance for nutritional and physiological traits in both irrigation treatments. However, CCN 51 showed a better response for tolerance to water stress. Despite the scarcity of studies on the physiological and biochemical response of these genotypes, it is noteworthy that the Ecuadorian clone CCN 51 displays high rates of CO2 assimilation (A) under open cultivation conditions compared to shaded agroforestry systems. When subjected to high radiation, CCN 51 tends to increase non-photochemical quenching of Chlorophyll a53. Furthermore, CCN 51 maintains high stomatal conductance rates (between 250-350 mmol m-2 s−1)54, exhibiting water use efficiency around 2.3 mmol mol−1. In physiological response to water stress, it is highlighted that CCN 51 shows low cadmium (Cd) absorption55, which is significant as elevated levels of this mineral can trigger competition with other essential nutrients such as Zn, Mn, Fe, and Cu56. These characteristics reinforce the adaptability and resilience of CCN 51 to adverse water stress conditions. Thus, while the CP 43 genotype showed superior performance in certain aspects, such as nutritional characteristics, it is CCN 51 that stands out in terms of water efficiency and stress tolerance, showcasing physiological adaptations that favor its performance under specific cultivation conditions57. These findings underscore the importance of considering not only overall performance but also specific responses to environmental factors when evaluating the suitability of different genotypes for specific agricultural practices.

It is worth remembering that cacao has photosynthetic characteristics of a shade-tolerant species, with a net carbon assimilation rate (A) at radiation levels ranging from 200 to 750 μmoL \({\text{m}}^{-2} {\text{s}}^{-1}\), with light compensation point ranging from 5 to 57 μmoL \({\text{m}}^{-2} {\text{s}}^{-1}\), and maximum value of A ranging from 1 to 8 μmoL \({\text{m}}^{-2} {\text{s}}^{-1}\). For this reason, most cacao plantations are carried out in shaded environments58,59. Soil water deficit reduces cocoa production, decreases seed size, and influences leaf renewal and flowering18. According to Carr and Lockwoods60, premature leaf fall, yellowing of basal leaves, wilting, small leaves and reduced growth are visible symptoms caused by drought in cocoa trees.

The physiological variables A, gs, Ci/Ca and E had higher values in the irrigated environment. The reduction in the Ci/Ca rate may be associated with the reduction in the stomatal conductance rate (gs) of the plants61. As shown in this study, there are cacao genotypes with alleles for drought tolerance, and the species has wide genetic variability. Therefore, it is possible and desirable to create breeding programs aimed at resistance to water stress. It is also worth remembering that water is a scarce and finite resource, which needs to be rationally used. In conclusion, genotypes CP 41, CP 43 and CCN 51 showed promise in terms of tolerance to water stress and accumulation of nutrients. In the comparative analysis between genotypes, it was observed that CP 41 showed superiority in various characteristics, specifically P, A/E, A/gs, Mn, and S when cultivated under irrigated conditions. Conversely, under non-irrigated conditions, CP 43 exhibited better performance in the nutritional features Mn, Mg, and Zn. Notably, the genotype CCN 51 excelled in key physiological traits, A/Ci, A/gs, and A/E, in both environments, indicating remarkable tolerance to adverse environmental conditions. As a result, this study provides an in situ assessment of these cacao genotypes, followed by a simultaneous selection of physiological and nutritional traits through the GT biplot approach. Grounded in multivariate analysis, this method identifies the strengths and weaknesses of each genotype, enhancing the selection process with greater robustness and efficiency. The obtained results have the potential to offer valuable insights into these genotypes, aiding informed decision-making for cultivation on farms or in large production areas.