Molecular, Physiological and Biochemical Responses of Theobroma cacao L. Genotypes to Soil Water Deficit

Six months-old seminal plants of 36 cacao genotypes grown under greenhouse conditions were subjected to two soil water regimes (control and drought) to assess, the effects of water deficit on growth, chemical composition and oxidative stress. In the control, soil moisture was maintained near field capacity with leaf water potentials (ΨWL) ranging from −0.1 to −0.5 MPa. In the drought treatment, the soil moisture was reduced gradually by withholding additional water until ΨWL reached values of between −2.0 to −2.5 MPa. The tolerant genotypes PS-1319, MO-20 and MA-15 recorded significant increases in guaiacol peroxidase activity reflecting a more efficient antioxidant metabolism. In relation to drought tolerance, the most important variables in the distinguishing contrasting groups were: total leaf area per plant; leaf, stem and total dry biomass; relative growth rate; plant shoot biomass and leaf content of N, Ca, and Mg. From the results of these analyses, six genotypes were selected with contrasting characteristics for tolerance to soil water deficit [CC-40, C. SUL-4 and SIC-2 (non-tolerant) and MA-15, MO-20, and PA-13 (tolerant)] for further assessment of the expression of genes NCED5, PP2C, psbA and psbO to water deficit. Increased expression of NCED5, PP2C, psbA and psbO genes were found for non-tolerant genotypes, while in the majority of tolerant genotypes there was repression of these genes, with the exception of PA-13 that showed an increased expression of psbA. Mutivariate analysis showed that growth variables, leaf and total dry biomass, relative growth rate as well as Mg content of the leaves were the most important factor in the classification of the genotypes as tolerant, moderately tolerant and sensitive to water deficit. Therefore these variables are reliable plant traits in the selection of plants tolerant to drought.

tolerant, moderately tolerant and sensitive to water deficit. Therefore these variables are reliable plant traits in the selection of plants tolerant to drought.
in the activity of oxidative stress enzymes, resulting in an increase in biomass [23]. Additionally, under conditions of low soil water availability there may be a shortage of Mg 2+ and alteration in the biomass allocation from roots to shoots [24].
In general, ROS production intensifies when plants are subjected to biotic and abiotic stresses, resulting in oxidative stress [11,12]. Antioxidative metabolism enzymes use Zn, Cu and Mn as cofactors [1,18]. Changes in the activity of these enzymes to remove ROS increase the plant's drought tolerance [25,13,26]. In addition, ROS play a fundamental role in the regulation of gene expression [27,28], perception and signal transduction [29].
Perception and signal transduction by plants under water stress conditions are driven by two distinct pathways, the ABA dependent and independent routes. During abiotic stress, ABA may be synthesized via the carotenoids biosynthetic pathway, in which the cleavage of cis-xanthophylls is catalyzed by a family of 9cis-epoxicarotenoide dioxygenases (NCED) [30,31], and acts as a messenger in endogenous stress responses [32,33]. In addition, some genes are negative regulators of ABA-dependent pathways, such as the family PP2C, encoding phosphatases, which in turn inhibit kinases and thus gene expression, and promote activation of anion (SLAC1) and cation channels [34].
In addition to genes known to be involved with water stress tolerance, over expression and/or repression of those involved in biosynthetic proteins routes, especially the pathway associated with carbon assimilation, are of great importance since they are related to the yield production of cultivated species [35]. The D1 protein, encoded by psbA, a component of PSII involved in photosynthetic electron transport, can be easily degraded and is continuously synthesized under stress [35]. On the other hand, the psbO protein, involved in the stabilization and oxygen evolution in the Mn cluster at PSII, has a fundamental role in photosynthesis [36] and performs a protective function for photosynthetic apparatus during abiotic stresses [37]. However, the high stability of PSII during drought observed in Festuca arundinacea, a highly drought tolerant species, is not associated with the accumulation of psbO, although its degradation affects the destabilization of the oxygen evolution complex under drought conditions [37].
The objectives of this study were to evaluate growth, chemical composition and oxidative stress of a sample of 36 cacao genotypes of different geographical origins subjected to water stress (drought). Also, to evaluate the expression of genes related to drought tolerance and biosynthesis of psbO and psbA proteins in genotypes identified in this study as tolerant and non-tolerant to water stress, aiming to elucidate possible mechanisms of drought tolerance and offer support for selection of genotypes to be grown in soils with low water storage capacity and/or in regions with irregular rainfall.

Plant material and growth conditions
A sample of 36 cacao genotypes, belonging to genetic groups Forastero, Criollo and Trinitario was selected for this study ( Table 1). As no information is available regarding the level of drought resistance of these genotypes, we selected original clonal accessions collected from different geographical regions used as progenitors in breeding programs and hybrids to compose the sample. Seminal seedlings were prepared from open pollinated seeds collected from clonal accessions at the Cacao Germplasm Bank of the Cacao Research Center (CEPEC), the research facility of the Executive Commission of the Cacao Farming Plan (CEPLAC), Ilhéus, Bahia. Five fruits were collected from each of the 36 genotypes, the seeds of each genotype were mixed and a randomly composed sample of 40 seeds were planted in 16 L pots containing soil as the substrate. Chemical and physical analyses of the soil were performed and fertilized according to the crop requirements during the seedling production [38]. The experiment was conducted in a greenhouse at CEPEC/CEPLAC, Ilhéus, Bahia, Brazil (14˚47'S, 39˚16'W, 55 m ASL).
During the time of the experiment, temperature and relative humidity were recorded (Fig. 1) using a thermo-hygrograph (Kipp & Zonen, model 836); and photosynthetically active radiation (PAR) was measured using a quantum meter (Model-QMSS SUN-1350 Apogee, City, USA). The maximum values of PAR inside the greenhouse ranged from 800 to 1200 mmol photons m 22 s 21 . Six months-old plants were divided into two groups and one group was subjected to drought by gradually reducing the soil water content by reducing water addition until the dawn leaf water potential (Y WL ) reached 22.0 to 22.5 MPa, these leaf water potentials were reached approximately 40-60 days after the beginning of the drought cycle. The second group of plants were used as controls and irrigated daily to maintain soil moisture near field capacity and Y WL between 20.1 to 20.5 MPa. Measurements of Y WL were done at the second or third mature leaf from the apex of the orthotropic axis between 2:00 and 4:00 am, using a pressure chamber (Model 1000, PMS Instrument Company, Albany, OR, USA) [39].

Growth parameters
Plant samples were collected at the beginning of the drought cycle (six months-old plants), when the Y WL of all genotypes was between 20.1 to 20.5 MPa and the soil moisture was near field capacity, and at 40 to 60 days after the beginning of the drought cycle when the Y WL of the different genotypes reached 22.0 to 22.5 MPa. Just before harvest, measurements were made for total leaf area per plant (TLAP), stem diameter (SD), plant height (PH), and leaf number per plant (LNP). The features SD and PH were measured using a digital caliper and ruler, respectively. At harvest the plants were divided into roots, stem and leaves.
Leaf area was measured by Li-Cor model Li-3100 leaf area meter (Li-Cor, inc. Lincoln, Nebraska, USA). Root area (ARS) was estimated after limiarization in the Gimp 2 software and subsequent analysis with the Sigma Scan Pro 5 program and root volume was estimated after immersion of roots in a known water volume and observing its displacement. Different plant parts were placed in paper bags and dried at 75˚C in a forced air circulation over to obtain total dry mass of the plant and its parts. From the dry biomass data of the different plant parts (root-RDB, stem-SDB, leaf-LDB) and total leaf area per plant (TLAP) several indices were

Macro and micro mineral nutrients
The leaf content of mineral macro and micronutrients was determined in all 36 genotypes studied. Approximately 200 mg of ground dry biomass was used for nitropercloric digestion (3:1). After digestion, Ca, Mg, Fe, Zn, Cu and Mn values were determined by atomic absorption spectrophotometery, P by colorimetry and K by flame emission photometry [43]. Nitrogen was determined by the Kjeldahl method after sulphosalicylic digestion [44]. Leaf mineral content was expressed as g plant 21 for each genotype and treatment.
The samples were immersed in liquid nitrogen, stored in a freezer at 280˚C and subsequently lyophilized. Extraction of enzymes and determination of their activities were performed following methodology described by Pirovani et al. [45]. Conversion of absorbance data (470 nm min 21 g 21 DW) to guaiacol consumption in mmol g 21

Gene expression
RNA was extracted from the second or third mature leaf from the apex of the orthotropic axis of six cacao genotypes [CC-40, C. SUL-4 and SIC-2 (nontolerant) and MA-15, MO-20, and PA-13 (tolerant)], identified during the data analysis. The leaf samples were immersed in liquid nitrogen, stored at 280˚C and subsequently lyophilized for gene expression analyses. For this study we used four genes: two candidate genes related to drought tolerance, involved in the ABA dependent pathway, NCED5 (9-cis-epoxycarotenoid dioxygenase 5) and PP2C (protein phosphatase-2C) and two genes related to proteins biosynthesis of PS II (psbA and psbO) ( Table 2). Approximately 0.02 g of each leaf sample was macerated in liquid nitrogen for RNA extraction with the RNAqueous kit (Ambion) following the manufacturer's recommendations. Samples of RNA were used for first-strand cDNA synthesis with RevertAid H Minus M-MuLV Reverse Transcriptase (Fermentas), according to the manufacturer's instructions using oligo d(T) 18 primers. The reactions were incubated at 65˚C for 5 min, 37˚C for 5 min, 42˚C for 60 min and 70˚C for 10 min. The primers were designed after analysis of conserved sequences in T. cacao ( Table 2). The q-PCR was performed in a RT-PCR thermocycler (Applied Biosystems, 7500 model) using the nonspecific detection sequence (fluorophore) SYBR Green I. The mix for the reaction was composed of cDNA as template, 0.5 mM of each primer and 12.5 mL of Maxima SYBR Green/ROX qPCR Master Mix 2x. Quantification of relative expression of genes were calculated as a percentage of the control treatment using the 2 -DDCt method [46] and the btubulin as endogenous control in order to detect changes in transcript number ( Table 2). leaf contents of N, P, K, Ca, Mg, Fe, Zn, Cu and Mn) were standardized as we measured them in different units (g, cm as well as ratios between them). The standardization was performed by the equation: Zij 5(Xij -Xj)/Sj, where Xij is the value of the i-th observation of the variable Xj; and Xj and Sj is the mean and standard deviation of the variable Xj, respectively. The 28 standardized variables were submitted to cluster analysis and factor analysis, using Statistica version 7 (Statsoft, Inc.Tulsa, OK, USA.). Nine of those variables made the greatest contribution to the formation of the first factor of the factorial analysis. These variables were submitted to colinearity analysis, based on tolerance and on the variance inflation factor (VIF), considering greater than 0.1 and less 10 [47], respectively, as the threshold for variable inclusion in the cluster and principal component analysis, using SPSS (SPSS, Inc., Chicago, IL). From the colinearity analysis it was found that eight variables were not collinear (TLAP, RDB, SDB, LDB, TDB, RGR, and leaf contents of Ca and Mg). These variables were used for cluster and principal components analyses. Cluster analysis was performed based on Euclidean distance and the dendograms constructed using the hierarchical agglomerative method [48].

Statistical analysis
We used a completely randomized design with 144 treatments [36 genotypes, two water regimes (control -Y WL between 20.1 to 20.5 MPa and drought -Y WL between 22.0 to 22.5 MPa) and two sampling times of plant material -baseline and 60 days of stress] and six replications (plants) for collecting RGR and NAR variables; with 72 treatments (36 genotypes and two water regimes) and six to eight replications to assess growth, oxidative stress and chemical composition; and with 12 treatments [six genotypes and two water regimes] and four replications (grounded pooled leaves of two plants) for gene expression assessment. Results were subjected to comparisons of treatment means using the Student t-test (P,0.05 and 0.01). Based on the results of the Student t-test we grouped the genotypes into three types: (i) tolerant genotypes, those that had from 0 to 10 significant variables; (ii) moderately tolerant genotypes, those that had from 11 to 15 significant variables; and (iii) sensitive genotypes, those that had above 16 significant variables.

Accumulation and partitioning of dry biomass
Soil water deficit significantly (P,0.05) influenced biomass production, reducing dry weight in all plant parts for most of the evaluated cacao genotypes, except EET-53, ICS-9, MA-15, OC-77, PA-150, PS-1319 and SPA-5 (Table 3). Significant reductions (P,0.05) in root (RDB), stem (SDB), leaf (LDB), shoot (SB) and total (TDB) dry biomass were found in 42, 50, 50, 58 and 64% of the genotypes, respectively, in relation to their controls, for each of these variables. Decreases in LDB, SB, SDB and RDB were observed mainly in drought sensitive genotypes (Table 3). Soil water deficit significantly (P,0.05) reduced leaf area per plant (TLAP), individual leaf area (ILA) and leaf number per plant (LNP) for most of the genotypes evaluated (Table 3). Significant reductions (P,0.05) were observed mainly for the LNP and PH variables in drought sensitive genotypes (Table 3).
In general, the cacao genotypes evaluated showed significant reductions (P,0.05) in stem diameter (SD), root volume (RV) and root area (ARS), with the exception of some tolerant genotypes (Table 3, Fig. 2). Overall in all genotypes tested, soil water deficit significantly reduced (P,0.05) growth variables such as SD, RV and ARS in 55, 75 and 81%, respectively, compared to the controls. Furthermore, no significant (P,0.05) intergenotypic reductions for R/S, SLB and LAR (Table 3) under water deficit conditions were observed. On the other hand, 42% of the evaluated genotypes showed significant reductions (P,0.05) for NAR and RGR, especially in sensitive genotypes, with decreases of 54 and 57%, respectively (Table 3).

Macro and micro minerals nutrients
Soil water deficit significantly (P,0.01) reduced leaf macro and micro nutrient content for most of the evaluated genotypes, except for some tolerant and moderately tolerant ones (Table 4). Reductions in leaf content of N, P, K, Ca and Mg were found for 28, 22, 22, 69 and 56%, respectively, of all the genotypes subjected to soil water deficit.
Water deficit sensitive genotypes when subjected to soil water deficit showed the highest significant (P,0.01) reductions in leaf N, P and K content, compared to control plants (Table 4).      The vast majority of the genotypes evaluated also showed changes in foliar micronutrient content when subjected to soil water stress, except for tolerant genotypes (MA-14, PA-13 and SIAL-169). There were significant reductions (P,0.05) in foliar contents of Fe, Zn, Cu and Mn in 53, 50, 81 and 69% of the genotypes evaluated, respectively (Table 4).

Enzyme activity
Overall, soil water deficit (drought) increased the activity of oxidative stress enzymes for most cacao genotypes evaluated, except for the tolerant genotype PA-13. The increase in peroxidase (GPX) activity was observed in 81% of the genotypes subjected to soil water deficit. Higher variations (P,0.01) were   (Fig. 3).
Regarding polyphenol oxidase (PPO) activity, there were significant changes (P,0.01) observed in 75% of the studied genotypes under water stress. The   highest values for the activity of PPO was found in moderately tolerant and susceptible genotypes (Fig. 4).

Identification of tolerant genotypes based on multivariate analysis
A multivariate analysis was performed to determine if the growth parameters, chemical composition and activities of oxidative stress (GPX and PPO) enzymes could provide information regarding selection of the most tolerant genotypes to water stress. Initially a cluster analyses based on the similarity of these variables was performed, using the differences (D) between control and water stressed plants within genotypes. The D values were used to construct a similarity matrix and a dendrogram was constructed based on similarity data (Fig. 5). The results showed the formation of three distinct groups (Fig. 5). The first group (I) was represented by 14 genotypes, the second (II) by seven and the third (III) by 15 (Fig. 5). There was a relationship between the groups formed and the number of significant variables for the different genotypes (Table 5). Furthermore, there was an association observed between the similarity, based on the analyzed variables and drought tolerance. Thus, genotypes PA-13, MA-15, OC-77, MO-20, PS-1319 and MA-14 were grouped as being tolerant to water stress, with lower D compared to their respective controls. They were part of the third group, whereas the second group was formed by CC-40, C. SUL-4, SIC-4 and SIC-17, considered nontolerant to water deficit, had higher D in relation to their controls (Fig. 5).
Next, from the factor analysis and colinearity test, we observed that the variables TLAP, RDB, SDB, LDB, TDB, RGR, Ca and Mg had the greatest contribution on the formation of the first factor and showed no colinearity among them. By submitting the D data of the non collinear variables to a cluster analysis and performing a dendrograma, four main groups were formed (Fig. 6). These results were similar to those groupings observed when a cluster analysis was performed using all growth variables, oxidative stress (GPX and PPO) and chemical composition. Thus, it can be suggested that the eight non-collinear variables are sufficient to separate the contrasting T. cacao genotypes in relation to tolerance to soil water deficits tolerance.
Principal components analysis formed groups, separating the more contrasting T. cacao genotypes regarding tolerance to soil water deficit (Fig. 7). Furthermore, the results agreed with cluster analysis by the agglomerative method of Ward (49). The first and second principal component explained 61 and 14%, respectively, of the total variance with a cumulative eigenvalue of 75% (Table 6). From the eigenvectors values, we observed that the variables that had the higher contribution in the formation of the first component were, TDB, RGR, LDB and foliar Mg content while the variable SDB and TLAP had the highest contribution in the second component. The remaining components explained 11, 7, 3, 2 and 1%, respectively, of the total variance (Table 6).
According to the first component, tolerant genotypes (Fig. 7) showed the greatest intergenotypic distinction. These genotypes had the lowest D values for linear combinations of the analyzed variables. Moreover, sensitive genotypes were

Gene expression
We observed increased expression of drought tolerance candidate genes in the studied genotypes. Genes associated with ABA biosynthesis and genes related to biosynthesis of proteins of PSII were expressed in genotypes considered as nontolerant to soil water deficit and repression of these genes was observed for tolerant genotypes, compared to controls (Fig. 8). Furthermore, regarding the number of psbO transcripts, there was a significant two fold increase (P,0.01) in the expression of the non-tolerant genotype C. SUL-4, whereas for the tolerant genotypes MO-20 and MA-15 there was a significant suppression (P,0.01) by 0.9 and 0.5 times, respectively (Fig. 8 A). Furthermore, there was a significant increase (P,0.01) in the number of psbA transcripts for the tolerant genotype PA-13 and the non-tolerant genotypes CC-40 and SIC-2 of 36, 12 and 2 times, respectively, compared to controls, while MA-15 showed repression of that gene by 0.8 times (Fig. 8 B). A significant increase (P,0.01) in the expression of NCED5 was found, mainly in non-tolerant genotypes C. SUL-4 and CC-40, which corresponded to 14 and 3 times, respectively, to that of control plants. Furthermore, for tolerant genotypes MA-15 and PA-13, we observed a significant suppression (P,0.01) by 0.4 and 0.2 fold, respectively, in the expression of that gene (Fig. 8C). Also, there was an over expression of PP2C, especially in non-tolerant genotypes C. SUL-4, CC-40 and SIC-2, with increases of 8, 3 and 2 times, respectively, while for tolerant genotypes PA-13 and MA-15 no significant increases were found (Fig. 8D).

Discussion
Soil water shortage is considered a major limiting factor in the production of many crops throughout the world. Physiological, biochemical and molecular responses in plants subjected to drought can be used as selection criteria for crop tolerance to this abiotic stress [27,20,49]. In genotypes with no tolerance to drought, soil water deficit promotes significant alterations in growth and   [50]. Similar responses have been reported for Hippophae rhamnoides [51] and Populus spp. [52,8], which also showed significant reductions in total biomass accumulation and root/shoot ratio. Of the 36 T. cacao genotypes evaluated, sensitive genotypes showed the greatest damage at the leaf level when subjected to water deficit, with sharp reductions in TLAP, LNP and ILA. On the other hand, tolerant genotypes showed no alterations in these variables under water stress conditions ( Table 3).
Reductions of TLAP, LNP and ILA promote, among other factors a decrease in photosynthesis and contributes significantly to the inhibition of plant growth [53]. In T. cacao, reductions in growth rates of leaf area and of total leaf area can be considered one of the earliest plant responses to stress as a result of the reduction in cell turgor and net photosynthetic rate [54,55]. In clones of Populus subjected to cycles of soil dehydration and rehydration, changes in TLAP were explained by differences in the number of leaves and the further expansion of ILA [56].
Drought conditions induced significant reductions in RGR and NAR (42% for both variables) in the studied cacao genotypes (Table 3). It is known that, in tree species, in general, NAR and RGR are differently affected by low soil water availability, which indicates that responses to water stress are complex, heterogeneous and may be consistent with the geographical distribution of each species [57,58].
In the present study, the cacao's responses to drought conditions in relation to height, SD, RV and ARS were quite varied among the genotypes, but the nontolerant genotypes showed a marked reduction in the values of these variables. On the other hand, for drought tolerant genotypes these changes were not similar to results found in Quercus sp [59] and Populus sp [60,61,8]. The genotypes that showed marked reductions for the RV and ARS variables also showed decreased SB (Table 3, Fig. 2), suggesting that plants sensitive to water stress show reductions in both the root and the shoot growth. Furthermore, limitation of the root system of these genotypes influenced the absorption of water and nutrients, thereby affecting the plant water status. We have also observed that cacao genotypes tolerant to drought maintained a root growth similar to the control plants, showing higher amounts of fine roots (Fig. 8). In contrast, in genotypes that showed significant reductions in growth variables, the proportion of fine roots also showed reductions. Silva and Kummerow [62] found, under field conditions, that plants of T. cacao produced large numbers of fine roots (diameter ,1 mm), which renewed quickly between one and 10 days, and growth were dependent on the frequency of rainfall. The dynamics of growth and renewal of roots, among other factors, can affect plant growth [63,64]. Tschaplinski et al. [65] in studies with Populus found that the clones most tolerant to water stress showed phenotypic plasticity in relation to greater carbon allocation to the roots, favoring increased root density and, consequently, occupying a greater soil volume, thereby restoring the water balance in the plant.
The responses of plants to drought at the mineral nutrition level are still poorly studied [16], although mineral macro and micronutrients have specific functions and may be required in large amounts by plants [66]. In the present study water deficit resulted in significant decreases in the mineral nutrient contents of leaves, a similar response of mineral nutrient reduction was observed in Fagus sylvatica when subjected to drought [16]. The cacao genotypes that were more tolerant to soil water stress showed no significant differences in leaf N, P and K contents between water deficit and control (Table 4). Usually, high concentrations of N-NO 3 are deposited in the vacuole, contributing significantly to the maintenance of cellular turgor, thus conferring tolerance to drought conditions [66]. Furthermore, changes in P concentrations can have positive effects by increasing water use efficiency and stomatal conductance [67]. Moreover, under water stress, activation of several transcription factors and regulation of gene expression depend on phosphorylation of protein mediated by protein kinases [66]. For K, an essential macronutrient for plant growth and development, accounting for nearly 70% of nutrients in the cacao xylem sap [68], a decrease in foliar nutrient content was found mainly for sensitive genotypes that also showed significant reductions of TDB and NAR. Potassium acts to regulate osmotic potential, required for enzyme activity and protein and carbohydrate syntheses, and helps in the process of stomatal opening and closure, and participates in water relations and cell elongation. Potassium deficiency slows plant growth, promotes leaf chlorosis, necrotic spots and shortening of internodes [69,66].
Although the content of macronutrients showed differences among genotypes, Ca and Mg content exhibited the greatest reductions with decreases of 69 and 56%, respectively (Table 4). However, tolerant genotypes maintained the content of these elements similar to controls. Maintaining high Ca and Mg content in these genotypes may have contributed to the increase in biomass and leaf area [66], activation of protein kinases, osmotic regulation and the opening and closing of stomata [20,70]. On the other hand, the marked deficiency of Ca and Mg found in sensitive genotypes may have influenced the highly significant reduction in shoot biomass [71].
Under water stress conditions, plants may exhibit micronutrient deficiency [15] that causes damage at the metabolic cellular level, since micronutrients have an important role in the protection against oxidative stress and are involved in the regulation and activation of enzymes that remove ROS [18]. In this study, the effects of water stress reduced Fe, Zn, Cu and Mn content for most genotypes, indicating that water stress influenced the uptake of these micronutrients by the cacao plants. Furthermore, the deficiency of these minerals may have interfered in photosynthesis and nitrogen fixation [1,72], and consequent biomass accumulation, and in the activities of peroxidases and polyphenol oxidases, enzymes responsible for elimination of ROS [18]. Micronutrients act as cofactors for enzymes of the antioxidative metabolism, Fe 2+ for catalases and peroxidases [73], Zn for superoxide dismutase and other enzymes of the antioxidative metabolism [74,75,18], Cu for polyphenol oxidase, and Mn activates superoxide dismutase [18], enzymes contributing to drought tolerance in plants. It is suggested that in addition to water deficit per se, the reduction in area and volume of the root system contributes to the poor uptake and promotes the deficiency of these elements, aggravating the response of the genotypes to drought [75].
Under conditions of soil water deficit, plants tend to increase the production of ROS, as one of the first plant responses to stress, due to stomatal closure and reduction in CO 2 fixation, which leads to excess excitation energy not being dissipated by the plant protection mechanisms [8,76]. Most cacao genotypes in our study showed significant increases in GPX and PPO activities. It is inferred that Fe deficiency may have contributed to the reduction of GPX activity for some moderately tolerant genotypes (Fig. 3), since, as mentioned above, this element acts as cofactor of peroxidase enzymes [73]. Oxidative stress enzymes are activated to remove ROS, which can promote cell damage, senescence and leaf abscission under water stress conditions [76] and induce programmed cell death [25]. Polyphenol oxidase promotes removal of hydrogen peroxide (H 2 O 2 ) [25,13]. Studies have shown a relationship between changes in peroxidase activity and stress tolerance and this may be an adaptation mechanism of plant tissues to stresses [77,78].
From the results of PPO activity it was not possible to separate cacao genotypes contrasting tolerance to soil water deficit. PPO enzymes are found in thylakoids and plastids, but there is not much information about the effects of changes in the activity of these enzymes during plant growth in response to water stress [79]. In most studies addressing the activity of PPO, there is a relationship of this enzyme to physiological damages. Polyphenol oxidase activity increases in response to different stresses [80,81,82].
Plants under abiotic stresses show changes in gene expression and regulation, in both the short and long term, as tolerance responses to unfavorable conditions [83]. In this study, cacao genotypes tolerant to water stress showed no changes in gene expression, contrasting with results found in Arabidopsis thaliana [84]. However, this is most likely due to the fact that the duration and intensity of the drought stress imposed in our study were applied gradually and over a longer period of time, which could be the reason transcription of some genes may have stabilized in tolerant genotypes (Fig. 8). The large accumulation of psbO transcripts for sensitive genotypes and repression for tolerant genotypes (Fig. 8A) suggests that its accumulation cannot be directly linked to drought tolerance, although the degradation of psbO protein probably destabilizes the oxygen evolution complex under drought conditions [37] and its reduction may limit plant growth and the concentration of other proteins encoded by both psbA and PSBP [36]. Moreover, the increase in the number of psbA transcripts, which encodes the D1 protein of the reaction center of PSII, may indicate that the protein is differentially expressed and easily damaged under water stress conditions [85].
Tan et al. [84] studying five genes of the NCED family in Arabidopsis reported that over a period of 35 h there was increased NCED5 expression in flowers and leaves under water stress conditions. Chao et al. [86] found that Mg deficiency resulted in an increase in ABA concentrations in leaves of Oryza sativa. This was also found in the current study with the increased expression of NCED5 in sensitive genotype C. SUL-4 ( Fig. 8 C). However, there are few studies related to the function and expression of NCED5, mostly performed in Arabidopsis, with increased expression of this gene under stress conditions [84,87]. Our results suggests that over expression of PP2C in genotypes susceptible to drought may indicate inactivation of protein kinases, and the consequent blocking of signal transduction in pathways dependent on ABA, phosphorylation, activation of transcription factors and expression of genes that confer drought tolerance [88,35].

Conclusions
Soil water deficit affected the majority of the physiological and biochemical variables as well as gene expression in the cacao genotypes evaluated in this study. Multivariate analysis showed that growth variables LDB, TDB, RGR and TLAP as well as the content of Mg in leaves were the most important variables in the separation of the genotypes as tolerant, moderately tolerant and sensitive to soil water deficit, therefore these traits are important in the selection of plants tolerant to drought.
Supporting Information S1