Total RNA was extracted according to previously reported protocols (Chang et al., 1993) with some modifications described previously (Osorio Zambrano et al., 2021). The purity and integrity of the RNA were verified with a Bioanalyzer 2100 (Agilent Technologies^®). The RNA samples of the three individuals (n = 3) per clone and treatment that presented the highest values of RNA integrity (RIN > 6) were selected for the construction of the cDNA sequencing libraries (Supplementary Table 8). A total of 12 cDNA libraries were indexed and generated after polyA RNA enrichment using the NEBNext^® Ultra™ kit (New England BioLabs) and were subsequently sequenced with paired ends of 150 bp using a HiSeq platform (Illumina, Inc.).

The expression of DEGs that were identified as tolerance candidate genes based on their molecular and cellular functions, as well as their log₂FC values and statistical significance, was validated via RT-qPCR. RT-qPCR analysis of the selected genes was performed according to previously reported protocols (Osorio Zambrano et al., 2021) using a Quant Studio 3 Real-Time PCR System (Applied Biosystems, CA, USA). The relative expression levels of the selected genes were normalized with respect to the expression level of the ubiquitin gene via the comparative Ct method (2 ^−ΔΔCt) (Pfaffl, 2001). The experiment was carried out in five independent biological replicates (individuals) per treatment (n = 5) (Supplementary Table 6). The list of primers designed and used in this validation is presented Supplementary Table 3. The reference gene used was selected after three-gene stability analysis using RefFinder platform which integrates the ΔCt method, geNorm, NormFinder, and BestKeeper (Xie et al., 2023). Each algorithm generated a stability ranking based on variation among Cq values, and RefFinder calculated the geometric mean of the ranking scores to obtain a comprehensive stability value, which was used to identify the most stable reference gene for normalization and was consistent with results previously obtained in cocoa plants (Gallego et al., 2018) (Supplementary Figure 6). Data from the validation of candidate drought-tolerance genes is available in Supplementary Tables 3, 6, 7, 9.

The Ψ_leaf values for the WW group ranged from -0.3 to -0.4 MPa, indicating sufficient hydration. In contrast, under DS, the Ψ_leaf dropped to an average of -2.5 MPa. All clones showed significant differences in their Ψ_leaf between treatments (P ≤ 0.05). This was accompanied by a significant decrease in soil VWC to 25% (Supplementary Figure 2A). At D52, the Ψ_leaf between clones varied: TSH565 and ICS60 had significantly greater Ψ_leaf (-2.3 MPa) compared to EET8 (-2.93 MPa). After rehydration (D59), the Ψ_leaf of the clones under DS recovered to values like those of WW plants indicating that although the three clones experienced a severe water deficit of 52 days, they were able to recover to almost their initial hydric state, which is a trait of drought tolerance (Figure 1A).

For WW plants, the RWC was greater (62%) for TSH565 than for EET8 (56%) and ICS60 (58%). At D52, only clones EET8 and ICS60 showed significantly reduced RWC values (P ≤ 0.05) compared to their respective control group (WW): 47% and 49% decrease respectively, whereas TSH565 showed a RWC like unstressed plants. At D59, the EET8 and ICS60 plants subjected to DS reached RWC values like those of their controls, with no differences between clones or between treatments (Figure 1B).

A and g_s were significantly lower (P ≤ 0.05) in DS plants than in the WW plants (Figures 2A, B). In WW plants, A was similar between clones: 4.61; 5.0 and 4.87 µmoles of CO₂ m^-2s^-1 for EET8, ICS60 and TSH565, respectively, whereas under DS, A was almost completely inhibited by 97% (0.13 µmol of CO₂ m^-2s^-1) in EET8 and ICS60 clones, while was 91% reduced (0.32 µmol of CO₂ m^-2 s^-1) in TSH565. Interestingly, after rehydration at D59, A recovered around 50% of the values exhibited by their respective control plants in the EET8 and ICS60 clones, while TSH565 exhibited a full recovery, with similar photosynthetic rates to those of its respective controls (Figure 2A). On the other hand, the g_s was reduced on average by 85% under DS with respect to the WW plants: dropping in average from 0.113 mol H₂O m^-2s^-1 to an average of 0.015 m^-2s^-1. As was observed for A, at D59, EET8 and ICS60 presented g_s values of approximately 50% of the value recorded for the control plants, whereas TSH565 recovered similar g_s values as the WW plants (Figure 2B).

Likewise, at D52, the WUE_i was significantly lower (P ≤ 0.05) in the clones subjected to DS than in WW, whose values were 42 µmol of CO₂ mol H₂O^-1 for EET8 and ICS60, and 45 µmol of CO₂ mol H₂O^-1 for TSH565. Under DS, WUE_i was reduced on average by 81% in EET8 and ICS60, whereas TSH565 showed only 29% reduction, with significant differences (P ≤ 0.05) between the clones. At D59, the WUE_i of all clones recovered to the values of the WW plants, without differences between clones or water treatment groups (Figure 2C).

At D52, F_v/F_m decreased to 0.537 (P ≤ 0.05) in EET8 compared with F_v/F_m of WW plants (0.725). A similar reduction pattern was observed for ICS60, with a F_v/F_mdropping from 0.757 (WW) to 0.632 (DS). Although both values indicate photoinhibition, an F_v/F_m close to 0.5 (EET8) reflects a more critical physiological state than 0.6 (ICS60). Interestingly, in TSH565, F_v/F_m was not significantly reduced (LSD, P > 0.05); between WW and DS groups. Similarly, at D59, the F_v/F_m of EET8 and ICS60 clones recovered in average 92% of the baseline value of WW, whereas TSH565, recovered the reference values of the control plants (Figure 1C).

Under DS, the accumulation of DWt was significantly reduced for clones EET8 and ICS60, with values 43% and 53% lower, respectively, compared to WW plants. This decrease was approximately 12% for clone TSH565 and was not statistically significant. For EET8 and ICS60 plants subjected to DS, the DWr, DWs and DWl were all reduced, indicating affectation of growth in all their organs, whereas for TSH565, only DWl was reduced, passing from 7.55 g to 4.57 g (39.5% reduction). Comparatively, for EET8, this decrease was 67%, ranging from 10.08g (WW) to 3.28g (DS), while ICS60 showed 70% reduction, ranging from 11.64g (WW) to 3.42g (DS). For these clones, the DWr, also decreased 30% and 40% respectively under DS, exhibiting a similar reduction of their DWs (28% and 35%, respectively) compared to WW plants. In contrast with TSH565, DWr and DWs did not differ between treatments (Figure 2D).

To identify differences in the responses of each clone to DS, the DEGs were further explored by applying a minimum threshold on the absolute value of the log₂FC > 2. With this filter, there were 328 upregulated genes in TSH565 and 404 genes in EET8. Among this set of genes, 136 were common to both clones, whereas 192 and 268 genes were unique to TSH565 and EET8, respectively. The total number of genes downregulated in clone TSH565 was 699, and 782 in EET8. Among these, 332 genes were common to both clones, whereas 367 were unique to clone TSH565 and 450 were unique to clone EET8.

The upregulated and downregulated genes in response to DS were analyzed via KEGG pathway ORA analysis: interestingly, the phenylpropanoid biosynthesis pathway was upregulated in both clones (Figure 4B). Additionally, in the EET8 clone, different metabolic pathways, the metabolism of amino acids and nucleotides, and the metabolism of starch and sucrose were also well represented among up-regulated genes. With respect to the DS -downregulated genes, the plant–pathogen interaction pathway was represented among the common downregulated DEGs. More specifically, the plant hormone signal transduction pathway was downregulated in EET8, whereas in TSH565, the biosynthesis of flavonoids was significantly represented among the downregulated genes (Figure 4B).

Using the filtered genes, the GO analysis of the common upregulated genes revealed that the biological process terms response to stress, response to stimuli, defense response, response to levels of oxygen, and biosynthesis of flavonoids were the most represented. The enriched molecular function terms were binding to other molecules and catalytic functions. Overall, the genes with the greatest change in expression, which are shared by both clones, were related to the stress response (Supplementary Figure 3A).

Regarding downregulated biological processes common to both clones, GO ORA analysis revealed genes related to terms like cell morphogenesis, cell wall biogenesis, response to osmotic stress and abiotic stimulus and response to salt stress; the enriched cellular component term was the extracellular region (Supplementary Figure 3B). In accordance with these findings, DS caused downregulation of genes associated with growth processes, as well as a response to other types of stress, suggesting that DS negatively affects these processes similarly in both clones.

The GO analysis of the 192 upregulated genes in TSH565 revealed only two overrepresented biological process terms: response to stimuli and response to stress. The enriched molecular function terms were related to binding unfolded proteins and the activity of protein kinases (Figure 5A). Several genes within these categories are annotated as components of ABA-related signaling pathways.

The GO analysis of the downregulated genes in TSH565 revealed that the biological process terms flavonoid metabolism, shoot development, response to jasmonic acid, leaf morphogenesis, and responses to external stimuli, among others, were overrepresented (Supplementary Figure 4A).

On the other hand, in EET8 clone, ORA of GO terms revealed up-regulation of genes related to biological process like response to chemical stimulus, response to osmotic stress, response to hydrogen peroxide, response to lack of water, response to saline stress, response to abiotic stimulus, response to stress, response to stimulation by ABA, defense response, response to growth regulators such as ethylene and gibberellins, and secondary metabolic processes, among others (Figure 5B). The enriched molecular function terms associated with these processes included binding to other molecules, catalytic functions, and the activity of transcription factors. Importantly, same genes could be involved in various processes, such as stress response, response to stimulation by ABA, and response to lack of water and abiotic stress. For instance, several transcription factors were found among the upregulated DEGs for this clone: NAC, WRKY, MYB, ERF/AP2, SRG1, CRF2, bHLH, Zinc finger_C2H2, ARF5, IAA6, and ERF114, which are all associated with the response to water deficit in plants and the ABA-dependent and ABA-independent pathways.

The GO analysis of the downregulated genes in this clone revealed that the biological process terms regulation of hormone levels, transport of hormones, response to auxins, ethylene, and stimuli, endogenous developmental processes, tropism and gravitropism, and cell wall biogenesis were overrepresented (Supplementary Figure 4B).

Finally, considering the high number of DEGs identified and the great diversity of biological processes involved in the response to DS, either in common or unique to each clone, we performed a non-exhaustive selection of the most functionally relevant DEGs associated with these GO categories and KEGG pathways, which supported the response of tolerance to DS in T. cacao (Supplementary Tables 1, 5). We corroborate these functions through orthologous relationships with genes from other model plant species, and the literature, to propose them as drought-stress tolerance candidate genes in cacao.

Thus, the genes with the greatest log₂FC under DS were related to osmoprotection, cell growth, synthesis of the ABA precursor, transport, signaling, defense, transcription factors, and protection from oxidative damage via the antioxidant system. Furthermore, a more directed search was conducted on genes related to the antioxidant system, osmolyte biosynthesis and PSII function, based on both the results of this research and previous physiological characterizations of drought response in cacao plants (Supplementary Table 2).

Additionally, the differential expression of the genes reported in the literature, as either target genes or interacting partners of those selected based on their relatively high log₂FC values (Supplementary Tables 1, 2) was verified. Among the DEGs identified as most relevant (based on significance, Log₂FC, or literature support), eight genes were selected to validate the RNA-seq expression profile using RT-qPCR (Supplementary Figure 5A). The RT–qPCR results were consistent with the RNA-seq data, except for one gene (FER, Supplementary Figure 5C). Changes in relative expression showed a positive correlation (r = 0.63; P = 0.01) between the two methods (Supplementary Figure 5B), thereby validating the workflow and analytical parameters applied in the DE analysis.

Protein kinases and phosphatases play a fundamental role in stress recognition and signal transduction. In both clones, several of these genes were found to be upregulated and downregulated (Figure 6; Supplementary Table 1), including receptor-like kinases (RLKs) involved in hydraulic signals perception, wall-associated protein kinases (WAKs), linked to cell wall extension; proline-rich extension-like kinases (PERKs), essential for maintaining cell wall integrity, and histidine kinases (AHKs), responsible for sensing turgor changes (Christmann et al., 2013; Shinozaki et al., 2017). The overexpression of these receptors highlights their role in mediating the drought response in T. cacao and underscores the complexity of signaling machinery deployed by these clones to perceive and trigger a response to DS. Conversely, the downregulation of some receptor genes aligns with previous reports describing the complex and dynamic regulation -both positive and negative- within the signaling networks that orchestrate drought stress responses (Zenda et al., 2019).

Among the genes that are most strongly induced by DS in both clones was a histidine phosphotransferase (HP; Tc09v2_g024060) (Supplementary Tables 1, 5). HPs are key elements of two-component signaling systems, transferring phosphorylation signals of histidine kinases (HKs) to response regulators. This DS-induced gene is an ortholog of the rice gene OsAHP1, a positive regulator of cytokinin-mediated signaling pathways that contribute to drought and salinity tolerance and promote root elongation (Sun et al., 2014; Liu et al., 2017).

Similarly, in both clones, two genes encoding FERONIA receptor kinases (Tc10v2_g001630 in EET8 and Tc02v2_g028060 in TSH565) were highly upregulated (Supplementary Table 1). FER plays a critical role in vegetative growth regulation through its interaction with the rapid alkalinization factor (RALF) peptide (Supplementary Table 2) (Wang et al., 2020a). It also acts as a signaling hub integrating pathways related to cell expansion, energy metabolism and stress response, making it a versatile regulator of growth and survival (Liao et al., 2017). A mechanism involving FER activation by ABA and interaction with RALF has been proposed to coordinate growth and stress signaling (Chen et al., 2016).

Interestingly, while FER was upregulated in EET8 and TSH565 clones, RALF expression diverged: it was upregulated in EET8 (Tc01v2_g024780) but downregulated in TSH565 (Tc05v2_g025930). This differential regulation suggests that the FER-RALF interaction in EET8 may have contributed to reduced root growth under DS, whereas in TSH565, FER induction combined with RALF repression supported continued root development (a 7.5% increase in DWr), enhancing water uptake and maintaining cellular turgor. Given FER´s multifunctional role in stress adaptation, it emerges as a strong candidate gene for mediating drought tolerance in T. cacao (Figure 7A).

In the MAPK signaling pathway, MAPK protein cascades are sequentially activated through the phosphorylation of at least three mitogen-activated protein kinases (MAPKKK, MAPKK, and MAPK). This pathway is a key regulator of plant development and response to stress (Dinesh et al., 2016). Here, the gene encoding MAPKKK13(Tc09v2_g029520), involved in the first step of this cascade, was highly expressed in the TSH565 clone (Supplementary Table 1). This gene is orthologous to Arabidopsis thaliana MAPKKK14, a Kinase implicated in ABA-mediated signaling pathway (Sah et al., 2016). Consistent with findings in other species the upregulation of MAPK cascade genes in response to stress highlights their central role in adaptative responses and underscores their potential as targets for improving drought tolerance in crops (Wang et al., 2015; Jia et al., 2016; Ye et al., 2017; Muhammad et al., 2019).

This study highlights the crucial role of abscisic acid (ABA) in the drought stress response of two T. cacao clones to DS, EET8 and TSH565. Both clones activated several ABA-related genes to cope with water deficit (Supplementary Tables 1 and 2). A key finding was the upregulated of beta-carotene 3-hydroxylase1 (BCH, chloroplastic), an enzyme essential for synthesizing zeaxanthin, which is a precursor to ABA. Zeaxanthin is a vital component of the xanthophyll cycle, a mechanism that helps protect thylakoid membranes and dissipates excess light energy as heat (Du et al., 2010). Studies in rice and Arabidopsis have shown that manipulating the BCH gene can improve drought tolerance by boosting xanthophyll cycle activity and ABA-mediated responses, suggesting a similar function in T. cacao (Davison et al., 2002).

Additionally, the gen Tc05v2_g012420, encoding glucan endo-1,3-beta-glucosidase (an ortholog of the Arabidopsis BG1 and BG2), was strongly upregulated in both clones, with higher expression in EET8 (7.5-fold) (Shinozaki et al., 2017). This enzyme likely hydrolyzes ABA-glycosyl ester (ABA-GE), a storage form of ABA, releasing active ABA during DS, contributing to stomatal closure for improved water conservation.

In TSH565 clone exhibited a more robust response, activating both ABA-dependent and independent pathways. It showed increased expression of genes involved in ABA synthesis and catabolism, including ZEP (zeaxanthin epoxidase), CYP707A2 (abscisic acid 8′-hydroxylase 2), and ABA4 (ABA DEFICIENT 4, chloroplastic), along with the extracellular ABA receptor PYL9 (Tc01v2_g028880). It also upregulates DREB2D, a transcription factor linked to ABA-independent DS in melon (Ansari et al., 2017) and clones of Camellia sinensis, a species taxonomically related to T. cacao (Parmar et al., 2019) (Figure 6; Supplementary Table 2).

Additionally, RBOHD (Respiratory burst oxidase homolog D) was induced in both clones, suggesting that ABA signaling in T. cacao, is dependent on ROS production. This is consistent with reports of NADH-dependent, calcium-activated oxidase generating superoxide during stress responses (Kwak et al., 2006; Chapman et al., 2019) and with evidence linking RBOHD activity to lateral root development (Yao et al., 2017).

During the DS response, the transport of synthesized ABA to target organs is essential. In both cocoa clones, the genes ABCG11 and ABCG36 (ATP-binding cassette family G), homologous to AtABCG25 (a key ABA transporter), along with several NRT1/PTR family genes, were upregulated (Figure 6; Supplementary Table 2) (Shinozaki et al., 2017; Kuromori et al., 2018; Parmar et al., 2019). Other ABC transporters also showed differential expressions: ABCB20 in EET8, associated with auxin transport and detoxification, and ABCC3 in TSH565, linked to glutathione conjugate transport (Pierman et al., 2018).

Anion channels, which play critical roles in cell signaling, osmoregulation, metabolism, nutrition and stress tolerance were also implicated (Chen et al., 2019). In particular, the SLAC1 anion channel (Tc03v2_g022100), which mediates ABA-dependent stomatal closure by promoting anion and K⁺ efflux from guard cells to reduce turgor and minimize water loss was notably downregulated in EET8 (log₂FC = -1.92) (Geiger et al., 2011). This downregulation suggests impaired stomatal regulation, potentially contributing to greater water loss, lower leaf water potential, reduced RWC in this clone (Figure 7A).

TFs are master regulators of gene expression (Osakabe et al., 2014; Yang et al., 2016). In EET8 and TSH565 clones, STOP1 (Tc09v2_g024900) and bHLH96 (Tc01v2_g008170), were overexpressed, while bZIP53 (Tc04v2_g024560) was specifically upregulated in TSH565. STOP1 regulates tolerance to multiple stresses including DS tolerance, hypoxia and heavy metal detoxification (Iuchi et al., 2007; Sadhukhan et al., 2017; Enomoto et al., 2019).

The bHLH TF family, one of the largest in plants, participates in developmental processes and stress signaling. They participate in jasmonic acid-mediated signaling pathways (Liu et al., 2014) and in ABA-mediated stomatal regulation through ABA kinases (AKs) substrates targeted by ABA-activated SNRK2 kinases (SNF1-related protein kinase 2s) during DS (Shinozaki et al., 2017). In maize, bHLH TF enhances drought tolerance by promoting root growth and ABA synthesis (Li et al., 2019). The strong upregulation of bHLH96 in both clones suggests a significant role in their drought response.

In TSH565, the TF bZIP53, a regulator of low energy signaling (Supplementary Table 1) was strongly upregulated by DS, potentially activating the asparagine biosynthesis pathway (ASN1, GLNS, ASP and PepCK) (Baena-González et al., 2007) (Supplementary Table 2). bZIP53 is also a target of SnRK1-like protein kinase (SNF1-related protein kinase 1), whose function is to regulate the energy balance that is essential for survival under stress (Dietrich et al., 2011). Interestingly, all these genes were upregulated in TSH565, counteracting the downregulation of other genes affecting energy metabolism under DS. These findings suggest that bZIP53 may help maintain energy homeostasis and contribute to enhanced DS tolerance observed in this clone.

Moreover, among the transcription factors induced in this clone, SRG1 (SCARECROW protein), which has been reported to be a repressor of the GA (gibberellic acid) signaling pathway, is involved in the negative control of plant growth (Yang et al., 2016).

Also, several TF were upregulated in EET8 clone indicating a strong and sustained activation of regulatory mechanisms mediated by multiple TF families. The maintenance of this transcriptional regulation at D52 suggests that the stress response in this clone is not transient but rather reflects a prolonged activation of regulatory pathways. This pattern is consistent with a strategy based on the continuous engagement of multiple signaling and regulatory mechanisms to cope with prolonged DS (Yang et al., 2016; Shinozaki et al., 2017).

Similarly, the complex response to drought involves endomembrane trafficking, which is linked to signaling pathways through changes in cellular processes and cargo delivery (Wang et al., 2020b). In TSH565, the CNIH4 (Tc01v2_g031670) gene was highly upregulated under DS (Supplementary Tables 1 and 5). This gene encodes a cargo receptor involved in transporting integral membrane proteins to their proper locations (Rosas-Santiago et al., 2015, 2017). In Arabidopsis, CNIHs regulate the sorting, transport and localization of glutamate receptor channels (GLRs) that mediate Ca²⁺ transport in various contexts (Wudick et al., 2018). In rice, OsCNIH1, interacts with a sodium transporter to ensure its correct localization on the Golgi membrane (Aridor and Traub, 2002). The fact that CNIH4 upregulation occurred exclusively in clone TSH565 highlights the need to investigate its interaction with transported proteins to better understand its role in this clone´s response to DS.

Aquaporins (AQPs) are a family of transmembrane proteins that play a critical role in transporting water and other substrates, making them essential for plant stress responses. These proteins facilitate the bidirectional flow of water and solutes across cell membranes (Afzal et al., 2016; Sun et al., 2024). Among them, TIPs (Tonoplast Intrinsic Proteins) and PIPs (Plasma Membrane Intrinsic Proteins) primarily transport water by forming channels in the vacuole and plasma membranes regulating intracellular water flow and consequently cellular turgor (Małgorzata Kurowska, 2020). Interestingly, both clones showed AQP gene downregulation under DS, while in TSH565, the TIP3–2 gene (Tc03v2_g005540) was strongly upregulated (log₂FC = 3.67) (Supplementary Tables 2 and 5). Interestingly, EET8 exhibited a greater reduction in RWC and ψ_leaf, whereas TSH565, showed smaller reduction in these parameters along with an increase in WUE_i. While AQP downregulation is a mechanism to reduce water loss during drought, their expression patterns can be highly variable (Małgorzata Kurowska, 2020). In some cases, AQPs are differentially regulated to redistribute water directly to critical tissues while limiting transport elsewhere (Merlaen et al., 2019). Overexpression of AQP genes in other species, such as tomato, resulted in high drought tolerance by improving the regulation of transpiration under stress (Sade et al., 2009).

These results suggest that differences in the regulation of AQP gene expression may underline the contrasting abilities of these clones to maintain cellular turgor and water flow within cells and tissues under DS.

Metabolites play a central role in plant drought responses, as genes involve in their biosynthesis are frequently upregulated under stress and correlate with their accumulation (Kumar et al., 2021; Li et al., 2024). In this study, several metabolic genes were highly upregulated in both clones under DS (Supplementary Table 1). Among them, GoIS, which catalyzes the first step of raffinose family oligosaccharide (RFO) biosynthesis, promotes the accumulation of galactinol and raffinose- osmoprotectants, that stabilize membranes, scavenge ROS, and protect the photosynthetic apparatus. Overexpression of GoIS2 has been linked with enhanced drought tolerance through increased endogenous levels of these metabolites (Taji et al., 2002; Xiao et al., 2023).

Similarly, P5CS, a key enzyme in proline biosynthesis and accumulation was induced in both TSH565 and EET8 (Kumar et al., 2021)(Supplementary Table 2). Proline contributes to osmotic adjustment, ROS scavenging, membrane integrity, and protein stabilization, thereby mitigating dehydration-induced damage (Ghosh et al., 2022). In T. cacao, proline accumulation has been linked to drought-tolerant genotypes displaying higher RWC and reduced stomatal opening under DS (Bae et al., 2009; Zakariyya et al., 2017; Dzandu et al., 2021). The expression profiles observed here suggest that both clones employ proline-mediated osmotic adjustment as part of their drought response.

Genes involved in the synthesis of additional compatible osmolytes, such as glycine-betaine (GB) and trehalose were also upregulated. These compounds protect thylakoid and plasma membrane, and help maintain photosynthetic efficiency under stress conditions (Basit et al., 2025) (Supplementary Table 2). Notably, TSH565 exhibited minimal alterations in photosynthetic machinery under DS, as reflected by stable F_v/F_m values, smaller reduction in the ψ_leaf and a higher WUE_i, whereas EET8 showed stronger reduction in these parameters. Although osmotic adjustment appears to be a common response in T. cacao, its activation in TSH565 may contribute more effectively to protecting the photosynthetic machinery from dehydration-induced damage.

Despite reductions in the ψ_leaf and F_v/F_m, the EET8 plants were able to withstand DS, potentially due to the strong upregulation of TPL (Thaumatin-like protein; Log₂ FC = 9.2) (Supplementary Table 1). Thaumatin - like proteins, also referred to as osmotin-like proteins, are associated with tolerance to biotic, osmotic and drought stress and accumulate in plants adapted to dry environments (de Jesús-Pires et al., 2020; Faillace et al., 2021). Their overexpression in crops like sesame and tobacco enhances drought and salinity tolerance, improves survival under DS, and contributes to pathogen resistance through physiological, biochemical, and molecular adjustments (Rajam et al., 2007; Chowdhury et al., 2017). The marked induction of TPL in EET8 suggests a compensatory role in mitigating reductions in ψ_leaf, highlighting its potential as a candidate gene for improving crop tolerance through breeding strategies (Manghwar and Hussain, 2021).

Cell wall remodeling also appears to contribute to drought response. Expansins, which regulate cell growth and expansion under abiotic stress were induced in both clones (Krishnamurthy et al., 2019; Li et al., 2011) (Supplementary Table 1). In particular, EXLB1, previously associated with drought responses in Brassica rapa, was highly upregulated, suggesting its involvement in modulating cell wall dynamics in T. cacao under DS (Krishnamurthy et al., 2019).

In parallel, drought stress triggered a robust activation of antioxidant system, which is essential for limiting ROS-induced damage to cellular components and maintaining redox homeostasis (Noctor et al., 2014; Jain et al., 2018). Both clones, showed strong upregulation of genes such as MIOX, COMT, and PPOs (Supplementary Table 1). MIOX contributes to ascorbate biosynthesis and in rice has been linked to enhanced drought tolerance through improved ROS scavenging and proline accumulation (Duan et al., 2012; Shi et al., 2017). COMT, involved in melatonin biosynthesis, regulates stress responses, and its overexpression has been shown to enhance drought tolerance and recovery in Arabidopsis (Yang et al., 2019), while exogenous melatonin in Chinese walnut improved growth recovery, photosynthetic efficiency, ROS elimination, and proline accumulation (Sharma et al., 2020). Increased PPO activity under DS has also been reported in T. cacao, contributing to H₂O₂ detoxification as part of an adaptive stress mechanism (Santos et al., 2014).

Additional antioxidant related components exhibit clone-specific patterns. Peroxidase genes, along with glutathione S-transferases and aldehyde dehydrogenase, were predominantly induced in EET8, indicating a broad and coordinated oxidative stress response. In contrast, LEA proteins and sHSPs showed stronger upregulation in TSH565. These proteins enhance drought tolerance by stabilizing membranes and proteins, preventing aggregation, and facilitating protein refolding, suggesting complementary antioxidant strategies between clones (Wang et al., 2004; Ul Haq et al., 2019) (Figure 6; Supplementary Table 2).

Finally, KEGG pathway analysis revealed upregulation of genes involved in phenylpropanoid biosynthesis in both clones under DS, highlighting the importance of non-enzymatic antioxidant defenses. Activation of this pathway reported previously in T. cacao promotes the accumulation of phenolic compounds, which play critical roles in plant acclimation and adaptation (Osorio Zambrano et al., 2021). Due to their antioxidant activity these compounds scavenge ROS, mitigating protein oxidation and lipid peroxidation in cell membranes, avoiding the oxidative stress during drought stress (Sah et al., 2016; Ye et al., 2017). Together, these results underscore the coordinated contribution of enzymatic and nonenzymatic antioxidant systems in regulating intracellular ROS and sustaining cellular homeostasis during DS (Passardi et al., 2005; Mishra et al., 2023).

In accordance with observed results at physiological level, DS induced changes in the expression of genes related to photosynthetic machinery. In TSH565, PsbP which encodes extrinsic PSII subunit involved in the normal oxidation of water molecules (Che et al., 2020) were upregulated (Supplementary Table 2). Studies in tobacco and Arabidopsis confirmed that PsbP is essential for PSII assembly, stabilization and photoautotrophy; loss-of-function mutants exhibit reduced growth, pale leaves and markedly lower F_v/F_m (0.47) (Ifuku et al., 2005; Yi et al., 2007). PsbP also contributes to stress tolerance under water deficit by regulating stomatal movement (Hong et al., 2020). The upregulation of these genes in TSH565 likely stabilized PSII during DS, and, together with GB-mediated osmoprotection, maintained photosynthetic function, suggesting that the lower reduction in A observed for this clone, was due to stomatal limitation rather than disruption of the photosynthetic machinery.

Conversely, in EET8, photosynthesis-related genes were downregulated (Supplementary Table 2), including PSBW, a core PSII component essential for structural stabilization and photosynthesis (Bishop et al., 2003; Soldatos et al., 2015). Loss of PSBW impairs PSII-LHCII complex formation, reducing energy transfer efficiency between PSII units and decreasing complex activity under excess light (García-Cerdán et al., 2011). In contrast, this gene was upregulated in drought-tolerant pearl millet clones submitted to drought (Dudhate et al., 2018).

Other photosynthetic genes were downregulated in EET8, (LHCs; PsaF; PsaH; Psak) that encodes PSI subunits, electron transport, and PSII subunits (Supplementary Table 2) indicating strongly repressed photosynthetic activity in this clone under DS. Proteomic and Transcriptomic studies in maize suggest that reduced LHC abundance under drought can serve as a protective strategy against photooxidative damage leading to reduced gas exchange and chlorophyll fluorescence (Dalal and Tripathy, 2018; Zhang et al., 2018). Therefore, the downregulation of these genes observed in EET8 may reflect a protective mechanism against the oxidative damage of its photosynthetic machinery.