Molecular study of drought response in the Mediterranean conifer Pinus pinaster Ait.: Differential transcriptomic profiling reveals constitutive water deficit‐independent drought tolerance mechanisms

Abstract Adaptation of long‐living forest trees to respond to environmental changes is essential to secure their performance under adverse conditions. Water deficit is one of the most significant stress factors determining tree growth and survival. Maritime pine (Pinus pinaster Ait.), the main source of softwood in southwestern Europe, is subjected to recurrent drought periods which, according to climate change predictions for the years to come, will progressively increase in the Mediterranean region. The mechanisms regulating pine adaptive responses to environment are still largely unknown. The aim of this work was to go a step further in understanding the molecular mechanisms underlying maritime pine response to water stress and drought tolerance at the whole plant level. A global transcriptomic profiling of roots, stems, and needles was conducted to analyze the performance of siblings showing contrasted responses to water deficit from an ad hoc designed full‐sib family. Although P. pinaster is considered a recalcitrant species for vegetative propagation in adult phase, the analysis was conducted using vegetatively propagated trees exposed to two treatments: well‐watered and moderate water stress. The comparative analyses led us to identify organ‐specific genes, constitutively expressed as well as differentially expressed when comparing control versus water stress conditions, in drought‐sensitive and drought‐tolerant genotypes. Different response strategies can point out, with tolerant individuals being pre‐adapted for coping with drought by constitutively expressing stress‐related genes that are detected only in latter stages on sensitive individuals subjected to drought.

, as well as in drought adaptation (see also Eveno et al., 2008;Grivet et al., 2011). Molecular studies developed during the last decades have provided some additional insights.
Thus, the use of cDNA-AFLPs  and microarrays (Perdiguero, Barbero, Cervera, Collada, & Soto, 2013;Perdiguero et al., 2012) has led to the identification of a few hundred candidate genes related to water use efficiency, including genes encoding dehydrins and genes related to cuticular wax biosynthesis. These studies found that, in general, although the variation in response to water deficit was conserved in aerial organs and roots, some genes were expressed in a time-dependent manner in specific organs (Plomion et al., 2016).
The current study is focused on the genetic analysis of maritime pine drought response at the whole plant level. With this aim, we developed a transcriptomic analysis of roots, stems, and needles of siblings selected from a full-sib cross between progenitors according to their contrasting response to drought (drought-tolerant versus drought-sensitive genotypes). Pinus pinaster is considered a recalcitrant species for vegetative propagation in adult phase (Greenwood, 1987), and therefore, most of the studies have been carried out analyzing single trees. The analysis was, however, designed using ramets (three clonal replicates of each genotype) from F 1 progeny individuals to study general mechanisms associated with sensitive or tolerant responses in the family. We identified organ-specific genes, genes differentially expressed in sensitive and tolerant genotypes in response to drought as well as drought-related genes, expressed in control conditions, that may be involved in the mechanisms determining pine tolerance to hydric stress.

| Plant material and experimental design
Two drought-sensitive (4, 147) and two drought-tolerant (132, 144) genotypes were selected from Gal1056 x Oria6, a reference fullsib family designed to study drought tolerance in maritime pine (de Miguel et al., 2012). These genotypes were vegetatively propagated as previously described in de Miguel et al. (2012). For water stress treatment, 24 2-year-old ramets (six clonal replicates of each genotype) were grown in a growth chamber under controlled conditions (watering to full capacity, 70% relative humidity, 20-25°C, 16/8 photoperiod, 13h of maximum light intensity at 800 μmol photons m −2 s −1 ) during an establishment phase of two months. Then, half of the ramets were subjected to moderate water stress. Well-watered plants (WW plants) were kept at a VWCs (soil volumetric water content) higher than 20 vol.%, while in plants submitted to water deficit (WD plants) soil water content was depleted down to 5 vol.% in 19 days and kept at this stage for 43 additional days (Sánchez-Gómez, Mancha, Cervera, & Aranda, 2017). At the end of drought period, roots, stems, and needles were individually harvested from each ramet and stored at −80°C.

| RNA extraction, quality determination, cDNA library preparation, and sequencing
For RNA extraction, each frozen tissue sample was individually powdered in a CryoMill (Retsch) using liquid nitrogen. A total of 12 pools were made bulking equal amount of each grinded root, stem, or needle tissue from WW ramets (pools 1-3) and WD ramets (pools 4-6) of the sensitive genotypes and from WW ramets (pools 7-9) and WD ramets (pools 10-12) of the tolerant genotypes ( Figure 1a). Total RNAs were extracted from needles and stems using PureLink Plant RNA Reagent Kit (Invitrogen). Extraction of total RNA from roots was performed according to Chang, Puryear, and Cairney (1993).
A total of 100 μg RNA was digested with RQ1 RNase-free DNaseI (Promega) during 30 min at 37°C and purified using the Amicon Ultra 0.5 ml (Millipore Corporation). RNA concentration and purity were estimated determining the spectrophotometric absorbance of the samples and the A260/A280 nm and A260/A230 nm ratios using at 230, 260, and 280 nm a spectrophotometer (Thermo Scientific; NanoDrop 1000) and integrity analyzed by agarose gel electrophoresis before and after DNaseI digestion. Finally, approximately 10 μg of each total RNA was further cleaned and concentrated using RNA MinElute CleanUp Kit (Qiagen) and RNA concentration was determined using the 2100 Bioanalyzer and the RNA 6000 Nano Kit (Agilent). RNA samples with an RNA integrity number (RIN) higher than 7 were used for library preparation. Additionally, using Plant/ Fungi Total RNA Purification Kit (Norgen Biotek Corp.), a total of 72 RNAs were extracted from all samples, representing each of the three organs of the three WW ramets and three WD ramets of each genotype (four genotypes × two treatments × three organs × three biological replicates), to be used as template for quantitative realtime PCR (qRT-PCR).
Twelve cDNA libraries were constructed (Figure 1a)

| Sequence processing and improvement of reference transcriptome
The quality of reads was checked during sequence processing with FastQC software (Andrews, 2010). Reads were trimmed and filtered removing reads with a quality score lower than 20, with similarity with common microorganisms or plastids, or low complexity reads, using SeqTrimNext software (Falgueras et al., 2010). When processed reads in a library showed overrepresented sequences or high proportion of Kmer content according to FastQC, a second trimming step rejecting 50 pb at the end of the sequences was performed using the FastX toolkit (http://hanno nlab.cshl.edu/ fastx_toolkit). CLC Genomics Workbench v.10 software (CLC Bio) was used for the successive steps ( Figure 1b). In order to obtain an improved P. pinaster transcriptome, including specific transcripts from sequenced libraries which could be of interest in this analysis, reads were mapped against all the transcripts contained in the reference P. pinaster transcriptome (http://www.proco gen.eu) available in plaza website (Proost et al., 2009). Unmapped reads were recovered and de novo assembled to identify putative transcripts higher than 200 bp, which were added to transcriptome for the successive steps.

| Functional annotation of the generated reference transcriptome
The new reference transcriptome was re-annotated to increase the overall total coding length and update function assignment to the new coding regions identified. All transcripts were used as query for  (Conesa et al., 2005) in order to infer gene ontology (GO) terms from the different GO ontologies (biological process, molecular function, and cellular component) F I G U R E 1 Experimental design and data analysis. (a) Experimental design. Twelve cDNA libraries were constructed and sequenced using as template RNAs extracted from different organs (roots, stems, needles) from ramets of two sensitive genotypes (4 and 147) and two tolerant genotypes (132 and 144). These genotypes were vegetatively propagated and grown under well water (WW, pools 1, 2, 3, 7, 8, and 9) and water deficit conditions (WD,pools 4,5,6,10,11,and 12). (b) Pipeline used for data analysis. Schematic overview of the pipeline used to process raw sequence files to identify constitutive and differentially accumulated transcripts. The software and files used or generated in each step are indicated

| Mapping and differential gene expression analysis
Reads from each library were mapped to previously generated transcriptome using default scoring values using CLC Genomics Workbench v.10 (CLC Bio), ignoring the reads with multiple mapping positions. Count data matrix was normalized using quartile normalization in order to avoid differences in deep sequencing between libraries.
Once sequencing results were validated, a pairwise comparison analysis was performed pursuing two main objectives. On one hand, the identification of genes differentially expressed between WW plant and WD plants at the organ level, both in sensitive (1vs4, 2vs5, and 3vs6) and tolerant genotypes (7vs10, 8vs11, and 9vs12; see Figure 1a). On the other hand, the identification of constitutively

| Single enrichment and gene set enrichment analysis
Blast2GO results, including GO term annotation and KEGG pathway information, were used for functional analysis of differentially expressed genes. GO term enrichment was highlighted for every analysis (drought stress response as well as constitutive accumulation) by comparison between functionalities identified in genes significantly upregulated and downregulated. GO term enrichment was analyzed by using FatiGO (Al-Shahrour et al., 2007) implemented in the OmicsBox software. Significant enrichment of GO terms was considered for p-values <.05. Venn diagrams were drawn using Venny 2.1 (Oliveros, 2015). KEGG pathway enrichment analysis was carried out using GSEA software (Subramanian et al., 2005). All genes preranked according to Kal's statistics were included in order to analyze every comparison as a global system. Analyses were run with 1,000 permutations of gene sets. Pathways with p-values <.05 were considered positively or negatively correlated.

| Validation by quantitative real-time PCR (qRT-PCR)
In order to validate the transcriptomic study, expression analysis of a set of five DEGs was analyzed by real-time qRT-PCR. Gene specific primers were designed using the NCBI Primer-Blast Tool (http:// www.ncbi.nlm.nih.gov/tools/ primer-blast/); sequences and transcript IDs are listed in Figure 2a.

| Differential response of Pinus pinaster genotypes to water stress
Phenotypic characterization of these plants has been described by Sánchez-Gómez et al. (2017). Physiological analyses indicate that water deficit had a general negative impact on leaf photosynthetic performance and osmotic potential. Yet, the four studied genotypes displayed two contrasting physiological sensitivities to water deficit.
While the genotypes 132 and 144 performed as drought-tolerant

| Mapping and identification of differentially expressed genes
Filtered reads from each library were independently mapped to the PpDR transcriptome. A total of 73,246 transcripts with at least one read in any of the cDNA libraries were detected. Roots showed the highest number of transcripts (>25,000), followed by stems and needles (Table 1).
To preliminary validate the quality of sequencing data, a differential expression analysis was carried out using three data sets generated grouping all data by organ. The results clearly highlight organ-specific expression patterns. The use of a highly restrictive threshold (FDR p <.001) led to identify a reduced group of transcripts with strong organ specificity. Needles showed the largest group of specific genes (99), followed by roots (25) and stems (8). Single enrichment analysis based on the needle-specific genes showed a significant enrichment in GO terms related to biosynthetic and lipid metabolic processes, photosynthesis, and generation of precursor metabolites and energy, all of them functionalities associated with this organ, which support the quality of sequencing. Once sequencing was validated with a preliminary differential organ-specific expression analysis, pairwise comparison was performed to analyze differential expression analysis in response to water stress. Thus, a total of 6,215 DEGs were iden- and 64% in needles of sensitive and tolerant plants, respectively), roots of WD-tolerant plants showed a significant higher percentage tolerant-specific genes than WD-sensitive plants (71% and 49%, respectively). It is important to highlight that in all organs but stems, sensitive-specific genes showed broader functional diversity than tolerant-specific genes.
Comparison between organs also allowed the identification of DEGs shared between genotypes that showed similar trends (upregulation, downregulation, or nonsignificant variation in response to water stress; Table 2). Genes specifically upregulated or downregulated in a single organ were the most common trend observed for both sensitive and tolerant genotypes. Only two transcripts encoding a RING-H2 finger protein (isotig37224) and a NAC transcription factor (unigene10311) were upregulated, while one transcript encoding a high-affinity nitrate transporter (unigene30176) was downregulated in all organs of WD-sensitive genotypes. When analyzing the organs of WD-tolerant genotypes, we found no DEGs sharing the same expression trend. Two genes, a suppressor protein SRp40 (isotig43193) and a D-tyrosyl-tRNA deacylase (unigene27014), were highly upregulated in all organs of WW-tolerant genotypes.

| Functional analysis of genes differentially expressed in roots, stems, and needles in response to drought
Pairwise comparisons from sensitive and tolerant genotypes at the organ level allowed the identification of DEGs associated with water stress response. Sensitive genotypes showed higher number of     It is important to highlight that six upregulated genes in roots of WW-tolerant genotypes (Table 2) were also upregulated in roots from both types of genotypes in response to water stress: a LHY protein isoform X1 (unigene28840); a chaperone regulator 6 isoform X2 (unigene796); a putative calcium-binding protein CML25 (iso-tig33124); two transcripts (isotig42238 and isotig75236) showing homology with a transposon type-TNT 1-94; and finally, a P. taeda protein-coding gene of unknown function (isotig31114).

| Gene expression analysis by quantitative realtime PCR
Expression analysis of five DEGs was analyzed on three ramets from each of the four genotypes by qRT-PCR, in order to validate this study. Alpha-dioxygenase 1 (ALPHA-DOX1) and disease resistance protein at5g63020 (DRP) were analyzed in roots, and

Drought-related DEGs in sensitive genotypes Drought-related DEGs in tolerant genotypes
Drought-related genes constitutively expressed in sensitive genotypes

STEMS
showed results in agreement with earlier transcriptomic analysis ( Figure 2b).

| D ISCUSS I ON
This study provides new information about molecular strategies underlying differential response to a moderate water deficit intensity of roots, stems, and needles from four P. pinaster F 1 progenies with contrasting responses to drought (de Miguel et al., 2012(de Miguel et al., , 2014. Since plant response to drought stress is regulated by intensity, duration, and rate of progression of imposed drought ( Comparison between sensitive and tolerant genotypes revealed that the latter showed higher levels of transcriptional activity in the three organs but in WW needles (Table 1). However, the number of DEGs associated with water stress response was significantly higher in all organs from sensitive genotypes. Although the lowest number of DEGs was detected in the needles of both genotypes, they were significantly enriched in GO terms related to "response to stress," as expected. These results may point to a basal activation of stress-responsive mechanisms in tolerant genotypes that allow them to rapidly face frequent droughts. Other studies have also described sensitive genotypes that exhibit hyper-response to drought stress compared to tolerant genotypes (Janiak et al., 2018;Muthusamy , Uma, Backiyarani, Saraswathi, & Chandrasekar, 2016;Pucholt, Sjödin, Weih, Rönnberg-Wästljung, & Berlin, 2015;Yates et al., 2014 ;You et al., 2019), indicating absence of some stress avoidance mechanisms in these sensitive genotypes that attenuate drought effects.
Functional analysis of genes exclusively expressed in sensitive and tolerant plants also supported this hypothesis, which highlights the differences with higher number of genes involved in primary metabolic process, nitrogen compound metabolic process, and different GO terms associated with response to stress mainly in roots and needles of tolerant plants while higher functional diversity identified in sensitive genotypes. The presence of specific cell signaling processes in stems from tolerant genotypes, which were not present in the sensitive ones, may indicate an induction of organ-specific transfer of information, which may be involved in response of tolerant genotypes to cope more effectively with drought fluctuations and its effects.
The analysis of specific pathways revealed interesting differential behavior between sensitive and tolerant genotypes. Thus, different pathways related to flavone and flavonol biosynthesis or carotenoid biosynthesis showed significant upregulated genes in roots of tolerant versus sensitive genotypes in response to water stress. Flavonoids play different molecular functions in stress protection, including inhibition of polar auxin transport, that interferes with hormone signaling (Dao, Linthorst, & Verpoorte, 2011), and antioxidant defense (Gill & Tuteja, 2010). Flavonoids accumulate at the site of lateral root formation in drought-stressed Arabidopsis plants, which could suggest a positive effect on lateral root formation (Shojaie , Mostajeran, & Ghannadian, 2016) also found in poplar (Dash et al., 2017).
Only six genes showing higher expression levels in WW-tolerant than WW-sensitive plants, which were also upregulated in WDtolerant and WD-sensitive plants, were detected in roots (Table 3).
Among them, transcripts promote abiotic stress tolerance such as chaperone, protecting structures (Park & Seo, 2015), and a dormancy auxin family-associated protein (hormone involved in protein modification, signal transduction, and in drought tolerance (Peleg & Blumwald, 2011)). In addition, a late elongated hypocotyl protein (LHY), a MYB-type transcription factor commonly associated with circadian clock control (Sanchez,Shin, & Davis, 2011), was also identified. In Arabidopsis, mutations in this gene produced hypersensitivity to ROS-generating agents, which indicates a function during detoxification processes (Park, Kwon, Gil, & Park, 2016). Finally, a putative calcium-binding protein CML25, a Ca 2+ sensor involved in regulating plant responses to abiotic stresses (Zeng et al., 2015), was also detected. This set of transcripts, involved in protection, signaling, and regulation of gene expression, could also contribute to maintain tolerant genotypes in a constant alert state.
Pre-adaptation of tolerant genotypes was also observed when analyzing genes involved in drought response out of the top-30 highest constitutively upregulated genes, with genes mainly acting in: structural protection, such as osmoprotectants, aquaporins, chaperones, late embryogenesis abundant and heat shock proteins; detoxification of reactive oxygen species (ROS); and cell wall metabolism and protection of subcellular structures (Le Gall et al., 2015) ( Figure 4) Tenhaken, 2014). Considering that lipids are the major cell membrane components and a source for signaling molecules, upregulation of sec14 cytosolic factor family protein, a phosphatidylinositol/phosphatidylcholine transfer protein located in the Golgi membrane, that functions as signal precursor inducing stress-responsive genes, phospholipids, and galactolipids (Liu et al., 2013), may promote membrane stability conferring stress tolerance (Larsson, Nyström, & Liljenberg, 2006). Additionally, ABAregulated genes and a SCF E3 ligase (phloem protein 2-b11), recently described as a negative regulator of drought response involved in ABA-independent signaling pathway in Arabidopsis (Li et al., 2014), were observed to be constitutive and highly overrepresented.
In stems, only three out of the 30 highest constitutively expressed genes were associated with drought response in sensitive plants: a cysteine proteinase inhibitor, involved in regulation of proteolysis (Kidrič, Kos, & Sabotič, 2014); and two genes involved in signaling, a zinc finger A20 and AN1 domain-containing stress-associated protein 8-like (whose overexpression in rice reduced stress-induced injuries such as chlorosis and cell death, improving recovery from stress (Vij & Tyagi, 2006)), and a CBL-CIPK involved in decoding Ca 2+ signals from calcineurin B-like proteins. Evidence also indicates that a high number of CIPKs participate in various stress responses as well as in other ABA responses (Mao et al., 2016). As in roots, the number of constitutively expressed genes associated with drought response also increases up to 12 in stems of tolerant plants, with genes involved in protection, such as beta-galactosidase 8 transcript, which may be associated with break of cell wall polysaccharides to direct sugars to cytoplasm to maintain of cell turgor under water loss (Gupta, Rai, Gayali, Chakraborty, & Chakraborty, 2016).
Upregulation of components of the Ca 2+ -permeable channel and the sugar phosphate/phosphate translocator (Jarzyniak & Jasiński, 2014) could be associated with the enhancement of the transport of these compounds across membranes. Cell wall reorganization (endoglucanase 17 and UDP-N-acetylglucosaminedolichyl-phosphate N-acetylglucosamine-phosphotransferase genes) and adjustment of membrane lipids (alpha/beta hydrolase family protein and chloroplastic lipoxygenase) were also upregulated, considering that the two latter genes are also involved in signaling (Bae et al., 2016;Hamiaux et al., 2012). Upregulation of bifunctional 3-dehydroquinate dehydratase/shikimate chloroplastic, involved in biosynthesis of aromatic amino acids and different aromatic secondary metabolites (alkaloids, flavonoids, lignins, and aromatic antibiotics) with important roles in plant stress response, was also detected.
Changes in leaf water content occur together with the inhibition of stem expansion. Auxin influences stem elongation and regulates the formation of the plant shoot architecture (Gallavotti, 2013). Highly upregulated genes involved in auxin regulation were identified, such as auxin response factor 6, a transcriptional activator that regulates the expression of auxin response genes (Liu et al., 2014), and an IAAamino acid hydrolase ILR1-like 4, that hydrolyzes IAA-amino acid conjugates to produce free IAA (Ludwig-Müller, 2011). Among hormone-regulated genes, upregulation of a component of SCF (COI1) coreceptor in plants stands out. It has been reported to be involved in JA perception and signal transduction (Larrieu & Vernoux, 2016).
However, the most significant difference was found in needles, where none of the 30 highest constitutively expressed genes were associated with drought response in sensitive plants, versus 15 genes in tolerant genotypes clustered into different relevant functional groups (Figure 4), among them, genes with protective functions: preservation of the correct folding of RNA molecules and proteins during stress (i.e., chaperone, heat shock cognate 70 kDa), osmolyte biosynthesis (an alpha-galactosidase-like protein), and osmotic regulation (i.e., an aquaporin, a YGGT family protein and a two-pore calcium channel protein 1, also constitutively accumulated in stems of tolerant genotypes). Protective effect of chloroplast-targeted chaperone protein DnaJ on photosystem II has been previously described in transgenic plants subjected to stress, in which chloroplast heat shock protein 70 was also identified as a partner of this chaperone (Kong et al., 2014). In this study, a plasma membrane intrinsic protein (PIP) gene was also constitutively expressed. This subfamily of aquaporins is involved in water transport and, depending on the member family, small solutes transport, playing an active role in drought stress response (Forrest & Bhave, 2007), regulating biotic stress responses, and involved in regulating root water uptake and transpiration rates (Afzal, Howton, Sun, & Mukhtar, 2016 concentration, regulating, for example, stomatal aperture in guard cells (Song et al., 2008) and also highlight the upregulation of a group of genes involved in cell wall remodeling (a xyloglucan endotransglucosylase hydrolase and a polygalacturonase noncatalytic subunit 2-like) as well as a group that included drought-inducible genes, such as RD22 and a member of NHL family, which are mediated by ABA.
All these results also support that tolerant genotypes exhibit permanent activation of mechanisms for cell protection and overexpression of stress pathways that pre-adapt them to respond more efficiently and rapidly to water stress. which is regulated by ABA). Stomatal closure associated with drought results in changes of rates of photosynthesis due to the decreased CO 2 availability and production of reactive oxygen species (ROS), such as superoxide radicals (Osakabe et al., 2014). Increased photorespiratory activity during drought is also accompanied by elevated levels of glycolate oxidase activity, resulting in H 2 O 2 production. In this study, seven out of the 23 upregulated genes were associated with enzymes that detoxify active oxygen species, such as catalase, glutathione S-transferase omega-like 2, glutamine amidotransferase YLR126C, phospholipid hydroperoxide glutathione peroxidase, and gamma-glutamylcyclotransferase At3g02910, a member of the peptide methionine sulfoxide reductase, and a blue copper protein ( Figure 4). Also, three genes with photosynthetic function, two of them related to the Calvin and Benson cycle (ribulose bisphosphate carboxylase small chain chloroplastic, fructose-bisphosphate aldolase chloroplastic) and a chloroplast protein that acts as an auxiliary component of the photosystem II (oxygen-evolving enhancer protein chloroplastic-like), were highly upregulated in needles of sensitive plants. Additionally, three genes were associated with signaling: G-type lectin S-receptor-like serine threonine-protein kinase At2g19130, sporulation protein RMD1, and universal stress protein a-like protein. Genes encoding proteins with USP domain are useful in stress signal perception and seem to promote drought tolerance (Sinha et al., 2016). Finally, a component of a multiprotein complex transcription factor (mediator-associated protein 1-like) and genes involved in polyamine metabolism (ornithine decarboxylase-like) and plant growth regulation were also highly upregulated. In the case of the needles of tolerant plants subjected to stress, the five highly upregulated DEGs were mainly involved in cell protection: a chaperone protein DnaJ chloroplastic-like and a Hsp70 nucleotide exchange factor FES1, both protecting folding of molecules under stress; an anthranilate chloroplastic-like involved in osmolyte biosynthesis that contributed to osmotic adjustment and thereby enhanced drought stress tolerance in plants (Liu, Shen, & Huang, 2015); and a protein YLS9-like involved in oxidative response. The remaining fifth gene was a transcription factor C2H2-type zinc finger protein, which is involved in regulation of drought response (Kiełbowicz-Matuk, 2012).

| CON CLUS IONS
One of the most interesting outcomes of this study was the finding that drought-tolerant genotypes expressed a high number of genes related to stress even before the water deficit took place as opposed with the drought-sensitive genotypes. This finding indicates that constitutive expression of drought-related genes, specifically hormone-regulated genes, genes involved in signaling pathways, as well as those involved in stress protection, can provide functional advantages to cope with an eventual water deficit. Interestingly, a significant number of genes related to the Calvin and Benson cycle and regulation of photosystem II were highly expressed in the sensitive genotypes but not in the tolerant genotypes when subjected to water deficit conditions. Regardless

CO N FLI C T O F I NTE R E S T
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.