Rice (Oryza sativa L. cv. “Saechucheongbyeo”) seeds were imbibed and sterilized with 5 L of distilled water containing 2.5 mL of a seed sterilizer (Kimaen, Farm Hannong, Korea) for 24h at room temperature and germinated in an incubator (darkness, 28°C) for 5 days. The uniformly grown seedlings (10 plants per treatment) were carefully transplanted into a hydroponic system containing 1/2 strength (0.5×) of Hoagland nutrient solution (Hoagland and Arnon, 1950) and grown until reaching at fifth leaf stage (3 weeks) with replacing a liquid media every week. The growth condition of a growth chamber (VS-91G09M-2600, VISION SCIENTIFIC Co. Ltd., Korea) was fixed with a 14/10h photoperiod, 60% of relative humidity (RH) and 27/27°C (day/night) temperature.

To develop the gradient of drought stress, water potential (Ψs) was firstly adjusted with 0 (0% of PEG-6000), −1.0 (8%), −1.5 (10%), and −3.1 (20%) MPa using polyethylene glycol (PEG)-6000 with 1/2 strength (0.5×) of Hoagland nutrient solution. Rice seedlings were exposed for 7 days with different strength of drought, and the shoot and root were carefully harvested and stored at −80°C until further analysis (six to eight plants per treatment).

The relative water content (RWC) was calculated as the following formula. RWC (%) = (fresh weight − dry weight)/(turgid weight − dry weight) × 100. The fresh weight of fully developed leaves of each plant was measured, and the turgid weight was measured from leaf disc (1 × 1 cm) immersed in distilled water at 4°C for 3h to minimize respiratory loss. The dry weight was determined after dryness for 48h at 80°C.

The proline was analyzed with a previous method (Carillo and Gibon, 2011). The fresh samples (0.5 mg) from rice shoots and roots were mixed with 1 mL 40% (w/v) ethanol at 4°C for 24h, centrifuged at 14000g for 5 min, and the supernatant was extracted. The supernatant (500 µL) was immediately incorporated into a reaction solution containing 3% sulfosalicylic acid, 1 mL of glacial acetic acid and 1 mL of acidic ninhydrin, incubated at 96°C for 1h, and the reaction was terminated on ice. Toluene (1 mL) was added to the reaction mixture, and the supernatant was used as an analytical sample for measuring proline using a spectrophotometer (520 nm) (UV-1900i, Shimadzu, Japan) after vortexing for 20 s and standing for 5 min. The L-proline was used as a standard (Sigma, MO, USA). The proline content was calculated using a method by Augusto et al. (Ramírez-Godoy et al., 2018): Proline content (µg per g of fresh sample) = [(µg Proline/mL × mL Toluene)/115.5 µg/µmol]/[(g sample)/5].

Chlorophyll contents were analyzed according to the method of Arnon (Arnon, 1949). Fresh leaves (0.5 g) were immersed in 25 mL of 80% (w/v) acetone. Mixtures were vortexed and incubated for 24h at darkness. The absorbance of supernatants was measured at 645 nm and 663 nm using spectrophotometer (UV-1900i, Shimadzu, Japan). The chlorophyll content was calculated as following formula:

Suberin lamellae were microscopically observed from the crown roots of rice seedlings exposed for 7 days under different water potentials (0, −1.0, −1.5, and −3.1 MPa). Root was divided into three developmental zones, division to elongation (0 mm–30 mm from root apex), elongation to maturation (30 mm–50 mm), and maturation (50 mm–100 mm), and quickly fixed with methanol. Suberin lamellae-specific staining are as follows: (1) first staining for 1h in methanol including 0.01% of fluorol yellow 088 (w/v) in darkness with gently agitation, (2) rinsing with methanol and second staining in methanol including 0.5% aniline blue (w/v) for 1h at room temperature in darkness, and (3) rinsing with methanol and mounting on water. Specimens were observed as a green fluorescence excited by ultraviolet (UV) light using a florescence microscope (×10, Eclipse Ti, Nikon, Tokyo, Japan).

The extraction of fatty acids was carried out according to a previous procedure (Park et al., 2021). Briefly, root sample (0.1 g, powdered) was mixed with 0.7 mL of chloroform:methanol solution (2:1, v/v), 0.7 mL of 0.58% sodium chloride, and 0.1 mL of pentadecanoic acid in chloroform as IS (1 mg/mL). The sample was vortexed for 30 s, and centrifuged at 15,000g at 4°C for 5 min. The under layer was transferred into an e-tube (2 mL) and dried using a vacuum concentrator. After dryness, the extract was incubated with 0.1 mL of toluene, 0.18 mL of methanol, and 0.02 mL of 5 M sodium hydroxide at 85°C for 5 min. After cooling for 3 min, the mixture was catalyzed by mixing 0.3 mL of boron trifluoride at 85°C for 5 min, and 0.8 mL of pentane and 0.4 mL of distilled water were blended with the samples. The mixture was centrifuged for 15 min at 350g at 4°C, and the supernatant was transferred to an e-tube (2 mL). The concentrated sample was resolved in 0.1 mL of hexane. The methylated sample (1 μL) was directly injected into the gas chromatography-mass spectrometry (GC-MS, Shimadzu, Japan). For separation of medium and long chain fatty acids, a Rtx-5MS column (30 m × 0.25 mm id, film thickness 0.25 μm, Restek, Bellefonte, PA, USA) was used. The temperature was initiated with 40°C for 2 min, increased to 320°C at a rate of 1°C/min, and finally maintained at 320°C for 6 min. A flow rate of helium as a carrier gas was 1.42 mL/min, and full scan mode (m/z 45–500) was used. The GC-MS software (version 4.11; Shimadzu, Kyoto, Japan) was used. The results were adjusted with their retention times and mass spectra with reference to standard compounds and the in-house library. Quantitative analysis was conducted using the ratio of the analyte peak area to the IS peak area.

Total RNA was extracted from the roots of rice seedlings at 7 days after the initiation of drought, 0, −1.0, −1.5 and −3.1 MPa, respectively, using TRIzol reagent (Invitrogen, Carlsbad, CA) according to the manufacturer’s instructions. The purity and concentration of the extracted RNA were estimated using NanoDrop (Thermo Fisher Scientific, Madison, WI, USA) and checked on a 1.2% agarose gel. Total RNA (1 μg) and the RT PreMix Kit with Oligo (dT) primers were used to synthesize first-strand cDNA using following PCR condition: 60 min at 45°C to cDNA synthesis and 5 min at 95°C to Rtase inactivation step.

Quantitative real-time PCR was performed by using a Real-Time PCR machine (CFX Opus 96, Bio-Rad, Hercules, CA, USA) with technical triplicates according to the manufacturer’s instructions. The reaction mixture consisted of 1 μL of cDNA template, 2 μL each of 10 mM forward and reverse primer ( Supplementary Table S1 ) and 5 μL SYBR Green Q Master mix (Labopass, Cosmo Genetech, Seoul, South Korea). The PCR conditions consisted of pre-denaturation step at 95°C for 5 min, followed by 50 cycles of denaturation 95°C for 10 s, annealing temperature of each primer ( Supplementary Table S1 ) for 30 s and elongation (72°C, 20 s). This step was followed by a melting curve, ranging from 65°C to 95°C at a heating rate of 0.5°C/s. A quantification method (2^–ΔΔCt) was used (Livak and Schmittgen, 2001) and the variation in expression was estimated using triplicate for each cDNA sample. The rice actin gene was used as a reference in the qRT-PCR. Primer sequences used for qRT-PCR were designed by Primer 3 software (Rozen and Skaletsky, 2000).

Statistical analysis was conducted using R statistical software version 4.0.3 and R studio version 1.3.1093. Data are presented as mean and standard deviation. Data were compared using one-way ANOVA and Tukey’s honest significant difference (HSD) test. For the data from qRT-PCR and fatty acid analysis, statistics were performed using SAS software version 9.4 (SAS Institute, Cary, NC, USA), and the significance was tested with Fisher’s least significant difference (LSD) test at the 5% level of probability.

In the current study, we employed four levels of water potential (PEG-6000-derived), 0, −1.0, −1.5, and −3.1 MPa, respectively. The growth and wilting (dehydration) of rice seedlings were significantly affected by the strength of water stress ( Figure 1 ). A visible damage by water stress was marked in the shoot rather than the root ( Figures 1A, B ). Aerenchyma formation was shown at 50-mm distance from the crown root tip ( Figure 1C ). The roots at −1.0, −1.5, and −3.1 MPa were likely to develop more aerenchyma spaces than well-watering condition (0 MPa), although the aerenchyma did not represent the difference between the levels of water stress. The growth (fresh weight based) of rice shoot was significantly decreased at drought condition, whereas root was significantly decreased only at -3.1MPa ( Figures 1D, E ). The limited growth (dry weight-based) of rice seedlings was significant at −3.1 MPa ( Figures 1F, G ), which indicated 50% of growth reduction. Plant total height was remarkably affected in all treatment groups, 7% reduction at −1.0 and −1.5 MPa and 20% at −3.1 MPa ( Figure 1H ). The R/S ratio was a trend of a decrease under water stress although there was no significant difference between water potentials ( Figures 1A–I ).

Proline, an osmoprotectant, was markedly accumulated in both shoot and root by water stresses, and the highest level showed at −3.1 MPa in the shoot and −1.0 MPa in the root ( Figure 2A ). The level of proline was 45–120 µmol g⁻¹ (fresh weight, FW) and 30–40 µmol g⁻¹ in water stressed shoots and roots, respectively. The accumulation of proline in both organs did not depend on the drought strength. The RWC was not differed from all water potentials ( Figure 2B ). Chlorophyll contents were significantly decreased by water stress ( Figures 2C–E ), and the reduction was remarkable in chlorophyll b, indicating 50%–60% of decrease in all water potentials. Chlorophyll a content by water stress was 1.0–1.1 mg g⁻¹ (FW), whereas chlorophyll b showed a range of 0.3–0.5 mg g⁻¹.

Based on current knowledge, we analyzed the transcriptional variations of some key genes, which are directly involved in ABA biosynthesis and signaling, from rice roots under different water potentials ( Figure 3 ). Both water potentials, −1.0 and −1.5 MPa, resulted in the significant upregulation of OsNCED3 (ABA biosynthesis), which indicated 5.0- and 8.0-fold higher water potential, respectively. Moreover, OsABA2 and OsABA3 also showed noticeable increase, 1.3- to 4.0-fold. Meanwhile, OsAAO3 remained unchanged or slightly decreased compared with the control (0 MPa). The OsSAPK10 (encoding SnRK2), ABA-signaling gene, was significantly promoted (6.0- to 8.0-fold, log₂ scale) by limited water supply (−1.0 and −1.5 MPa).

Based on enhanced expression of ABA-biosynthetic and -signaling genes ( Figure 3 ), we tried to understand the inter-communication mechanism between ABA and suberization. Suberin lamella, which is a product by accumulation of suberin in the root endodermis, is known as a tolerant mechanism against osmotic stress (Kreszies et al., 2018). The suberization in the root endodermis was markedly developed by increasing water limitations (−1.0 and −1.5 MPa) compared with well watering (0 MPa) ( Figure 4 ). The −1.0 and −1.5 MPa of water potential showed the thicker layer (green fluorescence) than 0 MPa. In addition, the strength of fluorescence in the most negative water potential (−3.1 MPa) was similar to 0 MPa, which implied less suberization. Therefore, due to that −3.1 MPa did not show significant variations, which were considered as a senescence or death, our observations are mainly described and interpreted with physiological changes in −1.0 and −1.5 MPa. The selected fatty acids ( Figure 5A ) and suberization-associated genes ( Figures 5B, C ) in different water potentials experiencing rice root were analyzed to validate suberization by microscopic observation. The composition of fatty acids was greatly perturbed by decreasing water potentials ( Figure 5A ). The fatty acids (16:0, 18:1, and 18:2) in −1.0 MPa significantly decreased, whereas those in −1.5 MPa was not perturbed. On the other hand, α-linolenic acid (18:3), the highest unsaturated form, was obviously accumulated in both −1.0 and −1.5 MPa compared with the control (0 MPa). Therefore, the ratio of the unsaturated to saturated forms was markedly enhanced by limited water supply conditions. The expression of suberization-involved genes showed a similar tendency with microscopic observation and fatty acid composition ( Figures 5B, C ). The key genes, OsCYP86A1, OsCYP86B1, OsGPAT5, and OsGPAT16, to produce suberin monomer from fatty acids were significantly upregulated (2.0- to 10.0-fold-higher, log₂ scale) in −1.0 and −1.5 MPa of water potential, whereas those in −3.1 MPa remained unchanged. An OsABCG2, transporting suberin monomer into cell wall layer, was also markedly promoted in both water potentials. By contrast, OsESB1, restricting the suberin lamella deposition, was significantly downregulated in both water potentials.

The experience of drought during plant growth noticeably perturbs not only ABA-derived metabolisms including suberization but also AQPs response. However, there is still limited information and/or unclear information on metabolic cooperation between ABA and AQPs. The expression of selected seven genes encoding plasma membrane-intrinsic proteins (3, PIPs) and tonoplast-intrinsic proteins (4, TIPs) from rice roots was compared between different water potentials ( Figure 6 ). The transcriptional changes were the very similar to those in ABA ( Figure 3 ) and suberization ( Figure 5 ). Firstly, the PIPs-encoding genes, OsPIP2;3 and OsPIP2;7, significantly increased, and an abundance showed 2.0- to 3.0-fold greater in OsPIP2;3 and 10.0- to 32.0-fold higher in OsPIP2;7. Many studies reported the enhanced AQPs activity under varying drought conditions. Additionally, three TIPs-encoding genes, OsTIP3;1, OsTIP3;2, and OsTIP4;1, also markedly increased in both drought conditions (−1.0 and −1.5 MPa), which is being ranged from two- to fourfold higher (log₂ scale).

In order to validate our observation, we employed the PCA with variables, and unsaturated fatty acids and all genes except for OsESB1 were loaded on the same group by PC1, which is covering 57.9% of a validation ( Figure 7 ). Therefore, the current results imply that ABA biosynthesis and signaling, suberization, and AQPs activity are directly and/or indirectly associated under the limited water condition and precisely regulated by each other.

In our study, the development of aerenchyma was observed in all drought treatments but could not be quantified ( Figure 1C ). Indeed, drought strongly developed aerenchyma development in roots via a degradation of cortex (Singh et al., 2013; Fonta et al., 2022) and, in context of previous observations, rice roots under drought are likely to operate the drought-tolerant strategy via precise regulation of both apoplastic and symplastic water flux. The height and dry weight of plants were largely influenced (decrease) under the limited water supply (Nasrin et al., 2020). Our study showed that the total plant length and dry weight was significantly reduced in root and shoot of the −3.1MPa ( Figures 1F–H ). However, there was no significant difference in dry weight at −1.0 and −1.5 MPa ( Figures 1F, G ).

Proline and chlorophyll are physiological indicators to determine whether plants are experiencing water stress such as drought (Shi et al., 2017; Gu et al., 2019). Proline, an osmo-protectant, is accumulated (Yoshiba et al., 1997), whereas chlorophyll is decreased in drought stress (Li et al., 2019). The positive relationship between increasing drought conditions and proline accumulation was extensively demonstrated (Sun et al., 2011), and proline revealed greater abundance in shoots compared with roots, even though it was accumulated in both tissues (Pandey et al., 2004). Our study has shown that proline more accumulated in shoot than root, and chlorophyll was decreased in drought conditions ( Figure 2 ). Recently, it was suggested that proline plays a role as an energy adjuster on stress responses including drought, and ABA is closely involved in this mechanism (Alvarez et al., 2022). Furthermore, water scarcity resulted in a decrease in chlorophyll content as a result of chlorophyll degradation due to the destruction of photosynthesis apparatus by ROS generation. In particular, chlorophyll b is more sensitive to stress condition because of the conversion of chlorophyll b to chlorophyll a under drought, and thus a decrease in chlorophyll b is prominent response to drought (Rüdiger, 2002). Our observations showed the same pattern in proline and chlorophyll contents, and, therefore, it is evident that the perturbation of those biochemicals was typical physiological response to the limited water supply.

ABA, which is well-known as a key plant hormone to drought, is directly involved in an activation of a variety of transcription factors responsible for stress condition (Zhang et al., 2006; Singh and Laxmi, 2015). This is because most of the transcription factors activated in drought are responsive to ABA (Singh and Laxmi, 2015). In addition, ABA also triggers the metabolism for root suberization via a cascade of PYL/PYR/RCAR - PP2C - SnRK2 (Wang et al., 2020; Wan et al., 2021; Shiono et al., 2022). This has led to studies of increasing drought tolerance using ABA biosynthesis and signaling enzymes and has been conducted until recently (Lu et al., 2009; Verma et al., 2019). qRT-PCR was performed by selecting six enzyme genes involved in ABA biosynthesis and signaling process ( Figure 3 ). Four genes involved in ABA precursor- and ABA-biosynthesis and two ABA-signaling genes OsPYL6 (Yadav et al., 2020) and OsSAPK10 (Wang et al., 2017) known to be expressed in roots were analyzed. In our study, OsPYL6 was significantly down-regulated in −3.1MPa ( Figure 3 ). This suggest that drought tolerance mechanisms involving ABA are not fully activated in −3.1MPa (permanent wilting point). Whereas, in −1.0 and −1.5 MPa, the expression of ABA biosynthesis genes (OsNCED3, OsABA2, and OsABA3) and OsSAPK10 were upregulated. Among them, OsNCED3 and OsSAPK10 were particularly highly upregulated under mild stress. These results are the same conclusion as the previous study of OsNCED3 (Huang et al., 2018), and it can give more strength to the claim that OsNCED3 is responsible for ABA accumulation and indirectly regulates ABA-dependent genes during water stress. It was demonstrated that OsNCED3 and OsSAPK10 indirectly regulate ABA-dependent cascading pathway in rice roots experiencing drought stress.

Abiotic stresses including drought affect the composition of membrane lipids in plants, and thus an abundance in unsaturated forms increases (Upchurch, 2008). Various levels of water limitation resulted in an accumulation of the unsaturated fatty acids in the roots of grapevine (Zhang et al., 2020) and almond (Elloumi et al., 2014). The enhanced ratio of the unsaturated to saturated fatty acid is clearly observed as a drought-tolerant mechanism in many plants (Moradi et al., 2017; Jin et al., 2022) including rice (Lenka et al., 2011) and maize (Zi et al., 2022). Furthermore, ABA-mediated metabolism predominantly promoted the substances of suberization enhanced the activity of C18:3-ACP (Wang et al., 2022). By contrast, the extreme drought condition resulted in the reduced membrane fluidity causing less function of cell membrane (Rawat et al., 2021).

In this study, genomic and anatomical analyses were performed to compare the development of suberin lamellae according to drought level ( Figures 4 , 5 ). The genes used for qRT-PCR were selected with reference to the following. Enzymes required for biosynthesis of aliphatic suberin precursors include fatty acyl ω-oxidases of CYP family and acyltransferases of the GPAT family (Li-Beisson et al., 2013). AtCYP86A1 and AtCYP86B1 are required for biosynthesis of C16-C18 and C22-C24 ω-oxygenated fatty acid (Li et al., 2007; Molina et al., 2009) and have high expression correlation with AtGPAT5 (Yang et al., 2012). OsCYP86A1 and OsCYP86B1 were selected through the neighbor phylogram of protein sequences from AtCYP86A1 and AtCYP86B1 (Waßmann, 2015). OsGPAT16, which is phylogenetically closest to AtGPAT5, and OsGPAT5, which were expressed in the roots and significantly increased in salt stress, were selected among GPATs of rice (Safder et al., 2021). ATP-binding cassette, subfamily G (ABCG), known as suberin transporter, is co-expressed with suberin biosynthesis genes and is related to root suberization (Yadav et al., 2014). OsABCG2 was selected with reference to its use in suberin-related gene expression analysis in rice (Singh et al., 2021). The suppression of AtESB1 expression was confirmed to increase root suberization (Wang et al., 2019). OsESB1 was predicted to be putatively orthologous in rice from the model plants Arabidopsis thaliana using bioinformatics tools (Kreszies et al., 2018).

Similar to ABA-related gene expression results ( Figure 3 ), suberization-related genes were upregulated at −1.0 and −1.5 MPa ( Figure 5 ). To date, ABA-mediated suberization was broadly demonstrated in many plants, ABA-treated roots of Arabidopsis (Wang et al., 2020), and rice (Shiono et al., 2022) revealed huge accumulation of suberin lamella, which implies the close cooperation between ABA and suberization metabolisms. Hence, the current results are clearly in line with previous observations and provide evidence that the limited water supply leads to the regulatory network responses from ABA-signaling transduction to suberization via the modification in fatty acid composition. This provided evidence that the stress-tolerance mechanism of plants is activated under −1.0 and −1.5 MPa, not in −3.1 MPa. Therefore, our results imply that ABA-dependent metabolism in rice roots under drought condition is likely to be closely connected with suberization by promoting suberin monomer production (Kim et al., 2022).

There is a total of 33 aquaporins in rice, which are classified into four types (Sakurai et al., 2008). Among the four types, plasma membrane intrinsic proteins (PIPs) and tonoplast instrinsic proteins (TIPs) exist in the plasma membrane and vacuole membrane, respectively. The PIP2 isoforms, regulating water permeability by both symplastic and apoplastic routes (Kaldenhoff and Fischer, 2006), are closely involved in water deficit. The transgenic rice overexpressing OsPIP2;3, localized in root endodermis, improved an adaptability against drought condition (Sun et al., 2021). OsPIP2;5 is predominantly localized in root endodermis and metaxylem and promotes water transport from endodermis to central xylem (Sakurai-Ishikawa et al., 2011). Our results showed a decreasing trend between different water potentials, and this finding is assumed that water movement toward the shoot could be strongly restricted to maintain water potential in root tissue under drought condition. OsPIP2;7, located in the endodermis and epidermis of root (Li et al., 2008), is highly promoted in response to ABA in drought-treated roots (Guo et al., 2006), and it is in an agreement with our observation. Accordingly, PIPs isoforms are likely to play different roles to regulate the inward/outward water flux in rice roots. TIPs isoforms, which regulate vacuolar water potential, are essential to control cell distention. Nguyen et al. (Nguyen et al., 2013) demonstrated that TIPs isoforms in rice plant were upregulated by water deficiency. Based on the current result, it is suggested that increased TIPs expression precisely regulate the water potential between both compartments, vacuole and cytosol.

All key words (ABA, suberization, AQPs, and fatty acid) gather in the topic of water transport in cell membrane under drought stress in the roots. Through the previous discussion, a severe drought level may suggest that plants no longer develop tolerance mechanisms and die. Therefore, we focused on the −1.0 and −1.5 MPa results to suggest a hypothesis about the mechanism of water movement in cell membrane in drought. It has been demonstrated that the content of unsaturated fatty acids constituting the cell membrane increase in drought, creating spaces between lipids, which increases the fluidity of the membrane and ultimately activates the function of aquaporins (Martínez-Ballesta and Carvajal, 2016). However, the interaction of membrane constituent has not been fully known. As confirmed in the previous discussion, a stress tolerance mechanism occurred in −1.0 and −1.5 MPa. Taken together, this study suggests that ABA is involved as a signaling molecule in the interaction of aquaporins and suberin, and suberin lamellae formed on the membrane may interact with fatty acids and aquaporins to affect the fluidity and water transport. In Figure 7 , it can be seen that the genes belonging to each keyword are highly correlated. OsPIP2;5 was the variable accounting for 28.8% of PC2. Clearly, OsAAO3 and OsABA3 are negatively correlated with OsESB1 in PC2. The defect of atesb1 is compensated by ABA biosynthesis (Wang et al., 2019). The results are also supported with that plants use ABA to mitigate the adverse effects of suberization developed through suppression of OsESB1 expression.

Further studies will be required to elucidate the interaction of each pathway in more detail to ultimately enhance drought tolerance mechanism. Since this study focused only on seedlings and roots, it will be needed to study the changes in each mechanism and the effect on other organs such as to leaves and grains.