**4. Discussion**

Source-sink relationships largely determine the grain yield of cereal crops, with developing grains being primary sinks, while the top two leaves, the flag-leaf, in particular, serves as the primary source [60,61]. Source sink regulation is orchestrated through intricate metabolic signaling [62], of which key players in sucrose sensing (e.g., trehalose-6-phosphate) and signal integration (e.g., SnRK1) are beginning to emerge [62]. Drought stress affects these relationships by reducing both source and sink strengths. In source organs, limitations in carbon fixation and primary metabolism lead to reduced resource allocation to developing sinks, causing yield reduction characterized by suboptimal grain filling [63]. In sink organs, drought reduces fertility, causing yield reductions through suboptimal seed setting [64].

While DTY-QTLs have demonstrated effects to improve rice grain yields under RDS, knowledge about underlying molecular mechanisms is limited. Functional studies of *qDTY12.1* suggested an intricate pattern of below-ground contributions [18], while physiological studies of *qDTY1.1* suggested above-ground implications [59]. Though confined to a single time point at the late booting stage (close to anthesis) after two weeks of RDS, our study suggested that DTY controlled mechanisms improve yield under drought by acting at both source and sink organs. At the flag-leaf, a coordinated response to sustain primary metabolism through cell wall loosening and maintained photosynthetic rates seems to allow for sufficient carbon and energy allocation to the developing panicle, which in turn enable reproductive structures to invest in protective mechanisms, including protein stabilization and turnover, ROS scavenging and production of protective secondary metabolites. While proximate effects in the panicle are suggested as improved male fertility and improved sink strength under RDS, the ultimate effects are improved seed setting and grain filling, and consequently, drought-tolerant yield (DTY) (Figure S21).

### *4.1. Source E*ff*ects—Flag-Leaf-Specific Di*ff*erences between DTY-IL and Swarna*

Collectively WGCNA and DE analyses suggested a complex interplay of a range of processes in the flag-leaf to contribute to the observed differences in RDS tolerance between DTY-IL and Swarna. These included specific protein turn-over, cell wall loosening, efficient ROS scavenging, and maintenance of photosynthesis (Figures 4–6 and Figure S12).

A direct consequence of drought is impaired cell turgor [65], which is countered by the stiffening of cell walls to provide structural resistance [66,67]. Prolonged drought stress challenges plants to modify their cell walls, resulting in both cell wall tightening and loosening. Tightening occurs in tissues that are of relevance to structural integrity, while loosening occurs in tissues that need to be maintained in a growing and metabolically active mode [66].

Leaf rolling, a common indicator of drought stress in rice [68] was prominent in Swarna under drought but nearly absent in DTY-IL (Figure 1D). Leaf rolling likely relates to aberrant cell turgor and cell wall homeostasis and phenotypically reflects findings in the flag-leaf specific module M14.

Cell wall organization or biogenesis genes showed an increase in expression in FL-M14. A total of 12 cell wall-related genes were significantly upregulated in the DTY-IL and significantly downregulated in Swarna (Figure 6). These included two glycosyltransferase family 43 proteins, previously reported being involved in the synthesis of glucuronoxylan hemicellulose of secondary cell walls [69] and two expansin genes. Expansins facilitate loosening and extension of plant cell walls by disrupting non-covalent bonding between cellulose microfibrils and matrix glucans [70] and implications in response to dehydration are well documented [71–75] and rose [76].

Higher expression of cytoskeleton and cell cycle-related genes in DTY-IL (Figure 6) further supported the concept of maintained cell growth and stability in the tolerant flag-leaf tissue. Contrastingly, cytoskeletal genes (tubulin and formin) and a cell cycle gene (cyclin) were significantly downregulated in Swarna (Table S10-1).

Several classes of enzymes control ROS production in the cell wall, most prominently plasma membrane NADPH oxidases [77] and class III peroxidases (CIII Prxs) [78]. CIII Prxs are secreted in the extracellular space, where they perform either cell wall stiffening through the peroxidative cycle [79] or cell wall loosening through the hydroxylic cycle [80,81]. In the present study, three CIII Prxs (LOC\_Os03g13200, LOC\_Os07g01370, and LOC\_Os07g48020) were present in FL-M14 (Table S10-1), with LOC\_Os03g13200 and LOC\_Os07g48020 significantly upregulated in DTY-IL and significantly downregulated in Swarna (Figure 6). High CIII Prxs expression in DTY-IL could support the generation of •OH for cell-wall loosening through cleavage of cell wall polymers [67]. Interestingly, decreased expression of a calcium-dependent NADPH oxidase in Swarna and increased in activity of the DTY-IL in FL-M16 (Table S10-2) was also observed. It is also known as respiratory burst oxidase, and is a well-studied enzymatic source of superoxide [82,83], which had previously been implicated in drought and high-temperature stability [83]. Hence, loosening of the cell wall and synthesis of structural constituents together is suggested to contribute to tolerance of water-deficit in the flag-leaf of DTY-IL.

More effective ROS scavenging, in general, seemed to be an important mechanism differentiating drought responses of Swarna and DTY-IL. Higher expression of peroxiredoxin in DTY-IL (Table S10-2) suggests increased reduction capacity for H2O2, indicating enhanced detoxification in drought-stressed leaves.

A primary detrimental effect of water stress in source tissues is impaired photosynthesis [84]. Reduced abundance of photosynthesis-related proteins in response to RDS had previously been reported [85] and was indeed reflected in drought-stressed leaves of RDS-susceptible Swarna (FL-M16). Components of the light reaction (two photosystem genes, components of the core complex of photosystem II (PSII) involved the primary light-induced photochemical processes), the dark reaction (ribulose bisphosphatase and the fructose-1,6 bisphosphatase), and photorespiration (ribulose bisphosphate carboxylase large chain precursor) were found to be consistently downregulated in Swarna (Table S10-2), suggesting functional impairments of general photosynthesis. Protection of photosynthesis from photoinhibition through photorespiration is a well-characterized drought response and furthermore prevents ROS accumulation in green tissues [86]. In addition, Swarna showed a lower expression of two ferredoxin-NADP genes, involved in thylakoid electron transport, suggesting reduced capacity in regulating the relative amounts of cyclic and non-cyclic electron for ATP and redox homeostasis [69]. Consequently, it is argued that the physiological environment in DTY-IL under RDS supports relatively higher rates of photosynthesis, which in turn, might sustain higher rates of energy and carbon production to support primary metabolism and source strength, ultimately leading to improved grain filling.

### *4.2. Sink E*ff*ects—Panicle Specific Di*ff*erences between DTY-IL and Swarna*

Collectively WGCNA and DE suggested a number of distinct mechanisms to contribute to differences in RDS tolerance between DTY-IL and Swarna in panicles. They included protein stabilization and turnover, ROS scavenging, biosynthesis of secondary metabolites for protection of reproductive organs, and hormonal signaling presumably to adapt reproductive developmental processes to drought. Under field conditions, they resulted in an approximate doubling of yield under drought for DTY-IL as compared to Swarna (Figure 1A), at no significant difference in plant height (Figure 1B).

Dehydration stress enhances the production of ROS and ROS-associated peroxidation causing damage to cellular structures [87]. Being essential for cellular signaling, ROS homeostasis depends on the balance between ROS production and scavenging [82]. Analysis of genome-scale metabolic pathways in the DTY-IL revealed the up-regulation of genes involved in the biosynthesis of antioxidant enzymes and metabolites (Figures 8 and 9 and Figure S14).

Secondary metabolite production is crucial in stress-adaptive mechanisms [88]. Genes in pathways involved in secondary metabolite biosynthesis, lipid biosynthesis, redox homeostasis, amino acid metabolism, carbohydrate metabolism, and protein metabolism were upregulated at the maximum booting stage under RDS in DTY-IL and downregulated in Swarna for P-M10 and P-M15 (Figure 9). Several metabolic pathways found to be shared between the two modules include glutathione, terpenoid, and ascorbate metabolism.

De novo protein synthesis and turnover is fundamental for plants to cope with drought stress [85]. Translational efficiency is affected by ribosome composition, thus relative ribosomal protein abundance can modulate plant environmental responses [89]. Similarly, drought-responsive peptidases and heat shock proteins can alter the active proteome to cope with stress [85,90]. In P-M10 five ribosomal protein-related genes, six protein degradation-related genes (among them four peptidases), and two genes related to protein folding and repair displayed higher expression in DTY-IL (Figure 9; Table S12-1). Congruently reduced expression of two peptidases and three genes related to protein processing, including a heat shock protein was observed for Swarna in P-M15 (Figure 9; Table S12-2). Collectively this suggested that panicles of DTY-IL were more responsive and had the necessary energy to adapt its proteome to drought conditions than Swarna.

In P-M10, six genes involved in the ROS scavenging (two ascorbate peroxidases and four peroxidase precursors) had elevated expression profiles in the DTY-IL (Figure 9; Table S12-1). Efficient reduction of H2O2 by peroxidases had previously been implicated with drought-tolerance in rice [19]. Specifically, plant ascorbate peroxidases (APXs) are crucial for ROS homeostasis [91] and free radical detoxification though the ascorbate-glutathione cycle [92], and their functional role in rice drought tolerance was demonstrated through transgenic approaches [91]. In P-M15 three ROS scavenging genes (1 glutathione *S*-transferase, 1 glutathione peroxidase, and 1 stromal ascorbate peroxidase) had a lower expression profile in Swarna (Figure 9; Table S12-2). Glutathione peroxidase (GPX) catalyzes the reduction of H2O2 using thioredoxin (Trx), while glutathione *S*-transferases (GSTs) are key to the removal of xenobiotic compounds [85]. Ectopic expression of a GST in *Arabidopsis* [93] and a GPX in rice [94] resulted in enhanced drought tolerance, suggesting functional implications.

Interestingly, an ABC function gene with AP2 domain-containing protein (LOC\_Os07g22770) controlling floral organ identity was downregulated in Swarna RDS under P-M15 (Table S12-2), suggesting a link to aberrant Swarna floral development under drought [95]. TFs belonging to AP2 and MYB family are involved in panicle development as well as water-deficit stress response, implying that they may represent a crosstalk component between redevelopment and stress.

A negative regulator of plant drought tolerance in abscisic acid (ABA) signaling, protein phosphatase 2C (PP2C) [96], was upregulated in the DTY-IL in P-M10 (Table S12-1). PP2C inhibits the activity of sucrose non fermenting 1 related kinase 1 (SnRK1) [97], a central integrator of metabolic signaling and regulator of starvation response. Thus, the higher expression of PP2C in DTY-IL might correlate with reduced SnRK1 activity, indicative of anabolism rather than catabolism and thus growth rather than the starvation mode in panicles of DTY-IL.

Brassinosteroids (BRs) are growth-promoting steroid hormones important for male fertility and pollen development [98]. BR catabolism is controlled by BAS1, a cytochrome P450 monooxygenase [99]. BRs bind to the extracellular domain of a cell-surface receptor kinase, BRASSINOSTEROID INSENSITIVE1 (BRI1) to activate kinase activity [100,101]. In P-M10, a BAS1-orthologue and two BRI1 genes were found to be upregulated in DTY-IL (Supplementary Table S12-1), suggesting a role for BR signaling in the maintenance of male fertility as part of the *qDTY*1.1-mediated RDS responses.
