The plant materials used in this study were: indica rice variety Huanghuazhan (HHZ); mutant gddt (growth-depressed and drought-tolerant) generated from HHZ through radiation mutagenesis; japonica rice variety Zhonghua 11 (ZH11); F₂ populations derived from the crosses of gddt × ZH11 and gddt × HHZ, respectively.

Seeds of HHZ (wild-type control) and gddt were sown individually in 96-well hydroponic boxes containing Kimora B formula nutrient solution, which was refreshed every 7 days. Shoot length and root length of seedlings were measured at the three-leaf stage. Additionally, HHZ and gddt were directly seeded in the paddy field at our university’s Jinshan campus in Fuzhou, following standard water, fertilizer, and pesticide management practices. At the mature stage, plant height, tiller number, panicle length, and seed setting rate were examined.

Seeds of HHZ and gddt were planted in rectangular plastic trays (36 cm × 28 cm × 4.5 cm) filled with paddy soil. After reaching the three-leaf stage, seedlings were subjected to approximately 10 d of drought stress by withholding watering. Subsequently, seedlings were rehydrated (1.5 L/tray), and their survival rates were assessed.

Seedlings of HHZ and gddt were cultured in 96-well hydroponic boxes as previously described. When reaching the three-leaf stage, seedlings were exposed to 20% PEG6000 for either 0 or 3 d. Subsequently, a 0.5 cm segment was cut from the middle part of the third leaf of each seedling, fixed with glutaraldehyde, washed with buffer solution, subjected to ethanol gradient dehydration and tert-butanol replacement, and finally dried using a VFD-21S freeze-drying instrument. Samples were then observed and photographed using a Leica S-800 scanning electron microscope. Stoma observation was conducted across 10 different visual fields for each sample.

The hydrogen peroxide (H₂O₂) levels in the shoots of three-leaf seedlings of HHZ and gddt obtained through 96-well hydroponic box culture as previously described were quantified using a kit from Beijing Bioss Biotechnology Co., LTD after treatment with 20% PEG6000 for 0 or 3 d. Three biological replicates were analyzed, each comprising 5 seedlings. In addition, nitrogen blue tetrazole (NBT) staining analysis was conducted on the seedlings, utilizing the third leaf of each seedling. Five biological replicates were examined, with one seedling per replicate. Furthermore, MDA levels (assessed via the thiobarbituric acid method), DHAR activities (Beijing Solarbio Technology Co., LTD. Kit), SOD activities (NBT method), and CAT activities (UV absorption method) in the shoots of the seedlings under PEG treatment were monitored daily until the 5^th day. Once more, three biological replicates were established, each comprising 5 seedlings.

Seedlings of HHZ and gddt were cultured in 96-well hydroponic boxes according to the previously outlined method. Upon reaching the three-leaf stage, shoots were harvested from a portion of the seedlings, immediately frozen with liquid nitrogen, and stored in an ultra-low-temperature refrigerator. The remaining seedlings underwent treatment with 20% PEG6000 for 24 h before being similarly frozen with liquid nitrogen. RNA extraction was carried out using a plant RNA small amount extraction kit (Guangzhou Meiji Biotechnology Co., LTD.). Subsequent reverse transcription and real-time fluorescence quantitative PCR (RT-qPCR) were performed utilizing kits from Shanghai Yeasen Biotechnology Co., LTD. Each sample was subjected to three biological replicates, with the Actin gene serving as the internal reference. Primer sequences are detailed in Supplementary Table S1 .

An F₂ population consisting of approximately 1000 plants derived from the cross between gddt and ZH11 was cultivated in our university experimental field. Around 60 d post-sowing, a subset of the F₂ population (comprising 245 plants) was examined to determine the segregation ratio between normal and mutant plants. Subsequently, 193 mutant plants were chosen from the F₂ population to establish a mutant DNA pool (designated as MZ-pool). This DNA pool, along with DNA samples from the two parental lines, was dispatched to Xiamen Jointgene Biotechnologies Co., LTD. for deep sequencing using Illumina’s Hiseq 2000 platform. SNPs and short InDels between the two parents were identified from the sequencing data, with reference to the genome of the rice variety Nipponbare (version IRGSP-1.0). Leveraging SNP and short InDel markers, the genomic region harboring the target gene was identified using the block regression mapping (BRM) method (Huang et al., 2020).

Based on the findings and data derived from BSA-seq, InDel markers were developed within the genomic region predicted to harbor the target gene ( Supplementary Table S1 ). Employing these InDel markers, along with some available RM-series SSR markers (https://www.gramene.org/microsat/ssr.html), fine mapping of the target gene was conducted by analyzing the recombination events between each marker and the target gene in mutant plants selected from the aforementioned F₂ population.

To pinpoint the candidate gene, an F₂ population consisting of approximately 500 plants derived from the cross between gddt and HHZ was developed. From this population, 100 mutant plants were selected to form a pool (designated as MH-pool), which was then subjected to deep sequencing using the same methodology as described previously. Utilizing the sequencing data from this DNA pool, in conjunction with the single-pool BSA-seq data mentioned above, the polymorphic site displaying the most extreme allele frequency bias within the fine mapped interval was identified, and the gene containing this site was designated as the candidate gene.

To knockout candidate gene OsPDIL1-5 (LOC_Os06g06790), two targets within the gene were designed for CRISPR-Cas9 ( Supplementary Table S1 ; Supplementary Figure S1A ) using the CRISPR RGEN software (http://www.rgenome.net/), followed by the construction of CRISPR-Cas9 vectors ( Supplementary Figure S1B ) and transfer of the vectors into ZH11. To perform genetic complementation test, the coding sequence of OsPDIL1-5 was amplified from HHZ cDNA using Phanta Max Super-Fidelity DNA Polymerase (Nanjing Vazyme Biotechnology Co., LTD.) and inserted into the UBI-driven vector pRHVcGFP ( Supplementary Figure S1C ), which was subsequently introduced into the gddt mutant. The agronomic traits and drought tolerance of both OsPDIL1-5 knockout (KO) and complementation (COM) lines were assessed using the methodologies outlined above.

The expression of OsPDIL1-5 in different tissues of HHZ, including root, stem and leaf at the seedling stage, and spikelet (before flowering) and developing seed (7 d after flowering) during the heading period, was detected using RT-qPCR. The methods for RNA extraction and RT-qPCR were the same as described above ( Supplementary Table S1 ). To examine whether OsPDIL1-5 is localized in endoplasmic reticulum (ER) as usually expected for a PDI-like protein, the pRHVcGFP-OsPDIL1-5 vector, which carried the fused gene OsPDIL1-5-GFP ( Supplementary Figure S1C ), was co-transferred with ER-RFP marker vector (HDEL-AtWAK2-RFP) into rice protoplasts using the PEG-mediated method, using the empty pRHVcGFP vector as negative control. Fluorescence in the transformed protoplasts was observed, and images were captured using a confocal laser scanning microscope (TCS SP8, Leica) with excitation wavelengths of 488 nm for GFP and 561 nm for the ER-RFP marker.

Compared to HHZ, the gddt mutant exhibited a significant reduction in shoot length and root length at the seedling stage ( Figures 1A, D, E ) and in plant height, tiller number, panicle length, and seed setting rate at the maturity stage ( Figures 1B, C, F–I ), indicating that the mutation has a pleiotropic effect on both vegetative growth and reproductive development. Morphologically, the mutant could be easily identified by the naked eye based on the number of tillers ( Figure 1B ).

The performances of HHZ and gddt seedlings in the drought stress experiment are shown in Figures 2A–F . Following a period of no watering for 10 d, the HHZ seedlings wilted, while the gddt seedlings only displayed leaf curling ( Figure 2D ). After rewatering for 3 d, all the gddt seedlings recovered with leaves fully expanded, whereas only 3 (6%) HHZ seedlings recovered and survived ( Figure 2F ). These findings suggest that gddt is much more tolerant to drought than HHZ at the seedling stage.

Examination of abaxial leaf stomata using a scanning electron microscope revealed that gddt had smaller stomata than HHZ ( Figures 2G–L, N ). Additionally, gddt also exhibited a lower stomatal density than HHZ ( Figure 2M ). In regard to stomatal aperture, it could be classified into three categories in both gddt and HHZ: completely open (CO), partially open (PO), and completely closed (CC) ( Figures 2G–L ). Under normal conditions, gddt had a lower percentage of CO (denoted as CO%) but a higher CC% than HHZ, while the PO% was comparable between gddt and HHZ ( Figure 2O ). Under drought stress, the CO% decreased and the CC% increased in both gddt and HHZ compared to normal conditions, but the tendency of lower CO% and higher CC% in gddt than in HHZ was maintained, while the PO% was similar to that under normal conditions in both HHZ and gddt ( Figure 2O ). Overall, it can be inferred that gddt may primarily affect drought tolerance by causing a reduction in stomatal density, size, and aperture.

We analyzed five physiological indexes and the expression of five genes related to drought tolerance in both gddt and HHZ seedlings at the three-leaf stage. The MDA content was consistently lower in gddt compared to HHZ over the five-day stress period, with a notable difference on the 5^th day, aligning with the drought tolerance phenotype of gddt, although the difference was not statistically significant from the 1^st to the 3^rd day ( Figure 3A ). H₂O₂ content in gddt was significantly lower than in HHZ under both normal and drought conditions. Notably, after stress, H₂O₂ content increased in HHZ but decreased in gddt ( Figure 3B ), a trend confirmed by NBT staining ( Figure 3C ).

During the five-day drought stress period, DHAR activity initially increased and then decreased in both HHZ and gddt. However, the activity in gddt was consistently greater than or at least comparable to that in HHZ ( Figure 3D ). Similarly, SOD and CAT activities showed a pattern of initial increase followed by a decrease in both HHZ and gddt during drought stress, with gddt generally exhibiting higher values compared to HHZ ( Figures 3E, F ). RT-qPCR analysis indicated that the expression of SOD-encoding gene OsSODA increased in both HHZ and gddt after 24 h of drought stress, with a significantly higher relative expression level in gddt under drought conditions ( Figure 3G ), which aligns with the SOD activity observed. The relative expression levels of CAT isozyme genes OsCATA and OsCATC were higher in gddt than in HHZ under normal conditions, and OsCATA expression remained significantly higher in gddt than in HHZ after 24 h of drought stress ( Figure 3G ), paralleling the observed CAT activity. However, the expression pattern of the DHAR-related gene OsDHAR1 did not well align with the DHAR activity. While the expression of OsDHAR1 increased in both HHZ and gddt after 24 h of drought stress as expected, its relative expression level in gddt was lower than that in HHZ under both normal and drought conditions ( Figure 3G ). Nevertheless, despite this inconsistency, these findings suggest an overall activation of the ROS scavenging system in gddt to help mitigate drought stress.

Since tiller number is easily observed among all mutant traits, it was used as an index to analyze their heredity. The F₁ generation of gddt × ZH11 displayed normal tillering, indicating that the mutant tiller phenotype (decreased tiller number) is recessive. Among the 245 F₂ plants studied, 189 showed normal phenotype, while 56 displayed mutant phenotype, a ratio close to the theoretical 3:1 segregation (χ² = 0.60, P = 0.439), suggesting that a single gene, named GDDT, controls the mutant traits.

In the single-pool BSA-seq analysis, the mutant DNA pool (MZ-pool) and the parental DNA samples (gddt and ZH11) underwent deep sequencing, with average coverages of approximately 74.9, 137.2, and 141.2 times the rice genome, respectively. From the sequencing data, around 2.6 million SNPs and short InDels were identified between gddt and ZH11. BRM analysis using these markers revealed the lowest allele frequency in the MZ-pool around the 3 Mb position on chromosome 6 ( Figure 4A ), indicating a probable location for GDDT. Subsequent fine mapping based on 177 mutant plants from the F₂ population of gddt × ZH11 using InDel markers and RM-series SSR markers narrowed down GDDT to an interval of 282.5 kb between markers RM19410 and ID3341597 ( Figure 4B ). Deep sequencing of the mutant DNA pool (MH-pool) from the F₂ of gddt × HHZ revealed a single base deletion (TCA→TA) in the allele from gddt within the target interval at the position of 3,200,128 bp on chromosome 6. This mutation site displayed complete segregation distortion with only the deletion allele from gddt being detected in both MH-pool and MZ-pool, out of a total of 86 and 51 counts of the site, respectively. The deletion was located in the second exon of the OsPDIL1-5 gene ( Figure 4C ), causing a frameshift mutation and early translation termination. Sequencing of OsPDIL1-5 in gddt, HHZ, and three mutant plants randomly selected from the F₂ population confirmed this deletion as the only mutation in the gene, implicating OsPDIL1-5 as the most likely candidate gene for GDDT.

Two OsPDIL1-5 knockout lines, Ospdil1-5ko1 (KO1) and Ospdil1-5ko2 (KO2), were generated from ZH11 using CRISPR-Cas9 ( Supplementary Figure S1A ). In KO1, a 209-base segment of OsPDIL1-5 was deleted, starting from the 103^rd base in the first exon. In KO2, a one-base insertion occurred at the 312^th base in the first exon, leading to a frameshift mutation. Consequently, the function of OsPDIL1-5 was completely impaired in both KO1 and KO2. In comparison to ZH11, KO1 and KO2 exhibited significantly enhanced drought tolerance and reduced shoot length and root length during the seedling stage ( Figures 5A–D, G, H ), as well as decreased plant height, tiller number, panicle length, and seed setting rate at the maturity stage ( Figures 5E, F, I–L ), mirroring the phenotypes of the gddt mutant. These findings provide confirmation that OsPDIL1-5 is the target gene GDDT.

Genetic complementation (COM) lines were generated by introducing the vector pRHVcGFP containing the coding sequence of the OsPDIL1-5 gene into the gddt mutant ( Supplementary Figure S1 ). Relative to gddt, the COM lines exhibited notably weaker drought tolerance ( Figures 6A–C ) and significantly longer shoot and root lengths ( Figures 6D, G, H ) during the seedling stage, as well as greater plant height, tiller number, panicle length, and seed setting rate at the maturity stage ( Figures 6E, F, I–L ), akin to the wild-type HHZ. This outcome provides further validation that OsPDIL1-5 is indeed the target gene GDDT.

As expected, the merged images of green fluorescence from the OsPDIL1-5-GFP fusion protein and red fluorescence from the ER-marker observed under a confocal microscope clearly indicate that OsPDIL1-5 is specifically localized in the ER ( Figure 7A ).

RT-qPCR analysis revealed that OsPDIL1-5 is expressed in all five tissues/organs assessed, with the highest expression in spikelet, followed by developing seed, stem, and root, and the lowest expression in leaf ( Figure 7B ).