*2.3. Characterization of gw2-ko*

The *O. grandiglumis* segment harboring *qCC2* is approximately 1.4 Mb in size. In this region, *GW2* was selected as a candidate gene based on its function as a RING-type E3 ubiquitin ligase. A 1-bp deletion at position 316 in the coding region of the *O. grandiglumis GW2* allele, which causes a premature stop and leads to an increased grain width and thickness, was also observed in this study (Figure S4) [21]. In *Arabidopsis*, ubiquitin-activated peptidase DA1 regulates leaf senescence together with E3 ubiquitin ligase BB [18]. In addition, *DA1* controls seed and final organ size with interacting partner *DA2,* which is orthologous to *OsGW2*. Due to the previously identified pleiotropic effect of *GW2*, the *GW2* gene was selected as a candidate gene for the *qCC2* QTL.

To determine if *GW2* is associated with delayed senescence, a T-DNA insertional mutant (PFG\_1B-10017.R) was selected from a T-DNA insertional mutant library [25]. The T-DNA insertion was confirmed by two sets of PCR primers, and homozygous T-DNA lines were selected by genomic PCR (Figure 4a,b). The *OsGW2* knock-out mutant (*gw2-ko*) showed increased grain size, and the transcription of *OsGW2* in wild type (WT Dongjin) and *gw2-ko* was evaluated using semi-quantitative RT-PCR (Figure 4c,d). The *OsGW2* transcript level was examined during DIS in WT. The expression level was down-regulated at 2 DAI, whereas gene expression was up-regulated after 4 DAI, and a similar level of gene expression with 0 DAI was observed (Figure 4e). Consistent expression patterns of *OsGW2* and *OgGW2 (O. grandiglumis GW2)* were found in Hwaseong and CR2002, but CR2002 showed lower transcript levels than Hwaseong at 0 and 4 DAI (Figure S5).

**Figure 4.** *OsGW2* T-DNA insertion knock-out mutant (*gw2-ko*) in wild-type Dongjin genetic background. (**a**) Schematic diagram showing the location of T-DNA insertion of the PFG\_1B-10017.R mutant. (**b**) T-DNA insertion was confirmed using two sets of primers indicated in (**a**), and lines homozygous for the T-DNA insertion were used for this experiment. (**c**) Comparison of grain shape between wild type and *gw2-ko*. (**d**) Confirmation of *OsGW2* knock-out by semi-quantitative RT-PCR. *OsUBQ5* was used as a loading control. (**e**) Gene expression pattern of *OsGW2* during the dark-induced senescence. The transcript level of genes was determined by qRT-PCR. *OsUBQ5* was used for normalization.

The *gw2-ko* mutant showed significantly higher SPAD values than Dongjin from 85 to 130 DAT under the field conditions (Figure 5a,b). To characterize the phenotypic difference between Dongjin and *gw2-ko* during DIS, fully expanded detached leaves were incubated in 3 mM MES buffer (pH 5.8) at 27 ◦C in the dark (Figure 5c). Leaves of Dongjin turned yellow at 6 DAI, while *gw2-ko* remained green. Total chlorophyll contents of Dongjin and *gw2-ko* at 6 DAI were also significantly different, despite the total chlorophyll contents of *gw2-ko* at 0 DAI being slightly higher than Dongjin (Figure 5d). The *Fv*/*Fm* ratio remained higher in *gw2-ko* than Dongjin at 5 DAI (Figure 5e). These results demonstrated that leaf senescence was delayed in *gw2-ko*.

An F2 population (*n* = 107) derived from a cross between Dongjin and *gw2-ko* was evaluated for DIS to determine if the T-DNA insertion in *OsGW2* was responsible for delayed senescence. After five days of dark incubation, plants homozygous for *gw2-ko* genotype (KO/KO group) showed significantly higher total chlorophyll contents than those homozygous for Dongjin genotype (DJ/DJ group) (Figure 6a,b). This segregation analysis suggested that T-DNA insertion in *OsGW2* caused the delayed senescence exhibited by *gw2-ko*.

Endogenous expression levels of SAGs and CDGs were examined (Figure 7). Among CDGs, *SGR*, *NYC1*, and *NOL* were significantly down-regulated in *gw2-ko* at 6 DAI, but *PAO* and *RCCR1* showed no significant difference between Dongjin and *gw2-ko*. SAGs (*OsNAP*, *Osh36*, and *OsI57*) were also down-regulated in *gw2-ko*. The loss-of-function of *GW2* led to the down-regulation of SAGs and CDGs during DIS.

The physiological phenotypes and endogenous transcript levels during DIS indicated that *gw2-ko* showed delayed leaf senescence compared to wild type and that *GW2* positively regulated leaf senescence in rice.

**Figure 5.** *gw2-ko* mutant showed delayed senescence. Comparison of (**a**) plant morphology (at 138 days after transplanting) and (**b**) SPAD value (*n* = 30) of wild-type Dongjin and *gw2-ko*. (**c**) Wild type and *gw2-ko* were grown in a paddy field, and fully expanded flag leaves at the heading stage were used for DIS. (**d**,**e**) Total chlorophyll contents (*n* = 6) and *Fv*/*Fm* ratio (*n* = 10) were compared between wild type and *gw2-ko*. Error bars indicate a standard error, and more than six samples were used for each experiment. \* and \*\*\* indicate significant difference at *p* < 0.05 and *p* < 0.001 based on Student's *t*-test, respectively.

**Figure 6.** Boxplot of chlorophyll contents at (**a**) 0 day after incubation and (**b**) 5 days after incubation under the dark-induced senescence conditions in the F2 population derived from a cross between Dongjin and *gw2-ko*. Genotype was determined with two sets of primers, as indicated in Figure 4a—LP+RP and RB+RP. One-way ANOVA was conducted to determine the significant difference between genotypes. \*DJ/DJ: homozygous for Dongjin, DJ/KO: heterozygous, KO/KO: homozygous for *gw2-ko*.

**Figure 7.** Expression of senescence-associated genes and chlorophyll degradation genes compared in Dongjin and *gw2-ko*. qRT-PCR was conducted to determine the transcript level of genes. *OsUBQ5* was used for normalization. Error bars indicate the standard deviation of three replications. \*, \*\*, and \*\*\* indicate significant difference at *p* < 0.05, *p* < 0.01, and *p* < 0.001 based on Student's *t*-test, respectively. DAI: days after incubation.

#### *2.4. Transcriptome Analysis using Hwaseong and CR2002*

To gain a better understanding of the mechanism for delayed senescence under dark-treatment conditions, the transcriptome of leaves from Hwaseong and CR2002 plants before (control) and after dark-treatment for 3 days were investigated by RNA-seq. A total of 264 million reads were generated using the Illumina Hiseq 4000 platform, and adapter and low-quality read trimming were conducted (Table S2). A total of 262 million high-quality reads were mapped against the reference genome sequence, and an average mapping rate of 66% was observed (Table S2). A multi-dimensional scaling plot showed that three replicates were clustered, indicating samples were consistently generated (Figure S6).

Differentially expressed genes (DEGs) were identified in Hwaseong and CR2002 under dark treatment compared with control conditions. DEGs were determined based on the criteria of a greater than 2-fold change with significance at Benjamini-Hochberg false discovery rate adjusted *p* < 0.05. A total of 10,339 DEGs was detected in the comparison of 0 day and 3 days of dark treatment in Hwaseong, and 9885 DEGs were found in CR2002 genotype during the dark treatment (Figure 8a). A Venn diagram of the DEGs indicated that Hwaseong and CR2002 shared 7901 of them (Table S5). Among these 7901 genes, only 29 genes showed different regulation patterns between Hwaseong and CR2002. For example, Os09g0511600 was up-regulated in Hwaseong by 5.13 log2FC (fold-change), while down-regulated in CR2002 by -8.26 log2FC. In addition, a total of 2438 and 1984 genes were specifically identified as DEGs in Hwaseong and CR2002, respectively (Tables S3 and S4). For Hwaseong, 1240 and 1198 genes were specifically up- and down-regulated during the dark treatment, respectively (Figure 8a and Table S3). A total of 838 up-regulated genes and 1146 down-regulated genes were specifically identified during the dark treatment in CR2002 (Figure 8a and Table S4).

**Figure 8.** Summary of the numbers of differentially expressed genes (DEGs) upon incubation of leaves of two lines (Hwaseong and CR2002) in complete darkness. (**a**) A Venn diagram, showing the number of genes shared and distinct to each genotype. (**b**) Significantly enriched gene ontology (GO) terms for CR2002-specific DEGs. Blue and green bars indicate the input list of CR2002 and background/reference, respectively. H0, H3, G0, and G3 indicate Hwaseong day 0, Hwaseong day 3, CR2002 day 0, and CR2002 day 3 in dark condition, respectively.

To know functional information of the DEGs specifically detected in Hwaseong and CR2002, gene ontology (GO) analysis was performed with agriGO v2.0. For the 1984 CR2002-specific DEGs, a total of 33 GO terms was significantly enriched (Figure 8b). Among 33 GO terms, biological process, cellular component, and molecular function ontologies included 8, 21, and 4 GO terms, respectively. For Hwaseong-specific DEGs, a total of 33 GO terms were also significantly overrepresented. Of these GO terms, 6, 23, and 4 were included in the biological process, cellular component, and molecular function ontologies, respectively (Figure S7). Hwaseong-specific DEGs included intracellular signaling cascade, organelle membrane, cytosol, cytoskeleton, Golgi membrane, endomembrane system, and Golgi apparatus GO terms, while iron ion transport, cytoplasmic part, envelope, cellular response to chemical stimulus, cell cycle, microtubule-associated complex, and organelle envelope were included only in CR2002-specific DEGs.

To understand the mechanism involved in delayed senescence in CR2002, the SAGs, CDGs, and genes involved in phytohormone biosynthesis and signaling were examined from the DEGs. A total of 109 genes were identified (Table S6). Among these genes, we focused on those specifically detected in Hwaseong or CR2002 (Table 3). During the dark treatment, *OsNIT1*, *OsDWF1*, and *OsRR3* were up-regulated in CR2002, and *OsERF5*, *OsABI1*, *OsJAZ1*, and *OsGLU* were down-regulated. Phytohormone signaling and biosynthesis genes were mainly found in the CR2002 DEGs. When these seven DEGs were subjected to RT-qPCR using the same G0/G3 samples as for RNA-Seq, similar expression patterns were observed, suggesting the involvement of these DEGs in delayed senescence in CR2002 (Figure S8). For Hwaseong, 18 genes were identified from the DEGs, and 10 and 8 genes were down- and up-regulated, respectively. Hwaseong-specific DEGs included phytohormone biosynthesis and signaling genes, NAC transcription factors, and senescence-associated genes.


**Table**detected in CR2002 and Hwaseong.

