Three KO lines and three overexpression lines were used in this study. The three overexpression lines, in which the OsGIF1 (LOC_Os03g52320) was driven by the 2x35S promoter (Mao et al., 2005), were obtained by self-pollinating the T0 plants described in our previous report (Li et al., 2016). The japonica variety Nipponbare (Nipp) was used as control. All plants were planted in the experimental field at the Rice Research Institute, Sichuan Agricultural University, Wenjiang. Phenotypic data were collected at the maturing stage. The data were analyzed using Excel (Microsoft) for mean values and standard errors of mean (SEM). Statistical significance was assessed by conducting Student’s t-test.

The primer sequences for molecular cloning and constructions are listed in Supplementary Table 1. We verified OsGIF1 function by generating two gRNA constructs, in which the gRNA was driven by the rice U6 promoter, and the plant-optimized Cas9 was driven by the UBI promoter (Miao et al., 2013). These constructs were introduced into the WT (Nipp) (Hiei et al., 1997). Then, the transgenic plants were subjected to PCR and sequencing analysis to determine the occurrence of mutations. To verify the association between the mutation in OsGIF1 and the mutant phenotype, we performed segregation analysis in some populations generated by back-crossing these mutants with WT and confirmed the co-segregation of the mutant phenotype and mutations. To evaluate off-target effects, four putative off-target sites were identified by similar sequence searches within the rice genome (Supplementary Table 2). These sites were then sequenced for mutation analysis.

The grain length, width, and 1,000-grain weight were measured by an automatic seed-size analyzing system (SC-G, Wanshen, Hangzhou). Other traits were investigated using conventional methods at maturing stage. An environmental scanning electronic microscope (QUANTA 450, Nikon) was employed to observe the outer surface of the leaf, stem internode, and outer glume. For histological analysis, samples of the leaf and stem internode were placed in the FAA solution for 12 h at 4°C, dehydrated in a graded ethanol series, followed by substitution using 3-methylbutyl acetate (Li et al., 2016). The samples were dissected and observed under a microscope (80I, Nikon) for determinations of cell number and size.

To localize the transcripts of OsGIF1 in rice tissue, a OsGIF1pro::GUS construct in which expression of the GUS gene was driven by native promoters was generated. Then, the OsGIF1pro::GUS construct was introduced into the Agrobacterium tumefaciens strain EHA105 to transform WT plants. Histochemical GUS staining was performed in the transgenic plants as described previously (Li et al., 2013).

Total RNAs were isolated from various rice tissues at different developmental stages using the TriPure Isolation reagent (Roche). The cDNAs were then reverse-transcribed using the Transcriptor First-Strand cDNA Synthesis kit (Roche). qRT-PCR was conducted in a total volume of 10 μl, with 0.3 μl of the reverse-transcribed product, 0.08 mM gene-specific primers, and 5.0 ml of Sso Advanced TM SYBR Green Supermix (Bio-Rad) using a Bio-Rad CFX96 Real-Time PCR System according to the manufacturer’s instructions. Data were analyzed by the relative quantification method (Livak and Schmittgen, 2001). The rice Actin gene was used as internal control. Each measurement was determined for at least two biological samples using three replicates for each sample.

The full-length cDNA of OsGIF1 was cloned into the pA7-GFP vector to generate the 2x35S::OsGIF1-GFP cassette. This plasmid was then introduced into rice protoplast cells for transient expression (Li et al., 2013). GFP signals were visualized using a confocal scanning microscope (Nikon A1, Kanagawa, Japan) 24–48 h after transformation.

In order to investigate the function of OsGIF1, plasmids containing CRISPR-Cas9 guide RNAs (gRNAs) against two target sites within the first exon of OsGIF1 were made (Figure 1B). The plasmids were then transformed into the wild-type (WT) variety of Nipponbare (Nipp), and approximately 30 transgenic plants were obtained for each plasmid. Sequencing of PCR-amplified OsGIF1 genomic DNA from transgenic plants showed five types of homozygous mutations within the target sites: one in target site 1 (an A insertion) and four in target site 2 (3-base deletions, 18-base deletions, 10-base deletions, and an T insertion) (Figure 1B). Although no biallelic plants were obtained, several heterozygous transgenic plants of the five mutations were found, and no phenotypes were observed.

Interestingly, the phenotype of the homozygous osgif1 mutant differed from its mutation type. Slight effects were observed on plant height in the 3-/18-base-deletion plants. By contrast, the A/T insertion plants had severe effects on plant height, leaf development, and panicle construction. Additionally, the 10-base-deletion plants had moderate impact on these traits, compared with plants bearing the other two mutations (Figures 1A,B). These phenotypic differences are to some extent in accordance with the effects of each mutation on the protein sequence of OsGIF1. The 3-/18-base deletions only deleted 1 or 6 AAs without changing the coding frame of OsGIF1, while the A/T insertion and 10-base deletions resulted in evidently truncated OsGIF1 by introducing premature stop codons (Supplementary Figure 1). To elucidate whether these differences in phenotype were also linked with off-target effects of the gRNAs, we sequenced several potential off-target sites and found no off-target editing for both gRNAs (Supplementary Table 2). Further linkage analysis of a F2 populations (28 plants in total) generated by the crosses of the T408-6 mutants and the WT found that all the plants with homozygotic mutations were showed in the mutant phenotype (6 plants), whereas other plants carrying no (10 plants) or heterozygous mutations (12 plants) were all normal in phenotype, confirming the association between the mutation and the mutant phenotype. Thus, the phenotype was directly caused by OsGIF1 mutation, and the phenotypic diversity of the different mutation types might have resulted from varying degrees loss of OsGIF1 function.

The most noticeable phenotype in the OsGIF1 KO plants was reduction of plant height. Three types of mutant phenotypes were found (Figure 1A). Thus, we selected three representative mutant plants, namely, T408-14, T408-6, and T409-17 to analyze the phenotype and function of OsGIF1. The plant height of all the KO plants decreased highly significantly compared with that of the WT (Figures 2A,B). Consistent with decrease in height, most of the mutant panicles became smaller, and the number of grains per panicle in the mutants was reduced (Figures 2C,D). We also examined the stems and internodes of mutant plants, since rice plant height is primarily determined by internode length of the stem. The results showed that almost all mutant stems and their internodes were considerably shortened, compared with their WT counterparts (Figures 2A,E). These results demonstrated that functional loss of OsGIF1 led to reduced plant height of rice by shortening the internode length of stems.

OsGIF1 KO plants also shows phenotypes related to leaf development (Figure 2F). Leaf lengths and widths of all the mutant plants were significantly reduced compared with those of the WT (Figures 2G,H). Besides, these mutants also showed different degrees of leaf rolling. The leaf rolling indices of plants with the most severe mutations, namely T408-6 and T409-17, were sharply increased, with almost all leaves rolled compared with those of the WT (Figures 2F,I). These results indicated that OsGIF1 affected rice leaf development by regulating both leaf size and leaf rolling.

Rice is an important crop as a source of carbohydrate for more than half of the world population. Therefore, we investigated whether OsGIF1 affects the grain yield of rice. We found that grain weights of all mutant plants were significantly decreased compared with that of the WT (Figure 2J and Supplementary Figure 2). The reduced grain weights were further observed to be mainly caused by significant decreases in grain length and grain width of the OsGIF1 KO plants (Figures 2K,L). This indicated that OsGIF1 also functioned in rice seed development, which is consistent with our previous findings (Li et al., 2016).

Reductions in the panicle seed setting rate were also observed in the OsGIF1 KO plants (Figure 2M), suggesting a role of OsGIF1 in regulating the rice reproduction process. The degree of reduction was consistent with the mutant type. In the most severely mutated plant, T409-17, the seed setting rate was extremely low, and the plant became almost completely sterile (Figure 2M). The floral structures in T409-17 at the heading date manifested the following series of abnormalities in several whirls of floral organs: tightly wrapped spikelets, shorter, twisted, or ectopic paleas/lemmas, decreased stamens, increased pistils, and white-colored sterile anthers (Supplementary Figure 2). These results indicated that serious KO of OsGIF1 led to various floral organ abnormalities and apparent reductions in the seed setting rate, suggesting an important role of OsGIF1 in the determination of rice floral organs.

We performed detailed investigations of multiple organs of several representative overexpression plants designated as T381 1-3, T395 1-6, and T382 1-1. First, qRT-PCR was performed to confirm the overexpression of OsGIF1 within these representative plants (Figure 3C). The plant heights of the overexpressing plants were slightly or sometimes significantly increased (Figures 3A,D). The transgenic plants also exhibited larger panicles than the WT (Figure 3B) although the number of grains per main panicle did not differ significantly between the transgenic and WT plants (Figures 3E,L). Consistent with our previous work (Li et al., 2016), overexpression of OsGIF1 in these plants also significantly increased grain size and weight by synchronously increasing grain length and width (Figures 3B,F–H). Additionally, overexpression of OsGIF1 apparently affected leaf development, because all of these plants exhibited highly significantly increased leaf length and width (Figures 3I–K). These results confirmed that overexpression of OsGIF1 increased the size of multiple rice organs, indicating a positive role of OsGIF1 in regulating rice organ size.

To investigate the cellular effect of OsGIF1, we first performed paraffin sectioning of stem internodes of these plants. Longitudinal histological sectioning analysis showed a clear trend toward reduced cell volume in the KO plants, especially in the severe KO plants (Figure 4A). Overall, these results are consistent with the degree of organ size alternation in these plants (Figure 1A). To further address this, we then counted cell numbers per unit area and found significant increases in T408-14 (+19.92%), T408-6 (+47.39%), and T409-17 (+94.8%) when compared to WT (Figure 4D), suggesting that cell sizes of the KO plants were altered. Direct cell measurement showed that the cell perimeters were significantly reduced in T408-14 (-7.39%), T408-6 (-16.6%) and T409-17 (-30.77%) (Figure 4E). Transverse sectioning of these internodes (Figure 4B) further confirmed the significantly increased cell numbers per unit area (+28.86% in T408-6 and +75.03% in T409-17) (Figure 4F) and reduced cell size (-18.57% in T408-6 and -27.23% in T409-17) (Figure 4G) in the stems of severe mutants. Consistent with the result of stems, transverse section of flag leaves (Figure 4C) also showed significantly reduced cell size in T408-14 (-12.93%), T408-6 (-16.33%), and T409-17 (-23.06%) (Figure 4H). Since grain size is influenced by spikelet hull, we next performed scanning electron microscopy (SEM) analysis of the spikelet hulls of the overexpression and KO plants. Results showed that T382 1-1 exhibited significantly enlarged cell volume compared to WT, while T409-17 showed opposite effect of reduced outer glume cells (Supplementary Figure 3). Cellular measurement further showed that, compared with those in the WT, the length and width of epidermal cells of the outer glumes increased by 43.8 and 15.22% in T382 1-1, and decreased by 16.64 and 12.18% in T409-17, respectively (Supplementary Figure 3). Taken together, these results suggest that OsGIF1 enhances the size of multiple important rice organs predominantly by promoting cell enlargement.

We also observed other morphological abnormalities except for cell size alternation in these plants. Transverse section of internodes revealed that layers of sclerenchyma cells of the stem vascular bundle and the number and size of the stem vascular bundle were apparently decreased in the KO plants (Figure 4B). Development of the stem vascular bundle was also affected, with incomplete development of outer vascular bundle occurring in the most serious KO plants (Figure 4B). Besides, SEM analysis of flag leaf revealed that the morphologies and organization patterns of cells were also altered in T382 1-1 and T409-17 (Supplementary Figure 4). Stoma guard cells are typically arranged in the longitudinal direction and separated by a longitudinal line of dumbbell silicon cells in the WT. However, two longitudinal lines of dumbbell silicon cells and several incompletely developed lines of stoma guard cells were observed both in T382 1-1 and T409-17. Several cells from these regions were morphologically abnormal compared to the WT. Interestingly, the number of silicon papillae on the leaf stoma guard cells seemed to have increased both in the overexpressing and KO plants (Supplementary Figure 4). Moreover, compared to the WT, organization patterns of stem exterior cells were slightly changed both in overexpressing and KO plants. However, abnormal development of stoma guard cells was especially apparent in T409-17, with significantly fewer stoma guard cells in the stem (Supplementary Figure 4), which is consistent with the findings of a recent study in which a role of AN3/GIF1 in modulating stomatal density was reported (Meng and Yao, 2015). Notably, the number of bulliform cells in the leaves of the serious KO plants was increased (Figure 4C), which abolished the balance of the Ad-Ab patterning and consequently led to rolled leaves. Thus, our findings reveal a cellular mechanism of the leaf-rolling phenotype in the most severe KO plants (Figures 2F,I), consistent with the function of SEMI-ROLLED LEAF1, a gene that modulates rice leaf rolling by regulating the formation of bulliform cells (Xiang et al., 2012). These results suggest that OsGIF1 might also affect other cellular processes during rice organ or tissue development.

We generated a construct in which the GUS gene was driven by the approximately 1.8 kb OsGIF1 promoter to examine the temporal and spatial expression patterns of OsGIF1. Histochemical staining suggested that OsGIF1 was differentially expressed in various rice tissues. Expression was relatively weak in the root (Figure 5A) and mature glume (Figure 5H) but was strong in the internode (Figure 5B), node (Figure 5D), the developing spikelet (Figures 5E–G), and especially in the developing anther (Figures 5E–G). Results of qRT-PCR assays further confirmed this result by showing that OsGIF1 was highly expressed in developing stems but moderately in the root, young panicle, and flower (Figure 5I). Overall, the expression pattern of OsGIF1 is consistent with its function. However, both GUS staining and qRT-PCR analysis indicated extremely low expression of OsGIF1 in leaves approaching maturity (Figures 5C,I). Taken together with our results that OsGIF1 is involved in leaf development (Figures 2F–I, 3I–K), these findings point toward a role for OsGIF1 in the early development of rice leaf.

In order to investigate the subcellular localization of OsGIF1, we constructed an OsGIF1-GFP (green fluorescent protein) fusion construct with its expression driven by the CaMV 35S promoter. Transient expression in the rice protoplast cells showed that OsGIF1-GFP localized preferentially in the nucleus and was weakly expressed in the cytoplasm (Figure 6). This result is consistent with the idea that GIF proteins function as transcriptional co-activators by forming complexes with GRF transcription factors.