Transcriptional repression of GIF1 by the KIX-PPD-MYC repressor complex controls seed size in Arabidopsis

Seed size is a key agronomic trait that greatly determines plant yield. Elucidating the molecular mechanism underlying seed size regulation is also an important question in developmental biology. Here, we show that the KIX-PPD-MYC-GIF1 pathway plays a crucial role in seed size control in Arabidopsis thaliana. Disruption of KIX8/9 and PPD1/2 causes large seeds due to increased cell proliferation and cell elongation in the integuments. KIX8/9 and PPD1/2 interact with transcription factors MYC3/4 to form the KIX-PPD-MYC complex in Arabidopsis. The KIX-PPD-MYC complex associates with the typical G-box sequence in the promoter of GRF-INTERACTING FACTOR 1 (GIF1), which promotes seed growth, and represses its expression. Genetic analyses support that KIX8/9, PPD1/2, MYC3/4, and GIF1 function in a common pathway to control seed size. Thus, our results reveal a genetic and molecular mechanism by which the transcription factors MYC3/4 recruit KIX8/9 and PPD1/2 to the promoter of GIF1 and repress its expression, thereby determining seed size in Arabidopsis.


Results
The KIX-PPD complex represses seed growth. KINASE-INDUCIBLE DOMAIN INTERACTING 8/9 (KIX8/9) and PEAPOD1/2 (PPD1/2) were previously reported to form a KIX-PPD complex and regulate leaf size by influencing cell proliferation in Arabidopsis thaliana 21,22 . Here, we found an important function of the KIX-PPD complex in seed size control. The kix8-1 plants exhibited larger seeds than wild-type (Col-0) plants (Fig. 1a, c). Seed weight of kix8-1 plants was also heavier than that of wild-type plants (Fig. 1d). The size of cotyledons usually reflects changes in seed size [24][25][26] . Consistent with this, cotyledons of kix8-1 were larger than wild-type cotyledons (Fig. 1b, e). By contrast, seed size and weight and cotyledon area in kix9-1 plants were similar to those in the wild type ( Fig. 1a-e). The kix8-1 kix9-1 double mutant showed significantly larger and heavier seeds and bigger cotyledons than the kix8-1 and kix9-1 single mutant (Fig. 1a-e), indicating that KIX8 and KIX9 function redundantly to control seed size and weight in Arabidopsis.
Seed area and weight and cotyledon area in ppd2-1 plants were increased compared with those in wild-type plants, while seed area and weight and cotyledon area in ppd1-2 plants were comparable to those in wild-type plants ( Fig. 1a-e). Because the PPD1 gene (AT4G14713) is close to the PPD2 gene (AT4G14720) in the chromosome, we could not isolate the ppd1-2 ppd2-1 double mutants. We generated ppd2-cr mutation in the ppd1-2 mutant background and ppd1-cr mutation in the ppd2-1 mutant background to obtain the ppd1-2 ppd2-cr and ppd1-cr ppd2-1 double mutants using the CRISPR-Cas9 technology, respectively (Supplementary Fig. 1) 27 . The ppd1-2 ppd2-cr and ppd1-cr ppd2-1 mutants had similar phenotypes (Fig. 1a-e). Seed area and weight and cotyledon area in ppd1-2 ppd2-cr and ppd1-cr ppa2-1 mutants were significantly increased compared with those in ppd1-2 and ppd2-1 single mutants (Fig. 1a-e), indicating that PPD1 and PPD2 function redundantly to control seed size and weight. The ami-ppd and ppd2-1 plants have been reported to produce large and curvatured leaves with more cell number [19][20][21] . Interestingly, we observed that ppd1-2 ppd2-cr and ppd1-cr ppd2-1 showed strong curvature of leaves ( Supplementary Fig. 2a), indicating that ppd1-2 ppd2-cr and ppd1-cr ppd2-1 are strong alleles compared with ami-ppd. We measured the leaf area of 32d-old Col-0, ppd1-2 ppd2-cr and ppd1-cr ppd2-1 plants and found that the third to eight leaves of ppd1-2 ppd2-cr and ppd1-cr ppd2-1 were smaller than those of wild type, but the tenth and eleventh leaves of ppd1-2 ppd2-cr and ppd1-cr ppd2-1 were larger than those of wild type ( Supplementary Fig. 2b, c). Considering that ami-ppd leaves had more cells than Col-0 leaves, we further examined palisade cell size and number of ppd1-2 ppd2-cr and ppd1-cr ppd2-1 fifth leaves. The palisade cell size of ppd1-2 ppd2cr and ppd1-cr ppd2-1 plants was smaller than that of wild-type plants ( Supplementary Fig. 2d, e). By contrast, the palisade cell number in ppd1-2 ppd2-cr and ppd1-cr ppd2-1 leaves was higher than that in wild-type leaves ( Supplementary Fig. 2e), consistent with higher cell number in ami-ppd and ppd2-1 leaves [19][20][21] . These results supported that ppd1-2 ppd2-cr and ppd1-cr ppd2-1 promote cell proliferation in leaves, but decrease cell expansion. These data also suggest a possible compensation mechanism between cell number and cell size in ppd1-2 ppd2-cr and ppd1-cr ppd2-1. This compensation phenomenon has been observed in several mutants [28][29][30] .
In addition, reciprocal crossing experiments showed that kix8-1 or ppd2-1 single mutation acts maternally to influence seed size ( Supplementary Fig. 3). Together, these results indicate that the KIX-PPD complex regulates seed growth through the maternal tissue of mother plants.
As the integuments belong to maternal tissues in Arabidopsis, we investigated the development of the outer integuments in the wild type and kix8-1 kix9-1 ppd1-2 ppd2-cr. The kix8-1 kix9-1 ppd1-2 ppd2-cr plants had bigger ovule area and longer outer integuments than wild-type plants at 0 DAP (days after pollination) (Fig. 1h-j). We then counted the number of cells in wild-type and kix8-1 kix9-1 ppd1-2 ppd2-cr outer integuments and found that kix8-1 kix9-1 ppd1-2 ppd2-cr outer integuments contained more cells than wild-type outer integuments (Fig. 1k). By contrast, the length of cells in kix8-1 kix9-1 ppd1-2 ppd2-cr outer integuments was similar to that in wild-type outer integuments (Fig. 1l). These results indicate that the KIX-PPD module restricts cell proliferation in outer integuments before fertilisation. We then examined cell number and cell length in wild-type and kix8-1 kix9-1 ppd1-2 ppd2-cr outer integuments at 2, 4 and 6 DAP, respectively. The outer integument cell number and cell length of kix8-1 kix9-1 ppd1-2 ppd2-cr were increased compared with those of the wild type at 4 and 6 DAP ( Fig. 1k-l), thereby resulting in long outer integument and large seed size in the kix8-1 kix9-1 ppd1-2 ppd2-cr plants ( Fig. 1i-j). These results indicate that the KIX-PPD module limits both cell proliferation and cell elongation in outer integuments after fertilisation. We further compared the effect of kix8-1 kix9-1 ppd1-2 ppd2-cr on cell proliferation during ovule and seed developmental processes. As shown in Fig. 1k, the KIX-PPD module restrains cell proliferation of the outer integument during both ovule and early seed developmental stages.
The formation of the KIX8/9-PPD1/2-MYC3/4 complex. PPD1/2 and 12 JAZ proteins belong to TIFY class II protein family 31 [32][33][34][35] . We therefore used the split luciferase complementation assays to test whether PPD proteins could interact with these transcription factors. We found that PPD1 and PPD2 interacted with MYC3 and MYC4 (Fig. 2a), but not with MYC2 and other transcription factors ( Supplementary  Fig. 4). By contrast, we did not detect the interactions between KIX8/9 and MYC2/3/4 in split luciferase complementation assays ( Supplementary Fig. 5). The interactions between PPD1/2 and MYC3/4 were further verified by forster resonance energy transfer and fluorescence lifetime imaging microscopy analyses (FRET-FLIM). As shown in Fig. 2b, the CFP fluorescence lifetime of MYC3-CFP was significantly decreased by the PPD1-YFP or PPD2-YFP in N. benthamiana leaves. The CFP fluorescence lifetime of MYC4-CFP was also significantly decreased by the PPD1-YFP or PPD2-YFP in N. benthamiana leaves. The bimolecular fluorescence complementation assays also showed that nYFP-PPD1 and nYFP-PPD2 associated with cYFP-MYC3 and cYFP-MYC4, but not with the cYFP control ( Supplementary  Fig. 6). To determine whether PPD1/2 could directly interact with MYC3/4 in vitro, we performed pull-down analyses. As shown in respectively. Co-immunoprecipitation analyses (co-IP) showed that Myc-PPD1/2 associated with GFP-MYC3/4 but not with the GFP control (Fig. 2d), indicating that PPD1/2 and MYC3/4 form a complex in Arabidopsis. The interactions between Myc-PPD1/2 and GFP-MYC2 were not found by co-immunoprecipitation analyses in Arabidopsis ( Supplementary Fig. 7).
MYC3 and MYC4 function redundantly to regulate seed size. As MYC3/4 could interact with PPD1/2, we asked whether MYC3/4 influence seed size. As shown in Fig. 3a-c, e, myc3 (GK_445B11) and myc4 (GK_491E10) plants produced larger seeds and cotyledons than the wild-type plants, consistent with a previous study 36 . The myc3 and myc4 plants also had heavier seed weight than the wild-type plants (Fig. 3d). The myc3 myc4 double mutant produced larger and heavier and bigger cotyledons than myc3 and myc4 single mutant (Fig. 3a-e), indicating that MYC3 and MYC4 function redundantly to regulate seed size and weight. In addition, overexpression of GFP-MYC3/4 fusion proteins driven by the CaMV 35S promoter in wild-type plants resulted in small and light seeds compared with the wild type ( Supplementary Fig. 10), indicating that MYC3/4 limit seed growth in Arabidopsis.
To investigate whether MYC3/4 function maternally or zygotically to control seed size, we performed reciprocal crossing experiments between the myc3 myc4 and wild type plants. As   The seeds (a) and 8-day-old seedlings (b) of Col-0, myc3, myc4, and myc3 myc4. c-e The relative seed area (c, n = 100), 100 seed weight (d, n = 10), and cotyledon area (e, n = 30) of Col-0, myc3, myc4 and myc3 myc4. Seeds from the third to seventh silique on the stem of six plants were used for analysis. Cotyledons from the 8-day-old seedlings were used for analysis. f, g The relative area of F 1 seeds (f) and F 2 seeds (g) from the Col-0/Col-0 (C/C), Col-0/myc3 myc4 (C/mm), myc3 myc4/Col-0 (mm/C), and myc3 myc4/myc3 myc4 (mm/mm) plants (n = 100). h Ovules of Col-0 and myc3 myc4 plants at 0 DAP (days after pollination). i-l The seed area (i), outer integument length (j), outer integument cell number (k), and outer integument cell length (l) of Col-0 and myc3 myc4 plants at 0, 2, 4, and 6 DAP (n = 33). Ovules and seeds from six siliques, which were from the fourth silique on the stem of six plants, were used for analysis. Scale bars, 0.5 mm (a), 0.2 cm (b), and 50 μm (h). Error bars represent ±SE. Different lowercase letters above the columns indicate the significant difference among different groups, one-way ANOVA Pvalues: P < 0.05. * indicates significant difference from the Col-0, one-way ANOVA P-values: *P < 0.05 and **P < 0.01. shown in Fig. 3f, the F 1 seed area of Col-0 plants pollinated with the pollen of myc3 myc4 plants was similar to that of selfpollinated Col-0 plants, and the F 1 seed area of myc3 myc4 plants pollinated with the pollen of Col-0 plants was comparable to that of self-pollinated myc3 myc4 plants. In addition, the size of Col-0/ myc3 myc4 and myc3 myc4/Col-0 F 2 seeds was similar to that of Col-0/Col-0 F 2 seeds and smaller than that of myc3 myc4/myc3 myc4 F 2 seeds (Fig. 3g). These results indicate that MYC3/4 act maternally to control seed size.
We then examined the development of wild-type and myc3 myc4 outer integuments. myc3 myc4 plants had larger ovules and longer outer integuments than the wild-type plants at 0 DAP ( Fig. 3h-j). The number of cells in myc3 myc4 outer integuments was increased compared with that in wild-type outer integuments, while the outer integument cell length of myc3 myc4 was similar to that of the wild type at 0 DAP (Fig. 3k, l). These results indicate that MYC3/4 limit cell proliferation in the ovules before fertilisation. We further examined cell number and cell length in wild-type and myc3 myc4 outer integuments at 2, 4 and 6 DAP. The outer integument cell number and length of myc3 myc4 were significantly increased compared with those of the wild type at 4 and 6 DAP, thereby resulting in longer outer integument and larger seed size in the myc3 myc4 plants (Fig. 3i-l). These results indicate that MYC3/4 limit both cell proliferation and cell elongation in outer integuments after fertilisation, consistent with the role of the KIX-PPD complex. We further compared the effect of myc3 myc4 on cell proliferation during ovule and seed developmental processes. As shown in Fig. 3k, MYC3/4 restrict cell proliferation in the integuments during both ovule and early seed developmental stages.
The KIX-PPD-MYC complex represses GIF1 expression. We previously reported the SAP-KIX-PPD signalling pathway has an important role in leaf size control 22,23 , and performed the RNAseq analysis using the first pair of leaves of 9-day-old myc3 myc4 and ppd1-2 ppd2-cr seedlings. 149 genes with significantly changed expression were found in both myc3 myc4 and ppd1-2 ppd2-cr plants (Supplementary Data 1). One of them was the transcriptional coactivator GIF1 (GRF-INTERACTING FACTOR 1), which has been reported to control the size of leaves, flowers, seeds, and cotyledons [8][9][10][11]13,37,38 . The expression of GIF1 was significantly upregulated in both myc3 myc4 and ppd1-2 ppd2-cr seedlings (Supplementary Data 1). We also found that expression levels of GIF1 were significantly higher in kix8-1 kix9-1, ppd1-2 ppd2-cr, and myc3 myc4 siliques than those in wild-type siliques at 0, 2, and 4 DAF (days after flowering) (Fig. 4a). By contrast, expression levels of GIF1 were decreased in the 2 DAF siliques of  Fig. 11). In addition, the LUC activity of GIF1pro:LUC was significantly reduced by overexpressing Myc-KIX8/9, Myc-PPD1/2 and Myc-MYC3/4 in the Col-0 protoplast (Fig. 4b). These results indicate that the KIX-PPD-MYC complex represses GIF1 expression. The KIX-PPD module limits leaf development by the repressor TOPLESS (TPL) 21,22 . Overexpression of Myc-TPL also reduced the LUC activity of GIF1pro:LUC in the Col-0 protoplast (Fig. 4b). Plant cis-acting regulatory DNA element analysis showed that there was a typical G-box sequence (5′-CACGTG-3′) at the -425 bp site in the 2 kb promoter region of GIF1 (https:// www.dna.affrc.go.jp/PLACE/?action=newplace) (Fig. 4c). MYCs and PPDs had been reported to associate with the G-box sequence to regulate target gene expression 21,39 . Furthermore, down-regulation of PPDs orthologs in legume Medicago truncatula and legume soybean leads to significant increases in expression of MtGIF1 and GmGIF1 40 . These results imply that GIF1 might be a target gene of the KIX-PPD-MYC repressive complex.
To investigate whether MYC3/4 could directly bind to the Gbox cis-acting element in the promoter of GIF1, we performed the electrophoretic mobility shift assays (EMSA). As shown in Fig. 4e-g, MBP-MYC3 and MBP-MYC4 bound to the biotinlabelled probe A from the GIF1 promoter containing the typical G-box (5′-CACGTG-3′) but not to the mutated biotin-labelled probe A (A-m). The binding ability of MBP-MYC3 and MBP-MYC4 to the probe A was decreased by adding the biotinunlabelled probe A. These results indicate that MYC3 and MYC4 directly bind to the GIF1 promoter.
GIF1 acts maternally to control seed size. The gif1 plants (SALK_150407) produced smaller leaves, seeds, and cotyledons than wild-type plants (Fig. 5a-c, e, and Supplementary Fig. 13a), consistent with previous studies 8, 10,11 . The seed weight of gif1 plants was also significantly lower than that of wild-type plants (Fig. 5d). In addition, the fertility of gif1 plants was lower than that of wild-type plants (Supplementary Fig. 13b). By contrast, overexpression of GIF1 (35S:GIF1) led to bigger and heavier seeds and bigger cotyledons than the wild type (Fig. 5a-e). These results indicate that GIF1 is required for normal seed and other organs development.
To investigate whether GIF1 functions maternally or zygotically to control seed size, we conducted reciprocal crossing experiments between the wild type and gif1. As shown in Fig. 5f, the F 1 seed size of Col-0 plants pollinated with the pollen of gif1 plants was similar to that of self-pollinated Col-0 plants, and the F 1 seed size of gif1 plants pollinated with the pollen of Col-0 plants was comparable to that of self-pollinated gif1 plants. In addition, the size of Col-0/gif1 and gif1/Col-0 F 2 seeds was similar to that of Col-0/Col-0 F 2 seeds and bigger than that of gif1/gif1 F 2 seeds (Fig. 5g). These results support that GIF1 acts maternally to control seed size.
We then investigated the outer integument cell number and cell length before and after fertilisation. The gif1 plants had shorter outer integuments with fewer and shorter cells than the wild type at 0 DAP (Fig. 5h, j-l). We further examined the outer integument cell number and cell length after fertilisation. The gif1 showed shorter outer integuments and smaller seeds than the wild type before 6 DAP (Fig. 5i, j). The outer integument cell number of gif1 was significantly decreased compared with that of the wild type at 2, 4 and 6 DAP (Fig. 5k). The cells in gif1 outer integuments were shorter than those in wild-type outer integuments at 2 DAP, while they were longer than those in wildtype outer integuments at 6 DAP (Fig. 5l), suggesting a compensation phenomenon between cell proliferation and cell elongation 28,41,42 . In addition, we observed that gif1 decreases cell proliferation in outer integuments during both ovule and early seed developmental stages (Fig. 5k).
Partially overlapping expression of KIX-PPD-MYC. As KIX8, KIX9, PPD1, PPD2, MYC3, MYC4 and GIF1 function in a signalling pathway to regulate seed size, we asked whether they have the similar expression patterns during ovule and seed development. To test this, we generated the KIX8pro:KIX8-GFP, KIX9pro:KIX9-GFP, PPD1pro:PPD1-GFP, PPD2pro: PPD2-GFP, MYC3pro:MYC3-GFP, MYC4pro:MYC4-GFP, and GIF1pro:GIF1-GFP transgenic lines. As shown in Fig. 6 and Supplementary Fig. 14, KIX8, KIX9, PPD1, PPD2, MYC3, MYC4, and GIF1 expressed in the integument and chalazal region of ovules before fertilisation. As shown in Fig. 6, PPD1, KIX9 and GIF1 strongly expressed in the nuclei of outer integument cells before 2 DAP and became weak from 3 DAP during seed development. MYC4 strongly expressed in the nuclei of outer integument cells before 3 DAP and became weak from 4 DAP during seed development. PPD2, KIX8, and MYC3 expressed in the nuclei of outer integument cells before 6 DAP. As shown in Supplementary Fig. 14, PPD1, KIX9 and GIF1 also expressed in the chalazal domain cells of seeds before 2 DAP and were not observed at 4 DAP. KIX8, PPD2, MYC3, and MYC4 expressed in the chalazal domain cells of seeds before 4 DAP. In addition, MYC3 and GIF1 expressed in endosperms before 4 DAP. These results indicate that KIX8, KIX9, PPD1, PPD2, MYC3, MYC4 and GIF1 have overlapped expression patterns during ovule development and possess partially overlapped expression patterns during seed development, supporting that they function in a common pathway to control seed size. Moreover, the GFP fluorescence in the epidermal cells of GIF1pro:GIF1-GFP;myc3 myc4 outer integuments was observed at 4 and 5 DAP, which was not observed in GIF1pro:GIF1-GFP plants (Fig. 6), supporting that MYC3/4 repress GIF1 expression. As the KIX-PPD-MYC complex associates with the promoter of GIF1 through MYC3/4 and represses its expression, we asked whether GIF1 act in a common pathway with the KIX-PPD-MYC module to control seed size. We crossed gif1 with myc3 myc4, kix8-1 kix9-1, and ppd1-2 ppd2-cr to generate gif1 myc3 myc4, gif1 kix8-1 kix9-1, and gif1 ppd1-2 ppd2-cr triple mutants, respectively. As shown in Fig. 7a-c, e, the gif1 mutation entirely suppressed the large seed and cotyledon phenotypes of myc3 myc4, indicating that gif1 is epistatic to myc3 myc4 with respect to seed and cotyledon size. Similarly, gif1 is also epistatic to myc3 myc4 with respect to seed weight (Fig. 7d). The large size of seeds and cotyledons of kix8-1 kix9-1 and ppd1-2 ppd2-cr plants was strongly but not entirely suppressed by the gif1 mutation ( Fig. 7a-c, e), indicating that KIX8/9 and PPD1/2 are strongly but not entirely dependent on GIF1 to control seed size. Similarly, the gif1 mutation strongly but not entirely suppressed the seed weight phenotype of kix8-1 kix9-1 and ppd1-2 ppd2-cr plants (Fig. 7d). These genetic analyses indicate that GIF1 acts in a common pathway with the KIX-PPD-MYC module to control seed size and weight. The ppd1-2 ppd2-cr and kix8-1 kix9-1 plants produced wider siliques than wild-type plants. By contract, gif1 plants had narrower siliques than wild-type plants ( Supplementary  Fig. 15). The gif1 mutation entirely suppressed the wide silique phenotype of kix8-1 kix9-1 plants and strongly suppressed the wide silique phenotype of ppd1-2 ppd2-cr plants (Supplementary Fig. 15). In addition, the silique phenotype of gif1 myc3 myc4 was similar to that of gif1 (Supplementary Fig. 15). These genetic analyses indicate that GIF1 acts genetically with the KIX-PPD-MYC module to control silique development. The gif1 and ppd1-2 ppd2-cr plants had lower fertility than wildtype plants (Supplementary Fig. 16). The myc3 myc4 and kix8-1 kix9-1 had similar fertility to wild-type plants, but the fertility of gif1 myc3 myc4 and gif1 kix8-1 kix9-1 was similar to that of gif1 plants (Supplementary Fig. 16). Additionally, the gif1 mutation decreased the fertility of ppd1-2 ppd2-cr plants ( Supplementary Fig. 16). All of the genetic analyses indicate that GIF1 acts with the KIX-PPD-MYC module to control multiple biological processes.
GIF1 and SAP function in a common pathway to control seed size. STERILE APETALA (SAP/SUPPRESSOR OF DA1, SOD3) acts as a part of the E3 ubiquitin ligase complex to control organ size by regulating the stability of the KIX-PPD complex 22,23 . The sod3-1 mutants produce small leaves, while 35S:SAP plants have large leaves 22,23 . We found that 35S:SAP plants also produced bigger seeds and cotyledons than the wild-type plants (Fig. 7a-c, e). Similarly, 35S:SAP plants produced heavier seeds than wild-type plants (Fig. 7d). The gif1 mutation strongly suppressed the large seed and cotyledon phenotype of 35S:SAP plants (Fig. 7a-c, e). The gif1 mutation also strongly suppressed the heavy seed phenotype of 35S:SAP plants (Fig. 7d). Additionally, 35S:SAP plants produced wider siliques than wild-type plants, while it was significantly suppressed by the gif1 mutation ( Supplementary  Fig. 15). Although 35S:SAP plants had similar fertility to wildtype plants, the fertility of 35S:SAP;gif1 plants was similar to that of gif1 (Supplementary Fig. 16). Moreover, the GIF1 expression was obviously higher in the 3 DAP siliques of 35S:SAP plants than that in wild-type plants (Supplementary Fig. 17). These data indicate that GIF1 acts as a downstream factor of SAP to control seed size.

Discussion
Seed size is the key agronomic trait that greatly determines the grain yield of plants. Although several factors have been reported to affect seed size in plants 4,5 , the genetic and molecular mechanisms that determine seed size remain elusive.
In this study, we discover a genetic and molecular mechanism that the transcription factors MYC3/4 recruit the repressor complex KIX8/9-PPD1/2 to the promoter of GIF1 and repress its expression, thereby determining seed size in Arabidopsis. The seeds (a) and 8-day-old seedlings (b) of Col-0, gif1, myc3 myc4, gif1 myc3 myc4, kix8-1 kix9-1, gif1 kix8-1 kix9-1, ppd1-2 ppd2-cr, gif1 ppd1-2 ppd2-cr, 35S:SAP, and 35S:SAP;gif1 plants. c-e The relative seed area (c, n = 100), 100 seed weight (d, n = 10), and cotyledon area (e, n = 30) of Col-0, gif1, myc3 myc4, gif1 myc3 myc4, kix8-1 kix9-1, gif1 kix8-1 kix9-1, ppd1-2 ppd2-cr, gif1 ppd1-2 ppd2-cr, 35S:SAP, and 35S:SAP;gif1 plants. f The TPL-KIX-PPD-MYC complex associates with the G-box sequence of GIF1 promoter and represses its expression. g The transcriptional repression of GIF1 is relieved by SAP modulating the KIX-PPD module for 26S proteasome degradation. Without the KIX-PPD complex, the binding ability of MYC3/4 with the promoter of GIF1 is decreased and the GIF1 expression is increased. Seeds from the third to seventh silique on the stem of six plants were used for analysis. Cotyledons from the 8-day-old seedlings were used for analysis. Scale bars, 0.5 mm (a) and 0.2 cm (b). Error bars represent ±SE. Different lowercase letters above the columns indicate the significant difference among different groups, one-way ANOVA P-values: P < 0.05. Previous studies reported that PPD1 and PPD2 act redundantly to regulate leaf size and shape by influencing both the primary and the secondary mitotic arrest fronts 20,21,23 . Considering that ppd2-1 had large seeds, while ppd1-2 did not obviously affect seed size, ppd1-2 ppd2-1 double mutant will help understand the role of PPD1/2 in seed size control. However, the PPD1 gene (AT4G14713) is close to the PPD2 gene (AT4G14720) in the chromosome, we could not isolate ppd1-2 ppd2-1 double mutant. We therefore generated the ppd2-cr mutation in the ppd1-2 mutant background and the ppd1-cr mutation in the ppd2-1 mutant background to obtain the ppd1-2 ppd2-cr and ppd1-cr ppd2-1 double mutants using the CRISPR-Cas9 technology, respectively (Supplementary Fig. 1) 27 . The ppd1-2 ppd2-cr and ppd1-cr ppa2-1 mutants produced larger and heavier seeds than ppd1-2 and ppd2-1 single mutants (Fig. 1a, c, d), indicating that PPD1 and PPD2 function redundantly to control seed size and weight. KIX8/9 interact with PPD1/2 to form a transcriptional repressor complex and control leaf size 21,22 . However, it is unclear whether the KIX-PPD complex affects seed growth in Arabidopsis. Here, we found that kix8-1 kix9-1 ppd1-2 ppd2-cr plants produced significantly larger seeds than the wild type (Fig. 1a, c). Reciprocal crossing experiments showed that the KIX-PPD complex functions maternally to control seed size (Fig. 1f,  g). Cellular observation indicated that the KIX-PPD complex predominantly represses cell proliferation and also slightly limiting cell elongation in the integuments (Fig. 3k, l). These results reveal that the KIX-PPD complex negatively regulates seed growth in Arabidopsis.
The transcriptional repressor complex usually interacts with the transcription factors to regulate gene expression 43,44 . We found that PPD1/2 could directly interact with the transcription factors MYC3/4 in vitro and in vivo, but not interact with MYC2. Although MYC3 and MYC4 share lots of overlapping functions with MYC2, distinct functions among them have been reported. For instance, MYC3 and MYC4 recognise similar cis-acting sequences (i.e. G-box and its variants) to MYC2, while the DNAbinding affinity of MYC3 and MYC4 differs from that of MYC2. MYC2 and MYC4 but not MYC3 interact with the JAZ4 protein.
The expression levels of VEGETATIVE STORAGE PROTEIN 2 (VSP2) and PLANT DEFENSIN 1.2 (PDF1.2), two of JA marker genes, are significantly different in myc2, myc3, and myc4 single mutants when treated with JA 33,34,45 . We further reveal that KIX8/9, PPD1/2, and MYC3/4 can form a complex in Arabidopsis (Fig. 2). Like kix8-1 kix9-1 and ppd1-2 ppd2-cr mutants, myc3 myc4 mutants produced bigger seeds than the wild type (Fig. 3a, c), consistent with a previous study 36 , further suggesting that MYC3/4 have the overlapped function with KIX8/9 and PPD1/2 in seed size control. Reciprocal crossing experiments indicate that MYC3/4 act maternally to limit seed growth. Cellular observations show that MYC3/4 influence both cell proliferation and cell elongation in the integuments, consistent with the role of the KIX-PPD complex in seed growth control. Therefore, the KIX-PPD-MYC module is crucial for seed size control in Arabidopsis.
To identify the targets of the KIX-PPD-MYC module in seed growth, we performed the RNA-seq and found that PPD1/2 and MYC3/4 repress the expression of GIF1. The expression levels of GIF1 in 0, 2, and 4 DAF siliques from kix8-1 kix9-1, ppd1-2 ppd2cr and myc3 myc4 plants were significantly higher than those of wild-type plants (Fig. 4a). Consistent with this, overexpression of Myc-KIX8/9, Myc-PPD1/2, Myc-MYC3/4, and Myc-TPL could reduce the activity of GIF1pro:LUC (Fig. 4b). In addition, EMSA experiments showed that MYC3 and MYC4 directly bind to the G-box sequence in the promoter of GIF1 (Fig. 4f, g). ChIP-qPCR analyses showed that KIX8/9 and PPD1/2 associate with the promoter of GIF1 through MYC3/4 (Fig. 4d). These findings indicate that MYC3/4 may recruit the transcriptional repressor complex TPL-KIX-PPD to the promoter of GIF1 to repress its expression (Fig. 7f). Overexpression of KIX8, KIX9, PPD1, PPD2, MYC3, or MYC4 led to small seeds (Supplementary Figs. 5 and 6), consistent with the result of PPD1OE (PPDOE) 19 . It is possible that overexpression of KIX8, KIX9, PPD1, PPD2, MYC3, or MYC4 in Arabidopsis might have more probability to form the TPL-KIX-PPD-MYC complex that represses the expression of GIF1, thereby resulting in small seeds. Interestingly, MYC proteins recruit the TPL-NINJIA-JAZ transcriptional repressor complex to regulate gene expression 43,46,47 . Overexpression of JAZ13 alone attenuates JA-induced defence responses in Arabidopsis leaves. Overexpression of NINJIA promotes root length when treated with MeJA 43 . Overexpression of MYC2, MYC3, or MYC4 accelerates JA-induced leaf senescence 46 . These results indicate that overexpression of the single complex components can cause phenotypes. The expression of GIF2 and GIF3 in the 3 DAP siliques of kix8-1 kix9-1, ppd1-2 ppd2-cr, and myc3 myc4 plants was also upregulated compared with that in the wild type ( Supplementary Fig. 18). Down-regulation of PPDs orthologs in legume Medicago truncatula and legume soybean leads to high expression of MtGIF1 and GmGIF1 in leaves, stipules, and seeds, respectively 40 . These results indicate that the expression of GIFs regulated by PPDs might be a common mechanism in dicotyledon plants. The loss-of-function mutation in GIF1 produced smaller seeds and cotyledons than the wild type ( Fig. 5a-c, e), consistent with previous studies 10,11 . The gif1 mutant has fewer cells and longer cells in the integuments than the wild type (Fig. 5k, l), suggesting a compensation mechanism between cell proliferation and cell elongation. This phenomenon has been observed in several seed size mutants 28,41,42 . In addition, GIF1 was reported to play significant roles in leaf, flower, and root development in Arabidopsis [8][9][10][11][12][13][14] , indicating that GIF1 is required for normal plant organ growth. Surprisingly, a previous study showed that one mutant allele of GIF1 (an3) promotes seed growth 48 . In this study, we have sufficient evidence to support that GIF1 is a positive regulator of seed size in Arabidopsis. For example, loss-of-function of GIF1 formed small seeds, while overexpression of GIF1 produced large seeds (Fig. 5a, c). The gif1 mutation completely suppresses the large and heavy seed phenotypes of myc3 myc4 (Fig. 7c, d). By contrast, the gif1 mutation strongly but not entirely suppresses the seed size and weight phenotypes of kix81-1 kix91-1 and ppd1-2 ppd2-cr (Fig. 7c, d), implying that KIX8/9 and PPD1/2 might have other mechanisms that act independently of GIF1 to control seed development. These genetic analyses also reveal that GIF1 functions in a common pathway with the KIX-PPD-MYC module to control seed size in Arabidopsis. Consistent with this, KIX8, KIX9, PPD1, PPD2, MYC3, MYC4 and GIF1 have overlapped expression patterns during ovule development and possess partially overlapped expression patterns during early seed developmental stages ( Fig. 6 and Supplementary Fig. 14).
STERILE APETALA (SAP/SUPPRESSOR OF DA1, SOD3) acts as a part of the E3 ubiquitin ligase complex to control organ size by regulating the stability of the PPD-KIX complex 22,23 . The expression of GIF1 was obviously higher in 35S:SAP plants than that in wild-type plants ( Supplementary Fig. 17), indicating that the transcriptional repression of GIF1 is released by the F-box protein SAP that modulates the KIX-PPD complex for 26S proteasome degradation (Fig. 7g). However, SAP did not modulate the stability of MYC3/4 proteins (Supplementary Fig. 19). Without the KIX-PPD complex, MYC3/4 could bind to the promoter of GIF1, but the binding ability is significantly decreased (Fig. 4d). Genetic analyses showed that the gif1 mutation strongly suppresses the large and heavy seed phenotypes of 35S:SAP (Fig. 7c, d), suggesting that SAP and GIF1 act in a common pathway to regulate seed growth. Based on these genetic and biochemical analyses, we build up a genetic and molecular framework for the SAP-KIX-PPD-MYC-GIF1 modulemediated control of seed size and weight in Arabidopsis (Fig. 7f, g).
Seed size is one of the important targets for plant breeding. In this study, we found that the KIX-PPD-MYC-GIF1 pathway is crucial for seed size control in Arabidopsis. Interestingly, loss-offunction of PPD orthologs in legume Medicago truncatula and legume soybean increases seed size and weight as well as leaf size 40 . In pea, mutations in the PPD ortholog ELEPHANT-EAR-LIKE LEAF 1 or the KIX ortholog BIGGER ORGANS cause large flowers and leaves 49 . In rice, overexpression of OsGIF1 results in large grains, leaves, and stems, while suppression of OsGIF1 leads to small grains and organs 16,17 . These findings suggest that the KIX-PPD-MYC-GIF1 pathway may possess a conserved function in different plant species. Thus, it will be interesting to investigate the roles of the KIX-PPD-MYC-GIF1 pathway in crops and utilise their homologues to improve seed size in key crops.
The CDS of GIF1 was obtained from the total RNA of the Col-0 plants with FastQuant RT Super Mix kit (TIANGEN, KR108) and cloned into the Kpn I and Spe I sites of pMDC32 vector to generate pMDC32-35S:GIF1 constructs with EZfusion kit (Genera, GR6086). pMDC32-35S:GIF1 constructs were transferred into the Col-0 plants by agrobacterium tumefaciens-mediated transformation 50 . 35S:GIF1 transgenic plants were screened out with 30 μg ml −1 hygromycin.
Seeds were sterilised with ethanol (75% v/v) for 3 min, bleach (10% v/v) for 15 min, and washed with sterilised water three times, and then plated to Murashige and Skoog (MS) medium. After storing in dark for 4 days at 4°C, seeds were grown at 22°C with 16 h light (28 W/6500 K)/8 h dark.
Morphological and cellular analysis. Seeds were harvested from the third to seventh silique on the stem of plants. Cotyledons were harvested from the 8-day-old seedlings. Siliques were harvested at 14 DAF (days after flowering). Parameters of seeds, cotyledons, and siliques were measured by ImageJ software after photographing. For seed weight, 100 seeds were weighed at each experiment by Mettler Toledo XP6 (Mettler Toledo, Switzerland). As for seed integument observation, the stamens in the fourth flower on the stem of plants were removed before flowering. The plants were pollinated with their own pollens. The seeds were harvested at 0, 2, 4 and 6 DAP (days after pollination). Ovules from six siliques, which were from the fourth silique on the stem of six plants, were used for analysis. At least five representative ovules were analysed. Seeds from six siliques, which were from the fourth silique on the stem of six plants, were used for analysis. At least five representative seeds in a silique were used for analysis. Seeds were firstly cleared with FAA solution (90 mL 70% ethanol, 5 mL acetic acid, and 5 mL 37% formaldehyde), and then dealt with Hoyer's solution (7.5 g gum arabic, 100 g chloral hydrate, 5 mL glycerol, 5 mL phenol, and 25 mL water) before used to observation under the differential interference contrast microscope (DIC, Leica DM2500). The integument cell number of seeds was counted in DIC. The integument length of seeds was matured by ImageJ.