GSK2399872A | Mll Signal

Control of grain size and rice yield by GL2-mediated brassinosteroid responses
Ronghui Che1†, Hongning Tong1†, Bihong Shi2, Yuqin Liu3, Shanru Fang3, Dapu Liu1, Yunhua Xiao1, Bin Hu1, Linchuan Liu1, Hongru Wang1, Mingfu Zhao3* and Chengcai Chu1*

Given the continuously growing population and decreasing arable land, food shortage is becoming one of the most serious global problems in this century1. Grain size is one of the determining factors for grain yield and thus is a prime target for genetic breeding2,3. Although a number of quantitat- ive trait loci (QTLs) associated with rice grain size have been identiﬁed in the past decade, mechanisms underlying their functions remain largely unknown4,5. Here we show that a grain-length-associated QTL, GL2, has the potential to improve grain weight and grain yield up to 27.1% and 16.6%, respectively. We also show that GL2 is allelic to OsGRF4 and that it contains mutations in the miR396 targeting sequence. Because of the mutation, GL2 has a moderately increased expression level, which consequently activates brassinosteroid responses by upregulating a large number of brassinosteroid- induced genes to promote grain development. Furthermore, we found that GSK2, the central negative regulator of rice brassinosteroid signalling, directly interacts with OsGRF4 and inhibits its transcription activation activity to mediate the speciﬁc regulation of grain length by the hormone. Thus, this work demonstrates the feasibility of modulating speciﬁc brassinosteroid responses to improve plant productivity.
To identify a grain-size-associated locus, we developed an F2 population from a cross between two indica rice varieties, BobaiB (BBB) and RW11, with signiﬁcant difference in grain size (Fig. 1a; Supplementary Figs 1 and 2). In contrast to BBB with a 1,000- grain weight of 19.3 g, RW11 has extra-large grains with a 1,000- grain weight of 40.2 g and a grain length 36.6% larger than BBB (Supplementary Fig. 1). Genetic investigation of 186 recurrent back- cross (BC3F2) individuals indicated that the long grain phenotype is controlled by a single semi-dominant locus (Fig. 1b). Using 899 BC3F2 short-grain progeny, the QTL was mapped to an approximately 700 kb region between two molecular markers, RM13792 and RM8248, on chromosome 2 and was designated GL2 (Fig. 1c). Because of the lack of polymorphic markers in this region between the two indica parents, we constructed a new BC3F2 mapping population by crossing RW11 with a japonica small grain variety Nipponbare (1,000-grain weight, 20.06 ± 1.14 g). Using a total of 3,891 individuals, the gene was eventually narrowed down to a 21 kb region between two markers RM13838 and GLS-91, which contains only one candidate gene, LOC_Os02g47280 (Fig. 1c).
Near-isogenic line carrying GL2 (NIL-GL2, BC9F2) plants have grains 24.2% longer and 15.8% wider than BBB (Fig. 1d,e), suggesting that GL2 regulates both the grain length and the grain width; the grain ﬁlling rate was also greatly increased (Supplementary Fig. 3), resulting in a signiﬁcant increase (27.1%) of the 1,000-grain weight (Fig. 1e). To verify the cloning result, a complementary vector containing

the whole gene derived from RW11 (a 6,399 bp genomic DNA including the 2,600 bp promoter region) was introduced into Nipponbare. Transgenic plants had obvious increases in grain length and weight to different extents, and the changes were consist- ent with the gene expression levels (Fig. 1f–h), demonstrating that the gene is responsible for the elongated grain phenotype.
This LOC_Os02g47280 gene encodes OsGRF4 (Growth Regulating Factor 4), belonging to a plant-speciﬁc family of transcription factors (Supplementary Fig. 4), members of which have been suggested to be involved in regulating cell proliferation and size in Arabidopsis6–10. The protein is speciﬁcally localized to the cell nucleus (Supplementary Fig. 5), and the gene is preferentially expressed in young panicles, with higher expression in the separated grain husk, but with much lower expression in other tissues (Fig. 2a; Supplementary Fig. 6). Importantly, compared with BBB, NIL-GL2 has an obviously elevated level of OsGRF4 transcripts in all the tissues tested; in young panicles with the highest GRF4 expression, the increase is approximately threefold (Fig. 2a).
Sequence analysis revealed multiple polymorphisms in the gene promoter and introns among the three mapping parents, whereas in the coding region, only a 2 bp substitution was identiﬁed. The sub- stitution leads to an amino acid change from serine in both BBB and Nipponbare to lysine in RW11 (Fig. 1c). It is well documented that the growth regulating factor (GRF) genes are targets of miR3967,8, and OsGRF4 contains a putative miR396 recognition sequence
(Fig. 1c; Supplementary Fig. 7). An RLM-RACE (5′ RNA ligase-
mediated rapid ampliﬁcation of cDNA ends) analysis showed that
miR396 could directly cleave OsGRF4 mRNA in vivo at the site that pairs with the 11th nucleotide of the 5′ end of the miR396 pairing region (Fig. 2b). Notably, the 2 bp substitution located in
the putative miR396 binding sequence of GL2 would perturb the cleavage by miR396, resulting in increased expression of GRF4.
Overexpression of the RW11-derived GL2 allele leads to an obviously enlarged grain size corresponding to the OsGRF4 expression and protein levels (Fig. 2c–e). Because the mutation also made an amino acid change, to conﬁrm that the perturbation of miR396 regulation on OsGRF4 is the cause of large grain size, an miR396-resistant variant of OsGRF4 that disrupts the miR396 recognition but does not cause an amino acid change (designated rGRF4, Fig. 2f) was introduced into Nipponbare under the control of a CaMV 35S promoter. The transgenic plants had an obviously larger grain size corresponding to the gene expression level (Fig. 2g,h), suggesting that the abolished transcriptional regu- lation of OsGRF4 by miR396, but not the amino acid alteration, is the cause of the large grain phenotype.
NIL-GL2 had a signiﬁcantly enlarged cell volume, as demon-
strated by both histological sectioning analysis and observation by

1 State Key Laboratory of Plant Genomics, National Centre for Plant Gene Research (Beijing), Institute of Genetics and Developmental Biology, Chinese Academy of Sciences, Beijing 100101, China. 2 College of Life Sciences, Fujian Normal University, Fuzhou 350007, China. 3 Rice Research Institute, Fujian Academy of Agricultural Sciences, Fuzhou 350018, China. †These authors contributed equally to this work. *e-mail: [email protected]; [email protected]
NATURE PLANTS | VOL 2 | JANUARY 2016 | www.nature.com/natureplants 1

a c
Markers

Chr. 2 35
Markers

18 0 4 9

82 90
n = 899

1 cM

b 25

BBB RW11

30 18 9 4 3

1 01
21 kb

2 2 3 6

n = 3,891
50 kb

20
15
10 ATG
5
0

GL2

miR396 binding sequence

LOC_Os02g47280

TGA

Grain length (mm)

A C A G T

T A AA (RW11-GL2)

C T del. T C

d e P = 1.71 × 10−18
12

3.5
3.0
2.5
2.0
1.5
1.0
0.5
0.0

P = 2.06 × 10−14

P = 6.54 × 10−10
30

BBB NIL-GL2

BBB NIL-GL2 BBB NIL-GL2

f g h

NP
8
C1
C2 4
C3
0

NP C1 C2 C3

0
NP C1 C2 C3

800

600

400

200

NP C1 C2 C3

Figure 1 | Cloning and veriﬁcation of GL2. a, Grain morphology of BBB and RW11. b, Distribution of plants with different grain length in the BC3F2 population. c, Map-based cloning of GL2. d, Grain morphology of BBB and NIL-GL2. e, Statistical data of grains of BBB and NIL-GL2. f, Grain morphology of NP and the three complementary lines. g, Statistical data of the indicated plants. h, Expression level of GRF4 in the plants. Values are shown as means ± s.d. (n = 30 in e,g; n =3 in h). P values were calculated by the Student’s t-test.

scanning electron microscopy of the grain husk (Fig. 2i–k). In addition, although NIL-GL2 had obviously increased seedling height, no change was observed in the plant height at the mature stage (Fig. 3a,b; Supplementary Table 1). These features to a large extent resembled those of plants with enhanced brassinosteroid bio- synthesis (for example m107) or signalling (for example Gi-2)11–13 (Supplementary Fig. 8). Consistently, physiological analyses of coleoptile tissues and of the leaf sheath revealed that NIL-GL2 had an obviously increased sensitivity to brassinolide, a bioactive brassinosteroid, but tended to be resistant to brassinazole (BRZ), an inhibitor of brassinosteroid biosynthesis (Fig. 3c; Supplementary Fig. 9a). In addition, both the coleoptile and the internode elongated much longer in NIL-GL2 than in BBB in darkness (Supplementary Fig. 9b). Further analysis of three rGRF4 overexpression lines

revealed that the plants had gradually enhanced brassinosteroid sen- sitivities together with increased gene expression, excluding the possible effect of other genes in the NIL background (Fig. 3d). These results demonstrated that GL2 is involved in brassinosteroid signalling and that NIL-GL2 has constitutively activated brassinosteroid responses.
Consistent with our very recent report that brassinosteroid stimulates cell elongation by promoting gibberellin (GA) biosyn- thesis in rice seedlings11, the GA biosynthetic genes, including GA20ox-1, GA20ox-2 and GA3ox-2, showed increased expression, whereas the GA inactivation gene GA2ox-3 showed decreased expression in NIL-GL2 compared with BBB (Fig. 3e). Further GA quantiﬁcation conﬁrmed the increased GA1 level in NIL-GL2 seed- lings (Supplementary Table 2). In contrast, both the expression of

2 NATURE PLANTS | VOL 2 | JANUARY 2016 | www.nature.com/natureplants

a 1.2

0.8

b
RLM-RACE

GRF4 5’

CCGTTCAAGAAAGCCTGTGGAAA 3’

0.4

0.0

<1 1

4 8 12 16 20

GL2

3’ GCAAGUUCUUUCGGACACCUU 5’ miR396d

5’ CCGTAAAAGAAAGCCTGTGGAAA 3’

c
NP OE1 OE2 OE3

Panicle length (cm)

f rGRF4 5’
GRF4 5’ NP
rOE1 rOE2 rOE3

CGCTCCAGGAAACCGGTCGAG 3’
* * * * * * *
CGTTCAAGAAAGCCTGTGGAA 3’

d e g h
450

8.6
8.2
7.8
7.4
7.0
6.6

NP OE1 OE2 OE3

300

150

α-Flag

NP OE1 OE2 OE3

5
NP 1 2 3

rOE

400
300
200
100
0

NP 1 2 3

rOE

i j P = 2.11 × 10−23
10
8
6
4
2
0

600

400

200

BBB
k

NIL-GL2

Palea Lemma

Figure 2 | GL2 escapes from miR396 suppression to regulate cell volume. a, GRF4 expression pattern. b, miR396 cleavage site and GRF4 mutation sites. c–h, Grain morphology, statistical data and GRF4 expression in GL2 overexpression plants (c–e) and in plants transformed with miR396-resistant GRF4 (f–h). i, Cross-sections of grain husk. j, Statistical data of the cell length and cell number. k, Scanning electron microscopic observation of spikelet lemma. Values are shown as means ± s.d. (n =3 in a,e,h; n = 30 in d,g; n = 15 in j). P values were calculated by the Student’s t-test.

brassinosteroid biosynthetic genes and the brassinosteroid content showed no signiﬁcant changes (Supplementary Fig. 10). In addition, a GA sensitivity test showed that the response of the second leaf sheath elongation to GA was the same in NIL-GL2 and BBB (Supplementary Fig. 11). Taken together, these results demonstrated that GL2 functions by activating brassinosteroid responses, but not brassinosteroid biosynthesis or GA signalling.
Very importantly, however, NIL-GL2 did not exhibit a loose plant structure like m107 and Gi-2 which have enlarged leaf angles. Instead, the NIL-GL2 plant seedlings had an obviously impaired sensitivity in the brassinosteroid-induced lamina bending

response (Fig. 3f). This is consistent with our recent report that increased GA biosynthesis as one of the downstream brassinosteroid responses will inhibit the brassinosteroid-induced lamina bending process in a feedback manner11. Moreover, our previous study suggested that enhanced brassinosteroid biosynthesis or signalling could induce GA inactivation to inhibit plant growth11. Consistent with this speculation, in young panicles, NIL-GL2 had decreased levels of both GA1 and GA4 compared with BBB (Supplementary Table 3) as well as increased expression of the GA inactivation genes (Supplementary Fig. 12), which explained the unaltered plant height of NIL-GL2 at the ﬁnal mature stage.

NATURE PLANTS | VOL 2 | JANUARY 2016 | www.nature.com/natureplants 3

a b c

1.8

1.2

0.6

BBB NIL-GL2 BBB NIL-GL2

0.0
0 −9 −8 −7

−6 −5

BL concentration (log M)

d 1.8

1.2

0.6

0.0

e
4

0
0 −8 −6
BL concentration (log M)

f 120

0 10

100 1000
BL (ng)

g h i

GL2

Expansin Extensin

Kinesin

GL2

BL m107 (log2)

300
250
200
150
20

Figure 3 | GL2-activated brassinosteroid responses. a–b, Phenotypes of BBB and NIL-GL2 at one week old (a) and at reproductive stage (b). c, Coleoptile elongation of BBB and NIL-GL2 in response to brassinolide. d, brassinolide response of rGRF4 overexpression plants. e, Expression level of GA metabolic genes. f, Lamina inclination in response to brassinolide. g, Overlapping DEG numbers between NIL-GL2/BBB and m107/NP. h, Heatmap of ‘expansin’, ‘extensin’ and ‘kinesin’ DEGs. i, Expression of the ‘expansin’ genes by qRT–PCR. Values are shown as means ± s.d. (n = 12 in c,d,f; n =3 in e,i). *P < 0.05;
**P < 0.01 (t-test).

To explore the GL2 functional mechanism further, we performed RNA-sequencing analysis to compare the global transcriptional pro- ﬁles of BBB and NIL-GL2 using both seedlings and young panicles as materials. Consistent with the expression of OsGRF4 in young panicles, 3,934 DEGs (differential expressed genes, ratio ≥2) were identiﬁed in panicle tissues, whereas only 214 DEGs were identiﬁed in seedlings (Supplementary Fig. 13a). In both tissues, the number of the upregulated genes was about 1.8 times the number of the downregulated genes. Enrichment analysis of panicle DEGs ident- iﬁed ‘plant hormone signal transduction’ as the top annotated pathway (Supplementary Fig. 13b). Notably, 120 (55.8%) of the seedling DEGs and 1,401 (35.7%) of the panicle DEGs were also differentially regulated in the seedlings of m107 compared with the wild-type Nipponbare (Fig. 3g; Supplementary Fig. 14a). Similarly, 112 (52.1%) and 990 (25.2%) DEGs of two tissues

overlapped with those of the brassinolide-treated plant seedlings (Supplementary Fig. 14b).
Signiﬁcantly, there were eight DEGs encoding ‘expansin’-like proteins and seven encoding ‘extension’-like proteins that were
upregulated in both NIL-GL2 and m107 (Fig. 3h; Supplementary Table 4). We veriﬁed the increased expression of the ‘expansin’ genes in NIL-GL2 by quantitative real-time polymerase chain reaction (qRT–PCR) (Fig. 3i). These results strongly suggested that GL2 activates the expression of these genes to promote cell expansion and grain size. Consistent with the idea that brassinoster- oid regulates microtubule reorientation to facilitate cell growth14,15, there were 26 ‘kinesin’ genes induced by both GL2 and brassinoster- oid (Fig. 3h; Supplementary Table 4). Kinesin-like proteins are involved in microtubule dynamics, and some of the proteins have been demonstrated to regulate cell elongation and seed length16–18.

4 NATURE PLANTS | VOL 2 | JANUARY 2016 | www.nature.com/natureplants

Vector

BD + AD

Leu-Trp-

Leu-Trp- Ade-His-

BD + AD

Leu-Trp-

Leu-Trp- Ade-His-

GRF4-F 1
GRF4-Q 1
GRF4-W GRF4-T GRF4-C
GRF4-ΔT 1
GRF4-ΔQ

120

119

259

258

379

259

394

GSK2 + AD BD + GRF4 GSK2 + GRF4
c

YFPc + YFPn−GRF4

YFPc−GSK2 + YFPn−GRF4

YFP

YFP + BF

d e

IP:
α–Flag

Input

Flag + GSK2−GFP

f
45
GRF4−Flag + 40
GSK2−GFP 35
30
α–GFP 25
20
–GFP 15
10
5
α–Flag 0

30% input GST pull-down, a-MBP
Figure 4 | GSK2 interacts with GRF4 and inhibits its activity. a, GRF4 transcription activation analysis in yeast. QLQ (Gln, Leu, Gln) and WRC (Trp, Arg, Cys) domains are two highly conserved regions in GRF proteins. b, GSK2 interacts with GRF4 in yeast. c, Bimolecular ﬂuorescence complementation (BiFC) analysis of GSK2–GRF4 interaction in tobacco leaf. BF, bright ﬁeld. d, GST pull-down analysis of GSK2–GRF4 interaction. e, Detection of GSK2–GRF4 interaction by co-immunoprecipitation analysis. GFP-tagged GSK2 and Flag-tagged GRF4 were expressed in tobacco leaves for the analysis. f, Luciferase reporter analysis of GRF4 activity and effect of GSK2 on the activity (means ± s.d. are shown; n = 3). ***P < 0.001 (t-test). ns, non-signiﬁcant.

Combined with the increased brassinosteroid sensitivity of NIL- GL2, these results strongly suggested that GL2 promotes grain size by activating brassinosteroid responses.
OsGRF4 has transcription activation activity, and the activation domain is located in its carboxy (C)-terminal region according to the analysis in yeast (Fig. 4a). We deleted the C-terminal activation domain of GRF4 and used the truncated protein (GRF4-ΔT) as bait to perform a yeast two-hybrid screening. Interestingly, we obtained a number of candidate interacting proteins, including both GSK2 and SLR1 (SLENDER RICE1) as well as three GIF (GRF-interacting factor) family members that have been reported to interact with GRF proteins10,19. The interactions were further conﬁrmed using the full-length cDNAs in yeast (Fig. 4b; Supplementary Fig. 15). GSK3- like kinases, including Arabidopsis BIN2 (BRASSINOSTEROID- INSENSITIVE2) and rice GSK2, were reported to phosphorylate a number of transcription factors to regulate downstream brassinoster- oid responses, and GSK2 knockdown plants have greatly enlarged grain size (Supplementary Fig. 8b). Indeed, BiFC (bimolecular ﬂuorescence complementation) and glutathione S-transferase (GST) pull-down analyses demonstrated that GSK2 can strongly interact with GRF4 (Fig. 4c,d; Supplementary Fig. 16), and the interaction was further conﬁrmed by co-immunoprecipitation analysis (Fig. 4e; Supplementary Fig. 17). Moreover, consistent with the central negative role of GSK2 in regulating brassinosteroid signalling and grain size, luciferase reporter analysis showed that GSK2 could

repress both GRF4 and GL2 transcription activation activity to a large extent (Fig. 4f).
Furthermore, by crossing a GSK2-1ox line13, Go-3, to NIL-GL2, we found that introduction of GSK2 into NIL-GL2 can largely sup- press its large grain size, as the Go-3 NIL-GL2 showed a very similar grain size to Go-3, demonstrating that GSK2 inhibits GL2 function in plants (Supplementary Fig. 18). Although how GSK2 biochemi- cally inhibits GL2 remains unclear in detail, these results strongly suggested that GSK2 interacts with GL2 and inhibits its function.
In summary, we propose that wild-type GRF4 was suppressed at the transcription level by miR396 and at the protein level by GSK2 (Supplementary Fig. 19a). The GL2 allele escaped from the suppres- sion by miR396 but was still repressed by GSK2, leading to a subtle but moderate increase in GRF4 activity combined with the preferen- tial expression of GRF4 in panicles, resulting in a strikingly increased grain size (Supplementary Fig. 19b). Thus, GL2, as one of the GSK2 substrates, could function through activating speciﬁc brassinosteroid responses to regulate grain length.
Consistent with the roles of brassinosteroid in regulating many important agronomic traits20, NIL-GL2 has many beneﬁcial traits including rapid germination (Supplementary Fig. 20a), enhanced growth of source leaves (Supplementary Fig. 20b,c), a rapid grain ﬁlling rate, unaltered plant architecture and seed number, and sig- niﬁcantly increased grain size and weight. Abundant studies have demonstrated the roles of GRFs in leaf growth21–23, and the

NATURE PLANTS | VOL 2 | JANUARY 2016 | www.nature.com/natureplants 5

signiﬁcantly increases of the source leaves rendered by GL2 could improve grain ﬁlling to sustain the enlarged sink size. Despite the decreased seed-setting, statistical analysis showed that the grain yield per plant of NIL-GL2 is 16.6% greater than that of BBB (Supplementary Table 1). A QTL pyramiding combining GL2 and GS324 showed that the two loci had an additive effect on cell elongation and grain length (Fig. 2k; Supplementary Fig. 21). Coincidentally, a recent article has reported the cloning of the same locus from a landrace25. Despite this, extensive sequence analysis of GRF4 in 203 rice minicore accessions worldwide and 517 landraces26 failed to identify a similar polymorphism in the miR396 recognition sequences, implying a tremendous potential for GL2 in hybrid breeding for further improvement of rice yield.

Methods
Plant materials and cultivation. Rice plants were cultivated under ﬁeld
conditions at three different experimental stations located in Shaxian (Fujian Province, 26°41′N), Lingshui (Hainan Province, 18°30′N) and Beijing (40°06′N). For physiological analyses, culture of rice seedling was performed in a growth
chamber at 30 °C under 10 h/14 h light/dark cycle using 1/2 MS as nutrient source. All chemical treatment assays in this study were performed as previously described in detail11.

Map-based cloning. A list of the markers used for QTL analysis and positional cloning is given in Supplementary Table 5.

Microscopy observation. Brieﬂy, the young spikelet hulls were ﬁxed, dehydrated and embedded in parafﬁn, and 10 µm thick sections were obtained. Cell number and cell length in the outer parenchyma layer of the spikelet hulls were measured by Olympus stream software. For scanning electron microscopy observation, the outer surfaces of spikelet glumes were observed with a scanning electron microscope
(S-3000N; Hitachi) after pre-treatment as previously described27. A confocal
ﬂuorescence microscope (Leica TCS SP6) was used to observe ﬂuorescence.

RNA extraction, cDNA preparation and qRT–PCR. Total RNAs were isolated using the TRIzol reagent (Invitrogen). cDNAs were synthesized using a reverse transcription kit (TOYOBO). qRT–PCR experiments were carried out using Chromo4 real-time PCR detection system (BIO-RAD, CFX96). The rice Ubiquitin gene was used as the internal control in all analyses. A list of the markers used for qRT–PCR was given in Supplementary Table 6.

5′ modiﬁed rapid ampliﬁcation of cDNA ends. Poly (A)+ mRNA was puriﬁed from total RNA using the PolyA kit (Promega). RNA ligase-mediated 5′ RACE was performed using the GeneRacer kit (Invitrogen). Relevant primers are listed in
Supplementary Table 6.

Vector constructions and plant transformation. Relevant primers are listed in Supplementary Table 6. Various DNA fragments were ampliﬁed and introduced into different binary vectors for plant transformation. For rGRF4 overexpression vector construction, speciﬁc primers carrying the mutations 7mGL2-1F and 7mGL2-1R were synthesized for amplifying the CDS (coding sequence); the full- length CDS was then introduced into the binary vector pCAMBIA1300-35S-Flag. All the ﬁnal constructs were sequenced to ensure the correctness of introduced segments and then were used to transform rice plants or to inﬁltrate tobacco leaves by Agrobacterium tumefaciens-mediated methods.

Yeast two-hybrid, BiFC, GST pull-down and luciferase reporter analyses. Various segments of OsGRF4 cDNA were ampliﬁed and cloned into the pGBKT7 vector for transcriptional activation activity analysis. Yeast two-hybrid screening was performed according to Matchmaker Gold Yeast Two-Hybrid System (Clontech) using a truncated protein (GRF4-ΔT) as bait. Full-length cDNAs of the candidate interacting proteins were ampliﬁed and cloned into the pGBKT7 or pGADT7 vector for further veriﬁcation. For BiFC assays, the full-length cDNA of OsGRF4 was cloned into the pVYNE(R) vector, and OsGSK2 was cloned into the pVYCE(R) vector28. The plasmids were co-expressed in tobacco leaf epidermis cells by Agrobacterium-mediated inﬁltration29. Luciferase reporter analysis was performed as previously described30 in rice protoplasts. GST pull-down assay was conducted as described previously13.

Co-immunoprecipitation (Co-IP) analysis. Two vectors pCAMBIA2300–35S– GSK2–eGFP and pCAMBIA1300–35S–GRF4–Flag were constructed and used to express GFP-tagged GSK2 or Flag-tagged GRF4 proteins in tobacco leaves, respectively. Co-IP analysis was performed based on previous reports with some modiﬁcations31. The proteins from 6 g leaf samples were extracted in 20 ml of IP buffer, then centrifuged and ﬁltered through two layers Miracloth (Calbiochem), and mixed with 20 μl of anti-Flag magnetic beads (Sigma). After a 2 h incubation,

the beads were washed twice with IP buffer, and then another two times with a buffer containing only 100 mM Tris HCl (pH 7.4) and 150 mM NaCl. The beads were further boiled in 50 µl of 2 × SDS sample buffer and then subjected to immunoblotting analysis.

RNA-sequencing analysis. One-week seedling shoot or young panicles (<1 cm in length) were sampled for RNA-sequencing analyses. The brassinolide treatment was performed 24 h before sampling, using 1 μM brassinolide added into culture medium. The mRNA was enriched using the oligo(dT) magnetic beads. Agilent 2100 Bioanaylzer and ABI StepOnePlus Real-Time PCR System were used to qualify and quantify of the sample library. The library products were ready for sequencing via Illumina HiSeq 2000. Differentially expressed genes were deﬁned by a twofold expression difference with FDR (false discovery rate) value <0.001.

Hormone measurements. Quantiﬁcation of endogenous GAs was performed as described32. Quantiﬁcation of brassinosteroid (CS) was performed as described previously33. About 4 g of rice seedlings shoots were harvested for GA and brassinosteroid measurements.

Received 31 July 2015; accepted 12 November 2015;
published 21 December 2015; corrected 13 January 2016
References
1. Rosegrant, M. W. & Cline, S. A. Global food security: challenges and policies.
Science 302, 1917–1919 (2003).
2. Xing, Y. Z. & Zhang, Q. F. Genetic and molecular bases of rice yield. Annu. Rev. Plant Biol. 61, 421–442 (2010).
3. Ikeda, M., Miura, K., Aya, K., Kitano, H. & Matsuoka, M. Genes offering the potential for designing yield-related traits in rice. Curr. Opin. Plant Biol. 16, 213–220 (2013).
4. Zuo, J. & Li, J. Molecular genetic dissection of quantitative trait loci regulating rice grain size. Annu. Rev. Genet. 48, 99–118 (2014).
5. Miura, K., Ashikari, M. & Matsuoka, M. The role of QTLs in the breeding of high-yielding rice. Trends Plant Sci. 16, 319–326 (2011).
6. Horiguchi, G., Kim, G. T. & Tsukaya, H. The transcription factor AtGRF5 and the transcription coactivator AN3 regulate cell proliferation in leaf primordia of Arabidopsis thaliana. Plant J. 43, 68–78 (2005).
7. Wang, L. et al. miR396-targeted AtGRF transcription factors are required for coordination of cell division and differentiation during leaf development in Arabidopsis. J. Exp. Bot. 62, 761–773 (2011).
8. Liu, D. M., Song, Y., Chen, Z. X. & Yu, D. Q. Ectopic expression of miR396 suppresses GRF target gene expression and alters leaf growth in Arabidopsis. Physiol. Plant 136, 223–236 (2009).
9. van der Knaap, E., Kim, J. H. & Kende, H. A novel gibberellin-induced gene from rice and its potential regulatory role in stem growth. Plant Physiol. 122, 695–704 (2000).
10. Kim, J. H., Choi, D. S. & Kende, H. The AtGRF family of putative transcription factors is involved in leaf and cotyledon growth in Arabidopsis. Plant J. 36, 94–104 (2003).
11. Tong, H. N. et al. Brassinosteroid regulates cell elongation by modulating gibberellin metabolism in rice. Plant Cell 26, 4376–4393 (2014).
12. Zhu, X. L. et al. Brassinosteroids promote development of rice pollen grains and seeds by triggering expression of Carbon Starved Anther, a MYB domain protein. Plant J. 82, 570–581 (2015).
13. Tong, H. et al. DWARF AND LOW-TILLERING acts as a direct downstream target of a GSK3/SHAGGY-like kinase to mediate brassinosteroid responses in rice. Plant Cell 24, 2562–2577 (2012).
14. Wang, X. L. et al. Arabidopsis MICROTUBULE DESTABILIZING PROTEIN40 is involved in brassinosteroid regulation of hypocotyl elongation. Plant Cell 24, 5193–5193 (2012).
15. Yamamuro, C. et al. Loss of function of a rice brassinosteroid insensitive1 homolog prevents internode elongation and bending of the lamina joint. Plant Cell 12, 1591–1606 (2000).
16. Li, J. A. et al. Mutation of rice BC12/GDD1, which encodes a kinesin-like protein that binds to a GA biosynthesis gene promoter, leads to dwarﬁsm with impaired cell elongation. Plant Cell 23, 628–640 (2011).
17. Kitagawa, K. et al. A novel Kinesin 13 protein regulating rice seed length. Plant Cell Physiol. 51, 1315–1329 (2010).
18. Fujikura, U. et al. Atkinesin-13A modulates cell-wall synthesis and cell expansion in Arabidopsis thaliana via the THESEUS1 pathway. PLoS Genet. 10, e1004627 (2014).
19. Kim, J. H. & Kende, H. A transcriptional coactivator, AtGIF1, is involved in regulating leaf growth and morphology in Arabidopsis. Proc. Natl Acad. Sci. USA 101, 13374–13379 (2004).
20. Tong, H. & Chu, C. Brassinosteroid signaling and application in rice. J. Genet. Genomics 39, 3–9 (2012).
21. Nelissen, H. et al. Dynamic changes in ANGUSTIFOLIA3 complex composition reveal a growth regulatory mechanism in the maize leaf. Plant Cell 27, 1605–1619 (2015).

6 NATURE PLANTS | VOL 2 | JANUARY 2016 | www.nature.com/natureplants

22. Kim, J. H. & Tsukaya, H. Regulation of plant growth and development by the GROWTH-REGULATING FACTOR and GRF-INTERACTING FACTOR duo. J. Exp. Bot. 66, 6093–6107 (2015).
23. Omidbakhshfard, M. A., Proost, S., Fujikura, U. & Mueller-Roeber, B. Growth- regulating factors (GRFs): a small transcription factor family with important functions in plant biology. Mol. Plant 8, 998–1010 (2015).
24. Fan, C. H. et al. GS3, a major QTL for grain length and weight and minor QTL for grain width and thickness in rice, encodes a putative transmembrane protein. Theor. Appl. Genet. 112, 1164–1171 (2006).
25. Hu, J. et al. A rare allele of GS2 enhances grain size and grain yield in rice. Mol. Plant 8, 1455–1465 (2015).
26. Huang, X. H. et al. Genome-wide association studies of 14 agronomic traits in rice landraces. Nature Genet. 42, 961–967 (2010).
27. Jin, Y. et al. An AT-hook gene is required for palea formation and ﬂoral organ number control in rice. Dev. Biol. 359, 277–288 (2011).
28. Waadt, R. et al. Multicolor bimolecular ﬂuorescence complementation reveals simultaneous formation of alternative CBL/CIPK complexes in planta. Plant J. 56, 505–516 (2008).
29. Sparkes, I. A., Runions, J., Kearns, A. & Hawes, C. Rapid, transient expression of ﬂuorescent fusion proteins in tobacco plants and generation of stably transformed plants. Nature Protocols 1, 2019–2025 (2006).
30. Guo, X. et al. The rice GERMINATION DEFECTIVE 1, encoding a B3 domain transcriptional repressor, regulates seed germination and seedling development by integrating GA and carbohydrate metabolism. Plant J. 75, 403–416 (2013).
31. Fiil, B. K., Qiu, J. L., Petersen, K., Petersen, M. & Mundy, J. Coimmunoprecipitation (co-IP) of nuclear proteins and chromatin immunoprecipitation (ChIP) from Arabidopsis. Cold Spring Harb. Protoc. 2008, pdb.prot5049 (2008).

32. Chen, M. L. et al. Highly sensitive and quantitative proﬁling of acidic phytohormones using derivatization approach coupled with nano-LC-ESI-Q- TOF-MS analysis. J. Chromatogr. B 905, 67–74 (2012).
33. Ding, J., Mao, L. J., Wang, S. T., Yuan, B. F. & Feng, Y. Q. Determination of endogenous brassinosteroids in plant tissues using solid-phase extraction with double layered cartridge followed by high-performance liquid chromatography- tandem mass spectrometry. Phytochem. Anal. 24, 386–394 (2013).

Acknowledgements
This work was supported by grants from National Natural Science Foundation of China (91435106, 91335203, 31170715), Ministry of Agriculture of China (2014ZX08009), Natural Science Foundation of Fujian Province (B0420002) and Youth Innovation Promotion Association CAS (2015076).

Author contributions
R.C. and H.T. designed the research, performed the experiments, analysed the data and wrote the paper. B.S, Y.L., S.F., Y.X., D.L., B.H., L.L. and H.W. performed the experiments.
C.C. and M.Z. supervised the project, designed the research and analysed the data.

Additional information
Supplementary information is available online. Reprints and permissions information is available online at www.nature.com/reprints. Correspondence and requests for materials should be addressed to M.Z. and C.C.

Competing interests
The authors declare no competing ﬁnancial interests.GSK2399872A