The homeotic gene long sterile lemma (G1) specifiessterile lemma identity in the rice spikeletAkiko Yoshida, Takuya Suzaki, Wakana Tanaka, and Hiro-Yuki Hirano1

Department of Biological Sciences, Graduate School of Science, University of Tokyo, Tokyo 113-8654, Japan

Edited by Martin F. Yanofsky, University of California at San Diego, La Jolla, CA, and approved October 2, 2009 (received for review July 17, 2009)

The mechanism of floral organ specification is principally con-served in angiosperms, as demonstrated by the ABC model. Bycontrast, mechanisms that regulate the development of organs orstructures specific to a group of species remain unclear. Grasseshave unique inflorescence units, comprising spikelets and florets.In the genus Oryza (rice), the single spikelet consists of a fertilefloret subtended by a lemma and a palea, two sterile lemmas, andrudimentary glumes. Each sterile lemma is a tiny glume-like organwith a smooth surface. Here, we have examined a long sterilelemma1 (g1) mutant, in which the sterile lemma is enlarged like thelemma. Detailed phenotypic analysis reveals that the large sterilelemma in the g1 mutant appears to be caused by homeotictransformation of the sterile lemma into a lemma, suggesting thatG1 is involved in the repression of lemma identity to specify thesterile lemma. Gene isolation reveals that G1 is a member of aplant-specific gene family that encodes proteins with a previouslyuncharacterized domain, named here ALOG (Arabidopsis LSH1 andOryza G1). G1 mRNA is expressed in sterile lemma primordiathroughout their development, and G1 protein is localized in thenucleus. A trans-activation assay using the yeast GAL4 systemsuggests that G1 is involved in transcriptional regulation. Repres-sion of lemma identity by G1 is consistent with a hypothesisproposed to explain the morphological evolution of rice spikelets.We also show that a wild rice species, Oryza grandiglumis, thatforms large sterile lemmas has serious mutations in the G1 gene.

flower � grass � Oryza � ALOG � morphological evolution

A lthough angiosperms produce diverse forms of flowers,from beautiful and entomophilous to inconspicuous and

anemophilous, the genetic programs directing floral develop-ment are fundamentally conserved. Floral organ specification isexplained by the ABC model, and genes constituting this modelare likely to be functionally conserved in a wide range offlowering plants (1–3). In contrast to our deep understanding offloral organs common to many angiosperms, genes that regulatethe development of floral organs and/or flower-associated struc-tures distinctive to a group of plant species remain poorlyelucidated.

Grass species, such as Oryza sativa (rice) and Zea mays(maize), bear a unique inflorescence consisting of spikelets andflorets (4, 5). Each spikelet produces a defined number of floretsdepending on species, and is subtended by a pair of glumes. Thefloret comprises the flower proper (carpels, stamens, and lodi-cules) and a pair of additional structures (a palea and a lemma)that subtend the flower. The lodicule, an organ homologous tothe petal in ordinary flowers, is small and semitransparent, andacts to open the palea and lemma for anthesis. Molecular geneticstudies in rice and maize have revealed that the function of Bclass MADS-box genes is conserved in grasses: these genesspecify the identities of the lodicule and stamen similar to theway in which B class genes specify the identities of the petal andstamen in eudicots (4, 6–9). In addition to the crucial role of Cclass MADS-box genes in carpel specification in eudicots, theidentity of the carpel is determined by the YABBY gene DROOP-ING LEAF (DL) in rice (10). Spatial expression patterns ofDL-related genes from maize and other grasses suggest the

conserved function of these genes in grasses (11). Stamenidentity and floral meristem determinacy, which are controlledby a single C class gene, AGAMOUS, in Arabidopsis, are regu-lated separately by two or more C class genes that were dupli-cated before divergence of the grass species (12, 13). The paleaand lemma are organs specific to grasses, and enclose inner floralorgans (14–16). The LEAFY HULL STERILE1 (LHS1) gene isinvolved in partially specifying the identity of the lemma andpalea in rice, in addition to regulating spikelet meristem deter-minacy (17, 18). Although LHS1 orthologs are expressed in thepalea and lemma in other grasses, the expression patterns ofthese genes suggest that they are also involved in various aspectsof the development and morphological diversification of spike-lets in grasses (19). The relationship of the lemma/palea to floralstructures in other monocots and eudicots is not yet fullyunderstood, although comparative morphological studies sug-gest that there is a correlation between the lemma/palea and theouter tepals of a typical monocot flower (20, 21).

Inflorescence architecture and spikelet organization variesfrom species to species within the grass family. For example, thenumber of florets varies from one to 40 depending on the species(22). Determinacy of the spikelet meristem, which is closelyassociated with floret number, is regulated by genes encodingAPETALA2 (AP2)-like transcription factors both in rice andmaize (23–25). Rice produces a single fertile floret within aspikelet. The floret is f lanked by a pair of glume-like organs,which are not usually observed in the spikelets of other grassessuch as maize and wheat. A hypothesis has been proposed thatthese glume-like organs are derived from the lemmas of twosterile florets that have been reduced during the evolution ofOryza (26). According to this hypothesis, these organs should becalled sterile lemmas rather than empty glumes, although con-fusingly both terminologies have been used in the rice literature.However, the size and identity of the sterile lemmas differ greatlyfrom those of lemmas. Despite this attractive hypothesis basedon morphological changes during evolution, to our knowledgethere have been no studies on the gene responsible for devel-opment of the sterile lemma and its evolutionary origin.

In this study, we have analyzed a rice mutant, called long sterilelemma1 (G1), that forms a large sterile lemma (LSL) (27–29).Detailed morphological analyses revealed that the LSL in g1 haslemma identity that clearly differs from the identity of the sterilelemma in WT. We isolated the G1 gene by map-based cloning.G1 encodes a protein that has a conserved domain flanked by anuclear localization signal and is specific to angiosperms. Cel-

Author contributions: A.Y. and H.-Y.H. designed research; A.Y., T.S., W.T., and H.-Y.H.performed research; and A.Y. and H.-Y.H. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Freely available online through the PNAS open access option.

Data deposition: The sequences reported in this paper have been deposited in the DNAData Bank of Japan. For a list of accession numbers, see Table S1.

1To whom correspondence should be addressed. E-mail: hyhirano@biol.s.u-tokyo.ac.jp.

This article contains supporting information online at www.pnas.org/cgi/content/full/0907896106/DCSupplemental.

www.pnas.org�cgi�doi�10.1073�pnas.0907896106 PNAS � November 24, 2009 � vol. 106 � no. 47 � 20103–20108

PLA

NT

BIO

LOG

Y

lular localization and trans-activation analysis by a yeast systemraised the possibility that G1 may be involved in transcriptionalregulation. Collectively, these results suggest that G1 is requiredfor the repression of lemma identity in the sterile lemma in rice,and might have been involved in its morphological modificationduring the evolution of rice.

ResultsThe LSL in the g1 Mutant has Lemma Identity. The rice spikelet hasone fertile floret in which a pistil, six stamens, and two lodiculesdevelop (Fig. 1A) (4, 14, 26). Glumes are highly reduced in rice,being observed as only vestigial structures at the base of the spikelet,and are thereby called rudimentary glumes (Fig. 1D). A pair ofsterile lemmas, which have the appearance of tiny glumes, is formedbetween the floret and the rudimentary glumes (Fig. 1 B and D).

In a recessive mutant, g1–1, we observed that the sterile lemmais larger than that of WT (Fig. 1 B and C). Close examination ofthe base of the spikelet revealed that the LSL in g1 was formedat the expense of the sterile lemma (Fig. 1 D and E). In addition,no abnormality was detected in other spikelet organs such as thepaleas and rudimentary glumes, suggesting that the g1 mutationspecifically affects the sterile lemma.

To address the effect of the g1 mutation in detail, we examinedmorphological differences in the epidermal cells between the g1LSL and the sterile lemma of WT. In WT, the sterile lemma hasa smooth surface on the abaxial side and rarely forms trichomeson its surface except for the marginal region (Fig. 1F). Bycontrast, the abaxial surface of the g1 LSL is very rough (Fig.1H). Convex structures called tubercles (30) are arranged in

parallel and the cell files with a tubercle are very clear. Inaddition, trichomes are formed on the abaxial surface. Thisepidermal morphology in the g1 LSL is highly similar to that ofthe lemma and palea (Fig. 1G), suggesting that epidermal cellsin the g1 LSL have an identity similar to those in the lemma/palea. Observations during the developmental process of thespikelet revealed that the epidermal cell identity of the LSL ing1 is determined at an early stage (Fig. S1). The surfacemorphologies of the g1 LSL and the WT sterile lemma were notdistinguishable from each other before initiation of the stamenprimordia (Fig. S1 G and P). The surface of the g1 LSL becamemore similar to that of the lemma than that of the sterile lemmaat the stage when the stamen primordia were clearly observed(Fig. S1 E, H, N, and Q). By contrast, the size of the g1 LSL wasnearly the same as that of the sterile lemma of a WT flower atthis stage (Fig. S1 B and K), suggesting that timing of thespecification of cell identity was not coupled to cell elongationin the LSL of the g1 spikelet.

We next observed a cross-section of the spikelet and foundthat the LSL in g1 seems to resemble the lemma rather than thepalea. First, the number of vascular bundles is five in both the g1LSL and the lemma, whereas it is three in the palea (Fig. 1 J andK). Second, the marginal regions of the g1 LSL curl inward likethose of the lemma and differ from those of the palea (Fig. 1 Land M). These phenotypic analyses suggest that the LSL in g1 isnot only larger but also has morphological and structural featuressimilar to those of the lemma. Thus, it is likely that G1 is involvedin specification of the identity of the sterile lemma.

G1 Encodes a Unique Protein with an Uncharacterized Domain. Toelucidate its molecular function, we isolated the G1 gene by usinga map-based strategy. Using F2 plants from a cross between g1–1(japonica) and Kasalath (indica), we found that G1 was restrictedto a 33.7-kb region between the markers g1M32 and g1M19 onthe short arm of chromosome 7 (Fig. 2A). We then found thata segment of approximately 12 kb was deleted between the twomarkers in g1–1. Within this deleted region, there were threepredicted genes [LOC�Os07g04650 (Os07g0139000),LOC�Os07g04660 (Os07g0139100), LOC�Os07g04670

A B C D

E

N

F G H I

J K L

M

Fig. 1. Spikelet phenotypes of the g1 mutant. (A) Schematic representationof the spikelet of WT rice. (B) A WT spikelet. (C) A g1 spikelet showing the LSL.Basal region of the spikelet in WT (D) and g1 (E). Epidermal surface of thesterile lemma (F) and the lemma in WT (G). (H) Epidermal surface of the LSL ing1. (I) A spikelet in a rescued g1 mutant, which was transformed with agenomic fragment encompassing the G1 locus. Cross section of the spikelet inWT (J) and g1 (K). Close-up view of the marginal region of the WT lemma (L)and g1 LSL (M). Red stars indicate vascular bundles (J and K). Arrows indicatethe peripheral structure of the lemma and LSL (L and M). (N) A spikelet of O.grandiglumis (W0613). le, lemma; pa, palea; rg, rudimentary glume; sl, sterilelemma. (Scale bars: 1 mm in B, C, I, and N; 200 �m in J and K; 100 �m in D, E,F, G, and H; and 50 �m in L and M.).

Fig. 2. Map-based cloning and protein structure. (A) Physical map of the G1locus. Numerals under maps indicate the number of recombinants among1,380 meioses. Pink open box indicates the 12-kb chromosomal deletion foundin g1–1. (B) Protein structure of G1 and its related proteins. Green and orangeboxes indicate the ALOG domain and nuclear localization signal, respectively.Gray boxes indicate amino acid insertions in the ALOG domain in G1.

20104 � www.pnas.org�cgi�doi�10.1073�pnas.0907896106 Yoshida et al.

(Os07g0139300)] in the rice database [TIGR release 5 (RAP2build 4.0)]. Expression analysis by RT-PCR revealed that onlyone gene, LOC�Os07g04670, was expressed in the young inflo-rescence in WT (Fig. S2E). Next, we determined the genomicsequence of LOC�Os07g04670 from three putative g1 mutants(g1–2, g1–3, g1–4) that had the LSL (Fig. S2). By a geneticcomplementation test, g1–4 was shown to be allelic to g1–1. Wefound missense and nonsense mutations in all three mutantsexamined (Fig. 2B and Fig. S3). A 3.2-kb genomic fragmentcontaining LOC�Os07g04670 from WT rescued the mutantphenotype of the g1 spikelet (Fig. 1I). Taking these resultstogether, we concluded that LOC�Os07g04670 is the G1 gene.

We found that G1 encodes a protein with a conserved domainof unknown function (Fig. 2B and Figs. S3 and S4). In the ricedatabase, this domain is called DUF640 (domain of unknownfunction 640). A nuclear localization signal f lanks the C-terminalregion of this domain. There are 10 genes that encode proteinswith DUF640 in both rice and Arabidopsis (Fig. 3). The only genefunctionally characterized so far in this family is LIGHT-DEPENDENT SHORT HYPOCOTYLS1 (LSH1), which is in-volved in phytochrome-dependent light signaling in Arabidopsis(31). Here, we named DUF640 the ALOG (Arabidopsis LSH1and Oryza G1) domain. Genes encoding proteins containing anALOG domain were also found in the genome of Physcomitrellapatens and Selaginella moellendorffii (32, 33), but not in thegenome of algae, fungi, or animals, suggesting that ALOGdomain genes may be specific to land plants (Fig. 3).

In contrast to other ALOG domains identified, the ALOGdomain in G1 is interrupted with two insertions consisting of 9and 23 aa (Fig. 2B and Figs. S3 and S4). Proteins with aninterrupted ALOG domain like G1 are restricted to the grassspecies: that is, an extensive similarity search did not identifyproteins with an interrupted ALOG domain in the genome ofother plants, including plants for which the whole genomesequence is available. Missense mutations are present in theALOG domain of g1–2 and g1–3 (Fig. 2B and Fig. S3). Theseverity of the spikelet phenotypes of these two alleles iscomparable to that of the phenotypes of g1–1 and g1–4, in whichG1 is completely defective, suggesting that amino acid substitu-tions in this domain are critical for protein function.

The Oryza genus has 23 species including two domesticatedspecies. Among them, Oryza grandiglumis, which has the allotet-raploid genome CCDD, forms an LSL similar to that of the g1mutant of O. sativa (Fig. 1N); indeed, the name of this species

is derived from this characteristic (34, 35). The number ofvascular bundles and the morphology of the peripheral region ofthe LSL in O. grandiglumis were similar to those of the LSL inthe g1 mutant (Fig. S2 F-H). To know whether the LSL in O.grandiglumis is associated with the function of G1, we sequencedG1 orthologs from this species. We found substitutions of aninvariant amino acid in the ALOG domain or nonsense muta-tions in both genomes of this allotetraploid species (Fig. S5).These results indicate that the LSL in O. grandiglumis is probablycaused by loss of function of G1.

Spatial Expression Patterns of G1 mRNA and Molecular Features of G1Protein. To address the molecular function of G1, we firstperformed in situ hybridization analysis (Fig. 4 A and B). G1 wasexpressed in the primordia of the sterile lemma from an earlystage of spikelet development. G1 expression in the sterilelemma primordia persisted throughout the development. Noexpression was detected in the lemma, although G1 signals weredetected in the basal region of the palea. Continuous expressionof G1 in the sterile lemma primordia is consistent with themutant phenotype of g1.

Next, we examined the subcellular localization of G1 protein.Constructs producing a fusion protein of G1 and GFP wereintroduced into onion epidermal cells by particle bombardment.The results clearly showed that the G1–GFP fusion protein waslocalized in the nucleus (Fig. 4 C–E). Similar results were obtainedin an experiment using a GFP-G1 fusion protein (Fig. S6).

Last, we addressed the function of G1 protein by using atrans-activation assay system based on yeast GAL4 (36). Theprotein-coding region of G1 cDNA was fused in-frame to asequence encoding the GAL4 DNA-binding domain (GAL4DB) and placed under control of the caulif lower mosaic viruspromoter (i.e., effector). As a reporter, the luciferase gene under

S. moellendorffii G1-like1P. patens G1-like1

OsG1L9OsG1L7

OsG1L8AtLSH5AtLSH6

AtLSH9AtLSH7

AtLSH8AtLSH1

AtLSH2AtLSH3

AtLSH4

AtLSH10

OsG1L6

OsG1BdG1

ZmG1SbG1

OsG1L1OsG1L2

OsG1L5OsG1L3

OsG1L40.05

945

648742

576

693

973767

1000

994390

589

690

1000

814201

784994

981470

483

G1

clad

e

Fig. 3. Phylogenic tree of proteins containing a ALOG domain. Amino acidsof the ALOG domain were compared (Fig. S4) and the tree was constructed bya neighbor-joining method (44). Numbers indicate bootstrap values.

C

D

E

F

G

GAL4-LUC 5xGAL4-TATA- LUC -Nos

GAL4DB 35S- - GAL4DB -Nos

GAL4DB-G1 35S- - GAL4DB G1 -Nos

GAL4BD

Relative luciferase activity

GAL4BDGAL4BD-G1

0 5 10 15 20

GAL4BD-G1Exp.2

Exp.1

Reporter

Effectors

G1-GFP

B le

pa

sl

st

sl

A fm

le

pa

sl

sl

DAPI

merged

GFP

Fig. 4. Spatial expression pattern of G1 mRNA and function of G1 protein.(A and B) in situ localization of G1 mRNA. fm, floret meristem; le, lemma; pa,palea; sl, sterile lemma; st, stamen. (C) Subcellular localization of the G1-GFPfusion protein. (D) DAPI staining of the nucleus. (E) Merged image of C and D.(F) Subcellular localization of GFP. (G) Trans-activation analysis of G1 by a yeastGAL4 system. GAL4BD, DNA-binding domain of GAL4; 5� GAL4, 5 copies ofGAL4-upstream activating sequence. (Scale bars: 50 �m.)

Yoshida et al. PNAS � November 24, 2009 � vol. 106 � no. 47 � 20105

PLA

NT

BIO

LOG

Y

the control of five copies of binding sites for GAL4 was used.Both effector and reporter plasmids were delivered to Arabi-dopsis leaves by particle bombardment. We found that the fusionprotein of GAL4BD and G1 enhanced transcriptional activity inArabidopsis leaves compared with GAL4BD alone (Fig. 4G).This observation suggests that G1 is involved in transcriptionalregulation.

DiscussionG1 Is Involved in the Specification of Sterile Lemma Identity. Ricebears a single fertile floret within a spikelet, in which the glumesare highly degenerated and visible only as two diminutiveoutgrowths (called rudimentary glumes). Sterile lemmas areproduced between the fertile floret and the rudimentary glumes(4, 14, 26). The sterile lemma has not been genetically charac-terized and is often mistaken for a glume. The present studyshowed that the G1 gene is specifically involved in specificationof the sterile lemma. In the g1 mutant, the LSL has structuralfeatures resembling those of the lemma. The size, shape, andarrangement of the abaxial epidermal cells in the LSL in g1 wereindistinguishable from those in the lemma. Trichomes wereformed on the abaxial surface of the g1 LSL as they are in thelemma. In addition, the number of vascular bundles and themarginal structure were shared by the LSL and the lemma. Thus,the LSL develops at the expense of the sterile lemma, suggestingthat the sterile lemma is homeotically transformed into thelemma in the g1 mutant. Because no such clear transformationhas previously been reported, G1 seems to be a unique homeoticgene that regulates the development of spikelet organs in grasses.

We found that G1 is a member of an uncharacterized genefamily that is shared by land plants. Furthermore, subcellularlocalization and trans-activation analysis by a yeast GAL4 systemsuggest that G1 is involved in transcriptional regulation. G1 isexpressed in the primordia of the sterile lemma throughout itsdevelopment. It is therefore likely that G1 specifies the identityof the sterile lemma by repressing lemma identity via theregulation of downstream target genes.

Evolution of the Spikelet Architecture and the Role of G1 in RiceSpikelets. The number of florets in a spikelet is a key feature ofspikelet architecture in grasses. Two florets are formed in amaize spikelet, although the lower floret is degenerated infemale spikelets (4, 5, 23). Interestingly, the number of floretsper spikelet is associated with ploidy level in wheat (37).

All extant rice species in the genus Oryza have a single fertilefloret in the spikelet. From comparative morphological studies,it has been hypothesized that a putative ancestor of Oryza hadthree florets and that, during evolution, the two lateral f loretsdegenerated, leaving only the lemma (26). The sterile lemma inthe extant Oryza species seems to be derived from morphologicalmodification of the remaining lemma. In line with this hypoth-esis, the default state of the sterile lemma would seem to be thelemma. Consistent with this idea, we revealed that the sterilelemma has the potential to transform into the lemma if G1activity is lost. Thus, it seems an attractive idea that G1 may havebeen recruited to specify the sterile lemma by repressing theidentity of the lateral remaining lemma during the evolution ofOryza (Fig. 5).

O. grandiglumis has LSLs similar to those of the g1 mutant (34,35). It is possible that the spikelet of O. grandiglumis correspondsto an intermediate type of spikelet with an LSL before evolutionof the repression of lemma identity, and that other rice specieshave lost the lemma identity of the sterile lemma in independentlineages during rice evolution. Alternatively, it is also possiblethat the spikelet phenotype in O. grandiglumis is caused bymutations that occurred in the G1 locus after structural featuresof the sterile lemma were established. Sequencing of the G1 genein two accessions of O. grandiglumis clearly supported the latter

idea. The two accessions shared a substitution of an invariantamino acid, which is completely conserved in all ALOG domainproteins, in one genome. In the other genome, each accessionhad a nonsense mutation at a different position. Thus, the G1gene might have been compromised sequentially during theevolution of this tetraploid species, although we need to confirmthat the amino acid substitution in the ALOG domain results inloss of G1 function. In addition, we are interested in the processof how the tetraploid species O. grandiglumis accumulated themutation in G1 in both genomes. It is possible that close relativesof O. grandiglumis with a CCDD genome, O. alta and O. latifolia(38), have a mutation in G1 in one of the genomes, and thecombination of two genomes with the mutation may be associ-ated only with the lineage of the O. grandiglumis.

Analyses of O. grandiglumis support the idea that G1 is a keyregulator of sterile lemma development. Based on the hypothesis ofArber (26), the spikelets of both O. grandiglumis and the g1 mutantof O. sativa appear to be a reversal. If so, it is surprising that thesterile lemma of both O. sativa and ancestral species of O. grandi-glumis has maintained the potential to express genes responsible forlemma identity for a long evolutionary period. The genus Leersia,which, like Oryza, belongs to Oryzeae, lacks the sterile lemma (16).It would be interesting to know whether G1 activity is involved indevelopmental arrest of the sterile lemma in this genus. Sterilelemmas are widely distributed in the grass family, which suggeststhat the sterile lemma either evolved in several independent lin-eages or was lost multiple times after its establishment near the baseof the grass clade. It is possible that G1 orthologs may be involvedin morphological modification of the sterile lemma in lineages otherthan those of Oryza evolution. Alternatively, G1 function may berestricted to the sterile lemma of Oryza, because the epidermalconvex structures (i.e., tubercles), which are suppressed by G1 in thesterile lemma, are a synapomorphy of the Oryza genus (16, 30)and/or G1-related proteins may have diverse functions as a result ofthe insertion of amino acids, which are highly variable amonggrasses.

Characteristics of G1 Protein. G1 has a unique conserved domain,named ALOG, that is f lanked by a nuclear localization signal.The 125-residue ALOG domain is rich in basic amino acids,especially arginine, and is highly conserved among land plants:53 of 125 aa were found to be completely conserved in 25proteins from seven species including rice, Arabidopsis, andPhyscomitrella. G1 is unique in that its ALOG domain isinterrupted by two insertions. This interruption is observed onlyin proteins closely related to rice G1 in grasses (i.e., G1 clade).Despite these insertions, G1 seems to be functional in rice, as

hypothetical ancestor 2 Oryza

G1

mutationsin G1

O. grandiglumis

hypothetical ancestor 1

lemma

lemma lemma

lemmasterilelemma

lemma

Fig. 5. Model of hypothesized changes in spikelet architecture during Oryzaevolution (26) and putative function of G1.

20106 � www.pnas.org�cgi�doi�10.1073�pnas.0907896106 Yoshida et al.

shown here by genetic and molecular biological studies. Con-servation of amino acids in the ALOG domain among proteinsin the G1 clade also supports this idea, because mutations mighthave accumulated if such G1-related proteins had no function.The restricted distribution of this interrupted ALOG domainraises the possibility that ALOG domain proteins in the G1 cladeare associated with developmental regulation specific to grasses.The role of the ALOG domain is unknown; thus, it would beinteresting to elucidate the biochemical function of this domain,together with the effect of the insertions in the G1 clade, infuture studies.

Materials and MethodsPlant Materials. Taichung 65 was used as the WT strain for standard spikeletand for in situ analysis. Strains HO680 (g1–1), H-138 (g1–2), and 7578-RD-3(g1–3) were provided by A. Yoshimura (Fukuoka, Japan), I. Takamure (Sap-poro, Japan), and H. Kitano (Nagoya, Japan), respectively (28, 29). g1–4(ND4295) was identified in the TOS17 mutant panel and provided by H.Hirochika and A. Miyao (39) (Tsukuba, Japan).

Isolation of G1. The primers used for PCR amplification are shown in Table S2.The G1 locus was mapped by using F2 plants from a cross between g1–1 andKasalath (ssp. indica). First, the G1 locus was mapped to a region betweenCAPS markers C261 and S20268 on the distal end of the short arm of chromo-some 7 by using 690 g1 homozygotes. Next, the locus was narrowed to a33.7-kb region between two closely linked CAPS markers, g1M32 and g1M19,which were designed by comparing the genomic sequences of japonica andindica (Table S2). A chromosomal deletion was found in this region, the extentof which was restricted to 12 kb. Three putative genes were located in this12-kb region. For complementation analysis, a 3.7-kb DNA fragment encom-passing the G1 locus was isolated from BAC clone OSJNBa0088E18 (ClemsonUniversity Genome Institute) and cloned into a gateway vector (Invitrogen).By an LR recombination reaction, the genomic fragment was transferred to abinary vector, pBI-Hm12-GW, which was constructed by introducing a Gate-way rfC cassette into pBI-Hm12 in this study. The resulting chimeric plasmidcarrying the G1 gene was used for transformation of g1–4 by the Agrobac-terium-mediated method (40). G1 cDNA was amplified by using a cDNA poolsynthesized from total RNA isolated from the young panicles. After sequenc-ing of the RT-PCR product, the ORF was predicted.

Cytological and Molecular Analyses. For in situ hybridization, a 552-bp frag-ment in the 3� region of G1 cDNA was amplified (Table S2) and cloned into aT-vector by TA cloning (Novagen). Probe synthesis, preparation of sections, insitu hybridization, and microscopic observation were performed as describedpreviously (41, 42). For analysis of subcellular localization, a DNA fragmentcorresponding to the coding region of G1 was amplified and inserted into aGateway binary vector, pGWB5 or pGWB6 (43), by recombination to producea G1–GFP or GFP–G1 fusion protein (Invitrogen). The resulting construct wasbombarded into onion epidermal cells by a particle-mediated DNA deliverysystem (PDS1000/He; Bio-Rad). After overnight incubation at 22 °C in the dark,the florescence of GFP was detected by a confocal micro-scanning lasermicroscope (Nikon). Last, cells were stained with DAPI to visualize the nucleus.The vectors used for the GAL4 trans-activation system were kindly provided byM. Ohme-Takagi (36) (Tsukuba, Japan). A DNA fragment amplified from G1cDNA (Table S2) was digested with EcoRV and SalI and inserted into a vectorcontaining the cauliflower mosaic virus CaMV35S promoter and a sequenceencoding the for GAL4 DNA binding domain (GAL4DB) to make a G1-GAL4BDfusion protein (i.e., effector). The resulting plasmid was introduced intoArabidopsis leaves by particle bombardment as described earlier. In co-transfection assays, we used 1.6 �g of reporter construct and 1.2 �g of effectorconstruct for each bombardment. As an internal control, an RLUC reportergene from Renilla (0.4 �g) under the control of 35S promoter (Promega) wasco-bombarded with both reporter and effector plasmids. After bombard-ment, the leaves were put on agar medium containing Murashige and Skoogmedium (pH 5.7) and incubated at 22 °C for 12 h. Luciferase assays wereperformed with the Dual-Luciferase Reporter Assay System according to theinstruction manual (Promega). After measurement of LUC and RLUC activity,the activity of LUC relative to RLUC was estimated.

ACKNOWLEDGMENTS. We thank Drs. A. Yoshimura, H. Kitano, and I. Taka-mure for their kind distribution of g1 alleles; Drs. Y. Takagi and T. Nakagawafor providing vectors for GAL4 activation analysis and gateway binary vectors,respectively; Drs. Y. Fujimoto, H. Takanashi, and N. Tsutsumi for technicalsupport; and Ms. K. Ohsawa for technical assistance. This work was supportedin part by Grants-in-Aid for Scientific Research from MEXT (20380005,20061006) and the Program of Basic Research Activities for Innovative Bio-sciences (PROBRAIN) (to H.-Y.H.), the Global COE Program (Integrative LifeScience Based on the Study of Biosignaling Mechanisms) from the Ministry ofEducation, Culture, Sports, Science and Technology (to A.Y. and W.T.), and aSasakawa Scientific Research Grant from The Japan Science Society (to A.Y.).

1. Coen ES, Meyerowitz EM (1991) The war of the whorls: genetic interactions controllingflower development. Nature 353:31–37.

2. Theissen G, et al. (2000) A short history of MADS-box genes in plants. Plant Mol Biol42:115–149.

3. Lohmann JU, Weigel D (2002) Building beauty: The genetic control of floral patterning.Dev Cell 2:135–142.

4. Bommert P, Satoh-Nagasawa N, Jackson D, Hirano H-Y (2005) Genetics and evolutionof inflorescence and flower development in grasses. Plant Cell Physiol 46:69–78.

5. Zanis MJ (2007) Grass spikelet genetics and duplicate gene comparisons. Int J Plant Sci168:93–110.

6. Ambrose BA, et al. (2000) Molecular and genetic analyses of the silky1 gene revealconservation in floral organ specification between eudicots and monocots. Mol Cell5:569–579.

7. Nagasawa N, et al. (2003) SUPERWOMAN 1 and DROOPING LEAF genes control floralorgan identity in rice. Development 130:705–718.

8. Whipple CJ, et al. (2004) Conservation of B-class floral homeotic gene function be-tween maize and Arabidopsis. Development 131:6083–6091.

9. Thompson BE, Hake S (2009) Translational biology: from Arabidopsis flowers to grassinflorescence architecture. Plant Physiol 149:38–45.

10. Yamaguchi T, et al. (2004) The YABBY gene DROOPING LEAF regulates carpel speci-fication and midrib development in Oryza sativa. Plant Cell 16:500–509.

11. Ishikawa M, et al. (2009) The spatial expression patterns of DROOPING LEAF orthologssuggest a conserved function in grasses. Genes Genet Syst 84:137–146.

12. Mena M, et al. (1996) Diversification of C-function activity in maize flower develop-ment. Science 274:1537–1540.

13. Yamaguchi T, et al. (2006) Functional diversification of the two C-class genes, OS-MADS3 and OSMADS58, in Oryza sativa. Plant Cell 18:15–28.

14. Chang T-T, Bardenas EA (1965) The morphology and varietal characteristics of the riceplant. (IRRI, Los Banos, Philippines).

15. Kellogg EA (2001) Evolutionary history of the grasses. Plant Physiol 125:1198 –1205.

16. Kellogg EA (2009) The evolutionary history of Ehrhartoideae, Oryzeae, and Oryza. Rice2:1–14.

17. Jeon JS, et al. (2000) leafy hull sterile1 is a homeotic mutation in a rice MADS box geneaffecting rice flower development. Plant Cell 12:871–884.

18. Prasad K, Parameswaran S, Vijayraghavan U (2005) OsMADS1, a rice MADS-box factor,controls differentiation of specific cell types in the lemma and palea and is anearly-acting regulator of inner floral organs. Plant J 43:915–928.

19. Malcomber ST, Kellogg EA (2004) Heterogeneous expression patterns and separateroles of the SEPALLATA gene LEAFY HULL STERILE1 in grasses. Plant Cell 16:1692–1706.

20. Sajo MG, Longhi-Wagner HM, Rudall PJ (2008) Reproductive morphology of theearly-divergent grass Streptochaeta and its bearing on the homologies of the grassspikelet. Plant Syst Evol 275:245–255.

21. Preston JC, Christensen A, Malcomber ST, Kellogg EA (2009) MADS-box gene expres-sion and implications for developmental origins of the grass spikelet. Am J Bot96:1419–1429.

22. Clifford HT, Soderstrom TR, Hilu KW, Campbell CS, Barkworth ME (1987) Spikelet andfloral morphology. Grass Systematics and Evolution, eds Soderstrom TR, Hilu KW,Campbell CS, Barkworth ME (Smithsonian Institution Press, Washington, DC), pp21–30.

23. Chuck G, Meeley RB, Hake S (1998) The control of maize spikelet meristem fate by theAPETALA2-like gene indeterminate spikelet1. Genes Dev 12:1145–1154.

24. Lee DY, Lee J, Moon S, Park SY, An G (2007) The rice heterochronic gene SUPERNU-MERARY BRACT regulates the transition from spikelet meristem to floral meristem.Plant J 49:64–78.

25. Chuck G, Meeley R, Hake S (2008) Floral meristem initiation and meristem cell fate areregulated by the maize AP2 genes ids1 and sid1. Development 135:3013–3019.

26. Arber A (1934) The Gramineae: a study of cereal, bamboo, and grasses (Cambridge UnivPress, Cambridge, UK).

27. Kinoshita T (1984) Gene analysis and linkage map. Biology of Rice, eds Tsunoda S,Takahashi N (Japan Sci Soc Press/Elsevier, Tokyo/Amsterdam).

28. Takamure I, Kinoshita T (1992) Modification of glume characters due to genic inter-actions and environmental conditions. Rice Genet Newslett 9:85–88.

29. Nagato Y, Yoshimura A (1998) Report of the committee on gene symbolization,nomenclature and linkage groups. Rice Genet Newslett 15:13–74.

30. Terrell EE, Peterson PM, Wergin WP (2001) Epidermal features and spikelet micromor-phology in Oryza and related genera (Poaceae: Oryzeae). Smithsonian Contr Bot91:1–50.

31. Zhao L, et al. (2004) Overexpression of LSH1, a member of an uncharacterised genefamily, causes enhanced light regulation of seedling development. Plant J 37:694 –706.

Yoshida et al. PNAS � November 24, 2009 � vol. 106 � no. 47 � 20107

PLA

NT

BIO

LOG

Y

32. Rensing SA, et al. (2008) The Physcomitrella genome reveals evolutionary insights intothe conquest of land by plants. Science 319:64–69.

33. Banks JA (2009) Selaginella and 400 million years of separation. Annu Rev Plant Biol60:223–238.

34. Vaughan DA (1994) The wild relatives of rice: a genetic resources hand book (IRRI, LosBanos , Philippines).

35. Zamora A, Barboza C, Lobo J, Espinoza AM (2003) Diversity of native rice (Oryzapoaceae) species of Costa Rica. Genet Res Crop Evol 50:855–870.

36. Ohta M, Ohme-Takagi M, Shinshi H (2000) Three ethylene-responsive transcrip-tion factors in tobacco with distinct transactivation functions. Plant J 22:29 –38.

37. Shitsukawa N, Kinjo H, Takumi S, Murai K (2009) Heterochronic development of thefloret meristem determines grain number per spikelet in diploid, tetraploid andhexaploid wheats. Ann Bot 104:1–9.

38. Bao Y, Ge S (2004) Origin and phylogeny of Oryza species with the CD genome basedon multiple-gene sequence data. Plant Syst Evol 249:55–66.

39. Miyao A, et al. (2007) A large-scale collection of phenotypic data describing aninsertional mutant population to facilitate functional analysis of rice genes. Plant MolBiol 63:625–635.

40. Hiei Y, Ohta S, Komari T, Kumashiro T (1994) Efficient transformation of rice (Oryzasativa L.) mediated by Agrobacterium and sequence analysis of the boundaries of theT-DNA. Plant J 6:271–282.

41. Suzaki T, et al. (2004) The gene FLORAL ORGAN NUMBER1 regulates floral meristemsize in rice and encodes a leucine-rich repeat receptor kinase orthologous to Arabi-dopsis CLAVATA1. Development 131:5649–5657.

42. Suzaki T, Yoshida A, Hirano H-Y (2008) Functional diversification of CLAVATA3-relatedCLE proteins in meristem maintenance in rice. Plant Cell 20:2049–2058.

43. Nakagawa T, et al. (2007) Development of series of gateway binary vectors, pGWBs, forrealizing efficient construction of fusion genes for plant transformation. J BiosciBioeng 104:34–41.

44. Saitou N, Nei M (1987) The neighbor-joining method: a new method for reconstructingphylogenetic trees. Mol Biol Evol 4:406–425.

20108 � www.pnas.org�cgi�doi�10.1073�pnas.0907896106 Yoshida et al.