DTH8 suppresses flowering in rice, influencing plant height and yield potential simultaneously.

The three most important agronomic traits of rice (Oryza sativa), yield, plant height, and flowering time, are controlled by many quantitative trait loci (QTLs). In this study, a newly identified QTL, DTH8 (QTL for days to heading on chromosome 8), was found to regulate these three traits in rice. Map-based cloning reveals that DTH8 encodes a putative HAP3 subunit of the CCAAT-box-binding transcription factor and the complementary experiment increased significantly days to heading, plant height, and number of grains per panicle in CSSL61 (a chromosome segment substitution line that carries the nonfunctional DTH8 allele) with the Asominori functional DTH8 allele under long-day conditions. DTH8 is expressed in most tissues and its protein is localized to the nucleus exclusively. The quantitative real-time PCR assay revealed that DTH8 could down-regulate the transcriptions of Ehd1 (for Early heading date1) and Hd3a (for Heading date3a; a rice ortholog of FLOWERING LOCUS T) under long-day conditions. Ehd1 and Hd3a can also be down-regulated by the photoperiodic flowering genes Ghd7 and Hd1 (a rice ortholog of CONSTANS). Meanwhile, the transcription of DTH8 has been proved to be independent of Ghd7 and Hd1, and the natural mutation of this gene caused weak photoperiod sensitivity and shorter plant height. Taken together, these data indicate that DTH8 probably plays an important role in the signal network of photoperiodic flowering as a novel suppressor as well as in the regulation of plant height and yield potential.

As food for more than half of the world's population, rice (Oryza sativa) has always been a hot spot in plant science. Its yield is not only determined by spike number, grain weight, and number of grains per panicle but also affected by plant height and flowering time. Recent studies have shown much progress on exploring the mechanism for regulation of yield potential, plant height, and flowering time in rice. Some genes regulating the rice yield trait have already been identified (Li et al., 2003;Ashikari et al., 2005;Fan et al., 2006;Song et al., 2007;Shomura et al., 2008;Wang et al., 2008;Weng et al., 2008). The control of rice plant height is mostly related to the synthesis and regulation of the phytohormone, such as GA and brassinolide (Ashikari et al., 1999;Yamamuroa et al., 2000;Honga et al., 2003;Sasaki et al., 2002Sasaki et al., , 2003Itoh et al., 2004;Tanabe et al., 2005). Photoperiod determines flowering time as the most important environment signal in plants. Arabidopsis (Arabidopsis thaliana), the model for long-day (LD) species, has been used to reveal how plants control flowering time induced by photoperiod (Corbesier et al., 2007;Kobayashi and Weigel, 2008). A series of genes about phytochrome, clock, and flowering have been found and a signal network has been built up. In LD conditions, CONSTANS (CO) gene in Arabidopsis, being a transcription factor and regulated by GIGANTEA (GI) gene, which is one of the circadian clock genes (Park et al., 1999;Huq et al., 2000;Sothern et al., 2002), activates the expression of FLOWERING LOCUS T (FT; Kardailsky et al., 1999;Kobayashi et al., 1999). After the FT protein moves to the shoot apex, it binds to FD, which is another transcription factor, to trigger genes that determine floral organ identity and then induce flowering (Abe et al., 2005;Wigge et al., 2005;Corbesier et al., 2007). Meanwhile, the short-day (SD) plant rice has a similar photoperiodic control of flowering pathway. Signals from light and circadian clocks are sent to OsGI (a rice ortholog of Arabidopsis GI) and OsGI regulates the expression of Heading date1 (Hd1, a rice ortholog of Arabidopsis CO) and OsMADS51 (Yano et al., 2000;Hayama et al., 2003;Kim et al., 2007). When rice is in SD conditions, Hd1 activates flowering by upregulating Hd3a (a rice ortholog of Arabidopsis FT; Kojima et al., 2002). And OsMADS51 activates Hd3a through Early heading date1 (Ehd1), in which Ehd1 encodes a B-type response regulator and has no obvious counterpart in Arabidopsis (Doi et al., 2004;Kim et al., 2007). By contrast, Hd1 represses flowering by down-regulating Hd3a expression under LD conditions (Kojima et al., 2002). Additionally, RID1/OsID1/ Ehd2, which is homologous to maize (Zea mays) Inde-terminate1 (ID1), promotes flowering in both SD and LD conditions by up-regulating Ehd1 (Matsubara et al., 2008;Park et al., 2008;Wu et al., 2008). Ghd7, encoding a CCT domain protein, has been proven to have major effects on an array of traits in rice, including number of grains per panicle, plant height, and flowering time (Xue et al., 2008); it represses Ehd1 and Hd3a expression and then delays flowering in LD conditions.
As one of the most common promoter elements found in a large number of genes in animals, fungi, and plants, the CCAAT box is recognized by the HAP complex, which consists of three subunits: HAP2 (NF-YA/CBF-B), HAP3 (NF-YB/CBF-A), and HAP5 (NF-YC/CBF-C; Maity and de Crombrugghe, 1998;Mantovani, 1999). In yeast (Saccharomyces cerevisiae) and mammals, each of the HAP subunits is encoded by a single gene (Mantovani, 1999), different from those in plants, in which all subunits are encoded by gene families (Gusmaroli et al., 2001(Gusmaroli et al., , 2002. The Arabidopsis genome encodes 10 AtNF-YA, 13 AtNF-YB, and 13 AtNF-YC subunits (Gusmaroli et al., 2002). The rice genome encodes 10 OsHAP2, 11 OsHAP3, and seven OsHAP5 subunits (Thirumurugan et al., 2008). Each subunit of the HAP complex contains an evolutionary conserved domain that mediates DNA binding or protein-protein interaction. Recent research suggests that the functional domain conserved within HAP2 and the CCT domain of CO share important residues, which are critical to the function of the CCT domain. CO might replace AtHAP2 in the HAP complex to form a trimetric CO/AtHAP3/AtHAP5 complex and flowering was delayed by overexpression of AtHAP2 or AtHAP3 in Arabidopsis (Wenkel et al., 2006). Also, two NF-YB/ HAP3 genes, LEAFY COTY-LEDON1 and LEC1-LIKE, were shown to be important regulators of embryogenesis (Lotan et al., 1998;Kwong et al., 2003). In rice, OsHAP3A or its homologs, OsHAP3B and OsHAP3C, were shown to control chloroplast biogenesis (Miyoshi et al., 2003). However, the functions of other members of the NF-YB/HAP3 gene families and all members of the NF-YA/HAP2 and NF-YC/HAP5 in Arabidopsis and rice are still unknown.
In this study, we identified a rice HAP3 gene, DTH8, by map-based cloning. Our data showed that DTH8 suppressed rice flowering under LD conditions and we surmise that it probably plays an important role in the regulation of plant height and yield potential in rice.

A Chromosome Segment of DTH8 Has a Remarkable Effect on Flowering Time, Plant Height, and Yield
Quantitative trait loci (QTL) analysis for days to heading was conducted over 5 years in natural LD (NLD) conditions (daylength . 14 h) based on a recombinant inbred line population (RIL), derived from two parents, cv Asominori (japonica) and cv IR24 (indica; Supplemental Fig. S1). One major QTL, DTH8, near marker R2976 on the short arm of chromosome 8 has been detected in all 5 years (Supplemental Table S1). Several chromosome segment substitution lines (CSSLs), such as CSSL8 and CSSL61, carrying a chromosome segment of IR24 on DTH8 under the Asominori genetic background, showed earlier heading and its plant height, number of grains per panicle, yield, and dry weight per plant decreased significantly in NLD condition compared with Asominori ( Fig. 1; Table I; Supplemental Fig. S2; Supplemental Table  S2), but no significant differences in these agronomic traits have been found in natural SD (NSD) conditions (daylength about 11.6 h; Fig. 1; Supplemental Table  S3). There was no difference in tiller number between CSSL61 and Asominori under either NLD or NSD conditions ( Fig. 1; Supplemental Tables S2 and S3).

DTH8 Encodes a Putative HAP3 Subunit of the CCAAT-Box-Binding Transcription Factor
An analysis of two secondary F2 populations, derived from the cross between CSSL8, CSSL61, and Asominori, showed that earlier heading and later heading plants of the two secondary F2 populations segregated as 62:164 and 56:150, respectively (x 2 = 0.71, 0.52 , x 2 2,0.05 = 3.84, P . 0.05; Supplemental Fig. S3), which suggests that DTH8 is a single Mendelian factor. Then, map-based cloning of DTH8 was performed. A high-resolution linkage analysis demonstrated that DTH8 is delimited within the interval between two insertion-deletion (Ind) markers, Ind8-47 and Ind8-15. This region was about 47 kb in the bacterial artificial chromosome contig OSJNBa0054L03 (http://www. gramene.org; Fig. 2A).
The genome annotation software, RiceGAAS (http:// ricegaas.dna.affrc.go.jp), predicts 12 open reading frames (ORFs) in this region ( Fig. 2A). One of them, ORF4, encodes a putative HAP3/ NF-YB/CBF-A subunit of the CCAAT-box-binding transcription factor, and has been annotated as a flowering-time gene, Hd5, in the database (Fig. 2C). There are 11 homologous genes (OsHAP3A-OsHAP3K) encoding HAP3 subunits in the genome of rice and ORF4 is named OsHAP3H (Thirumurugan et al., 2008;Fig. 2D;Supplemental Fig. S4). Without any intron, the reading frame of OsHAP3H totals 894 bp, encoding a protein of 297 amino acids in cv Asominori (Fig. 2, B and C). However, in cv IR24, 1-bp deletion was found in a position 322 bp away from start codon, causing a frameshift and premature termination of translation and leading to a mutant protein with only 125 amino acids (Fig. 2B).
To confirm that OsHAP3H is responsible for the phenotype of DTH8, the OsHAP3H cDNA from cv Asominori was expressed under CSSL61 background. The cDNA was inserted into pPZP2H-lac binary vector (Fuse et al., 2001) and 13 independent transgenic T0 lines were generated, which were planted in NSD conditions. The segregation ratio between transgenepositive and transgene-negative plants in T1 families from T0 plants with single-copy transgene was about 3:1 (Table I). Perfect cosegregation between the transgene and the phenotype was also observed ( Fig. 3; Table I): comparing with transgene-negative plants, days to heading, plant height, and number of grains per panicle of all transgene-positive plants increased significantly. Taken together, DTH8, which encodes a putative HAP3/NF-YB/CBF-A subunit of the CCAATbox-binding transcription factor, has a large effect on flowering time, plant height, and yield potential in rice.

Expression Pattern and Subcellular Localization of DTH8
Quantitative real-time PCR (QRT-PCR) analyses were done with total RNA isolated from leaves, leaf sheaths, culms, young panicles, and roots from cv Asominori grown under NLD conditions (Fig. 4A).
The results revealed that DTH8 was expressed in all examined tissues: The expressions were higher in roots, young panicles, and leaves than those in leaf sheaths and culms. To reconfirm the results above, the GUS gene driven by the DTH8 promoter was transformed into rice (Fig. 4B). Staining of transgenic plants verified again the DTH8 expression in the same tissues examined with QRT-PCR. The staining also showed more accumulation in roots, young panicles, and leaves, which matched the QRT-PCR findings. Subcellular localization of DTH8 protein was carried out by fusing DTH8 and GFP, which was driven by the cauliflower mosaic virus 35S promoter (35S:: DTH8:GFP) and delivered into onion (Allium cepa) epidermal cells by particle bombardment. The GFP signal was localized in the nucleus (Fig. 4C), indicating that DTH8 is a nuclear protein. This result also supports the hypothesis that DTH8 acts as a subunit of the CCAAT-box-binding transcription factor. To identify potential downstream genes that are regulated by DTH8, QRT-PCR was performed to Figure 1. Phenotypes of Asominori and CSSL61. A, Photos of Asominori and CSSL61, taken when CSSL61 reached maturity. B, Photos of Asominori and CSSL61, taken when Asominori reached maturity. C, Main panicles of Asominori and CSSL61. D, Grains from the main panicles of Asominori and CSSL61. A to D, Asominori and CSSL61 were grown under NLD conditions. E, Photos of Asominori and CSSL61, taken under NSD conditions. A to E, The left is Asominori and right is CSSL61. F, Days to heading of Asominori and CSSL61 in different daylength conditions. G, The genotype of CSSL61.
analyze the expression levels of DTH8 and seven flowering-related genes (Hd1, Ehd1, Hd3a, Ghd7, OsGI, OsMADS51, and RID1) in both Asominori and CSSL61. Leaf samples from 40-d-old plants grown under SD, LD, and NLD conditions were checked. No differences in DTH8 expression were found between Asominori and CSSL61 under all three conditions (  Table S3).
Also, the diurnal expression patterns of DTH8, OsGI, Hd1, Ehd1, Hd3a, and Ghd7 were checked in both Asominori and CSSL61 (Supplemental Fig. S6). The diurnal pattern of DTH8 is similar to that of OsGI in both LD and SD conditions. As in the single measurements at the onset of the light phase, severe reductions in Ehd1 and Hd3a mRNA were detected in both day and night cycles under LD conditions in Asominori. Meanwhile, little change was found in the diurnal patterns of OsGI, Hd1, Ehd1, Hd3a, and Ghd7 expressions between Asominori and CSSL61 in both LD and SD conditions.

The Expression of DTH8 Is Not Influenced by Hd1 and Ghd7
Previous studies show that the expression of Ehd1 and Hd3a could be down-regulated by Ghd7 (Xue et al., 2008) and Hd3a can also be suppressed by Hd1 under LD conditions (Kojima et al., 2002;Hayama et al., 2003). In this study, our data showed that under LD conditions, DTH8 could negatively influence the expression of Ehd1 and Hd3a but not Hd1 and Ghd7 To examine whether the expression of DTH8 is influenced by Hd1 and Ghd7, DTH8 expression was examined in Hd1 and Ghd7 near-isogenic lines (NILs) grown for 40 d under LD conditions. As shown in Figure 6A and Supplemental Figure S7A, the expression levels of DTH8 and Ghd7, compared with those in Nipponbare, were not changed in NIL (hd1) grown under LD conditions. However, Ehd1 and Hd3a were suppressed by Hd1 in Nipponbare. Similarly, no changes were found in the levels of DTH8 and Hd1 mRNA in the Ghd7 NILs under LD conditions. But the expressions of Ehd1 and Hd3a were also suppressed by Ghd7 ( Fig. 6B; Supplemental Fig.  S7B). These results indicate that DTH8 suppresses rice flowering by down-regulating the expression of Ehd1 and Hd3a, and that the expression of DTH8, Hd1, or Ghd7 is independent of each other under LD conditions.

DTH8 Regulates the Stem Growth
During maturity, the average height of CSSL61 and transgene-negative plants was approximately 84% of Asominori and transgene-positive plants (Figs. 1B and 3B; Table I). The internode elongation patterns between CSSL61 and Asominori were compared (Fig. 7;Supplemental Fig. S8) and the data showed that the panicles and internodes of CSSL61 were remarkably shorter than those of Asominori. Meanwhile, the diameter of stems between them was almost the same. To determine whether the dwarf phenotype Table I. Phenotypes of Asominori, CSSL61, transgene-positive, and transgene-negative segregants in the T1 generation derived from T0 plants with single-copy transgenes under NLD conditions Family 1 and 2 are the T1 plants derived from two independent T0 plants with single-copy transgene. Data are mean 6 SE. (+) and (2)  of the CSSL61 results from defective cell division and/or cell elongation, longitudinal sections of internodes III and IV of CSSL61 culms were compared with their Asominori counterparts. As shown in Figure 7, C, F, and G, the total cell number in the y axes in internodes III and IV of CSSL61 was not significantly less than that of Asominori, while the cell length of CSSL61 was significantly shorter. This result suggests that internode cell elongation is the main reason for the dwarf phenotype of CSSL61 and that DTH8 probably plays an important role in stem growth and development. Vertical lines without labels represent single-base substitutions between Asominori and IR24. Small rectangular boxes and arrowheads represent deletions and insertions, respectively. C, DTH8 cDNA and predicted amino acid sequence. Letters with an asterisk indicate 1-bp deletion IR24, and the amino acids with gray background represent the conserved domains for DNA binding or protein-protein interaction. F.S. and STOP represent frameshift mutation and create premature stop codon, respectively. D, A phylogenic tree of HAP3 protein families in rice, Arabidopsis, yeast (ScHAP3, NP_009532), and human (HsNF-B, L06145). The unrooted tree was generated based on the amino acid sequences of the conserved domain using the program DNAMAN. The scale represents the substitution percentage per site, and the similarity to DTH8 in the conserved domain of each HAP protein is indicated in parentheses. We investigated the diversity of the coding sequences of DTH8 among 40 rice varieties from a wide geographic range in Asia (Supplemental Table  S4). The DTH8 alleles in these varieties were grouped into nine types and eight distinct proteins were identified (Fig. 8). Frameshift mutations or premature stops caused by the deletions of some nucleotides were found in types 7, 8 (IR24), and 9 (Fig. 8). So, the DTH8 proteins of these three allele types seem nonfunctional. For other types, the protein of type 2 is same as that of type 1 (Asominori), and types 3 to 6 just have several SNPs and indels that could cause single amino acid substitution, addition, or deletion (Fig. 8). Therefore, the proteins encoded by types 2 to 6 are probably functional. To confirm this, the DTH8 alleles of type 4 (from cv Kasalath), type 5 (from cv 93-11), and type 6 (from cv IR64) were transformed into the CSSL61 (with nonfunctional DTH8 allele from IR24) background. The results showed that all transgene-positive plants in the T1 families were tall and late heading with large panicles, whereas all transgene-negative plants have phenotypes with opposite features (Supplemental Fig.  S9; Supplemental Table S5). We can thus separate the cultivars into two groups, functional and nonfunctional DTH8 alleles ( Fig. 8; Supplemental Table S4). The photoperiod sensitivity (PS) of flowering time in these varieties was also analyzed using PS index (PS index = [DTH NLD 2 DTH SD ]/DTH NLD ; DTH, NLD, and SD represent days to heading, NLD, and SD conditions, respectively; Supplemental Table S4). The results showed that cultivars that carry the functional DTH8 alleles tended to have larger PS indexes and that their flowering times were more sensitive to photoperiod, whereas those carrying nonfunctional DTH8 alleles tended to show smaller PS indexes and less photoperiod-sensitive flowering times ( Fig. 8; Supplemental Table S4).

DTH8 Has Pleiotropic Effects
Yield, plant height, and flowering time are the three most important agronomic traits in rice. Grain yield is mainly determined by factors such as spike/tiller  A, Tissue-specific expression pattern revealed by QRT-PCR. RNA was isolated from leaves (L), leaf sheaths (S), culms (C), young panicles (P), and roots (R) from Asominori when it was heading in NLD conditions. B, GUS expression in different tissues driven by DTH8 promoter under NLD conditions. a to f, Root, culm, node, leaf blade, leaf sheath, and young panicles. C, Nuclear localization of DTH8. a and e, Bright-field images of onion epidermal cells. b, Nuclear localization of DTH8-GFP fluorescence. c, The same cells stained with DAPI. d, The merged image of a, b, and c. f, Onion epidermal cells bombarded with the construct having GFP alone as the control. g, The merged image of e and f. number, number of grains per panicle, and grain weight. Recently, some genes related to yield have been cloned, e.g. MOC1, which initiates axillary buds and plays an important role in the control of rice tillering (Li et al., 2003); Gn1a controls the number of grains per panicle ; GS3, GW2, and qSW5/GW5 regulate grain size (grain length and grain width) and grain weight (Fan et al., 2006;Song et al., 2007;Shomura et al., 2008;Weng et al., 2008); and GIF1 controls rice grain filling . More genes for plant height in rice have been identified (Ashikari et al., 1999;Yamamuroa et al., 2000;Sasaki et al., 2002Sasaki et al., , 2003Honga et al., 2003;Itoh et al., 2004;Tanabe et al., 2005). The molecular genetic pathway for flowering time has been under elucidation in rice since the isolation of many correlative genes or QTLs (Yano et al., 2000;Kojima et al., 2002;Doi et al., 2004;Kim et al., 2007;Wu et al., 2008). Although there are close correlations among yield, plant height, and flowering time, most of these genes were reported to be regulating only one of the three traits. Recently, one QTL, Ghd7, has been proven to have large pleiotropic effects on an array of traits, including grain number, flowering time, and plant height (Xue et al., 2008). In this study, a rice HAP3 gene, DTH8, was isolated by map-based cloning. Its large pleiotropic effects were likewise demonstrated. DTH8 also delayed rice flowering by negatively influencing the expression of Ehd1 and Hd3a, being independent of Hd1 under LD conditions. Simultaneously, it plays an important role in regulating plant height and grain number. Meanwhile, the differences between DTH8 and Ghd7 were also very obvious. Tiller number in Ghd7 NILs with functional or nonfunctional alleles differed from each other (Xue et al., 2008); those between Asominori and CSSL61 are the same ( Fig. 1; Supplemental Tables S2 and S3). Ghd7 regulates the height through total cell number in the y axes of internodes and it also influences culm thickness, whereas DTH8 regulates height by altering cell Figure 5. The expression of DTH8 and other flowering-time genes in Asominori and CSSL61. QRT-PCR was performed with total RNA from leaves of 40-d-old plants under SD and LD conditions. Samples were collected at the initiation of the light phase (ZT 0 h). These experiments were repeated at least three times. Figure 6. The expression of DTH8 and other flowering-time genes in the NILs of Hd1 and Ghd7 under LD conditions. A, The expression in cv Nipponbare (Nip) and NIL (hd1; carries hd1 introgressed from cv Kasalath in a cv Nipponbare genetic background). B, The expression in NIL (Ghd7; carries functional Ghd7 from Minghui63) and NIL (ghd7; carries nonfunctional ghd7 from Zhen-shan97). QRT-PCR was performed with total RNA from leaves of 40-dold plants under LD conditions. Samples were collected at the initiation of the light phase (ZT 0 h). These experiments were repeated at least three times. elongation in the internodes, but keeps culm thickness stable (Fig. 7).

DTH8 Is a Novel Flowering Suppressor
Rice is a SD plant and most rice cultivars could be induced to flower more rapidly under SD conditions than under LD conditions. In SD conditions, signals from light and circadian clocks are received by OsGI, and it regulates the expression of Hd1 and OsMADS51 (Izawa et al., 2003;Kim et al., 2007). Hd1 induces rice flowering by up-regulating Hd3a expression (Kojima et al., 2002), while OsMADS51 promotes flowering by activating the expression of Hd3a through Ehd1 (Doi et al., 2004;Kim et al., 2007). In LD conditions, the expression of Hd3a is suppressed by Hd1 (Hayama et al., 2003), and Ehd1 and Hd3a are also suppressed by Ghd7 (Xue et al., 2008), which accounts for the late flowering in rice. Additionally, RID1/OsID1/Ehd2 promotes flowering under both SD and LD conditions by up-regulating Ehd1 (Matsubara et al., 2008;Park et al., 2008;Wu et al., 2008). In this study, DTH8 was found to delay rice flowering time by negatively influencing the expression of Ehd1 and Hd3a in LD conditions ( Fig. 5; Supplemental Fig. S5). But, under SD conditions, the expression of Hd3a and rice flowering time are not influenced by DTH8, and neither are the expressions of RID1, OsGI, Hd1, and Ghd7 in LD conditions ( Fig. 5; Supplemental Fig. S5). Meanwhile, the expression of DTH8 is independent of Hd1 and Ghd7 in LD conditions ( Fig. 6; Supplemental Fig. S7), suggesting that DTH8 is a flowering suppressor by down-regulating the expression of Ehd1 and Hd3a under LD conditions. There are many studies to have reported that Ehd1 locates in the upstream of Hd3a and could promote its expression (Doi et al., 2004;Kim et al., 2007;Matsubara et al., 2008;Park et al., 2008;Wu et al., 2008). At the same time, some other reports show that the flowering suppressor could repress Hd3a expression by down-regulating Ehd1 and then inhibits flowering (Xue et al., 2008). So we propose that DTH8 is upstream of Ehd1, which in turn is upstream of Hd3a in the pathway (Fig. 9). The diurnal expression pattern of DTH8 is similar to that of OsGI in both LD and SD conditions (Supplemental Fig. S6), suggesting that DTH8 probably locates downstream of OsGI in the pathway that regulates photoperiodic flowering time (Fig. 9). Because there is no Ehd1 ortholog in Arabidopsis (Doi et al., 2004), the DTH8-Ehd1-Hd3a relation probably represents a distinct pathway that does not exist in Arabidopsis.
Previous research shows that Hd1 and Ehd1 regulate the expression of Hd3a through two different pathways (Doi et al., 2004). However, the results from this study showed that the expression of Ehd1 is suppressed by Hd1 in LD conditions ( Fig. 6A; Supplemental Fig. S7A). So Ehd1 may be the integrator of different pathways in the regulation of flowering, which is down-regulated by Hd1, Ghd7, and DTH8 but up-regulated by RID1. Then, the expression of Hd3a or its orthologs was activated by Ehd1 and the transition from vegetative development to floral development in rice was induced under LD conditions (Fig. 9).

Possible Reason for the Pleiotropic Effects of DTH8
The protein encoded by DTH8 is a putative HAP3/ NF-YB/CBF-A subunit of the HAP complex, Os-HAP3H (Fig. 2), which binds to CCAAT box that is the cis-acting element in about 25% of eukaryotic promoters (Maity and de Crombrugghe, 1998;Mantovani, 1999). The HAP complex consists of three subunits: HAP2/NF-YA/CBF-B, HAP3/NF-YB/CBF-A, and HAP5/NF-YC/CBF-C. Each subunit of the HAP complex contains a conserved domain that is responsible for DNA binding or protein-protein interaction. Although many genes encoding HAP subunits have been found so far (Edwards et al., 1998;Gusmaroli et al., 2001Gusmaroli et al., , 2002Thirumurugan et al., 2008), only a few of them have been studied (Lotan et al., 1998;Meinke, 1992;West et al., 1994;Kwong et al., 2003;Miyoshi et al., 2003). In this study, a rice HAP3 gene, DTH8, has been proven to play an important role in the regulation of flowering time, plant height, and yield potential. Because of the existence of multiple HAP2 and HAP5 subunit orthologs, the HAP complexes OsHAP2/DTH8/OsHAP5 should not be the only one that may bind to the CCAAT boxes in those genes working in different metabolic pathways. The genes controlling the number of primary branches and second branches are regulated by the OsHAP2/ DTH8/OsHAP5 complexes, which could influence the number of grains per panicle. Rice height was affected by DTH8 or OsHAP complexes mainly through the regulation of internode cell elongation (Fig. 7). According to previous reports, the control of rice height is mostly related to the synthesis and regulation of phytohormones (Ashikari et al., 1999;Yamamuroa et al., 2000;Sasaki et al., 2002Sasaki et al., , 2003Honga et al., 2003;Itoh et al., 2004;Tanabe et al., 2005). The OsHAP2/DTH8/OsHAP5 complexes may be involved in the biosynthesis or degradation of phytohormones.
In this study, we found that DTH8 delayed rice flowering time by down-regulating the expression of Ehd1 and Hd3a in LD conditions (Fig. 5). But the expressions of DTH8, Hd1, and Ghd7 were not influenced by each other under LD conditions (Figs. 5 and 6), suggesting that the down-regulation of Ehd1 and Hd3a by DTH8 may be independent of Hd1 and Ghd7 under these conditions (Fig. 9). Recent research has shown that the conserved domain of HAP2 and the CCT domain of CO are similar to each other, where the most conserved residues are very important for the functioning of the CCT domain. CO could replace AtHAP2 in the HAP complex to form a trimetric complex, CO/AtHAP3/AtHAP5 (Wenkel et al., 2006). In LD conditions, flowering was postponed by overexpression of AtHAP2 or AtHAP3 in Arabidopsis. Meanwhile, FT expression was going down while the level of CO remained stable. The mRNA abundance of AtHAP3a shows a diurnal rhythm and is regulated by the flowering-time gene GI (Wenkel et al., 2006). The rice flowering genes Hd1 and Ghd7 also have the CCT domains (Yano et al., 2000;Xue et al., 2008) so they may  be able to replace OsHAP2 in the HAP complex. Thus, we speculate that the down-regulation of Ehd1 and Hd3a by DTH8, Hd1, and Ghd7 might be mediated by forming the possible complexes Hd1/DTH8/OsHAP5 and Ghd7/DTH8/OsHAP5. Of course, more evidences are needed to test this hypothesis.

Relationship between DTH8 and PS
A high degree of polymorphism in the DTH8 sequences was identified, some of which cause frameshift mutations or create premature stop codons. The DTH8 alleles in the core collection were grouped into nine types and eight distinct proteins were identified (Fig. 8). All the cultivars were separated into groups with functional or nonfunctional DTH8 alleles, depending on the results of transgenic analyses (Fig. 8;Supplemental Fig. S9;Supplemental Tables S4 and S5). Being a SD plant, the flowering time of rice is sensitive to variation in photoperiod induced by SD conditions and blocked by LD conditions. But with the long period of introduction and domestication, some rice varieties have become insensitive to photoperiod variation and flowering time of rice show little or even no difference when they are grown under either SD or LD conditions. Varieties with weak PS are usually grown in high-latitude area or in mid-and low-latitude area as early season rice.
The PS of flowering time for 40 varieties was also analyzed by looking at the PS index (Supplemental Table S4). The results showed that flowering times of cultivars that carry functional DTH8 alleles are more sensitive to photoperiod, whereas those with nonfunctional DTH8 alleles are less sensitive ( Fig. 8; Supplemental Table S4). However, there are still some exceptions, such as cv Jingnong1, cv XX-4, and cv 93-11. The flowering times of these varieties are weakly sensitive to photoperiod, even though they carry the functional DTH8 alleles (Supplemental Table S5). The possible reason is that these cultivars may contain other unknown genes that can weaken their PS. The cultivars having nonfunctional DTH8 alleles and showing weaker PS are grown as early season rice in mid-and low-latitude areas (Supplemental Table S5). Thus, it might be a better choice to breed rice cultivars with nonfunctional DTH8 alleles in high-latitude areas or as early season rice in mid-and low-latitude areas. The cultivars with weaker PS would be able to adapt well to LD conditions. Similarly, while breeding single or late-season rice for the mid-or low-latitude areas, cultivars with functional DTH8 alleles should be chosen because the flowering of these cultivars could be induced by the SD conditions and their grain yield could be increased by DTH8.

Plant Materials and Growth Conditions
A RIL and several CSSLs derived from two parents, cv Asominori (japonica) and cv IR24 (indica; Kubo et al., 1999;Supplemental Fig. S1), were grown in Nanjing NLD conditions (daylength . 14 h). The materials that include Asominori, CSSL61, Nipponbare, NIL (hd1), NIL (Ghd7), and NIL (ghd7) for the expression of DTH8 and other flowering genes were grown in NLD, LD (15 h light/9 h dark), and SD (9 h light/15 h dark) conditions. Asominori and CSSL61, for the analysis of diurnal expression patterns of flowering genes, were first grown in NLD conditions for 30 d and then transferred to LD or SD conditions for 10 d. The transgene-positive T0 and T1 plants were grown in NSD (about 11.6 h light/12.4 h dark) conditions (Hainan Province, China) and NLD conditions, respectively. The 40 varieties from Asia for genotyping the DTH8 locus were grown under NLD and SD conditions.

Mapping for DTH8
The RILs and several CSSLs derived from Asominori (japonica) and IR24 (indica; Tsunematsu et al., 1996;Kubo et al., 1999) were used for primary mapping of the DTH8 locus (Supplemental Fig. S1). The secondary F2 population containing 15,000 plants was constructed from the cross between CSSL61 and Asominori and 2,000 earliest plants were selected to fine map the DTH8 locus, using the approach described by Zhang et al. (1994). DTH8 was first localized to a 2.3-cM interval between simple sequence repeat (SSR) markers RM22475 and RM5556, based on 200 earliest plants. After more than nine SSR or InDel markers were developed between RM22475 and RM5556, 40 recombinant plants were selected and DTH8 was finally narrowed to the 47-kb region between Ind8-47 and Ind8-15 ( Fig. 2A). The molecular markers including SSR and InDel markers used for fine mapping DTH8 are shown in Supplemental Table S6.

Vector Construction and Transformation
The full-length coding region of DTH8 was isolated by PCR with primers 8 to 16 (Supplemental Table S7) from Asominori, Kasalath, 93-11, and IR64, then subcloned into the pPZP2H-lac binary vector (Fuse et al., 2001). The resultant plasmid was introduced into CSSL61 by means of Agrobacteriummediated transformation (Hiei et al., 1994). The genotype of transgenic plants was determined by PCR amplification of the hygromycin phosphotransferase gene (hpt) and the analysis of hygromycin resistance.
A 2.5-kb sequence of DTH8 promoter region was isolated by PCR amplification with primers Pro-14 (Supplemental Table S7) from Asominori and then subcloned into the pCAMBIA1381Z binary vector. The resulted plasmid was transformed into rice (Oryza sativa) through Agrobacterium-mediated transformation, and the transgenic plants were analyzed by GUS staining assay as described (Scarpella et al., 2003).

RNA Extraction and QRT-PCR
Total RNA from leaves, leaf sheaths, culms, roots, and young panicles were isolated using an RNA extraction kit following the manufacturer's instruction (Beijing Dingguo Biotechnology Co. Ltd., http://www.dingguo.com). Firststrand cDNA was reverse transcribed from DNase I-treated RNA with oligo (dT) as the primer. Gene expression was measured by QRT-PCR and semiquantitative RT-PCR using Ubiquitin and Actin gene as internal controls. The QRT-PCR and semiquantitative PCR primers for Actin, DTH8, OsGI, Hd1, Hd3a, Ehd1, Ghd7, RID1, and OsMADS51 are listed in Supplemental Table S8. The QRT-PCR was carried out in a total volume of 25 mL containing 2 mL of the reverse-transcribed product above, 0.2 mM of each primer and 13 SYBR green PCR master mix (TaKaRa Co. Ltd., http://www.takara.com.cn). The PCR was performed with a Bio-Rad iCycler (http://www.bio-rad.com/) using the following program: 95°C for 30 s, then 40 cycles of 95°C for 5 s, 60°C for 34 s. Changes in gene expression were calculated via the DDCT method. The semiquantitative PCR conditions were 95°C for 3 min, then 26 cycles (cycle numbers for Actin were 26; DTH8 and OsGI were 28; others were 30-35) of 95°C for 30 s, 58°C for 30 s, and 72°C for 30 s.

Subcellular Localization
The DTH8 ORF without a termination codon was cloned into the pA7-GFP vector, placing the DTH8 gene upstream of the GFP coding sequence to create in-frame fusion of DTH8 cDNA and GFP reporter gene. The fusion constructs, as well as the control (pA7-GFP), were transformed into onion (Allium cepa) epidermal cells by particle bombardment. The transformed onion epidermal cells were incubated at 22°C on Murashige and Skoog plates in the dark for 48 h. Then, the cells were examined under confocal fluorescence microscopy.

Scanning Electron Microscopic Observation
Culms of Asominori and CSSL61 were harvested 10 d after flowering and fixed in 2.5% v/v glutaraldehyde. Then, the samples were post fixed in 2% w/v OsO4 for 2 h, dehydrated through an ethanol gradient, infiltrated, and embedded in butyl methyl methacrylate. The samples conducted with critical point dry and sputter coated with platinum and observed using a scanning electron microscope (S-3000N, Hitachi).
Sequence data from this article can be found in the GenBank/EMBL data libraries under accession numbers HM775396 and HM775397.

Supplemental Data
The following materials are available in the online version of this article.
Supplemental Figure S1. The flow chart of genetic study of DTH8 in this article.
Supplemental Figure S2. Flowering times of Asominori and CSSLs carrying the chromosome segments including DTH8 from the donor parent IR24 with Asominori background under NLD conditions.
Supplemental Figure S3. Distribution of days to heading under NLD conditions of two secondary F2 populations derived from the cross between CSSL8, CSSL61, and Asominori.
Supplemental Figure S5. The expressions of DTH8 and other flowering genes in Asominori and CSSL61.
Supplemental Figure S7. The expression of DTH8 and other flowering genes in the NILs of Hd1 and Ghd7 under LD conditions.
Supplemental Figure S8. Plant height phenotype of cv Asominori and CSSL61.
Supplemental Table S1. DTH8 was mapped over 5 years in NLD conditions based on a RIL population derived from two parents, cv Asominori (japonica) and cv IR24 (indica).
Supplemental Table S2. Phenotypes of Asominori and CSSLs carrying the chromosome segments including DTH8 from the donor parent IR24 with Asominori background under NLD conditions.
Supplemental Table S3. Phenotypes of Asominori and CSSL61 under NSD conditions.
Supplemental Table S4. DTH8 alleles and its PS of flowering time among 40 rice varieties from a wide geographic range of Asia.
Supplemental Table S5. Phenotypes of transgenic and nontransgenic segregants in the T1 generation derived from T0 plants with single-copy gene grown under NLD conditions.
Supplemental Table S6. The sequences of primers used for mapping DTH8.
Supplemental Table S7. Primers for vector constructions in this article.
Supplemental Table S8. Primers for QRT-PCR and semiquantitative RT-PCR in this article.