Temporal and spatial regulation of DROOPING LEAF gene expression that promotes midrib formation in rice

Authors

  • Yoshihiro Ohmori,

    1. Department of Biological Sciences, Graduate School of Science, The University of Tokyo, Hongo, Bunkyo-ku, Tokyo 113-8654, Japan
    Search for more papers by this author
  • Taiyo Toriba,

    1. Department of Biological Sciences, Graduate School of Science, The University of Tokyo, Hongo, Bunkyo-ku, Tokyo 113-8654, Japan
    Search for more papers by this author
  • Hidemitsu Nakamura,

    1. National Institute of Agrobiological Sciences, Kannondai, Tsukuba 305-8602, Japan
    Search for more papers by this author
    • Present address: Department of Applied Biological Chemistry, Graduate School of of Agricultural Life Sciences, The University of Tokyo, Yayoi, Bunkyo-ku, Tokyo 113-8657, Japan.

  • Hiroaki Ichikawa,

    1. National Institute of Agrobiological Sciences, Kannondai, Tsukuba 305-8602, Japan
    Search for more papers by this author
  • Hiro-Yuki Hirano

    Corresponding author
    1. Department of Biological Sciences, Graduate School of Science, The University of Tokyo, Hongo, Bunkyo-ku, Tokyo 113-8654, Japan
    Search for more papers by this author

(fax +81 3 5841 4056; e-mail hyhirano@biol.s.u-tokyo.ac.jp).

Summary

Genes involved in the differentiation and development of tissues and organs are temporally and spatially regulated in plant development. The DROOPING LEAF (DL) gene, a member of the YABBY gene family, promotes midrib formation in the leaf and carpel specification in the flower. Consistent with these functions, DL is initially expressed in the central region of the leaf primordia (presumptive midrib) and in the presumptive carpel primordia in the meristem. To understand the regulatory mechanism underlying DL expression, we tried to identify cis-regulatory regions required for temporal and spatial expression of this gene. We found that the cis region responsible for the presumptive midrib-specific expression in the leaf primordia is located in intron 2. Next, we confined the region to a sequence of about 200 bp, which corresponds to a conserved non-coding sequence (CNS) identified by phylogenetic footprinting. In addition, a sequence termed DG1, incorporating a 5′ upstream region of about 7.4 kb, and introns 1 and 2, was shown to be sufficient to induce DL in the presumptive midrib, and to suppress it in other regions in the leaf primordia. By contrast, the regulatory region required for carpel-specific expression was not included in the DG1 sequence. We modified Oryza sativa (rice) plant architecture by expressing an activated version of DL (DL-VP16) in a precise manner using the DG1 sequence: the resulting transgenic plant produced a midrib in the distal region of the leaf blade, where there is no midrib in wild type, and formed more upright leaves compared with the wild type.

Introduction

Plant development is orchestrated by a number of genes involved in fundamental processes of cell activity, such as transcriptional regulation, signal transduction and cell-to-cell communication. The expression of genes functioning in these processes is temporally and spatially regulated in various stages and tissues in plant development. Floral homeotic genes, for example, are specifically expressed in a restricted region of the meristem, the presumptive region where the floral organs develop and in the primordia of those floral organs, and they promote the expression of downstream genes to elaborate floral organ structures (reviewed in Lohmann and Weigel, 2002; Jack, 2004; Bommert et al., 2005; Kater et al., 2006; Thompson and Hake, 2009). Positive and negative regulators of stem cell maintenance are expressed in specific domains of the meristem, govern communication among cells, such as stem cells and cells in the organizing center in the meristem, and maintain stem cell identity (Clark et al., 1997; Mayer et al., 1998; Fletcher et al., 1999; Suzaki et al., 2006, 2008).

Cis-regulatory elements responsible for the temporal and spatial expression of several important developmental genes in Arabidopsis thaliana have been well studied. For example, APETALA3 (AP3), a class-B gene involved in the ABC model, has binding sites for transcription factors, such as LEAFY (LFY), APETALA1, AGAMOUS (AG), PISTILATA (PI) and AP3 itself, in its promoter region (Hill et al., 1998; Tilly et al., 1998). AP3 expression is initially promoted by LFY in the presence of UNUSUAL FLORAL ORGANS (UFO), an F-box protein (Parcy et al., 1998). Later, AP3 expression is maintained by an autoregulatory circuit including AP3 and PI, whereby an AP3-PI heterodimer can bind to the CArG-box in the promoter region of AP3, and continues to facilitate the expression of AP3 (Parcy et al., 1998; Tilly et al., 1998). Regulatory elements are located not only in promoter regions but also in introns. For example, the class-C gene AG has regulatory sequences in its second intron, where binding sites for the transcription factors LFY and WUSCHEL (WUS) are located (Sieburth and Meyerowitz, 1997; Busch et al., 1999; Lohmann et al., 2001). LFY activates the expression of AG by cooperating with WUS (Lohmann et al., 2001).

Although much progress has been recently made in our understanding of the function of genes regulating monocot development (reviewed in Bommert et al., 2005; Kater et al., 2006; Thompson and Hake, 2009), the mechanism underlying the temporal and spatial regulation of these genes is poorly understood. The DROOPING LEAF gene plays two distinct roles in Oryza sativa (rice) development: DL is required for the formation of the midrib in the leaf and for the specification of the carpel in the flower (Nagasawa et al., 2003; Yamaguchi et al., 2004; Ohmori et al., 2008). A loss-of-function mutation in DL results in a leaf lacking a midrib, the strong structure formed in the central region of the leaf, resulting in drooping leaf phenotypes. In the flower, carpels are homeotically transformed into stamens in the null dl mutant. Consistent with these phenotypes in the dl mutant, DL is specifically expressed in the central region of the leaf primordia (presumptive midrib) from the P1 stage, and in a presumptive region of carpel initiation in the flower meristem (Yamaguchi et al., 2004). These specific temporal and spatial expression patterns of rice DL are also observed for DL orthologs in other grasses, including Zea mays (maize), Triticum aestivum (bread wheat) and Sorghum bicolor (sorghum), suggesting that the function of DL-related genes is conserved in grasses (Ishikawa et al., 2009).

DL encodes a putative transcription factor containing a zinc-finger domain and a YABBY domain (Yamaguchi et al., 2004). CRABS CLAW (CRC) is an ortholog of DL in Arabidopsis. CRC plays a role in carpel identity and nectary development in the flower, and is expressed both in the abaxial domains of the carpel primordia and in the nectary (Bowman and Smyth, 1999; Baum et al., 2001). Unlike rice DL, CRC appears to have no function in vegetative development in Arabidopsis. Thus, whereas the regulation of carpel development seems to be an ancestral function of DL and CRC, the role of DL in midrib formation and that of CRC in nectary development may be functions derived within the lineage of rice and Arabidopsis, respectively. CRC has binding sites for LFY and MADS domain proteins in its 5′ upstream region, and the latter site is likely to be essential for the transcriptional regulation of CRC (Lee et al., 2005). MADS box genes such as B- and C-class floral homeotic genes and SEPALLATA genes are required for the activation of CRC.

In this study, we tried to identify the cis-regulatory regions that temporally and spatially regulate DL expression to better understand the function of DL in both leaf and flower development. We found that intron 2 is essential for the expression of DL in the presumptive midrib. In addition, a sequence termed DG1 that incorporates a 5′ upstream region of about 7.4 kb, and introns 1 and 2, was sufficient to induce DL expression in the presumptive midrib from the P1 stage, and to suppress its expression in other parts of the leaf primordia, such as the vascular bundles. By contrast, the regulatory region responsible for carpel development was not included in the DG1 sequence. Using DG1 and an activated version of the DL protein, we have been able to modify rice plant architecture to make leaves more erect by enhancing midrib formation.

Results

Midrib formation and DL expression in the leaf

The midrib (thick central vein) was clearly observed in the central region of the wild-type leaf blade, whereas no such central vein was detected in a severe mutant (dl-sup1) of the DL gene (Figure 1a,b). Because of the lack of the midrib, the leaves were drooping in dl-sup1 (Figure 1d). Thus, formation of the midrib is likely to be essential for the erectness of the leaves. A cross section showed that the midrib consisted of two locules, the central septum and the adaxial small vascular bundle (Figure 1e). The locules constituted hollow thin cylinders aligned from proximal to distal in the wild-type leaf, and these cylindrical structures seemed to support the whole leaf in an erect manner in the wild type. No structures, such as the locules or septum, were found in the dl-sup1 leaf (Figure 1f).

Figure 1.

 Plant phenotypes and in situ localization of the DL transcript.
(a, b) Abaxial surface of the leaf blade in the wild type (a) and in dl-sup1 (b).
(c, d) Plant architecture of the wild type (c) and dl-sup1 (d). The arrow indicates a drooping leaf.
(e, f) Cross section of the leaf blade of the wild type (e) and dl-sup1 (f).
(g, h) Cross section of the shoot apex of the wild type (g) and dl-sup1 (h).
(i) Localization of DL transcripts in developing leaves.
Bracket indicates the midrib. Plastochron numbers are labeled P1–P4; asv, adaxial small vascular bundle; clv, central large vascular bundle; se, septum.
Scale bars: 100 μm.

Prior to formation of the midrib, the central region of the leaf primordia was thickened by cell proliferation along the adaxial–abaxial axis from the P2 to the P5 stage (Figure 1g). By contrast, this thickening was not observed in the leaf primordia of dl-sup1 (Figure 1h). In situ analysis revealed that DL was specifically expressed in the central region of the leaf primordia, i.e. the presumptive region of the midrib, from the P1 to the P4 stage (Figure 1i; Yamaguchi et al., 2004; Ohmori et al., 2008). This specific expression of DL in the presumptive midrib may be responsible for promoting cell proliferation in the central region of the leaf primordia. In addition, DL was expressed in the carpel primordia of the flower (Figure 5c,d), consistent with a defect in carpel development in the dl-sup1 flower (Yamaguchi et al., 2004; Ishikawa et al., 2009). No expression was detected in other parts of the rice plant. Thus, DL expression was strictly regulated in an organ-specific manner. Especially in the leaf, DL expression was restricted to a special type of cell that subsequently differentiates into the midrib.

Figure 5.

 GUS staining patterns of the DG1::GUS transgenic plants and in situ localization of the DL transcripts in the flower.
(a, b) GUS staining pattern in transgenic lines (T0 generation) carrying the DG1::GUS construct.
(c, d) Spatial expression pattern of DL in the wild-type flower.
The red arrowheads indicate GUS staining (b) and DL transcripts (d) in the midrib region of the lemma.
ca, carpel; fm, floral meristem; le, lemma; ov, ovule, st, stamen.
Scale bars: 20 μm.

Identification of conserved non-coding sequences

To identify the regulatory regions responsible for organ- or tissue- specific expression of DL, we first performed phylogenetic footprinting (Figure 2). Our previous study showed that the temporal and spatial expression patterns of DL and its orthologs are highly similar to each other in rice, maize, wheat and sorghum, suggesting that a common genetic mechanism may underlie the regulation of the expression of these genes (Ishikawa et al., 2009). We then analyzed the genomic sequences of DL-related genes from these four species using mVISTA (Mayor et al., 2000).

Figure 2.

 Phylogenetic footprinting of the DL locus from rice, maize, sorghum and wheat by mVISTA analysis.
A schematic structure of the rice DL locus is shown in the uppermost panel. The protein coding region and untranslated region (5′ and 3′ UTR) are indicated by black and white boxes, respectively. The 5′ and 3′ flanking regions and introns are indicated by unbroken lines. Numbers indicate the position from the putative transcription initiation site (+1). mVISTA analysis was performed using a 50-base sliding window. Regions with >75% identity shared by each comparison are indicated by color: blue for coding regions, and pink for non-coding regions. Five CNSs found by comparing the four DL-related genes are indicated by orange lines and labeled with their names (CNS-A to CNS-E).

The DL gene consists of seven exons and six introns (Yamaguchi et al., 2004). As expected, the exons are highly conserved among the four species (Figure 2). Out of six introns, no conserved sequence was detected in introns 3–6. By contrast, two distinct CNSs, referred to as CNS-D and CNS-E, were found in introns 1 and 2, respectively (Figures 2 and S1). Three CNSs were found in the upstream region: CNS-C was located just upstream of the putative transcription initiation site, and CNS-A and CNS-B were located more than 2 kb from this site. In addition, several CNSs were identified far upstream of the DL-related genes by the comparison of rice, maize and sorghum, although no sequence information was available in this region in wheat.

The regulatory region responsible for specific DL expression in the presumptive midrib

To detect organ- or tissue-specific expression of DL, we made various constructs incorporating the upstream region, introns 1 and 2, termed DG1–DG8, and fused each construct to the GUS reporter gene (Figure 3a). The resulting chimeric plasmids were introduced into the calli, and transgenic plants were regenerated. We generated more than 20 transgenic plants for each construct. About 50–90% of transgenic plants showed similar GUS expression patterns, although intensities of the signal varied among transgenic plants.

Figure 3.

 Schematic representation of the DL::GUS reporter constructs and GUS staining patterns.
(a) The white and black boxes indicate the 5′ untranslated region and exons, respectively. Unbroken lines indicate non-coding sequences such as the 5′ upstream region, and introns 1 and 2. Deleted regions in each construct are shown by the dotted lines. Vertical broken lines with numbers indicate the position from the putative transcription initiation site (a red broken line, +1). The modified DL sequences (DG1–DG8) are fused to the GUS gene encoding β-glucuronidase (GUS) from Escherichia coli (translational fusion). The CNSs in Figure 2 are indicated by the orange boxes (A–E).
(b–i) GUS expression patterns in a cross section of the shoot apex of transgenic lines (T1 generation) carrying DG5::GUS (b), DG4::GUS (c), DG3::GUS (d), DG6::GUS (e), DG7::GUS (f), DG8::GUS (g), DG1::GUS (h) and DG2::GUS (i).
(j, k) Close-up view of the vascular region in DG1::GUS (j; red dashed box in h) and DG2::GUS (k; blue dashed box in i).
Outlines of the leaf primordia (P1–P3) are traced by unbroken lines.
as, abaxial sclerenchyma; v, vascular bundle.
Scale bars: (b–i) 100 μm; (j, k) 30 μm.

We first analyzed transgenic plants carrying the DG3::GUS, DG4::GUS and DG5::GUS constructs, which contained about 2.2 kb of the promoter region (hereafter, we refer to these plants as DG3::GUS, DG4::GUS and DG5::GUS plants). In the DG5::GUS plant, the transgene of which lacked both introns 1 and 2, no GUS signals were detected in P1 and P2 primordia, and very weak signals were detected in P3 and P4, in addition to the vascular bundle (Figure 3b). In the DG4::GUS plant, the transgene of which lacked only intron 2, the expression levels of GUS in P3 and P4 were enhanced as compared with the DG5::GUS plant (Figure 3c), suggesting that intron 1 may be associated with quantitative gene expression. Although GUS expression expanded to the whole region in P3, no expression was detected in P1 and P2 in the DG4::GUS plant. Thus, the GUS expression pattern in both DG5::GUS and DG4::GUS did not reflect the expression pattern of the DL transcript. In the DG3::GUS plant, by contrast, the transgene of which has both introns 1 and 2, GUS signals were detected from the P1 stage (Figure 3d). In addition, GUS-expressing cells were restricted to the central region of the leaf primordia (P1–P4). These results suggested that intron 2 is essential for the expression of DL in the presumptive midrib from the initial stage (P1).

Next, we deleted the region upstream of the DG3 construct and analyzed the effect of the length of the promoter region. Both the DG6::GUS and DG7::GUS plants exhibited GUS expression patterns similar to those of the DG3::GUS plant (Figure 3e,f), suggesting that the regions deleted in DG6 or DG7, as compared with DG3, were not essential for the control of DL expression. DG7 contains only a 147-bp sequence starting from the initial nucleotide of the longest cDNA. Therefore, CNS-C may be associated with transcriptional initiation, and may be unrelated to the spatial and temporal control of DL.

Removal of intron 1 from DG7 (DG8::GUS plants) resulted in the downregulation of GUS expression (Figure 3g). This result confirmed our above hypothesis that intron 1 is associated with quantitative regulation.

Suppression of other GUS expression

Although the DG3::GUS, DG6::GUS and DG7::GUS plants showed specific expression in the presumptive midrib, other GUS signals were detected in the presumptive regions of the vascular bundles and abaxial sclerenchyma in the leaf primordia (Figure 3d–f). Because the DL transcript is detected only in the central region of the leaf primordia, and is not detected in other regions, it is possible that DG3 lacks a putative region that represses DL expression in other regions. To examine this idea, we made constructs that had a longer upstream region (DG1 and DG2). GUS expression in the vascular bundles and abaxial sclerenchyma disappeared in the DG1::GUS plant, although ectopic DL expression was still detected in the DG2::GUS plant (Figure 3h–k). Therefore, a putative negative element that represses DL expression in the vascular bundle and abaxial sclerenchyma may be located at a distant region (from −7390 to −3300 bp) from the coding region of DL.

Molecular dissection of intron 2

To narrow down the region responsible for the specific expression of DL in the presumptive midrib, we analyzed intron 2 in detail. First, we examined the effect of CNS-E by making the constructs DG7ΔFS (flanking sequences) and DG7ΔCNS-E (Figure 4g). GUS expression was detected in the central region of the leaf primordia from P1 to P3 in the DG7ΔFS::GUS plant, whereas no expression was observed in P1 or P2 in the DG7ΔCNS-E::GUS plants (Figure 4b,c). This result suggests that CNS-E has an important role in the specific expression of DL in the presumptive midrib.

Figure 4.

 Molecular dissection of intron 2.
(a–f) GUS staining patterns in a cross section of the shoot apex of transgenic lines: DG7::GUS (a), DG7ΔFS::GUS (b), DG7ΔE::GUS (c), DG7ΔE1::GUS (d), DG7ΔE2::GUS (e) and DG7ΔE3::GUS (f). (a–c) T1 generation; (d–f) T0 generation. Outlines of the leaf primordia (P1–P3) are traced by unbroken lines.
Scale bars: 50 μm.
(g) Schematic representation of constructs lacking various parts of intron 2.
Five constructs (DG7ΔFS::GUSDG7ΔE3::GUS) were generated on the basis of DG7::GUS. The lines and boxes are defined in the legend of Figure 3a. The CNSs are indicated by orange boxes (C–E). CNS-E in intron 2 is divided into three regions (E1, E2 and E3).

Next, we dissected CNS-E into three parts (E1, E2 and E3) and made constructs, each of which lacked one of the three parts. GUS was roughly expressed in the central region from P1 to P3 in the DG7ΔΕ1::GUS plant, as in the DG7::GUS plant, suggesting that the E1 region may not be essential for the precise expression of DL (Figure 4d). By contrast, the central region-specific expression of DL was disturbed in the DG7ΔΕ2::GUS plant: GUS signals were detected weakly in the whole region in P1 and in the peripheral region in P2 (Figure 4e). The removal of E3 (DG7ΔΕ3::GUS plants) resulted in a more pronounced defect: namely, the disappearance of the GUS signal in P1 (Figure 4f). These results suggest that E2 and E3 in CNS-E may be required for the specific expression of DL in the presumptive midrib in the leaf primordia.

Expression in the flower

Lastly, we examined GUS expression in the flower. Although the DL transcript is expressed in the carpel throughout its development, no GUS activity was detected in the carpel in the DG1::GUS plant (Figure 5). The DG1 construct contained a longer upstream region (about 7.4 kb), and introns 1 and 2. Thus, the regulatory region that promotes DL expression in the carpel seems to be located at either an upstream region far from the DL coding sequence or a region downstream of it.

However, GUS signal was detected in the lemma (Figure 5b). This signal corresponded to the central vein of the lemma, which is thought to be equivalent to the midrib of the leaf blade. No signal was detected in the vascular bundles in any floral organ (Figure 5a,b). These results confirmed the above conclusion that DG1 is responsible for inducing DL expression in the presumptive midrib, and for repressing ectopic DL expression in the vascular bundles of the leaf primordia.

Enhancement of DL activity

Because DL activity is related to the size of the midrib (Yamaguchi et al., 2004; Ohmori et al., 2008), it is expected that reinforcement of DL function will promote the formation of a larger midrib and make leaves more erect. Simple overexpression of DL, however, results in a seedling-lethal phenotype, because the midrib-like structures that form in the lateral region enclose firm younger leaves, and prevent their growth (Yamaguchi et al., 2004).

In this study, we tried to express an activated version of a chimeric DL protein using DG1, which would allow expression of the chimeric DL protein in only the central region of the leaf primordia. To enhance the function of DL, we used a viral transcription factor, VP16, which has strong transcriptional activation activity. We made a fusion gene encoding DL protein and the VP16 transcriptional activation domain, and expressed this DL-VP16 fusion gene under the control of DG1 (DG1::DL-VP16; Figure 6a).

Figure 6.

 Enhancement of DL activity.
(a) Diagram of the DG1::DL-VP16 construct. The DG1 sequence is fused in-frame to DL cDNA (exons 3–7) and the VP16 activation domain.
(b) Wild type (60 days old).
(c–e) Phenotype of the plant architecture in DG1::DL-VP16 (about 60 days after regeneration).
(f, g) and (i, j) Cross sections of the leaf blade in the wild type (f, g) and DG1::DL-VP16 (i, j). Distal (f, i) and proximal (g, j) regions are shown.
(h, k) Phenotypes of the pistil in the wild type (h) and DG1::DL-VP16 (k).
Scale bars: (b–e) 10 cm; (f, g, i, j) 100 μm; (h, k) 500 μm.

As compared with the wild type, the overall plant architecture was changed in DG1::DL-VP16 transgenic lines (Figure 6b–e). The leaves in the transgenic lines were more erect, although the erectness of the leaves varied among the transgenic lines. In the wild type, several leaves drooped in the distal region (Figure 6b). By contrast, leaves were erect even in the distal region in DG1::DL-VP16 plants (Figure 6c–e). In addition, the leaves stood upright with more sharp angles in the lamina joint region in lines 2 and 3 of the DG1::DL-VP16 transgenic plants (Figure 6d,e).

Cross sections of the leaves suggested that midrib formation was reinforced in the DG1::DL-VP16 lines. Wild-type leaves lacked a midrib in the distal region (about 20% of the way from the tip) of the leaf blade (Figure 6f). The DG1::DL-VP16 lines generated a clear midrib structure consisting of two locules and the septum in the distal region (Figure 6i). In the proximal region, the midrib seemed to be slightly larger in the transgenic lines than in the wild type (Figure 6g,j). In contrast to the leaf, no differences were observed in the carpel between the DG1::DL-VP16 and the wild type (Figure 6h,k). This result was consistent with the finding that DG1 does not contain a regulatory region to induce DL gene expression in carpel development.

We compared the phenotypes of the leaf primordia among DG1::DL-VP16, a weak allele of dl (dl-1) and the wild type. The central region of the leaf primordia (P3) was thinner in dl-1 than in the wild type (Figure 7a,b). This difference resulted from the number of cells in this region (Figure 7a,b,d,e). These observations confirmed our previous hypothesis that DL promotes cell proliferation in the central region (Ohmori et al., 2008). In the DG1::VP16 plant, by contrast, the central region was slightly thicker than that of the wild type (Figure 7c). The number of cells in the central region was much higher in the DG1::VP16 plant than in the wild type (Figure 7c–e). The difference was more evident in the P3 stage than in the P2 stage. These results suggest that an activated version of DL protein promotes strong cell proliferation, as compared with the wild-type protein. This enhanced cell proliferation is likely to be associated with the larger size of the midrib and the leaf erectness observed in DG1::VP16 plants.

Figure 7.

 Effect of DG1::DL-VP16 on cell proliferation in the central region of the leaf primordia.
(a–c) Central region (presumptive midrib) of the P3 stage leaf primordia: wild type (a), dl-1 (b), DG1::DL-VP16 (c). Cells in a column of the central region are falsely colored.
(d, e) Number of cells in the central region of the leaf primordia.
The cell numbers were counted in two independent plants. Scale bars: 100 μm.

Discussion

In this study, we examined the mechanisms that temporally and spatially regulate the expression of DL in rice development. The results indicate that a regulatory region responsible for specific expression in the presumptive region of the leaf primordia is located in intron 2, and that a DG1 sequence including a 7.4-kb promoter region, and introns 1 and 2, is sufficient to induce a GUS reporter gene in this leaf region, and to suppress it in other parts of the leaf primordia, such as the vascular bundles. In addition, we found that leaf erectness can be manipulated by controlling midrib size using the DG1 regulatory sequence and an activated version of DL protein.

Regulatory regions required for temporal and spatial regulation of DL expression

DL is expressed in the presumptive midrib in leaf development, and in the presumptive carpel primordia in flower development. It is of great interest to know how DL expression is regulated in these distinct organs.

DG1, which incorporates a long promoter region in addition to introns 1 and 2, seems not to be associated with carpel-specific expression in the flower, because no GUS expression was detected either in the presumptive region of carpel initiation in the flower meristem or in the carpel primordia itself. The region required for carpel-specific expression may be located far upstream of DL or in its downstream region. The dl-1 allele, which shows a weak drooping leaf phenotype, has no mutation in the coding region, but has a large deletion in the far upstream region of DL (Figure S2). The flower of dl-1, however, shows no homeotic changes in the carpel (Figure S2; Nagasawa et al., 2003; Ohmori et al., 2008). These observations suggest the possibility that a putative regulatory region responsible for carpel-specific expression is located downstream of the DL coding region. Thus, specific expression of DL in the carpel or in the presumptive midrib is regulated by distinct and separate regions.

Regulatory sequences in introns

DG1 contains several regions required for the precise expression of DL in the leaf (Figure 8). Intron 2 seems to be essential for expression in the presumptive midrib, whereas intron 1 seems to be related to quantitative gene regulation. The upstream promoter region in DG1 may function to repress ectopic expression in the vascular bundle and abaxial sclerenchyma.

Figure 8.

 Summary of the putative regulatory regions in DL.

Molecular dissection revealed that E2 and E3 within CNS-E in intron 2 are likely to be indispensable for temporal and spatial expression of DL in the leaf primordia. Removal of E3 resulted in the loss of GUS expression in P1, whereas removal of E2 resulted in deregulation of the specific expression of GUS in the central region of the leaf primordia: namely, the expansion of GUS expression in P1 and loss of that in P2. Therefore, E3 may be responsible for the initial activation of DL in the leaf primordia P1, and E2 may be required for restricting DL expression to the central region, i.e. the presumptive midrib. The upstream region seems to be dispensable for specific DL expression in the presumptive midrib because sequential deletion of this segment from DG1 had no obvious effect on the expression patterns of GUS in the leaf primordia.

Intron 1 seems to be associated with quantitative regulation. Removal of intron 1 (DG5 and DG8) resulted in a severe reduction of GUS staining; however, the spatial expression patterns in DG5 and DG8 plants were unchanged from those in plants carrying constructs that were the same except for the presence of intron 1. Therefore, intron 1 may contain an enhancer-like element. Our previous study showed that insertion of the retrotransposon TOS17 into intron 1 of DL causes severe defects in the leaf and flower (Yamaguchi et al., 2004). Taken together, these findings indicate that both introns 1 and 2 are likely to play an important role in the regulation of DL expression.

Proper combination of the promoter region and introns seems to be required for the exact control of DL expression

In Arabidopsis, the expression of AG is regulated by cis-regulatory elements located in intron 2 (Sieburth and Meyerowitz, 1997; Busch et al., 1999; Lohmann et al., 2001). Intron 2 of AG placed upstream of a minimal promoter of the cauliflower mosaic virus (CaMV) 35S promoter can confer spatial expression patterns in flower development similar to those of the AG transcript observed by in situ hybridization. Thus, intron 2 is sufficient to express AG in the flower. By contrast, intron 2 of DL did not induce GUS expression when it was placed upstream of the 35S minimal promoter (Figure S3a,c). In addition, even when intron 2 of DL was placed at a region upstream of the putative DL promoter (from −506 to +169 bp), no midrib-specific expression was observed (Figure S3b,d). These results suggest that both the combination and relative position of the DL promoter and intron 2 are important for inducing DL in a midrib-specific manner.

The CRC gene is an ortholog of DL in Arabidopsis, and is responsible for partial carpel identity and nectary development (Alvarez and Smyth, 1999; Bowman and Smyth, 1999). In contrast to DL, CRC has no CNSs in either intron 1 or intron 2 when compared with CRC orthologs from species in Brassicaceae (Lee et al., 2005). Several CNSs are found in the 5′ upstream region, and some of these sequences are important for CRC expression in carpels and nectaries. Thus, the distribution of regulatory sequences seems to differ between the two orthologous genes DL in rice and CRC in Arabidopsis. In addition, no CNSs were detected between whole regions of DL and CRC. Therefore, it is likely that the arrangement of cis-regulatory regions and their nucleotide sequences have highly diverged in the two genes during the evolution of rice and Arabidopsis.

Manipulation of leaf erectness

Erect leaves are an important agronomical trait. First, if leaves in the upper part of a plant are erect, then leaves in the lower part can perceive sunlight efficiently, and photosynthesis is promoted even in the lower leaves. Second, the number of plants that can be cultivated within an area should be increased, because rice plants with erect leaves require a smaller growing area than those with normal leaves. These two characteristics are associated with a higher yield of rice grains.

An attempt to make a rice plant with erect leaves was previously made using osdwarf4, a mutant defective in brassinosteroid biosynthesis (Sakamoto et al., 2006). This mutant shows more erect leaves because of a partial reduction in the level of brassinosteroid, and Sakamoto et al. succeeded in increasing biomass production and grain yield by the dense planting of osdwarf4.

Here, we tried to produce rice plants with erect leaves by generating a large midrib in leaves. To achieve this, it was necessary to express an activated version of DL protein specifically in the presumptive midrib, because its expression in other regions of the leaf primordia or in the carpel primordia can cause inhibition of normal vegetative growth or abnormal fruit development, respectively (Yamaguchi et al., 2004). Fortunately, DG1 appeared to satisfy these requirements. As expected, the DG1::DL-VP16 construct promoted midrib formation without any other abnormality. DG1::DL-VP16 lines developed morphologically normal carpels and produced many seeds, suggesting that normal fertile ovules were formed within these carpels.

The viral transcription factor VP16 is often used to enhance the function of transcription factors, because it has strong transcriptional activity. For example, the LFY-VP16 fusion protein strongly induces floral homeotic genes, such as AP1 and AG (Parcy et al., 1998). Because DL encodes a putative transcriptional factor belonging to the YABBY protein family, we tried to enhance DL activity using the VP16 activation domain. As a result, we succeeded in producing rice plants with more erect leaves by expressing a DL-VP16 fusion protein using the DG1 sequence. A midrib was formed in the distal region of the leaf blade, where no midrib is usually formed in the wild type; in addition, there was a slight increase in midrib size in the middle and basal regions of the leaf blade. The function of DL is likely to be conserved among grasses (Ishikawa et al., 2009). Accordingly, our result suggests that the genetic manipulation of DL-related genes may contribute to the modification of plant architecture in other important crops in the grass family.

Experimental procedures

Materials

The rice strain used in this study was O. sativa japonica Taichung 65. dl-1 has been described previously (Nagasawa et al., 2003; Ohmori et al., 2008). The CaMV 35S minimal promoter (Hobo et al., 1999) was kindly provided by Dr Hattori (Nagoya University).

Comparative sequence analysis

The genomic sequences of rice, maize and sorghum orthologs encompassing DL were identified by standard BLAST searches of The Rice Annotation Project Database (RAP-DB; http://rapdb.dna.affrc.go.jp/tools/blast) for rice, the Maize assembled Genomic Island (MAGI; http://magi.plantgenomics.iastate.edu/blast.do?blastGroup=BAC) for maize, and the Sorghum bicolor Plant Genome Database (SbDGB; http://www.plantgdb.org/SbGDB/cgi-bin/blastGDB.pl) for sorghum, using rice DL as a query. The wheat ortholog genomic sequences were determined from the wheat genomic clone WGF1 (Ishikawa et al., 2009; accession number AB570372).

The search for CNSs was carried out by pair-wise analysis of rice with maize, sorghum and wheat, using mVISTA (Mayor et al., 2000; http://genome.lbl.gov/vista/index.shtml). For comparison, the following regions of the genomic sequences were used (the positions are numbered according to the translation initiation site of each DL ortholog): rice (from −6646 to +11 516 bp), maize (from −7881 to +9021 bp), sorghum (from −8016 to +10 979) and wheat (from −4017 to +5231). For mVISTA, the parameters were set as follows: AVID alignment modality with a ‘calculated window’ of 50 bp, a ‘minimum consensus width’ of 30 bp and a ‘consensus identity’ of 75%.

Construction of plasmids

Outlines of the processes of plasmid construction are schematically represented in Figures S4–S6. The primers used for construction are shown in Table S1. In some cases, recognition sequences for restriction enzymes were attached to the sequences for primer annealing sites. The amplified fragments were first introduced into T vectors, and the sequences of fragments were verified by DNA sequencing.

To make the DG3::GUSDG7::GUS constructs, each DNA fragment (A, B, C, D, G and H) was amplified from genomic DNA and introduced into a T vector such as pGEM T easy (Promega, http://www.promega.com) or pTA2 (Toyobo, http://www.toyobo.co.jp) to give pGEM-A, pGEM-B, pGEM-C, pTA-D, pTA-G and pTA-H (Figure S4). Next, each fragment (fA, fB, fC, fD, fG and fH in Figure S4) was isolated after digestion of the corresponding plasmid with restriction enzymes and ligated into the binary vector pSMAHdNH627-M2GUS (H. Nakamura et al., unpublished data), which has a promoter-less GUS gene or its derivative (DGfA::GUS or DG3::GUS). For construction of DG1::GUS and DG2::GUS, the fE and fF fragments were isolated from a cloned DL EcoRI fragment (Yamaguchi et al., 2004) and ligated into DG3::GUS in place of the original region (Figure S4).

To make DG7ΔFS::GUS, DG7ΔCNS-E::GUS, DG7ΔΕ1::GUS, DG7ΔΕ2::GUS and DG7ΔΕ3::GUS, the fJ fragment was amplified by PCR and subcloned into pT7Blue (Novagen, now part of Merck, http://www.merck-chemicals.com) (pT7-J in Figure S5). The regions K, L, M, O, P or Q were removed from pT7-J by an inverse PCR technique, and named pT7JΔK, pT7JΔL-DM, pT7JΔO, pT7JΔP or pT7JΔQ, respectively (Figure S5). After isolation of the fragment from the corresponding plasmid, each fragment (fJΔK, fJΔL-ΔM, fJΔN, fJΔO, fJΔP and fJΔQ) was cloned into DG7::GUS in place of the original region (Figure S5).

To make the DG1::DL-VP16 construct, DL cDNA was amplified and inserted into a DL DraII fragment, which was isolated from a cloned DL EcoRI fragment (Yamaguchi et al., 2004; Figure S6). The modified cDNA was then ligated into pBI121, which contained a VP16-GR cassette made by using the XhoI site to fuse the VP16 activation domain (pBI121-DL/XhoI-VP16-GR). The region i1-i2-VP16 was obtained by PCR amplification from pBI121-DL/XhoI-VP16-GR and subcloned into pT7Blue (Novagen) (pT7-DL-i1-i2-VP16 in Figure S6). After isolation of an ApaI-PmeI fragment from pT7-DL-i1-i2-VP16, this fragment was cloned into DG1::GUS in place of the original region [DG1::DL-VP16 (-nosT) in Figure S6]. Lastly, DG1::DL-VP16 was generated by inserting the nopaline synthase terminator (nosT) from pSMAHdN627-M2GUS into the PmeI site.

All constructs were individually transformed into the Agrobacterium tumefaciens strain EHA101, and then transformed into rice as described previously (Hiei et al., 1994).

GUS staining

Shoot tips obtained about 60 days after regeneration (T0 plants), or from sowing (T1 plants) or inflorescences (carpel developing stage (T0 plants), were fixed in 90% ice-cold acetone for >16 h, and rinsed twice with a GUS working solution [50 mm phosphate buffer, pH 7.0, 5 mm K3Fe(CN)6, 5 mm K4Fe(CN)6, 1 mm EDTA] for 20 min each time. After rinsing, 5-bromo-4-chloro-3-indoyl β-d-glucuronide cyclohexylammonium salt (Wako, http://www.wako-chem.co.jp) was added to the GUS working solution to a final concentration of 0.5 mg ml−1, and placed under a vacuum (about 0.08 MPa) for 15 min, twice. After the vacuum, the tissue was incubated at 37°C overnight.

Histological observations

For histological observations, samples were fixed in a solution of 4% paraformaldehyde and 0.25% glutaraldehyde (w/v) in 50 mm phosphate buffer (pH 7.2) for at least 24 h at 4°C, and then dehydrated in a graded ethanol series. After substitution with xylene, the samples were embedded in Paraplast Plus (McCormick Scientific, http://www.mccormickscientific.com) and sectioned at a thickness of 20 μm using a rotary microtome. Sections were observed with a light microscope (BX50; Olympus, http://www.olympus-global.com).

In situ hybridization

To detect DL transcripts by in situ hybridization, a DL probe was prepared as described previously (Yamaguchi et al., 2004). Microtome sections (10 μm) were mounted on glass slides coated with silane (Muto Pure Chemicals, http://www.mutokagaku.com). RNA probes were labeled with digoxigenin using a DIG RNA Labeling Mix (Roche, http://www.roche.com) according to the manufacturer’s instructions. In situ hybridization and immunological detection of the signals were performed as described previously (Suzaki et al., 2004).

Acknowledgements

We thank T. Hattori for providing a vector containing the CaMV 35S minimal promoter, and M. Ishikawa and K. Ohsawa for technical assistance. This research was supported in part by Grants-in-Aid for Scientific Research from MEXT (20380005) (to H.-Y.H.) by a grant from MAFF (Genomics for Agricultural Innovation, GPN-0002) (to H.-Y.H.). Y.O. was supported by the Global COE Program (Integrative Life Science Based on the Study of Biosignaling Mechanisms) from MEXT.

Ancillary