Compensation and interaction between RISBZ1 and RPBF during grain filling in rice

Authors

  • Taiji Kawakatsu,

    1. Transgenic Crop Research and Development Centre, National Institute of Agrobiological Sciences, Tsukuba, Ibaraki 305-8602, Japan
    Search for more papers by this author
    • These authors contributed equally to the paper.

  • Masayuki P. Yamamoto,

    1. Transgenic Crop Research and Development Centre, National Institute of Agrobiological Sciences, Tsukuba, Ibaraki 305-8602, Japan
    Search for more papers by this author
    • These authors contributed equally to the paper.

    • Present address: Graduate School of Science and Engineering for Research, University of Toyama, Toyama 930-8555, Japan.

  • Satoru M. Touno,

    1. Transgenic Crop Research and Development Centre, National Institute of Agrobiological Sciences, Tsukuba, Ibaraki 305-8602, Japan
    2. Graduate School of Science and Technology, Chiba University, Matsudo, Chiba 271-8510, Japan
    Search for more papers by this author
  • Hiroshi Yasuda,

    1. Transgenic Crop Research and Development Centre, National Institute of Agrobiological Sciences, Tsukuba, Ibaraki 305-8602, Japan
    Search for more papers by this author
    • §

      Present address: Research Team for Crop Cold Tolerance, National Agricultural Research Centre for Hokkaido Region, Sapporo, Hokkaido 062-8555, Japan.

  • Fumio Takaiwa

    Corresponding author
    1. Transgenic Crop Research and Development Centre, National Institute of Agrobiological Sciences, Tsukuba, Ibaraki 305-8602, Japan
    Search for more papers by this author

*(fax +81 29 838 8397; e-mail takaiwa@nias.affrc.go.jp).

Summary

The rice (Oryza sativa L.) basic leucine Zipper factor RISBZ1 and rice prolamin box binding factor (RPBF) are transcriptional activators of rice seed storage protein (SSP) genes in vivo. To ascertain the functions of these trans-activators in seed development, knock-down (KD) transgenic rice plants were generated in which the accumulation of RISBZ1 and RPBF was reduced in an endosperm-specific manner by co-suppression (KD-RISBZ1 and KD-RPBF). The accumulation of most SSPs changed little between individual KD mutants and wild-type plants, whereas a double KD mutant (KD-RISBZ1/KD-RPBF) resulted in a significant reduction of most SSP gene expression and accumulation. The reduction of both trans-activators also caused a greater reduction in seed starch accumulation than individual KD mutants. Storage lipids were accumulated at reduced levels in KD-RISBZ1 and KD-RISBZ1/KD-RPBF seeds. KD-RPBF and KD-RISBZ1/KD-RPBF seeds exhibited multi-layered aleurone cells. Gene expression of DEFECTIVE KERNEL1 (OsDEK1), CRINKLY4 (OsCR4) and SUPERNUMERARY ALEURONE LAYER 1 (OsSAL1) rice homologues was decreased in the KD mutants, suggesting that these genes are regulated by RISBZ1 and RPBF. These phenotypes suggest that combinatorial interactions between RISBZ1 and RPBF play an essential role during grain filling. The functional redundancy and compensation between RISBZ1 and RPBF possibly account for weak effects on the SSP levels in single KD mutants, and help maintain various processes during seed development in rice. Physical interaction between RISBZ1 and RPBF may ensure that these processes are carried out properly.

Introduction

Plant seed storage proteins (SSPs) are specifically synthesized during seed maturation, then stored in the seed as a nitrogen source for the germinating seedling, and thus serve as nutrients for humans and livestock. Aside from their physiological and agricultural importance, SSP genes provide an important model system for investigating the temporal and spatial patterns of gene regulation in plants (Goldberg et al., 1994; Kroj et al., 2003). A hallmark of SSP expression is that it is restricted to maturing seeds (i.e. cotyledons and embryos in dicots, or the endosperm in cereals). Additionally, as SSP genes are expressed at extremely high levels during specific stages in seed development, much of what is known about seed-specific expression can be applied to the production of high-value recombinant proteins in the seeds of transgenic crops (Stoger et al., 2005; Takaiwa et al., 2007).

Most cereals, such as maize (Zea mays L.), barley (Hordeum vulgare L.) and wheat (Triticum aestivum L.), produce prolamins as their major SSP (Shewry and Tatham, 1999), but rice (Oryza sativa L.) accumulates glutelins, which are homologous with the 11S globulins of legumes (Takaiwa et al., 1999), as the major component. Irrespective of the evolutionary divergence between the prolamin and glutelin genes, cereal SSP genes are expressed in the endosperm. Several motifs, including GCN4, a prolamin box (P box), AACA and ACGT, have been identified as cis-regulatory elements involved in the endosperm-specific expression of cereal SSP genes (Zheng et al., 1993; Takaiwa et al., 1996; Washida et al., 1999; Wu et al., 2000). Among these elements, the GCN4 motif is recognized by Opaque2-like basic leucine zipper (bZIP) activators, including maize Opaque2 (O2), barley BLZ2, wheat SPA1 and rice RISBZ1 (Albani et al., 1997; Vicente-Carbajosa et al., 1998; Wu et al., 1998; Conlan et al., 1999; Onate et al., 1999; Onodera et al., 2001), and the P box is bound by DNA binding with one finger (DOF) zinc-finger transcription factors such as the prolamin box binding factor (PBF) subfamily (maize PBF, barley BPBF, wheat WPBF and rice RPBF; Vicente-Carbajosa et al., 1997; Mena et al., 1998; YamamoTo et al., 2006) and the barley SAD protein (Diaz et al., 2005). Additionally, direct interactions between maize O2 and PBF have been reported (Vicente-Carbajosa et al., 1997).

The RISBZ1 (rice seed b-Zipper 1) and RPBF regulatory factors are coordinately expressed with rice SSP genes, and activate transcription from SSP gene promoters in transient expression assays (Onodera et al., 2001; Yamamoto et al., 2006). In addition to the major storage protein glutelins, which account for 60–70% of the total seed protein, rice seeds accumulate other SSPs such as prolamins, albumins (14–16-kDa rice seed allergenic proteins; RAs) and α-globulin as minor components (Takaiwa, 1999). Most of the rice SSP genes that have been examined are trans-activated by RISBZ1 and RPBF, except for the RA gene, which is under the control of RPBF alone (Yamamoto et al., 2006). Furthermore, we have demonstrated in a transient expression assay that combinatorial interactions between RISBZ1 and RPBF in the transcriptional apparatus confer a synergistic effect on many rice SSP genes, encoding glutelins, prolamins and α-globulin (Yamamoto et al., 2006), an observation that points to the potential of rice seed for extremely high SSP gene expression.

The characterization of these activators, however, has been mainly carried out by in vitro assays such as electrophoretic mobility shift assays (EMSAs), or by in vivo transient expression assays. Inactivation in planta has, to our knowledge, not yet been reported, except for maize o2 mutants, which have reduced levels of the maize storage prolamin 22-kDa α-zein (Gibbon and Larkins, 2005). To better understand the role of these activators in the endosperm, we obtained knock-down (KD) transgenic rice plants in which the accumulation of RISBZ1 and RPBF was reduced, and their effects on seed development were assessed. Here, we show that the combination of RISBZ1 and RPBF play critical roles in global processes, including SSP gene expression, during seed development.

Results

Production of the RISBZ1 and RPBF KD plants

To examine the function of RISBZ1 and RPBF in the rice endosperm, we first searched for retrotransposon Tos17-induced insertion mutants and T-DNA insertion mutants that disrupt these transcription factors by screening the Rice Annotation Project Database (RAP-DB, http://rapdb.dna.affrc.go.jp), and by the PCR screening of a Tos17 knock-out rice population (Miyao et al., 2003). However, we could not identify any mutant lines that possessed Tos17 or T-DNA insertions within the exons of RISBZ1 and RPBF. Therefore, we analysed transgenic rice plants with a reduced accumulation of these transcription factors in the maturing endosperm. A ProRISBZ1:RISBZ1ΔAD plasmid (Figure 1a) was originally constructed for the expression of a truncated form of the RISBZ1 protein (RISBZ1ΔAD) lacking the N-terminal activation domain (Onodera et al., 2001), under the control of the RISBZ1 promoter. We generated transgenic lines harboring the ProRISBZ1:RISBZ1ΔAD construct, which segregated into two groups according to RISBZ1 accumulation. One group accumulated exogenous RISBZ1ΔAD (data not shown), and the other (KD of RISBZ1, KD-RISBZ1) reduced endogenous RISBZ1 (Figure 1c), probably as a result of an RNA-mediated gene silencing mechanism co-suppressing both the endogenous gene and the transgene. It should be noted that no transformants were recovered when the CaMV 35S promoter (Pro35S) was used in place of the RISBZ1 promoter (data not shown). A KD-RPBF mutant was obtained from transgenic plants transformed with ProNRP33:RPBF (Figure 1b). ProNRP33:RPBF was originally constructed to overexpress RPBF cDNA under the control of NRP33, a rice 13-kDa prolamin promoter (Qu and Takaiwa, 2004). Unexpectedly, RPBF was greatly reduced in every transgenic plant (Figure 1c and data not shown).

Figure 1.

 Generation of RISBZ1 and RPBF knock-down (KD) lines.
(a) and (b) Schematic representation of the constructs for KD-RISBZ1 (a) and KD-RPBF (b).
(a) ProRISBZ1:RISBZ1ΔAD comprised a RISBZ1 cDNA lacking an N-terminal activation domain (AD; amino acid positions 1–40) under the control of the RISBZ1 promoter.
(b) ProNRP33:RPBF comprised a full-length RPBF cDNA under the control of the 13-kDa prolamin (NRP33) promoter.
(c) Accumulation levels of RISBZ1, RPBF (upper) and seed storage proteins (lower) during grain filling in wild type, KD-RISBZ1, KD-RPBF and KD-RISBZ1/RPBF seeds. Protein samples were extracted from seeds at 7, 14 and 21 days after flowering (DAF), and from mature seed. Glutelins (precursor, acidic and basic subunits), α-globulin and 13-kDa prolamins are indicated.
(d) Expression analysis of RISBZ1 and RPBF in wild type, KD-RISBZ1, KD-RPBF and KD-RISBZ1/RPBF seeds by quantitative real-time PCR (qRT-PCR). Total RNA was extracted from developing seeds 7, 14, 21 and 28 DAF. The average values of three measurements were normalized to control RNA (17S rRNA). The expression level of each gene at 7 DAF of the wild type is set to one. The error bars represent the range of expression levels calculated from the standard deviations of the respective genes.

Next, we crossed KD-RISBZ1 and KD-RPBF, and obtained a RISBZ1 and RPBF double-KD line (KD-RISBZ1/KD-RPBF). We isolated KD-RISBZ1, KD-RPBF and KD-RISBZ1/KD-RPBF homozygous lines by repeated self-pollination for use in this study.

In wild-type seeds, the accumulation of RISBZ1 was very high until 14 days after flowering (DAF), then radically declined to low or undetectable levels (Figure 1c). RPBF accumulation was detectable throughout the endosperm development, and reached a maximum at 14 DAF (Figure 1c). RISBZ1 in KD-RISBZ1 and RPBF in KD-RPBF were very weak, but detectable, at 7 DAF, and became undetectable afterwards (Figure 1c). On the other hand, RPBF was steadily expressed throughout the development of KD-RISBZ1 seed, and the expression levels were greater than in wild-type seed at 7, 21 and 28 DAF (Figure 1c). In KD-RPBF seed, RISBZ1 was expressed steadily throughout development, although expression levels were lower than wild-type levels at 7 and 14 DAF (Figure 1c). In KD-RISBZ1/KD-RPBF seed, both RISBZ1 and RPBF were detectable only at 7 DAF (Figure 1c).

RISBZ1 and RPBF expression levels were also determined by quantitative RT-PCR (qRT-PCR). In wild-type seed, RISBZ1 was highly expressed until 7 and 14 DAF (Figure 1d). RPBF expression reached a maximum at 14 DAF, declining afterwards to approximately the same level as found at 7 DAF (Figure 1d). These expression patterns were consistent with our previous northern analysis (Yamamoto et al., 2006), and with the western analysis in this study (Figure 1c). In KD-RISBZ1 seed, RISBZ1 expression was much greater at 7 DAF, but was subsequently lower than that of wild-type seed (Figure 1d). RPBF expression was slightly higher at 7 DAF, but was lower at 14 and 21 DAF than in wild-type seed (Figure 1d). RISBZ1 expression was similar in KD-RPBF and KD-RISBZ1 seeds (Figure 1d). RPBF expression was much greater at 7 and 14 DAF, but rapidly decreased to lower levels than that of the wild type (Figure 1d). It is notable that RISBZ1 and RPBF were expressed at significantly higher levels at 7 DAF in KD-RISBZ1 and KD-RPBF seeds. This artificially high expression by the RISBZ1 and NRP33 promoters may have caused the co-suppression of RISBZ1 and RPBF. In KD-RISBZ1/KD-RPBF seed, the expression of both RISBZ1 and RPBF was lower than in wild-type seed throughout development (Figure 1d).

RISBZ1 and RPBF regulate SSP gene expression

KD-RISBZ1 and KD-RPBF seeds accumulated slightly less total SSP than the wild-type seed, but the SDS-PAGE patterns of most of the SSPs, such as the glutelins, prolamins and α-globulin, looked similar in the three lines (Figures 1c and 2a). In contrast to single-KD seeds, there was a significant reduction of SSP in the double-KD seeds (Figures 1c and 2a). Thus, for clarity, we investigated the SSP levels in KD lines by immunoblot (Figure 2). In both KD-RISBZ1 and KD-RPBF seeds, glutelin GluA was slightly decreased, and the level of GluB was comparable with that of the wild type. The GluD level was enhanced in KD-RISBZ1 and completely lost in KD-RPBF. α-Globulin was relatively decreased in KD-RISBZ1, and was further decreased in KD-RPBF. The Cys-rich 13-kDa prolamin RM1 was slightly decreased in KD-RISBZ1, and was relatively decreased in KD-RPBF. The Cys-poor 13-kDa prolamin Prol14 was comparable in KD-RISBZ1 and KD-RPBF. The 10-kDa prolamin RP10 was slightly increased in both KD-RISBZ1 and KD-RPBF. The 14–16-kDa RA proteins were greatly increased in KD-RISBZ1, and were markedly diminished in KD-RPBF. In KD-RISBZ1/KD-RPBF seeds, almost all of the SSPs were significantly decreased, except for RP10, which was slightly increased. Quantitative RT-PCR using RNA extracted from developing seeds at 14 DAF revealed that the SSP gene expression levels were closely correlated with their protein level in each of the KD lines (Figures 2b and 3). The expression of all of the SSP genes examined was significantly suppressed in KD-RISBZ1/KD-RPBF seeds (Figures 2b and 3), suggesting that RISBZ1 and RPBF act redundantly, and play a crucial role in the expression of rice SSP genes in the endosperm. Notably, the RP10 expression level and RP10 accumulation level were not correlated in these KD lines (Figures 2b and 3).

Figure 2.

 Seed storage proteins and OsPPDKB level in wild-type, KD-RISBZ1, KD-RPBF and KD-RISBZ1/RPBF seeds.
(a) SDS-PAGE of seed proteins extracted from mature seeds. Glutelins (precursor, acidic and basic subunits), α-globulin, 14–16-kDa rice allergen (RA), and 13-kDa prolamins are indicated. GluD-1 is also indicated (asterisk).
(b) Western blot of seed storage proteins (SSPs). The same samples as in (a) were used. Glutelins (GluA, GluB and GluD), α-globulin (Glb), 13-kDa prolamins (RM1 and Prol14), 10-kDa prolamin (RP10) and 14–16-kDa rice allergen (RA) were analyzed.
(c) Western blot of OsPPDKB. Proteins were extracted from developing seeds at 14 days after flowering.

Figure 3.

 Expression analyses of SSP genes, starch and lipid-metabolizing enzyme genes, and aleurone layer number-related genes, by quantitative RT-PCR.
RNA was extracted from developing seeds at 14 days after flowering (DAF) for SSP genes (GluA, GluB, GluD, Glb-1, RM1, Prol14, RP10 and RAG-1), starch (GBSS I, SS I, SS IIa, SS IIIa, OsAGPL2, OsAGPS2b and OsPPDKB) and lipid (Os08g0510400, Os09g0505300 and OsPPDKB) metabolizing enzyme genes, and at 7 DAF for aleurone layer number-related genes (OsCR4, OsDEK1 and OsSAL1). Light-grey, black, white and dark-grey bars represent the expression levels in wild-type, KD-RISBZ1, KD-RPBF and KD-RISBZ1/KD-RPBF seeds, respectively. The relative expression levels were normalized to that of 17S RNA.

Storage compounds

To examine the effect of RISBZ1 and RPBF on seed quality, we measured the total protein, starch and total lipid contents. As observed on SDS-PAGE analysis (Figure 2a), the protein concentrations were lower in KD lines (Table 1). Notably, KD-RISBZ1/KD-RPBF seed contained significantly lower protein levels, indicating a synergistic effect between RISBZ1 and RPBF on SSP accumulation (Table 1). The starch content per grain was moderately decreased in KD-RISBZ1 seeds compared with the wild type (Table 1). In contrast, there was much less starch in KD-RPBF seeds (Table 1). KD-RISBZ1/KD-RPBF seeds contained only half the starch of wild-type seeds, suggesting that RISBZ1 and RPBF regulate starch biosynthesis synergistically (Table 1). Although the expression level of granule-bound starch synthase gene GBSS I was decreased in KD-RISBZ1 and KD-RPBF seeds, and severely decreased in KD-RISBZ1/KD-RPBF, the expression of other starch metabolizing enzyme genes was increased in KD-RISBZ1, and was approximately the same as that of the wild type (except for the starch synthase gene SS IIa and pyruvate orthophosphate dikinase gene OsPPDKB) in KD-RPBF seed (Figure 3). In KD-RISBZ1/KD-RPBF seeds, the expression of these genes (SS I, SS IIa and SSIIa) tended to be similar to that in KD-RISBZ1, whereas the expression of the ADP glucose pyrophosphorylase gene OsAGPS2b was decreased (Figure 3).

Table 1.   Storage components of wild-type and kncok-down line seeds
LinesProtein (mg/grain)Starch (mg/grain)Lipid (mg/grain)
  1. Values are means ± SDs. All measurements were performed in triplicate.

Wild type2.60 ± 0.0213.4 ± 0.10.55 ± 0.01
KD-RISBZ12.50 ± 0.0012.8 ± 0.50.31 ± 0.09
KD-RPBF2.24 ± 0.049.9 ± 0.2 0.54 ± 0.04
KD-RISBZ1/KD-RPBF1.86 ± 0.006.7 ± 0.10.37 ± 0.00

The total lipid content per grain was also lower in KD-RISBZ1 and KD-RISBZ1/KD-RPBF seeds, but was similar to the wild type in KD-RPBF. Thus, only RISBZ1 was suggested to be involved in lipid biosynthesis or metabolism. As lipid metabolizing enzymes have not been well characterized in rice grain, we examined the expression levels of the rice ketoacyl-ACP reductase gene Os08g0510400 and the acyl-ACP thioesterase gene Os09g0505300, which are known to play a critical role in lipid metabolism in Arabidopsis, and are rice homologues of Arabidopsis At1g62610 and At3g25110, respectively (Mu et al., 2008). The expression of Os08g0510400 was slightly increased in KD-RISBZ1 and KD-RPBF seeds, and was markedly increased in KD-RISBZ1/KD-RPBF seed (Figure 3). The expression level of Os09g0505300 was similar in the wild type and KD lines (Figure 3). The expression of OsPPDKB was relatively increased in KD-RISBZ1 and KD-RISBZ1/KD-RPBF, and was relatively decreased in KD-RPBF seeds (Figure 3), whereas that of OsPPDKB was increased in the KD lines (Figure 2c).

KD seed phenotypes

The seed phenotypes of KD-RISBZ1, KD-RPBF and KD-RISBZ1/KD-RPBF were opaque and wrinkled (Figure 4a–h), with KD-RISBZ1 being the least affected, and KD-RISBZ1/KD-RPBF being severely affected (Figure 4a–h). The Seed dimensions were largely unchanged, except that KD-RISBZ1/KD-RPBF was somewhat narrower than the others, and all three of the KD lines were thinner and lighter than the wild type (Table 2).

Figure 4.

 Phenotypes of wild-type, KD-RISBZ1, KD-RPBF and KD-RISBZ1/RPBF seeds.
(a), (e), (i), (m), (q) and (u) Wild type.
(b), (f), (j), (n), (r) and (v) KD-RISBZ1.
(c), (g), (k), (o), (s) and (w) KD-RPBF.
(d), (h), (l), (p), (t) and (x) KD-RISBZ1/RPBF.
(a–h) Mature seeds.
(i–l) Cross sections of mature endosperm.
(m–p) SEM of the peripheral area of mature endosperm, with the cross sections indicated by a blue square in (i–l).
(q–t) SEM of the central area of mature endosperm cross sections, indicated by a red square in (i–l), respectively. Arrows and arrowheads in (m–t) indicate polygonal and round starch granules, respectively. Areas in which starch granules are tightly packed are indicated by asterisks.
(u–x) Aleurone cells in mature endosperm are indicated by a green square in (i–l). Yellow lines indicate the boundary between aleurone cells (upper) and the inner region (lower). The arrow in (w) indicates multi-layered aleurone cells. Scale bars: 5 mm in (a–d), 2 mm in (e–h), 1 mm in (i–l), 10 μm in (m–t) and 100 μm in (u–x).

Table 2.   Grain size and weight of wild-type and knock-down line seeds
 Seed sizea 
LinesHeight (mm)Width (mm)Thickness (mm)Grain weightb (mg)
  1. aValues are means ± SDs.

  2. bGrain weights were claculated by total seed weight/number of seeds.

Wild type (n = 36)4.6 ± 0.22.8 ± 0.22.0 ± 0.119.2
KD-RISBZ1 (n = 34)4.7 ± 0.22.7 ± 0.21.8 ± 0.216.8
KD-RPBF (n = 37)4.8 ± 0.22.7 ± 0.21.7 ± 0.115.4
KD-RISBZ1/KD-RPBF (n = 53)4.7 ± 0.32.2 ± 0.21.6 ± 0.210.6

In wild-type seeds, the surfaces of the endosperm cross sections were crystallized, with a glossy sheen (Figure 4i). KD-RISBZ1 seed had a floury-white core from the centre to the ventral region, with the remainder being like the wild-type (Figure 4j). In KD-RPBF seed, almost the entire cross-sectional area was floury, but with some glossy sheen at the centre (Figure 4k). KD-RISBZ1/KD-RPBF seed had a completely floury cross-sectional phenotype (Figure 4l).

To further investigate the nature of the opaque phenotype, we analysed starch morphology using scanning electron microscopy (SEM). In the wild-type endosperm, starch granules were similar in size and were polygonal (Figure 4m,q), and tightly packed starch granules were abundant in both the peripheral and central areas of the endosperm cells (Figure 4m,q). In the KD-RISBZ1 endosperm, although round-shaped granules were sometimes observed in the central area, similarly sized and polygonal starch granules were mainly distributed across the whole area, but they were relatively loosely packed (Figure 4n,r). The KD-RPBF endosperm contained similarly sized, but loosely packed, polygonal starch granules across the whole area (Figure 4o,s). However, tightly packed starch granules could sometimes be observed near the central area (Figure 4s). In KD-RISBZ1/KD-RPBF endosperm, starch granules were round and smaller than in the wild type (Figure 4p,t). The packaging of starch granules in KD-RISBZ1/KD-RPBF endosperm was loose (Figure 4p,t).

The aleurone layer contains lipids and enzymes that are required for germination and seedling development. Wild-type endosperm had mostly single-layered aleurone cells (Figure 4u), except for the multilayers (two or three layers) in the dorsal region. Similarly, KD-RISBZ1 had orderly single-layered aleurone cells (Figure 4v). In the KD-RPBF endosperm, however, the arrangement of the aleurone cells was sometimes disordered and multi-layered (Figure 4w). The KD-RISBZ1/KD-RPBF aleurone layer was distinct from the others, being composed of multilayered, enlarged and disorderly aleurone cells (Figure 4x).

RISBZ1 and RPBF positively regulate the aleurone layer number-related genes

In KD-RPBF and KD-RISBZ1/KD-RPBF seeds, the number of aleurone layers was increased, as described above (Figure 4). Previous studies on maize and Arabidopsis, CRINKLY 4 (CR4), DEFECTIVE KERNEL 1 (DEK1) and SUPERNUMERARY ALEURONE LAYER 1 (SAL1) have shown them to be key players in determining the number of aleurone layers (Becraft et al., 1996; Lid et al., 2002; Shen et al., 2003). To examine whether the increase in the number of aleurone layers was caused by the altered expression of these genes, we performed qRT-PCR with RNA extracted from developing seeds at 7 DAF. The expression of both positive (OsCR4 and OsDEK1) and negative (OsSAL1) determining genes for the aleurone layer number was decreased in KD-RISBZ1 and KD-RPBF, and significantly decreased in KD-RISBZ1/KD-RPBF seeds, suggesting that both RISBZ1 and RPBF are positive regulators of these genes (Figure 3).

RISBZ1 and RPBF are localized in the nucleus, and interact in rice protoplasts

The synergism between RISBZ1 and RPBF is observed not only in vivo, but also in planta (Yamamoto et al., 2006; also in this study). Such synergism is presumably to the result of interaction between RISBZ1 and RPBF. To investigate whether they interact, we examined subcellular localization of RISBZ1 and RPBF in rice protoplasts. Full-length RISBZ1 and RPBF cDNA, with double HA tags and a 6xHis tag (2HA-6His) at their C termini, were under the control of the enhanced Cauliflower mosaic virus (CaMV) 35S promoter, carrying the first intron of Shurunken, with an internal deletion (ShΔ). These HA-tagged RISBZ1 and RPBF expression constructs were transfected into protoplasts derived from rice calli and transiently expressed. When immunofluorescently detected with anti-HA and 4′,6-diamidino-2-phenylindole (DAPI) staining, it was shown that both RISBZ1 and RPBF were localized in the nucleus (Figure 5a). Next, co-precipitation analysis was carried out to examine whether these transcription factors interacted with each other in the nucleus. When Pro35S:RISBZ1 and Pro35S-ShΔ:RPBF-2HA-6His were transfected independently, or co-transfected into rice protoplasts, RISBZ1 or RPBF was detected with a specific antibody in the input protein extracts containing these constructs. In contrast, when RPBF-2HA-6His was purified with a His-select Ni-NTA affinity column from these extracts, RISBZ1 was only detected in the extract containing both RISBZ1 and RPBF-2HA-6His (Figure 5b), demonstrating that the RISBZ1 and RPBF are co-localized in the nuclei, and interact with each other in rice protoplasts. Notably, the detected RISBZ1 signal was only modest, suggesting that this interaction may be transient and physically weak.

Figure 5.

 Interaction between RISBZ1 and RPBF in rice calli protoplast.
(a) Localization of RISBZ1 and RPBF in rice protoplasts. Pro35S-ShΔ:RISBZ1-2HA-6His or Pro35S-ShΔ:RPBF-2HA-6His was transfected into rice protoplasts by electroporation. The immunofluorescence for RISBZ1-2HA-6His (upper) or RPBF-2HA-6His (lower) with anti-HA, and DAPI staining were merged.
(b) Affinity co-purification of RISBZ1 with RPBF-2HA-6His from rice protoplasts. Pro35S:RISBZ1 and Pro35S-ShΔ:RPBF-2HA-6His were transfected individually or in combination into rice protoplasts by electroporation. Protein extracts (input) were applied to a Ni column. After washing, the elution (His-tag purified) was collected. RISBZ1 and RPBF-2HA-6His were detected by immunoblot using anti-RISBZ1 and anti-HA, respectively.

Discussion

RISBZ1 and RPBF regulate SSP gene expression in planta

We have reported that RISBZ1 and RPBF are involved in the regulation of rice SSP genes (Onodera et al., 2001; Yamamoto et al., 2006). This result is supported by evidence that they are coordinately expressed with SSP genes, and that they are able to bind to the corresponding cis-elements located in the SSP gene promoters (RISBZ1/GCN4 motif [TGA(G/C)TCA] and RPBF/P-box [TG(T/A/C)AAAG] or AAAG/CTTT sequence motifs (Onodera et al., 2001; Yamamoto et al., 2006). It has also been shown that RISBZ1 and RPBF activate the transcription of several different rice SSP gene promoters in transient assays in a synergistic manner (Yamamoto et al., 2006). Therefore, to ascertain whether these activators actually function in the endosperm of stable transgenic rice plants, we produced KD mutant plants in which the accumulation of these activators was suppressed. Surprisingly, neither KD-RISBZ1 nor KD-RPBF exhibited apparent SSP reduction, although o2, a maize orthologue of RISBZ1, exhibited significant reduction of 22-kDa α-zein (Gibbon and Larkins, 2005). In contrast, KD-RISBZ1/KD-RPBF had significantly reduced levels of SSPs (Figure 2; Table 1). Reduced SSP levels were observed at the mRNA level, indicating that RISBZ1 and RPBF play a critical and redundant role in the activation of SSP gene transcription. RP10 expression is induced very early, and peaks at approximately 8 DAF (Su et al., 2004), when RISBZ1 and RPBF are still detectable at a reduced level in KD-RISBZ1/KD-RPBF. Such a temporal expression pattern of RP10, and partial reduction of RISBZ1 and RPBF, at an early stage may allow normal RP10 accumulation in KD-RISBZ1/KD-RPBF seeds.

Most of the major SSP proteins were not greatly affected by the single KD mutations, but the 14–16-kDa RA genes and one kind of glutelin (GluD-1) were apparently dependent on the level of RPBF protein present (Figures 1c and 2). The RPBF-dependent expression patterns of RA and GluD-1 in planta are consistent with those in vivo. RPBF trans-activates the expression of RAG-1 (one of the RA genes) and GluD-1, whereas RISBZ1 does not trans-activate the expression of the RAG-1 promoter, and weakly trans-activates the GluD-1 promoter in rice protoplast. These may be attributed to the RAG-1 promoter having several P boxes, but no typical GCN4 motif (Adachi et al., 1993), whereas the GluD-1 promoter also has several P boxes and a GCN4-like motif that is weakly recognized by RISBZ1 (Kawakatsu et al., 2008). This is in contrast with most rice SSP genes, which have both cognate GCN4 motifs and P boxes in their promoter regions (Wu et al., 1998, 2000; Yamamoto et al., 2006).

RISBZ1 and RPBF regulate starch and lipid content indirectly

The suppression of RISBZ1 and RPBF caused opacity and wrinkling in seeds (Figure 4). The endosperm opacity arises from the floury texture caused by loosely packed starch granules, rather than from decreased starch biosynthesis (Kang et al., 2005; Fujita et al., 2007). The starch content in the endosperm depends on ADP-glucose pyrophosphorylase (Ohdan et al., 2005). The Expression of OsAGPS2b was decreased in KD-RISBZ1/KD-RPBF seed, which may be one of the direct causes of significantly reduced starch content. In contrast, the expression of both OsAGPL2 and OsAGPS2b was increased in KD-RISBZ1, and unchanged in KD-RPBF seed, and the decreased level of OsAGPS2b in KD-RISBZ1/KD-RPBF seed was milder than the SSP genes, suggesting that these genes are not primarily regulated by RISBZ1 and RPBF. The reduction of starch content in KD-RPBF cannot be explained by the expression of starch synthase genes alone.

The nutrient materials used for the production of storage compounds are transported from the dorsal vascular bundle into the endosperm via an enclosing pericarp during the early grain-filling stage, and immediately after by the differentiation of the aleurone layers (Hoshikawa, 1989). The distance of the floury regions of the KD-RISBZ1 seeds from the influx zone (Figure 4) may indicate that the transport of storage-component precursors has been disturbed. In contrast, starch biosynthesis starts in the centre of the endosperm, and then gradually spreads throughout the whole endosperm. The relatively normal central region of KD-RPBF seeds may indicate that starch biosynthesis was normal early in grain filling, but then gradually diminished.

Lipid levels were decreased in KD-RISBZ1 and KD-RISBZ1/KD-RPBF seeds, but not in KD-RPBF seeds (Table 1), suggesting that RISBZ1 alone is involved in the regulation of lipid accumulation-related genes (i.e. biosynthesis). Although lipids are one of the major storage compounds in the endosperm, there have been few studies on lipid metabolizing enzyme genes in rice. Thus, we examined the expression levels of several important enzyme genes involved in lipid metabolism in Arabidopsis that are known to be expressed in the rice developing endosperm using preliminary microarray analysis (Mu et al., 2008; data not shown). Ketoacyl-acyl carrier protein (ACP) reductase is responsible for the second step of the type-II fatty acid elongation system, catalyzing the NADPH-dependent reduction of β-ketoacyl-ACP to generate β-hydroxyacyl-ACP and NADP(+) (Mekhedov et al., 2000). Acyl-ACP thioesterases primarily hydrolyse oleoyl-ACP (Mekhedov et al., 2000). The expression of these enzyme genes was not decreased in KD lines, suggesting that neither RISBZ1 nor RPBF primarily regulates the expression of these genes.

The inconsistency between storage compound (starch and lipid) accumulation and related gene expression might be explained by the activity of cytosolic pyruvate orthophosphate dikinase (cyPPDK) and substrate distribution. This is because cyPPDK catalyzes a reversible conversion between pyruvate and phosphoenolpyruvate (PEP) during seed development, and cyPPDK is a key biochemical interface factor between C and N metabolism (Meyer et al., 1982; Aoyagi and Bassham, 1984). Furthermore, rice cyPPDK, FLO4/OsPPDKB, modulates carbon metabolism, including starch and fatty-acid biosyntheses (Kang et al., 2005). Thus, bidirectional modulation caused the alteration in starch and lipid accumulation in the KD lines.

Although we have shown that the promoter of OsPPDKB is activated by RISBZ1 and RPBF (Yamamoto et al., 2006), the expression of OsPPDKB was slightly increased in KD-RISBZ1 and KD-RISBZ1/KD-RPBF seeds, and was relatively decreased in KD-RPBF seeds. OsPPDKB was accumulated at increased levels in the KD lines. Thus, RISBZ1 and RPBF are one of the modulators of OsPPDKB, rather than primary regulators, as is the case with the maize O2 and PBF orthologues (Gallusci et al., 1996; Maddaloni et al., 1996; Yanagisawa, 2000).

RISBZ1 and RPBF regulate the number of aleurone layers

In rice, the endosperm is surrounded by mostly single-layered aleurone cells, except for the multilayered (two or three layers) dorsal region. Although the number of aleurone layers was unchanged in KD-RISBZ1 seed, patchy multilayered aleurone layers were observed in KD-RPBF seed. It is notable that the aleurone layers were composed of abnormal supernumerary layer cells in the KD-RISBZ1/KD-RPBF seed. The extensive increase in the aleurone layer number in KD-RISBZ1/KD-RPBF seed suggests that both RISBZ1 and RPBF are involved in this process, with the contribution by RPBF being markedly greater than that by RISBZ1. A similar multiple-layer phenotype has been observed in maize supernumerary aleurone layer 1 (salI) mutants (Shen et al., 2003). SAL1 is a homologue of CHMP1, encoding a class-E vacuolar sorting protein, which mediates the trafficking of cargo proteins from multivesicular endosomes to lysosomes, for degradation. In contrast, maize defective kernel 1 (dek1) and crinkly4 (cr4) displayed absent and defective aleurone layers, respectively. DEK1 and CR4 are a calpain-like proteinase and a receptor-like kinase, respectively (Becraft et al., 1996; Lid et al., 2002). Both DEK1 and CR4 bind to the plasma membrane, and are involved in the production of the aleurone layer specification signal. The multiple aleurone layer phenotype in sal1 is thought to be caused by an overabundance of aleurone membrane signaling proteins, such as DEK1 and CR4 (Tian et al., 2007). As expected, the expression of OsSAL1 was significantly decreased in KD-RISBZ1/KD-RPBF seed. Although the expression of OsSAL1 was also decreased in KD-RISBZ1 and KD-RPBF seeds, their levels were higher than the threshold level. Interestingly, the expression of OsCR4 and OsDEK1 was more powerfully decreased in KD lines. In the promoters of OsCR4 and OsSAL1, there are many P boxes, including the typical ones recognized by the RPBF, found in the GluB-1 promoter. In the OsDEK1 promoter there are many P boxes and two typical GCN4 motifs. Thus, RISBZ1 and RPBF may directly regulate the expression of these genes. We speculate that as both the RISBZ1 and RPBF products were detectable in our KD lines at a very early stage, OsCR4 and OsDEK1 may already be synthesized and maintained at a sufficient level, even in the absence of OsSAL1. Although the expression levels of OsCR4, OsDEK1 and OsSAL1 were similar in KD-RISBZ1 and KD-RPBF, only KD-RPBF exhibited patchy multilayered aleurone layers, suggesting that only RPBF is involved in such regulation, with the aid of another, unknown factor in this process. Furthermore, the altered accumulation level of storage compounds in KD lines may be caused by the aberrant distribution of aleurone layer specifying signals, because a number of the mutants with a disorganized aleurone layer displayed disturbed, starchy endosperm growth, with opaque and shrunken apparatus (Becraft and Acuncion-Crabb, 2000;Shen et al., 2003; Lid et al., 2004; Olsen et al., 2008).

The combination of RISBZ1 and RPBF plays a significant role during grain filling

In contrast to the maize o2 mutants, neither KD-RISBZ1 nor KD-RPBF showed a significant reduction of most of their SSPs. Mutations and KDs of transcription factors often fail to result in informative phenotypes, which can be explained by the functional redundancy of transcription factors belonging to the same family (Bouche and Bouchez, 2001; Hiratsu et al., 2003; Hirschi, 2003; Zhang, 2003). Thus, it seems reasonable that KD-RISBZ1 or KD-RPBF may be at least partially replaced by redundant trans-activators (RISBZ2-5 for RISBZ1, Onodera et al., 2001; OsSAD for RPBF, Washio, 2001). However, the compensatory mechanism for the KD of RISBZ1 and RPBF seems to be more complicated, so that some, but not all SSP genes were affected by single KD mutants. KD-RISBZ1/KD-RPBF displayed a much more severe phenotype than single KD mutants, indicating that the RISBZ1 bZIP and RPBF DOF transcription factors play redundant global roles during grain filling. In the KD-RISBZ1 seed, RPBF was slightly upregulated at early stages, and RPBF was maintained at a higher levels than that in the wild type at a later stage, and vice versa, suggesting that RISBZ1 and RPBF compensate for the reduced activities of their counterpart at both the transcriptional and post-transcription level, and mainly at the latter. There must be an as yet uncharacterized sensor and balancing system for these transcription factors. Transcription factors belonging to different types also redundantly regulate seed maturation, including SSP accumulation, in Arabidopsis (To et al., 2006). To our knowledge, such compensatory activity among those Arabidopsis transcription factors, or between maize O2 and PBF, has not been reported. This compensatory mechanism, rather than partial reduction of RISBZ1 and RPBF, may be responsible for the weak effects on SSP observed in the individual KD mutants, although we cannot rule out the latter possibility. We also demonstrated the physical interaction between RISBZ1 and RPBF in rice protoplasts. Combined with the synergism between them found both in vivo and in planta, this interaction may ensure fully proper rice seed development. An interaction between O2 and PBF in vitro has previously been reported (Vicente-Carbajosa et al., 1997). Thus, the importance of the combination of RISBZ1/O2 and RPBF/PBF is found in both rice and maize. Previously, we have shown that a combination of qualitative cis-element (GCN4) and quantitative elements (P box, AACA motif and ACGT motif) is necessary for endosperm-specific expression of the minimal GluB-1 promoter (Wu et al., 2000). Recently, the RY-repeat has been reported to be involved in SSP gene expression in cereals, as well as in Arabidopsis (Moreno-Risueno et al., 2008; Santos-Mendoza et al., 2008). Thus, combinatorial interactions not only between RISBZ1/O2 and RPBF/PBF, but also with other transcription factors, may be required. Such interaction has been reported in barley (HvGAMYB-BPBF/SAD, HvFUS3-BLZ2) and Arabidopsis (ABI3-AtbZIP10/25) (Diaz et al., 2002, 2005; Lara et al., 2003; Moreno-Risueno et al., 2008).

Further study on the regulatory mechanisms of these transcriptional activators, including compensation and interaction, will help advance the understanding of how rice seed development is established and maintained.

Experimental procedures

Transgenic rice plants

For the construction of ProRISBZ1:RISBZ1ΔAD, a partial RISBZ1 cDNA corresponding to the RISBZ1 protein without its N-terminal 40 amino acid residues (between positions 323 and 1513, accession number AB05347) was amplified by PCR using forward and reverse primers containing NcoI and BamHI recognition sites, respectively. The PCR product was digested with NcoI and BamHI, and then linked to the RISBZ1 promoter between positions −1674 and +213 (Onodera et al., 2001). For the construction of ProNRP33:RPBF, the entire coding sequence of RPBF cDNA (between positions 97 and 1218, accession number AK107294) was amplified by PCR and linked to ProNRP33 (Qu and Takaiwa, 2004). These constructs were then inserted into the modified binary vector pGPTV-35S-HPT (Yasuda et al., 2005).

Rice plants (O. sativa cv. Kita-ake) were transformed with the plasmids described above by Agrobacterium-mediated transformation, as described previously (Qu and Takaiwa, 2004).

SDS-PAGE and western blot

SDS-PAGE and western blots were carried out using total proteins from single seeds as described previously (Yamamoto et al., 2006). SDS-PAGE gels were stained with Coomassie Brilliant Blue G-250 (Nacalai Tesque, http://www.nacalai.co.jp). Anti-Prol14 rabbit antibody was produced against the synthetic peptide YGAPSTITTLGGVL from the C-terminal region between positions 135 and 148 of Prol14 (accession number M23744). Anti-α-globulin antibody was raised against the synthetic peptide DRQLTGRERFQ between positions 53 and 63 of α-globulin (accession number AK287940). Other primary antibodies used in western blots were described previously [anti-RISBZ1 and anti-RPBF antibodies (Yamamoto et al., 2006), anti-GluA, anti-GluB, anti-13-kDa prolamin (RM1), anti-10-kDa prolamin antibodies (Takagi et al., 2006), anti-14–16-kDa RA antibody (Matsuda et al., 1988) and anti-GluD antibody (Kawakatsu et al., 2008)].

Quantitative real-time PCR

Total RNA was extracted as described previously (Yasuda et al., 2005). After RNase-free DNase I treatment (Takara, http://www.takara-bio.co.jp), cDNA was synthesized using the Superscript III First-Strand Synthesis System for qRT-PCR (Invitrogen, http://www.invitrogen.com). qRT-PCR was performed in a volume of 20 μl using the SYBR Premix Ex TaqII (Perfect Real Time) kit (Takara) on an ABI Prism 7000 HT Sequence Detector (Applied Biosystems, http://www.appliedbiosystems.com). Reactions were performed following the manufacturer’s protocol. Triplicate assays were carried out. The expression levels of RISBZ1 and RPBF were normalized to 17S RNA using the expression levels in the wild type at 7 DAF as the reference. Expression levels of SSP genes, starch and lipid-metabolizing enzyme genes and aleurone layer number-related genes were normalized to 17S RNA. The primers used for qRT-PCR are listed in Table S1.

Scanning electron microscopy

Mature seeds were dissected into approximately 1-mm-thick sections with a razor blade, and fixed in PFA (4% paraformaldehyde, 0.2% glutaraldehyde and 50 mm sodium phosphate buffer, pH 7.2). After washing in sodium phosphate buffer, the samples were dehydrated in a graded ethanol series. Dehydrated samples in 100% ethanol were infiltrated with acetone, and then dried in air. Samples were observed under a scanning electron microscope (S-2380N; Hitachi, http://www.hitachi-hitec.com) at an accelerating voltage of 3 kV.

Histological analysis

Mature seeds were dissected into approximately 1-mm-thick sections with a razor blade. Sections were stained in a 0.1% solution of Sudan black B, dissolved in 70% ethanol, washed with 50% ethanol and then observed.

Analysis of the protein, starch and total lipid contents

Dehulled rice grains were ground to powder. To measure the protein, starch and lipid contents, 100 mg, 25 mg and 1.5 g of powder were used for the assay, respectively. All measurements were performed in triplicate. Nitrogen content was measured by a modified Dumas method using a nitrogen analyzer (FP-528; Leco, http://www.leco.com). The protein content was calculated from the nitrogen content (with a protein conversion factor of 5.95). The starch content was measured using a starch assay kit (Sigma-Aldrich, http://www.sigma-aldrich.com), following the manufacturer’s protocol. The lipid content was determined by the Folch method (Folch et al., 1957). Total lipid was extracted with chloroform/methanol (2:1) three times, and then filtered (DISMIC-25; Advantec, http://www.advantecmfs.com). To remove contaminants, distilled water was added and centrifuged, and then the chloroform fraction containing the lipid was collected and weighed after evaporation.

Transient assays, immunocytochemistry and affinity co-purification

Pro35S:RISBZ1 was constructed as described previously (Onodera et al., 2001). Pro35S-ShΔ:RISBZ1-2HA-6His and Pro35S-ShΔ:RPBF-2HA-6His were constructed by cloning RISBZ1 and RPBF full-length cDNA into Pro35S-ShΔ:2HA-6His vector, kindly provided by Dr Kagaya (Mie University). The preparation of rice protoplasts and transfection were carried out as described previously (Yamamoto et al., 2006). Transfected protoplasts were fixed in PBS (140 mm NaCl, 2.7 mm KCl, 10 mm Na2HPO4 and 1.8 mm KH2PO4), containing 3.7% formaldehyde, for 30 min at room temperature. After washing with PBS three times, protoplasts were permeated with PBS containing 0.3% Triton X-100 for 30 min at room temperature, and then blocked in blocking buffer (PBS containing 1% BSA and 0.02% Tween20) for 1 h at room temperature. Anti-HA monoclonal mouse antibody (Sigma-Aldrich) was added up to a 1:1000 dilution, and incubated for 30 min at room temperature. After washing with PBS three times, samples were incubated with 1:1000 diluted Alexa488 conjugated anti-mouse IgG goat antibody (Invitrogen) in blocking buffer for 1 h. After washing with PBS three times, samples were mounted with Antifade solution (Invitrogen), containing 1 μg ml−1 DAPI, and observed under fluorescence microscopy (BX-50; Olympus, http://www.olympus.com). Image processing was performed with Photoshop CS3 (Adobe, http://www.adobe.com). For affinity co-purification, proteins were extracted from protoplasts with extraction buffer (10 mm Tris, pH 8.0, 200 mm NaCl, 0.1% Triton X-100, 10% glycerol, 20 mm imidazol and Complete mini EDTA-free protease inhibitor cocktail (Roche, http://www.roche.com). His-tagged proteins were affinity purified with Ni-NTA Spin Columns (Qiagen, http://www.qiagen.com), washed three times with wash buffer (extraction buffer containing 40 mm imidazol instead of 20 mm, without Triton X-100) and eluted with elution buffer (wash buffer containing 250 mm imidazol instead of 40 mm).

Acknowledgements

We thank Dr T. Matsuda (Nagoya University) and Dr Y. Kagaya (Mie University) for providing the anti-14–16-kDa RA antibody and the Pro35S-ShΔ:2HA-6His vector, respectively, Dr Y. Kobayashi and Dr T. Hattori (Nagoya University) for valuable suggestions and technical advice about the transient assay using rice callus protoplasts, Dr Kawano and Dr Nagao (National Food Research Institute) for technical assistance in the measurement of protein and lipid contents, respectively, and Prof. K. Yamada (University of Toyama) for encouragement. This work was supported by research grants from the Ministry of Agriculture, Forest and Fisheries of Japan (Functional analysis of genes relevant to agriculturally important traits in rice genome, IP2001, Genomics and Agricultural Innovation, GMC0008) to FT, and by a Grant-in-Aid for Young Scientists from the Ministry of Education, Culture, Sports, Science and Technology of Japan (19780011 to TK and 20880013 to MY).

Ancillary