Reducing Rice Seed Storage Protein Accumulation Leads to Changes in Nutrient Quality and Storage Organelle Formation 1[W][OA]

Rice ( Oryza sativa ) seed storage proteins (SSPs) are synthesized and deposited in storage organelles in the endosperm during seed maturation as a nitrogen source for germinating seedlings. We have generated glutelin, globulin, and prolamin knockdown lines and have examined their effects on seed quality. A reduction of one or a few SSP(s) was compensated for by increases in other SSPs at both the mRNA and protein levels. Especially, reduction of glutelins or sulfur-rich 10-kD prolamin levels was preferentially compensated by sulfur-poor or other sulfur-rich prolamins, respectively, indicating that sulfur- containing amino acids are involved in regulating SSP composition. Furthermore, a reduction in the levels of 13-kD prolamin resulted in enhancement of the total lysine content by 56% when compared with the wild type. This observation can be mainly accounted for by the increase in lysine-rich proteins. Although reducing the level of glutelins slightly decreased protein storage vacuoles (PSVs), the simultaneous reduction of glutelin and globulin levels altered the inner structure of PSVs, implicating globulin in framing PSV formation. Knock down of 13-kD prolamins not only reduced the size of endoplasmic reticulum-derived protein bodies (PBs) but also altered the rugged peripheral structure. In contrast, PBs became slightly smaller or unchanged by severe suppression of 10- or 16-kD prolamins, respectively, indicating that individual prolamins have distinct functions in the formation of PBs. Extreme increases or decreases in sulfur-poor prolamins resulted in the production of small PBs, suggesting that the ratio of individual prolamins is crucial for proper aggregation and folding of prolamins.

Seed storage proteins (SSPs) are stored nitrogen sources for germinating seedlings and thus serve as a nutrient source for humans and livestock. Since SSPs account for a substantial portion of the total protein in seeds, SSP content and composition are related to the nutrient quality of seeds (Shewry and Halford, 2002). Improving the nutrient composition of seeds is a major target of molecular breeding programs; however, since SSP genes constitute multigene families, even several mutations in a few structural genes have little effect on seed protein content and amino acid composition.
In most cereals, prolamins are the major SSP, whereas rice (Oryza sativa) and oat (Avena sativa) preferentially accumulate SSPs belonging to the 11S-type globulin family, glutelin in rice and 12S globulin in oat (Shewry and Halford, 2002). In rice, 60% to 80% of total seed protein is composed of glutelins, encoded by 15 genes in the rice genome, and 20% to 30% of total seed protein are prolamins that are encoded by 34 genes (Kawakatsu et al., 2008;Xu and Messing, 2009). Glutelins can be classified into four groups (GluA, GluB, GluC, and GluD) based on amino acid sequence similarity (Kawakatsu et al., 2008). Prolamins are classified into three groups (10, 13, and 16 kD) by their molecular mass according to their electromobility on SDS-PAGE gels. The 13-kD prolamins can be further classified into three subgroups (class I, II, and III) by their Cys residue content (Muench et al., 1999). Based on amino acid sequence similarity, 16-and 13-kD prolamins correspond to maize g 2 -zein, and 10-kD prolamins correspond to d-zeins (Xu and Messing, 2009).
Glutelins and globulin are deposited into protein storage vacuoles (PSVs), whereas prolamins accumulate in endoplasmic reticulum (ER)-derived protein bodies (PBs) that are formed within the lumen of the rough ER (Krishnan et al., 1986;Li et al., 1993). PBs are spherical (diameter 1-2 mm) with a low electron density structure surrounded by an ER-origin membrane with polysomes (Yamagata and Tanaka, 1986). PSVs have an irregularly shaped and electron-dense structure with a diameter of 2 to 4 mm. Glutelins are localized in the inner region of PSVs as crystalloids with lattice structures, and a-globulin is largely sequestered at the peripheral matrix surrounding the glutelins (Krishnan et al., 1992).
Maize (Zea mays) opaque2 (o2) is a classic mutant of the maize kernel. O2 encodes a basic Leu zipper transcription factor, and mutations in this gene result in a severe reduction in 22-kD a-zein accumulation and alternation to an opaque and floury kernel phenotype (Hartings et al., 1989;Schmidt et al., 1990). The o2 kernel also contains an elevated level of Lys (Mertz et al., 1964). When 22-kD a-zeins were specifically suppressed in endosperm by RNA interference, transgenic maize kernels changed to the opaque and floury phenotype, and the Lys content was elevated in a similar manner as o2, suggesting that these changes are due to a reduction in 22-kD a-zeins (Segal et al., 2003). Functions of individual zeins in PB formation and opacity have recently been examined using transgenic maize (Wu and Messing, 2010). Although individual reductions of band g-zeins had little effect on PB formation, their simultaneous reduction severely distorted PB structure, suggesting a redundant function of individual zeins for stabilizing PBs. Furthermore, PBs of plants with a substantially reduced 22-kD a-zein content produced budding structures, indicating that the major 22-kD a-zeins are crucial for the maintenance of normal PBs; however, d-zeins may play a minor role in PB formation. Furthermore, not only a-zeins but also band g-zeins are important for the vitreous kernel phenotype. Compositional changes in amino acid composition by reducing b-, g-, and d-zeins were not examined.
We have generated transgenic rice plants in which different SSP genes were suppressed (H. Yasuda, Y. Wakasa, T. Kawakatsu, and F. Takaiwa, unpublished data). In this study, we analyzed transgenic seeds with reduced SSP(s) to decipher the effect of reducing SSPs on seed phenotype, including protein storage organelle formation during grain filling. Our results revealed that reductions in SSP expression were compensated for by increases in other SSPs at the mRNA transcript and protein accumulation levels. Notably, a reduction in the 13-kD prolamins resulted in an enrichment of Lys, an essential amino acid, without a corresponding change in grain phenotype. Analysis of the intracellular structures in these gene-silenced plants suggests that globulin, or 13-kD prolamins and the ratio of individual prolamins, play important roles for PSV or PB formation, respectively. Irrespective of SSP reduction, the endosperms of SSP lines were vitreous and different from maize kernels, indicating that the change to the opaque or floury phenotype is caused by different mechanisms in rice and maize.
Next, we investigated SSP levels in detail by immunoblot analysis using specific antibodies. Our anti-GluA antibody reacts with GluA-1 and GluA-2, but not with GluA-3, whereas the anti-GluB antibody reacts with GluB-1 and GluB-2. Levels of GluA-2 were remarkably suppressed in Glu-less, whereas levels of GluA-1 were slightly suppressed in GluBless (Fig. 1B). In other constructs, GluA levels were not changed (Fig. 1B). Levels of GluB-1 and GluB-2 were significantly suppressed to below detectable levels in Glu-less, GluB-less, and GluB·Glb-less (Fig. 1B). In contrast, accumulation of GluB-1 and GluB-2 was enhanced in 13-, 10-, and 16-kD Pro-less (Fig. 1B). GluC levels increased in all SSP-less constructs (Fig.  1B). GluD-1 levels decreased in GluB-less but increased in other constructs (Fig. 1B). The 13-kD prolamins are classified into three classes (I-III) based on their deduced amino acid composition, with the most prominent difference being the Cys content. The class I prolamins (e.g. RM2 and RM4) are poor in Cys (one Cys residue per molecule), the class II prolamins (e.g. RM9) contain the highest number of Cys residues (nine Cys residues per molecule), and the class III prolamins (e.g. RM1) contain an intermediate level of Cys residues (five Cys residues per molecule). The 10-kD prolamin, RP10, and the 16-kD prolamin, RP16, are rich in Cys, containing 11 and 13 Cys residues per molecule, respectively. In 13-kD Pro-less transgenic rice, levels of RM1, RM2, and RM4 were significantly suppressed, but RM9 levels were reduced only slightly. In glutelin-less lines, increases in the levels of Cys-rich RM1 and RM9 were very modest, whereas levels of Cys-poor RM2 and RM4 were markedly enhanced (Fig. 1B). It is likely that the Cys-poor prolamins largely accounted for the increased levels of 13-kD prolamins in Glu-less. In contrast, Cys-rich RM1 and RM9 levels increased in 10-and 16-kD Proless, and Cys-poor RM2 and RM4 levels were reduced in 10-kD Pro-less and unchanged or slightly increased in 16-kD Pro-less (Fig. 1B). RP10 levels were significantly suppressed in 10-kD Pro-less and relatively enhanced in other SSP-less lines. RP16 levels were greatly reduced in 16-kD Pro-less, increased in glutelinless lines, slightly increased in 13-kD Pro-less, and significantly increased in 10-kD Pro-less (Fig. 1B). The reduction of Cys-rich prolamin levels tended to be compensated by other Cys-rich prolamins. Glb-1 levels were greatly suppressed in GluB·Glb-less but increased in the other SSP-less lines (Fig. 1B). In conclusion, reduced SSP levels were compensated by increases in the accumulation of other SSPs. Compensation effects are summarized in Table II.
To examine whether changes in SSP accumulation are reflected by altered mRNA levels, we performed quantitative reverse transcription (qRT)-PCR with RNA extracted from maturing seeds at 10 d after flowering (DAF). In all SSP-less lines, expression of targeted SSP genes was effectively down-regulated (Fig. 2). In contrast, expression levels of most nontargeted SSP genes were up-regulated, presumably as a consequence of suppressing the target gene(s), with some exceptions. RM4 transcript levels were greatly decreased in 10-kD Pro-less transformants (Fig. 2). Since RM9 and RM1, which share a higher homology to RP10 than RM4, were not suppressed, down-regulation of RM4 was suggested to be a side effect of RP10 suppression. RM9 transcript levels slightly decreased in 13-kD Pro-less seeds, consistent with the relatively low similarity between RM9 and trigger genes (RM1 and RM2; Fig.  2). These mRNA expression levels basically paralleled the SSP accumulation levels, indicating that compensation for reduced SSPs by increasing other SSPs is primarily regulated at the transcriptional level.

Grain Phenotypes
We examined the effect of reducing the levels of several SSPs on grain phenotype. Brown seed weight slightly decreased in Glu-less, GluB-less, 13-kD Proless, and 16-kD Pro-less and, on the other hand, slightly increased in GluB·Glb-less and 10-kD Pro-less (Table III). Starch content was similar in the glutelin-less lines and slightly increased in 10-and 16-kD Pro-less when compared with the wild type (Table III). Nitrogen content slightly decreased in Gluless, GluB-less, 10-kD Pro-less, and 16-kD Pro-less (Table III). There were some differences in brown seed weight, starch content, and total nitrogen content among the wild type and the SSP-less lines. This result is in contrast to analyses of grains containing reduced levels of transcription factors regulating SSP genes, such as RISBZ1 and RPBF, in which the nitrogen and starch contents were significantly reduced ).
Next, we examined seed phenotypes. Grain size (height, width, and thickness) was statistically different in some SSP-less lines examined here (Table III), but the degrees were within negligible levels. Rice endosperm is vitreous and easily transmits light. Since a reduction in zein content causes an opaque and floury phenotype in maize, we investigated whether SSP reduction causes opacity in rice endosperm; however, SSP-less lines did not show any apparent opaque phenotype (Fig. 4). In maize, unfolded protein reaction (UPR) caused by uncleaved signal peptides of zein is also related to the opaque and floury phenotype (Coleman et al., 1997;Kim et al., 2004). A reduced starch content loosens the packaging of starch granules and causes opacity in rice ). Since our SSP-less lines did not exhibit a strong UPR and contained starch at normal levels, their endosperms were vitreous.

Total Amino Acid Contents
SSPs represent a substantial portion of all amino acids in the endosperm. Therefore, compensatory synthesis of additional SSP(s) after reducing the accumulation of one SSP is expected to change the amino acid composition of total seed proteins. To examine the total amino acid profile, bulked seeds from more than 20 panicles were ground into a fine powder and used for amino acid analyses. The sum of total amino acid content increased in Glu-less, 13-kD Pro-less, and 16-kD Pro-less (Supplemental Table S1). In Glu-less, the levels of Glx (Glu and Gln), Ser, Gly, Ala, Tyr, Leu, and Pro increased, and the level of Trp decreased ( Fig. 5A; Supplemental Table S1). In 13-kD Pro-less seeds, the levels of all amino acids, except for CY2 (cystine), Met, and Trp, increased ( Fig. 5A; Supplemental Table S1). Notably, the Lys content increased by 56% in 13-kD Pro-less ( Fig. 5A; Supplemental Table S1). In 16-kD Pro-less seeds, Asx (Asp and Asn), Glx, Ser, Gly, Thr, Arg, Ala, Tyr, Phe, Leu, and Pro levels increased, and CY2 decreased ( Fig. 5A; Supplemental Table S1). In GluB-less, GluB·Glb-less, and 10-kD Pro-less, the sum of total amino acid contents was similar to that in the wild type. In GluB-less, Asx, Arg, Tyr, Met, Trp, and Phe levels decreased. In GluB·Glb-less, Ala, Tyr, Ile, Leu, and Pro levels increased, and Met and Trp levels decreased. In 10-kD Pro-less, CY2 and Met levels Figure 1. SSP accumulation in wild-type and transgenic plants. A, SDS-PAGE of rice seed proteins. Bands for glutelin precursors (approximately 50 kD), glutelin acidic subunits (approximately 30 kD), glutelin basic subunits (approximately 22 kD), a-globulin (approximately 24 kD), and 13-kD prolamins (approximately 13 kD) are indicated. Acidic subunits of glutelins (GluB-4, GluA-2, GluA-1, GluA-3, GluB-2, and GluB-1, from top to bottom) can be detected on Coomassie Brilliant Blue-stained SDS-PAGE gels (Supplemental Fig. S1), but bands corresponding to GluB-2 and GluB-1 are not separated in our SDS-PAGE conditions. B, Immunoblot using specific antibodies. Four antibodies for glutelins (GluA, GluB, GluC, and GluD), four antibodies for 13-kD prolamins (RM1, RM2, RM4, and RM9), and antibodies for 10-kD prolamin (RP10), 16-kD prolamin (RP16), and a-globulin (Glb-1) were used. Numbers in parentheses indicate the number of Cys residues per molecule.   Table S1). In summary, although individual amino acid levels changed, reducing the levels of SSPs did not affect the amino acid compositions drastically (Fig. 5A), a result that is in contrast with observations of transgenic maize, which had significantly reduced levels of both 22and 19-kD a-zeins (Huang et al., 2006).

Free Amino Acid Contents
Although Low glutelin content-1, a rice mutant with a highly reduced glutelin content, contained free amino acids at levels similar to the wild type, globulin reduction increased the overall free amino acid content by approximately 50% (Ashida et al., 2006). Consistently, maize kernels with reduced amounts of zein proteins contain highly elevated levels of free amino acids (Huang et al., 2006). Reducing SSP accumulation in rice had more subtle effects (Fig. 5B). The sum of free amino acids increased in Glu-less and GluB·Glbless, decreased in 10-kD Pro-less and 16-kD Pro-less, and was similar in GluB-less and 13-kD Pro-less, when compared with that of the wild type (Supplemental Table S2). Asp, Glu, Asn, and Arg largely contributed to the changes in the sum of free amino acids ( Fig. 5B; Supplemental Table S2). However, the degrees of increase and decrease were much lower than those observed in maize kernels with reduced levels of zein proteins (Supplemental Table S2; Huang et al., 2006). These findings suggest that surplus amino acids derived from suppressed SSP accumulation were efficiently incorporated into other proteins, especially other SSPs, in our transgenic rice lines. Increases in free Lys (50 nmol g 21 dry weight) were dwarfed in comparison with the increase in total Lys (15 mmol g 21 dry weight), indicating that increases in the total Lys content were due to an increase in the Lys-rich non-13-kD prolamin proteins in 13-kD Pro-less mutants (Supplemental Tables S1 and S2).

Storage Organelle Formation
To investigate the effect of reducing SSP levels on PB formation, subcellular structures in the subaleurone layer of rice endosperm cells were examined. Thick sections (100 mm) at 15 to 20 DAF were stained with rhodamine B to localize ER-derived PB (magenta) organelles and immunostained with anti-OsTIP3 (for tonoplast intrinsic protein) antibody to visualize PSV (green) structures and then subjected to analysis by confocal laser microscopy. Glutelins and globulin are deposited in PSVs, and prolamins localize into PBs. In wild-type endosperm, PBs were smaller and more spherical than PSVs, although PSVs were more abundant (Fig. 6A). In the glutelin-less lines, Glu-less, GluBless, and GluB·Glb-less, PB-IIs were relatively fewer and smaller (Fig. 6, A-D). This result indicates that glutelins are responsible for the expansion of PSVs, and compensation by Glb-1 is not sufficient for recovery to the normal-sized PSVs. In Glu-less and GluBless, small PBs were widely distributed, in addition to normal-sized PBs (Fig. 6, B and C). In GluB·Glb-less, the number of relatively small PBs increased (Fig. 6D). This difference may reflect an increase in levels of 13-kD prolamin (Fig. 1). In 13-kD Pro-less, rhodaminelabeled PBs could barely be detected, and PBs became very small and changed to dot-like structures, reflecting the importance of 13-kD prolamins for the structure of wild-type PBs (Fig. 6E). In contrast, both larger and smaller PSVs increased in 13-kD Pro-less, reflecting the increase in glutelins and globulin levels (Figs. 1 and 6E). In 10-kD Pro-less, PBs were smaller than in the wild type, PSVs were slightly larger, and the number of small PSVs also increased in a manner similar to 13-kD Pro-less (Fig. 6F). In 16-kD Pro-less, PBs were smaller than in the wild type and the number of additional small PSVs increased, but the degree of these alterations was much lower than for 13-kD Proless and 10-kD Pro-less, reflecting the notion that 16-kD prolamin is a minor component (Fig. 6G). Next, we observed the organization of PBs in the developing endosperm at 15 DAF by transmission electron microscopy. In addition to size, PBs and PSVs can be distinguished by their electron densities (PBs, low density; PSVs, high density; Fig. 7, A and B). As observed by confocal laser microscopy, fewer small PSVs and incomplete PSVs with empty space were observed in Glu-less, GluB-less, and GluB·Glb-less (Fig. 7, C-H). Small but normal PBs were also observed (Fig. 7, C-H). It should be noted that PSVs appeared cracked in the micrographs irrespective of their normal size in GluB·Glb-less ( Fig. 7H; Supplemental Fig. S5), indicating that Glb-1 alone or in combination with glutelins is important for the proper organization of PSVs. Although the rhodamine B signals were very low in 13-kD Pro-less, there were many very small PBs (Figs. 6F and 7I). The majority of these small PBs might not have been detected by confocal laser microscopy or were not stained by rhodamine B. The peripheries of these small PBs were distorted (Fig. 7J). Additionally, elongated and abnormally shaped PSVs were also observed (Fig. 7I). In 10-and 16-kD Pro-less, smaller PBs were observed, but their peripheries were smooth (Fig. 7, K-N). In prolamin-less lines, small PSVs were also observed. Figure 2. SSP gene expression in wild-type and transgenic plants. Expression levels of SSP genes were analyzed by qRT-PCR using RNA extracted from maturing grains at 10 DAF. Expression levels are normalized to 17S rRNA and are represented relative to the expression level in wild-type grain at 10 DAF. Error bars indicate the SD of three replicates. Light blue, red, yellow, green, purple, orange, and brown bars indicate the wild type, Glu-less, GluB-less, GluB·Glb-less, 13-kD Pro-less, 10-kD Pro-less, and 16-kD Pro-less, respectively. Asterisks indicate statistical difference from the wild type by Dunnett's test (P , 0.05).

Loss of SSP Is Compensated by Other SSPs
Cereals are an important source of carbohydrates as well as nitrogen that are reserved as starch and SSPs. SSPs have been extensively studied because cereal SSP content and composition are related to the nutritional quality of seeds (Shewry and Halford, 2002). Furthermore, recombinant protein production utilizing SSPs has become more popular due to high levels of expression and stability (Takaiwa, 2007). Since cereal SSPs are encoded by multigene families, only a few SSP-deficient mutants have been induced by mutagenesis using chemical mutagens or irradiation (Iida et al., 1997(Iida et al., , 1998Kusaba et al., 2003). In this study, we investigated the effects of SSP reduction not only for glutelins and globulins but also for prolamins. Reduction of each individual SSP was compensated by the increase of almost all other SSPs (Fig. 1), indicating the existence of a consistent compensation mechanism. In maize, reductions in zein levels are not compensated by increases in other types of zeins, except that the suppression of 22-kD a-zein was compensated for by 19-kD a-zein, and vice versa (Huang et al., 2004;Wu and Messing, 2010). Compensation for reductions in SSP levels was achieved at the mRNA expression level and was well correlated with protein accumulation levels, indicating that the accumulation levels of endogenous SSPs are mainly regulated at the transcriptional or posttranscriptional levels. Perceiving available free amino acids, the activities of transcription factors that generally regulate SSP gene expression, such as RISBZ1 and RPBF, may be increased.
When sulfur-rich RP10 was suppressed, the levels of other sulfur-rich prolamins were enhanced, but levels of sulfur-poor prolamins were depressed (Table II). Regarding these amino acid redistributions, it is important to consider that there may be a regulatory mechanism for maintaining the total nitrogen content. When sulfur-rich sunflower (Helianthus annuus) seed albumin accumulated to high levels in transgenic rice seeds, the total amount of sulfur-containing amino acids, Met and Cys, was unaffected (Hagan et al., 2003). In these seeds, the total seed protein content remained nearly constant, whereas individual SSP compositions changed: native sulfur-rich SSPs decreased, whereas sulfur-poor SSPs increased. Therefore, a surplus of available sulfur-containing amino acids and their key metabolic intermediates may activate sulfur-rich, or inhibit sulfur-poor, prolamin synthesis at the transcriptional or posttranscriptional level. Conversely, when it comes to a shortage, such redistribution is reversed. In other words, the availability of sulfur may determine which prolamin, sulfur rich or sulfur poor, should be newly synthesized as a storage reserve to maintain the appropriate amino acid balance. This hypothesis is also supported by the more significant increase in sulfur-poor 13-kD prolamin levels, rather than the slight increase in sulfur-rich prolamin levels in the glutelin-less lines, since the reduction in levels of relatively sulfur-poor glutelins gave rise to subtle increases in levels of available sulfur-containing amino acids. Such dominant regulation by sulfur-containing amino acids is due to the unavailability of inorganic sulfur in rice endosperm (Hagan et al., 2003).   The 13-kD Pro-less Is a High-Lys Rice The 13-kD Pro-less seed contained a higher content (56%) of total Lys than the wild type (Supplemental Table S1), corresponding to an increase of 26% in the Lys-protein ratio. A reduction in a-zein content by the o2 mutation or RNA interference knockdown also resulted in an enhancement of the total Lys content of maize grains (Mertz et al., 1964;Segal et al., 2003;Huang et al., 2006). This result was attributed to increases in the levels of nonzein Lys-rich proteins, such as elongation factor 1A (Habben et al., 1993). As shown in Figure 1, suppression of 13-kD prolamins was compensated for by an increase in the glutelins (Fig. 1). Glutelins contain more than 10 Lys residues per molecule, in contrast to the prolamins, which contain almost no Lys residues. Therefore, the increase in glutelin levels may contribute to the high-Lys content of 13-kD Pro-less. Although differences were not significant, Lys decreased in the glutelinless lines and increased in the prolamin-less lines ( Fig.  5A; Supplemental Table S1). Two rice mutants that contained lower amounts of 13-kD prolamins had a relatively increased Lys-protein ratio (Krishnan, 1999). The glutelin-prolamin ratio may be one determinant for Lys content in seeds. This idea is consistent with the observation that the glutelin-prolamin ratio is reversed in o2 compared with the wild type (Mertz et al., 1964). Prolamins are major SSPs in most cereals, whereas glutelins and globulins are major SSPs in rice and oat. Rice and oat grains are better sources of Lys than other cereal grains, suggesting that the glutelin content is crucial for the Lys content in cereal seeds (Juliano, 1993).
We also found three candidate seed proteins that likely contribute to Lys content. Lys-protein ratios of rice embryo globulin 1 (REG1), REG2, and BiP are 3.4%, 3.8%, and 9.4%, respectively. Expectedly, levels of these proteins increased in 13-kD Pro-less ( Fig. 3;  Supplemental Fig. S2). Rice high-Lys mutants also contain increased levels of embryo globulins (Schaeffer and Sharpe, 1987). A similar situation has been observed in maize kernels. Maize orthologs of embryo globulins and BiP increased in high-Lys opaque mutants (Zhang and Boston, 1992). Therefore, not only the glutelin-prolamin ratio, but also embryo globulins and BiP, are likely key components for seed Lys content. This hypothesis is consistent with the evidence that Lys content was not drastically decreased in glutelinless lines. Increases in embryo globulins and BiP may compensate for glutelin reduction. Considering that the ratio of endosperm protein amount to embryo protein amount is quite high (more than 80%), and glutelins and most BiP are endosperm proteins, Lys increased largely in the endosperm (Supplemental Fig. S3).
Lys is an essential amino acid and is generally deficient in cereal grains. Therefore, exogenous Lys is supplemented in livestock feed. Increasing the Lys content in total digestible nitrogen is an important target of breeding programs for Lys-deficient cereal grains. Several approaches have been tried to increase the Lys content in cereal grains. Classically, o2 and floury2 mutants, in which Lys-deficient zeins were reduced, were identified as high-Lys maize lines (Mertz et al., 1964;Nelson et al., 1965), and the high-Lys phenotype was mimicked by suppression of a-zeins (Segal et al., 2003;Huang et al., 2006). High-Lys mutants have been also identified in rice (Schaeffer and Sharpe, 1987). However, these high-Lys mutants had chalky and floury phenotypes, and their agronomic traits were inferior to the wild type owing to brittleness and insect susceptibility. The 13-kD Pro-less contained a high Lys content due to the increase in glutelins, globulins, and BiP. A large part of the increased Lys is likely to be contained in these digestible proteins. The 13-kD Pro-less had a vitreous endosperm and contained starch at a similar level to the wild type. The difference, opaque in maize but vitreous in rice, may be caused by differences in starch grain morphologies. Rice has a compound starch grain consisting of variant starch granules in an amyloplast, while maize has a simple starch grain consisting of a single starch granule in an amyloplast (Matsushima et al., 2010). Interactions between PBs and starch granules are related to the vitreous kernel phenotype (Gibbon et al., 2003). Since starch granules are already packed into starch grains, interaction between PBs and starch granules may not be required for vitreous endosperm in rice. Therefore, 13-kD Pro-less is a good breeding resource for high-Lys rice.
severely suppressed in the Glu-less, GluB-less, and GluB·Glb-less lines, PSVs became smaller and fewer ( Fig. 6; Table II). In micrographs, PSVs in GluB·Glb-less looked cracked, although PSVs in Glu-less and GluBless appeared normal ( Fig. 7; Supplemental Fig. S5), indicating that PSV structure is not sufficiently maintained by filling with glutelins alone. The role of Glb-1 during PSV formation may be to construct a frame for PSVs, and glutelins may be the filling inside the frame, so that in the absence of Glb-1, glutelins were improperly packed into PSVs.
In contrast to maize PB formation (Lending and Larkins, 1989), rice PB formation is not well characterized. In developing maize endosperm, the band g-zeins are first deposited to PBs, followed by the accumulation of aand d-zeins in the inner region. Finally, the PB is fully filled with aand d-zeins, and band g-zeins are localized in the periphery. Suppression of 22-kD a-zein levels caused protuberances from the PB, indicating that 22-kD a-zein not only expands the PB but also maintains the shape of the PB (Wu and Messing, 2010). PBs without either bor g-zeins exhibited a slightly abnormal and underdeveloped shape (Wu and Messing, 2010). Reductions of both band g-zeins drastically changed PB shape, resulting in PBs with jagged peripheries, as shown in rice 13-kD Pro-less ( Fig. 7J; Wu and Messing, 2010). Therefore, maize band g-zeins and rice 13-kD prolamins may Figure 5. Total and free amino acid contents in mature grain of wild-type and transgenic plants. A, Total amino acid contents. B, Free amino acid contents. Error bars indicate the SD of three replicates. Light blue, red, yellow, green, purple, orange, and brown bars indicate the wild type, Glu-less, GluB-less, GluB·Glb-less, 13-kD Pro-less, 10-kD Pro-less, and 16-kD Pro-less, respectively. For actual values and statistical analyses, see Supplemental Tables S1 and S2. DW, Dry weight. provide a similar function during PB formation. This idea is consistent with phylogenic analysis in which band g-zeins in maize and 13-and 16-kD prolamins belong to group II (Xu and Messing, 2009). However, PBs in 13-kD Pro-less are much smaller than in the wild type, in contrast to the relatively normal-sized PBs in the double knockdown of band g-zeins. This observation is reasonable in light of the fact that 13-kD prolamins represent a very significant proportion of the rice prolamins. Individual 13-kD prolamins may play distinct roles in keeping a spherical shape and in expansion, perhaps depending on Cys residues. Since band g-zeins contain a number of Cys residues, disulfide bonds formed between them may be important for maintaining normal PB shape. Therefore, Cysrich 13-kD prolamins may play a role in PB formation corresponding to band g-zeins, whereas Cys-poor 13-kD prolamins may be involved in enlarging PBs. Although 16-kD Pro-less did show little alteration in PB formation, this result may be explained by the low level of 16-kD prolamin and compensation by other prolamins.

Effects of PSV and PB Protein Reduction on PB and PSV Formation, Respectively
Although both PSV proteins and PB proteins are synthesized as secretory proteins on rough ER, PSV proteins are transported into PSVs via Golgi or directly as PAC vesicles, whereas PB proteins aggregate to form PBs in the ER lumen (Kawakatsu and Takaiwa, 2010). Cereal grains contain either PSVs or ER-derived PBs as the deposition sites of SSPs (Shewry and Halford, 2002). However, rice seed is unique in that both PSVs and PBs coexist in the same endosperm cell, providing interesting material for studying the relationship between their structure and SSP accumulation. From the analysis of transgenic lines deleting PSV proteins (glutelins and globulin) and PB proteins (prolamins), the reduction of SSPs was compensated by other SSPs depending on the available free amino acid source, irrespective of any SSP. In PSV proteindeficient lines, in which glutelins and/or globulin were suppressed, sulfur-poor 13-kD prolamins were preferentially increased (Table II). As a result, the number of small PBs numerously increased (Fig. 6), suggesting that the ratio of Cys-poor to Cys-rich prolamins may be important for proper aggregation/ folding of prolamins. This observation is consistent with the results obtained for 13-kD Pro-less, in which almost all the Cys-poor prolamins were deficient ( Fig.  7I; Table II). This decline likely caused UPR, resulting in BiP induction. Another explanation is that BiP was up-regulated to compensate for lost BiP in the ER, since BiP was captured in the matrix of the increased numbers of PB (Li et al., 1993). Increases in Cys-rich 13-kD prolamins might assist proper aggregation/folding of prolamins and provide something of a release from UPR in PSV protein-deficient lines (Table II).
In PB protein-deficient lines, PSV proteins increased to compensate for the loss of PB proteins. In contrast to the increase in PB proteins, an increase in PSV proteins was not likely to induce UPR, suggesting that PSV proteins were properly folded and deposited into PSV, or that the ER loading was not sufficient to induce UPR. It is difficult to evaluate the effect of increased PSV proteins on PSV formation owing to its irregularly formed structure. However, it is notable that elongated PSVs were observed in 13-kD Pro-less, and additional small PSVs were also observed in PB protein-deficient lines ( Fig. 6; Table II). Empty PSVs were not observed in PSV protein-deficient lines, although there were incomplete PSVs with an empty space, suggesting that vacuole formation or development was inhibited ( Fig.  7; Table II; Supplemental Fig. S4). Therefore, PSV formation and/or development may be under feedback regulation by PSV protein amounts.
Although recent studies have advanced understanding of protein trafficking into storage organelles in rice endosperm, less availability of SSP mutants prevented the analysis of the formation of storage organelles. In this study, we showed the contribution of SSP composition to nutrient quality of rice grain and to storage organelle formation in rice endosperm by using various SSP knockdown lines. Our data also indicate the distinct roles of individual SSPs on storage organelle formation. Further studies, including transgenic lines with multiple suppressed genes and timecourse analyses of SSP localization during PSV and PB formation, will clarify the roles of these proteins in protein organelle formation, and this knowledge will help storage organelle engineering for the production of high-value recombinant proteins in rice endosperm.

Plant Materials and Growth Conditions
We have identified a modified Glucagon-like peptide-1 sequence as an RNA silencing-inducible sequence (RSIS) in rice (Oryza sativa) and developed a new RNA-silencing system, which is induced by simply linking RSIS to a target gene sequence (Yasuda et al., 2005;H. Yasuda, Y. Wakasa, T. Kawakatsu, and F. Takaiwa, unpublished data). In this system, two target genes can be simultaneously suppressed by linking a unique target sequence to either the 5# or 3# end of RSIS. Vector construction is summarized in Table I. Rice (cv Kita-ake) was transformed by the Agrobacterium tumefaciens-mediated method (Goto et al., 1999). Wild-type and transformant plants were grown at 25°C/20°C (12/12 h) under natural light conditions in pots (12 cm diameter) containing a commercial soil mixture (Bonsol No. 1; Sumitomo Chemicals).

Seed Protein Extraction and Immunoblotting
Extraction of total proteins from rice seeds, SDS-PAGE, and immunoblotting were performed as described previously (Kawakatsu et al., 2008). Total seed proteins were extracted independently from four seeds per line. After identical SSP compositions were confirmed by SDS-PAGE (data not shown), they were mixed and analyzed. Embryo proteins and endosperm proteins were extracted after the embryo was excised with a razor blade. Equal volumes were loaded, meaning that proteins per grain could be compared. Anti-RM2 and anti-RM9 peptide antibodies were produced against synthetic peptides YSAPDSITTLGGVLY and CGIYPSYNTAP, respectively. Other antibodies were prepared as described previously (Takagi et al., 2006;Kawakatsu et al., 2009;Yasuda et al., 2009). Globulins were extracted with globulin extraction buffer (0.5 M NaCl and 10 mM Tris-HCl [pH 6.8]) containing Complete Mini Protease Inhibitor (Roche) at 4°C. Relative accumulation levels of chaperone proteins were calculated from the intensities of immunoblot bands on x-ray films using ImageJ software (http://rsbweb.nih.gov/ij/).

RNA Extraction and qRT-PCR
Total RNA was extracted as described previously (Takaiwa et al., 1987). After RNase-free DNase I treatment (Takara), cDNA was synthesized using the SuperScript III First-Strand Synthesis System for qRT-PCR (Invitrogen). qRT-PCR was performed in a volume of 20 mL using the SYBR Premix Ex TaqII (Perfect Real Time) kit (Takara) on an ABI Prism 7000 HT Sequence Detector (Applied Biosystems). Triplicate reactions were performed following the manufacturer's protocol. The expression levels were normalized to 17S RNA using the expression levels of the wild type at 10 DAF as the reference. The primers used for qRT-PCR are listed in Supplemental Table S3.

Measurement of Total Starch and Nitrogen
Total starch content was measured using a total starch assay kit (Megazyme) following the manufacturer's protocol. The total nitrogen content was measured by the Improved Dumas Method using a LECO528 nitrogen/ protein analyzer (LECO).

Measurement of Amino Acid Composition
Mature seeds (more than 500 grains of each homozygote) were ground into powder (bulked rice flour) with a mill. For total amino acid analysis, 2 mg of bulked rice flour was hydrolyzed with 1 mL of 6 N HCl under vacuum at 110°C for 22 h, then aliquots were dried down and dissolved in 0.1 N HCl. For total Trp analysis, base hydrolysis was performed prior to o-phthalaldehyde (OPA) derivatization. Samples (5 mg) were hydrolyzed with 1 mL of 5 N NaOH under vacuum at 110°C for 22 h, then neutralized with 6 N HCl and 0.1 N HCl. For total CY2 and Met analysis, 2 mg of bulked rice flour was treated with performic acid at 4°C overnight, prior to HCl hydrolysis. Free amino acids were extracted from 200 mg of bulked rice flour with 2 mL of 5% (w/v) trichloroacetic acid at 4°C overnight. Samples were filtered and derivatized with OPA and 9-fluorenylmethyl chloroformate. OPA-and 9-fluorenylmethyl chloroformate-derivatized samples were injected onto a reverse-phase HPLC column (ZORBAX Eclipse-AAA, 4.6 3 150 mm, 3.5 mm; Agilent Technology) controlled by an Agilent 1100 HPLC system equipped with a fluorescence detector.

Immunocytochemistry
Developing seeds at 20 DAF were sampled for immunohistochemical analysis as described previously. Anti-OsTIP3 antibody, a kind gift from Dr. Takehiro Masumura, was used as a primary antibody for staining PSVs at a 1:1,000 dilution (Takahashi et al., 2004). Alexa488-conjugated goat anti-rabbit IgG (Invitrogen) was used at a 1:500 dilution as the secondary antibody. Rhodamine B was used to stain PBs. The specimens were observed with a confocal laser scanning microscope (Radiance 2000; Bio-Rad) using GFP and rhodamine filter sets.

Transmission Electron Microscopy
Immature 15-DAF seeds were fixed in 4% paraformaldehyde, 0.2% glutaraldehyde, and 50 mM sodium phosphate buffer (pH 7.2) overnight at 4°C. After washing in sodium phosphate buffer, the samples were dehydrated in a graded ethanol series and embedded in LR White resin (London Resin). Ultrathin sections, prepared using an ultramicrotome and mounted on copper grids, were stained with uranyl acetate and lead citrate. Sections were observed with a transmission electron microscope (H-7100; Hitachi) at an accelerating voltage of 75 kV.

Supplemental Data
The following materials are available in the online version of this article.
Supplemental Figure S1. Enlarged view of glutelin acidic subunits on a SDS-PAGE gel.
Supplemental Figure S2. Embryo globulin accumulation in wild-type and transgenic plants.
Supplemental Figure S3. Embryo proteins and endosperm proteins in wild-type and transgenic plants.
Supplemental Figure S4. Transmission electron microscopy of the subaleurone layer cell in wild-type and transgenic plants at low magnification.
Supplemental Table S1. Total amino acid content of mature rice grains.
Supplemental Table S2. Free amino acid content of mature rice grains.