To investigate the role of subunit assembly in the intracellular deposition of multimeric recombinant proteins, we expressed a partially humanized secretory immunoglobulin in rice endosperm cells and determined the subcellular locations of the assembled protein and its individual components. Transgenic rice plants expressing either individual subunits or all the subunits of the antibody were generated by particle bombardment, and protein localization was determined by immunoelectron microscopy. Assembly of the antibody was confirmed by immunoassay and coimmunoprecipitation. Immunolocalization experiments showed no evidence for secretion of the antibody or any of its components to the apoplast. Rather, the nonassembled light chain, heavy chain and secretory component accumulated predominantly within endoplasmic reticulum-derived protein bodies, while the assembled antibody, with antigen-binding function, accumulated specifically in protein storage vacuoles. These results show that the destination of a complex recombinant protein within the plant cell is influenced by its state of assembly.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Subcellular localization has a significant influence over the structure, stability and yield of recombinant proteins expressed in plants. Particularly in view of recent data indicating the surprising deposition pattern of a recombinant LTB (Escherichia coli heat-labile enterotoxin B) in maize (Chikwamba et al., 2003), the need to confirm the actual intracellular location of recombinant proteins has been highlighted (Hood, 2004). This is particularly true for multimeric proteins, which need to be folded and assembled before they can achieve their native state. A widely used strategy is to target the components of such proteins to the endomembrane system, which is a rich source of molecular chaperones. Targeting is achieved using transit signal peptides, which facilitate the import of the recombinant polypeptides into the endoplasmic reticulum (ER). Here, they fold, acquire N-linked glycans and assemble into the multimeric protein. It is generally believed that, in the absence of any additional sorting information, proteins in the endomembrane system are secreted by default to the apoplast (Denecke et al., 1990). However, where further sorting signals are present, the protein may be directed to alternative intracellular destinations, such as the vacuole. In some cases, the signals comprise short generic peptide tags that bind to specific receptors, such as the KDEL and HDEL tags that ensure the retrieval of many ER-resident proteins (Denecke et al., 1992). In other cases, the signals are more elaborate and consist of structural information rather than a peptide sequence (Kirsch et al., 1996; Neuhaus, 1996). Intracellular transport is also influenced by protein conformation, oligomeric state and the structure of any attached carbohydrates (Ellgaard and Helenius, 2003).
Protein trafficking within the secretory pathway is particularly noticeable in the cells of storage organs, where protein storage organelles are generated from the endomembrane system. The cereal endosperm has evolved as a dedicated storage tissue and contains storage proteins of different solubilities, i.e. the salt-soluble globulins and alcohol-soluble prolamins. These two types of storage protein accumulate in separate phases in most cereals (Shewry, 1995). In rice, the separation is represented by the formation of two very distinct types of protein storage organelle: protein bodies (PBs), which contain mainly prolamins and are formed within the ER, and protein storage vacuoles (PSVs), which contain mainly glutelins (Krishnan et al., 1986).
Prolamin mRNA has been detected on distinct ER membranes, mainly on the surface of prolamin PBs, indicating that storage protein mRNAs are targeted to and translated at specific subdomains of the ER (Li et al., 1993; Hamada et al., 2003).
Rice prolamins are retained in the ER even though they lack the typical H/KDEL ER-retrieval signal. It has been suggested that retention is achieved by binding to the ER-resident molecular chaperone BiP, which facilitates the folding and assembly of prolamins into insoluble aggregates (Li et al., 1993). The assembly process itself may also contribute to ER retention (Shewry, 1995). In the case of maize prolamins (zeins), an N-terminal sequence has been identified that is responsible for ER retention (Geli et al., 1994).
Glutelins, on the other hand, are homologous to the 11S globulins of legumes, and are transported from the cisternal ER through the Golgi to the PSV, where they are processed and accumulate as large macromolecular complexes (Katsube et al., 1999).
It has been recognized that plant cells are capable of assembling complex mammalian molecules, such as immunoglobulin G (IgG) (Hiatt et al., 1989), IgM (Baum et al., 1996) and secretory IgA (sIgA) (Ma et al., 1995). The synthesis of functional antibody molecules requires the assembly of several polypeptide chains, a process that occurs with high efficiency in the ER of tobacco cells (Frigerio et al., 2000; Hadlington et al., 2003). For sIgA, 10 polypeptide chains (comprising four distinct types of polypeptide) need to be assembled and linked by disulphide bonds. In mammalian systems, this assembly is a two-step process involving two distinct cell types. First, polymeric IgA (pIgA) is generated by the assembly of the heavy chains (HC), light chains (LC) and joining chain (JC). pIgA is then secreted into the blood by plasma cells, captured by the pIg receptor at the basal site of epithelial cells and transcytosed to the apical site. There the pIg receptor is proteolytically cleaved and the ectoplasmic domain constitutes the secretory component (SC), which stays attached to the pIgA. In transformed plants, where the processed form corresponding to mature SC is expressed, the entire assembly process takes place within a single cell.
We have investigated the fate of this multisubunit protein in the storage cells of rice endosperm. As antibodies are heteromultimeric proteins, individual components can be detected using standard immunological methods, whilst antigen binding is a clear and sensitive indicator for the assembled state of HC and LC. Taking advantage of these properties, we used transgenic rice lines expressing either individual subunits or all four subunits of a chimeric secretory immunoglobulin (sIgA/G) to determine whether the four subunits could assemble within rice endosperm cells, and whether the process of assembly influenced the subcellular localization of the recombinant protein. We found that nonassembled components of the antibody accumulated mainly within ER-derived PBs, whilst the assembled antibody accumulated specifically in PSVs.
Selection of transgenic rice lines
Transgenic rice plants were produced by particle bombardment using five plasmids (Figure 1), one carrying the selectable marker hpt and the remainder carrying cDNAs encoding the four components of the sIgA/G, i.e. LC, HC, JC and SC. All five plasmids were simultaneously introduced by cobombardment to achieve the expression of assembled Guy's 13 secretory IgA specifically binding to surface adhesin I/II of Streptococcus mutans, a major causative agent of dental caries (Ma et al., 1998).
Enzyme immunoassays were carried out using primary antibodies specific for human IgG or human SC, so that transgenic lines expressing individual components of the sIgA/G complex could be identified. Transgenic lines, such as A418-2 expressing LC, B112a expressing HC and L82d expressing SC, were selected for further analysis because of their relatively high expression levels (Figure 2A–C). No antibody was available for the immunological detection of JC, and so line B111a containing transcripts for the J-chain was identified by Northern blot analysis (data not shown).
The minimal requirement for antigen binding is the association of LC and HC. Transgenic rice lines expressing two, three or all four immunoglobulin components leading to the assembly of functional antibody complexes were selected through enzyme immunoassays, in which S. mutans surface adhesin I/II (the antigen recognized by the assembled antibody) was used as the capturing agent. Assembled moieties consisting of HC and LC were detected with specific antisera to human LC and HC (Figure 3A,B). Detection of sIgA/G, i.e. the fully assembled antibody comprising all four immunoglobulin components, was achieved with antiserum specific to human SC (Figure 3C). The detection of SC in the functional antibody complex is indirect proof that JC is also present, as this is a prerequisite for the recruitment of SC to the complex (Mestecky and McGhee, 1987; Ma et al., 1995). From a population of 64 transgenic rice lines bombarded with the five separate plasmids, 12 lines (19%) expressed all four immunoglobulin components in addition to the selectable marker. None of the lines expressing single immunoglobulin components showed cross-reaction with the antigen (Figure 3A–C).
The presence of JC in lines B48d (where the three other components were detected with antibodies) and A419-2 (where HC and LC, but not SC, were present) was also inferred by Northern blot analysis (data not shown). Thus, we concluded that line A419-2 probably contained dimeric IgA/G and line B48d contained the entire sIgA/G. Line B48d was the highest expressing sIgA/G line and was selected for further immunolocalization experiments. Lines B48d and A419-2 were employed to confirm the assembly of functional antibodies by coimmunoprecipitation using an antihuman alpha-chain antibody coupled to agarose beads. Subsequent immunoblot analysis confirmed the coprecipitation of intact LC in lines B48d and A419-2 (Figure 4A,B) and SC in line B48d (Figure 4C), whereas, in line B112a, only HC was detected. Line L82d, expressing SC alone, was used as a negative control.
Ultrastructure of rice endosperm cells
LR (London Resin) white resin was used to embed the specimens for ultrastructural analysis and immunoelectron microscopy. Although compromising the structural integrity of the sections to a certain extent, the use of this resin was necessary to optimize antibody–antigen recognition. Immature rice endosperm cells were filled with a dense cytoplasm rich in ribosomes, starch granules and organelles. The organelles included prolamin-containing PBs derived from the ER, which appear round and pale, and glutelin-containing PSVs, which appear irregular and electron dense (Krishnan et al., 1986). We have previously employed antibodies against prolamin and glutelin to confirm the morphological criteria used to distinguish between the two types of protein storage organelle (Torres et al., 2001).
Localization of individually and simultaneously expressed immunoglobulin components
To determine the subcellular distribution of single immunoglobulin chains in the endosperm of transgenic seeds, kappa LC, gamma/alpha HC and human SC were localized within the cellular ultrastructure using specific primary antisera and secondary antibodies conjugated to 10-nm gold particles. The experiments were performed using ultrathin sections of immature seeds from transgenic lines A418-2, B112a and L82d (which expressed LC, HC and SC alone, respectively), and from transgenic line B48d, where all four components were present in the same cell and at least a proportion of the antibody was assembled into sIgA/G (Figures 3C and 4). To determine whether the localization of the immunoglobulin components was affected by protein assembly, the relative intracellular distribution of the components was determined and compared.
None of the immunoglobulin components was detected in the apoplast (in any of the transgenic lines). Instead, we observed intracellular retention within both ER-derived PBs and PSVs. However, the relative levels of the recombinant proteins in these different protein storage organelles varied significantly. Abundant 10-nm gold particles were observed on both PBs and PSVs when LC was detected in line A418-2 (Figure 5A). In line B48d, labelling was also evident on both PBs and PSVs, with a slightly increased relative abundance of gold particles in PSVs (Figure 5B).
When HC was detected in line B112a, gold particles were observed predominantly in ER-derived PBs, with little labelling on PSVs (Figure 6A). In contrast, in line B48d, a significant proportion of the labelling for HC was found within PSVs (Figure 6B).
A similar situation was found for SC. When expressed individually, SC was predominantly sequestered in PBs, with little labelling on PSVs (Figure 7A). In contrast, in line B48d, a significant proportion of the labelling for SC was found within PSVs (Figure 7B). In all cases, no background was observed on sections from wild-type plants (data not shown).
In order to measure the distribution of single immunoglobulin components in the protein organelles, we carried out a quantitative analysis of the gold labelling. The density (number of gold particles/µm2) was calculated in PBs and PSVs for wild-type plants and transgenic lines A418-2, B112a, L82d and B48d, using all the different antibody probes. The results, expressed as the relative percentage labelling on PBs and PSVs, are presented in Table 1. In lines expressing the antibody components individually, HC and SC showed a more striking segregation than LC to PBs (87% and 91% for HC and SC, respectively, vs. 62% for LC). In contrast, in line B48d, where all four components were present in the same cell, a significant proportion of the labelling for all antibody chains was found within PSVs. The relative amount of HC-specific label within PSVs shifted from 13% to 40% when all components of the antibody were expressed, and the relative amount of SC-specific label in PSVs shifted from 9% to 30% (Table 1). The relative distribution of LC appeared to be less affected by the presence of the other immunoglobulin components, as indicated by a shift from 38% to 52% (Table 1).
Table 1. Relative distribution (%) of gold labelling within protein bodies (PBs) and protein storage vacuoles (PSVs) of transgenic rice endosperm cells
Relative labelling (%) in ER-derived PBs
Relative labelling (%) in PSVs
ER, endoplasmic reticulum; HC, heavy chain; JC, joining chain; LC, light chain; SC, secretory component.
Light chain detection
Line A418-2 (LC)
Line B48d (LC, HC, JC, SC)
Heavy chain detection
Line B112a (HC)
Line B48d (LC, HC, JC, SC)
Secretory component detection
Line L82d (SC)
Line B48d (LC, HC, JC, SC)
Localization of the assembled antibody
The change in the distribution of HC and SC observed when lines B112a and L82d (individual components) and line B48d (all four components) were compared appeared to be related to the potential for antibody assembly. This suggested either that HC and SC were transported more efficiently to PSVs as part of an assembled complex, or that they had a greater stability within PSVs once assembled. We investigated the subcellular compartmentalization of the functional IgA/G within endosperm cells using the antigen for immunodetection. This strategy discriminates between unassembled (non-functional) antibody components and assembled (functional) antibody. We incubated ultrathin sections of wild-type seeds and seeds from transgenic lines B48d (sIgA/G), B112a (HC) and L82d (SC) with a tagged antigen epitope (S. mutans surface adhesin I/II). Intense and specific labelling was found on PSVs in line B48d, whereas other cellular organelles, including the ER-derived PBs, were devoid of gold particles. No significant background was observed on sections from wild-type plants (Figure 8A) or transgenic lines B112a or L82d expressing individual components (data not shown). Using the antigen for detection clearly demonstrated that the functional fraction of the antibody was restricted to PSVs (Figure 8B).
We investigated the subcellular localization of a complex multimeric protein and its individual subunits in the endosperm cells of transgenic rice seeds. This was achieved by producing transgenic plants expressing individual components of the antibody and plants expressing all four polypeptide chains. Using immunoassays and coimmunoprecipitation, we confirmed that the antibody could assemble in rice endosperm cells, and we demonstrated that the recombinant multisubunit antibody was functional in that it retained its ability to bind to the surface adhesin of S. mutans.
We studied the localization of HC, LC and SC using transgenic lines expressing these individual components alone. Similar experiments with JC were not carried out because a suitable antibody was not available. Our results showed that HC and SC were concentrated mainly in the ER-derived PBs (where prolamins are usually deposited), i.e. these subunits accumulated in an ER-derived structure. Only a minor portion of the label (approximately 10%) was observed in PSVs (the site for glutelin accumulation). We have previously reported similar results for a recombinant scFv antibody fragment that was intentionally retained in the ER using an added C-terminal KDEL sequence (Torres et al., 2001). We suggested that the presence of a small amount of the antibody in PSVs represented inefficient retention and leakage through the Golgi network, or perhaps an alternative route from the ER to the PSVs. Similarly, Frigerio et al. (2001) have reported that a minor fraction of phaseolin with an added KDEL sequence is transported to the vacuole without passing through the Golgi network. They speculated that this could be a common route for the slow turnover of ER-resident proteins.
It is important to note, however, that none of the antibody chains used in the present study contained a recognized signal for ER retention. Although it is conceivable that the polypeptides were retained by passive trapping within prolamin aggregates, this has not been reported for other heterologous proteins, such as bean phaseolin, which was mainly localized in glutelin PSVs when expressed in transgenic rice seeds (Zheng et al., 1995). An assembly-defective mutant form of phaseolin was transiently retained in the ER of tobacco leaf cells, where it was shown to interact with BiP before being degraded by the quality control machinery (Pedrazzini et al., 1997). In tobacco protoplasts, the unassembled HCs of a murine IgG were also retained in the ER and an interaction with BiP was demonstrated (Nuttall et al., 2002). In lima bean, phaseolin forms aggregates in the ER because of misfolding when glycosylation is prevented by tunicamycin (Sparvoli et al., 2000). These examples show that the structural features of a polypeptide and protein interactions can play key roles in intracellular transport. A similar situation is found in mammalian cells where, in the absence of LC, the CH1 domain of HC remains unfolded and therefore acts as a substrate for BiP. Subsequent binding to LC promotes the folding of CH1 and dissociation from BiP (Lee et al., 1999; Vanhove et al., 2001).
Like HC, free SC is also retained in the ER of rice endosperm cells and accumulates predominantly in PBs. Various mechanisms for the ER retention of newly synthesized proteins have been documented, which may favour interactions with chaperones or membranes within the ER. These include features such as hydrophobicity, non-native conformation and the presence of exposed core peptide segments (Ellgaard and Helenius, 2001, 2003). In this respect, it is important to realize that SC constitutes only the ectoplasmic domain of pIgR and normally does not pass through the ER of mammalian cells in a free and soluble form. Instead, it is synthesized and secreted by epithelial cells as a membrane-bound precursor (pIgR), which undergoes maturation on assembly with pIgA during vesicular transport across the epithelial membrane (Mestecky and McGhee, 1987). In sIgA-producing myeloma and CHO cells transfected with cDNA encoding SC, optimal secretion of SC was observed only after intracellular binding to dIgA (Chintalacharuvu and Morrison, 1997; Berdoz et al., 1999).
Unlike HC and SC, LC was also detected in the storage vacuole, such that only approximately 50% of the label was located in ER-derived PBs and the remainder in PSVs. The absence of any label in the apoplast is somewhat surprising, as there are no specific targeting signals on LC and exit from the ER would be expected to result ultimately in secretion (Denecke et al., 1990). In mammalian cells, most LCs homodimerize in the absence of HC and are secreted in this form (Leitzgen et al., 1997). Hadlington et al. (2003) reported that LC monomers were efficiently secreted from tobacco protoplasts. In contrast, we did not find evidence for the secretion of LC to the apoplast of rice endosperm cells. Although we cannot rule out the possibility that proteolytic activity in the apoplast of endosperm cells may prevent protein accumulation in this compartment, this still does not alter the fact that a major proportion of LC is present in both ER-derived PBs and PSVs.
In line B48d, expressing all four components of the antibody simultaneously, the relative distribution of HC and SC in PBs and PSVs was clearly different from that in lines expressing individual polypeptides. One possible explanation is that the assembled antibody is more stable than the free components, and is therefore able to accumulate in PSVs. In contrast, free HC and SC would be degraded rapidly in PSVs and would not be detectable. Although we cannot exclude differential stability as a possible explanation, our results are also consistent with a model in which unassembled antibody chains are retained in the ER and do not travel further along the secretory pathway until assembly is achieved. Thus, unassembled chains would remain in the ER for a prolonged period of time and become incorporated into prolamin PBs, whereas assembly would promote their efficient transport to PSVs. The assembly of sIgA/G subunits was not particularly efficient in rice endosperm cells, as indicated by the presence of unassembled antibody chains (HC, LC, SC) in prolamin PBs of line B48d. Therefore, although the assembled antibody was detected using the functional antigen-binding assay, both HC and LC were also present as individual polypeptides in this line. The hybrid IgA/G HC used is an engineered molecule with no natural counterpart. We cannot exclude the possibility that this might affect assembly. However, the junction between the gamma and alpha moieties of HC is far away from the CH1 domain and is unlikely to affect interactions with LC. Indeed, efficient assembly has been demonstrated in tobacco (Hadlington et al., 2003).
The in situ antigen-binding assay provided conclusive evidence that the pool of assembled functional antibody is almost exclusively located in glutelin PSVs. This is consistent with the assumption that, following assembly, export from the ER is highly efficient. A number of published examples support the general conclusion that unfolded or uncomplexed polypeptide chains with exposed hydrophobic residues or free cysteine residues remain in the ER until a suitable interacting partner is found (Ellgaard and Helenius, 2003). Assembly itself may facilitate the completion of folding and this may trigger the release of chaperones, allowing the mature oligomer to leave the ER, as has been shown for HC (Lee et al., 1999). Similar events appear to occur for endogenous plant proteins. For example, trimer formation is a prerequisite for the export of pea legumin and bean phaseolin from the ER to the vacuole (Jung et al., 1997; Frigerio et al., 2001).
It remains to be determined why the assembled chimeric sIgA/G accumulates in PSVs and does not reach the apoplast, as would be expected. Frigerio et al. (2000) found that the murine version of the same IgA/G was partly retained in the ER of tobacco leaf protoplasts. Furthermore, a proportion of the protein complex was diverted from the secretory pathway and delivered to the central vacuole by means of a cryptic sorting signal present in the tailpiece of the alpha domains present in the hybrid HC (Hadlington et al., 2003). In the present study, such an assumption would have to include sequences in the chimeric LC to explain its delivery to PSVs. Furthermore, there appears to be a difference between tobacco leaf protoplasts and rice endosperm cells in terms of sIgA/G localization. In tobacco, a proportion (which was implied to be the functional proportion) of the sIgA/G and monomeric IgA/G pools was secreted, while degradation products accumulated in the lytic vacuole (Frigerio et al., 2000). In rice endosperm, we cannot exclude the possibility that part of the immunoreactive material in glutelin PSVs represents degradation products with intact antigen-binding sites. However, we could not identify any signal in the apoplast of rice endosperm that could be attributed to the functional and intact antibody fraction detected using immunoblot assays and enzyme-linked immunosorbent assays (ELISAs). Taken together, our observations may imply a plant- or tissue-specific difference, reflecting a trend for deposition into PSVs in endosperm cells.
In summary, our data indicate the following: (i) the assembly of all four components of the sIgA/G molecule takes place within the rice endosperm; (ii) the destination of the components is affected by their state of assembly; (iii) neither the assembled antibody nor any of its components follows the default pathway (secretion to the apoplast); and (iv) the localization of the antibody and its components is distinct in rice endosperm and tobacco leaves, suggesting that the intracellular trafficking of recombinant multimeric proteins may be different in different expression hosts and tissues. Such issues may have far-reaching implications in the choice of expression hosts and strategies for molecular farming, particularly for the production of multimeric antibodies.
Vectors and plant transformation
Plasmids containing cDNAs encoding the components of a partially humanized sIgA/G specific for S. mutans surface antigen I/II were introduced into mature rice embryos (Oryza sativa L. cv. Bengal) by particle bombardment. The components comprised: (i) a partially humanized chimeric LC; (ii) a partially humanized chimeric HC; (iii) a human JC (GenBank accession number NM_144646); and (iv) the processed form of the human SC (GenBank accession number NM_002644), kindly provided by Planet Biotechnology (Hayward, CA). The LC contained the murine variable domain and the human kappa constant domain. The HC consisted of a murine variable domain, CH1 and CH2 from human IgG1, and CH2 and CH3 from human IgA1. The cDNAs were cointroduced into rice with the hpt marker encoding hygromycin phosphotransferase (Sudhakar et al., 1998). In all cases, transformation constructs were designed such that the recombinant immunoglobulin components were expressed under the control of the constitutive maize ubiquitin-1 promoter (with first intron), and contained either their native human N-terminal signal peptide (JC, SC) or the native N-terminal signal peptide from the murine immunoglobulin heavy chain cDNA (LC, HC) to target the gene products to the endomembrane system.
Protein extraction and enzyme immunoassays
Rice seeds (10 g) were crushed to a fine powder and mixed 1 : 1 (w/v) with protein extraction buffer 200 mm Tris-HCl, pH 7.5, 5 mm ethylenediaminetetraacetic acid (EDTA), pH 8.0, 0.1% Tween-20, 0.002% sodium azide, 10 µg/mL leupeptin]. Insoluble debris was removed by two rounds of centrifugation at 4 °C.
For the detection of LC, Maxisorp 96-well microplates (Greiner, Germany) were coated with 50 µL of goat antihuman IgG-Fab fragment (Fab-specific) antiserum [Sigma, Poole, UK, diluted to 5 µg/mL in Tris-buffered saline (TBS), pH 8.0] for 2 h at 37 °C, washed with TBS containing 0.1% Tween-20 and blocked with 5% dry milk in TBS for 1 h at room temperature, 200 µL per well. After washing, the plates were incubated with a twofold dilution (from neat) of rice seed extracts in TBS (pH 8.0) for 2 h at 37 °C and washed. Plates were incubated with 50 µL per well of alkaline phosphatase (AP)-conjugated goat antihuman kappa LC (Sigma) antiserum, diluted 1 : 10 000 in TBS (pH 8.0), for 2 h at 37 °C. To detect AP activity, para-nitrophenylphosphate (Sigma) was dissolved to 1 mg/mL in substrate buffer (1.05% diethanolamine, 0.02% MgCl2·6H2O, pH 9.8) and 50 µL was added to every well. Plates were read at 405 nm after at least 30 min of incubation. Measurements from three independent replicates were combined to define a mean value.
For the detection of HC, sheep antihuman IgG1 antiserum (Biogenesis, Poole, UK), diluted to 5 µg/mL in TBS (pH 8.0), was added to microplates (50 µL per well) for 2 h at 37 °C, and washed and blocked as above. Following the incubation of rice seed extracts, AP-conjugated goat antihuman IgA (alpha-chain-specific) antiserum (Sigma), diluted to 1 : 1000 in TBS (pH 8.0), was used to detect the expression of HC.
For the detection of SC, goat antihuman SC antiserum (Sigma), diluted to 5 µg/mL in TBS (pH 8.0), was used to coat microplates (50 µL per well) for 2 h at 37 °C, and washed and blocked as above. Following incubation with rice seed extracts, mouse monoclonal antihuman SC (Sigma) was used as the primary antiserum and goat anti-mouse IgG (H + L)-horseradish peroxidase (The Binding Site, Birmingham, UK) was used as the secondary antiserum, both diluted to 1 : 1000 in TBS (pH 8.0). Peroxidase activity was detected by the addition of 50 µL per well of tetramethylbenzidine dihydrochloride (Sigma) dissolved to 0.1 mg/mL in substrate buffer (citrate-phosphate buffer with 0.004% hydrogen peroxide, pH 5.0). After 10 min at room temperature, the enzymatic reaction was stopped with 50 µL per well of 2 m H2SO4 and the optical density (OD) was determined at 450 nm.
For the detection of the assembled IgA/G, Maxisorp 96-well microplates (Greiner, Germany) were coated with 50 µL of a soluble whole cell extract from recombinant E. coli expressing an E-tagged fragment (residues 39–983) of S. mutans surface antigen I/II, diluted to 1 : 1000 in phosphate-buffered saline (PBS), pH 7.4, overnight at 4 °C. The plates were washed with distilled H2O and blocked with 200 µL per well of PBS containing 2.5% bovine serum albumin (BSA) for 2 h at 37 °C. Plates were then washed with distilled H2O containing 0.1% Tween-20, air-dried and incubated with a twofold dilution (from neat) of rice seed extract in PBS containing 2.5% BSA and 0.1% Tween-20 for 2 h at 37 °C. Finally, the plates were washed with distilled H2O containing 0.1% Tween-20. Plates were incubated with 50 µL per well of the appropriate primary antiserum, diluted to 1 : 1000 in PBS containing 2.5% BSA and 0.1% Tween-20, for 2 h at 37 °C, and then washed. The primary antisera comprised sheep/goat antihuman kappa LC (The Binding Site, Birmingham, UK), goat antihuman alpha HC (Sigma) and goat antihuman SC (The Binding Site, Birmingham, UK). A donkey antisheep/goat IgG-AP conjugate (Stratech Scientific Ltd, Luton, UK) was then used as a secondary antiserum for 2 h at 37 °C, and then washed. AP activity was detected as above. Functional sIgA/G was detected with goat antihuman secretory component antiserum.
One millilitre of plant extract was mixed with 50 µL of agarose beads coated with goat antihuman alpha-chain antibody and incubated on a shaker for 1 h. The agarose beads were washed three times and loaded on to a 10% reducing polyacrylamide gel. After blotting on to nitrocellulose membranes, the detection of immunoselected proteins was carried out according to standard procedures (Sambrook et al., 1989).
Specimen processing for electron microscopy
Immature seeds (14 days after fertilization) from wild-type and transgenic rice plants were cut into small pieces with a razor blade under PBS (pH 7.4). The samples were fixed overnight in 4% formaldehyde in PEM buffer [50 mm piperazine-1,4-bis(2-ethanesulphonic acid) (PIPES)/KOH, pH 6.9, 5 mm ethyleneglycol-bis(aminoethyl)-tetraacetic acid (EGTA), 5 mm MgSO4, pH 7.4] at 4 °C, washed in PBS and subsequently immersed in 30% ethanol at 4 °C for 1 h. They were then transferred to a cold box (Thor Industrial, Cryogenics, Oxford, UK) for further processing at −20 °C, where the specimens were dehydrated through an ethanol series (50%, 70% and 100%, for 1 h each) and infiltrated in LR white resin (London Resin Company Ltd, Reading, Berkshire, UK), containing 0.5% benzoin methyl ether as catalyst, in a series 1 : 1, 1 : 2, 1 : 3 ethanol : resin, for 1 h each, and 100% resin overnight. The specimens were transferred to precooled Beem capsules (Agar Scientific, Stanstead, UK), filled with freshly made resin, and allowed to polymerize under indirect UV light at −20 °C for 24 h. Polymerization was completed for 16 h under indirect UV light at room temperature. Finally, 100-nm sections were prepared using a Leica Ultracut E ultramicrotome and collected on 4% pyroxylin and carbon-coated 200 mesh grids.
Immunogold labelling of immunoglobulin components
Grids carrying ultrathin sections were floated on drops of PBS and 3% BSA in PBS. Sections were incubated with adsorbed primary antisera (see next section) diluted in PBS : goat antihuman kappa LC (The Binding Site, Birmingham, UK; 1 : 300), PBS : goat antihuman alpha HC (Sigma; 1 : 300) and PBS : goat antihuman SC (The Binding Site, Birmingham, UK; 1 : 100) for 1 h at room temperature. After washing with PBS, the sections were incubated with a secondary rabbit antigoat antisera conjugated to 10-nm gold particles (BioCell, Cardiff, UK), diluted 1 : 25 in PBS and incubated at room temperature for 45 min. Finally, the grids were washed with PBS and water, air-dried and stained with an aqueous 2% uranyl acetate solution for 30 min. Electron microscopy was carried out on a TEM Jeol 1200 running at 80 kV.
Antisera adsorption to prevent nonspecific binding
Rice seeds from wild-type plants were ground in a coffee grinder and mixed with an excess of cold acetone at 4 °C. The solution was then filtered through Whatman paper and cold acetone was added to the powder until it was white in colour. Air-dried powder was then stored at −20 °C. Primary antisera were diluted to 1 : 10 in PBS (pH 7.4), mixed with rice protein powder at a ratio of 1 mL per 100 mg, and incubated for 2 h with gentle stirring at room temperature. Samples were then centrifuged and the supernatant was mixed with more rice protein powder (100 mg), incubated and centrifuged as above. The supernatant was then filter sterilized through a 0.2-µm filter and the adsorbed antisera were divided into aliquots and stored at −20 °C.
Functional antibody immunolocalization
To determine the subcellular compartments containing assembled forms of the sIgA/G antibody, ultrathin sections of seeds from non-transformed and transgenic lines (B48d, B112a, L82d) were incubated with E-tag labelled S. mutans surface antigen, diluted 1 : 5000 in 3% BSA in PBS, for 1 h at room temperature. The sections were then washed in PBS and the antibody was detected with mouse anti-E-tag IgG (Amersham Biosciences, Chalfont St Giles, UK), diluted 1 : 50 in PBS for 1 h at room temperature, followed by goat anti-mouse IgG conjugated with gold particles (BioCell, Cardiff, UK), diluted 1 : 25 in PBS for 45 min at room temperature. Finally, the grids were washed with PBS and water, air-dried and stained with an aqueous 2% uranyl acetate solution for 30 min. Electron microscopy was carried out on a TEM Jeol 1200 running at 80 kV.
Quantitative evaluation of the distribution of immunogold labelling in rice endosperm
For each transgenic line, 10–15 micrographs originating from three independent labelling experiments were used to estimate the density of gold particles (number of particles/µm2) on ER-derived PBs and PSVs. The shapes of these organelles were outlined on transparencies and their areas were estimated by counting the number of squares on a 0.25 µm × 0.25 µm grid. The number of gold particles was hand-counted on the selected areas and the density of gold particles was calculated. The results were expressed as the percentage of total density per labelling experiment and are given in Table 1.
We thank Dr M. Parker for helpful discussions and Drs R. Casey and R. M. Twyman for critical reading of the manuscript.