We have studied the transport of proricin and pro2S albumin to the protein storage vacuoles of developing castor bean (Ricinus communis L.) endosperm. Immunoelectron microscopy and cell fractionation reveal that both proteins travel through the Golgi apparatus and co-localize throughout their route to the storage vacuole. En route to the PSV, the proteins co-localize in large (>200 nm) vesicles, which are likely to represent developing storage vacuoles. We further show that the sequence-specific vacuolar sorting signals of both proricin and pro2SA bind in vitro to proteins that have high sequence similarity to members of the VSR/AtELP/BP-80 vacuolar sorting receptor family, generally associated with clathrin-mediated traffic to the lytic vacuole. The implications of these findings in relation to the current model for protein sorting to storage vacuoles are discussed.
Developing castor bean (Ricinus communis L.) seeds accumulate lipids and storage proteins in their endosperm. The castor bean endosperm is a living tissue which, upon germination, undergoes programmed cell death to ensure full mobilization of the storage material to the cotyledons (Gietl and Schmid, 2001; Schmid et al., 2001). In developing endosperm, the major storage proteins: 7S lectins [namely ricin (RCAII) and R. communis agglutinin (RCAI)], 2S albumins and 11S globulins, accumulate in protein storage vacuoles (PSV, previously named protein bodies or aleurone grains; Tully and Beevers, 1976; Youle and Huang, 1976). The R. communis PSV contains phytin globoids and a single large protein crystalloid within a soluble protein matrix. The insoluble crystalloid is composed of 11S globulin while the protein matrix is a mixture of 7S lectins and 2S albumins (Fukusawa et al., 1988; Tully and Beevers, 1976). Each of these storage proteins is co-translationally translocated into the lumen of the endoplasmic reticulum (ER) and subsequently transported to the PSV. N-glycosylated lectins travel through the Golgi apparatus, where their glycans are modified and, in the case of ricin, specific fucose residues are added (Lord and Harley, 1985).
We have recently studied the intracellular sorting of both castor bean ricin and 2S albumin in heterologous tobacco mesophyll protoplasts (Brown et al., 2003; Frigerio et al., 1998). Ricin is a heterodimeric glycoprotein composed of a catalytic, ribotoxic A chain (RTA) disulphide-linked to a galactose-binding B chain (RTB). It is made as a single polypeptide precursor in which RTA and RTB are joined by a 12 amino acid linker that is removed upon vacuolar delivery. 2S albumin is also synthesized as a single precursor protein, subsequently processed in PSV into two, different heterodimeric storage proteins, each composed of a large and a small subunit (Brown et al., 2003; Irwin et al., 1990). In transfected protoplasts, as in the native tissue, both proteins reach the vacuole, where they are processed to their mature forms (Brown et al., 2003; Frigerio et al., 1998).
We have characterized the vacuolar sorting signals of both the ricin and albumin precursors. For proricin, the signal resides within a 12-residue linker propeptide that lies between the A and the B chain (Frigerio et al., 2001). For proalbumin, the signal is again contained within an internal propeptide (PPII), in this case separating small subunit I from large subunit I (Brown et al., 2003). NMR structural analysis shows that this propeptide resides in a relatively unstructured loop which is exposed on the surface of the protein (Pantoja-Uceda et al., 2003). Both the proricin and proalbumin sorting signals appear to be of the sequence-specific type (ssVSS; Matsuoka and Neuhaus, 1999), which have previously only been associated with Golgi-mediated traffic to the lytic vacuole (LV). ssVSS contain a hydrophobic residue in the context of an NPIR-like sequence which has been shown to interact with receptor proteins of the vacuolar sorting receptor (VSR) family (Ahmed et al., 1997; Kirsch et al., 1994; Paris and Neuhaus, 2002; Shimada et al., 1997). Accordingly, mutation of a single isoleucine (for proricin) or leucine (for proalbumin) residue within these motifs causes the precursors to be secreted by tobacco protoplasts. Both sorting signals are also sufficient to target reporter proteins to the vacuole (Brown et al., 2003; Frigerio et al., 2001; Jolliffe et al., 2003).
A novel transport route for non-glycosylated proteins including 2S albumin has been described in developing castor bean endosperm and pumpkin cotyledons (Hara-Nishimura et al., 1998). In this route, large precursor-accumulating vesicles (PAC) containing proalbumin bud directly from the ER, seemingly bypassing the Golgi complex to fuse with PSV. A working model for targeting in castor bean endosperm has correspondingly been proposed whereby 2S albumin exits the ER in PAC, which subsequently receive Golgi-modified glycoproteins, such as proricin, into their periphery before they reach PSV (Hara-Nishimura et al., 1998). Indeed, two putative receptors of the VSR class, PV72/82, have been isolated from the membranes of PAC (Shimada et al., 1997, 2002). PV72 has been shown to be present in the Golgi (Shimada et al., 2002), and to interact with various sequences derived from pumpkin 2S albumin (Shimada et al., 1997, 2002; Watanabe et al., 2002). These findings were rationalized by ascribing a role to PV72 in salvaging any 2S albumin that escapes aggregation in the ER (Shimada et al., 2002). However, the presence of an ssVSS in castor bean 2S albumin suggests that transport of this protein to PSV may be solely receptor-, and thus Golgi-, mediated. In addition, we have recently provided indirect evidence for transport of proalbumin through the Golgi in tobacco protoplasts (Brown et al., 2003). The disparity of these data prompted us to investigate in more detail the transport of proricin and proalbumin in their native tissue.
In this work we show that both proricin and proalbumin indeed co-localize and travel through the Golgi apparatus. Both proteins subsequently appear in large vesicles which are likely to represent developing storage vacuoles. These are distinct from PAC as their lumina contain proteins carrying Golgi-modified glycans. We further show that the sequence-specific sorting signals of both proricin and proalbumin bind in vitro to a protein that shares homology with the VSR/AtELP/BP-80 vacuolar sorting receptor family, previously associated with the clathrin coated vesicle (CCV)-mediated pathway to LV (Bassham and Raikhel, 2000; Vitale and Raikhel, 1999).
We have recently investigated the vacuolar targeting of both ricin and 2S albumin by transient expression in tobacco mesophyll protoplasts and wanted to further characterize the fates of these proteins in their native tissue. To ascertain whether proricin and proalbumin are transported along the same, Golgi-mediated route to PSV, we performed immunoelectron microscopy on ultrathin sections of high-pressure frozen developing castor bean endosperm.
Ricin and 2S albumin co-localize in the Golgi and PSV
We isolated developing castor bean endosperm at the stage of testa maturation corresponding to the peak of synthesis of all major classes of storage proteins (Lord, 1985b). The endosperm is characterized by regularly shaped cells containing toluidine blue-reactive structures (Figure 1a). Ultrastructural analysis reveals abundant, electron translucent lipid bodies (Figure 1b,c) and a number of more electron-opaque, spherical structures of variable size corresponding to the toluidine blue-reactive structures. These were previously identified as PSV (Figure 1) (Tully and Beevers, 1976; Youle and Huang, 1976). Although both globoids and crystalloids have previously been detected at this stage of development after chemical fixation (Gifford et al., 1982), only the globoids were visible in high pressure frozen sections of the same tissue (Figure 1c, arrowheads).
Based on substantial biochemical evidence, ricin and 2S albumin are classified as storage proteins of the PSV matrix (Irwin et al., 1990; Lord, 1985b; Tully and Beevers, 1976; Youle and Huang, 1976). This was confirmed by immunolabelling endosperm sections with anti-ricin A chain (anti-RTA) and anti-2S albumin (anti-2SA) antisera. Figure 2 (lanes 1 and 2) shows that both rabbit and sheep anti-RTA antisera specifically recognized the proricin precursor (grey arrow) and two glycoforms of ricin A chain (empty arrowheads) (Fulton et al., 1986) from a preparation of total endosperm proteins. However, it has previously been demonstrated that the rabbit anti-2SA antiserum, in addition to recognizing 2S albumin (black arrowhead), also recognizes the ricin A chain glycoforms (empty arrowheads) (Figure 2, lane 3; also Brown et al., 2003). We therefore removed the contaminating antibodies by passage through a column bearing immobilized ricin, generating an antiserum which recognized only 2S albumin (Figure 2, lane 4).
To immunologically assess the identity of PSV in developing castor bean endosperm, we initially labelled sections with antibodies against α-TIP, a well-established marker for the tonoplast of the PSV (Jauh et al., 1999; Johnson et al., 1990). Anti-α-TIP coupled with 10 nm gold particles clearly not only decorated the outer rim of the electron opaque bodies (Figure 3a), but also the Golgi apparatus (Figure 3b). This pattern has been previously observed in developing pea cotyledons (Hillmer et al., 2001). We therefore conclude that the anti-α-TIP-lined organelles are PSV. Both anti-RTA (Figure 3c, 10 nm gold) and anti-2SA antisera (Figure 3d) label the lumen of the electron opaque bodies. Double immunogold labelling using anti-RTA (coupled with 10 nm gold) and anti-α-TIP (coupled with 20 nm gold) confirmed that the matrix of the PSV was the site of ricin and 2S albumin deposition (Figure 3c, arrowheads indicate α-TIP labelling). Most importantly, both ricin and 2S albumin are also found in the Golgi apparatus (Figure 3, panels e and f). Whilst previously reported for ricin, where the acquisition of Golgi-modified glycans is well documented (Lord, 1985a; Lord and Harley, 1985), these findings now directly demonstrate that proalbumin is also transported through the Golgi complex en route to PSV in castor bean endosperm. Indeed, 94% of the Golgi stacks observed in this study (73 of 78) were labelled with anti-2SA. Double immunogold labelling with anti-RTA (10 nm gold) and anti-2SA (20 nm gold) confirmed the co-localization of these two proteins in both the Golgi and PSV (Figure 4a,b, respectively). Background labelling was consistently very low (see Table S1). In addition to reproducibly detecting 2S albumin in the Golgi, we also observed immunogold labelling in the ER (Figure 4c). Strikingly, the ER in developing endosperm displayed a homogeneous, tubular morphology, in which dense aggregates of the type postulated to form PAC vesicles were never observed. Furthermore, 2S albumin labelling within the ER was always uniformly distributed, in contrast to the clusters observed in the Golgi cisternae (Figures 3f and 4a). This is consistent with the sorting event occurring in the Golgi, rather than the ER. Thus, we conclude that in developing castor bean endosperm both proricin and proalbumin are transported from the ER to the Golgi en route to PSV.
Ricin and 2S albumin accumulate in large, post-Golgi vesicular structures
As well as the Golgi and large PSV, the antibodies raised against RTA (coupled with 10 nm gold) and 2S albumin (coupled with 20 nm gold) also label the lumen of large vesicles of 150–400 nm in diameter (Figure 5a). These vesicles possessed a peripheral electron-translucent layer, and were often seen close to PSV [Figure 5b(inset),c]. Ribosomes were never observed around the periphery of these structures, suggesting that they did not originate directly from rough ER. In addition, anti-complex glycan antiserum revealed a distribution indistinguishable from anti-RTA (Figure 5b, 10 nm gold), labelling throughout the lumen of these vesicles (Figure 5b, main panel and inset, 20 nm gold), further indicating that the lumenal cargo has travelled via the Golgi. As these structures appear to have the same distribution of lectins and albumins as PSV, they may therefore represent precursors to storage vacuoles. In further support of this, anti-α-TIP labelled the membrane of these vesicles (Figure 5c). The vesicles were also often found in close proximity to one another and/or to a large storage vacuole, suggesting imminent homotypic fusion. Figure 5(c) shows such an example, where five α-TIP-lined vesicles (black stars) are clustered together near a PSV.
Ricin and 2S albumin co-fractionate with VSR-like proteins
The sequence-specific nature of the vacuolar sorting signals of ricin and 2S albumin suggests that both proteins may be ligands for a sorting receptor(s) of the VSR family (previously named BP-80 or ELP; Ahmed et al., 1997; Kirsch et al., 1994; Shimada et al., 1997). As the interaction between such receptors and clathrin adaptor proteins has been demonstrated in vitro (Sanderfoot et al., 1998), this in turn makes them candidates for recruitment into CCV. As neither the clathrin heavy-chain antiserum nor the anti-BP-80 antisera (MAbs 17F9 or 14G7) (Cao et al., 2000) were compatible with the EM methodology described here (data not shown), we instead investigated the potential co-localization of these proteins by subcellular fractionation and Western blotting.
Developing castor bean endosperm was homogenized and subjected to fractionation on a continuous 15–55% isopycnic sucrose gradient. Fractions were resolved by either reducing or non-reducing SDS-PAGE before immunoblotting with a panel of antisera (Figure 6). The majority of the precursor (Figure 6) and mature (data not shown) forms of ricin and 2S albumin are detected at the top of the gradient. This is expected, as fragile PSV break during homogenization in aqueous buffer. There is also a degree of breakage of other endomembranes, as testified by the fact that anti-BiP labels fractions at the top of the gradient, in addition to the microsome peak. Both the ricin and 2S albumin precursors, however, also peak at a fraction of density 1.21 g ml−1. These fractions are clearly distinct from the ER microsomes, as indicated by the position of the molecular chaperone BiP (Figure 6). Significantly, proteins reacting with anti-BP-80 monoclonal antibody 17F9 also coincide with ricin and 2S albumin. Furthermore, this 1.21 g ml−1 fraction is also labelled by an anti-clathrin heavy chain antibody. We conclude that a proportion of proricin and proalbumin is found in cellular fractions that are enriched in both BP-80-like species and clathrin.
The sorting signals of ricin and 2S albumin bind to VSR-like proteins in vitro
The cell fractionation data raise the possibility that proricin and proalbumin may be ligands of VSR (BP-80)-like receptors. To test this hypothesis, we produced affinity columns carrying immobilized peptides corresponding to the proricin linker and proalbumin PPII, both in their wild-type form or with the respective, crucial isoleucine/leucine changed to glycine (Figure 7a). A potential receptor would be expected to bind to the wild-type, but not the mutated propeptides. As a positive control, a column bearing the proaleurain VSS was also generated, as previously used to select BP-80 from pea (Kirsch et al., 1994). Clarified CHAPS extracts of membranes prepared from developing castor bean endosperm were applied to the columns, washed, and eluted at low pH. Eluates were separated by non-reducing SDS-PAGE and visualized with Coomassie (Figure 7b). A protein migrating below the 64 kDa marker bound to the wild-type, but not the I271G ricin linker peptide column (Figure 7b, empty arrowhead, compare lanes 1 and 2). Likewise, a species of comparable mobility bound to the 2S albumin PPII, but not the L58G peptide column (Figure 7b, compare lanes 4 and 5). A protein with an apparent molecular mass of around 50 kDa was also eluted in every case (Figure 7b, black arrowhead) as well as from a column prepared without coupled peptide (data not shown). This protein, identified by mass spectrometry as calreticulin (data not shown), is a contaminant which binds to the column matrix itself.
Interestingly, the proteins released from the control proaleurain column were indistinguishable from those released by the wild-type ricin linker or 2S albumin PPII columns at the resolution provided by this SDS-PAGE (Figure 7b, compare lanes 1, 3 and 4). These data suggest that a protein from castor bean, shown to bind the targeting signals of these endogenous PSV residents in an isoleucine/leucine-dependent manner, also bound the ssVSS responsible for targeting proaleurain to LV, and thus may be BP-80-like. This possibility was investigated by Western blotting. Figure 7(c) shows immunoblots of gels loaded with column eluates from castor bean endosperm as above, using a primary antibody raised against a non-reduced epitope that spans the ‘unique’ N-terminal domain of pea BP-80 (MAb17F9) (Cao et al., 2000). The lanes loaded with protein eluted from the linker-, PPII- and proaleurain-affinity columns yielded immunoreactive bands (Figure 7c, empty arrowheads), whereas there were no detectable immunoreactive bands in the eluates from either column bearing the mutated propeptides (Figure 7c, lanes 2 and 5). This concurs both with an earlier biochemical demonstration of the inability of these signals to confer vacuolar targeting in tobacco protoplasts (Brown et al., 2003; Frigerio et al., 2001), and with the Coomassie-stained gels presented in Figure 7(b). In order to provide a size marker for BP-80, total castor bean and pea (Pisum sativum) extracts were immunoblotted with the same antibody. Figure 7(d) shows that a single immunoreactive band from pea and three to four bands from castor bean extracts were revealed that migrated within the range of the castor bean proteins eluted from the ricin and albumin VSS-bearing columns. To corroborate our findings, we also performed immunoblots using a different monoclonal antibody (Mab 14G7) that detects the third EGF repeat in pea BP-80 (Cao et al., 2000). The 14G7 antibody recognizes one of the castor bean polypeptides around 64 kDa both in the column eluate (Figure 7e, lane 1) and in total castor bean extracts (Figure 7f). As two different monoclonal antibodies detect bands in the same size range, this confirms that the castor bean species are indeed BP-80-like. In most gel systems, BP-80 proteins resolve with an apparent Mr of approximately 80 kDa. The apparent size discrepancy detected in our experiment is due to our SDS-PAGE system, which is optimized for the separation of storage globulins and uses a 200:1 acrylamide/bisacrylamide ratio (see Experimental procedures for details). Indeed, when total castor bean homogenates were resolved on conventional SDS-PAGE, (acrylamide/bisacrylamide ratio of 37:1), immunoblotting with 17F9 revealed a compact band migrating at around 80 kDa (see Figure S2).
From the data shown in Figure 7 we conclude that the ssVSS of both ricin and 2S albumin can bind BP-80-like receptor(s) in vitro. More than one immunoreactive species is bound to each wild-type peptide column (Figure 7c, empty arrowheads), with the number and immunoreactivity of these species differing between columns. This suggests the presence of multiple VSR isoforms in castor bean endosperm. We focused on the polypeptide that bound to both the ricin linker and 2S albumin PPII peptides and that produced the strongest Coomassie-stained band (Figure 7b, lane 1, empty arrowhead). After subjecting the excised polypeptide to tryptic digestion, the peptide mixture was analysed by nano-electrospray tandem MS. Several multiple charged peptides were fragmented in order to resolve their amino acid sequence. Three peptides presented high sequence similarity to several known members of the VSR family (Table 1). Specifically, one peptide was identical to the Arabidopsis vacuolar sorting receptor AtELP1 (VSRAt-1; Ahmed et al., 1997) and the remaining two peptides showed more than 80% sequence identity with the same protein. The three peptides span the lumenal domain of the receptor (Table 1). Intriguingly a fourth peptide, WGGYDC, with no apparent homology to other known VSR, was also identified.
Table 1. Sequence similarity of three peptides from the major proricin linker-binding protein to members of the vacuolar sorting receptor family
The deposition of both proricin and proalbumin in the PSV would therefore appear to be a receptor-mediated event, in a manner at least analogous to CCV-mediated traffic to LV.
We have systematically analysed the transport of two major storage proteins to the PSV of developing castor bean endosperm. The data presented here not only confirm the predictions we made whilst studying the fate of ricin and albumin in tobacco protoplasts (Brown et al., 2003; Frigerio et al., 1998), but also provide further insights into the sorting of seed storage proteins in the native tissue. Namely, we demonstrate a direct link between the targeting of two storage protein precursors to PSV and the interaction of their sorting signals with receptors of the VSR family.
In this study, developing castor bean endosperm was subjected to high-pressure freezing and freeze-substitution, to prevent any artefactual alterations to structure and protein location. The immunogold labelling presented here clearly shows that castor bean proalbumin is present in the Golgi complex.
Further, EM data reveal that the distribution of 2S albumin – a non-glycosylated protein – and the previously characterized glycoprotein, ricin, is identical using these methods. In addition to observing these storage proteins together in the PSV, co-localization was also seen in smaller structures (150–400 nm, Figure 5). These structures could be labelled with antibodies raised against α-TIP, a PSV marker, and were frequently found close to PSV. We believe that the presence of complex glycan-bearing proteins throughout the lumen of these vesicles is sufficient to distinguish them from PAC. Indeed, the data presented here suggest that the entire content of these large vesicles is derived from the Golgi. Furthermore, throughout our study, ribosomes were never seen on their delimiting membranes, again indicating that these structures did not originate directly from the ER. Moreover, we never observed aggregates within the ER, or indeed any evidence of budding from this compartment. Accordingly, these large vesicles should instead be regarded as early PSV, or possibly pre-vacuolar compartments, although further work is required to clarify their precise role in storage protein deposition.
Our data firmly support the idea that in developing castor bean endosperm, all 2S albumin is trafficked to PSV via the Golgi apparatus. The co-localization of 2S albumin and ricin throughout the secretory pathway indicates that both proteins follow the same route to PSV. Indeed, the evidence presented here formally rationalizes, in the native tissue, our previous demonstration that both proteins possess ssVSS (Brown et al., 2003; Frigerio et al., 2001). A family of VSR has now been characterized, of which pea BP-80 (Kirsch et al., 1994) is the prototype. Members of this family are integral membrane proteins which bind with high specificity to NPIR-like vacuolar sorting signals at the trans-Golgi level and recruit clathrin coats through an adaptin-binding tyrosine motif present in their cytosolic tail (Paris and Neuhaus, 2002). VSR have been shown to exist in pea, maize, rice, pumpkin and Arabidopsis (Paris and Neuhaus, 2002). In accordance with this model, castor bean proteins recognized by antibodies raised against BP-80 and clathrin heavy chain co-localize with the precursors of both ricin and 2S albumin when resolved on a sucrose gradient (Figure 6). Furthermore, the vacuolar sorting signals of both ricin and 2S albumin bind in a sequence-specific manner to a BP-80-like protein. The demonstration of a specific interaction with a sorting receptor, and co-localization of ricin and 2S albumin with this receptor and with a clathrin coat protein, may explain the apparent paradox of two bona fide PSV residents possessing ssVSS.
Although VSR have been previously implicated in the transport of other members of the albumin family (Kirsch et al., 1996; Shimada et al., 1997, 2002; Watanabe et al., 2002), the roles of the peptides used to bind the receptors in vitro were not always investigated fully in vivo. Here we have clearly shown that the well-characterized VSS of two castor bean storage proteins (Brown et al., 2003; Frigerio et al., 2001; Jolliffe et al., 2003) interact with a novel member of the recognized VSR family, thus adding R. communis to the growing list of plants in which these receptors have been identified, and fuelling the debate as to the breadth of their function. Indeed a recent report by Shimada et al. (2003) has shown that genetic knockout of one of the seven Arabidopsis isoforms of VSR, namely VSRat-1, led to partial secretion of the precursors of all major classes of storage proteins into the apoplast of mature Arabidopsis seeds. This new finding further strengthens the link between VSR-mediated transport pathways and the sorting of proteins to the PSV in seeds. Together, these findings potentially call for a re-evaluation of the working model for seed storage protein deposition and for the role of VSR in this process. In this respect it will be important to characterize the transport of the remaining class of storage proteins – the 11S globulins – to the PSV of Ricinus endosperm.
Ricinus communis L. plants were grown from seed in a greenhouse at 15°C and under a 16 h/8 h light/dark cycle. Endosperm tissue was excised from the ripening seeds during testa formation (typically 8 weeks after flowering) at a developmental stage when the lectins and storage proteins are rapidly synthesized (Lord, 1985b).
The following antibodies were used for biochemical analysis and immunoelectron microscopy: mouse monoclonal anti-clathrin heavy chain (BD Biosciences, Oxford, UK), mouse monoclonal anti-BP-80 antibody (MAb17F9, Cao et al., 2000); rabbit polyclonal anti-bean α-TIP (Johnson et al., 1990); rabbit polyclonal anti-tobacco BiP (Pedrazzini et al., 1997), rabbit anti-recombinant ricin A-chain (anti-RTA, Frigerio et al., 1998); sheep anti-recombinant RTA, rabbit polyclonal anti-castor bean 2S albumin (anti-2SA); rabbit polyclonal anti-complex glycans (Lainéet al., 1991).
2S albumin antibody purification
Crude rabbit polyclonal antisera raised against 2S albumin (Brown et al., 2003) was purified to remove contaminating antibodies recognizing ricin agglutinins by passage through a column bearing 1 ml of agarose-bound R. communis agglutinin II (Vector Laboratories, Peterborough, UK). The column was washed with PBS containing 0.1 m lactose, and the antisera applied in 0.1 m lactose to prevent interaction of antibody glycans with the lectin subunit of the immobilized ricin. Flow-through fractions, which contained the anti-2SA antibodies, were collected.
Sucrose gradient fractionation
Endosperm isolated from maturing castor bean seeds was homogenized in an ice-cold mortar with ice-cold 100 mm Tris-Cl, pH 7.8, 10 mm KCl, containing 12% (w/w) sucrose and 2 mm MgCl2, using 6 ml of buffer per gram of fresh leaf tissue. The homogenate was centrifuged for 10 min at 1000 g at 4°C; 600 μl of the supernatant was loaded on a 11-ml linear 15–55% (w/w) sucrose gradient. After centrifugation at 35 000 rpm, 4°C for 2 h in a Beckman SW40 rotor (Beckman Coulter, High Wycombe, UK) (154 400 g average), fractions of 600 μl were collected. Immunoblot analysis of 60 μl aliquots of these fractions was performed as described (Pedrazzini et al., 1997), using the indicated antisera.
Electrophoretic separation of proteins
Proteins were separated by sodium dodecylsulphate (SDS)-polyacrylamide gel electrophoresis using the method described by Laemmli (1970), and using an acrylamide-bisacrylamide ratio in the separating gel of 200:1, which has been found to be optimal for the resolution of certain storage globulins (Bollini and Chrispeels, 1979).
A 2.5 ml volume [2 mg ml−1 in coupling buffer (50 mm Tris; 5 mm EDTA, pH 8.5)] of sulphydryl-containing peptides (synthesized by Alta Biosciences, University of Birmingham, UK) were immobilized on ‘SulfoLink’ coupling gel columns (Pierce, Perbio Science UK, Tattenhall, UK), and the non-specific binding sites blocked, according to the manufacturer's instructions. Endosperm was homogenized in a Petri dish on ice by continuous chopping for 15 min using a hand-held razor blade, after addition of cold grinding buffer (150 mm Tricine, pH 7.5; 1 mm EDTA-Na, pH 7.5; 10 mm KCl; 1 mm MgCl2; 100 mm lactose), supplemented immediately before use with ‘complete’ protease inhibitor cocktail (Roche Diagnostics, Lewes, UK). Pea homogenates were prepared using an ice-cold pestle and mortar by grinding in the same buffer. The resulting suspensions were filtered through four layers of mira cloth, and centrifuged at 10 000 g for 15 min at 4°C. The floating lipid layers were discarded and the underlying supernatants carefully removed to a clean tube. Membranes were pelleted from post-mitochondrial supernatants by centrifugation in 3 ml thick-walled polycarbonate tubes at 100 000 g (Beckman TL-100 ultracentrifuge, TLA-100.3 rotor) at 4°C for 30 min, and the supernatant discarded. Pellets were resuspended in CHAPS buffer (Kirsch et al., 1994: 20 mm Hepes NaOH, pH 7.1; 150 mm NaCl; 1 mm MgCl2; 1 mm CaCl2; 1% (w/v) 3-[(3-cholramidopropyl) dimethylammonio]-1-propanesulphonate (CHAPS)], and incubated at room temperature for 30 min to solubilize integral membrane proteins. Residual membranes were removed by re-centrifugation of the same tubes, as in the first step, and the supernatant was carefully removed to a clean tube. Two-millilitre aliquots of membrane protein extracts were applied to the affinity columns, previously brought to room temperature and equilibrated according to the manufacturer's instructions. Column gel beds were resuspended, and incubated with gentle agitation for 30 min, before being allowed to settle for 30 min. Columns were washed according to the manufacturer's instructions. Bound proteins were released by addition of 10 ml elution buffer [25 mm NaOAc, pH 4.0; 150 mm NaCl; 1 mm EGTA; 1% (w/v) CHAPS], and the eluate collected in 1 ml aliquots. Protein was precipitated from fractions by the addition of an equal volume of cold 30% (w/v) TCA. After mixing, the sample was incubated on ice for a minimum of 30 min. The precipitant was harvested by centrifugation at 17 000 g for 15 min at 4°C, the supernatant discarded, and the pellet washed twice with 1 ml of cold acetone, centrifuging between each wash for 5 min as before. The acetone-washed pellet was then dried in a vacuum desiccator, resuspended in non-reducing loading buffer, and separated by SDS-PAGE. Subsequent protein detection was either by Coomassie-blue staining, or immunoblot as described (Pedrazzini et al., 1997), using monoclonal BP-80 antibody.
The band of interest was excised from a Coomassie-stained gel and digested overnight with trypsin as described (Anelli et al., 2002). The unseparated peptide mixture was concentrated and desalted over a capillary column packed with POROS R2 material and eluted directly into the electrospray needle. Nanoelectrospray tandem MS experiments were performed on a Q-Star Pulsar (PE SCIEX Instruments, Toronto, Canada). Several multiple charged peptides were fragmented in order to read the amino acid sequence. One peptide (FVGDGYTHCK) was identical to AtELP (Accession number sptrembl: P93026). Two other sequenced peptides, namely CIGDTEADVDNPVLK and LNIPSALITK, showed high homology to CIGDPEADVENPVLK and IPSALITK, also present in P93026.
Transmission electron microscopy
Samples for transmission electron microscopy observation were chosen from castor bean endosperm at the testa maturation stage. Thin slices of endosperm (<0.5 mm) were dissected, then 2 mm diameter discs were cut out of the slices using a disposable biopsy punch (Stiefel Laboratories Ltd, Wooburn Green, UK), and placed in an aluminium sample holder. Prior to capping the sample holder, the sample was covered with hexadecene in order to remove any remaining air between the sample and the holder. Pairs of holders were clamped together, and samples were immediately frozen using a BAL-TEC HPM 010 high-pressure freezer (BAL-TEC AG, Balzers, Liechtenstein).
Freeze-substitution was carried out in a Reichert AFS (Leica, Vienna, Austria) freeze-substitution system. Sample holders were split open under liquid nitrogen and placed into plastic porous specimen pots containing the substitution medium previously frozen in liquid nitrogen. For standard ultrastructural observations, 2% osmium tetroxide in acetone dried over molecular sieve was used. Plastic specimen pots were put into an universal aluminium container onto the surface of the frozen substitution medium and transferred into the Reichert AFS pre-cooled to −160°C. Sample temperature was increased to −85°C over 5 h. Freeze-substitution was carried out as follows: −85°C for 26 h, 2°C h−1 temperature increase over 12.5 h, −60°C for 8 h, 2°C h−1 temperature increase over 15 h, −30°C for 9 h (Steinbrecht and Müller, 1987; Studer et al., 2001). At the end of the freeze-substitution run, samples were warmed up to 20°C with a temperature increase of 4°C h−1. All subsequent steps were carried out at room temperature.
For immunogold labelling, samples were subjected to the same treatment except that the freeze-substitution medium was 0.5% uranyl acetate in ethanol dried over molecular sieve and after the freeze substitution run, samples were warmed up to −20°C with a temperature increase of 1°C h−1. All subsequent steps were carried out at this temperature.
Embedding and sectioning
Samples for standard ultrastructural observations were rinsed in anhydrous acetone, carefully removed from the sample holders, embedded stepwise in Spurr's medium resin (TAAB Laboratories Equipment Limited, Aldermaston, Berkshire, UK) (Spurr, 1969), and polymerized at 70°C for 9 h.
Samples for immunogold labelling were rinsed in cold ethanol and, after careful removal from the sample holders, embedded stepwise in LR White medium resin (Agar Scientific, Stansted, UK) over 1 week. Polymerization was under UV light for 24 h at −20°C and for another 24 h at 0°C.
Ultrathin sections were cut with an RMC MTXL ultramicrotome (Boeckeler Instruments, Tucson, AZ, USA) and collected on formvar-coated 300 mesh hexagon copper grids (Agar Scientific).
Immunogold labelling, staining and observation
For immunogold labelling, sections were first blocked for 30 min with a 1:30 dilution of pre-immune serum in TRIS (20 mm TRIS, 15 mm NaN3, 225 mm NaCl), pH6.9, supplemented with 1% BSA (TRIS-BSA). They were then treated with 0.1% Tween20 in TRIS-BSA and 0.02 m glycine in TRIS-BSA, each for 15 min, washed three times for 10 min in TRIS-BSA, and incubated in a dilution of primary antibody at 4°C overnight. The following primary antibodies were used in this study: anti-α-TIP (dilution 1:1000, 1:2000), rabbit anti-RTA (dilution 1:200), sheep anti-RTA (dilution 1:2000, 1:4000), rabbit anti-2SA (dilution 1:2000, 1:8000), rabbit anti-complex glycans (dilution 1:2000, 1:4000).
After washing in TRIS-BSA (three times 15 min), sections were treated with a 1:20 dilution (in TRIS-BSA) of the corresponding secondary antibodies (goat anti-rabbit or donkey anti-sheep) conjugated to 10 nm gold (BBInternational, Cardiff, UK) for 1.5 h at room temperature. Finally, after thorough washing in TRIS-BSA and ultrapure water, sections were dried on filter paper. Occasionally, 1% fish gelatin (Sigma, Dorset, UK) was added to the primary and secondary antibody solutions as well as to the TRIS-BSA washings as an additional blocking agent.
For simultaneous double immunogold labelling, the same protocol was applied except that sections were exposed to mixtures of pre-immune serum (both at a 1:30 dilution in TRIS-BSA) and to mixtures of the primary antibodies (for dilutions see above). Secondary antibodies were also used as a mixture (both at a 1:20 dilution in TRIS-BSA) with one conjugated to 10 nm gold particles and the other one to 20 nm gold particles (see Figures 3–5).
For controls of both single and double immunogold labelling, samples were subjected to the same treatment except that the corresponding pre-immune serum substituted the primary antibody.
All sections were stained with 2% uranyl acetate in 70% ethanol for 15 min, followed by lead citrate for 15 min, and examined with a JEOL JEM-1200EXII (JEOL UK, Hertfordshire, UK) at 80 or 120 kV.
We thank Barry Martin for technical assistance with high-pressure freezing and Angela Cattaneo for tandem MS peptide sequencing. We are grateful to Martin Chrispeels for the gift of the anti-α-TIP antiserum and John Rogers for the 17F9 and 14G7 MAbs. This work was supported by the BBSRC (grants 88/C17404 to LF/LMR, C15728 to CH and a studentship to JCB).
Table S1 Analysis of background immunogold labelling Sets of electron micrograph negatives were digitalized and the areas of Golgi stacks, protein storage vacuoles (PSV), and surrounding cytoplasm were calculated using the ImageJ analysis package (publicly available at http://www.rsb.info.nih.gov/ij/index.html). Labelling of the Golgi and/or PSV was compared with background labelling and is shown as the average number of gold particles per square micrometre ± standard deviation. (*) number of gold particles for α-TIP was calculated for the areas of both Golgi and PSV
Figure S1 Gel mobility of castor bean VSR-like proteins. Total extract from castor bean endosperm was resolved by SDS-PAGE either using either standard (37:1) or 200:1 acrylamide/bisacrylamide ratio and immunoblotted with monoclonal anti-BP-80 17F9 antibody. Immunoreactive bands were visualized by ECL. Numbers on the left indicate molecular mass markers in kilodaltons. Note the gel-type-dependent difference in mobility of the immunoreactive polypeptides.