Sequence-specific, Golgi-dependent vacuolar targeting of castor bean 2S albumin

Authors


For correspondence (fax +44 24 7652 3568; e-mail lroberts@bio.warwick.ac.uk).

Summary

The targeting of the castor bean (Ricinus communis) 2S albumin precursor has been investigated by expressing cDNA in transformed tobacco (Nicotiana tabacum) leaf cells and by following biosynthesis in the native tissue. Correct targeting in both tissues was accompanied by processing of the precursor. Delivery to vacuoles was sensitive to brefeldin A (BFA) treatment in both tissues and to perturbation of COPII function in tobacco, supporting the view that transport through the Golgi is required. The targeting signal for this Golgi-dependent routing lies within the propeptide of the first heterodimer of proalbumin. This propeptide directed a normally secreted reporter protein to the vacuoles of tobacco cells in a Golgi-dependent manner in vivo, unless a critical Leu residue was mutated, supporting the view that a sequence-specific signal was needed to target a seed storage protein to the vacuoles of a vegetative cell.

Introduction

The castor bean 2S albumin precursor is translated as a single pre-proprotein that can be processed into two different heterodimeric storage proteins upon transport and delivery into the protein storage vacuoles (PSV) of castor bean endosperm (Irwin et al., 1990). Pre-proalbumin is converted to proalbumin by removal of an N-terminal signal peptide in the endoplasmic reticulum (ER) lumen, and is then trafficked to the PSV, reportedly by a unique transport pathway that involves the packaging of proalbumin into large vesicles that bud directly from the ER and that subsequently fuse with the PSV (Hara-Nishimura et al., 1998). Arrival in the vacuole is accompanied by cleavage of three propeptides: one at the N-terminus, one from within the first heterodimer, and one from within the second heterodimer together with a cleavage between the two heterodimers (Irwin et al., 1990). In comparison with other members of the albumin family, the castor bean 2S albumin is unique both in its greater size and in generating two distinct heterodimers instead of one.

The majority of seed storage proteins and lectins have been shown to be transported to the PSV via the Golgi complex (Greenwood and Chrispeels, 1985; Hohl et al., 1996; Lord, 1985a). However, the existence of precursor accumulating vesicles (PACs) has been demonstrated in developing castor bean endosperm and in maturing pumpkin seeds and induced in vegetative cells of Arabidopsis (Hara-Nishimura et al., 1998; Hayashi et al., 1999). PACs arising from the ER could provide a method of efficiently transporting, in a Golgi-independent manner, large quantities of non-glycosylated seed storage proteins that exhibit a propensity to aggregate at the site of their initial deposition and folding (Hara-Nishimura et al., 1998). Studies of a number of different plant vacuolar proteins have revealed that positive sorting information is required in order to reach the vacuole (Vitale and Raikhel, 1999). To date, three different classes of vacuolar sorting signal (VSS) have been identified: sequence-specific (ssVSS), thought to be receptor-mediated, C-terminal (ctVSS) and physical structure (psVSS) (Matsuoka and Neuhaus, 1999; Vitale and Raikhel, 1999).

The finding of two PAC membrane proteins with homology to the vacuolar sorting receptors BP-80 from pea (Kirsch et al., 1994) and AtELP from Arabidopsis (Ahmed et al., 1997) suggested that they too serve as sorting receptors for proteins destined for the PSV (Shimada et al., 1997). Indeed, using immobilised peptides derived from both native and mutant internal (2S-I) and C-terminal (2S-C) regions of the pumpkin 2S proalbumin, PV72/82 was shown to interact with a 2S-I sequence containing an RRE amino acid motif and a 2S-C sequence containing the sequence NLPS (Watanabe et al., 2002). In addition, the C-terminus of Brazil nut 2S albumin has been implicated both in vacuolar targeting (Saalbach et al., 1996) and in binding BP-80 (Kirsch et al., 1994).

Here, we have investigated the targeting of the castor bean 2S albumin precursor both by expressing precursor cDNA in transformed tobacco (Nicotiana tabacum) leaf cells and by following biosynthesis in the native tissue. In tobacco and castor bean endosperm, correct targeting is accompanied by processing of the precursor, and is sensitive to brefeldin A treatment. Further, in tobacco, we show that vacuolar targeting is blocked by the perturbation of COPII function, and that the targeting signal lies in a propeptide within the first heterodimer. This propeptide directed a normally secreted reporter to the vacuoles of tobacco cells in a Golgi-dependent manner, unless a critical Leu residue was mutated. The possible relevance of this with respect to targeting of the storage albumin in castor bean endosperm is discussed.

Results

The linear arrangement of subunits in the 29.3 kDa 2S albumin precursor of castor bean endosperm is shown schematically in Figure 1 as 2SA, and is based on the designation of subunits given in Irwin et al. (1990). The 21-residue signal peptide is followed by a 17-residue propeptide (PPI), after which lies the small subunit (SS1; expected MW 4.2 kDa), an intervening 14-residue propeptide (PPII) and the large subunit (LS1; expected MW 8.0 kDa) of the first heterodimer. The second heterodimer directly follows the first, and is composed of a small subunit (SSII; expected MW 4.0 kDa), an intervening 3-residue propeptide (PPIII) and a large subunit (LSII; expected MW 7.4 kDa).

Figure 1.

Schematic representation of the constructs expressed in tobacco protoplasts.

All coding sequences were under the control of the CaMV 35S promoter. SP, pre-pro2S albumin signal peptide; SP, prephaseolin signal peptide; white boxes represent propeptides I, II and III, respectively, SSI/II, small subunits I/II; LSI/II, large subunits I/II. 2SA refers to the castor bean 2S albumin precursor, whilst PHSL refers to bean phaseolin. Derivatives of these proteins are indicated schematically and described in the text.

Immunoprecipitates of 2S albumins from castor bean endosperm tissue that had been metabolically labelled for 1 h and chased for periods up to 24 h detected the presence of an approximately 30 kDa precursor (Figure 2a, black arrowhead) that was almost completely converted to processed products with the expected mobilities of approximately 4–8 kDa after 24-h chase (Figure 2a, vertical line). The cross-reacting 14-kDa band (highest open arrowhead) was a contaminant whose processing to a slightly faster species (lower open arrowhead) was sensitive to monensin whereas the 2S albumin was not (data not shown). Interestingly, the rabbit anti-albumin serum was also able to immunoselect the approximately 66 kDa lectin (ricin and agglutinin) precursors (Figure 2a, left panel, solid dot), and their processed subunits (asterisks), at 0- and 6-h chase: the two A chains (apparent overlapping MWs of 30 kDa), and the ricin and agglutinin B chains (with apparent MWs of 32 and 34 kDa, respectively). This co-precipitation probably arose as a result of lectin contamination in the original antigen preparation. As the routing of proricin has been well characterised (Frigerio et al., 1998a; Lord, 1985b), this could provide a useful reference point and control in subsequent experiments using the native tissue. That the trafficking of the proalbumin could be reproduced in tobacco leaf protoplasts is shown in Figure 2(b), where transformed cells expressed DNA encoding 2SA (Figure 1). Again, a precursor–product relationship was seen upon chase (Figure 2b), although the exact processing pathway appeared to be slightly different within tobacco vacuoles compared with that in castor bean PSV, as indicated by the presence of a different processing intermediate in the former (Figure 2a,c). Once again, the unidentified immunoreactive 14-kDa contaminant is not seen in tobacco protoplasts. The smallest albumin fragments of approximately 4–8 kDa are the same in both tissues, although in tobacco, there is also a larger 2SA intermediate equivalent in size to a single heterodimer (Figure 2c, asterisk) that is only occasionally seen in the native tissue. The consistent appearance of this intermediate in tobacco may reflect differences in the kinetics of processing 2SA. The key point here, however, is that processing to final products of approximately 4–8 kDa can be taken to indicate targeting to vacuoles, as it is for other vacuolar proteins such as ricin (Lord, 1985b) and phaseolin (Ceriotti et al., 1995). It should be noted that fragments generated in tobacco vacuoles are degraded, so the steady-state population is always low and it is not possible to quantitatively recover such fragments. By contrast, 2SA precursor en route to vacuoles, or the extracellular medium (below), is stable, making recovery more reliably quantitative. The appearance of a small amount of 2SA precursor in the medium during the chase (Figure 2b, lanes 10–12) was reminiscent of the behaviour of phaseolin under identical conditions (Frigerio et al., 1998b). This phenomenon is correlated with the sudden burst of protein synthesis that occurs upon transient expression (Frigerio et al., 1998b).

Figure 2.

The fate of 2S albumin in R. communis endosperm and tobacco leaf protoplasts.

(a) Maturing castor bean endosperm halves were pulse-labelled for 1 h, and chased for the indicated period of time in the presence of unlabelled amino acids. Cell homogenates were immunoprecipitated with anti-2S albumin antiserum and analysed by reducing SDS–PAGE. The black arrowhead indicates the position of the 2S albumin precursor, and the vertical line the position of vacuolar fragments. The dot and asterisks indicate the position of the Ricinus lectin precursors and processed subunits, respectively. The open arrowheads indicate an unidentified contaminant and its processed form. Numbers at the left indicate the molecular mass markers in kilodaltons.

(b) Tobacco protoplasts were transfected with 40 µg of the plasmid encoding 2S albumin (2SA), or 40 µg of the empty vector (pDHA), pulse-labelled for 1 h and chased for the indicated times in the presence of unlabelled amino acids before immunoprecipitation from homogenates or media and analysis as described for (a). The black arrowhead indicates the position of the 2S albumin precursor, and the vertical line the position of vacuolar fragments. Numbers at the left indicate the molecular mass markers in kilodaltons.

(c) A comparison of the fragments produced in castor bean and tobacco, indicating an immunoreactive contaminant in the former (open arrowheads) and a processing intermediate in tobacco (asterisk). The vertical line indicates the position of the processed 2SA fragments. Numbers at the right indicate the molecular mass markers in kilodaltons.

The vacuolar targeting of the castor bean 2S albumin precursor both in the native tissue and in tobacco leaf protoplasts was sensitive to the Golgi-disrupting reagent brefeldin A (BFA), as shown by the persistence of intracellular precursor, presumably trapped in the ER throughout the chase (Figure 3a,b, black arrowheads). Albumin processing is not seen in samples treated with BFA, whereas vacuolar fragments are visible in the untreated controls. As pointed out above, such fragments are turned over within vacuoles, so it is not possible to quantify the BFA effect by analysis of processed protein. It is therefore important to gauge the effect of such treatments by examining the fate of the precursor with time. As expected for proteins that normally traffic through the Golgi, BFA also blocked transport of the lectin precursors in the native tissue (Figure 3a, solid dot).

Figure 3.

Effect of BFA on the trafficking of 2S albumin in R. communis endosperm and tobacco leaf protoplasts.

(a) Maturing castor bean endosperm halves were pulsed for 1 h in the presence (+) or absence (−) of 10 µg ml−1 BFA and chased for the indicated periods of time, as above. Homogenates immunoprecipitated with anti2S albumin antiserum were analysed by reducing SDS–PAGE.

(b) Tobacco protoplasts were transfected with 40 µg of the plasmid encoding 2S albumin and then pulse-labelled for 1 h in the presence (+) or absence (−) of 10 µg ml−1 BFA and chased for the indicated period of time. Cell homogenates (lanes 1–6) and incubation media (lanes 7–12) were immunoprecipitated with anti-2S albumin antiserum and analysed by reducing SDS–PAGE. The black arrowheads indicate the position of the 2S albumin precursor, and the vertical lines indicate the position of vacuolar fragments. The dot and asterisks in (a) indicate the position of the Ricinus lectin precursors and processed subunits, respectively. Numbers at the left indicate the molecular mass markers in kilodaltons.

A similar phenotype was seen when tobacco cells were co-transfected with DNA encoding Sec12 (Figure 4a). The rationale for using Sec12 is that overexpression of this protein will titrate the pool of Sar1 to prevent the formation of COPII vesicles involved in anterograde transport from ER to Golgi (Phillipson et al., 2001). The block, therefore, can never be complete because the effect of Sec12 in a particular cell will be determined by its level of expression. Again, by observing the fate of precursors during the chase (black arrowheads) rather than their processed products, which are unstable in the vacuole, it is clear that Sec12 expression significantly blocked not only proalbumin transport (Figure 4a) but also the phaseolin control (Figure 4b), a protein that is known to traffic to vacuoles in a Golgi-dependent manner (Ceriotti et al., 1995; Pedrazzini et al., 1997). The persistence of these precursors within the cells throughout the chase period is a strong evidence for their retention in the early secretory pathway. The glycosylated phaseolin precursor had a slightly faster mobility on chase consistent with glycan modification (Figure 4b, lane 4). It can be rationalised that the block in COPII-dependent anterograde transport may permit such glycan modification if Golgi glycosylation enzymes that normally cycle through the ER (Cole et al., 1998; Storrie et al., 1998) now become redistributed there by COPI-mediated retrograde transport.

Figure 4.

ER to Golgi transport of 2S albumin and phaseolin is inhibited by overexpression of Sec12 in tobacco leaf protoplasts.

Protoplasts were transfected with 40 µg of the plasmid encoding (a) 2SA or (b) phaseolin (T343F), supplemented with either 40 µg of the Sec12 DNA (+) or the empty vector, pDHA (−). Cells were pulse-labelled for 1 h and chased for the indicated period of time. Homogenates (lanes 1–4) and incubation media (lanes 5–8), immunoprecipitated with appropriate antiserum, were analysed by reducing SDS–PAGE. The bottom of (a) is a longer exposure of this section of the fluorograph to allow clearer detection of the transient vacuolar fragments. The black arrowheads indicate the position of the precursor, and the vertical lines indicate the position of vacuolar fragments. Numbers at the left indicate the molecular mass markers in kilodaltons.

Vacuolar targeting signals are usually found within propeptide sequences that are removed once delivery is accomplished. We therefore decided to investigate the location of any albumin targeting signal by systematically eliminating the various propeptides. Whereas elimination of the first and third propeptides made no difference to the vacuolar targeting of 2SA (data not shown), when tobacco cells were transfected with DNA encoding preproalbumin DNA carrying a deletion of the second propeptide (PPII) (designated ΔPPII in Figure 1), vacuolar targeting of the precursor was abolished (Figure 5, lanes 4–6). Indeed, in the absence of this 14-residue propeptide, there was no evidence of any intracellular processing of ΔPPII that would indicate vacuolar delivery (Figure 5, lower panel, lanes 4–6), and the precursor was quantitatively secreted into the medium during the chase (Figure 5, upper panel, compare lanes 4–6 with lanes 13–15). Clearly, PPII is necessary for vacuolar targeting. Interestingly, a region within this internal propeptide (LRMP) bears some similarity to the NPIRL consensus of ssVSS (Matsuoka and Nakamura, 1999; Matsuoka and Neuhaus, 1999). It has previously been demonstrated that Ile to Gly substitution within this consensus disrupts vacuolar targeting. We therefore mutated the Leu residue of the LRMP sequence in PPII to a Gly residue, to generate 2SAL58G (Figure 1). We found that this substitution caused a significant loss of vacuolar targeting, with the 2SAL58G precursor now being secreted (Figure 5, upper panel, compare lanes 7–9 and lanes 16–18). Indeed, the secretion of the point mutant 2SAL58G was qualitatively the same as the secretion of 2SA lacking the entire PPII domain. It therefore seems that PPII is functioning as a ssVSS (Matsuoka and Nakamura, 1999; Matsuoka and Neuhaus, 1999).

Figure 5.

Delineation of the vacuolar sorting signal of 2S albumin.

Tobacco protoplasts were transfected with 2 µg of the plasmid encoding 2SA, ΔPPII, or 2SAL58G. In each case, transfection mixes contained 38 µg of the empty vector (pDHA). Cells were pulse-labelled for 1 h and chased for the indicated period of time. Cell homogenates (lanes 1–9) and incubation media (lanes 10–18) were immunoprecipitated with anti-2S albumin antiserum and analysed by reducing SDS–PAGE. The black arrowhead indicates the position of wild-type 2SA precursor, and the vertical line indicates the position of vacuolar fragments. The bottom section is from a longer exposure of this section of the fluorograph to allow clearer detection of the transient vacuolar fragments. Numbers at the left indicate the molecular mass markers in kilodaltons.

In order to test whether the proalbumin PPII was not only necessary but also sufficient to promote vacuolar targeting, fusions were created with a secretory version of phaseolin, Δ418, lacking its ctVSS (designated PPIIΔ418 in Figure 1) (Frigerio et al., 1998b). As the L58G mutation of PPII in its native protein resulted in secretion, we created a version of PPIIΔ418 bearing this same substitution (designated PPIIL58GΔ418 in Figure 1). As a control for vacuolar targeting, a singly glycosylated version of native phaseolin, T343F (Figure 1; Pedrazzini et al., 1997), was used. In tobacco leaf protoplasts, unlike the native tissue, targeting of phaseolin trimers is accompanied by processing into a set of low molecular weight fragments, the appearance of which is diagnostic of vacuolar delivery (Bagga et al., 1992; Pedrazzini et al., 1997). The results of pulse-chase experiments are shown in Figure 6a. Clearly, the transient vacuolar fragments are only evident in cells expressing T343F (Figure 6a, lane 2). Some T343F was recovered in the medium (Figure 6a, lanes 9 and 10), as has been previously observed by Jolliffe et al. (2003). As expected, the version of phaseolin lacking its vacuolar targeting signal (Δ418) accumulated in the medium during the chase (lane 12), with no visible fragments inside the cells (lane 4). Strikingly, PPIIΔ418 was not found in the medium at all (lanes 13–14), whilst PPIIL58GΔ418 was secreted to the same extent as Δ418 (compare lanes 12 and 16). Phaseolin fragments from PPIIΔ418 were, however, not recovered within the cells (Figure 6a, lane 6), or within isolated vacuoles (data not shown). Such alteration of Δ418 may render the protein extremely susceptible to intracellular proteolysis. Indeed, fusion of PPII onto the N-terminus of T343F appeared to cause an altered processing of the protein, as indicated by an absence of the largest fragment in the vacuoles upon chase (Figure 6a, compare lanes 2 and 18). This has also been seen for Δ418 bearing the proricin linker at its N-terminus (Jolliffe et al., 2003).

Figure 6.

Influence of the native and mutated 2S albumin PPII on the intracellular fate of the phaseolin reporter.

Tobacco protoplasts were transfected with 2 µg of the plasmid encoding wild-type phaseolin (T343F), secretory phaseolin (Δ418), and the fusions PPIIΔ418 or PPIIL58GΔ418 (left panel) or with PPIIT343F (right panel). In each case, transfection mixes contained 38 µg of the empty vector (pDHA). Cells were pulse-labelled for 1 h and chased for the indicated period of time.

(a) Cell homogenates (lanes 1–8, 17–18) and incubation media (lanes 9–16, 19–20) were immunoprecipitated with antiphaseolin antiserum and analysed by reducing SDS–PAGE. Numbers between the panels indicate the molecular mass markers in kilodaltons.

(b) Immunoprecipitates were treated with (+) or without (−) endoglycosidase H and analysed by reducing SDS–PAGE. The black arrowhead indicates the position of intact wild-type phaseolin, and the vertical line indicates the position of the transient vacuolar fragments. Numbers at the left indicate the molecular mass markers in kilodaltons.

To rule out ER retention and/or ERAD (Di Cola et al., 2001) as a possible explanation for the lack of vacuolar fragments within cells expressing PPIIΔ418, we analysed the single glycan on the phaseolin reporter. Analysis of the unprocessed PPIIΔ418 showed that, like T343F, it had acquired endoglycosidase H resistance by 6 h chase (Figure 6b, lanes 4 and 8, respectively). This indicated that PPIIΔ418 was not retained in the ER, as would be the case if it was held up by quality control, but that it had instead been transported through the Golgi, where terminal glycosylation occurs. Together, these data clearly demonstrate that PPII is both necessary for the vacuolar targeting of 2S albumin and sufficient to re-direct an otherwise secreted reporter protein.

Discussion

In the present study, the vacuolar sorting signal for the castor bean 2S albumin precursor has been delineated to the second of three propeptides (PPII), in which a sequence-specific signal resides (Figure 6a). A number of different regions in 2S albumin have been implicated in vacuolar targeting (Matsuoka and Neuhaus, 1999). However, interaction of these putative 2S albumin signals with targeting receptors has been largely inferred from in vitro peptide binding experiments using approaches that do not take account of the peptide's normal structural accessibility, or the physiological microenvironment in vivo (Shimada et al., 1997; Watanabe et al., 2002). The in vivo approach taken here however provides unambiguous evidence for a sequence-specific targeting determinant in the castor bean 2S albumin precursor in tobacco. This vacuolar sorting signal contains an LRMP sequence in which the Leu residue is functionally important. This is similar to the IRPV intramolecular targeting signal with its key hydrophobic Ile residue that has been found in the intramolecular propeptide of the ricin precursor, proricin (Frigerio et al., 2001; Jolliffe et al., 2003).

The apparent similarity of the castor bean 2S albumin and ricin vacuolar targeting signals is of interest for two reasons. Firstly, these sequence-specific signals resemble the type more usually found on proteins destined for lytic vacuoles (LV) that predominate in vegetative cells. Such signals usually contain an NPIR-like motif and often occur as N-terminal propeptides, as in sporamin (Matsuoka and Nakamura, 1999) and barley aleurain (Holwerda et al., 1992), although this location is not obligatory (Frigerio et al., 2001; Jolliffe et al., 2003; Koide et al., 1997). However, ricin and 2S albumin are clearly proteins of the PSV in castor bean endosperm. Targeting to the PSV, found mainly in seed and other storage tissues, often involves C-terminal signals that reside within propeptides, as in barley lectin (Dombrowski et al., 1993), common bean phaseolin (Frigerio et al., 1998b) and tobacco chitinase A (Neuhaus et al., 1994). These signals show no apparent sequence homology (Matsuoka and Neuhaus, 1999; Vitale and Raikhel, 1999). Secondly, the ricin signal is known to direct proteins through the Golgi in both tobacco leaf protoplasts and the native tissue. That 2S albumin also follows a Golgi-dependent route is supported by the effects of BFA (Figure 3), suggesting that trafficking through the Golgi is required for both ricin and 2S albumin. However, in native tissue, it has been reported that the 2S albumins of pumpkin and castor bean are transported to the PSV in a process involving the budding of large 200–400 nm vesicles directly from the ER (Hara-Nishimura et al., 1998). These PACs are believed to package and carry aggregated storage proteins in a route that completely bypasses the Golgi (Hara-Nishimura et al., 1998). These large vesicles can seemingly occur in cells of vegetative tissues too (Hayashi et al., 1999), although in this published example, PAC-like vesicles were induced by the ectopic expression of a 2S albumin but their fusion with vacuoles was not observed. It is possible, however, that PAC transport may be sensitive to the effects of BFA treatment. For example, the engorgement of the ER with Golgi contents may interfere with the aggregation of PAC cargo. Vacuolar delivery of 2S albumin was, however, also prevented in tobacco cells when subjected to a more specific block in membrane transport. Overexpression of Sec12, a guanine nucleotide exchange factor for the small GTPase Sar1, can titrate Sar1 to block COPII-dependent ER–Golgi transport (Phillipson et al., 2001). Under these conditions, the transport of phaseolin to the vacuoles was also disrupted (Figure 4). Further, during a chase, the acquisition of endoglycosidase H resistance by the fusion protein, PPIIΔ418 (in which the 2S albumin-targeting signal was fused to a secretory phaseolin reporter, Figure 6), again implied the involvement of the Golgi in the routing of the 2S albumin precursor.

These data are consistent with the idea that in tobacco at least, targeting of 2S albumin is Golgi- and receptor-mediated. A family of BP-80 receptors (Ahmed et al., 1997; Hadlington and Denecke, 2000; Kirsch et al., 1994; Shimada et al., 1997) has been recognised for proteins destined for vacuoles. Indeed, the membranes of PAC in maturing pumpkin seeds are known to contain at least two BP-80-like proteins, PV72/PV82 (Shimada et al., 1997). PV72 has been demonstrated to interact with specific storage albumin propeptides in vitro (Shimada et al., 1997; Watanabe et al., 2002). Furthermore, a recent study has shown that PV72 contains complex-type glycans and has a Golgi location (Shimada et al., 2002). If 2S albumin can be sorted by such a receptor, it is not implausible that BP-80 could fulfil this role in the heterologous expression system of tobacco protoplasts.

So what do these data imply for the trafficking of the castor bean 2S proalbumin? Data from pumpkin has placed 2S albumin in both PACs (Hara-Nishimura et al., 1998) and the Golgi (Shimada et al., 2002). While our data does not exclude either possibility, we can provide a rationale for passing through the Golgi. The sequence specificity of the Golgi-dependent vacuolar delivery in tobacco is consistent with the idea of a receptor-mediated step. Thus, it is possible that such a pathway could exist in castor bean endosperm also. Developing endosperm is a dedicated storage tissue with large-scale synthesis and deposition of storage protein being the predominant process. In this situation, PAC formation itself may not require sequence-specific targeting signals at all. Rather, it may be the process of aggregation (Hillmer et al., 2001; Holkeri and Vitale, 2001) around a nucleation site, similar to that occurring in the sorting of regulated secretory proteins in animal cells (Tooze et al., 2001), that triggers the vesicle budding process. Whether this is the predominant pathway taken by non-glycosylated proalbumin in the native tissue is as yet unresolved. However, under these conditions, it is conceivable that a proportion of 2S albumin may escape aggregation and, with bulk flow being the default pathway, it would be selectively advantageous to the plant to capture any errant protein. Closer analysis of the putative sequence-specific targeting signal of 2S albumin indeed reveals a similarity to the NPIRL consensus, although this is less evident than with the proricin signal. Thus, in the absence of lytic vacuoles, the albumin may have evolved a sequence-specific signal to permit salvage by Golgi-located BP-80-like receptors such as PV72.

Protein sequence alignment shows that the castor bean 2S albumin has a larger size compared to that of other albumins (data not shown), suggesting that it arose from a gene duplication event. Indeed, the second internal propeptide of castor bean albumin appears to be a truncated version of the first in that two of the three residues are the same P1′ and P1 amino acids that occur after and before the scissile bonds in PPII. Given that the albumin may have evolved a ssVSS signal to permit salvage, this abbreviated propeptide, PPIII, may therefore be a vestigial targeting signal. Clearly, further studies using the native tissues are now urgently warranted to resolve the significance of the various pathways to PSV for endogenous proteins, and to identify the molecular machineries and transport intermediates involved.

Experimental procedures

Recombinant DNA

Constructs used in this work are shown in Figure 1 (not to scale). All protein-coding sequences were fused downstream of the CaMV 35S promoter in expression vector pDHA (Tabe et al., 1995). The preproalbumin cDNA has been described previously by Irwin et al. (1990). Phaseolin constructs T343F and Δ418 are as described by Lupattelli et al. (1997) and Frigerio et al. (1998b), respectively, and preproricin as described by Frigerio et al. (1998a). The 2S albumin sequence was cloned into pDHA as an XbaI/PstI fragment using the oligonucleotides 5′-CTGCAGTTAGAACCGGCA-3′ and 5′-TCTAGAATGGCAAAGCTC-3′, from the preproalbumin cDNA described previously by Irwin et al. (1990). The phaseolin constructs T343F and Δ418 have been described previously by Frigerio et al. (1998b) and Lupattelli et al. (1997). The propeptide deletion was introduced into 2S albumin by using the oligonucleotides 5′-GCAATCAAGTTCAAGAAGACAGCAGCAGG-3′ and 5′-CCTGCTGCTGTCTTCTTGAACTTGATTGC-3′; the Leu 58 to Gly mutation was introduced using the oligonucleotides 5′-GAGAAGAAGTGGGAAGGATGCCTGGAGATGAAAACC-3′ and 5′-GGTTTTCATCTCCAGGCATCCTTCCCACTTCTTCTC-3′; the fusion of the second propeptide of 2S albumin to the N-terminus of both Δ418 and T343F was generated using the oligonucleotides 5′-CTGCCTCATTTGCCTCAACAGGAGAAGAAGTGTAAGGATGCCTGGAGATGAAAACACTTCACTCCGG-3′ and 5′-CCGGAGTGAAGTGTTTTCATCTCCAGGCATCCTTAACACTTCTTCTCCTGTTGAGGCAAATGAGGCTG-3′, and the fusion of the mutated (L58G) propeptide to the N-terminus of Δ418 was achieved using the oligonucleotides 5′- CTGCCTCATTTGCCTCAACAGGAGAAGAAGTGGGAGGATGCCTGGAGATGAAAACACTTCACTCCGG-3′ and 5′-CCGGAGTGAAGTGTTTTCATCTCCAGGCATCCTTCCCACTTCTTCTCCTGTTGAGGCAAATGAGGCTG-3′, all using the QuickChange in vitro mutagenesis system (Stratagene, La Jolla, CA, USA), following the manufacturer's instructions. The plasmid encoding Sec12 (Phillipson et al., 2001) was kindly provided by Dr J. Denecke (Leeds, UK).

Protoplast transfection

Protoplasts prepared from axenic leaves of tobacco (N. tabacum cv. Petit Havana SR1) were subjected to polyethylene glycol-mediated transfection as described by Pedrazzini et al. (1997), and incubated overnight at 25°C in the dark before pulse-chase labelling.

In vivo labelling of protoplasts and analysis of expressed polypeptides

Pulse-chase labelling of protoplasts using Pro-Mix (a mixture of 35S-Met and 35S-Cys; Amersham, Buckinghamshire, UK) was performed as described by Pedrazzini et al. (1997). Where indicated, a 2 mg ml−1 stock of brefeldin A (Boehringer Mannheim, Basel) dissolved in ethanol was added to the incubation medium to a final concentration of 10 µg ml−1 1 h before radiolabelling. Homogenisation of protoplasts and incubation media was performed by adding, to the frozen samples, two volumes of ice-cold homogenisation buffer (150 mm Tris–HCl, 150 mm NaCl, 1.5 mm EDTA, and 1.5% (w/v) Triton X-100 (pH 7.5)) supplemented with complete protease inhibitor cocktail (Roche). Immunoprecipitation of expressed polypeptides was performed as described previously by Frigerio et al. (1998b) using rabbit polyclonal antisera raised against either mature castor bean 2S albumin or phaseolin. Where indicated, immunoprecipitates were treated with 10 mU of endoglycosidase H (Roche, Germany; 5 mU µl−1) at 37°C overnight, after being re-suspended in 20 µl sodium citrate buffer (0.25 m sodium citrate (pH 5.5); 0.2% SDS) and boiled for 5 min.

In vivo labelling of Ricinus communis tissue

Ricinus communis plants were grown from seeds in a greenhouse at 15°C and under a 16 h light/8 h dark cycle. Endosperm tissue was excised from the ripening seeds during testa formation at a developmental stage when the lectins and storage proteins are rapidly synthesised (Lord, 1985b). Intact endosperm halves were each treated with 5 µl (25 µCi) of 35S-Promix, which was added directly to the abaxial surface. After a pulse of 1 h at room temperature, 10 µl of 2.5 mm unlabelled cysteine and methionine was added to each half. At each time point, one endosperm half was removed, frozen and ground to a fine powder in liquid N2. The frozen powder was mixed with 1.0 ml of 1% (v/v) NP-40 containing 0.2 m galactose, supplemented with complete protease inhibitor cocktail. After standing for 1 h at room temperature, membranes were removed from the suspension by centrifugation at 100 000 g for 30 min. Duplicate aliquots (25 µl) of the supernatant were added to 1.0 ml of cold 10% (w/v) trichloroacetic acid. The precipitated protein was collected by filtration onto a Whatman GF/A filter disc and washed with 10% trichloroacetic acid, and its radioactivity content was determined by scintillation counting after immersing the disc in 5 ml of scintillant. A volume of supernatant containing 106 c.p.m. was taken at each stage for immunoprecipitation. Where indicated, 5 µl of a 2 mg ml−1 stock of BFA (prepared as described above) was layered onto each half for 1 h prior to labelling.

Immunoprecipitation from R. communis tissue

Crude endosperm homogenates were dispersed in an equal volume of immunoprecipitation buffer (1% (v/v) NP-40; 10 mm Tris–HCl, pH 7.5; 150 mm NaCl; 2 mm EDTA; 0.4 m galactose) supplemented with complete protease inhibitor cocktail. Immunoprecipitation was carried out as described above, using rabbit polyclonal antisera raised against mature castor bean 2S albumin. It should be noted that the rabbit antiserum to the 2S albumins also contains antibodies against the Ricinus lectins. The ricin precursor and its processed subunits can therefore be immunoselected as contaminants along with the albumins.

Protein resolution and detection

Immunoselected proteins were analysed by 15% (w/v) reducing SDS–PAGE (0.075% (w/v) bisacrylamide or 0.5% (w/v) bisacrylamide for albumin, in order to permit detection of the lower molecular weight fragments) and fluorography.

Acknowledgements

This work was supported by UK Biotechnology and Biological Sciences Research Council by a studentship to JCB and grant 88/C17404 to LF/LMR. We thank Aldo Ceriotti for critical reading of the manuscript.

Ancillary