Localization of Vacuolar Transport Receptors and Cargo Proteins in the Golgi Apparatus of Developing Arabidopsis Embryos

Authors

  • Giselbert Hinz,

    1. Department of Cell Biology, Heidelberg Institute for Plant Sciences, University of Heidelberg, 69120 Heidelberg, Germany
    Search for more papers by this author
    • These authors contributed equally to the paper.

  • Sarah Colanesi,

    1. Department of Cell Biology, Heidelberg Institute for Plant Sciences, University of Heidelberg, 69120 Heidelberg, Germany
    Search for more papers by this author
    • These authors contributed equally to the paper.

  • Stefan Hillmer,

    1. Department of Cell Biology, Heidelberg Institute for Plant Sciences, University of Heidelberg, 69120 Heidelberg, Germany
    Search for more papers by this author
  • John C. Rogers,

    1. Division of Molecular and Cellular Biosciences, National Science Foundation, Arlington, VA 22230, USA
    Search for more papers by this author
  • David G. Robinson

    Corresponding author
    1. Department of Cell Biology, Heidelberg Institute for Plant Sciences, University of Heidelberg, 69120 Heidelberg, Germany
    Search for more papers by this author

David G. Robinson, david.robinson@urz.uni-heidelberg.de

Abstract

Using immunogold electron microscopy, we have investigated the relative distribution of two types of vacuolar sorting receptors (VSR) and two different types of lumenal cargo proteins, which are potential ligands for these receptors in the secretory pathway of developing Arabidopsis embryos. Interestingly, both cargo proteins are deposited in the protein storage vacuole, which is the only vacuole present during the bent-cotyledon stage of embryo development. Cruciferin and aleurain do not share the same pattern of distribution in the Golgi apparatus. Cruciferin is mainly detected in the cisand medial cisternae, especially at the rims where storage proteins aggregate into dense vesicles (DVs). Aleurain is found throughout the Golgi stack, particularly in the transcisternae and transGolgi network where clathrin-coated vesicles (CCVs) are formed. Nevertheless, aleurain was detected in both DV and CCV. VSR-At1, a VSR that recognizes N-terminal vacuolar sorting determinants (VSDs) of the NPIR type, localizes mainly to the transGolgi and is hardly detectable in DV. Receptor homology-transmembrane-RING H2 domain (RMR), a VSR that recognizes C-terminal VSDs, has a distribution that is very similar to that of cruciferin and is found in DV. Our results do not support a role for VSR-At1 in storage protein sorting, instead RMR proteins because of their distribution similar to that of cruciferin in the Golgi apparatus and their presence in DV are more likely candidates. Aleurain, which has an NPIR motif and seems to be primarily sorted via VSR-At1 into CCV, also possesses putative hydrophobic sorting determinants at its C-terminus that could allow the additional incorporation of this protein into DV.

Plants are unique among the eukaryotes in being able to store large amounts of protein. This normally not only takes place during seed formation but may also occur in vegetative tissues (1). The recipient organelle for these proteins is the vacuole, which may develop de novoto accommodate this function (2). Storage proteins are usually deposited in aggregated form in the protein storage vacuole (PSV), and their condensation may be seen earlier in the endomembrane system of plant cells (3–5). However, vacuoles like lysosomes also contain acid hydrolases, and even the PSV contains hydrolytic processing enzymes (6). Thus, different types of vacuolar proteins traffic through the plant secretory pathway, and as plants are now considered to possess multiple vacuole types (7,8), even in the same cell, there is an obvious requirement for selective recognition and sorting of these proteins.

Previous work on the identification of vacuolar sorting determinants (VSDs) underlines the complexity of vacuolar protein sorting in plants (9–11). Basically, three different types of VSDs have been elucidated: a signal-specific VSD – usually the NPIR motif, which is often located at the N-terminus (12) but occasionally at the C-terminus (13) and sometimes even internally (14). A second type of VSD of variable length, but often containing hydrophobic residues, is frequently present at the C-terminus of storage proteins [C-terminal vacuolar sorting determinant (ctVSD), (15)]. In addition to these, more or less well-defined VSDs, some storage proteins, e.g. legumin in peas and phytohemagglutinin in beans, appear to require the presence of large internal stretches of polypeptide for successful sorting to the PSV (16,17).

Two different types of vacuolar sorting receptor (VSR) are currently in discussion in the plant literature. The first to be identified on the basis of its specific binding to the NPIR motif present in the thiol protease aleurain was named BP-80 and was found to be enriched in clathrin-coated vesicle (CCV) preparations isolated from developing pea cotyledons (18). Arabidopsis has seven BP-80-type VSRs, the homologue of pea BP-80, being termed AtELP (now VSR-At1; (8,19)]. The second receptor is called as receptor homology-transmembrane-RING H2 domains (RMR) on account of the presence of a RING-H2 domain in its cytosolic tail (20). Although having a lumenal domain, which is similar to the PA domain of BP-80, RMR does not bind to aleurain. Instead, it interacts with the C-terminal VSDs in barley lectin, bean phaseolin and tobacco chitinase (21–23).

It had become convenient to postulate that acid hydrolases, such as aleurain, were sorted in the plant Golgi apparatus via BP-80-type VSRs and packaged into CCV for transport to the prevacuolar compartment (PVC). This was supported by three lines of evidence. First, the identification of a tyrosine motif in the cytoplasmic domain of BP-80, which was shown to interact with the μ-adaptin of an adaptor protein type 1 complex (AP-1) (24). Second, the discovery that multivesicular bodies (MVBs), representing the plant PVC (25), are enriched in VSRs (26). Third, VSRs also interact with VPS35, a component of the recycling retromer complex, which also locates to MVB (27). Thus, with BP-80, an appealing sorting and recycling scenario was developed, which was analogous to that operating in mammalian cells with the mannosyl 6-phosphate receptor and in yeast cells with the VSR Vps10p. At the same time, strong cytological evidence became available for a different pathway for storage proteins. This pathway known as the ‘dense vesicle’ (DV) pathway is based on the observation that the segregation of globulin-type storage proteins in developing pea cotyledons is initiated in the ciscisternae of the Golgi apparatus, leading to their aggregation at the periphery of the cisternae in vesicles with an osmiophilic/electron-dense content (4). As a consequence, the sorting mechanisms for acid hydrolases and storage proteins were considered to be separate events occurring at different sites in the plant Golgi apparatus (10,28). This notion is further supported by the recent findings of Rojo and coworkers (29) who have demonstrated that the two vacuolar sorting pathways in Arabidopsis are dependent on two distinct SNARE proteins, VTI11 and VTI12, respectively, where VTI11 is involved in targeting into the lytic vacuole and VTI12 in sorting into the storage vacuole.

Several reports, however, have recently challenged the notion that BP-type VSRs function exclusively to sort acid hydrolases. Hara-Nishimura and coworkers (30,31) have generated Arabidopsis VSR-At1 knockouts. The lesion is not lethal, but vacuolar protein sorting in the embryos of these plants is perturbed and results in a significant quantity of the storage proteins being secreted into the cell wall. In agreement with this, Otegui et al. (32) have reported the presence of VSR-At1 in the DV of Arabidopsis embryos. This has prompted us to reinvestigate the subcellular distribution of VSR-At1 in Arabidopsis embryos and to compare it with the distribution of RMR proteins on the one hand and on the other hand with aleurain and cruciferin as lytic and storage vacuole cargos, respectively. Whereas the distribution of RMR is conformed with its postulated role in storage protein sorting, our data do not support a direct role for VSR-At1 in this process. However, the distribution of aleurain – the original ligand for BP-80-type receptors – does not allow for a clear-cut association with VSR-At1 and CCV. Aleurain is sorted in part into DV and may be delivered by these vesicles to the PSV. We therefore believe that although the two sorting mechanisms are spatially separate and deal with different ligands, they are nonetheless interdependent, so that lesions in the one will ultimately affect the correct functioning of the other.

Results

Immunogold localization of cruciferin and aleurain in developing Arabidopsis embryos

As derived from in silicodata [AtGenExpress, (33); Figure S1], both the storage protein cruciferin and the cysteine protease aleurain are expressed throughout Arabidopsis embryo development. However, whereas the expression of cruciferin is strongly upregulated during embryo development, that of aleurain is not. Arabidopsis has three cruciferin genes that encode 50 kDa precursor proteins, which are then processed upon reaching the PSV to two polypeptides of ∼30 and ∼20 kDa (34). Like other 11S globulins, cruciferin exits the endoplasmic reticulum (ER) as an oligomeric 11S protein complex (34) and is deposited in the PSV (Figure 1A). Aleurain also gets processed: in barley aleurone, a 42-kDa proform chases into a 32-kDa mature form (35), but in Arabidopsis roots and suspension-cultured cells, the pro- and mature forms are 39 and 29 kDa, respectively (36). However, using a polyclonal antiserum generated against a recombinant C-terminal 127 aa peptide from Arabidopsis aleurain At5g60360 (36), a major polypeptide at 36 kDa (with an underlying minor band at 35 kDa) is detected in Western blots of mature embryos and desiccated seeds (Figure 2, lane 1). Assuming that this polypeptide is indeed aleurain, it means that at the ‘bent-cotyledon’ stage of embryo development, both storage (cruciferin, Figure 1A) and proteolytic proteins (aleurain, Figure 3A) accumulate in the same compartment, the PSV.

Figure 1.

Figure 1.

Immunogold electron microscopy of Arabidopsis embryos: cruciferin. A) Overview showing specificity of labeling. Cruciferin is present in PSVs and in putative PVCs (*). P, plastid. B–D) Gallery of Golgi stacks showing cruciferin gold label at the cisternal periphery (B), at the periphery and in a DV (C) and concentrated in an attached DV but not in a CCV (D). Dense vesicles are signified with arrows and CCV with small arrowheads. Bars = 250 nm (A) and 150 nm (B–D). c, cisGolgi faces; t, transGolgi faces.

Figure 2.

Figure 2.

Protein gel blots of Arabidopsis tissues with aleurain antibodies. Lane 1: imbibed dry seeds. Lane 2: cotyledon tissue from 3-day-old germinating seeds. Lane 3: leaves. Lane 4: cell suspension cultures. Total membrane fractions were prepared as given in Materials and Methods.

Figure 3.

Figure 3.

Immunogold electron microscopy of Arabidopsis embryos: aleurain. A) Overview showing specificity of labeling. Aleurain is concentrated in the PSV. B) Positive aleurain labeling of Golgi stack-associated DV. C) Aleurain gold labeling of a CCV. Nu = Nucleus. Bars = 200 nm (A and B) and 100 nm (C).

High-pressure freezing/freeze substitution (HPF/FS) prevents possible artifactual relocation of antigens during chemical fixation and therefore permits the detection of non-aggregated storage proteins in the lumen of Golgi cisternae and ER in developing pea cotyledons (37). Although the membranes of the Golgi cisternae were only weakly stained in HPF/FS samples of Arabidopsis embryos, the polarity of the Golgi stack was still recognizable (Figure 1). Cruciferin antibodies specifically labeled the Golgi apparatus (Figure 1B–D), but the labeling was not distributed evenly across the stack. About 85% of the immunogold label located to the cisand medial cisternae, with only 15% in the transGolgi (Table 1), and 89% of the medial and cislabel was concentrated at the rims of the cisternae (Figure 3B–D; Tables 1 and 2). This distribution of immunogold labeling therefore indicates that cruciferin is concentrated at putative DV-budding sites and is supported by the observation that 60% of the sectioned DV also labeled positively for cruciferin (Figure 3B–D; Table 3). These results are in agreement with our earlier data obtained on pea cotyledons for the storage globulins legumin and vicilin (37).

Table 1. Immunogold labeling of VSR and cruciferin between cis/median Golgi and transGolgi in Arabidopsis bent-cotyledon stage embryos
CruciferinaRMRaVSR-At1a
cis/medialtranscis/medialtranscis/medialtrans
  • a

    Sixteen Golgi stacks were evaluated for the cruciferin label, 15 for the RMR label and 25 for the AtVSR1 label.

  • b

    Total GPs counted.

11318b67b44b30b118b
85%15%60.5%39.5%20.3%79.7%
Table 2. Immunogold labeling of VSR and cruciferin within the cis/median Golgi cisternae in Arabidopsis bent-cotyledon stage embryos
CruciferinaRMRaVSR-At1a
RimbCenterbRimbCenterbRimbCenterb
  • a

    Sixteen Golgi stacks were evaluated for the cruciferin label, 15 for the RMR label and 25 for the AtVSR1 label.

  • b

    The rim is defined as the outer 20% of the cisternae, so center means 60% of the cisternae.

  • c

    Total GPs counted.

101c12c58c9c18c12c
89%11%86%14%60%40%
Table 3. Immunogold labeling of receptors and cruciferin in DV in Arabidopsis bent-cotyledon stage embryos
CruciferinRMRAtVSR1
  • a

    Ratio labeled/total DV.

64/107a74/231a11/119a
60% labeled32% labeled9% labeled

In comparison with cruciferin, immunogold labeling of Golgi stacks with aleurain antibodies was weaker (compare Figures 3B and 1B–D; Tables 1 and 4). But in contrast to cruciferin, the aleurain labeling was evenly distributed across the stack, with about 40% in the cishalf and about 60% in the transhalf of the stack (Table 4). Both CCV (Figure 3C) and DV (Figure 3B) labeled positively for aleurain, but the degree of labeling of DV depended on the developmental stage: in early bent-cotyledon stages, when the PSV had a flocculent rather than a solid content, only about 21% of the DVs were labeled (Table 5). At the end of embryo development, about 43% of the DVs labeled positively for aleurain (Table 5). This contrasts with the relatively constant DV labeling density for cruciferin between the two developmental stages, although the total cruciferin labeling of the Golgi almost doubles during this period (Table 6).

Table 4. Immunogold labeling of aleurain across the Golgi stack in Arabidopsis bent-cotyledon stage embryos
cismedialtrans
  • Fifty-two Golgi apparatus were evaluated and 22 were not labeled.

  • a

    Total GPs counted.

33a20a28a
40%25%35%
Table 5. Comparison of immunogold labeling of cruciferin and aleurain labeling in DV at different stages of bent-cotyledon embryo development
CruciferinAleurain
Young bentOld bentYoung bentOld bent
  • a

    Ratio labeled/total DV.

74/138a64/107a34/158a64/146a
53.6%60%21.5%43.8%
1.3 DV/Stack1.5 DV/Stack1.4 DV/Stack1.4 DV/Stack
Table 6. Comparison of immunogold labeling of aleurain and cruciferin in Golgi stacks at different stages of bent-cotyledon development
CruciferinAleurain
Young bentOld bentYoung bentOld bent
  • a

    Total Golgi stacks evaluated.

  • b

    Total GPs evaluated.

  • c

    Ratio of GPs per Golgi stack.

86a16a71a64a
393b131b161b150b
4.5c8.2c2.3c2.3c

Cruciferin colocalizes with RMR but not with VSR-At1 in Golgi stacks and DV

We compared the relative distribution of cruciferin with that of the two putative storage protein receptor proteins, VSR-At1 and RMR, in DV and within the Golgi stack. Because the relative labeling densities obtained with the different antibodies used were quite similar [with 7.5 gold particles (GPs)/stack for RMR, 5.9 GP/stack for VSR-At1 and 8.2 GP/stack for cruciferin, respectively], we consider the comparisons to be valid. About 80% of the VSR-At1 label was restricted to the transGolgi cisternae and the transGolgi network (TGN) (Figure 4A,B; Table 1). In the cisand medial cisternae, 60% of VSR-At1 was in the rims and 40% was in the center of the cisternae (Table 2). However, only 9% of the DVs were labeled with the VSR-At1 antibody (Table 3).

Figure 4.

Figure 4.

Immunogold electron microscopy of Arabidopsis embryos: VSR1 and clathrin. A and B) VSR1 label is restricted to the transGolgi where it is seen in budding vesicles with clathrin coats. C–H) Gallery of vesicle profiles at the transGolgi. Some have clathrin coat present either as a cap (C, E and G) or as a diffuse coating around the vesicle (F and H), and others do not have a discernible clathrin coat. The majority of these vesicles have an electron-dense content. The larger ones (C, D, G and H) are DV, but the smaller ones (E and F) may be CCV. Dense vesicles are signified with arrows. Clathrin (CL) is indicated with small arrowheads. Bars = 100 nm throughout. c, cisGolgi faces; t, transGolgi faces.

With an average diameter of 120 nm (without coat), DVs in Arabidopsis are less than half the diameter of those present in the pea cotyledons. Nevertheless, true CCV (see arrowheads in Figures 1D and 3C) can still be distinguished from DV as they have a diameter (without coat) of around 50 nm. Although clathrin can sometimes be difficult to detect in the HPF/FS samples, most of the VSR-At1-labeled DVs showed a partial clathrin-like coat structure (see arrowheads in Figure 4C–H). Because of their smaller size, the proportion of the DV membrane surface in Arabidopsis that is coated with clathrin is much higher than is the case with pea cotyledon DV. In order to positively identify clathrin at the surface of DV in the Golgi stack, immunogold labeling with clathrin heavy chain antibodies was performed. Ninety-four percent of the clathrin label was limited to the transGolgi (Table 7), but only 16% of the total Golgi-associated DV labeled positively for clathrin (Figure 4C–H). However, whereas about 32% of the transGolgi-localized DVs were clathrin positive, only one of 47 DVs attached to the cis/medial cisternae showed clathrin labeling (Table 8).

Table 7. Immunogold labeling of clathrin within the Golgi stacks of Arabidopsis bent-cotyledon embryos
Cis/medialatransa
  • a

    Thirty-nine Golgi stacks were evaluated.

  • b

    Total GPs counted.

5b82b
5.7%94.3%
Table 8. Immunogold labeling of DV with clathrin in Arabidopsis bent-cotyledon embryos
cis/medialatransa
  • a

    Thirty-nine Golgi stacks with 87 DV were evaluated.

  • b

    Ratio labeled/total DV.

1/47b13/40b
2.1%32.5%

The second putative receptor is the RMR protein, and in Arabidopsis, there are six RMR genes (20,22,23). Two RMR proteins defined by the complementary DNAs JR700 and JR702 have been investigated so far. These differ primarily in the length of the cytoplasmic domain, with JR700 having a calculated molecular mass of 34 kDa and JR702 of 48 kDa (20,38). The lumenal domains of both proteins are highly conserved, with about 41% identity, and the overall structural features are similarly highly conserved. As judged from microarray expression data (www.arabidopsis.org), the gene for JR702 (At1G71980) is expressed throughout embryo development as an abundant transcript. In essentially all tissues, it is expressed at more than twofold higher levels than that for JR700 (At5G66160). The antibody used here was raised against the lumenal domain of the JR702 RMR protein (20).

Because RMR was not detectable in specimens embedded in lowicryl resin HM20 (data not shown), the more hydrophilic K2M resin was used instead. Ultrastructural preservation of the specimens was not as good in that resin, but the polarity of the Golgi stacks and the DV attached to the stacks was discernible. In a manner similar to cruciferin, the larger part of the RMR label was found in the cisand medial Golgi cisternae (60%) as against 40% in the transcisternae (Figure 5B,C; Tables 1–3). This is a statistically different result from that for VSR-At1 (p ≤ 0.001). Unlike VSR-At1 (p ≤ 0.01), RMR antibodies preferentially labeled the rims of the cisternae: 86% of the label was at the periphery and only 14% of the label centrally localized (Figure 5B,C; Table 2). About 32% of DVs were labeled with RMR antibodies (Figure 5D,E; Table 3); again a result very different from that obtained for VSR-At1 (p ≤ 0.001).

Figure 5.

Figure 5.

Immunogold electron microscopy of Arabidopsis embryos: RMR. A) Overview showing specificity of labeling. RMR is seen in the PSV and in the Golgi apparatus, where it is seen in peripheral DV. Otherwise, label is not seen in the cytoplasm or on the cell wall (CW). B and C) Distribution of RMR in Golgi stacks: label is at the periphery and concentrated in DV. D and E) Individual DV with high RMR labeling density. Arrows point to DV. Bars = 200 nm (A) and 100 nm (B–E). c, cisGolgi faces; t, transGolgi faces.

The VSR-At1- and RMR-mediated sorting pathways merge in a joint PVC

Prevacuolar compartments in the seed coat, showing the typical morphological appearance of an MVB, were labeled with VSR-At1 antiserum, where GPs were predominantly associated with the limiting membrane (Figure 6A). In the embryo, PVCs with clear internal vesicles are difficult to recognize; nevertheless, VSR-At1 antibodies labeled predominantly the boundary membrane of organelles having the size of MVB (Figure 6B,D). These structures were completely filled with aggregated proteins and labeled positively for cruciferin (Figure 6D) and aleurain (Figure 6C) [see also (32); Figure 6B]. They also labeled positively for RMR (Figure 6E), thus indicating that the two transport pathways merge at the level of the PVC as suggested by Otegui et al. (32). Interestingly, RMR labeling was predominantly internal, consistent with previous findings that RMR protein is internalized into lumenal contents during its traffic to the PSV (20,23).

Figure 6.

Figure 6.

Immunogold electron microscopy of PVCs in Arabidopsis embryos. A) An MVB in a non-storage protein-expressing cell from the seed coat stained positively with VSR-At1 immunogold. B–E) Similar-sized PVCs in the cotyledon parenchyma of the embryo labeled with VSR-At1 (B and D), aleurain (C), cruciferin (D) and RMR (E). D is from a double immunogold-labeled section, with VSR-At1 as 10-nm GPs and cruciferin as 5-nm GPs. Arrows point to 10 nm gold, arrowheads to 5 nm gold. Bars = 100 nm.

However, our data, which indicate the presence of both RMR and VSR proteins in one type of PVC, seem to contradict those of Park et al. (23) who showed that the reporter constructs of the two VSRs (BP80 and RMR) when coexpressed were frequently found in separate organelles having the size of PVCs. Unfortunately, we were unable to perform double immunogold labeling for VSR-At1 and RMR because both antisera were generated in rabbits, so that we cannot comment on the possibility that there might be two populations of PVCs, each carry a different type of VSR. However, all the PVCs observed were filled with storage proteins. In addition, we should point out that a single PVC in developing Arabidopsis embryos is in keeping with the existence of a single vacuole type (see above), whereas in the suspension-cultured tobacco cells used by Park et al. (23), there are likely to be both storage and lytic vacuoles and as a consequence two separate PVC types with different VSRs.

Discussion

During the bent-cotyledon stage of embryo development, hydrolytic and storage proteins accumulate in the same vacuole

The bent-cotyledon stage is the last stage of embryo development in Arabidopsis before seed desiccation and represents the major period for storage protein expression (34,39). In agreement with Otegui et al. (32), only one vacuole is seen to be present in the cells of the cotyledons at this time, and this vacuole, because of the presence of copious amounts of storage proteins in particular the 11S globulin cruciferin, is nominally classified as a PSV. However, recent investigations in our laboratory on the heart and torpedo stages of embryo development indicate that this PSV arises de novoand, as previously shown for pea cotyledons (2), supercedes a typical lytic vacuole (C. Wei, G. Hinz, D. G. Robinson, unpublished data). It should be noted, however, that others have raised the question of whether prevacuolar organelles and PSVs themselves might be MVBs, where the vesicles define a separate compartment (40,41).

Our immunogold labeling data suggest that the PSV in developing Arabidopsis also contains large amounts of aleurain. However, we have grounds to question whether the protein detected really is aleurain. In agreement with Poxleitner et al. (42), we have detected an extra 29-kDa polypeptide in Western blots of 3-day-old germinating Arabidopsis seedlings (Figure 2, lane 2). According to these authors, there is no precursor–product relationship between the 36- and 29-kDa polypeptides, which are recognized by the aleurain antibodies. This strongly suggests that at least two developmentally different but closely similar cysteine proteases are produced by Arabidopsis embryos. Alternatively, the unique vacuolar environment during embryo development may generate processing of proaleurain to the 36-kDa product, a result that has been duplicated in vitrousing a processing protease different from the protease that yields 29-kDa mature aleurain (43). Despite the unclear nature and function of the aleurain-type antigen in the Arabidopsis embryos, a general principle seems to be emerging that two physiologically different types of protein accumulate in PSVs during cotyledon development, namely storage proteins and proteases. Thus, in addition to the data of Otegui et al. (32) on an aspartic protease, it has recently been reported for soybean cotyledons that a subtilisin-like serine protease also accumulates in the PSV and only becomes active through gradual acidification of the PSV during the first 48 h of germination (44).

The distribution of VSR-At1 in the Golgi apparatus of Arabidopsis embryos is not compatible with its postulated role in storage protein transport

If VSR-At1 were to be responsible for the recognition and segregation of storage proteins, its distribution in the Golgi apparatus of developing Arabidopsis embryos should reflect the site where sorting occurs, but this is not the case. Although there are several functionally different isoforms of VSR in Arabidopsis, which may also be detected by the VSRAt-1 antiserum (8,19,45,46), our immunogold labeling studies nevertheless show that the VSR antigens and clathrin have a similar distribution within Golgi stacks of developing Arabidopsis embryos, being principally located to the transcisterna and TGN. In contrast, the storage globulin cruciferin is mainly detected in the cisand medial cisternae, especially at the rims where DVs are formed. In fact, only a small percentage of DV (<10%) labeled positively for VSR-At1. These results are therefore conform with the previous data obtained by Hinz et al. (47) on developing pea cotyledons, which showed that the BP-80 receptor was enriched in CCV-containing fractions but not in DV fractions. However, our observations contradict those recently published by Otegui et al. (32). These authors presented micrographs of individual DV labeled with VSR-At1 but gave no information on the position of these DVs within the stack nor the relative incidence of this labeling. Moreover, the quality of their micrographs was such that clathrin coats were not recognizable, although the same cryopreparation procedures as employed here were used.

As demonstrated by Hillmer et al. (37) for pea cotyledons and confirmed here for Arabidopsis, the sorting of storage globulins and their packaging into DV are initiated very early as they transit the Golgi stack. Based on its different location in the Golgi apparatus, it is therefore difficult to see how VSR-At1 can participate in storage protein transport to the PSV, but this process is indeed perturbed in Arabidopsis VSR-At1 knockouts (30,31), although a substantial proportion of the storage proteins were still sorted to the PSV. How can this be explained? Two related sets of observations on mammalian cells need to be taken into consideration. In mammals, a condensation-based protein sorting mechanism operates in cells exhibiting regulated secretion and involves the gradual transformation of immature to mature secretory granules (48,49). Accompanying this maturation process is the attachment of clathrin to the membrane of the immature secretory granule and the formation of CCV (50,51). The function of CCV in this case is considered to be the removal of selected lumenal and membrane proteins (52–54). These CCVs have been shown to contain the mannosyl 6-phosphate receptor (MPR) as well as the adaptor molecules AP-1 and Golgi-localized gamma-adaptin ear homology containing Arf-binding proteins (GGAs) (55,56). Interestingly, interfering with GGA function reduces the formation of CCV, and this in turn downregulates secretory granule maturation (57). We have previously pointed out that DV may also undergo an analogous maturation, as CCV also develop and bud off from their surface (4,10,58). Rather than being involved in sorting storage globulins into DV, VSR-At1 may selectively remove proteins from DV, and this process might be essential for the correct functioning of the DV pathway to the PSV. If this interpretation is correct, and because GGAs are apparently not present in plants, we would predict that interfering with the functioning of the Arabidopsis μ-adaptin gene (24) should give rise to a similar phenotype as the VSR-At1 knockouts of Shimada et al. (30).

The second consideration involves the important role of receptors in governing membrane flow in cells. LeBorgne and Hoflack (59) demonstrated that mannose-6-phosphate receptors, because they are the major transmembrane proteins sorted toward endosomes, determine the number of CCV formed at the TGN. In the absence of these receptors, there was a major redistribution of adaptor proteins and clathrin within the cells. A similar result might be expected if VSR protein was absent. Membrane that would normally partition into CCV either accumulates in the Golgi or flows to alternative destinations. As VSR protein is normally abundantly present in prevacuolar organelles (26), the inability to direct membrane via CCV to those organelles might well impair their number or function and cause mistargeting of a variety of proteins. In this regard, an important unanswered question is whether RMR proteins were also missorted in the VSR knockout mutants.

Despite these considerations, the central concept that evolves from our present and previous (37) results is the importance of spatial regulation of sorting. Formation of storage protein aggregates early in the Golgi, and their recognition by RMR proteins would direct them to DV. However, if aggregation was a relatively inefficient process, some storage proteins would escape to reach the transGolgi where they might be caught by VSR proteins. These in turn might then direct these ‘escaped’ molecules into CCV for transport to a single mixed function PVC as appears to be present in Arabidopsis embryos. However, the numerous published experiments demonstrating that VSR proteins can bind sequences found on storage proteins do not prove that VSR proteins are the single and most important sorting receptor for storage proteins.

Are RMR proteins more likely candidate VSRs for storage proteins?

Park et al. (22) demonstrated that an RMR protein interacted with the C-terminal AFVY motif of the 7S globulin phaseolin and was essential for its correct targeting to the vacuole in Arabidopsis mesophyll protoplasts. Although not tested for in their paper, Otegui et al. (32) considered the participation of RMR proteins in storage protein transport in Arabidopsis embryos as unlikely because 7S globulins are not or hardly expressed in this tissue. However, there are six RMR genes in Arabidopsis, and the JR702 RMR gene [different from the RMR gene employed by Park et al. (22)] is expressed in Arabidopsis embryos. Immunogold labeling with JR702 RMR-specific antibodies reveals a distribution similar to that of cruciferin in the Golgi stack, with a much higher proportion of label being detected in DV than with VSR-At1 antibodies (p ≤ 0.001). Thus, from a structural point of view, RMR is more likely to be a receptor for storage protein sorting than VSR-At1. This opinion is in accord with the recent bimolecular fluorescence complementation studies of Park et al. (23), which showed that RMR proteins colocalize with a green fluorescent chitinase ctVSD reporter construct, while VSR-At1 did not. It is also supported by direct binding studies, which showed that while VSR-At1 recognized the sequence-specific vacuolar sorting determinant (ssVSD) in proaleurain, it did not interact with the ctVSD in chitinase. In contrast, and vice versa, RMR binds strongly to the ctVSD of chitinase but only if presented with a free C-terminus and only weakly to the ssVSD of proaleurain.

The RMR proteins were originally discovered as components of the membrane-containing crystalloid in tobacco seed PSV (20), and it has been shown by Park et al. (23) that the RMR–cargo interaction is not dissociated at low pH. Both observations suggest that unlike MPR and BP-80-type receptors, RMR proteins are not recycled from an acidic PVC (10,25) and must accompany their attached cargo to the PSV. Superficially, this might seem a costly way to sort proteins, but the sorting of storage globulins also entails aggregation/condensation, as has been demonstrated both morphologically (27) and biochemically (47). As shown by Castelli and Vitale (60), the aggregation of storage protein precursor proteins depends on the presence of the ctVSD. Furthermore, aggregation seems to be an early event in the Golgi apparatus starting in the cisrather than in the transGolgi cisternae (60). This latter result is in accordance with the localization of the sorting receptors to the cisrather than to the transhalf of the Golgi stack [(37) and this report].

There are therefore two possibilities as to how RMR proteins might participate in this event. Individual storage proteins could bind to RMR proteins and act as nuclei for further storage protein condensation. Alternatively, storage protein microaggregates could present on their surfaces free ctVSDs for RMR protein interaction. The essential difference between these two possibilities is whether storage protein aggregation occurs before or after the interaction with RMR proteins. Either way, a relatively small number of RMR proteins could interact with a large number of storage protein molecules and provide the cytoplasmic motifs for interactions with as yet uncharacterized coat proteins that would drive DV formation. By acting in small numbers in the formation/sorting of protein aggregates, RMR proteins would therefore still be able to function efficiently as receptors even if they could not recycle back to the Golgi apparatus.

Clathrin-coated vesicles or DV: aleurain contains two different types of VSDs

Interestingly, although the N-terminal NPIR-containing sequence of aleurain was used successfully as bait to fish for BP-80 (18), aleurain itself contains at least one additional type of putative VSD. As shown in Figure S2A, aleurain belongs to a family of conserved proteins all of which contain an NPIR-like motif at their N-terminus (61). However, they also contain a highly conserved hydrophobic C-terminal motif similar to other known ctVSDs (10,11). Significantly, in closely related proteases that are secreted, like EP-B in Hordeum vulgare, both motifs are absent [(35); Figure S2B]. In contrast, neither the vacuolar processing enzyme (β-VPE) nor the aspartic protease (AP-A1), investigated by Otegui et al. (32), possesses a clearly defined VSD (see Figure S3 for sequence alignments). Proteins of the VPE family have a conserved isoleucine at their N-terminus, and at their C-terminus, they bear a GXSA sequence. Aspartic protease-A1 lacks an isoleucine-containing motif at its N-terminus but does have a conserved putative C-terminal hydrophobic motif (AA, AV). The C1 serine protease in the PSV of soybean cotyledons also lacks an NPIR motif but does have a number of hydrophobic residues at its C-terminus […RSPIVVFNTA; (44)].

Speculations have been made concerning the potential role of hydrophobic sequences of this type in the sorting of storage proteins (10) and have been fueled by the demonstration that the C-terminal AFVY sequence of bean phaseolin (a 7S vicilin-type globulin) is essential for the correct targeting of this protein to the vacuole (15). However, recent experiments performed on the 11S globulin of soybean indicate that a C-terminal hydrophobic sequence alone is not sufficient for targeting to the PSV. Two additional determinants, including an N-terminal-located signal-specific type VSD, were found to be necessary (62). A similar conclusion was reached by He et al. (44) who noted that some other C1-type serine proteases also have similar C-terminal hydrophobic residues but are secreted [see also (63)]. Therefore, there are no consensus sorting motifs for proteases that accumulate in PSVs in maturing cotyledons along with the storage proteins, which become their substrates during seed germination. It is entirely possible that depending on the nature of the dominant VSD in a particular seed protease, it could be sorted into either a CCV or a DV. However, and apparently irrespective of the type of Golgi exit vesicle, proteases and storage proteins are brought together at the PVC [Figure 6; see also (32)].

In contrast to Otegui et al. (32) who showed that neither β-VPE nor AP-A1 exited the Golgi apparatus separately to storage proteins, our data suggest that aleurain is present in DV, as well as in CCV. However, as mentioned above, the nature of the antigen(s) detected by the aleurain peptide antibodies is unclear. In order to shed further light on this matter, we have performed a sequence blast using the Arabidopsis thalianatigr database against the C-terminal 127 aa, which were used for the original immunization. We identified 28 cysteine proteases with a score of 200 or more and having at least 32% identity and 52% positives in the C-terminus as putative immunopositive proteins. Out of these, we further selected for proteins expressed in developing seeds having a signal peptide and a molecular mass of 36–40 kDa. We identified five cysteine proteases having a putative hydrophobic ctVSD and three having a putative N-terminal ssVSD (Figure S4). Significantly, and in contrast to aleurain, these proteins appear to have only a single VSD. Based on this information, we interpret the immunogold labeling of both DV and CCV as reflecting the sorting of two different aleurain-like cysteine proteases on the basis of a ctVSD and a N-terminal ssVSD, respectively.

Materials and Methods

Immunoelectron microscopy

Eight- to 10-day-old Arabidopsis embryos were removed from the siliques. The seed coat was removed after submerging individual embryos in hexadecene. Five to eight submerged embryos were then mounted onto planchettes and frozen in a high-pressure freezer (HPF010; Bal-Tec). Freeze substitution was then performed in a Leica AFS freeze substitution unit (Leica) in dry acetone supplemented with 0.1% uranyl acetate at −85°C for 3 days before gradually warming up to −35°C over an 18-h period. After washing twice in 100% ethanol for 60 min, the seeds were infiltrated and embedded in either Lowicryl HM20 or K4M at −35°C and polymerized for 2 days at the same temperature with ultraviolet (UV) light in the FS apparatus (37). To increase sectioning quality, the blocks were then hardened with UV light for another 3 days at room temperature. Ultrathin sections were cut on a Leica Ultracut S (Leica) and incubated with antibodies against cruciferin (protein A purified serum at 1:50 dilution; H. Bäumlein, IPK Gatersleben, Germany), aleurain [protein A purified serum at 1:50 dilution (36)], VSR-At11 (affinity purified at 1:2 dilution) (25), RMR (at 1:5 dilution) (20) and clathrin (monoclonal clathrin heavy chain antibody C610500 at 1:25 or 1:50 dilution; BD Bioscience), followed by incubation with 10-nm gold-coupled secondary antibodies (BioCell GAR10 or GAM10 in the case of clathrin; BioCell) at a dilution of 1:50 in PBS supplemented with 1% BSA.

For double immunolabeling with antisera against VSR-At1 and cruciferin, sections were first placed on a solution with the first antibody (VSR-At1), followed by the incubation with gold-coupled secondary antibody (BioCell GAR5). The section was then transferred onto a grid with the antibody-covered face toward the grid surface. The grid was than inverted and floated on a drop of the second antibody (cruciferin), followed by the incubation with the secondary antibody (BioCell GAR10). Buffers and dilutions were as described above.

Sections were poststained with aqueous uranyl acetate/lead citrate and examined in a Philips CM10 transmission electron microscope operating at 80 kV. Negatives were scanned with an Epson Perfection 4990 Photo by the Epson Photoshop Plugin on an Apple MacIntosh computer using settings individually optimized for visibility of cell structures and GPs.

Protein isolation and protein determination

Ten grams of Arabidopsis cultured cells (27) was homogenized on ice using an Elvehjem homogenizer (Braun) in 20 mL buffer A containing 40 mm HEPES-KOH, pH 7.4, 10 mm KCl, 3 mm MgCl2, and 0.5 mm ethylenediaminetetraacetic acid, and the protease inhibitors 2 μg/mL leupeptin, 2 μg/mL aprotinin, 1 μg/mL transepoxysuccinyl-l-leucylamido-(4-guanido)-butane (E-64), 0.7 μg/mL pepstatin and 1 mm o-phenanthroline. The homogenate was filtered through two layers of paper fleece (Schleicher & Schuell), centrifuged for 5 min at 2000 × g, the supernatant was centrifuged for 20 min at 20 000 × g and the supernatant was centrifuged for 1 h at 100 000 × g, yielding a cytosol (supernatant) and a total membrane fraction, respectively. Two grams of rosette leafs of green-house-cultured Arabidopsis plants (27) was homogenized in a Waring blender (Braun) in 4-mL buffer A, and the homogenate was filtered as described above and cleared by a low-speed centrifugation for 10 min at 2000 × g. For germination, dry seeds were cultured on a moist paper fleece in a Petri dish at room temperature in the dark. One hundred milligrams of each developing seeds, dry mature seeds and 3-day germinating seedlings, respectively, was homogenized in a mortar on ice in 2-mL of buffer A and sea sand and cleared by a low-speed centrifugation for 10 min at 2000 × g. Again, total membranes were sedimented at 100 000 × g. Protein was measured according to the method of Bradford (64).

Gel electrophoresis and protein gel blotting

Prior to SDS–PAGE, proteins were precipitated with methanol/CHCl3 according to Wessel and Flügge (65). SDS–PAGE was performed according to Laemmli (66). Proteins were electroblotted onto nitrocellulose using a BioRad semidry blotting device in a transfer buffer (0.025 m Tris and 0.192 m glycine) containing 10% (w/v) methanol for up to 90 min at 2 mA/cm2. After blocking (1% BSA and 5% fat-free dry milk), the blots were probed with cruciferin antibodies at a dilution of 1:5000 or with aleurain antibodies at a dilution of 1:5000. Secondary antibodies were coupled to horseradish peroxidase (Sigma). Visualization of the bound antibodies was with an enhanced chemiluminescence kit (Pierce).

Gene alignments and expression studies

The sequence alignment was performed using the clustalw program (67) provided by the embl-ebi database and search engine. Accession numbers for the sequences were as follows – aleurain family proteins: H. vulgarealeurain BAF02548; Lolium multiflorumcysteine proteinase CAB71032; Oryza sativaoryzain gamma chain precursor P25778; Zea mayscysteine protease CAA68192; Z. mayscysteine protease 2 Q10717; Brassica napuscysteine protease ABA71255; Nicotiana tabacumcysteine protease BAA96501; Brassica oleraceasenescence-associated cysteine protease AAL60582; Nicotiana benthamianacysteine protease aleurain type AAZ32410; N. tabacumNTCP23-loke cysteine protease AAK07729; Pisum sativumcysteine protease ABF18679; P. sativumearly leaf senescence abundant cysteine protease CAC41636; Solanum lycopersicumcysteine proteinase 3 precursor Q40143; A. thalianathiol protease aleurain precursor AtAleu Q8H166, At5g60360; A. thalianaputative cysteine proteinase AALP ANN31820; A. thalianacysteine-like endopeptidase AALP-like NP_001032106; A. thalianathiol protease aleurain-like precursor Q8RWQ9, At3g45310; A. thalianacysteine-type endopeptidase NP_001030812; and A. thalianacysteine protease-like protein CAB72480.

Expression studies were performed using the AtGenExpress Visualization Tool (33)– cruciferin: A. thalianaCAB81440, At4g28520; P15455, At5g44120; AAP81800, At1g03880; NP_171885, At1g03890 and aleurain AtAleu Q8H166, At5g60360.

Acknowledgments

We are most grateful to Steffi Gold for technical assistance. This study was supported by funds from the German Research Council, DFG Ro 440/11-3. We are indebted to Dr Natasha Raikhel (Riverside, USA) for the gift of the aleurain antibody and Dr Helmut Bäumlein (IPK Gatersleben, Germany) for the gift of the cruciferin antibody.

Ancillary