Sorting of plant vacuolar proteins is initiated in the ER

Authors


For correspondence (fax +49 7071 295797; e-mail Peter.Pimpl@zmbp.uni-tuebingen.de).

Summary

Transport of soluble cargo molecules to the lytic vacuole of plants requires vacuolar sorting receptors (VSRs) to divert transport of vacuolar cargo from the default secretory route to the cell surface. Just as important is the trafficking of the VSRs themselves, a process that encompasses anterograde transport of receptor–ligand complexes from a donor compartment, dissociation of these complexes upon arrival at the target compartment, and recycling of the receptor back to the donor compartment for a further round of ligand transport. We have previously shown that retromer-mediated recycling of the plant VSR BP80 starts at the trans-Golgi network (TGN). Here we demonstrate that inhibition of retromer function by either RNAi knockdown of sorting nexins (SNXs) or co-expression of mutants of SNX1/2a specifically inhibits the ER export of VSRs as well as soluble vacuolar cargo molecules, but does not influence cargo molecules destined for the COPII-mediated transport route. Retention of soluble cargo despite ongoing COPII-mediated bulk flow can only be explained by an interaction with membrane-bound proteins. Therefore, we examined whether VSRs are capable of binding their ligands in the lumen of the ER by expressing ER-anchored VSR derivatives. These experiments resulted in drastic accumulation of soluble vacuolar cargo molecules in the ER. This demonstrates that the ER, rather than the TGN, is the location of the initial VSR–ligand interaction. It also implies that the retromer-mediated recycling route for the VSRs leads from the TGN back to the ER.

Introduction

It is generally acknowledged that the receptors responsible for sorting acid hydrolases into the lysosome of mammalian cells (the mannosyl 6-phosphate receptor, MPR) (Braulke and Bonifacino, 2009) and into the vacuole of yeast cells (Vps10p) (Bowers and Stevens, 2005) recognize their ligands in a late Golgi compartment: the trans-Golgi network (TGN). In both classes of organisms, transport away from the TGN towards a pre-lysosomal/pre-vacuolar compartment is clathrin-dependent (Deloche et al., 2001; Puertollano et al., 2003), involving an interaction between monomeric adaptors (GGAs, Golgi-localized, γ-ear-containing ARF-binding proteins) and dileucine-containing motifs in the cytosolic tails of the receptors (Doray et al., 2007; Abazeed and Fuller, 2008). Upon arrival, receptor–ligand complexes dissociate, and the ligands are delivered to the lysosome/vacuole through fusion of a pre-lysosomal/pre-vacuolar intermediate with the lytic compartment (Bright et al., 2005), while the receptors are recycled back to the TGN in a retromer-dependent manner (Nothwehr et al., 1999; Arighi et al., 2004; Bonifacino and Hurley, 2008; Schellmann and Pimpl, 2009). Retromer consists of a heterotrimeric cargo-recognition complex and two sorting nexins (SNXs) that constitute the small retromer subunit (Seaman et al., 1998). SNXs are characterized by the presence of two domains: a phox (phagocyte NADPH oxidase complex) homology (PX) domain that allows binding to phospholipids, and a BAR (Bin, amphiphysin, Rvs) domain that enables sensing of membrane curvature, and thus drives the formation of transport carriers (Cullen, 2008; Lemmon, 2008).

The above paradigm for post-Golgi protein trafficking has often been used as a guideline for plant scientists in discussions on vacuolar protein transport (for recent reviews, see Richter et al., 2007; Bassham and Blatt, 2008; Foresti and Denecke, 2008). However, this paradigm does not withstand critical evaluation of the differences between the situation in plants and the other two eukaryotic cell types. Firstly, sorting of lysosomal acid hydrolases in mammalian cells is a highly coordinated process in the Golgi apparatus involving cargo recognition and cryptic signal transformation in the cis-Golgi, followed by signal unveiling and receptor recognition in the TGN. Cargo recognition occurs on the basis of a tertiary structure-based signal patch that involves two or three lysine residues (Tikkanen et al., 1997). This is recognized by the enzyme GlcNac-1-phosphotransferase, whose action results in blockage of the processing of N-linked oligosaccharides on these hydrolases. Exposure of the key mannosyl 6-phosphate groups for recognition by MPRs then occurs in the TGN through the action of a second enzyme known as the uncovering enzyme (N-acetylglucosamine-1-phosphodiester N-acetyl glucosaminidase) (Rohrer and Kornfeld, 2001). In contrast, soluble proteins destined for the lytic vacuole in plants, and also for the yeast vacuole, are sorted on the basis of a primary amino acid sequence motif. For plants, this is most commonly of the NPIR-type and is usually situated at the N-terminus (Paris and Neuhaus, 2002; Robinson et al., 2005), but targeting peptides at the C-terminus are also known (Bednarek et al., 1990; Shimada et al., 2002). Vacuolar sorting receptors (VSRs) of the BP80 family recognize both types of sorting motif, and it has been suggested that the specificity for NPIR-binding is conferred by the combined action of the N-terminal RMR homology domain together with the central domain, while the other motif binds to the central domain, which is stabilized by epidermal growth factor (EGF) repeats (Cao et al., 2000). In short, there are no similarities in the luminal domains of the mammalian MPRs, yeast Vps10p and plant VSRs, and the mechanism of cargo recognition is completely different in each case.

Secondly, the cytoplasmic tails of the three classes of receptors also differ significantly with respect to their structures, except for the presence of a tyrosine sequence. Although the tyrosine motif in the cytoplasmic domain of VSRs is capable of interacting with a clathrin adaptor (Happel et al., 2004), this endocytosis motif is used for internalization of MPR–ligand complexes at the plasma membrane, and does not appear to be important for biosynthetic trafficking towards the lysosome (Ghosh et al., 2003). Instead, di-leucine motifs, usually located at the C-terminus of the MPRs, are recognized by the GGAs, causing their sequestration into clathrin-coated vesicles (CCV) (Doray et al., 2007). In yeast, the tyrosine sequence is necessary for the efficient cycling of Vps10p between the TGN and the pre-vacuolar compartment (PVC) (Cooper and Stevens, 1996), but it is unclear whether exit from the Golgi is dependent upon it (Bowers and Stevens, 2005). On the other hand, vacuolar protein sorting is perturbed in GGA-deficient cells (Hirst et al., 2001), and GGAs are absolutely required for TGN-to-PVC transport in a cell-free yeast system (Abazeed and Fuller, 2008). However, Vps10p lacks a canonical DXXLL signal, so that, in the case of this receptor, the GGAs must recognize some other motif in the cytosolic tail (Abazeed and Fuller, 2008). BP80-type VSRs also lack a di-leucine motif in their cytoplasmic tails, but, in contrast to mammalian and yeast cells, GGAs are not encoded in the Arabidopsis genome.

Thirdly, plant retromer, which has been shown to interact with VSRs in vitro (Oliviusson et al., 2006), had previously been localized to the PVC (Oliviusson et al., 2006; Jaillais et al., 2008; Yamazaki et al., 2008), and was therefore judged to recycle VSRs back to the TGN (Foresti and Denecke, 2008). However, a recent re-examination of retromer localization in plants has placed this recycling protein complex at the TGN (Niemes et al., 2009). Moreover, as in mammalian cells, vacuolar cargo transport beyond the early endosome, which in plants is the TGN, appears to be receptor-independent (Niemes et al., 2009).

In this paper, we demonstrate that VSRs interact with their soluble ligands in the lumen of the ER. We further show that retromer-mediated recycling is required for the specific ER export of VSRs and their ligands. As this transport is independent of the COPII-mediated bulk flow of secretory cargo, we postulate that VSRs and their ligands are selectively exported from the ER, and, after completion of transport, VSRs recycle back to the ER in a retromer-dependent manner.

Results

Co-expression of SNX mutants or RNAi knockdown of SNXs traps BP80 in the ER

We have previously reported that perturbation of retromer-mediated transport through expression of sorting nexin 1 (SNX1) or SNX2a mutants in tobacco mesophyll protoplasts led to a shift in the steady-state distribution of the VSR reporter GFP–BP80 from the PVC to the TGN (Niemes et al., 2009). SNX mutants lacking the N-terminus (ΔN) are thought to prevent recruitment of the large retromer subunit onto the membranes, while mutants lacking the coiled-coil domain (ΔCC) fail to drive membrane curvature for carrier formation. However, although individual expression of SNX1 or SNX2a mutants altered transport of the VSR, vacuolar delivery of soluble cargo was only marginally perturbed (Niemes et al., 2009). This raised the question as to whether retromer function is completely inhibited under these conditions, as it has been reported that mammalian SNXs may be capable of forming homodimers to compensate for the loss of a single SNX type (Rojas et al., 2007). Therefore, we analysed the effects of simultaneous expression of SNX1 and SNX2 mutants on VSR trafficking.

Tobacco protoplasts co-expressing either the two SNX1/SNX2a-ΔN mutants (Figure 1d–f) or the two SNX1/SNX2a-ΔCC mutants (Figure 1g–i) with the VSR reporter GFP–BP80 and the Golgi marker Man1–RFP led to drastic accumulation of the VSR reporter in the ER compared to the control (Figure 1a–c). However, this accumulation is not a result of over-production of the reporter. Although such an effect may be observed with Golgi markers, e.g. ST–YFP, and already occurs when the optimal amount of expressed plasmid DNA is increased only two- to threefold, over-production of GFP–BP80 does not result in accumulation in the ER, even if the amount of DNA used for transfection is increased 10-fold (see Figure S1). The typical punctate pattern of the co-expressed Golgi marker (Figure 1b,c) was not affected by expression of the SNX mutants (Figure 1e,f,h–i), indicating that the observed effect is specific for the VSR reporter.

Figure 1.

 Co-expression of SNX1 and SNX2a mutants trap VSR reporter in the ER.
Transient expression in tobacco protoplasts. Effect of sorting nexin mutants on distribution of the Golgi marker Man1–RFP (red) and the PVC marker GFP–BP80 (green). Co-expression of markers alone (a–c), or together with SNX1-ΔN and SNX2a-ΔN mutants (d–f) or SNX1-ΔCC and SNX2a-ΔCC mutants (g–i). Note: simultaneous expression of two SNX mutants shifts the distribution of the PVC marker to the ER, whilst the Golgi marker remains punctate. Scale bars = 5 μm.

As an alternative strategy to inhibit retromer function by tackling both SNXs simultaneously, we performed RNAi knockdown experiments. By expressing an RNAi construct that is directed against the first coiled-coil domain of SNX2a but is also expected to affect the other two sorting nexins due to the high degree of similarity within this domain, we were able to induce post-translational gene silencing of all SNXs (Niemes et al., 2009). RNAi knockdown in protoplasts isolated from tobacco leaves (Figure 2a–c) or suspension-cultured Arabidopsis cells (Figure 2d–i) also resulted in a drastic accumulation of the VSR reporter in the ER (compare Figure 2a with Figure 1a, and Figure 2g with Figure 2d), and again the Golgi signal remained unaffected in all cases (compare Figure 2b with Figure 1b, and Figure 2h with Figure 2e). Because of the unchanged pattern of the Golgi marker, we hypothesized that accumulation of the VSR reporter in the ER does not cause a general inhibition of ER export. To prove this hypothesis, we performed quantitative protein transport studies in tobacco protoplasts, using α-amylase as a reporter to monitor protein transport towards the cell surface (Phillipson et al., 2001). Consistent with the continued presence of functional Golgi stacks during the live-cell imaging analysis, the secretion of α-amylase was not inhibited by co-expression of the two SNX1/SNX2a-ΔN mutants, or co-expression of the two SNX1/SNX2a-ΔCC mutants or by RNAi knockdown (see Figure S2).

Figure 2.

 SNX RNAi knockdown traps GFP–BP80 in the ER.
(a–c) Transient expression in tobacco protoplasts. Effect of SNX RNAi expression on distribution of the Golgi marker Man1–RFP (red) and the PVC marker GFP–BP80 (green). Note: co-expression of the markers and the SNX RNAi construct shifts the distribution of the PVC marker to the ER, whilst the Golgi marker remains punctate.
(d–i) Transient expression in Arabidopsis protoplasts. Co-expression of markers alone (d–f) and together with the SNX RNAi construct (g–i). Note: expression of the SNX RNAi knockdown shifts the distribution of the PVC marker to the ER, whilst the Golgi marker remains punctate.
Scale bars = 5 μm.

BP80 accumulation in the ER is due to selective inhibition of ER export

In order to determine the origin of the BP80 molecules that accumulated in the ER after expression of the RNAi construct for 24 h, we performed a time-course experiment using tobacco protoplasts. At early stages (6–12 h), the PVC marker GFP–BP80 and the cis-Golgi marker Man1–RFP were found in close proximity to each other in control protoplasts (Figure 3a,c) as well as in protoplasts expressing the SNXCCRNAi construct (Figure 3b,d). This pattern most likely represents a Golgi/TGN marker distribution, rather than the expected Golgi/PVC distribution. Individual GFP–BP80 signals were also present in both cases, and may represent either non-Golgi-associated TGNs (Foresti and Denecke, 2008) or PVCs (Tse et al., 2004). However, in the control protoplasts, the Man1–RFP and GFP–BP80 signals were clearly separate after longer periods of incubation (20–24 h), yielding the typical distribution pattern of Golgi/PVC markers (Figure 3e,g). This probably reflects the gradual passage of VSR molecules from the TGN to the PVC via a maturation event, as discussed previously (Niemes et al., 2009).

Figure 3.

 SNX knockdown inhibits ER export of de novo synthesized VSR reporter.
(a–j) Transient expression in tobacco protoplasts. Effect of SNX RNAi expression on distribution of the Golgi marker Man1–RFP (red) and the PVC marker GFP–BP80 (green) in a time-course experiment. Co-expression of markers alone (first column), and together with SNXCCRNAi (second column) after 6–24 h. Note: in control protoplasts (a, c, e, g), GFP–BP80 is in close proximity to the Golgi marker Man1–RFP at 6 and 12 h, and gradually separates from the Golgi marker (20 and 24 h). Separation of the Golgi and PVC markers is delayed when the SNX RNAi construct is expressed (b, d), and an ER pattern for GFP–BP80 becomes visible after 20 h, and is very prominent after 24 h of expression (f, h).
(i, j) Inhibiting protein synthesis prevents the appearance of GFP–BP80 in the ER. Control (i) and SNX RNAi-expressing (j) protoplasts were incubated with the protein synthesis inhibitor cycloheximide for 12 h after 12 h of expression without the inhibitor.
Insets show a higher magnification of regions of interest. Scale bars = 5 μm.

In sharp contrast to the control, the PVC marker GFP–BP80 started to accumulate additionally in the ER after 20 h of incubation in protoplasts expressing the SNXCCRNAi construct (Figure 3f,h). However, punctate GFP–BP80-positive structures also remained visible. We have previously shown that expression of this construct for 16 h drastically reduces the amount of endogenous SNXs, and, as a consequence, GFP–BP80, although present in a large number of Golgi-independent punctate structures, did not reach the ARA6-positive, wortmannin-sensitive PVC (Niemes et al., 2009). Therefore, some of the remaining punctate structures seen after 20 h expression of the SNXCCRNAi construct might also represent TGNs, rather than PVCs. As shown previously (Figure 2a–c), the Golgi signal remained unaffected even after 24 h of expression (Figure 3h). When the protein synthesis inhibitor cycloheximide was added to the protoplasts after 12 h of expression, followed by incubation for a further 12 h (Figure 3i,j), the GFP–BP80 signals remained punctate and ER accumulation of the VSR reporter did not occur in the protoplasts expressing the SNXCCRNAi construct (compare Figure 3j with Figure 3h). In addition, it also appeared that a large proportion of the GFP–BP80 signals remained immediately adjacent to the Golgi marker signal (Figure 3j), and this was not quite as evident in the control (Figure 3i). This indicates that the appearance of GFP–BP80 in the ER is the result of newly synthesized reporter molecules that accumulate there because they cannot be exported. This inhibition appears to be specific for the VSR reporter, as ER export of the fluorescent Golgi marker was not perturbed under these conditions.

Co-expression of SNX mutants or RNAi knockdown of the SNXs also traps soluble vacuolar cargo in the ER

As serious perturbation of retromer function ultimately caused an accumulation of the VSR reporter GFP–BP80 in the ER, we assessed the fate of two fluorescently tagged vacuolar cargo molecules (GFP–sporamin and aleurain–GFP). Irrespective of the strategy employed (co-expression of SNX1/SNX2a mutants or RNAi knockdown), both BP80 ligands followed the fate of the receptor (Figures 1 and 2) and accumulated in the ER (Figure 4c–h) rather than in the lytic vacuole (Figure 4a,b). Thus, irrespective of the strategy used to completely inhibit sorting nexin function, such inhibition results in the accumulation of both receptor and ligands in the ER.

Figure 4.

 Complete retromer inhibition results in accumulation of VSR ligands in the ER.
(a–h) Transient expression in tobacco protoplasts.
(a, b) Expression of vacuolar cargo molecules alone shows a vacuolar pattern for GFP–sporamin (a) and aleurain–GFP (b).
(c–f) Effects of co-expression of the two mutants SNX1-ΔN and SNX2a-ΔN (c, d) or SNX1-ΔCC and SNX2a-ΔCC (e, f) on the transport of GFP–sporamin (c, e) and aleurain–GFP (d, f). Note: in all cases, cargo molecules were trapped in the ER.
(g, h) Effects of SNX RNAi knockdown on the transport of GFP–sporamin (g) and aleurain–GFP (h). Note: The cargo molecules were trapped in the ER.
Scale bars = 5 μm.

ER-localized derivatives of BP80 do not cause inhibition of ER export

Complete inhibition of retromer-mediated transport always resulted in accumulation of the VSR reporter GFP–BP80 and soluble vacuolar cargo in the lumen of the ER, without affecting COPII-mediated transport of Golgi markers or the soluble secretory reporter α-amylase. This raised the question as to why soluble vacuolar cargo is capable of accumulating to high levels in the ER against the bulk flow. We hypothesized that this is due to interaction with the endogenous functional VSR BP80, as the VSR reporter construct lacks the entire luminal ligand-binding domain (LBD) of the functional VSR (Figure 5a) and therefore cannot interact with these cargo molecules. However, this also raised the question as to whether such an interaction occurs at all in the ER, as it is commonly believed that VSRs bind their ligands in the TGN. We therefore hypothesized that, if VSR–ligand interaction does indeed occur in the ER, an ER-localized functional VSR should also lead to ER accumulation of soluble vacuolar cargo molecules. To prove this hypothesis, we constructed various derivatives of the VSR BP80 that would remain in the ER (Figure 5a).

Figure 5.

 Molecular tools to localize VSR–ligand interaction.
(a) Schematic illustration of the VSR BP80 and calnexin, the fusion proteins GFP–BP80 and GFP–CNX, the chimeras BP80–CNX and BP80–CNX–cerulean and the cerulean-tagged full-length BP80. Domains of BP80 are shown in red; those of calnexin are shown in blue. The fluorophore is indicated as XFP. Abbreviations: C, cytosol; L, lumen; LBD, luminal ligand-binding domain; TMD, transmembrane domain; CT, cytosolic tail; ceru, cerulean; FL, full-length)
(b–j) Transient expression in tobacco protoplasts.
(b–g) Co-expression of BP80–CNX–cerulean (green) and markers (red) for the ER (p24–RFP) (b–d) and the Golgi (Man1–RFP) (e–g). Note: The cerulean-tagged BP80–CNX chimera consisting of the cytosolic tail and transmembrane domain of calnexin with the luminal binding domain of BP80 is localized to the ER and has no affect on the distribution of the Golgi marker Man1–RFP.
(h–j) Co-expression of BP80FL–cerulean (green) and the ER marker p24–RFP (red). Note: cerulean-tagged full-length BP80 localizes to the ER.
Scale bars = 5 μm.

It was previously shown that residence of the chaperone calnexin (CNX) in the ER is due to its transmembrane domain (TMD) and cytosolic tail (CT). The GFP–calnexin–TMD/CT fusion (GFP–CNX) is nowadays an established ER marker for live-cell imaging analysis (daSilva et al., 2005). Based on these observations, we fused the sequence encoding the LBD of BP80 to the sequence encoding the TMD/CT of calnexin (BP80–CNX) in order to anchor the LBD of BP80 in the ER membrane. For live-cell imaging, we added a fluorescent tag by fusing the coding sequence of either cerulean (BP80–CNX–cerulean) or mKate (BP80–CNX–mKate) to the CT of the chimera. We also fused the cerulean coding sequence directly to the C-terminus of the full-length BP80 coding sequence (BP80FL–cerulean). We speculated that, in this construct, any ER export signal within the extremely short tail (35 amino acids) of BP80 would be shielded by addition of the 239 amino acids of the fluorophore.

Co-expression with the ER marker p24–RFP or the Golgi marker Man1–RFP in tobacco protoplasts shows that the BP80–CNX–cerulean chimera exhibits the expected ER localization (Figure 5b–g). This is also true for the BP80FL–cerulean construct (Figure 5h–j), and supports the notion that addition of a fluorescent protein might also have severe consequences for the tagged molecule in the in vivo situation.

It has been previously reported that use of fluorescent membrane proteins such as GFP–calnexin as marker molecules for the ER facilitates the formation of enlarged ER cisternae, but does not affect ER functionality (Irons et al., 2003; Runions et al., 2006). As expression of the ER-anchored VSR derivatives also appeared to cause this phenotype, we wished to eliminate any possibility that these constructs might influence trafficking pathways, e.g. leading to a general accumulation in the ER. Although no apparent effect on the transport of the Golgi marker Man1–RFP was observed during expression of the ER-localized BP80–CNX–cerulean (compare Figure 5e–g with Figure 1b), we also analysed the influence of all ER-localized BP80 derivatives on the ER export of the quantifiable secretory reporter α-amylase. If strong accumulation of the BP80 derivatives were to inhibit ER export in general, one would expect to see a drastic decrease in the secretion index of the amylase reporter. This is observed upon inhibition of COPI/II-mediated transport caused by the drug brefeldin A or expression of COP-specific GTPase mutants (Pimpl et al., 2003). However, as Figure 6(a,b) shows, none of the BP80 chimeras caused a drastic reduction in amylase secretion, demonstrating that a general inhibition of ER export did not occur.

Figure 6.

 Biochemical characterization of ER-anchored VSRs.
(a, b) Transport analysis in tobacco protoplasts using the reporter α-amylase alone (first column) and co-expressed with BP80–CNX–cerulean, BP80–CNX–mKate and BP80FL–cerulean in various amounts, given in μg plasmid DNA as indicated on the x axis. Error bars are the SD of five independent experiments. Statistical analysis was performed using a homoscedastic two-sided t test; the data show no significant differences compared to the control.
(c) Transport analysis in tobacco protoplasts using the vacuolar reporter α-amylase–sporamin alone and untreated (first column) or treated with 10 μm wortmannin for 20 h after electroporation (all following columns). BP80wt, PVC marker GFP–BP80, BP80FL–cerulean and the chimera BP80–CNX were expressed in various amounts, given in μg plasmid DNA as indicated on the x axis. Note: wortmannin dramatically increases the secretion index of α-amylase–sporamin, GFP–BP80 enhances the secretion index even more, but BP80wt, BP80FL–cerulean and BP80–CNX significantly weaken the affect of wortmannin. Error bars are the SD from five independent experiments. Statistical analysis was performed using a homoscedastic two-sided t test. First, data were combined to blocks. Different capital letters indicate significant differences, the same capitals indicate similarity. Second, the t test was performed for each experiment in comparison with the 10 μm Wortmannin data. Asterisks indicate significance
(*< 0.05; **< 0.01; n.s., no significance).

In terms of transport, the VSR reporter GFP–BP80 has previously been shown to compete with endogenous VSRs. As the VSR reporter lacks ligand-binding capabilities, transport of the VSR ligand α-amylase–sporamin is inefficient when the VSR reporter is expressed, and results in induced secretion of the ligand to the cell surface (daSilva et al., 2005). These authors also demonstrated that, under these conditions (as well as in the presence of the drug wortmannin), the transport of ligands is limited by the availability of endogenous VSRs, as co-expression of wild-type receptors partially reduced the transport of the vacuolar-targeted ligands to the cell surface (daSilva et al., 2005). We have adapted this elegant in vivo assay to determine whether our ER-anchored BP80 constructs are capable of ligand binding. We induced a VSR-limiting situation by applying the drug wortmannin, resulting in detectable secretion of α-amylase–sporamin. Under these conditions, co-expression of molecules capable of binding vacuole-destined ligands will reduce the wortmannin-induced secretion, as has been shown for the wild-type VSR (daSilva et al., 2005). Figure 6(c) shows the induced secretion of α-amylase–sporamin after wortmannin incubation (10 μm) compared to the control. This wortmannin background was kept constant in all samples to determine the influence of BP80FL–cerulean and BP80–CNX compared to the VSR BP80 or the non-ligand-binding VSR reporter GFP–BP80 as standards. As Figure 6(c) shows, neither BP80FL–cerulean nor BP80–CNX additionally enhance the secretion of α-amylase–sporamin as seen for GFP–BP80, and instead alleviate the wortmannin-induced secretion as seen with the wild-type receptor. Therefore, we conclude that both ER-anchored VSR derivatives possess ligand-binding capability.

ER export of soluble vacuolar cargo is inhibited by ER-anchored BP80

We next investigated whether the ER-resident VSR derivatives could cause ER retention of soluble vacuolar cargo molecules, similar to that observed upon complete inhibition of retromer-mediated recycling (Figure 4). Co-expression of BP80–CNX–cerulean with either aleurain–GFP or GFP–sporamin in tobacco protoplasts yielded accumulation of these vacuolar cargo molecules in the ER (Figure 7). Co-expression of BP80–CNX–mKate with GFP–sporamin demonstrated that this effect was independent of the fluorophore (see Figure S3). The typical vacuolar location of these BP80 ligands (compare Figure 4a,b) was never observed when BP80–CNX–cerulean was detected in the ER (Figure 7). To demonstrate BP80–CNX–cerulean-induced ER localization of the ligand aleurain–GFP (Figure 7a–e), the ER marker p24-RFP was coexpressed in addition and the merged image of these signals is shown in Figure 7d. To rule out the possibility that ER retention of the cargo molecules was due to over-production of the ER marker, additional control experiments were performed to demonstrate that aleurain–GFP exhibits its natural vacuolar pattern even if p24–RFP is co-expressed (Figure 7f). Although the fluorophore of the BP80 derivatives faced the cytosol, we also wished to ensure that the fluorophores did not influence the interaction of the LBD with the ligands. Therefore, we also co-expressed the untagged BP80–CNX chimera with either aleurain–GFP (Figure 8d–f) or GFP–sporamin (Figure 8j–l) together with the Golgi marker Man1–RFP as an internal control for unperturbed COPII-mediated ER export. Again, none of the ligands exhibited a typical vacuolar localization compared to the controls (Figure 8a–c,g–i), whilst the Golgi marker was not affected under these conditions.

Figure 7.

 ER-localized BP80 results in accumulation of ligands in the ER.
(a–i) Transient expression in tobacco protoplasts.
(a–f) Co-expression of the vacuolar cargo marker aleurain–GFP and the ER marker p24–RFP alone (f) and with BP80–CNX–cerulean (a–c). (d) Merged image of aleurain–GFP (b) and p24–RFP (c) if co-expressed with BP80–CNX–cerulean. (e) Merged image of BP80–CNX–cerulean (a), aleurain–GFP (b) and p24–RFP (c).
(g–i) Co-expression of the vacuolar cargo marker GFP–sporamin and BP80–CNX–cerulean.
Scale bars = 5 μm.

Figure 8.

 ER accumulation of VSR ligands does not cause collapse of ER export.
(a–i) Transient expression in tobacco protoplasts.
(a–f) Co-expression of the vacuolar cargo marker aleurain–GFP and the Golgi marker Man1–RFP alone (a–c) or together with the chimera BP80–CNX (d–f).
(g–l) Co-expression of the vacuolar cargo marker GFP–sporamin and the Golgi marker Man1–RFP alone (a–c) and together with the chimera BP80–CNX (d–f).
Note: Both vacuolar cargo markers show an ER distribution if co-expressed with BP80–CNX, but the Golgi marker is not affected.

Even though the results obtained for BP80–CNX and BP80–CNX–cerulean were indistinguishable, these derivatives represent chimeric proteins that consist of the BP80 LBD and the TMD/CT of calnexin. To rule out the possibility that the TMD/CT of calnexin influences the function of the LBD of BP80 in addition to the altered localization, we also tested the influence of BP80FL–cerulean on the transport of both ligands and the Golgi marker. Figure 9(a–h) shows that expression of BP80FL–cerulean also led to an ER accumulation for both ligands, without any recognisable perturbation of trafficking of the Golgi marker, as just seen with the BP80–CNX fusions. These results suggest that the presence of the LBD of BP80 in the lumen of the ER is capable of accumulating ligands in the lumen of the ER despite the ongoing export of Golgi markers and secretory cargo molecules. However, we also wished to test whether the ER-anchored VSR derivatives exhibit a general block of vacuolar transport. To address this question, we co-expressed BP80–CNX–cerulean or BP80FL–cerulean with the VSR reporter GFP–BP80 (Figure 9i–l). In both cases, the VSR reporter was found to localize in a typical punctate PVC pattern, and was not affected by the presence of the ER-anchored VSR derivatives. This clearly shows that ER export of vacuolar cargo molecules is not generally blocked. Therefore, the accumulation of VSR ligands in the ER is due to their specific interaction with the ER-anchored VSRs, which trap them in the ER against COPII-mediated bulk flow and ongoing vacuolar transport.

Figure 9.

 Over-production of full-length BP80 does not interfere with COPII-mediated transport.
(a–t) Transient expression in tobacco protoplasts.
(a–h) Co-expression of the vacuolar cargo markers aleurain–GFP (a–d) or GFP–sporamin (e–h) and the Golgi marker Man1–RFP together with BP80FL–cerulean. Note: both vacuolar cargo markers show an ER distribution if co-expressed with BP80FL–cerulean, but the Golgi marker is not affected.
(i–l) Co-expression of the ER-anchored VSR derivatives BP80–CNX–cerulean (i, j) or BP80FL–cerulean (k, l) with the VSR reporter GFP–BP80. Note: the ER-anchored VSR derivatives do not cause a general block of the route to the PVC, as they do not influence transport of the VSR reporter.
(m–t) Expression of the ER marker GFP–HDEL (m, q), the vacuolar cargo molecule aleurain–GFP and Man1–RFP (n, r), the PVC marker GFP–BP80 (o, s) and the TGN marker YFP–SYP61 (p, t) in control protoplasts (m–p) or co-expressed with BP80wt (q–t). Note: BP80wt changes the distribution of aleurain–GFP and GFP–BP80 into an ER pattern, whilst ER, Golgi and TGN markers show the same distribution as in control protoplasts.

BP80 does not compete for COPII-mediated ER export

The results presented here explain the accumulation of VSR ligands in the ER upon complete inhibition of retromer recycling. However, the question remains as to why VSRs are inhibited from leaving the ER, even though a Golgi marker remains unaffected. A possible explanation is that the VSR utilizes different transport machinery. It has previously been shown that, in terms of transport, non-ligand-binding GFP–BP80 is capable of competing with the wild-type VSR (daSilva et al., 2005). In this case, identical cytosolic tails compete with each other. As a consequence, a proportion of competing cargo molecules will accumulate while awaiting the next carrier. To identify putative competitors for the VSR, we co-expressed the wild-type VSR with GFP–HDEL, a marker for COPII export and COPI retrieval, the ligand aleurain–GFP, the Golgi marker Man1–RFP, the known competitor GFP–BP80 and the TGN marker YFP–SYP61 (Figure 9q–t). Compared to the controls (Figure 9m–p), only localization of GFP–BP80 and aleurain–GFP were altered into an ER pattern, indicating competition for ER export but all other putative COPII cargos remained unaffected. Therefore, we propose that the ER export of vacuolar cargo does not occur via the COPII machinery (Figure 10).

Figure 10.

 Receptor-mediated protein sorting to the lytic vacuole in plants.
The secretory pathway starts at the ER. Vacuolar proteins are recognized by the VSR BP80 in the lumen of the ER, and the receptor–ligand complexes exit the ER (blue arrow) and reach the TGN where they dissociate. BP80 receptors are then recycled back to the ER by retromer-dependent transport (blue arrow). These two transport routes for the receptor BP80 are independent of COPI/II-mediated transport (red arrows). Vacuolar cargo molecules continue to move towards the lytic vacuole in a receptor-independent manner. Abbreviations: N, nucleus; PM, plasma membrane; PVC, pre-vacuolar compartment; LV, lytic vacuole.

Discussion

Inhibition of retromer function affects ER export of VSRs and their ligands

We have previously identified the TGN as being the location at which retromer-mediated recycling of VSRs occurs. When retromer function is perturbed by expression of mutants of sorting nexins 1 or 2a (SNX1/2a), VSRs accumulate at the TGN (Niemes et al., 2009). However, when both SNXs are inhibited simultaneously, either by co-expression of SNX1/2a mutants or RNAi knockdown, VSRs accumulate in the ER instead. At first sight, these different effects on localization of the VSRs are difficult to reconcile with inhibition of a single transport step mediated by only one carrier. However, the time-course analysis of RNAi-induced knockdown of SNX function revealed spatio-temporal effects: VSRs initially accumulate at the TGN, and then, at later time points (20–24 h), in the ER. Use of the protein synthesis inhibitor cycloheximide allowed us to analyse the phenomenon of ER accumulation of the VSR reporter GFP–BP80. As use of this drug does not alter the presence of previously synthesized and post-Golgi-localized VSRs, the prevention of ER accumulation of VSRs under conditions of SNX knockdown can only be explained by an inhibition of ER export. Such inhibition of ER export appeared to be specific to VSRs and soluble vacuolar cargo molecules, i.e. the ligands of the VSRs, and did not have an effect on ER export of the secretory bulk-flow marker α-amylase or fluorescent Golgi markers. Therefore, ER retention of these soluble ligands against the bulk flow can only be explained by retention of the respective VSRs in the ER.

VSR–ligand interactions occur in the lumen of the ER

Two previous sets of observations indirectly suggested that vacuolar cargo molecules bind to BP80-type receptors in the ER. Firstly, when a construct consisting of the luminal ligand-binding domain (LBD) of BP80 tagged with the ER retrieval motif HDEL (BP80LBD–HDEL) was expressed in Arabidopsis, the transport of vacuolar enzymes was inhibited and instead they were found to be present in greater amounts in ER fractions (Watanabe et al., 2004). A similar result was obtained by daSilva et al. (2005), who also demonstrated microsomal retention of a chimeric vacuolar reporter construct (α-amylase–sporamin) when BP80LBD–HDEL was expressed in tobacco protoplasts. Under these conditions, wortmannin, which normally causes the secretion of the α-amylase–sporamin construct, had no such effect.

Although this elegant assay unequivocally identified α-amylase–sporamin as a BP80 ligand, it did not pinpoint the location of the receptor–ligand interaction, as BP80LBD–HDEL cycles between the ER and the Golgi via the HDEL receptor ERD2, which is distributed uniformly across the stack (Boevink et al., 1998). For this reason, we prepared an ER-anchored derivative of BP80 (BP80–CNX) to analyse VSR–ligand binding in the lumen of the ER. This was achieved by exchanging the TMD and the CT of the VSR for those of the ER-resident chaperone calnexin, which was shown to be sufficient to confer ER retention of fusion proteins such as the commonly used ER marker GFP–calnexin (Irons et al., 2003). Expression of this construct, both as fluorescently tagged and untagged versions, led to ER retention of vacuolar cargo. This allowed direct proof of the ER localization of receptors, and additional monitoring for fluorescent ligands and the Golgi marker.

In an alternative approach, ER retention of a full-length VSR was achieved simply by fusing cerulean to the CT of the receptor (BP80FL–cerulean). This construct also conferred ER localization of the ligands, but the LBD of the VSR is not deprived of its natural context. However, this drastic alteration of VSR localization due to (X)FP-tagging also demonstrates the risk involved in using this technique for newly discovered proteins of unknown function. It should also be noted that none of these ER-localized VSR derivatives blocked the transport route from the ER towards the PVC in a general way, as judged by co-expression with the PVC marker GFP–BP80. Therefore, we conclude that the ligands were trapped in the ER due to a specific interaction with the LBD of the ER-anchored VSRs, rather than non-specific blockage of either the vacuolar route or the COPII-mediated bulk flow.

The second piece of evidence in favour of the ER as the compartment where BP80–cargo interactions occur comes from studies on the ligand-binding properties of VSRs. The luminal tail of BP80 has three EGF repeats, one of which shows high-affinity calcium binding (Cao et al., 2000). Calcium binding stabilizes the receptor–ligand complex, even at low pH. Indeed, it would appear that calcium concentration is more important for ligand binding than pH (Watanabe et al., 2002). In this respect, the ER is an ideal location for the initial receptor–ligand interaction, as calreticulin, one of the most abundant molecules in the lumen of the ER, also confers calcium homeostasis of this compartment (Christensen et al., 2008). Furthermore, it is known that – at least in mammalian cells – the concentration of Ca2+ is generally much higher in the ER than in the Golgi apparatus (Montero et al., 1997; Pinton et al., 1998). Although reliable data for calcium concentrations in the plant TGN are not available, if they are lower than in the ER, this could well be the cause of ligand dissociation in this post-Golgi compartment.

These results put earlier data on PV72, a BP80-related isomer in maturing pumpkin seeds, in another light. PV72 binds to an internal peptide sequence of the storage protein 2S albumin (Shimada et al., 1997), and is present in the limiting membrane of so-called PAC (precursor-accumulating vesicles) (Hara-Nishimura et al., 1998). Interestingly, these large (200–400 nm diameter) vesicles bud off the ER, and fuse directly with the vacuole, thus bypassing the Golgi apparatus (Hara-Nishimura et al., 1993). PV72 is therefore a special but very convincing example of a BP80-type VSR interacting with its ligand in the ER.

Retromer-mediated recycling leads back to the ER

Despite morphological differences across the kingdoms, the mechanism of receptor-mediated sorting follows a common principle: ligand recognition in a donor compartment, transport of receptor–ligand complexes to the acceptor compartment, dissociation of these complexes, and finally recycling of the receptors back to the donor compartment for further rounds of transport. We have positively identified the TGN as the starting point of the retromer route (Niemes et al., 2009), and have shown that the VSR–ligand interaction is initiated in the lumen of the ER in vivo. However, inhibition of retromer function also specifically inhibited the ER export of VSRs and their ligands, without perturbing COPII-mediated transport. As VSRs accumulate in the ER under these conditions, the possibility that this export block is simply due to a lack of VSRs can be ruled out. Therefore, one has to assume that other components of the trafficking machinery become limiting. This is also supported by the observed inhibition of GFP–BP80 transport caused by expression of the untagged, full-length VSR. These molecules have been shown to compete with each other for transport (daSilva et al., 2005). However, the VSR did not compete with the transport of markers for the Golgi or the TGN. Therefore, we postulate that ER export of VSRs occurs via a different transport mechanism than COPII (Figure 10). According to this hypothesis, inhibition of retromer-mediated transport would cause the corresponding ER export route to cease, without affecting COPII trafficking. A similar effect has been reported previously for COPI- and COPII-mediated transport between the ER and the Golgi. In this bi-directional vesicle trafficking system, the COPII-mediated route collapses when the COPI route is perturbed (Pimpl et al., 2003; Stefano et al., 2006). The molecular mechanism by which this occurs remains unclear, but a loss of SNARE recycling could be the cause. Inhibition of retromer did not cause a collapse of the COPII-mediated route; therefore, it is assumed that COPI-mediated recycling from the Golgi was also not perturbed. This suggests that the limitation at the ER leading to VSR accumulation cannot be compensated for by the still functional COPI-mediated transport. Proteins such as the p24 proteins which are known to recycle from the Golgi via COPI-mediated transport, possess a di-lysine signal (Langhans et al., 2008) that is absent in the CT of the BP80 family members. Therefore, we postulate that retromer-mediated transport of the VSRs leads directly to the ER (Figure 10), and that the VSRs and their vacuolar cargos leave the ER by a route other than the COPII-dependent Golgi marker and the secreted proteins. The mechanism and target of this route remain to be identified.

Experimental procedures

Plant material

Tobacco plants (Nicotiana tabacum var. SR1) were grown as previously described (Pimpl et al., 2006). Suspension cultures of Arabidopsis thaliana var. Landsberg erecta PSB-D were cultivated as described previously (Miao and Jiang, 2007) and analysed 3 days after sub-culturing.

Recombinant plasmid production

Arabidopsis coding sequences were amplified by PCR from first-strand cDNA prepared from 3-day-old seedlings (Pimpl et al., 2003), using the oligonucleotides shown in Figure S4. Recipient plasmids were cut according to the restriction sites of the fragments, and dephosphorylated prior to ligation.

The plasmids encoding markers/reporters have been described previously as indicated: ST–YFP (Brandizzi et al., 2002), Man1–RFP (Nebenfuhr et al., 1999), GFP–BP80 and GFP–sporamin (daSilva et al., 2005), aleurain–GFP (Humair et al., 2001), α-amylase (pAmy), α-amylase–sporamin (Pimpl et al., 2003), p24–RFP and GFP–HDEL (Langhans et al., 2008) and YFP–SYP61 (Niemes et al., 2009). The following sorting nexin mutants were used: SNX1-ΔN, SNX1-ΔCC, SNX2a-ΔN, SNX2a-ΔCC and the RNAi construct SNXCCRNAi (Niemes et al., 2009).

For generation of the untagged BP80–CNX (pML1) construct, the backbone was obtained from GFP–CNX (daSilva et al., 2005) and the 35S promoter was extracted from SNX2aM1 (pSN6) (Niemes et al., 2009). The LBD of BP80 was amplified from wild-type BP80 (daSilva et al., 2005). The backbones of fluorophore-tagged BP80–CNX–mKate and BP80–CNX–cerulean were obtained from RTLN–mKate (pIC3) (Langhans et al., 2009) or SNX1–cerulean (pSN18) (Niemes et al., 2009). The BP80–CNX fragment was amplified from pML1. The BP80FL–cerulean (pML2) construct was ligated with the backbone from SNX1–cerulean (pSN18) (Niemes et al., 2009). Full-length BP80 was obtained from wild-type BP80 (daSilva et al., 2005).

Isolation of protoplasts and transient gene expression

Protoplasts were isolated from Arabidopsis suspension-cultured cells 3 days after sub-culturing. The cells were centrifuged at 80 g for 10 min and resuspended in TEX buffer (see Foresti et al., 2006) supplemented with 0.2% macerozyme R10 and 0.4% cellulase R10 (Yakult Honsha Co. Ltd, http://www.yakult.co.jp), and incubated at 25°C for 20 h. Protoplasts were washed four times with electroporation buffer (see Bubeck et al., 2008) by flotation at 80 g for 10 min followed by removal of the liquid medium using a peristaltic pump. Mesophyll protoplasts were isolated from leaves of 6–8-week-old tobacco plants as previously described (Bubeck et al., 2008). Unless otherwise stated, 1–30 μg of plasmid DNA was transfected and expressed for 20 h.

Secretion assays

α-amylase activity was assayed in culture medium and cells using a Megazyme α-amylase reagent kit (http://www.megazyme.com). The secretion index was calculated as the ratio of secreted activity compared with cellular activity as described by Bubeck et al. (2008).

Confocal microscopy and immunofluorescence labelling

Imaging was performed using a Zeiss Axiovert LSM 510 Meta confocal laser scanning microscope with a C-Apochromat 63 x/1.2 W corr water immersion objective (http://www.zeiss.com/). Main beam splitters (HFT) 458/514 and 488/543 were used. The following fluorophores (excited and emitted by frame switching in the multi-tracking mode) were used: GFP (488 nm/496–518 nm), YFP (514 nm/529–550 nm), RFP (543 nm/593–625 nm), cerulean (458 nm/464–486 nm) and m-Kate (543 nm/614–646 nm). Pinholes were adjusted to 1 Airy unit for each wavelength. Protoplasts were analysed accordingly. Post-acquisition image processing was performed using the Zeiss LSM 510 image browser (4.2.0.121) and CorelDrawX4 (14.0.0.567; http://www.corel.com).

Acknowledgements

The financial support of the Deutsche Forschungsgemeinschaft (grants PI 769/1-1 and RO 440/11-3/14-1) is gratefully acknowledged.

Ancillary