The secretory pathway of eukaryotic cells comprises a network of organelles that connects three large membranes, the plasma membrane, the vacuole and the endoplasmic reticulum. The Golgi apparatus and the various post-Golgi organelles that control vacuolar sorting, secretion and endocytosis can be regarded as intermediate organelles of the endocytic and biosynthetic routes. Many processes in the secretory pathway have evolved differently in plants and cannot be studied using yeast or mammalian cells as models. The best characterized organelles are the Golgi apparatus and the prevacuolar compartment, but recent work has shed light on the role of the trans Golgi network, which has to be regarded as a separate organelle in plants. In this study, we wish to highlight recent findings regarding the late secretory pathway and its crosstalk with the early secretory pathway as well as the endocytic route in plants. Recently published findings and suggested models are discussed within the context of known features of the equivalent pathway in other eukaryotes.
In contrast to prokaryotes, eukaryotic cells employ an internal membrane system that controls the secretion of biomolecules and also mediates uptake of substances and delivery to specific intracellular locations. It offers a practical solution to the much lower surface/volume ratio of the larger eukaryotic cells and permits specialized biochemical reactions to occur in distinct cell compartments. This endomembrane system, in this study termed the secretory pathway, evolved early in the history of life on earth (1), and its presence is one of the key features of eukaryotes.
Compartments are interconnected by vesicle transport (2), but they can also change composition, gradually mature over time and migrate within a cell. Some organelles can serve as short-term or long-term storage compartments, whereas others function as cross-junctions and show very low steady-state levels of cargo molecules that pass through them. The minimal secretory pathway consists of the nuclear envelope (NE) and endoplasmic reticulum (ER), the Golgi apparatus, various post-Golgi intermediate compartments, the vacuoles/lysosomes and the small vesicular transport carriers that shuttle between these compartments (Figure 1). One can think of the NE/ER membrane, the plasma membrane and the vacuolar compartments as the three corners of the endomembrane universe, while all other compartments are intermediates. This basic set-up has been elaborated and adapted to the individual needs in different organisms and cell types and can thus contain further compartments and specific differentiations that are not conserved among eukaryotic kingdoms (1). This divergence is mostly noticeable for the post-Golgi compartments, but even the early secretory pathway exhibits specialization at the ER–Golgi interface that must have occurred after the last common eukaryotic ancestor.
When considering its biosynthetic role in manufacturing and delivering lipids, proteins and polysaccharides, the ER is the first organelle, followed by the Golgi apparatus and a range of further intermediate compartments en route to either the lysosomes/vacuoles or the plasma membrane. This transport direction is usually referred to as anterograde or biosynthetic transport, and the Golgi apparatus plays a central role in sorting molecules to the two major end locations or to recycle essential components back to the ER.
However, the same pathway can be considered from the opposite perspective of internalizing plasma membrane proteins and associated ligands to intracellular compartments (3,4). In this case, the plasma membrane is the starting membrane and through endocytosis or phagocytosis, material from outside the cell can reach the intermediate post-Golgi compartments. From there, retrograde traffic will ultimately reach either the lytic organelles or the ER (5), but by analogy to the early biosynthetic pathway, a recycling route also exists that returns essential components rapidly back to the plasma membrane (6). The route from the plasma membrane to the lytic compartment is used to downregulate cell surface receptors, while the route to the ER is misused by some bacterial toxins to ultimately enter the cytosol. In addition to certain toxins, pathogens such as viruses, bacteria and protozoa also use the endocytic route to enter eukaryotic cells and often manipulate trafficking machinery of the host cell to satisfy their own purpose (4).
As illustrated in Figure 1, some, if not all, of the organelles of the secretory pathway can be shared by both transport routes. Because of intense research on entry of pathogens and endocytosis of cell surface receptors in mammalian cells, post-Golgi intermediate organelles are now widely called ‘endosomes’. In plants and yeast, the secretory pathway has been studied mostly from the biosynthetic perspective, and the term prevacuole or prevacuolar compartment (PVC) is used for at least one of this highly plastic and dynamic group of organelles (7).
An important principle in the transport between individual compartments of the secretory pathway is that many steps are cyclic by nature. This reversibility has been recognized for a long time (5) and permits recycling of membranes and machinery components to maximize efficiency. However, as a consequence of this complexity, many transport steps depend on the reciprocal recycling route, and gene function is difficult to assign by analysis of loss-of-function mutants alone because they often exhibit pleiotropic effects, the most noticeable of which may not necessarily be the primary. It is extremely difficult to distinguish between primary, secondary or indirect consequences of the gene defects such as developmental disorders. Therefore, a better understanding of the mechanisms mediating proper segregation of cargo at the various cross-junctions in this pathway remains a major challenge in cell biology and will require diverse and complementary experimental approaches.
In plants, biosynthetic transport through the secretory pathway has been intensely investigated and includes a large portfolio of characterized protein sorting signals, receptors as well as regulatory components that control individual transport steps within the pathway (8). In contrast, experimental work on the endocytic route has been hampered by a lack of suitable markers and practical limitations to experimental approaches, but recently, some of the early work has been taken to new frontiers.
Unique Features of the Plant ER–Golgi System
The events comprising protein biosynthesis and subsequent export from the ER are perhaps the most conserved among kingdoms. Yet, a number of plant-specific features have been identified that justify further analysis. In plant cells, the ER forms a highly dense network of thin tubules that undergo constant remodelling and include highly mobile regions as well as fixed points at the level of the plasmodesmata where the ER forms ‘wormholes’ to adjacent cells (9–11). Tubular ER is occasionally connected to cisternal ER sheets that are variable in size and more pronounced in cells exhibiting high secretory capacity. In vivo imaging studies revealed that individual Golgi stacks are highly mobile and move along the ER tubules (9). Because of this high mobility, the subject of protein trafficking between the ER and the Golgi stacks has particularly fascinated the field.
An important feature of the plant secretory pathway is that in addition to its role in protein sorting and protein modification, the Golgi apparatus plays a central role in the biosynthesis of cell wall polysaccharides (12). The cell wall is a rigid but yet continuously remodelled structure and requires organized delivery of de novo synthesized components over the entire plasma membrane surface. Perhaps for this reason, the plant Golgi apparatus is distributed as individual stacks continuously migrating over the tubular cortical ER network (9) and in close association with the so-called ER export sites (ERES) (13).
In yeast, formation of vesicles transporting cargo from the ER to the Golgi depends on the guanosine triphosphatase (GTPase) Sar1 that controls recruitment of the peripheral membrane protein Sec16 (14–16) and a coat protein complex [termed coat protein (COP)II] from the cytosol to shape a coated vesicle (17–20). Recent evidence suggests that the coat also marks the vesicle for fusion with the correct target membrane because a subunit of the transport protein particle I complex that receives COPII vesicles at the Golgi membrane binds the COPII coat component Sec23 (21). This unexpected finding explains why Sec23 overproduction strongly inhibits anterograde ER-to-Golgi transport (22) and is toxic to wild-type cells (23).
In plants, the GTPase Sar1 or COPII components can be used to visualize ERES, specialized ER subdomains that form nucleation points for Sar1-mediated COPII recruitment (13,24). Sec16 homologues exist in Arabidopsis thaliana (AT5G47480 and AT5G47490) but have not been analysed yet. ERES are closely associated (and move together) with individual Golgi stacks and appear to be induced by membrane but not by soluble cargo (13). Also in mammalian cells, selective cargo can influence the turnover of COPII subunits at ERES (25). Although the close apposition between ERES and Golgi bodies can be slightly different in other cell types (26), it does not look as if COPII vesicles have to engage in long-distance traffic in plants.
It should be noted that any reports on an apparent spatial separation of Golgi stacks from ER membranes are from different tissues (i.e. root meristem cells) analysed by electron microscopy (27). This technique allows the visualization of ER cisternae (or sheets) studded with ribosomes, which only form a small portion of the cortical ER and may indeed be often separated from Golgi bodies. Such studies can therefore not exclude close connections with ER tubules. The typical tubular cortical ER seen with fluorescence microscopy is probably devoid of ribosomes and will be almost impossible to make out with electron microscopy. Perhaps the recent identification of plant reticulons, a group of wedge-like membrane-embedded proteins that specifically enrich tubular ER (10) could provide immunological detection strategies of such structures by electron microscopy, but it should be noted that the diameter of tubular ER can be often so thin that it excludes most of the lumen and will thus be extremely difficult to visualize and label in sections. The plant ER–Golgi interface has been subject to a number of excellent reviews that discuss these discrepancies in observations in detail (27–29).
If COPII vesicles exist in vivo also remains to be demonstrated. A decade of intense research on the ER–Golgi interface, using immunogold electron microscopy with well-characterized and specific anti-plant COPII sera, has so far failed to demonstrate convincingly that vesicular structures play a predominant role in the ER-to-Golgi route in vivo. Instead, tubular connections between ER and Golgi cisternae have been visualized in thick sections using electron microscopy (27). This does not exclude the possibility that tubular connections can be COPII coated or that COPII is involved in the initial formation of tubules, and this will not be easy to test at the ultrastructural level using immunogold labelling, for the same reasons as discussed above. Also, any transport model has to take into account that the Golgi stacks move very fast on tubular ER, and an effective long-distance vesicular traffic appears to be difficult to imagine.
Another potentially plant-specific adaptation of the early secretory pathway is the possibility that ER export could also be COPII independent. The formation of large ER-derived transport carriers, the so-called precursor-accumulating (PAC) vesicles or KDEL-protease (KV) vesicles (30,31), is thought to constitute a Golgi-independent route to the storage vacuoles of pumpkin seeds. COPII inhibition experiments using titration of Sar1 or dominant-negative mutants of Sar1 not only demonstrated COPII dependence of bulk flow markers (32) and ER residents (33) but also revealed that some secreted proteins are resistant to the effectors (34).
A third difference that can be noted in comparison with mammalian cells is in the strategy for recycling of essential components back to the ER. In mammalian cells, the first stages after ER export appear to be just dependent on homotypic fusions of COPII vesicles to form the intermediate compartment, also referred to as vesiculo tubular clusters (VTCs) that initiate retrograde cargo sorting and control further transport to the Golgi (21,35,36). VTCs offer the first opportunity for recycling ER residents in mammals (37,38), but plants and yeasts seem to rely on specific retrieval from the individual Golgi cisternae to ensure recycling of export machinery components or to return ER residents that have escaped by bulk flow. Plants do not contain a defined intermediate compartment (29), and the Golgi stack alone must ensure proper cargo segregation. In tobacco cells, soluble proteins carrying ER retention signals are progressively depleted from cis to trans Golgi cisternae (32) and can be detected in isolated Golgi-derived retrograde transport vesicles (39). These data suggest that the plant ER–Golgi transport system is more comparable to the early secretory pathway in yeast (40) rather than in mammals.
It has been observed that bulk flow of soluble proteins is efficient in plants (32,33) and that recycling of soluble chaperones from the Golgi is a major strategy for their effective accumulation in the ER (33,39,41). From a theoretical perspective, it must be appreciated that tubular connections would permit Golgi movement along ER tubules and maintain bulk flow by providing a ‘hose-pipe’ for lumenal fluid-phase transport. In contrast, coated vesicles of 50 nm would have very little lumen in their centre if membrane-spanning proteins, sorting receptors and their ligands occupy much of the internal periphery. COPII-coated tubules would still support a degree of specificity for inclusion or exclusion of defined membrane-spanning proteins. Evidence for ER export signals has only been obtained for membrane-spanning proteins (42).
Retrograde vesicle budding depends on a similar group of COPs, called coatomer or COPI, and in contrast to COPII vesicles, COPI vesicles can be isolated from plant material (39). COPI assembly is controlled by the GTPase ADP ribosylation factor (ARF), ARF1-specific GTPase-activating proteins and nucleotide exchange factors (GTP exchange factors, GEFs). The ARF-GEF family is more complex and less conserved among plants (43), as illustrated by the differential sensitivity to the drug Brefeldin A (BFA). Like mammalian cells, plant species contain ARF-GEFs involved in Golgi-to-ER recycling that are sensitive to this drug, which can be detected by the rapid redistribution of Golgi membranes to the ER (44). But, A.thaliana roots form an exception in exhibiting a BFA-insensitive ARF-GEF that renders the plant relatively resistant to the drug with respect to transport events in the early secretory pathway (45,46). In plants, ARF1 is thought to control retrograde Golgi-derived COPI transport, which is indirectly required to maintain effective ER-derived anterograde COPII transport (47,48). The importance of the recycling principle for the ER–Golgi interface was illustrated in plants by showing that ERES integrity depends on COPI function (49).
A very exciting step forward in the study of the plant Golgi apparatus arose from a high-throughput screen for abnormal localization patterns of a well-characterized membrane-spanning fluorescent Golgi marker in transgenic A. thaliana(50). Several mutants were found in which the Golgi marker is either partially mistargeted to post-Golgi membranes or partially retained in the ER, suggesting a vast portfolio of mild sorting mutants affected in pathways from and towards the Golgi apparatus. A similar mutant screen in A. thaliana using soluble vacuolar reporter proteins has shed light on key elements in the retrograde route to the ER. The maigo2 mutant (51) identifies a putative plant homologue of yeast TIP20, a component of the Dsl1p complex for tethering COPI vesicles to the ER membrane (52–54), that was recently shown to prevent back fusion of COPII vesicles with the ER (55). The Arabidopsis maigo2 gene product interacts with the ER SNAREs Sec20 and Ufe1 in vitro(51) and may be part of the Golgi-to-ER import machinery.
It remains to be shown if the sites for COPI vesicle fusion are specialized ER subdomains (ER import sites, ERIS) and if these are closely associated with ERES. Plants form an excellent system to tackle this biological question because of the disperse nature of individual Golgi bodies spread over the cortical ER network of epidermis cells, allowing high-quality fluorescence micrographs to be recorded routinely (56). A close spatial apposition between ERES and ERIS would permit effective membrane recycling mechanisms, but it would also cause problems to segregate recycling receptors (such as ERD2) from their soluble ligands (such as the abundant chaperone BiP) to avoid that these immediately leak out again through the COPII-mediated anterograde route.
Maintenance of the Golgi Apparatus may Depend on Post-Golgi Compartments
In all kingdoms, the Golgi apparatus is responsible for the targeting of hydrolases to the lytic compartments, but plants also contain storage vacuoles that have unique functions in seeds and other tissues and are used for the accumulation of proteins, polysaccharides and lipids for a variety of applications. Although it has been a popular view that plant cells contain functionally distinct vacuoles that are supported by completely distinct trafficking routes (57–59), recent research suggests that at least Golgi-derived lytic and storage vacuoles are possibly derivatives of the same precursor vacuole with different fates in different tissues (58,60–63). While the last word has not been spoken about this issue (64–66), it is also clear that species differences as well as tissue specializations will leave plenty of room for specific adaptations of the vacuolar transport system in plants. We will return to the debate on the two vacuoles (see below) but first discuss the implications of post-Golgi traffic on the maintenance, biogenesis and polarity of the Golgi stack in plants.
One key finding that has stood the test of time is that seed storage tissues of legumins such as beans or peas produce Golgi-derived dense vesicles (DVs) that are clearly distinct from clathrin-coated vesicles or any other type of vesicle in plants. A potential cytosolic coat-driven DV budding mechanism has not been identified to date. DVs are thought to segregate from early Golgi compartments because storage proteins are progressively depleted from cis to trans cisternae, whereas the vacuolar sorting receptor BP80 is found in clathrin-coated vesicles and its concentration within the Golgi stack increased from cis to trans cisternae (67).
The receptor homology-transmembrane-RING H2 domain protein (RMR) is a good candidate for a sorting receptor for storage proteins. It was shown to recognize the C-terminal vacuolar sorting signal for the bean storage protein phaseolin (68,69) and localize to DVs (70). Interestingly, BP80 and RMR also show different intra-Golgi gradients. While BP80 is mostly concentrated on the trans Golgi cisternae, RMR shows a high concentration at the rims of the cis and medial Golgi where storage proteins are thought to aggregate and partition into DVs.
Gradients of soluble cargo molecules in Golgi stacks indicate several modes of exit from the Golgi apparatus. While the constitutively secreted cargo α-amylase is evenly distributed along the stack of cisternae, an ER-retained HDEL-tagged α-amylase derivative forms a gradient that progressively depletes this molecule from cis to trans cisternae (32). This is consistent with the model that COPI vesicles specifically retrieve retrograde cargo from Golgi cisternae and exclude bulk flow markers. As a consequence, the ratio between retrograde cargo and anterograde cargo decreases progressively, giving rise to cisternal maturation. The continuous depletion of storage proteins from cis to trans Golgi cisternae in pea cotyledons (67) suggests that also DV budding may contribute to cisternal maturation, but the vesicles do not move back to the ER but reach a storage vacuole by an as yet unknown pathway.
In theory, the same rules about selective depletion and cisternal progression are also applicable to membrane-spanning proteins. However, selective retrograde transport alone is insufficient as a mechanism to explain how the vacuolar sorting receptor BP80 accumulates in trans cisternae. A progressive increase in the absolute levels can only be explained by a dynamic equilibrium between de novo synthesis/anterograde transport and recycling from post-Golgi compartments back to the trans Golgi cisternae (see further discussion on the late secretory pathway). Therefore, the Golgi stacks cannot be just regarded as a ‘traffic jam’ that forms when exported material from the ER has to be sorted to various distal locations or to return to the ER. It is becoming increasingly evident that it is a complex and polarized structure containing gradients of molecules that transit through or reside within this organelle and that traffic from upstream as well as downstream organelles in the secretory pathway contributes to Golgi identity and possibly biogenesis.
Evidence from work with protists suggests that a proper Golgi stack may not be able to form de novo by homotypic fusions of ER-derived COPII vesicles alone (71). In Pichia pastoris, Golgi stacks have been proposed to form de novo(72,73), but experimental tools cannot rule out that post-Golgi compartments participate in forming a ‘template’ for Golgi biosynthesis. Recently, a role for clathrin in Golgi biogenesis has been proposed (74). When the Golgi apparatus was dispersed through 1-butanol treatment, an effect that is normally reversible upon wash-out, knockdown of clathrin heavy chain prevented Golgi reassembly and restoration of the normal stack formation. This would suggest that Golgi bodies cannot form de novo by ER export alone but that Golgi biogenesis depends on transported membranes from distal locations. A potential role for clathrin in Golgi biogenesis may be surprising at first. However, Golgi export could be closely linked to the various recycling routes back to the Golgi by analogy to the linkage between ER export and recycling from the Golgi (49). Therefore, recycling from post-Golgi compartments to Golgi stacks may be interrupted when anterograde clathrin-mediated transport from the Golgi is impaired.
Golgi biogenesis in plants was studied by monitoring its regeneration upon BFA treatment followed by drug wash-out (75). The authors observed that Golgi stacks initiate as clusters of ARF1-positive vesicles and undergo cisternal growth, suggesting that Golgi regeneration is COPI dependent. However, the presence of ARF1 alone is not exclusively indicative of COPI involvement because ARF1 has been detected on post-Golgi structures (76) that were devoid of COPI coats (49,77,78). ARF1 could lead to recruitment of clathrin adaptors through interaction with distinct populations of ARF-GEFs (79). Consistent with this, the GNOM gene product, one of the various Arabidopsis ARF1 GEFs, was shown to reside on post-Golgi endosomes, and its sensitivity to BFA was shown to lead to a defect in the recycling of the auxin efflux facilitator PIN-FORMED1 (PIN1) to the plasma membrane (80), a process that is probably not COPI dependent.
It is not known if GNOM also controls the biosynthetic vacuolar transport route, but first evidence for an involvement of ARF1 in post-Golgi biosynthetic vacuolar transport arose from the finding that a GTP-trapped mutant of ARF1 causes induced secretion of soluble vacuolar cargo and simultaneously inhibited constitutive secretion (48). Inhibition of secretion could be explained by a slow and indirect effect of COPI traffic interference on the closely linked COPII-dependent ER export route. In contrast, inhibition of vacuolar sorting was more sensitive to the ectopic expression of the dominant mutant and could be because of a faster and direct effect on clathrin-mediated Golgi export. A simpler model would be that interference with selective retrieval of proteins through COPI vesicles impairs Golgi maturation and that proper Golgi maturation is a prerequisite for clathrin-coated vesicle budding for the biosynthetic route to the vacuole. Whether this step occurs at the level of the trans Golgi cisternae or the trans Golgi network (TGN) remains to be shown. Further work has to be carried out to determine whether Golgi bodies reform from the trans or the cis end during BFA wash-out, and more markers, in particular from potential post-Golgi residents, need to be included in the analysis.
The Trans Golgi Network in Plants
The plant TGN is not simply the trans-most Golgi compartment but has to be regarded as a completely separate organelle. The earliest characterized markers for the plant TGN are the SNAREs SYP61 and SYP41 (plant homologues of yeast SNAREs TLG1 and TLG2), and in protoplasts, these markers label punctate structures that are often physically separated from the Golgi bodies (81). We have investigated the morphology of the TGN in the apical tobacco leaf epidermis cortex and obtained the same results (Figure 2). SYP61 labels organelles that (i) exhibit a more variable size than Golgi bodies and (ii) can either be overlapping or be closely associated (open arrows) or (iii) found several microns separated from the nearest Golgi stack (white arrows). Together with the rab11 GTPase group, one of the vacuolar H(+)-adenosine triphosphatases (V-ATPases) and the secretory carrier membrane proteins, several markers for the TGN are now available (81–84), but its precise role in the pathway has yet to be analysed in detail.
An excellent model protein from the biosynthetic perspective is the plant vacuolar sorting receptor BP80, which is a type I membrane-spanning protein that is thought to shuttle between the Golgi and the PVC to target soluble vacuolar proteins to the plant vacuoles. It was initially purified from clathrin-coated vesicles (85), and it concentrates on trans Golgi cisternae (7,67,70,86) even though its highest steady-state levels are found in the PVC in plants (87,88), the presumed final sorting station before the vacuole (7). The ultrastructural detection of BP80 in the Golgi stacks, clathrin-coated vesicles and the PVC, as well as the localization of a μ-adaptin for clathrin on the Golgi stacks (89), has led to the popular notion that BP80 leaves the Golgi in clathrin-coated vesicles. Indeed, clathrin-coated buds and vesicles are often seen very close to the periphery of the Golgi stacks (67).
However, past evidence from very elegant ultrastructural studies suggested that clathrin-coated vesicles bud from a post-Golgi organelle in plants. It was termed the partially coated reticulum (PCR) to indicate its unique morphology and possibly a separate function in the pathway (90–93). First descriptions mention it as a clathrin-coated tubular network adjacent to or isolated from individual Golgi stacks (92), and it is now considered to be the TGN of plants (83,84,94). For these reasons, it could be postulated that BP80 should pass through this compartment and possibly be more abundant in PCRs than in Golgi stacks, but this has not been reported in the literature. Compared with Golgi stacks, PCRs are much more difficult to identify morphologically as the appearance depends on the angle of the sections and may just look like a few clustered clathrin-coated vesicles.
The first powerful experimental approach to monitor endocytosis through cationized ferritin uptake time–courses led to convincing evidence suggesting that the PCR could be an early compartment within the endocytic route from the plasma membrane (95,96) that leads to the multivesicular body and the vacuole (97). Close proximity of putative PCRs with Golgi stacks could explain why endocytic cargo has been detected on Golgi stacks in the past (95,96,98,99) and also why BP80 appears to label the trans Golgi face of the stacks.
Endocytic cycling of plasma membrane proteins has recently become a new focus in plant biology because of its potential role in cell polarity and development (80,100,101). The use of the styryl dye FM4-64 as a fluorescent endocytic tracer has been a popular method to trace endocytic compartments of plant cells (102) because it is technically less demanding than detection of cationized ferritin by electron microscopy. FM4-64 labels every endomembrane of the secretory pathway depending on incubation conditions (103), but carefully timed uptake experiments revealed that the plant TGN is labelled earlier than the multivesicular bodies (83). This corresponds to the much earlier observations using ferritin uptake that also placed the PCR before the multivesicular bodies in the endocytic route (95–97).
Inspired by the fact that mammalian rab5 GTPases are found on early endosomes, the plant rab5 group of GTPases including the two nearly identical GTPases, RhaI and ARA7, and the plant-specific N-myristoylated ARA6 (104,105) were first assumed to control endocytosis. It soon became clear that plant rab5 GTPases are involved in the biosynthetic route to the vacuole as ARA7 (106,107) and an ARA6 homologue of Mesembryanthemum crystallinum(108) localize to the plant PVC. The PVC is thought to be the equivalent of the mammalian late endosome (7) and was recently shown to be identical to multivesicular bodies (88), which are also labelled with all plant rab5 GTPases (109). So, while mammalian rab5 is a landmark of the early endosome, plant rab5 GTPases reside on the PVC that is indeed not the first organelle to be visited by the endocytic tracer FM4-64 (83,84).
It is also important to note that there is disagreement on the localization of the plant rab5 GTPases, including ARA7 (104,106) as well as ARA6 (104,108), and none of the protein transport studies using manipulated rab5 homologues has systematically compared all three GTPases. Although the most recent study indicates that all three rab5 GTPases can label multivesicular bodies (109), these suggestions were drawn from immunogold electron microscopy. Similar to the experimental difficulties in quantifying Golgi versus TGN localization, these studies may suffer from a bias towards multivesicular bodies, which are easier to identify morphologically than TGNs, particularly when the TGNs are far from the Golgi stacks.
In spite of the early labelling of the plant TGN by FM4-64, it may thus be premature to propose the TGN as the plant equivalent of the mammalian early endosome (94). Recent experimental evidence from carefully conducted time-resolved uptake studies confirmed that FM4-64 indeed labels the TGN before the PVC, but at 15-min incubation, approximately 60% of the punctate structures highlighted by FM4-64 did not colocalize with the TGN marker (84). The true plant equivalent of the early endosome could thus be identical to these 60% of the structures (the majority), and because rab5 GTPases label the PVC in plants, a marker for these ‘very early FM4-64 compartments’ remains to be found. It should also be noted that the timing of FM4-64 labelling alone does not demonstrate vectorial transport. There is no proof that the dye moves from the plasma membrane through the TGN to the PVC. It could simply be that the dye reaches the TGN by a different and faster route. The molecules of the dye that appear at a later time-point in the PVC may never have passed through the TGN. This alternative scenario is just as plausible; it illustrates that the experiments are not at all easy and must be interpreted with an open mind.
Rab GTPases bring further confusion to the table because recent subcellular localization studies of rab11 GTPases show that these are localized to the plant TGN and the growing cell plate (82). The rab11 group is a landmark for recycling endosomes in mammalian cells, an organelle distinct from the TGN and early endosomes (6,110). Recycling endosomes are thought to be responsible for the recycling of endocytic cargo back to the plasma membrane rather than transport to late endosomes and ultimately the lysosomes and connect the endocytic pathway to the exocytic pathway. It is possible that this recycling function is carried out by the TGN in plants. More dramatically, the colocalization of rab11 with the cell plate syntaxin Knolle (82) suggests that the cell plate is not a specialized form of plasma membrane but rather a specialized TGN, supporting the plausible notion that de novo secretion rather than endocytosis lies at the origin of the new cell plate during cell division (111). De novo secretory activity during cell plate formation was also supported by the accumulation of the ER marker calreticulin at the cell plate (112), an early observation that has not been explored by the plant cytokinesis field.
Towards an In-Depth Exploration of the Endocytic Route in Plants
To reach further insight on endocytosis, it is crucial to expand our tools to monitor endocytic trafficking to identify drugs that interfere with distinct transport steps as well as the genes that control it. An important step forward in the field is the recent discovery that the compound endosidin 1 (ES1) interferes with endocytic cargo sorting and gives rise to accumulation of the auxin translocators PIN2 and AUX1 as well as the brassinosteroid receptor BRI1 within structures that also contained the TGN markers SYP61 and the V-ATPase subunit VHA-a1 (113). The specificity of the effect was shown by the fact that other plasma membrane proteins such as PIN1 and PIN7 were unaffected. It is likely that ES1 interferes with the recycling towards the plasma membrane, providing a further argument of a role of the plant TGN as a recycling endosome. In addition, two Eps15 homology domain proteins from Arabidopsis have recently been shown to influence endocytosis after modifying expression levels (114).
The plant homologue of the DnaJ domain-containing receptor-mediated endocytosis-8 (RME-8) gene was found in a genetic screen and first localized to a novel compartment distinct from Golgi, TGN and PVC (115). However, a follow-up study localized Arabidopsis RME-8 on the PVC, and it was suggested to control a late step in vacuolar sorting (116). These data indicate that endocytic routes and biosynthetic routes to the vacuoles may be functionally connected. Interestingly, a systematic evaluation of the properties of a sterol biosynthesis mutant suggested that changes in the membrane composition can also affect endocytic recycling, leading to defective polarized PIN2 localization (117).
In addition to the dye FM4-64, a fluorescently modified phospholipid, 1,2-bis-(4,4-difluoro-5,7-dimethyl-4-bora-3a,4a-diaza-s-indacene-3-undecanoyl)-sn-glycero-3-phosphocholine has recently been introduced to monitor endocytic transport and to follow lipid turnover in growing pollen tubes (118). Pollen tubes represent one of the best characterized polarized cell systems to study endocytosis and exocytosis (119). But, also the lower epidermis of leaf cells could potentially be regarded as a polarized cell type system, and because it has become a popular tool for in situ protein sorting assays after Agrobacterium-mediated leaf infiltration, investigators should realize that apical and basolateral plasma membrane compositions could be fundamentally different.
The field is thus rapidly expanding, and it will be important to investigate new effectors and drugs with the most complete list of molecular markers for intracellular compartments and to seriously consider membrane lipids for future strategies to explore key players.
The Biosynthetic Route to the Vacuole
Two distinct type I membrane-spanning proteins have been implicated as receptor molecules in transport of soluble proteins to plant vacuoles, the BP80 gene family (120) and the RMR gene family (68–70,121). They share some sequence homology at the N-terminus in their lumenal parts but are otherwise quite distinct. In vivo analysis of BP80 in plants was achieved by replacing the lumenal ligand-binding domain by the green fluorescent protein (GFP-BP80), and it was shown that such a construct interferes with the recycling of endogenous BP80, leading to induced secretion of vacuolar proteins (122). Importantly, overexpressed wild-type receptor could restore vacuolar sorting by out-competing constant levels of the competitor GFP-BP80, providing a positive and most sensitive receptor–ligand-binding assay in vivo. It also shows that the retrograde transport step is limiting and that anterograde transport through to the vacuole has a very high capacity in plants, which in turn explains why in plants the highest steady-state levels of BP80 are found in the PVC (88).
Exploring the rapid and quantitative competition and reconstitution assays led to convincing evidence suggesting that the tail is important for ER export, Golgi-to-PVC traffic as well as recycling from the PVC (123). Mutagenesis of a variety of conserved amino acids in the short cytosolic tail of BP80 led to the identification of several motifs. The best established of those is the so-called Yxxφ motif that binds to clathrin adaptors in vitro(7,89). The Y612A mutation converts the YMPL motif to AMPL and leads to mistargeting of GFP-BP80 to the plasma membrane. But, evidence for extra Golgi punctate structures as well as residual leakage to the vacuoles and weak processing to a soluble GFP core fragment suggested that GFP-BP80 (Y612A) could possibly cycle through the plasma membrane to the PVC and ultimately the vacuole (123).
Mistargeting to the plasma membrane was confirmed for the full-length BP80 (Y612A) variant, which also induces cosecretion of vacuolar cargo, again providing a positive assay for receptor–ligand interactions. It is highly unlikely that wild-type BP80 visits the plasma membrane in significant amounts because the plant apoplast is an acidic/lytic environment and vacuolar cargo would be dissociated. This is supported by the fact that overexpression of wild-type BP80 is easily achieved in transgenic plants, while the overexpression of full-length BP80–Y612A leads to few transgenic plants with low expression levels and dwarf phenotypes (unpublished data), probably because of cosecretion of vacuolar hydrolases (123) that interfere with cell wall proteins causing unspecific developmental disorders. However, endocytosis of BP80 could be a back-up system to clear up mistargeted receptors. The fact that GFP-BP80 (Y612A) can still endocytose and reach the vacuole was shown by a rapid wortmannin-induced redistribution of plasma membrane-localized fusion protein to the vacuolar lumen (123). These results may suggest that BP80 can enter a non-clathrin-mediated route from the plasma membrane, but this remains to be tested.
While it is generally accepted that BP80 releases its ligands in the PVC, this has not been formally demonstrated by experiments in vivo. It is also totally unclear how soluble vacuolar cargo is finally delivered to the vacuole. BP80 is thought to recycle from the PVC in a retromer-mediated process, but it is not clear if BP80-depleted PVCs fuse directly with the vacuole or if a further vesicle-mediated transport step is involved. The retromer components VPS35, VPS29 and VPS26 have been identified in plants by sequence homology with yeast vacuolar protein sorting (VPS) genes and shown to interact with the BP80 C-terminal cytosolic tail (124). A putative Arabidopsis Vps5 homologue was identified in the same manner, but knock-out mutants were not analysed with respect to vacuolar sorting (125). Although the authors postulated that the identified VPS5 member defines a novel structure, it is likely to be localized to the same PVC as the other retromer gene products. Further work on mutant VPS29 plants (126) and VPS35 knock outs (127) addressed specifically the knock-on effect on vacuolar cargo, and an alternative study suggested a role of VPS29 in organ initiation in plants (128). However, none of the knock-out strains or mutants has been used to test the direct binding partner of retromer and verified if BP80 is mistargeted to the vacuole.
The question regarding the last step in the vacuolar sorting route is relevant because receptor recycling alone is insufficient to guarantee that only cargo and not receptors become exposed to the lytic lumen of the vacuole. If PVCs fuse directly with the vacuole, how does the sorting machinery monitors simultaneously that cargo is present in the lumen but receptors are depleted so that it is safe to initiate membrane fusion with the vacuole? This question relates to the entire problem of BP80 sorting based on a short cytosolic tail. In the Golgi or TGN, the tail has to signal to cytosolic components to arrange for adaptor binding and clathrin recruitment but only when receptors are loaded with ligands. In the PVC, the same tail has to signal to cytosolic components to interact with retromer but only when ligands have been dissociated. It is likely that further membrane-spanning proteins may contribute to BP80 sorting and mediate conditional display/activity of defined sorting determinants. The role of the lipid environments may also be explored.
Turnover of BP80 could be achieved by incorporation into internal PVC vesicles that are expected to end up in the vacuole lumen. Members of the three endosomal sorting complex required for transport (ESCRT) (I, II and III) complexes have been identified in plants (129), and vesicle formation was shown to be influenced by the biological activity of the AAA ATPase VPS4 (109). The latter observation represents an important step forward and could be explored to test if BP80 transport includes specific partitioning into PVC internal vesicles and if these could include even different types of vesicles as suggested for mammalian cells (130,131). An excellent starting point for studying PVC maturation in the late plant vacuolar route arose from recent studies using the pH-sensitive fluorescent dye LysoTracker Red (132). The next step have to be made with live imaging to capture discrete PVC–vacuole fusion events if they take place.
Open Questions on Vacuolar Sorting
It may not be possible to use mammalian cells as tools to understand plant endocytosis and in particular vacuolar sorting because the vacuolar route exhibits plant-specific evolutionary traits that have emerged after the last common eukaryotic ancestor (1) and explains why the unusual N-myristoylated rab5 GTPase ARA6 is only found in the plant kingdom. The labelling of rab11 GTPases on the TGN (82) and the rab7 GTPase on the tonoplast (133) would support the notion that vacuolar sorting is not conserved in all eukaryotes. The most popular argument for the uniqueness of plants arose from the model that plant cells can contain functionally different vacuoles that coexist in the same cell and thus require different transport pathways and machineries (59). Because this is currently subject to some debate, it is worthwhile to summarize the most important arguments in the matter.
The first evidence arose from studies suggesting that membrane-spanning vacuolar proteins reach the tonoplast by a novel route as its arrival was not impaired by the drug BFA, which effectively blocked the arrival of soluble cargo (134). A little later, the phosphoinositide 3-kinase inhibitor wortmannin was reported to induce secretion of soluble vacuolar proteins with C-terminal signals but not sequence-specific (NPIRL motif containing) proteins (135). The plant vacuolar sorting receptor BP80 was then found to bind to NPIRL-containing proteins but not the so-called C-terminal vacuolar sorting variants (136). Finally, using antibodies specific to known vacuolar markers, fluorescence microscopy detected different putative vacuolar structures in plant cells (59).
One of these key findings failed to be reproduced in later studies that indicated a clear wortmannin sensitivity of proteins carrying NPIRL-like vacuolar sorting motifs (48,137) by preventing the recycling of the NPIRL-receptor BP80 (122). Furthermore, differences in the sensitivity to drugs alone are not conclusive because dosage effects could be explained by different affinities of ligands to the same receptor. For the sake of argument, if wortmannin inhibits the recycling of the plant vacuolar sorting receptor BP80 (122), it will lead to a progressive depletion of receptors at the level of the Golgi/TGN. Under these conditions, weak BP80 ligands will be the first to be secreted as they are driven off by stronger ligands that will be transported normally until receptor depletion is very advanced (requiring higher levels of the drug and longer incubation times).
Initially, BP80 was considered to be the vacuolar sorting receptor for the lytic vacuole because of its binding to sequence-specific sorting signals. However, the proven role of a BP80 homologue in storage protein transport in pumpkin (138–141) and the binding of BP80 proteins to other storage proteins carrying sequence-specific sorting signals (142) required a re-evaluation of the general concept. In addition, the same argument on receptor affinity could be raised for the evidence that partial BP80 depletion through gene knock out leads to discriminate secretion of storage proteins but not lytic vacuolar cargo (62). In the most recent studies, it appears that at least the model plant A. thaliana exhibits only a single vacuole that hosts all previously used markers (61) and that reduced BP80 levels affect both types of vacuolar sorting signals (143).
In conclusion, the debate about the two-vacuole theory has shifted back in favour towards the single vacuole theory with cell-type-specific adaptations (66), but this does not mean that all vacuolar proteins reach their destination in the same manner. It is also clear that two different vacuoles can be found in the same cells (144), but it is possible that these are still derived from a common precursor. To test a two-pathway theory, it is not enough to find an inhibitor that affects one type of cargo and not the other; it is equally important to identify the reciprocal inhibitor that affects the other but not the one. The existence of DVs in addition to clathrin-coated vesicles remains a proven fact (70), and for cereals, it has even been suggested that the ER itself, or structures derived from it, can simply become a storage ‘vacuole’(145). We should even be careful with the term ‘vacuolar marker’ because such markers may not label exclusively the vacuoles in all cell types, particularly at those developmental stages where real vacuoles have not yet developed. When can we call a structure a vacuole, and when is it still a PVC undergoing maturation and preparing for homotypic fusion?
Secretion of Soluble Proteins
Little is known in plants about the transport carriers that mediate delivery of secretory proteins and polysaccharides from the Golgi apparatus to the plasma membrane. In yeasts, Golgi export for secretion is thought to be vesicle mediated, and two distinct routes have been discovered that can replace each other under certain conditions (146–148). In mammalian cells, a role of specific lipids in Golgi-to-plasma membrane transport has recently been demonstrated (149), suggesting that secretory carriers are not simply a ‘left over’ compartment of the Golgi apparatus that emerges when all specific cargos have been selectively removed in all directions. In plants, a first report indicated that cell wall polysaccharides and proteins may reach the plasma membrane through different pathways (150). However, no ultrastructural data on transport vesicles or other membrane-bounded carriers have been reported that could shed light on the late secretory pathway. Likewise, mutants in which secretory cargo has been trapped in a post-ER compartment have yet to be isolated, but this may be a difficult task because proteins could be mistargeted to the vacuoles instead.
In the absence of hard data, we can only speculate on the pathway(s) of secreted proteins. Arguments can be found that native α-amylase could reach the cell surface directly from the Golgi apparatus, the TGN or the PVC. Interference with the PVC-to-vacuole transport by overdose of the syntaxin SYP21 (plant homologue of yeast SNARE pep12) had no effect on Golgi morphology or α-amylase secretion (151). The TGN and PVC fuse to a supercompartment after incubation of cells with wortmannin (84), but the drug has no effect on the secretory marker α-amylase (48) or the morphology of the Golgi stacks (84). These findings would suggest that native α-amylase exits the Golgi stack and reaches the plasma membrane without passing through the TGN. However, SYP21 overexpression caused partial secretion and accumulation of soluble vacuolar cargo in clustered/enlarged PVCs. Also, the drug wortmannin leads to secretion of vacuolar proteins. These data suggest that the plasma membrane can be reached either from the TGN or from the PVC. It is also possible that secretion is robust because several alternative routes can be used.
Constitutive secreted α-amylase is not concentrated along the Golgi stack (32), consistent with the idea that it moves passively by bulk flow. Sadly, this principle has not been confirmed with other secretory markers and should still be regarded as a working model. In general, it has been difficult to monitor soluble cargo in Golgi stacks by immunogold labelling either through electron microscopy or through fluorescent techniques. In fact, most recent studies on Golgi maturation in other systems have taken advantage of abundant membrane-spanning residents of the Golgi cisternae (152). The high turnover of cargo in transit, en route either to the cell surface or to the vacuoles, causes steady-state levels of these molecules to be low and thus below the detection limit of current techniques.
Solving the mystery of how secreted proteins reach the plasma membrane remains a formidable challenge for the plant field, and it is hopeful that recently developed genetic approaches using constitutively secreted fluorescent markers (153) will contribute to knowledge gained in other systems and reveal possible plant unique features. In addition, the use of a variety of secretory cargo molecules (soluble protein, cell wall polysaccharides, glycosyl-phosphatidylinositol-anchored molecules and membrane-spanning proteins) should be used to maximize the chances for new findings.
It has become clear that the plant secretory pathway exhibits many exciting features that deserve the attention of the research community. Based on the current data, we propose the following model to describe the well-established plant-specific adaptations of the secretory pathway (Figure 3).
The early secretory pathway consisting of the ER/NE is the most conserved, and except for morphological changes in different kingdoms, the biosynthetic starting point of the pathway must have been present in the last common eukaryotic ancestor in almost the same form. Plants do not contain an intermediate compartment between the ER and the Golgi and contain individual stacks of Golgi cisternae that sort retrograde cargo back to the ER and mature gradually. These stacks are mobile and move with ERES, although nascent ERES without Golgi stacks can been seen as well.
Of the post-Golgi compartments, only two structures have been well described. The TGN or PCR has to be regarded as a separate organelle from the Golgi stacks (Figure 2), and the PVC or the multivesicular body is the last sorting station before reaching the vacuoles. Although the TGN has been proposed as an early endosome, it could be regarded as a recycling endosome, while evidence for a putative earlier compartment on the endocytic route has been obtained, but it remains to be characterized with appropriate markers together with a full complement of other organelle markers of the secretory pathway.
Although constitutive secretion of soluble proteins is the default pathway in plants, it is currently not known if it involves any post-Golgi compartments. From the available date, the plasma membrane could potentially be reached from the Golgi stack directly, the TGN and if possible even the PVC, and further work with dominant-negative inhibitors an in particular combinations of inhibitors may help to dissect the various alternative transport routes in the pathway. Last but not least, we still need to learn how PVC-to-vacuole transport is achieved so that sorting receptors are rescued and cargo is deposited.