What Is Moving in the Secretory Pathway of Plants?

The secretory pathway of eukaryotic cells is a fascinating system of membrane compartments in dynamic equilibrium from the constant budding and fusion of vesicles and other membrane-enclosed transport carriers. Compared to yeasts, the plant protein machinery controlling vesicle budding, tethering, and fusion appears to be more complex and comparable to that of multicellular metazoans, confirming prior conceptions about the great complexity of the secretory pathway in plants. Moreover, phylogenetic analysis of trafficking components from plants, fungi, and animals suggest that many of those components were derived from duplications occurring after the last common eukaryotic ancestor, implying that numerous features of the endomembrane system evolved separately in the different kingdoms (Vernoud et al., 2003; Mouratou et al., 2005; Dacks et al., 2008). Indeed, some particular subfamilies of key trafficking factors are expanded in number in plants or may even be unique to these organisms. Examples of these are the putative plant-specific SYP7 group of syntaxins (Sanderfoot, 2007), or the expanded number of Rab GTPases (Rutherford and Moore, 2002; Nielsen et al., 2008) and GBG type of ADP-ribosylation factor guanine-nucleotide exchange factors (ARF-GEFs; Mouratou et al., 2005). A measure of these differences is given by the specific characteristics of the endomembrane system in plants, such as the large size and specialized functions of plant vacuoles (Marty, 1999), the rapid movement of Golgi stacks (Nebenfuhr and Staehelin, 2001), and the unique organization of endosomal compartments (Ueda et al., 2001, 2004; Chow et al., 2008). This justifies a closer look at protein trafficking pathways in plants because it may reveal novel diversification or the existence of distinct solutions for the same functions that may remain otherwise unnoticed.

In this Update, we will first review the ongoing efforts to genetically characterize the machinery involved in biosynthetic trafficking in plants. In recent years, many new genes implicated in different steps of trafficking in the plant biosynthetic pathway have been discovered (Table I). We will then discuss how the functional analysis of these components has allowed more detailed characterization of the trafficking pathways for soluble and membrane proteins in this route. Finally, we will discuss the recent data on the convergence/intersection of the biosynthetic and endocytic pathways in plants and comment on the current debate regarding the existence of two fundamentally different trafficking routes to vacuoles within the same cell.

Table I.

Summary of genes functionally linked to trafficking in the biosynthetic pathway

Data collected from published reports. The Arabidopsis Genome Initiative (AGI) code for Arabidopsis genes is shown. When known, localization and site of action are shown; type of genetic evidence as well as the experimental system are also shown. Abbreviations not defined in the text: ERES, ER export sites; CCV, clathrin-coated vesicle; Vac, vacuole; CP, cell plate; ND, not determined; Mut, mutant in endogenous genes; DN, expression of dominant-negative constructs; OE, overexpression; CA, constitutively active construct; At, Arabidopsis; Nt, Nicotiana tabacum; Nb, Nicotiana benthamiana; Os, Oryza sativa; Spo, short peptide from sweet potato (Ipomoea batatas) sporamin containing a sequence-specific vacuolar sorting signal; Amy, barley α-amylase; Aleu-GFP, first 143 amino acids from barley aleurain containing the signal peptide and the sequence-specific vacuolar sorting signal fused to GFP; ST, rat sialyl transferase; GFP-2SC, GFP with a signal peptide fused to the C-terminal propeptide from pumpkin 2S albumin; GFP-CT24, GFP with the signal peptide and the C-terminal 24 amino acids of soybean (Glycine max) (R)-conglycinin; PHT1-GFP, Arabidopsis phosphate transporter PHT1;1 fused to GFP; Inv-GFP, tobacco secreted invertase fused to GFP; PAT, phosphinothricin acetyl transferase; SynA-GFP, synthetic arabinogalactan protein fused to GFP; GFP-CTPPBL, GFP with a signal peptide fused to the C-terminal propeptide from barley lectin; BL, barley lectin; GFP-CHI, GFP with a signal peptide fused to the vacuolar sorting signal from tobacco chitinase-A.

Gene	AGI Code	Localization	Trafficking Step	Markers Affected	Evidence
ADL6	At1g10290	Golgi	Vacuolar trafficking	Spo-GFP, GFP-EBD	DN: At
AGD7	At2g37550	Golgi	Golgi to ER	Spo-GFP, phaseolin, Inv-GFP, g-COP, ST-GFP, GH	OE: At
ARF1	At2g47170	Golgi, endosomes	Golgi to ER	Amy, Amy-Spo; ST-GFP, Erd2-GFP, Spo-GFP; H-ATPase-GFP; GFP-Rer1b, GFP-sed5	DN: Nt, At
BET11	At3g58170	Golgi	PM trafficking	SecYFP	OE: Nt
EPSIN1	At5g11710	Golgi, PVC, actin	Vacuolar trafficking	Spo-GFP	Mut: At
GFS10	At4g35870	ND	Vacuolar trafficking	Seed storage protein, GFP-CT24	Mut: At
GNL1	At5g39500	Golgi	Golgi to ER, vacuolar trafficking	SecGFP, Aleu-GFP, Pin1, Pin2	Mut: At
GNOM	At1g13980	LE, Golgi	Endosome to PM	Pin1, Pin2	Mut: At
GRV2/KAM2	At2g26890	Punctate	Vacuolar trafficking	Seed storage proteins, GFP-2SC	Mut: At
KAM1/MUR3	At2g20370	Golgi	Vacuolar trafficking	GFP-2SC	Mut: At
MAG1/VPS29	At3g47810	PVC	Vacuolar trafficking	Seed storage proteins, GFP-CT24	Mut: At
MAG2	At3g47700	ER	ER to Golgi	Seed storage proteins	Mut: At
Memb11	At2g36900	Golgi	ER to Golgi	SecYFP, ST-YFP, ERD2-YFP	OE: Nt
PHF1	At3g52190	ER	ER export	PHT1-GFP	Mut: At
OsAGAP	N/A		PM trafficking	AUX1	OE: Os, At
OsGAP1	N/A	TGN or PVC	Vacuolar and PM trafficking	H-ATPase-GFP, Ca-ATPase-GFP and Aleu-GFP	DN: At
RABmc	N/A	Golgi, PVC	Vacuolar trafficking	Aleu-GFP	DN: Mc
RABA2A	At1g01200	EE, CP	CP formation	ND	DN: At
RABD2A	At1g02130	ND	ER to Golgi	SecGFP, ST-GFP	DN: Nt
RABE1D	At5g03520	Golgi	Golgi to PM	SecGFP	DN: Nt
RABF2A/RHA1	At5g45130	PVC	Vacuolar trafficking	Aleu-GFP, Spo-GFP	DN: At
RABF2B/ARA7	At4g19640	PVC, Golgi	Vacuolar trafficking	Aleu-GFP	DN: Nt
Rab11	N/A	Secretory vesicles	PM trafficking	SecGFP, SynAGP-GFP, Inv-GFP	DN, CA: Nt
RHD3	At3g13870		ER to Golgi	SecGFP, ST-GFP	Mut: At
RMR1	At5g66160	Golgi, DV, PVC, Vac	Vacuolar trafficking	Phaseolin	DN: At
SAR1A, SAR1B	At1g09180, At1g56330	Cytosol, ERES	ER to Golgi	Amy, Amy-HDEL, ERD2-GFP, GFP-RER1b, Spo, Spo-GFP, calreticulin, secGFP, PAT	DN: At, Nt
SEC12	At2g01470	ER	ER to Golgi	Amy, Amy-HDEL, ERD2-GFP, Phytepsin, calreticulin, PAT	OE: Nt
SEC22	At1g11890	ER, ERES	ER to Golgi	SecYFP, ST-YFP, ERD2-YFP	OE: Nt
SYP21	At5g16830	PVC	Vacuolar trafficking	Amy-Spo, BopTIP-GFP, YFP-SYP22, TIP1;1-YFP	OE, DN: Nt
SYP22/SGR3	At5g46860	PVC, Vac	Vacuolar trafficking	TTG2	Mut: At
SYP31/SED5	At5g05760	Golgi	ER to Golgi	SecYFP, ERD2-YFP	OE: Nt
SYP121	At3g11820	PM	PM trafficking	SecGFP	DN: Nt
SYP122	At3g11820	PM	PM trafficking	SecGFP	DN: Nt
SYP132	At5g08080	PM	PM trafficking	Pathogenesis-related proteins	VIGS: Nb
SYP71	At3g09740	PM	PM trafficking	SecGFP	DN: Nt
TFL1	At5g03840	PM, Vac, punctate	Vacuolar trafficking	VacPerox, GFP-CTPPBL, VAC2	Mut: At
VAMP711	At4g32150	Vac	Vacuolar trafficking	H2O2-containing vesicles	AS: At
VAN3/SFC	At5g13300	TGN	ND	PIN1-GFP	Mut: At
VCL1-1	At2g38020	Vac, PVC	Vacuolar trafficking	AtAleu	Mut: At
VPS9	At3g19770	ND	ND	PIN1-GFP, GFP-RABF2B, RabF1-GFP	Mut: At
VPS35A, VPS35B, VPS35C	At1g75850, At2g17790, At3g51310	PVC	Vacuolar trafficking	Seed storage proteins	Mut: At
VSR1/GFS1	At3g52850	Golgi, CCV, DV, PVC	Vacuolar trafficking	Seed storage proteins, GFP-CT24	Mut: At
VTI11/ZIG1	At5g39510	PVC	Vacuolar trafficking	Aleu-GFP	Mut: At
VTI12	At1g26670	TGN	Vacuolar trafficking	Seed storage proteins, VacPerox, GFP-CHI, BL, VAC2	Mut: At

LOSS-OF-FUNCTION MUTANTS AFFECTING THE TRAFFICKING MACHINERY IN THE BIOSYNTHETIC ROUTE

Relatively few trafficking components have been genetically characterized through analysis of loss-of-function mutants. Redundancy within the large gene families encoding many of these components may account for this and, indeed, this has already been shown in a number of cases. Arabidopsis (Arabidopsis thaliana) has three genes homologous to the yeast retromer component VPS35, and trafficking phenotypes can be observed by combining null alleles in AtVPS35b and AtVPS35c, or null alleles in AtVPS35a and AtVPS35c and a weak allele in AtVPS35b. These double and triple mutants showed defects in growth and in the vacuolar targeting of storage proteins in seeds (Yamazaki et al., 2008). The VTI11 and VTI12 SNAREs are a well-characterized example of conditional redundancy. VTI11 and VTI12 have distinct localizations, interacting SNARE partners, and functions. However, in the absence of one of them, the other takes over in the respective complex and substitutes it functionally, albeit only partially (Surpin et al., 2003; Sanmartin et al., 2007). Moreover, a single amino acid change in VTI12 enhances its capacity to substitute for VTI11, probably by increasing its affinity for the corresponding SNARE complex (Niihama et al., 2005). Another interesting example of redundancy was recently reported between GNL1 and GNOM, two Arabidopsis ARF-GEFs (Richter et al., 2007; Teh and Moore, 2007). Through mutant analysis and engineering of brefeldin A-resistant and -sensitive forms of these proteins, it was demonstrated that GNL1 and GNOM are redundantly involved in COPI-coated vesicle formation at the Golgi stacks. Moreover, GNL1 functions with an unidentified brefeldin A-sensitive ARF-GEF in endocytosis of pin-formed (PIN) efflux carriers, whereas GNOM functions in their recycling from endosomes to the plasma membrane (PM).

Genetic redundancy may lead to subtle phenotypes difficult to discover through the reverse-genetic approaches that are commonly used in plants such as Arabidopsis. In cases where redundancy is not present, knockout of trafficking genes may result in early lethality (Rojo et al., 2001; Sanderfoot et al., 2001). Only in rare cases, T-DNA insertions provoke a weak loss of function that permits the analysis of essential trafficking genes (Goh et al., 2007). Forward-genetic screens are powerful tools to study processes where redundancy or lethality is a problem because they allow the enrichment and isolation of informative alleles. In recent years, a number of genetic screens to dissect trafficking to the PM and the vacuole have been reported.

Moore and collaborators reported a genetic assay to detect Arabidopsis mutants blocked in secretion based on the enhanced stability and fluorescence of a secreted GFP construct (secGFP) when retained in intracellular compartments. With this assay, two genes necessary for export of cargo (including secGFP) from the endoplasmic reticulum (ER) have been identified: RDH3, which encodes a GTP binding protein, and GNL1, which encodes an ARF-GEF (Zheng et al., 2004; Teh and Moore, 2007).

Several genetic assays to dissect trafficking to plant vacuoles have also been described. A laborious screen for maigo (mag) mutants that accumulate seed storage protein precursors, which is indicative of a block in their vacuolar targeting, led to the isolation of a number of Arabidopsis mutants. MAG1 encodes the retromer component VPS29 (Shimada et al., 2006), which may participate in retrograde trafficking from the prevacuolar compartment (PVC) to the trans-Golgi network (TGN). Although not directly tested, it strongly suggests that plant vacuolar sorting receptors (VSRs) deplete at the level of the TGN in mag1 plants because they are probably degraded in the vacuoles. This would lead to the secretion of their ligands, the storage proteins, to the extracellular space. It should be possible to rescue the phenotype by simply overexpressing VSRs and it should also be easy to test receptor turnover in wild-type and mutant plants, but these experiments have yet to be reported.

MAG2 encodes a protein related to Rint1/TIP20 (Li et al., 2006). In yeast, TIP20 was shown to prevent back fusion of ER-derived COPII-coated vesicles with the ER (Kamena and Spang, 2004) and is part of a docking complex for Golgi-derived COPI vesicles (Cosson et al., 1997; Frigerio, 1998) that is termed the Dsl1p complex. This putative tethering complex is composed of Dsl1, Dsl3 (Sec39), TiP20, and the trimeric Use1p-, Sec20p-, and Ufe1p-t-SNARE protein acceptor complex for the v-SNARE Sec22 on Golgi-derived COPI vesicles (Kraynack et al., 2005). The MAG2 gene product shows direct interactions with Sec20 and Ufe1, but not a putative plant Dsl1 homolog (Li et al., 2006). Further work will have to be carried out to understand the exact functioning of these putative ER import sites. If the DSL1 complex is compromised in the mutants, ER-derived COPII vesicles could mistarget to the ER, COPI vesicles might not properly fuse, and the entire ER-Golgi interface might collapse due to the intimate dependence of anterograde and retrograde transport pathways. However, two potential isoforms of MAG2 exist in plants that could partially compensate for each other, hence the possibility to obtain viable mutants. mag2 mutants accumulate large amounts of storage proteins in a novel compartment that is probably derived from the ER. It will be interesting to see whether these structures are related to the well-described PAC vesicles found in pumpkin (Cucurbita spp.; Hara-Nishimura et al., 1998). A one-way traffic strategy that does not require continuous recycling could be adequate for nonequilibrium systems, such as developing seeds that simply accumulate storage proteins until dormancy. It will also be interesting to test the effect of the mag2 mutant on the properties of vegetative cells to transport cargo between the ER and the Golgi apparatus.

Recently, Hara-Nishimura and collaborators developed a more efficient method for isolating mutants in vacuolar sorting. This screen is based on detection of the SP-GFP-CT24 vacuolar marker in the apoplasm, where the marker is stabilized, leading to a great increase in fluorescence. In this way, more than 100 green fluorescent seed (gfs) mutants were isolated. Analysis of 10 of those mutants showed that they corresponded to seven complementation groups (Fuji et al., 2007). Four mutants were novel alleles of the plant vsr1 (Shimada et al., 2003), whereas another was allelic to katamari2 (kam2; Tamura et al., 2007). The kam mutants were isolated in a screen for plants that accumulate in intracellular compartments the vacuolar SP-GFP-2SC construct, which is degraded in the vacuole in a light-dependent manner (Tamura et al., 2003). The kam1 and kam2 mutants were shown to correspond to mutations in MUR3, an integral Golgi membrane protein that interacts with actin (Tamura et al., 2005), and GVR2, a peripheral membrane-associated protein of unknown function (Tamura et al., 2007).

Another vacuolar mutant screen has recently been described (Sanmartin et al., 2007; Sohn et al., 2007). The screen is based on secretion of a vacuolar marker (VAC2), which in the apoplasm negatively regulates stem cell proliferation, leading to early termination of meristems. In this way, six genes required for trafficking of vacuolar storage proteins have already been identified: the TGN SNARE VTI12, TFL1, and four other components that may be implicated in regulating SNARE complex formation, in vesicle budding, and in protein sorting (J. Zouhar, A. Muñoz, O. Delgadillo, and E. Rojo, unpublished data). VTI12 may be involved in retrograde vesicle fusion at the TGN, whereas the role of TFL1 is not known, although it closely colocalizes with the δ-adaptin of the AP-3 adaptor complex and may therefore be involved in vacuolar cargo exit from an endosomal compartment (Newell-Litwa et al., 2007).

It will be important to devise further screens with more elaborate reporter systems to study plant-specific phenomena, such as Golgi morphology and motility, and to discover those key players of the pathway that evolved after the last common eukaryotic ancestor.

DOMINANT AND SEMIDOMINANT MUTANTS AFFECTING BIOSYNTHETIC TRAFFICKING

In addition to mutants, a large body of genetic data on the function of different trafficking components has been attained through expression of dominant transgenes. Although loss-of-function mutants are a first step toward gene identification, further knowledge can often be gained through the use of gain-of-function mutants. These can range from dominant-negative mutants to simply overexpressed wild-type proteins, which lead to titration of other components of the system and shed light on the network of possible interactions within a living cell.

Through the available trafficking assays, three major gene families have been studied by transient expression of dominant constructs: GTPases of the Rab and ARF families, ATPases of chaperones and other key regulators of organelle traffic, and SNAREs. In the case of the GTPases, it is possible to engineer mutated versions of a particular Rab or ARF locked in the GTP or GDP bound forms. Nucleotide-free mutants have also been instrumental with Rab GTPases and are thought to titrate their exchange factors, although this has not been formally tested in plants. In the case of ER export, it was experimentally shown that overdose of the exchange factor Sec12 titrates its target GTPase Sar1 and prevents ER export of soluble cargo (Phillipson et al., 2001). Similar experiments with ARF exchange factors combined with quantitative transport assays remain to be done.

Curiously, nucleotide-free rab mutants are expected to exhibit a dominant effect and reveal their role in membrane trafficking even in the presence of the wild-type gene (Table I) because they lead to stable complexes with their nucleotide exchange factors, thus titrating those and preventing wild-type Rabs from functioning even when these are overexpressed (Jones et al., 1995; Richardson et al., 1998). For these reasons, it is not expected to help to express more wild-type protein even though such scenarios have been suggested (Batoko et al., 2000; Kotzer et al., 2004). It should be noted that, in those experiments, levels of the dominant-negative effector were not quantified, so it is possible that coinfiltration of a second Agrobacterium impaired the transfection efficiency of the strain encoding the dominant-negative mutant. In other instances, the highly dominant nature of mutants of the GTPases ARF1 (Pimpl et al., 2003) and Sar1 (J. Denecke, unpublished data) prevented complementation even with a 10-fold excess of wild-type protein. In these cases, naked DNA transfection was carried out in protoplasts, which allows more quantitative cotransfection and control of the expression levels of the effectors.

In the case of SNAREs, overexpression of truncated molecules lacking the transmembrane domain (Sp2 fragment) can exhibit dominant-negative effects (Tyrrell et al., 2007), possibly through titration of binding partners. It has not been shown which binding partners are titrated, but expression of wild-type SNARE alleviated the effect, suggesting only weak dominance. It was shown that effectiveness of SP2 fragments strongly depends on the type of SNARE and, in the case of the PVC SNARE SYP21, overexpression of the wild-type molecule inhibited traffic, whereas an expressed SP2 fragment had no measurable effect under conditions where effector molecules were quantified in addition to the test cargo molecules (Foresti et al., 2006). Similar results were obtained with the Golgi syntaxin SYP31, which caused inhibited secretion of a secYFP marker when it was overexpressed as a full-length protein (Chatre et al., 2005). Further work must be carried out to test whether discrepancies in the results are due to the model system or the type of SNARE.

Expression of engineered forms of a protein that have an altered sensitivity to a drug is a very powerful system to study gene function. This approach was used to prove the specific function of the ARF-GEFs GNOM and GNL1 in ER-Golgi trafficking and in endocytosis and recycling to the PM (Geldner et al., 2003; Richter et al., 2007; Teh and Moore, 2007). Elegant recent work on the GNOM gene product established a novel heterotypic interaction of large ARF-GEFs that could control membrane association and subsequent ARF activation (Anders et al., 2008). With the advent of chemical genomics studies, many more drugs affecting different steps of trafficking will be isolated and their targets identified (Rojas-Pierce et al., 2007), allowing fine tuning and, in some cases, reversible fine tuning of protein activity in a particular transport reaction.

An advantage of transient expression experiments is the possibility of testing particular constructs that would be deleterious to stable transformed plants. Moreover, the speed of the assays permits much deeper characterization of the particular trafficking steps studied. An example of this is the thorough functional analysis of the Bp80/VSR1 protein, which demonstrated that recycling of VSR1 to the PVC is essential for efficient transport of vacuolar cargo and saturable (daSilva et al., 2005). By replacing the lumenal domain with GFP, an artificial transport competitor (GFP-BP80) was created that titrated endogenous BP80 at the recycling step, leading to BP80 depletion and secretion of vacuolar proteins. Unlike ARF1 mutants, the competitor is not dominant and effective complementation by overexpressed wild-type BP80 was demonstrated. The system allowed the identification of sequences in the cytosolic tail that are required for its targeting (daSilva et al., 2006) and may reveal specific sorting determinants for ER export, Golgi-to-PVC traffic, and recycling.

It is interesting to note that silencing techniques, such as RNA interference (RNAi) or virus-induced gene silencing (VIGS), have not been widely used yet to analyze trafficking, although they could prove very helpful, especially in transient assays. Similar to the titration of a gene product by a permanently binding mutant molecule, this technique simulates loss-of-function mutations and can be timed by the inducible expression of RNAi. Recently, evidence linking Syp132 to secretion of pathogenesis-related proteins and resistance to pathogens was obtained through VIGS silencing of the tobacco (Nicotiana tabacum) gene (Kalde et al., 2007). Through antisense silencing, the VAMP7C family of SNAREs was implicated in fusion of H₂O₂-containing vesicles with the tonoplast (Leshem et al., 2006). Limitations of the technology may be the time frame of turnover of the endogenous molecule after successful induction of RNAi, but this is a matter of optimization and is certainly a valuable strategy to be explored.

A caveat to many of the genetic data available is that only one or, at best, a few trafficking markers were tested for each mutant or transgenic line analyzed. This is particularly relevant for experiments using transient expression because, in that case, there is no previous selection against deleterious side effects, as in the case of stable mutants. Analyzing multiple markers increases the chances to dissect between a general block in endomembrane trafficking and specific branches of a pathway, even if a specific effector influences more than one pathway (Pimpl et al., 2003). Moreover, deep characterization of transport defects will allow distinction between primary effects and downstream consequences of loss-of-function mutations. For instance, it has recently been proposed that the VPS29 gene product controls cell polarity and organ initiation (Jaillais et al., 2007). Alternatively, it is possible that a compromised retromer simply mistargets vacuolar hydrolases to the apoplast by analogy to the earlier observed defects in storage protein traffic in vps29 mutants (Shimada et al., 2006). Therefore, it cannot be ruled out that a fundamental defect in vacuolar sorting may simply cause inappropriate turnover of a range of cell surface proteins and secreted cell wall proteins, leading to unspecific developmental defects.

TRANSPORT OF SOLUBLE CARGO

Transport of soluble proteins in the biosynthetic pathway is probably the best-characterized trafficking process in the endomembrane system of plants. It is now widely accepted that the default destination for soluble proteins in the secretory pathway is the apoplasm and that no export signals are required to achieve transport rates of native secretory proteins like amylase (Denecke et al., 1990; Crofts et al., 1999; Phillipson et al., 2001; Pimpl et al., 2006). To be retained in intracellular compartments, proteins require positive sorting information in their sequence. Signals for retention in the ER and for targeting to the vacuole have been thoroughly characterized (Vitale and Hinz, 2005). In contrast, soluble proteins retained in the Golgi or in endosomal compartments have not been reported. Moreover, very little is known about the process of secretion of soluble cargo to the PM. For example, the secretory vesicles have not been characterized, nor is it known whether the last compartment that soluble proteins pass before reaching the PM is the Golgi apparatus or the TGN. In fact, it cannot even be ruled out that the PVC can serve as a sorting station for transport to the PM during certain conditions. An added level of complexity of secretion in plants is that cell wall polysaccharides appear to use different pathways than proteins (Leucci et al., 2007).

Due to the agronomic importance of storage proteins accumulated in vacuoles and protein bodies of plants, many studies have been focused in this process. Comparatively, a lot of data in plant vacuolar trafficking have been generated and fueled intense debate. An important point of discussion in the field of vacuolar trafficking is the nature of the receptors that recognize the vacuolar sorting determinants in soluble cargo to package them into vesicles destined for the vacuole. Two families of proteins have been put forward as candidates to fulfill that role. The VSR family of proteins, made up of seven members in Arabidopsis, has many of the characteristics expected for sorting receptors. They are type I transmembrane proteins with a large lumenal domain that interacts with vacuolar sorting determinants. The small cytosolic tail harbors motifs that are involved in interaction with adaptor complexes of vesicle coats. Moreover, VSRs are present on the TGN and PVC (Sanderfoot et al., 1998), and recycling from the PVC to the TGN is essential for its function (daSilva et al., 2005). The role of VSRs in trafficking of vacuole cargo is firmly established by a number of studies (Shimada et al., 2003; Jolliffe et al., 2004; daSilva et al., 2005, 2006; Fuji et al., 2007; Craddock et al., 2008). However, it is still unclear whether their main function is in trafficking of storage or lytic cargo. Analysis of single mutants of the VSR genes has shown that only mutants in VSR1 secrete seed storage proteins. However, the bulk of storage cargo was still correctly targeted to the vacuole in the single vsr1 mutant. This has been interpreted by some as evidence for VSRs not being the main sorting receptors for storage proteins. An alternative explanation is that other VSR genes have minor activity in seeds compared to VSR1, but sufficient to take over in its absence. A prediction of this hypothesis is that secretion of storage proteins would increase by combining the vsr1 mutation with mutations in other VSR genes. A second family of proteins encoding the receptor homology-transmembrane-RING H2 domain protein (RMR) has also been proposed to act as VSRs for storage proteins. AtRMR1 was shown to interact with the sorting determinant of the storage protein phaseolin and chitinase, and dominant-negative versions blocked exit of phaseolin from the Golgi in Arabidopsis protoplasts (Park et al., 2005, 2007). A functional relation between RMR and storage proteins is further supported by their coregulated expression (Pearson correlation coefficient r = 0.605 for AtRMR1 and the cruciferin gene At4g28520), as analyzed in published microarray experiments (Manfield et al., 2006). A prediction of the hypothesis that RMRs are the main VSRs is that loss-of-function mutations should cause defects in storage protein trafficking, at least at a similar level as the vsr1 mutant displays. However, such phenotypes have not been reported yet, suggesting either that functional redundancy is complicating the analysis, or that RMR proteins are not sorting receptors but are involved in other aspects of storage protein accumulation. Recently, the localization of endogenous storage protein cargo and VSR and RMR proteins in embryos of Arabidopsis was analyzed by immunoelectron microscopy (Hinz et al., 2007). However, these analyses did not settle the dispute because the results were conflicting. Hinz and collaborators found colocalization of storage proteins with RMRs at the rims of the Golgi stacks, already from the cis-Golgi cisternae, and in dense vesicles (DVs). In contrast, VSRs were found separated from storage cargo, mainly at the trans-Golgi cisternae, after storage protein aggregation has occurred. However, Otegui and collaborators reported marked colocalization of storage proteins and VSRs at the Golgi stacks and in DVs. A possible model to reconcile the genetic and localization data is that RMRs may be involved in storage protein aggregation and VSRs in targeting those aggregates to vesicles destined for vacuoles. A thorough analysis of vsr and rmr mutants should shed light into this matter in the coming years.

TRANSPORT OF MEMBRANE PROTEINS

Initially, progress in investigating the transport of plant membrane proteins was slow compared to soluble proteins, mainly due to the lack of good markers and simple assays. With the development of live imaging with GFP fusions and other spectral variants, advances on understanding the sorting of membrane-spanning proteins within the plant secretory pathway is now fast and furious and many principles are starting to emerge. Whereas secretion appears to be the default for soluble proteins, the rules regarding the exit of membrane-spanning proteins from the ER and subsequent sorting to other locations appear to be more complex, as suggested by conflicting reports. It was first suggested that the tonoplast is the default membrane (Hofte and Chrispeels, 1992). This hypothesis was further tested with type I membrane-spanning reporter fusions constructed from ER residents and mutagenesis of the known ER retention signals. One of these studies confirmed the tonoplast as default (Barrieu and Chrispeels, 1999), whereas the other suggested the PM (Benghezal et al., 2000).

Further complexity was introduced by a study that suggested a critical role of the length of the transmembrane domain in the sorting of type I membrane-spanning proteins (Brandizzi et al., 2002). Short transmembrane domains (17 amino acids) would be unable to leave the ER, intermediate length would lead to Golgi retention, whereas long transmembrane domains would reach the PM. In this study, the tonoplast lumen or the extracellular medium was not analyzed. The first ER export signal for membrane-spanning proteins was identified through deletion and transplantation analysis and established the well-known DXE signature for COPII-dependent ER export (Hanton et al., 2005).

Most recently, a systematic analysis of the sorting of p24 proteins revealed that deletion of a di-Lys motif for COPI-mediated ER retention led to transport to the PVC and a soluble degradation product was detected in the vacuole lumen (Langhans et al., 2008). This matches perfectly with earlier in vitro data (Contreras et al., 2004a, 2004b), but, interestingly, the previously identified diaromatic motif for in vitro COPII interaction was dispensable for ER export in vivo, despite the short length of the transmembrane domains (16 amino acids). ER export of membrane-spanning proteins is either very complex or current predictions of the transmembrane domain length is unreliable with available software.

The problem with many of these studies is that the reporter fusion is proteolytically cleaved and leads to the detection of the soluble reporter in either the vacuole (Barrieu and Chrispeels, 1999; Langhans et al., 2008) or the extracellular medium (Benghezal et al., 2000). Depending on the position in the secretory pathway, cleavage can either lead to secretion from the Golgi or vacuolar transport from the PVC, but it would be ideal to have a reporter that is stable.

As part of a systematic analysis of the sorting of the plant VSR BP80, the lumenal portion of BP80 was replaced by GFP and the C-terminal tail was mutagenized (daSilva et al., 2005, 2006). Interestingly, a deletion of most of the cytosolic tail results in a protein that is significantly retained in the ER despite an intermediate transmembrane domain length, but also reaches the Golgi, PVC, and a soluble cleavage product (termed GFP-core) can be detected in the vacuole. The cytosolic tail of BP80 contains a conserved DXE motif and a Tyr-based YxxΦ signal for clathrin-mediated transport that is required for in vitro binding to μ-adaptin (Happel et al., 2004). Mutagenizing the Tyr residue (Y612A) results in partial mistargeting to the PM and partial progress to post-Golgi punctate in vivo and yields an intermediate level of vacuolar GFP-core (daSilva et al., 2006), suggesting that it still reaches the vacuolar membrane. In the case of GFP-BP80, it was firmly established that GFP-core is only detectable in the vacuole, and microsomes containing ER, Golgi, and post-Golgi endosomes (including the PVC) were devoid of GFP-core (daSilva et al., 2005, 2006). This suggests that the fusion proteins reach the tonoplast intact and are then cleaved.

Rules regarding the sorting of membrane-spanning proteins are thus far from established in plants. It should also be noted that comparable BP80 transmembrane domains were used in two studies, but different results were obtained (Brandizzi et al., 2002; daSilva et al., 2006). Possible differences can be due to minor changes in the reporter, such as the presence of a peptide at the N terminus of GFP carrying a functional N-glycosylation consensus site (daSilva et al., 2006). Therefore, it appears that a black-and-white rule regarding the transmembrane domain length and its influence on membrane protein sorting cannot be maintained.

Finally, results obtained with type I membrane-spanning proteins could not necessarily apply to all membrane-spanning proteins. For instance, the sorting of type II membrane-spanning proteins could be dependent on both the transmembrane length and on motifs exposed in the cytosolic peptides (Saint-Jore-Dupas et al., 2006). Likewise, the sorting of tail-anchored proteins like most of the SNAREs may even deviate as early as the ER import step. They represent a subset of type II membrane-spanning proteins in which the transmembrane domain is so close to the C terminus that translation terminates before the signal recognition particle can arrest translation. Tail-anchored molecules are thought to be transported in a signal recognition particle-independent manner and depend on a cytosolic and highly conserved 40-kD ATPase for insertion into the ER membrane (Stefanovic and Hegde, 2007). First attempts to use tail anchoring in plants have resulted in stabilization of a recombinant protein (Barbante et al., 2008), but further work needs to be carried out to assess how targeting of tail-anchored molecules is regulated. An interesting gene family will be the SNAREs that almost all share similar overall structures and yet partition to specific target membrane domains throughout the secretory pathway (Uemura et al., 2004). It is particularly interesting for this group of proteins because many of the members are thought to recycle between different compartments and also require conditional activation, depending on the membrane that they visit.

THE ORGANIZATION OF THE LATE SECRETORY PATHWAY

Compared to the knowledge on the early secretory pathway in plants and the route to the vacuole, much less is known about endocytic transport routes or the route taken by secretory proteins. Where do secretory proteins segregate from vacuolar proteins? Are there several routes to the PM? One of the difficulties with research on endocytosis is related to the fact that much of the research was inspired by analogy from processes in animals and yeasts. For instance, the Rab5 group was initially assumed to label the early endosome (EE), and colocalization with endocytic tracers, such as FM4-64, was taken for granted as evidence for endocytic compartments. Later evidence demonstrated that the Rab5 group labels the PVC or the multivesicular body (Kotzer et al., 2004; Haas et al., 2007), the suggested plant equivalent of the late endosome (LE).

It is clear that colocalization with FM4-64 alone is totally insufficient as an indication for involvement in endocytosis, because, depending on the time of incubation, cell type, and experimental conditions, the tracer can label almost any compartment of the cell (Bolte et al., 2004). As a consequence, carefully timed experiments revealed that the plant TGN may behave as an EE because it is labeled by FM4-64 prior to the PVC, and it is therefore thought to be the first site of intersection between the biosynthetic and endocytic pathways (Dettmer et al., 2006; Lam et al., 2007a). However, punctate structures other than the TGN are labeled early after addition of the tracer and could represent an even earlier compartment in the endocytic route (Lam et al., 2007a). If this can be confirmed and followed by identification of independent markers for these structures, it has to be considered that the TGN acts as a recycling endosome in plants, and the localization of the GTPase Rab11 on these structures seems to point in this direction (Chow et al., 2008). By similarity to other systems, plants may also have a recycling endosome, distinct from the TGN/EE, from where endocytosed proteins, and possibly biosynthetic cargo, could be delivered to the PM. However, a recycling endosome has not been identified yet in plants, although it may be the compartment through which PIN proteins cycle in their way back to the PM (Lam et al., 2007b).

DIFFERENT VACUOLES AND DIFFERENT SORTING ROUTES WITHIN THE SAME CELL

During the last decade, it has been a popular concept to single out plants for their extraordinary complexity of the vacuolar transport. This was inspired by published evidence suggesting that soluble and membrane-spanning proteins reach the vacuoles by different routes (Gomez and Chrispeels, 1993), by a difference in sensitivity to wortmannin when comparing two soluble vacuolar cargos (Matsuoka et al., 1995), and by reports showing that compartments containing lumenal and membrane markers characteristic of lytic and storage vacuoles can be differentiated in certain cell types (Hoh et al., 1995; Paris et al., 1996; Di Sansebastiano et al., 2001; Park et al., 2004). Moreover, in Mesembryanthemum crystallinum, functionally different vacuoles within a single leaf mesophyll cell can be clearly recognized (Epimashko et al., 2004), and in pea (Pisum sativum) cotyledon parenchyma cells from embryos PSV and lytic vacuoles are also morphologically distinguishable (Hoh et al., 1995). However, recent work has not found evidence of distinct vacuoles in Arabidopsis embryos, pea, and barley (Hordeum vulgare) roots (Otegui et al., 2006; Hunter et al., 2007; Olbrich et al., 2007). It is as yet unclear whether this reflects species differences, but further work should be carried out to investigate this in more detail. Even if plant cells contain a single vacuole type, different routes to that vacuole may exist, similarly to the situation in yeast or mammals. Indeed, there is evidence of segregation of vacuolar cargo to different subdomains of the Golgi and distinct vesicles (Otegui et al., 2006). Differences in the sensitivity to drugs (Matsuoka et al., 1995) or to dominant-negative inhibitors, such as GTP-trapped ARF1 (Pimpl et al., 2003), could be used as arguments to justify different routes. However, the results could also be explained if two different vacuolar sorting signals bind the same receptor, but with different affinities. Under impaired transport conditions, strong ligands might still reach the vacuole while weak ligands are secreted. This may explain why some in vitro binding studies are above or below the detection limit (Kirsch et al., 1996) and also why inhibition of transport leads to differential dose-response characteristics (Pimpl et al., 2003). In addition, the knockdown of Arabidopsis BP80 has led to secretion of some storage proteins, but not all vacuolar cargo (Shimada et al., 2003), but this could be due to the fact that storage proteins are more abundant and have a lower affinity to BP80 because they can use alternative ways to target the vacuoles (i.e. by aggregation and incorporation into DVs). Indeed, an alternative study using artificial ligands and expressed at more equivalent levels revealed that BP80 knockdown leads to induced secretion of different types of vacuolar cargo without discrimination between different classes of sorting signals (Craddock et al., 2008). In summary, if two different vacuolar transport routes exist, it is important to identify reciprocal specificity of inhibitors or mutations so that one route can be inhibited without affecting the other and vice versa. This has not been accomplished yet.

OUTLOOK

The secretory pathway of plants is an important production platform of important biomolecules and deserves our full attention in the current climate of depleting energy resources that are easy to exploit. In addition, the complexity of the pathway is particularly challenging from a fundamental perspective because it contains many circular processes that require molecular switches, signal transduction, and feedback loops. Finally, it has now become clear that the plant secretory pathway is by no means identical to the pathway in yeasts or mammals and that many plant-specific features remain to be explored. This is best documented by a comparison of Rab GTPases and their subcellular localizations. Mammalian Rab11 is localized on the recycling endosomes, but plant Rab11 is found on the TGN (Chow et al., 2008). Mammalian Rab5 is a key molecule of EE, whereas in plants this group is found on the PVC (Kotzer et al., 2004; Dettmer et al., 2006; Haas et al., 2007) and contains a unique plant-specific N-myristoylated Rab5 derivative (Ueda et al., 2001). Finally, mammalian Rab7 is found on LEs, whereas in plants it decorates the tonoplast (Saito et al., 2002). If Rab GTPases would be used as signatures for organelle identities, we could thus suggest that the phragmoplast is not a derivative of the PM, but simply a specialized form of TGN (Chow et al., 2008) that perhaps later matures into a PM. Likewise, plant vacuoles may be equivalents of mammalian LEs rather than lysosomes. Clearly, we still have much to learn about the intricacies of the plant secretory pathway.