L. Jiang, State (China) Key Laboratory for Agrobaiotechnology, The Chinese University of Hong Kong, Shatin, New Territories, Hong Kong, China Fax: +852 2603 5646 Tel: +852 2609 6388 E-mail: email@example.com
In receptor-mediated sorting of soluble protein ligands in the endomembrane system of eukaryotic cells, three completely different receptor proteins for mammalian (mannose 6-phosphate receptor), yeast (Vps10p) and plant cells (vacuolar sorting receptor; VSR) have in common the features of pH-dependent ligand binding and receptor recycling. In striking contrast, the plant receptor homology-transmembrane-RING-H2 (RMR) proteins serve as sorting receptors to a separate type of vacuole, the protein storage vacuole, but do not recycle, and their trafficking pathway results in their internalization into the destination vacuole. Even though plant RMR proteins share high sequence similarity with the best-characterized mammalian PA-TM-RING family proteins, these two families of proteins appear to play distinctly different roles in plant and animal cells. Thus, this minireview focuses on this unique sorting mechanism and traffic of RMR proteins via dense vesicles in various plant cell types.
Eukaryotic cells share a common organization of organelles within their endomembrane systems, where each is a membrane-bound compartment which defines a separate environment for specific functions, and different organelles communicate with each other via transport vesicles. In general, a unique type of vesicle is required for each step in traffic, and transmembrane receptor proteins that are specific for one vesicle type recruit cargo that will be transported from one organelle to another in the step mediated by that vesicle [1–6]. A general principle that applies across eukaryotic species defines vesicle specificity: the cytoplasmic coat proteins that cause a vesicle to bud from its organelle source interact with the specific receptor proteins and cause them to partition with their cargo into the budding vesicle [7,8]. Thus, in general terms, a sorting receptor is specific for one vesicle type that traffics in one specific step between two endomembrane organelles.
The endomembrane systems for animal, yeast and plant cells have in common the presence of an organelle with an acidic lumenal pH that serves as a digestive compartment, the lysosome or vacuole [9,10]. In general, soluble proteins within the lumen of the endomembrane system that are destined for the lysosome/vacuole traffic through the Golgi apparatus where they are recruited at the trans-face into clathrin-coated vesicles (CCVs) by receptors that are unique for each type of organism but share in common the ability to bind a specific feature on the ligand protein at neutral pH, and then release the ligand protein upon encountering an acidic pH when the transport vesicles fuse with the target endosome or prevacuolar compartment (PVC) [11–16]. The sorting receptors then recycle back to the Golgi apparatus in a second type of CCV [17,18]. This recycling mechanism makes the sorting process efficient, in that one receptor can participate in sorting multiple ligand molecules.
In plant cells, the vacuolar sorting receptor (VSR) proteins that participate in this trafficking step belong to the BP-80 protein family [19–25]. The best-studied members of the family recognize a protein sequence on ligand molecules that contain a central NIPR (Asn-Pro-Ile-Arg) or similar motif [14,26]. The ligand-binding specificity of one BP-80 protein was studied by expressing in insect cells and then purifying from the culture medium the BP-80 lumenal domain (termed tBP-80). The most N-terminal ∼ 100 residues defined a domain that is also highly conserved in the lumenal sequences of what we termed receptor homology domain-transmembrane sequence-RING-H2 (RMR) proteins . Results from ligand-binding studies were consistent with a model in which the ligand-binding domain was contained within the N-terminal unique region, and where the RMR domain contributed to ligand binding . The receptor homology domains found within BP-80 and RMR proteins were subsequently designated the protease-associated (PA) domain which is important for substrate or ligand binding [28,29]. These experiments provided a reason to hypothesize that RMR proteins themselves might have a function in binding different types of ligands and might also serve as sorting receptors. However, the native ligands for most VSRs and RMRs remain to be identified and characterized in plants .
This possibility was subsequently considered in light of observations indicating that plant cells could contain two different types of vacuoles, a lytic or digestive vacuole and a vacuole that stored proteins [1,16,31], that the storage vacuole was served by an intracellular pathway different from that trafficked by BP-80 [13,32], and that so-called ‘dense vesicles’ (DVs) trafficked specifically to storage vacuoles .
The RMR protein family in plants
The identification and characterization of PA-TM-RING proteins in plants were not achieved until recently. The plant RMR was first identified by homology search using the pea VSR BP-80 N-terminal amino acid sequence. JR700 (Arabidopsis RMR1 or AtRMR1) and JR702 (Arabidopsis RMR2 or AtRMR2) from Arabidopsis were subsequently cloned and characterized . Further genomic analysis indicated that the Arabidopsis genome contains five RMR genes (AtRMR1–5), whereas the rice (Oryza sativa) genome has two RMR genes (OsRMR1–2). All of these RMRs share high amino acid sequence similarity (Fig. 1D), but relatively little is known about their subcellular localization and function as well as their potential ligands in plants. Structurally, similar to VSR, RMR is predicted to be a type I integral membrane protein that contains a typical N-terminal signal peptide, followed by a PA domain likely responsible for protein–protein interaction  and a single transmembrane domain. In contrast to the short cytoplasmic tail of VSR, the plant RMR has a long cytoplasmic tail (CT) with a typical C3H2C3 RING-H2 domain (Fig. 1A).
The transmembrane domain (TMD) and CT sequences of the Arabidopsis AtRMR1/2 and the rice OsRMR1/2 are quite similar to the corresponding regions of the PA-TM-RING proteins from mice, chicken and humans, in particular, their TMD and RING-H2 domain sequences are highly conserved, with similar spacing between the domains (Fig. 1B), indicating the probability of a similar function among these proteins. The C3H2C3 RING-H2 domain is associated with different biological functions in proteins from both mammalian cells and plants, such as functioning as transcriptional regulators [35,36] and as a ubiquitin–protein ligase [37–40]. In mammalian cells, the function of the RING-H2 domains of the PA-TM-RING proteins has been relatively well studied [41–43], but the function of the RMR RING-H2 domains in plants remains elusive.
RMR proteins traffic in a pathway different from that of BP-80
In order to gain insight into the function of RMR proteins, the intracellular localization and trafficking of RMR was studied in different plant cells and tissues [33,34,44–46]. Immunofluorescence and immunoelectron microscopic studies with purified antibodies raised either to a recombinant protein containing part of the RMR lumenal domain, or to a peptide representing a unique sequence in the RMR protein cytoplasmic tail gave similar results. In sections of tomato seeds where protein storage vacuoles (PSVs) are large and easily visualized, RMR proteins were present within PSVs and localized to large intravacuolar structures termed ‘crystalloids’; two other integral membrane proteins also colocalized to crystalloids . Biochemical analysis of purified crystalloid demonstrated a high ratio of lipid to protein. All of these observations were consistent with the concept that crystalloid represented intravacuolar arrays of lipid bilayers into which both integral membrane proteins and soluble proteins were packed . Subsequent studies using PSVs from plants in the Brassicaceae family, which lack microscopically defined crystalloids, demonstrated that their PSVs also contained an internal, covalently cross-linked network of integral membrane proteins, including RMR proteins . Thus the concept that formation of PSVs in plant seed embryos involves internalization of membranes containing specific integral protein markers may be generally applicable.
A second experimental approach was used to define the pattern of RMR protein organelle traffic . In these experiments, a chimeric integral membrane reporter protein comprised of a lumenal proaleurain reporter domain linked to AtRMR2 transmembrane sequence and cytoplasmic tail (designated Re-R-R). Proaleurain is the precursor of aleurain, a vacuolar cysterine protease from barley that is processed into its mature form in lytic PVC. The report was expressed, and its traffic was compared with that of a similar reporter proteins, but in one case containing the pea BP-80 transmembrane sequence and cytoplasmic tail (designated Re-B-B, known to traffic from ER to Golgi to lytic PVC), and in a second case containing the BP-80 transmembrane sequence but with the alpha-tonoplast intrinsic protein (a PSV marker) cytoplasmic tail (designated Re-B-alpha, known to traffic directly from ER to a PSV PVC). The proaleurain reporter moiety would be proteolytically processed by a specific maturase [48,49] if it reached the lytic PVC, and traffic into the Golgi would be assessed by evaluating whether the reporter protein acquired complex modifications to its two Asn-linked oligosaccharide chains . The results are summarized in Table 1, and document that the RMR reporter protein entered the Golgi apparatus because it acquired complex glycans, but it did not traffic to the lytic PVC . Thus, RMR proteins trafficked through the Golgi apparatus in a pathway distinct from that of BP-80, and were directed to a protein storage vacuole equivalent in the suspension cultured protoplasts that also contained alpha-tonoplast intrinsic protein, whereas in plant seed embryos the RMR proteins were concentrated in internal membrane arrays in PSVs.
Table 1. Exploration and determination of RMR or VSR protein trafficking via reporter fusion protein. TMD, transmembrane domain;CT, cytoplasmic tail; α-TIP, alpha-tonoplast intrinsic protein; LIVSR, lily vacuolar sorting receptor; LISCAMP, lily secretory carrier membrane protein; GFP, green fluorescent protein; TGN, trans-Golgi network; ER, endoplasmic reticulum; NA, not determined.
Lumenal proaleurain reporter domain + BP-80 TMD and CT (Re-B-B)
ER to Golgi to lytic PVC
Lumenal proaleurain reporter domain + AtRMR TMD and CT (Re-R-R)
ER to Golgi to lytic PVC to lytic vacuole in pollen tube
LlSCAMP + GFP
PM to apical endocytic vesicels to TGN to lytic vacuole in pollen tube
The growing pollen tube is an ideal single-cell model system to study protein trafficking and their functions in the secretory and endocytic pathways in plants. The dynamics and function of BP-80 and secretory carrier membrane protein (SCAMP) were also recently characterized in growing lily (Lilium longiflorum) pollen tube . SCAMP localized to early endosomes, plasma membrane and cell plate in plant cells . Both lily BP-80/VSR and SCAMP cDNAs (termed LIVSR and LISCAMP respectively) were cloned and used to make green fluorescent protein (GFP) fusions for transient expression under the control of the pollen specific promoter ZM13 in germinating lily pollen tubes via particle bombardment for protein trafficking studies. Interestingly, GFP–LISCAMP was mainly concentrated in the tip region (Fig. 2A), which is enriched with plant endocytic vesicles and early endosomes 50–200 nm in size . By contrast, GFP–BP-80/GFP–LIVSR were found to locate throughout the pollen tubes except the apical clear zone region (Fig. 2B) and were concentrated in ∼ 0.2-μm diameter punctate organelles that represent prevacuolar compartments for the lytic vacuole. In addition, microinjection of VSR or SCAMP antibodies significantly reduced the growth rate of the lily pollen tubes . Because VSRs mediate vacuolar protein transport , whereas SCAMPs may play roles in endocytosis [50,52] as well as cell plate formation , these results together suggest that both VSR and SCAMP are required for pollen tube growth, likely working together in regulating protein trafficking and membrane flow in the secretory and endocytic pathways which need to be coordinated in order to support pollen tube elongation.
RMR proteins may also function in pollen tube growth because microarray data analysis of gene expression in Arabidopsis (GENEVESTIGATOR, https://www.genevestigator.com/gv/index.jsp) shows that AtRMR3 is highly expressed in pollen compared with other AtRMRs in various tissues (unpublished results). We have thus recently taken a similar approach to study the dynamics and distribution of GFP-tagged RMR proteins using the same pollen tube transient expression system. As shown in Fig. 2C, when transiently expressed in a tobacco pollen tube, a weak GFP–AtRMR3 signal was diffusely distributed throughout the length of the growing pollen tube but missing from the tip region, and concentrated within some large ∼ 1–2-μm organelles (Fig. 2C) that were mobile (data not shown), a pattern that was different from those of GFP–LlSCAMP (Fig. 2A). Given the known association of RMR proteins with protein storage vacuoles or their PVCs in other plant systems, we tentatively identify these structures as pollen tube PSVs or their PVCs, although a firm identification will require further colocalization studies with markers for other organelles and/or immunogold EM studies.
Role of RMR proteins as sorting receptors
The ability of the AtRMR2 lumenal domain to bind potential protein ligands was evaluated using the recombinant protein expressed in insect suspension culture cells from which it was secreted into the culture medium and purified [44,45]. It should be noted that all RMR lumenal domains contain two conserved sites for Asn-linked glycosylation (Fig. 1A), and use of the insect cell expression system allowed assurance that proper glycosylation would be achieved . This consideration was relevant because the relatively large size of such glycans would impose steric limitations on interactions of the relatively small RMR protein with potential ligands.
The experimental approach evaluated interactions with two distinct types of known vacuolar sorting determinant sequences. The first type is the NPIR (Asn-Pro-Ile-Arg) motif recognized by the VSR proteins, whereas the second type is demonstrated by two different C-terminal propeptide sequences representing the class of targeting signals that have no apparent sequence conservation but the function of which requires placement at the C-terminus of ligand proteins [31,53]. It had been hypothesized that the latter directed proteins into the pathway to PSVs , and subsequent studies using genetic approaches in Arabidopsis identified a specific SNARE complex, important for membrane fusion in eukaryotes, to be essential for traffic through the pathway required for vacuolar targeting of ligands carrying C-terminal vacuolar sorting determinants (defined as the PSV pathway), but not the pathway for traffic to a lytic vacuole [55,56].
Park et al.  assessed binding of the AtRMR2 lumenal domain to synthetic peptides of defined sequences that were coupled to agarose beads. AtRMR2 bound specifically to known C-terminal vacuolar sorting determinant sequences, but only if they were presented with a free C-terminus. Interestingly, binding of the RMR protein to these C-terminal sorting determinant sequences was not pH dependent; in contrast to the interaction of BP-80 with its sequence-specific ligands, the RMR protein could not be eluted from the peptide–agarose beads by treatment at pH 4. In addition, specific binding was blocked by the C-terminal addition of two Gly residues, a modification known to prevent function in vacuolar sorting . Specific binding to peptides carrying sequence-specific sorting determinants was not observed. Thus, RMR proteins specifically bind to peptides corresponding to sorting determinants for the PSV pathway, which is distinct from the pH-dependentBP-80/AtVSR1 sorting pathway to the lytic vacuole.
The functional relevance of these in vitro binding results was further tested by expressing pairs of recombinant proteins, each representing a receptor lumenal domain and a soluble ligand carrying complementary halves of the GFP molecule, by transient expression in tobacco suspension culture protoplasts. In this bimolecular fluorescence complementation assay , reconstitution of a fluorescent GFP molecule occurs only when the two halves are brought into close proximity through interaction of the recombinant proteins to which they are attached. The obtained results indicated that BP-80 preferentially interacted with the vacuolar targeting sequence of lytic vacuole marker proaleurain rather than the C-terminal propeptide of the PSV marker chitinase . Conversely, AtRMR2 preferentially interacted with the chitinase C-terminal propeptide but not with the proaleurain targeting sequence. These results were consistent with the in vitro binding assay results and indicated that the AtRMR2 lumenal domain could interact in a specific manner with the chitinase C-terminal vacuolar sorting determinant in vivo.
In a separate series of experiments, the reporter protein Re-R-R with either GFP or monomeric red fluorescent protein (mRFP) inserted into its cytoplasmic tail was transiently expressed in the suspension culture protoplasts. Consistent with previous findings that endogenous RMR proteins were internalized into PSVs in developing seed embryos, Re-R-R tagged with either fluorescent molecule was present in small punctate cytoplasmic organelles, but also was internalized into the lumen of the protoplasts’ central vacuoles. Thus, traffic of these proteins, which as previously shown  was determined by sequences in the AtRMR2 cytoplasmic tail, resulted in the cytoplasmic tails containing the fluorescent tags being transferred from the cytoplasm to the vacuole lumen.
A different study used the lumenal domain of AtRMR1 expressed in bacterial system for binding studies . Those authors found that the AtRMR1 protein bound to C-terminal vacuolar sorting sequences but not to sequence-specific sorting sequences, and that binding was pH dependent and was abolished at pH 4. In addition, they presented data that argued for recycling of the AtRMR1 protein in transient expression experiments in Arabidopsis suspension culture protoplasts. These results and those obtained for AtRMR2, as well as experiments localizing endogenous RMR proteins in vivo [33,34], appear to be contradictory. However, the possibility remains that AtRMR1 has substantially different ligand-binding properties and patterns of traffic within cells. Future genetic study using knockout mutants of individual AtRMRs or coexpression of AtRMR1 and AtRMR2 in the same cells may be able to address these differences.
Spatial regulation of ligand sorting by RMR proteins in the Golgi apparatus
From studies using developing pea cotyledons, Hinz et al. have provided elegant data to argue for traffic of seed-storage proteins in DVs, separate from BP-80 proteins which are predominantly present in CCVs. Using quantitative analyses at the electron microscope level, those authors demonstrated that globulin-type storage proteins form aggregates in the cis-Golgi that partition at the periphery of cisternae and then move sequentially towards the trans-face where they bud off as DVs. By contrast, BP-80 receptors were localized predominantly at the trans-Golgi and were associated with CCVs . They therefore proposed a novel model whereby spatial regulation of sorting within the Golgi apparatus might explain how traffic of storage proteins to PSVs could be separated from traffic of proteins destined to be carried by CCVs to the lytic PVC.
In a subsequent study, those authors quantitatively analyzed the distribution of AtRMR2, Arabidopsis AtVSR proteins (BP-80 homologs) and the storage protein cruciferin in the Golgi apparatus and vesicles during Arabidopsis embryo development . In contrast to Otegui et al. (2006) , but consistent with prior results in the pea system, cruciferin was present predominantly at the periphery in the cis and medial cisternae and in DVs. AtVSR labeling was predominantly at the trans-face and in CCV, with very small amounts associated with DVs. By contrast, labeling for AtRMR2 in the Golgi and DVs was very similar to that for cruciferin. These results were interpreted to support the concept that RMR proteins were associated with sorting of storage protein aggregates into DVs. Consistent with findings from other studies, labeling for AtRMR2 on organelles representing PVCs was predominantly internal, providing further support that these proteins are internalized into organelle lumens during their traffic to the PSV. Such internalization would remove the possibility that AtRMR2 could recycle back to the Golgi apparatus to participate in more than one round of ligand sorting.
How could RMR proteins serve as efficient sorting receptors if they do not recycle? The aggregation model for storage protein sorting  may provide an explanation. By interacting with an aggregate of many storage protein molecules as the aggregate is sorted into a DV, a limited number of RMR proteins could participate in DV coat protein formation and efficiently promote sorting [33,44].
The process of internalization of RMR proteins into prevacuolar organelles would result in removal of cytoplasmic tails of the proteins from the cytoplasm. The RING-H2 domain found in mammalian RMR protein homologs has been shown to function as a ubiquitin–protein ligase [37,38]. There is no direct experimental evidence that plant RMR proteins have a similar ubiquitin–protein ligase activity although it is reasonable to hypothesize so. Consistent with this possibility, our preliminary studies using immunogold EM labeling with anti-ubiquitin sera on ultrathin sections of cells from developing tobacco seed embryos prepared by high-pressure freezing/frozen substitution demonstrated positive labeling in the protein storage vacuole crystalloid (Fig. 3). This result would be consistent with the concept that RMR proteins are internalized into the PSV as it develops, and intermolecular ubiquitination might help explain the observation that ‘crystalloid’ proteins from Brassica napus were cross-linked in a manner that resisted treatment with disulfide reducing agents . Such hypothesis of ubiquitin-mediated cross-linking during internalization of proteins into the PSV could be tested in future experiments by isolation and biochemical analysis of PSVs.
Both the mammalian GRAIL and RNF13 proteins affect complex functions in cells where they are expressed. In the case of the RNF13 protein, the cytoplasmic tail is cleaved from attachment to the TMD during traffic to endosomes; the now free cytoplasmic tail with its ubiquitin–protein ligase activity has been postulated to provide a mechanism for activation of signaling pathways that would affect cell functions and fate . Although genes encoding the beta and gamma secretase proteases that are hypothesized to participate in such a cleavage process  are not present in plant genomes , it is possible that some other mechanism for cleavage of plant RMR protein cytoplasmic tails within the basic region separating the transmembrane and RING-H2 domains conserved in both plant and mammalian proteins (as indicated in Fig. 1A,B) might exist. Thus there may be an advantage to the cell to have these relatively abundant proteins removed from the cytoplasm as they reach the terminus of their trafficking pathway. Whether the free tail would participate in some signaling process remains to be tested experimentally.
Conclusion and future perspective
In conclusion, Fig. 4A summaries the subcellular localization, trafficking and possible function of RMRs in developing seeds, where RMR-mediated storage protein sorting is achieved via concentration sorting in storage PVC [or dark intrinsic protein (DIP) organelles] or DVs (Fig. 4A). In addition, the three integral memebrane proteins, RMR, VSR and SCAMP, show distinct patterns of subcellular localization and dynamics in the same growing pollen tubes (Fig. 4B), indicating their distinct functional roles and transport pathways in plants. However, the native ligands or cargo proteins for most VSRs and RMRs in plants remain elusive.
To identify native cargo proteins for the Arabidopsis AtVSRs and AtRMRs, we have recently developed and used a transgenic Arabidopsis suspension culture cell system expressing the N-terminus of VSR or RMR (lacking its TMD and CT) so that the secreted truncated VSR or RMR proteins would bring along their native cargoes into the culture media to be identified by LC-MS/MS analysis , however, this approach would be difficult to carry out for RMR cargo identification if RMR binds to aggregates. Such a biochemical/cell biology approach for functional characterization of VSR and RMR, as well as their cargo proteins in plants, will likely generate novel information to complement genetic approaches.
Published studies of the luminal domain of plant RMRs suggest that these proteins function as sorting receptors for transporting storage proteins to PSVs in plants. However, the functional roles of the RMR C-terminal RING-H2 domain remain largely unknown compared with that of the mammalian PA-TMD-RING proteins. Because the RING domains are highly conserved between the plant RMR and the mammalian PA-TMD-RING proteins (Fig. 1A–C) and because ubiquitin was localized in the PSV crystalloid where RMR proteins are concentrated (Fig. 3), it is reasonable to hypothesize that plant RMR proteins may also have a similar ubiquitin–protein ligase activity. Such hypothesis can be tested via in vitro ubiquitin–protein ligase activity analysis in future experiments.
Our work has been supported by grants from the Research Grants Council of Hong Kong (CUHK488707, CUHK465708, CUHK466309, CUHK466610 and HKUST6/CRF/08), UGC-AoE, CUHK Schemes B/C.