Endocytosis plays an important role in plant physiology, but how endocytic organelles are organized remains unknown. We present the evidence that endosomes are functionally differentiated in Arabidopsis cells. Two types of Rab5-related GTPases are localized on distinct population of endosomes in a partially overlapping manner. Ara7 and Rha1 are on an early type of endosomes with AtVamp727, where recycling of plasma membrane proteins occurs. In contrast, the plant-unique Rab5, Ara6, resides on distinct endosomes with the prevacuolar SNAREs. Partially overlapping localization of Ara6 and Ara7/Rha1 with reciprocal gradients suggests maturation of endosomes from one to the other.
In the animal system the endocytic pathway plays an essential role in a variety of cell processes such as signal transduction, recycling of synaptic vesicles, and uptake of nutrients. The organelles involved in this endocytic pathway, collectively called endosomes, are roughly classified into four classes in mammalian cells, that is early endosomes, late endosomes, recycling endosomes, and lysosomes (Mellman, 1996). Among them, early endosomes and recycling endosomes are regarded as the site of recycling of internalized proteins and membranes to the plasma membrane (PM). In higher plants, endocytic organelles are involved in plant functions of a higher order such as pattern formation (Geldner et al., 2001), but our understanding of how plant endosomes are organized and functionally differentiated remains very limited.
Rab GTPases, key regulators of vesicular transport, are useful organelle markers because they strictly localize to the specific organelles where they function. Whole genome sequencing revealed that the Arabidopsis genome encodes 57 Rab GTPases, which can be classified into eight subgroups in accordance with the similarity to mammalian Rab GTPases. Four of the eight subgroups are categorized as Rab5, Rab7, Rab11, and Rab18 subgroups, which function in the endocytic pathway in mammalian cells. No homologs are found in the Rab4 and Rab9 subgroups, which are also involved in endocytosis in mammals. For the subgroup of Rab5, Arabidopsis has three members, Ara6, Ara7 and Rha1. Ara6 is a plant-unique type Rab5 homolog and has several interesting features in its structure (Ueda et al., 2001). Ara6 lacks the C-terminal Cys-motif to be isoprenylated, which is conserved in most Rab GTPases and is known to be essential for their membrane binding and function. Instead, Ara6 harbors an N-terminal amino acid stretch where this protein is fatty-acylated. Furthermore, the sequence of the effector domain is totally different from that of other Rab5 group members, although the overall sequence similarity to the Rab5 group is considerably high. The GTPase cycle of Ara6 should be regulated in a different way from conventional Rab GTPases, because Ara6 seems to be recycled independently of a general regulator of Rab proteins, Rab GDP dissociation inhibitor (Ueda et al., 2001). On the contrary, Ara7 and Rha1 are conventional-type Rab5 orthologs. Their C-terminal region contains the Cys-motif, and the sequence of the effector domain is identical to that of Rab5 group members in other organisms (Anuntalabhochai et al., 1991; Ueda et al., 2001). Rab5 is a key player in the homotypic fusion of early endosomes in mammalian cells. Arabidopsis Rab5 homologs have also been shown to function in the endocytic pathway. The expression of GTP-freeze mutants of Ara6 and Ara7 causes aggregation of deformed endosomes in Arabidopsis cells (Ueda et al., 2001). Using these Rab5 homologs and endosomal SNARE proteins as markers, we demonstrate that there are at least two populations of endosomes in Arabidopsis cells, which are functionally differentiated.
Results and discussion
We previously reported that Ara6 is localized on the subpopulation of endosomes stained by FM4-64, which is a tracer of the endocytic pathway (Ueda et al., 2001). To examine whether other Rab5 homologs, Ara7 and Rha1, are also on the subpopulation of endosomes or not, we stained Arabidopsis suspension cells expressing GFP-tagged Ara7 or Rha1 with FM4-64. As indicated in Figure 1(a), GFP fluorescence was always observed on a subpopulation of dotty organelles stained by this dye, as in Ara6-GFP. This indicates that plant endosomes are heterogeneous and perhaps differentiated. To examine whether these Rab5 homologs are on the same population of endosomes or not, they were tagged with different colors of fluorescent proteins, expressed in Arabidopsis cells, and observed with a color CCD camera. When GFP-tagged Ara6 and monomeric red fluorescent protein (mRFP) (Campbell et al., 2002)-tagged Ara7 were simultaneously expressed in the same cells, rapidly moving punctate organelles were observed with various colors in the cells (Figure 1b, see Supplementary material, Movie S1). Interestingly, most of them bore intermediate colors between red and green, such as orange, yellow, or yellowish green, whereas red and green dots were also observed. A similar result was observed when Ara6-GFP and mRFP-Rha1 were coexpressed (Figure 1b, Supplementary material, Movie S2), but the coexpression of GFP-Ara7 and mRFP-Rha1 gave only yellow staining, indicating that Ara7 and Rha1 are on the same population of endosomes (Figure 1b, Supplementary material, Movie S3). These results indicate that the endosomes in Arabidopsis cells differentiate into at least two populations and Ara6 and Ara7/Rha1 are localized on partially overlapping but different populations of endosomes with reciprocal gradients (Figure 4).
To further clarify the differentiation of endosomes, we employed SNARE molecules as another type of organelle markers. The Arabidopsis genome contains 54 genes encoding SNAREs, some of which have already been characterized (Sanderfoot et al., 2000). First, we examined whether known SNARE proteins colocalize with any of the Rab5-related proteins. The SNARE members were tagged with Venus, a modified bright version of the yellow fluorescent protein (Nagai et al., 2002), and Rab GTPases were tagged with GFP. Neither AtSyp31 (AtSed5, Golgi marker, data not shown) nor AtSyp41 [AtTlg2a, trans-Golgi network (TGN) marker, Figure 2b] colocalized with any Rab5-related proteins, ruling out the involvement of these organelles. AtSyp21 (AtPep12) and AtSyp22 (AtVam3) are known to localize and function in the prevacuolar compartment (PVC), a synonym of late endosomes (Sanderfoot et al., 2001). When Venus-tagged AtSyp21 or AtSyp22 are expressed in Arabidopsis cells, these fluorescent chimeric proteins are also observed on punctate organelles (AtVam3 is also found on the vacuolar membrane). The majority of these punctate organelles are neither Golgi nor TGN, because Venus-AtSyp21 or Venus-AtSyp22 rarely colocalize with GFP-AtSyp41 or GFP-AtSyp31 (Uemura et al., 2004). As shown in Figure 2(a), these PVC SNAREs showed good colocalization only with Ara6-GFP but not with Ara7 or Rha1, indicating that Ara6-GFP mainly resides on late endosomes. This observation also confirms that Ara6 and Ara7/Rha1 are on different populations of endosomes.
Which SNARE protein colocalizes with Ara7/Rha1 then? To answer this question, we systematically analyzed the subcellular localization of all Arabidopsis SNARE proteins, and found that one member of the VAMP family, AtVamp727, expresses a localization pattern very similar to that of Rab5-related proteins. When Arabidopsis cells expressing GFP-AtVamp727 were stained by FM4-64, GFP-AtVamp727 localized on a subpopulation of FM4-64-stained organelles, as in Rab5-related GTPases (Figure 1a). Thus we coexpressed Venus-AtVamp727 and GFP-tagged Rab5s in Arabidopsis cells. As shown in Figure 2(a), Ara7 and Rha1 colocalized well with this SNARE protein, whereas a small population of Ara6-residing endosomes still overlapped with AtVamp727.
The next important question is how these two types of endosomes could functionally differentiate in plant cells. The Arabidopsis gnom mutant gave us a clue on this issue. The gnom mutant plants display pleiotropic defects, among which the anomaly in pattern formation along the apical–basal axis is remarkable. Some severe gnom alleles show a complete loss of the apical–basal axis and form a round embryo (Jürgens, 1992). GNOM encodes a GDP/GTP exchange factor for the Arf GTPases (Arf GEF) (Steinmann et al., 1999). Arf GTPase regulates vesicle formation and sorting of cargo molecules in multiple steps of vesicular traffic including endocytosis in yeast and mammalian cells (D'Souza-Schorey et al., 1995; Yahara et al., 2001). Guanine nucleotide exchange is an essential step for the activation of this molecule. GNOM and some other members of the Arf GEF family are known as targets of a widely used inhibitor of vesicular traffic, Brefeldin A (BFA). By the use of a BFA-resistant mutant version, GNOM has recently been demonstrated to reside on endosomes and be essential for the recycling of an auxin efflux regulator, AtPin1 (Geldner et al., 2003). Contrary to the severe growth arrest of the gnom mutant plants, the suspension-cultured cells generated from the gnom mutant proliferate normally like wild-type cells. This provided us with a useful recycling-defective cell line to investigate the functional differentiation of endocytic organelles in Arabidopsis.
In the gnom mutant cells, Ara7-positive endosomes lose their fine dotty shape and appear as larger patches or clustering ring-shape structures (Geldner et al., 2003) (Figure 3). This suggests that the Ara7-positive endosomes are the site of GNOM-dependent recycling of PM proteins such as AtPin1. As Rha1 and AtVamp727 colocalize with Ara7, they are also expected to localize on the abnormally deformed endosomes when expressed in gnom-mutant cells. This was indeed the case as shown in Figure 3. In contrast, Ara6-GFP was found on such abnormal endosomes much less frequently, once again consistent with the distinct nature of the two types of endosomes. Other organelle markers, GFP-AtSyp41 (TGN), GFP-AtSyp21 and GFP-AtSyp22 (late endosomes), and GFP-AtSyp31 (Golgi), showed indistinguishable subcellular localization patterns in gnom and wild-type cells (Figure 3). From these observations, we conclude that the organelle where Ara7, Rha1, and AtVamp727 are mainly residing is the sole place where GNOM-dependent traffic takes place.
As described above, Ara6, AtSyp21 and AtSyp22 are enriched in the same type of endosomes, which is distinct from Ara7/Rha1-rich endosomes. Ara6-rich endosomes are likely to be located at a later stage of the endocytic pathway, because AtSyp21 and AtSyp22 are reported to function in membrane fusion in late endosomes (PVC) and/or the vacuole in the vacuolar transport pathway (Sanderfoot et al., 2001; Sato et al., 1997; da Silva Conceicao et al., 1997). Studies of yeast cells show that the endocytic pathway merges with the vacuolar transport pathway at late endosomes prior to reaching the vacuole, the final destination of the endocytic pathway. In plants the routes of endocytosis and vacuolar transport merge at the PVC (Tse et al., 2004). The Ara7/Rha1-rich endosomes probably represent an earlier compartment which is important for recycling to the PM. These two functionally different endosomes are morphologically indistinguishable at the confocal microscope level, and the two types of Rab5 homologs partially colocalize in various ratios on these endosomes. These findings imply that the two differentiated populations of endosomes are not completely stable compartments but can gradually change their components and properties as is the case of Golgi cisternae proposed in the cisternal maturation model (Pelham and Rothman, 2000). The partial functional overlap of the two types of endosomes may also explain the apparent discrepancy between our results and those recently published by Sohn et al. (2003). They showed that the overexpression of the GDP-freeze form of Ara7 or Rha1 blocks the transport of sporamin-GFP and aleurain-GFP to the vacuole and causes their missecretion, and proposed that Ara7 and Rha1 regulate the transport between the PVC and the vacuole. Although only limited populations of Ara7 and Rha1 were found on the PVC, it is possible that they still play an important role in the function of late endosomes/PVC, which may be perturbed by the GDP-freeze version of Ara7 and Rha1. Alternatively, the apparent discrepancy in our results may be due to differences, in cell type, protein markers, expression strategies, and so on, in the experimental systems used.
Dynamic rearrangement of Rab GTPases on endosomes has also been discussed for mammalian endosomes. Multiple combinations of three endocytic Rab GTPases, Rab5, Rab4 and Rab11, compose three major domains on a continuous endosome, one that contains only Rab5, and others with Rab4 and Rab5 or Rab4 and Rab11. Rab5 functions in the transport from the PM to early endosomes and the homotypic fusion of early endosomes. On the contrary, Rab4 and Rab11 are implicated in the fast recycling pathway from early endosomes and rather slow recycling from recycling endosomes, respectively (Somsel Rodman and Wandinger-Ness, 2000). The absence of Rab4 homologs in plants suggests that plant cells have evolved a unique system for rapid recycling between early endosomes and the PM. The Ara7-rich population of endosomes, where GNOM-dependent recycling takes place, may represent the functional counterpart of the recycling domain of early endosomes where Rab4 functions in mammalian cells. It has been proposed for animal endosomes that the divalent effector proteins of Rab GTPases are central components of the machinery regulating the subcompartmental organization (de Renzis et al., 2002). Studies of effector molecules for the Rab5-family proteins in plants will provide clues to unveil how and why the reciprocal distribution of these molecules are established and maintained (Figure 4).
In addition to Rab4, the Arabidopsis genome lacks Rab9 homologs, which are known to regulate traffic from late endosomes to the TGN in mammalian cells. By contrast, many diverse members of Arabidopsis Rabs (26 of 57) are scored as Rab11 homologs (Vernoud et al., 2003). Considering this unique composition of plant Rab proteins and their unique distribution, together with the unique property of the vacuolar system (Jauh et al., 1999; Paris et al., 1996), endocytic and post-Golgi transport pathways in plant cells should be organized and operated in a different manner from other organisms. Such unique features in vesicular traffic in plants may serve as the molecular basis of plant-unique functions of a higher order such as responses to environment (Kato et al., 2002), establishment of body axis (Friml et al., 2003), and organogenesis (Benkova et al., 2003).
In the experiments presented here, we used fluorescent protein-tagged Rab and SNARE proteins as organelle markers. We were aware of the possibility that overexpression of such fusion proteins might cause mislocalization. However, as already published, Ara6-GFP is correctly localized on the same organelles as endogenous Ara6 (Ueda et al., 2001). The localization of GFP-AtSyp22 is also indistinguishable from the endogenous AtSyp22 (Uemura et al., 2002). Furthermore, GFP-AtSyp22 complements the defect of gravitropism of the sgr3–1 mutant (M.T. Morita and M. Tasaka, Nara Institute for Science and Technology, Ikoma, Japan, personal communication), indicating that GFP-AtSyp22 is fully functional in vivo. Many published results with yeast and mammalian cells demonstrate that Rabs and SNAREs tagged with GFP or its analogs can remain functional and show proper localization if the experiments are performed carefully. It is not always easy to prove that every fusion protein used is completely functional, but we should avoid artifact of overproduction as much as possible, for example by looking at only weakly or moderately fluorescent cells.
Plasmid construction and transient expression
cDNAs for SNARE proteins were isolated by RT-PCR using total RNA prepared from Arabidopsis suspension-cultured cells as a template. cDNAs for Ara6 and Ara7 were isolated as described previously (Ueda et al., 2001), and cDNA for Rha1 was amplified by RT-PCR using total RNA prepared from suspension cells as a template. Open reading frames (ORFs) of ARA6, ARA7, and RHA1 were fused to the ORF of sGFP (provided by Yasuo Niwa of Shizuoka University) or mRFP (provided by Roger Y. Tsien of University of California at San Diego) in the directions of 5′- ARA6 XFP-3′, 5′- XFP ARA7–3′, and 5′- XFP RHA1–3′, respectively, and subcloned into the pHTS13, a derivative of pBSIIKS+, which contains the CaMV 35S promoter and the Nos terminator. ORFs for SNARE proteins were fused to the ORF of sGFP or Venus (provided by Atsushi Miyawaki of RIKEN) in the direction of 5′- XFP SNARE-3′, and subcloned into the expression vector which is derived from pUC18 and contains the CaMV 35S promoter and the Nos terminator. Transient expression of chimeric proteins was performed as described previously (Ueda et al., 2001). For the cotransformation with two plasmids, the ratio of plasmids added in the cells was changed according to the efficiency of transformation. In the co-expression experiments of two XFP-tagged Rabs, 20 μg of each plasmid was used in a single transformation procedure. In the co-expression experiments of the Rab and SNARE, 40 μg of plasmid containing XFP-tagged RAB and 1 μg of plasmid containing XFP-SNARE were used for a single transformation procedure to yield equivalent fluorescence, because promoter activity driving XFP-SNARE was much stronger than that driving XFP-tagged RAB.
Confocal laser scanning microscopy
For double labeling with sGFP and FM4-64, Arabidopsis cells expressing GFP-tagged Ara6, Ara7, or AtVamp727 were incubated in MS cell culture medium plus 0.4 m mannitol and 50 μm FM4-64 for 10 min. Cells were washed twice, incubated for 15 min at 23°C, and observed with a Zeiss confocal laser scanning microscope, LSM510 (Carl Zeiss Co., Ltd, Tokyo, Japan).
For the observation of cells expressing both GFP-tagged and Venus-tagged proteins, we used the META device attached to LSM510, which enabled us to separate GFP fluorescence from Venus fluorescence. Arabidopsis cells expressing two Rab5-related proteins, which were tagged with sGFP and mRFP, respectively, were observed with a confocal laser microscope system we developed: Olympus BX52 fluorescence microscope equipped with a confocal scanner Model CSU10 (Yokogawa Electric Corp., Tokyo, Japan) and an ORCA-3CCD color camera (Hamamatsu Photonics, Hamamatsu, Japan). Images were taken and processed with the Aquacosmos software (Hamamatsu Photonics).
Cells expressing only sGFP-tagged Rab or SNARE protein were observed with a confocal laser microscope system: Olympus BX52 fluorescence microscope equipped with CSU10 and a cooled CCD camera, ORCA-I (Hamamatsu Photonics). Images were obtained by the IPLab software (Scanalytics, Fairfax, VA, USA) and projection images were constructed using the VoxBlast software (VayTek, Fairfield, IA, USA).
We are grateful to Miyo Morita-Terao and Masao Tasaka of the Nara Institute for Science and Technology for the gnom suspension-cultured cell line, to Yasuo Niwa of Shizuoka University for sGFP, to Roger Y. Tsien of University of California at San Diego for mRFP, and to Atsushi Miyawaki of RIKEN Brain Science Institute for Venus. This work was supported by Grants-in-Aid for Scientific Research on Priority Areas and research grants from the Bioarchitect Project and the Technological Development for Imaging Dynamics of Biological Nano-machines Project of RIKEN, the Dynamic Bio Project of the New Energy and Industrial Technology Development Organization, and the Human Frontier Science Program.