Evidence for a sorting endosome in Arabidopsis root cells

Authors

  • Yvon Jaillais,

    1. Reproduction et Développement des Plantes, Institut Fédératif de Recherche 128, Centre National de la Recherche Scientifique, Institut National de la Recherche Agronomique, Université Claude Bernard Lyon I, Ecole Normale Supérieure de Lyon, 46 allée d’Italie, 69364 Lyon cedex 07, France
    Search for more papers by this author
  • Isabelle Fobis-Loisy,

    1. Reproduction et Développement des Plantes, Institut Fédératif de Recherche 128, Centre National de la Recherche Scientifique, Institut National de la Recherche Agronomique, Université Claude Bernard Lyon I, Ecole Normale Supérieure de Lyon, 46 allée d’Italie, 69364 Lyon cedex 07, France
    Search for more papers by this author
  • Christine Miège,

    1. Reproduction et Développement des Plantes, Institut Fédératif de Recherche 128, Centre National de la Recherche Scientifique, Institut National de la Recherche Agronomique, Université Claude Bernard Lyon I, Ecole Normale Supérieure de Lyon, 46 allée d’Italie, 69364 Lyon cedex 07, France
    Search for more papers by this author
  • Thierry Gaude

    Corresponding authorSearch for more papers by this author

*(fax +33 4 72 72 86 00; e-mail thierry.gaude@ens-lyon.fr).

Summary

In eukaryotic cells, the endocytic and secretory pathways are key players in several physiological processes. These pathways are largely inter-connected in animal and yeast cells through organelles named sorting endosomes. Sorting endosomes are multi-vesicular compartments that redirect proteins towards various destinations, such as the lysosomes or vacuoles for degradation, the trans-Golgi network for retrograde transport and the plasma membrane for recycling. In contrast, cross-talk between the endocytic and secretory pathways has not been clearly established in plants, especially in terms of cargo protein trafficking. Here we show by co-localization analyses that endosomes labelled with the AtSORTING NEXIN1 (AtSNX1) protein overlap with the pre-vacuolar compartment in Arabidopsis root cells. In addition, alteration of the routing functions of AtSNX1 endosomes by drug treatments leads to mis-routing of endocytic and secretory cargo proteins. Based on these results, we propose that the AtSNX1 endosomal compartment represents a sorting endosome in root cells, and that this specialized organelle is conserved throughout eukaryotes.

Introduction

Endocytosis corresponds to the internalization of molecules from the plasma membrane (PM) and their subsequent intracellular trafficking within the cell. In animals, endocytosed molecules are sorted among several intracellular compartments named endosomes, which are differentiated into four major classes: early endosomes, recycling endosomes, late endosomes and lysosomes (Perret et al., 2005; Raiborg et al., 2003). However, the classical view of endosomes as stable and distinct vacuolar compartments has been challenged by new findings that show endosomes as a continuum of compartments that undergo maturation (Miaczynska and Zerial, 2002; Perret et al., 2005; Rink et al., 2005). Thus, in yeast and mammalian cells, early endosomes mature into late endosomes through intermediate compartments that contain intra-lumenal vesicules, known as multi-vesicular bodies (MVB). During this maturation process, sorting of cargo proteins towards the trans-Golgi network (TGN), PM or lytic compartments occurs through a vast network of tubules. Specific associations of proteins with these tubules regulate protein sorting, such as the retromer complex for retrograde transport to the TGN, EHD1/RME1 and Rab11 for slow recycling to the PM, and the ESCRT (endosomal sorting complex required for transport) machinery for degradation to vacuoles/lysosomes (Bonifacino and Rojas, 2006).

Endocytic pathways are far less well understood in plants (Samaj et al., 2005). Indeed, for a long time, endocytosis was thought not to be possible in plant cells because of the existence of a cell wall and high turgor pressure. This dogma was recently rejected as intact plant cells can internalize PM proteins (Geldner et al., 2001), fluorescent dyes or sterols (Grebe et al., 2003). To date, only a few endosomal proteins have been described in plants, and endosome differentiation and functions are not well defined. The first plant proteins shown to associate with the membrane of endosomes were the Arabidopsis Rab GTPases AtRABF1 (also known as Ara6) and AtRABF2b (also known as Ara7) (Ueda et al., 2001, 2004). RABF1 and RABF2b are two of three Arabidopsis homologues of the mammalian small GTPase Rab5. Rab GTPases are key regulators of vesicular transport and vesicle fusion (Zerial and McBride, 2001). In the protoplasts of Arabidopsis suspension-cultured cells, AtRABF1 and AtRABF2b were located to two distinct populations of endosomes, with some overlap (Ueda et al., 2001, 2004). By following internalization of the fluorescent lipophilic styryl dye FM4-64, AtRABF1 was localized mainly to a late type of endosome, whereas AtRABF2b was localized on early endosomes (Ueda et al., 2004). It was suggested that the partially overlapping localization of RABF2b and RABF1 reflects maturation of endosomes from one type to the other. In contrast, in tobacco leaf epidermal cells, AtRABF2b was localized to the late endosomal/pre-vacuolar compartment (PVC) (Kotzer et al., 2004). Localization of RABF proteins to the PVC was also seen in other cell types (Bolte et al., 2004a; Haas et al., 2007; Lee et al., 2004; Sohn et al., 2003). The contradiction regarding the localization of RABF1 and RABF2b, which is probably due to the various model systems employed, precludes the use of these proteins as specific markers to distinguish between early and late endosomes/PVC. However, both proteins have been used in several studies as general endosomal markers (Dettmer et al., 2006; Dhonukshe et al., 2006; Geldner et al., 2003; Grebe et al., 2003; Jaillais et al., 2006; Takano et al., 2005). Recently, we demonstrated the existence of at least two distinct endosomal compartments in Arabidopsis root cells, depending on the presence of either the GNOM or AtSNX1 protein (Jaillais et al., 2006). GNOM is a GDP/GTP exchange factor for small G-proteins (ARF-GEF) that is required for cycling of some auxin transporters (Geldner et al., 2003). AtSNX1 is the first-described plant sorting nexin and is homologous to human SNX1 (Vanoosthuyse et al., 2003). HsSNX1 is a member of the retromer complex in mammalian cells (Seaman, 2005). The retromer complex has a conserved structure and function from yeasts to mammals. It is composed of five proteins involved in the recycling/retrograde transport of vacuolar sorting receptors (VSRs) from endosomes to the TGN (Seaman, 2005). VSRs are transmembrane proteins that transport lytic enzymes from the TGN to the lytic compartments (lysosome/vacuole).

By searching genome databases, we found that all members of the mammalian retromer complex are conserved in Arabidopsis (Vanoosthuyse et al., 2003). Recently, Oliviusson et al. (2006) showed that three proteins of the Arabidopsis retromer complex, namely VPS35 (At3g51310), VPS29 (At3g47810) and VPS26 (At5g53530), co-localize with VSRs in the PVC. PVC has been described as a multi-vesicular compartment enriched in VSRs (Tse et al., 2004). The dissociation between VSRs and their ligands is thought to take place within the PVC due to the low luminal pH of this compartment. Then VSRs are recycled back to the TGN for an additional round of sorting, while lytic enzymes are routed towards the lytic vacuoles (daSilva et al., 2005). As AtSNX1 is a putative component of the Arabidopsis retromer complex, we surmised that AtSNX1 would also be localized to the PVC, like the other three proteins of the retromer analysed so far. Here we show, by using transgenic Arabidopsis lines expressing diverse endomembrane markers, that endosomes labelled with AtSNX1 overlap with the pre-vacuolar compartment in Arabidopsis root cells. Furthermore, we demonstrate that alteration of the routing functions of AtSNX1 endosomes leads to mis-routing of both endocytic and secretory cargo proteins in planta. Previous work based on the analysis of FM4-64 internalization in tobacco BY-2 cells suggested that the PVC lies on the secretory and endocytic pathways (Tse et al., 2004). Here we confirm the role of the PVC as a crossroads merging the secretory and endocytic pathways by investigating the intracellular trafficking of fluorescent-tagged plasma membrane proteins in root cells. Moreover, internalization of known recycling plasma proteins through AtSNX1 endosomes indicates that the AtSNX1 endosomal/PVC compartment might function as a sorting endosome from which endocytosed plasma membrane proteins and secretory proteins are sorted to diverse destinations. This suggests that sorting endosomes are conserved throughout eukaryotic cells.

Results

AtSNX1 endosomes and PVC define a similar compartment in Arabidopsis root cells

To label the PVC in plant cells, we generated Arabidopsis transgenic lines that constitutively express a fluorescent PVC marker (see Supplementary Table S1). This marker is a truncated form of the binding protein of 80 kDa (BP80), an extensively studied pea VSR (Brandizzi et al., 2002; Paris and Neuhaus, 2002; daSilva et al., 2005; Tse et al., 2004). The intra-luminal part of BP80, involved in interaction with the transported lytic enzymes (e.g. the protease aleurain), was removed and replaced by a fluorescent protein (YFP). The remaining transmembrane and cytosolic domain of BP80 retains all the trafficking signals required for normal routing of the protein, and is a bona fide PVC reporter when constitutively over-expressed in BY-2 cells (Tse et al., 2004). Localization of GFP/YFP fusion proteins has been shown to correlate with that of the native proteins (Tse et al., 2004; Xu et al., 2006). It is reasonable to consider that the behaviour of fluorescent proteins reflects the genuine location and dynamics of endogenous proteins, although we cannot exclude the possibility that the fluorescent-tagged proteins might have altered location or stability. We decided to use Arabidopsis root tips as a model system for our analysis, because endocytosis has been extensively studied in this tissue (Abas et al., 2006; Dettmer et al., 2006; Dhonukshe et al., 2006; Geldner et al., 2001, 2003; Grebe et al., 2003; Jaillais et al., 2006; Paciorek et al., 2005; Russinova et al., 2004). To compare the location of YFP–BP80-labelled organelles with compartments of the endocytic pathways, we used the fluorescent dye FM4-64 as an endocytic tracer (Dettmer et al., 2006; Geldner et al., 2003; Grebe et al., 2003; Jaillais et al., 2006; Paciorek et al., 2005). We detected YFP–BP80 labelling in punctate compartments that were also stained with the FM4-64 dye after 15 min of treatment (Figure 1a–c). This observation confirms previous data obtained in BY-2 cells (Tse et al., 2004). Next, we looked at the effects of the fungal toxin brefeldin A (BFA), which is known to interfere with vesicle trafficking, on the location of YFP–BP80. In Arabidopsis root tip cells, BFA causes the TGN, GNOM and AtSNX1 endosomes to aggregate, with endocytosed FM4-64 being located inside an aberrant compartment (the ‘BFA compartment’) surrounded by Golgi stacks (Dettmer et al., 2006; Geldner et al., 2003; Grebe et al., 2003; Jaillais et al., 2006). Surprisingly, while the YFP–BP80-labelled PVC is insensitive to BFA in BY-2 cells (Tse et al., 2004), we found that YFP–BP80-containing organelles accumulated into FM4-64-labelled BFA compartments in Arabidopsis root cells (Figure 1d–f, Figure S1). Similarly, we observed that GFP–RABF2b compartments were labelled with FM4-64 (Figure 1g–i) and accumulated into BFA compartments (Figure 1j–l). Such cell selectivity of the BFA effect has already been described, for example between root and mesophyll cells (Russinova et al., 2004), or between epidermal and stele cells (Paciorek et al., 2005).

Figure 1.

 YFP–BP80 localizes to endosomes in Arabidopsis root cells.
(a–f) Root tip cells of YFPBP80 seedlings after 15 min of FM4-64 uptake and treated concomitantly with mock (a–c) or BFA treatment (d–f).
(g–l) Root tip cells of GFP–RABF2b seedlings after 15 min (g–i) or 30 min (j–l) of FM4-64 uptake and treated concomitantly with mock (g–i) or BFA treatment (j–l). Scale bars = 5 μm.

Inhibitors of phosphatidylinositol-3-OH (PI-3) kinase such as wortmannin (Wm) or LY-294002 induce an enlargement of AtSNX1 endosomes (Jaillais et al., 2006). In BY-2 cells, Wm was previously reported to lead to vacuolation of YFP–BP80-bearing organelles (Tse et al., 2004), and a similar effect was recently described for the TGN (Lam et al., 2007). To investigate which organelles in Arabidopsis root cells are similarly sensitive to Wm, we treated a collection of Arabidopsis transgenic lines expressing fluorescent-tagged organelle markers with Wm (see Supplementary Table S1). Some of these lines had the organelle marker protein driven by the strong 35S promoter. Although such lines have been used extensively in previous studies, some artefactual effects might be caused by the high expression levels of the fluorescent-tagged proteins. To limit this drawback, we selected transgenic lines that expressed only medium levels of the fluorescent markers. In addition, we employed several organelle markers to label the same compartment, and, when available, used the relevant endogenous promoters. As in BY-2 cells, YFP–BP80-labelled organelles were sensitive to Wm, as were AtSNX1–GFP-, RABF1–GFP- and GFP–RABF2b-labelled compartments (Figure 2a–h). The latter two endosomal markers were shown to co-localize with AtSNX1 endosomes in Arabidopsis root cells (Jaillais et al., 2006). By contrast, the TGN labelled either by TLG2a–GFP (Figure 2i,j) or a functional VHA-a1–GFP (Figure 2k,l) was insensitive to Wm. We also found the cis-Golgi labelled by ERD2–GFP (Figure 2m,n) or the trans-Golgi labelled by ST–GFP (Figure 2o,p) was not sensitive to Wm.

Figure 2.

 Sensitivity to wortmannin of various organelles in Arabidopsis root cells.
Root tip cells of seedlings expressing diverse fluorescent-tagged endomembrane markers after 60 min of mock treatment (a, c, e, g, i, k, m, o) or after 60 min of Wm treatment (b, d, f, h, j, l, n, p). The PVC labelled by YFP–BP80 (a, b) and endosomal compartments (c–h) are sensitive to Wm, whereas the TGN and Golgi apparatus are insensitive (i–p). Arrows indicate Wm compartments. Scale bars = 5 μm.

To obtain more direct clues as to whether AtSNX1 endosomes also display the BP80 PVC marker, we crossed an Arabidopsis line expressing a functional AtSNX1–mRFP fusion protein under the control of the AtSNX1 promoter (Jaillais et al., 2006) with the YFP–BP80-expressing line. In hybrid seedlings, we found a very high level of co-localization for AtSNX1–mRFP and YFP–BP80, as defined by the presence of the two markers on the same punctate structure (Figure 3a–c and Supplementary Figure S2). To confirm these results, we used another endosomal marker line that constitutively expresses mRFP–RABF2b (Takano et al., 2005). We observed that the mRFP–RABF2b fusion protein co-localized with AtSNX1 but not with the Golgi, TGN or GNOM endosome markers (Supplementary Figure S3). In addition, mRFP–RABF2b exhibited the same BFA and Wm sensitivity as AtSNX1 (Supplementary Figure S4). In hybrid seedlings co-expressing YFP–BP80 and mRFP–RABF2b, we found that mRFP–RABF2b co-localizes extensively with YFP–BP80 (Figure 3d–f). When we treated root tips of hybrid seedlings co-expressing AtSNX1–mRFP and YFP–BP80 with Wm, we observed that Wm compartments were labelled with both fluorescent-tagged organelle markers (Figure3j–l). Interestingly, YFP–BP80 was not found to co-localize with the TGN marker VHAa1–mRFP in the absence or presence of Wm (Figures 3g–i and 3m–o). Thus, BP80, which is known to cycle between the TGN and PVC, is mainly detected at the PVC, indicating that BP80 dynamics are in favour of a TGN to PVC displacement. Altogether, our results indicate that AtSNX1 endosomes and the PVC are overlapping endosomal compartments in Arabidopsis root cells that are both labelled by FM4-64 and sensitive to BFA and Wm. Our data also show that RABF1, RABF2b, AtSNX1 and BP80 label the same compartment. In addition, consistent with the work of daSilva et al. (2005), our data suggest that Wm blocks the retrograde transport of BP80 from the PVC to the TGN. Although relevant for root cells, we cannot exclude the possibility that the co-localization we observed between RABF proteins, SNX and BP80 might differ in other tissues or in isolated cultured cells.

Figure 3.

 SNX1 endosomes and PVC are overlapping compartments in Arabidopsis root cells.
Root tip cells of hybrid seedlings co-expressing AtSNX1–mRFP and YFP–BP80 (a–c), mRFP–RABF2b and YFP–BP80 (d–f), or mRFP–RABF2b and VHAa1–mRFP and YFP–BP80 (g–i). The merged images show a high level of co-localization between endosomal and PVC markers (c, f) but no localization between the TGN and PVC markers (i). After wortmannin treatment of hybrid seedlings co-expressing AtSNX1–mRFP and YFP-BP80 (j–o), root cells exhibit Wm compartments that are labelled both by AtSNX1–mRFP and YFP-BP80 (j–l). In hybrid seedlings co-expressing VHAa1–mRFP and YFP–BP80, only YFP–BP80 was detected in Wm compartments (m–o). Scale bars = 5 μm.

Acidic nature of AtSNX1 endosomes

Along the endocytic pathways of mammalian and yeast cells, early endosomes differentiate into endosomal compartments (late endosomes, multi-vesicular bodies or PVC), the luminal pH of which becomes more acidic the closer compartments are to the lysosomes or lytic vacuoles (Perret et al., 2005). To investigate the acidic nature of AtSNX1 endosomes, we used the acidotropic probe LysoTracker Red (LT). LT is a permeable red fluorochrome that accumulates in the membrane of acidic organelles (Wubbolts et al., 1996). In Arabidopsis root cells, AtSNX1–GFP did not co-localize with LT (Figure 4a–c). We obtained similar results with other endosome or PVC markers that had been previously shown to co-localize with AtSNX1 (Supplementary Figure S5). Interestingly, we found that, after Wm treatment, the Wm compartments were labelled by LT, suggesting that these have become more acidic (Figure 4d–f). This acidification might originate from fusion of the AtSNX1/PVC compartments with vacuolar-derived vesicles that might be induced by Wm. Another possibility is that, after Wm treatment, AtSNX1 endosomes accumulate mis-targeted proteins involved in organelle acidification, notably proteins whose final destination is the lytic vacuoles. Although LT labelling seems to be less specific for lytic compartments of plant cells compared with animal cells, the relatively low acidification of AtSNX1 endosomes, as deduced from the lack of LT accumulation in these compartments, is in favour of AtSNX1 endosomes being part of an early or immature late endosomal compartment. It is worth noting that the acidic nature of Wm compartments is deduced only from their staining by LT. As very little is known about the specificity of LT staining with regard to compartment acidification in plant cells, it is wise to mention that staining by this reagent does not necessarily indicate the acidic character of stained organelles in plant cells.

Figure 4.

 Acidic nature of AtSNX1 endosomes.
Root tip cells of AtSNX1–GFP seedlings in the presence of 50 nm LysoTracker Red (LT) with mock treatment (a–c) or after 60 min of wortmannin treatment (d–f). Arrowheads indicate Wm compartments, asterisks indicate vacuoles. Scale bars = 5 μm.

AtSNX1 endosomes are involved in cargo sorting in the secretory pathway

Next, we investigated whether AtSNX1 endosomes are involved in the sorting of proteins from the TGN to the lytic vacuoles. Using transiently transfected tobacco protoplasts, daSilva et al. (2005) showed that long-term Wm treatment inhibits retrograde transport of a truncated GFP–BP80 reporter from the PVC to the TGN, and leads to its degradation inside the vacuoles. When we submitted Arabidopsis seedlings expressing YFP–BP80 to a 3 h Wm treatment, YFP–BP80 was found both in Wm compartments and in the vacuoles of root cells (Figure 5a,b). Thus, long-term Wm treatment seems to interfere similarly with VSR retrograde trafficking from the PVC to TGN in the two cell types.

Figure 5.

 AtSNX1 endosomes are involved in proper trafficking of proteins to the vacuoles.
(a, b) Root tip cells of YFPBP80 seedlings with mock treatment (a) or after 180 min of Wm treatment (b).
(c) Western blot analysis of proteins extracted from aleurain–GFP seedlings after 1 or 3 h of mock treatment (lanes 1 and 2, respectively), after 1, 2 or 3 h of BFA treatment (lanes 3, 4 and 5, respectively) and after 1, 2 or 3 h of Wm treatment (lanes 6, 7 and 8, respectively). The same amounts of total protein were loaded on each lane, and the blot was immunostained using anti-tubulin (as a control for equal loading) and anti-GFP antibodies.
(d–i) Root tip cells of aleurain–GFP seedlings after mock treatment (d, g), or after 180 min of BFA (e), CHX (f) or Wm (h) treatment. The inset (i) shows a Wm compartment from the root cell of seedlings co-expressing AtSNX1–mRFP and aleurain–GFP. AtSNX1–mRFP is clearly visible at the surface of the Wm compartment, the lumen of which is filled with aleurain–GFP.
(j) Root tip cells of aleurain–GFP seedlings after 180 min of combined Wm and CHX treatment. Arrowheads indicate Wm compartments, asterisks indicate vacuoles and arrows indicate BFA compartments. Scale bars = 5 μm.

We then studied trafficking of the BP80 ligand, the vacuolar protease aleurain (Fluckiger et al., 2003; Paris and Neuhaus, 2002). We raised a transgenic line expressing a GFP-tagged form of aleurain (Fluckiger et al., 2003). In root cells, we found that aleurain–GFP is localized in the vacuoles, but also accumulates in the cytoplasm as punctate compartments, which probably represent an intermediate trafficking compartment used by aleurain–GFP on its way to the vacuoles (Figure 5d). We also found these intermediate compartments to be partially labelled by FM4-64 and to co-localize with both mRFP–RABF2b and AtSNX1–mRFP (Supplementary Figure S6). These observations indicate that vacuolar proteins (e.g. aleurain) might transit through AtSNX1 endosomes before reaching their final destination. After a 3 h BFA treatment, we detected aleurain–GFP in BFA compartments, but no labelling was visible in the vacuoles (Figure 5d,e). The absence of aleurain–GFP in the vacuoles was not due to a potential activating effect of the drug on aleurain–GFP degradation, as seedlings treated with BFA for increasing periods of time (up to 3 h) showed roughly constant levels of total aleurain–GFP (Figure 5c). In addition, after a 3 h treatment of seedlings with the protein synthesis inhibitor cycloheximide (CHX), the aleurain–GFP signal had almost totally vanished, indicating an approximate turnover of 3 h for aleurain–GFP (Figure 5f). Together, these data indicate that the absence of aleurain–GFP in the vacuoles of BFA-treated seedlings results from the degradation of aleurain–GFP in the vacuoles concomitantly with the inhibition of its trafficking from the BFA compartments towards the vacuoles. Thus, BFA seems to block the SNX1 endosome/PVC to vacuole pathway, a result which is consistent with the recent finding that BFA inhibits degradation of the plant brassinosteroid receptor BRI1 (Geldner et al., 2007).

When seedlings were treated with Wm for 3 h, aleurain–GFP accumulated in Wm compartments but was still detected within the vacuoles (Figure 5g–i). Interestingly, a soluble synthetic BP80 ligand (amy-spo) was found to be secreted at a high level after Wm treatment in tobacco cells (daSilva et al., 2005). Under our experimental conditions, we were not able to detect the secretion of aleurain–GFP in the extracellular space, but we cannot exclude the possibility that such a secretion might also partly occur in the root tip cells. Wm treatment did not alter the degradation capacities of vacuolar hydrolases, as aleurain–GFP completely disappeared after a 3 h Wm treatment in the presence of CHX (Figure 5j). In addition, seedlings treated with Wm for increasing periods of time (up to 3 h) showed roughly similar levels of total aleurain–GFP (Figure 5c). These results indicate that Wm maintains proper trafficking of aleurain–GFP towards the vacuoles.

To summarize, our results indicate that a prolonged Wm treatment blocks the retrograde transport of YFP–BP80 from AtSNX1 endosomes to the TGN, leading to a massive routing of YFP–BP80 to the vacuoles, but the anterograde transport of aleurain–GFP from AtSNX1 endosomes to the vacuoles is not affected. It is likely that the normal routing of aleurain–GFP to the vacuoles is maintained under these conditions by continuous de novo synthesis of endogenous Arabidopsis VSRs. Together, our results reveal that AtSNX1 endosomes are directly involved in the routing of lytic enzymes to the vacuoles.

AtSNX1 endosomes are involved in endocytic cycling and down-regulation of PM proteins

To identify PM proteins that might be routed through AtSNX1 endosomes, we took advantage of the selective effect of Wm on endosomes. We previously applied this strategy to differentiate between the routes used by two auxin carriers, PIN FORMED 1 (PIN1) and PIN2 (Jaillais et al., 2006). To study the co-localization of PM proteins with AtSNX1 endosomes, we crossed the AtSNX1mRFP or mRFPRABF2b lines with various marker lines expressing GFP-tagged PM proteins that are known to cycle between the PM and endosomes and/or to be routed to vacuoles for degradation (see Supplementary Table S1). To ascertain that the fluorescence of tagged proteins only marked endocytic and not biosynthesis pathways, we treated roots with the protein synthesis inhibitor cycloheximide. BRASSINOSTEROID INSENSITIVE 1 (BRI1) is the receptor of the brassinosteroid phytohormones. In Arabidopsis roots, a functional BRI1–GFP translational fusion is localized at the PM and endosomes (Geldner et al., 2007; Russinova et al., 2004). The labelling of endosomes by BRI1–GFP is BFA-sensitive and ligand-independent, suggesting that BRI1–GFP cycles constitutively between the PM and endosomes (Geldner et al., 2007; Russinova et al., 2004). We found that BRI1–GFP-positive endosomes were labelled with AtSNX1–mRFP and mRFP–RABF2b, in the absence or presence of Wm (Figure 6a,b, and see also Supplementary Figure S7). This suggests that BRI1–GFP can transit through AtSNX1 endosomes during recycling towards the PM. Two other PM proteins, PLASMA MEMBRANE INTRINSIC PROTEIN 2a (PIP2a) and LOW TEMPERATURE INDUCIBLE PROTEIN 6b (LTi6b), are believed to cycle between the PM and endosomes on the basis of BFA sensitivity (Grebe et al., 2003; Paciorek et al., 2005). In seedlings co-expressing GFP–PIP2a and AtSNX1–mRFP or mRFP–RABF2b, we only detected GFP–PIP2a labelling at the plasma membrane (Figure 6c, and see also Supplementary Figure S7). However, after Wm treatment, GFP–PIP2a was found at the plasma membrane and in the Wm compartment, implying that GFP–PIP2a, like BRI1–GFP, may cycle constitutively between the PM and AtSNX1 endosomes (Figure 6d, and see also Supplementary Figure S7). In seedlings co-expressing GFP–LTi6b and AtSNX1–mRFP or mRFP–RABF2b, we found GFP–LTi6b at the PM but also in intracellular compartments that only occasionally co-localized with AtSNX1–mRFP or mRFP–RABF2b, both in the presence or absence of Wm (Figure 6e,f, and see also Supplementary Figure S7). These observations indicate that GFP–Lti6b traffics only partially through the Wm-sensitive AtSNX1 endosomes, and that GFP–Lti6b uses other endocytic pathways that are Wm-insensitive, such as those involving GNOM endosomes (Jaillais et al., 2006) or the TGN. Indeed, the TGN has recently been proposed to be an early endosomal compartment, as it is the first compartment to be labelled with FM4-64 both in Arabidopsis root cells (Dettmer et al., 2006) and cultured tobacco BY-2 cells (Lam et al., 2007). HIGH BORON REQUIRING 1 (BOR1) is a boron transporter that is localized at the PM in the absence of boron, but is rapidly routed to vacuoles for degradation in the presence of boron (Takano et al., 2002, 2005) (Figure 6g,i, and see also Supplementary Figure S7). This down-regulation is mediated by internalization of BOR1 into the cell through RABF2b-containing endosomes, and subsequent sorting towards the lytic vacuoles (Takano et al., 2005). After Wm treatment, we observed that BOR1–GFP co-localized with AtSNX1–mRFP or mRFP–RABF2b, either in the absence or presence of boron (Figure 6h,j, and see also Supplementary Figure S7). These data indicate that BOR1–GFP trafficking relies on AtSNX1 endosomes both when en route to the lytic vacuoles in the presence of boron and when it is recycled to the PM in the absence of boron. Taken together, our co-localization analyses show that AtSNX1 endosomes are involved in the trafficking of several PM proteins.

Figure 6.

 AtSNX1 endosomes are transit compartments for PM proteins during their endocytic trafficking.
Root tip cells of hybrid seedlings co-expressing AtSNX1–mRFP and BRI1–GFP (a, b), GFP–PIP2a (c, d), GFP–LTi6b (e, f) and BOR1–GFP (g–j). Seedlings were treated (b, d, f, h, j) or untreated (a, c, e, g, i) with Wm for 90 min. BOR1–GFP seedlings were grown for 7 days on a low-boron medium. Root tip cells were then mounted on either a low-boron medium (−boron) (g, h) or LM medium with boron (+ boron) (i, j) for 30 min before observation. Scale bars = 5 μm.

Discussion

The AtSNX1 endosomal compartment corresponds to the PVC/MVB in root cells

Our work shows that, in Arabidopsis root cells, the AtSNX1 endosomes also contain the vacuolar sorting receptor BP80, which is a marker of the PVC/MVB (Tse et al., 2004). The very high level of co-localization between AtSNX1- and BP80-fluorescent-tagged proteins, as well as the same BFA and Wm sensitivity exhibited by AtSNX1- and BP80-bearing organelles, indicate that AtSNX1 endosomes and the PVC constitute one and the same endosomal compartment in Arabidopsis root cells. This compartment is distinct from GNOM endosomes (Geldner et al., 2003; Jaillais et al., 2006) and the TGN. This is deduced from the lack of co-localization between AtSNX1–mRFP or YFP–BP80 and specific markers of GNOM endosomes and the TGN, and by the insensitivity of these organelles to Wm. Interestingly, in contrast to our observations, Lam et al. (2007) recently reported that, in tobacco BY-2 cells, the PVC/MVB is not found in BFA compartments, although it is sensitive to Wm, and that the TGN fuses to the PVC/MVB after Wm treatment. This discrepancy between our data and those of Lam et al. underlines the variability that endomembrane systems may display depending on cell types. This is reminiscent of the cell selectivity previously described for the effect of BFA, for example between root and mesophyll cells (Russinova et al., 2004), or between epidermal and stele cells (Paciorek et al., 2005). These observations strongly suggest that endocytic pathways are likely to be cell-type-specific in plants. As endocytosis emerges as a key regulator of numerous signalling pathways controlling cell polarity and development in plants (Geldner and Jurgens, 2006), it seems that this cell-type specificity of endocytosis might constitute a fine regulatory system used by plant cells to sense environmental cues differently according to their tissue location. This result suggests that caution should be exercised in extrapolating data obtained from a particular tissue or species to other systems. Interestingly, our work shows that Wm specifically affects SNX1 endosomes/PVC/MVB in Arabidopsis root cells, but has no apparent effect on the other organelles analysed, such as the TGN, GNOM endosomes or the Golgi apparatus. Thus, Wm appears to be a very powerful tool to study the localization of proteins in the PVC/MVB as well as the dynamics of proteins passing through this organelle. In contrast, BFA, which has been extensively used in protein trafficking studies in Arabidopsis root cells (Abas et al., 2006; Dettmer et al., 2006; Dhonukshe et al., 2006; Geldner et al., 2001, 2003; Grebe et al., 2003; Jaillais et al., 2006; Paciorek et al., 2005; Russinova et al., 2004), has a much broader effect on endomembrane compartments, as the TGN, GNOM endosomes, AtSNX1 endosomes/PVC/MVB and Golgi apparatus accumulate in or around the BFA compartments in Arabidopsis root cells (this work and Jaillais et al., 2006; for example). BFA is known to block recycling of PM proteins (Abas et al., 2006; Dettmer et al., 2006; Dhonukshe et al., 2006; Geldner et al., 2001, 2003; Grebe et al., 2003; Jaillais et al., 2006; Paciorek et al., 2005; Russinova et al., 2004). Here, we show that BFA is also capable of preventing the transfer of aleurain–GFP from the BFA compartment to the lytic vacuoles (see Figure 5), suggesting that BFA might have a general inhibitory effect on the transport of proteins from the PVC/MVB to the lytic vacuoles.

Recently, Oliviusson et al. (2006) showed that three components of the larger subunit of the plant retromer complex, namely VPS26, VPS29 and VPS35, are associated with membranes of the PVC/MVB in BY-2 cells. Our data indicate that AtSNX1, which is another putative component of the plant retromer, is localized at the PVC/MVB in root cells. Although we have no evidence that AtSNX1 directly interacts with the other three VPS proteins in the same multi-protein complex, we may assume that the plant retromer is localized to the PVC/MVB in root cells.

The PVC/MVB is a crossroads between the secretory and endocytic pathways in Arabidopsis root cells

Our analysis of VSR protein sorting and trafficking of endocytosed plasma membrane proteins reveals that cargos of both the secretory pathway (such as YFP–BP80 or aleurain–GFP) and endocytic pathway (such as BRI1–GFP, BOR1–GFP or GFP–PIP2a) are routed through AtSNX1 endosomes. Moreover, alteration of AtSNX1 endosome functions by long-term Wm treatment, which leads to mis-routing of YFP–BP80 (see Figure 5a,b) and accumulation of recycling PM proteins in Wm compartments, indicates that AtSNX1 endosomes are involved in the proper sorting of these cargos. Although we present no direct evidence that recycling of PM proteins is altered following Wm treatment, our data suggest that AtSNX1 endosomes are sorting endosomes located at the crossroads between the secretory and endocytic pathways. Tse et al. (2004) proposed that the PVC lies on the secretory and endocytic pathways, based on the analysis of FM4-64 internalization in tobacco BY-2 cells. However, uptake and distribution of FM4-64 might not reflect the genuine trafficking of plasma membrane proteins in plant cells (Bolte et al., 2004b). Our data provide compelling evidence that endocytosed PM proteins as well as cargos of the secretory pathway merge at the PVC/MVB.

In yeast and mammals, the composition of proteins at the PM is maintained and adjusted by a ubiquitin-dependent endocytic pathway that removes transmembrane proteins from the PM and delivers them into the lumen of the lysosomes/vacuoles for degradation (Babst, 2005). The sorting signal for entering into this pathway is mono-ubiquitination of the cargo proteins, which, once mono-ubiquitinated, are then routed through MVB to the lytic compartments of the cell (Gruenberg and Stenmark, 2004; Raiborg et al., 2003). If internalized PM proteins are not ubiquitinated, they are recycled to the PM via recycling endosomes. The initial step of recognition of mono-ubiquitinated cargo proteins for entry into the MVB, the vesicle formation of the MVB, and the sorting of endosomal cargo proteins into these vesicles are dependent on the function of three conserved heteromeric protein complexes called ESCRT-I, ESCRT-II and ESCRT-III (Babst, 2005). ESCRT-I has been proposed to serve as a sorting receptor that recognizes mono-ubiquitinated proteins at the early endosome and initiates their proper sorting via the ESCRT machinery. Interestingly, a comparative analysis of fully sequenced genomes revealed that components of the ESCRT machinery are highly conserved in yeast, mammals, plants (e.g. A. thaliana and Oryza sativa), Drosophila melanogaster and Caenorhabditis elegans (Winter and Hauser, 2006). Recently, the Arabidopsis homologue of the yeast Vps23 protein (Vps23p), which is a key component of ESCRT-I in yeast and animals, has been characterized (Spitzer et al., 2006). The Arabidopsis VPS23 protein, also known as ELC (product of the ELCH gene or At3g12400), is a component of the plant ESCRT-I complex and functions in cytokinesis. By using fluorescent-tagged proteins, Spitzer et al. (2006) showed that ELC co-localizes with RABF1 and RABF2b in protoplasts of Arabidopsis cultured cells. These findings, together with our observations that the PVC/MVB in root cells receives endocytosed as well as secretory cargo proteins, are consistent with our assumption that AtSNX1 endosomes are sorting compartments in which cargo proteins are routed to diverse destinations via endocytic or secretory pathways.

Position of the PVC/MVB in the endosomal pathway in Arabidopsis root cells

We found that RABF2b co-localized extensively with AtSNX1 in the PVC/MVB. Three other groups have previously shown that the over-expression of dominant negative RABF proteins interferes with the correct sorting of lytic enzymes towards the vacuoles (Bolte et al., 2004a; Kotzer et al., 2004; Sohn et al., 2003). More recently, over-expression of a dominant negative form of RABF2b in Arabidopsis roots was reported to inhibit endocytosis and normal cell plate formation (Dhonukshe et al., 2006). Altogether, these data allow us to propose that RABF2b acts early in the endocytic pathway, and, like its Rab5 mammalian counterpart, regulates endocytic vesicle fusion with early endosomes. Recently, the TGN was reported to be the first compartment labelled with FM4-64 in Arabidopsis root cells (Dettmer et al., 2006). This was confirmed in cultured tobacco BY-2 cells (Lam et al., 2007). Whether FM4-64 transits first through AtSNX1 endosomes to then reach the TGN or directly accumulates in the TGN without trafficking via endosomal structures remains an unsolved question. If FM4-64 passes first through AtSNX1 endosomes, an absence of accumulation of FM4-64 may occur if the dye leaves AtSNX1 endosomes much faster that it leaves the TGN. FM4-64 could then label AtSNX1 endosomes later, after fusion of FM4-64-stained vesicles derived from the TGN or other secretory structures with AtSNX1 endosomes. It would be of particular interest to decipher the relationships between these two key compartments in the endocytic pathways.

Conclusions

Our current view of endocytic pathways in plants is mainly based on the data and concepts derived from animal and yeast studies. However, we are still lacking experimental data in plant systems to establish the accuracy of this assumption. Here, we provide evidence for the existence of sorting endosomes at the crossroads between the endocytic and secretory pathways in root cells. These compartments contain cargo proteins originally from the PM and TGN, and are able to redirect these cargos towards various destinations such as the PM, TGN or the vacuoles. Our data, which indicate that AtSNX1 endosomes display protein markers and functions that have been attributed to either early or late endosomes in earlier studies, are consistent with the recent view that endosomes form a continuum of compartments (Bonifacino and Rojas, 2006; Perret et al., 2005). We show here that sorting endosomes are present in plant cells, and thus sorting endosomes are conserved organelles in eukaryotic cells from yeast to man including plants. However, it remains to be determined whether the endocytic sorting machinery is functionally conserved as well. The relationships that probably exist between the TGN, GNOM and AtSNX1 endosomes in root cells also remain to be determined. The challenge for the future will be to elucidate the molecular basis of endocytic sorting in plants and to determine its implication in plant cell signalling and development.

Experimental procedures

Material and growth conditions

Plants of the Columbia (Col-0) accession were grown on MS medium (4.2 g/l MS salts, 1% sucrose, 8 g l–1 plant agar, pH 5.8) under short-day conditions (8 h light, 16 h dark) at 21°C for 7 days before observation. The plasmidspAtSNX1::AtSNX1–mRFP in the snx1 mutant, pAtSNX1::AtSNX1–GFP in the snx1 mutant, pGNOM::GNOM–GFP in the gnom mutant and pBRI1::BRI1–GFP in the bri1 mutant, and the pVHA-a1::VHA-a1–GFP, pVHA-a1::VHA-a1–mRFP, p35S::ERD2–GFP, p35S::ST–GFP, p35S::TLG2a–GFP, p35S::RABF1–GFP, p35S::GFP–RABF2b, p35S::mRFP–RABF2b, p35S::GFP–PIP2a, p35S::GFP–LTi6b and p35S::BOR1–GFP lines have been described previously (Dettmer et al., 2006; von der Fecht-Bartenbach et al., 2007; Geldner et al., 2003; Grebe et al., 2003; Jaillais et al., 2006; Paciorek et al., 2005; Russinova et al., 2004; Takano et al., 2005). YFP–BP80 and aleurain–GFP constructs were transformed into the Col-0 accession and analysed as described previously (Jaillais et al., 2006).

Microscopy and drug treatments

Co-localization analyses were performed on F1 or F2 hybrid seedlings stably co-expressing the GFP- and mRFP-tagged markers. Each experiment was repeated at least three times, and more than 10 roots were analysed per experiment. Roots of 7-day-old seedlings grown on MS medium were mounted in LM medium (2.1 g l–1 MS salts, 1% sucrose, pH 5.8), and analysed on a LSM-510 laser scanning confocal microscope (Zeiss, http://www.zeiss.com/). For FM4-64 treatments, seedlings were incubated for 2 min at room temperature in LM medium containing FM4-64 (3.5 μm, Invitrogen; http://www.invitrogen.com), and rinsed three times in LM medium before observation. Brefeldin A (100 μm in DMSO/EtOH, Sigma; http://www.sigmaaldrich.com) and wortmannin (33 μm in DMSO, Sigma) treatments were also performed in LM medium for the times indicated before observation. For LysoTracker Red (LT, Invitrogen) treatments, seedlings were mounted in LM medium containing 50 nm LT (in DMSO), and observed for 5–15 min after dye addition. When integral transmembrane proteins (YFP–BP80, BRI1–GFP, GFP–PIP2a, GFP–LTi6b, BOR1–GFP) were analysed, seedlings were incubated in the presence of the protein synthesis inhibitor cycloheximide (50 μm in DMSO) for 60 min before and concomitantly with the other treatments. This cycloheximide treatment allows the localization of only previously synthesized transmembrane proteins.

Western blot analysis

Proteins were extracted and analysed as previously described (Cabrillac et al., 2001; Giranton et al., 2000).

Acknowledgements

We thank Dr C. Hawes (Oxford University, UK) for providing ERD2GFP and STGFP plasmids, Dr J.-M. Neuhaus (Neuchatel University, Switzerland) for providing the aleurain–GFP plasmid, Dr N. Paris (Institut National de la Recherche Agronomique, Université de Montpellier, France) for providing the YFPBP80 construct, Dr J. Takano (Tokyo University, Japan) for supplying the BOR1–GFP and mRFPRABF2b hybrid seeds, Dr J. Chory (SALK Institute, USA) for the gift of the BRI1–GFP line, Dr K. Schumacher (Tübingen University, Germany) for the gift of the VHA-a1GFP and VHA-a1mRFP lines, Dr G. Jürgens (Tübingen University, Germany) for the gift of the GNOMGFP line, and the Nottingham Arabidopsis Stock Centre (NASC) for supplying the GFPLTi6b and GFPPIP2a lines. We thank Dr N. Paris for helpful discussions, and C. Lionnet and F. Simian from the Institut Fédératif de Recherche (IFR) and Plateau Technique d’Imagerie/Microscopie (PLATIM) for technical assistance in confocal microscopy. This work was supported by the ‘Action Concertée Incitative Développement’ from the French Ministry of Research.

Ancillary