An Arabidopsis GRIP domain protein locates to the trans-Golgi and binds the small GTPase ARL1


  • Maita Latijnhouwers,

    1. Cell-to-Cell Communication Programme, Scottish Crop Research Institute, Invergowrie, Dundee DD2 5DA, UK,
    Search for more papers by this author
  • Chris Hawes,

    1. Research School of Biological & Molecular Sciences, Oxford Brookes University, Headington, Oxford OX3 BP, UK, and
    Search for more papers by this author
  • Claudine Carvalho,

    1. Research School of Biological & Molecular Sciences, Oxford Brookes University, Headington, Oxford OX3 BP, UK, and
    Search for more papers by this author
  • Karl Oparka,

    1. Cell-to-Cell Communication Programme, Scottish Crop Research Institute, Invergowrie, Dundee DD2 5DA, UK,
    Search for more papers by this author
    • Present address: Institute of Molecular Plant Sciences, University of Edinburgh, Mayfield Road, Edinburgh EH9 3JH, UK.

  • Alison K. Gillingham,

    1. Medical Research Council Laboratory of Molecular Biology, Cambridge CB2 2QH, UK
    Search for more papers by this author
  • Petra Boevink

    Corresponding author
    1. Cell-to-Cell Communication Programme, Scottish Crop Research Institute, Invergowrie, Dundee DD2 5DA, UK,
      (fax +44 1382 562426; e-mail
    Search for more papers by this author

(fax +44 1382 562426; e-mail


GRIP domain proteins are a class of golgins that have been described in yeast and animals. They locate to the trans-Golgi network and are thought to play a role in endosome-to-Golgi trafficking. The Arabidopsis GRIP domain protein, AtGRIP, fused to the green fluorescent protein (GFP), locates to Golgi stacks but does not exactly co-locate with the Golgi marker sialyl transferase (ST)-mRFP, nor with the t-SNAREs Memb11, SYP31 and BS14a. We conclude that the location of AtGRIP is further to the trans side of the stack than STtmd–mRFP. The 185-aa C-terminus of AtGRIP containing the GRIP domain targeted GFP to the Golgi, although a proportion of the fusion protein was still found in the cytosol. Mutation of a conserved tyrosine (Y717) to alanine in the GRIP domain disrupted Golgi localization. ARL1 is a small GTPase required for Golgi targeting of GRIP domain proteins in other systems. An Arabidopsis ARL1 homologue was isolated and shown to target to Golgi stacks. The GDP-restricted mutant of ARL1, AtARL1-T31N, was observed to locate partially to the cytosol, whereas the GTP-restricted mutant AtARL1-Q71L labelled the Golgi and a population of small structures. Increasing the levels of AtARL1 in epidermal cells increased the proportion of GRIP–GFP fusion protein on Golgi stacks. We show, moreover, that AtARL1 interacted with the GRIP domain in a GTP-dependent manner in vitro in affinity chromatography and in the yeast two-hybrid system. This indicates that AtGRIP and AtARL1 interact directly. We conclude that the pathway involving ARL1 and GRIP domain golgins is conserved in plants.


The Golgi apparatus is central to the secretory pathway in eukaryotic cells. Proteins, glycoproteins and lipids enter this organelle from the endoplasmic reticulum (ER) to be processed. On exiting, cargo is packaged into vesicles and sent off to the correct cellular destination. The Golgi apparatus is also likely to play a role in the endocytic pathway, where it is responsible for recycling of membranes and proteins back into the ER. The organization of the mammalian Golgi apparatus is significantly different from its counterpart in plants. Whereas in animal cells the Golgi stacks are generally arranged side-by-side in a ribbon structure around the nucleus, the individual stacks in plants are motile structures distributed throughout the cytoplasm. Their movement is closely associated with the ER network and is dependent on the presence of actin (Boevink et al., 1998; Brandizzi et al., 2002; Saint-Jore et al., 2002).

In mammals and yeast, the importance of golgins or Golgi matrix proteins in maintaining the Golgi structure and tethering vesicles to target membranes is becoming increasingly clear (Gillingham and Munro, 2003; Whyte and Munro, 2002). Golgins are large integral or peripheral membrane proteins with extensive coiled-coil regions, motifs that are known to fold into a rod-like shape. The name ‘matrix protein’ originates from the fact that these proteins form part of a proteinaceous skeleton that remains after the Golgi membranes have been extracted with detergent (reviewed by Barr and Short, 2003). In plants golgins remain unexplored, although the presence of a ribosome-excluding matrix around each Golgi stack has been observed (Staehelin and Moore, 1995) and intercisternal elements have been detected by electron microscopy (reviewed by Hawes, 2005).

A subset of golgins is characterized by a C-terminal domain called the GRIP domain. This 42-residue domain was shown to be sufficient for targeting of the human golgin-245 to the Golgi apparatus, although including the flanking amino acids greatly increased the efficiency of Golgi targeting (Kjer-Nielsen et al., 1999). The GRIP domain is present in four mammalian proteins and one yeast golgin (Kjer-Nielsen et al., 1999; Munro and Nichols, 1999), and has been identified in other animals and the protist Trypanosoma brucei (McConville et al., 2002). Recently, a protein possessing a GRIP domain was identified in Arabidopsis, and the GRIP domain, fused to GFP, was shown to be located to the plant Golgi apparatus (Gilson et al., 2004). This protein was named AtGRIP.

All four mammalian GRIP proteins are localized to the trans-Golgi network (TGN), an extensive membrane network on the trans side of the Golgi stack, which functions in sorting proteins for secretion or transport to lysosomes and in receipt of material from endosomes. Overexpression of the GRIP domain of golgin-245 led to disruption of the TGN and mislocalization of the TGN protein TGN46, presumably due to displacement of the endogenous protein (Yoshino et al., 2003). Recently, it was shown that perturbation of the function of the mammalian GRIP domain protein golgin-97 blocked endosome-to-TGN transport (Lu et al., 2004). Similarly, the yeast GRIP domain protein IMH1 has been implicated in protein sorting in the late Golgi or endosomes (Tsukada et al., 1999).

Golgi localization of GRIP domain proteins is dependent on the presence of a member of the small GTPase family called Arf-like1 (ARL1) (Lu and Hong, 2003; Panic et al., 2003a; Setty et al., 2003). ARL1 locates to the trans side of the Golgi or the TGN (Lu et al., 2001), and was shown to bind specifically to the GRIP domain when in the active, GTP-bound state (Lu and Hong, 2003; Panic et al., 2003a; Setty et al., 2003; Van Valkenburgh et al., 2001). Yeast arl1 mutants have minor defects in protein sorting in the TGN (Rosenwald et al., 2002) and in ion homeostasis (Love et al., 2004; Munson et al., 2004a,b). Knock-down of ARL1 in HeLa cells using RNAi resulted in a block in the transport of Shigatoxin B fragment from endosomes to the TGN (Lu et al., 2004). These results suggest the involvement of ARL1 in endosome-to-Golgi trafficking.

Here we show that a fusion of the full-length Arabidopsis GRIP domain protein AtGRIP with GFP locates to the distal side of the trans-Golgi, and that a mutation of a conserved tyrosine in the GRIP domain to alanine disrupts Golgi localization. An Arabidopsis ARL1 (AtARL1) was identified and localized to Golgi stacks. Mutations that lock AtARL1 in the GDP-bound (T31N) or the GTP-bound (Q71L) form altered the location of the protein. We also provide evidence for a direct interaction between ARL1-GTP (but not ARL1-GDP) and AtGRIP. These results suggest that the function of GRIP domain golgins and ARL1 are conserved in plants.


AtGRIP is an Arabidopsis golgin

blast searches of the Arabidopsis genomic database ( with the mammalian Golgi matrix proteins GCC185 and golgin-97 identified an ORF (At5g66030) which shared significant sequence similarity with the query sequences. Further inspection revealed that the region of sequence similarity is centred around the C-terminus of both query sequences and corresponds to the C-terminus of At5g66030. This Arabidopsis ORF encodes a 788-aa protein with extensive coiled-coil regions in the N-terminal 693 amino acids, based on a prediction by the ARABI-COIL Database (; Rose et al., 2004). The C-terminus of this protein contains a GRIP domain with 50 and 47% identity to the GRIP domains of GCC185 and golgin-97, respectively. This Arabidopsis ORF has been described previously (Gilson et al., 2004), and its GRIP domain was shown to localize to Golgi stacks.

A full-length AtGRIP cDNA clone was isolated from an Arabidopsis cell culture cDNA pool by PCR amplification using gene-specific primers. The cDNA clone was inserted into binary vectors to create translational fusions with GFP in either orientation (Figure 1). The fusion proteins AtGRIP–GFP and GFP–AtGRIP were expressed transiently in tobacco CB137 by agro-infiltration. CB137 is a transgenic line expressing the transmembrane domain of rat sialyl transferase (STtmd) fused to mRFP under control of the enhanced cauliflower mosaic virus 35S promoter (Figure 2a). The infiltrated tissue was examined by confocal microscopy 2 days later. AtGRIP–GFP and GFP–AtGRIP were both observed as green, moving bodies in the cytoplasm of epidermal cells. When AtGRIP–GFP (or GFP–AtGRIP) and STtmd–mRFP were imaged together, the green and red signals were clearly located on the same bodies (Figure 2b–e; Supplementary Material, Movie 1). STtmd is well established as a Golgi marker in plants (Boevink et al., 1998; Saint-Jore et al., 2002; Wee et al., 1998). The co-localization of AtGRIP–GFP or GFP–AtGRIP with STtmd–mRFP therefore indicates that AtGRIP was Golgi-localized. Closer examination of the confocal images clearly showed that the fluorescent signals from AtGRIP–GFP (or GFP–AtGRIP) and STtmd–mRFP were not completely co-localized. STtmd was previously shown to reside predominantly in the medial and trans Golgi by immunogold labelling (Boevink et al., 1998). To determine in which part of the Golgi stack AtGRIP resides, AtGRIP–GFP was expressed in CB137 leaves in combination with YFP fusion proteins of three different Arabidopsis t-SNAREs, Memb11 (Membrin1), SYP31 (Sed5), and BS14a (Bet11). YFP–Memb11 and YFP–SYP31 are thought to be predominantly located in the cis-Golgi, whereas YFP–BS14a is mainly detected in the trans-Golgi, based on differences in brefeldin A (BFA) responses (Uemura et al., 2004; P. Moreau and co-workers, University of Bordeaux, Bordeaux, France, personal communication). In Figure 2(f–h), GFP is shown as green, mRFP as red, and YFP as blue. In cells expressing AtGRIP–GFP in combination with YFP–Memb11 and STtmd–mRFP, as well as in cells expressing AtGRIP–GFP in combination with YFP–SYP31 and STtmd–mRFP, three-coloured Golgi stacks were detected; each of the three fusion proteins located to a different part of the stack. In some stacks a clear pattern was distinguished: STtmd–mRFP was detected in the centre, with AtGRIP–GFP on one side and YFP–Memb11 or YFP–SYP31, respectively, on the other side. Not all stacks appeared the same, which is presumably due to the fact that Golgi stacks are in motion and constantly tumble around their axes (Supplementary Material, Movie 1), with the result that different stacks will be at different orientations to the plane of scanning. In cells expressing AtGRIP–GFP, YFP–BS14a and STtmd–mRFP, STtmd–mRFP co-located with YFP–BS14a, whereas AtGRIP–GFP fluorescence was adjacent to this area (Figure 2h). Because AtGRIP–GFP was adjacent to STtmd–mRFP and YFP–BS14a, and on the opposite side of the stack to YFP–Memb11 and YFP–SYP31, it was concluded that AtGRIP was located to the far side of the trans-Golgi stacks or to the TGN.

Figure 1.

Constructs of AtGRIP domains and full-length AtGRIP fused to GFP used in this study.

Figure 2.

AtGRIP locates to the far trans-Golgi.
Confocal images of leaf epidermal cells of wild-type tobacco and STtmd–mRFP (in red)-expressing tobacco line CB137, 2 days after Agrobacterium inoculation. Strains carrying AtGRIP–GFP and GFP–AtGRIP (both in green) were inoculated at OD600 = 0.1, and those carrying the t-SNARE constructs YFP–Memb11, YFP–SYP31 and YFP–BS14a (in blue) at OD600 = 0.01. (a) Uninoculated leaf of CB137; (b) AtGRIP–GFP; (c) AtGRIP–GFP and STtmd–mRFP; (d) GFP–AtGRIP; (e) GFP–AtGRIP and STtmd–mRFP; (f) AtGRIP–GFP, YFP–Memb11 and STtmd–mRFP; (g) AtGRIP–GFP, YFP–SYP31 and STtmd–mRFP; (h) AtGRIP–GFP, YFP–BS14a and STtmd–mRFP. Bar = 3 μm.

Golgi localization of the GRIP domain is abolished by mutation of a conserved tyrosine to alanine

The N-terminal 710 aa of AtGRIP were fused to GFP and expressed in CB137 tobacco leaves by agro-infiltration. The fusion protein GFP–Nterm showed staining throughout the cell, a pattern most reminiscent of cytosolic staining (Figure 3a,b). The protein was also found in the nucleus (not shown). We expected that the GRIP domain of AtGRIP would locate to Golgi stacks. To verify this, the 45-aa GRIP domain of AtGRIP (amino acids 711–755) was fused to GFP. Unexpectedly, this fusion protein (GFP–GRIP1) was detected almost exclusively in the cytoplasm and nucleus. However, in some cells weak co-localization with STtmd–mRFP was observed in punctate structures (Figure 3c,d, arrows). A second GRIP domain fusion protein, GFP–GRIP2 (amino acids 605–788), consisting of the C-terminal 185 aa of AtGRIP fused to GFP, located much more efficiently to Golgi stacks (Figure 3e,f), although a considerable amount of the fusion protein was still detected in the cytoplasm. These results indicate that the residues adjacent to the GRIP domain in AtGRIP greatly enhance Golgi localization of the GRIP domain.

Figure 3.

The GRIP domain containing C-terminal 185 aa of AtGRIP target the Golgi.
Confocal images of tobacco CB137 leaf epidermal cells 2 days after Agrobacterium inoculation at OD600 = 0.1. (a) GFP–Nterm (green image); (b) GFP–Nterm + STtmd–mRFP (red, merged image), projection of 18 scans (18.11 μm depth). (c) GFP–GRIP1 (green image); (d) GFP–GRIP1 + STtmd–mRFP (merged image), projection of 10 scans (11.37 μm depth). (e) GFP–GRIP2 (green image); (f) GFP–GRIP2 + STtmd–mRFP (merged image), 18 scans (18.36 μm depth). (g) GFP–GRIP2-Y717A (green image); (h) GFP–GRIP2-Y717A + STtmd–mRFP (merged image), 10 scans (12.05 μm depth). Bar = 10 μm.

Previous studies have shown that mutation of a conserved tyrosine in the GRIP domain to alanine disrupts Golgi localization (Kjer-Nielsen et al., 1999; Munro and Nichols, 1999). The corresponding tyrosine residue in AtGRIP was identified as Y717 and replaced by an alanine. The GRIP domain with the Y717A mutation (GFP–GRIP2-Y717A) failed to localize to Golgi stacks, and was instead distributed to the cytosol (Figure 3g,h). This demonstrates that tyrosine 717 in AtGRIP is crucial for Golgi localization in plants.

An Arabidopsis ARL1 homologue localizes to Golgi stacks

In animals and yeast, the GRIP domain binds to a small GTPase, ARL1. To find out whether ARL1 is conserved as GRIP domain receptor in plants, the Arabidopsis genomic database was searched using the blast algorithm with human ARL1 (HsARL1) as a query sequence. Many of the Arabidopsis ARFs and ARF-like proteins share significant similarity with HsARL1. Alignment of ARL1 proteins from rat, fly, worm and budding yeast with ARF1 from rat and budding yeast revealed several residues that are conserved in all ARL1 proteins, but not in ARF1 (Q38, K62, T72, S73, Y77, C80, A87, D98, S159 and K162; Lu and Hong, 2003). From the group of Arabidopsis proteins identified in the blast search with human ARL1, one was found to possess all but one (S159) of the residues that are characteristic for ARL1 and is 65% identical to human ARL1 (Figure 4). This protein has been described previously (Lebas and Axelos, 1994) and, although the authors noticed the sequence similarity with the ARL1 family of proteins, it was designated ARF3. The cDNA encoding this protein, corresponding to ORF At2g24765, was amplified from an Arabidopsis cell-culture cDNA pool. The cDNA was inserted into a binary vector to create a translational fusion to the 5′ end of the GFP gene. The resulting protein was expressed transiently in tobacco CB137, and the infiltrated leaves were examined by confocal microscopy. Green moving bodies were detected in epidermal cells. The same bodies were labelled with STtmd–mRFP, showing that this protein localizes to Golgi stacks (Figure 5a,b). Based on its similarity with human ARL1 and its localization to Golgi stacks, this protein was renamed AtARL1. AtARL–YFP was co-expressed with AtGRIP–GFP in tobacco CB137 epidermal cells. When both proteins were expressed at low levels, ARL1 and AtGRIP located to the same part of the Golgi stacks (Figure 5g,h). When ARL1–YFP was expressed at higher levels, the fusion protein was also detected in other parts of the stacks.

Figure 4.

Alignment of AtARL1 with orthologues across kingdoms.
Alignment of AtARL1 with ARL1 proteins from Drosophila melanogaster (Dm), Caenorhabditis elegans (Ce), Homo sapiens (Hs) and Saccharomyces cerevisiae (Sc), and with Arabidopsis thaliana ARF1 (At2g47170). Sequences were aligned using clustalw and are shaded where more than half the residues are related (grey) or identical (black). ARL1-specific residues are indicated by an asterisk.

Figure 5.

AtARL1 locates to Golgi stacks, while GDP- and GTP-restricted mutants show different patterns and AtARL co-locates with AtGRIP.
Confocal images of tobacco CB137 leaf epidermal cells 2 days after Agrobacterium inoculation. (a) AtARL1–GFP (green image); (b) AtARL1–GFP + STtmd–mRFP (merged image), 10 scans (9.72 μm depth). (c) AtARL1-T31N–GFP (green image); (d) AtARL1-T31N–GFP + STtmd–mRFP (merged image), 10 scans (10 μm depth). (e) AtARL1-Q71L–GFP (green image); (f) AtARL1-Q71L–GFP + STtmd–mRFP (merged image), 10 scans (10.68 μm depth). (g, h) AtARL1–YFP (blue) + AtGRIP–GFP (green) + STtmd–mRFP (red): (g) merged image of green and red channel; (h) merged image of blue and red channel. AtARL1–GFP, AtARL1-T31N–GFP, AtARL1-Q71L–GFP and AtGRIP–GFP were infiltrated at OD600 = 0.1. AtARL1–YFP was infiltrated at OD600 = 0.05. Bar = 10 μm (a–f) or 1 μm (g, h). Bar insets = 3 μm (a–f) or 1 μm (g, h).

To investigate differences between AtARL1 in the GDP- and GTP-bound forms, amino acid changes were introduced that were previously reported to result in GTP- and GDP-restricted forms of small GTPases (Dascher and Balch, 1994; Lu et al., 2001; Zhang et al., 1994). To create the GDP-restricted form, a threonine residue at position 31 was replaced by asparagine (T31N). To create the GTP-restricted form, a glutamine residue at position 71 was replaced by leucine (Q71L). ARL1-T31N and ARL1-Q71L were fused to the N-terminus of GFP and the fusion proteins were transiently expressed in tobacco leaves. ARL1-T31N–GFP localized to Golgi stacks and was detected in the cytosol or ER (Figure 5c,d). Wild-type ARL1–GFP was also occasionally found in the cytosol, but this was observed only in cells that expressed very high levels of the fusion protein. ARL1-T31N–GFP, in contrast, was detected in the cytosol or ER, even when levels of expression were very low. ARL1-Q71L–GFP similarly located to Golgi stacks, but was also detected on a large number of small, irregularly shaped, moving bodies (Figure 5e,f and inset). These bodies occasionally associated with Golgi stacks, forming irregularly shaped complexes (Supplementary Material, Movies 2 and 3). We attempted to identify the nature of the bodies and co-expressed ARL1-Q71L–GFP with proteins known to be involved in post-Golgi trafficking. The plant Rab proteins AtRabF1 (Ara6) (from Arabidopsis) and m-Rab, its homologue from Mesembryanthemum crystallinum, label pre-vacuolar compartments (Bolte et al., 2004; Ueda et al., 2004). AtARL1-Q71L–GFP was co-expressed with AtRabF1 (results not shown) and m-Rab (Figure S1), but neither of these proteins co-located with the AtARL1-Q71L–GFP-specific bodies, showing that these bodies are unlikely to be pre-vacuolar compartments.

Overexpression of ARL1 creates more binding sites for GRIP domains on Golgi stacks

Untagged AtARL1 was co-expressed in tobacco epidermal cells together with GFP–GRIP2. To ensure that ARL1 was expressed in most cells, the Agrobacterium strain carrying the construct from which ARL1 is expressed was infiltrated at a high concentration (OD600 = 0.2). The strain carrying the construct from which GFP–GRIP2 is expressed was infiltrated at a lower concentration (OD600 = 0.05). GFP–GRIP2 on its own is detected on Golgi stacks and in the cytosol (Figure 6a,c). In the presence of overexpressed ARL1, all GFP–GRIP2 was detected on Golgi stacks and there was hardly any GFP–GRIP2 present in the cytosol (Figure 6b,e). When ARL1-T31N was used instead of wild-type ARL1, GFP–GRIP2 was found predominantly in the cytosol (Figure 6g). In contrast, GFP–GRIP2 in an ARL1-Q71L background was found mainly on Golgi stacks, as was the case with wild-type ARL1 (Figure 6i). However, some GFP–GRIP2 was detected on small bodies similar in size to those detected with ARL1-Q71L–GFP (Figure 6i, inset). To find out if binding of the GRIP domain was specific, ARL1 was co-expressed with GFP–GRIP2-Y717A. All GFP–GRIP2-Y717A was cytosolic in the presence of ARL1, and this pattern did not change when ARL1 was replaced by ARL1-T31N or ARL1-Q71L (Figure 6d,f,h,j). This demonstrates that ARL1, and the mutants ARL1-T31N and ARL1-Q71L do not bind GFP–GRIP2-Y717A, and that binding is specific for GFP–GRIP2.

Figure 6.

Overexpression of AtARL1 creates more binding sites on the Golgi for the GRIP domain of AtGRIP.
Confocal images of tobacco leaf epidermal cells 2 days after Agrobacterium infiltration. Strains carrying AtARL1, AtARL1-T31N, AtARL1-Q71L were inoculated at OD600 = 0.2; those carrying GFP–GRIP2 or GFP–GRIP2-Y717A at OD600 = 0.05. (a) GFP–GRIP2, 25 scans (50 μm depth). (b) ARL1 + GFP–GRIP2, 25 scans (44 μm depth). (c) GFP–GRIP2, eight scans (7.3 μm depth). (d) GFP–GRIP2-Y717A, 14 scans (15.63 μm depth). (e) ARL1 + GFP–GRIP2, eight scans (8.34 μm depth). (f) ARL1 + GFP–GRIP2-Y717A, 10 scans (9.59 μm depth). (g) ARL1-T31N + GFP–GRIP2, 10 scans (11.23 μm depth). (h) ARL1-T31N + GFP–GRIP2-Y717A, 10 scans (8.76 μm depth). (i) ARL1-Q71L + GFP–GRIP2, 10 scans (9.04 μm depth). (j) ARL1-Q71L + GFP–GRIP2-Y717A, 10 scans (10.79 μm depth). Arrows mark small structures associated with AtARL1-Q71L expression. Bar = 50 μm (a, b) or 5 μm (c–j) Bar insets = 3 μm.

AtARL1 binds directly to the GRIP domain of AtGRIP in a GTP-dependent manner

To investigate whether the interaction between AtARL1 and the GRIP domain of AtGRIP was direct and could be reconstituted in vitro, AtARL1 and the GRIP domain fragment of AtGRIP (GRIP2) were both expressed in Escherichia coli. AtARL1, ARL1-T31N and ARL1-Q71L were fused with glutathione-S-transferase (GST) to facilitate protein purification. Like most members of the ARF family, AtARL1 possesses a glycine at position +2 that is myristoylated, allowing the protein to be associated with the membrane in its active conformation (reviewed by Burd et al., 2004). To improve solubility of the GST–ARL1 fusion protein, the first 14 aa of AtARL1 were removed. Previous experiments with an ARF mutant lacking the first 17 aa showed that such removal of the N-terminus did not affect effector binding (Panic et al., 2003a; Paris et al., 1997). GRIP2 and GRIP2-Y717A were tagged with six histidines (His6-tag) on the N-terminus, to allow detection of the proteins on Western blot using anti-His6 antibodies. GST–ARL1 fusion protein was immobilized on glutathione sepharose beads in the presence of GDP or the non-hydrolysable GTP analogue, GTPγS. Bacterial lysates of strains expressing His6-GRIP2 or His6-GRIP2-Y717A were applied onto the GST–ARL1 beads, again in the presence of GDP or GTPγS. The bound proteins were eluted in a buffer containing the opposite nucleotide (GTPγS for the ARL-GDP beads and GDP for the ARL-GTP beads). The eluted proteins were analysed on Coomassie blue-stained SDS–PAGE (Figure 7a) and on Western blot incubated with an anti-His6 antibody (Figure 7b). As shown in the top panel of Figure 7(a), there was clear binding between the His6-tagged GRIP domain and GST–ARL1 in the presence of GTPγS. In the presence of GDP, no binding was observed. There was only a low level of binding between His6-GRIP2-Y717A and ARL1-GTP and ARL1-GDP. The interaction between the GTP-restricted mutant, ARL1-Q71L and His6-GRIP2 was comparable with wild-type ARL1, whereas the GDP-restricted mutant ARL1-T31N showed no binding. Again, His6-GRIP2-Y717A showed no significant binding to either mutant. The lower bands in the Coomassie-stained gels and on the Western blots presumably correspond to a breakdown product of the full-length protein.

Figure 7.

The GRIP domain of AtGRIP binds directly to AtARL1 in a GTP-dependent manner.
(A) Coomassie-stained SDS–PAGE gel of total lysates (lys) of Escherichia coli strains expressing His6-GRIP2 or His6-GRIP2-Y717A and of proteins that bound to GST-AtARL1, GST-AtARL1-T31N or GST-AtARL1-Q71L. GST-AtARL1 was preloaded with GDP or the non-hydrolysable analogue of GTP, GTPγS. Bound proteins were eluted in elution buffer with the opposite nucleotide and applied to SDS–PAGE.
(B) Western blot of the gel shown in the top panel of (a) probed with an anti-His antibody. Arrows indicate the band of the size of full-length GFP–GRIP2. The band below GFP–GRIP2 is a possible breakdown product of the full-length protein.
(C) Results of yeast two-hybrid analysis. Yeast transformants were plated on selection plates. The presence of two interacting partners allows yeast to grow in the absence of histidine and adenine. GAL4BD = GAL4-binding domain; GAL4AD = GAL4-activation domain. (a) GAL4BD–AtGRIP (full length) + GAL4AD-ARL1. (b) GAL4BD–AtGRIP (full length) + GAL4AD-ARL1-Q71L. (c) GAL4BD–GRIP2 + GAL4AD-ARL1. (d) GAL4BD–GRIP2 + GAL4AD-ARL1-Q71L. (e) GAL4BD–GRIP2-Y717A + GAL4AD-ARL1. (f) GAL4BD–GRIP2-Y717A + GAL4AD-ARL1-Q71L.

Next, binding between AtGRIP and AtARL1 was analysed using the yeast two-hybrid system (Chien et al., 1991). Wild-type ARL1 and ARL1-Q71L were fused at their N-termini with the activation domain of GAL4, again excluding the first 14 aa of ARL1. AtGRIP and the GRIP domain fragments GRIP2 and GRIP2-Y717A were fused to the C-terminus of the GAL4-binding domain. A prey and bait construct was sequentially transformed into yeast and the transformants were plated on selection plates. Only the combination of GAL4BD–AtGRIP or GAL4BD–GRIP2 with GAL4AD-ARL1-Q71L produced colonies (Figure 7c) and these colonies tested positive in the X-gal assay (results not shown), indicating that these two fusion proteins interact. Neither AtGRIP nor GRIP2 interacted with wild-type ARL1 in this assay, and GRIP2-Y717A tested negative with both ARL1 and ARL-Q71L.


Plant Golgi stacks were first visualized in living cells by Boevink et al. (1998) using a fluorescently tagged fragment of a rat Golgi enzyme, sialyl transferase (STtmd) and the Arabidopsis homologue of the HDEL receptor (AtERD2). Numerous bodies were highlighted, all of similar size, that were closely associated with the ER. Electron microscopy confirmed that these bodies were Golgi stacks. This report also confirmed the striking differences in the morphologies of the plant Golgi apparatus and this organelle in mammals, which generally is stationary and clustered around the nucleus. A number of plant Golgi enzymes and receptors have been identified since, and vesicle trafficking in plants is being extensively studied regarding the roles of Rabs, ARFs, COPs and SNAREs (reviewed by Hawes, 2005; Nebenführ and Staehelin, 2001).

Golgins play an important role in vesicle transport. They typically form homodimers that can bind vesicles to tether them in close proximity to the target membrane. This then enables the v- and t-SNARE helices to form complexes and to complete the vesicle-fusion event (Whyte and Munro, 2002). A range of non-tethering functions has also been suggested for the golgins (Barr and Short, 2003), which makes them interesting candidates for studying plant Golgi organization. Gilson et al. (2004) first identified AtGRIP and demonstrated the Golgi localization of its GRIP domain. The identification and Golgi localization of this protein show that the GRIP domain class of golgins is conserved in plants. AtGRIP is similar to its animal and yeast homologues in that extensive coiled-coil domains are followed by a GRIP domain at the C-terminus of the protein.

A YFP fusion protein of the t-SNARE Memb11 and SYP31 were reported to relocate to the ER after BFA treatment (Uemura et al., 2004; P. Moreau, personal communication), which is indicative of proteins located in the cis or medial cisternae (Nebenführ et al., 2002). YFP–BS14a showed an ‘aggregate-type pattern’ in the cytoplasm, suggesting that this protein is located further to the trans-side (Uemura et al., 2004). To determine the sub-Golgi location of AtGRIP, YFP-tagged SYP31 and Memb11 were used to highlight the cis face of Golgi stacks in the presence of STtmd–mRFP, a well established medial to trans-Golgi marker (Boevink et al., 1998; Saint-Jore et al., 2002; Wee et al., 1998). Co-expression of these t-SNAREs with AtGRIP–GFP showed that AtGRIP locates to the poles of the stacks opposite SYP31 and Memb11, suggesting that AtGRIP locates even further to the trans side of the stack than does STtmd–mRFP. Gilson et al. (2004) also observed that the GRIP domain of AtGRIP located to a position adjacent to the α-Man I marker and envisaged a trans-Golgi location for this protein. In mammals, GRIP proteins are found in the TGN, the network of tubules on the trans side of the mammalian Golgi apparatus. It remains uncertain whether AtGRIP–GFP locates to the far trans-Golgi or to the TGN. Electron microscopic studies have shown that in plants some, but certainly not all, Golgi stacks possess a TGN-like structure (Saint-Jore-Dupas et al., 2004). Moreover, fluorescent protein expression of SNARE constructs suggests that TGN-like structures are occasionally in close proximity to Golgi stacks, but are also often seen separated from the stacks (Uemura et al., 2004; M. Latijnhouwers, unpublished data). This description of the plant TGN does not correspond with the pattern of AtGRIP–GFP in tobacco leaves, which was not normally found dissociated from the Golgi. Throughout this study, AtGRIP–GFP was imaged in cells expressing the protein at low levels to minimize effects of protein overexpression. However, the possibility cannot be excluded that its distribution or the structure of the Golgi are still affected by increased levels of AtGRIP. Further work will be required to address this issue.

Domain localization using GFP fusion proteins showed that the GRIP domain on its own (the 45-aa construct GRIP1) shows only very weak, barely detectable Golgi labelling. However, Gilson et al. (2004) showed that the same 45-aa region extended with the C-terminal 32 aa of AtGRIP-labelled Golgi stacks in Nicotiana plumbaginifolia suspension cells (Gilson et al., 2004). The 3D structure of the GRIP domain has demonstrated that it binds to ARL1 as a dimer (Luke et al., 2005). The amino acids adjacent to the GRIP domain are thought to aid dimerization of the GRIP domain through the formation of coiled-coils. This explains why our 185-aa GRIP2 construct targeted the Golgi much more efficiently than the GRIP domain alone. This fusion protein, however, still showed high cytoplasmic background, probably due to high levels of expression. Overexpression of GRIP domain fragments may also displace endogenous, full-length GRIP domain proteins (Kjer-Nielsen et al., 1999). In HeLa cells, this led to altered distribution of TGN components and inhibition of vesicle transport from the TGN to the plasma membrane in HeLa cells (Yoshino et al., 2003).

GRIP domain proteins are effectors for the small GTPase ARL1 in animals and yeast, and the presence of Golgi-located ARL1 is a requirement for the GRIP domain to target the Golgi (Panic et al., 2003a; Setty et al., 2003). A protein was identified in the Arabidopsis genome that shows similarity to the ARF family of small GTPases, and that possesses nearly all ARL1-specific residues. This protein, previously named ARF3 (Lebas and Axelos, 1994), was demonstrated to locate to Golgi stacks, and we renamed it AtARL1. Based on the level of autofluorescence, cells expressing very high levels of AtGRIP–GFP generally seemed unhealthy, so only cells expressing low levels were imaged. In contrast, high expression levels of AtARL1–GFP did not noticeably harm cells.

When the GFP-labelled GRIP domain fragment GRIP2 was expressed independently, Golgi labelling was accompanied by high cytoplasmic fluorescence. In contrast, when untagged ARL1 was expressed in the same cells as GFP–GRIP, nearly all GFP–GRIP2 fluorescence was collected on Golgi stacks. This suggests that ARL1 recruits GFP–GRIP2 to the Golgi, and that increasing the amount of ARL1 increases the number of binding sites for GFP–GRIP2 on Golgi stacks. In a background of overexpressed ARL1-T31N, the GDP-locked form of ARL1, GFP–GRIP2, was not detected on Golgi stacks. This suggests that ARL1-T31N may interfere with binding of the GRIP domain to the Golgi. As with wild-type AtARL1, overexpression of ARL1-Q71L, the GTP-locked form of ARL1, increased the amount of GFP–GRIP2 on Golgi stacks, but was also visible on small, irregularly shaped structures. This suggests that ARL1-Q71L also recruits GFP–GRIP2 to these structures.

The results of affinity chromatography using recombinant GRIP domain and AtARL1 protein, as well as the yeast two-hybrid analysis, demonstrate that the interaction between the two proteins is direct and that no other proteins are required. As was previously shown for yeast ARL1 (Panic et al., 2003a; Setty et al., 2003), only AtARL1 in the GTP-bound state has the ability to bind the GRIP domain. Studies of the structure of ARL1 have revealed that, in the GDP-bound state, the switch I region (one of the regions in the protein involved in binding nucleotide-exchange factors and effectors) forms a conformation that provides severe steric hindrance to binding of the GRIP domain (Amor et al., 2001; Panic et al., 2003b; Wu et al., 2004). GTP binding brings about a conformational change that removes the steric interference and exposes all residues that are required for interaction with the effector. In the yeast two-hybrid system, the interaction between the prey (ARL1) and bait (AtGRIP) takes place in the nucleus, while the exchange factors for ARL1 will be located on the Golgi apparatus. ARL1 is therefore likely to be predominantly in the GDP-bound form, which explains the lack of interaction between GRIP2 and wild-type AtARL1 in this assay. Previous reports also described that the binding between the GRIP domain and wild-type ARL1 was less efficient or non-existent compared with the GTP-restricted form of ARL1 (Lu and Hong, 2003; Van Valkenburgh et al., 2001).

A GDP-restricted form of human ARL1 was shown to locate to the cytosol in CHO cells (Lu et al., 2001). In our experiments, overexpression of AtARL1-T31N–GFP in tobacco epidermal cells resulted in higher levels of the fusion protein in the cytosol or in the ER compared with wild-type AtARL1–GFP. However, while overexpression of the human ARL1-T31N caused the Golgi to disintegrate, some of the Arabidopsis ARL1-T31N was observed on Golgi stacks that were also labelled by STtmd–mRFP, showing that the stacks are (at least partially) intact. The GTP-restricted mutant of human ARL1, ARL1-Q71L, causes expansion of the Golgi apparatus in CHO and NRK cells (Lu et al., 2001; Van Valkenburgh et al., 2001). Expression of the GTP-restricted form of the Arabidopsis ARL1 fused to GFP (AtARL1-Q71L–GFP) in tobacco cells resulted in green-labelled Golgi, and a large number of labelled structures that were smaller than Golgi stacks but of very irregular shape and size. These structures occasionally associated with the Golgi and appeared to form complexes with the stacks, but the identity of these structures remains to be determined.

ARL1 and GRIP domain proteins are thought to be involved in post-Golgi trafficking. In vivo and in vitro experiments showed that, in mammalian systems, transport of a fragment of Shigatoxin B from endosomes to the Golgi apparatus was dependent on the presence of the GRIP domain protein golgin-97 and on ARL1. This shows that both proteins play an important role in endosome-to-Golgi trafficking (Lu et al., 2004). Unfortunately, there is to date no marker for endosome-to-Golgi transport in plants, and preliminary experiments suggest that, in cells overexpressing AtARL1 or the mutant forms of the protein, there is no inhibition of the trafficking of post-Golgi markers. Interestingly, in yeast, deletion of either the GRIP domain protein IMH1 or ARL1 causes only mild phenotypes, and it is hypothesized that a pathway involving the yeast RAB6 orthologue YPT6 is partially redundant with the ARL1 pathway (Graham, 2004). Recently it has become clear that ARL1 is recruited to the Golgi by another ARF-like GTPase, ARL3 (yeast) or ARFRP1 (human), although the mechanism behind it is unknown. ARL3 binds an integral membrane protein located in the TGN, called SYS1 (Behnia et al., 2004; Setty et al., 2004). Both ARL3 and SYS1 have potential homologues in Arabidopsis, and an obvious next step will be to find out if their functions in the localization of the GRIP domain and ARL1 are conserved in plants.

Experimental procedures

Construction of AtGRIP and ARL1 expression plasmids

Standard molecular techniques were used as described by Ausubel et al. (1999). Primers were obtained from MWG Biotech (Ebersberg, Germany). Restriction enzymes were from New English Biolabs (Hitchin, UK). Expand HiFi polymerase (Roche, Basel, Switzerland) was used for PCR.

Total RNA from an Arabidopsis Col-0 cell suspension culture was used in a first-strand cDNA synthesis reaction (Superscript III; Invitrogen, Paisley, UK) using oligo dT20 primer. The cDNA of AtGRIP was amplified from this cDNA pool using the primers GC3-F AGGCGCGCCAAAAATGTCCGAAGACAAGGA and GC3-R AAGGAAAAAAGCGGCCGCTGAAAACGAGAATCTTGA. ARL1 cDNA was amplified using the primers ARL1-F AGGCGCGCCAAAAATGGGAATCTTATTCACGCG and ARL1-R GGAAAAAAGCGGCCGCGCCACTTCCCGACTTCAA. The AscI site and the Not1 site in the forward and reverse primers, respectively, are underlined. Using these sites the cDNAs were cloned into pENTR-1a-MCS. This vector was modified from pENTR 1A (Invitrogen) by replacing the ccdB gene with the AscI–NotI linker (CCTGCAGGCGCGCCATATGCGGCCGCACCGGTG). The cDNAs encoding the t-SNAREs Memb11 (At2g36900), SYP31 (At5g05760) and BS14a (At3g58170), AtRabF1 (Ara6) and AtRabF2b (Ara7) were similarly amplified using gene-specific primers possessing AscI and NotI sites, and cloned into pENTR-1a-MCS. From this vector the cDNAs were transferred to binary vectors using the Gateway system following instructions provided by the manufacturer (Invitrogen). The binary vectors pMDC83 and pMDC43 (Curtis and Grossniklaus, 2003) were used for fusions to the N- or C-terminus of mGFP5 (Haseloff et al., 1997), respectively, whereas pB7YWG2 and pB7WGY2 (Karimi et al., 2002) were used for fusions to the N- or C-terminus of EYFP (Clontech, Mountain View, CA, USA), respectively. pMDC32 (Curtis and Grossniklaus, 2003) was used for expression of untagged proteins.

The GRIP domain fragment GRIP1 was amplified from AtGRIP with the primers GRIP1-F AGGCGCGCCAAAAATGGAAAGAAAACAGAAGAGAG and GRIP1-R AAGGAAAAAAGCGGCCGCTGTTGTTGTTGCCGCTGTTG. The larger GRIP domain fragment GRIP2 was created using GRIP2-F AGGCGCGCCAAAAATGACAAATCTTCGAAAATCTA and GC3-R. The N-terminal fragment (N-term) was amplified from AtGRIP using GC3-F and Nterm-R AGAAAAAGCGGCCGCTTATCTCTTCTGTTTTCTTTCCAT. GRIP1, GRIP2 and N-term were cloned into pENTR-1a-MCS and subsequently introduced into pMDC43.

Mutations were created using PCR mutagenesis. ARL1-T31N was created from ARL1 using the primers ARL1-TtoN-F CACAATCCTCTATCGG and ARL1-TtoN-R CCGATAGAGGATTGTGTTTTTTCC. ARL1-Q71L was created from ARL1 using primers ARL1-QtoL-F AACAAGCATCAGGCCATA and ARL1-QtoL-R TATGGCCTGCTGCTTGTTAGTCCA. GRIP2-Y717A was created from GRIP2 using primers GRIP2-YtoA-F ATATGACAGCCCTAAAGAATG and GRIP2-YtoA-R CATTCTTTAGGGCTGTCATAT. The mutagenic bases are underlined.

ARL1 was amplified using CGGGATCCAACAAAGAAGCTCGAATCC and CCCAAGCTTAGCCACTTCCCGACTTCAA and cloned into the BamHI and HindIII sites of pRP265 (Smith and Johnson, 1988) to create fusions with GST. Similarly, GRIP2 and GRIP2-Y717A were amplified using primers with BamHI and HindIII sites. His6-tagged GRIP2 and GRIP2-YtoA were created by cloning the PCR fragments into the BamHI and HindIII sites of pQE-30 (Qiagen, Hilden, Germany). All PCR products were sequenced using the Big Dye Terminator ver. 3.1 cycle sequencing kit (Applied Biosystems, Foster City, CA, USA).

Transient expression in plants

Wild-type Nicotiana tabacum and N. tabacum line CB137 plants were grown in a glasshouse at 22°C (day temperature) and 18°C (night temperature) with a minimum 16 h light. Each construct was transformed to Agrobacterium tumefaciens GV3101 or LBA4404 by electroporation. For Agrobacterium transient expression, strains were grown overnight in a shaking incubator (200 rpm) at 28°C in 5 ml Luria–Bertani (LB) medium + 0.5% glucose (w/v) supplemented with the appropriate antibiotics. The overnight cultures were diluted 1:10 in the same growth medium and left to grow under identical conditions for 6 h. Cultures were then centrifuged (6 min, 2600 g) and the pellets resuspended in 5 ml (10 mm MgCl2, 10 mm MES pH 5.6, 150 μm acetosyringone). The bacterial suspension was diluted with the same buffer to adjust the inoculum concentration to the final OD600 value (see Figure legends). Infiltrations were performed as described by Batoko et al. (2000). The inoculum was delivered to tobacco leaves by gentle pressure infiltration through the stomata of the lower epidermis, using a 1-ml syringe without a needle. For experiments requiring co-infection of more than one construct, bacterial strains containing the constructs were mixed before performing the leaf infiltration, with the inoculum of each construct adjusted to the required final OD600. The infected area of the leaf was delimited and labelled with an indelible pen.

Stable transformation of tobacco

Three days after infiltration with Agrobacterium carrying the appropriate binary vector [in this case pVKH18En6 (Batoko et al., 2000) with 35S-STtmd–mRFP], tobacco leaves were removed, placed in sterilization solution [1:1 hyperchlorite solution:water, 0.01% (v/v) Tween 20] for 5 min, washed three times in sterile distilled water, cut into small pieces and placed on shooting media [0.5 × MS, 0.8% agar, 3% (w/v) sucrose, 0.1 mg l−1 indole butyric acid (IBA) (Sigma Aldrich, Gillingham), 0.8 mg l−1 6-benzylaminopurine (BAP) (Sigma Aldrich, Chelsworth), 0.1 mg l−1 carbenicillin (Melford, Suffolk, UK), 0.2 mg l−1 Ticarcillin/clavulanic acid (Duchefa Biochemie BV, Haarlem, the Netherlands) and hygromycin] to select for the binary vector. The leaf discs were left for 3–4 weeks for shooting to occur, shoots were removed using sterile technique and placed on rooting media (as shooting media without BAP and selection for binary vector, 0.5 mg l−1 IBA) for approximately 10 days. Plantlets were transferred to larger growth containers for screening.

Confocal imaging

All imaging was conducted on a Leica TCS-SP2 AOBS using an HCX APO 63x/0.90 w water-dipping lens. GFP was imaged using 488-nm excitation and its emission was collected from 500–520 nm, or 500–510 nm if imaged in combination with YFP. For mRFP, excitation at 568 nm was used and emission collected at 600–620 nm. The excitation wavelength for YFP was 514 nm and its emission was recorded at 535–545 nm. GFP and mRFP were imaged simultaneously, whereas GFP and YFP were imaged sequentially using a line-by-line mode. The optimal pinhole diameter was set at 1 Airy unit (<0.9 μm for GFP; <1.0 μm for YFP) in all cases. Post-acquisition image processing was done using photoshop ver. 8.0 software (Adobe).

Affinity chromatography with immobilized GTPases

GST–ARL1 was purified, loaded with GDP or GTPγS (Sigma) and bound to glutathione sepharose beads (Amersham Biosciences, Buckinghamshire) as previously described (Gillingham et al., 2004). In brief, E. coli JM109 cells containing the appropriate plasmids were induced at OD600 = 0.7 for 14 h with 0.2 mm isopropyl-β-d-thiogalactopyranoside at 17°C. Bacterial lysates were prepared by sonication in 20 ml lysis buffer [20 mm Tris pH 8.0, 110 mm NaCl, 1% Triton X-100, 5 mmβ-mercaptoethanol, 1 mm phenylmethylsulphonylfluoride (PMSF) and protease inhibitors (Roche)] with 200 μm GDP. The lysates were centrifuged (20 min, 5000 g) and the supernatant incubated with glutathione sepharose at 4°C for 30 min. The beads were washed in lysis buffer with 200 μm GDP, followed by NE buffer (20 mm Tris pH 8.0, 110 mm NaCl, 1 mm EDTA, 5 mm MgCl2, 5 mmβ-mercaptoethanol, 1 mm PMSF and protease inhibitors) with 10 μm either GDP or GTPγS. They were subsequently incubated three times for 30 min at room temperature (22°C) in NE buffer with either 1 mm GDP or GTPγS and 20 min in NS buffer (NE buffer without EDTA) containing the same nucleotides.

Lysates of bacterial strains expressing His6-GRIP2 and His6-GRIP2-Y717A were prepared as described above, except that GDP was omitted from the lysis buffer. The lysates were incubated with GST–ARL1-GDP or GST–ARL1-GTP beads for 2 h at 4°C. Beads were then washed in NS buffer with 10 μm GDP or GTPγS, and the binding proteins eluted in elution buffer (20 mm Tris pH 8.0, 1.5 m NaCl, 2 mm EDTA, 5 mmβ-mercaptoethanol and 1 mm of the opposite nucleotide (GTPγS or GDP). In the case of GST–ARL1-T31N and GST–ARL1-Q71L, the lysis buffers contained 200 μm GDP or GTP, respectively, and the nucleotide-loading step was omitted.

The eluates were analysed on SDS–PAGE and stained with Coomassie brilliant blue, or blotted onto nitrocellulose and probed with an anti-His6 antibody conjugated with HRP (Sigma) in PBS (0.1% Tween 20, 5% milk powder). ECL™ (ECL; Amersham Biosciences) was used for detection.

Yeast 2-hybrid analysis

Full-length AtGRIP, the GRIP domain fragment GRIP2 and GRIP2-Y717A were amplified by PCR using gene-specific primers, both containing SmaI restriction sites. The PCR products were cloned into the SmaI site of the pGBKT7 bait vector (BD Biosciences Clontech, Mountain View, CA, USA). AtARL1 and AtARL1-Q71L were amplified by PCR and cloned into the pGADT7-Rec prey vector, again using the unique SmaI restriction site in the vector. Yeast strain AH109 was transformed sequentially with pGBKT7-AtGRIP, pGBKT7-GRIP2 or pGBKT7-GRIP2-Y717A and pGADT7-Rec-ARL1or pGADT7-Rec-ARL1-Q71L, using a lithium acetate method (Gietz et al., 1995). Colonies were selected on synthetic plates lacking histidine, tryptophan, leucine and adenine, up to 7 days. Positive yeast transformants were replated on plates containing X-α-gal (Clontech) to test the expression of the reporter gene MEL1.


We are grateful to Dr Mark Curtis and the University of Zurich for providing the Gateway binary vectors pMDC43, pMDC83 and pMDC32, the University of Ghent for providing the vectors pB7YWG2 and pB7WGY2, and Roger Y. Tsien for making the mRFP construct available. We thank Susanne Bolte for the m-Rab-YFP construct and Federica Brandizzi for her work on the STtmd–mRFP plants. BBSRC and SEERAD are acknowledged for their financial support, and Sean Munro for his hospitality and advice.