Dynamic behavior of clathrin in Arabidopsis thaliana unveiled by live imaging

Subcellular localization of GFP-tagged clathrin light chain

To visualize clathrin in living cells, we created transgenic plants that expressed green fluorescent protein (GFP) fused to the CLC encoded by the Arabidopsis Genome Initiative locus identifier, At2g40060. This CLC was shown to interact with the hub domain of mammalian CHC (Scheele and Holstein, 2002), and CLC–GFP chimeric proteins have been shown to colocalize with immunodetectable endogenous CHC (Robert et al., 2010). Furthermore, At2g40060 expression is under the control of regulatory elements located in the 5′-flanking, 3′-flanking, and intron sequences of the CLC gene (Figure 1a). Thus, it is likely that the localization of GFP-tagged CLC reflects the localization of endogenous clathrin in structures like clathrin-coated pits and CCVs. The CLC–GFP fusion protein was localized to the plasma membrane, intracellular structures, and the cytosol, consistent with previous observations (Figure 1b; Fujimoto et al., 2010; Konopka et al., 2008; Van Damme et al., 2011). To identify the location of cytoplasmic CLC–GFP-positive points, we stained plant cells with an endocytosis tracer, FM4-64, which indicated that these compartments were endosomal in nature (Figure 1b). The CLC signal on the plasma membrane was detected at cell plates, which were also stained with FM4-64, as reported previously (arrowhead in Figure 1b; Konopka et al., 2008; Van Damme et al., 2011).

Distinct distribution of dynamin and clathrin during cell plate formation

At the beginning of cell plate formation, Golgi-derived vesicles are shown to accumulate at the cell division plane, form tubules, and fuse into tubular networks, which gradually grow into fenestrated sheets (Jürgens, 2005). Electron microscopy and tomographic analyses have shown that many clathrin-coated buds appear on the cell plates as they form (Otegui et al., 2001; Samuels et al., 1995; Seguí-Simarro et al., 2004). Dynamin-like rings have also been observed at constricted tubular regions of hourglass-shaped vesicles and on the fenestrated sheets (Otegui et al., 2001; Seguí-Simarro et al., 2004). Consistent with those observations, the dynamin-related proteins of A. thaliana, DRP1A and DRP2B, accumulate at newly synthesized parts of cell plates as they form in BY-2 cells (Fujimoto et al., 2008). However, to date, a direct comparison of the spatiotemporal localization of clathrin and dynamin has not been made to determine whether clathrin and dynamin act in the same trafficking events. To obtain a further insight into the distribution and dynamics of these proteins during cell plate formation, we examined dividing cells in transgenic plants that expressed both DRP1A–GFP and CLC fused to monomeric Kusabira-Orange (CLC-mKO; Karasawa et al., 2004). We found that DRP1A–GFP accumulated uniformly at the division plane during the early stage of cell plate formation, as reported previously (Fujimoto et al., 2008; Figure 2a,m). On the other hand, CLC-mKO did not localize to the cell plate at this stage (Figure 2e,m). This finding indicates that dynamin acts independently of clathrin during early cell plate development. Subsequently, DRP1A–GFP disappeared from the central part of expanding cell plates and accumulated at the newly synthesized regions, near the edges of cell plates as they formed (Figure 2b,c,n,o). At this stage, CLC-mKO seemed to be excluded from the edges of cell plates, but was predominantly localized in the central region (Figure 2j,n, arrowheads). At the final stage of cytokinesis, both CLC-mKO and DRP1A–GFP signals diminished in intensity in the central region (Figure 2g,o, arrows), but increased in intensity at the edges of the cell plates (Figure 2o, blue arrowheads). Thus, dynamin and CLC partially overlapped throughout cell plate formation. This suggests that dynamin plays a role in cell plate formation that is additional to its interaction with CLC in the formation of CCVs (Otegui et al., 2001; Seguí-Simarro et al., 2004).

A major proportion of clathrin localizes on the trans-Golgi network (TGN)

To determine the precise subcellular localization of CLC–GFP, we compared its localization with several known organelle markers. Because CLC–GFP-positive organelles were shown endosomal in nature (Figure 1b), we established transgenic plants that co-expressed CLC–GFP and a marker for the TGN or the MVE, which are involved in the endocytic trafficking pathway (Dettmer et al., 2006; Ebine et al., 2011; Uemura et al., 2004). These included an mRFP-tagged syntaxin of plants 43 (mRFP-SYP43), which is a marker for the TGN (Ebine et al., 2011; Uemura et al., 2004); the Venus-RHA1, a marker for a subpopulation of MVEs (Anuntalabhochai et al., 1991; Haas et al., 2007); and ARA6-mRFP, a marker for another subpopulation of MVEs (Ebine et al., 2011; Haas et al., 2007). We also generated transgenic plants that co-expressed CLC–GFP and ST-mRFP, a marker for the Golgi trans-cisternae (Boevink et al., 1998). As shown in Figure 3, a major proportion of the CLC–GFP signal overlapped with mRFP-SYP43 (Figure 3c). We also frequently observed that some CLC-positive domains had partially merged or were associated with other organelle markers (Figure 3a,b,d).

For a more quantitative comparison of the subcellular localization of CLC and other organelle markers, we constructed a macro in Metamorph software, which semi-automatically measures the distance from the center of each GFP signal to the center of the nearest RFP or Venus signal in acquired images (Bouttéet al., 2006). We classified the resulting distances into three categories: (i) colocalized: a distance between two centers that was below the resolution limit of the objective lens (0.24 μm in this study); (ii) associated: a distance less than the sum of the radii of two punctate signals (<1 μm in this study); and (iii) independent: a distance larger than the sum of the radii of two punctate signals (>1 μm) (Figure 4a). In the positive control (ARA6–GFP and ARA6-mRFP in Figure 4b), about 70% of the GFP signal was classified as colocalized, and 20% was classified as associated. Among the organelle markers examined, CLC colocalized most closely with SYP43; furthermore, over 50% of the CLC-positive points were classified as colocalized (Figure 4b). This result indicates that CLC predominantly localized at the SYP43-positive TGN; however, a substantial fraction of CLC was also localized in close proximity to RHA1- or ARA6-positive MVEs and the ST-positive trans-Golgi cisternae (Figure 4b).

Dynamic behavior of CLC–GFP in root epidermal cells

We then analyzed the dynamics of clathrin and the organelle markers in root epidermal cells. These cells possess a relatively thin cytoplasmic layer surrounding the vacuole. Most of the CLC–GFP points moved in concert with mRFP-SYP43 points (arrowheads in Figure 5a; Movie S1). We also occasionally observed CLC-positive, SYP43-negative tubules that protruded from the TGN, and vesicles that pinched off from these tubules (arrow in Figure 5a). We also observed that two or more mRFP-SYP43 points were transiently bridged by CLC-positive tubules (blue arrowheads in Figure 5b). These observations suggest that clathrin-positive vesicular and/or tubular carriers mediate traffic between the TGNs, or between the TGNS and other organelles. In addition, we noticed that there were smaller CLC–GFP points that frequently overlapped with the subdomains of ARA6 points (Movie S2); these points moved together in the cytosol (white arrowheads in Figure 6), which suggests a stable association between ARA6-endosomes and a subpopulation of CLC–GFP. Finally, we observed faint ARA6-mRFP signals on larger CLC points (white arrows in Figure 6).

To compare the localizations of CLC, SYP43, and ARA6 in the same cell, we generated transgenic plants that expressed CLC–GFP, mRFP-SYP43, and ARA6-Venus. As observed above, the major proportion of the CLC signal was colocalized with SYP43-positive points (white arrows in Figure 7; Movie S3). We also frequently observed ARA6-positive points independent of CLC–GFP and mRFP-SYP43 points (magenta arrowheads in Figure 7). Furthermore, we occasionally observed a CLC signal independent of either mRFP-SYP43 or ARA6-Venus (white arrowhead in Figure 7). We previously reported that ARA6-localizing MVEs and SYP43-localizing TGNs were independent from each other in root tip cells (Ebine et al., 2011); however, in elongating root epidermal cells, we sometimes observed some membrane domains that bore all three markers (cyan arrowhead in Figure 7). Thus, the majority, but not all, CLC was localized on the TGN, and a subpopulation of CLC resided on the subdomain of MVEs. This probably reflects the multiple functions of clathrin in post-Golgi trafficking pathways.

Effects of BFA on clathrin distribution and dynamics

We next examined the effects of drugs on CLC–GFP distribution and dynamics. BFA inhibits activation of ARF GTPase, which acts in multiple trafficking pathways (Chardin and McCormick, 1999; Nielsen et al., 2008; Robinson et al., 2008). BFA treatment generally causes a redistribution of Golgi-localized proteins to the endoplasmic reticulum (Sciaky et  al., 1997; Saint-Jore et al., 2002). Arabidopsis thaliana possesses a BFA-insensitive ARF GEF, known as GNOM-LIKE 1, which functions at the Golgi apparatus (Richter et al., 2007; Teh and Moore, 2007). Therefore, the major effect of BFA treatment in A. thaliana is the formation of large organelle aggregates that include the TGN and MVE in plant cells; these are known as ‘BFA bodies’ (Ebine et al., 2011; Grebe et al., 2003; Jaillais et al., 2008; Figure 8d). Upon BFA treatment, most of the cytoplasmic CLC–GFP was trapped inside BFA bodies, which surrounded central regions that contained SYP43-positive domains (Figures 8e,f and Figure S1). Treatment with dimethyl sulfoxide (DMSO) did not exert this effect (Figure 8a–c).

We then examined the effect of BFA on CLC–GFP at the plasma membrane by VIAFM. The fluorescence-tagged CLC molecules were observed as punctate foci at the plasma membrane, as reported previously (Fujimoto et al., 2010; Konopka et al., 2008; Naramoto et al., 2010). It was also shown that the CLC constructs colocalized with DRPs and the regulator proteins for ARF GTPases at the plasma membrane. Thus, CLC–GFP most likely labeled clathrin-coated pits/newly-forming CCVs at the plasma membrane. As shown in Figure 8g, the density of clathrin-coated pits was significantly reduced with BFA treatment (mean ± standard deviation): DMSO-treated cells had 118 ± 9.06 foci per 100 μm² area, and BFA-treated cells had 38.3 ± 6.63 foci per 100 μm² area (n = 10 areas; P < 0.01, Student’s t-test; Figure 8h). We also noticed that the lateral movement of some clathrin pits was more prominent in BFA-treated cells than in DMSO-treated cells (arrowheads in Figure 8i; Figure S2; Movie S4, S5). These results indicate that BFA affects CLC distribution and dynamics on the plasma membrane, the TGN, and MVEs.

Effects of wortmannin on clathrin localization and dynamics

We next examined the effect of wortmannin (Wm), a phosphatidylinositol 3-kinase inhibitor, on CLC–GFP. ARA6-positive endosomes were swollen to form vacuolated ring-like structures as previously described (Grebe et al., 2003; Jaillais et al., 2008; Ebine et al., 2011). Regarding the TGN, a TGN protein, secretory carrier membrane protein 1 (SCAMP1) was reported to localize on dilated ring-like structures in Wm-treated BY2 cells (Lam et al., 2007; Wang et al., 2009); however, the mRFP-SYP43-labeled TGN was insensitive to Wm treatment in our experimental system (Ebine et al., 2011; Figure 9a). Similarly, most CLC–GFP points were not markedly affected by Wm; this was consistent with the notion that CLC–GFP was predominantly localized on the TGN (Figure 9b,c). We also found that noticeable amounts of CLC–GFP and GFP-SYP43 were closely associated with (or localized on the subdomains of) ring-shaped ARA6-positive structures (arrows in Figure 9a,c).

The effect of Wm was more evident at the plasma membrane, when observed with VIAFM. Wm caused aggregation of clathrin patches at the plasma membrane (white arrowheads in Figure 9d; Movie S6). This finding may be related to the inhibitory effect of Wm on plant endocytosis (Dettmer et al., 2006; Ebine et al., 2011; Emans et al., 2002; Reichardt et al., 2007; Robatzek et al., 2006; Yamada et al., 2005). We then compared the resting times of CLC molecules in Wm-treated and DMSO-treated conditions. We randomly chose 100 CLC-positive points observed in the first frame, generated kymographs over 16 sec for each point of CLC–GFP (data not shown), and examined the time taken to disappear from the focal plane by sequestration into the cytoplasm. In DMSO-treated cells, the resting time of each CLC point was highly variable (dark purple bars in Figure 9e). This finding was probably due to the variable time required for loading cargos into clathrin-coated pits. Alternatively, it may have been due to the variable time required to form clathrin-coated pits. On the other hand, in Wm-treated cells, 92 out of the 100 points analyzed remained at the plasma membrane for the entire 16 sec (light purple bars in Figure 9e). Thus, the resting time of CLC was extended by Wm treatment. We next examined changes in the distribution pattern of clathrin-coated domains at the plasma membrane over 15 sec. We superimposed two images of one region taken 15 sec apart, where clathrin domains were colored green (time 0 sec) or red (time 15 sec) (Figure 9f). In DMSO-treated samples, most of the CLC dots did not overlap. This indicated that the distribution pattern of clathrin-coated domains was almost completely rearranged within 15 sec (Figure 9f, upper panels). In sharp contrast, the green and red images of Wm-treated samples overlapped quite well. This indicates the stabilization of clathrin-coated pits by Wm treatment at the plasma membrane (Figure 9f, lower panels). Accordingly, Pearson correlation coefficients (Rr) calculated for 10 independent images were significantly higher for Wm-treated cells (0.89 ± 0.022) than for DMSO-treated cells (0.38 ± 0.045; P < 0.01, Student’s t-test, Figure 9f). These results suggest that the inhibitory effect of Wm on plant endocytosis was exerted by aggregation and stabilization of clathrin-coated pits at the plasma membrane.

Intracellular localization of clathrin light chain

To date, apparently inconsistent results have been reported on the intracellular localization of clathrin (Dhonukshe et al., 2007; Kang et al., 2011; Konopka et al., 2008; Stierhof and El Kasmi, 2010). This prompted us to use a robust method to perform a detailed, quantitative study to compare the subcellular localizations of clathrin and organelle markers. In this study, we visualized clathrin with CLC–GFP in A. thaliana and detected its intracellular localization and dynamics in living cells. We found that a major population of cytoplasmic CLC–GFP was localized on the SYP43-positive TGN. Moreover, live imaging indicated that CLC–GFP moved in concert with SYP43-positive membrane domains in root epidermal cells. In addition, CLC–GFP signals were also frequently observed close to ST-positive Golgi trans-cisternae, but colocalization of CLC–GFP and ST-mRFP was rarely observed. Previous studies that reported clathrin localization on the Golgi trans-cisternae might be explained by the close association between the TGN and the trans-cisternae of the Golgi (Dettmer et al., 2006; Kang et al., 2011; Viotti et al., 2010). The quantitative localization method used in this study represents a powerful tool for reliable evaluation of colocalization.

We also found that smaller CLC–GFP signals were tightly associated with ARA6-positive MVEs. In animal and yeast systems, clathrin has been shown to participate in the formation of intraluminal vesicles in MVEs (Sachse et al., 2002). Additionally, in plant cells, clathrin plaques and membrane invaginations at the rims of clathrin plaques were observed on the limiting membranes of MVEs by electron microscopy (Robinson et al., 2011; Stierhof and El Kasmi, 2010); this implies involvement of clathrin in the formation of intraluminal vesicles in the MVEs of plant cells. However, in animal cells, it was also reported that CCV formation occurred at the endosomes/lysosome interface (Brodsky et al., 2001). Further studies are needed to elucidate the precise function of clathrin on MVEs of plant cells.

Dynamin and clathrin at the forming cell plate

In plants, the dynamin-related proteins, DRP1A and DRP2B, have been shown to colocalize with clathrin at the plasma membrane. This suggests that the functions of dynamin and clathrin are coordinated in CCV formation, as proposed for the animal system (Konopka et al., 2008; Fujimoto et al., 2010). It was also reported that both dynamin and clathrin were localized on cell plates as they formed (Fujimoto et al., 2008; Konopka et al., 2008; Otegui et al., 2001; Samuels et al., 1995; Seguí-Simarro et al., 2004; Tahara et al., 2007; Van Damme et al., 2011). However, unlike their activities at the plasma membrane, we found that dynamin and clathrin behaved differently in cell plate formation (Figure 2). This finding indicates that these molecules contributed to distinct cellular events in cell plate formation.

Cell plate formation processes were intensively analyzed previously with electron tomography (Otegui et al., 2001; Samuels et al., 1995; Seguí-Simarro et al., 2004). In the early stages of cell plate formation, Golgi-derived vesicles fused at the cell plate assembly matrix, and this gave rise to a new cell plate (Seguí-Simarro et al., 2004). The fused vesicles first took on an hourglass-like shape. Dynamin-like helices were observed at the neck of these hourglass-shaped vesicles, and they seemed to cause tubulation of the neck region. This resulted in a deformation from the hourglass-shaped to a dumbbell-shaped structure (Otegui et al., 2001; Seguí-Simarro et al., 2004). On the other hand, clathrin-coated buds and CCVs were not observed during the early stages of cell plate formation. At later stages, these structures began to appear at the central region of the tubulo-vesicular network, which was formed by the fusion of the dumbbell-like vesicles (Otegui et al., 2001; Samuels et al., 1995; Seguí-Simarro et al., 2004). In this study, we demonstrated sequential recruitment of DRP1A and CLC to cell plates as they formed in living cells. DRP1A was recruited to the cell plates during the early stages, and CLC was observed at cell plates at the later stages, when DRP1A was concentrated at the edges of cell plates. These observations are consistent with the model proposed previously based on electron tomography; dynamin and clathrin act in different processes during cell plate formation. However, our observation of faint, but significant dynamin localization at the center of cell plates suggests that dynamin may also be involved in CCV formation at the subregions of cell plates.

Effects of BFA on clathrin localization and dynamics

We showed that clathrin patches at the plasma membrane were sensitive to BFA and Wm. This result appeared to contradict a previous report by Konopka et al. (2008), who argued that BFA and Wm did not exert any effects on clathrin dynamics. This apparent discrepancy could be explained by the different treatment times employed in these experiments; we treated seedlings for 1 and 2 h with 50 μm BFA and 33 μm Wm, respectively; in contrast, Konopka et al. (2008) treated samples for 20 min with 100 μm BFA and 33 μm Wm. Under our experimental conditions, BFA caused aggregation of the TGN and MVEs into BFA bodies, as previously reported (Ebine et al., 2011; Jaillais et al., 2008; Uemura et al., 2004); and Wm treatment caused MVE deformation (Ebine et al., 2011).

With VIAFM, we found that BFA treatment significantly reduced the number of clathrin patches at the plasma membrane (Figure 8g,h). Based on the observation that BFA caused the aggregation of cytoplasmic CLC into BFA bodies, a plausible interpretation was that sequestering CLC into BFA bodies resulted in a reduction of the amount of CLC available for recruitment to the plasma membrane. Alternatively, BFA might accelerate clathrin-mediated endocytosis. This notion would be consistent with the previous report that uptake of a fluorescent endocytosis tracer, FM1-43, was promoted by BFA treatment in tobacco BY-2 cells (Emans et al., 2002). However, it has been also reported that a mutation in GNOM, a BFA-sensitive ARF GEF of A. thaliana, causes perturbation of endocytosis (Naramoto et al., 2010). Because BFA is known to have different effects on different plant species and tissues (Robinson et al., 2008), further studies are needed to determine its mechanism of action in other plants and tissues.

The reduced number of clathrin-coated pits might also be explained by the reduction of endocytic cargo proteins on the plasma membrane. BFA is known to perturb recycling from endosomes to the plasma membrane in A. thaliana (Geldner et al., 2003), and this resulted in an accumulation of plasma membrane proteins (like PIN1) in BFA bodies (Dhonukshe et al., 2007; Geldner et al., 2003; Robatzek et al., 2006; Takano et al., 2005). BFA is also known to interfere with the secretory pathway (Satiat-Jeunemaitre et al., 1996). This situation could also lead to depletion of cargo molecules on the plasma membrane. Decreased trafficking activity would also affect the supply of new membrane to replenish the plasma membrane. Thus, the amount of cargo and/or membranes may be insufficient to evoke endocytosis on the plasma membrane of BFA-treated cells.

Another noteworthy BFA effect on clathrin patches at the plasma membrane was the acceleration of lateral movement (Figure 8i). This result suggests that the ARF GEF responsible for CCV formation at the plasma membrane may also stabilize clathrin-coated pits by restricting lateral movement.

Wortmannin inhibition of endocytosis in plant cells

Wm, a compound known to inhibit phosphatidylinositol 3-kinase (PI3K), is widely used in the plant cell biology field to inhibit endocytosis (Dettmer et al., 2006; Ebine et al., 2011; Emans et al., 2002; Reichardt et al., 2007; Robatzek et al., 2006; Yamada et al., 2005). However, the mechanism of the inhibitory effect on plant endocytosis is unknown. In this study, a VIAFM assessment has shown that Wm has striking effects on the distribution and dynamics of clathrin-coated pits on the plasma membrane; with Wm treatment, clathrin-coated pits formed large aggregates and displayed prolonged resting times (Figure 9d,e). Although further studies are needed to elucidate the precise site of action of Wm, these results strongly suggest that Wm affects the later steps of CCV formation, but not earlier events, like clathrin assembly at the plasma membrane. In the animal system, dynamin, which depends on phosphoinositide metabolism, recruits components and regulates CCV formation (Brodsky et al., 2001; Di Paolo and De Camilli, 2006). Similarly, in plants, the dynamin-related proteins, DRP1A and DRP2A, have been shown to bind PI3P and PI4P (Backues and Bednarek, 2010; Lee et al., 2002) and also assemble at clathrin-coated pits during CCV formation (Fujimoto et al., 2010; Konopka et al., 2008). In Vicia faba, Wm was shown to inhibit production of both PI3P and PI4P (Jung et al., 2002). Taken together, these lines of evidence suggest that Wm may prevent recruitment of dynamin to clathrin-coated pits by altering phosphoinositide metabolism in the plasma membrane. However, other factors involved in endocytosis, including clathrin adaptors, also require specific phosphoinositides for localization and functions (Di Paolo and De Camilli, 2006; Owen et al., 2004). Future studies on how Wm affects these components in CCV formation would be necessary to clarify the precise effect of this drug on plant endocytosis.

Plant materials

The following organelle markers were introduced into plants, as described previously: CLC–GFP, Venus-RHA1, ARA6-mRFP, ARA6-Venus, mRFP-SYP43, and ST-mRFP (Ebine et al., 2008; Fujimoto et al., 2010; Ebine et al., 2011). CLC-mKO was generated by replacing GFP with mKO. For DRP1A–GFP, the cDNA encoding GFP was inserted at the end of the cDNA for DRP1A, just in front of the stop codon. The cDNA for DRP1A was fused to 1.5 kb of the regulatory 5′-flanking region of the DRP1A gene. ST-mRFP was a gift from K. Shoda (RIKEN). Transgenic plants that expressed combinations of GFP, mRFP, and Venus fusion proteins were generated by cross-pollination, and F1 or F2 generations were used for microscopic observations. All plants were grown on 1× Murashige Skoog (MS; Murashige and Skoog, 1962) medium in plates at 23°C under continuous light.

Confocal laser scanning microscopy

Multi-color observations were carried out with an LSM710 confocal microscope (Carl Zeiss, http://www.zeiss.com/). For labeling with FM4-64, A. thaliana roots were soaked in 1/2× MS liquid medium that contained 4 μm FM4-64 at 23°C for 2 h. For drug treatments, BFA (50 μm; Sigma-Aldrich) and Wm (33 μm; Sigma-Aldrich, http://www.sigmaaldrich.com/) were applied for 1 and 2 h, respectively. The colocalization analysis was essentially performed as described previously (Bouttéet al., 2006). In brief, the distance from the center of a GFP signal to the center of the closest mRFP signal was measured with Metamorph software (Molecular Devices, http://www.moleculardevices.com/). Noise was removed by the Median Filter, and signals were enhanced with the Morphological Top Hat Filter. Then, images were converted into binary images and noise was removed again. The coordinates of the centers of objects were measured with Integrated Morphometry Analysis; these were then compared in two processed images (GFP and mRFP, or GFP and Venus; Appendix S1). Images were captured with an LSM710 confocal microscope with oil immersion lens (×63, NA = 1.40). We analyzed at least four images from at least three independent roots, which included between 622 and 1093 GFP points.

VIAFM

Roots of transgenic A. thaliana seedlings were placed on glass slides (76 × 26 mm; Matsunami, http://www.matsunami-glass.co.jp/), covered with a 0.12-0.17-mm-thick coverslip (24 × 60 mm; Matsunami), and root epidermal cells were observed under a fluorescence microscope (Nikon Eclipse TE2000-E and a CFI Apo TIRF × 100 H/1.49 numerical aperture objective, http://www.nikon.com/) equipped with a Nikon TIRF2 system. GFP and mRFP were simultaneously excited with 488 nm and 561 nm lasers, respectively. The fluorescence emission spectra were separated with a 565LP dichroic mirror and filtered through either a 515/30 (GFP) or 580LP (mRFP) filter with a Dual View filter system (Photometrics, http://www.photometrics.com/). All images were acquired with an Andor iXonEM EMCCD camera. The acquired images were analyzed with Photoshop CS4 extended, version 11.0.2 (Adobe Systems, http://www.adobe.com/) and Image J (National Institutes of Health, http://www.nih.gov/). The Pearson correlation coefficient was calculated with Image Pro Plus (Media Cybernetics, http://www.mediacy.com/).

Accession numbers for the sequence data

The Arabidopsis Genome Initiative locus identifiers for the genes mentioned in this article are At2g40060 (CLC2), At3g05710 (SYP43), At3g54840 (ARA6/RABF1), At5g45130 (RHA1/RABF2a), and At5g42080 (DRP1A/ADL1A).