The endoplasmic reticulum (ER) is a highly versatile membrane compartment that extends throughout the cytoplasm of eukaryotic cells. Probably the most important of its numerous functions is that it acts as the port of entry for newly synthesized proteins that are destined for distribution and transport among the various organelles constituting the secretory and endocytic pathways. This characteristic feature was first established more than 30 years ago in classic studies on the intracellular transport of secretory proteins in mammalian cells (Palade, 1975) and has been continually elaborated on ever since (amongst numerous recent reviews, for animal cells, see Klumperman, 2000; Lee et al., 2004; for plants, see Jürgens, 2004; Ward and Brandizzi, 2004).

It is universally accepted that ER-to-Golgi protein transport in mammalian cells is mediated by the sequential action of COPII- and COPI-coat protein complexes (Duden, 2003; Lee et al., 2004). This is because a pleiomorphic structure known alternatively as the ER–Golgi intermediate compartment (ERGIC) and vesicular tubular clusters (VTCs) transits along microtubules from the ER to the perinuclear-located Golgi apparatus with the help of a dynein/dynactin motor (Murshid and Presley, 2004). Characteristically, VTCs have COPI coats (Horstmann et al., 2002). However, there is general agreement that only the COPII machinery is responsible for the actual transport of cargo out of the ER (Barlowe, 1998, 2003).

In mammalian cells, export competent soluble and transmembrane cargo molecules collect at ER exit sites (ERES), which are defined by the presence of COPII-coat proteins, the ER-Golgi SNAREs Sed5, Bos1, Sec22, and Bet1, and several integral membrane proteins, including members of the p24 family and the Erv41/46 complex (Otte et al., 2001; Miller et al., 2002; Mossessova et al., 2003). The coat proteins are the GTPase Sar1p and two dimeric protein complexes: Sec23/24 and Sec13/31 at the cytosolic surface of the membrane (Antonny and Schekman, 2001). COPII-coat protein recruitment starts by the binding of Sar1p to the guanine nucleotide exchange factor (GEF) Sec12p, an integral ER membrane protein (Barlowe and Schekman, 1993), and is followed by the sequential attachment of Sec23/24 and then Sec13/31 dimers (Aridor and Balch, 2000). The Sec23/24 dimer has been implicated in the selection of cargo molecules into COPII vesicles (Bi et al., 2002; Miller et al., 2002) by interacting with diacidic (DXE) and diaromatic (FF) motifs in the cytoplasmic tails of transmembrane cargo molecules (Aridor et al., 2001; Otte and Barlowe, 2002).

In mammalian gland cells engaged in regulated secretion, ERES collect at specialized domains of the ER known as transitional ER (tER). Such domains are characterized by a high density of vesicle/tubule budding profiles in thin sections (for example, see Sesso et al., 1994; Bannykh et al., 1996; Ladinsky et al., 1999). The presence of COPII-coat proteins at these sites has been confirmed by immunogold labeling (Orci et al., 1991; Tang et al., 2000, 2001; Horstmann et al., 2002). tER is also often recognized in microorganisms. A clear example is that of the model alga Chlamydomonas reinhardtii, where tER and adjacent Golgi stacks are held in an ER amplexus attached to the nuclear envelope (Zhang and Robinson, 1986). Another well-known case is that of the fission yeast Pichia pastoris, which in contrast with Saccharomyces cerevisiae possesses a stacked Golgi apparatus (Mogelsvang et al., 2003). Pichia has several discrete tER domains each lying juxtaposed to a Golgi stack (Rossanese et al., 1999). The reason for such aggregations of COPII budding sites was thought to lie in the oligomerization status of Sec12p, which in P. pastoris has been shown to have large lumenal tails allowing for the interaction of adjacent molecules (Bevis et al., 2002). However, other scaffolding proteins, possibly Sec16p (Supek et al., 2002), now seem to be required for this event because COPII budding sites in P. pastoris still form when the localization of Sec12p to the tER is disrupted (Soderholm et al., 2004). In cultured cells and those mammalian cells exhibiting constitutive secretion, ERES are randomly located on the surface of the ER (Hammond and Glick, 2000; Stephens et al., 2000).

Like the yeasts, higher plant cells have a polydisperse Golgi apparatus and do not possess VTCs (Pavelka and Robinson, 2003). In addition, the Golgi apparatus moves along actin filaments that run parallel and close to the ER (Boevink et al., 1998; Ward and Brandizzi, 2004). Despite these clear morphological differences in the early secretory pathway, COPI/COPII vesiculating machineries appear to be quite conserved amongst the various eukaryotic cell types. Thus, COPI homologs can be found in the Arabidopsis database (Andreeva et al., 1998), and some have been identified in plant extracts (Movafeghi et al., 1999; Contreras et al., 2000, 2004). The presence of COPI proteins at the surface of vesicles budding from the periphery of Golgi cisternae has also been demonstrated by immunolabeling at both light (Ritzenthaler et al., 2002a) and electron microscopy levels (Pimpl et al., 2000). Many plant COPII homologs have also been detected and partially characterized (Sec12, Bar-Peled and Raikhel, 1997; Sar1 and Sec23, Movafeghi et al., 1999). Moreover, a functional Sar1 has been shown to be necessary for successful ER-to-Golgi transport in plant cells (Takeuchi et al., 2000; Phillipson et al., 2001).

tER in higher plant cells is poorly characterized, and vesiculation profiles at the ER in thin sections have only rarely been recorded in the literature (e.g., Craig and Staehelin, 1988; Staehelin, 1997; Ritzenthaler et al., 2002a), suggesting that ERES in this cell type are short lived and randomly distributed. To visualize ERES in tobacco (Nicotiana tabacum) BY-2 cells, we have employed two different approaches: (1) direct visualization of endogenous COPII proteins (Sar1, Sec13, and Sec23) by immunofluorescence microscopy in cell lines stably expressing ER- and Golgi-localized green fluorescent protein (GFP) markers and (2) visualization of ER-bound Sec13 by expression of a LeSec13:GFP construct in cells transiently expressing ER- and Golgi-localized red fluorescent protein (RFP) markers. In both cases, COPII is seen as punctate fluorescence over the surface of the ER. These point sources considerably outnumber Golgi stacks, although some are seen to associate with the rims of Golgi stacks. COPII labeling does not change or disappear with BFA, despite considerable morphological changes in the Golgi apparatus. Prevention of ER export through expression of a Sar1 mutant locked in the GDP state leads to disturbances in the ability to visualize COPII at the ER.

Figure 1. Cross-Reactivities of Antisera Raised against Recombinant Arabidopsis COPII Proteins.

We also tested the COPII antisera on fractions collected from a linear sucrose density gradient of Arabidopsis membranes to verify their association with the ER. As seen in Figure 1B, AtSec12, AtSec13, and AtSar1 have distribution profiles very similar to the ER marker calnexin and are clearly different to that of the Golgi marker reversibly glycosylated polypeptide. These results are consistent with our previous demonstration that the behavior of AtSec23-bearing membranes from cauliflower inflorescence in response to Mg²⁺ ions is typical for ER in sucrose gradients.

Figure 2. Confocal Immunofluorescence Images of BY-2 Cells Labeled with AtCOPII Antisera.

In BY-2 cells expressing the Golgi marker GmMan1:GFP, immunostaining with anti-AtSar1 gave rise to a homogeneous punctate image in cortical optical sections (Figures 2D to 2F). The density of this labeling was found to be 3.9, 4.4, and 3.5 point sources/μm² in Figures 2F, 2I, and 2L, respectively, with diameters of 540 ± 174 nm, 568 ± 230 nm, and 512 ± 166 nm (n > 100), respectively. By comparison, Golgi stacks with an average apparent diameter of 880 ± 164 nm (n = 77) had a density of only 1.5/μm² (Figure 2F). COPII antibody staining did not appear to be preferentially associated with cisternae of the cortical ER network (Figures 2G to 2L). At high magnifications, individual punctate COPII fluorescent signals were often found sitting directly on strands of ER (see arrows in Figure 2M). Using the same parameters for the estimations obtained from Figure 2F, the average diameter of these signals was 456 ± 18 nm (n = 43). Interestingly, and in agreement with immunostaining for COPII in mammalian cells (Rust et al., 2001), the colocalization (yellow) of green (GFP:HDEL) and red (anti-AtSar1) signals is restricted to narrow semicircular profiles on the ER. Punctate fluorescent signals of a similar size were also seen lying adjacent to the ER tubules. These could represent either ER-bound COPII out of the plane of section or individual released COPII vesicles, as suggested by Rust et al. (2001).

Figure 3. Expression of Sec13-GFP Is without Effect on Secretion.

Beginning ~24 h after exposure to dexamethasone, the BY-2 cells were expressing the Sec13:GFP fusion construct in sufficient quantities to allow for detection by confocal microscopy. In optical sections taken through the cortex (Figure 4A), a dense punctate image was obtained, similar to the antibody staining generated with COPII antisera (Figure 2). When cells expressing the GFP fusion construct were fixed and immunostained with AtSec13 antibodies, almost a perfect colocalization was obtained (Figures 4B to 4D), confirming the Sec13 identity of the protein carrying the GFP signal. During cytokinesis, the LeSec13:GFP signal in mitotic BY-2 cells is clearly seen to aggregate in and around the phragmoplast (Figure 4E), a region in which Golgi stacks are excluded (Figures 4H and 4I). Interestingly, when the LeSec13:GFP signal is compared with that of coexpressed BiP:DsRed (immunoglobulin binding protein cognate, a lumenal marker of the ER), there are discrete differences in the distribution of the two (Figures 4F and 4G). Although there is a clear colocalization of the two fluorescent signals in the phragmoplast, BiP:DsRed is absent from the reformed nuclear envelope. In contrast with LeSec13:GFP, BiP:DsRed is localized to the division plane, suggesting that the ER that gets trapped in the plasmodesmata of the cell plate lacks the capacity to bind COPII.

Figure 4. Visualization of COPII in BY-2 Cells Expressing AtSec13:GFP.

Median optical sections from BY-2 cells expressing LeSec13:GFP reveal, in addition to punctate fluorescence throughout the cytoplasm, an intense staining of the nuclear envelope and a diffuse staining of the nuclear matrix except for the nucleoli (Figure 4J). This, for us initially unexpected result, is fully in keeping with the known behavior of Sec13 in other eukaryotic cells. In yeast, it has been shown that, in addition to being a COPII-coat protein, Sec13p is also incorporated into the nuclear pore complex Nup84p (Siniossoglou et al., 2000). Moreover, in mammalian cells, Sec13 has recently been shown to shuttle between the cytosol and nuclear matrix (Enninga et al., 2003). As shown in Figure 5, we have performed on the LeSec13:GFP BY-2 cell line the same kind of fluorescence recovery after photobleaching (FRAP) experiments that were performed by the latter authors. Recovery of LeSec13:GFP at the nuclear rim was still not observed after 60 min when a region of the nuclear envelope that also included part of the ER, cytosol, and nuclear matrix was photobleached (Figures 5D and 5F). By contrast, the pool of cytosolic and ER-associated LeSec13:GFP recovered very rapidly (Figure 5C). When most of the area of the nucleus was photobleached, a near complete recovery of intranuclear LeSec13:GFP was detected after 10 min with a half-time of ~2 min (Figures 4G to 4J). This was significantly slower than the recovery of free GFP (half-time of 30 s). These results suggest a very rapid exchange of LeSec13:GFP molecules between the cytoplasm and the ER, a slower exchange between the nucleus and the cytoplasm, and a high stability of the nuclear pore complex-bound LeSec13:GFP. More importantly, they are almost identical to those obtained by Enninga et al. (2003), underlining yet again the credibility of the fluorescence signal we obtained with the LeSec13:GFP construct.

Figure 5. Analysis of LeSec13:GFP Dynamic by FRAP.

To gain further insights into the dynamics of the LeSec13:GFP fluorescence, LeSec13:GFP expressing cells were visualized by time-lapse microscopy (Figure 6). Because of the rapidity at which the fluorescent points appear and disappear, Nipkov disk–based confocal microscopy proved superior to conventional confocal laser scanning microscopy for this analysis. With this imaging technique we were able to continuously observe changes in cortical LeSec13:GFP for several minutes without any apparent photobleaching (Figure 6; see Supplemental Video 1 online). Taken separately, individual frames appeared very similar to the images obtained with fixed cells (cf. Figures 5 and 4A): numerous punctate fluorescent structures that averaged 430 ± 220 nm (n > 900) in diameter with a density of 5.0 punctate structures·μm⁻² were observed. A comparison between successive frames revealed that only very few point light sources could be followed for any significant length of time, making an estimation of their movement impossible. Thus, on successive frames (arrowheads and circles, Figure 6A), the point light sources appeared immobile. We therefore conclude that COPII binding has a very short half-life (at the most 10 s). This results in a blinking rather than moving appearance of LeSec13:GFP fluorescence.

Figure 6. Analysis of LeSec13:GFP Dynamic by Time-Lapse Microscopy.

Figure 7. Effects of Sar1[T39N] Secretory Inhibitors on LeSec13:GFP.

The macrocyclic lactone brefeldin A (BFA) is often used to block secretion (reviewed in Nebenführ et al., 2002). In addition to perturbing the Golgi apparatus, it has been suggested that BFA might also directly affect ER export (Brandizzi et al., 2002). This claim has recently found support in the observation that the distribution of punctate Sar1-YFP structures on the ER of tobacco epidermis is changed upon addition of BFA (daSilva et al., 2004). We therefore decided to investigate what effects BFA has on the COPII structures we have visualized in BY-2 cells by applying the drug to the LeSec13:GFP cell line expressing the cis-Golgi marker GmManI-RFP. Treatment with 10 μg·mL⁻¹ lead to a complete redistribution of the RFP-tagged protein into the ER but appeared to have no effect on the LeSec13:GFP punctate structures (Figures 8A to 8F). Similarly, BFA had no effect on COPII visualization when performed by immunostaining with AtSec13 antibodies (Figures 8G to 8I).

Figure 8. Effect of BFA on LeSec13:GFP.

Figure 9. Labeling of ER and Golgi in BY-2 Cells Expressing LeSec13:GFP.

To confirm the validity of GmMan1-RFP as a Golgi marker, we bombarded transgenic BY-2 cells expressing GmMan1:GFP with GmMan1-RFP. Again, a perfect colocalization with both Golgi markers was obtained (data not shown). The LeSec13:GFP cell line was then bombarded with GmMan1-RFP. Optical sections in the cortical region of a cell are presented in Figures 9G and 9H. As with the antibody staining data presented above, the LeSec13:GFP fluorescent points greatly outnumbered the Golgi stacks (Figure 9I). However, some of the Golgi stacks were densely surrounded at their periphery by LeSec13:GFP fluorescence, giving rise to a kind of halo (see the group of Golgi stacks in the middle of Figure 9I, and Figure 9K), whereas others were relatively free (Figure 9J).

Figure 10. Analysis of LeSec13:GFP in Relation to the Golgi Apparatus by Time-Lapse Microscopy.

Figure 11. Analysis of Golgi COPII Association in Relation to Golgi Speed.

Although the interaction between Sec12 and Sar1-GTP is pivotal to the formation of ERES, Sec12 is excluded from COPII vesicles induced in vitro (Barlowe et al., 1994; Barlowe, 2002). In experiments when Sec12-GFP is expressed transiently, it would appear that the COPII GEF is distributed uniformly throughout the ER and not concentrated at ERES as might be expected. This has been described for S. cerevisiae (Rossanese et al., 1999), for mammalian cells (Weissman et al., 2001), and most recently for tobacco epidermal (daSilva et al., 2004) and BY-2 cells (this article). However, such fluorescent patterns are not in accordance with the discrete punctate visualization of COPII coat components as seen here and reported on numerous occasions elsewhere (Shugrue et al., 1999; Hammond and Glick, 2000; Stephens et al., 2000; Rust et al., 2001). An explanation for this discrepancy is not immediately apparent.

In contrast with Sec12, which is an integral membrane protein, Sar1 and the other COPII-coat proteins cycle on and off the ER membrane. As a consequence, these proteins can be detected at the surface of the ER (at ERES), in the cytosol as individual coat proteins, and theoretically on released COPII vesicles. Because of their lower local concentrations and higher diffusion rates, fluorescently labeled cytosolic COPII proteins give a weak, uniform, and diffuse signal, as seen in the nuclear matrix for Sec13 (Figures 4J and 5H to 5J). A similar diffuse cytosolic fluorescence can be observed for COPI components, especially after their release from Golgi membranes upon BFA treatment (Ritzenthaler et al., 2002a). This type of fluorescence is different to the punctate fluorescence seen for all COPII components. Those punctae not on or immediately adjacent to the ER may well represent released COPII vesicles, as suggested by Rust et al. (2001).

A consistent observation in our investigation, and one which contradicts the data presented by daSilva et al. (2004), is that COPII binding sites that we have visualized at the surface of the ER greatly outnumber that of the Golgi stacks. Although Golgi stacks temporarily associate with COPII principally at their rims, it is not clear whether the visualization of COPII binding is sufficient to allow these sites to be defined as ERES. This can only be done by showing in vivo that cargo molecules (membrane or lumenal) collect and exit from the ER at these sites. For the moment, we can therefore only regard the punctate COPII sites as being putative ERES. If each were an ERES, this would mean that COPII vesicles are formed and released with extreme rapidity because our live cell imaging data suggest that COPII cycles on and off the ER within seconds. However, the continual formation and release of COPII vesicles does not appear to be very plausible because it is difficult to understand how mobile Golgi stacks could efficiently collect this released cargo. An alternative scenario is that the punctate COPII sites are potential ERES, but the completion of vesicle budding and release is only triggered upon arrival and docking of a Golgi stack. As a consequence, this would mean that a large portion of the COPII-coat proteins are involved in futile cycles of binding and dissociation. At the least, this would indicate that Sec13, and the other COPII-coat proteins for that matter, are not a limiting factor in the ER-to-Golgi transport in plants. Indeed, their overexpression, as demonstrated here and by daSilva et al. (2004), is without effect on secretion.

Randomly distributed ERES in mammalian cells are relatively immobile (displacement time of 5 to 15 μm·h⁻¹; Stephens, 2003) in comparison with the rate of cargo transport between the cortical ER and the perinuclear Golgi apparatus (0.5 to 1 μm·s⁻¹; Stephens et al., 2000). As previously mentioned, ERGIC/VTCs are responsible for this long-range transport, and these are generally considered to arise from the homotypic fusion of COPII vesicles (Stephens and Pepperkok, 2001; Duden, 2003). It would appear that each ERGIC/VTC is formed from a single ERES (Stephens et al., 2000). Upon completion of mitosis in mammalian cells, ERES form de novo (frequency: 2 h⁻¹·100 μm⁻²) and continue to do so during interphase and remain visible for several minutes (Stephens, 2003). During this time, COPII proteins continually cycle on and off the membrane but with different kinetics for each of the three major components (Sar1, Sec23/24, and Sec13/31; R. Forster, D. Stephens, and R. Pepperkok, unpublished data). According to Stephens (2003), ERES can also fuse with one another and divide. In common with mammalian cells, ERES in BY-2 cells are quite stationary, but they appear to be more dynamic structures than their mammalian counterparts: individual ERES were rarely visible for periods longer than a few seconds. Because of this property, it was not possible to ascertain whether plant ERES aggregate and/or divide. It could also be the reason why the visualization of COPII budding in plants by electron microscopy has been so elusive.

It is well known that during mitosis the Golgi apparatus in mammalian cells breaks down into vesicles (Shorter and Warren, 2002). It has been claimed that these vesicles, together with Golgi matrix proteins that are required as a scaffold for the reconstitution of the Golgi apparatus at the onset of the subsequent interphase, lie in close proximity to ERES whose function is arrested during mitosis (Prescott et al., 2001; Seemann et al., 2002). Immunostaining with COPII antisera has suggested that ERES were nonetheless still visible during mitosis (Prescott et al., 2001). However, the recently published data of Stephens (2003) indicates that this visualization is artifactual in nature: live cell imaging with YFP-Sec23 clearly showed a displacement of COPII into the cytosol during mitosis. In plants, the Golgi apparatus does not fragment during mitosis, and in BY-2 cells, many Golgi stacks appear to be immobilized in the immediate vicinity of the mitotic spindle (Nebenführ et al., 2000). However, during telophase, the plant Golgi apparatus is particularly active in secreting to the forming cell plate. As shown by Segui-Simarro et al. (2004), Golgi stacks enter the phragmoplast between the daughter nuclei during late telophase, where increasing amounts of ER are also to be found. Such a stage is comparable to that depicted in Figures 3F and 3G. Consistent with this are our results, obtained by live cell imaging with LeSec13:GFP, which demonstrate that ERES are visible within the phragmoplast and are presumably functionally intact.

ARF-GEFs have so far not been reported at the ER in any cell type, so claims that BFA can act at the level of ERES (Brandizzi et al., 2002; daSilva et al., 2004; Hawes and Brandizzi, 2004) require experimental substantiation. Our results showing that BFA has little effect on the ability to recognize COPII binding sites are in agreement with anti-Sec31 staining data obtained on NRK cells (Puri and Linstedt, 2003). However, as such these results say nothing about the export competence of the ERES so visualized. Ward et al. (2001) previously showed that COPII components still cycle at ERES after addition of BFA. More recent FRET measurements performed on Vero cells indicate that BFA interferes with the kinetics of the interaction between Sec23 and Sec31, whereas the interaction between Sar1 and Sec23 remained unaltered (R. Forster, D. Stephens, and R. Pepperkok, unpublished data). However, treatment with BFA for short periods, during which time COPI assembly was inhibited, did not alter the steady state distribution of any COPII component. Thus, it seems likely that any effect of BFA on ER export is an indirect one resulting from a perturbation in the fine tuning of the interdependent COPI and COPII machineries (Stephens et al., 2000; Ward et al., 2001), upon whose maintenance successful ER–Golgi transport depends.

Support for this latter model has recently been presented by daSilva et al. (2004), who have investigated the distribution of AtSar1-YFP, AtERD2-GFP, and the Golgi marker ST-GFP by transient expression in tobacco epidermis. According to these authors, these two constructs were constantly located together into “distinct but overlapping structures,” and that this tandem structure was mobile in an actin-dependent manner. A degree of permanency for this structure was suggested by selective photobleaching of the YFP signal that inevitably recovered in the immediate vicinity of the GFP signal. In complete contradiction with the secretory unit concept, our results show that several ERES can attach to a single Golgi stack at any one time and that ERES are not constantly associated with the Golgi apparatus, assuming of course that each punctate COPII labeling represents one ERES (see above). Moreover, Golgi-ERES associations are not permanent but are continually changing in number and position at the rims of the stack as the Golgi moves. Another important distinction between our data and that of daSilva et al. (2004) relates to the relative apparent sizes of the Golgi and ERES images. In the latter article, these two structures appear to be of similar size, but in our work, ERES appear significantly smaller. Thus, our observations are more in keeping with a kiss-and-run model for ER-to-Golgi transport.

A feature common to the vacuum cleaner and stop-and-go models is that ERES outnumber Golgi stacks and are relatively stationary. This is supported by the data presented here. However, in the sense that ERES are seen at the rims of both stationary and moving Golgi stacks in BY-2 cells, our data do not exactly conform with the latter model. Indeed, we have demonstrated that a stationary Golgi stack can be visualized over a 20-s period both with and without associated peripheral ERES. Thus, Golgi motility per se does not seem to be a precondition for successful ER-to-Golgi transport, and this is in agreement with FRAP measurements dealing with the recovery of photobleached Golgi marker proteins on immobilized (Brandizzi et al., 2002) and moving (Brandizzi and Hawes, 2004) Golgi stacks.

Mobility is a crucial feature of the mobile secretory unit model, even though it remains unclear as to how a Golgi stack and its associated ERES remain together during movement. However, this is important because if true it would mean that a Golgi stack would have to drag its ERES through a lipid membrane. In our opinion, this can only be achieved through direct physical continuities (tubular ER-Golgi connections) or through the existence of some kind of scaffolding elements linking the two together. The former possibility has previously been suggested (Brandizzi and Hawes, 2004; Hawes and Brandizzi, 2004), but always in neglect of retrograde COPI vesicle transport, which does seem to exist in plants (Pimpl et al., 2000; Ritzenthaler et al., 2002a). Evidence for the latter is lacking. With regard to Golgi motility we would like, in addition, to point out some inconsistencies in the data published using the tobacco leaf epidermis and BY-2 cell systems. According to Boevink et al. (1998), Golgi stacks in epidermal cells move with speeds of up to 0.76 μm·s⁻¹ along stationary cortical ER and in excess of 2.2 μm·s⁻¹ within transvacuolar cytoplasmic strands. In the article by daSilva et al. (2004), the speed of Golgi stacks lies between 0.1 and 0.3 μm·s⁻¹ (calculated from the values given in Figure 9 of that article). These latter values contrast with those determined for Golgi stacks in BY-2 cells (~3 μm·s⁻¹) by Nebenführ et al. (1999) and confirmed here. In fact, such low velocities are in the range of the almost stationary wiggling motion described by Nebenführ et al. (1999) for BY-2 cells. Because our data suggest that the degree of Golgi-COPII association increases the slower the Golgi moves, which is understandable, we would therefore plead for more caution in the interpretation of the temporal aspect of ERES-Golgi associations.

The differences in the depiction of the ERES and Golgi stacks as given in our article and that of daSilva et al. (2004) may well lie in the relative secretory status of the cell types employed in the two studies. Tobacco BY-2 cells represent a rapidly growing and dividing cell system with a high rate of secretion. Tobacco leaf epidermal cells, on the other hand, hardly grow, do not divide, and will obviously be secreting at a much lower level. Thus, one might consider the leaf epidermis to represent a kind of minimal system, with membrane trafficking to the cell surface and within the endomembrane system being reduced to a housekeeping role. In this situation, ER export will not be comparable to that in BY-2 cells: the number of ERES may well be reduced to a level where their number approximates that of the Golgi stacks. This being the case, it is probably more efficient to have a Golgi stack hovering in the vicinity of a more or less stationary ERES than to be rapidly wandering across the surface of an ER with few exit sites. By contrast, a situation where ERES greatly outnumber Golgi stacks would have advantages for a rapidly secreting system, such as BY-2 cells, because it would allow more material to be sorted at the ER and secreted per unit of time. It also would be more robust toward distortions because missing a few events would not matter, whereas the secretory unit model would require much more stringent regulatory mechanisms.

Transformation of BY-2 cells was done by coculturing a 3-d-old BY-2 culture with a 36-h-old LBA4404 culture. The coculture was incubated in the dark without shaking for 2 d at 25°C. The cells were pelleted and washed three times in fresh medium containing 250 μg/mL of carbenicilin. After a final wash, the suspension was poured on the solid medium containing carbenicilin and hygromycin. Positive calli (obtained after several weeks) were screened by fluorescence microscopy to select the best cell lines. Expression of LeSec13:GFP in the BY-2 cells was initiated by the addition of 10 μM dexamethasone (Sigma-Aldrich). After 24 to 30 h of incubation, samples were removed for microscopy.

Sequence data from this article have been deposited with the EMBL/GenBank data libraries under accession numbers A1776423 and BAB09140.

We gratefully acknowledge the financial support of the Deutsche Forschungsgemeinschaft. We thank Masaki Takeuchi and Jürgen Denecke for providing us with the GDP- and GTP-fixed Sar1 mutants. We also thank Inhwan Hwang for giving us the BiP clone and Chris Hawes for the Sec12-YFP construct. The technical assistance of Sara Duval is gratefully acknowledged. We also appreciate having had useful discussions at various stages during this investigation with Jürgen Denecke, Andreas Nebenführ, and Christiane Stussi-Garaud. The Inter-Institute Zeiss LSM510 confocal microscopy platform was cofinanced by the Centre National de la Recherche Scientifique, the Université Louis Pasteur, the Région Alsace, the Association de la Recherche sur le Cancer, and the Ligue Nationale contre le Cancer.