Vesicles versus Tubes: Is Endoplasmic Reticulum-Golgi Transport in Plants Fundamentally Different from Other Eukaryotes?

What Can EM Tell Us?

Back in the 1970s and 1980s, prior to the discovery of COPI and COII vesicles, various electron microscope (EM) studies of the plant endomembrane system suggested that there may be direct membrane connections between the ER and Golgi bodies, which were often termed dicytosomes in those days (Mollenhauer at al., 1976; Harris, 1979). To enable selective enhancement of the membranes of these organelles and to permit studies in three dimensions using thick sections and high-voltage electron microscopy, osmium impregnation techniques combined with either zinc iodide or potassium ferricyanide, to enhance the deposition of osmium on membranes and within the lumen of EM tubules and ER/Golgi cisternae, were often applied (Harris, 1979; Hepler, 1981). This resulted in a number of reports suggesting that the ER was directly connected to Golgi bodies via membrane-bounded tubules and that Golgi bodies themselves presented many tubules at the margins of their cisternae.

In our laboratory, we undertook a limited survey of ER-Golgi interactions in a number of tissues using the zinc iodide osmium impregnation technique (Juniper et al., 1982) and concluded that in maize (Zea mays) root caps, bean (Phaseolus vulgaris) root tips, and leaves (probably erroneously), the ER and Golgi were not connected. However, in the heavily protein-secreting glands of the Venus flytrap (Dionea muscipula), after stimulation of secretion, we observed numerous tubular extensions from cisternal rims, including cis-cisternae, that appeared to interconnect with the ER. At the same time, a study of the Golgi apparatus in developing wheat (Triticum aestivum) endosperm (Parker and Hawes, 1982), using high-voltage electron microscopy of thick sections of osmium-impregnated tissue, reported fine peripheral cisternal tubules connecting the cis-Golgi as well as other cisternae to the ER. This was also reported in developing bean and mung bean (Vigna radiata) seeds (Harris, 1979; Harris and Oparka, 1983). In wheat, these fine tubules were often much thinner than an ER tubule and, unless heavily osmium impregnated, under normal contrasting conditions in the EM, would unlikely scatter sufficient electrons to be readily visible. More recently, we reported ER-Golgi connections in tobacco leaves using osmium impregnation (Brandizzi et al., 2002b). Of course, the presence of tubules in electron micrographs does not prove they are involved in transport, retrograde, or anterograde but does show perhaps a more intimate relationship between the two organelles than is often assumed. Likewise, it has been shown that the ER surface itself is highly motile (Runions et al., 2006). Therefore, conceptually, there should be no problem in envisaging membrane flow from the ER through to the Golgi.

In contrast to these results, there are relatively few reports of vesicles budding from the ER or existing at the interface between the ER and Golgi, although Schnepf and Christ (1980) suggested that a vesicle flow exits in the epithelial cells of the nectaries of Asclepias curassavica. Only in a few publications using ultra-rapid high-pressure freeze fixation followed by freeze substitution and electron tomography have COPII vesicles been shown budding from the ER (Kang and Staehelin, 2008; Hawes, 2012; Donohoe et al., 2013), mainly in root meristem cells. An example in developing Arabidopsis endosperm tissue can be seen in Figure 1, C to G, of this article. Unfortunately, three-dimensional analysis and reconstruction by electron tomography of data such as these still relies on manual tracing of images, as autosegmentation algorithms cannot as yet differentiate the subtle differences in contrast in such electron micrographs to permit totally unbiased autosegmented reconstructions. However, it is possible to observe tubular connections between ER and cis-Golgi in tomograms of osmium-impregnated root material (C. Hawes, unpublished data; Fig. 3A). At this stage, it should be noted that lack of COPII vesicles is not restricted to plants but has been reported in several microorganisms. Likewise, depletion of COPII components does not always inhibit cargo transport from the ER in yeasts and animals (for references, see Mironov, 2014).

• Download figure
• Open in new tab
• Download powerpoint

Figure 3.

A, Maximum-intensity projection in negative contrast of a stack of thin sections from a tomogram of a pea (Pisum sativum) root tip Golgi body and associated ER impregnated by the osmium zinc iodide technique. The reconstruction is presented at an angle to show a clear tubular connection between the ER and cis-Golgi. B, Inside face view of a dry-cleaved carrot (Daucus carota) suspension culture cell. The cell had been fixed on a coated EM grid, dehydrated, and critical point dried prior to dry cleaving on double-sided tape. The view onto the plasma membrane shows dark mitochondria (M), complete Golgi stacks in face view (G), cisternal ER (CER), and tubular ER (arrows). Note the huge difference between the diameter of a Golgi body and ER tubules.

An argument has been made on the lack of COPII EM images based around the calculation that, in reality, there are very few COPII vesicles at any one time in the ER-Golgi interface and that, in any one thin section, at most, only one vesicle would be seen (Langhans et al., 2012). This argument of course only holds true if such vesicles do exist. If they don’t, then they obviously would not be seen. Likewise, if tubules are transient in nature, they would rarely be seen in conventional electron micrographs and, when caught in cross section, would appear to be vesicular! An obvious experiment where COPII vesicles should be seen is in the reformation of Golgi after brefeldin A (BFA)-induced reabsorption into the ER. One such study on tobacco Bright Yellow2 (BY2) cells reported buds on the ER surfaces, which were infrequent, and tubular vesicular clusters, representing the earliest observable stage of stack regeneration cells (Langhans et al., 2007). These clusters immunolabeled for COPI coat components but not for COPII proteins. Considering the number of Golgi stacks in such cells, which is in the hundreds, if COPII vesicles exist, then it is surprising that none were seen in these experiments.

Of course, the get-out-of-jail card and perhaps a lazy answer to all of these discrepancies in the ultrastructural literature is simply to state that data from chemically fixed material is artifactual in nature due to slow fixation rates and only data from ultra-rapidly frozen freeze-substituted material is acceptable. This of course ignores the fact that, in chemically fixed material, it is easy to visualize both COPI and clathrin coats. So, why not COPII coats, especially as they are relatively easy to visualize in chemically fixed cells of algae such as C. noctigama (Hummel et al., 2007)? Two other golden rules of thin-section transmission electron microscopy also have to be remembered: (1) A thin section presents a two-dimensional image, and thus a tubule in cross section can easily be misinterpreted as a vesicle; and (2) Any biological material has to scatter sufficient electrons to form an image. Thus, a membrane in transverse section, spanning 70 nm of resin, scatters sufficient electrons to form a classic unit-membrane image, whereas the same stained membrane in face view may not present sufficient heavy-metal stain molecules and thus be electron lucent and not form an image; thus, fine tubules and membranes in face view can be missed. Selective-membrane staining techniques overcome this latter limitation. Of course, other EM techniques exist such as freeze-fracture or freeze-fracture deep etch, which should reveal structured exit sites on ER and COPII coats, but as far as I am aware, apart from the occasional image showing clathrin-coated vesicles and COPI vesicles (Coleman et al., 1987; Andreeva et al., 1998, no such images of COPII structures have been published in plants.

Has Live-Cell Imaging Helped?

Our initial observations on Golgi and ER in living leaf epidermal cells let us observe, for the first time, the dynamic nature of the organelles and the fact that Golgi bodies in leaves appeared to move over the surface of the ER (Boevink et al., 1998). This led us to propose the hoover model of Golgi bodies traveling over the ER surface sucking up vesicles produced by the ER, thus making the serious, but all too common, mistake of assuming that the plant ER-Golgi interface would function exactly the same as the mammalian ER in the production of COPII vesicles. However, over the past decade or so, we have refined our ideas and developed the secretory unit concept of ERESs and Golgi bodies traveling as single units around the cell with the motile surface of the ER (daSilva et al., 2004; Langhans et al., 2012). Such advances were made possible by the revolution in live-cell imaging offered by fluorescent protein technology and direct organelle labeling combined with techniques such as photobleaching and photoactivation of fluorescent probes. This enabled a range of experiments to be undertaken on the ER-Golgi interface, and contrary to what is often stated, it was shown there is no real evidence that transport between the ER and individual Golgi bodies only takes places when stacks are stationary (the stop-and-go model; Brandizzi et al., 2002b). We demonstrated that in a fluorescence recovery after photobleaching experiment, recovery of fluorescence of a Golgi membrane protein could be demonstrated in moving Golgi, indicating a continual transfer of protein from ER to Golgi (daSilva et al., 2004). Subsequently, laser manipulation of Golgi has demonstrated that when captured and translated through the cytoplasm by an infrared laser beam, Golgi bodies almost always drag a tubule of ER behind them (Sparkes et al., 2009c). One of the conclusions from this work was that the ER and Golgi are closely associated and tethered together, not via the cytoskeleton, but most likely by a number of structural proteins such as the Golgins/Golgi matrix proteins (Hawes et al., 2008). In studies of Golgi destruction and reformation in tobacco leaf cells, we showed that with either a genetic block of secretion or inhibition by BFA, reabsorption of the Golgi back into the ER was an ordered process, starting with the release of translocated Golgins into the cytoplasm followed by sequential reabsorption of trans-, medial, and cis-membranes into the ER. However, a cis-located matrix component, CASP, remained associated with ERESs, and we suggest that this protein may be part of the tethering complex holding the Golgi to the ER (Osterrieder et al., 2010; Schoberer et al., 2010). How this ordered membrane transport back to the ER is mediated is not known; perhaps some of the cisternal associated Golgi-ER tubules are involved in this pathway.

Interestingly, Golgi bodies in vacuolated tissue such as leaf epidermal cells are only ever associated with curved membranes of the ER such as tubules or more rarely the edges of cisternae (Sparkes et al., 2009b) and are never seen on the flat faces of ER cisternae. This situation has also been reported for yeast, where ERESs were associated with high-curvature domains containing the membrane-curving ER protein reticulon1 (Okamoto et al., 2012), and we have preliminary evidence that in Arabidopsis, reticulons may interact with SEC12, the Sar1-guanine nucleotide exchange factor that recruits the GTPase to the ER membrane as part of the COPII coat-building process (Kriechbaumer and Hawes, unpublished data). Thus, we can conclude that plant ERESprobably requires a curved ER surface on which to form.

Due to the nature of fluorescence and the diffraction limit of the microscopies used, an artificial impression is given in typical confocal micrographs of the diameter of ER tubules, compared with the diameter of a Golgi body (around 1 µm), giving a ratio of approximately 1:2. However, when observed by EM techniques such as dry cleave to show the cortical ER, whose tubes are in reality are around 30 to 90 nm, and associated Golgi, it is obvious that this ratio is more like 1:8, even taking into account the fact that cis-Golgi cisternae tend to be smaller in diameter than the rest of the stack (Fig. 3B). Thus, ERESs have to form on a relatively restricted area of highly curved membrane and not a flat surface, and perhaps the requirement for membrane bending proteins to produce a bud is lessened. Also, an ERES would need to be linear in structure along roughly 200 nm of ER tubule. Could such a structure produce sufficient 70-nm vesicles to transfer the required protein and membrane to a Golgi body? Such calculations have not yet been made.

Of course, it is generally accepted that COPII components are required in plants for some, if not all, transport between the ER and Golgi. As described above, it is easily demonstrated that inhibition of the formation of a COPII coat by expressing a nonhydrolysable form of the coat-initiating GTPase SAR1p results in the disruption of Golgi stack homeostasis and resorption of Golgi membrane back into the ER (Osterrieder et al., 2010). However, coating of a membrane patch to promote curvature and concentrate cargo at the tip of a transiently produced tubule is a distinct possibility. It has been shown in vitro that COPII components can tubulate liposome membrane (Bacia et al., 2011), and it has even been suggested that COPII coat components have sufficient flexibility to form 300-nm tubular structures that can accommodate large filamentous cargo such as procollagen fibrils (Miller and Schekman, 2013). Such tubes could easily span the narrow greater-than-300-nm interface between ER and cis-Golgi in plants.

The Mobile Secretory Unit: What Are the Consequences for Bidirectional ER-Golgi Traffic?

In the frequently used leaf epidermal system, it is well established that COPII-fluorescence labeling always colocalizes with fluorescent Golgi markers, whether the Golgi stacks are mobile or not (daSilva et al., 2004; Hanton et al., 2009). For most researchers, punctate fluorescent COPII signals on the surface of the ER are synonymous with ERESs, but this may not necessarily be so because the visualization of COPII binding does not actually reveal the actual exit event. Langhans et al. (2012) have also questioned the fidelity of COPII-fluorescence labeling in recording ERESs on the surface of the ER and have suggested instead that the signals may instead represent prefusion COPII vesicles lying in the interface between ERESs and the cis-Golgi. Despite these caveats, it seems that anterograde traffic from the ER is restricted to a domain of the ER immediately adjacent to a Golgi stack and is embodied in the concept of the secretory unit (daSilva et al., 2004; Hanton et al., 2009). It now appears that retrograde traffic between the Golgi and the ER is also spatially restricted, because it has been demonstrated that the SNAREs on target membrane (t-SNAREs) required for COPI vesicle fusion with the ER localize to ER domains immediately underneath Golgi stacks (Lerich et al. 2012). This unique feature of the secretory pathway in plants probably serves the purpose of control and regulation, because the problem of stochastic release of COPII vesicles into the cytosol and their subsequent capture by Golgi stacks is avoided.

The concept of the secretory unit received great support from laser-trapping technology, which demonstrated that ER tubules moved with individual Golgi stacks when the latter were displaced (Sparkes et al., 2009a). This key observation could be interpreted as proof of direct membrane continuities between the ER and the cis-Golgi, but in the opinion of the majority, it is a consequence of the existence of tethering/matrix proteins that not only anchor the Golgi stack to the ER surface (Latijnhouwers et al., 2005; Osterrieder, 2012), but also maintain the integrity of the Golgi stack (Ito et al., 2014). An important feature of these experiments is that the Golgi stacks were immobilized by actin inhibitors, obviously a prerequisite for capturing otherwise mobile Golgi stacks. In this situation, interlocking matrix protein interactions between the ER and the Golgi are probably comparable to the real-life situation of a stop period in Golgi travel. However, it remains unclear as to whether the actual exit of anterograde cargo from the ER (i.e. vesicle budding [or COPII tube formation]) is restricted to the stationary phase or occurs continuously during Golgi movement. We also do not know whether antero- and retrograde transport are separate or coordinated, synchronized events. Lerich et al. (2012) have suggested that prefusion clusters of COPII and COPI vesicles might accompany mobile Golgi stacks, but the colocalization of Golgi stacks with the SNARE of the Qc-type syntaxin of plants72 (SYP72) of the t-SNARE fusion complex only in the immobilized condition suggests that entry of COPI retrograde cargo into the ER is restricted to stationary Golgi stacks.

Why Are COPII Vesicles Difficult to Visualize in Higher Plants?

COPI vesicles have been detected at the periphery of Golgi cisternae in higher plants (Pimpl et al., 2000; Donohoe et al., 2007). Although their positive identification as retrograde carriers to the ER in plants awaits confirmation, it is very likely that they do fulfill this function. This being so, there seems to be no a priori reason for questioning the participation of COPII vesicles in anterograde ER-Golgi traffic, yet their visualization in higher plant cells has proved difficult. In mammalian cells, there is a transport intermediate between the ER and the perinuclear immobile Golgi complex known as the ER-Golgi intermediate compartment (ERGIC; Appenzeller-Herzog and Hauri, 2006). It is this structure rather than the Golgi apparatus that is engaged in bidirectional COPII/COPI-mediated trafficking. During its formation (apparently through homotypic COPII vesicle fusion) and before it begins moving in the direction of the Golgi complex, it lies close to the surface of the ER (200–500 nm distant) in the immediate vicinity of ER export sites. Nevertheless, free COPII-coated carriers/vesicles have been reported on several occasions in the ER/ERGIC interface (Zeuschner et al., 2006; Hughes et al., 2009; Witte et al., 2011). In support of these observations, it has been recently demonstrated that loss of function of a cytosolic protein complex (transmembrane tyrosine receptor kinase-fused gene), which forms aggregates that temporarily trap COPII carriers in the ER/ERGIC interface, causes free COPII carriers to accumulate throughout the cytoplasm (Johnson et al., 2015).

A similarly narrow interface exists between ER and the cis-Golgi (the probable ERGIC equivalent in plants), so why the problem in seeing COPII vesicles in thin sections of higher plant cells? Langhans et al. (2012) have attributed this to the relatively small numbers of transport vesicles (at the very most 20) in the interface at any one time, which extrapolates to only a single vesicle in a thin section. There may be other contributing factors, for example, the speed of vesicle transport and, obviously if bidirectional transport is restricted to the stationary phase of Golgi movement, the timing of transport. If so, the chances of visualizing transport vesicles in the ER-Golgi interface will be seriously affected by the mobile status of the Golgi stack at the moment of fixation.

Have Gap-Spanning Tubules Been Observed in the Interface between the ER and cis-Golgi?

This can be answered with a categorical no, despite opposing claims made in some recent reviews (Sparkes et al., 2009a; Stefano et al., 2014). A careful scrutiny of the electron micrographs in the papers cited in these reviews (for details, see “Chris Hawes: Let It Be Tubes”) in support of direct membrane continuity at the ER/Golgi interface reveals that the connections are not at the interface but are, in fact, lateral connections between undefined but probably median Golgi cisternae and the ER. Moreover, the frequency of such continuities is enhanced under nonphysiological conditions, e.g. cold (Mollenhauer at al., 1976) or BFA (Ritzenthaler et al., 2002) treatments. The physiological significance of such lateral connections remains obscure, especially because it is not in harmony with the glycoprotein processing reactions that are supposed to occur in a sequential manner through the Golgi stack (from cis to trans). Also, lateral connections of this type are difficult to reconcile with Golgi stacks gliding over the surface of the ER. Finally, if tubular connections between the ER/nuclear envelope and the cis-face of a Golgi stack do exist, one would expect that the chances of visualizing them would be greater in those cases where the Golgi is immobile. However, such structures have never been seen in algal cells with naturally stationary Golgi stacks nor in higher plant cells where the Golgi has been immobilized through actin inhibitors.

What Are the Advantages of Vesicles?

Vesicles allow organelles to communicate among themselves and with the cell exterior (via the plasma membrane). With such transport vectors, cargo molecules can be transferred between organelles without disturbing their integrity. By excluding certain molecules and including others, vesicles also allow for a high degree of selectivity in intracellular transport. The efficient sorting of proteins into vesicles is to a great extent related to coat proteins at their surface that interact with several different types of integral transmembrane proteins: cargo receptors, helper proteins, such as the p24 proteins, and SNAREs. All of the coat proteins (COPI, COPII, and clathrin) can polymerize to form spherical structures (cages) and thus have the inherent ability to form vesicles, although it is true that COPII does it differently to COPI and clathrin (Hughson, 2010). However, the special properties of the COPII coat allow the incorporation of cargos of different sizes, something that cannot be achieved with clathrin or COPI vesicles. Therefore, COPII vesicles of different sizes can accommodate export of different cargo from the ER (Miller and Schekman, 2013) without the need for direct tubular connections. Finally, by concentrating SNAREs into a small amount of membrane, the efficiency of vesicle fusion with a specific target compartment is increased. The inhibition of vesicle formation can lead to uncontrolled fusion of organelles mediated by SNAREs on vesicle membrane and t-SNAREs as seen for the Golgi-ER hybrid structures formed in the presence of BFA (Elazar et al., 1994; Ritzenthaler et al., 2002).

By contrast, tubular contacts (irrespective of their longevity), or direct contacts between organelles culminating in fusion and then fission (as, for example, in hug-and-kiss models; Kurokawa et al., 2014), have the inherent caveat that an unspecific mixing of organelle contents may occur. Another problem is that if there is direct contact between the lumina of two adjacent compartments, how can pH differences be maintained? This is particularly important in the case of the endoplasmic reticulum protein retention (ERD2) receptor recognizing the tetrapeptide motif lys-asp-glu-leu (KDEL) receptor, which, in mammalian cells, is thought to bind to its KDEL-ligands at an acid pH (pH 6.7) in the cis-Golgi and to release them at the higher pH (pH 7.2) of the ER lumen (Wilson et al., 1993; Majoul et al., 1998; Paroutis et al., 2004). As is the case with the mannosyl-6 P receptor in mammalian endosomes, apparently only a relatively small shift in pH is sufficient to cause ligand release. A similar pH gradient of about 0.5 pH units between the ER and the Golgi also exists in plants (Martiniére et al., 2013), and it has recently been shown that the binding of COPII proteins to a plant ERD2 homolog is favored at neutral pH conditions, whereas the recruitment of COPI proteins to ERD2 is more optimal at an acidic pH (Montezinos et al., 2014).

Bidirectional protein trafficking between the ER and the Golgi apparatus also entails the movement of membrane. In the anterograde direction, this serves the purpose of replenishing membrane lost at the trans-face of the Golgi stack through release of the trans-Golgi network (Viotti et al., 2010) and continually drives the process termed cisternal maturation. In the retrograde direction, it is a consequence of receptor recycling. While vesicles clearly fulfill this requirement, it is less easy to see how tubes can achieve it. If ERES and cis-Golgi cisternae briefly touch each other, fuse, and rapidly separate, there can be no net movement of membrane at all. This is a stringent interpretation of the hug-and-kiss model of Kurokawa et al. (2014). However, if there is an active uptake of membrane (a patch of previously COPII-coated ERESs) together with soluble cargo as the cis-Golgi docks onto ERES, the model should perhaps be renamed hug and bite.

As new data becomes available, researchers are often compelled to revise their standpoints on particular issues. My initial interpretation of ER-Golgi traffic, based on immunofluorescent studies in tobacco BY2 cells (Yang et al., 2005), led me to believe that ERESs were greatly in excess of Golgi stacks. Switching to leaf epidermal cells and perhaps more stringent localization criteria via transient expression of X-fluorescent protein tagged, convinced me of the validity of the secretory unit concept. Nevertheless, I was never in doubt about vesicles, and I maintain that the unique features of the secretory unit, i.e. mobile Golgi stacks with a narrow interface to the ER, are not necessarily an impediment to COP vesicles as mediators for bidirectional protein traffic between the ER and the Golgi apparatus in higher plants. By contrast, direct membrane continuities between the ER and the Golgi have physiological drawbacks, and there is no convincing evidence for their existence, even in organisms where Golgi stacks remain permanently stationary.