High-pressure freezing provides insights into Weibel-Palade body biogenesis

Weibel-Palade bodies (WPBs), first described 40 years ago (Weibel and Palade, 1964), are among the most striking objects to be seen in cells. As long cigar-shaped granules, these lysosome-related organelles of endothelial cells play a central role in the initiation of inflammation and in haemostasis. Apart from their overall shape, their hallmark has been the presence of internal striations running parallel to their long axis that are revealed as tubules in transverse section (Weibel and Palade, 1964). These tubules are von Willebrand Factor (VWF), a key haemostatic mediator, which is stored within the organelle (Wagner et al., 1982). We have demonstrated that the storage of VWF as tubules is important for this protein's ability to trap platelets (Michaux et al., 2006). Only when the tubules are intact can the recently described long strings of platelet-catching VWF (Dong et al., 2002) unfurl efficiently from the endothelial cells upon exocytosis. Deliberate disruption of the VWF tubules not only reduces platelet recruitment, but also causes rounding of the organelles (Michaux et al., 2006). We have therefore defined the VWF tubule as key to both WPB ultrastructure and VWF function.

Despite their importance in driving the formation of WPB shape, the relationship between tubule formation and WPB biogenesis is poorly understood. A great deal is known about the complex biosynthesis of VWF, including its cleavage into a pro-peptide plus mature protein by furin, and the need for interactions between these two components for tubule formation to occur (reviewed in Hannah et al., 2002; Michaux and Cutler, 2004; van Mourik et al., 2002); however, even simple questions, such as exactly where along the secretory pathway the tubules start to drive the formation of this uniquely-shaped organelle, remain unanswered.

A second area of interest concerns the role of clathrin coats in WPB formation. Although we have recently proposed a novel scaffolding role for an AP-1/clathrin coat in initial WPB formation (Lui-Roberts et al., 2005), the relationship between the AP-1/clathrin coat and the formation of tubules themselves has yet to be investigated. This has become crucial since we identified tubule formation as the key to WPB formation. In addition, clathrin-coated vesicles are used for removing missorted material from the post-trans-Golgi network (TGN) but are still immature in endocrine and neuroendocrine cells (Dittie et al., 1999; Klumperman et al., 1998). It is thought that control of entry into forming granules at the TGN ('sorting by entry') is not completely effective at excluding proteins destined for other post-TGN destinations, such as the endosomal system or constitutive exocytosis. The post-budding sorting step ('sorting by retention') allows the removal of such missorted proteins (reviewed in Arvan and Castle, 1998; Borgonovo et al., 2006). There is no evidence that clathrin-mediated removal of missorted proteins occurs from immature WPBs during their biogenesis. If it does not occur, then that would imply that sorting-for-entry is more important than sorting-by-retention in WPB formation.

A third area where the process of WPB biogenesis is unclear is in changes to the cargo and to overall granule morphology after budding from the TGN. This is in contrast to dense-core secretory granule (DCG) maturation, where an increase in density is accompanied by membrane remodelling and even crystallisation of internal cargo (Eaton et al., 2000; Greider et al., 1969; Kuliawat and Arvan, 1992). Whereas recruitment of Rab27a (Hannah et al., 2003) and CD63 (Vischer and Wagner, 1993) after WPB budding from the TGN are clearly maturation-specific events during WPB biogenesis (Harrison-Lavoie et al., 2006), corresponding structural change in WPB content has not been established.

Finally, since the WPB can be 1-5 μm long (Weibel and Palade, 1964) with a highly organised internal content, one further intriguing question is: how do these enormous organelles manoeuvre within the crowded environment of cells that average no more than 20 μm in diameter and are almost flat?

To address these questions, it was necessary to analyse WPB formation in human umbilical vein endothelial cells (HUVECs) by transmission electron microscopy (EM). By using high-pressure freezing (HPF) and freeze substitution (Reipert et al., 2004; Studer et al., 2001) rather than traditional chemical fixation in order to preserve the organelles in as near to a living state as possible, we have begun to answer many of the questions raised above.

VWF tubules are essential to the function of WPBs in haemostasis and their intra-organellar distribution, running parallel to the long axis of the organelle, immediately suggests a close link between tubules and organelle shape. We therefore used VWF tubules as the major criterion in identifying WPBs throughout this study. We have investigated the relationships between VWF tubules and the initial formation of WPBs in HUVECs.

We began a morphological analysis of the ultrastructure of forming WPBs using EM following conventional chemical fixation. We focused on structures connected to or closely apposed to the Golgi, especially those containing tubules and those that are clathrin coated, since these features are characteristic of immature WPBs (Lui-Roberts et al., 2005; Weibel and Palade, 1964). Although it is possible to find potential WPB precursors, we are rarely able to see any connections between them and the Golgi.

Fig. 1A shows two swollen cisternal ends reminiscent of condensing vacuoles that could represent the earliest stage of WPB biogenesis. Their possible connections to the Golgi stack are interrupted, perhaps where they have not survived fixation. However, although they contain electron dense material, it is difficult to distinguish whether this is characteristic VWF tubules.

By contrast, in Fig. 1B we see two potential WPBs in which the VWF tubule content is partially preserved and which also have the characteristic coated outer surface. Although they are adjacent to the Golgi stack, we cannot see a direct connection for either organelle. In Fig. 1C there are multiple candidates for immature WPBs, at least some of which contain tubules, and a further bulbous structure with a more amorphous content that is substantially clathrin coated. These structures are in close proximity to the adjacent Golgi, but no direct connections are observed.

Finally, Fig. 1D shows an elongated structure, which contains ordered VWF tubules in the main body of the organelle. It is partially coated at one end where a tubular extension leads toward a fenestrated element of the Golgi, which is probably the TGN. This contains all the elements likely to be indicative of an immature WPB. However, such an image is very rare, and almost unique in the thousands of electron micrographs that we have obtained in the course of examining chemically fixed samples. More typical are the images in Fig. 1A-C, which are suggestive of what forming WPBs look like. However, they lack information and cannot provide a basis for a thorough characterisation of the immature WPB. Since these images suggest that the VWF tubules, the external coating and the putative tubular connections between the forming WPB and the TGN are labile and difficult to preserve by chemical fixation, we turned to HPF for further analyses of these critical elements.

Fig. 2 shows a significant improvement in preservation of the HUVECs, with less extracted cytoplasm, better-defined, perhaps more rounded, membranes (e.g. the MVB seen in Fig. 5A), more distinct clathrin coats and, most importantly, excellent preservation of the relationship between Golgi membrane protrusion and VWF tubule formation obtained by HPF. A low-power view of the peri-Golgi area of a HUVEC reveals numerous electron-lucent immature structures that are clearly WPBs adjacent to the Golgi stack (Fig. 2A). WPBs have characteristic striations reflecting their content of VWF tubules and are typically between 1-5 μm long, as assessed by immunofluorescence, and 100-250 nm wide. The image illustrates the size of these secretory granules compared with the Golgi from which they originate, each immature WPB being about the same length as the stretch of the Golgi ribbon seen in this image.

Fig. 2B shows that the VWF is able to tubulate within the TGN, the latter identified as such by its fenestrations, clathrin coat and proximity to the Golgi (Harrison-Lavoie et al., 2006). The presence of a single short tubule within the membrane protruding from the TGN, strongly suggests that this, rather than the structures shown in Fig. 1, is the earliest stage of WPB formation. We have not observed tubulation occurring within the other cisternae of the Golgi. The clathrin coat seen together with the forming tubule is suggestive of a role for clathrin not only in forming the cigar-shaped organelle, as we previously reported (Lui-Roberts et al., 2005), but also potentially in tubule formation itself. It is also noteworthy that the curvature of the coated bud is dramatically different to that of the adjacent conventional coated vesicles, which indicates that the presence of VWF may be affecting this coat's ability to control the curvature of the membrane.

Using HPF we often see structures with multiple tubules, as seen in Fig. 2C,D, that are still attached to the Golgi via a stalk. In Fig. 2C the TGN is seen en face and with an irregularly shaped WPB identified by its characteristic overall shape and internal tubules. In Fig. 2D we see two long WPBs filled with multiple tubules but still with membrane stalks emerging from their Golgi-proximal ends. These are presumably more-mature organelles than the protrusion containing a single tubule seen in Fig. 2B. The connecting stalks always lack VWF tubules and are much narrower than the WPBs, having a similar width to the Golgi cisternae (Fig. 2C,D). The stalk is rarely seen pinching off the centre of the end of the organelles but is usually displaced to one side. These structures do not appear as simple swollen ends to the trans-most cisternae. Rather, there is an abrupt transition from the large tubule-containing immature granule to the tubule-free stalk, and the forming WPB is rather separated from the cisterna. Overall, the resolution of the connections between forming WPBs and the TGN and the relative ease of finding these connections are much greater than for the images we obtained using chemical fixation.

Although we have previously shown that tubulation is necessary to form the characteristic elongated structure of WPBs (Michaux et al., 2006), it is apparent from the micrographs of HPF cells that the membrane is not simply wrapped directly around the tubule(s) in immature WPBs (Fig. 2, Fig. 3A,B). Indeed there even seems to be a consistent spacing between the membrane and the tubule along the long axis, perhaps even suggesting repulsion between membranes and tubules (Fig. 3B). By contrast, at the tip of the WPB the tubules are able to abut the membrane. Additionally the tubules themselves may repulse each other, as they tend to be evenly spaced and rarely touch each other along the long axis. Presumably as a consequence of these facts, the width of the immature WPB in a given section correlates with the number of tubules present (Fig. 3E), although there appears to be a minimum width so that the mean width of structures containing a single tubule is 120 nm.

The relationship between tubule length and WPB length is more complicated. There is a high degree of disorganisation in these early WPBs and by no means do all tubules run parallel to each other along the full length of the cigar. However, the tubules do seem to overlap such that in these elongated structures there are no areas that are completely devoid of tubules (Fig. 3C,D).

The maturation of WPBs is poorly understood. However, the use of HPF has given us new insights into this process. An interesting observation is the dual role for clathrin. Firstly, in the less mature WPB (or those with fewer tubules), it seems to play a supportive role as described by Lui-Roberts et al. (Lui-Roberts et al., 2005). In Fig. 4A the distinctive clathrin coat is seen at the tip, possibly promoting the bending of the membrane, though the degree of curvature is distinct from that seen on a clathrin-coated vesicle. Fig. 4B shows clathrin on the long axis of the WPB. This scaffold is often apparent in regions where the tubules are disorganised. The second role for clathrin is indicated in Fig. 4C, which shows a Weibel-Palade body prepared by HPF with a clathrin-coated bud that we postulate may be removing missorted material that is not destined for regulated exocytosis via the WPB.

A second maturation-related phenomenon, changes in the structure of content proteins, was also unknown for WPBs. We have found that HPF reveals a striking difference between immature and mature WPBs, not previously observed in HUVECs when using conventional chemical fixation: there is a dramatic increase in the electron density of the organelles (Fig. 5A-D). A low magnification view (Fig. 5A) shows that, while the immature WPB still attached to the Golgi is electron-lucent and filled with disorganised, discrete tubules, the mature WPB at the periphery of the cell is dramatically more electron dense. Further examples of, by our definition, mature WPBs showing increased electron density and found in the periphery of the cell are shown in Fig. 5B-D.

The increase in density is analogous to the concentration of cargo observed for other secretory organelles (Arvan and Castle, 1998). It could be explained by the compaction of the tubules. They have seemingly overcome the repulsion, to the membrane and each other, observed in the immature organelles. The compaction of tubules results in the narrower diameter of 151 nm observed in the mature WPB, as opposed to the 222 nm diameter in the immature WPB. In addition, it appears that electron-dense material is laid down between the tubules (Fig. 5D). Thus the loss of electron-lucent space must be a combination of these two factors. However, as with the immature WPB, there is a range in organelle width, in part due to the differences in numbers of tubules incorporated (Fig. 5E).

In conjunction with the compaction and deposition of electron dense material, the tubules become more regularly arranged, with the majority of tubules running continuously from one end of the WPB to the other through a single section plane (Fig. 5B,C).

In addition to the WPB containing tubules of VWF, we observe membrane-bound organelles containing short structures, distinct in morphology from cross-sections through tubules, albeit of the same dimensions (Fig. 6A-C). Such structures have not been observed using conventional chemical fixation. We believe that they are VWF multimers organised as tubular subunits. These organelles containing the subunits can be large and are often more rounded than elongated, which is consistent with the observation of Michaux et al. that tubules are required for elongation (Michaux et al., 2006). The putative subunits present as short sections of tubule, seen as parallel lines, spaced from each other at the same distance as the tubule edges in a full-length tubule (note the similarities between subunits and tubules within the separate organelles shown in panel B). The ends of subunits are routinely open, but sometimes curved, and the subunits are distinguishable from cross-sections of tubules. Fig. 6D shows an example of a structure containing both subunits and cross-sections of either tubules or subunits, whereas Fig. 6E contains cross-sections only. The cross-sections appear as circles in the micrographs. It is unclear whether these structures are still attached to the Golgi, nevertheless they are located close to cisternae as shown in Fig. 6B,C.

The high degree of VWF organisation coupled with the sheer size of these organelles could make manoeuvring around the cell difficult. However, a cursory look by confocal microscopy at endothelial cells stained for endogenous VWF shows that not all WPBs are straight. Fig. 7A shows a gallery of the different shapes of WPB observed. Although they do not conform to the conventional view of WPBs as highly organised linear structures, they do not appear to be curved but have an angled or `zig-zag' shape, indicating that, despite being generally rigid, they may fold at hinges. This suggests that WPBs could manoeuvre within cells by sharply bending at hinges rather than curling around obstacles. To observe bending WPBs in live cells over time, we took advantage of a VWF construct where the A2 domain was replaced by GFP (Romani de Wit et al., 2003). Expression of this construct in HUVECs revealed that most bending WPBs had only one break point or hinge. However, in one instance, we observed a WPB for over 24 minutes and identified up to three potential hinges (Fig. 7B, see also supplementary material Movie 1), which would allow more versatility of movement.

To understand how WPBs can bend at hinges, we examined their internal structure by electron microscopy. The abrupt bending of the WPB correlates with disordered content in the organelles (Fig. 7C). This indicates that the VWF tubules themselves are not able to curve and, consequently, are perhaps able to break at the hinges of the WPB, although the increase in electron density makes it difficult to discern exactly what is happening to the tubules. In addition to bending, the images reveal that the WPB can twist and branch further, aiding the mobility of the organelle (Fig. 5D and Fig. 7D respectively).

We have employed morphological approaches, primarily high-pressure freezing and freeze substitution, to analyse the formation of WPBs in HUVECs. This analysis is particularly timely because two new findings have recently appeared. The first was the discovery by Michaux et al. of the functional importance of the storage of VWF as tubules. If the tubules are key to fully functional WPBs, then an analysis of WPB formation that is focused on issues associated with tubules would be required. Second, Lui-Roberts et al. have recently found a key role for AP-1/clathrin in the formation of WPBs, but no analysis of which membranes are coated, especially in the context of tubulation, has hitherto been carried out.

We now demonstrate that, like conventional secretory granules, VWF undergoes progressive condensation, leading to denser and denser WPBs. Our results also suggest that clathrin has a dual role, in retrieving missorted material as well as in WPB formation. Finally, we used both EM and movies to demonstrate that VWF tubules are rigid but that WPBs can bend at hinges where the tubules are snapped.

Aspects of some of these issues are addressed in the early literature (increasing density, appearance of tubules in the Golgi) (Matsuda and Sugiura, 1970) but in no system have all the intimately related aspects been examined together. Crucially, we show here that the use of HPF allowed us to capture more information than previously possible on the biogenesis and maturation of WPBs in a well-characterised cell culture system, which is accessible to subsequent experimental investigation by modern molecular tools

The earliest point along the secretory pathway at which we see short but unmistakeable VWF tubules is within the TGN. Further, this location is consistent with our own light microscopic identification of immature WPBs associated with/forming within TGN-46-positive attenuated tubules extending from the TGN (Lui-Roberts et al., 2005). This is consistent with the importance of the interaction between the pro-peptide and mature VWF, in forming WPBs. Since the interactions are pH-dependent and the cleavage of the VWF by furin likely occurs within the TGN, a first appearance of the tubules shortly thereafter is in line with these observations. The shape of the membrane surrounding the tubule shown in Fig. 2B is of interest - it does not tightly wrap around the tubule. There must be some factor(s) controlling its dimensions beyond the visible VWF tubule.

We recently discovered a novel role for a clathrin/AP-1 coat in the formation of WPBs. We speculated that the coat was needed to act as an external scaffold in forming the structures (Lui-Roberts et al., 2005). We now see two kinds of coating revealed by HPF. One is a lattice along the axis or around the end of an elongated structure - the immature WPB, where it resembles the coating seen by chemical fixation and identified as clathrin/AP-1. This sometimes very extensive, partially curved coating is neither the flat lattices seen on the inner face of the plasma membrane nor the tightly curved coated buds seen on the TGN and the plasma membrane. One particularly interesting example of this is the presence of coating on the projection seen in Fig. 2B. Whether its failure to deform the TGN membrane into a classical budding profile is prevented by the presence of the VWF tubule is an attractive hypothesis, but as yet unconfirmed. This coat extends the known forms that clathrin can take within cells.

The other form of WPB-associated coat is on small traditionally shaped buds, probably involved in removal of material from the maturing granule, as has been found in other secretory granule model systems. The machinery used to recruit the coat is as yet unknown for this system, as are the contents. If these structures are used similarly to those observed elsewhere then they might be involved in removal of proteins such as the mannose 6-phosphate receptor or furin (Klumperman et al., 1998; Dittie et al., 1999). Whether they are particularly labile is not known, but they have been observed in other systems using chemical preservation; perhaps they are much less common in WPBs and this particular trafficking pathway plays a minor role in maturation of the endothelial granule.

Most secretory granules undergo ultrastructural changes as they mature after budding. This can include an increase in density and changes in shape or the development of highly ordered, even crystalline cargo. Such changes have been reported for WPBs in endothelial cells of the rabbit eye, where electron-lucent WPBs - reminiscent of our immature WPBs - are reported to be adjacent to the Golgi, whereas in other WPBs the tubules are `embedded in a moderately dense matrix of fine particles' (Matsuda and Sugiura, 1970). These latter appear similar to the mature WPBs seen at the periphery in endothelial cells after HPF. Interestingly, based on the distribution of the two types of WPB in different cells, Matsuda and Sugiura suggest that the organelles take more than 2 weeks to mature. Following preparation of samples by HPF, we see a dramatic difference between the pericentriolar electron-lucent structures and the more peripheral WPBs, which have an electron-dense interior in which the tubules can hardly be discerned. The latter are very different to the WPBs commonly seen in samples prepared by chemical fixation. Exactly how this change occurs is not understood. This apparent compaction may be a result of the compression required to produce functional VWF strings upon exocytosis (Michaux et al., 2006). The presence of electron-dense WPBs in rapidly growing HUVECs in culture, plus the time taken for WPBs to acquire membrane protein markers of maturation, such as Rab27 (Hannah et al., 2003), suggests that maturation time is a few hours rather than 2 weeks. This estimate would be in line with the time taken for WPBs to acquire membrane protein markers of maturation such as Rab27 (Hannah et al., 2003). The exact correlation in time or any functional relationship between acquisition of CD63, Rab27 and the shift in morphological density are not yet clear. However, the intracellular distribution of the electron-lucent versus the dense WPBs strongly implies that the former will be among the Rab27-negative population and the latter will be the mature, Rab27-positive population.

We found that VWF tubules within immature WPBs do not curve, even where there is plenty of space for them to do so, which suggests they are stiff. This rigidity could impede movement within the cell (especially since WPBs are 10 to 30 times bigger than most large secretory organelles, e.g. melanosomes); however, our observations of WPBs bending at hinges clearly indicate that the longest WPBs have some flexibility, although it requires VWF tubules to be discontinuous. Live observation of WPBs has been reported before (Romani de Wit et al., 2003) and a careful viewing of these data, plus helpful discussions with the authors has confirmed that their data is consistent with our interpretation. The mechanism underlying the formation of the hinges is unknown. One possibility is that tubules form in parallel during WPB formation, and an interruption to this process could lead to a hinge. Alternatively, homotypic fusion of already-budded WPBs, as described for other secretory granules (Tooze et al., 2001) could explain the presence of hinges, although such fusion events have not been described for WPBs. A third possibility is that tubules, albeit stiff, might be brittle, and thus easily mechanically `broken'. The cytoskeleton might generate an appropriate force for bending via motor-dependent interactions. Regardless of the mechanism used, it might be necessary for these stiff organelles to bend at hinge-points in order for WPBs to migrate from their pericentriolar site of formation to the cell periphery.

In some structures such as those seen in Fig. 6, we see very short lengths of parallel lines spaced apart by the same distance as the width of the tubules. Often in clusters within the vesicles, they sometimes also show connections that link them together. Altogether, they look remarkably like pre-tubules, subunits of VWF that have not yet come together to form long tubules. However, we note that since the structures are not tubules, we cannot formally disprove that they represent some other oligomerising molecule such as multimerin. If, as we surmise, the subunits are VWF, it is not yet possible to say what biochemical stage of multimerisation they represent, since no real model to map the biochemical data on VWF onto the tubular structure has yet been formulated. It is difficult to place these images within a series of WPBs of increasing maturity: we have not yet seen such subunits clustered within elements of the TGN, especially adjacent to a forming tubule, as might be expected if they were precursors to the type of image seen in Fig. 2B. Rather they are seen in detached organelles, sometimes of similar diameter to the immature WPBs and often with extensive coating as associated with tubule-formation. Possibly these precursors will never form mature WPBs but are destined for constitutive release as low molecular weight multimers, or for degradation. Indeed, it has been reported that between 90% (Tsai et al., 1991) and 10% (Sporn et al., 1986) of VWF is secreted constitutively, and that such material is of lower molecular weight than that released from WPBs by regulated exocytosis. Subunits may, alternatively, be able to organise themselves into tubules after they have left the Golgi. Were these subunit-containing structures to be WPB precursors, then post-budding fusion of immature WPBs must be a significant feature in the formation of WPBs, as is the case for other granules. However, whereas in Fig. 6C there are apparent connections between the individual subunits, we do not see the presence of subunits in the newly forming WPBs seen in Fig. 2, indicating that the subunits seem to be organised into tubules at an earlier stage.

Our data reveal answers to some of the questions raised in the introduction, yet the first detailed look at WPB formation by HPF also reveals many more intriguing features that demand further studies. Whether the use of HPF will be as productive for studying all organelles is not yet clear, but the preservation achieved by this method makes it a good choice for anyone interested in cell morphology/organelle formation and function. In our system, HPF has allowed us to confirm that tubulation of VWF occurs in the TGN and that the forming WPBs often remain intimately connected with the TGN while already in their elongated form. We also identify two additional maturation steps: the presence of clathrin-coated buds, which we surmise to be involved in retrieval of missorted proteins; and the concentration of cargo, observed by an increase in electron density. Finally, for the first time we see subunits in structures that may mature into WPBs or could be destined for constitutive release.

Human umbilical vein endothelial cells (HUVECs) were purchased from TCS Cellworks (Botolph Claydon, UK). The cells were grown in M199 (Gibco-BRL, Grand Island, NY) with 20% fetal calf serum (BioWest, Nuall, France), 10 U/ml heparin (Sigma, St Louis, MO), 50 μg/ml gentamicin (Gibco-BRL) and 30 μg/ml endothelial cell growth supplement (Sigma).

Cells were grown on carbon-coated, 1.4 mm diameter sapphire coverslips (Leica Microsystems UK). The coverslips were placed cell-side up into 1.5 mm diameter flat specimen holders under tissue culture medium and transferred to the EMPACT high-pressure freezer (Leica Microsystems UK) as quickly as possible and frozen as described previously (Reipert et al., 2004; Studer et al., 2001). The specimens were stored under liquid nitrogen until freeze substitution was carried out.

The Leica AFS freeze substitution unit was pre-cooled to -90°C and samples were transferred into the chamber under liquid nitrogen into pre-cooled fine-mesh-based plastic capsules in universal containers (Leica Microsystems UK). Samples were freeze substituted for 22 hours in 0.5% osmium tetroxide (TAAB) in extra dry acetone (Fisher Scientific). The osmium tetroxide was removed and, after a 2 hour acetone wash, replaced with 0.1% tannic acid (TAAB) in acetone for 8 hours. The AFS was then warmed at 5°C per hour to -60°C, held at -60°C for 8 hours, followed by a warm up to -30°C. The samples were held at -30°C for 8 hours, after which the tannic acid/acetone was exchanged for acetone. The samples were then warmed to 0°C at 5°C per hour, held for 1 hour and quickly warmed to 20°C in 1 hour to prevent evaporation of the acetone. Following two 10 minute washes at room temperature in propylene oxide, samples were infiltrated with propylene oxide:epon (1:1) for 1 hour followed by two changes of epon for two hours each. The sapphire coverslips, still in the flat specimen holders, were placed into flow-through capsules inside gelatin capsules (Leica Microsystems UK), topped up with epon and baked at 60°C overnight.

Following polymerisation, the epon blocks were cut out of the plastic/gelatin capsules with a razor blade. The specimen holders and sapphire coverslips were snapped off from the epon stubs by plunging into liquid nitrogen and then `flicking off' with a razor blade. Ultrathin sections were cut on an Ultracut UCT ultramicrotome (Leica Microsystems UK), post-stained with lead citrate and observed using a Tecnai G2 Spirit transmission electron microscope (FEI, Eindhoven, Netherlands). Images were collected using a Morada CCD (Olympus Soft Imaging Systems).

For immunofluorescence studies, antibody staining and confocal microscopy were carried out as previously described (Blagoveshchenskaya et al., 2002). Sheep polyclonal anti-human VWF (Serotec) was used in conjunction with secondary antibodies coupled to fluorophores (Jackson).

Cells were transiently transfected with VWF-GFP, as described previously (Romani de Wit et al., 2003), using Nucleofection (Amaxa, Cologne, Germany) according to the manufacturer's instructions. The movie in Fig. 8 was obtained 2 days after transfection. Cells were plated on a glass bottom dish (Intracel) and timelapse microscopy was performed using a Zeiss Axiovert Openlab fluorescence timelapse system enclosed at 37°C. Each image is the projection of Z-stacks. Fifty pictures were acquired over 24 minutes (about every 28 seconds).

This work was funded by the Medical Research Council of the UK. We thank members of the Cutler lab for critical comments. VWF-GFP was a kind gift from Jan Voorberg (Sanquin Research, Amsterdam, The Netherlands).