Efficient coupling of Sec23-Sec24 to Sec13-Sec31 drives COPII-dependent collagen secretion and is essential for normal craniofacial development

COPII-dependent budding from the endoplasmic reticulum (ER) is required for the efficient export of secretory proteins (Barlowe et al., 1994). COPII assembly is initiated by a transmembrane protein within the ER membrane, Sec12, which catalyzes guanine nucleotide exchange on the small GTPase Sar1. Sar1-GTP subsequently recruits a heterodimeric complex of Sec23 and Sec24 (Sec23/24), which is responsible for the capture of cargo proteins into nascent buds. COPII assembly proceeds through recruitment of Sec13 and Sec31 (Sec13/31), which form a structural cage around the budding vesicle. In elegant FRET assays of COPII assembly and cargo capture, Sato and Nakano (Sato and Nakano, 2005) have shown that Sec23/24 can stably assemble onto membranes in the presence of cargo through direct protein-protein interactions; this maintains Sec23/24 on the membrane during the Sar1-GTP cycle. In other in vitro data, Sar1 in its GTP-bound form has been directly implicated in the generation of membrane curvature through integration of its N-terminal amphipathic helix (Bielli et al., 2005; Lee et al., 2005). Sec13/31 can self-assemble into cage-like structures that could conceivably provide the driving force for membrane deformation (Stagg et al., 2006). Indeed, widely proposed models of Sec13/31 function (Mancias and Goldberg, 2005), including our own (Watson and Stephens, 2005), typically implicate Sec13/31 in the stabilization of the membrane association of Sec23/24 and accumulation of cargo, in the initiation and propagation of membrane curvature, and in the final scission event. Recent data have shown that, at least in vitro, Sec13/31 is not required for the formation of the pre-budding complex containing Sar1, Sec23/24 and associated cargo (Sato and Nakano, 2005). Although it is clear that the pre-budding complex of Sar1-Sec23/24 and cargo can form in vitro, the precise role of Sec13/31 in stabilizing the recruitment of cargo and in the deformation of the membrane has not been explored in intact cells.

In Saccharomyces cerevisiae and humans, good evidence exists that Sec13 is required for the generation of transport vesicles from the ER (Salama et al., 1993; Tang et al., 1997). Sec31 is similarly required for ER-to-Golgi transport in many assays (Salama et al., 1997; Tang et al., 2000). However, in yeast, Hsp150 was shown to be secreted in SEC13-1 mutant cells (Fatal et al., 2002). A series of `bypass Sec13' mutants have been isolated (Elrod-Erickson and Kaiser, 1996). These gene products negatively regulate COPII vesicle formation and are required for proper discrimination of cargo from ER residents in budding assays. These data suggest that, in some cases in which Sec13-function is impaired but probably not abolished (e.g. through temperature-sensitive mutation), ER export of some cargo can continue. We have used siRNA-mediated suppression of Sec13 and a combination of single-cell imaging, fluorescence photobleaching, electron microscopy (EM) and zebrafish genetics to define the role of Sec13/31 in vivo. Although our approaches do not lead to a complete loss of Sec13 from cells, we show that, in cells possessing less than 10% of their normal level of Sec13 expression, the ER is distended and decorated with large, open-necked budding profiles, but transport of small transmembrane cargo is normal; by contrast, collagen secretion is markedly inhibited by Sec13 depletion. Suppression of Sec13 expression in zebrafish results in defective proteoglycan secretion and in defects in craniofacial development, mimicking genetic loss of Sec23A function. It is important to note that, in both experimental systems, the expression of Sec13/31 is depleted but not abolished. These data and other recently published work (Bi et al., 2007; Fromme et al., 2007) suggest key roles for Sec13/31 in the completion of COPII-mediated budding through stimulation of the GTPase activity of Sar1. Together, these data show that sufficient levels of Sec13/31 and efficient assembly of the full COPII coat are required for effective secretion of selective components, notably of the extracellular matrix, and that this has direct implications for craniofacial development.

Depletion of Sec13 or Sec31 was achieved using siRNA duplexes transiently transfected into HeLa cells. Fig. 1 shows the efficacy of depletion by immunoblotting. Lamin-A/C was used as a targeted siRNA control with α-tubulin as loading controls. Notably, we observed a concomitant decrease in Sec31 expression upon Sec13 depletion (Fig. 1A). We were also able to effectively suppress expression of Sec31A from cells (Fig. 1B). We were unable to achieve a high level of suppression of expression of the second human Sec31 homologue, Sec31B (Fig. 1C), and this was further complicated by the detection of multiple bands by the Sec31B-specific antibody, which could reflect splice forms, degradation products of Sec31B (Stankewich et al., 2006) or non-specific detection. Loss of Sec31A from cells correlated with efficiency of Sec13 silencing (Fig. 1D), with the most-effective depletion of Sec13 and concomitant loss of Sec31A from cells occurring 72 hours after siRNA transfection. By contrast, loss of Sec31A alone (with limited suppression of Sec31B expression) did not cause concomitant loss of Sec13 (supplementary material Fig. S1). Unlike with other COPII components, there is only one Sec13 isoform in humans and, thus, coupled with concomitant loss of Sec31A upon Sec13 suppression, we anticipated that efficient suppression of Sec13 alone is likely to have key implications for COPII function.

Following depletion of Sec13/31 expression, we determined the intracellular localization of markers of the ER-Golgi boundary in cells (Fig. 2). Depletion of Sec13 causes a clear reduction of Sec31A-positive labelling (Fig. 2A), consistent with the results of the immunoblotting experiments shown in Fig. 1. Furthermore, Sec13 depletion caused a relocalization of COPII (Sec24C and Sec16) and cargo (ERGIC-53) to the juxtanuclear region, which is very reminiscent of the effects of expression or microinjection of a GTP-restricted mutant of Sar1 (e.g. Stephens and Pepperkok, 2004). In addition, the Golgi (GM130-labelling) becomes more compact in these cells yet retains its characteristic juxtanuclear localization. Similar (although not so dramatic) effects were seen following depletion of Sec31A and Sec31B together (Fig. 2B; using the most effective siRNA duplexes against each isoform, Sec31A-2 and Sec31B-2, see Fig. 1); we attribute the lesser severity of the phenotype to the less effective depletion of these proteins. Images in each column of Fig. 2A were taken at the same imaging settings and therefore reveal this dramatic accumulation, particularly that of the otherwise more diffuse ERGIC-53 labelling, in the juxtanuclear region. This accumulation of COPII and ERGIC-53 is not due to an upregulation of protein synthesis, as was determined by immunoblotting (Fig. 2C). Fig. 2D shows the localization of ERGIC-53 at a lower exposure in Sec13-suppressed cells; here, one can clearly see the juxtanuclear accumulation of both Sec16 and ERGIC-53 following Sec13 suppression compared with lamin-A/C-suppressed controls. During the early secretory pathway, COPII function is closely coupled to COPI (Scales et al., 1997). Immunolabelling shows the classical Golgi and peripheral punctate labelling seen for COPI (Fig. 2E, here localizing the β-COP subunit), and the global loss of COPI from membranes in Sec13-suppressed cells with some remaining localization to both Golgi-like structures and peripheral punctae. Effective suppression is evidenced by loss of Sec31A labelling. Despite this clear perturbation of the early secretory pathway in Sec13-suppressed cells, localization of human β1,4 galactosyltransferase I (GalT, a marker of the trans-Golgi) was unchanged compared with lamin-A/C-suppressed controls, retaining its characteristic juxtanuclear localization (data not shown).

To investigate the nature of these juxtanuclear accumulations of COPII and cargo in more detail, we performed fluorescence recovery after photobleaching (FRAP) to determine whether COPII could continue to cycle on and off these structures [as would be predicted from the work of Sato and Nakano (Sato and Nakano, 2005)]. FRAP experiments (Fig. 3) showed a statistically significant increase (P=0.0072) in the half life of recovery of YFP-Sec23A in Sec13-depleted cells compared with controls (Fig. 3A). This indicates an increased residency time on the ER membrane. The immobile fraction of YFP-Sec23A at ER exit sites (ERES) shows a statistically significant increase (P=0.0097) following depletion of Sec13 compared with controls (Fig. 3B), indicating that a larger fraction of the protein is stably associated with ERES membranes at any one time. Furthermore, GFP–ERGIC-53 (Ben-Tekaya et al., 2005) showed rapid recovery into photobleached areas, indicating cargo entry into and exit from these structures, consistent with continuity with the underlying ER membranes (not shown).

To further define the requirement for Sec13 in ER export, we carried out a transport assay in cells expressing the temperature-sensitive glycoprotein tsO45-G-YFP (Gallione and Rose, 1985; Keller et al., 2001). Sec13 is required for ER export in many systems but, much to our surprise, depletion of Sec13 did not inhibit the transport of tsO45-G-YFP through the secretory pathway (Fig. 4). In these assays, cells depleted of Sec13 for 48 hours were infected at 39.5°C with an adenovirus expressing tsO45-G-YFP (Keller et al., 2001); under these conditions, tsO45-G-YFP is retained in the ER. After a further 16 hours, cells were shifted to 32°C to allow export of tsO45-G-YFP. Fluorescence microscopy showed that tsO45-G-YFP was equally delivered through the secretory pathway at 32°C in lamin-A/C-depleted cells and Sec13-depleted cells (Fig. 4A). Also evident in Fig. 4A (total tsO45-G-YFP) is the juxtanuclear condensation of labelling following Sec13 depletion, as would be predicted from the data in Fig. 2. Arrival of tsO45-G-YFP at the plasma membrane (shown by immunolabelling with an antibody directed against the extracellular epitope in non-permeabilized cells, Fig. 4A, lower panels) is indistinguishable between lamin-A/C- and Sec13-suppressed cells. Quantification of the delivery of tsO45-G-YFP to the plasma membrane showed no significant difference in Sec13-suppressed cells compared with lamin-A/C-suppressed controls (Fig. 4B). Knockdown efficiency was confirmed by Sec31A immunolabelling (data not shown). In order to define any subtle kinetic differences in ER-to-Golgi transport, we measured the delivery of the protein to the Golgi by monitoring acquisition of EndoH resistance by tsO45-G-YFP; this was similarly unaffected by Sec13 depletion (Fig. 4D) compared with controls (Fig. 4C) (one representative example is shown). This provides a more sensitive detection of ER-to-Golgi transport than delivery to the plasma membrane; despite a very strong suppression of Sec13 expression in these experiments, no significant differences were seen in the delivery of tsO45-G-YFP to the Golgi. Visual inspection of time-lapse sequences also showed no differences in delivery of tsO45-G-YFP to the juxtanuclear region (data not shown). Given the essential nature of Sec13/31 for cell function, we were very surprised to find that transport of tsO45-G-YFP from the ER to Golgi and on to the plasma membrane was not significantly slowed in cells depleted of Sec13 compared with wild type. A huge array of evidence argues against Sec13/31 being dispensable for function in COPII vesicle biogenesis (e.g. Pryer et al., 1993; Salama et al., 1997; Tang et al., 1997; Matsuoka et al., 1998), and we interpret our data as indicating that the small quantity of Sec13/31 remaining following siRNA suppression is sufficient for efficient transport of tsO45-G-YFP. Our data suggest that the quantity of Sec13/31 normally expressed within HeLa cells is not required for all ER export. Because our experiments only provide a reduction in Sec13/31 expression and do not entirely abolish it, our data are consistent with an essential role for small amounts of Sec13/31 in ER export.

In order to examine the ultrastructure of ER and associated buds in Sec13-depleted cells, we performed EM of high-pressure frozen and freeze-substituted samples (Fig. 5). In control cells (Fig. 5A), we routinely observed budding profiles from ribosome-free areas of ER in close proximity to Golgi membranes (Fig. 5A, arrowhead). Following Sec13 depletion, distended ER was evident throughout cells, and frequently in the vicinity of Golgi membranes (Fig. 5B and enlarged in Fig. 5C). This distended region was largely ribosome-free, which is consistent with the enlargement of transitional ER and perhaps reflects the increased fluorescence intensity in this region seen in Fig. 2. Distended ER was also evident at peripheral sites within the cell, away from the Golgi (data not shown). These ER structures were also decorated with budding profiles that were reminiscent of COPII buds at transitional ER. Budding profiles in close proximity to Golgi membranes could alternatively be COPI coated. These buds were larger than the vesicular profiles seen in control cells (Fig. 5A, arrow). In addition to these large structures, we also observed more-typical coated vesicle profiles, which were in association with presumed ERES (Fig. 5B, arrows). Budding of coated vesicles from the Golgi appeared to be unaffected and, from inspection of multiple images, the Golgi did not appear to be significantly distended following Sec13 suppression. These data suggest that at least a subset of COPII-dependent budding from the ER is perturbed upon Sec13 depletion.

These vesicle bud profiles and distended ER are highly reminiscent of those seen in patient fibroblasts from sufferers of cranio-lenticulo-sutural dysplasia (CLSD; OMIM #607812) (Boyadjiev et al., 2006; Fromme et al., 2007). One key feature of this disorder is the accumulation of collagen in fibroblasts. We therefore examined the effect of suppressing Sec13 expression on collagen secretion from human primary dermal fibroblasts. Fig. 6A shows that control (lamin-A/C suppressed) cells show strong deposition of collagen, as shown by labelling for the C-terminal telopeptide (using antibody LF-67, arrowheads in Fig. 6A indicate robust deposition of extracellular collagen fibrils); these cells also show relatively uniform labelling for Sec31A. By contrast, cells depleted of Sec13 show much less deposition of extracellular collagen fibrils (compared with lamin-A/C-suppressed cells) and loss of Sec31A labelling (consistent with our data from HeLa cells). Semi-automated image analysis was used to quantify these data. Thresholding images to detect only high pixel intensities allowed selective detection of collagen fibrils, which were then counted by measuring their skeletal length in large-format montage images of multiple fields of view acquired using an automated X-Y scanning stage (supplementary material Fig. S2). Further examination of intracellular procollagen using an antibody directed against the N-terminal propeptide of type I procollagen (LF-39) showed clear accumulation of unprocessed procollagen in cells depleted of Sec13/31 (supplementary material Fig. S2B, asterisks). These data revealed a >threefold loss of fibrillar-collagen deposition in Sec13-suppressed cells compared with lamin-A/C-suppressed controls (Fig. 6B). Immunoblotting confirmed effective suppression of Sec13 expression and concomitant loss of Sec31A from these cells (Fig. 6C). Multiple species of Sec31 were seen in primary fibroblasts, probably owing to the presence of multiple splice forms and/or post-translationally modified forms of the protein (Tang et al., 2000); all forms were suppressed by Sec31A siRNA transfection. Furthermore, the distribution of GalT in Sec13-suppressed fibroblasts was indistinguishable from that of control cells (data not shown). Thus, suppression of Sec13 expression reduces the secretion and deposition of collagen fibrils without significantly affecting the secretion of tsO45-G-YFP or the accumulation of GalT in the Golgi. These data reflect observations made in patient fibroblasts from CLSD cases (Fromme et al., 2007).

Many phenotypic aspects of CLSD are recapitulated in crusher mutant zebrafish (Lang et al., 2006); both CLSD and crusher are caused by a mutation in Sec23A (Boyadjiev et al., 2006; Lang et al., 2006). To examine the role of Sec13/31 in vivo, we used morpholino oligonucleotide-based suppression of Sec13 expression in embryos of the zebrafish Danio rerio. Two non-overlapping translation-blocking morpholinos were used for these experiments. Sec13 morphants were viable [at least until 5 days post fertilization (dpf)] but were smaller overall, had smaller heads and eyes, and showed pigmentation defects (Fig. 7A). Fig. 7B shows that, at 5 dpf, Sec13 morphant embryos have defects in craniofacial development and the pectoral fins are kinked (arrows in wild-type, arrowheads in Sec13 morphants). This phenotype shows some similarity to the Group-II classification (mutations affecting cartilage differentiation and morphogenesis) of Neuhauss and colleagues (Neuhauss et al., 1996), which includes the crusher mutant. Alcian-blue staining (which primarily stains proteoglycans of extracellular-matrix components) to reveal the skeleton shows that, at 4 dpf, there is a noticeable absence of cartilage formation at the front of the head (e.g. Meckel's cartilage; Fig. 7C, mc) in Sec13 morphants. The defect is more pronounced at 5 dpf, which reveals extensive disorganization of structure within the neurocranium and defects in the pectoral fin, which is kinked and often malformed. Additional images obtained from experiments using alternative morpholino oligonucleotides are shown in supplementary material Fig. S3. Scanning electron microscopy (Fig. 7D, showing dorsal, lateral and ventral views) shows clearly the defects in craniofacial (Fig. 7D, arrowhead) and pectoral-fin (Fig. 7D, arrow) development. Notably, a clear defect in eye morphogenesis is also evident from these images; we are currently investigating this further. Immunoblotting of lysates of dechorionated embryos confirms the reduction in Sec13 expression (Fig. 7E). Furthermore, the embryos show a concomitant loss of Sec31A (Fig. 7E), as we observed for HeLa cells (Fig. 1). Although these blots show that suppression of expression of Sec13 in zebrafish embryos is less effective than in tissue-culture cells, our data show clearly that the deposition of proteoglycan is impaired and that this correlates with clear defects in craniofacial development.

Our data provide key information regarding Sec13/31 function in intact cells and during embryonic development. We found that Sec13 is required to maintain the structural stability of Sec31. Depletion of Sec13/31 causes a clustering of pre-budding complexes containing Sec16, Sec23/24 and cargo in a juxtanuclear area, which is highly reminiscent of the phenotype following expression of a GTP-restricted form of Sar1. The remaining COPII-coat associated with these pre-budding complexes is stabilized in Sec13-depleted cells. Sec13/31 suppression causes distension of the ER and accumulation of budding profiles on the ER membrane. Collagen secretion is defective but transport of smaller cargo is unperturbed. Finally, suppression of Sec13 in zebrafish causes defects in proteoglycan secretion and craniofacial-development defects.

The finding that Sec13 is required to maintain the stability of Sec31 is entirely consistent with the X-ray crystallographic structure of the Sec13/31 complex, in which part of Sec31 is seen to form one blade of the Sec13 WD40 propeller structure (Fath et al., 2007). In line with data obtained from in vitro reconstitution experiments (Matsuoka et al., 1998), we found that >90% depletion of Sec13/31 expression does not perturb the association of Sec23/24 or Sec16 with the membrane. Thus, pre-budding complexes are able to form, recruit cargo and accumulate in the juxtanuclear area. Recent kinetic data from in vitro binding experiments show that the association of Sec23/24 with cargo on membranes can be maintained despite ongoing GTP hydrolysis by Sar1 (Sato and Nakano, 2005). The presence of Sec23/24 on the membrane can be maintained by interactions with cargo proteins (Sato and Nakano, 2005) and the continual GTP loading of Sar1 by Sec12 (Futai et al., 2004). Indeed, the rate of GTPase stimulation by Sec23 is balanced by the rate of GTP loading by Sar1 such that a continual supply of Sar1-GTP is maintained. The binding and further stimulation of GTP hydrolysis by Sec13/31 results in a net loss of active Sar1 from the membrane (Antonny et al., 2001). Our FRAP experiments provide support for this model in living cells. We conclude that it is the recruitment of Sec13/31 that is required to stimulate Sar1 GTPase activity in vivo, either by acting on the conformation of the Sec23/24 complex in association with Sar1, or (as indicated by other recent data) through a direct effect on Sar1 itself (Fath et al., 2007).

Our EM data indicate that suppression of Sec13/31 does not prevent either the generation or stabilization of membrane curvature. Thus, these properties probably lie in other COPII components, which could include Sec16, which, in concert with Sar1, can form large curved structures in cells (Watson et al., 2006), and/or Sec23/24, which is proposed to adopt a curved conformation (Bi et al., 2002). Curvature of the membrane can result from activation of Sar1 to its GTP-bound form, directing association of its amphipathic N-terminus with the membrane (Bielli et al., 2005; Lee et al., 2005). GTP hydrolysis by Sar1 could result in scission of the vesicle from the membrane through perturbation of a bilayer structure (Bielli et al., 2005; Lee et al., 2005). In Sec13-depleted cells, the loss of Sec13/31 GAP activity could limit the potential for membrane scission, resulting in a decrease in carrier formation. It is also possible that Sec13/31-stimulated Sar1 GTPase activity is required for faithful incorporation of certain cargoes. Functions have been shown for Sar1 GTPase activity in the fidelity of cargo sorting into COPII vesicles (Sato and Nakano, 2004), and differing requirements for Sar1-GTPase activity in terms of cargo incorporation in to pre-budding complexes have been found (Stephens and Pepperkok, 2004). The defined coat in our EM images probably reflects accumulation of Sar1-Sec23/24 complexes on the membrane; this is largely absent in images from patient fibroblasts (Fromme et al., 2007), which could either reflect a difference in preparation method or could suggest that pre-budding complexes containing Sec23A-F382L are less stably associated with the ERES membrane than those containing wild-type Sec23A.

Defects in COPII-dependent budding have been identified as the underlying cause of CLSD (Boyadjiev et al., 2003; Boyadjiev et al., 2006; Lang et al., 2006; Fromme et al., 2007). CLSD results from a missense mutation in Sec23A [F382L (Boyadjiev et al., 2006)] that lies at the point of Sec23A that directly contacts Sec31 (Bi et al., 2007). This disease is characterized by facial dysmorphisms in humans (Boyadjiev et al., 2003) and is largely recapitulated by the crusher mutation in the zebrafish D. rerio (Neuhauss et al., 1996; Lang et al., 2006). The crusher mutation is in fact a nonsense mutation in Sec23A (Lang et al., 2006), almost certainly resulting in non-functional protein. Both CLSD patients and crusher mutant zebrafish have defects in the export of both collagen and complex proteoglycans (Boyadjiev et al., 2006; Lang et al., 2006).

Recent data have suggested that the selective defect in CLSD patients and crusher zebrafish (i.e. that only certain cell types show a phenotypic defect and indeed that the patients and zebrafish are viable) is due to the presence of a second Sec23 isoform, Sec23B, in both humans (Boyadjiev et al., 2006; Fromme et al., 2007) and zebrafish (Lang et al., 2006). Low expression of Sec23B in calvarial osteoblasts (which are primarily responsible for the primary ossification of the cranial suture) could mean that this isoform cannot compensate for defects in Sec23A and therefore these specific cell types show defects in secretion (Fromme et al., 2007). However, suppression of either Sec23A or Sec23B in zebrafish individually, or in combination, results in embryos that show craniofacial development defects (Lang et al., 2006). The efficacy of knockdown in these experiments was not shown, but these data show that neither Sec23A nor Sec23B is alone sufficient for proper craniofacial development in zebrafish. These data are consistent with those presented here if one considers that the expression of both Sec23A and Sec23B in calvarial osteoblasts together provide sufficient Sec23 to couple efficiently to the outer, Sec13/31, layer of the COPII coat. Our data reveal very similar phenotypes to those of CLSD patients and crusher zebrafish, and suggest that we are recapitulating the defect in the coupling of Sec13/31 to Sec23/24. Consequently, we conclude that, rather than a specific role for Sec23A (e.g. in the capture of specific cargo), it is the efficient and effective coupling of Sec23/24 to Sec13/31 that is the key requirement for craniofacial development.

Sec13 is expressed from the end of gastrulation onwards in zebrafish (Thisse et al., 2001) and, notably, during the first 48 hours of embryonic development, expression is highest in the notochord and during this time increases in the cleithrum, optic tectum and pectoral-fin musculature (Thisse et al., 2001), which correlates well with the developmental defects that we observe in Sec13 morphants (particularly those evident in scanning electron micrographs). Intriguingly, expression is also high in the otic vesicle and retina. Sec31 is also highly expressed in notochord, optic tectum, otic vesicle, cleithrum and fin (Thisse et al., 2001), again correlating well with our findings. This suggests an unusually high secretory load in these tissues at this stage of development, which requires highly efficient COPII function.

The dysmorphisms of the cranial structure that we observe in our Sec13 morphants is in fact most similar to that described for the bulldog, Jekyll, mr hyde and particularly round mutant zebrafish embryos (Neuhauss et al., 1996). Jekyll is caused by a mutation in UDP-glucose dehydrogenase that results in defective proteoglycan secretion (Walsh and Stainier, 2001). Taken together, our data and those of other laboratories (e.g. Fromme et al., 2007) suggest that Sec23A is not the sole component that is relevant to the secretion of collagen or proteoglycan, i.e. it is not the specific cargo selector involved. Rather, any defect in coupling of the inner and outer layers of the COPII coat appears to result in defects in secretion of large macromolecular cargo from the ER. Clearly, it will be of interest to determine whether any other craniofacial-development mutants in zebrafish or other clinical cases with craniofacial dysmorphisms are caused by mutation in Sec13, Sec31, other COPII subunits or indeed other components of the ER export machinery.

Surprisingly, despite a robust inhibition of collagen secretion and deposition, we did not observe any significant inhibition of transport of tsO45-G-YFP in cells depleted of Sec13/31, the Golgi remained largely unperturbed, and GalT remained localized to the Golgi cisternae, indicating that there was not a general defect in the transport of glycosylation enzymes to the Golgi. Other work strongly indicates that Sec13/31 is required to provide the final stimulus of Sar1 GTPase activity to trigger vesicle fission (Fath et al., 2007; Fromme et al., 2007). At first glance, this might be difficult to reconcile with our own work, in which we show that a near-complete loss of Sec13 has no significant effect on the trafficking of GalT or tsO45-G-YFP. However, we would argue that our data are in fact entirely consistent with this model, in particular if one invokes the intriguing possibility of selective recruitment of the remaining Sec13/31 to the bud neck to drive fission. Alternatively, it is possible that the remaining Sec31 is alone sufficient to stimulate Sar1 GTPase activity (because more Sec31 than Sec13 remains in each of our experiments, including following direct Sec31 suppression).

We hypothesize that normal expression levels of Sec13/31 are required to scaffold the formation of larger carriers emerging from the ER (such as those containing assembled, fibrillar collagen) but not for the secretion of smaller cargoes (tsO45-G-YFP or GalT). Scaffolding activity could require more-efficient coupling of Sec13/31 to Sec23/24 than does recruitment for a scission event. Thus, inefficient coupling of the inner and outer layers of the COPII coat through depletion of Sec13/31 from cells would result in a selective defect in secretion of large, macromolecular cargo. Our data therefore argue that efficient formation of the full COPII coat is of greater importance for collagen secretion than any direct role for Sec23A in, for example, cargo concentration. In addition to a defect in the secretion of fibrillar collagens and complex proteoglycans, this could include chylomicron retention diseases (OMIM #246700), in which the assembly of large apoB-containing chylomicrons is defective owing to mutation of Sar1B. Fromme and colleagues provide further evidence that differential coupling of Sar1 isoforms to Sec13/31 might underlie this disease (Fromme et al., 2007). In summary, our data indicate a selective requirement for non-limiting quantities of Sec13/31 in COPII-dependent export from the ER, and reveal a clear requirement for highly efficient COPII-coat assembly in craniofacial development.

All reagents were purchased from Sigma-Aldrich (Poole, UK) unless otherwise indicated. HeLa cells (ATCC CRL-2) were maintained in DMEM supplemented with 10% FCS (Invitrogen, Paisley, UK) and 1% glutamine. Primary human dermal fibroblasts were purchased from Cascade Biologics/Invitrogen and were cultured in Medium 106 supplemented with low-serum growth supplement. At 24 hours prior to the start of the experiments, cells were seeded onto either 22-mm coverslips or glass-bottom dishes (MatTek, Ashland, MA).

Monoclonal mouse anti-ERGIC-53 was from Alexis Biochemicals (Nottingham, UK), anti-GM130 was from BD Transduction Laboratories (Cambridge, UK), anti-tubulin from Neomarkers (Fremont, CA) and anti-GFP from Covance (Harrogate, UK). Polyclonal antibodies against Sec16A were from Bethyl Laboratories (Montgomery, TX), Sec13 antibodies were generous gifts from Wanjin Hong (IMCB, Singapore) and Beatriz Fontoura (Southwestern Medical Center, Dallas, TX), anti-lamin-A/C from Cell Signaling Technology (Hitchin, UK), mouse monoclonal anti-Sec31A was from BD Biosciences, and anti-β-COPI (maD) was from Sigma-Aldrich. Secondary antibodies were from Jackson ImmunoResearch Laboratories (PA). Anti-human GalT was from CellMab (Gothenburg, SE). Collagen-I antibodies [LF-39 directed against the human collagen α1 (I) amino-propeptide (Fisher et al., 1989) and LF-67 recognizing the human collagen α1 (I) carboxy-telopeptide (Bernstein et al., 1995)] were very generously provided by Larry Fisher (NIH, Bethesda, MD). Anti-Sec24C and -Sec31A antibodies were raised against synthetic peptides synthesized by Graham Bloomberg, University of Bristol and coupled to KLH before immunization into rabbits. Antibodies were affinity purified using peptide coupled to sulfolink resin (Pierce, Cramlington, UK) according to the manufacturer's protocols. Peptide sequences were as follows: Sec24C: MNVNQSVPPVPPFGQ(C); Sec31A: MKLKEVRTAMQAWS(C), in which C-terminal cysteines were added for coupling.

siRNA duplexes against Sec13 were designed using the online algorithms of, and synthesized by, MWG-Biotech (London, UK). These sequences were: Sec13-1, 5′-GCACUCAUGUUACGAGGAA-3′; Sec13-2, 5′-GGAGGAGCAGAAGCUAGAA-3′; Sec31A-1, 5′-ACACAGGAGAGGUGUUAUA-3′; Sec31A-2_3′UTR, 5′-GCUUCUCUUCCACUCAAUA-3′; Sec31B-1, 5′-AGAUAUUGAUGGACUCCUA-3′; and Sec31B-2, 5′-UUGGAGCCAUAUUAAGGAA-3′.

HeLa cells were transfected with siRNA duplexes using a calcium-phosphate method as previously described (Watson and Stephens, 2006). Primary human dermal fibroblasts were transfected using geneFECTOR (VennNova, FL) according to the manufacturer's instructions. Times indicated for siRNA suppression (typically a total of 72 hours) refer to the time from the addition of siRNA duplexes.

For immunofluorescence, cells were first washed with phosphate buffered saline (PBS) and then fixed with methanol at –20°C for 4 minutes, followed by blocking with PBS containing 3% bovine serum albumin (BSA) for 30 minutes. Primary antibodies were diluted to varying concentrations in PBS containing 3% BSA. Cells were incubated with primary antibodies for 1 hour at room temperature. They were subsequently washed with PBS and incubated with secondary antibodies, either Cy2 or Cy3, for 1 hour at room temperature. Cells were counterstained with DAPI and mounted onto slides using Mowiol. For collagen labelling, cells were fixed in paraformaldehyde for 20 minutes, permeabilized with 0.1% Triton X-100 for 5 minutes, blocked in 3% BSA/PBS and labelled with either LF-67 or LF-39 in combination with mouse monoclonal anti-Sec31A. Available Sec13 antibodies were not found to be suitable for immunofluorescence. Fixed cells were imaged on an Improvision 3DM system (Improvision, Coventry, UK) comprising an IX-81 microscope (Olympus Microscopes, London, UK), with ASI PZ-2000 XYZ stage (Applied Scientific Instruments, Eugene, OR) DG-4 illumination system (Sutter, Novato, CA, USA) with Brightline `Pinkel' filter sets with single-band exciters and multi-band dichroic and emission filters (Semrock, Rochester, NY). Images were acquired using a Hamamatsu Orca-ER cooled CCD camera with Volocity version 4.2 (Improvision). Images were resized using Adobe Photoshop and montages were created using Adobe Illustrator CS (Adobe).

Cells were seeded onto live-cell dishes and siRNA transfected for 72 hours using oligofectamine (Invitrogen) as per the manufacturer's instructions. Cells were subsequently transfected with plasmid DNA using FuGENE, as per the manufacturer's instructions, 48 hours after siRNA transfection. Immediately before imaging, the media was aspirated from cells and replaced with imaging media (DMEM containing 30 mM HEPES). YFP-Sec23A (Stephens, 2003) and GFP-ERGIC-53 (Ben-Tekaya et al., 2005) have been previously described.

Cells were live imaged in a 37°C heated Perspex box (Life Imaging Services, Reinach, CH) on a Leica SP5 confocal imaging system (Leica Microsystems, Milton Keynes, UK) using 488 nm and 514 nm lines of an argon laser. A pinhole size of 3 Airy disk was used to take images using a Leica DMI 6000 inverted microscope. Cells were subjected to five pre-bleach frames, eight bleach frames using three lasers at 100% power and 140 post-bleach frames. Data was processed and statistical significance determined using GraphPad Prism (San Diego, CA).

Cells were grown on either 6- or 3-cm dishes until 80% confluent. Media was removed and cells were washed with ice-cold PBS before addition of ice-cold lysis buffer (50 mM Tris-HCl pH 7.4, 150 mM NaCl, 1 mM phenylmethyl sulphonylfluoride, 1% Triton X-100, 1× protease inhibitor cocktail). Cell scrapers were used to remove cells from the dish. Lysates were then centrifuged at 20,000 g at 4°C for 20 minutes to remove cell debris. Protein assays were carried out using a BCA protein assay kit (Pierce). Typically, between 10-20 μg of protein was loaded onto precast SDS-PAGE gels (Invitrogen). Rabbit anti-Sec13 (Wanjin Hong) and rabbit anti-Sec31A (described herein) were used for immunoblotting of lysates from both HeLa cells and fibroblasts.

Cells were plated onto 3-cm dishes and siRNA transfected as above. At 48 hours after transfection, cells were infected with YFP-tsO45-G-virus for 2 hours at 37°C, 5% CO₂. The virus was then removed and fresh media added before shifting the cells to a 39.5°C, 5% CO₂ incubator for 16 hours. The dishes were placed in a 32°C water bath and, after each time point (0, 10, 20, 30, 60 and 90 minutes), cells were removed and washed with ice-cold PBS before addition of lysis buffer (50 mM Tris-HCl pH 7.4, 150 mM NaCl, 1 mM PMSF, 1% Triton X-100, 1× protease inhibitor cocktail). Cells were removed from the plates using a cell scraper and then transferred to a microcentrifuge tube and centrifuged at 20,000 g, 4°C for 20 minutes to remove cell debris. The protein concentration was determined using a BCA protein assay kit (Pierce). Protein sample (10 μg) was heated at 95°C for 5 minutes in denaturing buffer (Invitrogen). Reaction buffer was added to each sample before adding 250 units of EndoH (NEB, Hitchin, UK) and incubating at 37°C for 16 hours. Following this, SDS sample buffer was added to the samples and heated at 95°C for 5 minutes. Subsequently, the samples were run on Bis-Tris polyacrylamide gels (Invitrogen) and probed with an anti-GFP antibody (Covance).

Sec13- or lamin-A/C (control)-suppressed HeLa cells were grown on 1.5-mm sapphire discs. They were briefly rinsed in DMEM containing 30 mM HEPES, 10% FCS and 20% BSA. The latter two components were added to prevent freezing artefacts (McDonald et al., 2007). The sapphire disc was then placed in a 0.1-mm-deep membrane carrier and high-pressure frozen using the EMPACT2 with rapid-transfer system [Leica Microsystems (Verkade, 2008)]. The samples were subjected to a short-freeze-substitution protocol (Verkade, 2008). Briefly summarized, samples were placed in 1% OsO₄ and 0.1% uranyl acetate in ultra-pure acetone and left for 5 hours at –90°C. The temperature was slowly increased (5°C per hour) to 0°C and the samples were washed with ultra-pure acetone. They were gradually infiltrated with Epon and eventually embedded in moulds. The Epon was hardened overnight and ultrathin sections were counterstained and analyzed with a FEI Tecnai12 Biotwin (FEI, Eindhoven, The Netherlands) equipped with a bottom-mount 4K EAGLE CCD camera.

Zebrafish (D. rerio) AB strain were raised and maintained as described previously (Westerfield, 1995). Two non-overlapping, translation-blocking morpholino oligonucleotides targeting D. rerio Sec13 (BC153483.1) were designed and synthesized by GeneTools LLC (Philomath, OR). The sequences used were MO1: 5′-CACTGTGTTAATGACCGAAACCATG-3′ and MO2: 5′-TTTGCTTATATCCCTCAACAACCTC-3′. Morpholino knockdown of Sec13 protein level and expression levels of Sec31 and α-tubulin were confirmed by western blotting using antibodies against mammalian proteins that showed cross-reactivity to D. rerio orthologues. Embryos at 4 and 5 dpf were fixed in 4% PFA for 2 hours then mounted in 1% methylcellulose for photography. Alcian-blue staining of 4- and 5-dpf embryos was performed as described previously (Neuhauss et al., 1996); samples were mounted in 100% glycerol and all photographs were taken using Zeiss Axioplan2 5× or 10× lens (Zeiss, Welwyn Garden City, UK).

For scanning electron microscopy, morpholino-injected zebrafish (5 dpf) were fixed in 4% PFA at 4°C, then post-fixed in 1% OsO₄ in dH₂O for 1 hour, dehydrated in graded ethanol, and critical-point dried (100% ethanol/carbondioxide). The specimens were sputter-coated with gold (Polaron Sputter Coater) and viewed on a Philips 501B emission scanning electron microscope.

We thank Harry Mellor, Jon Lane, Pete Watson, Pete Cullen and members of the Stephens lab for helpful discussions throughout this project, and Paul Martin for facilitating the work with zebrafish. We also thank Kai Simons, Patrick Keller, Hans-Peter Hauri, Larry Fisher, Wanjin Hong and Beatriz Fontoura for kind gifts of antibodies and plasmids. We thank the MRC for funding this work through a Senior Non-Clinical fellowship from (to D.J.S.), a research grant from the BBSRC, and the MRC, Wolfson Foundation for providing support to establish and develop the Wolfson Bioimaging Facility.