Structure of the complete, membrane-assembled COPII coat reveals a complex interaction network

Joshua Hutchings, Viktoriya G. Stancheva, Nick R. Brown, Alan M.C. Cheung, Elizabeth A. Miller, Giulia Zanetti

Preprint posted on June 24, 2020

Flexible cages are the sum of their parts: The complete structure of the COPII coat reveals numerous weak interactions cumulatively contribute to assembly

Selected by Nicola Stevenson

Categories: biochemistry


Anterograde transport between the endoplasmic reticulum (ER) and Golgi apparatus is mediated by transport carriers formed at specialised domains of the ER called ER exit sites (ERES). Generation of these carriers requires complex protein machinery, at the core of which is the COPII coat. In fact, COPII is so integral to this process its discovery, published in 19941, was recognised as part of the 2013 Nobel Prize in Physiology or Medicine awarded to James Rothman, Randy Schekman, and Thomas Südhof for defining vesicular traffic.

The COPII coat is a cage-like structure, which assembles in two layers on the cytosolic surface of the ER to recruit cargo and form transport carriers. During assembly, Sar1, a small GTPase, is recruited to the ER membrane by its GTPase activating protein Sec12. Here it inserts an amphipathic helix into the membrane following GDP-GTP exchange and recruits the Sec23/Sec24 subcomplex to form the inner coat. Polymerisation of the inner coat initiates membrane deformation for transport carrier generation, whilst Sec24 acts to recruit cargo to the bud site. Sec23, meanwhile, recruits the Sec13/31 subcomplex to form the outer coat layer and mature transport carriers are created.

Twenty-six years on from its discovery, research into COPII biology is as active as ever, including a recent burst of new ideas implicating liquid-liquid phase separation in COPII assembly2 and a renewed debate around the structure of the COPII-generated transport carriers3. Such prolific study has in part been fuelled by the continuing discovery of new COPII regulators, but also by key technological advances, not least those made in structural biology with the advent of electron cryo-tomography. In this study, Hutchings et al take full advantage of this technique to build on their previous model of COPII structure4 and define previously unknown interactions between COPII components within the fully assembled coat.

Key findings

In this study, purified yeast COPII proteins were incubated with giant unilamellar vesicles (GUV) to reconstitute fully assembled COPII coats in association with membranes. These were then analysed by cryo-electron tomography and subtomogram averaging to generate high resolution structures for evaluation. Note, COPII assembly pulls GUVs into tubular membrane structures with variable diameters in these conditions due to the presence of non-hydrolysable GTP.

The outer coat vertices, formed of the β-propeller domains of four adjacent Sec31 subunits, were resolved to 12 Å. Each β-propeller can be seen to form slightly different interactions with its neighbours, producing a two-fold rather than four-fold symmetry. The relative position of these domains is also shifted compared to the published structure of human COPII coats (solved by single particle cryo-EM OF Sec13/315), suggesting evolutionary divergence in COPII structure. At the Sec31 β-propeller interface, the authors also identify a novel negatively charged loop that eluded discovery in previous crystal structures due to its disordered nature. Deletion of the Sec31 β-propeller domains does not prevent inner coat assembly and GUV tubulation, but mutant yeast are not viable and vesicle budding is precluded.

The Sec13/31 rods connecting these outer coat vertices in a left-handed direction were resolved to 11Å. Here a previously un-resolved density was discovered and demonstrated to be the putative helical C-terminal domain of Sec31 (CTD). Sec31ΔCTD yeast were not viable unless secretory load was reduced and Sec31ΔCTD could only generate vesicles in vitro in the presence of non-hydrolysable GTP. Therefore the CTD may be important when coat is turning over. Visualisation of the Sec13/31 rods connecting the outer coat vertices in a right-handed direction (resolved to 13-15A) revealed the occasional presence of a Sec13/31 heterotetramer bridging two rods or a rod and a vertex instead of two vertices. These appeared to be used to resolve mismatches in the lattice.

The inner coat was resolved to 4.6Å, allowing the identification of interactions between the inner and outer coat. The WN residues of the active peptide of Sec31 and the Sec31 PPP region contact Sec23. A single density is also clearly visible extending from both sides of the prolines and bridging two adjacent inner coat subunits. Furthermore, a negatively charged region of 9-10 residues on the Sec23 surface was identified, which binds a positively charged disordered domain and the PPP domain of Sec31. The two coats therefore interact at multiple points. Overall, the positioning of structural features in the outer coat relative to those of the inner coat was found to be random, consistent with a flexible linkage between the two layers.

Neighbouring Sec23 subunits interact over a large surface area. At this interface, the authors map phosphorylation site S742, known to be important for regulating the coat6. A novel, flexible 17 residue L-loop is also identified on Sec23, which seems to become partially ordered upon interaction with Sar1. Deletion of this loop has no significant effect on membrane tubulation, however when paired with a Sec31ΔPPP, this results in multi-budded structures. Thus, when the inner coat is weakened, the outer coat can compensate by promoting a spherical morphology.

Overall these discoveries build up a picture of the COPII coat coming together through a complex network of partially redundant interactions which permits both strength and flexibility in the structure.



ERES can be highly pleiomorphic in structure, with budding membranes adopting a mixture of vesicular and tubular conformations. The COPII coat must therefore be flexible enough to encompass a range of membrane curvatures whilst maintaining the strength to shape the membranes into carriers. A number of the features described in this study go a long way towards explaining how this is achieved through the combined action of multiple redundant interactions and flexible protein regions. The links made between structure and function here are also important when we consider how these proteins interact in the crowded and complex environment of the ERES, which contain numerous regulators that act upon these interfaces. It will be exciting to see in the future how different cargo or regulator protein binding sites map onto and impact this structure as we build up an increasingly fuller picture of COPII biology.

I chose this preprint because of the technical advance as much as the biological one. This study is a really good example of how cryo-electron microscopy can be used to determine the structure of ever more complicated protein assemblies. Its ability to discern and describe flexible and disordered regions in proteins is also invaluable in our pursuit to understand the functional relevance of these domains.


Posted on: 8th July 2020


Read preprint (No Ratings Yet)

  • Author's response

    Giulia Zanetti shared

    1. Under the circumstances of your experiments, assembly of the COPII coat draws the GUV membranes into long tubules, however in cells, the membranes observed at ERES are much more heterogeneous. Have you been able to observe COPII structures on membranes with different curvatures – for example how does COPII look at the end of the tubules where the membrane may perhaps more closely resemble the curvature seen in ‘classic’ vesicles?

    We have indeed (see supplementary figure 3)! There are two main points to make in answering this very important question. 1. We see extensive tubulation in our reaction due to the fact that the coat is stabilised by using non-hydrolysable GTP analogues. The assembly interfaces we describe would be shorter lived in conditions of GTP hydrolysis, where coat is turning over.  We expect, as you suggest, that in physiological settings the dynamic coat assembly and disassembly would lead to a much more variable morphology.  2. Because in our conditions spherical coated membranes are relatively rare, we don’t have enough data to perform high resolution analysis of coat architecture. However, when manually picking and averaging outer coat vertices we notice these have a very similar shape to the ones we obtain from the tubular regions.


    1. Post-translational modifications of COPII components have been proposed to regulate the size of COPII carriers by affecting the structure of the coat. For example, Sec23 and Sec31 can be ubiquitinated. Would this be something you could look at in your model and do you have any predictions as to how any known modifications would affect the structure?

    This would be very nice to do. We have not attempted to perform ubiquitylation of Sec31 or Sec23 in vitro, but this is definitely a possible direction forward. We also do not know the site of ubiquitylation so are not able to predict how it would affect the assembled coat structurally.

    In general, we have noticed that most variations of the coat at assembly interfaces have little effect on budding, and we need to combine multiple defects to see any phenotype (supporting the idea that the coat functions through a complex and partially redundant interaction network). It is possible that, likewise, post-translational modifications would have only a mild effect in our in vitro system, but we will not know until we try.


    1. What effect may cargo have on this structure and is anything known about the cargo-Sec24 interface?

    This is a very interesting question and one that we are pursuing at the moment. We do not have an answer yet but we hope to reconstitute budding in the presence of cargo and analyse its effects. There are crystal structures of Sec24 bound to short cargo-derived peptides, and we have shown before that these binding sites are compatible with the polymerisation of the inner coat. Whether cargo affects coat architecture, and whether this forms part of a regulatory mechanism, we don’t know yet.


    1. Sec23 has a lot of interactors like the TRAPP complex, TANGO1/cTAGE5 etc despite the fact it forms the inner coat. Based on your model, do you think the flexible relationship between the inner and outer coat and the ‘gappy’ structure of the Sec13/31 coat is sufficient to allow interactors to reach through the outer coat to bind Sec23 or do you envisage such interactions to be limited to the edges of the coat?

    Regarding TANGO1/cTAGE5 (and Sec16), we believe they probably bind Sec23 before the outer coat does, and are eventually displaced as the coat grows. For what concerns TRAPP and other later factors such as TFG, this is a difficult one to answer, as we don’t really know much about the dynamics of uncoating. Does the outer coat ‘come off’ first? Or do these factors reach through the gaps? Cryo-tomography is possibly not the best technique to answer this, and we eagerly await dynamic data to shed light on the mysterious uncoating process!

    Have your say

    Your email address will not be published. Required fields are marked *

    This site uses Akismet to reduce spam. Learn how your comment data is processed.

    Sign up to customise the site to your preferences and to receive alerts

    Register here