To the best of our knowledge, this is the first report detailing ER membrane sheet organization in a native environment, i.e. unfixed, hydrated cells. To date, the reports of attempts to visualize the ultrastructure of the ER membranes included freeze fracture electron microscopy of the crystalloid ER (tubular OSER) in the UT-1 cells, which showed macromolecular complexes of approximately 10 nm in diameter [28
], and a freeze fracture/deep etch electron microscopy analysis of endomembranes of various cells, which revealed existence of 'bridges' connecting the membranes of different organelles [29
]. However, both reports had common problems inherent to the employed methodology: the electron microscopy was performed on the dehydrated and/or fixed specimen, after shadowing with platinum and carbon - the replicas were subjected to imaging. This implies that the observed structures could be artefacts, e.g. due to aggregation of material during the etching process, demeaning the value of those observations.
We used CEMOVIS methodology to preserve the native ultrastructure and to image the intracellular membranes of the HEK293 cells by cryo-EM at unprecedented level of detail. A sub-compartment of ER, OSER, was used as a model system, which allowed us to sample a sufficient amount of stacked ER sheet micrographs for analysis. Using projection images of the OSER membranes we have identified ordered cytosolic and luminal macromolecular arrays supporting ER membrane stacking. The imaging of the OSER membranes may not allow us to measure directly the interaction between OSER membranes and thus we may not unequivocally rule out the possibility that the ordering of the molecules is secondary to ER stack formation. However, we propose that the OSER membranes may be stabilized by these arrays.
We have shown that induction of ER sheet stacks by YFP-tagged calnexin overexpression is not driven by fluorescent protein dimerization. Monomeric fluorescent protein-tagged calnexin induces OSER formation; the stacked membranes are too far apart for direct in trans
interaction between overexpessed proteins. It is unlikely that the discrepancy between our results and those published previously [10
] resulted from usage of different cell types, cell culture conditions, reagents and DNA constructs. Mammalian cells were used throughout and similar vectors were used for transfections in both studies; the proteins employed to induce OSER (cytochrome b(5), Sec61[10
] and calnexin [15
]) are known to exist in a complex with each other [30
], and are therefore likely to reside in the same compartments. In addition, OSER-like structures have been observed in mammalian cells in the absence of any protein overexpression, with or without GFP involved [15
]. It is possible that under certain circumstances, depending on the structure of a particular fusion construct, the presence of GFP may indeed be conducive to trans
-interaction between the ER membranes leading to membrane tethering. It is worth noting that a broad range of ER-localized proteins have been known to induce OSER formation, including: cytochrome b5 [31
], HMG-CoA reductase [28
], inositol 1,4,5-trisphosphate receptor [32
], cytochrome P-450 [33
], microsomal aldehyde dehydrogenase [34
], torsinA [17
] and others. Therefore, it can be anticipated that more than one element in the structure of these very different proteins should affect their ability to cause rearrangements in the ER ultrastructure.
Several intracellular compartments, both in prokaryotes and eukaryotes, have been studied by CEMOVIS previously with a comparable level of detail [35
]. Ordered arrangement of trans-cleft molecular complexes in the neuronal synapses has been visualized using this technique [42
]. Recently published tomographic reconstruction of desmosomal vitreous sections revealed quasi-crystalline arrays of cadherin molecules connecting the skin cells [25
]. We propose a model whereby OSER membranes are stabilized in a similar way, i.e., adjacent ER membranes are connected by extended arrays of intracellular 'adhesion' molecules, although our results indicate that the OSER-stabilizing structures are more disordered, compared to the highly organized arrays of the desmosomes [25
A number of questions remain open, the most important being the identity of the proteins involved in OSER stabilization. We have ruled out the obvious candidates, represented by the coiled-coil domain-containing Climp63, and the LINC complex proteins that have been proposed to perform a similar role in the NE. This shows that, although the OSER membranes may originate from the NE at least in some cases [10
], the molecular machinery involved in OSER stabilization is likely different from that specialized in NE maintenance. Because the ordered areas could be observed between both luminal and cytosolic ER membrane faces, it is possible that more than one protein component may be involved in ER stacking interactions. It is possible that proteins involved are not integral membrane proteins, as is the case with the members of GRASP family (e.g., GRASP65), which are peripherally inserted into the Golgi membranes and stack the Golgi cisterns upon oligomerization [44
]. Also, it remains to be determined whether the formation of the OSER stacks is primed by the interactions between adjacent ER membranes via the ordered macromolecular complexes, or whether the protein crowding and array formation is preceded by the formation of a membrane stack, which is in turn brought about by an independent mechanism.