Structure of the connexin-43 gap junction channel in a putative closed state

Reviewer #1 (Public Review):

Gap junctions, formed from connexins, are important in cell communication, allowing ions and small molecules to move directly between cells. While structures of connexins have previously reported, the structure of Connexin 43, which is the most widely expressed connexin and is important in many physiological processes was not known. Qi et al used cryo-EM to solve the structure of Connexin 43. They then compared this structure to structures of other connexins. Connexin gap junctions are built from two "hemichannels" consisting of hexamers of connexins. Hemichannels from two opposing cells dock together to form a complete channel that allows the movement of molecules between cells. N-terminal helices from each of the 6 subunits of each hemichannel allow control of whether the channels are open or closed. Previously solved structures of Cx26 and Cx46/50 have the N-termini pointing down into the pore of the protein leaving a central pore and so these channels have been considered to be open. The structure that Qi et al observed has the N-termini in a more raised position with a narrower pore through the centre. This led them to speculate whether this was the "closed" form of the protein. They also noted that, if only the protein was considered, there were gaps between the N-terminal helices, but these gaps were filled with lipid-like molecules. They therefore speculated that lipids were important in the closure mechanism. To address whether their structure was open or closed with respect to ions they carried out molecular dynamics studies, and demonstrated that under the conditions of the molecular dynamics ions did not traverse the channel when the lipids were present.

Strengths
The high resolution cryo-EM density maps clearly show the structure of the protein with the N-termini in a lateral position and lipid density blocking the gaps between the neighbouring helices. The conformation that they observe when they have solved the structure from protein in detergent is also seen when they reconstitute the protein into nanodiscs, which is ostensibly a more membrane-like environment. They, therefore, would appear to have trapped the protein in a stable conformational state.
The molecular dynamics simulations are consistent with the channel being closed when the lipid is present and raises the possibility of lipids being involved in regulation.
A comparison of this structure with other structures of connexin channels and hemichannels gives another representation of how the N-terminal helix of connexins can variously be involved in the regulation of channel opening.

Weaknesses
While the authors have trapped a relatively stable state of the protein and shown that, under the conditions of their molecular dynamics simulations, ions do not pass through, it is harder to understand whether this is physiologically relevant. Determining this would be beyond the scope of the article. To my knowledge there is no direct evidence that lipids are involved in regulation of connexins in this way, but this is also an interesting area for future exploration. It is also possible that lipids were trapped in the pore during the solubilisation process making it non-physiological. The authors acknowledge this and they describe the structure as a "putative" closed state.
The positions of the mutations in disease shown in Figure 4 is interesting. However, the authors don't discuss/speculate how any of these mutations could affect the binding of the lipids or the conformational state of the protein.

It should also be noted that a structure of the same protein has recently been published. This shows a very similar conformation of the N-termini with lipids bound in the same way, despite solubilising in a different detergent.

Reviewer #2 (Public Review):

The manuscript from Qi et. al. provides novel structures for connexin 43 (Cx43) gap junction channels (GJCs) and hemichannels, which they claim correspond to the closed conformations of these channels. This leads the authors to propose a mechanism of gating that implicates the existence of lipids in the pore, which could stabilize the N-terminal domain as the gate region within the pore. The authors performed a lipidomic assay in their structures and identified a dehydroepiandrosterone (DHEA), a sterol compound specifically enriched in their Cx43 purified samples. However, at the current structural resolution, they cannot conclude whether DHEA is the small lipid-like density found at the pore of closed channels. Further studies, including functional studies, are needed to determine whether DHEA is a gating intermediary. Interestingly, other recently published structures of large-pore channels support the notion that lipids are found inside the pore. However, this evidence is only supported by Cryo-EM structures and is an issue generating major controversy in the field, particularly when these molecules are implicated in the gating mechanisms. The finding of putative lipids-pore interactions is a very intriguing observation, but it should be interpreted carefully. A major concern is that channel reconstitution is performed in an excess of lipids and detergents that could lead to artifacts. Thus, these lipid-like densities observed in Cx43 (and other structures) after single particle analysis could not represent native lipid-protein interactions. Subsequently, all conclusions for the role of lipids in gating could rely on a potential protein purification-induced artifact. Also, it is hard to visualize how the lipids can move in/out of the pore during gating, particularly from this putative lipids-pore conformation to an open conformation.

Another important aspect of this work is that provided structures for both Cx43 GJCs and hemichannels. As expected, there are differences in extracellular loops rearrangements between these two structures. One issue, however, is that the resolution for Cx43 hemichannels is still low (3.98 Å), thus interpretations need to be taken with caution. In addition, the intracellular domains that are important for the gating and regulation of Cx43, including the intracellular loop and the carboxyl-terminal domain were not resolved in these structures. Nevertheless, this is a common issue for other connexin Cryo-EM structures reported in the literature.