Discovery of lipid binding sites in a ligand-gated ion channel by integrating simulations and cryo-EM

Reviewer #1 (Public Review):

The authors use a combination of structural and MD simulation approaches to characterize phospholipid interactions with the pentameric ligand-gated ion channel, GLIC. By analyzing the MD simulation data using clusters of closed and open states derived previously, the authors also seek to compare lipid interactions between putative functional states. The ultimate goal of this work is to understand how lipids shape the structure and function of this channel.

The strengths of this article include the following:

1. The MD simulation data provide extensive sampling of lipid interactions in GLIC, and these interactions were characterized in putative closed and open states of the channel. The extensive sampling permits confident delineation of 5-6 phospholipid interaction sites per subunit. The agreement in phospholipid binding poses between structures and the all-atom MD simulations supports the utility of MD simulations to examine lipid interactions.

2. The study presents phospholipid binding sites/poses that agree with functionally-important lipid binding sites in other pLGICs, supporting the notion that these sites are conserved. For example, the authors identify interactions of POPC at an outer leaflet intersubunit site that is specific for the open state. This result is quite interesting as phospholipids or drugs that positively modulate other pLGICs are known to occupy this site. Also, the effect of mutating W217 in the inner leaflet intersubunit site suggests that this residue, which is highly conserved in pLGICs, is an important determinant of the strength of phospholipid interactions at this site. This residue has been shown to interact with phospholipids in other pLGICs and forms the binding site of potentiating neurosteroids in the GABA(A) receptor.

Weaknesses of this article include the following:

1. The authors describe in detail state-dependent lipid interactions from the MD simulations; however, the functional significance of these findings is unclear. GLIC function appears to be insensitive to lipids, although this understanding is based on experiments where GLIC proteoliposomes were fused to oocyte membranes, which may not be optimal to control the lipid environment. Without functional studies of GLIC in model membranes, the lipid dependence of GLIC function is not definitively known. Therefore, it is difficult to interpret the meaning of these state-dependent lipid interactions in GLIC.

2. It is unlikely that the bound phospholipids in the GLIC structures, which are co-purified from e. coli membranes, are POPC. Rather, these are most like PE or PG lipids. While it is difficult to accommodate mixed phospholipid membranes in all-atom MD simulations, the choice of POPC for this model, while practically convenient, seems suboptimal, especially since it is not known if PE or PG lipids modulate GLIC function. Nevertheless, it is striking that the overall binding poses of POPC from the simulations agree with those identified in the structures. It is possible that the identity of the phospholipid headgroup will have more of an impact on the strength of interactions with GLIC rather than the interaction poses (see next point).

3. The all-atom MD simulations provide limited insight into the strength of the POPC interactions at each site, which is important to interpret the significance of these interactions. It is unlikely that the system has equilibrated within the 1.7 microseconds of simulation for each replicate preventing a meaningful assessment of the lipid interaction times. Although the authors report exchange of up to 4 POPC interacting at certain residues in M4, this may not represent binding/unbinding events (depending on how binding/interaction is defined), since the 4 Å cutoff distance for lipid interactions is relatively small. This may instead be a result of small movements of POPC in and out of this cutoff. The ability to assess interaction times may have been strengthened if the authors performed a single extended replicate up to, for example, 10-20 microseconds instead of extending multiple replicates to 1.7 microseconds.

Reviewer #2 (Public Review):

The authors convincingly show multiple inner and outer leaflet non-protein (lipid) densities in a cryo-EM closed state structure of GLIC, a prokaryotic homologue of canonical pentameric ligand-gated ion channels, and observe lipids in similar sites during extensive simulations at both resting and activating pH. The simulations not only corroborate structural observations, but also suggest the existence of a state-dependent lipid intersubunit site only occupied in the open state. These important findings will be of considerable interest to the ion channel community and provide new hypotheses about lipid interactions in conjunction with channel gating.