Spontaneous Lipid Binding to the Nicotinic Acetylcholine Receptor in a Native Membrane
AAIP/123-QED
Spontaneous Lipid Binding to the Nicotinic Acetylcholine Receptor in a NativeMembrane
Liam Sharp and Grace Brannigan
1, 2 Center for Computational and Integrative Biology, Rutgers University-Camden,Camden, NJ Department of Physics, Rutgers University-Camden, Camden,NJ
The nicotinic acetylcholine receptor (nAChR) and other pentameric ligand-gated ion chan-nels (pLGICs) are native to neuronal membranes with an unusual lipid composition. Whileit is well-established that these receptors can be significantly modulated by lipids, the un-derlying mechanisms have been primarily studied in model membranes with only a fewlipid species. Here we use coarse-grained molecular dynamics (MD) simulation to probespecific binding of lipids in a complex quasi-neuronal membrane. We ran a total of 50microseconds of simulations of a single nAChR in a membrane composed of 36 species oflipids. Competition between multiple lipid species produces a complex distribution. Wefind that overall, cholesterol selects for concave intersubunit sites and PUFAs select forconvex M4 sites, while monounsaturated and saturated lipids are unenriched in the nAChRboundary. In order to characterize binding to specific sites, we present a novel approachfor calculating a “density-threshold affinity” from continuous density distributions. Wefind that affinity for M4 weakens with chain rigidity, which suggests flexible chains mayhelp relax packing defects caused by the conical protein shape. For any site, PE head-groups have the strongest affinity of all phospholipid headgroups, but anionic lipids stillyield moderately high affinities for the M4 sites as expected. We observe cooperative ef-fects between anionic headgroups and saturated chains at the M4 site in the inner leaflet.We also analyze affinities for individual anionic headgroups. Combined, these insightsmay reconcile several apparently contradictory experiments on the role of anionic phos-pholipids in modulating nAChR. 1 a r X i v : . [ c ond - m a t . s o f t ] F e b . INTRODUCTION The nicotinic acetylcholine receptor (nAChR) is a well studied excitatory pentameric ligandgated ion channel (pLGICs). nAChRs are found at high density in post-synaptic membranesand the neuromuscular junction in mammals, and the electric organ in
Torpedo electric rays.The nAChR is activated by the binding of agonists such as nicotine or acetylcholine to the or-thosteric site in the extra-cellular domain (ECD). When post-synaptic nAChRs are activated en-mass they stimulate an action potential. Thus nAChRs play a critical role in both cogni-tion and memory and neuromuscular function . nAChR and the greater pLGIC superfamilyplay various roles in neurological diseases related to inflammation , addiction , chronic pain ,Alzheimer’s Disease ,spinal muscular atrophy , schizophrenia and neurological autoim-mune diseases .nAChRs are highly sensitive to their local lipid environment. nAChR poorly conducts ionsin model phosphatidylcholine (PC)-only membranes, but can conduct a current with the addi-tion of cholesterol or anionic lipids , though too much cholesterol can also cause a loss offunction . Functional studies using Xenopus oocytes require lipid additives such asasolectin or lipids from synaptic membranes to recover native levels of nAChR ionflux. Understanding the mechanism of modulation requires understanding how and where themodulating lipid interacts with the receptor, and these interactions may themselves be dependentupon the rest of the lipid composition.Mammalian neuronal membranes have unique compositions compared to other mam-malian membranes . Neuronal membranes are more similar to the membrane of the Tor-pedo electric ray’s electric organ than the average mammalian membrane . The neuronalmembrane is rich in lipids in which one or both chains are polyunsaturated fatty acids (PU-FAs), particularly the n − n − ∼ −
25% of theacyl chains of neuronal phospholipids, and are involved in secondary signaling and neuronaldevelopment . PUFAs are linked to a number of neurological diseases and disorders that over-lap nAChR related diseases. PUFAs play a roll in major depressive and bipolar disorder ,schizophrenia , and Alzheimer’s Disease .Functional experiments have focused on the role of anionic lipids and cholesterol as modulatorsof pLGICs (the role of polyunsaturation has received comparatively little attention due2o common challenges with oxidation of polyunsaturated chains). Such experiments have beenoverwhelmingly consistent with a role for direct binding of lipids as a modulatory mechanism.As for most membrane proteins, it is experimentally challenging to capture the boundary lipidcomposition of pLGIC because lipids are small molecules that may remain partly fluid even in theirbound state. Numerous structures of pLGICs have revealed a conserved arrangement for both theTMD and the ECD. In the TMD, each subunit has four membrane helices (M1-M4) with the fivesubunits forming a “star” shape around a central pore (Figure 1A). The M2 helix lines the pore, theM1 and M3 helices form a middle ring that includes the intersubunit cavities, and the M4 helicesform the tip of the star. Structural methods have resolved potential cholesterol molecules andphospholipids bound to subunit interfaces, but crystallographic disorder introduced by lipidstypically precludes identification of lipid species. Mass spectrometry has revealed specific bindingof anionic lipids, with additional mutagenesis studies suggesting localized sites in the inner leafletnear the M4 helices. Molecular dynamics (MD) simulations are particularly useful for visualizing and characteriz-ing microscopic interactions within a fluid system. Given a putative cholesterol or lipid bindingmode, atomistic simulations can be used to probe stability of the lipid binding mode. For pen-tameric channels, such approaches have primarily demonstrated stability of bound cholesterol ,particularly at intersubunit sites . Unfortunately, fully atomistic simulations suffer from slowdiffusion of lipids within the membrane, which prevents spontaneous lipid sorting by proteins overaccessible simulation time scales.Coarse-grained MD simulations use simplified molecular models that can reveal sponta-neous lipid sorting, domain formation, and protein partitioning over simulation timescales .Coarse-grained MD simulations have been used previously to probe interactions of pLGICs withpropofol as well as spontaneous lipid binding in model membranes . In previous work,we found that nAChR embedded in multiple domain-forming model membranes partitioned tothe PUFA-rich liquid disordered domain , rather than to the cholesterol-rich liquid-ordered or“raft” domain that was suggested by cholesterol modulation. We observed that cholesterol stilloccupies embedded sites on the nAChR TMD, where it is shielded from the liquid disordereddomain. However, native membranes are primarily composed of heteroacidic lipids with twodistinct chains, where each chain has a different domain preference; such lipids will naturallydestabilize domains. In non-domain forming model membranes composed of heteroacidic lipids,two classes of five-fold symmetric sites emerged: an intersubunit site and the M4 site (Figure3B). Cholesterol and saturated chains were enriched at the inter-subunit interfaces and n-3 PUFAacyl-chains were enriched around the M4 helices . These results were consistent with binding tominimize packing defects: the rigid lipids could fill in the concave regions at the intersubunit siteswhile the flexible chains would easily deform around the “star points” of the M4 helices. Yet itwas not clear whether these same patterns would be upheld in the more complex environment ofa native neuronal membrane, which has many more options for minimizing any packing defect.Neuronal membranes also contain a sizeable fraction of anionic lipids in the inner leaflet .With collaborators in the Cheng lab, we recently showed that anionic headgroups bind prefer-entially to the pLGIC Erwinia ligand-gated ion channel (ELIC), when the same acyl chains areused for both headgroups. Through coarse-grained MD, we found specific binding sites for 1-palmitoyl-2-oleoyl phosphatidylglycerol POPG in the intersubunit sites (inner leaflet); these sitescontained basic amino acids that were also implicated through mutagenesis . In nAChR the high-density of basic amino acids are in the M4 site (inner leaflet) rather than the intersubunit site (innerleaflet), so we would expect a shift for nAChR even in model membranes, due purely to the proteinsequence. The relative roles of headgroup charge vs acyl chain saturation in driving affinity areunknown.The use of complex quasi-realistic membranes in coarse-grained MD simulations is growingmore feasible. In 2014, Ingólfsson et al simulated an “average mammalian” membrane contain-ing 63 lipids species, followed in 2017 by a coarse-grained neuronal membrane . Multiple ac-cessible and realistic membranes have been developed for comparison of protein-lipid interactionsbetween model and quasi-native membranes . To our knowledge, no such coarse-grainedMD simulations using quasi-native membranes have been used with pLGICs.While the model membranes we used previously are useful for identifying putative sites, theyhave critical limitations. As stated previously, model membranes typically vary headgroup chargeor acyl chain saturation, not both. Model membranes also do not allow for identification of morespecific chemical variations within general saturation classes (i.e. n-3 PUFAs like DHA vs n-6PUFAs like α -linolenic acid) or like-charged head groups (PC vs PE, or phosphoserine (PS) vs phosphoinositol (PI)). For this work, we embed the neuromuscular nAChR in a coarse-grainedneuronal membrane . To test whether the predictions we developed from model membranes holdfor native membranes, we develop a new method for quantifying affinities for partially-occupiedbinding sites.The remainder of this paper is organized as follows. Section II presents our simulation and4nalysis approach, including introduction of the density-threshold affinity. Section III presentsresults and discussion of site selectivity of neutral lipids, followed by a reoriented discussion of thesame data that is focused on lipid preferences of individual sites. We then consider selectivity ofanionic lipids in the inner leaflet and finally consider the effects of specific headgroup differences.Section IV concludes. II. METHODSA. Simulation Composition
All simulations used the coarse-grained MARTINI 2.2 topology and forcefield. nAChR co-ordinates were based on a cryo-EM structure of the αβ γδ muscle-type receptor in native torpedomembrane (PDB 2BG9 ). This is a medium resolution structure (4Å) and was further coarse-grained using the martinize.py script; medium resolution is sufficient for use in coarse-grainedsimulation, and the native lipid environment of the proteins used to construct 2BG9 is criticalfor the present study. The secondary, tertiary and quaternary structure in 2BG9 was preservedvia soft backbone restraints during simulation as described below, so any inaccuracies in localresidue-residue interactions would not cause instability in the global conformation.nAChR was embedded in a coarse-grained neuronal membrane based on Ingólfsson et al . Theneuronal membrane from described by Ingólfsson contains phospholipids, sterols, diacylglycerol,and ceramide. Membranes presented in this paper only consider phospholipids and cholesterol,for a total of 36 unique lipid species, see Table SI 1.Coarse-grained membranes were built using the MARTINI script insane.py , which was alsoused to embed the coarse-grained nAChR within the membrane. The insane.py script randomlyplaces lipids throughout the inter- and extra-cellular leaflets, and each simulation presented in thismanuscript was built separately. All simulation box sizes were 40 x x
35 nm with ∼ , − ,
000 lipids and total ∼ ,
000 beads.
B. Simulations
Molecular dynamics simulations run using the MARTINI 2.2 forcefield and GROMACS ε r =15 and electrostatic cutoff length at 1.1 nm. Energy minimization was5erformed for 1000000 steps, but energy minimization tended to concluded after ∼ − × − bar − and a pressure coupling constant set to 3.0 psfor the NPT ensemble.Molecular dynamics simulations were run using a time step of 20 fs for 5 µ s for 10 replicas.Simulations were conducted in the NPT ensemble, by using the velocity rescaling to a tempera-ture of 323 K with a coupling constant set to 1 ps. Semi-isotropic pressure coupling was set toParrinello-Rahman with compressibility at 3 × − bar − and pressure coupling constant set to3.0 ps.Secondary structures restraints with MARTINI recommendations were constructed by themartinize.py script and imposed by GROMACS . The nAChR conformation was preservedby harmonic bonds between backbone beads separated by less than 0.5 nm and calculated usingthe ElNeDyn algorithm associated with MARTINI with a coefficient of 900 kJ · mol − . a Outer Leaflet Inner Leaflet Å Å M1M2M3M4 α β δ α c P ( N ) N0 P site P bulk P > P < N peak b M4Intersubunit M4 Intersubunit ɣ FIG. 1. Binding site boundaries and distribution definitions. (a) Structure of the nAChRTMD , viewedfrom the extracellular domain. Helices are colored by subunit ( α :green, β :purple, γ : cyan, δ :grey). (b)Boundaries of the pseudo-symmetric intersubunit (black) and M4 (white) sites. The angular componentsare determined by the location of the M1 and M3 alpha-helices for either two adjacent subunits (intersubunitsites) or a single subunit (M4 sites), and are listed in Table S5. Circles correspond to the helices shown inpanel A. (c) The distributions P site ( n ) (blue) and P bulk ( n ) (dashed red) represent the probability distributionsfor number of beads of a certain lipid species in the site or in an analogously-sized area of the bulk, respec-tively. The value n peak maximizes P bulk . P < (pink) is the area under P site to the right of n peak while P > (lightblue) is the area under P site to the right of n peak . . Calculation of Polar Density Distributions As in our previous work , the two-dimensional density distribution ρ B of the beads withina given lipid species B around the protein was calculated on a polar grid: ρ B ( r i , θ j ) = (cid:10) n B ( r i , θ j ) (cid:11) r i ∆ r ∆ θ (1)where r i = i ∆ r is the projected distance of the bin center from the protein center, θ j = j ∆ θ isthe polar angle associated with bin j, ∆ r = 10Å and ∆ θ = π radians are the bin widths in theradial and angular direction respectively, and (cid:10) n B ( r i , θ j ) (cid:11) is the time-averaged number of beadsof lipid species B found within the bin centered around radius r i and polar angle θ j . In order todetermine enrichment or depletion, the normalized density ˜ ρ B ( r i , θ j ) is calculated by dividing bythe approximate expected density of beads of lipid type B in a random mixture, x B s B N L / (cid:104) L (cid:105) ,where s B is the number of beads in one lipid of species B, N L is the total number of lipids in thesystem, and (cid:104) L (cid:105) is the average projected box area:˜ ρ B ( r i , θ j ) = ρ B ( r i , θ j ) x B s B N L / (cid:104) L (cid:105) (2)where the expected density is derived at the first frame of the simulation. Python software forthese calculations are under active development and are located at .This expression is approximate because it does not correct for the protein footprint or anyundulation-induced deviations of the membrane area. The associated corrections are small com-pared to the membrane area and would shift the expected density for all species equally, withoutaffecting the comparisons we perform here. For a given lipid species or class, analysis excludedany replicas in which fewer than 5 lipids of the species/class were in the leaflet at any point in thesampled simulation. D. Calculation of the density-threshold affinity
Although lipids to occupy clearly detectable hot-spots, binding to these sites are not straight-forward to describe by a traditional two-state model. Lipids are chains that may partially occupyor fully occupy a site, and they may share a site with another lipid that is partly or fully occupyingthe site. While the standard affinity can be determined from the probability of single occupancy,7he density-threshold affinity is determined from the probability that a site is occupied by morebeads than would be expected based on bulk density.For a given site, consider two probability distributions: the probability P site ( n ) of finding n beads within the site and the probability P bulk ( n ) of finding n beads within a region of equivalentarea in the bulk, respectively. For a lipid that has no affinity for this binding site, we expect P site ( n ) = P bulk ( n ) , while P site ( n ) should be right-shifted for a strong affinity and left-shifted in thepresence of competition. We calculate the degree of right or left shift by first finding the numberof beads n peak that corresponds to the peak of the density distribution in the bulk. As illustrated inFigure 1 C, we then integrate P site on both the left and right side of the threshold n peak to yield P < and P > respectively: P < ≡ ∑ n ≤ n peak P site ( n ) (3) P > ≡ ∑ n > n peak P site ( n ) (4)Note that this step breaks the distribution into two macrostates on either side of the threshold,allowing clear analogy with a classic binary binding model. The free energy difference betweenthe two macrostates is ∆ G = − RT ln P > P < (5)where R is the gas constant and T is temperature. We term this free energy difference the “density-threshold affinity”. In the special case of binary occupancy, P site ( n ) = ( + K D / [ L ]) − , if n = ( + [ L ] / K D ) − , n = K D is the dissociation constant and [ L ] is the ligand concentration. In a dilute solution thevolume per ligand is typically much larger than the site volume, so P bulk ( n ) = n = n >
0, so n peak =
0. Consequently, for this special case, P < = ( + [ L ] / K D ) − and P > = ( + K D / [ L ]) − . Then Equation 5 reduces to the classic form for the chemical potential RT ln K D − RT ln [ L ] . 8 . Binding Site Definition and Occupancy Calculations We consider two classes of site: intersubunit sites and M4 sites. Each pLGIC has ten of eachsite (five in the outer leaflet and five in the lower leaflet) for a total of twenty sites (Figure 1B). Theboundaries for each site were drawn to correspond to the localized binding hot spots observed forheteroacidic membranes , and are non-overlapping. Inter-subunit sites include bins with angularcomponents between the M1 and M3 alpha-helices of two adjacent subunits, and a radial compo-nent satisfying 10 < r ≤ < r ≤ P site ( n ) , a distribution was taken across frames at 10 ns intervals. Forany frame, the beads of a given lipid or chain type were binned onto a fine polar grid with ∆ r = ∆ θ = π . The bins falling within the site boundaries were then summed to calculate theoccupancy n . This approach allowed for straightforward adjustment of site boundaries if neededwithout needing to re-bin the whole trajectory. F. Calculation of Accessible Area
Calculation of P bulk requires determining the accessible site area in order to calculate the densi-ties in a bulk region of similar area. The area A accessible to the lipids is the difference between thetotal site area A tot and the area A ex excluded by the protein: A = A tot − A ex A tot is straightforwardto calculate by summing over the areas of the bins i within the site boundaries: A tot = ∑ i r i ∆ r i ∆ θ i .Calculating A ex is less straightforward, and although there are many possible ways to do this, forself-consistency we used the same tools from our primary analysis.In a single lipid membrane, P site ( n ) = P bulk ( n ) as long as P bulk ( n ) is calculated using the properarea A . We exploit this identity to calculate A for each site, by running a single nAChR in puredi-palmitoyl phosphatidylcholine (DPPC) for ∼ A for each sitesuch that P site ( n ) and P bulk ( n ) have the same peak. These areas are reported in Table SI 4.9 µs 1 µs 2 µs3 µs 4 µs 5 µs Inter subunitM4
Cholesterol Monounsaturatedn-3 PUFA Saturated a b
FIG. 2. A molecular perspective of coarse-grained simulation results. a) Multiple frames from a singlesimulation replica over 5 µ s . The nAChR TMD is shown in surface representation and colored as in Figure1. Cholesterol and acyl chains within 15 Å of nAChR are shown as beads, and colored by chain type:saturated lipids: blue, monounsaturated lipids:orange, n-6 PUFAs:pink, n-3 PUFAs: beige, and cholesterol:red. Each phospholipid color includes several lipid species of the same type, and simulations included alarger membrane and the ECD, which is not shown. b) Representative poses of lipids for individual sites,colored as in A, but viewed from within the membrane looking at the TMD surface. Cholesterol selects forthe intersubunit site while monounsaturated lipids have a particularly low affinity for this site. PUFAs selectfor the M4 site, while saturated lipids have a particularly low affinity. III. RESULTS AND DISCUSSIONA. Effect of acyl chain on site selectivity among neutral lipids
Representative frames from a typical trajectory of boundary lipids are shown in Figure 2A,with representative poses shown in Figure 2B. In order to quantitatively compare the lipid distri-butions for the native system to our previous model system, we plotted the enrichment of boundarydensity relative to bulk density on a two-dimensional polar heat map centered around the protein.This enrichment is shown in Figure 3A for cholesterol and various acyl chains grouped by satu-ration. Saturated and monounsaturated acyl chains are not significantly depleted or enriched inthe boundary of the protein. Regions of cholesterol density are much more localized than in themodel membrane (Figure 3C) , with pockets of high enrichment very close to the protein andweak depletion in the remainder of the boundary region. Both n-6 and n-3 PUFA chains yieldfive-fold symmetric enrichment around the M4 alpha-helices, as observed for n-3 PUFAs in themodel membrane. In the neuronal membrane, however, this enrichment is less well-defined and10 uter Inner SatCholMonon-6 PUFAn-3 PUFA NeutralAnionic Outer Inner a b
Heteroacids HomoacidsSaturatedPUFACholesterol log ( ˜ ρ ) log ( ˜ ρ ) c SatCholn-3 PUFA Heteroacids HomoacidsSaturatedPUFACholesterol log ( ˜ ρ ) log ( ˜ ρ ) Heteroacids HomoacidsSaturatedPUFACholesterol log ( ˜ ρ ) log ( ˜ ρ ) FIG. 3. Lipid density enrichment around a central singular nAChR. (a) and (b) Density enrichment ˜ ρ a for lipids in a neuronal membrane, calculated using eq 2 for both outer and inner leaflets, averaged over 10replicas for 2.5 µ s each. The maximum radius from the nAChR pore is 60 Å. Depletion relative to a randommixture (log ˜ ρ a <
0) is blue while enrichment (log ˜ ρ a >
0) is red. Lipids are organized by acyl chain (a) orheadgroup (b). Acyl chain density includes only the relevant chain of a heteroacidic lipid, while headgroupdensity includes the whole lipid. Helices are represented as circles colored as in Figure 1. Intersubunit(solid line) and M4 (dashed line) site boundaries are marked. (c) Equivalent analysis for nAChR in a modelmembrane of 2:2:1 n-3 PUFA:saturated:cholesterol, using previously published trajectories . spreads into the intersubunit regions. In particular, additional pockets of significant enrichmentare apparent in the β − δ subunit interface in the outer leaflet. The overall area of the regions ofPUFA-enrichment decrease in the inner leaflet, where n-3 PUFAs are enriched around M4 helices,but n-6 PUFA density is not five-fold symmetric and has weak enrichment. Overall, the loss ofdefinition in site boundaries diverges from the well-defined five fold enrichment for n-3 PUFAswe saw in model membranes .In order to reduce these distributions to affinities that are more straightforward to interpret,we calculated the density-threshold affinity ∆ G for various lipid species as defined in Eq. 5. Weorganize this information in two different ways: Figure 4 provides the “lipid’s perspective” and isorganized to identify the preferred site for a given lipid type (the lipids’ “site selectivity”), while11 at Mono n-6 n-3 Δ G ( kc a l / m o l ) Δ G ( kc a l / m o l ) N eu t r a l A n i on i c Chol
Outer Inter-Subunit Outer M4 Inner Inter-Subunit Inner M4
FIG. 4. Density-threshold affinities organized to reveal site selectivity. The density-threshold affinities( ∆ G )are calculated using Equation 5, where error bars are the standard error (n=10 independent replicas).Density-threshold affinitiesare colored by site; in the outer leaflet: intersubunit (blue) and M4 (green), andfor the inner leaflet: intersubunit (orange) and M4 (red). Values are separated by headgroup charge (rows)and acyl chain type (columns). More negative values indicate stronger affinities, while more positive valuesindicate more displacement of the lipid by other lipid species. Data incorporates 10 replicas averaging overthe last half of the 5 us trajectory, with five-fold averaging over each type of pseudosymmetric site. Figure 5has an alternate representation of the same data. Figure 5 provides the “site’s perspective” and is organized to identify the most favorable lipids fora given site (the sites’ “lipid specificity”).We first consider site selectivity for neutral lipids. Affinities for neutral lipids and cholesterolare shown in Figure 4A, where more negative values of ∆ G indicate a stronger density-thresholdaffinity and more positive values indicate more displacement by other lipids. Overall, as shownin Figure 4A, saturated lipids have similar density-threshold affinities across all sites, which isconsistent with the generally flat distribution observed in Figure 3. Yet saturated lipids do yielda slightly stronger affinity for intersubunit sites, at least in the outer leaflet, which may drive thehigh amount of saturated enrichment observed at these sites in model membranes. Outer leafletmonounsaturated lipids are slightly more unfavorable in intersubunit sites than M4 sites, and thisdifference grows in the inner leaflet.In contrast to saturated and monounsaturated lipids, PUFAs and cholesterol are highly selectivefor particular sites. As shown in Figure 4A, neutral PUFAs have significantly stronger affinities for124 sites than for innersubunit sites in the same leaflet. Such selectivity is consistent with the PUFAenrichment density in Figure 3A, where n-3 PUFAs can occupy most regions of the TMD but haveparticularly high levels of enrichment around M4. It is further consistent with our expectationsfrom model membranes (Figure 3C). Regardless of the site class, PUFAs favor the outer leaflet siteover the inner leaflet site, but both sets of M4 sites are more favorable than both sets of intersubunitsites. Conversely, cholesterol has significantly stronger affinities for innersubunit sites than for M4sites, which is also consistent with the enrichment density in Figure 3A and our expectations frommodel membranes (Figure 3C). For cholesterol, however, the leaflet is a bigger determinant ofaffinity than the site; cholesterol has a stronger affinity for either outer leaflet site compared toeither inner leaflet site. B. Lipid preferences of intersubunit and M4 sites
We now switch perspectives to considering which neutral lipids are most favorable for partic-ular sites. As shown in Figure 5 A and B, intersubunit sites in both leaflets prefer cholesterolto phospholipids, which is expected based on the results from model membranes. Upon visualinspection, this result may appear to diverge from the cholesterol polar density plots in neuronalmembranes (Figure 3 A). The present results show that while the overall footprint of cholesterolenrichment in (Figure 3 A) is small, this small region actually reflects a tight and persistentlyoccupied binding site. The highly right-shifted distributions for cholesterol are shown in FigureS1.PUFA chains yield affinities for the intersubunit site that are approximately > > n-6 > saturated > monounsaturated. Thus, even though PUFAchains prefer M4 sites to intersubunit sites, and saturated chains prefer intersubunit sites to M4sites, PUFAs have a stronger affinity for either site type than do saturated lipids.For intersubunit sites, monounsaturated lipids have the weakest affinities ( > . > n-6 > monounsaturated > saturated. This is consistent with a rolefor PUFAs in minimizing unfavorable membrane deformations caused by the pLGIC’s conical-starshape. Surprisingly, cholesterol had a stronger affinity for M4 sites than any acyl chains otherthan n-3 PUFAs. Cholesterol is rigid, small, and has asymmetric sides (rough and smooth) whichpotentially allows it to embed between alpha-helices and compete with n-3 PUFAs for binding.Any cholesterol bound within the grooves of the subunit interface (as hypothesized based on atom-istic simulations and observed in β subunits of nAChR (using coarse-grained simulations ), willalso get counted within the M4 site. Δ G ( kc a l / m o l ) Δ G ( kc a l / m o l ) O u t e r Lea f l e tI nne r Lea f l e t Inter-Subunit Site M4 Site
Neutral Anionic Neutral Anionic
Saturated Monounsaturated n-6 PUFA n-3 PUFA Cholesterol ab cd
FIG. 5. Density-threshold affinities organized to reveal lipid preferences by site. Data shown is identicalbut reorganized and recolored from Figure 4. Here, density-threshold affinitiesare colored by chain type(Saturated:blue, Monounsaturated:pink, n-6 PUFAs:orange, n-3 PUFAs:tan, Cholesterol:red), and separatedby leaflet (rows) and site (columns). . Effect of Head Group Charge on Affinity Depends on Leaflet and Binding Site Figure 3b compares the density enrichment for anionic headgroups with that of neutral head-groups. Data is shown for the inner leaflet only, because anionic lipids are not present in the outerleaflet at the start of simulations and few anionic lipids flip flop to the outer leaflet.In the inner leaflet, the anionic lipids are expected to select for sites that are lined with basicamino acids, which are in different locations depending on subunit (Figure ?? ) As shown in Figure3b, anionic lipids are generally enriched around the M3/M4 helices for the α γ , γ , δ , and β subunits.Anionic lipids are enriched at intersubunit sites and around M4 sites for all subunits but the α subunits. Non- α nAChR subunits have basic amino acids closer to M4 alpha-helices, as shownin Figure SI 2a. We incorporate data from all five pseudo-symmetric sites to obtain the density-threshold affinities reported in Figure 4B, which suggest that anionic lipids have significantlystronger affinities for M4 sites on average. The average anionic affinity difference between inter-subunit and M4 sites is ∼ − . ∼ ABLE I. Density-threshold affinities() of neutral lipids for both sites in the outer leaflet, by head group.Errors are standard errors (n=10 independent replicas).
Intersubunit Sites M4 Sites ∆ G (kcal/mol) ∆ G (kcal/mol)PE -0.2 ± ± ± ± D. Role of Individual Lipid Headgroups in Determining Affinity
Neutral and anionic are bulk terms that categorize numerous lipid head-groups by charge. Tounderstand the role of the chemical distinctions between head groups of like charge, we brokethe headgroup affinities down by headgroup species in Table I. In the outer leaflet, lipids containa mixture of PE and PC headgroups. The small neutral PE head group has the strongest affinityacross all headgroups for both inter-subunit and M4 sites, -0.2 ± ± ∼ > . ,so it is possible that this affinity simply reflects the high affinity of PUFA chains. However, evenfor identical chains, both experimental and simulation data suggests stronger PE-ELIC than PC-ELIC interactions.Table II shows specific head group affinities in the inner leaflet. As in the outer leaflet,lipids with PE headgroups still have the strongest affinity of all lipids, but in the inner leafletwe are also able to distinguish affinities for anionic species. For the intersubunit site, PI,PS, and PC have similar affinities (within statistical error), and have significantly strongeraffinities for these sites than the phosphoinositides (PIPS) PIP1, PIP2, PIP3, which have asignificantly stronger affinity than phosphatidic acid (PA). Thus, from strongest to weakest,PE > PI ∼ PS ∼ PC >> PIP1 ∼ PIP2 ∼ PIP3 >> PA for the intersubunit site. In contrast, at the M4site, more significant differences among moderate affinity headgroups emerge. PI has sig-nificantly stronger affinity than PS (a difference of 0.3 ± ± > PI > PS > PC >> PIP1 ∼ PIP2 ∼ PIP3 ∼ PA for the M4 site.16
ABLE II. Density-threshold affinities( ∆ G) of neutral lipids for both sites in the inner leaflet, by head group.Values are sorted by strength of affinity for intersubunit sites. Errors are standard errors (n=10 independentreplicas).
Inner Inter Sites Inner M4 Sites ∆ G (kcal/mol) ∆ G (kcal/mol)PE 0.3 ± ± ± ± ± ± ± ± . ± ± ± ± ± ± ± ± IV. CONCLUSIONS
Using coarse-grained simulations of the nAChR within a quasi-neuronal membrane containingover thirty lipid species, we have observed spontaneous lipid binding and quantified lipid speci-ficity for two types of sites in the protein TMD. These two site classes represent the most concave(intersubunit site) and convex (M4 site) portions of the star-shaped nAChR and were initially ob-served as “hot spots” in our previous simulations of model membranes. Compared to classicligand binding sites, these sites are superficial and have a large volume. The “ligands” occupyingthem are also non-traditional: lipids are flexible chain molecules that may only partially occupythe site and are likely to share the site with other partially-occupying ligands. While our labhas developed promising alchemical approaches for calculating traditional affinities of atomisticlipids for more highly localized, well-defined sites, these hot spots required a different approach.Here we have proposed a softer “density-threshold affinity” for characterizing these affinities fromspontaneous, unbiased coarse-grained simulations. While we restrict the use of this method here tonAChR, it should be straighforward to extend to any other transmembrane proteins with detectableregions of density enrichment.Our results are summarized graphically in Figure 6. Based on our results from model mem-branes, we had hypothesized that PUFAs would select for the convex M4 sites and that raft-forming lipids like cholesterol and saturated lipids would select for the concave inter-subunit sites.Overall, our results were consistent with this expectation. Yet although lipids containing PU-FAs do prefer the M4 site to the intersubunit site, their affinity for even the intersubunit sites arestronger than that of all other phospholipids. This result underscores the reliable partitioning of17 uter Leaflet Inner Leaflet SaturatedPUFA Cholesterol Neutral Anionic
FIG. 6. Cartoon of expected boundary lipids for the nAChR in a native membrane for both leaflets. Proteinis shown in the center of both leaflets in a cyan floral shape. Grey and black outlines depict sites favorablefor neutral and anionic lipids respectively. Fill color represents the lipids most likely to occupy each site(red: cholesterol, blue: saturated, beige: PUFA) and outline represents headgroup charge (gray: neutral,black: anionic). nAChR to PUFA-rich, liquid-disordered domains that we observed in homoacidic, domain formingmembranes , and suggests PUFAs may have been absent from the intersubunit site in heteroacidicmembranes because of the constraints of the lipid topology. In the latter simulations, all lipidscontained one saturated chain and one PUFA chain, so binding of the PUFA chain to its preferredM4 site requires the saturated chain to find the most favorable location nearby (in the intersubunitsite) and may block binding of other PUFA chains to that site. These constraints are relaxed inthe native neuronal membrane, which has a more diverse lipid composition with multiple differentchain pairings; about 6% of the phospholipids in our simulated membranes contain no saturatedchain at all. Nonetheless, our previous results using simplified binary heteroacidic/cholesterolmembranes played a key role in identifying the natural site boundaries.As expected, within each leaflet cholesterol has the strongest affinity for the inter-subunitsites, although the affinity of cholesterol for the M4 sites was second only to that of n-3 PU-FAs. Combined, these results are consistent with an overwhelming amount of evidence spanningfour decades that suggests direct interactions between cholesterol and nAChRs, regardless of thephospholipid composition of the membrane. One surprise for cholesterol was the role of the leafletin determining affinity: cholesterol has a stronger affinity for either outer leaflet site compared toeither inner leaflet site. This result may reflect competition with anionic saturated lipids in theinner leaflet, which would be consistent with multiple experiments , suggesting that anioniclipids can partially or fully compensate for a loss of cholesterol. This result is also consistent with18holesterol embedded in the outer TMD (which has numerous gaps in the amino acid density)but not the inner TMD.Based on our results using ELIC , we had expected that anionic lipids would select for sites onthe inner leaflet lined with basic residues. In the homomeric ELIC, these residues are symmetricly-arranged, while in the heteromeric nAChR they vary by subunit(Figure ?? a), with the M4 sitecontaining the most such residues on most subunits. The present results support that expectation:anionic lipids have a stronger affinity for M4 than inter-subunit sites.For both outer and inner leaflets, neutral lipids with smaller head groups (PE) have strongeraffinity than the larger PC headgroup. It is unclear why PE is more favorable than other neutrallipids at this time, though this is consistent with previous work , and the most straightforwardexplanation is that the smaller headgroup introduces fewer clashes with the protein TMD.Among anionic lipids in the inner leaflet, regardless of the site, PS and PI have an affin-ity greater than or equal to PC, and much greater than the other anionic lipid headgroups(PIP1,PIP2,PIP3, and PA). The lipid headgroups PS and PI both have a charge of -1, whilePA in the MARTINI forcefield carries a charge of -2, and PIP1, PIP2, and PIP3 have chargesof -3,-5, and -7. These results suggest that the inner leaflet sites select for monoanionic head-groups, while multianionic headgroups are highly unfavorable. Due to the limitations of thecoarse-grained model, future atomistic calculations are required to validate and understand theapparent preference of the M4 site for PI over PS.The present results highlight the utility of model membranes for developing hypotheses of spe-cific lipid-protein interactions, and the need to test those hypothesis within more complex nativemembranes. The present results could be tested and aid in interpretation of experiments carried outin more complex membranes. For instance, we would expect that mutations of the basic residuesfacing the inner leaflet would reduce binding of saturated phospholipids with anionic headgroups,which would be replaced with bound cholesterol. We would also predict that if PUFAs cause gainof function via binding to the intersubunit site, this gain would be enhanced by replacing someheteroacidic lipids with homoacidic lipids while keeping the total fraction of PUFA chains con-stant. In general the present results provide valuable insight into how to predict lipid competition,which is one of the primarily challenges of interpreting experiments in complex membranes.19 CKNOWLEDGMENTS
GB and LS were supported by the Busch Biomedical Foundation. This project was supportedby generous allocation through the Rutgers University Office of Advanced Research Computing(OARC), which is supported by Rutgers University and the state of New Jersey. We are gratefulto Dr. Jérôme Hénin for helpful input and suggestions.
DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from the corresponding authorupon reasonable request. Scripts for polar density analysis and plotting scripts can be found ongithub: https://github.com/BranniganLab/densitymap.
REFERENCES C. M. Hénault, J. Sun, J. P. D. Therien, C. J. B. DaCosta, C. L. Carswell, J. M. Labriola,P. F. Juranka, J. E. Baenziger, The role of the M4 lipid-sensor in the folding, trafficking, andallosteric modulation of nicotinic acetylcholine receptors, Neuropharmacology 96 (2015) 157–168. N. Mukhtasimova, C. J. B. DaCosta, S. M. Sine, Improved resolution of single channel dwelltimes reveals mechanisms of binding, priming, and gating in muscle AChR, The Journal ofGeneral Physiology 148 (2016) 43–63. D. Kalamida, K. Poulas, V. Avramopoulou, E. Fostieri, G. Lagoumintzis, K. Lazaridis,A. Sideri, M. Zouridakis, S. J. Tzartos, Muscle and neuronal nicotinic acetylcholine recep-tors: Structure, function and pathogenicity, FEBS Journal 274 (2007) 3799–3845. A. Taly, P.-J. Corringer, D. Guedin, P. Lestage, J.-P. Changeux, Nicotinic receptors: allosterictransitions and therapeutic targets in the nervous system, Nature Reviews Drug Discovery 8(2009) 733–750. H. Patel, J. McIntire, S. Ryan, A. Dunah, R. Loring, Anti-inflammatory effects of astroglial α κ B pathway andactivation of the Nrf2 pathway., Journal of neuroinflammation 14 (2017) 192. G. T. Yocum, D. L. Turner, J. Danielsson, M. B. Barajas, Y. Zhang, D. Xu, N. L. Harrison,G. E. Homanics, D. L. Farber, C. W. Emala, GABA A receptor alpha4-subunit knockout en-20ances lung inflammation and airway reactivity in a murine asthma model, American Journalof Physiology - Lung Cellular and Molecular Physiology 313 (2017) L406—-L415. J. Egea, I. Buendia, E. Parada, E. Navarro, R. León, M. G. Lopez, Anti-inflammatory role ofmicroglial alpha7 nAChRs and its role in neuroprotection, 2015. G. L. Cornelison, N. C. Pflanz, M. E. Tipps, S. J. Mihic, Identification and characterizationof heptapeptide modulators of the glycine receptor, European Journal of Pharmacology 780(2016) 252–259. W. Xiong, T. Cui, K. Cheng, F. Yang, S.-R. Chen, D. Willenbring, Y. Guan, H.-L. Pan, K. Ren,Y. Xu, L. Zhang, Cannabinoids suppress inflammatory and neuropathic pain by targeting alpha3glycine receptors, The Journal of Experimental Medicine 209 (2012) 1121–1134. J. Walstab, G. Rappold, B. Niesler, 5-HT(3) receptors: role in disease and target of drugs,Pharmacology & Therapeutics 128 (2010) 146–169. R. Picciotto, Marina, Neuroprotection via nAChRs: the role of nAChRs in neurodegenerativedisorders such as Alzheimer’s and Parkinson’s disease, Frontiers in Bioscience 13 (2008) 492. C. M. Martin-Ruiz, J. A. Court, E. Molnar, M. Lee, C. Gotti, A. Mamalaki, T. Tsouloufis,S. Tzartos, C. Ballard, R. H. Perry, E. K. Perry, Alpha4 but not alpha3 and alpha7 nicotinicacetylcholine receptor subunits are lost from the temporal cortex in Alzheimer’s disease, Jour-nal of Neurochemistry 73 (1999) 1635–1640. A. S. Arnold, M. Gueye, S. Guettier-Sigrist, I. Courdier-Fruh, G. Coupin, P. Poindron, J. P.Gies, Reduced expression of nicotinic AChRs in myotubes from spinal muscular atrophy Ipatients, Laboratory Investigation 84 (2004) 1271–1278. S. N. Haydar, J. Dunlop, Neuronal nicotinic acetylcholine receptors - targets for the devel-opment of drugs to treat cognitive impairment associated with schizophrenia and alzheimer’sdisease., Current topics in medicinal chemistry 10 (2010) 144–152. V. A. Lennon, L. G. Ermilov, J. H. Szurszewski, S. Vernino, Immunization with neuronalnicotinic acetylcholine receptor induces neurological autoimmune disease, Journal of ClinicalInvestigation 111 (2003) 907–913. S. Kumari, V. Borroni, A. Chaudhry, B. Chanda, R. Massol, S. Mayor, F. J. Barrantes, Nico-tinic acetylcholine receptor is internalized via a Rac-dependent, dynamin-independent endo-cytic pathway, Journal of Cell Biology 181 (2008) 1179–1193. J. E. Baenziger, J. A. Domville, J. P. D. Therien, The role of cholesterol in the activation ofnicotinic acetylcholine receptors., Current topics in membranes 80 (2017) 95–137.21 A. W. Dalziel, E. S. Rollins, M. G. McNamee, The effect of cholesterol on agonist-inducedflux in reconstituted acetylcholine receptor vesicles, FEBS Letters 122 (1980) 193–196. J. F. Ellena, M. A. Blazing, M. G. McNamee, Lipid-Protein Interactions in ReconstitutedMembranes Containing Acetylcholine Receptor, Biochemistry 22 (1983) 5523–5535. M. Criado, H. Eibl, F. J. Barrantes, Corrections: Effects of Lipids on Acetylcholine Receptor.Essential Need of Cholesterol for Maintenance of Agonist-Induced State Transitions in LipidVesicles: (Biochemistry (1982) 21(15) (3622–3629) (10.1021/bi00258a015)), Biochemistry 22(1983) 524. T. M. Fong, M. G. McNamee, Correlation between acetylcholine receptor function and struc-tural properties, Biochemistry 25 (1986) 830–840. T. M. Fong, M. G. McNamee, Stabilization of Acetylcholine Receptor Secondary Structureby Cholesterol and Negatively Charged Phospholipids in Membranes, Biochemistry 26 (1987)3871–3880. O. T. Jones, M. G. Mcnamee, Annular and Nonannular Binding Sites for Cholesterol Associ-ated with the Nicotinic Acetylcholine Receptor, Biochemistry 27 (1988) 2364–2374. C. Sunshine, M. G. McNamee, Lipid modulation of nicotinic acetylcholine receptor function:the role of membrane lipid composition and fluidity, Biochim Biophys Acta 1191 (1994) 59–64. C. J. DaCosta, S. A. Medaglia, N. Lavigne, S. Wang, C. L. Carswell, J. E. Baenzinger, Anioniclipids allosterically modulate multiple nicotinic acetylcholine receptor conformational equilib-ria, Journal of Biological Chemistry 284 (2009) 33841–33849. S. B. Mantipragada, L. I. Horváth, H. R. Arias, G. Schwarzmann, K. Sandhoff, F. J. Barrantes,D. Marsh, Lipid-protein interactions and effect of local anesthetics in acetylcholine receptor-rich membranes from Torpedo marmorata electric organ, Biochemistry 42 (2003) 9167–9175. F. J. Barrantes, Cholesterol effects on nicotinic acetylcholine receptor: Cellular aspects, Sub-Cellular Biochemistry 51 (2010) 467–487. C. J. Baier, J. Fantini, F. J. Barrantes, Disclosure of cholesterol recognition motifs in trans-membrane domains of the human nicotinic acetylcholine receptor, Scientific Reports 1 (2011)69. Y. Zhou, M. E. Nelson, A. Kuryatov, C. Choi, J. Cooper, J. Lindstrom, Human alpha4beta2acetylcholine receptors formed from linked subunits., The Journal of neuroscience : the officialjournal of the Society for Neuroscience 23 (2003) 9004–9015.22 G. Gamba, W. G. Hill, N. M. Southern, B. Maciver, E. Potter, G. Apodaca, C. P. Smith, M. L.Zeidel, G. Warren, Isolation and characterization of the Xenopus oocyte plasma membrane: a new method for studying activity of water and solute transporters, American Journal ofPhysiology-Renal Physiology 15261 (2005) 217–224. Q. Chen, M. N. Kinde, P. Arjunan, M. M. Wells, A. E. Cohen, Y. Xu, P. Tang, Direct PoreBinding as a Mechanism for Isoflurane Inhibition of the Pentameric Ligand-gated Ion ChannelELIC, Scientific Reports 5 (2015) 13833. N. Kouvatsos, P. Giastas, D. Chroni-Tzartou, C. Poulopoulou, S. J. Tzartos, Crystal structure ofa human neuronal nAChR extracellular domain in pentameric assembly: Ligand-bound alpha2homopentamer, Proceedings of the National Academy of Sciences 113 (2016) 9635–9640. M. Nys, E. Wijckmans, A. Farinha, Ö. Yoluk, M. Andersson, M. Brams, R. Spurny, S. Peigneur,J. Tytgat, E. Lindahl, C. Ulens, Allosteric binding site in a Cys-loop receptor ligand-bindingdomain unveiled in the crystal structure of ELIC in complex with chlorpromazine, Proceedingsof the National Academy of Sciences 113 (2016) E6696—-E6703. L. Polovinkin, G. Ghérici Hassaine, J. Perot, E. Neumann, A. A. Jensen, S. N. Lefebvre, P.-J.Corringer, J. Neyton, C. Chipot, F. Dehez, G. Schoehn, H. Nury, Conformational transitions ofthe serotonin 5-HT 3 receptor, Nature (2018). S. X. Moffett, E. A. Klein, G. Brannigan, J. V. Martin, L-3,3 (cid:48) ,5-triiodothyronine and preg-nenolone sulfate inhibit Torpedo nicotinic acetylcholine receptors, PLoS ONE 14 (2019) 1–18. P. Kumar, Y. Wang, Z. Zhang, Z. Zhao, G. D. Cymes, E. Tajkhorshid, C. Grosman,Cryo-EM structures of a lipid-sensitive pentameric ligand-gated ion channel embedded in aphosphatidylcholine-only bilayer, Proceedings of the National Academy of Sciences of theUnited States of America 117 (2020) 1788–1798. L. Conti, A. Limon, E. Palma, R. Miledi, Microtransplantation of cellular membranes fromsquid stellate ganglion reveals ionotropic GABA receptors, Biological Bulletin 224 (2013)47–52. C. Cotman, M. L. Blank, A. Moehl, F. Snyder, Lipid Composition of Synaptic Plasma Mem-branes Isolated from Rat Brain by Zonal Centrifugation, Biochemistry 8 (1969) 4606–4612. R. Taguchi, M. Ishikawa, Precise and global identification of phospholipid molecular speciesby an Orbitrap mass spectrometer and automated search engine Lipid Search, Journal of Chro-matography A (2010). 23 W. C. Breckenridge, I. G. Morgan, J. P. Zanetta, G. Vincendon, Adult rat brain synaptic vesiclesII. Lipid composition, BBA - General Subjects 320 (1973) 681–686. H. I. Ingólfsson, T. S. Carpenter, H. Bhatia, P. T. Bremer, S. J. Marrink, F. C. Lightstone,Computational Lipidomics of the Neuronal Plasma Membrane, Biophysical Journal 113 (2017)2271–2280. T. G. McEvoy, G. D. Coull, P. J. Broadbent, J. S. M. Hutchinson, B. K. Speake, Fatty acidcomposition of lipids in immature cattle, pig and sheep oocytes with intact zona pellucida, JReprod Fertil 118 (2000) 163–170. J. Y. Kim, M. Kinoshita, M. Ohnishi, Y. Fukui, Lipid and fatty acid analysis of fresh and frozen-thawed immature and in vitro matured bovine oocytes, Reproduction 122 (2001) 131–138. G. van Meer, A. I. P. M. de Kroon, Lipid map of the mammalian cell, Journal of Cell Science124 (2010) 5–8. J. H. Lorent, K. R. Levental, L. Ganesan, G. Rivera-Longsworth, E. Sezgin, M. Doktorova,E. Lyman, I. Levental, Plasma membranes are asymmetric in lipid unsaturation, packing andprotein shape, Nature Chemical Biology 16 (2020) 644–652. H. I. Ingólfsson, M. N. Melo, F. J. Van Eerden, C. Arnarez, C. A. Lopez, T. A. Wassenaar,X. Periole, A. H. De Vries, D. P. Tieleman, S. J. Marrink, Lipid organization of the plasmamembrane, Journal of the American Chemical Society 136 (2014) 14554–14559. F. J. Barrantes, The lipid environment of the nicotinic acetylcholine receptor in native andreconstituted membrane, Critical Reviews in Biochemistry and Molecular Biology 24 (1989)437–478. O. Quesada, C. González -Freire, M. C. Ferrer, J. O. Colón -Sáez, E. Fernández-García, J. Mer-cado, A. Dávila, R. Morales, J. A. Lasalde-Dominicci, Uncovering the lipidic basis for thepreparation of functional nicotinic acetylcholine receptor detergent complexes for structuralstudies, Scientific Reports 6 (2016) 32766. R. K. McNamara, M. Ostrander, W. Abplanalp, N. M. Richtand, S. C. Benoit, D. J. Clegg,Modulation of phosphoinositide-protein kinase C signal transduction by omega-3 fatty acids:Implications for the pathophysiology and treatment of recurrent neuropsychiatric illness,Prostaglandins Leukotrienes and Essential Fatty Acids 75 (2006) 237–257. R. K. McNamara, R. Jandacek, T. Rider, P. Tso, K. E. Stanford, C. G. Hahn, N. M. Richtand,Deficits in docosahexaenoic acid and associated elevations in the metabolism of arachidonicacid and saturated fatty acids in the postmortem orbitofrontal cortex of patients with bipolar24isorder, Psychiatry Research 160 (2008) 285–299. M. Maekawa, A. Watanabe, Y. Iwayama, T. Kimura, K. Hamazaki, S. Balan, H. Ohba,Y. Hisano, Y. Nozaki, T. Ohnishi, M. Toyoshima, C. Shimamoto, K. Iwamoto, M. Bundo,N. Osumi, E. Takahashi, A. Takashima, T. Yoshikawa, Polyunsaturated fatty acid deficiencyduring neurodevelopment in mice models the prodromal state of schizophrenia through epige-netic changes in nuclear receptor genes, Translational Psychiatry 7 (2017) 1–11. R. M. Adibhatla, J. F. Hatcher, Role of lipids in brain injury and diseases, 2007. M. Schneider, B. Levant, M. Reichel, E. Gulbins, J. Kornhuber, C. P. Müller, Lipids in psychi-atric disorders and preventive medicine, 2017. N. Koga, J. Ogura, F. Yoshida, K. Hattori, H. Hori, E. Aizawa, I. Ishida, H. Kunugi, Alteredpolyunsaturated fatty acid levels in relation to proinflammatory cytokines, fatty acid desaturasegenotype, and diet in bipolar disorder, Translational Psychiatry 9 (2019). K. Hamazaki, M. Maekawa, T. Toyota, B. Dean, T. Hamazaki, T. Yoshikawa, Fatty acid com-position of the postmortem prefrontal cortex of patients with schizophrenia, bipolar disorder,and major depressive disorder, Psychiatry Research 227 (2015) 353–359. M. Peet, Eicosapentaenoic acid in the treatment of schizophrenia and depression: Rationaleand preliminary double-blind clinical trial results, Prostaglandins Leukotrienes and EssentialFatty Acids 69 (2003) 477–485. C. Bushe, C. Paton, The potential impact of antipsychotics on lipids in schizophrenia: Is thereenough evidence to confirm a link?, Journal of Psychopharmacology 19 (2005) 76–83. G. E. Berger, S. Smesny, G. P. Amminger, Bioactive lipids in schizophrenia, InternationalReview of Psychiatry 18 (2006) 85–98. J. A. Conquer, M. C. Tierney, J. Zecevic, W. J. Bettger, R. H. Fisher, Fatty acid analysis of bloodplasma of patients with alzheimer’s disease, other types of dementia, and cognitive impairment,Lipids 35 (2000) 1305–1312. G. Di Paolo, T.-W. Kim, Linking lipids to Alzheimer’s disease: cholesterol and beyond, NatureReviews Neuroscience 12 (2011) 284–296. S. A. L. Bennett, N. Valenzuela, H. Xu, B. Franko, S. Fai, D. Figeys, Using neurolipidomics toidentify phospholipid mediators of synaptic (dys)function in Alzheimer’s Disease, Frontiers inPhysiology 4 JUL (2013) 1–16. R. S. Yadav, N. K. Tiwari, Lipid Integration in Neurodegeneration: An Overview ofAlzheimer’s Disease, 2014. 25 P. V. Escribá, Membrane-lipid therapy: A historical perspective of membrane-targeted therapies— From lipid bilayer structure to the pathophysiological regulation of cells, Biochimica etBiophysica Acta - Biomembranes 1859 (2017) 1493–1506. G. H. Addona, H. Sandermann, M. A. Kloczewiak, S. S. Husain, K. W. Miller, Where doescholesterol act during activation of the nicotinic acetylcholine receptor?, Biochimica et Bio-physica Acta - Biomembranes 1370 (1998) 299–309. D. Laverty, P. Thomas, M. Field, O. J. Andersen, M. G. Gold, P. C. Biggin, M. Gielen, T. G.Smart, Crystal structures of a GABA A -receptor chimera reveal new endogenous neurosteroid-binding sites, Nature Structural and Molecular Biology 24 (2017) 977–985. M. M. Budelier, W. W. L. Cheng, Z. W. Chen, J. R. Bracamontes, Y. Sugasawa, K. Krishnan,L. Mydock-McGrane, D. F. Covey, A. S. Evers, Common binding sites for cholesterol andneurosteroids on a pentameric ligand-gated ion channel, Biochimica et Biophysica Acta -Molecular and Cell Biology of Lipids 1864 (2019) 128–136. S. Basak, N. Schmandt, Y. Gicheru, S. Chakrapani, Crystal structure and dynamics of alipid-induced potential desensitized-state of a pentameric ligand-gated channel, eLife 6 (2017)e23886. C. M. Hénault, C. Govaerts, R. Spurny, M. Brams, A. Estrada-Mondragon, J. Lynch,D. Bertrand, E. Pardon, G. L. Evans, K. Woods, B. W. Elberson, L. G. Cuello, G. Branni-gan, H. Nury, J. Steyaert, J. E. Baenziger, C. Ulens, A lipid site shapes the agonist response ofa pentameric ligand-gated ion channel, Nature Chemical Biology (2019). J. J. Kim, A. Gharpure, J. Teng, Y. Zhuang, R. J. Howard, S. Zhu, C. M. Noviello, R. M.Walsh, E. Lindahl, R. E. Hibbs, Shared structural mechanisms of general anaesthetics andbenzodiazepines, Nature 585 (2020) 303–308. A. Tong, F. F. Hsu, P. A. Schmidpeter, C. M. Nimigean, L. Sharp, G. Brannigan, W. W. Cheng,Direct binding of phosphatidylglycerol at specific sites modulates desensitization of a Ligand-gated ion channel, eLife 8 (2019). G. Brannigan, J. Henin, R. Law, R. Eckenhoff, M. L. Klein, Embedded cholesterol in thenicotinic acetylcholine receptor, Proceedings of the National Academy of Sciences 105 (2008)14418–14423. J. Hénin, R. Salari, S. Murlidaran, G. Brannigan, A predicted binding site for cholesterol onthe GABAA receptor, Biophysical Journal 106 (2014) 1938–1949.26 J. Doma ´nski, S. J. Marrink, L. V. Schäfer, Transmembrane helices can induce domain formationin crowded model membranes, Biochimica et Biophysica Acta - Biomembranes 1818 (2012)984–994. M. Chavent, A. L. Duncan, M. S. Sansom, Molecular dynamics simulations of membraneproteins and their interactions: from nanoscale to mesoscale This review comes from a themedissue on Biophysical and molecular biological methods, Current Opinion in Structural Biology40 (2016) 8–16. T. S. Carpenter, C. A. López, C. Neale, C. Montour, H. I. Ingólfsson, F. Di Natale, F. C. Light-stone, S. Gnanakaran, Capturing Phase Behavior of Ternary Lipid Mixtures with a RefinedMartini Coarse-Grained Force Field, Journal of Chemical Theory and Computation 14 (2018)6050–6062. H. I. Ingólfsson, H. Bhatia, T. Zeppelin, W. F. Bennett, K. A. Carpenter, P. C. Hsu, G. Dharu-man, P. T. Bremer, B. Schiøtt, F. C. Lightstone, T. S. Carpenter, Capturing Biologically Com-plex Tissue-Specific Membranes at Different Levels of Compositional Complexity, The journalof physical chemistry. B 124 (2020) 7819–7829. T. T. Joseph, J. S. Mincer, Common internal allosteric network links anesthetic binding sites ina pentameric ligand-gated ion channel, PLoS ONE 11 (2016) 1–20. L. Sharp, R. Salari, G. Brannigan, Boundary lipids of the nicotinic acetylcholine receptor:Spontaneous partitioning via coarse-grained molecular dynamics simulation, Biochimica etBiophysica Acta - Biomembranes 1861 (2019) 887–896. K. Woods, L. Sharp, G. Brannigan, Untangling Direct and Domain-Mediated InteractionsBetween Nicotinic Acetylcholine Receptors in DHA-Rich Membranes, Journal of MembraneBiology 252 (2019) 385–396. S. J. Marrink, V. Corradi, P. C. Souza, H. I. Ingólfsson, D. P. Tieleman, M. S. Sansom, Com-putational Modeling of Realistic Cell Membranes, Chemical Reviews 119 (2019) 6184–6226. K. A. Wilson, H. I. MacDermott-Opeskin, E. Riley, Y. Lin, M. L. O’Mara, Understanding theLink between Lipid Diversity and the Biophysical Properties of the Neuronal Plasma Mem-brane, Biochemistry 59 (2020) 3010–3018. J. Lorent, L. Ganesan, G. Rivera-Longsworth, E. Sezgin, K. Levental, E. Lyman, I. Levental,The Molecular and Structural Asymmetry of the Plasma Membrane, bioRxiv (2019) 698837. N. Unwin, Refined structure of the nicotinic acetylcholine receptor at 4 Å resolution, Journalof Molecular Biology 346 (2005) 967–989.27 D. H. de Jong, G. Singh, W. F. D. Bennett, C. Arnarez, T. A. Wassenaar, L. V. Schäfer, X. Peri-ole, D. P. Tieleman, S. J. Marrink, Improved parameters for the martini coarse-grained proteinforce field., Journal of chemical theory and computation 9 (2013) 687–697. T. A. Wassenaar, H. I. Ingólfsson, R. A. Böckmann, D. P. Tieleman, S. J. Marrink, Computa-tional lipidomics with insane: A versatile tool for generating custom membranes for molecularsimulations, Journal of Chemical Theory and Computation 11 (2015) 2144–2155. H. J. Berendsen, D. van der Spoel, R. van Drunen, GROMACS: A message-passing parallelmolecular dynamics implementation, Computer Physics Communications 91 (1995) 43–56. M. J. Abraham, T. Murtola, R. Schulz, S. Páll, J. C. Smith, B. Hess, E. Lindah, Gromacs: Highperformance molecular simulations through multi-level parallelism from laptops to supercom-puters, SoftwareX 1-2 (2015) 19–25. X. Periole, M. Cavalli, S. J. Marrink, M. A. Ceruso, Combining an elastic network witha coarse-grained molecular force field: Structure, dynamics, and intermolecular recognition,Journal of Chemical Theory and Computation 5 (2009) 2531–2543. G. B. Liam Sharp, Densitymap, github, 2020. G. Brannigan, F. L. Brown, Contributions of Gaussian curvature and nonconstant lipid volumeto protein deformation of lipid bilayers, Biophysical Journal 92 (2007) 864–876. K. S. Kim, J. Neu, G. Oster, Curvature-mediated interactions between membrane proteins,Biophysical Journal 75 (1998) 2274–2291. N. Dan, P. Pincus, S. A. Safran, Membrane-Induced Interactions between Inclusions, Langmuir9 (1993) 2768–2771. M. Goulian, Inclusions in membranes, Current Opinion in Colloid and Interface Science 1(1996) 358–361. M. Goulian, R. Bruinsma, P. Pincus, Long-range forces in heterogeneous fluid membranes, Epl23 (1993) 125–128. J. B. Fournier, P. Galatola, High-order power series expansion of the elastic interaction betweenconical membrane inclusions, European Physical Journal E 38 (2015) 1–8. R. Salari, T. Joseph, R. Lohia, J. Henin, G. Brannigan, A streamlined, general approach forcomputing ligand binding free energies and its application to GPCR-bound cholesterol, Journalof Chemical Theory and Computation (2018). J. E. Baenziger, M.-l. Morris, T. E. Darsaut, S. E. Ryan, Effect of Membrane Lipid Compo-sition on the Conformational Equilibria of the Nicotinic Acetylcholine Receptor*, Journal of28iological Chemistry 275 (2000) 777–784. J. J. Wenz, F. J. Barrantes, Nicotinic acetylcholine receptor induces lateral segregation of phos-phatidic acid and phosphatidylcholine in reconstituted membranes, Biochemistry 44 (2005)398–410. A. K. Hamouda, M. Sanghvi, D. Sauls, T. K. Machu, M. P. Blanton, Assessing the lipid re-quirements of the Torpedo californica nicotinic acetylcholine receptor, Biochemistry 45 (2006)4327–4337.100