The functional role of the a M4 transmembrane helix in the muscle nicotinic acetylcholine receptor probed through mutagenesis and coevolutionary analyses

The activity of the muscle-type Torpedo nicotinic acetylcholine receptor (nAChR) is highly sensitive to lipids, but the underlying mechanisms remain poorly understood. The nAChR transmembrane a -helix, M4, is positioned at the perimeter of each subunit in direct contact with lipids and likely plays a central role in lipid sensing. To gain insight into the mechanisms underlying nAChR lipid sensing, we used homology modeling, coevolutionary analyses, site-directed mutagenesis, and electrophysiology to examine the role of the a -subunit M4 ( a M4) in the function of the adult muscle nAChR. Ala substitutions for most a M4 residues, including those in clusters of polar residues at both the N and C termini, and deletion of up to 11 C-terminal residues had little impact on the agonist-induced response. Even Ala substitutions for coevolved pairs of residues at the interface between a M4 and the adjacent helices, a M1 and a M3, had little effect, although some impaired nAChR expression. On the other hand, Ala substitutions for Thr422 and Arg429 caused relatively large losses of function, suggesting functional roles for these specific residues. Ala substitutions for aromatic residues at the a M4- a M1/ a M3 interface generally led to gains of function, as previously reported for the prokaryotic homolog, the Erwinia chrysanthemi ligand-gated ion channel (ELIC). The functional effects of individual Ala substitutions in a

The activity of the muscle-type Torpedo nicotinic acetylcholine receptor (nAChR) is highly sensitive to lipids, but the underlying mechanisms remain poorly understood. The nAChR transmembrane a-helix, M4, is positioned at the perimeter of each subunit in direct contact with lipids and likely plays a central role in lipid sensing. To gain insight into the mechanisms underlying nAChR lipid sensing, we used homology modeling, coevolutionary analyses, site-directed mutagenesis, and electrophysiology to examine the role of the a-subunit M4 (aM4) in the function of the adult muscle nAChR. Ala substitutions for most aM4 residues, including those in clusters of polar residues at both the N and C termini, and deletion of up to 11 C-terminal residues had little impact on the agonist-induced response. Even Ala substitutions for coevolved pairs of residues at the interface between aM4 and the adjacent helices, aM1 and aM3, had little effect, although some impaired nAChR expression. On the other hand, Ala substitutions for Thr422 and Arg429 caused relatively large losses of function, suggesting functional roles for these specific residues. Ala substitutions for aromatic residues at the aM4-aM1/aM3 interface generally led to gains of function, as previously reported for the prokaryotic homolog, the Erwinia chrysanthemi ligand-gated ion channel (ELIC). The functional effects of individual Ala substitutions in aM4 were found to be additive, although not in a completely independent manner. Our results provide insight into the structural features of aM4 that are important. They also suggest how lipid-dependent changes in aM4 structure ultimately modify nAChR function.
Lipids likely influence pLGIC function by altering interactions between M4 and M1/M3 (30,31). One model proposes that altered M4-M1/M3 interactions modulate contact between the M4 C terminus and the b6-b7 loop (the Cys loop in eukaryotic pLGICs), a structure at the interface between the extracellular domain (ECD) and TMD that plays an important role in translating agonist binding into channel gating (32)(33)(34)(35). This model is supported by the observation that a lipid bridges interactions between M4 and the b6-b7 loop/Cys loop in the prokaryotic Gloeobacter violaceus ligand-gated ion channel, GLIC, with mutations of implicated residues impairing both function and expression (36). The M4 C terminus is also important in the expression and/or function of the a4b2 nAChR, the 5-HT 3A receptor, the r1 GABA A receptor, and a a7/5-HT 3A receptor chimera (37)(38)(39)(40), and has been implicated in neurosteroid-induced potentiation of both the neuronal a4b2 nAChR and the GABA A receptor (41,42). On the other hand, specific interactions between M4 and the b6-b7 loop/Cys loop are not critical to function in the Erwinia chrysanthemi ligand gated ion channel, ELIC, suggesting that distinct mechanisms underlie lipid sensing in different pLGICs (see, for example, reference 43).
To better understand lipid sensing, we and others have used Ala-scanning mutagenesis to identify residues in M4 that play an important role in function (36,39,(44)(45)(46)(47). Whereas these studies have been informative, they have all focused on homomeric pLGICs. An intriguing question that remains to be addressed is how the M4 a-helix from different subunits of a heteromeric pLGIC influences channel function and, thus, participates in lipid sensing. As a first step to addressing this question, we use here a combination of mutagenesis, evolutionary couplings, and two-electrode voltage clamp electrophysiology to explore the functional role of M4 from the a subunit (aM4) of the human adult muscle nAChR (a 2 bed). Our data reveal several unique features regarding the role of aM4 in function and provide a structural basis for understanding the lipid sensitivity of the muscle-type nAChR.

Structural models of the human muscle a subunit
To properly interpret functional data derived from mutations in aM4 of the human adult muscle nAChR, a 3D structure that accurately defines the interactions that occur between aM4 and both aM1/aM3 and lipids is required. In the absence of such a structure, we first generated homology models based on the 4-Å-resolution apo structure of the Torpedo muscle-like nAChR (referred to as the 2BG9 Torpedo model), a revised Torpedo nAChR structure with a corrected sequence register in aM2/aM3, and a 3.3-Å-resolution nicotine-bound structure of the neuronal a3b4 nAChR (28,48,49). After the first submission of this work, an a-bungarotoxin-bound structure of the Torpedo nAChR was solved at 2.7-Å resolution (29). The new 6UWZ Torpedo structure reveals several novel features relevant here, including a corrected register in the fitting of the amino acid sequence into aM2/aM3 (50), a short aMX helix at the periphery of the aM4 N terminus, and an aM4 helix that is both contiguous with the intracellular aMA helix and rotated about its long axis to define different side-chain projections toward both aM1/aM3 and lipids. The rotation of aM4 is surprising, because five residues (Cys412, Met415, Cys418, Thr422, and Val425) that are labeled by the lipophilic photoreactive probe 3-trifluoromethyl-3-(m-[ 125 I]iodophenyl)diazirine ([ 125 I]-TID) and that define the aM4-lipid interface in the Torpedo 2BG9 structure (Fig. S1) (51) now project toward aM1/aM3. Whereas the positions of these residues in the new 6UWZ Torpedo structure are difficult to reconcile with the [ 125 I]-TID labeling, they are supported by the structures of other neuronal nAChRs (52,53) as well as by an identified interaction that occurs between aM1 Ser226 and aM4 Thr422 during channel function (24,43). The hydrogen bond donor/ acceptor distances between Ser226 and Thr422 are 11.1 and 7.1 Å in the 2BG9 and 6UWZ Torpedo structures, respectively. Thus, the new 6UWZ Torpedo structure requires a smaller reorientation of aM1/aM4 upon channel gating to bring Ser226 and Thr422 in close contact.
Note that the 2BG9 Torpedo structure was solved from cryoelectron microscopic images obtained from native Torpedo synaptosomes, whereas the 6UWZ Torpedo structure was solved from images obtained using nAChRs that were detergent solubilized from native Torpedo membranes, affinity purified, and then reconstituted into lipid nanodiscs. Given the high resolution of the 6UWZ Torpedo structure and structures of other neuronal nAChRs, we reinterpreted our mutagenesis data primarily in terms of a homology model based on the 6UWZ Torpedo structure (referred to as the 6UWZ Torpedo model), as shown in Fig. 1 and 4. Given the strength of the [ 125 I]-TID labeling data, however, we considered a speculative possibility that aM4 undergoes a reorientation during solubilization/affinity purification/nanodisc reconstitution and, thus, have taken into account the orientation of aM4 defined in the 2BG9 Torpedo model.
A coevolutionary map of aM4-aM1/aM3 interactions We used a coevolutionary approach to test further our two homology models and to identify potential interactions at the aM4-aM1/aM3 interface that play a role in channel function. In the coevolutionary approach, primary amino acid sequences of related proteins across the evolutionary timeline are aligned to identify pairs of residues that have coevolved. A probability score is given to every pair of residues in the sequence, with each probability score related to the proportion of sequences in which that residue pair has evolved together through the evolutionary timeline. This approach has been shown to successfully predict residue pairs in close physical proximity in a variety of proteins (54,55).
Nine coevolved pairs with probability scores of greater than 90% were identified at the aM4-aM1/aM3 interface: four pairs at the aM1-aM4 interface ( . Only one residue, Thr133, in the Cys loop was found to have coevolved to some extent with a residue on aM4, Val413, but the probability score for the Thr133/Val413 pair is only 14.7%, suggesting that interactions between aM4 and the Cys-loop have not evolved a conserved role in muscle nAChR function. Each of the four coevolved pairs of residues at the aM1-aM4 interface are relatively close to each other in both homology models, although Ile219/Val425 and Phe233/Phe414 are slightly closer in the 2BG9 Torpedo model, whereas Gly230/ Leu411 and Tyr234/Leu411 are closer in the 6UWZ Torpedo model ( Fig. 2 and Table S1). In contrast, the 6UWZ Torpedo model places four of the five coevolved pairs of residues at the aM3-aM4 interface in much closer proximity than the 2BG9 Torpedo model. The large distance between coevolved pairs of residues at the aM3-aM4 interface in the Torpedo 2BG9 model provides further evidence that there is an error in the register of the amino acid sequence in both aM2 and aM3, as confirmed in the 6UWZ Torpedo structure.

Ala scan of aM4
To probe the role of aM4 in nAChR function, we first generated Ala mutations at every position along aM4 and examined the functional effects of these mutations using two-electrode voltage clamp electrophysiology. Given the uncertainty of our original homology models, we were generous in the choice of residues that were mutated. In total, 36 Ala mutations were generated, starting at Trp399 and ending at Gly437.
The WT human adult muscle nAChR expresses robustly in Xenopus oocytes and responds in a concentration-dependent manner to acetylcholine with an EC 50 of 7.61 6 1.25 mM (n = 50) (Fig. 3). All 36 aM4 Ala mutants expressed, with the resulting EC 50 values for the measured dose responses of each mutant, are summarized in Table 1. Although each measured EC 50 value reflects a weighted average of all rate constants associated with both agonist binding/dissociation and channel opening/closing, the studied mutations are distant from the agonist binding site. Observed changes in EC 50 likely reflect primarily changes in the opening/closing rate constants (i.e. channel gating), as opposed to effects on agonist binding/ dissociation (i.e. agonist affinity) (24,56). Note that changes in the rates of desensitization can also influence the measured EC 50 values, although the mutants investigated here have, at most, subtle effects on desensitization rates (Fig. 3). The reported changes or lack of reported changes in EC 50 values do not correlate with altered rates of desensitization. Regardless, we have presented concentration-dependent whole-cell trace electrophysiology recordings for a comprehensive set of mutations in Fig. S2.
The changes in EC 50 value for every mutant relative to the WT are heat mapped onto aM4 in Fig. 4, where the aM4 heat map is compared with similar heat maps generated for the prokaryotic homologs, GLIC and ELIC. Our initial discussion focuses on comparisons of the aM4 Ala scan with those of GLIC and ELIC, because all data sets were recorded under comparable conditions (36). The comparisons are broadened to include other pLGICs in the Discussion.
Of the 36 Ala mutations, 13 lead to statistically significant changes in the EC 50 values relative to the WT, although two of these (Lys400 and Tyr401) likely should be considered as being part of aMA and project toward the cytoplasm (Fig. 1). The proportion of mutations on aM4 (13 of 36) that lead to significant changes in the measured EC 50 values is lower than that in either GLIC (17 of 26) or ELIC (26 of 29), likely reflecting the fact that the a subunit and, thus, each mutation, is present only twice per nAChR pentamer, whereas each mutation in both GLIC and ELIC is repeated in all five subunits. Also, a larger proportion of the generated substitutions in the human aM4 are conservative changes from a larger to a smaller aliphatic side chain (18 mutations) or from Gly to Ala (4 mutations).
On the other hand, even though fewer Ala mutations in aM4 lead to statistically significant changes in the measured EC 50 values than in either GLIC or ELIC, three mutations, R429A, T422A, and F426A, lead to changes in EC 50 values (5.3-fold, 4.1-fold, and 3.8-fold, respectively) that are similar in magnitude to the largest changes in EC 50 values observed with M4 Ala substitutions in GLIC/ELIC, despite the fact that the aM4 mutations occur in only two of the five subunits. In fact, only a few Ala mutations in GLIC/ELIC lead to comparable changes in EC 50 values. To put this observation in perspective, if the effects of the R429A mutation were independent and conserved across all five subunits, then the simultaneous Ala substitution for all five residues would lead to a 60-fold change in the measured EC 50 value, a value greater than that of any functional change observed with Ala mutations in the homopentameric GLIC and ELIC. This calculation highlights the functional sensitivity of the human adult muscle nAChR to mutations along aM4.
Finally, the nature and pattern of the observed changes in the EC 50 values have both similarities with and differences to those observed in GLIC and ELIC. In aM4, Ala mutations lead to a mix of both loss (8/13)-and gain (5/13)-of-function phenotypes. In contrast, the majority of the Ala mutations (15 of 17) of M4 residues in GLIC led to loss-of-function phenotypes, whereas all 26 Ala mutations in M4 of ELIC led to gain-of-function phenotypes. In addition, the vast majority of the Ala substitutions that influence GLIC function, and those with the largest gain-of-function phenotypes in ELIC, are located at the M4-M1/M3 interface. Similar to GLIC and ELIC, the 6UWZ Torpedo model places most (8/13) of the changes in function, including the most impactful, at the aM4-aM1/aM3 interface. In contrast, the 2BG9 Torpedo model places the latter residues at the aM4-lipid interface, with several being directly labeled by [ 125 I]-TID (51).
Polar residues at the N and C termini of aM4 aM4 exhibits clusters of polar/charged residues at both its N (Asp407 and His408) and C (Arg429, Glu432, Asn434, Gln435, and Gln436) termini ( Fig. 1) that could form functionally important interactions with either lipids or adjacent residues in the TMD and/or ECD. At the aM4 N terminus, Asp407 projects toward Lys242 on the aM1-aM2 loop (6UWZ Torpedo model) or toward Thr237 on M1 (2BG9 Torpedo model), whereas His408 projects toward lipids, forming a hydrogen with the phosphate of a bound lipid in one a subunit of the 6UWZ Torpedo structure. Regardless, the Ala substitution of either residue has little effect (1.3-fold gain and 1.5-fold loss of function, respectively) on the measured EC 50 value, suggesting that neither residue forms interactions that are critical for expression or function.
At the aM4 C terminus, the 6UWZ Torpedo model projects Arg429 and Glu432 toward His134 of the Cys-loop with the remaining polar residues (Asn434, Gln435, and Gln436) unresolved. In contrast, the 2BG9 Torpedo model projects Glu432 and Gln435 toward Phe137 of the Cys-loop, with Arg429, Asn434, and Gln436 facing the lipid. Surprisingly, Ala substitutions of Glu432, Asn434, Gln435, and Gln436 have no effect on channel function, suggesting that none of these residues form functionally significant interactions that are critical to expression or function. On the other hand, R429A leads to a relatively large 5.3-fold loss of function, suggesting an important functional interaction between aM4 and the Cys-loop. Despite this important result, the interpretation that Arg is functionally important is clouded by the lack of an effect on the measured EC 50 value upon deletion of this residue, as discussed below.
The hydrophobic core of aM4 The central core of aM4 is formed mainly by aliphatic residues, although these are interspersed with two aromatic residues, Phe414 and Phe426, and the potential hydrogen-bonding residues, Cys418 and Thr422. The conservative Ala substitution of the five large aliphatic residues in the hydrophobic core of aM4, Leu410, Leu411, M415, and Val425, each leads to a statistically significant change in the EC 50 value, although all changes are less than 2-fold. The Ala substitutions for the two aromatic residues, Phe414 and Phe426, lead to 1.7-fold and 3.8-fold gains of function, respectively, suggesting detrimental effects on function. The latter are of particular interest given the importance of aromatic interactions at the M4-M1/M3 interfaces in the folding and function of other pLGICs (31,44). The functional roles of aromatic residues at the aM4-aM1/aM3 interface are explored in more detail below.  The Ala substitution of Thr422 leads to a relatively large 4.1fold loss of function. Thr422 faces aM1 in the 6UWZ Torpedo model but projects tangentially to aM1 interacting with lipid in the 2BG9 Torpedo model. Consistent with our data showing a loss of function, single-channel measurements indicate that T422A in the mouse adult muscle nAChR leads to a 5-fold decrease in channel open times. In addition, a mutant cycle analysis suggests an interaction between Thr422 and the nearby aM1 Ser226 stabilizes the open state (43). Interestingly, although the adjacent C418A mutation leads to only a small 1.4-fold loss of function, C418W leads to a relatively large 16to 25-fold gain of function and a congenital myasthenic syndrome (43,57). Channel function in the adult muscle nAChR appears to be sensitive to the structure of residues in the vicinity of Thr422. This region of aM4 may be particularly responsive to changes in the surrounding lipid environment.

The aM4 C terminus
Previous studies with other pLGICs have highlighted a role for the M4 C terminus in function, possibly through interactions with the Cys-loop (36,38,39,45). In contrast, Ala mutations of most residues at the aM4 C terminus that could interact with the Cys-loop or other structures in the ECD have little effect on the measured EC 50 values, as discussed above, suggesting that such interactions are not critical in the muscle nAChR. To further probe the functional role of interactions between aM4 and the Cys-loop, we generated aM4 Cterminal deletions. Deleting up to four residues at the C terminus of aM4 has no effect on channel function ( Table 2). Deleting from five to eleven residues leads to a gradual increase in the measured EC 50 values corresponding to a gradual loss of function, but the maximal loss of function is only 3-fold despite a major disruption in the structure of aM4. Deleting twelve residues or the entire aM4 a-helix (data not shown) leads to a loss of functional expression. These data suggest that interactions between aM4 and the Cys-loop are not critical.
Note that the deletion of nine C-terminal residues (Arg429 to Gly437) led to essentially no further change in the measured EC 50 value than the deletion of eight C-terminal residues (Leu430 to Gly437), despite the fact that the R429A mutant leads to a relatively large 5.3-fold loss of function. The functional importance of Arg429 may only be evident in the presence of the adjacent C-terminal polar/anionic residues Glu432, Asn434, Gln435, and/or Gln436.
Aromatic residues at the aM4-aM1/aM3 interface Given their importance in other pLGICs (30,31,39,44,46), we examined the functional role of aromatic residues at the aM4-aM1/aM3 interface in the nAChR (Fig. 5 and Table 3). In the intracellular leaflet of the bilayer, aM4 Phe414 interacts with aM1 Phe233 and aM3 Phe284 in the 6UWZ Torpedo model but is also close to aM1 Tyr234 in the 2BG9 Torpedo model. As noted, Ala mutation of Phe414 leads to a small 1.7fold gain of function, whereas the F233A and F284A mutations lead to 2.9-fold and 1.6-fold gains of function, respectively. The F414A/F233A double mutant leads to a 7-fold gain of function, slightly more than what would be expected if the two mutations are independent and additive. These three bulky aromatic residues are detrimental to channel function and may sterically prevent optimal aM4-aM1/aM3 interactions, as is observed with aromatic residues at the M4-M1/M3 interface in ELIC.
In contrast, the Ala mutant of Tyr234 does not express, although both Y234L and Y234F give nearly WT EC 50 values, indicating that a bulky hydrophobic residue is required at this position for proper folding and/or function. Intriguingly, Tyr234 lies along the base of a lipid binding site in the 6UWZ structure, where it interacts extensively with a bound lipid and shapes the lipid binding site. The role of this putative lipid binding aromatic residue warrants further exploration.
In the extracellular leaflet of the bilayer, aM4 Phe426 projects toward aM3 Tyr277/Phe280 in the 6UWZ Torpedo model. As noted, F426A mutation leads to a relatively large 3.8-fold gain of function. In contrast, the two aM3 aromatic mutations had minimal effects, with Y277A and F280A leading to a 1.5-fold loss and a 1.8-fold gain of function, respectively. Even the double mutant, F280A/F284A, had little effect on the measured EC 50 value.
We also probed the role of aM1 Phe225 on channel function, as our early models suggested that it projects toward aromatic residues located on aM3. The Functional role of coevolved residues at the aM4-aM1/aM3 interface We used mutant cycles to test whether interactions between coevolved pairs of residues play a role in channel function. In a mutant cycle, the individual effects of mutating two potentially interacting residues are compared with the effect of the double mutant to assess whether the residues influence function independently or through a common mechanism; the latter would be the case if they were involved in a direct interaction. The independence of the functional effects is evaluated using a parameter, V (see Experimental procedures), where an V value of 1 indicates that the two mutations influence channel function independently, whereas V values of increasingly greater than or less than 1 indicate increasing degrees of mutual dependence. Although mutant cycles are ideally used when the rate constants for gating can be defined, this approach has proven informative at the whole-cell level (43,58). We generally consider up to ;2-fold changes in V values as being within the error limits of our measurements (58). Both individual and double Ala mutations were created for each coevolved pair of residues at both the aM1-aM4 (Ile219/ Val425, Gly230/Leu411, Phe233/Phe414, and Tyr234/Leu411) and aM3-aM4 interfaces (Tyr277/Thr422, Phe280/Gly421, Thr281/Cys418, Phe284/Val417, and Thr298/Met406) (Table  S2). At the aM1-aM4 interface, the effects of the single and double mutations are all minimal, suggesting that these coevolved pairs do not play a critical role in either folding or function. Similar effects are also observed with Ala mutations of coevolved residues at the aM3-aM4 interface, except that T281A and T298A do not functionally express. In addition, although the Ala mutations of the evolutionarily coupled residues aM3 Tyr277 and aM4 Thr422 lead to 1.5-fold and 4.1fold losses of function, respectively, the Y277A/T422A double mutant does not express. Although the latter findings lend further support to the hypothesis that Thr422 plays an important role in aM4 function, the collective data suggest that coevolved residues at the aM1-aM3/aM4 interface generally do not play a critical functional role.

The additivity of mutations along aM4
Finally, we asked whether the functional effects of Ala mutations along aM4 are additive. We first created two . If each mutation impacts function via changes only to the local environment surrounding that residue, then we would expect the mutations to be independent and, thus, additive, leading to a 28-fold loss of function and a 10-fold gain of function, respectively. The R429A1 T422A1L411A triple mutant led to only a 6-fold loss of function, close to 5 times less than expected (Table S3). Thus, the effects of these three mutations are not completely independent. Unfortunately, the F426A1F414A1 D407A triple mutant gave no functional expression.
We next created two triple Ala mutants involving residues where the individual Ala mutations led to a mix of loss-and gain-of-function phenotypes. If the mutations are independent and, thus, additive, we would expect the triple mutants, F426A1M415A1H408A and R429A1F414A1D407A, to lead to a 2-fold gain of function and a 2-fold loss of function, respectively. The F426A1M415A1H408A triple mutant gave a 3-fold gain of function, suggesting that these three mutants influence function independently. In contrast, the R429A1F414A1D407A triple mutant did not give functional expression.

Discussion
The long-term goal of this work is to understand the mechanisms underlying the exquisite functional sensitivity of the muscle nAChR to its surrounding membrane environment. The M4 a-helix located at the lipid-exposed periphery of the transmembrane domain of each subunit plays a central role in lipid sensing. M4 likely translates changes in the surrounding membrane environment to altered channel function via its interactions with the adjacent transmembrane a-helices, M1 and M3, although the underlying pathway(s) remains unclear. An important first step in the elucidation of these pathways is to understand the functional role of the M4 a-helices in the different subunits of the a 2 bgd (fetal) or a 2 bed (adult) pentamers. Here, we focused on the functional role of aM4 (i.e. M4 from the a subunit) from the human adult muscle nAChR.
We first examined the functional effects of Ala substitutions at each residue along the entire length of aM4 and found that the effects generally differ both quantitatively and qualitatively from those observed with M4 Ala mutations in other pLGICs, such as GLIC and ELIC. In the nAChR, only 13 of 36 substitutions in aM4 lead to statistically significant changes in the agonist-induced response, whereas a large majority of the Ala substitutions in GLIC (17 of 26) and ELIC (26 of 29) alter channel function. In the nAChR, Ala mutations in aM4 lead to a mix of both loss-and gain-of-function phenotypes, whereas in GLIC and ELIC the vast majority of M4 Ala mutations lead to either loss-of-function or gain-offunction phenotypes, respectively. In the nAChR and ELIC, Ala substitutions for aromatic residues at the aM4-aM1/ aM3 interface generally lead to improved channel function, whereas similar mutations in GLIC lead to losses of function. Finally, although a smaller number of Ala substitutions in aM4 lead to changes in function, the magnitudes of the changes in function of the most impactful mutations are comparable with those observed in the homomeric GLIC and ELIC, even though the mutations in aM4 are repeated in only two of the five subunits.
Collectively, these results suggest that aM4 impacts channel function in a manner that differs from the functional impact of M4 in either GLIC or ELIC. In GLIC, aromatic residues on M1, M3, and M4 interact with each other, creating a tight, complementary network at the M4-M1/M3 interface. Ala substitutions to almost any of these residues lead to a reduction in or a complete loss of expression and/or function, with double aromatic substitutions being especially damaging. A tight, complementary aromatic interface between M4 and M1/M3 is not only essential for both folding and function but also leads to tighter interactions (than those in ELIC) that render GLIC functionally insensitive to its surrounding membrane environment (30,44). A complex network of interacting aromatic residues at the M4-M1/M3 interface is also observed in anionselective pLGICs (59), with Ala substitutions of aromatic residues in these pLGICs generally leading to losses in function and/or expression (44,45).
In contrast, there are relatively few bulky aromatic residues at the M4-M1/M3 interface of ELIC, with Ala mutations of most residues leading to gains in function. Even double aromatic-to-Ala alterations are tolerated and lead to enhanced channel function. It appears that the bulky aromatic side chains prevent the M4-M1/M3 interactions that are required for optimal channel function. They also likely create a more malleable M4-M1/M3 interface that renders ELIC more sensitive to its surrounding membrane environment (30). In fact, increased dynamics in the C-terminal half of M4, which may only be possible in the absence of aromatic residues leading to strong M4-M1/M3 interactions, play a key role in lipid binding that shapes the agonist-induced response (27).
As in ELIC, the nAChR, along with other cation-selective pLGICs, such as the 5-HT 3 R and the neuronal a7 homo- pentameric nAChR, lacks the complex network of aromatic residues at the aM4-aM1/aM3 interface that is found in GLIC. In the a subunit of the muscle nAChR and the a7 nAChR, Ala mutations of aromatic residues at the M4-M1/ M3 interface generally lead to modest gains in function (47). These bulky aromatic residues may prevent optimal M4-M1/ M3 interactions for channel function. A more malleable aM4-a M1/aM3 may also contribute to the exquisite lipid sensitivity of the nAChR, as has been extensively characterized for the Torpedo homolog (5). On the other hand, Ala substitutions along M4 in 5-HT 3A R either have no effect or lead to losses in expression and/or function (46). In this context, it is important to note that the introduction of aromatic interactions at the M4-M1/M3 interface of ELIC to create the same complex network of aromatic interactions that occur in GLIC both leads to an enhancement in channel function and reduces the functional sensitivity of ELIC to lipids (30,60). Whereas similar experiments have not been performed in the muscle nAChR, Trp residues have been introduced along aM4 in the Torpedo nAChR at positions originally thought, based on [ 125 I]-TID labeling, to be exposed to lipids. Although most of these Trp substitutions were detrimental to folding and/or function (22,61), three Trp substitutions now positioned at the aM4-aM1/aM3 interface in the 6UWZ Torpedo nAChR structure lead to relatively large gains of function. It is possible that the addition of an extra bulky aromatic residue at select positions along this interface in the Torpedo nAChR helps to create optimal interactions between aM4 and aM1/ aM3 to enhance channel function. Further work is required to assess the effects of aromatic residues at this interface in both function and lipid sensing.
In contrast to ELIC, however, the aM4 Ala-scan suggests that two residues, Arg429 and Thr422, both play a specific role in channel function, with the R429A and T422A mutations leading to relatively large loss-of-function phenotypes. It is possible that these residues contribute to specific interactions that ultimately drive how aM4 both influences function and senses its surrounding lipid environment. Similarly, in the homopentameric 5-HT 3A R, three mutations along M4, D434A, Y441A, and W459A, lead to losses in expression and/or function. Of particular note, the D434A mutant does not express, possibly as a result of a critical interaction with Arg251 on the M1-M2 loop, which in turn forms a second hydrogen bond with the backbone of Leu244 on M2 (46). The equivalent mutation in the a7 nAChR, D446A, also does not express (47). In contrast, Ala mutation of the equivalent residue in aM4 of the muscle nAChR, Asp407, expresses robustly and produces a slight gain of function. It appears that the pathways by which M4 influences channel function differ in the 5-HT 3A R, the neuronal a7 nAChR, and the a subunit of the muscle nAChR.
Finally, we assessed whether the effects of individual Ala mutations of residues located along aM4 are independent and, thus, additive. Although our data are limited, particularly as two of the four triple mutants do not express, they suggest that the effects of multiple mutations along M4 are additive, although not in a completely independent manner. In other words, although one of the triple mutants gives a loss of function larger than any of the individual mutations, the loss of function is less than expected if the individual effects were simply added together. These data suggest that a change in the orientation of the entire aM4 could lead to changes in a number of local interactions that add up to a significantly larger impact on channel gating. Similarly, lipid-dependent changes in the orientation of aM4 could alter nAChR function through the addition of multiple, subtle changes in local interactions at the aM4-aM1/aM3 interface. Further experiments are required to both test this hypothesis and assess how the M4 a-helix from other subunits in the heteromeric muscle nAChR contribute to channel function.

Experimental procedures
Homology model and coevolutionary analysis Homology models were created using the Swiss-Model online server (RRID:SCR_018123). Evolutionary couplings were determined using the online EVolutionary couplings server (RRID:SCR_018745).

Molecular biology and electrophysiology
Mutants were created from WT human a1, b1, d, and e nAChR sequences in the pcDNA3 vector using QuikChange TM site-directed mutagenesis kits (Agilent) and were verified by sequencing (43). The resulting vectors were linearized and capped cRNA produced by in vitro transcription using the mMESSAGE mMACHINE® T7 kit (Ambion).
Stage V-VI oocytes were injected with 5 ng of mutated a1 subunit cRNA along with 2.5 ng each of WT b1, d, and e subunit cRNA and allowed to incubate for 1 to 4 days at 16°C, as described elsewhere (31). Whole-cell currents were measured in response to acetylcholine concentration jumps using a twoelectrode voltage clamp (TEVC) apparatus (OC-725C oocyte clamp; Holliston, MA) in the presence of 1 mM atropine to prevent the activation of endogenous calcium-activated chloride channels via muscarinic acetylcholine receptors. Whole-cell currents were recorded in HEPES buffer (96 mM NaCl, 2 mM KCl, 1.8 mM BaCl 2 , 1 mM MgCl 2 , and 10 mM HEPES, pH 7.3), with the transmembrane voltage clamped at voltages between 220 mV and 280 mV, depending on the levels of protein expression. Dose responses for each mutant were acquired from at least two different batches of oocytes. Each individual dose response was fit with a variable-slope sigmoidal doseresponse curve. Plots were created using GraphPad Prism, and the individual EC 50 values and Hill coefficients from each experiment were averaged to give the presented values 6 standard deviations. For the presented dose-response curves, the individual dose responses were normalized, and then each data point was averaged. Curve fits of the averaged data are presented, with the error bars representing the standard errors. Statistical significance was tested using a one-way analysis of variance (ANOVA), followed by Dunnett's post hoc test.
EC 50  where WT is the WT control, mut 1 is the first mutant, mut 2 is the second mutant, and mut 1,2 is the double mutant.