Structural switch in acetylcholine receptors in developing muscle

Abstract

During development, motor neurons originating in the brainstem and spinal cord form elaborate synapses with skeletal muscle fibers¹. These neurons release acetylcholine, which binds to nicotinic acetylcholine receptors on the muscle, initiating contraction. Two types of acetylcholine receptors are present in developing muscle cells, and their differential expression serves as a hallmark of neuromuscular synapse maturation^2–4. The structural principles underlying the switch from fetal to adult muscle receptors are unknown. Here, we present high resolution structures of both the fetal and adult muscle nicotinic acetylcholine receptors, isolated from bovine skeletal muscle in developmental transition. These structures, obtained in the absence and presence of acetylcholine, provide a structural context for understanding how the fetal versus adult receptor isoforms are tuned for synapse development versus the all or none signaling required for high fidelity skeletal muscle contraction. We find that ACh affinity differences are driven by binding site access, channel conductance is tuned by widespread surface electrostatics, and open duration changes result from intrasubunit interactions and structural flexibility. The structures further reveal pathogenic mechanisms underlying congenital myasthenic syndromes.

Skeletal muscle contraction is initiated by the release of the neurotransmitter acetylcholine (ACh) from motor neurons onto ACh receptors (AChR) at specialized synapses. Early in development, AChRs are diffusely prepatterned across the surface of muscle cells^5,6. Motor neurons spontaneously release small amounts of ACh^7,8, causing electrical activity in the muscle and repressing receptor expression in extrasynaptic nuclei. Innervation induces receptor accumulation at nerve-muscle contact sites, and formation of neuromuscular junctions (NMJ)^9,10 around the time of birth. In adults, ACh binding to the postsynaptic receptor causes ion channel opening, cation permeation, and local depolarization of the muscle fiber, triggering strong muscle contraction¹¹.

Muscle AChRs undergo a developmental switch in subunit composition at the time of motor neuron innervation (Fig. 1a, b)^2,5,12. Fetal muscle receptors have a high sensitivity to ACh^13–15, lower conductance^3,16,17 and, longer open channel duration¹⁶. These properties enable muscle activity from spontaneously released neurotransmitter, which is essential for the differentiation of the NMJ and ontogenetic death of motor neurons^7,18,19. Substitution of the fetal AChR with the adult isoform leads to abnormal NMJ development^7,20, because of the adult isoform’s low sensitivity to ACh^13,14, higher conductance¹⁷, and briefer openings¹⁶. However, these distinct properties render adult receptors ideally tailored for robust and reproducible high-frequency neuromuscular transmission at mature synapses²¹, where ACh reaches millimolar concentrations²².

Figure 1: NMJ and AChR structures in fetal and adult stages of development.

a, b, Cartoons of immature (a) and mature (b) NMJ synapses illustrating differences in biophysical properties of fetal and adult AChRs, including ACh sensitivity and ion conductance. c, d, Cryo-EM maps of fetal AChRs bound to neurotoxin in a resting-like state (c) and ACh in a desensitized state (d), top and side views. e, f, Cryo-EM maps of adult AChRs bound to neurotoxin in a resting-like state (e) and ACh in a desensitized state (f), top and side views. Unique glycosylation for γ subunit (γN52-glycan) and ε subunit (εN86-glycan) assignment is indicated as cyan colored densities.

Muscle AChRs are composed of five homologous subunits arranged in a ring to form a cation-selective ion channel through the cell membrane^23,24. Fetal receptors contain two α1 subunits, and one each of β1, γ, and δ. In adults, the γ subunit is replaced by ε^2,25, in which most of the mutations causing congenital myasthenic syndromes are found^26–28. To date, there has been no structural information for a bona fide muscle AChR. Recent structures of the Torpedo electric organ AChR^24,29,30, homologous to the mammalian fetal muscle receptor, are the current best model, but fail to address mature receptor physiology, and congenital myasthenia. Here we present structures of native fetal and adult muscle AChRs, revealing how the developmental switch alters channel biophysics and pharmacology to enable NMJ maturation. These findings further illuminate how mutations in the adult receptor cause myasthenic syndromes.

Tissue in developmental transition contains both receptor isoforms

While the density of AChRs at the mature NMJ is high, their overall amount in skeletal muscle is only ~6 μg per 100 g of tissue³¹. We used fetal bovine skeletal muscle for AChR isolation, as bovine muscle is readily obtainable in quantity, and fetal tissue contains higher receptor levels than adult³². We engineered a 3-finger peptide neurotoxin for affinity purification of the receptor on a kilogram tissue scale (see Methods). Briefly, skeletal muscle was homogenized and fractionated by centrifugation. Membranes were solubilized in detergent with phospholipids and cholesterol added throughout purification to stabilize the receptor (Extended Data Fig. 1a). The approach was iteratively optimized and the receptor was tracked using fluorescently-labeled α-bungarotoxin (Extended Data Fig. 1c, d, e). Final yields were ~30 μg of AChR per kg muscle; this purified receptor was applied to electron microscopy grids, frozen, and imaged (Extended Data Figs. 1, 2, 4).

To our surprise, we identified both fetal and adult muscle receptors from this tissue, by mass spectrometry (Extended Data Fig. 1i), and in high-resolution cryo-EM 3D-reconstructions (Fig. 1, Extended Data Fig. 3). Precocial animals that can walk on the day they are born likely complete the developmental switch in subunit composition, to form mature NMJs, well before birth³. Accordingly, even the early-stage fetal skeletal muscle contains both fetal and adult AChRs, serendipitously revealing structures of both from the same cryo-EM sample.

Native receptor structures in different functional states

We refined the toxin-bound fetal and adult receptor cryo-EM maps to 2.1 Å overall resolution (Fig. 1c, e, Extended Data Fig. 2, 5, Video 1). These structures serve as references for a non-conducting, resting-like state of the receptor^24,33,34. To probe neurotransmitter recognition and ACh-induced conformational changes, we displaced toxin with ACh (Extended Data Fig. 1b, h). This competition experiment resulted in structures of ACh-bound fetal and adult receptors, both at 2.5 Å resolution (Fig. 1d, f, Extended Data Fig. 4, 5, Video 1). At equilibrium, saturating ACh results in receptor desensitization; these ACh-bound structures represent non-conducting desensitized states. The high quality of the density maps enabled de novo model building of most of the receptor as well as toxin, ACh, and lipids. The five receptor subunits are arranged as α_γ-γ-α_δ-δ-β (fetal) and α_ε-ε-α_δ-δ-β (adult) around the channel axis (Fig. 1c–f, Extended Data Fig. 6a–e). Each subunit has a typical Cys-loop receptor ECD fold of mostly β strands and a transmembrane domain (TMD) containing four helices, M1-M4. The intracellular domain (ICD) includes an amphipathic post-M3 helix (MX) and an MA helix that is continuous with M4 (Extended Data Fig. 6e). In contrast to the short helical C-terminus in α subunits, the γ, ε, δ, and β subunits have extended C-termini stabilized through interactions with their respective ECDs (Extended Data Fig. 6k–o). In addition to a common, large N-linked glycan emanating from a conserved asparagine in the Cys-loop of all muscle receptor subunits, the γ, ε, and δ subunits each have a unique glycosylation site (Extended Data Fig. 3, 6f–j). The combination of distinctive C-termini and N-glycosylation patterns allowed for the successful 3D classification and proper subunit identification during cryo-EM data processing (Extended Data Fig. 2–4).

Neurotransmitter sensitivity

The fetal AChR has a 30-fold higher affinity for ACh compared to the adult isoform¹³, which enables detection of the sparse ACh released during the earliest stages of NMJ development^7,35. Comparison of the fetal and adult receptor structures bound to ACh reveals differences underlying this sensitivity, which stem from the unique αγ and αε interfaces (Fig. 2a, b). The quaternary ammonium group of ACh is tightly encapsulated by five conserved aromatic core residues from the α (αY113, αW169, αY210, αY218) and γ/ε (γW77/εW75) subunits (Fig. 2c, d). The ACh acetyl group is positioned to form a hydrogen bond with a water molecule coordinated by the main-chain atoms of γN129 on β5 and γL141 on the β6 strand (equivalent to εN127 and εL139) (Fig. 2c, d). In addition, γL141/εL139 are close to the aromatic core and should contribute to stabilization of ACh. Notably, these direct interactions are conserved between the fetal αγ and adult αε ACh sites, suggesting other factors must tune ACh affinity^13–15.

Figure 2: Structural basis of ACh sensitivity.

a, b, Cryo-EM densities highlight the fetal αγ (a) and adult αε (b) interfaces for ACh binding; top views. The ACh density in each interface is shown in the upper-left corners. c, d, Cartoon structures of αγ (c) and αε (d) ACh binding pockets. ACh molecules are shown as yellow sticks. Interacting protein residues are shown as sticks and the interacting water molecules are shown as cyan spheres. e, f, Surface representations of αγ and αε interfaces reveal different compactness of ACh binding pockets. The accessible pocket volume is 91.8 ų in the fetal αγ interface, and 201.2 ų in the adult αε interface; also see Extended Data Figure 7e and 7f.

The structures reveal two striking features that differ between the αγ and αε interfaces to modulate ACh affinity. First, the ACh binding pocket in the αγ interface is more compact than in the αε interface, where the ACh acetyl group is solvent exposed (Fig. 2e, f). This difference in solvent access leads to slower ACh dissociation from the αγ site than from the αε binding site, consistent with predictions from simulations^14,15 (Extended Data Fig. 7e, f). The compact αγ site results at least in part from two bulky substitutions, γY139/εS137 and γE79/εG77 (Fig. 2c, d). γY139 provides a hydrophobic cover, locking ACh in a more extensive aromatic box, and interacts with γE79 to occlude one side of the pocket (Fig. 2c). Second, the ACh sites differ in structural stability. The αγ interface harbors two additional interactions compared to the αε interface. One is an unpredicted disulfide bond between γC137 and γC83 (Extended Data Fig. 7a). Sequence comparisons suggest this clearly resolved disulfide is conserved in mammals, while it is absent in the Torpedo receptor (Extended Data Fig. 7c). This disulfide covalently links the β6 and β2 strands, both of which directly interact with ACh. Earlier simulations support the relative flexibility of the region being a determinant of fetal versus adult AChR sensitivities to neurotransmitter¹³. A final stabilizing element is a hydrogen bond interaction between the main-chain of αN115 and γN75 just adjacent to two core aromatic residues, αY113 and γW77 (Extended Data Fig. 7a, b). This interaction further stabilizes the aromatic box of the αγ ACh site and is absent in the αε site due to a substitution at εS73 (Extended Data Fig. 7d). These two distinguishing interactions endow the αγ interface with greater structural stability, explaining how the αγ site is better organized and less dynamic than αε¹⁴. Together, we conclude that the higher ACh sensitivity of the fetal receptor stems from the unique αγ interface being both more compact and more rigid than the αε interface of adult receptor.

Conductance difference

In a mature muscle cell at rest, the membrane potential is maintained near −85 mV. AChR opening results in non-selective cation permeation; at the strongly polarized potential, conductance is dominated by an influx of Na⁺ ions as well as Ca^{2+ 36}. The adult receptor has a 50% higher single channel conductance than the fetal receptor (~60 pS vs. ~40 pS in bovine muscle³) and a higher relative Ca²⁺ permeability^37,38. Comparisons of the fetal and adult AChR structures explain these distinctions underlying proper functioning of the adult NMJ. Negatively charged residues in the ion permeation pathway, including the polar ECD and ICD linings, and the largely hydrophobic TMD pore, are expected to be major determinants of channel conductance^17,39,40 (Fig. 3a). Indeed, the adult receptor possesses a much more strongly electronegative surface through both the ECD channel vestibule and the ICD fenestrations compared to the fetal receptor (Fig. 3b–e). Lining the ECD vestibule, ε has four anionic residues that are either neutral or hydrophobic in γ (Extended Data Fig. 7g, i). The more negatively charged surface in the ε ECD vestibule would concentrate permeant cations⁴¹. The ε ICD also contains four anionic substitutions, specifically in the MA helix that frames intracellular portals where ions exit (Fig. 3d, e, Extended Data Fig. 7h, j). We found that neutralizing these four charges in the ε ICD decreased the single-channel conductance of the adult muscle receptor by ~20% (Fig. 3i, j), consistent with an observation in the α3β4 ganglionic AChR⁴². Interestingly, the γ ICD contains four positively charged substitutions (Extended Data Fig. 7k–m), similar to the 5-HT_3A receptor, which has an unusually low single-channel conductance (Extended Data Fig. 7n)⁴³.

Figure 3: Electrostatics endow the adult receptor with high conductance.

a, Side view of adult receptor with β subunit removed shows the ion permeation pathway (yellow dashed lines with arrows) that determines channel conductance. b–e, Electrostatic potential of ECD (b, c) and ICD (d, e) inner surfaces of fetal and adult receptors reveal different charge properties. Red, negative; blue, positive. f, Sequence alignment of γ and ε subunits. g, Superposition of the pore regions of fetal and adult structures reveals the K/Q at M2 20′ influences cation flux. h, Structural comparison reveals the residue differences on M2 helices. γV18′ and εI18′ insert into a pocket surrounded by hydrophobic residues. Corresponding residues in g and h are shown as sticks. i, Representative single-channel recordings from adult wild type and εE/K mutant (εD436K/εE438K/εE442K/εE443K) at different voltages. O, open; C, closed. j, Single channel conductance statistics for wild type (n = 5 independent cells, 53.28 ± 1.52 pS) and εE/K mutant (n = 7, 42.27 ± 2.14 pS). Error bars denote the mean ± s.e.m.; Welch’s t-test was used.

The M2 helices from each subunit line the TMD pore. These segments, given prime numbers counting from bottom to top of M2, differ at only four positions between ε and γ (Fig. 3f). γK20′ at the top of the pore orients inward in both the resting-like and desensitized states (Fig. 3g), ideally positioned to tune cation flux. Acidic substitutions here enhance both overall conductance and relative Ca²⁺ permeability in multiple cationic Cys-loop receptors^39,44,45. This repelling, positive charge in γ is neutralized in ε (εQ20′) (Fig. 3g). In the intracellular mouth of the pore, γK268 is replaced by εQ267 (Fig. 3f). Basic residues at these two positions are structurally unfavorable for cation flux, and dampen muscle receptor conductance¹⁷. Both γV18′/εI18′ and γA4′/εS4′ on M2 orient away from the pore axis; how these substitutions affect channel function is more speculative. The 18′ substitution could enhance the hydrophobic interactions with surrounding residues (Fig. 3h) and stabilize the M2 helix conformation during channel opening, which could in turn increase conductance of the adult receptor. The change at 4′ in combination with two K-Q replacements is known to regulate conductance of both muscle receptor isoforms¹⁷. Together, structural and functional data define specific charge reversals and neutralizations along the entire ECD-TMD-ICD permeation pathway that result in the higher conductance of the adult muscle AChR.

State transition reveals an extensive gate

Comparison of the adult resting-like and desensitized state structures reveals asymmetric conformational changes beginning in the ACh binding sites and extending to the channel pore (Fig. 4a–d, Extended Data Video 2). The conformational transitions in the M2 helices result in overlapping gates blocking ion flux in the two observed non-conducting states. In the resting-like state, two hydrophobic constrictions connected by three rings of hydrophobic residues shut the upper half of the pore (Fig. 4e, g). The upper constriction at 16′ results in a minimal diameter (d_min) of 2.3 Å. Interestingly, a phenylalanine (βF6′) orients into the pore, and with the L9′ ring together form the lower constriction point (d_min 1.7 Å) (Fig. 4g). In the desensitized state, the pore diameter at the 16′ gate widens from 2.3 to 6.5 Å, sufficient for hydration of the extracellular ~third of the pore (Fig. 4f). Movement of M2 helices away from the pore axis results in rotation or translation of L9’ side chains and βF6′, in a subunit-dependent manner (Fig. 4d), and an increase in diameter at the lower constriction (1.7 to 4.0 Å; Fig. 4f, g). This increased diameter is not expected to allow permeation of hydrated cations⁴⁶. Accordingly, in the muscle receptor, the L9′ ring and βF6′ act in concert to form the desensitization gate while also contributing to the more extensive resting-state gate.

Figure 4: State transitions and gates in adult AChR.

a, Comparison of adult resting-like (light grey) and desensitized (colored) state structures reveals ECD conformational changes upon ACh binding. Toxin removed for clarity. b, c, Conformational changes in the coupling region (b) and the TMD regions (c) induced by ACh binding; colors as in a. d, Close-up top view of L9′ and F6′ orientations in different states; protein residues are shown as sticks; colors as in a. e, Permeation pathway in the resting-like state pore region; red dots indicate pore diameters less than 2.3 Å. f, As in e, but for the adult desensitized state; red dots here indicate pore diameters less than 4.1 Å. Pore lining residues in e and f are shown as sticks except βG2′, shown as a sphere. g, Pore profile of adult resting-like and desensitized structures. h, Comparison of whole cell patch clamp electrophysiology of adult wild type and mutants. i, Apparent desensitization rates (τ) of adult wild type (n = 5 independent cells, 0.40 ± 0.08 s), βF6′S (n = 6, 0.24 ± 0.02 s), βL9′S (n = 6, 1.84 ± 0.13 s), βF6′S/L9′S (n = 10, 0.87 ± 0.07 s) mutants. Welch’s analysis of variance (ANOVA) with Dunnett’s multiple comparisons test was used. Mean ± s.e.m.; boxes indicate the 25th to 75th percentiles, whiskers indicate the minimum and maximum values, and the central line shows the median; all data points are shown. j, Representative single-channel recordings from adult wild type and βF6′S mutant. O, open; C, closed.

The 6′ phenylalanine is strikingly rare among Cys-loop receptors, found only in mammalian muscle receptor β subunits and in the glycine receptor β subunit (Extended Data Fig. 9h, i)^47,48. While the L9′ ring has been studied extensively^49,50, βF6′ appears to have escaped notice; to see it occluding the permeation pathway in both the resting-like and desensitized states suggests it could provide a muscle receptor-specific control on permeation or gating. In macroscopic patch-clamp electrophysiology recordings, we found that the adult wild-type receptor desensitized rapidly, while the βF6′S mutation resulted in a slight increase in the apparent rate of desensitization (Fig. 4h, i). We were curious whether the strong L9′ gate masked effects from the βF6′S mutation, and thus tested single and double mutants including βL9′S. The βL9′S single mutant desensitized slowly and was more sensitive to ACh, consistent with previous reports^49,50 (Extended Data Fig. 8a). Remarkably, the βF6′S+βL9′S double mutant largely restored the characteristically fast desensitization of the wild type receptor (Fig. 4h, i). In single channel analysis, we found that the βF6′S mutation results in briefer openings, a kinetic defect consistent with the macroscopic result (Fig. 4j, Extended Data Fig. 8b, c). These findings suggest that replacement of βF6′ with a polar residue, as found in nearly every other subunit, results in a fast-desensitizing channel. Our structural and functional data point to the L9′ ring being a major pore element controlling desensitization, with the βF6′ residue counteracting the stability of the desensitization gate. How might a bulky hydrophobic residue, found to occlude the pore in two non-conducting states, in fact enhance open channel duration? We suggest that the local environment of βF6′ is the key; all other subunits have a polar side chain at 6′, creating an unfavorable environment for the β subunit’s phenylalanine. An activated-state structure of the glycine receptor suggests that 6′F rotates out of the pore and into a hydrophobic pocket (Extended Data Fig. 9l, m)⁴⁸. If the same occurs in the muscle AChR, we propose that βF6′ can impede conductance in resting and desensitized states through steric pore block, while also stabilizing the activated state by virtue of moving out of the pore and into a more favorable hydrophobic environment. This regulation is especially important in the muscle receptors that must respond to ACh with precise timing; openings that are either too brief or prolonged can be pathogenic, causing congenital myasthenic syndromes⁵¹.

Importantly, we found that the conformational transitions between fetal and adult muscle receptors were similar to each other (Extended Data Fig. 9a–g). Indeed, the resting-like and desensitized-state structures between fetal and adult have a root-mean-square deviation (RMSD) under 0.5 Å, suggesting high structural similarity (Extended Data Fig. 9j, k), consistent with desensitization not differing between these two isoforms^21,52. Thus, although the fetal and adult muscle receptors are distinct in many channel properties, they appear to share a common gating mechanism.

Channel open duration

The fetal receptor has a longer open channel duration than the adult isoform^3,11. This kinetic tuning allows fetal miniature end-plate currents (MEPCs) to trigger spontaneous contractions in the developing muscle fiber. Adult MEPCs do not cause contractions, as the AChR channels open too briefly⁷, but in contrast respond and then shut quickly during evoked transmission in the mature NMJ²⁰. Comparison of the fetal and adult AChR structures suggests determinants of open time underlying development versus coordinated muscle contraction. We observed an electrostatic interaction network mediated by γT316 in the fetal receptor TMD, stabilizing the helical bundle (Fig. 5b). This network is disrupted by an alanine substitution (γT316 to εA315) in the adult AChR (Fig. 5c). We found that the γT316A mutant subtly decreases mean channel open time (Fig. 5d, Extended Data Fig. 8e, f), suggesting that multiple factors are required to explain the difference in channel open duration. Indeed, other residues, including γM482, regulate channel open time¹⁶. We find that γM482 inserts into a local hydrophobic pocket surrounded by several residues from M1, M3 and M4 (Fig. 5e). This interaction that stabilizes TMD helical packing is abolished by an alanine substitution in the adult (εA462) (Fig. 5f). Further, many congenital myasthenic syndromes (CMS) mutations, like εL289F and εV279F, increase TMD hydrophobic interactions, and have a longer open time (Extended Data Fig. 8j, k). These results suggest that stronger intrasubunit TMD interactions in the fetal vs. adult receptors act together in regulating channel open time.

Figure 5: Channel open time.

a, Structural comparison of fetal and adult TMD and ICD regions. b, c, Close-up views of different interactions in fetal and adult TMD regions. Interacting residues are shown as sticks. d, Representative single-channel recordings from fetal wild type, γT316A and γG461A mutants. O, open; C, closed. e, f, Close-up views of two key residues from the M4 helices and their interactions.

A 30-residue segment was found to tune the open durations of the two muscle receptor isoforms¹⁶. This segment corresponds to the MA helix, which forms the ICD portals and is continuous with M4 (Fig. 5a). The most striking difference between fetal and adult MA helices is in electrostatic potential (Extended Data Fig. 7k–m, q). However, a charge-reversal mutant did not change the channel open time (Fig. 3i). In activated-state structures of the α7 nicotinic and 5-HT_3A receptors, the MA-M4 junction is suggested to unwind, to enlarge ICD portals for ion permeation and accommodate outward movement of TMD helices^54,55 (Extended Data Fig. 7o, p). A glycine in the MA helix of the 5-HT_3A receptor introduces greater flexibility at the MA-M4 junction to facilitate their disruption⁵⁴. We observed more glycines in the MA-M4 junction in the fetal receptor compared to the adult (Extended Data Fig. 7q). Indeed, we found that glycine mutants, like γG461A, decrease the channel open time compared to wild type (Fig. 5d, Extended Data Fig. 8g, h). Together, we conclude that the difference in channel open duration between the fetal and adult receptors is governed by multiple factors, including the extensive intrasubunit interactions present in the fetal receptor that are absent in the adult isoform, and an increase MA structural flexibility.

ε subunit mutations cause congenital myasthenic syndromes

Mutations in the muscle AChR are the most common cause of CMS⁵¹. Approximately 80% of the mutations are found in the ε subunit^26–28, and fall into two groups: those whose major effect is to alter channel kinetics; and those whose major effect is to reduce AChR expression. Here we leverage the structure of the ε-containing adult AChR to understand mechanisms of CMS pathogenesis (Fig. 6, Extended Data Fig. 8i–l). Among the kinetic mutants, fast-channel syndromes are typified by abnormally brief ion channel openings. Mutations concentrate in the ECD and ICD regions (Fig. 6a). εD195 is located on loop F and forms the only hydrogen bond across the interface with loop C (to αA211), pinning loop C shut around ACh (Fig. 6b). The εD195N mutation either abolishes or weakens this interaction, explaining the increased rate of ACh dissociation^56,57. Slow-channel syndromes are typified by enhanced channel opening or delayed channel closing, resulting in longer-lived open states⁵⁷. Slow-channel mutations cluster in the TMD, particularly in the pore-lining M2 helix (Fig. 6a). εV285A is the only pathogenic mutation on the εM2 helix wherein the side chain orients toward the channel axis to directly influence ion permeation. This 13′V position is conserved among the α, β, γ, δ and ε subunits, and forms part of the hydrophobic upper half of the pore and the strong barrier to hydration in the resting-like state (Fig. 6c). Pruning the side-chain to alanine would increase open probability by destabilizing the resting state; indeed, εV285A was found to increase the rate of channel opening by 6-fold⁵⁸.

Figure 6: Structural context for congenital myasthenic syndromes.

a, CMS mutations mapped on adult ε subunit; blue, cyan, and yellow spheres indicate fast-, slow-channel, and receptor deficiency mutations respectively. For additional deficiency mutations also see Extended Data Figure 8j. b, Interaction between loops C and F in the desensitized state. ACh is shown as yellow sticks. c, εV285 at 13′ of εM2 forms a ring with four other valines to influence ion permeation. Gray arrow represents permeation pathway. d, The ε C-terminus forms a disulfide bond with the ECD. Corresponding residues in panel b, c, and d are shown as sticks. e, Sequence alignment indicates a conserved disulfide bond in the ε C-terminus among mammalian muscle receptors.

Many CMS patient mutations lead to primary receptor deficiency, a catch-all term for mutations that result in ~90% loss of AChRs on the motor endplate (Extended Data Fig. 8l)⁵¹. Among them, C490 is essential for cell surface expression and replacement of C490 by serine did not restore the receptor deficiency⁵⁹. The pathogenic basis of this seemingly benign substitution has remained mysterious. Surprisingly, we found that this ε subunit C-terminal cysteine forms a disulfide bond with εC210, and would thereby stabilize the conformation of the C terminus and perhaps loop F as well (Fig. 6d). This disulfide bond interaction is conserved among mammalian muscle AChRs but absent in the preceding models of the Torpedo AChR (Fig. 6e). Together, mapping disease mutants on the muscle receptor structures provides new confidence in defining CMS pathological mechanisms.

Conclusions

Here we sought to determine the structure of the muscle AChR and how changes in it enable proper NMJ development. By isolating receptors from muscle tissue in a developmental transition, we were able to directly compare the structures of mammalian fetal and adult receptors. We found that a looser and more flexible ACh binding site in the adult receptor explains its adaptation to the high ACh concentrations at the synapse. We found that differences in electrostatic surfaces endow the two isoforms with different ion conductances geared toward their roles first in synapse formation and then in fast chemical neurotransmission. We further uncovered a pattern of extensive intrasubunit contacts specific to the fetal receptor that, when broken, result in shorter channel open times, like in the adult receptor. A surprise in the transmembrane pore revealed an extensive gate that includes a bulky hydrophobic residue tuning desensitization kinetics. These findings, integrated with decades of functional work on this prototypical receptor, and new electrophysiological tests, provide a detailed view of receptor changes in NMJ development and pathological mechanisms underlying CMS.

Methods

Preparation of the toxin affinity reagent

To extract receptor from tissue, we modified a consensus short chain neurotoxin⁶⁰ to have two affinity tags, a polyhistidine tag for purifying the toxin itself from bacteria, and a FLAG tag to extract native receptors from muscle tissue. The pQE60-6his-3FLAG-ScNtx plasmid was transformed into K-12 derived SHuffle Express Competent B E. coli cells (NEB) for expression. The positive SHuffle cell colonies were first tested for toxin expression. For small scale screening of expression, the individual colonies were selected to grow in 5 mL of LB medium containing 100 μg/mL of ampicillin and 1 mM IPTG (isopropyl β-D-thiogalactoside; Sigma) at 37 °C overnight in a shaking incubator at 250 rpm. Cells were harvested and lysed using 30 μL plasmid extract lysis buffer. Expression was evaluated qualitatively by SDS-PAGE and Western blotting using anti-his-tag and anti-flag tag antibodies. The best expressing clone was used to perform large overexpression experiments.

For large scale expression and purification, an initial bacterial culture was grown overnight at 30 °C at 200 rpm in 5 mL LB plus ampicillin. Cells were diluted to 8 L in modified medium⁶⁰ and grown at 37 °C to an OD₆₀₀ of 1. Bacteria were then induced with 0.1 mM IPTG at 16 °C and 210 rpm for 20 h. Cells were next centrifuged at 5,000 × g, 20 min, 4 °C and the pellet was resuspended in protein lysis buffer (50 mM Tris-HCl pH 7.4, 500 mM NaCl, 15 mM imidazole and 1 mM PMSF (phenylmethylsulfonyl fluoride). The cell suspension was then passed four times through an Avestin C5 cell disruptor at 5,000-10,000 psi. The lysate supernatant was collected after centrifugation at 30,000 rpm, 4 °C for 30 min, and loaded onto a Ni–NTA affinity column, washed, and eluted. The peak fractions were concentrated and loaded onto a Superdex 75 Increase 10/300 GL column (Cytiva) for further purification. Purified toxins were stored at −80 °C.

ACh receptor preparation from bovine skeletal muscle tissue

One kg flash frozen fetal bovine skeletal muscle tissue at the 1^st-2^nd trimester (Animal Technologies) was weighed and then broken with a hammer into little smaller pieces. All further steps were performed at 4 °C except for the filtration step to remove fat (see below). The tissue was brought to ~4.2 L with buffered solution (50 mM Tris-HCl pH 7.4, 150 mM NaCl, 10 mM EDTA, 10 mM imidazole, 10 mM iodoacetamide, 1 mM PMSF). The tissue was homogenized using a Ninja blender at the max speed for 2 min (coarse blending), then transferred into a Waring blender for 2 min at speed 5 (for fine blending). The insoluble material was then separated by centrifuging at 11,000 g for 2 h, resulting in a hard pellet, and the supernatant. The ~4.2 L supernatant was then centrifuged at 40,000 rpm for 1 h to precipitate the membrane pellet, the so-called soft pellet. Both the hard pellet and soft pellet contained ACh receptor protein based on the fluorescent-bungarotoxin FSEC assay; they were combined together for further extraction and purification.

M2 FLAG resin (Sigma, 2.5 mL) was washed using 200 mL cold TBS. The purified toxins were loaded onto the resin and recirculated at 4 °C for 5 h. Then, flow-through fractions were collected and the resin was washed using cold TBS again until no free toxin was detected in the flow-through fractions. The hard and soft pellets were resuspended in 2.2 L buffer containing 50 mM Tris-HCl pH 7.4, 150 mM NaCl, 1 mM EDTA, 1 mM PMSF and cOmplete protease inhibitor cocktail (Sigma), and homogenized again for 30 s using the Waring blender. Next, Triton X-100 (Sigma) was added to a final concentration of 2.8%. After solubilization for 2 h nutating at 4 °C, the extract was centrifuged at 10,000 g for 1 h. The lipid layer and small fat drops were removed by passing through regularly exchanged filters (~30 exchanges per preparation, 185 mm Whatman filter papers, Cat No 1001-185) that were cut into circles and fitted into a bottle top vacuum filter holder. The clarified detergent extract was loaded onto the FLAG resin column pre-bound with toxin.

The ~2 L of extract was passed through the column using a peristaltic pump at 4 mL/min and 4 °C. To ensure sufficient binding, the flow through was recirculated through the resin for 40 h. After binding, the resin was washed with 200 mL washing buffer (50 mM Tris-HCl pH 7.4, 150 mM NaCl, 1 mM EDTA, 1 mM PMSF, 1 mM LMNG (Lauryl Maltose Neopentyl Glycol, Anatrace) and 10 μM soy polar lipids (Avanti Polar Lipids) plus cholesterol (Sigma) and DDM (n-dodecyl-β-D-maltoside, Anatrace) in a 4:1:1.84 w:w ratio). Wash fractions were analyzed by FSEC using intrinsic tryptophan signal to confirm that major contaminants were removed. ACh receptors bound to toxin were eluted using wash buffer supplemented to contain 200 μg/mL 3X FLAG peptide (Sigma), 0.25 mM LMNG and 10 μM soy lipids plus cholesterol. Samples were analyzed by FSEC using intrinsic tryptophan signal and fluorescently-labeled α-bungarotoxin (Thermo Scientific), pooled, and concentrated for preparative SEC (in 50 mM Tris pH 7.4, 150 mM NaCl, 0.25 mM LMNG, and 10 μM soy lipids plus cholesterol) using a Superose 6 Increase 10/300 GL column (Cytiva) and SDS-PAGE analysis. Peak SEC fractions were pooled and prepared for cryo-EM or alternately further treated to replace bound toxin with ACh, as described below.

Preparation of native receptors bound to ACh

We extended our toxin-based purification method to prepare the native receptors bound to acetylcholine. First, excess free toxin eluted with the receptor from flag resin was removed using a Superose 6 SEC column. The peak fractions containing the receptors were collected, pooled, and concentrated to 1 mL, and then this sample was diluted into 25 mL modified SEC buffer containing 1 M ACh, 0.25 mM LMNG, and 10 μM soy lipids plus cholesterol. The diluted sample was loaded onto a 1 mL FLAG resin column that was freshly washed with cold TBS. The sample was recirculated through the column for 3 days at a speed of 2 mL/min at 4 °C. After this binding step, the receptors that remained toxin-bound would re-bind to the FLAG resin, while the flow-through fractions should contain mainly ACh-bound receptors. The flow-through fractions were pooled, concentrated, and loaded onto a Superose 6 SEC column with a mobile phase containing 50 mM Tris-HCl pH 7.4, 150 mM NaCl, 0.25 mM LMNG, 2 mM ACh and 10 μM soy lipids plus cholesterol, and the peak fractions were pooled and concentrated for further analysis and cryo-EM sample preparation.

Cryo-EM sample preparation

Three μL of purified proteins at a concentration of 0.2 mg/mL were added to a freshly glow discharged (30 mA, 10 s) 300-mesh copper grid covered with a 2 nm carbon film (R2/1, Quantifoil). Grids were immediately blotted for 3.5 s under 100% humidity at 4 °C with bolt force −15. Grids were plunge-frozen into liquid ethane using a Vitrobot. Grid quality was screened on the UCSD 200 kV Talos Arctica microscope. The grids with good ice, particle distribution, and particle density were saved.

Cryo-EM data collection and image processing

The toxin-bound sample EM data were comparatively straightforward to process. 8,050 dose-fractioned images were collected on the UCSD Titan Krios 2 at 300 kV with a total dose of 50 e⁻/Ų and a magnification of 130,000x and 0.935 Å/pixel. The defocus range was set from −1.6 to −2.2 μm. All data processing was done using cryoSPARC v4.3 and 4.4⁶¹. Dose-fractionated images were gain-normalized, motion corrected, and CTF estimation was done using default parameters in cryoSPARC. Particles were picked using the blob picker and extracted with a box size of 400 pixels and binned to 200 pixels. After several rounds of 2D classification, ~1.5 million particles were selected from classes with clear secondary structural features. Twenty thousand particles were randomly selected to generate a 3D model using ab initio reconstruction. All particles were then submitted into a 3D Homogeneous Refinement job, which revealed compositional heterogeneity at the γ/ε position. 3D classification was performed with 10 classes resulting in two good classes with stronger density and clear ICD and 8 classes with very weak ICD or/and TMD density. These two good classes exhibited very clear and distinctive features in their ECDs and M4 C-terminal extensions. These two good classes were re-extracted at full pixel size and subjected to Non-uniform (NU) Refinement⁶², optimizing per-particle defocus and per-exposure-group CTF parameters in each refinement iteration. NU refinement generated maps at a resolution of 2.09 Å for the adult receptor (209,258 particles) and 2.14 Å for the fetal receptor (168,720 particles), respectively. DeepEMhancer⁶³ was used to improve the map quality before model building.

For the ACh-bound sample, ~30,000 images were collected using the settings as above except for a defocus range from −1.6 to −2.0 μm. All data processing was again performed using cryoSPARC v4.3 and 4.4, except for mask generation performed in Relion v4.0⁶⁴. After several rounds of 2D classification, ~3 million particles were kept and subjected to 3D Homogeneous Refinement using an initial model generated from randomly selected particles. The 3D reconstruction suggested that a small fraction of the particles contained bound toxin, and many particles had a damaged ICD (Extended Data Fig. 4). To first remove the toxin-bound particles, a soft mask focused on the ECD was generated using Relion_mask_creation based on the 3D homogeneous refinement map. The output particles from the initial refinement were then subjected to 3D classification (4 to 5 classes) using the ECD mask, then all particles without toxin bound were selected. To separate the fetal and adult receptor particles, a second smaller soft mask focused on the γ/ε ECD position was generated. The selected particles were further classified using this mask and all fetal and adult receptor particles were selected. To separate the complete particles from those without ICD density, a third soft mask focused on the TMD and ICD was generated. Both fetal and adult particles were classified using this focused mask; only one class for the fetal and one class for the adult receptor had strong and clear ICD density. To further improve the ICD region of γ and ε subunits, an additional 3D classification was performed for both fetal and adult particles. These classes were selected, re-extracted at full size and refined using cryoSPARC NU Refinement to a resolution of 2.45 Å for the adult receptor (167,900 particles) and 2.45 Å for the fetal receptor (259,740 particles).

Model building, refinement, and validation

Because of the high map resolution, we tested ModelAngelo⁶⁵ in building models de novo into the toxin-bound fetal and adult ACh receptor maps. This auto building software generated two models that fit the maps well using supplied sequences, including the toxin, except in selected regions with worse local resolution. These two starting models, which included several fragments, were linked and checked manually residue by residue in Coot (version 0.9.8.93)⁶⁶ along with global real space refinement in Phenix (version 1.20.1)⁶⁷ with secondary structure and Ramachandran restraints. Model geometry and clash scores were checked using MolProbity⁶⁸.

For the ACh-bound datasets, the refined toxin-bound structures were used as the starting models, and roughly fitted into the new density maps using UCSF Chimera (version 1.16)⁶⁹. The pore and loop C regions, which exhibited larger conformational changes compared to the toxin-bound structures, fitted poorly and were manually rebuilt. ACh molecules were placed into unambiguous density in the classical neurotransmitter sites. We observed non-continuous, non-protein densities at the ECD-TMD junction in the channel vestibule, which were not present in the toxin-bound structures. These features were unmodeled due to ambiguity in composition. The initial models were refined against the two new ACh-containing maps in Phenix through several cycles of real space refinement. The improved models were then checked manually residue by residue in Coot along with global real space refinement in Phenix with secondary structure and Ramachandran restraints. Model geometry and clash scores were checked with MolProbity. Sequences for model building were downloaded from the UniProt database and sequence alignments were made using Clustal Omega⁷⁰.

The pore diameter was calculated by HOLE 2⁷¹. The map and structural figures were generated using UCSF Chimera, ChimeraX (version 1.7.1)⁷² and PyMOL (version 2.5.5, Schrodinger, LLC). The main text figures were made from density maps after post processing with DeepEMhancer. Density consistent with several cholesterol and phospholipids was present in each map, in positions also identified in a Torpedo receptor structure²⁹. These densities were not very strong, but were built based on consistency with the earlier models.

Mass spectrometry analysis

Proteins were identified after SDS-PAGE separation using standard methods⁷³ in the UCSD proteomics core. Protein bands near 50 kD were cut out from a gel stained with Coomassie blue using a clean blade. Gel slices were cut into 1 mm by 1 mm cubes and treated with 100 mM ammonium bicarbonate and acetonitrile to destain. The samples were then dried, reduced, alkylated, dehydrated, and subjected to trypsin digestion. After overnight digestion, peptides were extracted, and the combined extractions were analyzed using liquid chromatography and tandem mass spectrometry (LC-MS/MS) with electrospray ionization. The TimsTOF 2 pro mass spectrometer (Bruker) was utilized with a nano-scale reversed-phase UPLC system (EVOSEP ONE). The mass spectrometry settings included a PASEF method, with mobility-dependent collision energy ramping and target intensity per precursor. Peaks Studio X software was used for protein identification and label-free quantification.

Electrophysiology

Human α, β, δ, γ and ε gene sequences were downloaded from the UniProt database, synthesized by GenScript, and subcloned into the pEZT-BM vector⁷⁴. Mutations were introduced by site directed mutagenesis. Whole cell and single channel voltage-clamp recordings were made from adherent HEK293S GnTI⁻ cells transiently transfected with human α, β, δ, and either γ or ε (or their mutants) plasmids. Each 10 mm well of cells in a 12-well dish was transfected with 0.1–0.4 μg plasmid mixture of each subunit or their mutants in a ratio of 2α:1β:1δ:10γ/ε, using Lipofectamine 2000 reagent (Invitrogen). The transfected cells were incubated at 30 °C. After 48 h post-transfection, cells were re-plated on 35 mm dishes and allowed to settle for at least 3 hours. Recordings were made 48-96 h after transfection.

For whole cell voltage-clamp recording, the bath solution contained (in mM): 140 NaCl, 2.4 KCl, 4 MgCl₂, 4 CaCl₂, 5 HEPES and 10 glucose pH 7.3. Borosilicate pipettes were pulled and polished to an initial resistance of 2–4 MΩ, filled with the pipette solution containing (in mM): 100 CsCl, 30 CsF, 10 NaCl, 10 EGTA, and 20 HEPES pH 7.3. Whole cell currents were recorded with an Axopatch 200B amplifier, sampled at 20 kHz, and low-pass filtered at 2 kHz using a Digidata 1550B (Molecular Devices). Cells were held at −75 mV. Acetylcholine was prepared in bath solution from a 1 M concentrated stock stored at −80 °C and thawed twice per day. Solution exchange was achieved using a gravity driven RSC-200 rapid solution changer (Bio-Logic). Whole cell currents data were collected with pClamp 11 and analyzed with Clampfit 11 software (Molecular Devices). Desensitization curves were fitted with a single-phase exponential equation in Clampfit 11 software.

Single channel recordings were obtained in the cell-attached configuration. Bath solution contained (in mM) 140 KCl, 2.4 NaCl 4 MgCl₂, 4 CaCl₂, 5 HEPES and 10 glucose pH 7.3. Pipettes with a resistance ranging from 3–5 MΩ were filled with bath solution plus 10 μM acetylcholine. The single channel recording data were collected with an Axopatch 200B amplifier, sampled at 50 kHz, and low-pass filtered at 10 kHz using a Digidata 1550B (Molecular Devices). A command voltage ranging from 50 mV to 125 mV was applied to the interior of the patch pipette to establish the membrane potential. Detection of single channel open and closing, and the analysis of open duration times and single channel amplitude at a given commend voltage, were performed using TAC4.3.3 software (BruxtonCorp.).

Statistics and Reproducibility

For the whole-cell and single-channel electrophysiology experiments, statistical analysis was performed using GraphPad Prism 10.2.0 software (GraphPad software, Inc, La Jolla, CA). Data are expressed as mean ± s.e.m., the Welch’s t-test was used for Fig. 3j, and Welch’s analysis of variance (ANOVA) with Dunnett’s multiple comparisons test for Fig. 4i. In Fig. 4i, boxes indicate the 25th to 75th percentiles, whiskers indicate the minimum and maximum values, and the central line shows the median; all data points are shown. Replicate numbers n in each figure are the number of independent cells that the recordings are taken from.

Extended Data

Extended Data Figure 1: Isolation and identification of ACh receptors from bovine skeletal muscle.

a, Overview of methods for isolation of AChRs from one kg bovine muscle tissue. b, A method for preparation of ACh-bound receptors. c, Schematic of the fluorescently-labeled α-bungarotoxin detection assay. d, e, Identification of receptor from both hard pellet (d) and soft membrane pellet (e) using the fluorescently-labeled α-bungarotoxin detection assay. f, A final sample reveals strong tryptophan and α-bungarotoxin FSEC peaks. For d–f, chromatograms were obtained using a Sepax SRT SEC-500 column on a Shimadzu HPLC, flow rate of 0.35 mL/min. g, h, SDS-PAGE gels of purified bovine receptor bound with either short chain toxin (g) or ACh (h); for gel source data, see Supplementary Figure 1. This gel is representative of two gels run on independently purified receptors. i, Mass spectrometry identification of bovine receptor subunits.

Extended Data Figure 2: Cryo-EM data processing of toxin-bound receptors.

All steps were carried out in Cryosparc v4 (see Methods for details).

Extended Data Figure 3: Map quality enables subunit assignment.

a, b, Different map features of γ (a) and ε (b) subunits. Three types of structural features help to clearly distinguish the fetal and adult receptors and likely enabled facile classification of particles in cryo-EM data processing. First, glycosylation patterns are distinct. The γ subunit is glycosylated at Asn52 (γN52-glycan) located in the extracellular domain (ECD) β-sheet region (β5), while the ε is glycosylated at Asn86 (εN86-glycan) in a short loop adjacent to the N-terminus of the η2 helix on the apex of the ECD. Second, the C-termini of the γ and ε subunits differ in both length and conformation. The C-terminal extension in γ is larger and stabilized mainly by hydrophobic interactions with the ECD (c), while the C-terminus of the ε subunit is short, and stabilized by a disulfide bond (d). Third, there are amino acid differences in the initial N-terminal helix of each subunit, for example the fetal isoform has a glycine (γG31), and the adult has a histidine residue (εH29) at the corresponding position. To show glycosylation density clearly here, the threshold for the glycosylation map regions is lower than for the protein residues.

Extended Data Figure 4: Cryo-EM data processing of ACh-bound receptors.

All steps were carried out in Cryosparc v4 (see Methods for details).

Extended Data Figure 5: Cryo-EM FSC plots, local resolution, and angular distribution.

a–c, Masked FSC curve, local resolution, and angular distribution of particles for fetal ACh receptor + toxin. d–f, as a–c but for adult ACh receptor + toxin. g–I, As a–c but for fetal ACh receptor + ACh. j–l, As a–c but for adult ACh receptor + ACh.

Extended Data Figure 6: Structures of fetal and adult ACh receptors in different states.

a–d, Structures of fetal and adult receptors bound to either toxin or ACh; top and side views. e, αε interface labeled as an example to indicate domains, helices, glycans, and ACh binding pocket. f–j, ECD of each subunit to show the glycans. k–o, Different C-terminus from each subunit. p, Toxin densities in the αε and αδ interfaces. Density is shown for three residues on the tip of the toxin molecules inserting into the ACh binding pocket to illustrate confidence in interpreting interactions. q, r, Representative lipid densities in the TMD region. q, A phospholipid binds to the bottom of the TMD in the adult resting-like structure. The surrounding residues around the phospholipid are shown as sticks in the β-α interface. r, A phospholipid binds to the top of the TMD in the adult resting-like structure. The surrounding residues around the phospholipid are shown as sticks in the α-δ interface.

Extended Data Figure 7: Structural basis of ACh sensitivity and conductance.

a, b, Close-up views of fetal αγ (a) and adult αε (b) ACh binding pockets. Residues are shown as sticks and waters as cyan spheres. The αγ interface has two extra interactions, a disulfide bond and a hydrogen bond, compared to αε. c, d, Sequence alignments of the residues involved in these interactions. e, f, Accessible area and volume of ACh binding pocket in fetal and adult interfaces calculated using the program CASTp (http://sts.bioe.uic.edu/castp). The accessible volumes are shown as semi-transparent red chambers. g, h, Structures of fetal ECD (g) and ICD (h) show the non-negatively charged residues; γQ35, γN38, γT105, γY127 in ECD and γQ456, γT458, γS462, γG463 in ICD. i, j, Structures of adult ECD (i) and ICD (j) show the negatively charged residues at the corresponding positions of fetal receptor; εD33, εD36, εE103, εE125 in ECD and εD436, εE438, εE442, εE443 in ICD. k, l, Electrostatic potential of ICD outer surfaces of fetal (k) and adult (l) receptors reveal differences from MA helices; red, negative; blue, positive. m, n, ICD structures of fetal muscle receptor (m) and 5-HT_3A receptor⁵⁴ (n, PDB 6BE1) show that both have several positively charged resides in their MA helices; in the fetal receptor, γR452, γR454, γH459, and γK465 have been replaced by εS432, εT434, εA439, and εS445 in adult. o, p, Single subunit structure of the α7 nicotinic ACh receptor (o) and 5-HT_3A receptor^54,75 (p) in both resting and activated states show that the M4-MA junctions are disrupted during channel opening. q, Sequence alignments of MA helices of bovine and human muscle isoforms reveal differently charged residues; negative, red; positive, blue.

Extended Data Figure 8: Electrophysiology analysis of gating residues and congenital myasthenic syndrome sites in the ε subunit.

a, Representative whole cell patch-clamp electrophysiology recordings from the adult wild type AChR and mutants. b, c, Representative single channel cluster duration histograms fitted by a sum of exponentials for adult wild type (b) and βF6′S mutant (c) AChRs; fitted curves for each component (dashed lines) and their sum (bold lines, the peak is indicated by a black line) are displayed; n ≥ 5 cells for each recording. d, Representative single channel recordings from βL9′S and βF6′S/L9′S mutants. e–h, As in b, c, but for fetal wild type (e, n = 3), γT316A mutant (f, n = 3), γG461E mutant (g, n = 4) and γG461A mutant (h, n = 3). The γT316A mutant subtly shortens the longer duration component of opening and the γG461E mutant slightly reduces the frequency of the longer duration component of opening while the γG461A mutant decreases the channel open time. i, Coupling region of ε subunit showing a site of two mutations causing fast channel CMS in patients. εE204 and εR238 form an electrostatic interaction network with εE65 on the β1β2-loop and εD158 on the Cys-loop, which couple ACh binding to channel opening. Both εE204K and εR238W mutations would destabilize this network and thereby diminish coupling efficiency, which would slow the rate of channel opening, leading to fast-channel syndrome⁷⁶. j, k, Slow channel mutant examples at the TMD interface (top view) of α and ε subunits. εL289 (j) and εV279 (k) on M2 nestle into a pocket formed by several hydrophobic residues at the α-ε TMD interface. εL289F and εV279F mutations would result in clashes with nearby residues in the observed resting-like conformation, which could thereby destabilize the resting state in favor of a longer-lived activated state, consistent with the slow-channel CMS phenotype^77,78. l, Examples of point mutants that cause receptor deficiency CMS⁵¹, shown as yellow spheres.

Extended Data Figure 9: State transition and gates in fetal AChR.

a, Comparison of fetal resting-like (light grey) and desensitized (colored) state structures reveals ECD conformational changes upon ACh binding. b, c, Conformational changes in the coupling region (b) and the TMD regions (c) triggered by ACh binding; colors as in a. d, Close-up top view of L9′ and F6′ changes in different fetal receptor states; colors as in a. e, Permeation pathway in the resting-like state pore region; red dots indicate pore diameters less than 2.2 Å. f, As in e, but for the fetal desensitized state; red dots here indicate < 4.1 Å. g, Pore profiles of fetal resting-like and desensitized structures. h, i, Sequence alignments of 6′ and 9′ at M2 helices among Cys-loop family receptors; h, among different species; i, in humans. j, k, Structural comparisons of fetal and adult receptors at different state reveal high similarity. l, m, Structures and conformations of the 6′ and 9′ residues in M2 helices in the apo (l) and open (m) heteromeric glycine receptors⁴⁸.

Extended Data Video 1: Cryo-EM map-model agreement in the adult receptor.

Download video file ^{(40.9MB, mp4)}

Maps shown are those generated through DeepEMhancer postprocessing, except for the pore region in the ACh-bound adult receptor, where the unsharpened map is displayed.

Extended Data Video 2: Adult receptor state transitions.

Download video file ^{(39.5MB, mp4)}

In both α subunits, ACh binding coincides with loop C clamping down ~9 Å toward the channel axis, forming an aromatic box around the agonist (Fig. 4a). Loop F from the complementary subunits (ε and δ) moves up and toward the closed loop C, positioned to form an H-bond between the loop C backbone and an aspartic acid conserved in all complementary subunits (Fig. 4a). These local conformational changes coincide with twisting of the α subunit ECDs, approximately as rigid bodies, such that upon ACh binding, the Cys-loop and β1β2 loops move toward the channel axis. This movement in the ECD component of the coupling region results in the α subunit M2-M3 loops swinging outward, pulling the extracellular ends of α-M2 away from the channel axis (Fig. 4b). The coupling region transitions in the ε and δ subunits are similar to each other but distinct from the α subunits in that the whole intra-subunit coupling region largely behaves as a rigid body. Here, β1β2-loop interactions with the M2-M3 loop remain intact and both pivot outward and away from the channel axis (Fig. 4b, c, Extended Data Video 2). The conformational transition of the β subunit is the subtlest; as seen in the Torpedo receptor²⁹, this subunit appears to act like a semi-rigid scaffold.

Extended Data Table 1.

Cryo-EM data collection, refinement and validation statistics

	Fetal receptor + neurotoxin (EMD-43926) (PDB 9AWK)	Adult receptor + neurotoxin (EMD-43924) (PDB 9AVV)	Fetal receptor + ACh (EMD-43923) (PDB 9AVU)	Adult receptor + ACh (EMD-43925) (PDB 9AWJ)
Data collection and processing
Magnification	130,000×		130,000×
Voltage (kV)	300		300
Electron exposure (e−/Å2)	50		50
Defocus range (μm)	−1.6 to −2.2		−1.6 to −2.0
Pixel size (Å)	0.935		0.935
Symmetry imposed	C1		C1
Initial particle images (no.)	~1.5 million		~3 million
Final particle images (no.)	168,720	209,258	259,740	167,900
Map resolution (Å)	2.14	2.09	2.45	2.45
FSC threshold	0.143	0.143	0.143	0.143
Refinement
Initial model used (PDB code)	N/A	N/A	9AWK	9AVV
Model resolution (Å)	2.42	2.31	2.63	2.67
FSC threshold	0.5	0.5	0.5	0.5
Map sharpening B factor (Å2)	−14.5	−40.1	−40.1	−29.7
Model-to-map, CCmask	0.82	0.80	0.80	0.82
Model composition
Non-hydrogen atoms	18,111	18,176	17,065	17,003
Protein residues	2,167	2,176	2,047	2,041
Ligands	45	45	42	43
Water	0	0	5	5
B factors (Å2)
Protein	61.10	39.13	80.05	75.02
Ligand	63.04	40.97	75.66	73.44
Water	-	-	49.11	48.60
R.m.s. deviations
Bond lengths (Å)	0.006	0.006	0.005	0.005
Bond angles (°)	0.882	0.947	0.817	0.805
Validation
MolProbity score	1.14	1.15	1.14	1.13
Clashscore	3.50	3.57	2.90	3.09
Poor rotamers (%)	0.30	0.60	0.32	0.43
Ramachandran plot
Favored (%)	97.99	98.09	97.73	97.87
Allowed (%)	2.01	1.91	2.27	2.13
Disallowed (%)	0.00	0.00	0.00	0.00

Data availability

All atomic models and cryo-EM maps have been deposited in the Protein Data Bank and Electron Microscopy Data Bank: fetal resting-like (PDB 9AWK, EMD-43926); adult resting-like (PDB 9AVV, EMD-43924); fetal ACh-bound (PDB 9AVU, EMD-43923); and adult ACh-bound (PDB 9AWJ, EMD-43925). All other data are available upon request.