Muscle and neuronal nicotinic acetylcholine receptors

Abstract

Nicotinic acetylcholine receptors (nAChRs) are integral membrane proteins and prototypic members of the ligand-gated ion-channel superfamily, which has precursors in the prokaryotic world. They are formed by the assembly of five transmembrane subunits, selected from a pool of 17 homologous polypeptides (α1–10, β1–4, γ, δ, and ε). There are many nAChR subtypes, each consisting of a specific combination of subunits, which mediate diverse physiological functions. They are widely expressed in the central nervous system, while, in the periphery, they mediate synaptic transmission at the neuromuscular junction and ganglia. nAChRs are also found in non-neuronal/nonmuscle cells (keratinocytes, epithelia, macrophages, etc.). Extensive research has determined the specific function of several nAChR subtypes. nAChRs are now important therapeutic targets for various diseases, including myasthenia gravis, Alzheimer's and Parkinson's diseases, and schizophrenia, as well as for the cessation of smoking. However, knowledge is still incomplete, largely because of a lack of high-resolution X-ray structures for these molecules. Nevertheless, electron microscopy studies on 2D crystals of nAChR from fish electric organs and the determination of the high-resolution X-ray structure of the acetylcholine binding protein (AChBP) from snails, a homolog of the extracellular domain of the nAChR, have been major steps forward and the data obtained have important implications for the design of subtype-specific drugs. Here, we review some of the latest advances in our understanding of nAChRs and their involvement in physiology and pathology.

Abbreviations

A-AChBP

Aplysia californica AChBP

AAN

autoimmune autonomic neuropathy

ACh

acetylcholine

AChBP

acetylcholine-binding protein

AChR

acetylcholine receptor

AD

Alzheimer's disease

ADHD

attention deficit hyperactivity disorder

B-AChBP

Bulinus truncatus AChBP

α-Bgtx

α-bungarotoxin

α-Cbtx

α-cobratoxin

CICR

calcium-induced calcium release

CMS

congenital myasthenic syndrome

CNS

central nervous system

α-Ctx

α-conotoxin

DA

dopamine

EAAN

experimental autoimmune autonomic neuropathy

EAMG

experimental autoimmune Myasthenia gravis

ECD

extracellular domain

ER

endoplasmic reticulum

ERK/MAPK

extracellular signal-regulated mitogen-activated protein kinase

GST

glutathione S-transferase

5-HT

5-hydroxytryptamine

IFN-γ

interferon-γ

IL

interleukin

L-AChBP

Lymnaea stagnalis AChBP

LBD

ligand-binding domain

LGIC

ligand-gated ion channels

M1–4

transmembrane fragments 1–4

mAb

monoclonal antibody

mAChR

muscarinic AChR

MAPK

mitogen-activated protein kinase

MG

myasthenia gravis

MIR

main immunogenic region

MuSK

muscle-specific receptor tyrosine kinase

nAChR

nicotinic acetylcholine receptor

NF-κB

nuclear factor-κB

NMDA

N-methyl-d-aspartate

NMJ

neuromuscular junction

PD

Parkinson's disease

PNS

peripheral nervous system

TNF

tumor necrosis factor

VOCCs

voltage-operated Ca²⁺ channels

VTA

ventral tegmental area.

Ion channels

The cell membrane is the main barrier for ion movement, so specific proteins, such as ion channels, have evolved to transport ions across it. Ion channels are gated pores that permit the passive flow of ions down their electrochemical gradients. Because of their important functional roles, their membrane location and structural heterogeneity and the restricted tissue expression of some channel types, ion channels are ideal drug targets. They share a common structural pattern: the central water-filled pore, through which the ions move, is usually formed by four or five transmembrane α helices (tetrameric or pentameric channels, respectively), arranged like the staves of a barrel. These different pore-forming helices can be parts of different subunits.

Access to the pore is governed by a ‘gate’, which can be opened or closed by electrical, chemical or mechanical signals, depending on the type of channel. Ligand-gated ion channels (LGICs) form a group of ion channels in which ‘gate’ opening is controlled by the binding of a chemical messenger. They share several common physiological and structural features and, because of their oligomeric structure, have a number of ligand-binding sites. The ligand-binding sites and the ‘gate’ lie at a considerable distance from each other, the ligand-binding sites quite commonly being located at the interface between two adjacent subunits, which is energetically more favorable given that domain interfaces undergo larger conformational changes upon ligand binding.

Nicotinic acetylcholine receptors

Acetylcholine receptors (AChRs) are integral membrane proteins that respond to the binding of acetylcholine (ACh), which is synthesized, stored and finally released by cholinergic neurons. Like other transmembrane receptors, AChRs have been classified according to either their pharmacological properties or their relative affinities for various molecules, and can therefore be further divided into: (a) nicotinic AChRs (nAChRs, the ‘ionotropic’ AChRs), which are particularly responsive to nicotine [1,2]. nAChRs are the prototype members of the LGICs superfamily; and (b) muscarinic AChRs (mAChRs, the ‘metabotropic’ AChRs), which are particularly responsive to muscarine [3,4]. mAChRs are members of the membrane-bound G-protein-coupled receptor superfamily.

In the N-terminal extracellular domain (ECD) of each LGIC subunit there is a conserved sequence of 13 residues flanked by covalently bonded cysteines, forming a loop located between the ligand-binding domain (LBD) and ion-channel domain. Because of this unique common characteristic feature, LGICs are also called the Cys-loop receptor superfamily. nAChRs are composed of five homologous subunits organized around a central pore and are further divided into two groups: (a) the muscle type, found in vertebrate skeletal muscles, where they mediate neuromuscular transmission at the neuromuscular junction (NMJ), as well as in fish electric organs; and (b) the neuronal type, found mainly throughout the peripheral nervous system (PNS) and central nervous system (CNS), but also in non-neuronal tissues.

cDNAs for 17 types of nAChR subunits have been cloned from several species. These consist of α subunits (α1–10), which carry the main parts of the ligand-binding sites, β (β1–4), γ, δ, and ε subunits. Based on their different ligand-binding properties, the diverse group of the nAChRs has been divided into two main classes: (a) the α-bungarotoxin (α-Bgtx)-binding nAChRs, which can be either homopentamers of α7, α8 or α9 subunits or heteropentamers (e.g. α₂βε(γ)δ); and (b) nAChRs which do not bind α-Bgtx, contain the α2–α6 and β2–β4 subunits, exist only as heteropentamers and bind agonists with high affinity [2].

Most of the early studies on nAChR subunit composition and structure were performed on molecules isolated from the electric organs of the electric fish, Torpedo and Electrophorus, a tissue extremely rich in nAChRs (many milligrams of nAChR per kilogram of Torpedo electric organ). Torpedo nAChR, which shows a high degree of homology with vertebrate skeletal muscle nAChR [5], has been studied extensively; it has two ACh-binding sites located at interfaces between subunits, a transmembrane ion channel, and all the structural elements required for activation and desensitization processes [6,7]. Using recombinant DNA techniques, close homologies between nAChR subunit sequences from fish electric organs and skeletal muscle from higher vertebrates, including humans, have been revealed [8]. Therefore, all these nAChRs are often named muscle-type nAChRs.

Early in the study of the nAChR, investigators attempted to characterize the receptor using a number of ligands, but the great advance occurred when small polypeptides from the venom of poisonous snakes were tested and found to block neuromuscular transmission in vertebrates. α-Bgtx from Bungarus multicinctus venom was historically the first toxin used to characterize nAChRs and the demonstration that it causes almost irreversible block of Torpedo[9] and Electrophorus[10] nAChRs was a great step forward, as it made possible the isolation and characterization of the molecule.

Extensive studies on nAChRs from various species have demonstrated that each nAChR subunit consists of: (a) the N-terminal ECD of each subunit which is ∼ 210–220 amino acids long and bears the LBD for agonists and competitive antagonists [11,12]; (b) four small (15–20 amino acids long) hydrophobic transmembrane segments (M1–M4) and two small hydrophilic loops, linking segments M1–M2 and M2–M3; (c) a larger loop varying in size (100–150 residues) and sequence between subunits, which lies between M3 and M4, and bears phosphorylation sites [13]; and (d) the C-terminal end of each subunit which bears a small (4–28 amino acids) hydrophilic extracellular segment.

Several post-translational modifications contribute to the structure and function of the molecule. An unusual disulfide bond, common to all α subunits, is formed between two adjacent Cys residues that correspond to amino acids 192 and 193 of the Torpedoα1 subunit. This particular bond contributes to the agonist-binding site. Glycosylation sites, the number and location of which differ between subunits, are located in the ECD. As mentioned above, phosphorylation sites have also been identified in the intracellular domain between the M3 and M4 transmembrane segments.

A breakthrough in our understanding of the structure and function of nAChRs came from solving the crystal structure of the acetylcholine-binding protein (AChBP) [14,15]. This protein is a naturally occurring homolog of the ECD of nAChR which can be produced in large quantities and its water solubility facilitates crystallization and X-ray analysis. Subsequent solution of the X-ray structures of the AChBP complexes with agonists and antagonists [16–18] provided excellent models for the LBDs of LGICs and their complexes with putative drugs through homology modeling. Recently, prokaryotic homologs to LGICs have been shown to exist [19] which can also be obtained in large amounts. The cloning and expression of the Gloeobacter violaceus channel was achieved and homology modeling in parallel with electrophysiological experiments revealed key residues for its function [20].

nAChR 3D structure

Structural information on nAChRs has been derived from electron microscopy studies on 2D arrays of the Torpedo nAChR [7,21,22]. These earlier studies revealed the dimensions and shape of the molecule, defined the arrangement of the subunits and the boundaries between them and provided an insight into the location of the binding sites and the organization of the ion channel. Subsequent 4.6 and 4 Å resolution electron microscopy studies provided an insight into the structure of the extracellular and transmembrane subunit domains [23,24]. Extracellularly, the ACh-binding pockets were shown to be surrounded by seven-stranded β sheets forming a larger β-sandwich structure. These data confirmed CD studies [25–30], which indicated that the LBD consists predominantly of β sheets. It was clearly shown that the four membrane-spanning segments of each subunit (M1–M4) are α helices [24].

Detailed information on the atomic structure of an ACh-binding domain first became available following elucidation of the crystal structure at 2.7 Å resolution of the AChBP from the glial cells of the mollusc, Lymnaea stagnalis[14] (Fig. 1). This protein, a soluble homopentamer of a 210 amino acid subunit, is a structural and functional homolog of the ECD of Cys loop receptor subunits, showing up to 24% sequence identity with human nAChR ECDs and 15–18% with ECDs of other LGICs, and is used as a model for the LBD of nAChRs. The resolved dimensions of the AChBP (a cylinder, 62 Å high, with a diameter of 80 Å and a central hole 18 Å in diameter) are in good agreement with those estimated by electron microscopy for the Torpedo nAChR [23]. When viewed along the fivefold axis, the AChBP homopentamer resembles a toy windmill with blade-like protomers (Fig. 1A). Each protomer consists of an α helix close to the N-terminal, two short 3₁₀ helices and 10 β strands (β1–β10), connected through equal number of loops (Fig. 1B). The β strands are arranged in two sets joined through a cysteine disulfide bridge, the Cys loop, forming a β-sandwich hydrophobic core.

Figure 1

Crystal structure of Lymnaea stagnalis AChBP, which is homologous to nAChR ECD at 2.7 Å resolution. (A) Top view. Each subunit of the homopentamer is shown in a different color and symbolized with a different letter (A–E). The five ligand-binding sites between the subunits are shown in ball-and-stick representation. (B) Side view of AChBP protomer from outside the pentameric ring. The side of the protomer, bearing the conserved Cys loop is called the principal (plus) side. Also shown, is the AChBP region corresponding to the MIR epitope of the α subunit of the muscle nAChR. Reproduced from Brejc et al.[14], with permission.

To date, no 3D X-ray structure of any nAChR or any other LGIC is available. Although very small, nondiffracting 3D crystals of Torpedo nAChR were generated as early as in 1988 [31,32], several attempts to obtain high-quality crystals of this large membrane protein or even of its smaller (mainly extracellular) fractions have met with little success.

Based on the crystal structure of AChBP, the atomic model of Torpedo muscle-type nAChR at 4 Å resolution electron microscopy studies (Fig. 2), allowed a detailed description of the whole receptor in its closed-channel form at a chemical level [33]. More recently, crystal structures of various molluscan AChBPs with known cholinergic ligands and toxins have provided detailed information on the conformational changes induced upon ligand binding.

Figure 2

Electron microscopy structure of Torpedo nAChR at 4 Å resolution. (A) Ribbon diagrams of the whole nAChR from Torpedo electric organ, as viewed from the synaptic cleft. Also shown are the locations of αTrp149 (gold) and the MIR epitope. (B) Side view of the α subunit. The location of the functionally important A, B, C, β1–β2 and Cys loops is shown. Also, the MIR epitope and the membrane (E, extracellular; I, intracellular) are presented. Reproduced from Unwin [33], with permission.

nAChR overview

In the 4 Å model, the Torpedo nAChR has a total length of ∼ 160 Å normal to the membrane plane and is divided into three domains: an N-terminal extracellular LBD, a membrane-spanning pore and an intracellular domain (Fig. 2B). The receptor subunits in the LBD are each organized around a curled β-sandwich hydrophobic core, consisting of 10 β strands which are joined through the Cys loop and contain one N-terminal α helix, like the protomers of the closely related AChBP [14]. The hydrophobic core of this receptor domain consists of conserved residues, equivalent to those of the AChBP, with the exception of the αLeu6 near the end of the α helix. This domain also contains several loop regions, i.e. loops A, B, C, Cys loop and β1–β2 loop, which are critical for receptor function, identified as components of the LBD (Fig. 2B).

Early biochemical studies involving site-directed mutagenesis and affinity labeling indicated that two separate parts of the nAChR-LBD are involved in the formation of the agonist/competitive antagonist-binding site [34–36]. One is called the ‘principal component’ of the binding site and resides on the α subunit, whereas the other is called the ‘complementary component’ and resides on the adjacent non-α subunit. The ACh-binding pocket of the nAChR is formed between loops A, B and C on the α subunit and strands β5 and β6 of the β-sandwich core of the adjacent γ or δ subunit and lies ∼ 40 Å above the membrane surface and on opposite sides of the channel pore [33]. The key residues of the loops implicated in the formation of the ACh binding site are Tyr190, Cys192 and Tyr198 of the C loop, which is incorporated in the β9–β10 hairpin and Trp149 of the B loop (Fig. 3A).

Figure 3

(A) ACh-binding region of the closed channel of Torpedo nAChR at the interface between α and γ subunits. Loops B and C of the α subunit and the adjacent β5 and β6 strands of γ subunit are shown. (B) (a) simplified C^α traces of the ligand-binding region of the α subunit in the closed channel; (b) equivalent region of L-AChBP complexed with carbamylcholine [16]; (c) superimposition of the two regions, revealing the closure of B and C loops around the bound agonist. Reproduced from Unwin [33], with permission.

α1-ECDs also contain the main immunogenic region (MIR) [37–40] (Fig. 2), a region of overlapping epitopes (including amino acids 67–76 of the α1 subunit), against which a large fraction of autoantibodies against nAChR is directed in the autoimmune disease myasthenia gravis (MG). The critical segment of the MIR that serves as the epitope of autoantibodies is localized to residues αTrp67–αAsp71, with αAsn68 and αAsp71 being the most important. All five residues appear to be exposed to the solvent, because of the wide separation between the loop they form (part of the β2–β3 loop) and the N-terminal α helix of the LBD. The equivalent loops of the other subunits (non-α) seem to have the same fold, but these are closer to the respective helices and this results in the partial burial of the residues aligning with αTrp67–αAsp71. There are considerable differences in the amino acid sequence of the 67–76 segment between α1 and the neuronal α subunits. Smaller differences occur between α1 and α3/α5 and this is in agreement with observations that some anti-MIR mAbs bind, in addition to α1, to α3 and α5 but not to other neuronal α subunits [41,42] (S. J. Tzartos et al., unpublished data). Although the corresponding region in the AChBP (residues 65–72) has no sequence or functional similarity to the MIR on the α1 subunit, the crystal structure of the AChBP shows that this is located in a highly accessible area at the ‘top’ of the pentamer (Fig. 1B) [14], as would normally be expected for a very antigenic region.

In the membrane-spanning domain, the α helices of each subunit (M1–M4) are arranged symmetrically, forming the channel pore [24]. The five M2 helical segments of all subunits line the pore, forming an inner ring, whereas the other 15 helical segments in M1, M3 and M4 from all five subunits coil around each other, forming an outer ring, which shields the inner ring of M2 domains from the membrane lipids. In the closed channel, the M2 helices come together near the middle of the membrane, forming a hydrophobic girdle, which is considered to be the gate of the channel, functioning as an energetic barrier to ion permeation [33]. Apart from the α-helical regions M1–M4, the small loops M1–M2 (intracellular) and M2–M3 (extracellular) are also considered to be part of the membrane-spanning region (Fig. 2B). These loops are functionally important, as the transmembrane domain interacts through the M2–M3 loop with the β1–β2 loop and the Cys loop of the ECD. Moreover, the membrane-spanning domain is joined covalently to the LBD through an extension of the β10 strand into M1 helix.

The intracellular region consists of the M3–M4 loop, which contains a curved α helix (MA) immediately before the M4 transmembrane region (Fig. 2B). Most of the loop immediately after the M3 transmembrane region (the M3–MA loop) seems to be rather disordered and it is not resolved in the electron microscopy structure. Each subunit contributes one MA α helix, to shape the wall of the vestibule. These five helices form an inverted pentagonal cone which has five intervening spaces (‘windows’) of a width < 8 Å (thus comparable with the diameter of a hydrated sodium or a potassium ion), surrounded by negatively charged side chains. These windows represent obligatory ion pathways, because no alternative routes exist for transport inwards or outwards the intracellular vestibule. Therefore, they constitute a charge and size ‘selectivity filter’, which facilitates cation transport, but prevents anions and large ions from going through. Both the extracellular and intracellular vestibules of the channel are lined mainly by negatively charged side-chain groups, which form an electrostatic environment that stabilizes and thus favors the influx of cations (Na⁺, K⁺ and Ca²⁺) [33].

In the LBD, the Torpedo subunits were shown to interact mainly through polar side chains [43], similar to the interactions seen in the AChBP. Charged side chains on both α1 subunits form ion pairs with side chains on neighboring subunits. The subunit–subunit interactions of the membrane-spanning domain are mainly attributed to hydrophobic side chains projecting from the helices M1, M2 and M3, implicating relatively few residues on M1 and M3. Intracellularly, the subunit–subunit interactions involve contacts between the M1–M2 loops as well as between the M3–M4 and MA regions of neighboring subunits. Because of the incomplete resolution of the electron microscopy structure in this region, no detail of these interactions is available.

Atomic structure of the ligand-binding site

In the crystal structure of AChBP [14] each ligand-binding site is found in a cavity at each interface between the five subunits, lined by aromatic and hydrophobic residues, previously shown to be involved in ligand-binding in nAChRs by mutational analyses and site-directed labeling [11,36,44–46]. The ligand-binding site is formed by the contribution of highly conserved, through the LGIC family, residues from loops A (Tyr89), B (Trp143, Trp145) and C (Tyr185, the double cysteine 187–188 and Tyr192) of the principal (plus) side of one subunit and by less conserved residues from loops D (Trp53, Gln55), E (Arg104, Val116, Leu112 and Met114) and F (Tyr164) of the complementary (minus) side of the adjacent subunit (Fig. 4A). When the protomer is viewed by the side and perpendicular to the fivefold axis, the plus side of each protomer is the one that bears the Cys loop (Fig. 1B).

Figure 4

Representation of the ligand-binding site and of the interactions upon agonist binding in (A) Lymnaea stagnalis AChBP (Hepes bound) [14], (B) Lymnaea stagnalis AChBP (nicotine bound) [16], (C) Lymnaea stagnalis AChBP (carbamylcholine bound) [16], (D) Aplysia californica AChBP (lobeline bound) [55], and (E) Aplysia californica AChBP (epibatidine bound) [55]. Reproduced from Brejc et al.[14], Celie et al.[16] and Hansen et al.[55], with permission.

The structure of AChBP also revealed the interactions between residues of each protomer in the dimer interface. The subunit interface consists entirely of loop regions on the plus side and mainly of secondary structure elements (α1, β1–3, β5–6 and L9) on the minus side. Its buried nature leads to a preference of hydrophobic residues, including only a bifurcated salt bridge formed between Glu149 on the plus side of one subunit and Arg3 and Arg104 on the minus side of the adjacent subunit. In contrast to the well-conserved residues stabilizing the protomer structure by the formation of hydrophobic cores, the interface residues important for the pentameric formation and some for ligand binding are not well conserved between the members of the LGICs, thus creating specificity for different ligands. The only highly conserved interface residues are those forming the plus side of the ligand-binding site.

Differences in the subunit interfaces of AChBPs from different species, lead to distinct ligand-binding properties. This is more stressed by the comparison of the subsequent crystal structure of Bulinus truncatus AChBP (B-AChBP) [47] with that of Lymnaea stagnalis AChBP (L-AChBP). Although their structures were found to be very similar, the residues of the subunit interface have not been conserved in B-AChBP except for the ligand-binding residues. These differences in subunit interface contacts between B-AChBP and L-AChBP have a major effect on the stability and ligand-binding properties of these proteins. L-AChBP is more stable than B-AChBP as shown by thermal stability studies using CD [47], probably because L-AChBP has on average one extra hydrogen bond and two additional salt bridges in each interface. Apparently, the mode of pentamer formation is poorly conserved. Analyzing the AChBP structures for interface regions that could improve the formation of stable pentamers in nicotinic structures can prove beneficial, because expression of properly assembled nAChR LBDs has proven remarkably difficult, and in many cases a mixture of monomers, dimers, and multimers was obtained [26,48–50]. In addition, B-AChBP has a 5–10-fold higher affinity for the binding of cholinergic ligands [nicotine, acetylcholine, carbamylcholine, and (+)-tubocurarine], compared with L-AChBP, but interestingly does not bind significantly to α-Bgtx [47]. The differences in ligand-binding were attributed to three nonconserved residues on the complementary side of the ligand-binding of the two AChBPs (L-AChBP Arg104, Leu112 and Met114 changed to B-AChBP Val103, Ile111 and Val113, respectively) and probably to alternative residues which may be part of the subunit–subunit interfaces. Mutation of these three L-AChBP residues to the B-AChBP corresponding ones led to a triple mutant L-AChBP, which showed a clear gain-of-affinity for binding both nicotine and (+)-tubocurarine, equivalent to that of the B-AChBP [47]. However, the remaining differences in ACh binding and in entropic and enthalpic contribution showed that regions outside the binding site (and probably in the interface) also contribute to the affinity [47]. Interestingly, B-AChBP contains a lysine residue at position 183 (adjacent to the conserved Tyr184 of loop C), which seems to be responsible for abolishing α-Bgtx binding, similar to the human α2, α3 and α4 subunits that do not bind the toxin. NMR and mutagenesis studies have shown that nAChRs that bind to α-Bgtx have a Tyr or Phe residue at this position [51–53] and replacement of this Tyr by Lys completely prevents α-Bgtx binding [53,54].

Although the initial crystal structure of L-AChBP was obtained in the absence of any cholinergic ligand, the Hepes molecule used in the crystallization buffer, mimics known cholinergic ligands, as it contains a positively charged quaternary ammonium group. Indeed, this molecule was detected well-placed in each ligand-binding site of L-AChBP, stacked on the highly conserved Trp143 by cation-π interations, as expected for nicotinic agonists. This revealed for the first time the critical role of Trp143 in ligand binding.

More recently, cocrystallization of various molluscan AChBPs with several agonists and antagonists [16–18,55,56] revealed details of the atomic interactions between ligands and specific residues in the ligand-binding site. The precise identification of the molecular determinants in these crystal complexes and the understanding of the interactions with ligands are expected to contribute to the development of highly selective (subtype-specific) nAChR modulators and to the design of specific drugs for several nAChR-related disorders, such as Alzheimer's disease (AD) and Parkinson's disease (PD), as well as nicotine addiction. In addition to the Ach-binding site, other binding sites on nAChRs have also been identified. There is a binding site for positive allosteric modulators around and including Lys125 of the N-terminal extracellular region of the α1 subunit [57], a steroid-binding site at an allosteric site distinct from both the ACh-binding site and the ion channel [58] and two binding sites for noncompetitive blockers or negative allosteric modulators [59]; the first site is located on the M2 transmembrane segments of the nAChR within the pore and the second site appears to lie at the interface between the nAChR protein and membrane lipids. Interestingly, a recently performed molecular docking of three known allosteric modulators of nAChRs (galanthamine, codeine and eserine) with AChBP and with models of human α7, α3β4 and α4β2, based on the X-ray structure of AChBP, identified three distinct binding sites, located in the channel pore [60]. The first site seems to be located between the L1 and L4 loops (α/+ subunit) and the β3 and β5 strands (β/– subunit), situated exactly opposite the agonist-binding site, at a distance < 12 Å. The second site appeared to be situated between loops L4 (α/+ subunit) and L5 (β/– subunit) and finally the third site was shown to be close to L5 loop and β7 and β2 strands (β/– subunit). Considering the position of the third site at the bottom of the LBD, this might be implicated in the modulation of the motions for opening or closing the gate of the channel.

Agonist binding

Crystal structures of L-AChBP with bound nicotine or carbamylcholine [16] and more recently of Aplysia californica AChBP (A-AChBP) with bound lobeline or epibatidine [55] revealed agonist binding to a pentameric LGIC. The architecture of the ligand-binding pocket, with the aromatic nest made of Tyr93, Trp147, Tyr188 and Tyr195 from the plus face and Tyr55 from the minus face of the interface (A-AChBP numbering), is highly conserved in the AChBPs (Fig. 4B–E). In all agonist-bound AChBP structures, the agonist is fully enveloped by the protein through hydrogen bonds, π-cation, dipole-cation and van der Waals forces. A highly conserved Trp143 (L-AChBP numbering) from the principal face of the subunit at the center of each complex makes strong aromatic π-cation interactions with the agonists (Fig. 4B). The highly conserved Asp85 (L-AChBP numbering) also has a major structural role. It is situated behind the central Trp143 and polarizes its main-chain carbonyl group, further stabilizing the ligand positively charged group. The vicinal disulfide is in contact with all four agonists, mostly through Cys187 with carbamylcholine and through Cys188 with nicotine (L-AChBP numbering). In all cases, ligand-binding is further supported by interactions with residues of the less-conserved minus face, creating ligand-binding specificity. Both lobeline and epibatidine exhibit additional interactions with A-AChBP Trp147 (L-AChBP Trp143). The central piperidine ring of lobeline exhibits a stacking interaction with A-AChBP Trp147, favoring a hydrogen bond between the tertiary amine and the Trp carbonyl oxygen, whereas the pyridine amine of epibatidine binds to A-AChBP Trp147 and Ile118 (Leu112 in L-AChBP) through a solvent molecule. Tyr185 in L-AChBP (Tyr188 in A-AChBP) makes aromatic contacts with carbamylcholine and epibatidine (Fig. 4B), but not with nicotine, and forms van der Waals interactions with the lobeline methyl group. Tyr192 in L-AChBP (Tyr195 in A-AChBP) forms aromatic contacts with all agonists and L-AChBP Tyr89 (A-AChBP Tyr93) contacts the ligands with its methyl group. This Tyr93 is further involved in the formation of a second hydrogen bond between the bridge ring of epibatidine and A-AChBP (Fig. 4E). Leu112 and Met114 in L-AChBP (Met116 and Ile118, in A-AChBP, respectively) make hydrophobic contacts with nicotine, carbamylcholine and lobeline, whereas Ile118 is involved in the first hydrogen bond between epibatidine and A-AChBP. The L-AChBP Arg104 (A-AChBP Val118) makes hydrophobic contacts with carbamylcholine and polar contacts with epibatidine. Interestingly, L-AChBP Leu102 and Met114 residues contribute significantly to nicotine binding, as they form a hydrogen bond with pyridine N1 through a water molecule. Nicotine binding is further enhanced by a second hydrogen bond between the pyrollidine N2 and the carbonyl group of L-AChBP Trp143.

Antagonist binding

Binding sites for competitive antagonists, such as snake neurotoxins and cone snail conotoxins are also located at α subunit interfaces [35,61]. Using synthetic peptides, α-neurotoxin-binding sites have been identified on both neuronal and muscle α subunits [62–66]. Specifically, we found that the main α-Bgtx-binding segment corresponds to residues 189–195 in the Torpedo[63] and to 186–197 in the human α7 subunit [66]. The α-Bgtx-binding segment in human α1 subunit corresponds to residues 185–196 [62].

Photoaffinity-labeling, site-directed mutagenesis and competition studies have shown that the α-Bgtx-binding sites on the neuronal α7 and muscle-type nAChRs have a similar conformation and interact with identical or overlapping sites on the α-Bgtx molecule. An aromatic amino acid in loop C (Phe187 in human α7) is necessary for α-Bgtx binding, judging by the reduced α-Bgtx binding affinity of human α1 and the absence of α-Bgtx binding to neuronal α3 and α4 receptors, which have nonaromatic residues at the homologous position [63,66–68]. Negatively charged amino acids in loop C are also important in maintaining an electrostatic interaction with positively charged amino acids in α-Bgtx [63,66,69]. Although the same amino acid sequences contribute to the agonist- and competitive antagonist-binding sites, difference in antagonist-binding affinity can be attributed to species-dependent amino acids. Glycosylation also affects α-Bgtx binding to some, but not all, nAChRs [26,50,70–74]. Deglycosylation of the ECD of the human α1 subunit (expressed in Pichia pastoris) abolishes α-Bgtx binding [75], but has no effect on recombinant ECDs of Torpedoα1 and human α7 expressed in Escherichia coli and P. pastoris[26,50,70], reflecting a differential role of glycosylation in protein folding.

The crystal structures of L-AChBP with α-cobratoxin (α-Cbtx) [56], and of A-AChBP with α-conotoxins (α-Ctxs) [17,18,55] and the alkaloid antagonist methyllycaconitine (MLA) [55], revealed the specific residues that are involved in toxin/antagonist-binding and the conformational changes induced in nAChR-LBD upon antagonist binding.

α-Cbtx is a long α-neurotoxin, with a characteristic three-fingered fold; four disulfide bonds form a β-structure hydrophobic core from which three loops (I, II and III) emerge. The α-helical loop II bears the fifth disulfide bond, essential for recognition of the respective neuronal nAChRs [76]. In the crystal structure of the complex L-AChBP–α-Cbtx, Cbtx loop II was inserted deeply into each of the five ligand-binding pockets with an angle ∼ 45° relative to the median axis of the AChBP ring [56], through essentially hydrophobic and aromatic contacts. Phe29 and Arg33 on the tip of Cbtx loop II interact with the highly conserved Trp53, Tyr185, Tyr192 and Trp143 of the principal side of the AChBP ligand-binding pocket (L-AChBP numbering) (Fig. 5A). Additional contacts include Cbtx residues surrounding the tip of loop II with Glu163, Glu55, Leu112, Met114 and Tyr164 on the complementary side. In nAChR models, amino acids involved in α-neurotoxin binding have been mapped to loop C in the principal component and to loop F in the complementary component [77]. Binding of α-Cbtx to AChBP induces a significant displacement of loops C and F by ∼ 10 Å, uncapping the pocket and stabilizing C loop in an ‘open’ conformation similar to that of the ligand-free AChBP structures [55].

Figure 5

Representation of the ligand-binding site and of the interactions upon antagonist binding in (A) Lymnaea stagnalis AChBP (α-cobratoxin bound) [56], (B) Aplysia californica AChBP (α-conotoxin PnIA (Ala10Leu D14Lys) bound) [17], (C) Aplysia californica AChBP (α-conotoxin ImI-bound) [55], and (D) Aplysia californica AChBP (MLA-bound) [55]. Reproduced from Celie et al. [17], Hansen et al. [55] and Bourne et al. [56], with permission.

α-Ctxs are short-peptide toxins in which a helical region, braced by two conserved disulfide bonds, defines a characteristic two-loop framework [78]. Because of their rigid structure and amino acid diversity, they bind distinct nAChR subtypes with unique selectivity. α-Ctx PnIA, a competitive inhibitor of neuronal nAChRs, shows higher affinity for α3β2 nAChRs than for the α7 subtype [61,78]. However, Ala10Leu substitution shifts its selectivity from the α3β2 to the α7 subtype [79]. The PnIA double mutant (Ala10Leu/Asp14Lys) [17] further increases the efficacy to chicken α7 nAChR compared with the PnIA mutant (Ala10Leu) [80]. α-Ctx-ImI, a specific and high-affinity inhibitor of the α7 nAChR [81], was recently reported to display even higher affinity for the human α3β2 nAChR [82]. In the AChBP–Ctx complexes, both PnIA double mutant and ImI bind deeply into each AChBP ligand-binding pocket, burying 60 and 75% of their solvent-accessible surface area, respectively. Upon binding of these α-Ctxs, loop C of A-AChBP is stabilized in an ‘open’ conformation similar to the A-AChBP–Cbtx complex [56]. In contrast, loop F does not undergo significant movement, probably because α-Ctxs are much smaller than α-Cbtx. The disulfide bond Cys2–Cys8 (loop I) of both α-Ctxs interacts with the highly conserved disulfide bond (in loop C) of AChBP (Fig. 5B,C). However, PnIA double mutant and ImI use different interactions to create a high-affinity and/or selective binding. The PnIA double mutant makes mainly hydrophobic contacts in the ACh pocket using Pro7, Ala9 and Leu10 as anchoring points, whereas ImI uses the tripeptide Asp5–Pro6–Arg7 in loop I as a binding determinant forming a salt bridge and more hydrogen bonds. The Leu10 side chain of PnIA, responsible for high affinity for α7 nAChR is positioned in a hydrophobic pocket formed by Val146 (principal side), Val106, Met114 and Ile116 (complementary side), whereas Lys14 of this double mutant is exposed to the solvent and does not affect binding to A-AChBP. The selective binding of ImI to nAChR subtypes is reflected in the specific inhibition of A-AChBP compared with L-AChBP [17,83]. The lack of affinity of ImI toward L-AChBP could, at least in part, be explained by a substitution of Asp75 (A-AChBP), which forms a unique interaction with ImI Trp10, by Gln in L-AChBP. In addition, Ile116 in A-AChBP is similar to the corresponding Leu119 of the α3β2 nAChR, recently identified as a key residue for α-Ctxs binding [84]. In L-AChBP, this residue corresponds to the larger Met114 which may interfere with Ctx binding. Finally, the substitution of Met114 (A-AChBP) by Leu112 in L-AChBP could further contribute to loss of affinity of ImI to L-AChBP.

The stabilization of the loop C conformation of A-AChBP in an ‘open’ state upon binding of α-Cbtx and α-Ctxs, was also observed in the MLA–AChBP complex (Fig. 5D) [55]. It therefore appears to be essential for antagonist activity.

Conformational changes induced by ligand binding

The structures of the agonist-bound AChBPs when compared with the Hepes-bound or even better with the apo-form (in the absence of any ligand or ligand-like molecule) A-AChBP crystal structure [55] revealed conformational changes induced by agonist binding which may trigger channel opening in intact nAChRs. Analysis of the main-chain differences indicated that loop C makes significant backbone movements, capping the entrance of the binding cavity and trapping the bound agonists. In the uncapped conformation of loop C, the conserved Tyr185 (equivalent to αTyr190 in the receptor) is 8 Å from the conserved Lys139 (equivalent to αLys145) in β7 strand, which forms a salt bridge with the conserved Asp194 (equivalent to αAsp200) in β10 strand. However, in the agonist-bound crystal structures of AChBP, C loop tilts inwards, with Tyr185 within 2–3 Å of Lys139 and Asp194 displaced from the Tyr/Lys pair (L-AChBP numbering). In the receptor, mutation of any of these three residues markedly impairs channel gating and all three are interdependent in contributing to gating [85]. So, when agonist binds, capping of loop C draws a conserved Tyr into register with a conserved Lys, forming a hydrogen bond which stabilizes the loop C conformation in its binding conformation.

Apparently, all AChBP crystal structures can be classified into two categories with respect to the loop C conformation: ‘open’ or ‘closed’. The first category includes the structures of the apo-form of A-AChBP as well as all the antagonist-bound AChBPs (α-Cbtx-, -α-Ctx–PnIA double mutant, α-Ctx–ImI- and MLA–A-AChBP), representing the resting state of nAChRs. The second category includes the AChBP structures with bound agonists (nicotine–L-AChBP, carbamylcholine–L-AChBP, lobeline–A-AChBP and epibatidine–A-AChBP) or buffer molecules which mimic a cationic agonist (Hepes–L-AChBP, Hepes–A-AChBP), representing the activated state of the nAChRs. Upon agonist binding, loop C clearly makes rigid-body motions and swings as much as 11 Å between the two extreme positions observed in the ImI and epibatidine complexes. Superposition of all structures also indicates rigid-body movements of the β1–β2 loop and the Cys loop, upon agonist binding.

Channel opening

Electron microscopy experiments on helical tubes from Torpedo postsynaptic membranes (generated using a specific technique to mimic the release of ACh and thus allowing the study of the channel in the open state) [7] first provided an insight into the gating mechanism of the nAChR channel. Ligand-binding triggers a tertiary conformation change in the protein, resulting in opening of the ion channel. Later studies of the whole conformational change [86], the transmembrane structure [24] and the closed channel [33] of the Torpedo nAChR, explained in more detail the mechanism of channel opening. The subunits of the closed-channel form have two alternative conformations: one characteristic of α subunits and the other characteristic of the three non-α subunits. In both α subunits, the inner β sheets of the LBD are rotated ∼ 10° relative to the non-α subunits about a normal to the membrane plane axis. The orientation of the outer β sheets in the α subunits is different compared with that in the non-α subunits. This special conformation of the α subunits is called ‘distorted’, because these subunits convert to a similar conformation to that of the non-α- and of the agonist-bound AChBP. This distorted conformation is stabilized by several interactions between the interfaces of both sides of α subunits and the adjacent subunits, implicating residues on loop B and on the inner β sheets.

By comparing the α-ECDs of the receptor in its closed-channel form with the structure of the agonist-bound AChBP [16,55], insight can be gained into the rearrangements that should take place during ACh binding, allowing the distorted α subunits to relax. In the closed-channel form of the receptor, loop C projects away from the body of the α subunit, in contrast with the ligand-bound AChBP, where loop C is closer to the A and B loop residues implicated in ACh binding (e.g. the C^α distance between Cys192 of loop C and Trp149 of loop B is 6 Å longer in the α subunits of the closed-channel form of the receptor than in ligand-bound AChBP). B loops of the α subunits come close to the β5 and β6 strands of the adjacent subunit (γ or δ) and are possibly stabilized in this tense arrangement by interactions across the subunit interface (e.g. salt bridge between αAsp152 and γArg78 or δArg81). All these imply that these loops must undergo quite large rearrangements to allow coordination of the binding residues to ACh. Both B and C loops would surround the ACh molecule so as to enable coordination of the relevant side chains. Loop B would rotate clockwise and loop C would both twist and rotate anticlockwise (Fig. 3B). Loop B plays an interesting role in this local rearrangement as it joins the outer and inner β sheets and therefore must participate directly in driving their relative displacements, leading to the opening of the channel.

In principle, several components of the nAChR-LBD could be involved in transmitting the conformational changes upon ACh-binding, to the membrane-spanning domain, where the channel gate is located. As shown in the Torpedo nAChR structure only the Cys loop, the β1–β2 loop and the covalently connected β10 strand of the LBD make direct contacts with the membrane-spanning domain (Fig. 2B). Both loops interact with amino acid residues of the M2–M3 segments; Cys loop near the N-terminus of M3 and β1–β2 loop near the C-terminus of M2 (Fig. 6). In the α subunits these loops are 2–3 Å closer to the end of M2, along the M2–M3 linker than in the non-α subunits. The main amino acid residues of the Cys loop that interact with residues Ile, Tyr and Phe (aligning with αIle274, αTyr277 and αPhe280) of M2–M3 linker are the consecutive Phe-Pro-Phe (common to all five subunits). In the β1–β2 loop, two residues seem to be important for interaction with the M2–M3; the residue aligning with αVal46 (Val44 in AChBP) and the adjacent highly conserved Gln (aligns with αGlu45). The side chains of these residues together make an arc embracing the M2–M3 backbone. In the non-α subunits, the side chains equivalent to Val46 seem slightly displaced from the ends of M2 helices and do not make equivalent contacts.

Figure 6

Representation of interactions between the Cys loop and β1–β2 loop of the LBD with M2–M3 linker of the transmembrane domain of Torpedo nAChR. The Cys loop, the β1–β2 loop, the extension of M1 into β10-strand, and components of the membrane-spanning domain are in red, blue, green and gray, respectively. Reproduced from Unwin [33], with permission.

In conclusion, the opening of the channel gate upon binding of ACh (or other agonist) to the LBD can be explained in two ways. One possibility is that, following the relaxing rearrangements of the structural elements of the α subunits, both displaced β1–β2 and Cys loops of the α subunits rotate back toward their relaxed non-α locations and therefore stop interacting with the relevant residues of the M2–M3 segment. Therefore, M2–M3 is no longer ‘locked’ to one place and is free to move due to its flexibility conferred by the conserved αGly275 at the end of the M3, allowing gating motions to occur (Fig. 6). Another possibility is that during the displacements of β1–β2 and Cys loops following ACh binding, the particular interaction involving αVal46 of the β1–β2 loop is maintained and thus loop movement drags the end of M2 away from the axis of the channel, destabilizing the weak hydrophobic interactions holding the M2 helices of the gate together, so that they break apart and the channel opens.

Interestingly, the recently characterized pentameric Gloeobacter violaceus (Glvi) protein, considered as the ancestor of eukaryotic pentameric LGICs, shares 20% sequence identity with human α7 nAChR and molecular modeling studies have predicted a very similar conformation for Glvi compared with AChBPs and Torpedo nAChR [20]. A striking feature of Glvi protein is that in addition to the absence of the N-terminal α helix and the large cytoplasmic loop, it also lacks the disulfide bridge forming the conserved Cys loop. Therefore, it is very interesting to understand the mechanism of channel opening in this protein, which seems to function as a proton-gated channel, and trace any contribution of the equivalent to the nAChR Cys-loop domain of this protein. Also very interesting would be to understand how this protein compensates for channel selectivity, due to the lack of the cytoplasmic loop (lack of selectivity ‘windows’), compared with Torpedo nAChR. Of course, all these issues remain to be addressed by the elucidation of the structure of the Glvi protein at high resolution, something that would be of great value, as this would reveal the core structure of a complete LGIC for the first time.

Muscle-type nAChRs

General features of muscle-type nAChRs: fetal and adult types

In fetal muscle prior to innervation or after denervation and in fish electric organs, the nAChR subunit stoichiometry is (α1)₂β1γδ[87], whereas, in adult muscle, the γ subunit is replaced by ε to give the (α1)₂β1εδ stoichiometry [88,89]. The γ/ε and δ subunits are involved, together with the α1 subunits, in shaping the ligand-binding sites and maintaining cooperative interactions between the α1 subunits [90]. The presence of different non-α subunits confers different affinities to the two binding sites [91]. The site in α_γ (the α1 subunit next to the γ subunit) is biochemically distinguishable from that in α_δ due to the fact that α_γ has a higher affinity for the competitive antagonist, (+)-tubocurarine [23,92]. Binding of ACh to the α_γ and α_δ sites induces conformational changes, predominantly in the α1 subunits, which are communicated to the transmembrane region, causing channel opening [86]. The β1 subunit is important for nAChR clustering, as shown by studies on hybrid muscle nAChRs, in which the β1 subunit was replaced by its neuronal counterpart [93]. Phosphorylation of unassembled γ subunits and glycosylation of the δ subunit are required for increased nAChR assembly efficiency [94,95].

The transition from the γ-type to the ε-type nAChR occurs synchronously at all endplates within a fast muscle, suggesting that neural activity causes suppression of the γ subunit gene through transcriptional activation of the ε subunit gene [96,97]. The fetal-type nAChR continues to be expressed in the thymus and in some extraocular muscles of the adult [96]. Transcriptional regulatory elements in the γ subunit direct muscle-specific, developmentally regulated, activity-dependent synapse-specific transcription in vivo[98].

Humans, but no other species, express equal amounts of mRNA for the normal α1 subunit and for the α1* subunit, which contains 25 additional amino acids between amino acids 58 and 59 [99], generated by alternative splicing of the primary RNA transcript. The α1* subunits do not undergo conformational maturation and do not form functional channels either alone or with other subunits [100]. The significance of this isoform is unknown.

Several mutations in genes encoding muscle-type nAChR subunits induce premature termination of translation or affect residues essential for nAChR assembly. Such mutations in the ε subunit are usually compensated by expression of the fetal γ subunit. Re-expression of the fetal-type nAChR occurs in many chronic neurogenic and some inflammatory myogenic disorders and is surprisingly restricted to type I muscle fibers (slow twitch or red) [101]. Type I fibers contain more mitochondria and myoglobin and rely on aerobic oxidation for energy [102]. Expression of fetal-type nAChR is therefore crucial for maintaining neuromuscular transmission in deficiencies caused by nAChR ε subunit mutations and is associated with upregulation of γ subunit mRNA [101]. No significant upregulation of α1 subunit mRNA has been seen in patients with neurogenic disorders, in contrast to animal studies [101,103,104]. This implies that the α1 and γ subunits are independently regulated. Alternatively, the α1 and other subunits might be present in excess [101].

In addition to their critical physiological role, muscle-type nAChRs are an important autoantigen involved in MG. MG is caused by failure of neuromuscular transmission as a result of the binding of autoantibodies to muscle nAChR, which causes loss of functional nAChRs, leading to defective signaling at the NMJ [105]. (see section: Acetylcholine receptor-associated diseases).

nAChR gene expression and regulation at the NMJ

During myogenesis, nAChR subunits are initially expressed at low levels in myoblasts, then their expression increases and they are assembled and inserted into the plasma membrane at a concentration of 10³ µm⁻²[106]. In mature muscles, nAChRs are present at high concentrations (10⁴ µm⁻²) in the postsynaptic membrane at the NMJ, but are almost absent from the remaining 99% of the muscle fiber surface [107]. Both transcriptional and post-translational processes contribute to this redistribution in mature muscle: (a) upon innervation, the diffusely distributed nAChRs migrate to synaptic areas, where they become anchored to the subsynaptic cytoskeleton; (b) nerve-derived factors cause transcriptional activation of the nAChR genes in those nuclei that directly underlie the synapse; and (c) muscle activity suppresses nAChR subunit gene expression in nonsynaptic nuclei and promotes expression by synapse-associated nuclei [96]. Synapse-specific nAChR transcription requires an enhancer sequence, the N-box motif, which is present in the promoter and/or intronic regions of nAChR subunit genes [108]. Expression of the ε subunit is therefore controlled by neurotrophic factors, whereas γ-subunit expression is regulated by both muscle electrical activity and neurotrophic factor(s) from the nerve, so that the γ subunit is expressed by extrasynaptic nuclei in developing, denervated and paralysed muscles and ε-subunit expression is confined to subsynaptic nuclei under all circumstances [89,96].

The multidomain proteoglycan, agrin, is a key factor in the clustering of pre-existing nAChRs and in local gene expression of postsynaptic proteins. It is synthesized by the nerve and released from motor nerve terminals in the synaptic cleft, where it activates the muscle-specific receptor tyrosine kinase (MuSK). As a result of alternate splicing, agrin exists in several isoforms that differ in their ability to induce nAChR clustering [109]. Agrin-deficient knockout mice have normal nAChR levels, but few postsynaptic clusters [106]. Following activation by agrin, MuSK undergoes autophosphorylation on several cytoplasmic tyrosine residues, which then act as docking sites for signal-transducing molecules and promote nAChR clustering through rapsyn [108–110]. Tyrosine phosphorylation of the nAChR β1 subunit links nAChR to the cytoskeleton and contributes to efficient clustering. Rapsyn interacts with the intracellular portion of nAChR subunits. Other effector molecules, such as neuregelin-1 and nitric oxide, have been implicated in promoting clustering and gene expression at the NMJ, but all are dependent on the agrin/MuSK signaling pathway [109,111,112]. Kukhtina et al. [113] demonstrated that the intracellular loop of the δ subunit is unfolded and predicted the presence of 12 functional binding motifs involved in protein–protein interactions. Thus, it is likely that other binding partners involved in nAChR assembly, trafficking and clustering will be discovered.

Electrophysiological studies have shown that the switch in the composition of nAChRs from fetal- to adult-type is a key process in the maturation of the NMJ, because it is accompanied by changes in the functional properties of endplate channels, which are the sites of neuromuscular contact [88]. Channel mean open time decreases from 5 to 1.5 ms during the first two postnatal weeks and mean channel conductance increases by 50%[114–117]. Fetal- and adult-type receptors have 39 and 59 pS single-channel conductance with increased Ca²⁺ influx and 10.4 and 5.3 ms open time, respectively [88]. Extracellular Mg²⁺ and Zn²⁺ affect differentially synaptic transmission of fetal- and adult-type receptors in response to ACh by reducing single-channel conductance of the former less potently than the latter [118]. Studies on recombinant T. californica nAChRs and chimeric fetal and adult mouse nAChRs, in which M1–M4 of the γ subunit were exchanged for those of the ε subunit and vice versa, suggested that the difference in conductance can be attributed to amino acid charge differences in segments M1–M4, especially in M2 of the γ and ε subunits [119]. By contrast, the difference in channel open time was attributed to M4 segment and the amphipathic helix between segments M3 and M4 [119,120]. This is in agreement with structural studies suggesting that cations (Mg²⁺ and Zn²⁺) interact strongly with the additional negative residues that line the ε-nΑChR's cytoplasmic and extracellular rings close to the membrane-spanning M2 segment [121]. The response of fetal and adult-type nAChRs to antagonists can also differ. For example, mouse fetal-type nAChRs are more sensitive to 5-hydroxytryptamine (5-HT), an open-channel blocker of muscle nAChRs, than adult-type nAChRs [122]. Subunit composition also affects the metabolic stability of nAChRs; the half-life of adult-type nAChRs at the adult MNJ is 10 days compared with 24 h for fetal-type nAChRs [123].

The γ- to ε-subunit switch coincides with other postnatal events in the maturation of the NMJ, such as transition from multiple to single axon innervation of the muscle fiber, elaboration of branched morphology by the endplates, formation of junctional folds, as well as localization and interaction with synaptic antigens, raising the possibility that these events might be related [97,117]. The functional significance of the switch has been investigated by studies on knockout mice for the ε-subunit gene [88,115,117,124,125]. Expression of fetal-type receptors partially compensated for the lack of adult-type receptors in these animals, which showed severe muscle weakness and died prematurely. Adult-type nAChRs are therefore essential for the maintenance of the highly organized structure of neuromuscular synapses in adult muscle [115,125]. Furthermore, the shorter mean open time of adult nAChRs contributes to the stabilization of the initial nerve–muscle contacts and synapse maturation in differentiated muscles [97,115]. Mutations in the pore-forming M2 region of the ε subunit causing high Ca²⁺ permeability by prolonging channel mean open time also demonstrated that the ε subunit is a key determinant of Ca²⁺ permeability of adult muscle and protects the synaptic region from the harmful effects of excessive calcium influx [116,117,126,127]. Mutant mice in which the fetal-type γ subunit was replaced with a chimeric one with ε-subunit-like functional properties had altered muscle innervation pattern, suggesting that fetal-type nAChRs ensure an orderly innervation pattern for skeletal muscle [115]. It has also been suggested that the long duration currents mediated by fetal-type nAChRs are necessary for developing spontaneous muscle contractile activity and stabilizing the clustering of nAChRs at the endplates [115,128].

Expression studies and mutational analysis have provided insights into the regulation of nAChR assembly. Processing and assembly of nAChR subunits is a slow process, taking ∼ 2 h to complete [129]. Torpedo nAChR assembly, in particular, is temperature sensitive and increases as the temperature decreases from 37 to 20 °C [130,131]. Two models have been proposed for nAChR assembly.

The ‘sequential model’ suggests that newly synthesized subunits either assemble rapidly into α1β1γ trimers or remain unassembled and are rapidly degraded [130,131]. The chaperone protein, calnexin, associates with 50% of newly synthesized subunits during the early events of nAChR assembly, but not after maturation [132], promoting stabilization of unassembled subunits into α1β1γ trimers and, by slowly dissociating from these, controls the rate of nAChR folding and oligomerization. A rate-limiting step in nAChR assembly is the formation of α-Bgtx-binding sites on α1β1γ trimers. Such trimers capable of binding α-Bgtx are short-lived intermediates and rapidly assemble into α1β1γδ tetramers, which can then bind ACh [133]. The addition of the α1 subunit to the α1β1γδ tetramers to form the second ACh binding site is another rate-limiting step. The ‘sequential model’ therefore suggests that subunits assemble at the interface between the β1 and γ subunits and that the second α1 subunit assembles at the interface between the γ and δ subunits. N-Linked glycosylation at residue 141 and the formation of disulfide bridges between cysteines 128–142 and 192–193 are also required for the correct maturation of the Torpedoα1 subunit and the subsequent formation of the α-Bgtx binding site, and are potential sites of regulation for nAChR assembly [70,134].

In the ‘heterodimer model’[135–137], the α1 subunits assemble with the γ or δ subunit, generating, respectively, α1γ or α1δ complexes and the ACh-binding sites are formed at the interface of the heterodimers. Following ACh-binding-site formation, the heterodimers assemble with β1 subunits into (α1)₂β1γδ heteropentamers. As in the ‘sequential model’, interactions with other subunits or chaperones are required to prevent misfolding. Evidence supporting this model came from experiments in which the α1 subunits were either expressed alone or in combination with one of the other subunits (β1, γ or δ) [133,138,139]. However, the ‘heterodimer model’ cannot account for the two distinguishable ACh-binding sites being formed at different times and on different subunit complexes during assembly [133].

Both models propose that discrete assembly intermediates are formed and confined in the endoplasmic reticulum (ER), whereas unassembled subunits or incorrectly processed and assembled complexes are recognized by the cellular machinery and degraded through proteasomes prior to reaching the Golgi apparatus, leading to the accumulation of only correctly assembled complexes [133]. The ubiquitin–proteosome system through ER-associated degradation is thought to be involved in the control of this process [140].

Expression of recombinant muscle-type nAChRs

The expression of recombinant polypeptides corresponding to functional domains of muscle nAChRs is essential for their detailed physiological and structural analysis. Because of the high yield of protein, prokaryotic expression systems were initially used. However, because of the lack of essential post-translational modifications, they are considered to be problematic, because aggregation of nAChR subunits occurs and denaturating conditions are therefore required to solubilize the protein. Eukaryotic expression systems, despite the usually lower yield, offer the advantage of post-translational modifications, such as glycosylation, so the expressed proteins may have near-native conformational features.

Full-length, truncated and mutant T. californica, mouse, rat and human muscle-type nAChR subunits have been expressed in E. coli, Saccharomyces cerevisiae, P. pastoris, baculovirus-infected insect cells, Xenopus oocytes, mouse fibroblasts, HEK 293 cells and COS cells. A far from exhaustive overview of some of these studies, which provided insights into the function of the individual subunits, is presented below.

Torpedo nAChR

The ECD of the T. marmorataα1 subunit (residues 1–209) was expressed in inclusion bodies in E. coli[26]. Following renaturation, it exhibited high α-Bgtx binding affinity, suggesting that it had a near-native conformation. The same fragment of the T. californicaα1 subunit ECD was also expressed in E. coli in inclusion bodies and was in vitro refolded [27]. The renatured protein bound α-Bgtx and mAbs raised against the intact T. californica nAChR [27]. Two fragments of the T. californicaα1 subunit, consisting of part of the ECD with and without the first transmembrane segment (residues 143–210 and 143–238, respectively) were expressed in E. coli and had a significant affinity for α-Bgtx [28]. This study was important in identifying the ECD determinants necessary for α-Bgtx binding [141].

Mouse nAChR

The ECD of the mouse nAChR α1 subunit (residues 1–210) fused with a glycosylphosphatidylinositol anchorage sequence has been expressed in mammalian cells (CHO cells) and Xenopus oocytes [25]. The purified protein bound α-Bgtx and had the conformation-specific binding sites expected for the correctly folded subunit. Furthermore, a similar construct (residues 1–211) has been expressed in the yeast, P. pastoris, as a soluble protein with a high affinity for α-Bgtx and conformation-dependent mAbs [29].

Human nAChR

We expressed the ECD of the human nAChR α1 subunit (residues 1–207) in E. coli and reconstituted it using the ‘artificial chaperone’ approach [142]. The reconstituted protein bound conformation-dependent anti-nAChR mAbs, but not α-Bgtx, apparently due to lack of N-glycosylation which seems to be necessary for high-affinity binding of α-Bgtx to mammalian muscle nAChR [75]. The slightly longer form of residues 1–210 has also been overexpressed, in our laboratory, as a water-soluble molecule in P. pastoris[75]. This glycosylated molecule bound α-Bgtx, conformation-dependent anti-nAChR mAbs and myasthenic patients' antibodies against nAChR [143]. A fusion between the ECD of human nAChR α1 subunit (residues 1–205) and maltose-binding protein was constructed and expressed in E. coli as a soluble protein [144]. The nAChR α1 fragment was recovered following protease cleavage. Both the fusion protein and the nAChR α1 fragment seemed to have a near-native conformation judging by the binding of conformation-dependent mAbs.

The ECDs of the human muscle non-α subunits, β1 (residues 1–221), γ (residues 1–218) and ε (residues 1–219), have also been expressed, by our group, in P. pastoris as soluble glycosylated proteins [48]. The proteins seem to have native-like conformational characteristics, as evidenced by the high binding affinity for conformation-dependent anti-nAChR mAbs and autoantibodies from the sera of MG patients [145]. Variants of these proteins with mutations and/or the addition of hydrophilic epitopes have also been constructed in our laboratory to increase their solubility and expression yield for structural trials (unpublished data).

Coexpressions

All four full-length T. californica subunits have been expressed in S. cerevisiae and had the antigenic properties of authentic Torpedo subunits [146]. Combinations of all four full-length T. californica subunits (α1, β1, γ and δ) have been stably expressed in Xenopus oocytes and mammalian cell lines and have provided insights into nAChR assembly and regulation [135,139,147–150]. These studies demonstrated that incorporation of the β1 subunit into the α1γδ complex is a rate-limiting step in nAChR assembly, that association of the α1 subunit with either the γ or the δ subunit is necessary for agonist binding, and nAChRs lacking the β1, γ or δ subunit show weak channel activity. Furthermore, coexpression of T. californica nAChR subunits carrying site-directed mutations at conserved sites (N-glycosylation and disulfide bond) and normal subunits in Xenopus oocytes led to intracellular retention of most of the assembled nAChR complexes and allowed the identification of regulation sites for nAChR assembly [71,72,151]. The ECDs of all four T. californica subunits have been coexpressed in baculovirus-infected insect cells and assembled into pseudo-pentameric native-like nAChR ECD complexes [49]. The intracellular loop of the Torpedoδ subunit has also been expressed in E. coli and, following urea treatment and efforts at refolding, was found to be unfolded by NMR analysis complemented by protein structure prediction algorithms, suggesting that an interaction with as yet unknown protein partners is required for it to assume an ordered conformation [113].

Hybrid bovine–T. californica and mouse–T. californica nAChRs have also been expressed in Xenopus oocytes as functional receptors [152–155]. These hybrids were useful for probing channel-gating determinants within the primary sequence of the individual subunits and assessing species differences in channel properties. Hybrid cat–T. californica nAChRs, expressed in Xenopus oocytes, have shown that species differences in the γ subunit primary sequence account for the observed differences in the properties of the receptors [156]. Torpedo and hybrid rat–Torpedo chimeras expressed in mammalian muscle cell lines provided insights into the temperature sensitivity of Torpedo nAChR assembly [157]. All muscle-type α1 subunits are encoded by a single gene, except in Xenopus, in which two genes have been identified; each of these was coexpressed with T. californicaβ1, γ, and δ subunits and the two nAChRs were found to have strikingly different affinities for α-Bgtx (∼ 1000-fold difference) [158]. Hybrid expression has also proved useful in structural studies. Expression of hybrid bovine–T. californica nAChRs in Xenopus oocytes showed that M2 and the bend between M2 and M3 of the δ subunit are important in determining the rate of ion transport through the open channel [159]. Hybrid T. californica α1–mouse β1, γ, δ nAChRs and human α1–T. californica β1, γ, δ nAChRs expressed in mammalian muscle cells (fibroblasts and TE671 rhabdomyosarcoma cells) revealed the immunodominance of antiα1 subunit sera in anti-MG sera [160,161].

Expression of combinations of mammalian subunits in Xenopus oocytes and muscle cell lines has allowed the measurement of the kinetics of subunit assembly, as well as the study of the electrophysiological and biochemical properties of the resulting complexes [100,136,137,162–168]. Coexpression of all four mammalian receptor subunits as functional receptors in muscle cell lines has also been carried out [100,116,168–170]. Furthermore, all four full-length mammalian receptor subunits have been expressed as fully assembled functional receptors in Xenopus oocytes and nonmuscle cell lines to determine the properties of the receptors in the absence of interactions with other muscle proteins [95,168,171–173]. The resulting receptors had the pharmacological, electrophysiological and metabolic properties of nAChRs normally found in myotubes and have proved valuable in studying nAChR synthesis, assembly and transport to the cell surface. Human α1 nAChR subunit exists in two isoforms (α1 and α1*), which differ in length, as mentioned previously. Each isoform has been coexpressed with the other three full-length fetal and adult subunits in Xenopus oocytes and the properties of the corresponding fetal and adult channels investigated [100]. These studies, complemented by site-directed mutagenesis of the two α1 isoforms, demonstrated that these isoforms differ in the pore-lining region and that only the α1 subunit, expressed in combination with the other subunits, is integrated into functional receptors. Differences in the electrophysiological properties of the corresponding fetal and adult receptors were also identified. Truncated fragments of the mouse nAChR α1 subunit containing the ECD and all, part or none of the first transmembrane segment and the δ subunit were coexpressed in COS cells, showing that the M1 segment is necessary for heterodimer formation [174].

Animal studies

Chimeric subunits consisting of the N-terminal of the γ subunit and the C-terminal of the ε subunit have been constructed by homologous recombination and expressed in mice [119]. Mice homozygous for the chimera were healthy and fertile and expressed nAChRs with the properties of the adult-type endplate throughout their lifetime; the authors concluded that one function of the fetal-type nAChR is to ensure the orderly innervation of the skeletal muscle [115].

Neuronal-type nAChRs

General features of neuronal-type nAChRs

Neuronal nAChRs are widely expressed in the nervous system in peripheral ganglia and certain areas of the brain, and in nonexcitable cells, such as epithelial cells and cells of the immune system. To date, nine α (α2–α10) and three β (β2–β4) subunit genes have been cloned. The α7–α10 subunits are found either as homopentamers (of five α7, α8 or α9 subunits) or as heteropentamers (of α7/α8 and α9/α10) [175]. By contrast, the α2–α6 and β2–β4 subunits form heteropentameric receptors, usually with a (α_x)₂(β_y)₃ stoichiometry. The α5 and β3 subunits cannot form functional receptors when expressed alone or in paired combinations with β or α subunits, respectively. They only form operational channels when coexpressed with other functional subunit combinations, and are thus commonly termed ‘orphan’ or ‘auxiliary’ subunits. Although these subunits were initially thought to have a structural role, emerging evidence showed that α5 incorporation into α6β4β3 receptors was required for high-affinity nicotinic ligand binding, suggesting that they play a critical part in the assembly and pharmacological properties of nAChRs and their role in ligand binding may have to be readdressed [176].

Earlier classification of the nAChRs was based upon their pharmacological properties. As a result, neuronal-type nAChRs were divided into two classes: (a) the high-affinity agonist-binding class (with nm affinities), which do not bind α-Bgtx, later found to be the heteropentameric nAChRs formed by α2–α6 and β2–β4 subunits; and (b) a second class that binds agonists with lower (µm) affinities and binds α-Bgtx with nm affinities (sometimes referred to as α-Bgtx–nAChRs), later shown to be usually homopentameric molecules formed by α7–α9 subunits [105]. The homopentameric nAChRs are thought to have five identical ACh-binding sites per molecule (one at each α subunit interface), whereas the heteropentameric nAChRs have two ACh-binding sites, located at the interface between an α and a β subunit.

nAChR subunit composition has been shown to be important in regulating the response to agonists and the subcellular localization of assembled channels. Ectopic expression of the β2 and β4 subunits with combinations of α2–α4 subunits in HEK 293 cells revealed that β2-containing nAChRs have a higher affinity for most ligands than β4-containing nAChRs [177]. Recently, Lukas's group [178] showed that α4 nAChRs containing either β2 or β4 subunits have distinct pharmacological and physiological properties, with α4β4 nAChRs having higher current amplitude and stronger responses to agonists than α4β2 nAChRs. Incorporation of an auxiliary subunit into an αβ nAChR alters its properties [179]; for instance, the α4α5β2 nAChR has a higher Ca²⁺ permeability and desensitization rate and higher EC₅₀ values than the α4β2 nAChR [180]. Moreover, when α5 is coexpressed with α3 and β2 subunits, the receptor formed shows increased ACh sensitivity compared with the α3β2 receptor [41]. Interestingly, this effect is not observed if β4 is used instead of β2, whereas an increase in desensitization rate and Ca²⁺ permeability is seen with both the β2- and β4-containing channels. Furthermore, Fischer and colleagues [181] have shown that deletion of α5 in mice has different effects depending on the localization of the nAChR. Somatic nAChRs from neurons of the superior cervical ganglion had different agonist potencies depending on the presence of α5, but the magnitude of the response was not affected, whereas, in the case of presynaptic receptors, the response was significantly higher in neurons not expressing α5. Incorporation of the α5 subunit has also been shown to affect the receptor affinity for agonists and antagonists in neurons of the autonomic nervous system [182].

Given the intricate role of each nAChR subtype at various positions in the cell and its contribution to signal transmission, nAChR targeting at specific sites is crucial. For instance, in the chick ciliary ganglion α3, α5 and β4 containing receptors are located at the interneuronal synapse, whereas α7 receptors are excluded from the synapse and are scattered perisynaptically. The cytoplasmic loop of the α3 subunit appears to regulate receptor targeting in this case, as exchanging the loop of α7 with that from α3, but not α5, causes α7 to aggregate in the synaptic region [183]. Studies in cultured hippocampal neurons showed that α7 nAChRs were distributed in the somatodendritic area, while α4β2 nAChRs were found both in the dendrites and axon. Further analysis using fusion proteins of CD4 or interleukin (IL)-2 receptor, normally universally distributed, with domains of the nAChRs intracellular loops reveiled that there is a 48 amino acid stretch of the α7 subunit loop conferring dendritic distribution and a 25 amino acid region in the α4 subunit loop responsible for axonal targeting [184].

Neuronal-type nAChR-mediated Ca²⁺ signaling

Neuronal-type nAChRs show a significant, subtype-dependent permeability to Ca²⁺ and a strong inward rectification (voltage dependency) [185,186]. The high Ca²⁺ selectivity of neuronal-type nAChRs is of pivotal physiological importance, because intracellular Ca²⁺ signals are involved in the rescue/demise of developing neurons and the modulation of their activity. Although, Ca²⁺ influx is often preceded by Na⁺ membrane depolarizing ion currents, within this part of the review, only the neuronal-type nAChR-mediated Ca²⁺ signaling are discussed.

The neuronal-type nAChR-mediated increase in the intracellular Ca²⁺ concentration results primarily from direct permeation of the nAChR pore, due to its high Ca²⁺ permeability. The Ca²⁺ influx causes membrane depolarization and subsequent activation of voltage-operated Ca²⁺ channels (VOCCs) [187,188] or, alternatively, triggers further Ca²⁺ release from ryanodine-dependent intracellular stores (calcium-induced calcium release, CICR), generating prolonged Ca²⁺ signals [188–191]. It seems that different neuronal-type nAChR subtypes are coupled with different Ca²⁺ pathways. The most Ca²⁺-permeable homomeric α7 nAChRs, albeit capable of activating VOCCs [192,193], mainly induce Ca²⁺ currents, which subsequently trigger the CICR machinery [188,194,195]. By contrast, neuronal nAChRs containing α3- and/or β2 subunits in the brain and ganglionic neurons operate primarily through Ca²⁺ signals coupled with the opening of VOCCs [188,190,194]. Thus, second messenger Ca²⁺ signals act as converters of the acute neuronal nAChR stimulation into sustained downstream effector functions, such as neurotransmitter release, gene expression and metabolism, which shape neuronal activity. Interestingly, a reciprocal relationship exists, with intracellular Ca²⁺ levels affecting neuronal-type nAChR activity [196,197].

Neuronal-type nAChR expression

Expression of neuronal-type nAChR subunits varies in different cell types and neurons located in different parts of the nervous system. nAChR α4, α5, α6, β2 and β4 subunits, expressed in mesostriatal dopaminergic neurons assemble in different nAChRs combinations that play a role in modulating the release of striatal dopamine (DA) [198,199]. α7 receptors, which account for the majority of the α-Bgtx-binding sites in the nervous system, are widely distributed and are present at a high concentration in the hippocampus, especially in GABAergic neurons [200–202]. α8-containing receptors have been found only in the chick nervous system as homopentamers or in combination with α7, while α9 receptors are mainly expressed in the cochlea and sensory ganglia of the nervous system [203]. Finally, α10 appears to serve as a structural subunit in the formation of functional receptors with α9 in cochlear mechanosensory hair cells; these receptors mediate transmission to efferent olivocochlear fibers [175,204].

Because different subunit combinations can lead to different pharmacological properties, the array of subunits expressed by any given cell affects its responses to stimuli. Dissection of the promoter of the α3 and α5 genes has revealed many different regulatory elements required to meet the needs of specific cells [205–208].

Once synthesized, nAChR subunits must assemble into pentameric channels and be transported from the ER through the Golgi to the cell surface. nAChRs that have not been properly assembled are targeted to the proteasome for degradation. Similar to muscle nAChRs, ubiquilin-1, a member of the ubiquitin protein family, is thought to be involved in controlling this process [209]. In the case of α7, it has been shown that, in COS cells, sequences within its M1 segment prevent its surface localization [210], whereas, in α4 and β2, a stretch of conserved hydrophobic amino acids in the cytoplasmic domain between M3 and M4 is required for efficient transport to the cell surface [211]. RIC-3, an ER/Golgi-localized protein, appears to have a dual effect on nAChR trafficking, retaining mature α4β2 nAChRs intracellularly and enhancing surface transport of functional α7 nAChRs [212,213]. α7 nAChRs form characteristic clusters in the somatic spines of ciliary neurons. Lipid rafts are essential for maintaining these clusters, as they promote the interaction between α7 nAChRs and rapsyn that is required for nAChR clustering [214,215]. Upon stimulation, α7 receptors are internalized and new receptors are targeted to the surface from intracellular pools, both processes being controlled by the soluble N-ethylmaleimide-sensitive factor attachment protein receptor (SNARE) [216]. Although this does not alter the number of surface nAChRs, it is necessary for maintaining the ability of the nAChR to respond to subsequent stimuli and propagate downstream signaling.

Exposure to nicotine results in an increase in the number of nicotinic ACh-binding sites. Early results showed that α4β2 nAChRs are upregulated in rat brains after chronic exposure to nicotine [217,218], and further investigation revealed that this upregulation is dependent on the nAChR subtype. More specifically, nicotine upregulates mostly high-affinity nAChRs such as α3β4 and α3β2 [218–221]. This upregulation is not dependent upon mRNA transcription or de novo protein synthesis, because there is no detectable increase in levels of subunit mRNA or total protein (surface plus intracellular), and protein synthesis inhibitors do not block the nicotine-induced upregulation. Furthermore, it appears that the flow of ions through the channel is not required either, because the channel blocker, mecamylamine, has an additive effect to nicotine [222]. Indeed, upregulation is due to an increase in the proportion of the nAChR that is expressed on the cell surface and this is thought to occur due to a nicotine-driven increase in the assembly of intracellular nAChR subunits and a decrease in the rate of their internalization [223,224]. Despite the increase in surface nAChRs, function is not enhanced and the upregulated channels have a decreased capacity for ion flow [219,221]. Upregulation is also observed in cells ectopically expressing nAChRs, indicating that this is not an adaptive mechanism to compensate for desensitized nAChRs [222,225].

Roles of neuronal-type nAChRs

Although neuronal-type nAChRs are major players in rapid synaptic transmission in the PNS [226,227], the relatively few examples of neuronal nAChR-mediated fast-signal propagation in the mammalian CNS [228–230] suggest that, in the brain, neuronal-type nAChRs act in a more sophisticated way, mainly as synaptic modulators [5]. Recently, the identification of several neuronal-type subtypes in non-neuronal cells and, most importantly, the discovery of the anti-inflammatory cholinergic pathway have attracted much attention to the role of the neuronal-type nAChRs in the periphery.

Neuronal nAChRs in the CNS

Modulation of neuronal transmission

Neuronal nAChRs are widely distributed in the brain. In terms of subcellular localization, they are found in pre-, peri- and postsynaptic areas [231,232] (Fig. 7). Presynaptic and perisynaptic nAChRs act as autoreceptors or heteroreceptors regulating the release of several important neurotransmitters, mainly ACh, DA, norepinephrine, glutamate, 5-HT and 4-aminobutyrate throughout the CNS. It is noteworthy that the synaptic release of a particular neurotransmitter can be regulated by different neuronal-type nAChR subtypes in different CNS regions. For instance, DA release from striatal and thalamic DAergic neurons is controlled, respectively, by the α4β2 subtype or both the α4β2 and α6β2β3 subtypes [233,234]. Interestingly, glutamatergic neurotransmission seems to be exempt from this ‘pluralistic’ neurotransmitter control network, being ubiquitously regulated by α7 nAChRs [233].

Figure 7

Neuronal nAChRs can be found at perisynaptic, presynaptic and postsynaptic areas. Depending on their distribution, they can exert a broad variety of functions modulating neuronal signaling at both the pre- and postsynaptic levels.

Presynaptic neuronal-type nAChRs modulate neurotransmitter release by promoting exocytotic mechanisms either through activation of VOCCs following membrane depolarization or by direct alteration of the intracellular Ca²⁺ concentration due to the intrinsic Ca²⁺ permeability of the neuronal-type nAChR pore. There is evidence suggesting that neurotransmitter release is affected by both mechanisms. DA release from striatal synaptosomes is mediated by non-α7 neuronal nAChR subtypes functionally coupled to VOCCs [235–237], whereas, at rat hippocampal glutamatergic terminals, Ca²⁺ entry through α7 nAChRs initiates CICR from presynaptic stores, inducing glutamate release and eventually eliciting bursts of miniature excitatory postsynaptic currents [238]. In addition to Ca²⁺-dependent exocytosis, neuronal-type nAChRs modulate neurotransmitter release through secondary messenger pathways, allowing them to exert more subtle regulatory actions. Ca²⁺-dependent protein kinase C has been implicated in striatal DA release by neuronal nAChRs [239] and extracellular signal-regulated mitogen-activated protein kinase (ERK/MAPK) stimulation by protein kinase C has been related to nicotine-evoked catecholamine release by adrenal medullary cells [240].

Postsynaptic nAChRs probably mediate a small proportion of the fast excitatory transmission in the CNS and are generally considered of less physiological importance than the presynaptic and perisynaptic nAChRs. Although it has been quite difficult to locate sites in the brain where ACh release produces fast postsynaptic nicotinic responses, the presence of postsynaptic α7, α4β2 and α3β4 subtypes has been demonstrated in several brain regions [233,241]. In particular, when the nicotinic postsynaptic excitation coincides with presynaptic glutamate release, postsynaptic nAChR activity has been shown to enhance the induction of synaptic potentiation in the hippocampus [242].

Regulation of gene expression by nAChRs

The expression of various neuronal-type nAChR subtypes early during embryogenesis [243–246] and the existence of a primary machinery enabling both ACh synthesis and the response to ACh [247] suggest a role for cholinergic signaling during early neuronal development. Although the importance of neuronal-type nAChR expression before neuronal differentiation remains obscure, the involvement of neuronal-type nAChRs in the regulation of early gene expression was originally suggested by the Ca²⁺-dependent regulation of c-fos gene transcripts after nAChR activation in PC12 cells [248]. Nicotine-induced upregulation of c-fos and junB gene expression has now been demonstrated in various brain regions [249–251]. Because these immediate-early genes, generally involved in mechanisms of abuse, function as transcription factors, their nicotine-mediated upregulation suggests that nicotine may modulate the expression of additional genes. In fact, exposure of SH-SY5Y cells to nicotine has been shown to alter the expression of a diverse set of proteins, including transcription factors, protein processing factors, RNA-binding proteins and plasma membrane-associated proteins, and there is evidence implicating activation of neuronal-type nAChRs in this gene profile alteration [252].

In addition, neuronal-type nAChR-mediated gene regulation provides a means for modulating neurotransmitter release per se, as demonstrated by the example of tyrosine hydroxylase [253,254]. This enzyme catalyzes the crucial rate-limiting step in catecholamine biosynthesis and is subject to various control mechanisms. Prolonged nicotine treatment upregulates tyrosine hydroxylase expression both in vivo and in cell culture [253], and stimulation of α7 nAChRs is implicated in the nicotine-triggered increase in tyrosine hydroxylase mRNA levels [254]. This effect is Ca²⁺ dependent and requires a sustained elevation of the intracellular Ca²⁺ concentration due to release from intracellular stores and downstream activation of protein kinase A or MAPK, which, in turn, phosphorylate the cAMP-response element-binding protein (CREB) [255–257].

Contribution to plasticity and memory functions

Neuronal-type nAChRs have been implicated in cognition, mainly because of the respective beneficial or adverse effect of neuronal-type nAChR agonists and antagonists on learning and memory. The hippocampus, generally thought to be the key brain area for the encoding and retrieval of memory, has attracted particular interest with respect to neuronal-type nAChR activity. In the hippocampus, at least three distinct functional neuronal-type nAChR subtypes (α7, α4β2 and α3β4) can be detected [258]. Synaptic function in the hippocampus seems to result from GABAergic (inhibitory) and glutamatergic (excitatory) inputs to the hippocampal interneurons, which are modulated by different nAChR subtypes exhibiting different desensitization rates upon agonist stimulation [242,259,260].

Recent significant progress in the field of the molecular pathways underlying human cognition has implicated the ERK/MAPK signaling cascades in various cognitive mechanisms [261]. Neuronal-type nAChRs mediate the Ca²⁺-dependent activation of ERK/MAPK and CREB in several neuronal cell lines [262–264]. In the hippocampus in particular, nicotine-evoked neuronal-type nAChR stimulation causes store-mediated Ca²⁺-influx, promoting activation of Ca²⁺-calmodulin-dependent protein kinase and ERK/MAPK and CREB phosphorylation, as shown in vitro[265,266].

Ca²⁺ signaling and the ERK/MAPK cascades are also involved in the neuronal-type nAChR-mediated neuroprotection of hippocampal neurons in vitro[267,268]. Neuronal-type nAChRs are generally involved in neuroprotection/neurodegeneration. This has been demonstrated both in animal models, in which increased neurodegeneration is observed in aged β2-subunit knockout mice [269], and in humans suffering from AD, in whom α4β2 nAChRs are markedly reduced in the brain [270] and α7 nAChRs seem to interact with the amyloid plaque-component, β-amyloid peptide [271], as discussed below. Interestingly, a recent study proposed an alternative mechanism in which β-amyloid peptide induces the pathology of AD by triggering a postsynaptic α7-mediated increase in N-methyl-d-aspartate receptor endocytosis [272]. Additional studies will be required to unravel this novel mechanism possibly leading to dementia.

Cholinergic mechanisms of reward and addiction

The ventral tegmental area (VTA) has been implicated in the rewarding motivational effects of a wide variety of addictive drugs, including nicotine [273,274]. There are two major DAergic routes extending from the VTA to the nucleus accumbens (mesolimbic pathway) or the prefrontal cortex (mesocortical pathway). Within the VTA, both DAergic and GABAergic neurons are implicated in signaling pathways of reward [274], the latter providing inhibitory input to DAergic neurons [275]. The early acute effects of nicotine predominantly affect GABAergic neurons, where 4-aminobutyrate release is modulated by rapidly desensitizing neuronal-type nAChR subtypes [198,276]. In fact only the α4 and β2 subunits are found on GABAergic neurons [277]. Hence, the stimulation and fast subsequent desensitization of α4β2 nAChRs produce prolonged stimulation of DAergic neurons by removal of the inhibitory 4-aminobutyrate input (Fig. 8). Nicotine also acts on presynaptic neuronal-type nAChRs located on VTA glutamatergic terminals which show a lower rate of desensitization after nicotine exposure, eliciting an excitatory glutamatergic input to the DAergic neurons (Fig. 8) [276,278]. Indeed, several studies have suggested that α7-containing nAChRs mediate the presynaptic actions of nicotine in the CNS, especially glutamate release [279–281]. Finally, DAergic neurons express nAChR α4, α3, α5, α6, β2 and β3 subunits [277] (Fig. 8). The somatodendritic nAChRs, in the form of various neuronal nAChR subunit combinations, on DAergic neurons of the VTA can directly excite these neurons by receiving cholinergic signals ascending from outside the VTA [282]. This process eventually results in transient DA responses that are terminated by nAChR desensitization [283–285].

Figure 8

Modulation of DA release in the VTA by nAChRs. Native nAChRs modulate DA release in DAergic neurons of the mesolimbic and nigrostriatal pathways both directly and indirectly. nAChR agonists exert direct modulation of DA release through presynaptic and preterminal nAChRs. Alternatively, activation of α7 nAChRs on glutamatergic terminals trigger release of Glu, which in turn stimulates ionotropic Glu receptors on DAergic terminals, finally inducing DA release. Additionally, desensitization of the α4β2 nAChR on GABAergic interneurons can remove the inhibitory 4-aminobutyrate (GABA) input on DAergic neurons, thereby indirectly eliciting DA release. Modified from Jensen et al.[234], with permission.

The role of neuronal-type nAChRs in non-neuronal tissues and cells

Apart from the importance of neuronal-type nAChRs in the nervous system, their expression has been demonstrated in ‘nonexcitable’ cells, namely lymphocytes, monocytes, macrophages, dendritic cells, adipocytes, keratinocytes, endothelial cells and epithelial cells of the intestine and lung [286–289]. The complex role of neuronal-type nAChRs in the periphery is reflected by their ability to attenuate or enhance the progression of several pathologies, e.g. attenuate ulcerative colitis and enhance Crohn's disease. In addition, different nAChR types are responsible for different functions. For example, it has recently been shown in the human lung that, although stationary cells express mainly α7 nAChRs [290], migrating bronchial epithelial cells express α3α5β2 receptors, indicating that they may be involved in the wound-repair process [291], whereas in keratinocytes, both α3- and α7-containing receptors are present and are responsible, respectively, for mediating keratinocyte chemokinesis and chemotaxis [292]. Most importantly, the key position of neuronal-type nAChRs as the intermediate link between the involuntary nervous system and inflammation has stimulated interest in the field of anti-inflammatory therapeutics. The role of neuronal-type nAChRs in inflammation and angiogenesis is discussed in detail below.

The cholinergic anti-inflammatory pathway

The vagus nerve is a ‘parasympathetic’ nerve that emanates from the cranium and innervates all major organs, creating a neuronal network of peripheral surveillance. It branches into both sensory (input) and motor (output) fibers, thus establishing a bidirectional connection between the brain and the immune system. Thus, once the sensory fibers of the vagus nerve are activated by stimuli resulting from persistent inflammation, the brain reacts to this information by activating the efferent fibers of the vagus nerve, which deliver ‘immuno-suppressive’ messages to the periphery.

Various findings suggested the existence of a cholinergic anti-inflammatory mechanism: upon stimulation, the vagus nerve reflexively releases ACh, and ACh significantly reduces the release of pro-inflammatory cytokines in lipopolysaccharide-stimulated human macrophage cultures through α-Bgtx-sensitive nAChRs [293]. However, the identity of the specific nAChR subtype on macrophages remained elusive until recently. It is now well-documented that physiological (vagus nerve-secreted ACh) or pharmacological (exposure to agonists, such as nicotine) stimulation of the homopentameric α7 nAChR, present on the surface of tissue macrophages, blocks the expression of the pro-inflammatory cytokines, tumor necrosis factor (TNF), IL-1, IL-6 and IL-18, as well as the secretion of the high-mobility group box 1 (HMGB1) protein [294–296] (Fig. 9). Interestingly, the α7 nAChR controls cytokine production at the post-transcriptional level, without affecting levels of mRNAs for TNF, IL-1, IL-6 and IL-18 [295]. The situation is different for HMGB1, the constitutive intracellular expression of which is crucial for the survival and normal transcriptional regulation of macrophages [297]. α7 nAChR blocks the secretion, rather than the translation of HMGB1, probably by inhibiting its translocation from the nucleus to the cytoplasm [295]. In terms of the subcellular mechanisms underlying the cholinergic anti-inflammatory machinery, several studies have provided evidence for the α7-mediated inhibitory action of nicotine on the nuclear factor-κB (NF-κB) pathway, which is crucial for macrophage activation and pro-inflammatory cytokine secretion [295,298,299]. The Jak2-STAT3 pathway is also involved, as shown both in vivo and in vitro, because α7 nAChR stimulation after nicotine binding induces phosphorylation of Jak2, which in turn activates STAT3 [300].

Figure 9

The ‘nicotinic anti-inflammatory pathway’. The vagus nerve can regulate inflammation through an α7 nAChR-mediated physiological pathway. Homomeric α7 nAChRs are present on the surface of tissue macrophages. Upon physiological (vagus nerve-secreted ACh) or pharmacological (via exposure to agonist, such as nicotine) stimulation, the production of the pro-inflammatory cytokines TNF-α, IL-1, IL-6 and IL-18 is inhibited and the inflammatory crisis controlled. Reproduced from Ulloa [296], with permission.

Implication in angiogenesis

Nicotine induces the proliferation of endothelial cells in vitro[301] and endothelial cells express functional neuronal-type nAChRs [302], which are mainly α7 nAChRs [290,303], but also α3-, α4-, β2- or β4-containing nAChRs [303]. The involvement of neuronal-type nAChR-mediated cholinergic pathways (mainly α7) in endothelial cell growth and angiogenesis, including tumor angiogenesis, has been demonstrated [303,304]. Formation of the cholinergically induced endothelial network is partially dependent on vascular endothelial growth factor and completely dependent on the phosphatidylinositol 3-kinase and MAPK pathways, eventually resulting in NF-κB activation [303]. According to a recent study, nicotine abrogated the apoptotic effect of several chemotherapeutic agents in nonsmall cell lung carcinoma cell lines. Its action is exerted through an α3-containing nAChR subtype and requires the activation of an Akt pathway [305,306].

Expression of recombinant neuronal-type nAChRs

The large number of nAChR subunits and the fact that many subunits are usually expressed in a given cell type complicate the study of their properties and functions. A strategy to avoid this problem is the ectopic expression of recombinant receptor subunits or combinations. Cells commonly used for the heterologous expression of neuronal-type nAChRs include Xenopus oocytes, epithelial cells and neural crest-derived cell lines. Expression of full-length α7 in the human epithelial cell line SH-EP1, which are null for nAChRs, showed that it retained the same pharmacological and functional properties with the native receptor, as evidenced by ligand binding and current responses and can therefore be used for functional studies [307,308]. However, some studies indicate that not all cells can form functional α7 nAChR channels. Studies on the ectopic expression of α7 in neural-crest-derived rat cells (PC12 and GH3), mouse fibroblasts (NIH 3T3) and insect ovary cells (Sf9) showed that only the cell lines of neural-crest origin were able to produce assembled receptors as indicated by ¹²⁵I-labeled α-Bgtx binding, although all transfected cells expressed the transgene as transcribed mRNA when transfected [309]. In a similar study, α7 was expressed in SH-EP1 and GH4C1 (rat pituitary) cells in varying amounts but not in CHO, CV1 (monkey fibroblasts) or SN56 (fusion of mouse septal neurons and a neuroblastoma line) cells, whereas α4 and β2 receptors were produced in all five cell lines [310]. These findings suggest that there are cell-specific factors regulating receptor formation and maturation at the post-translational rather than the transcriptional level and these should be taken into account when trying to ectopically express a nAChR.

Xenopus oocytes have been used for the production of truncated forms of α7, α4 and β2, consisting of their ECDs with or without the M1 segment. α4 and β2 ECD coexpression resulted in detectable epibatidine binding only in the presence of the M1, although both forms of the proteins were expressed and glycosylated [311]. By contrast, the α7 subunit ECDs had ligand-binding affinities and sedimentation velocities indicative of native-like receptor formation irrespective of the presence of the M1 [312]. However, the M1 was required for a production yield comparable to that of the full-length receptor, probably due to limitations in folding and assembly. This evidence suggests that, at least for some neuronal-type subunits, expression of the ECD together with the M1 segment may be required to allow efficient protein formation.

The α4 and β2 subunits have also been expressed in Xenopus oocytes as concatamers, in which the C-terminus of α4 was linked by a 18–36 amino acid linker peptide to the N-terminus of β2, and vice versa [313]. The resulting receptors were more efficient than those formed by unlinked subunits. Taking this a step further, Groot-Kormelink et al. [314] expressed α3β4 receptors as pentameric constructs connected by a flexible linker, resulting in the formation of receptors with a predetermined stoichiometry and subunit arrangement. This approach could prove valuable in dissecting variations between receptors with minor differences in subunit composition and organization.

Such systems are useful for the study of pharmacological and functional properties of nAChRs, but are not suitable for the production of large quantities of receptor protein, which would be useful in the elucidation of receptor structure. As previously mentioned, most attempts at large-scale production have been targeted at the muscle-type nAChR subunits, but some have been carried out on neuronal-type nAChR subunits. nAChRs are large integral membrane proteins and are therefore difficult to produce in large quantities. In addition, their transmembrane hydrophobic character makes structural studies very difficult. However, it may be possible to circumvent this by expressing only the ECD of the receptor subunits, because this contains all the elements of the ligand-binding site and folds independently of the rest of the molecule, as shown by chimeric nAChR α7-5HT3 receptors [315]. The α7 ECD has been expressed in E. coli as a fusion protein with maltose-binding protein and glutathione S-transferase (GST) [316,317]. A downside was that the protein was found in insoluble inclusion bodies, which need to be solubilized with urea or guanidine, but most of the resulting refolded protein was in the form of high molecular mass aggregates. However, when 0.1% SDS was added during the refolding procedure aggregates were greatly decreased [30]. In the same study, a mutant α7 ECD is described with a C116S mutation, which appeared to further decrease aggregate formation and showed a slightly improved α-Bgtx binding affinity. More recently, a double mutant form of the α7 ECD was expressed in P. pastoris by our group. In this mutant, in addition to the C116S, the α7 Cys loop was exchanged for the Cys loop of the soluble acetylcholine-binding protein (AChBP) [50]. The resulting protein had a greatly improved solubility as well as a higher affinity for α-Bgtx compared with the wild-type α7 ECD expressed in P. pastoris or E. coli. Further work has shown that mutation of additional hydrophobic residues exposed on the surface of the molecule, as identified from a model of the human α7 ECD we have constructed based on the recently solved structures of the L. stagnalis AChBP and the Torpedo muscle nAChR, offers extra solubility and enhanced α-Bgtx binding affinity (M. Zouridakis, P. Zisimopoulou, E. Eliopoulos, K. Poulas and S. J. Tzartos, in preparation). Electron microscopy studies (by N. Unwin, MRC, Cambridge, UK) showed that the recombinant protein forms particles similar to the expected ECD of the whole nAChR (Fig. 10). Currently, efforts are also being made by our group to produce receptor subunits in higher eukaryotic systems, insect and mammalian expression systems, aiming towards a more physiological glycocylation and maturation of the proteins.

Figure 10

Electron microscopy image of individual human nAChR α7 ECD oligomers expressed in P. pastoris. Negative-stain electron microscopic image of glycosylated human α7 ECD novel mutant (M. Zouridakis, P. Zisimopoulou, E. Eliopoulos, K. Poulas and S. J. Tzartos, manuscript in preparation). Arrowheads indicate an end-on view of the ECD oligomers. Images were kindly taken by N. Unwin.

nAChR-associated diseases

Muscle-type nAChR-associated diseases

The muscle-type nAChR is known to be the target in several inherited and acquired diseases, most of which lead to impaired neuromuscular transmission and muscle weakness. The acquired autoimmune disease MG, is the most common and best studied of these and is usually caused by autoantibodies to muscle nAChRs [2,105,318,319]. The rather rare inherited conditions, called congenital myasthenic syndromes (CMS), are associated with several abnormalities affecting ACh release, acetylcholinesterase activity, nAChR function and/or nAChR number [320].

Myasthenia gravis

MG is a remarkably heterogeneous noninherited autoimmune disease, usually characterized by the presence of circulating autoantibodies that bind to and destroy muscle nAChRs. Its hallmarks are muscular weakness and fatigability. It can either remain localized to a single muscle group (e.g. ocular MG) or can spread to several skeletal muscles (generalized MG). Autoantibodies reduce the number of available nAChRs, causing a defect in neuromuscular transmission and muscle weakness [319]. This weakness is prominent in muscles that are used frequently and repetitively, for example, the ocular, bulbar and facial muscles (resulting in ptosis and double vision), while other commonly affected muscles include those of the limbs, neck and shoulders. Although about 85% of MG patients have antibodies against nAChR, ∼ 15% of patients with generalized MG do not (seronegative MG) [321]. About 20–40% of seronegative MG patients have serum antibodies against the postsynaptic protein, MuSK [322]. Other autoantibodies against striated muscle tissue proteins (mainly titin and ryanodine receptors) have been found in subgroups of MG patients [323]. Current medications control MG by improving neuromuscular transmission or down-modulating the immune system. They include acetylcholinesterase inhibitors and steroidal or nonsteroidal immunosuppressants, and, in severely affected patients, administration of intravenous immunoglobulins or plasmapheresis. Surgical treatment consists of thymectomy. MG is the most common disease affecting the NMJ, with a prevalence reported to be higher than 70 per million [324].

The autoantibodies can reduce the number of functional nAChRs at NMJs by at least three mechanisms. First, being bivalent molecules, antibodies can cross-link nAChRs in the muscle postsynaptic membrane, thus stimulating internalization and degradation, leading to an overall loss of nAChRs, a process called antigenic modulation. A second, and probably the most important, mechanism is the complement-mediated focal lysis of the postsynaptic membrane. Antibodies bind to nAChR and trigger the complement cascade, resulting in the focal destruction of the postsynaptic membrane by the membrane attack complex. Third, antibodies against the nAChR-binding site can directly inhibit receptor function [40,325].

The autoimmune character of MG was first demonstrated when myasthenic symptoms were observed in rabbits immunized with nAChR [326] and, since that time, the pathogenic role of nAChR antibodies in MG patients has been established by several approaches (e.g. detection of circulating antibodies against nAChR, passive transfer of the disease from humans to animals, localization of immune complexes, IgG and complement on the postsynaptic membrane and the beneficial effects of plasmapheresis). Novel therapeutic approaches to MG are now being explored in animal models. Active immunization of animals with purified nAChR from different species induces acute or chronic experimental autoimmune MG (EAMG), with symptoms starting about 10 days after immunization. This EAMG model is characterized by a high antibody titer against nAChR, loss of more than half of the nAChRs in muscles and the presence of antibodies bound to the remaining receptors in the postsynaptic membrane. The process is thought to be mediated by both antibody and complement-mediated focal lysis of the postsynaptic membrane. The pathological mechanisms that impair neuromuscular transmission in MG and chronic EAMG are similar [327]. Chronic EAMG provides a good model for testing new therapeutic approaches. However, this model does not share the as yet unknown molecular and/or cellular mechanisms that sustain the autoimmune response to nAChRs in humans with MG. Alternatively, EAMG can be passively transferred to animals by injection of serum from MG patients or animals with EAMG or of anti-MIR mAbs [328,329].

More than half of the antibodies against nAChR in MG patients and animals with EAMG are directed against the MIR [40]. The MIR was originally defined by the ability of mAbs against the nAChR to inhibit the binding to the nAChR of other antibodies or mAbs against nAChR. It is now established that the MIR is not a single epitope, but rather a cluster of overlapping conformation-dependent epitopes, with residues 67–76 of the α1 subunit forming the core of the region. This segment is on the extreme synaptic end of each of the two α1 subunits. These extreme ends of the α subunits are the most accessible parts of the nAChR (Fig. 2) and this accessibility may be relevant to the high immunogenicity of the MIR [14,22]. mAbs against the MIR exhibit almost all of the pathological properties of serum autoantibodies in MG [40].

The mechanisms that induce and sustain the antibody-mediated autoimmune response to the muscle nAChR are not clear and may differ in various forms of the disease. MG onset requires a genetically favorable background combined with a number of environmental stimuli. In most MG patients, the immunogen is likely to be native nAChR, because the spectrum of autoantibody specificities is very similar to that seen in animals immunized with native muscle nAChR and different from that of animals immunized with denatured muscle nAChR [105]. The fetal form may be the immunogen in some MG cases, because MG sera often react with epitopes unique to this subtype [330].

Because of the frequent thymic abnormalities found in MG patients and the probable clinical improvement after thymectomy, it is believed that the thymus is implicated to the onset and/or maintenance of MG. In the thymus, both single nAChR subunits and whole nAChR molecules are normally expressed by thymic epithelial cells and myoid cells [331]. It has been proposed that an inflammatory response in the hyperplastic thymus modulates the expression of nAChR by thymic epithelial cells and myoid cells [332] and that this nAChR expression in the hyperplastic thymic environment, combined with a genetic predisposition, could be sufficient to prime nAChR-responsive CD4⁺ T cells, leading to autoimmunity [333]. Supporting evidence was provided by Poea-Guyon et al. [334], who demonstrated that a large number of IFN-γ- and TNF-α-regulated genes are highly expressed in the myasthenic thymus. Moreover, a recently proposed hypothesis by Bernasconi and colleagues [335] tries to correlate innate immunity and MG as a result of the increased levels of Toll-like receptor 4 (TLR4) mRNA seen in patients with thymitis, compared with patients with thymomas.

Thymomas are thymic epithelial tumors frequently associated with paraneoplastic autoimmunity. Although no whole nAChR molecules are found in thymomas, single nAChR subunits have been detected. However, it is uncertain whether thymomas actively sensitize, or simply fail to induce tolerance in, newly produced thymocytes autoreactive with the nAChR [336–338].

The understanding of the mechanisms underlying MG has led to the design of novel therapeutic strategies. Tolerance induction (oral or nasal) [339], T-cell receptor vaccination [340], protective antibody fragments against nAChR [340–344] and phosphodiesterase inhibitors [345] are some of the experimental approaches under investigation. Our group, having successfully expressed the ECDs of the human α1, β1, γ and ε subunits in P. pastoris in water-soluble form, is developing a specific antibody against subunit apheresis method [48,75]. The immobilization of each ECD on an insoluble carrier (CNBr–Sepharose) provides ‘immunoadsorbent’ columns that can be used for the ex vivo elimination of patients' antibodies against nAChR. In addition, ECD-immunoadsorbents have allowed the isolation of subunit-specific autoantibodies from the sera of MG patients and the study of their in vitro (cell cultures) and in vivo (EAMG) pathogenicity. To date, autoantibodies to the α1 and β1 subunits have been isolated and their potency has been compared with those of the untreated and antibody-depleted sera (unpublished data). The anti-α1 autoantibodies were found to be much more potent than the anti-β1 autoantibodies in terms of their modulating and pathogenic activity, and the depleted sera showed a dramatically reduced activity compared with the untreated sera (unpublished data), strongly suggesting that the antibody against nAChR is the main/sole pathogenic factor in the MG sera. The elucidation of the role of autoantibodies to each single nAChR subunit will shed light on the pathogenic mechanisms of MG and should lead to novel therapeutic approaches.

Congenital myasthenic syndromes

Congenital myasthenic syndromes (CMSs) are a highly heterogeneous group of inherited disorders, characterized by defective neuromuscular transmission, resulting in muscle fatigue. Following the identification of mutations in nAChR subunits, other genes encoding presynaptic, synaptic or postsynaptic proteins were also identified as candidate genes for CMSs [320,346–348]. Engel [348,349] initially classified these syndromes according to the lesion site (postsynaptic, synaptic or presynaptic) and pathophysiology. The classification of CMSs is still tentative, as further studies, including mutation analysis and chromosome studies, are likely to provide further information. To date, the majority of these disorders present abnormalities in postsynaptic function at the NMJ.

The first myasthenic symptoms of CMSs occur early in life, usually in the first two years after birth. In some rare cases, onset occurs in the second to third decade of life. The severity and course of CMSs are highly variable, ranging from minor symptoms to progressive disabling weakness and even death. Clinical diagnosis of CMSs is often possible on the basis of myasthenic symptoms involving fatigable weakness of the ocular, bulbar or limb muscles present since birth or early childhood [350]. On electromyography, a decremental compound action potential on repetitive low-frequency stimulation of the motor nerve and a negative test for antibodies against nAChR and calcium channel can help distinguish CMSs from MG and Lambert–Eaton syndrome.

To date, ∼ 60 mutations in genes encoding the different nAChR subunits (α1, β1, ε and δ) have been reported, and affect the ECDs, transmembrane segments M1–M3, and the cytoplasmic domain between M3 and M4 [350]. Despite the diversity of these mutations, we can group them into two major categories: kinetic and low-expressor mutations. Kinetic mutations fall into two subclasses according to whether they increase the response to ACh (as in slow-channel syndromes) or decrease the response to ACh (as in fast-channel syndromes).

Slow- and fast-channel syndromes

The main characteristics of these syndromes are kinetic abnormalities of nAChR function. The term ‘slow-channel syndrome’ originates from the abnormally slow decay of the synaptic currents caused by abnormally prolonged opening of the nAChR channel. As a consequence, the postsynaptic region is overloaded with cations, which causes endplate myopathy, loss of nAChR from the folds, widening of the synaptic space, vacuolar change and apoptosis. In most cases, slow-channel syndromes appear early in life and cause severe disability by the end of the first decade [351].

Slow-channel syndromes are caused by dominant gain-of-function mutations. At least 18 slow-channel mutations have been reported [351–353]. Croxen and colleagues [354] also recently described a recessive loss-of-function mutation resulting in a slow-channel syndrome. Although the majority of slow-channel syndromes result from mutations in M2 of the α1, β1, δ and ε subunits, some are caused by mutations in other functional domains of the subunits. Of particular interest are the αG153S mutation (near the ACh-binding site) and the αN217K mutation (in the N-terminal region of M1), which act mainly by increasing the affinity for ACh [353].

The term ‘fast-channel syndrome’ originates from the abnormally fast decay of the synaptic response, caused by abnormally brief channel opening due to decreased affinity for ACh, decreased gating efficiency or a decreased number of openings of the channel upon acetylcholine binding/occupancy [355]. The majority of fast-channel syndromes are caused by recessive loss-of-function mutations; however, a single missense mutation in the nAChR α1 subunit gene, causing replacement of Phe256 in M2 with leucine, has a dominant-negative effect [356]. The mutations are located in different functional domains of the α1, β1 and δ subunits. Mutations in the ECDs decrease the affinity for ACh, those in transmembrane sections impair gating efficiency and those in the long cytoplasmic loop of the ε subunit destabilize channel kinetics. Usually, the mutated allele causing the kinetic abnormality is accompanied by a null mutation in the second allele. At least 13 fast-channel mutations have been identified [320,347].

The clinical features of fast-channel syndromes resemble those of MG, but tend to be milder when they affect gating efficiency [357], moderately severe when channel kinetics are impaired [358] and severe when the affinity for ACh is affected [359,360]. In most fast-channel syndromes, therapy relies on combined treatment with 3,4-diaminopyridine and cholinesterase inhibitors.

nAChR deficiency with or without kinetic abnormalities

In patients with CMS, several homozygous or heterozygous recessive mutations in the nAChR subunits are found that result in a reduced number of functional nAChRs at the postsynaptic membrane. These low-expressor and null mutations have been reported in all subunits of the adult nAChR, but are more frequent in the ε subunit, particularly its long cytoplasmic M3–M4 linker. Patients with mutations in the ε subunit have milder symptoms than those with mutations in other subunits.

More than 50 ε subunit mutations have been reported. Some of these cause premature termination of translation by producing a nonsense or splice site [361] or frame-shift mutations [352,362,363]. In addition, some missense mutations alter residues essential for assembly (glycosylation sites, cysteine loop) or occur in the signal peptide, resulting in reduced gene expression. Point mutations of regulatory elements (N-box) or the promoter region of the ε subunit gene also result in low ε subunit gene expression [364,365]. Of particular interest is the 1369delG mutation, which results in the loss of the C-terminal cysteine, Cys470, crucial to both the maturation and surface expression of the adult receptor [366]. In addition, the frame-shifting ε1267delG mutation, occurring in Romany populations, results in nAChR deficiency at the endplate [362,363,367]. The prevalence of this mutation appears to be high due to a founder effect in the Romany population [368].

In nAChR deficiency, most patients respond quite well to anticholinesterase drugs, while others derive additional benefit from 3, 4-diaminopyridine [369].

nAChR deficiency caused by rapsyn mutations

Rapsyn is a 43 kDa postsynaptic protein that, together with agrin and MuSK, plays an essential role in nAChR clustering at the postsynaptic membrane. Mutations in rapsyn were discovered in patients diagnosed with CMS who showed endplate deficiency without any mutation in any nAChR subunit [370]. Endplate studies in these patients revealed decreased staining for rapsyn and nAChR, as well as impaired postsynaptic development.

Twenty-one rapsyn mutations have been identified [320,347] in the coding (missense, frame-shift, stop and splice site mutations) and promoter regions. The missense mutation, N88K, has been identified in all patients with mutation in the RAPSN gene and can be either homozygous or heterozygous. Other mutations in the coding region result in mutations in different domains of the protein. Patients carrying rapsyn mutations can have mild or severe symptoms. Most respond moderately well to anticholinesterase drugs, and some derive additional benefit from 3,4-diaminopyridine.

Mutations in the MuSK gene

MuSK plays a crucial role in the agrin–MuSK–rapsyn pathway by organizing the postsynaptic scaffold and nAChR aggregation. Recently, two heteroallelic mutations in the MuSK gene, a frame-shift mutation (c.220insC) and a missense mutation (V790M), were identified in a single individual with a CMS phenotype. The frame-shift mutation resulted in absence of MuSK expression. The missense mutation did not affect MuSK catalytic activity, but reduced the expression and stability of MuSK, leading to decreased agrin-dependent nAChR aggregation [371]. As described above (in the MG section), MuSK also plays an important role in autoimmune MG, in which 20–40% of anti-nAChR-seronegative MG patients have antibodies against MuSK, which are believed to be the pathogenic factor in these patients [322].

Neuronal-type nAChR-associated diseases

In contrast to the muscle-type nAChR, the pathophysiological functions of neuronal-type nAChRs are not well defined. It is well documented that brain nAChRs participate in complex functions, such as attention, memory and cognition, and clinical data suggest their involvement in the pathogenesis of several disorders (AD and PD, schizophrenia, depression, etc.).

Alzheimer's disease

AD, the most common cause of dementia, is characterized by a progressive decline in cognitive function, particularly affecting memory, attention and orientation, whereas motor and sensory abilities are usually undisturbed. It appears that many biochemical events crucial for neuronal communication and synaptic plasticity fail during the course of the disease [372]. One prominent hallmark of AD is an early and pronounced loss of cholinergic function (the ‘cholinergic hypothesis’) [373]. Muscarinic AChRs have attracted most attention because nAChRs are not as abundant as mAChRs in the brain, and the contribution of deficits in nicotinergic transmission in AD has not received as much attention as deficits in signal transduction in the muscarinic system. Nevertheless, the nicotinic system has been shown to modulate attentional processes and is involved in facilitation of memory. It is thus very likely that deficits in nicotinic signaling are also involved in the behavioural and cognitive deficits seen in AD.

A number of studies have reported reduced numbers of central nAChRs in the aged and in AD patients. It has been speculated that this reduction might be caused by a preferential presynaptic location on degenerating projection neurons [374]. Stimulation of nAChRs (especially α7 subtype) and treatment with nicotinic agonists are proved to protect neurons [375].

Because the deposition of brain amyloid plays a role in the neurodegeneration associated with AD, the relationship between amyloid deposition and cholinergic neuron activity is of great interest. Nicotine has been shown to inhibit the development of cellular toxicity induced by beta-amyloid peptides. Wang et al. [376] showed that both the 42-amino acid beta-amyloid peptide, Abeta(1–42), the predominant beta-amyloid peptide species in amyloid plaques, and the α7 nAChR are present in neuritic plaques and colocalize in individual cortical neurons. Using extracts of human brain tissue and cells that overexpress either α7 nAChR or amyloid precursor protein, Abeta(1–42) and α7 nAChR were coimmunoprecipitated by antibodies, suggesting that they are tightly associated. Abeta(1–42) and α7 nAChR bind with high affinity, and this interaction can be inhibited by α7 nAChR ligands. However, Lamb and co-workers [377] expressed various nAChR subtypes in Xenopus oocytes (e.g. α4β2, α2β2, α4α5β2 and α7) and showed that Abeta(1–42) blocks the binding of ligands to various non-α7 nAChRs, but not to α7 nAChRs. The block by Abeta(1–42) was dependent on the subunit makeup and stoichiometry of these receptors.

In addition, there is evidence that smokers have a reduced risk of developing AD [378] and preventive measures (including stimulants of nicotinic receptors) [379] have been suggested in the treatment of AD. To date, the acetylcholinesterase inhibitors have been the most widely used anti-AD drugs and have been partially successful in slowing loss of cognition [375].

Schizophrenia

Schizophrenia is a chronic psychotic complex disorder with a strong genetic predisposition. The exact molecular cause of any type of schizophrenia is still unknown. The incidence of smoking in schizophrenic patients is extremely high (80–90% versus 25–30% in the general population) [379]. Studies have shown that tobacco use transiently restores the schizophrenic patient's cognitive and sensory deficits, and cessation of smoking exacerbates the disease symptoms [380]. The α7 nAChR seems to participate directly in the pathophysiology of schizophrenic disturbances. Leonard et al. [381] reported a higher prevalence of functional promoter mutations in α7 nAChRs in schizophrenic subjects than in controls. Post-mortem binding studies have revealed a disturbance of nicotinic receptor expression affecting the α7 and α4β2 subunits in various cerebral areas [382].

Attention-deficit hyperactivity disorder

Attention-deficit hyperactivity disorder (ADHD) is a disease characterized by a persistent pattern of inattention and distractability and/or hyperactivity/impulsivity to such a degree that it impairs academic or occupational functioning. ADHD sufferers typically show evidence of this disorder in their childhood and ∼ 50% continue to demonstrate clinically significant symptoms into adulthood [383]. The incidence of cigarette smoking in individuals with ADHD has been found to be higher (∼ 40%) than in the general adult population (25–30%) [384]. Animal studies have shown that nAChR-related mechanisms are involved in attentional function [385]. The α4β2 and α7 nAChRs are critical for attention and working memory in rats [386,387]. Nicotine skin patches and a number of nAChR agonists improve clinical ADHD symptoms [388]. Wilens et al. [389] reported that the novel nicotinic cholinergic agent, ABT-418, a nicotinic agonist with selectivity for the α4β2 nAChR, may reduce impulsivity, hyperactivity and attentional deficits in adults with ADHD. Ueno et al. [390] used animal models of ADHD and showed that nicotine improves attention and memory in rats through activation of α4β2, but not α7, nAChRs.

Parkinson's disease

PD is a neurodegenerative movement disorder. In addition to the well-established loss of DAergic neurons, several studies have shown that nAChRs play a critical role in PD [391]. Epidemiological studies have shown that smoking protects against PD [392]. A number of α4β2 nAChR agonists have shown beneficial effects in PD [393]. Xie et al. [394] have shown that the neuroprotective effect of nicotine in PD is receptor independent and is due to its interaction with the mitochondrial respiratory chain and its antioxidant effects.

Pemphigus vulgaris

Pemphigus vulgaris is an autoimmune disease of keratinocytes in which the cells of the epidermis loose adherence (acantholysis), resulting in blistering of the skin or oral mucosa. An autoimmune response to α9 nAChRs has been identified [395]. Many neuronal-type nAChR subunits have been found at low levels in keratinocytes, where their responses to endogenous ACh and to nicotine influence cell adherence and motility in vivo and may influence development, wound healing and wrinkling in vivo[396]. Pemphigus patients occasionally develop MG and/or thymoma [397].

Autoimmune autonomic neuropathy

Autonomic neuropathies are inherited or acquired neuropathies in which the autonomic nerve fibers, both sympathetic and parasympathetic, are affected. The neuropathies can be autoimmune, idiopathic or due to diabetes, amyloidosis, drugs, etc. Autoimmune autonomic neuropathy (AAN) is sometimes associated with a neoplasm and the patients have high titres of antibodies to ganglionic nicotinic nAChRs [398]. Autoantibodies to α3 nAChRs have been found in ∼ 40% of patients with idiopathic or paraneoplastic dysautonomia [399,400]. An animal model of AAN has been developed and has shown the involvement of α3 nAChR in the pathogenesis of the disease. Rabbits immunized with the recombinant ECD of the human nAChR α3 subunit produced antibodies against nAChR and developed signs of experimental AAN [401]. The same group soon demonstrated that experimental AAN is an antibody-mediated disorder by documenting sympathetic, parasympathetic and enteric autonomic dysfunction in mice injected with rabbit IgG containing ganglionic nAChR antibodies. The autonomic signs were associated with reversible failure of nicotinic cholinergic synaptic transmission in the superior mesenteric ganglia. In addition, mice injected with IgG from two patients with AAN demonstrated a milder phenotype, with evidence of urinary retention and gastrointestinal dysmotility [402].

Hereditary epilepsies

Specific genes coding for ligand- and voltage-gated ion channels that are associated with hereditary epileptic phenotypes have been identified. Some rare idiopathic epilepsies are associated with mutations in genes coding for different neuronal-type nAChR subunits. Most mutations found to date are in the α4 subunit, the most abundant subunit in the CNS. Specifically, the identification of mutations in the α4 subunit in patients with human benign familial neonatal convulsions or autosomal dominant nocturnal frontal lobe epilepsy raise the possibility that the observed gene defects are causally linked with these two diseases or, alternatively, that α4 nAChR mutants increase the probability of epileptic discharges [403].

Autism

Autism is a developmental disorder associated with structural abnormalities of the brain. Cerebellar abnormalities have been identified by neuroimaging or neuropathology. The cholinergic neurotransmitter system has been implicated on the basis of nAChR loss in the cerebral cortex. Numbers of the α3 and α4β2 nAChRs have been found to be significantly reduced in autistic subjects compared with controls [404]. An increase in α7 subunits has also been observed [405].

Smoking addiction

Smoking is a major public health problem and the α4β2 and α7 subtypes of nAChRs, which are the most abundant subtypes in the brain, are closely associated with nicotine addiction and nicotine-induced behaviors [406]. α4β2 nAChRs have the highest sensitivity to nicotine and repeated nicotine exposure increases the functional nicotinic receptors in the brain. Functional upregulation of α4β2 nAChRs, observed in the brains of both smokers and animals chronically exposed to nicotine, is combined with a sensitization of the mesolimbic dopamine response to nicotine [407]. This response appears to be associated with the overall addictive properties of nicotine (but also of other drugs of abuse). The α7 nAChRs are also overexpressed in small cell lung carcinoma of smokers [408]; in this case, in vitro experiments have suggested that the malignant growth can be ceased using α-neurotoxins or α-conotoxins, by blocking these receptors [409]. Nicotine itself, in a number of commercially available forms (nicotine gum, transdermal patch, nicotine nasal spray, nicotine inhaler, etc.) is effective as part of a strategy to promote smoking cessation. However, compounds that could act as α4β2 agonists offer a more promising approach. Existing treatments have demonstrated only moderate efficacy in assisting smokers to quit. Varenicline, recently approved by the US FDA as an aid to smoking cessation treatment, has a novel mechanism of action, targeting α4β2 nAChR [410]. It has both agonistic and antagonistic properties that together are believed to account for the reduction of craving and withdrawal as well as blocking the rewarding effects of smoking. Its targeted mechanism of action, superior efficacy and excellent tolerability make varenicline a welcome and useful addition to the therapeutic options for smoking cessation.

Future perspectives

For several decades, the nAChR has served as the prototypic molecule for neurotransmitter receptors. In this review, we discuss its localization, structure, function and pathogenicity, topics which have been investigated, and many clarified, in the last half-century. Although an enormous amount of data has been accumulated, several fundamental questions remain unanswered. These include the exact channel function, the exact role of the different nAChR subtypes in different locations, the identification of the ‘key’ nAChR subtypes in the various diseases in which they are involved, the discovery or design of subtype-specific ligands, and the atomic structure of the whole molecule. The solution of the 3D structure (especially that of human neuronal nAChR subtypes) is a major challenge, which is, however, necessary for the development of novel therapeutics. The efforts currently being invested make us optimistic that considerable progress in this field will be made in the near future. The knowledge obtained will permit in-depth understanding of the mechanisms of channel activation and function and the design of selective ligands for therapeutic purposes.

Acknowledgements

Original studies in the authors' laboratories described in this review have been supported by grants from the European Commission, the MDA of USA and the Greek GSRT.