Introduction

An enduring goal is to understand how agonists activate their receptors1. For ligand-gated, or ‘agonist-activated’, ion channels, such as the nicotinic acetylcholine receptor (AChR), agonist binding culminates in activation/opening of an intrinsic ion-conducting pore2. Early models postulated that agonists are effective at opening the channel because they have a higher affinity for the open state than the closed state, and thus their binding shifts the equilibrium towards the open state3. In a landmark study, del Castillo and Katz proposed that agonism could be explained by a two-step process, where the agonist first binds to an inactive receptor, that then isomerises to an active/open conformation4. Within this framework, the ability of a given agonist to catalyse the isomerization step, and open the channel, was encoded in the chemical structure of the agonist, as well as its interactions with the receptor-binding site. Full vs. partial agonists differed in their ability to catalyse the isomerization step, with full agonists being more effective than partial agonists5.

The advent of the patch clamp technique and the development of single-channel analysis methods have made detailed investigations of the activation mechanism possible5,6,7. Early single-channel experiments were consistent with del Castillo and Katz’s two-step mechanism, however, to account for the two agonist sites on the AChR, the original scheme had to be extended to include two agonist-binding steps instead of one8. The emerging single-channel data also revealed additional complexity, including a class of brief closings, originally called ‘nachschlag shuttings’, which interrupted bursts of agonist-activated openings9. In some cases, the duration of these closings appeared to be agonist-independent, necessitating the introduction of an additional closed state in a more intricate scheme10. These observations, combined with studies of the related glycine receptor11 and continued technical improvements12, ultimately led to more elaborate mechanisms that incorporated intermediate closed states, termed ‘flipped’ or ‘primed’, which preceded channel opening3,13. These findings placed the origins of agonism before channel opening in the overall activation process3, and demonstrated that the ultimate opening and closing rates of the AChR were independent of the agonist used to elicit them; a somewhat counterintuitive finding for a channel ‘activated’ by agonist.

Despite the explanatory power of intermediate closed states, they have been inferred solely from kinetic analysis of single-channel measurements, and thus they remain largely phenomenological14. What is ‘flipping’? What is ‘priming’? To gain further insight we have taken advantage of a reconstructed ancestral muscle-type AChR β-subunit, called ‘βAnc’, which at the amino acid level shares more than 70% sequence identity with the human β-subunit15. In addition to forming hybrid ancestral/human AChRs, where βAnc replaces both the human β- and δ-subunits16, βAnc also forms homopentamers17. These βAnc homopentamers ‘prime’ before opening spontaneously and, reminiscent of the wild-type AChR, display steady-state burst behaviour with nachschlag-like shuttings (Fig. 1a)17. For βAnc to occupy both the β- and δ-subunit positions in hybrid AChRs, the (+) and (–) interfaces of βAnc must not only be compatible with each other, but also with the corresponding interfaces of their bracketing α-subunits (Fig. 1a). This suggests that heteropentamers composed of βAnc and the human α-subunit should be possible (Fig. 1b). If so, this would provide an opportunity to probe the function of the α-subunit, and specifically its role in agonism, in a unique context.

Fig. 1: Subunit composition and single-channel activity of wild-type and βAnc-containing acetylcholine receptors.
figure 1

a Subunit stoichiometry and arrangement of the human adult muscle-type acetylcholine receptor (muscle-type AChR; top), where the agonist-binding sites at the α-δ and α-ε subunit interfaces are indicated with asterisks (*). A reconstructed ancestral β-subunit (βAnc; purple) forms hybrid acetylcholine receptors (βAnc hybrid AChR; middle) where βAnc substitutes for the human β-subunit (β; orange) and supplants the human δ-subunit (δ; green). For this to happen, the principal (+) and complementary (–) interfaces of βAnc must be compatible with each other, as well as with the corresponding interfaces of their bracketing α-subunits (red highlights). βAnc also forms homomers (bottom; boxed), which open spontaneously (no agonist), and whose single-channel activity in the presence of agonist mirrors that of the muscle-type acetylcholine receptor. Recordings were obtained in the cell-attached patch configuration with a constant applied potential of –120 mV and filtered with a 5 kHz digital Gaussian filter. Unless otherwise indicated, all recordings were acquired in the presence of 60 μM acetylcholine, where openings represent inward cation currents, and are shown as upward deflections. The scale bar (20 ms, 10 pA) aligned to the top trace applies to all traces. b The compatibility of βAnc with bracketing α-subunits in hybrid AChRs predicts that α/βAnc heteromers are viable.

Full size image

Here, we show that α/βAnc heteromers are indeed viable, and that incorporation of the human α-subunit leads to an apparent α-subunit-dependent repression of spontaneous activity, which can be relieved by the addition of agonist. These findings provide insight into the origin of intermediate states that precede ligand-gated ion channel opening, and demonstrate how ‘agonist activation’ may instead be the apparent manifestation of agonist-induced derepression.

Results

The human α-subunit coassembles with βAnc

We previously showed that βAnc forms spontaneously opening homopentamers that at the single-channel level share hallmarks of human muscle-type AChR activity, including steady-state burst behaviour with nachschlag-like shuttings17. At concentrations of acetylcholine approaching saturation of the adult human AChR, bursts of single-channel activity from the two types of channels are essentially indistinguishable (Fig. 1a). This reflects both the high open probability of bursts of βAnc homopentamer activity that is independent of agonist, as well as the similar extent of open-channel block by acetylcholine in the two types of channels17.

To determine if the human muscle-type α-subunit can coassemble with βAnc to form α/βAnc heteromers, we measured cell-surface binding of radiolabeled α-bungarotoxin (α-Btx), an AChR competitive antagonist that can bind exclusively to determinants on the α-subunit18. When cells were cotransfected with cDNAs encoding the four human muscle-type AChR subunits, robust cell-surface binding of α-Btx was detected (Fig. 2a). By contrast, cells transfected with cDNA encoding only the human α-subunit displayed little or no binding of α-Btx, indicating that on its own, the α-subunit is incapable of forming α-Btx-binding sites that express on the cell surface. Similarly, when cells were transfected with cDNA encoding only βAnc, essentially no binding of α-Btx was detected. Given that single-channel activity of βAnc homopentamers is readily observed (Fig. 2c, top trace)17, the simplest interpretation is that despite their robust cell-surface expression, βAnc homopentamers do not bind α-Btx. Evidently, as with its extant β-subunit counterparts, βAnc lacks essential determinants of α-Btx binding. When cDNAs encoding βAnc and the human α-subunit are transfected together, robust cell-surface binding of α-Btx is restored, demonstrating that βAnc can shepherd the α-subunits and their associated α-Btx-binding sites to the cell surface, presumably by incorporating them into channels containing both types of subunits (Fig. 2a).

Fig. 2: The human α-subunit and βAnc can coassemble to form viable heteromeric channels that open spontaneously.
figure 2

a Relative cell-surface binding of [125I]-α-bungarotoxin (α-Btx) to cells transfected with the full complement of cDNAs encoding the human adult muscle-type acetylcholine receptor subunits (WT; white), the α-subunit alone (α; darkest grey), βAnc alone (βAnc; light grey), or a 1:1 (by weight) mixture of the α-subunit and βAnc (1:1; intermediate grey). Bar graphs represent the mean of two or three replicates (shown) from independent transfections, normalised to WT. b Frequency of spontaneous bursts of openings from cells transfected with cDNA encoding βAnc alone (0:1; light grey), or with increasing amounts of human α-subunit cDNA (1:1, 10:1; by weight; intermediate and dark grey, respectively), while the amount of βAnc cDNA is kept constant. Bar graphs represent the mean burst frequency averaged across ten separate single-channel patches (individual data points shown in each case), from at least three independent transfections. Difference between mean burst frequency is statistically significant (asterisks), as determined by one-way ANOVA (Tukey’s multiple comparison test; α level of 0.05; 0:1 vs. 1:1, p = 0.0015; 1:1 vs. 10:1, p = 0.0247). c Single-channel burst behaviour of patches from cells transfected with different ratios (0:1, 1:1, 10:1; by weight) of α-subunit and βAnc cDNA. In the absence of agonist, recordings were obtained in the cell-attached patch configuration with an applied potential of –120 mV and filtered with a 5 kHz digital Gaussian filter. In each case, openings are upward deflections, with the scale bar (2 s, 10 pA) beside the bottom trace applying to all zoomed out traces, and the inset boxes themselves representing 40 ms and 25 pA. d Burst duration histograms (see Methods) were manually fit (solid line) with a minimum sum of exponential components (dashed lines; labelled i and ii) containing the types of openings shown inset in c. Note that the dashed lines in the top panel are hidden by the solid line, as there is only a single exponential component that overlaps perfectly with the overall fit.

Full size image

Incorporation of α-subunits reduces spontaneous activity

To assess the functional consequences of replacing one or more βAnc subunit(s) in βAnc homopentamers with the human α-subunit, we cotransfected cells with cDNAs encoding both types of subunits and then examined the single-channel activity from cell-attached patches in the absence of agonist (Fig. 2b, c). Addition of α-subunit cDNA to a transfection mixture containing the same amount of βAnc cDNA resulted in an overall reduction in the spontaneous activity of patches. The extent of this reduction related to the amount of α-subunit cDNA added, and was best quantified as an α-subunit-dependent decrease in the frequency of spontaneous bursts (Fig. 2b).

Co-transfection with α-subunit cDNA also led to changes in the burst behaviour of the resulting channels (Fig. 2c, d). In the absence of the α-subunit, bursts were uniformly long, with a mean duration of ~10 ms (Fig. 2c, d; top). When cells were cotransfected with α:βAnc cDNAs at a ratio of 1:1 (by weight), the introduction of the α-subunit led to a second class of briefer events, bracketed by relatively long shut periods, which appeared as isolated openings (Fig. 2c; middle, inset ii) and contrasted with the bursts of closely spaced openings observed with βAnc homopentamers (Fig. 2c; top & middle, inset i). The two types of events were clearly distinguished in burst duration histograms (Fig. 2d; middle), where individual openings separated by closings briefer than a specified critical closed duration (τcrit; 2 ms throughout) have been combined into bursts19. The heterogeneous burst behaviour of channels in these patches suggests that two types of channels, βAnc homopentamers and α/βAnc heteropentamers, are present. Consistent with this hypothesis, increasing the ratio of cotransfected α:βAnc to 10:1 leads almost exclusively to isolated brief openings, essentially eliminating long bursts, and presumably βAnc homopentamers. Evidently, cotransfecting the human α-subunit with βAnc leads to a reduction in both the frequency and duration of spontaneous openings, and thus an apparent repression of the spontaneous activity.

Reduction in spontaneous activity depends on the number of α-subunits

The above experiments show that the spontaneous single-channel activity of patches containing both βAnc and the human α-subunit is reduced relative to those containing only βAnc homopentamers, and thus that the presence of the α-subunits appears to repress spontaneous activity. Given that the α-subunit only traffics to the cell surface in the presence of βAnc, a tacit assumption is that the reduced spontaneous activity is a result of the α-subunit replacing one or more βAnc subunits in α/βAnc heteromers. To determine how many α-subunits are present in spontaneously opening α/βAnc heteromers we used an electrical fingerprinting strategy, where the α-subunit was tagged with mutations that reduce its contribution to overall single-channel amplitude (Supplementary Figs. 1 and  2). A similar strategy has been employed with tetrameric potassium channels20, and pentameric ligand-gated ion channels21,22,23,24,25, including the muscle-type AChR16. Incorporation of one or more mutant low conductance α-subunits (αLC) would be expected to lead to a progressive decrease in the single-channel amplitude of resulting αLC/βAnc heteromers, thereby allowing us to directly register the number of incorporated αLC-subunits in individual channels.

Using the wild-type AChR background, we confirmed that channels incorporating αLC had a lower single-channel amplitude, but maintained a similar kinetic profile (Supplementary Fig. 2). This control demonstrated that the conductance altering mutations did not measurably affect other properties of the α-subunit. We then transfected cells with different ratios of αLC and βAnc cDNAs, and measured the distribution of single-channel amplitudes in patches from cells expressing the two types of subunits (Fig. 3). When cells were transfected with cDNAs encoding αLC and βAnc, a variety of single-channel amplitudes were observed in each patch (Fig. 3a, b). As expected, the relative proportion of high and low amplitude classes was dependent upon the ratio of transfected αLC to βAnc cDNAs (Supplementary Fig. 3). In mixtures of αLC and βAnc, a maximum of only three amplitude classes were detected (Fig. 3c), where the mean amplitude of the highest amplitude class matched that of βAnc homopentamers (Fig. 3c; inset 2, top). Thus, the simplest interpretation is that the intermediate and lowest amplitude classes originated from channels that have incorporated either one or two αLC subunits, respectively. If channels containing more than two αLC-subunits were present in these patches, they did not open spontaneously, and thus were not evident in these fingerprinting experiments.

Fig. 3: Electrical fingerprinting reveals the contribution of α-subunits to the apparent repression of spontaneous single-channel activity.
figure 3

a Co-transfection of a cDNA encoding the human α-subunit harbouring reporter mutations that reduce single-channel conductance (αLC) with cDNA encoding βAnc leads to reduced spontaneous activity, where b the amplitude and duration of single-channel events is variable. Indicated openings in (a; i–v) are shown expanded in b, with sight-lines overlaid to indicate the amplitudes of individual events. Openings are upward deflections, where the scale bars represent 1 s and 10 pA in a, and 10 ms and 10 pA in b. c Combining single-channel bursts from multiple recordings (see also Fig. S3) where cells were transfected at different cDNA ratios (by weight) shows how single-channel amplitudes segregate into three well-defined amplitude classes (overlaid gaussians; solid lines), where each amplitude class corresponds to openings from channels incorporating either 0, 1, or 2 αLC-subunits. Plotting the amplitude of individual bursts as a function of their duration reveals an apparent correlation (inset 1). d Dwell time analysis of selected bursts whose amplitude is within 1.5 standard deviations of the mean of each amplitude class (shaded regions from inset 2 in c) reveals the contribution of each successive αLC-subunit (dark grey subunits in schematics) to open (left) and burst (middle) durations, as well as in determining the number of successive openings occurring in each burst (right). The traces under the schematics (left) show openings (upward deflections) representative of each exponential component from visual fits of their corresponding duration histograms (right). Scale bar beside the bottom trace in d represents 5 ms and 10 pA. Recordings were obtained in the cell-attached patch configuration with an applied potential of –120 mV and filtered with a 5 kHz digital Gaussian filter.

Full size image

To assess the impact of each α-subunit on the spontaneous activity of βAnc-containing channels we first plotted an amplitude vs. duration scatter plot, where each data point represented the amplitude and duration of individual bursts in the fingerprinting experiments (Fig. 3c, inset 1). This analysis showed an apparent correlation between the amplitude and duration of bursts, with lower amplitude bursts being, on average, briefer in duration. This suggested that the incorporation of each successive αLC subunit led to a progressive decrease in burst duration. To quantify this effect, we collected openings whose amplitude was within 1.5 standard deviations of the mean amplitude of each amplitude class (Fig. 3c, inset 2), and then performed dwell time analysis on each set corresponding to spontaneous activity from channels with zero, one, or two α-subunits (Fig. 3d). Consistent with previous results17, spontaneous openings of βAnc homopentamers were best fit by a single exponential component, and occurred in quick succession, ultimately coalescing into bursts that had a mean duration of ~10 ms. In contrast, channels containing one or two α-subunits exhibited two classes of spontaneous openings, both of which were briefer than the single class of openings in βAnc homopentamers. When one α-subunit was present the longer of these two components was predominant, whereas when a second α-subunit was present the briefest component was predominant. Thus, with incorporation of each successive α-subunit the average duration of openings became briefer and briefer. In addition, when either one or two α-subunits were present, openings appeared as isolated events. This altered burst behaviour was quantified by binning bursts based upon the number of openings they contained, and showed a steep α-subunit-dependent drop off in the number of openings occurring within each burst (Fig. 3d; far right).

Agonist relieves apparent repression of spontaneous activity

From the perspective of βAnc homopentamers, the above data demonstrate that replacing one or more βAnc-subunit(s) with the human α-subunit leads to a reduction in spontaneous single-channel activity. Given that in wild-type AChRs, residues from the α-subunits form the principal (+) face of the agonist-binding sites, we wondered whether this apparent α-subunit-dependent repression could be relieved by the addition of agonist. We began by once again transfecting cells with various ratios of α:βAnc cDNAs, and then recorded spontaneous single-channel activity in the absence of agonist. To assess the impact of agonist, we identified cells that had a low, but quantifiable, spontaneous activity in the absence of agonist, and then returned to the same cell to record additional data from a new cell-attached patch, but this time in the presence of 300 μM acetylcholine. We reasoned that this high concentration of acetylcholine might be required to elicit a response given that α/βAnc heteromers, with βAnc providing the complementary (–) face of the agonist-binding site, lack important residues involved in agonist recognition16. Furthermore, by returning to the same cell to acquire paired recordings, first in the absence, and then in the presence of acetylcholine, we were able to limit patch-to-patch variability by controlling for the individual expression level of each transfected cell. The same strategy has been used to quantify calcium potentiation of the human α7 acetylcholine receptor26. While extensive open-channel block resulting from the high agonist concentration precludes a meaningful comparison of dwell times (Supplementary Fig. 4), we were able to assess the effect of agonist on burst frequency (Fig. 4). For βAnc homopentamers (0:1; α:βAnc), although there was some variability in the frequency of bursts plus or minus agonist in paired patches from the same cell, the overall frequency of bursts was not significantly different in the presence or absence of agonist. By contrast, in cells transfected with a 10:1 ratio of α:βAnc cDNAs (see also 1:1 in Supplementary Fig. 5), paired recordings from the same cell revealed that, in each case, the presence of agonist led to a marked increase in the frequency of bursts. On average, the increase in burst frequency was more than 10-fold, revealing that agonist was able to relieve this aspect of the apparent repression imparted by the α-subunits in α/βAnc heteromers (Fig. 4c; Supplementary Fig. 5).

Fig. 4: Agonist relieves apparent repression of α/βAnc heteromers.
figure 4

Single-channel activity from cells expressing a βAnc homopentamers (0:1; α:βAnc cDNA; by weight) and b α/βAnc heteromers (10:1) in the absence (–) and presence (+) of 300 μM acetylcholine. Recordings were obtained in the cell-attached patch configuration with an applied potential of –120 mV and Gaussian filter of 5 kHz. Single-channel openings represent inward cation currents, and are shown as upward deflections. The scale bar beside the bottom trace in b applies to all traces and represents 1 s and 10 pA. c Comparison of burst frequency in paired recordings from cells expressing either βAnc homopentamers (0:1), or α/βAnc heteromers (10:1). In each case, paired cell-attached recordings from the same cell were acquired first in the absence (–), and then in the presence (+), of 300 μM acetylcholine in the patch pipette. Box plots represent one standard deviation from the mean, with the internal horizontal line denoting the mean of the 10 recordings in each case. Maximum and minimum values are presented as box plot whiskers. As determined by a two-way ANOVA, the difference between mean burst frequency –/+ 300 μM acetylcholine is statistically significant (α level of 0.05) for the α/βAnc heteromers (10:1, p = 0.0011; asterisk), but not the βAnc homopentamers (0:1, p = 0.7189; ns).

Full size image

Discussion

Previously we showed that βAnc readily forms homopentameric channels that open spontaneously17. Here, we have shown that the human muscle-type α-subunit can coassemble with βAnc to form α/βAnc heteromers (Fig. 2a). Co-transfection of the α-subunit with βAnc led to a reduction in the frequency of spontaneous openings in cell-attached patches (Fig. 2b), as well as alterations in the single-channel behaviour consistent with inhibition of the spontaneous activity (Fig. 2c). In both cases, the degree of inhibition was proportional to the amount of α-subunit cotransfected (Fig. 2). Tagging the α-subunits with conductance/reporter mutations allowed us to determine that α/βAnc heteromers containing one or two α-subunits are viable, and open spontaneously (Fig. 3). Dwell time comparison of channels containing zero, one, or two α-subunits revealed that incorporation of each successive α-subunit progressively decreased the duration of both individual spontaneous openings, as well as bursts of closely spaced spontaneous openings (Fig. 3). Burst duration was further decreased through an α-subunit-dependent reduction in the mean number of openings occurring within each burst (Fig. 3). Finally, when both βAnc and the α-subunit were cotransfected, the presence of agonist in cell-attached patches led to a greater than 10-fold increase in the single-channel activity (Fig. 4).

What do these experiments involving unnatural, heterologously expressed, spontaneously opening channels based upon a reconstructed ancestral muscle-type acetylcholine receptor β-subunit tell us about pentameric ligand-gated ion channels found in nature today? In particular, our experiments provide possible insight into the function of the α-subunits in muscle-type AChRs. Relative to βAnc homopentamers, the reduced spontaneous activity of α/βAnc heteromers depends steeply upon the number of incorporated α-subunits, and thus it appears as though the α-subunits have the capacity to repress intrinsic spontaneous activity. However, a similar outcome would be expected if each βAnc subunit imparted an activating presence that is lost in α/βAnc heteromers. Determining the relative contribution of these two possibilities is not trivial, yet the implications of a possible repressive role for the α-subunits are profound. Usually viewed as the principal subunits contributing to agonist binding and activation27, we’ve shown that the α-subunits may also have the capacity to repress intrinsic spontaneous activity. While this was demonstrated in an unnatural channel, the incorporated α-subunits were wild-type human α-subunits, raising the possibility that the α-subunits perform the same function in their natural setting, and repress intrinsic spontaneous activity in the heteromeric muscle-type AChR.

From the classic pharmacological perspective, ‘activation’ refers to the transition of a receptor from an inactive to an active state, following application of agonist28. Implied within this definition is that the agonist directly drives the transition of the receptor from an inactive to an active state. For ion channels, their open/ion-conducting conformation is their active state, and thus ion channel activation is the opening of the ion channel driven by an external stimulus. For ligand-gated (or ‘agonist-activated’) ion channels, such as the AChR, agonists are expected to directly drive the transition from an inactive/closed state to an active/open state2. This is precisely why the ‘flip’ and ‘prime’ models of agonist activation are surprising. In both models, the ultimate opening and closing rates of the channel are independent of the agonist, begging the question: are agonists bona fide agonists if they do not directly activate the channel?

A repressive role for the α-subunits offers a way to reconcile these counterintuitive findings, and provides insight into phenomenological ‘flipped’ and ‘primed’ states. Rather than activating the channel, agonists could instead derepress it. In this framework, the α-subunits repress spontaneous/basal activity, and agonists remove this repressive influence leading to an ‘apparent activation’, which is in fact the restoration of intrinsic basal activity. To derepress the channel, agonists drive the transition of the channel from a repressed/closed state where openings are unlikely (e.g., ‘unflipped’ or ‘unprimed’), to a derepressed/closed state where openings are more likely (e.g., ‘flipped’ or ‘primed’). An important distinction between derepression and activation, is that in the derepression paradigm, agonists do not drive the transition from an inactive/closed state to an active/open state, but instead drive a transition between two distinct closed states that are both inactive. From this perspective, full vs. partial agonists differ in their ability to derepress the channel, but once derepressed, the channel opens and closes in the same way irrespective of the efficacy of the bound agonist. These predictions that (1) agonists drive a transition between two inactive/closed states, and (2) channel opening is independent of agonist efficacy, are both realised in the ‘flip’ and ‘prime’ models3,13.

Mirroring muscle-type AChRs that contain two α-subunits, our fingerprinting experiments detected a maximum of two α-subunits in spontaneously opening α/βAnc heteromers. While we cannot exclude the possibility that three (or more) α-subunits can exist in α/βAnc heteromers, for this to happen, at least two α-subunits would have to occupy neighbouring positions in the resulting heteropentamer. This would require that the (+) and (–) interfaces of the α-subunit be compatible with each other, thereby allowing multiple α-subunits to self-associate. There is no evidence that the human muscle-type α-subunit can self-associate, and consistent with this, we did not detect cell-surface binding of radiolabeled α-Btx when the α-subunit was transfected alone. The simplest interpretation is that α/βAnc heteromers contain at most two α-subunits, where again mirroring muscle-type AChRs, the two α-subunits are separated by at least one intervening subunit (in this case a βAnc subunit). Notably, the presence of two α-subunits in muscle-type AChRs is enough to ensure that spontaneous openings are infrequent29. Similarly, two α-subunits in α/βAnc heteromers is enough to reduce spontaneous activity, such that spontaneous openings occur infrequently in comparison to βAnc homopentamers.

The possibility that the α-subunits hold the channel shut, as opposed to being responsible for opening it, also has potential implications for how agonism evolved in this family of proteins. Given that several aspects of wild-type AChR activation are independent of agonist and preserved in βAnc homopentamers, it seems plausible that spontaneous channel activity existed before its regulation by agonist. In such a scenario, spontaneously opening channels evolved a subsequent layer of regulation through accumulating mutations that repressed their constitutive activity, which could then be restored by the presence of bound agonist. Essentially, constitutive/basal activity is inhibited by a repressor protein, in this case the α-subunits, which is then derepressed by the binding of an inducer, in this case an agonist (Fig. 5b). This is consistent with the observation that most AChR mutations alter the intrinsic tendency of the protein to open without changing the energetic contribution from bound agonist30. If regulation by agonist evolved to control pre-existing basal activity, then pentameric channels with significant basal activity and biologically relevant functions should still be found in nature today, and indeed related GABAA receptors containing the β3 subunit display spontaneous/basal activity that is thought to control neuronal excitability31.

Fig. 5: Parallels between induction of gene expression and agonism in a pentameric ligand-gated ion channel.
figure 5

a Constitutive expression of a gene (top) can be repressed by the binding of a repressor (yellow protein) to an operator sequence (middle), which can then be derepressed by the binding of an inducer (orange circle) to the repressor (bottom). b Constitutive activity of a pentameric ion channel (top; purple βAnc) can be repressed by incorporation of a repressor protein (middle; yellow α-subunits), which can then be derepressed by the binding of an inducer (bottom; orange agonist).

Full size image

In famous work involving the lac operon, Jacques Monod and François Jacob showed that gene expression can be controlled by a repressor protein that binds to DNA and inhibits constitutive expression of a target gene (Fig. 5a). Expression of the target gene could then be induced by inhibiting the repressor protein. In Monod’s words, ‘the inducer acts not by provoking’ target gene expression, but instead ‘by inhibiting an inhibitor’ of its expression32. Similarly, we have shown that rather than ‘provoking’ AChR openings, agonists may instead ‘inhibit the inhibition’ of constitutive activity, and thus ‘activation’ of the AChR, the prototypical ligand-gated ion channel33, may be the apparent manifestation of an analogous derepression mechanism (Fig. 5b).

Methods

Molecular biology

cDNA for the human AChR α1-subunit cloned into the pRBG4 plasmid was provided by Steven M. Sine, while cDNA encoding the ancestral β1 subunit (βAnc) was ordered as a synthetic oligonucleotide and cloned into the same pRBG4 plasmid15. Point mutations substituting three arginine residues present in the homologous human 5-HT3A subunit34 (Supplementary Fig. 1) into the human α1 subunit to produce the low conductance variant (αLC; αD389R/N393R/A397R; mapped onto Protein Data Bank ID: 7QKO35 in Supplementary Fig. 1) were introduced into cDNAs by inverse PCR36. Sanger sequencing confirmed the sequence of the entire reading frame.

Mammalian cell expression

cDNAs encoding human and ancestral subunits were transfected into BOSC 23 cells37 (CVCL_4401; ATCC number: CRL-11270). Cells were maintained in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% (by volume) fetal bovine serum at 37 °C, to a confluency of 50 to 70%. Calcium phosphate transfections were carried out for 3–4 h and terminated by media exchange. Experiments were performed 16–24 h after transfection. A separate plasmid encoding green fluorescent protein was included in all transfections to facilitate identification of transfected cells.

Cell line authentication and mycoplasma testing

Approximately five million confluent cells were harvested and their total DNA isolated (E.Z.N.A.® Tissue DNA Kit), and then submitted to The Centre for Applied Genomics Genetic Analysis Facility (The Hospital for Sick Children, Toronto, Canada) for STR profiling using Promega’s GenePrint® 24 System. A similarity search on the 8159 human cell lines with STR profiles in Cellosaurus release 42.0 was conducted on the resulting STR profile, which revealed that the cell line shares closest identity (88%, CLASTR 1.4.4 STR Similarity Search Tool score) with Anjou 65 (CVCL_3645). Anjou 65 is a child of CVCL_1926 (HEK293T/17) and is itself a parent line of CVCL_X852 (Bartlett 96). Bartlett 96 is the parent line of BOSC 2337. PCR tests confirmed that the cells were free from detectable mycoplasma contamination38.

Single-channel patch clamp recordings

Single-channel recordings from BOSC 23 cells transiently transfected with cDNAs encoding wild-type, ancestral, or low conductance subunits, were obtained in the cell-attached configuration with a membrane potential of –120 mV and a temperature maintained between 19 and 22 °C. The bath solution contained (in mM) 142 KCl, 5.4 NaCl, 0.2 CaCl2 and 10 HEPES (4-(2-hydroxyethyl)−1-piperazineethanesulfonic acid), adjusted to pH 7.40 with KOH. The pipette solution contained (in mM) 80 KF, 20 KCl, 40 K•aspartate, 2 MgCl2, 1 EGTA (ethylene glycol-bis(β-aminoethyl ether)-N,N,N′,N′-tetraacetic acid), and 10 HEPES, adjusted to a pH of 7.40 with KOH. Acetylcholine chloride (Sigma) was added to pipette solutions to the desired final concentration and stored at –80 °C. Patch pipettes were fabricated from type 7052 or 8250 nonfilamented glass (King Precision Glass) with inner and outer diameters of 1.15 and 1.65 mm, respectively, and coated with SYLGARD 184 (Dow Corning). Prior to recording, electrodes were heat polished to yield a resistance of 5 to 8 MΩ. Single-channel currents were recorded using an Axopatch 200B patch clamp amplifier (Molecular Devices), with a gain of 100 mV/pA and an internal Bessel filter at 100 kHz. Data were sampled at 1.0 μs intervals using a BNC-2090 A/D converter with a National Instruments PCI 6111e acquisition card and recorded by the program Acquire (v. 6.0.0; Bruxton).

Single-channel analysis

Single-channel event detection was performed using the program TAC (v. 4.3.3; Bruxton), where data were analysed with an applied 5 kHz digital Gaussian filter. Opening and closing transitions were detected using the 50% threshold crossing criterion, and corrected for instrument rise time using the R package scbursts (v. 1.6)39. Open and burst durations were placed into logarithmic bins40, and the sum of exponentials was manually fit to the open and burst duration distributions within the program TACfit (v. 4.3.3; Bruxton). Bursts of closely spaced openings were defined by a critical closed duration (τcrit), where individual openings separated by closings briefer than τcrit (2 ms throughout) were concatenated with their intervening closings to produce individual bursts19. Duration histograms were constructed using a universal 2 ms burst resolution, and by omitting bursts briefer than 100 μs. Supplementary Fig. 2 analysis includes burst open probability, which was determined within R (v. 4.0.4) using the packages scburst (v. 1.6)39, extremevalues (v. 2.3.3)41, and MASS (v. 7.3-55)42.

Electrical fingerprinting

Cells were transfected as described above, but with an αLC to βAnc ratio of 10:1, 1:1, or 0:1 (αLC:βAnc; by weight). Data were digitally filtered to 5 kHz within TAC and amplitudes were determined by setting the baseline and open-channel current for individual openings by eye. For each ratio, ~2000 events were collected and pooled from between 9–12 patches. A critical closed duration of 2 ms was uniformly applied throughout to define bursts of single-channel activity. To ensure that only fully resolved amplitudes were included in the analysis, events briefer than 100 μs were omitted. Pooled events were binned and plotted in event-based amplitude histograms, which were then fit with a set of Gaussian distributions. Data within 1.5 standard deviations of the mean amplitude encompassing each Gaussian, and representing each amplitude class, were then isolated for downstream dwell time analysis. Open and burst duration histograms were generated from events in each amplitude class in TACFit, and burst sorting and analysis was performed within R using the packages scbursts (v. 1.6)39 and MASS (v. 7.3-55)42.

Paired recordings from the same cell

Paired recordings, in the absence (–) and then presence (+) of agonist, were obtained from successive cell-attached patches of the same cell. A 5 min continuous recording in the absence of agonist was obtained, and then a second 5 min recording from a new cell-attached patch, on the same cell, was acquired, but this time in the presence of 300 μM acetylcholine within the patch pipette. Data were filtered to 5 kHz, and all events within the 5 min window post voltage application were analysed. A critical closed duration and a burst resolution of 2 ms, was uniformly applied throughout. Paired data plots were generated in R with the packages ggplot2 (v. 3.3.5)43 and ggpubr (v. 0.4.0).

Radioligand-binding experiments

AChR cell-surface expression was measured by binding of [125I]-labelled α-bungarotoxin to transfected cells. Approximately 850,000 BOSC 23 cells were plated onto 6 cm dishes and transfected with α, βAnc, or both α and βAnc (1:1; by weight) cDNA’s. Cells were harvested one day post-transfection and incubated for 1 h at room temperature with [125I]-α-bungarotoxin (25 nM; specific activity of 10 Ci/mmol) in potassium Ringer’s solution. Ringer’s solution contained (in mM), 140 KCl, 5.4 NaCl, 1.8 CaCl2, 1.7 MgCl2, 25 HEPES, and 30 mg L−1 bovine serum albumin, adjusted to pH 7.40 with KOH25. Cells were deposited onto 25 mm Whatman GF/C microfiber filter discs using a Hoefer filtration manifold and washed three times with 5 mL of Ringer’s solution. Nonspecific binding of [125I]-α-bungarotoxin to the filter discs was minimised by pre-incubating each disc in Ringer’s solution containing 1% bovine serum albumin for 1 h. Bound toxin was counted for 2 min in a Wizard2 1-Detector Gamma Counter (Perkin-Elmer).

Statistics and reproducibility

For single-channel recordings, each replicate represents data acquired from the same cell-attached patch, where each patch was from a different cell. For the paired replicates in Fig. 4 and Supplementary Fig. 5, consecutive cell-attached patches, first in the absence and then in the presence of 300 μM acetylcholine, were acquired from the same cell. For radioligand-binding experiments, each replicate constitutes data acquired from separate transfections. Statistical tests were performed using GraphPad Prism (v. 8.0.0). One-way ANOVA (Fig. 2) was performed using Tukey’s multiple comparison test. For paired replicates (Fig. 4 and Supplementary Fig. 5), separate two-way ANOVA tests were performed for each cDNA ratio. All significant differences had adjusted p-values below the predetermined α level of 0.05.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.