Ancestral acetylcholine receptor β-subunit forms homopentamers that prime before opening spontaneously

Human adult muscle-type acetylcholine receptors are heteropentameric ion channels formed from two α-subunits, and one each of the β-, δ-, and ε-subunits. To form functional channels, the subunits must assemble with one another in a precise stoichiometry and arrangement. Despite being different, the four subunits share a common ancestor that is presumed to have formed homopentamers. The extent to which the properties of the modern-day receptor result from its subunit complexity is unknown. Here, we discover that a reconstructed ancestral muscle-type β-subunit can form homopentameric ion channels. These homopentamers open spontaneously and display single-channel hallmarks of muscle-type acetylcholine receptor activity. Our findings attest to the homopentameric origin of the muscle-type acetylcholine receptor, and demonstrate that signature features of its function are both independent of agonist and do not necessitate the complex heteropentameric architecture of the modern-day protein.


Results 143
Our first inkling that a reconstructed ancestral b-subunit (βAnc) may be able to form 144 homomeric channels came from heterogeneity in single-channel recordings acquired after 145 alterations to our original transfection protocol. Typical cotransfection of cDNAs encoding 146 human α-, d-, and e-subunits with a cDNA encoding βAnc leads to robust cell surface 147 expression of βAnc-containing hybrid AChRs (10). In an attempt to lower overall AChR 148 expression with the purpose of facilitating single-channel analysis, we reduced the total 149 amount of subunit cDNA in our transfections (~6-fold), while maintaining the same 2:1:1:1 150 subunit cDNA ratio (by weight; α:βAnc:d:e). Reducing the amount of cDNA lowered overall 151 AChR expression as expected, but also led to a previously unseen heterogeneity in our 152 patches ( Fig. 2A). Instead of a single population of channels with a uniform amplitude of ~10 153 pA, and a burst behavior indicative of βAnc-containing AChRs ( Fig. 2A; left inset), we also 154 observed a second class of channels with a different kinetic signature, and an increased 155 amplitude of ~16 pA ( Fig. 2A; right inset). A similar trend was not observed when cDNA 156 encoding the wild-type human b-subunit was cotransfected instead of βAnc (Fig. S1), 157 indicating that the ancestral b-subunit was the source of the heterogeneity. Consistent with 158 this, lowering the proportion of βAnc cDNA in the transfection mixture reduced the fraction 159 of high amplitude channels (Fig. 2B,D), while transfecting exclusively with βAnc cDNA 160 resulted in patches where all channel openings had a uniformly high amplitude (Fig. 2C,D). 161 This demonstrated that when transfected alone, βAnc forms functional ion channels. 162 The traces in Figure 2 were recorded in the presence of agonist (30 µM acetylcholine). 163 In the human adult muscle-type AChR, the agonist-binding sites are located at the α-d and α-164 e interfaces, and the b-subunit is the only subunit that does not participate directly in agonist 165 binding ( Fig. 1) (21). We were therefore surprised to see single-channel activity in patches 166 from cells transfected exclusively with βAnc, as channels formed from a muscle-type b-167 subunit alone would not be expected to have intact agonist-binding sites. To determine if the 168 activity of βAnc-alone channels was dependent upon acetylcholine, we recorded single-169 channel activity in the absence of acetylcholine (Fig. 3). When no acetylcholine was present, 170 patches from cells that were transfected exclusively with βAnc cDNA still displayed single-171 channel activity, indicating that βAnc-alone channels open spontaneously under these 172 conditions (Fig. 3A). Furthermore, spontaneous activity occurred as bursts of closely spaced 173 openings, separated by brief closings (Fig. 3B) (14). The briefest of these intervening closings 174 were reminiscent of classic "nachschlag shuttings" (Fig. 3B, "ii" in inset), observed in early 175 patch clamp recordings from frog end-plate nicotinic receptors, and originally thought to 176 relate to agonist efficacy (22). Thus, despite being homomeric and lacking agonist-binding 177 sites, βAnc-alone channels display single-channel hallmarks of the muscle-type AChR. 178 To gain insight in the spontaneous activity of βAnc-alone channels, we performed 179 kinetic analysis of our single-channel data. First, we determined a critical closed duration 180 (τcrit) to define bursts arising from a single ion channel. Then we determined the minimum 181 number of components in our apparent open and closed dwell duration histograms by fitting 182 each with a sum of exponentials. Open duration histograms were fit well by a single 183 exponential component, while closed duration histograms required at least two components 184 (Fig. 3C). This suggested that a minimal scheme with a single open state and two closed states 185 is necessary and sufficient to describe the spontaneous activity of βAnc-alone channels. Based 186 on this, we then fit the sequence of single-channel dwells using the three possible kinetic 187 schemes, two linear and one cyclic, comprising a single open state and two closed states (Fig.  188  3D; Fig. S2). As a control we also fit a simplified two-state scheme, where a single open state 189 was connected to a single closed state, which, based on the relatively poor fit of the closed 190 durations, confirmed that inclusion of a second closed state was justified (Fig. S2). Overlaying 191 the resulting fits on top of duration histograms revealed that each of the possible three-state 192 schemes fits the observed dwells equally well, thus discriminating between the possible 193 kinetic schemes is not trivial (Fig. S2). We settled upon the simple linear scheme, where βAnc-194 alone channels transition from a closed state, "C", to an intermediate closed state, "C′", before 195 opening to "O′". The form of this scheme, with an intermediate closed state that precedes 196 channel opening, is guided by models of AChR activation that include a single "flipping" or 197 multiple "priming" steps (13, 23, 24). Given that for βAnc-alone channels there is a single 198 intermediate closed state that precedes channel opening, we refer to this state in our scheme 199 as "flipped" or "primed". 200 In the presence of acetylcholine, the single-channel current traces appeared different 201 of βAnc-alone homomers with wild-type AChRs, we determined the kinetics of acetylcholine 213 and QX-222 block for both types of channels ( Fig. 4; Fig. S3; Fig. S4). To fit our βAnc-alone 214 single-channel data recorded in the presence of a blocker, we introduced an additional open, 215 but blocked (i.e. non-conducting) state connected to our open state, where the forward rate 216 of block was dependent upon the concentration of the blocking molecule (Fig. 4A,B). We then 217 globally fit each of our βAnc-alone data sets encompassing between 0 and 100 µM 218 acetylcholine or QX-222. Initially, we restricted the rates of the core (C-C′-O′) scheme to 219 those inferred in the absence of blocker, however allowing all parameters to be estimated, 220 led to negligible changes in the inferred rates of block. For βAnc-alone and wild-type channels, 221 the rates of acetylcholine and QX-222 block were comparable (see Tables S1,S2), suggesting 222 that the structure of the open pore in the two types of channels is similar. 223 Given that βAnc-alone channels are expressed in the absence of other AChR subunits, 224 a reasonable hypothesis is that they are homopentamers. To determine the subunit 225 stoichiometry of βAnc-alone channels, we employed a single-channel electrical fingerprinting 226 strategy, where mutations altering unitary conductance are used to count the number of 227 individual βAnc subunits in βAnc-alone channels. A similar strategy has been employed with 228 tetrameric potassium channels (29), and other pLGICs (30), including both the 229 homopentameric α7 AChR (31-33) and the heteropentameric muscle-type AChR (7). The 230 approach relies on identifying high-conductance (HC) and low-conductance (LC) variants of 231 the βAnc subunit, and then co-expressing them to reveal a number of amplitude classes. 232 Openings in each amplitude class originate from channels incorporating the same ratio of HC 233 to LC subunits, and based on the total number of amplitude classes, the number of βAnc 234 subunits within βAnc-alone channels can be inferred. 235 When βAnc is expressed alone, the resulting channels exhibit a single, uniform 236 amplitude, distributed around a mean of ~16 pA (Fig. 5A,D), making the wild-type βAnc 237 subunit an ideal high-conductance (HC) subunit for electrical fingerprinting. To identify a 238 low-conductance (LC) variant of βAnc, we took advantage of a structural feature inherent to 239 eukaryotic pLGICs: as conducting ions exit the channel's transmembrane pore, they are 240 obliged to pass through one of five portals in the cytoplasmic domain (21, 34). Framed by 241 charged or polar residues from each subunit, these portals influence single-channel 242 conductance. Mimicking the homologous 5-HT3A receptor, which has an unusually low 243 single-channel conductance (35-37), we substituted three arginine residues (E420R, D424R, 244 E428R) into this region of βAnc. When βAnc harbouring three arginines in this region was 245 transfected by itself, the resulting channels exhibited a reduced single-channel amplitude 246 centered around ~1-2 pA (Fig. 5B,D). With its markedly reduced amplitude, βAnc harbouring 247 three arginine residues is a suitable LC subunit for electrical fingerprinting. 248 When cDNAs encoding HC and LC variants of βAnc were transfected together, a variety 249 of single-channel amplitudes were observed in each patch (Fig. 5C). The relative proportion 250 of channels with high versus low amplitude could, to some degree, be tuned by the ratio of 251 HC to LC βAnc cDNA used for transfection (Fig. S5). Constructing event-based amplitude 252 histograms, and pooling amplitudes from more than one recording, revealed that the 253 amplitudes segregated into as many as six amplitude classes, with the highest and lowest 254 amplitude classes matching that of the HC and LC forms of βAnc-alone channels. The 255 difference in amplitude between successive classes was somewhat regular, demonstrating 256 five approximately equal contributions to single-channel conductance (Fig. 5D), consistent 257 with the hypothesis that βAnc-alone channels are homopentamers. 258 As noted previously, reconstruction of βAnc was based upon a molecular phylogeny 259 that diverged from the accepted species phylogeny (7-10). Reconstruction of ancestral 260 protein sequences is based upon a best-fit model of amino acid evolution, a multiple 261 sequence alignment, as well as a phylogenetic tree relating the sequences within the 262 alignment (38). We therefore wondered if the ability of βAnc to form spontaneously opening 263 homomers was an artefact of the discordant tree used to reconstruct it. To test this, we took 264 advantage of an alternate ancestral β-subunit, called "βAncS", whose reconstruction was based 265 upon a molecular phylogeny that matched the accepted species phylogeny (7). Despite 67 266 substitutions and 6 indels relative to βAnc, when expressed alone, βAncS still formed 267 homomeric channels that opened in bursts in the absence of acetylcholine (Fig. 6). This 268 demonstrated that the ability of βAnc to form homomers that spontaneously open is not an 269 artefact of the phylogeny used to reconstruct it. Instead, this surprising ability of βAnc is 270 robust to the phylogenetic uncertainties inherent in ancestral sequence reconstruction, as 271 well as substantial variation in the amino acid sequence of the reconstructed β-subunits. 272 While βAncS forms homomers that open spontaneously, inspection of the βAncS single-273 channel activity revealed additional complexity not seen with the original βAnc subunit. For 274 βAncS, single-channel bursts were heterogeneous, displaying at least three distinct kinetic 275 behaviours (Fig. 6B). In some cases, the kinetic behaviour changed within a burst, 276 demonstrating that the different kinetics were possible within the same channel ( Fig. 6B; 277 boxed). This heterogeneity was also reflected in apparent open and closed duration 278 histograms, with each displaying a minimum of three exponential components (Fig. 6C). The 279 increased number of exponential components indicated additional open and closed states 280 relative to βAnc, and thus that the three-state scheme used to fit βAnc was insufficient to 281 describe the spontaneous activity of βAncS. To account for the additional states, we expanded 282 our original scheme to include two additional "priming" steps, where openings could occur 283 from one of three primed states (Fig. 6D). This scheme, with multiple priming steps, builds 284 directly upon the one used to fit muscle-type AChRs that had been engineered to open 285 spontaneously (24). Given the heterogeneity of the βAncS single-channel activity, and the 286 complexity of this scheme, we caution against over interpretation of the inferred rates. 287 Nevertheless, we note that in accord with the muscle-type AChR, the equilibrium gating 288 constants appear to increase (Table S3; compare Θ1, Θ2, Θ3), and thus the open states become 289 more and more favoured, for each successive priming step. In any case, the fits suggest that 290 a scheme of this form, with multiple stages of priming, is adequate to describe the complex 291 spontaneous single-channel activity of βAncS homomers. 292

Discussion 293
We have shown that a reconstructed ancestral acetylcholine receptor β-subunit (βAnc) 294 readily forms homopentameric channels. At first glance, this may seem unexpected. Modern-295 day muscle-type β-subunits do not appear to form homopentamers, and instead appear fully 296 entrenched within heteropentameric muscle-type AChRs. Since reconstruction of βAnc was 297 informed almost exclusively by modern muscle-type β-subunits, there was no reason to 298 expect that βAnc would behave differently (10). However, as mentioned previously, βAnc can 299 replace both the human muscle-type β-and d-subunits in hybrid ancestral/human AChRs 300 (7). For this to be possible the principal (+) and complementary (−) subunit interfaces of βAnc 301 must be compatible with each other (Fig. 1), leading to the logical hypothesis, confirmed 302 here, that βAnc can form homopentamers. This ability to revert the muscle-type β-subunit, 303 which is entrenched in a heteropentamer, back to a subunit capable of forming 304 homopentamers, attests to the homopentameric origins of all pentameric ligand-gated ion 305 channel subunits (39). 306 We have also shown that βAnc homopentamers open spontaneously. This is surprising. 307 What is even more surprising is that the spontaneous single-channel activity of βAnc 308 homopentamers resembles that of the agonist-activated muscle-type AChR. Some of the first 309 single-channel recordings of frog AChRs activated by agonists revealed that agonist- was open for more than 90% of the duration of the burst). Spontaneous openings of βAnc 315 homopentamers also occur in bursts, and despite the absence of agonist, activation appears 316 efficient, with the probability of being open within a burst also exceeding 0.90. The kinetic 317 structure of bursts of spontaneous βAnc openings also resembles that of the agonist-activated 318 AChR, with bursts containing several types of closings, the briefest of which are reminiscent 319 of classic "nachschlag shuttings". This is significant because it was originally proposed that 320 the duration of nachschlag shuttings was related to agonist efficacy (22). However, 321 subsequent work showed that nachschlag shuttings were independent of agonist, which 322 necessitated the introduction of an additional closed state appended to the agonist-activated 323 open state in early schemes of AChR activation (41). This latter finding was some of the 324 impetus for refined schemes, where the additional closed states preceded channel opening 325 and were referred to as "flipped" or "primed" (23, 24). Our kinetic analysis has shown that 326 the spontaneous activity of βAnc fits an analogous scheme, containing an intermediate closed 327 state that also precedes channel opening, but where the agonist binding steps have been 328 omitted due to the absence of agonist. Evidently, these functional hallmarks of AChR 329 activation do not arise from the complex heteropentameric architecture of the muscle-type 330 receptor, nor do they depend upon the presence of agonist. Instead, they are fundamental 331 properties preserved and encoded in the reconstructed amino acid sequence of βAnc. which lines the channel pore at its narrowest constriction, is a well-known determinant of 338 single-channel conductance and block, and is one of the more conserved regions across all 339 pentameric ligand-gated ion channel subunits (2). Accordingly, this site is largely conserved 340 in βAnc (9), thus the chemical composition and profile of the open pore in βAnc 341 homopentamers is expected to resemble that of the open pore in the wild-type AChR. 342 A degree of uncertainty is inherent in the reconstruction of any ancestral protein, and 343 to solidify evolutionary conclusions it is important to assess whether the functions of 344 putative ancestors are robust to these uncertainties. We have shown that an alternate 345 ancestral β-subunit (βAncS), with 67 substitutions and 6 indels relative to βAnc, was still able 346 to form homomers that opened spontaneously. Furthermore, spontaneous openings of βAncS 347 homomers also occurred in bursts that contained brief closings (i.e., "nachschlag shuttings"), 348 and had an open probability that exceeded 0.90. These findings demonstrate that the ability 349 of βAnc to form spontaneously opening homomers is robust to substantial variations in the 350 inferred ancestral β-subunit amino acid sequence. Thus, the ability of these reconstructed 351 ancestral subunits to form spontaneously opening homomers is deeply embedded in their 352 structure and evolutionary history. 353 The kinetic behaviour of βAncS homomers was more complex than originally observed 354 with βAnc. This increased complexity necessitated the expansion of the original scheme used 355 to fit βAnc to include two additional priming steps. Both the original βAnc and expanded βAncS 356 schemes parallel the scheme used to describe the muscle-type AChR, which included two 357 priming steps (i.e. "singly" and "doubly" primed) that each correlated with conformational 358 changes around the two agonist-binding sites (24). In the case of βAncS, which lacks agonist-359 binding sites, the simplest interpretation is that the different levels of priming correlate with 360 conformational changes occurring within individual subunits. Although they are labelled as 361 "singly", "doubly", and "triply" primed, assuming that βAncS also forms homopentamers, the 362 three priming steps in βAncS homomers could represent the three most terminal priming 363 steps, where three, four, or five βAncS subunits are primed. Within this framework, openings 364 from βAncS channels with zero, one, or two subunits primed, are presumably unstable, and 365 thus not observed in our single-channel recordings. In an alternate scenario, "singly" and 366 "doubly" primed could refer to βAncS channels with one or two primed subunits, respectively. 367 While "triply" primed could refer to channels with three or more primed subunits, but where 368 openings from βAncS channels with three, four, or five primed subunits are indistinguishable. 369 Applied to βAnc, these interpretations suggest that either (1)  place the roots of agonism at a stage in the activation process that precedes channel opening 376 (13, 23, 24). A consequence of these mechanisms is that the ultimate opening and closing 377 rates of the channel are independent of agonist. Here, we have shown that additional single-378 channel hallmarks of AChR function are also independent of agonist, as they occur in 379 spontaneously opening homopentameric channels that are devoid of agonist-binding sites, 380 and which are formed from reconstructed ancestral AChR β-subunits. Often overlooked, the 381 β-subunit is the least conserved of the four AChR subunits, and is the only subunit that does 382 not contribute residues to the two AChR agonist-binding sites. Despite these considerations, 383 hallmarks of AChR function remain deeply embedded in β-subunit sequence, structure, and 384 evolutionary history. Given that these functional hallmarks are independent of agonist, it is 385 tempting to speculate that they predate agonism, and thus that agonism evolved 386 subsequently as an additional layer of regulation in this family of pentameric ion channels. , where the agonist-binding sites at the α-δ and α-ε subunit interfaces are indicated with asterisks (*). A reconstructed ancestral β-subunit (β Anc ; purple) forms hybrid acetylcholine receptors (middle) where β Anc substitutes for the human β-subunit (β; orange) and supplants the human δ-subunit (δ; green). The principal (+) and complementary (−) interfaces of β Anc must be compatible for two β Anc subunits to sit side-by-side (red highlight), which predicts that homomers formed from multiple β Anc subunits should be possible (right, boxed).   Table S1. Openings are upward deflections. Recordings were obtained with an applied voltage of -120 mV. Data were filtered to 10 kHz (scale bars = 25 ms, 10 pA; applies to "A" and "B"). The sequence of dwells from each data set, encompassing the full concentration range of the blocker, was globally fit to the same three-state scheme used for β Anc , where an additional fourth state corresponding to the open/blocked channel was added (Scheme 2). Global kinetic fits were performed on three individual recordings for each concentration of blocker, from at least two separate transfections, corresponding to 15 total patches for each global fit. Note that the recordings in the absence of blocker are the same for each data set. Rate constants with error estimates are presented in Table S1. The amplitudes segregate into six well-defined amplitude classes (combined from the two patches in "C"), where the highest and lowest amplitude classes match that of the all-HC and all-LC classes, respectively. Plot of the mean amplitude of each class as a function of the presumed number of incorporated HC subunits (error bars = standard deviations of the mean but are smaller than the points themselves). Recordings were obtained with an applied voltage of -120 mV, and traces were digitally filtered to 1 kHz to facilitate amplitude detection (scale bar = 25 ms, 10 pA; applies to "A", "B", and "C"). The three-state scheme in Figure 3 (Scheme 1) was expanded to include additional priming steps ("singly", "doubly", and "triply" primed), each with their own connected open state (Scheme 3). Rate constants are shown, with error estimates provided in Table S3.

This PDF file includes:
Figures S1 to S5 Tables S1 to S3 Description of Source Data files  Table S2. Global kinetic fits were performed on three individual recordings for each concentration of acetylcholine, from at least two separate transfections, corresponding to 24 total patches for the entire global fit.  Figure S3, but with an additional open/blocked state (corresponding to QX-222 block) flanking the doubly-liganded open state (Scheme 5). All rates, except those describing QX-222 block, were fixed to that at the indicated acetylcholine concentration as estimated from original QX-222-free fits in Figure S3. For both concentrations of acetylcholine, estimates of the rate constants from global fitting of each data set are shown, with error estimates presented in Table S2. Global kinetic fits were performed on three individual recordings for each concentration of QX-222 at both concentrations of ACh, and from at least two separate transfections, corresponding to 15 total patches for each global fit.  Figure S5: Event-based amplitude histograms derived from a representative single patch where cells were cotransfected with cDNAs encoding for high-conductance (HC) and low-conductance (LC) variants of β Anc at a ratio (mass:mass) of (A) 1HC:4LC or (B) 4HC:1LC. In each case, cotransfection of HC and LC β Anc subunits led to a distribution of amplitudes that segregated into distinct amplitude classes, where the proportion of events in each class was biased by the relative proportion of each type of cDNA transfected. These two patches with overlapping amplitude classes were combined to produce the plot in Figure 5D.   Note: Rate constants were estimated from fitting Scheme 4 or Scheme 5 (below) presented in Figure S3 or Figure S4, respectively. Where "A" represents agonist, and "R", "R*", and "R*  Note: Rate constants were estimated from fitting Scheme 3 in Figure 6 (and below). Data were globally fit (three individual patches from two separate transfections) with rate constants and associated errors (parentheses) estimated within MIL (see Experimental Procedures). Gating equilibrium (θn) constants represent βn/⍺n. Rate constants are presented as s -1 .

Description of Source Data files
Source Data - Figure 3.zip • Source data for Figure 3. Detected single-channel event durations of spontaneously opening βAnc homomers. Compressed file includes three TAC event files (*.evt format) of the single-channel detections for the three recordings used in the presented kinetic analysis, as well as the associated R scripts (*.txt format) for defining and sorting bursts. Compressed file includes 24 TAC event files (*.evt format) of the single-channel detections for the three recordings for each acetylcholine concentration (fileA, fileB, fileC in each case) in the presented kinetic analysis, as well as the associated R scripts (*.txt format) for defining and sorting bursts.

(48 files total)
Source Data - Figure S4.zip • Source data for Figure S3. Detected single-channel event durations for QX-222 block of the wild-type adult muscle acetylcholine receptor. Recordings were in the presence of either 10 µM or 30 µM acetylcholine (ACh), and increasing concentrations of QX-222 (QX222; 10 µM, 30 µM, 60 µM and 100 µM). Compressed file includes 30 TAC event files (*.evt format) of the single-channel detections for three recordings for each condition (fileA, fileB, fileC in each case) included in the kinetic analysis, as well as the associated R scripts (*.txt format) for defining and sorting bursts.