The Journal of Physiology

Transcription

1 J Physiol (2014) pp Agonists binding nicotinic receptors elicit specific channel-opening patterns at αγ and αδ sites Patrick Stock 1, Dmitrij Ljaschenko 1, Manfred Heckmann 1 and Josef Dudel 2 1 Department of Neurophysiology, Institute of Physiology, University of Wuerzburg, Wuerzburg, Germany 2 Friedrich Schiedel Institute for Neuroscience, Technical University Munich, Munich, Germany The Journal of Physiology Key points High-resolution patch clamp currents evoked by epibatidine (Ebd), carbamylcholine (CCh) and acetylcholine (ACh) were compared. Ebd binds with 75-fold higher affinity at αγ than at αδ sites, whereas CCh and ACh prefer αδ sites of nicotinic ACh receptor (nachr) channels. Similar short (τo1 ), intermediate (τ O2 ) and long (τ O3 ) types of opening were observed. τ O2 openings were maximally prevalent at low Ebd concentrations, binding at αγ sites, whereas τ O1 openings appear to be generated at αδ sites. Short ( 0.75 ms) bursts of openings (τb1 )arisefromtheαγ site, and long (>10 ms) bursts (τ B2 ) arise from double liganded receptors. The duration of bursts and of openings within bursts depended on the agonist. Limited by the temporal resolution, the closings within bursts were invariant at 3 μs. Blocking αδ sites with α-conotoxin M1 (CTx) eliminated both τo1 and τ B2 and left only τ O2 and the short τ B1 bursts, as expected. Abstract Embryonic muscle-type nicotinic acetylcholine receptor channels (nachrs) bind ligands at interfaces of α- andγ- orδ-subunits. αγ and αδ sitesdifferinaffinity, buttheir contributions to opening the channel have remained elusive. We compared high-resolution patch clamp currents evoked by epibatidine (Ebd), carbamylcholine (CCh) and acetylcholine (ACh). Ebd binds with 75-fold higher affinity at αγ than at αδ sites, whereas CCh and ACh prefer αδ sites. Similar short (τ O1 ), intermediate (τ O2 ) and long (τ O3 ) types of opening were observed with all three agonists. τ O2 openings were maximally prevalent at low Ebd concentrations, binding at αγ sites. By contrast, τ O1 openings appear to be generated at αδ sites. In addition, two types of burst appeared: short bursts of an average of 0.75 ms (τ B1 ) that should arise from the αγ site, and long bursts of ms (τ B2 ) in duration arising from double liganded receptors. Limited by the temporal resolution, the closings within bursts were invariant at 3 μs. Corrected for missed closings, in the case of ACh the openings within long bursts lasted 170 μs and those in short bursts about 30 μs. Blocking αδ sites with α-conotoxin M1 (CTx) eliminated both τ O1 and τ B2 and left only τ O2 and the short τ B1 bursts, as expected. Furthermore we found desensitization when the receptors bound ACh only at the αγ site. When CTx was applied to embryonic mouse endplates, monoquantal current rise times were increased, and amplitude and decay time constants were reduced, as expected. Thus the αγ and αδ sites of nachrs elicit specific channel-opening patterns. P. Stock and D. Ljaschenko made equal contributions to this study. DOI: /jphysiol

2 2502 P. Stock and others J Physiol (Received 4 November 2013; accepted after revision 21 March 2014; first published online 24 March 2014) Corresponding authors M. Heckmann: Department of Neurophysiology, Institute of Physiology, University of Wuerzburg, Roentgenring 9, Wuerzburg, Germany. heckmann@uni-wuerzburg.de J. Dudel: Friedrich Schiedel Institute for Neuroscience, Technical University Munich, Biedersteinerstrasse 29, Munich, Germany. josef.dudel@lrz.tu-muenchen.de Abbreviations ACh, acetylcholine; CCh, carbamylcholine; CTx, α-conotoxin M1; DFP, diisopropylfluorophosphate; Ebd, epibatidine; LLR, logarithmic likelihood ratio; μ burst, mean burst duration; μ gap, mean shut time; μ open, true mean duration of openings within bursts; nachrs, nicotinic acetylcholine receptors; n gaps/burst, mean number of gaps per burst; PDFs, probability density functions; qepscs, quantal excitatory postsynaptic currents; r.m.s., root mean square; s.n.r., signal to noise ratio; t crit, critical shut time. Introduction Nicotinic acetylcholine receptors (nachrs) are cation-permeable ion channels that mediate neuronal signals at chemical synapses in response to presynaptically released acetylcholine (ACh). Muscle-type nachrs in particular transmit all fast synaptic excitation at endplates of voluntary muscles in vertebrates. Studied for over a century (Langley, 1906), they are among the best investigated synaptic proteins (Colquhoun, 2006) for which substantial quantitative structural knowledge has been obtained (Kistler & Stroud, 1981; Miyazawa et al. 1999; Unwin & Fujiyoshi, 2012). The receptors are composed of five homologous subunits, symmetrically arranged around a central membrane-spanning pore (Fig. 1A) In embryonic or denervated muscle, the arrangement of subunits as viewed from the extracellular space is α, β, δ, α, γ in clockwise order. Transmitter binds at the interface between an α-subunit and the neighbouring γ- orδ-subunit (Karlin, 1993; Sine et al. 1995; Xie & Cohen, 2001) and elicits openings of the central ion-conducting pore of the receptor. Agonists bind at both sites with different levels of affinity (Czajkowski et al. 1993; Sine, 2012). Since the binding gating problem was first addressed by del Castillo & Katz (1957), kinetic schemes have been proposed for the nachr and its relatives with the aim of quantitatively describing opening behaviour. The development of the patch clamp technique (Neher & Sakmann, 1976), the gigaseal and the various recording modes (Sigworth & Neher, 1980; Hamill et al. 1981) made it possible to observe single-channel currents at high temporal resolution and revealed the burst-like nature of channel activation (Colquhoun & Sakmann, 1981; Sine & Steinbach, 1986a, b)(fig.1c). Seminal investigations of nachr using different agonists delivered a first broad description of current patterns at a resolution of 40 μs (Colquhoun & Sakmann, 1985). When the temporal resolution of on-cell patch clamping was improved to 6 μs, it became clear that a minimum of three different open states exist (Parzefall et al. 1998; Hallermann et al. 2005). The two types of single opening were attributed to the binding of agonist to one of the binding sites, whereas bursts of openings required simultaneous binding at both sites (Fig. 1C). These investigations also defined 3 μs closings within bursts as a constant property of the channel and as independent of the nature of the agonist. However, the concentration on which the proportions of single and burst openings depended was poorly predicted. In particular, it remained unclear how the differing binding sites influence opening kinetics. Using mutated embryonic receptors, Akk et al. (1996) concluded that binding of agonist at the αγ site promotes gating more than does binding at the αδ site. Further studies suggested that after binding of agonists, the structure of the receptor is rearranged stepwise until the channel finally opens (Grosman et al. 2000; Chakrapani et al. 2004; Zhou et al. 2005; Purohit et al. 2007). Based on these findings, two groups recently introduced so-called flip or primed states into the scheme of receptor kinetics (Lape et al. 2008; Mukhtasimovaet al. 2009; Colquhoun & Lape, 2012). The relationship between binding sites and gating kinetics has not yet been resolved. Therefore, we combined ultra high-resolution single-channel recordings and agonists with differing binding site affinities as well as a site-specific blocker for mouse embryonic nachrs (Fig. 1). Methods Ethical approval All experiments were approved by the Veterinary Office of the Regierungsbezirk Unterfranken, Germany and were carried out in compliance with the guidelines of the Journal of Physiology (Drummond, 2009). Preparations Myotubes containing embryonic nachrs were prepared from interossal toe muscles of the hind feet of neonatal wild-type mice. In line with national guidelines, we ensured that animals did not suffer needlessly. Mice of either sex aged up to postnatal day 7 (Dudel & Heckmann,

3 J Physiol Different binding site contributions ) were decapitated and the muscles excised and incubated in enzyme solution for 1 h at 37 C to dissociate the cells and remove presynaptic nerve terminals (for details, see Költgen et al. 1991). To ensure that only embryonic-type receptors were exposed on the myotubes, cells were kept in culture for at least 7 days, as described in Franke et al. (1992). The cells were used for 1 2 h for patch clamp recordings in the cell-attached mode. For the measurement of quantal synaptic currents hemidiaphragms of mice of either sex aged < 6 days were superfusedinaplexi-glasschamber. Synaptic currents Quantal excitatory postsynaptic currents (qepscs) were measured extracellularly at single nerve terminals of mouse muscle while excitations were prevented by tetrodotoxin. A perfused macro-patch electrode with an opening at the tip of about 10 μm in diameter was placed over a terminal. Synaptic currents were elicited by passing 1 ms current pulses through the electrode to elicit quantal transmitter releases. The qepscs were recorded through the same electrode. For details, see Dudel (2007). On-cell configuration For recordings, cover slips were transferred to Petri dishes containing physiological solution (162.0 mm NaCl, 5.3 mm KCl, 2.0 mm CaCl 2,0.67mM NaH 2 PO 4, 15 mm Hepes; NaOH adjusted to ph 7.4). The same solution was used for the pipette tip, but in addition contained one of the agonists [ACh, carbamylcholine (CCh), epibatidine (Ebd)] in concentrations ranging from μm to 300 μm. All recordings were performed at room temperature (20 ± 2 C) using the on-cell patch clamp technique (Hamill et al. 1981). To improve the signal : noise ratio (s.n.r.), all patches were voltage clamped by 200 mv (resting potential about 60 mv). Low-noise modifications We used an Axopatch 200B amplifier (formerly Axon Instruments, now Molecular Devices, Inc., Sunnyvale, CA, USA) with a cooled input field-effect transistor and a capacitive feedback input. The set-up was modified to yield very low baseline noise (for details, see Parzefall et al. 1998; Hallermann et al. 2005). Borosilicate glass pipettes produce substantial noise because salt solutions creep up the inner and outer walls of the pipette. Quartz glass has superior electrical characteristics and its hydrophobic surface does not support the creeping of fluid films. Quartz glass tubes with 2 mm outside and 1 mm inside diameters (Heraeus Quarzglas GmbH, Hanau, Germany) were used to pull pipettes with a DMZ-Quartz puller (Zeitz Instruments GmbH, Martinsried, Germany) (Dudel et al. 2000). Acquisition The filters in the outputs of the Axopatch 200B amplifier were removed and signals were recorded at full bandwidth. After the removal of the original filter, the amplifier s internal bandwidth was determined to 130 khz at a cut-off of 3 db. The data stream was converted using either a micro 1401 digitizer at a sampling rate of 400 khz or a power 1401 digitizer at a sampling rate of 1 MHz (both: Cambridge Electronic Design Ltd, Cambridge, UK). Data were stored directly on a regular PC running Windows XP Professional (Microsoft Corp., Redmond, WA, USA), using Signal3 Version X.X (Cambridge Electronic Design Ltd). The data were exported as 16-bit integer formatted data. Further processing was performed using DC Analysis software (David Colquhoun, University College London; As the internal bandwidth of the Axopatch 200B was 130 khz ( 3 db), filtsamp (DC Analysis) was used to filter the data with a digital Gaussian filter of Hz ( 3 db) to obtain a final filter cut-off frequency (fc) of 30 khz ( 3 db). With an fc of 30 khz, the typical rise time is μs. Resulting data streams were saved in continuous sample (consam) file format. Likewise, filtsamp is a utility to omit data points. If data are acquired at a 1 MHz sampling rate, in order to achieve a resolution similar to that obtained with a lower sampling rate, only every third point is used (the effective sampling rate is 333 khz), which is also necessary for the evaluation program. Idealization For the reliable detection of events, an s.n.r. of at least 10 is necessary. With the optimized low-noise set-up, the mean ± S.E.M. root mean square (r.m.s.) noise of the recordings was 0.85 ± pa at 30 khz fc. The overall mean ± S.E.M. amplitude was ± 0.42 pa, resulting in a total mean s.n.r of For the idealization of recordings, the program SCAN (DC Analysis Programs) was used (Colquhoun & Sigworth, 1995). The recordings were analysed as described earlier (Hatton et al. 2003; Colquhoun et al. 2003). The idealized records were analysed using the EKDIST (DC Analysis Programs) utility with an imposed resolution of 6 μs. Exceptions were made for two of the records using 100 μm ACh as agonist. Given their amplitude, reduction values of 8 μs and 10 μs were considered to be safe in these records to prevent any events being missed at the imposed resolution. The problem of unresolved events will be addressed further in the Results section. Idealized records were checked for the stability of

4 2504 P. Stock and others J Physiol the fitted amplitudes and open probability. In the case of any observed instability, the whole recording was not used. Dwell time distributions Opentime,shuttimeandburstlengthdistributionswere calculated with EKDIST. All dwell time histograms are shown at logarithmic scale with logarithmic binning for the abscissa and root scale for the ordinate (Sigworth & Sine, 1987; McManus et al. 1987). Open time distributions were displayed as open period distributions to avoid the interpretation of baseline fluctuations of sufficient amplitude as individual openings. As long as the amplitude change was <2 pa, neighbouring dwells were considered as single openings. Consistent with our findings in the observation of amplitude histograms, nachrs opened to one conductance level only and no subconductance level was observed. Exponential probability density functions (PDFs) with two and three components were directly fitted to each of the distributions and the logarithmic likelihood ratio (LLR) was calculated (Hallermann et al. 2005). With an LLR value of >4.6, the probability that three components will be accepted by mistake although two components are sufficient is <1% and thus the procedure can be considered to be safe (Rao, 1973; Horn, 1987). Shut time distributions were fitted using PDFs with five or six components. If the same distribution was fitted several times with a mixture of five components delivering unstable results, a sixth component was used to comply with the broader dispersion of shut times in the distribution. Statistical analysis To test whether parameter groups (e.g. of different agonists) show diversity, ANOVA or, if there were only two groups to compare, Student s t tests were performed. The number of records carried out for each concentration differed between agonists and concentrations and the exact number will be given in the results when necessary. The non-parametric Mann Whitney rank sum test was used for statistical analysis unless stated otherwise. Data are reported as means ± S.E.M.; n indicates the sample number and P denotes the level of significance ( P < 0.05). Results High-resolution recordings Figure 2A shows original current traces, obtained with 1 μm ACh. Below the upper trace a heavy grey line indicates the stretch of the recording that is expanded in the following trace. Idealization of current traces was carried out as described in Colquhoun & Sigworth (1995) to ensure maximal resolution (Fig. 2B). Dwell time distributions (Fig. 2C and D)showdifferent components and their amplitudes for openings and closings. In the open period distribution it is quite obvious that a minimum of two components are required to fit the PDFs to the histogram. The peaks in the distributions indicate the mean open period for the component and the corresponding areas of the components, and add up to unity and thus represent a measure for the relative frequency of events. For the gaps (Fig. 2D), the most eye-catching observation is the high abundance of very short gaps at the resolution limit of 6 μs that is consistently present in all records. Although a high-resolution recording set-up was used, gaps in this group cannot be completely resolved. Fitting a mixture of PDFs predicts a high proportion of gaps beyond the resolution limit of 6 μs. The predicted number of gaps assumes that about 81% (total mean) of the gaps were constantly missed. For comparison, a proportion of short openings were similarly missed, but the numbers were substantially lower and thus about 16% of openings were missed by the detection method (Fig. 2C). Channel open times with ACh, CCh and Ebd Agonists for this study of binding site-related gating characteristics were chosen for their specific selectivity for structurally different binding sites: ACh, the natural transmitter, is 30-fold more site-selective for the αδ binding site (Zhang et al. 1995). CCh has similar but lower site selectivity. The third agonist, Ebd, shows an opposite preference in binding in comparison with ACh and CCh (Zhang et al. 1995). It selects the αγ site 75-fold more oftenthantheαδ site of the receptor (Prince & Sine, 1998a, b). We also employed a blocker of the αδ binding site, α-conotoxin M1 (CTx; see below), which shows a 10,000-fold preference for the αδ site over the αγ site (Sine et al. 1995). Figure 3A shows the mean durations of open time components for the agonists ACh, CCh and Ebd. Former electrophysiological investigations of nachrs proposed a minimum of two open period components to describe dwell time distributions (Colquhoun & Sakmann, 1985). The existence of three different components has been demonstrated with the ultra high-resolution recording technique for the agonists ACh and suberyldicholine (Parzefall et al. 1998; Hallermann et al. 2005). LLRs were calculated to test whether fitting a mixture of three versus two components (Fig. 2) gives significantly better results for the agonists used here, as indicated by the higher LLRs (Hallermann et al. 2005). For CCh and Ebd, the LLRs show a roughly indirect logarithmic concentration dependency that resembles the

5 J Physiol Different binding site contributions 2505 decrease in frequency observed in the intermediate open time component in three-component PDF fitting. Thus, although it may not be necessary for all of the records to use the three-component approach, general application seems to be appropriate. In three-component fits, none of the durations of opening show significant concentration dependency. Mean durations of components differ significantly (P < 0.05) for long openings (τ O3 ) with CCh and Ebd (Fig. 3A). Figure 3B compares the relative areas of the three components of open times as fitted at the respective agonist concentrations. The areas are equivalent to the relative frequency of openings from the respective component. As expected, the longest component (τ O3, green) increases with rising agonist concentration. The increase is similar with ACh and CCh, but significantly steeper with Ebd. Whereas with ACh and CCh the long component approaches 100% of the total area at a concentration of about 1 mm, Ebdismoreeffective, approaching saturation with a concentration of 10 μm, as seen also in receptor binding studies (Prince & Sine, 1998a, b). The concentration dependencies of short and intermediate components of open times [τ O1 and τ O2 (red and black, respectively)] are less similar in comparison with those of the long component. The area of τ O1 decreases concentration dependently for ACh. The areas of the intermediate open period components (τ O2 )seemtochange insignificantly with the concentration of ACh and CCh. By contrast, the area of this component decreases steeply with Ebd concentration. Concentration dependencies of the shorter open time components should be interpreted with some caution because relative frequencies have been plotted. The longest component (τ O3 ) is predominant and is larger than the other two except at the lowest Ebd concentration. Therefore, the slopes of the concentration dependencies of the τ O1 and τ O2 components must be considered in relation to the slope of τ O3. The problem is illustrated by the effects of Ebd. The relative frequency of τ O3 increases from 40% to almost 100% with the rising concentration. The relative frequency of τ O2 decreases with a similar but negative slope. This relationship may suggest that the absolute frequency of τ O2 openings almost does not change as the agonist concentration rises. Absolute frequencies of the three components could not be determined directly. With the high-resolution patch clamp technique, the effects of one agonist concentration could be evaluated only at one specific patch. Individual patches recorded from unknown numbers of active receptorsorchannels.inordertodetermineaverages at a certain concentration, normalization of individual data points was necessary and was achieved by calculating relative frequencies. Figure 2D presents an example of a distribution of shut times for 1 μm ACh. The distribution is dominated by a 3 μs component (τ C1 ), followed by a number of less frequent longer components. This picture is rather invariant with different agonists and concentrations, and findings will not be illustrated in detail, but listed exemplarily in the legend for Fig. 2. As a great number of all the detected gaps are of the shortest component and very similar length values of about 3 μs are found, the fitted PDFs can be expected to give good estimates for the predicted number of events. The total predicted number is derived from the number of detected gaps plus the number of predicted gaps beyond the resolution limit, which are found in the process of fitting PDFs. Only the 3 μs gap component is contained within bursts that represent individual entities. The longer gap components Figure 1. Embryonic muscle type nicotinic acetylcholine receptor (nachr) with αδ-subunit and αγ-subunit acetylcholine (ACh) binding sites A, the receptor, as seen from the synaptic cleft (structural data from Unwin, 2005). Dashed circles indicate the two ligand binding sites at interfaces of α- and neighbouring γ - and δ-subunits. Whereas epibatidine (Ebd) has a substantially higher affinity for αγ than αδ binding sites, carbamylcholine (CCh) and ACh show opposite and less selectivity, indicated by the font size of the agonist abbreviation. The antagonist α-conotoxin M1 (CTx) shows about 10,000-fold selectivity for αδ sites. B, structures of ACh, CCh and Ebd showing carbon (turquoise), nitrogen (blue), oxygen (red), chloride (green) and hydrogen (white). C, reaction scheme emphasizing the two binding sites.

6 2506 P. Stock and others J Physiol are generated by an unknown number of channels and cannot be interpreted as actions of one specific channel. Most importantly in the context of the present discussion, there is no indication of significant differences between the agonists for the briefest gaps. Bursts of channel openings with ACh, CCh and Ebd As Fig. 2A and B shows, openings and closings of the channels do not occur at random in time, but are arranged as single openings and bursts of openings (Colquhoun & Sakmann, 1985). How do the described components of open and shut periods fit into this pattern? Correlations of open times and the adjacent shut periods demonstrate their relationships (Magleby & Song, 1992). An example for 10 μm ACh is given in Fig. 9A. The plot shows a strong correlation of short gaps with long openings and vice versa. Obviously the longer open periods τ O3 and τ O2 in Fig. 3A together with the short component τ C1 of shut times make up the bursts. Short openings τ O1 create a band of longer shut times in Fig. 9A. Such correlations are consistent across all records and very similar results have been published previously for high-resolution data (Parzefall et al. 1998; Hallermann et al. 2005). Functionally the bursts of channel openings (Fig. 2A) are of prime interest because they provide almost the total current flow necessary for synaptic transmission. Further, when using the on-cell patch clamp technique, it is not possible to guarantee that all openings originate from the same receptor molecule, despite the fact that no overlapping openings are observed. Individual bursts are obviously generated by one receptor channel, but other bursts or single openings may each arise from a different receptor. As mentioned above, bursts are commonly defined as a series of openings separated by gaps shorter than a critical shut time (t crit ). Within the bursts, gaps of the τ C1 Figure 2. Current traces and dwell time distributions A, successive 10-fold temporal expansions of a low-noise trace obtained with acetylcholine (ACh). The heavy grey lines indicate the section to be expanded. B, idealization of the original data (blue) at a sampling rate of 333 khz and final evaluation (red) (see text). Top row: left, a 6 μs opening; right, a transition from shut to open (467 μs) followed by a 10 μs gap. Bottom row: a long opening with gaps of different durations. C, open period distribution for a 1 μm ACh recording (data in black) fitted with two and three components (continuous blue and dashed red lines, respectively). The maxima of the fits are at 5.6 μs and 662 μs for two components, covering 22.6% and 77.4%, respectively, of the total area, and 5.6 μs, 251 μs, 775 μs for three components, covering 23.5%, 15.2% and 61.3%, respectively, of the total area. D, shut time distribution fitted with a mixture of five components: 3.0 μs, 96.5%; 69.0 μs, 0.64%; 587 μs, 0.96%; 11.6 ms, 0.77%, and 153 ms, 1.1%. The shortest shut time (τ C1 ) is a stable component with all agonists and concentrations with mean durations of 3.3 ± 0.23 μs, 3.2 ± 0.11 μs and 3.0 ± 0.14 μs and 95.7 ± 0.8%, 96.6 ± 0.4% and ± 2.53% mean area, for ACh (n = 22 evaluated patches), carbamylcholine (CCh) (n = 14) and epibatidine (Ebd) (n = 19), respectively. Similar to τ C1 but with broader dispersion, the means for τ C2 are 40.8 ± 13.1 μs, 36.9 ± 6.7 μs and 54.4 ± 17.2 μs, with 1.3 ± 0.30%, 1.2 ± 0.13% and 1.5 ± 0.30% component areas for ACh, CCh and Ebd, respectively.

7 J Physiol Different binding site contributions 2507 type constitute the most stable and dominant parameter, and, in high-frequency resolution recordings, account for at least 80% of detected gaps. Furthermore, short gaps are highly correlated with long openings, whereas short openings are highly correlated with long gaps (Fig. 9A). Even in Fig. 2A the two groups can be easily identified by eye. We chose t crit to discriminate between these two groups; t crit is positioned in the shut time distributions rightafterthedominantτ C1 component (i.e. at 40 μs in Fig. 2D). In different recordings values of t crit were set between 30 μsand70μs. In Fig. 4 the distributions of the openings in Fig. 3 have been re-evaluated to define single openings and bursts. Openings separated by gaps shorter than t crit were grouped as bursts, and openings separated by longer gaps were counted as single openings. In Fig. 4A, the shortest component (about 5 μs; red) corresponds to the short τ O1 single openings in Fig. 3A. Thelongest component (>10 ms; orange) obviously represents long bursts of openings that contain the openings of 1 ms (green) in Fig. 3A. The ms component in Fig. 4A (blue) represents short bursts because no single openings of this length are left in Fig. 3A. The openings within these short bursts are those of about 200 μs (black)in Fig. 3A. The long and short bursts share the 3 μs gapsbetween openings: there is no other definite gap component within t crit. A large proportion of these gaps is not measured, but predicted. If the bursts contain more gaps than are measured, the openings within bursts must be shorter than they appear. Using the approach described by Colquhoun & Sakmann (1985), mean open periods (μ open ) within bursts that are corrected for missing gaps can be determined: μ open = μ burst μ gap n gaps/burst n gaps/burst + 1 Corrected mean durations (μ open ) were calculated from the mean burst duration (μ burst ) minus the shut times during the burst. The latter is the product of the mean shut time (μ gap ) and the mean number of gaps per burst (n gaps/burst, including the estimated missed gaps). The difference must be divided by the mean number of openings per burst, which is the mean number of gaps per burst (n gaps/burst ) plus 1. The calculated mean μ open values for long bursts are 171 ± 19.0 μs, 206 ± 23.4 μs and 146 ± 10.6 μs for ACh, CCh and Ebd, respectively (Fig. 5). With these μ open and gap durations of 3 μs, the long bursts contain on average about 140, 100 and 90 openings for ACh, CCh and Ebd, respectively. If analogous correction factors are used for the openings within short bursts, the μ open values will be 31 μs, 50 μs and23μs, respectively. The average number of openings within short bursts will be 24, 14 and 13 for ACh, CCh and Ebd, respectively. The correction to μ open is an approximation. A probably more significant correction could be reached if our data were introduced into a model of channel kinetics fitted through the HJCFIT programme (Hawkes et al. 1992). In the Discussion, we will explain why we cannot present such a model at present. We have not yet commented on the 65 μs component (black) in Fig. 4A. This does not seem to represent single Figure 3. Concentration and agonist-dependent channel openings A, durations of short (τ O1, red), intermediate (τ O2, black) and long (τ O3, green) openings for acetylcholine (ACh), carbamylcholine (CCh) and epibatidine (Ebd). With Ebd τ O3 is ± 0.04 ms and significantly (P < 0.05) shorter than its 1.19 ± 0.08 ms with CCh. Short and intermediate (τ O1 and τ O2 ) openings do not differ with means of 6.19 ± 0.71 μs, 7.43 ± 2.34 μs and 5.35 ± 0.76 μs forτ O1, and ± ms, ± 0.07 ms, ± ms for τ O2 for ACh, CCh and Ebd, respectively. The number of evaluated patches is as in Fig. 2. B, relative frequency of open period components as the percentage of the area of the fitted probability density functions at different ACh, CCh and Ebd concentrations with one to four evaluated patches per data point.

8 2508 P. Stock and others J Physiol openings because it has no counterpart in Fig. 3A. If this component were a burst, it could contain only short τ O1 openings. However, in Fig. 9A correlations of short gaps and short openings are very improbable. The 65 μs component may be related to the short bursts. Their openings are shorter than 65 μs, but the latter value may require correction for missed gaps. Therefore, we are inclined to add the black component in Fig. 4A to the short bursts. As for the differences in duration between the applied agonists, the single openings τ O1 (red) and τ O2 (black) are insignificantly different with overall mean values of 5.3 ± 0.54 μs (τ O1 ) and 61.6 ± 9.78 μs (τ O2 ). The bursts τ B1 (blue) and τ B2 (orange) have insignificantly different durations of about 0.75 ms and 25 ms, respectively, with the agonists ACh and CCh. With Ebd, the long bursts appear to be significantly shorter than the 25 ms of the other two agonists, whereas the difference in the short bursts is not significant. The relative frequencies of the different components at increasing concentrations of agonist in Fig. 4B draw a rather diverse picture: the proportion of the short single openings component (τ O1 ) of Ebd (39.22 ± 4.49%; red) is significantly smaller than those of the other agonists ACh (65.9 ± 1.96%) and CCh (55.2 ± 8.75%). The long single openings component (τ O2, black) is large only with Ebd, but declines at higher concentrations. The short bursts (τ B1, blue) component shows significantly different negative slopes for concentration dependencies of 1.91 ± 0.47%/μM, 5.01 ± 1.65%/μM and 15.9 ± 5.10%/μM, for ACh, CCh and Ebd, respectively. The areas of long bursts (τ B2, orange) increase slightly with the concentration of agonist, most strongly and highly significantly for Ebd (8.5 ± 1.60%/μM). Thus the concentration dependence of burst time distributions for Ebd elicited channel currents that differ substantially from those of the other agonists. Block by CTx As we stated in the Introduction (Fig. 1), ACh and CCh prefer to bind to the αδ site of the receptor channel, whereas Ebd prefers the αγ site. The results presented so far may indicate that the binding of an agonist to the αδ site elicits short single openings (τ O1 ), binding to the αγ site elicits longer single openings (τ O2 )aswellasshortbursts (τ B1 ), and simultaneous binding to both sites elicits long bursts (τ B2 ). (Detailed arguments for this will be given in the Discussion.) To test this interpretation, we introduced a blocker of the αδ site, CTx (Sine et al. 1995; Cortez et al. 2007; Azam & McIntosh, 2009) and, on application of ACh, expected to isolate channel openings elicited by binding to the αγ site of the receptor. Figure 6 shows that, in the continued presence of CTx, ACh elicited short openings, whereas the long bursts of openings (τ B2 ) were absent. In Fig. 6A, with 1 μm ACh, many short openings are spread over the uppermost trace. However, although we might expect to see many more short openings with 100 μm ACh and a 2 s time calibration, Fig. 6B shows long stretches without any channel openings, intersected by dense groups of openings that represent clusters of short openings from receptor channels that pop up from desensitization and last until the receptor desensitizes again. Note that this refers to desensitization of monoliganded nachr with agonist bound only at the αγ site. Open time distributions in Fig. 7 demonstrate that the 5 μs component (τ O1 ) that shows up with ACh alone Figure 4. Concentration and agonist-dependent single openings and bursts A, mean durations of single openings (τ O1,red;τ O2, black) and bursts (τ B1,blue;τ B2, orange) for acetylcholine (ACh), carbamylcholine (CCh) and epibatidine (Ebd). For Ebd τ B2 is significantly shorter than for CCh (P < 0.05). B, relative frequencies of the components in (A) at different ACh, CCh and Ebd concentrations as percentages of the total area of the fitted probability density functions. The number of evaluated patches is as in Fig. 3.

9 J Physiol Different binding site contributions 2509 is absent with CTx (Fig. 7B). In addition, the major ACh-elicited component τ O3 of approximately 1 ms is absent with CTx. Figure 7B mainly presents a 0.4 ms component in the presence of CTx that may be an equivalent of the 0.35 ms component in Fig. 7A.Figure7B also shows a 0.13 ms component that is absent in Fig. 7A, but may be an equivalent of the τ O2 long single openings of Fig. 3A or of the unresolved short bursts. The distributions of shut times obtained with ACh show a dominant 3 μs component (Fig. 2D) that represents the gaps between openings in bursts. A number of longer shut time components occur much more rarely and mostly represent intervals between single openings from an unknown number of channels within the patch. When CTx was applied together with ACh, the 3 μs shut times prevailed (Fig. 8). However, the longer intervals between openings occurred much more often than without CTx. The shift in the proportions of short and long shut periods indicates that the 3 μs gaps became less prevalent. This is expected because CTx eliminated the long bursts of openings (Fig. 6), which are the main source of short gaps. However, CTx did not eliminate the 3 μsgaps,andit remains to be seen whether these gaps are associated with longer openings forming bursts. The associations of open and short shut times can be demonstrated in correlation diagrams. Figure 9A presents a colour-coded diagram obtained with 10 μm ACh. The yellow strip in the upper left of the diagram shows an association of very short shut times with long openings: the typical representation of bursts (Parzefal et al. 1998; Hallermann et al. 2005). In Fig. 9B, whichreferstothe presence of 1 μm ACh and 1 μm CTx, the yellow red patch in the upper left corner is less extensive than that with ACh alone, but assures the association of short gaps and relatively long openings (i.e. the survival of short bursts). A short burst is seen in Fig. 6B at the end of the high-resolution trace. Quantal synaptic currents in embryonic mouse muscles In nature CTx blocks synaptic currents. The effects of 300 nm CTxwererecordedin17hemidiaphragmsin mice aged 2 6 days, which still expressed the embryonic nicotinic receptor type (cf. Fig. 1A in Dudel & Heckmann, 2002). A representative result is given in Fig. 10A, which shows samples of qepscs, controls and after superfusion with CTx. Figure 10B presents the time course of the experiment plotting amplitudes, rise times and decay time constants of the qepscs. A 5 min application of 300 nm CTx reduced the amplitude of the qepscs to 25%. Along with the reduction in qepsc amplitude, qepsc rise time was increased and the qepsc decay time constant was shortened. This block was incomplete; CTx binds slowly and if its application were to be extended to, for instance, 15 min, qepscs would vanish. The recovery of the qepscs from the effects of CTx during more than 2 h of washing is not attributable to the unbinding of CTx from nachrs. In the illustrated experiment, CTx was applied only to the 10 μm diameter space below the recording electrode; during washing, CTx-free receptors diffuse into this space. The recovery was included in order to demonstrate that the recording remained intact. The binding of CTx to postsynaptic receptors seems to be irreversible because no recovery occurs when CTx is applied in the electrode and the bath. Eight experiments of this type gave essentially the same results. In parallel to the reduction in amplitude, the qepsc decay time constant was usually approximately halved. In three of the experiments the cholinesterase was blocked by diisopropylfluorophosphate (DFP; Sigma-Aldrich Corp., St Louis, MO, USA). This lengthened the decay time constant of control qepscs by about 50%, but did not interfere with the usual effects of CTx. Discussion Figure 5. The corrected mean opening length within bursts The real number of short gaps in bursts is approximated by probability density function fitting of the distribution of shut times (Fig. 2D). This is used to calculate the real mean opening length within bursts, μ open (see text). The μ open elicited by epibatidine (Ebd) is significantly smaller than that elicited by carbamylcholine (CCh) (P < 0.05; factor 0.5). ACh, acetylcholine. Desensitization Figures 3 and 4 show the results of recordings at agonist concentrations ranging from μm to 300 μm, which reflect a 100,000-fold increase in concentration. In comparison, the proportions of different components of channel open time changed by less than 10-fold. This is especially astonishing in the case of the long burst component, which should increase in number in proportion to the square of the concentration.

10 2510 P. Stock and others J Physiol The relatively weak concentration dependence of the occurrence of bursts has been noticed before in both the first bursts to be resolved (Colquhoun & Sakmann, 1985; Sine & Steinbach, 1986a, b) and high-resolution recordings (Hallermann et al. 2005). Most patch clamp recordings of agonist-elicited channel currents use a continuous superfusion of the agonist. In most receptor channels the elicited single-channel currents decrease in number with the duration of agonist application: the receptors desensitize. This did not concern the investigators: even at high agonist concentrations channel openings would pop up from desensitization, and the kinetics of openings of channels were assumed not to be affected by desensitization. However, this assumption may have been incorrect, at least for the rapidly and deeply desensitizing nicotinic ACh receptor. Desensitization has been studied by applying pulses of ACh to outside-out patches of denervated ( embryonic ) mouse muscle (Franke et al. 1993). A reaction scheme (3) has been derived and simulated (see Fig. 10 in Franke et al. 1993). Desensitization from the unliganded receptor R leads to desensitized state D that rapidly binds agonist and reaches A 2 D. A 2 O openings in long bursts are elicited from A 2 D with the low rate d 1, which is independent of agonist concentration. In longterm applications of agonists, the probability of open channels p O rises with ACh concentration only at ACh of < 1 μm, and remains constant at p O = 0.01 to p O = at higher concentrations. In the present study, we found, in recordings from patches with ACh and CTx, that desensitization also takes off from monoliganded nachrs, with agonist (ACh) only at the αγ site, which might affect the relative contributions of the different types of opening. Thus, desensitization may explain why at ACh concentrations above 0.1 μm the proportion of bursts does not rise clearly in Fig. 4B. A burst starting from A 2 D will be terminated with the dissociation of one agonist to AR, from which short openings and bursts as well as long bursts may be generated, finally decaying to R and D. CCh and Ebd may not desensitize as rapidly and deeply as ACh, but no quantitative data have been published, to our knowledge. Figure 6. Original current traces in the presence of acetylcholine (ACh) and α-conotoxin M1 (CTx) A and B are from different patches. The presentation is as in Fig. 2. Note the absence of long bursts. Single liganded receptors The main aim of the present study is to identify the binding sites that generate the different types of opening. Ebd is the only agonist able to elicit long single openings (τ O2 ) at a higher frequency than short openings (τ O1 )at very low agonist concentrations (Fig. 3). Ebd binds with higher affinity to the αγ site of the receptor and we may conclude that occupation of this binding site elicits τ O2 single openings. By contrast, the short openings (τ O1 ) should arise from binding to the αδ site, which prefers binding of ACh and CCh over the binding of Ebd. This interpretation is supported by the results of the blocking of the αδ site by CTx: single τ O2 openings and τ B1 bursts remain, whereas τ O1 openings and long bursts disappear (Figs 6, 7 and 9B). As mentioned earlier, the 65 μs component of Fig. 4 may also represent single openings, the duration of which may be shortened if we consider missed closings. This component should also originate from binding to the αγ site and may be associated with short bursts. The relative frequency of short bursts (τ B1 in Fig. 4) declines when agonist concentrations increase. Short bursts are therefore suggested to derive from a single-ligand receptor state. Short bursts, similarly to long single openings, are the preferentially elicited channel activations when Ebd is used in low concentrations (Fig. 4). Further, short bursts and long single openings remain when the αδ binding site is blocked by CTx. This

11 J Physiol Different binding site contributions 2511 gating behaviour suggests binding of the agonist to the αγ site. Short gaps and long bursts of openings We have recorded very short gaps of 3 μsdurationbetween the openings in bursts. τ C1 is invariant when elicited by different agonists, but this may reflect the limit of the present time resolution of 6 μs. In future recordings with better time resolution, agonist-specific differences in the durations of the gaps may be revealed. The proportion of long bursts rises in frequency with increased agonist concentration (Fig. 4B). This is to be expected if these bursts emerge from a doubly liganded receptor state (Colquhoun & Sakmann, 1985; Hallermann et al. 2005; Lape et al. 2008; Mukhtasimova et al. 2009; Sine, 2012). However, long bursts also occur at low agonist concentrations (Parzefall et al. 1998) and, very rarely, when no agonist is applied (Jackson, 1986; Auerbach, 2010). As discussed above, the increase in the frequency of long bursts with rising agonist concentration is limited by desensitization. It might be argued that long bursts may also arise from singly liganded receptors, but long bursts disappear when the αδ site is blocked with CTx (Figs 6 and 9B). Note that the duration of openings within bursts has been corrected (Fig. 5). The mean durations of the openings within long bursts (Colquhoun & Sigworth, 1995) differ significantly between Ebd and CCh (Fig. 5). Apparently the agonist affects the rate of its unbinding (see also Colquhoun & Sakmann, 1985). Short and long bursts do not differ significantly between the agonists ACh and CCh, but the durations of long bursts with the agonist Ebd are only almost half as long as those with the other agonists (Fig. 4). Part,butnotall,ofthisshorteningmayreflecttherelatively short openings in bursts elicited with Ebd. The ability to keep the receptor locked in its burst state appears to depend on the nature of the agonist. Although the durations of openings and bursts thus are influenced by the agonist, these effects are relatively small. As Fig. 4A shows, each of the four types of opening occupies a separate range of durations. The opening and closing dynamics of the channel-opening gate seem to be fixed and to be relatively little affected by the bound agonist. Figure 7. Open period distributions change with α-conotoxin M1 (CTx) A, control recording with 1 μm acetylcholine (ACh) fitted with three components and the parameters: τ O1,6.5μs and 33.9%; τ O2, 354 μs and 9.2%, and τ O3, 1300 μs and 56.9%. B, another recording with 1 μm ACh plus 1 μm CTx again fitted with three components and the parameters: τ O1,5.9μs and 3.7%; τ O2, 132 μs and 39.0%, and τ O3, 412 μs and 57.3%. Reaction scheme Figure 11 shows a reaction scheme that may largely represent the complexity of the observed data. The scheme does not contain a desensitization loop. Neither does the scheme generally include the additional shut states between agonist binding and channel opening suggested by Lape et al. (2008) and Mukhtasimova et al. (2009) (see also Sine, 2012). Rather, where we have evidence for a kinetically necessary intermediate state, we have inserted an R state. Reaction schemes for a bursting nicotinic receptor channel were proposed by Colquhoun & Sakmann (1985) with reference to the frog, and by Sine & Steinbach (1986a, b) with reference to a mammalian channel. The first agonist binding step led to short openings, and the second binding step to double-ligand receptors that open the channel in long bursts. As we observed, the binding steps to one or the other of the two binding sites have different results. Consequently, the sequence of binding, whether first to the αγ or first to the αδ site, becomes important. Therefore, we suggest two parallel reactions (Fig.11).Inthereactionshownintheupperdiagram, the agonist binds first to R γ from which the channel can open to a τ B1 short burst. After binding the second agonist to the still free αδ site, the channel opens to a long burst in O γδ A 2. The binding is reversible and if the agonist at

12 2512 P. Stock and others J Physiol the αδ site is lost first, the reaction reverts to R γ A. In the reaction shown in the lower lane, the agonist binds first at R δ, and short openings τ O1 can start from R δ A. Binding the second agonist again leads to O γδ A 2,thelongburst.The unbinding of an agonist proceeds in a manner analogous to that in the upper scheme, but the first unbinding to R γ AortoR δ A is independent of the sequence of previous bindings. As the starting points R and the end-points O γδ A 2 are identical, the two schemes could be joined at both ends, but the separate alternative schemes stress the importance of the initial binding. The scheme can account for the observed binding site-specificities of agonists: Ebd prefers binding at the αγ site and primarily elicits the short bursts O γ A (comprising τ O2 single openings and τ B1 short bursts in Fig. 4B) that decrease in frequency with rising agonist concentration. CTx at low concentration blocks the αδ site at R δ CTx and short single openings R δ O disappear, as do the long, bursts O γδ A 2. ACh and CCh bind primarily to R δ,and the τ O1 short openings amount to about 70% of the events in Fig. 4B (less with Ebd). Even with 300 μm ACh, the proportion of τ O1 openings (red) is more than twice as high as that of long bursts (orange). At this high ACh concentration R δ A seems to live long enough to generate several short openings τ O1 per long burst; therefore the opening rate to O δ A must be in the same range as the agonist binding rate. The high proportion of τ O1 short openings probably also indicates that the long bursts prefer to finally lose the agonist at the αγ site to result in more τ O1 openings from R δ A, and consequently the proportion of openings from R γ A will decrease. At 10 nm ACh almost all events are short openings τ O1 or short bursts τ B1, with a small percentage of long bursts. The openings from single-ligand receptors have been largely disregarded so far, but they are prominent even at high agonist concentrations. This has puzzled many investigators (e.g. Colquhoun & Sakmann, 1985; Sine & Steinbach, 1986a, b; Hallermann et al. 2005). Figure 8. Shut time distributions with α-conotoxin M1 (CTx) A, data from a recording with 1 μm acetylcholine (ACh) and 1 μm CTx fitted with five components and the parameters: 3.2 μs, 84.2%; 85.5 μs, 0.5%; 835 μs, 0.1%; 12.9 ms, 0.6%, and ms, 14.6%. B, data from a recording with 100 μm ACh and 0.3 μm CTx fitted with five components and the parameters: 3.4 μs, 53.2%; 715 μs, 2.6%; 8.58 ms, 40.8%; 56.3 ms, 1.4%, and 1018 ms, 2.0%. In addition to the predominant brief gaps, longer shut times are apparent. Figure 9. Correlations of open periods and adjacent shut times The degree of correlations increase from negative (blue) to maximally positive (red). A, measured in the presence of 10 μm acetylcholine (ACh): relatively long openings associated with very short closings represent bursts of openings. Short openings are positively correlated with a wide range of shut times and represent single openings. B, measured in the presence of 1 μm ACh and 1 μm α-conotoxin M1 (CTx). Note some bursts and the absence of a large number of single short openings.