Physostigmine modulation of acetylcholine currents in COS cells transfected with mouse muscle nicotinic receptor

Neuroscience Letters 401 (2006) 20–24

Physostigmine modulation of acetylcholine currents in COS cells transfected with mouse muscle nicotinic receptor Lucie Svobodov´a a , Jan Kr˚usˇek a,∗ , Tom´asˇ Hendrych b , Frantiˇsek Vyskoˇcil a,c a

Institute of Physiology, Academy of Sciences of the Czech Republic, V´ıdeˇnsk´a 1083, 142 20 Prague, Czech Republic b Institute of Physics of Charles University, Division of Biophysics, Ke Karlovu 5, 121 16 Prague, Czech Republic c Department of Animal Physiology and Developmental Biology, Charles University, Viniˇ cn´a 7, 120 00 Prague, Czech Republic Received 4 October 2005; received in revised form 13 February 2006; accepted 22 February 2006

Abstract Physostigmine (Phy), a reversible inhibitor of acetylcholine (ACh) esterase (AChE), may also act as a low potency agonist and a modulator of the nicotinic receptor. The actions of Phy on mouse muscle nicotinic receptors in the COS-7 cell line were studied by the patch-clamp technique. Currents were recorded in the whole-cell mode 3–7 days after cell transfection by plasmids coding ␣␤␥␦ combination of receptor subunits. The application of ACh to cells clamped at −10 mV produced inward currents which displayed desensitization. The application of Phy in concentrations up to 1 × 10−3 M did not give reliable specific whole-cell membrane responses. The application of Phy in concentrations of 10−6 –10−4 M together with ACh modulated the amplitude; accelerated desensitization of currents induced by ACh and increased the final extent of desensitization in a concentration-dependent manner. This finding is in contrast to the suppression and slowing down of desensitization by Phy and 1-methyl-galanthamine observed in Torpedo receptors. © 2006 Elsevier Ireland Ltd. All rights reserved. Keywords: Nicotinic ACh receptor; Eserine; Desensitization; Allostery

Nicotinic types of acetylcholine receptors (nAChR) are a part of the superfamily of ligand gated ionic channels which are comprised of five subunits forming the central ion channel. nAChRs are activated from agonist binding sites localized on the boundary between ␣ and neighbouring subunits (for review see [2]). In embryonic muscle receptors, the secondary parts of two binding sites are formed by ␦ and ␥ subunits. Ligands acting on these binding sites are either agonists such as acetylcholine [3], nicotine and carbamylcholine or competitive antagonists as +tubocurarine. The binding of two agonist molecules to the binding sites is necessary to induce or stabilize conformational changes of the channel complex, which lead to opening of the channel [4,28]. The open channel is permeable to monovalent cations and also, to a different degree, to calcium ions. In the prolonged presence of an agonist the receptor enters into an inactive desensitized state [9,17]. In nicotinic receptors, there are several different time scales of desensitization ranging from tens of mil-

∗

Corresponding author. Tel.: +420 29644 2551; fax: +420 29644 2488. E-mail address: [email protected] (J. Kr˚usˇek).

0304-3940/$ – see front matter © 2006 Elsevier Ireland Ltd. All rights reserved. doi:10.1016/j.neulet.2006.02.065

liseconds to minutes. Significant examples of structural parts of the neuronal receptor controlling fast and slow desensitization were summarized by Giniatullin et al. [6]. The desensitization phenomenon is also very sensitive to modulation by different influences such as the cellular environment [7], intracellular free calcium concentration, phosphorylation [8,23] or the type of agonist [6,24]. Physostigmine is a prototype of a novel class of nicotinic receptor ligands which are generally known as reversible inhibitors of acetylcholine esterase, but also display direct modulating, activating and blocking effects on nicotinic acetylcholine receptors (for review see [12,16,26]). Examples of other such ligands are galanthamine, codeine and serotonine. Phy modulates nicotinic responses by acting on binding sites that differ from acetylcholine binding sites that are presumably localized on the ␣ subunit [13]. The Phy binding site is supposed to be localized around the position of Lys-125 of the N-terminal extracellular domain [13] and could be blocked by the monoclonal antibody FK1 raised against the Torpedo nAChR. In various subtypes of neuronal nicotinic receptors, positive allosteric effects on channel activation were observed and it has been proposed that this modulation could be essential in long-

L. Svobodov´a et al. / Neuroscience Letters 401 (2006) 20–24

term beneficial effects of certain anticholinesterases used in the treatment of Alzheimer’s disease [12,15,16]. The aim of our work was to study direct action of Phy on mouse muscle embryonic receptors expressed in COS cells and to compare these effects with results on other types of nAChR. We focused on the description of Phy effects on ACh activated receptors and on the modulation of the ACh current amplitude (desensitization) in particular. The experiments were performed on COS cells, transfected by plasmids which code the appropriate subunits by a lipofectamine 2000 procedure (Gibco BRL). COS cells were cultivated in a minimal essential medium which was supplemented with 10% of fetal calf serum (both from Sigma Chemical, St. Louis, MO). More than 72 h after the transfection procedure, the whole-cell patch-clamp measurements were performed using an Axopatch 200A amplifier (Axon Instruments, Foster City, CA). Successfully transfected cells were detected by cotransfection with CD-4 coding plasmid and Dynabeads M-450 CD4 (Dynal Biotech, Norway) aggregation control. Fire-polished glass micropipettes with outer diameter of about 3 ␮m were filled with a solution of the following composition (in 10−3 M): CsCl 140, KCl 2.5, CaCl2 0.5, MgCl2 1, EGTA 5, HEPES–CsOH 10, pH 7.2. The resulting resistances of the microelectrodes were 3–10 M
. The cell bath solution contained (in 10−3 M): NaCl 160, KCl 2.5, CaCl2 1, MgCl2 2, HEPES–NaOH 10, glucose 10, pH 7.3. Solutions of drugs (all from Sigma Chemical, St. Louis, MO) were applied using a rapid microcomputer controlled perfusion system [19] consisting of an array of 10 parallel glass tubes each approximately 400 ␮m in diameter. The tubes were positioned and different solutions were switched on under microcomputer control [18,34]. A complete change of the solution around the cell could be carried out in 30–60 ms. For signal recording and evaluation of data an Axon Instruments Digidata 1320A digitizer and PCLAMP-9 software package (Axon Instruments, Foster City, CA) were used. The values are given as means ± S.E.M. ACh responses were fitted to two exponentials and a constant plateau was determined using Clampfit 9 programme (Axon Instruments, Foster City, CA). Relative changes of parameters were calculated as a ratio of the values in the presence and absence of Phy. The rapid application of ACh to successfully transfected cells (10–90% from total in the dish), evoked desensitizing inward ionic currents in the whole-cell mode of the patch-clamp technique, as high as 5 nA (Fig. 1A). Application of Phy alone, up to 1 × 10−3 M, did not evoke reliable whole cell currents (not shown). On the basis of these results ACh and Phy were applied concomitantly. To test the speed of Phy action, it was applied together with ACh into plateaus produced by long application of ACh (not shown). The effect of Phy was complete in less than 2 s. To prevent the complications due to different times of the onsets of the ACh and Phy effects, the experimental protocol started with 5 s preapplication of Phy followed by coapplication of identical concentration of Phy together with 1 × 10−5 M ACh for 10 s. The cells were clamped at −10 mV to minimize the contribution of voltage-dependent open-channel block in Phy-induced

21

Fig. 1. Effect of physostigmine on amplitude and time course of ACh-induced responses. (A) Current responses of COS cells expressing the mouse muscle embryonic nicotinic receptor were induced by rapid application of 10 ␮M ACh as indicated by bars. 5 s preapplication and presence of physostigmine in concentrations as indicated reduced the amplitudes of the current responses and changed their time courses. (B) Decrease of 10−5 M ACh response amplitudes with increasing Phy concentration. (C) Dependence of Phy 10−4 M effects on the concentrations of ACh. Phy was applied 5 s before start of ACh + Phy application.

ACh response decrease [31], which might occur at more negative potential. Preapplication and co-application of different concentrations of Phy (3 × 10−6 M to 1 × 10−3 M) together with 1 × 10−5 M ACh, the amplitude of the ACh response at a HP = −10 mV with IC50 = 7 × 10−5 M (Fig. 1B), changed the time courses

22

L. Svobodov´a et al. / Neuroscience Letters 401 (2006) 20–24

of the responses in a concentration-dependent manner. Phy 1 × 10−3 M blocked the responses to ACh 1 × 10−5 M almost completely (Fig. 1A). To exclude the possibility that Phy could act on the binding site identical with ACh binding site, the dependence of blocking effect of 1 × 10−4 M Phy on concentration of ACh was studied (Fig. 1C). The degree of block around 50% of control response was constant for higher concentrations of ACh (1 × 10−5 M to 1 × 10−3 M). In low ACh concentrations (1 × 10−6 M to 1 × 10−7 M) responses were potentiated by 1 × 10−4 M Phy up to 160–300%. This increase was accompanied by more pronounced desensitization which was almost absent in low ACh responses. The time courses of current decay of the control responses to 1 × 10−5 M ACh were fitted by two exponential components plus a constant A(t) = A1 e−(t/τ1 ) + A2 e−(t/τ2 ) + C, where A1 , τ 1 and A2 , τ 2 are amplitudes and time constants of two components of desensitization. C is the amplitude of the fraction of response that does not show desensitization, i.e. the final plateau. In the presence of 1 × 10−3 M Phy, some small responses can be fitted by only one exponential component before they reach a constant plateau; this may reflect a lower discrimination of the curve fitting. The time constants τ 1 , τ 2 , normalized amplitude components a1 = A1 /Amax and a2 = A2 /Amax of two exponentials and normalized plateau c = C/Amax varied markedly between cells and also between the responses of a single cell. In response to 1 × 10−5 M ACh, the faster component of the exponential decay has a time constant of 0.15–0.6 s and the slower component varied between 1.3 and 4.4 s. To minimize the influence of such variability, the relative changes induced by different concentrations of Phy were expressed as a(Phy) /a(control) and used for evaluation. The control values were estimated as the means of the two control responses before and after application of Phy. The normalized amplitude components a1 and a2 , slow and fast decay time constants τ 1 and τ 2 and the normalized plateau c were diminished in the presence of Phy in a concentrationdependent manner. With increasing Phy concentration, the slower component (a1 ) of desensitization was more sensitive to Phy than the faster component (a2 ) (Fig. 2A). Both time constants τ 1 and τ 2 also decreased with increasing Phy concentration (Fig. 2B) as did the relative value of the plateau level c (Fig. 2C). The greater sensitivity of the slower component to Phy together with the reductions of both time constants and reduction of the plateau were manifested as the faster and more complete desensitization of the Phy effect on the ACh responses. The mechanism of Phy action on nicotinic receptors in many preparations is considered to be a complex process including direct activation of receptors, open channel blockade and the allosteric influence of receptor kinetics. In the present study, we concentrated on the influence of Phy on desensitization of embryonic mouse muscle nicotinic receptors expressed in COS cells where the situation is evidently different. The mouse muscle embryonic receptors expressed in COS cells did not show the direct massive activation of the receptor by Phy detectable in the whole-cell mode of the patch-clamp, which has been observed in some invertebrate nicotinic receptors [33]. This finding does

Fig. 2. Physostigmine affects parameters of the ACh (10−5 M) desensitization. The ACh desensitization response was fitted by two exponential components A1 (slower) and A2 (faster) with time constants τ 1 and τ 2 and a nondesensitizing final plateau C. Normalized amplitude values (individual response divided by maximal value Amax ) a1 , a2 and c together with τ 1 and τ 2 were compared in control responses and in responses under defined concentrations of Phy. (A) Decrease of normalized amplitude a1 and a2 components of ACh (10−5 M) responses under the influence of Phy. Phy influence on a1 , a2 amplitudes of double exponential function. As shown, both components a1 and a2 decrease as Phy was increased. Note that the decrease of the a2 component started at Phy concentrations two orders of magnitude higher than those for a1 . (B) Decrease of τ 1 and τ 2 of the 10−5 M ACh responses under the influence of different concentrations of Phy. Both time constants were reduced starting from 1 × 10−5 M Phy. (C) Reduction of normalized steady state responses c to ACh (10−5 M) in the presence of Phy. The reduction of relative plateau c was explicit at Phy concentrations 1 × 10−5 M and higher. At 1 × 10−3 M Phy, all components of ACh responses were almost completely blocked.

L. Svobodov´a et al. / Neuroscience Letters 401 (2006) 20–24

not, however, exclude a direct activation of single channel openings that was reported at the single channel level in Torpedo [22], frog muscle receptors [27,25], rat hippocampal receptors [25] and in PC12 nicotinic receptors [30]. The Phy-induced decrease in the amplitude of the ACh response is voltage dependent and is presumably connected at least partially with open channel block by the positively charged Phy molecule [27,31]. To minimize the effects of an open channel block we clamped cells at −10 mV. At this potential, the voltage-dependent component of nicotinic response reduction is minimal [31] and recording from COS cells is more stable than at positive membrane potentials. Because the blocking effect of physostigmine on ACh response amplitude is not diminished by high concentrations of ACh (Fig. 1C) it could be concluded that the mechanism is not competitive but allosteric. The potentiating effects of Phy at low concentrations of ACh are similar to the partial agonistic effect of +tubocurarine observed in embryonic receptors [10,29]. It could be speculated that it depends on the presence of large population of nicotinic receptors with single ligands which are opened, by the action of physostigmine. One possible mechanism could be the binding of physostigmine on a vacant acetylcholine binding site and its partial agonistic effect similar to +tubocurarine. Another possibility could be an allosteric effect of Phy from another binding site, causing an increase of the open channel probability of single liganded receptor. The effect of Phy on the inactivation of receptors activated by higher concentrations of ACh could generally be characterized as acceleration and deepening of desensitization. In our twoexponential component model it was manifested as shortening of both time constants and a decrease in the amplitudes of the components, which effect was more pronounced in the slower exponential component. The response plateau was also reduced in size. When Ach is applied for several seconds, desensitization is, in fact, the product of two processes, i.e. an entering into and recovering from the desensitized state. As we found in a previous study [31], the recovery from desensitization was not influenced by the presence of 1 × 10−5 M Phy. Phy-induced deepening of desensitization is therefore apparently due to the acceleration of entry of nAChR into desensitized state(s) and not by slowingdown of the recovery after ACh application. The acceleration of desensitization, which is described here, is in contrast with the reported removal of desensitization in the Torpedo [14]. Decrease of the response plateau caused by Phy also contradicts the reactivation of Torpedo receptors by Phy which had already been desensitized by high concentrations of ACh [11]. This difference could be caused by the high variability of desensitization time constants ranging from milliseconds to seconds [5] and even hours [20] in different preparations and molecular variants of nAChR. The number of exponential components detected is also variable and may reach up to five [5]. Both time constants of nAChR desensitization observed in COS cells are relatively slow in comparison with receptors in C2C12 or TE671 cells (Kr˚usˇek and Filip, unpublished observation).

23

The different effect of Phy on nAChRs in the Torpedo on the one hand and in COS on the other hand can also be explained by an incomplete homology of mouse embryonic muscle receptor subunits and subunits of receptor expressed by the Torpedo. For example, the homology between the Torpedo ␥ subunit and mouse muscle ␥ subunit is only 56%. When the similarities of the molecular properties among a part of non-conserved aminoacids are considered, such a corrected homology takes 77% [1]. The most outstanding dissimilarities between amino-acid sequences of ␥ subunits are located in the C-terminus segment and in the intracellular loop between M3 and M4 segments, which might be potentially responsible for entry of the receptor into a desensitized state. Furthermore, the posttranslation modifications, assembling, glycosylation or phosphorylation cannot be excluded and may also contribute to differences between cell inhered Torpedo receptors and muscle receptors expressed by COS-cells. The physiological role of direct modulation of nAChR by physostigmine in synaptic transmission can be expected mainly at immature mammalian endplates where, unlike in adults, the synaptic region has less AChE [21,36]. It might be well limited in adult skeletal muscles because of much higher sensitivity of cholinesterases to physostigmine action [32]. Thus, at least in the mature synaptic cleft the remaining role of physostigmine is the prolongation of the ACh presence and action (cf. [35]). Acknowledgements This work was supported by grants no. A501 1411, A100110501, AV0Z 50110509, 305/03/H148 and LC554. Authors are grateful to Dr. Charles Edwards for critical reading the manuscript. References [1] E.X. Albuquerque, M.D. Santos, M. Alkondon, E.F. Pereira, A. Maelicke, Modulation of nicotinic receptor activity in the central nervous system: a novel approach to the treatment of Alzheimer disease, Alzheimer Dis. Assoc. Disord. 15 (Suppl. 1) (2001) S19– S25. [2] H.R. Arias, Topology of ligand binding sites on the nicotinic acetylcholine receptor, Brain Res. Brain Res. Rev. 25 (1997) 133–191. [3] S. Ciani, C. Edwards, The effect of acetylcholine on neuromuscular transmission in the frog, J. Pharmacol. Exp. Ther. 142 (1963) 21– 23. [4] D. Colquhoun, B. Sakmann, Fast events in single-channel currents activated by acetylcholine and its analogues at the frog muscle end-plate, J. Physiol. 369 (1985) 501–557. [5] S. Elenes, A. Auerbach, Desensitization of diliganded mouse muscle nicotinic acetylcholine receptor channels, J. Physiol. 541 (2002) 367–383. [6] R. Giniatullin, A. Nistri, J.L. Yakel, Desensitization of nicotinic ACh receptors: shaping cholinergic signaling, Trends Neurosci. 28 (2005) 371–378. [7] F. Grassi, E. Palma, A.M. Mileo, F. Eusebi, The desensitization of the embryonic mouse muscle acetylcholine receptor depends on the cellular environment, Pflugers Arch. 430 (1995) 787–794. [8] R.L. Huganir, P. Greengard, Regulation of neurotransmitter receptor desensitization by protein phosphorylation, Neuron 5 (1990) 555–567. [9] B. Katz, S. Thesleff, A study of the desensitization produced by acetylcholine at the motor end-plate, J. Physiol. 138 (1957) 63–80.

24

L. Svobodov´a et al. / Neuroscience Letters 401 (2006) 20–24

[10] C. Kopta, J.H. Steinbach, Comparison of mammalian adult and fetal nicotinic acetylcholine receptors stably expressed in fibroblasts, J. Neurosci. 14 (1994) 3922–3933. [11] J. Kuhlmann, K.O. Okonjo, A. Maelicke, Desensitization is a property of the cholinergic binding region of the nicotinic acetylcholine receptor, not of the receptor-integral ion channel, FEBS Lett. 279 (1991) 216–218. [12] A. Maelicke, E.X. Albuquerque, Allosteric modulation of nicotinic acetylcholine receptors as a treatment strategy for Alzheimer’s disease, Eur. J. Pharmacol. 393 (2000) 165–170. [13] A. Maelicke, T. Coban, A. Schrattenholz, B. Schroder, S. ReinhardtMaelicke, A. Storch, J. Godovac-Zimmermann, C. Methfessel, E.F. Pereira, E.X. Albuquerque, Physostigmine and neuromuscular transmission, Ann. N. Y. Acad. Sci. 681 (1993) 140–154. [14] A. Maelicke, T. Coban, A. Storch, A. Schrattenholz, E.F. Pereira, E.X. Albuquerque, Allosteric modulation of Torpedo nicotinic acetylcholine receptor ion channel activity by noncompetitive agonists, J. Recept. Signal Transduct. Res. 17 (1997) 11–28. [15] A. Maelicke, M. Samochocki, R. Jostock, A. Fehrenbacher, J. Ludwig, E.X. Albuquerque, M. Zerlin, Allosteric sensitization of nicotinic receptors by galantamine, a new treatment strategy for Alzheimer’s disease, Biol. Psychiatry 49 (2001) 279–288. [16] A. Maelicke, A. Schrattenholz, M. Samochocki, M. Radina, E.X. Albuquerque, Allosterically potentiating ligands of nicotinic receptors as a treatment strategy for Alzheimer’s disease, Behav. Brain Res. 113 (2000) 199–206. [17] L.G. Magazanik, F. Vyskoˇcil, Dependence of acetylcholine desensitization on the membrane potential of frog muscle fibre and on the ionic changes in the medium, J. Physiol. (Lond.) 210 (1970) 507–518. [18] M.L. Mayer, L. Vyklick´y Jr., The action of zinc on synaptic transmission and neuronal excitability in cultures of mouse hippocampus, J. Physiol. 415 (1989) 351–365. [19] M.L. Mayer, L. Vyklick´y Jr., G.L. Westbrook, Modulation of excitatory amino acid receptors by group IIB metal cations in cultured mouse hippocampal neurones, J. Physiol. (Lond.) 415 (1989) 329–350. [20] E.L. Meyer, Y. Xiao, K.J. Kellar, Agonist regulation of rat alpha 3 beta 4 nicotinic acetylcholine receptors stably expressed in human embryonic kidney 293 cells, Mol. Pharmacol. 60 (2001) 568–576. [21] B. Nystrom, Histochemical studies of end-plate bound esterases in “slow-red” and “fast-white” cat muscles during postnatal development, Acta Neurol. Scand. 44 (1968) 295–317. [22] K.O. Okonjo, J. Kuhlmann, A. Maelicke, A second pathway of activation of the Torpedo acetylcholine receptor channel, Eur. J. Biochem. 200 (1991) 671–677. [23] K. Paradiso, P. Brehm, Long-term desensitization of nicotinic acetylcholine receptors is regulated via protein kinase A-mediated phosphorylation, J. Neurosci. 18 (1998) 9227–9237.

[24] K.G. Paradiso, J.H. Steinbach, Nicotine is highly effective at producing desensitization of rat alpha4beta2 neuronal nicotinic receptors, J. Physiol. 553 (2003) 857–871. [25] E.F. Pereira, M. Alkondon, T. Tano, N.G. Castro, M.M. Froes-Ferrao, R. Rozental, R.S. Aronstam, A. Schrattenholz, A. Maelicke, E.X. Albuquerque, A novel agonist binding site on nicotinic acetylcholine receptors, J. Recept. Res. 13 (1993) 413–436. [26] E.F. Pereira, C. Hilmas, M.D. Santos, M. Alkondon, A. Maelicke, E.X. Albuquerque, Unconventional ligands and modulators of nicotinic receptors, J. Neurobiol. 53 (2002) 479–500. [27] K.P. Shaw, Y. Aracava, A. Akaike, J.W. Daly, D.L. Rickett, E.X. Albuquerque, The reversible cholinesterase inhibitor physostigmine has channel-blocking and agonist effects on the acetylcholine receptor-ion channel complex, Mol. Pharmacol. 28 (1985) 527–538. [28] S.M. Sine, J.H. Steinbach, Activation of a nicotinic acetylcholine receptor, Biophys. J. 45 (1984) 175–185. [29] J.H. Steinbach, Q. Chen, Antagonist and partial agonist actions of dtubocurarine at mammalian muscle acetylcholine receptors, J. Neurosci. 15 (1995) 230–240. [30] A. Storch, A. Schrattenholz, J.C. Cooper, E.M. Abdel Ghani, O. Gutbrod, K.H. Weber, S. Reinhardt, C. Lobron, B. Hermsen, V. Soskic, E.F.R. Pereira, E.X. Albuquerque, C. Methfessel, A. Maelicke, Physostigmine, galanthamine and codeine act as ‘noncompetitive nicotinic receptor agonists’ on clonal rat pheochromocytoma cells, Eur. J. Pharmacol. 290 (1995) 207–219. [31] L. Svobodov´a, J. Kr˚usˇek, T. Hendrych, F. Vyskoˇcil, Allosteric modulation of the nicotinic acetylcholine receptor by physostigmine, Ann. N.Y. Acad. Sci. 1048 (2005) 355–358. [32] T. Thomsen, B. Kaden, J.P. Fischer, U. Bickel, H. Barz, G. Gusztony, J. Cervos-Navarro, H. Kewitz, Inhibition of acetylcholinesterase activity in human brain tissue and erythrocytes by galanthamine, physostigmine and tacrine, Eur. J. Clin. Chem. Clin. Biochem. 29 (1991) 487– 492. [33] I. van den Beukel, R.G. van Kleef, R. Zwart, M. Oortgiesen, Physostigmine and acetylcholine differentially activate nicotinic receptor subpopulations in Locusta migratoria neurons, Brain Res. 789 (1998) 263– 273. [34] L. Vyklick´y Jr., M. Benveniste, M.L. Mayer, Modulation of N-methyld-aspartic acid receptor desensitization by glycine in mouse cultured hippocampal neurones, J. Physiol. 428 (1990) 313–331. [35] F. Vyskocil, E. Nikolsky, C. Edwards, An analysis of the mechanisms underlying the non-quantal release of acetylcholine at the mouse neuromuscular junction, Neuroscience 9 (1983) 429–435. [36] F. Vyskocil, G. Vrbova, Non-quantal release of acetylcholine affects polyneuronal innervation on developing rat muscle fibres, Eur. J. Neurosci. 5 (1993) 1677–1683.