Muscarinic Receptors in Developmental Axonal Competition at the Neuromuscular Junction

In recent years, we have studied by immunohistochemistry, intracellular recording, and western blotting the role of the muscarinic acetylcholine receptors (mAChRs; M1, M2, and M4 subtypes) in the mammalian neuromuscular junction (NMJ) during development and in the adult. Here, we evaluate our published data to emphasize the mAChRs’ relevance in developmental synaptic elimination and their crosstalk with other metabotropic receptors, downstream kinases, and voltage-gated calcium channels (VGCCs). The presence of mAChRs in the presynaptic membrane of motor nerve terminals allows an autocrine mechanism in which the secreted acetylcholine influences the cell itself in feedback. mAChR subtypes are coupled to different downstream pathways, so their feedback can move in a broad range between positive and negative. Moreover, mAChRs allow direct activity-dependent interaction through ACh release between the multiple competing axons during development. Additional regulation from pre- and postsynaptic sites (including neurotrophic retrograde control), the agonistic and antagonistic contributions of adenosine receptors (AR; A1 and A2A), and the tropomyosin-related kinase B receptor (TrkB) cooperate with mAChRs in the axonal competitive interactions which lead to supernumerary synapse elimination that achieves the optimized monoinnervation of musculoskeletal cells. The metabotropic receptor-driven balance between downstream PKA and PKC activities, coupled to developmentally regulated VGCC, explains much of how nerve terminals with different activities finally progress to their withdrawal or strengthening.


Introduction
During the nervous system development, an overproduction of neurons and synapses generates an extensive connectivity that is corrected by an activity-dependent reduction that refines the specificity of the neuronal circuit [1][2][3][4]. The final specificity is attributable to the appropriate matching between the origin and the target of the nerve fibers. Thus, Hebbian competition between nerve endings with different activities leads to the elimination or strengthening of their synapses [5][6][7]. This developmental synaptic elimination occurs throughout the nervous system, representing a basic mechanism of sinaptogenesis [2,[8][9][10][11][12][13][14].
During development, skeletal myocytes start polyinnervated by several axons [15][16][17], but after their competitive interactions, neuromuscular junctions (NMJ) finally become innervated by only one axon [6,13,14,18]. There are many reviews about synapse elimination mainly focused on the NMJ [1,[18][19][20][21][22][23][24][25][26][27]. Several relevant cues of the molecular and cellular mechanisms involved in the elimination of supernumerary nerve terminals have been investigated and collected in the cited reviews. The non-competitive reduction in the number of nerve endings which a motoneuron can support has been studied and characterized. Moreover, it seems also that there is an initial and continuous exchange of the appropriate nerve terminals to produce homogenous fiber-type motor units. At this motorneuron level, [22,28] show that neuromuscular synapse elimination was accelerated in mutant mice lacking connexin 40, a developmentally regulated gap junction protein, expressed in motor and other spinal neurons, to facilitate electrical coupling.
A major factor, however, is the level of activity of each nerve terminal in a poliinnervated NMJ, and, on the whole, the axonal loss is retarded at low levels of activity and accelerated at increased levels. During the axonal competition, the various nerve endings in a NMJ have a mutual influence on one another and on the postsynaptic muscle cell and the terminal Schwann cell. In adults, terminal Schwann cells sense the release of ACh and ATP from the nerve through M 1 and A 1 receptors and in turn influence transmitter release [29,30]. Terminal Schwann cells have been involved also in axonal competition during development. Glial activity induces synaptic potentiation (through presynaptic adenosine 2 A receptors) of strong input in dual junctions but not in weak terminals [31]. Roche et al., [32] using mice lacking Nfasc155, a glial protein detected a delay in postnatal synapse elimination at the NMJ. Moreover, neuregulin-1 signaling between terminal axons and glia during development influences glial cell activation and interposition between the terminal and muscle [23] affecting axon loss. Recently, Jung et al. [33] proposed a model that reproduced synapse elimination showing that synapse elimination is accelerated by increased areas of teloglial cell vacancies. Activity-induced changes in the stability of the muscle cell postsynaptic nAChRs can contribute to reduce the synaptic efficacy of ruled-out nerve terminals [34][35][36]. Several pre-(calcitonin gene-related peptide, CGRP, [37]) and postsynaptic-derived signals (BDNF, [38,39]) can also influence supernumerary axonal loss by rewarding or punishing certain nerve terminals. Je et al. [40] using genetic manipulations and pharmacological studies show the involvement of endogenous proBDNF and mBDNF in synapse elimination.
Several presynaptic receptors-mainly muscarinic acetylcholine (ACh) autoreceptors (mAChR), adenosine autoreceptors (AR), and tropomyosin-related kinase B receptor (TrkB)-allow the multiple developing nerve terminals to communicate in the competition that leads to synapse loss in the NMJ [41][42][43]. This communication can occur directly or with the intermediation of the postsynaptic or Schwann cell components of the tripartite synapse. In particular, mAChRs in the motor terminals seem to permit direct competitive interaction between multiple nerve endings through differential activity-dependent ACh release in the shared narrow developing synaptic cleft. The more active endings may directly punish the less active ones and reward themselves [24,34,[43][44][45], and asynchronous activity seems to optimize this interaction to promote synapse elimination [46].
The presence of mAChRs in the motor nerve terminal presynaptic membrane is a clear example of an autocrine mechanism in which a secreted product of a cell can externally influence itself as a feedback modulation both during development and in the adult. Interleukin-2 is another example, being produced by and acting on T lymphocytes themselves (in addition of its paracrine action on target macrophages and other immune cells) [47,48]. In the case of the cholinergic autoreceptors, the presence of several muscarinic subtype molecules coupled to different downstream pathways can move the autoregulation of the neurosecretion in a broad range between positive and negative effects [49][50][51][52][53]. Because of all these reasons, mAChRs are a relevant component of the complex regulation through pre-and postsynaptic activities of the supernumerary synapse elimination conducting to the optimized monoinnervation of the musculoskeletal system.
We have been working in developmental axonal competition and synapse elimination since the late 1970s (for instance [54,55]), and the molecular mechanisms are far from being fully elucidated today though many questions have been answered [4,6,18]. In recent years, we have studied by immunohistochemistry, intracellular recording electrophysiology, western blotting, subcellular fractionation and co-immunoprecipitation, the involvement of the mAChR subtypes in the mammalian NMJ functionality during development and in the adult [3,[56][57][58][59][60][61]. In the adult, we characterized how M 1 and M 2 mAChRs regulate the PKA subunits (catalytic and regulatory), the PKC (PKCβI and ε isoforms), and their exocytotic targets (Munc18-1, SNAP-25, and MARCKS phosphorylation) showing a codependent balance between muscarinic auto receptors which orchestrates transmitter release regulation [62,63]. We analysed also the involvement of altered metabotropic receptor signaling in amyotrophic lateral sclerosis SOD1-G93A mice [64]. Here, we review previously published data to evaluate the relevance of mAChRs during development in synapse elimination and their crosstalk with other receptors, the downstream kinases, and the targeted voltage-gated calcium channels (VGCC).
Our contribution to the understanding of the synapse elimination process can be summarized in this review: (1) the characterization of the presumably most relevant membrane receptors involved, the mAChRs and adenosine receptor subtypes (both allowing a direct paracrine influence between activity-different neighbor competing nerve endings) and the trophic factor receptor TrKB allowing the retrograde influence through BDNF (2) the characterization of the changes in the downstream PKC/PKA ratio (cPKCβI and nPKCε favor axonal retraction) in the competing nerve terminals as a relevant point of the process, and (3) the specific involvement of several VGCC in determining the transmitter release capacity and the retraction of the nerve endings.

mAChR Subtypes in Transmitter Release During Development
We studied the effect of mAChR subtype modulation-mainly in the rodent Levator auris longus muscle (LAL)-by comparing neurotransmission at the newborn stage (P7-P9) versus in the monoinnervated mature NMJs (P30). Axonal elimination almost occurs during the first 2 weeks after birth, although a residual 3% of multiply innervated synapses remains at the end of the first month. The period P7-P9 corresponds to the middle of the axonal loss process and the nerve terminal elimination coincides with the morphological maturation of the postsynaptic component on the NMJ. At birth, nearly all of the NMJs (96%) in the LAL muscle are innervated by more than one axon, and at the end of the first week, the percentage of polyinnervation is reduced to nearly 50%. Newborn P7-P9 neurotransmission study includes developing monoinnervated NMJ and developing polyinnervated ones, mainly dual junctions passing the last process of axonal competition. In the latter, strong and weak nerve terminals can be identified by their evoked endplate potentials (EPP), with the least quantal content originating from the weak terminal in dually-innervated junctions [39,59,[65][66][67][68]. Intracellular recordings using muscarinic agonists and blockers show that some of the mAChR subtypes (M 1 , M 2 , and M 4 ) influence ACh release both in developing [58][59][60][61] and in adult NMJs [41,69,70]. Table 1 shows the list of substances used in the described experiments.
In the adult, the M 1 receptor increases ACh release because its selective block with pirenzepine [PIR] or muscarinic toxin-7 [MT-7] reduces it. On the contrary, the M 2 receptor reduces ACh release because its selective inhibition with methoctramine [MET] or AFX-116 increases it [41,52,58,71]. The M 3 (4-DAMP) and M 4 (tropicamide [TRO] and muscarinic toxin-3 [MT-3]) blockers do not produce any effect in transmitter release. Thus, in the adult, M 1 and M 2 receptors influence ACh secretion in a positive and negative feedback, respectively.
However, during development, all the selective M 1 and M 2 blockers were tested to reduce the release both in the recently monoinnervated NMJ and in the strongest endings of dual junctions ( Fig. 1) that are still in the competition at P7-P9. Interestingly, the effect of any M 1 antagonist was not additive with the effect of any M 2 antagonist, suggesting the operation of the same mechanism in both cases (i.e., the effect of PIR is completely prevented by a preincubation with MET and viceversa). This suggests the commitment of all muscarinic pathways to promote neurotransmission in these nerve endings during maturation. Nevertheless, in the weakest nerve contact in dual junctions, only the M 2 blockers reduce release whereas M 1 and M 4 blockers increase the EPP ( Fig. 1) [39,[58][59][60][61]. We sequentially added PIR and TRO to the muscle and found that their respective effects on the weak nerve terminals were additive, showing that M 1and M 4 -mediated pathways are different in these endings. However, like in the strong and in the monoinnervated junctions, the effect of the M 1 and M 2 drugs is still mutually exclusive and not additive. Thus, the two additive M 4 and M 1 mechanisms result in a powerful inhibitory modulation that probably overcomes the M 2 mechanism, which enhances neurotransmission in all kinds of endings during development. Finally, as stated, in the adult monoinnervated NMJs, M 2 and M 1 receptors change their coupling to regulate neurotransmission by negative and positive feedback, respectively.
In summary (Fig. 1), an advantage of the more differentiated nerve terminals seems to be the commitment of all mAChR subtypes (M 1 and M 2 in this case) to enhance ACh release, and using this autocrine mechanism, the strongest endings may reinforce themselves to win the competition. However, the weakest axon may be negatively influenced by ACh release from the strongest axons through the M 1 and M 4 subtype pathways. The described evidence argues in favor of the relevance of local factors explaining, in fact, how an axon that fails at one muscle endplate can win the competition at another [72]. So, a developmentally regulated specific expression of the mAChR subtypes seems to be a relevant mechanism in synaptogenesis.

mAChR in the Physical Withdrawal of Supernumerary Axons
How the described evidence of the developmental muscarinic modulation of ACh release can be related with supernumerary axon elimination from the NMJ? We favor the hypothesis of the final strengthening and consolidation of the strongest endings in dual junctions because their molecular and functional similarity with the solitary endings in the most mature NMJ. Thus, both M 1 and M 2 receptors are coupled to potentiate release in these endings along with the expression in these nerve terminals of a more differentiated VGCC stoichiometry and serinethreonine kinases coupling to ACh release (see later). To investigate this, we made axonal counts in confocal LAL preparations from B6. Cg-Tg (Thy1-YFP) 16 Jrs/J mice that express spectral variants of GFP (yellow-YFP) at high levels in motor neurons [43]. Muscles were processed to detect also the postsynaptic nicotinic acetylcholine receptors (nAChRs) with TRITC-α-BTX (Bungarotoxin). We counted the percentage of singly-, dually-, and triply-(or more) innervated synapses at P7, P9, and P15 after 2 (days 5, 6), 4 (days 5-8), and 10 (days 5-14) subcutaneous applications over the LAL muscle surface of the considered muscarinic substance [3,56,57].
In P7 mice, we observed that when M 1 or M 4 receptors are selectively blocked by PIR or MT3, axonal loss is accelerated (but not when M 2 is blocked with MET). Thus, at P7 (considering the effect on neurotransmission of the muscarinic receptors, see Fig. 1), M 2 favors ACh release and possibly the competitive force, related with more transmitter release, in all axons although not affecting the axonal elimination rate. However, M 1 increases release in the strong axon and decreases it in the weak one (together with M4 in this case), and these tonic effects resulting in a delay in axon loss (evidenced by a conspicuous acceleration when M 1 or M 4 receptors are selectively blocked) [43]. We do not know which one can be the prevalence of any of these muscarinic receptors in the different nerve endings at this period, but it is conceivable that mAChR subtypes would participate and even be involved in determining competitive interactions rather than speeding up axonal elimination around P7 [3]. Thus, it appears that at P7, mAChRs-mediated competitive axonal interactions (and also interactions mediated by AR and TrkB receptors-see later) are taking place with the result of an initial delay in synapse elimination because effective axon loss is not yet occurring in most synapses at this moment.
Two days later (P9), the continued exposition to PIR or MET (but not to the M 4 blocker MT3) for 4 days results in a clear delay in axon loss. This indicates that both M 1 and M 2 receptors acquire during this period the role of promoting the full sequence of axonal elimination. Interestingly, MET has a greater ability than PIR to delay the final monoinnervation, indicating the powerful effect of the M 2 on axon loss at P9, probably potentiating the strongest nerve endings [43]. Joint inhibition of M 2 and M 1 (MET + PIR) pathways show that their effect on axonal elimination is not additive, suggesting a shared downstream mechanism at this developmental stage (see below) and the commitment of all muscarinic receptors to promote axon elimination. In the adult, mAChRs show some G protein promiscuity [73,74] suggesting that M 2 -M 1 shared developmental mechanism may relate with G protein sharing. The effect on the ACh release of these two receptors may reinforce the increasingly stronger endings and be detrimental to the weak ones. The presence of the M 4 mAChR subtype in the weakest ending at P7 and their functional disappearance at P9 along with the shift of the M 1 function (from ACh release reduction in the weakest endings to favoring ACh release in the endings that become stronger) can be important changes in synapse elimination. Interestingly, the progressive change in M 4 and M 1 function during NMJ maturation coincides with the slightly later shift of the M 2 function that will change to negatively modulate ACh release around P15 and for the rest of the adult stage [43]. Therefore, the functional shift of all mAChR types during the maturation of the neuromuscular synapses argues in favor of their relevance in this process. The shift mechanism must be able to explain how M2 changes from positive to negative action on ACh release whereas M 1 changes reciprocally. The possibility that a change in the expression (protein level) of the M 1 and M 2 mAChRs themselves may contribute to explain their functional change needs to be seriously considered. However, the movement of both M 1 and M 2 receptors between the extreme positive and negative influences on ACh release during maturation raises some concerns about the possible relevance of the protein level change to fully explain these extreme changes in their downstream coupling. Alternatively, as stated, these metabotropic receptors are GPCR, and a developmentally regulated displacement of their coupling between Gs and Gi proteins seems to be an attractive hypothesis. This mechanism may be facilitated because several GPCR function within lipid raft plasma membrane microdomains, which may be important for regulating their signal transduction. In a previous study in the mature NMJ, however [75], we show that the disruption of lipid rafts (methyl-beta-cyclodextrin, 2%) does not change the normal coupling and mutual relations of adenosine receptors and mAChRs on ACh release.

Relation of the mAChR with AR (A 1 , and A 2A ) and TrkB Receptor on Developmental Axonal Elimination
Even the continued application of M 1 and M 2 inhibitors cannot stop axonal loss, which is completed around P15 [43].
This suggests the complex involvement of multiple other pathways including postsynaptic-derived factors [39,43,56,57]. We found that, in addition to the presynaptic mAChR M 1 , M 2 , and M 4 , at least adenosine receptors (AR; A 1 and A 2A ) and the tropomyosin-related kinase B receptor (TrkB), cooperate in synapse elimination [56,57]. It seems that this multiple signaling would define the conditioning factors of the axonal competition and thus, the final fate of the individual nerve terminals. However, the achievement of the monoinnervated NMJ may be constitutively regulated.
Concerning to transmitter release, AR is present in the motor terminals of the newborn and adult NMJs [76,77] and, during development, released adenosine from different components of the synapse may activate both A 1 R and A 2A R and have a facilitatory action on ACh release. Unfortunately, we do not know the specific involvement of these receptors in the strong and weak nerve endings. Neurotrophins and their receptors are also expressed in both development and adulthood [78][79][80][81][82][83]. Low doses of BDNF rapidly induce a TrkB-dependent potentiation at developing NMJs in culture [84] and, in ex vivo developing NMJ, BDNF increases ACh release in both the weak and strong endings around P7 [68]).
Concerning to developmental axonal elimination, specific inhibitors reveal that both AR delay axonal loss at P7 but accelerate it at P9. This effect is similar to that of mAChRs. The BDNF-TrkB pathway also plays a biphasic role because BDNF initially delays elimination and subsequently accelerates it at P9 [3]. Thus, several metabotropic receptors overlap and share the common function of modulating a major mechanism of synaptogenesis as can be the definition of the final matching of the synaptic partners. Interestingly, for all receptors, an initial delay in axonal elimination observed at P7 is followed by the acceleration at P9, pointing to the existence of a multifactorial and redundant mechanism aimed at ensuring the specific NMJ monoinnervation. For instance, all receptors (except M 4 ) directly accelerate axonal loss at P9. Ranked according to their importance (from more to less), these are M 2 -M 1 -A 1 -A 2A -TrkB [3].
Given the observed downstream shared effect of these receptors, we simultaneously applied two selective antagonists to reveal the cooperation between mAChR, AR, and TrkB receptors and the possible additive (synergistic) or occlusive (antagonic) crosstalk between them (Fig. 2) [4,43,56,57]. Studying the adult NMJ, we identified several links between purinergic receptors and mAChRs and found that the functional integrity of mAChRs coupling to the neurotransmission depends on normal purinergic receptors operation. This indicates the clear interaction between both receptor families in the adult [75]. In the newborn, the main results show a synergistic role of the M 1 mAChR, which potentiates the effect of both AR (A 1 , 58% and A 2A 36%) and TrkB (25%) on axonal elimination. On the other hand, though the M 4 subtype is not directly involved in axonal loss as previously stated, it strongly potentiates the effect of AR (A 1 , 33% and A 2A 32%) and TrkB (23%) thus acting similarly as to the M 1 receptor. Interestingly, a comparable effect of M 1 and M 4 is observed on the ACh release capacity of the weakest nerve terminals in dual junctions as shown above (see also Fig. 1). However, as previously stated, M 2 has the most powerful effect on axon loss and the inhibition of both AR or TrkB receptor does not affect their function. When the TrkB inhibitor TrkB-Fc is associated with one of the AR inhibitors (DPCPX for A 1 , or SCH58261 for A 2A ), the final effect is just the same as the individual effect of one of them on axon loss. When both ARs are blocked simultaneously, occlusion is complete, and the final result is no different from that of the untreated control [57].
Thus, taking these data into consideration, we represented in Fig. 2 the observed relations of the considered receptors that modulate developmental supernumerary axonal loss. In Fig. 2, the receptors that are related with a blue link (M 1 /M 4 with M 2 , A 1 with TrkB, and TrkB with A 2A ) seem to share the same pathway because their dual inhibition produces the same effect as their individual inhibition over axon loss. On the other hand, M 1 /M 4 show a synergistic additive behavior (green link) with TrkB, A 1 , and A 2A . Finally, both AR, A 1 , and A 2A show an antagonistic relationship and are mutually occlusive (red link). All these receptors are involved in promoting supernumerary axonal elimination, and we investigate their downstream links.

mAChR Coupling to Serine Kinases During Synapse Elimination
The mAChR downstream signaling converges in intracellular effector kinases, mainly the serine-threonine protein kinases A and C (PKA and PKC), which phosphorylate targets involved in synaptic function and axon loss. Receptors and kinases may coordinately regulate the developmental synapse elimination. Generally, M 1 operates by stimulating PKC whereas M 2 and M 4 inhibit PKA. It is known that, in most cells, A 1 , M 1 , and TrkB operate mainly by stimulating the phospholipase C (PLC) and, therefore, PKC and the inositol triphosphate (IP 3 ) pathway, whereas A 2A , M 2 , and M 4 inhibit the adenyl cyclase (AC) and PKA pathway [50,51,85,86].
At the left of Fig. 2, we represented the receptors coupled to activate PKC (M 1 , A 1 , and TrkB-we added M 4 due to a connection with M 1 -), and on the right, those that downregulate PKA (M 2 and A 2A ), and we investigated the hypothesis of the reciprocal involvement of these kinases in synapse elimination. In dually innervated, developing NMJs, the block of PKC (for instance, with calphostin C or chelerytrine) increases ACh release from the weakest nerve terminals (roughly 80%) but does not change ACh release from the strong nerve terminal or even from the more mature monoinnervated junctions. Moreover, after blocking PKC, the mean number of functional axon terminals per synapse increases by about 50%, indicating some recruitment of silent synapses at this time that are probably in the process of disconnection [39, 58-60, 67, 87]. Thus, PKC is involved in reducing neurotransmission in certain weak nerve endings which may facilitate axonal elimination. Recently, we found [88] that PKC favors axon loss through cPKCβI and nPKCε isoform activity (as judging by the effect of their general [Bry-1 or PMA] and specific activators [dPPA, FR236924] and inhibitors [βIV 5-3 and εV 1-2 ] respectively) whereas PKA-I and II activity (as judging by the effect of their specific blockers [H-89, Rp8-Br, and Rp-cAMPs] and activator [Sp8Br], respectively) delay axonal loss in P9 mice. Furthermore, no significant differences exist between the effects of PKA activators and PKC inhibitors, or between PKA inhibitors and PKC activators, on changing axon loss rate [25]. Moreover, a similar level of PKA inhibition and PKC potentiation (mainly of the cPKCβI and nPKCε isoforms that are strictly localized on the presynaptic site [89][90][91]) seems to be required to advance in axonal loss, clearly suggesting the complementarity of these kinases. On the contrary, the increase of the PKA activity, the reduction of the PKC activity, or, in most cases, both situations simultaneously can reduce synapse elimination [57]. Thus, a metabotropic receptor-driven balance between PKA and PKC activities seems to be involved in synapse elimination and axonal withdrawal as represented in Fig. 2.
It is known that reduction of the postsynaptic activity or contraction results in a delay in synapse elimination during NMJ development [18,[92][93][94][95]. In line with this, we made experiments blocking the muscle cell's contractile activity with μ-conotoxin GIIIB which blocks muscle cell sodium channel but preserves neurotransmission because does not influence the nAChR [26]. Accordingly, incubation with μ-conotoxin GIIIB also results in a delay in axon loss. Thus, a contractile activity-related retrograde influence from the postsynaptic site may contribute to the synapse elimination. The simultaneous application of one presynaptic cPKCβI or nPKCε activator and μ-conotoxin GIIIB fully prevents the postsynaptic contraction block effect on axon loss. Thus, the axonal loss can be altered by acting directly in presynaptic targets (and receptors like mAChRs). Possibly, the above-cited presynaptic PKCs may be modulated by retrograde control (for instance, through BDNF production). This argues in favor of a complex regulation through preand post-synaptic activity of the serine-threonine kinases as mediators of the synapse elimination. The regulation of these kinases by mAChR and neurotrophic receptors affecting their phosphorylating activity on targets of the exocytotic vesicular release apparatus (as synapsin I and the SNARE/ SM proteins Munc18-1 and SNAP-25) has been described by us in the adult NMJ [96,97].

mAChR Coupling to Calcium Channels During Synapse Elimination mAChR and Calcium Channels in Transmitter Release During Development
During development, P/Q, N, and α1D-L subtypes of the VGCC are present in the nerve terminals on the LAL muscle. The protein expression of all these channels increases during the development P5-P7 time period. Western blots at P30 show that the P/Q level is at its highest whereas α1D-L and N channel proteins stabilize at a lower level [26].
In dually innervated fibers during NMJ maturation (P7-P9), the block of any of these three VGCC reduces about 2/3 of the EPP produced by the strongest ending [65,66], indicating the multichannel dependence of calcium entry to promote ACh release in these endings. In the early monoinnervated endplates, the P/Q-type channel blocker ω-Aga-IVA and the N-type blocker ω-CgTx-GVIA still reduce the EPP amplitude (~ 80% and ~ 60%, respectively) whereas the L-type blocker nitrendipine does not anymore. Finally, in the adult (at P30), only the P/Q-type VGCC functionally persists being the only one that, when blocked, strongly inhibits ACh release [69]. However, in the weak ending of dual NMJs, the block of any VGCC channel results in an increase of the size of the evoked EPP, indicating that a part of the calcium entry through all channels can negatively influence transmitter release and even may contribute to disconnect these endings [65,66]. Figure 3 shows that in the strongest endings, there is a differential coupling of the calcium channels with the M 1 and the M 2 receptors. M 1 receptors need the P/Q-and the L-types VGCC whereas the M 2 effect needs the P/Q-and Fig. 3 Differential coupling of VGCC and mAChRs in the strong and weak endings in a common endplate. The effect of the mAChR and VGCC on ACh release is shown in green (potentiation) or red (depression). The links between receptors and channels indicate the mutual dependence between these molecules to produce the final effect. In the strongest nerve terminal (shadowed in green), M 1 receptors need the P/Q-and the L-type VGCC, whereas the M 2 need the P/Q-and N-type VGCC to potentiate ACh release. The weak nerve ending (in yellow) is represented with the mAChR (M 1 , M 2 , and M 4 ) and the VGCC (P/Q, N, and L) subtypes that are operative in these endings. The M 2 function (also potentiation in these endings), similarly depends on P/Q-and N-type channels, but not on the L-type channel. A leading role for the L channel in axonal loss seems to emerge because of their unique coupling to M 1 and M 4 mAChRs, which clearly reduces neurotransmission in the weak nerve terminals presumably in process of elimination. Moreover, the coupling of all muscarinic and metabotropic receptors studied to promote axonal elimination at P9 had a multichannel dependence (P/Q-and L-type) with a relevant role of the L channel N-types VGCC. As previously stated, in these strongest ending in dually-innervated synapses, both M 1 and M 2 mAChR have an ACh release potentiating effect (see Fig. 1). In the monoinnervated junctions, the ACh release potentiating effect of both M 1 and M 2 mAChR relies only on the P/Qtype VDCC because the effect of the receptors is occluded only when this channel is inhibited, even in the presence of high Ca 2+ concentration [59]. However, in the weak endings, the function of tropicamide-sensitive M 4 mAChRs did not depend on the P/Q-type VDCCs, although it did depend on the normal function of the L-and N-type channel [39,58,59,61]. In the same weak endings, the pirenzepine-sensitive M 1 mAChRs function had multichannel dependence (P/Q-, N-, and L-types), and the methoctramine-sensitive M 2 function also had a multichannel dependence (P/Q-and N-type channels but not the L-type channel). As previously stated, in the weakest nerve contact in dual junctions, only the M 2 has an ACh release potentiating effect whereas M 1 and M 4 reduce release (Fig. 1).
Thus, results indicate that the nerve ending that becomes strong during competition uses a specific coupling (different from the adult) of the M 1 and M 2 mAChR with a broad (the three VGCC are involved), well-defined VGCC stoichiometry that favors ACh release. This configuration courses with the maintenance of these axon terminals. On the contrary, it seems that in the weakest nerve ending, the M 2 release potentiating effect (linked to P/Q and N VGCC) is surpassed by the M 1 /M 4 effect (linked to the three VGCCs) that depresses ACh release and favors axon elimination.

mAChR and Calcium Channels in the Withdrawal of Supernumerary Axons
The L and P/Q-type (but not the N-type) channels tonically enhance synapse elimination because their block prolongs the multiinnervation of the developing NMJ, whereas their exogenous stimulation (Bay-K8644 for L channel and GV-58 for P/Q channels) results in a significant acceleration [26]. Moreover, the delaying effect on axon loss of the L and P/Q channels block is equal to that produced by intracellular calcium sequestration with BAPTA-AM. Thus, the calcium entry through these operative channels present in immature nerve endings results in their final loss. The [Ca 2+ ] i increase contributes both to transmitter release reduction in certain weak axons (as shown above) as well as nerve terminal loss, and this coincidence argues in favor of a shared mechanism relating transmitter release and axonal competition. It is tempting to speculate about the involvement in the neurite retraction or growth during axonal competition of the differentially expressed Ca(2 +) sensor proteins [NCS-1, Ca(2 +)/ calmodulin and several neuro-specific calmodulin-like Ca(2 +) sensor proteins as CaBP1] [98][99][100]. There is a clear relation between PKA, PKC, and VGCC for developmental axonal loss and synapse elimination. The result after the block of the L-channel (and also after intracellular calcium sequestration) is the same as the inhibition of cPKCβI [25] and from stimulation of PKA [88]. However, the inhibition of the nPKCε produces a greater delay in synapse elimination than the L or P/Q channel block or calcium sequestration. This suggests a VGCC-independent component of the PKC-induced enhancement of axonal withdrawal. In adult NMJ, we have seen that nPKCε promotes the phosphorylation of SNARE/SM proteins Munc18-1 and SNAP-25 in an activity-dependent manner [96,97]. Similar to the block of L channels, the block of P/Q-type channels produces retardation in axonal loss similar to that observed after cPKCβI inhibition. However, the effect of the P/Q block in delaying axonal loss is smaller than PKA activation, suggesting the relevant involvement of the PKA activity in axonal stabilization [25].

Concluding Remarks
Axonal competition for synaptic sites is a basic development process that is regulated to achieve optimal connectivity during neurogenesis. In the NMJ, supernumerary axon loss leads to the optimized monoinnervation of the voluntary muscle cells. This process involves activity-dependent autocrine, paracrine (between neighbor nerve terminals), and retrograde (from muscle cells) signalings impacting on the competing nerve terminals. The metabotropic mAChR (M 1 , M 4 , and M 2 subtypes), purinergic receptors (A 1 and A 2A ), and TrkB receptors ensure downstream changes in the balance between PKA (favors axonal strengthening) and presynaptic PKC isoforms (cPKCβI and nPKCε favor axonal retraction) activities. mAChRs and kinases pathways differentially couple to P/Q, N, and L subtypes of the VGCC to differently modulate ACh release in the developing nerve terminals (for instance, the strong and weak endings in a dual junction at a given moment during competition). Moreover, calcium inflow through L-and P/Q-type channels could affect nerve terminals depending on their activity, leading to their final withdrawal or strengthening.
Beyond the analyzed competitive interactions between the multiple axons, several uncertainties persist in the understanding of developmental synapse elimination as, for instance, the mechanism of functionality shifting of muscarinic receptors (and other receptors such as adenosine receptors) or the mechanism of calcium-induced retraction of the ruled-out axons along with the molecular mechanism of rewarding strong endings. The described findings contribute to understanding several aspects of the punishment-rewarding interactions between nerve endings and the contribution of postsynaptic retrograde involvement. It can be stated, however, that the inhibition of any one of these pathways only changes the rate of axonal elimination that finally is completed about 2-3 weeks postnatal indicating the complex multifactorial nature of the process. We think that the high number of molecules and different pathways in the cholinergic peripheral NMJ that is directed to the common objective of supernumerary synapse elimination suggests that some of these molecules may contribute to the same function in other neuronal systems. It seems that the multifactorial mechanism works with precision, though an alteration in many possible points may allow malfunctioning of receptors signaling, kinases ratio, or calcium channel balance resulting in the persistence of multiinnervation. In fact, this alteration has been suggested and even has been shown in a number of diseases such as autistic spectrum disorder (ASD) [101][102][103][104].

Data Availability
We believe that our data are not appropriate for the repository databases available in neuroscience.

Ethics Approval
The mice were cared for in accordance with the guidelines of the European Community's Council Directive of 24 November 1986 (86/609/EEC) for the humane treatment of laboratory animals. All experiments on animals have been reviewed and approved by the Animal Research Committee of the Universitat Rovira i Virgili (Reference Number: 0233).

Consent to Participate
Not applicable to that section.

Consent for Publication
Not applicable to that section.

Competing interests The authors declare no competing interests.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.