During the development of the nervous system, synapses are eliminated on a broad scale (Thompson, 1985; Bourgeois and Rakic, 1993). This allows connectivity to be refined on the basis of Hebbian activity-dependent axonal competition (Jansen and Fladby, 1990; Sanes and Lichtman, 1999). In newborn animals, the skeletal muscle fibers are polyinnervated in the neuromuscular junction area (NMJ; Redfern, 1970; Brown et al., 1976; Ribchester and Barry, 1994), but at the end of the axonal competition, endplates are innervated by a single axon (Benoit and Changeux, 1975; O’Brien et al., 1978; Jansen and Fladby, 1990; Sanes and Lichtman, 1999). This peripheral synapse has been studied extensively as a model for synapse development (Liu et al., 1994; Nguyen and Lichtman, 1996; Chang and Balice-Gordon, 1997; Sanes and Lichtman, 1999; Lanuza et al., 2001; Santafé et al., 2001; Herrera and Zeng, 2003; Wyatt and Balice-Gordon, 2003; Buffelli et al., 2004; Garcia et al., 2010). Several signaling molecules and presynaptic receptors play a role in the axonal competition, which means that the various nerve endings influence one another (Santafé et al., 2009a; Garcia et al., 2010; Nadal et al., 2016). Postsynaptic-derived trophic substances (Nadal et al., 2016) and the participation of glial cells (Lee et al., 2016; Yang et al., 2016) also make a decisive contribution.

In a previous study, we investigated how individual muscarinic acetylcholine receptor (mAChR) subtypes (M₁, M₂ and M₄), adenosine receptors (AR; A₁ and A_2A) and tropomyosin-related tyrosine kinase B (TrkB) receptors are involved in the control of synapse elimination in the mouse NMJ (Nadal et al., 2016). The data show that mAChR, AR and TrkB signaling lessen the initial chance of axonal elimination (around P5–P7) by extending the period of axonal competition but then increase (around P9) axonal loss rate (Nadal et al., 2016). The three receptor sets promote axonal disconnection at the beginning of the second postnatal week largely independently of the postsynaptic nicotinic acetylcholine receptor (nAChR) cluster maturation. In addition, a real cooperation between some of the mAChR and AR subtypes is observed. More specifically, preliminary results show that both AR subtypes (A₁ and A_2A) can add their independent effect on axonal loss to the effect of the M₁ muscarinic receptor, which leads to greater elimination because of the additive effect of the pathways (Nadal et al., 2016).

In this study, we investigate whether the mAChR subtypes and the TrkB receptor also work together, and whether the respective pathway inhibitors have any additive or occlusive effects and, therefore, whether there is any real cooperation between them in synapse elimination at the NMJ. The main result shows that, like the mAChR and AR relations, the effect of M₁ and TrkB receptors can be added to increase axonal loss rate at P9 but that the effect of M₂ is largely independent of the TrkB receptors. Thus, cooperative and non-cooperative signaling contribute to synapse elimination, which highlights the importance of axonal competition and loss in the development of neural connectivity.

The present experiments show evidence of the cooperation between the presynaptic M₁, M₂ and M₄ mAChR subtypes and the TrkB signaling to modulate the conditions of the developmental axonal competition and loss. In a previous study we found that these receptors (as well as presynaptic AR, A₁ and A_2A subtypes) separately contribute to accelerate synapse elimination around P9 in the mouse NMJ (Nadal et al., 2016). It was thought that the muscarinic autoreceptors of the transmitter acetylcholine (ACh) may allow direct competitive interaction between nerve endings through a differential activity-dependent ACh release. The more active axons may directly punish the less active ones or reward themselves (Santafé et al., 2009a). However, an axon that is eliminated at one NMJ may be successful at another (Tomàs et al., 2011), which suggests that other receptors and local postsynaptic- (and glial cell) derived factors are involved. The involvement of the TrkB signaling described may allow a postsynaptic-derived trophic substance such as Brain-derived neurotrophic factor (BDNF) or neurotrophin-4 (NT-4) to make a contribution (Yoshii and Constantine-Paton, 2010).

Interestingly, we observed that both the presynaptic-derived signal (ACh acting on axonal M₁ and M₂ mAChRs) and the TrkB-mediated signal (which may be originated by a postsynaptic-derived NT) have the same effect: namely, the acceleration of supernumerary nerve ending elimination. It seems that the outstanding regulatory resources in the NMJ synaptogenesis are committed to achieving monoinnervation. These presynaptic receptors converge in a common intracellular mechanism and a limited repertoire of effector kinases to phosphorylate protein targets and bring about structural and functional changes leading to axon loss. It is well known that in most cells M₁ and TrkB operate by stimulating the phospholipase C gamma (PLC gamma) and therefore the protein kinase C (PKC) pathway along with the inositol triphosphate (IP3) pathway, whereas M₂–M₄ inhibit the adenyl cyclase (AC) and protein kinase A (PKA) pathway (Caulfield, 1993; Felder, 1995; Caulfield and Birdsall, 1998; Nathanson, 2000). In all cases, however, common final changes such as intracellular calcium oscillations can occur (Santafé et al., 2006; Amaral and Pozzo-Miller, 2012). Both PKA and PKC activity changes have been shown to affect pre- and postsynaptic maturation (Lanuza et al., 2001, 2002). Our present data can be related with the intracellular coupling of the receptors to these serine kinases. Though the blocking efficacy of the selective inhibitors of the muscarinic receptors is not assessed here, M₂ increases the axonal loss rate most with a slight involvement of the M₄ receptor but independently of the M₁ and TrkB receptors. This suggests that downregulation of PKA activity through the couple M₂–M₄ is a key factor in synapse elimination. Concurrently, M₁ and TrkB also contribute separately to axonal loss, but their combined action has a potent summed effect similar to the effect of the M₂ receptor. This suggest that activation of the PLC gamma-PKC pathway through the couple M₁-TrkB may be the other key factor in this process. Thus, a displacement of the PKA/PKC activity ratio to lower values (inhibition of PKA and/or stimulation of PKC) in some nerve endings may have a leading role in synapse elimination. In this context, blockade of PKC in the newborn LAL muscle produces an initial blockade of synapse elimination and a subsequent delay (Lanuza et al., 2002).

In fact, these changes in the kinase activity leading to synapse elimination must occur at least (but not only) in the weakest axons during the competitive interactions. The neurotransmitter release capacity is an important factor in the competing capacity of the various nerve terminals in a NMJ. During development, in the polyinnervated NMJ several nerve endings with different levels of maturation and ACh release capacity get together and compete. The coupling to neurotransmitter release of the considered receptors and kinases is not the same in each of these various endings themselves and in the mature synapses in the adult (Tomàs et al., 2014). So, how does the specific coupling to ACh release of receptors and kinases in the weak and strong axons in competition contribute to axonal loss? As far as serine kinases are concerned, in the adult motor nerve endings both PKA and PKC potentiate ACh release when coupled to neurotransmission (Santafé et al., 2009b). Similarly, the same potentiation is observed in most neuromuscular synapses during development as, for instance, in those formed by the strongest axons (those that evoke the large endplate potential, EPP) in the polyinnervated junctions (Santafé et al., 2004). However, in the weakest endings the inhibition of PKC increases the evoked EPP size indicating that, in normal conditions without any inhibition, this kinase tonically couples to ACh release reduction in these low releasing synapses. Therefore, an M₁-TrkB-mediated increase in PKC activity in the weakest endings would debilitate further their ACh release capacity and competitive force and facilitate their elimination. In addition, an M₂-mediated PKA downregulation in all nerve endings in competition may differentially affect their ACh release and contribute to elimination. Thus, at this point, there is a significant agreement between the known involvement of these molecules in neurotransmission and axon loss.

However, when considering the real postnatal coupling to ACh release of the mAChR and the TrkB receptor in the different nerve endings (the strongest and the weakest) on developing synapses (Santafé et al., 2004; Garcia et al., 2010), additional interpretative keys are needed. In the mature NMJ, M₁ and M₂ subtypes modulate evoked transmitter release by positive and negative feedbacks, respectively (Santafé et al., 2003, 2006). However, during NMJ synaptogenesis, the functional significance of the subtypes is different from in the adult. M₂ receptors promote release in all nerve endings independently of their ACh release level or maturation state whereas an M₁- and M₄-mediated reduction in release is observed in the weakest endings on dual junctions (Santafé et al., 2001, 2002, 2003, 2004, 2007, 2009a). Similarly, the BDNF-TrkB pathway contributes to potentiate ACh release in different neuromuscular adult models but the potentiation is not observed in the weakest nerve endings during development and even some ACh release inhibition was observed in the strongest endings (Garcia et al., 2010). Therefore, interpreting the links and molecular relations between transmitter release and elimination of nerve terminals seems more complex than it seemed at first. The involvement of other signaling such as AR can contribute to this complexity (Todd and Robitaille, 2006; Nadal et al., 2016). However, some conclusions can be drawn on the basis of all the above data. First, it should be pointed out that, contrary to what happens in the adult, M₁ (and M₄) and PKC activity reduces ACh release in the weakest endings and promotes axonal loss. In fact, blocking mAChRs (M₁- and/or M₄-subtypes) or PKC or voltage-dependent calcium channels (VDCCs; P/Q-, N- or L-type or Ca²⁺ influx) can lead to similar percentage increases in the size of the synaptic potentials evoked by weak axons (Santafé et al., 2003, 2004, 2007, 2009a,b; Tomàs et al., 2011). Therefore, the M₁-PKC pathway may debilitate the ACh release capacity and competitive force of these synaptic contacts and facilitate their elimination. The final target molecules involved may be the VDCC, specially the L-type which is exclusively coupled to ACh release in these weak endings (Santafé et al., 2001) and may contribute to carry high calcium near the molecular mechanism directly involved in axon loss. Second, like M₁ signaling, BDNF-TrkB signaling accelerates axon loss. However, it is not so clear whether it is involved in the modulation of ACh release in the nerve endings that are in competition because it does not affect release in the weak axons. Because PKC effectively reduces release in these endings, the TrkB pathway may operate through the IP3 pathway to increase intracellular calcium and modulate the loss of axons. Third, M₂ promotes axonal elimination the most. However, unlike the adult, this muscarinic subtype promotes ACh release in all the endings that are in competition, including the weakest endings and the solitary ending that finally wins the competition. Therefore, there is a shift in the M₂ coupling during development but how this affects its relation with PKA and how this relates with axonal loss is not known.

LN: data collection, quantitative analysis, literature search and data interpretation; EH, AS and MT: data collection, NG: statistics; JT, NG and MAL: conception and design, literature search, data interpretation, confocal microscopy and manuscript preparation. JT, NG and MAL contributed equally to this work. All authors listed, have made substantial, direct and intellectual contribution to the work, and approved it for publication.

This work was supported by a grant from the Catalan Government (2014SGR344) and a grant from Ministerio de Economía y Competitividad (MINECO; SAF2015-67143-P).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.