1. Cell Biology
  2. Neuroscience
Download icon

A novel synaptic plasticity rule explains homeostasis of neuromuscular transmission

  1. Gilles Ouanounou  Is a corresponding author
  2. Gérard Baux
  3. Thierry Bal
  1. Centre National de la Recherche Scientifique, FRE 3693, France
Research Article
  • Cited 17
  • Views 1,343
  • Annotations
Cite this article as: eLife 2016;5:e12190 doi: 10.7554/eLife.12190

Abstract

Excitability differs among muscle fibers and undergoes continuous changes during development and growth, yet the neuromuscular synapse maintains a remarkable fidelity of execution. Here we show in two evolutionarily distant vertebrates (Xenopus laevis cell culture and mouse nerve-muscle ex-vivo) that the skeletal muscle cell constantly senses, through two identified calcium signals, synaptic events and their efficacy in eliciting spikes. These sensors trigger retrograde signal(s) that control presynaptic neurotransmitter release, resulting in synaptic potentiation or depression. In the absence of spikes, synaptic events trigger potentiation. Once the synapse is sufficiently strong to initiate spiking, the occurrence of these spikes activates a negative retrograde feedback. These opposing signals dynamically balance the synapse in order to continuously adjust neurotransmitter release to a level matching current muscle cell excitability.

https://doi.org/10.7554/eLife.12190.001

eLife digest

Nerve cells communicate with each other, and with targets such as muscle cells, at junctions called synapses. The nerve cell before the synapses releases a chemical called a neurotransmitter, which binds to receptors on the cell after the synapses. However, the first cell cannot determine by itself whether it is releasing the correct amount of neurotransmitter to activate its partner. For this, it requires feedback from the second cell.

This feedback is particularly important at synapses between nerve cells and muscle cells, which are known as neuromuscular junctions. The likelihood that a given amount of transmitter will activate a muscle cell can vary with age and after exercise. Muscle cells must therefore be able to instruct their nerve cell partners to increase or decrease neurotransmitter release to accommodate these changes.

Ouanounou et al. have now identified the mechanism by which muscle cells determine whether nerve cells are releasing an appropriate amount of neurotransmitter. Experiments in two distantly related animals – mice and embryos from a frog called Xenopus – revealed that muscle cells use two calcium-based signals. The first is the flow of calcium ions into the muscle cell in response to binding of neurotransmitter to receptors at the synapses: this tells the muscle cell how active the nerve cell is. The second is the release of calcium ions from internal stores inside the muscle cell: this occurs whenever neurotransmitter release is sufficient to activate the muscle cell.

In response to the first calcium signal, the muscle cell sends positive feedback to the neuron, telling it to increase neurotransmitter release further. In response to the second signal, the muscle cell sends negative feedback to reduce neurotransmitter release. Thus, when neurotransmitter release is not enough to activate the muscle, positive feedback dominates and neurotransmitter release increases. However, when the muscle is activated, the two types of feedback act in balance to maintain efficient communication across the synapse.

The next steps are to identify the cell signaling cascades that are mobilized by the two calcium signals, including the specific molecule (or molecules) that regulate neurotransmitter release.

https://doi.org/10.7554/eLife.12190.002

Introduction

In the nervous system, presynaptic neurotransmitter release, postsynaptic receptors, and postsynaptic excitability can be modulated to bi-directionally and durably change synaptic efficacy. There is a large diversity of plasticity processes, which together shape the neuronal network (Changeux and Danchin, 1976; Goda and Davis, 2003; Munz et al., 2014) and tune its properties (Nelson and Turrigiano, 2008). In the brain, concomitantly to various forms of associative plasticity which sustain memory and learning, homeostatic plasticity processes are proposed to restrain the mean level of neuronal activity within a physiological regime, and to maintain the stability of recurrent network activity that can be challenged by associative plasticity (Turrigiano and Nelson, 2004; Macleod and Zinsmaier, 2006; Marder and Goaillard, 2006; Turrigiano, 2007; Nelson and Turrigiano, 2008). Homeostatic plasticity has been extensively studied at the neuromuscular synapse, in particular at the glutamatergic neuromuscular junction (NMJ) in Drosophila (Frank, 2014), due to the robustness of the homeostatic control of synaptic transmission and the great accessibility of the experimental model to genetic manipulations.

In vertebrates, each skeletal muscle fiber is mono-innervated (Sanes and Lichtman, 1999), and each single presynaptic action potential (AP) induces one postsynaptic AP, corresponding to a unity synaptic gain (ratio between the numbers of post- and presynaptic APs equal to 1) (Wood and Slater, 2001). The stability of the gain implies that presynaptic neurotransmitter release and/or postsynaptic receptors adapt the effective synaptic strength to the excitability of the muscle fiber, which depends on the fiber characteristics, and presumably decreases with growth and exercise (Turrigiano, 2007). In adult vertebrate skeletal muscles, cholinergic nicotinic receptors are clustered in the synaptic region. Expression and location of nicotinic receptors have been shown to depend not only on agrin (McMahan, 1990; Hall and Sanes, 1993; Gautam et al., 1995; 1996; Sandrock et al., 1997) but also on activity (Lømo, 2003), suggesting their possible role as adjustment variables in the control of synaptic strength. In the recent years however, thorough studies of the glutamatergic NMJ in Drosophila revealed that presynaptic regulation of neurotransmitter release is certainly a major adjustment variable for synaptic homeostasis (Davis and Müller, 2015). In such 'presynaptic homeostasis', increase in neurotransmitter release counterbalances a genetically-induced decrease of the postsynaptic sensitivity to glutamate (Petersen et al., 1997; Davis et al., 1998; DiAntonio et al., 1999), a genetically-induced increase of postsynaptic input conductance (Paradis et al., 2001) or a pharmacological blockade of postsynaptic receptors (Frank et al., 2006), thus resulting in the strict maintenance of the level of evoked postsynaptic depolarization. Presynaptic compensation of postsynaptic excitability changes implies the existence of a retrograde feedback process from the innervated muscle fiber to the motor neuron. Particular attention was paid to the determination of the molecular targets of the homeostatic retrograde signaling and of the presynaptic cell signaling involved. It appeared that both the readily releasable vesicle pool (Müller et al., 2012; 2015) and the presynaptic voltage-gated Ca2+ channels (Müller and Davis, 2012) are targeted to promote glutamate release when postsynaptic excitability is reduced. The abundance of voltage-gated Ca2+ channels can also be decreased in response to a vesicular content increase (Gaviño et al., 2015), confirming the bi-directionality of the process. In Drosophila, endostatin is a candidate trans-synaptic factor for the retrograde feedback targeting presynaptic neurotransmitter release (Wang et al., 2014). Upstream from the retrograde factors, the postsynaptic kinases, CAMKII (Haghighi et al., 2003) and TOR (Penney et al., 2012), pathways are necessary for a functional homeostatic control at the Drosophila neuromuscular transmission.

An abundant literature is thus progressively assembling pieces of the signaling puzzle underlying the interactions between the motor neuron and the muscle cell for the control of synaptic strength. However, besides the nature of the retrograde feedback and its presynaptic targets, a mechanism for the evaluation of the synaptic strength should be present at the postsynaptic level. The nature and dynamics of these mechanisms, primarily sensing synaptic strength in the muscle cell and initiating retrograde feedback, are key issues of homeostatic plasticity that still remain enigmatic in both invertebrates and vertebrates.

Here we investigate this issue in vertebrate (Xenopus and mouse) cholinergic transmission. Motor neurons and muscle cells of Xenopus laevis embryos establish functional neuromuscular synapses (Tabti et al., 1998) in primary co-culture. Standard intracellular recording from muscle cells is possible in Xenopus culture owing to the small size of the cells (Figure 1A, perforated patch-clamp), while it is not possible in muscle cells from adult mice, due to macroscopic movements during contraction. To circumvent this issue and maintain intracellular recordings in moving tissue, we adapted a ‘floating electrode’ device from previous methods invented for muscles (Woodbury and Brady, 1956) or neurons (Kunze, 1998). We applied this method to mouse ex vivo nerve-skeletal muscle preparations (Figure 1B, Methods, and Figure 1—figure supplement 1). This approach preserved the physiological conditions, necessary to reveal the control exercised by the muscle cell over the neuromuscular synaptic transmission, a role that might have been previously overlooked in vertebrates due to commonly used high concentrations of the nicotinic receptor antagonist curare to prevent contraction.

Figure 1 with 3 supplements see all
Synaptic transmission is homeostatically regulated.

(A) Perforated patch-clamp on Xenopus neuron (N) and muscle cell (M) in primary culture. Presynaptic APs were triggered with current steps. Postsynaptic APs were recorded under current-clamp and nicotinic synaptic currents under voltage-clamp (-80 mV). (B) Intracellular recording in soleus muscle fibers from an adult mouse using a floating sharp electrode (see Materials and Methods and Figure 1—figure supplement 1). (C) Nicotinic conductance calculated from averaged ePSCs (n = 30 ePSCs for each dot) in different Xenopus muscle cells as a function of their input conductance. The black line shows the linear regression. (D) In mice, membrane potential reached by the ePSP in individual FDB muscle fibers after treatment with µ-conotoxin GIIIB, in absence of burst stimulation of the nerve (black dots, n= 88 fibers, 2 muscles, 2 mice), and in test preparations (grey dots, n = 108 fibers, 2 muscles, 2 mice) for which the nerve was burst stimulated prior to conotoxin treatment (15 bursts in 10 min, each of 120 events at 30 Hz). (E) Mean synaptic gain at Xenopus synapses (dots, n = 5 synaptic connections) and in mouse neuromuscular junctions (squares, n = 4 muscle fibers from different mice) during chronic bursts of presynaptic stimulation (bursts of 80 to 120 pulses, 30 Hz for 30 min).

https://doi.org/10.7554/eLife.12190.003

Results

A ubiquitous unity synaptic gain

Synaptic efficacy is linked to the ability of evoked postsynaptic potentials (ePSP) to reach firing threshold. In vertebrate muscle cells, the amplitude of PSPs depends on the ratio between the nicotinic conductance activated by acetylcholine (ACh) and the muscle input conductance, which is largely due to resting K+ leak currents. Fibers in flexor digitorum brevis (FDB), extensor digitorum longus and soleus mouse muscles have a wide distribution of input conductances (Figure 1—figure supplement 2). A similar distribution in Xenopus muscle cells (Figure 1C, X axis) is linked to the diversity of membrane surfaces with homogenous conductance density (196 ± 14 pS/pF, n=19, Figure 1—figure supplement 3). Despite these differences in muscle cell excitability, a single presynaptic AP triggered a single postsynaptic AP, and a contraction, in all tested cells from Xenopus and mouse (Figure 1). In Xenopus cultures, we found that the nicotinic conductance activated in muscle cells by spike-evoked ACh release varied linearly with the postsynaptic input conductances measured in a population of synaptic neuron/muscle cell pairs (Figure 1C), and resulted in comparable ePSP (evoked PSP) voltage amplitudes. Similarly, in mouse muscles, we assessed ePSP characteristics after exogenous application of µ-conotoxin GIIIB which selectively blocks Na+v1.4 channels involved in the muscle cell AP (Cruz et al., 1985; Li et al., 2003). This manipulation does not affect nerve APs, since this Na+ channel subtype is not critical for spiking in motor neurons. Single-spike-evoked PSPs exhibited a narrow range of voltage magnitudes (Figure 1D, in FDB muscle). The comparable amplitude of ePSPs, despite widely ranging resting conductances in the postsynaptic muscle cells, indicates that the neuromuscular synapse exhibits robust plasticity that regulates its functional strength within a narrow range.

Figure 2 with 1 supplement see all
Calcium signaling in Xenopus muscle cell and synaptic homeostasis.

The middle scheme depicts our model of homeostatic control of the synaptic strength. The nicotinic Ca2+ influx elicits a positive retrograde feedback signal (red arrow) on the presynaptic neurotransmitter release. The positive feedback is balanced by a negative retrograde feedback signal (blue arrow) triggered by the muscle AP-induced DICR. (DICR = depolarization-induced calcium release, AP = action potential, RyR = ryanodine receptor, SR = sarcoplasmic reticulum, R nico = nicotinic receptors). (A) Independence of the DICR signal from external Ca2+. AP (black trace, induced with a brief postsynaptic current step), Ca2+ dye (Fluo4) relative fluorescence in standard external medium (dark blue trace) and in Ca2+ free medium (light blue trace), and qualitative measure of the cell contraction (green trace, see Materials and methods). (B) Voltage-dependence of the DICR signal. The Fluo4 fluorescence was measured at the steady-state of the Ca2+ signal during a classical voltage-step protocol from a holding potential of -80 mV and averaged (n = 3 cells). (C) Blockade of the DICR signal by loading the muscle cell with 100 µM ryanodine. (D) Representative example of the DICR signal (blue trace) during repetitive muscle APs (black trace). (E) Ca2+ build-up upon nicotinic receptor activation. Fluo4 signal (upper traces) in control and in Ca2+ free medium, during iontophoretic ACh applications (5 pulses) under postsynaptic voltage-clamp (-80 mV). The lower trace shows the nicotinic currents. (F) Dependence of the nicotinic Ca2+ build-up on the nicotinic conductance with increasing iontophoretic ACh applications at a constant membrane potential (holding potential –80 mV).

https://doi.org/10.7554/eLife.12190.007

A stable synaptic gain

How does the neuromuscular synapse adjust its strength depending upon the excitability of the postsynaptic muscle fiber? Previous research reported that in Xenopus cell cultures, burst activation of presynaptic input results in marked enhancement of ACh release, in both voltage-clamp and current-clamp conditions (Wan and Poo, 1999). We reproduced this result in the specific condition of post-synaptic voltage-clamp (not shown) but not in that of current-clamp where we found instead absence of potentiation. To observe such absence in current-clamp, we believe that it is essential that all or most of the ePSPs evoked by the conditioning burst, reach threshold for spike initiation in the postsynaptic cell. This was not the case in the only example of current-clamp presented by Wan and Poo 1999 (see their Figure 1A(i)). Homeostatic plasticity rules suggest in fact that this form of plasticity should not occur in the presence of synaptic transmission that is effective in generating a muscle AP. To test this hypothesis, we initiated burst stimulation of motor neurons in both Xenopus and mice –during current-clamp recordings– in order to allow voltage responses and APs in the muscle cells. In these physiological conditions and despite the chronic burst activity, no significant change in the synaptic gain was visible. The probability for a synaptic event to induce a postsynaptic AP was close to unity and kept constant (Figure 1E). In Xenopus however, test evoked postsynaptic currents (ePSCs) were recorded under brief periods of postsynaptic voltage-clamp before and after the 30 min of chronic activity under current-clamp, and showed a slight relative decrease of 0.2 in the averaged ePSCs amplitude (see first bar in Figure 3C). In mice, the average ePSPs were not changed by high-frequency conditioning stimulations (Figure 1D and 3F). Altogether, these observations suggest that synapses able to trigger postsynaptic APs do not exhibit strong plasticity, and that the occurrence of muscle APs may be involved in the homeostatic mechanisms adjusting and maintaining the synaptic strength in concordance with the postsynaptic input conductance.

Figure 3 with 2 supplements see all
Nicotinic receptor activity induces a positive feedback on ACh release.

(A) In order to obtain the nicotinic calcium signal in absence of DICR in Xenopus, synaptic events were transiently kept subthreshold either with curare (decreased nicotinic conductance, Gnico) or by dynamic-clamp injection of gK+ leaks (increased input conductance, Gin), or left suprathreshold while DICR was blocked by ryanodine. (B) Effect of presynaptic burst stimulation and curare on spontaneous (upper trace, sPSC) and evoked synaptic currents (ePSC). ePSCs recorded under voltage-clamp were evoked at low rate (0.03 Hz) and averaged by 30–40 events (lower traces). During conditioning presynaptic stimulations (3 bursts of 5 events at 30 Hz), the postsynaptic potential was released from clamp and curare transiently applied (middle trace). Upper trace: for clarity, ePSCs were removed from the continuous trace in order to display the sPSCs only. (C), In Xenopus, mean ePSC relative change 30 min after control chronic bursting activity ('chronic activity', n = 5, illustrated in Figure 1E), 45 min after subthreshold synaptic activity ('Curare', n = 9, illustrated in B; 'Dynamic-clamp', n = 5, illustrated in Figure 3—figure supplement 2A), transient curare application in absence of stimulation ('curare No Burst', n = 3), sub- ('curare-ryanodine', n = 3) and suprathreshold synaptic activity ('ryanodine', n = 6, illustrated in Figure 3— figure supplement 3) in muscle cells preloaded with ryanodine. (D), Relative change in amplitude and frequency of sPSCs after potentiation in Xenopus. (E), In FDB mice muscles, voltage reached by ePSPs in ryanodine treated (black dots, n = 60 fibers, 2 mice), and in ryanodine treated and burst stimulated preparations (red dots, n = 70 fibers, 2 mice). (F) Mean ePSP amplitude in control ('no stim') and high frequency nerve stimulation ('Pre stim') shown in Figure 1D, in non-stimulated ('Ryanodine') and in high frequency stimulated ('Pre stim Rya') ryanodine treated preparations shown in E. *, p<0.05; **, p<0.01; ***, p<0.001; t-test.

https://doi.org/10.7554/eLife.12190.009

Activity-dependent Ca2+ signaling and synaptic homeostasis

How might the strength of synaptic transmission depend upon the occurrence of an AP in the postsynaptic muscle cell? We reasoned that there may be two distinct postsynaptic Ca2+ signaling pathways involved in this homeostatic regulation: a postsynaptic reporter of the presynaptic activity that signals the occurrence of a synaptic event, and a postsynaptic reporter that signals the occurrence of a postsynaptic AP. The postsynaptic Ca2+ build up (Figure 2E–F) due to a high Ca2+ permeability of the nicotinic receptor-channel (Fucile et al., 2006) (PCa2+/PNa+=0.23 in Xenopus, Figure 2—figure supplement 1) is a good candidate for signaling the occurrence of synaptic events. Furthermore, in vertebrates, the skeletal muscle cells exhibit a specific Ca2+ signal linked to the AP and involved in the excitation-contraction coupling, i.e. 'the depolarization-induced Ca2+ release' (DICR). The DICR signal (Figure 2A–D) is a Ca2+ release from the sarcoplasmic reticulum through ryanodine receptors, triggered by plasma membrane depolarizations above –40 mV, and thus triggered by AP in physiological conditions. Depolarization is detected by a voltage-sensor (the dihydropyridine receptor) located in the plasma membrane. This voltage-sensor derives from an L-Type Ca2+ channel; it is voltage-dependent but not ion permeable (Almers et al., 1981; Melzer et al., 1995). Functionally linked to the ryanodine receptor at the level of the T-tubule, its activation triggers the opening of the ryanodine receptor (Rios and Brum, 1987; Catterall, 1991; Franzini-Armstrong and Protasi, 1997). The resulting Ca2+ release is independent of the extra-cellular Ca2+ (Figure 2A), purely voltage-dependent, and its voltage-dependency follows the typical pattern of the activation curve of a 'High Voltage-Activated' Ca2+ channel (Figure 2B). Ryanodine fully blocks this Ca2+ signal (Figure 2C). We hypothesized that these two distinct Ca2+ signals, nicotinic Ca2+ and DICR, are used by the muscle cell to detect synaptic activity and its efficiency to elicit an AP, respectively, leading to retrograde control of the presynaptic ACh release (Figure 2 scheme). To validate this hypothesis, we examined each of these Ca2+ dependent mechanisms in isolation and their respective impact on synaptic efficacy.

Nicotinic Ca2+ triggers LTP

We first examined the specific role of nicotinic Ca2+ influx in the regulation of synaptic transmission. To trigger the nicotinic Ca2+ influx but not the DICR in Xenopus muscle cells, we transiently reduced synaptic strength below AP threshold during the conditioning presynaptic burst stimulations performed in postsynaptic current-clamp. We monitored subsequent synaptic conductance changes under postsynaptic voltage-clamp during low-rate single-pulse intracellular stimulation of the neuron. We reduced the synaptic efficacy either by decreasing the synaptic conductance with low-doses of the reversible nicotinic antagonist curare (Figure 3A–C), or by artificially increasing the muscle cell input conductance using dynamic-clamp (Sharp et al., 1993; Robinson and Kawai, 1993; Prinz et al., 2004) (Figure 3A,C, and Figure 3—figures supplements 1,2A). The lowered synaptic conductance mimicked the effects of low ACh release in developing synapses. Simulation and dynamic-clamp injection of a combination of a linear and an inwardly rectifying K+ leak conductances (Materials and methods and Figure 3—figure supplement 1) mimicked the increased endogenous input conductance associated with muscle cell growth. In both cases, presynaptic burst-stimulation (3 bursts of 5 pulses at 30 Hz) induced a strong, fast (within minutes) and long-term potentiation (LTP) of ePSCs (Figure 3B,C, and Figure 3—figure supplement 2A), while the transient application of curare in absence of presynaptic stimulation did not induce potentiation (Figure 3C, 'curare No Burst' bar). This form of LTP induced by subthreshold synaptic events depends solely on nicotinic receptor activity and therefore was insensitive to DICR blockade obtained by pre-loading the muscle cell with ryanodine (Figure 3C). Furthermore, when ePSPs were left intact but in presence of postsynaptic ryanodine, presynaptic burst-stimulation produced postsynaptic APs and nonetheless a strong LTP of the ePSCs (Figure 3C, and Figure 3—figure supplement 2B), while no potentiation occurred when DICR was functional (Figure 1E and Figure 3C). The LTP induced in our postsynaptic current-clamp protocols was accompanied with a strong increase of the frequency, but not amplitude, of spontaneous postsynaptic currents (sPSCs) (Figure 3B,D). As previously observed in voltage-clamp protocols (Wan and Poo, 1999), this indicates that LTP is due to an evoked ACh release increase and not to a postsynaptic receptor modulation. We found similar results in ryanodine-treated mouse muscles (Figure 3E,F): it is only when nicotinic channels were activated following conditioning bursts of nerve stimulations and ACh release —in the absence of DICR and contraction— that subsequent single test pulses revealed a strong LTP of the ePSPs. In contrast, as shown above, when DICR was functional and the nerve was stimulated, nicotinic ePSPs did not change and remained identical to those in the absence of nerve stimulation (Figure 1D and Figure 3F). Therefore the simultaneous occurrence of postsynaptic nicotinic receptor activation and DICR is essential to the stability of functional synaptic gain. Because the regulation targets ACh release, this implies the existence of retrograde feedback onto the presynaptic compartment.

DICR triggers LTD

In order to understand the specific role of DICR, we examined its effect on synaptic efficacy in isolation from the effect of nicotinic receptor activation. Direct stimulation of the Xenopus muscle cell induces postsynaptic APs and DICR, with no ACh release and no involvement of nicotinic receptors (Figure 4A). Conditioning bursts of suprathreshold current steps injected in the muscle cell induced a strong, fast and long-term depression (LTD) of ePSCs (Figure 4B,C). The average amplitude of sPSCs did not change, confirming the presynaptic locus of gain control, for negative modulations of synaptic strength just like for the positive modulations described above (Figure 4D). LTD induced by repetitive postsynaptic depolarizations has been previously reported (Lo et al., 1994; Dan et al., 1995). Here we show that this form of LTD relies strictly on DICR: it was insensitive to removal of external Ca2+ (Figure 4C and Figure 4—figure supplement 1A), and blocked by postsynaptic pre-loading with ryanodine (Figure 4C and Figure 4—figure supplement 1B). In adult mouse nerve-muscle preparations, external electrical stimulation elicits an AP in both the nerve and the muscle fibers, a situation in which both DICR and nicotinic Ca2+ influx occur. As expected, in the presence of external Ca2+, the external burst stimulation did not change the average ePSP (Figure 4E, F) compared to nerve stimulation only (Figure 1D and Figure 4F). When the external Ca2+ was transiently removed, the external stimulation still elicited APs in both the nerve and the muscle, but neither the ACh release nor its associated postsynaptic nicotinic Ca2+ influx. In these conditions, the external stimulation induced DICR alone and resulted in a strong LTD of the ePSPs (Figure 4E,F). This form of depression is reversible, since subsequent burst stimulation of the nerve induced a rapid potentiation of the ePSPs (Figure 4—figure supplement 2).

Figure 4 with 2 supplements see all
Muscle DICR induces a negative feedback on ACh release.

(A) In order to trigger the DICR signal in absence of nicotinic Ca2+ influx in Xenopus, postsynaptic APs were directly induced in the muscle cell with brief current steps, without presynaptic stimulation, in presence or absence of external Ca2+ (upper right scheme) and with ryanodine that blocks the DICR (bottom right scheme). (B) Effect of selective postsynaptic firing on synaptic currents. Same layout as in Figure 3B. The middle trace shows the muscle APs firing in response to positive current steps injection. (C) In Xenopus, mean ePSC relative change 45 min after control chronic bursting synaptic activity ('chronic activity', n = 5, see 1E), direct triggering of the muscle APs shown in B ('Direct AP', n = 5), APs triggered in a Ca2+-free medium ('Direct AP Ca2+-free', n = 6, illustrated in Figure 4—figure supplement 1), APs triggered in muscle cells loaded with ryanodine ('Direct AP Ryanodine', n = 3, illustrated in Figure 4—figure supplement 2). (D), Relative change in sPSCs amplitude and frequency in Xenopus after potentiation (red) or depression (green). (E), In FDB mouse muscles, voltage reached by ePSPs in externally stimulated preparations (15 bursts of 120 events at 30 Hz, 10 min) in presence (black dots, n = 75 fibers, 2 mice) and in absence of external Ca2+ (green dots, n = 120 fibers, 2 mice). (F), Mean ePSP amplitude in control ('no stim') and high frequency nerve stimulation ('Pre stim') shown in Figure 1D, in externally stimulated preparations shown in E in presence ('External stim-Ca2+') and absence of external Ca2+ ('External stim-0Ca2+'). ***, p<0.001.

https://doi.org/10.7554/eLife.12190.012

Plasticity orientation rule is homeostatic

In the previous experiments, we used the extreme cases of either fully subthreshold synaptic events or of postsynaptic APs in the absence of presynaptic neuron activity, to emphasize the roles of the nicotinic calcium and the DICR signals respectively. However, during normal synaptic activity the two calcium signals are combined and a unique plasticity orientation rule, reflecting the interaction between the potentiating and depressing synaptic processes, should be possible to define in terms of synaptic efficacy.

Patch-clamp on Xenopus cells in culture allows the examining of changes in the synaptic strength in individual neuromuscular synapses, and the assessing of synaptic conductance. In Figure 1E, we observed in non-treated Xenopus synapses that 30 min of chronic burst stimulation of the motor neuron did not drastically change the synaptic gain, but nevertheless induced a slight depression of the averaged ePSCs. Next, we set to closely examine these small synaptic changes in individual synapses. Given the variety of average synaptic conductance among neuromuscular synapses (Figure 1C, Y axis), and its strong correlation with the muscle input conductance (Figure 1C), the range of the ratio between the synaptic and the input conductances 'Gsyn/Gin' is much tighter than the synaptic conductance range. We reasoned that because this ratio, and not the synaptic conductance alone, determines the synaptic efficacy (ePSP amplitude), it should be the appropriate synaptic parameter targeted by the homeostatic process. Therefore, we recorded the ePSCs under brief test periods of postsynaptic voltage-clamp, before and after 20–30 min of chronic burst stimulation of the motor-neuron under postsynaptic current-clamp, and normalized these averaged ePSCs by the input conductance of the muscle cell. Figure 5A (black dots) shows that the slight depression was not homogenous among neuromuscular synapses. The degree of depression seems to depend on the distance of the initial Gsyn/Gin ratio from the mean ratio obtained after chronic activity (Figure 5A, dashed line). Consequently, standard deviation from the mean is reduced after chronic activity (Figure 5A, inset). This synaptic depression behavior can be interpreted as the convergence of the Gsyn/Gin ratios towards the set point of the homeostasis.

Homeostatic control of the synaptic efficacy.

(A) In Xenopus, ratios between averaged synaptic conductance ('Gsyn', calculated from 30–40 ePSCs) and muscle cell input conductance ('Gin') before and after 20–30 min of chronic burst stimulation of the motor neuron (burst of 20–60 events at a 20–30 Hz frequency, every 30–40 s) under postsynaptic current-clamp in non-treated (black dots) and low curare-treated (red dots) synapses. Green dots represent the Gsyn/Gin ratio before and after 1–3 bursts of 5 presynaptic stimulations at 30 Hz (green dots) in ryanodine loaded muscle cells (same data than in Figure 3C, ryanodine bar). Inset shows the mean ± standard deviation of the Gsyn/Gin ratios in the three conditions. The dotted lines show the averaged Gsyn/Gin ratio after chronic activity in non-treated synapses. (B) Degree of plasticity (relative change in Gsyn/Gin ratio) shown in A expressed as a function of the difference between the initial individual Gsyn/Gin ratio and the averaged ratio after chronic burst activity ('Distance to the set point'), in non-treated (black dots) and curare-treated (red dots) synapses. The solid line shows the theoretical homeostatic relationship between plasticity and the distance to a set point of 2.36, calculated as the mean Gsyn/Gin ratio after chronic activity in non-treated synapses. (C) Apparent averaged Gsyn/Gin ratios calculated in mouse FDB muscles from the data of Figure 3E, in non-stimulated and non-treated synapses (no burst, black dot), in burst-stimulated and non-treated synapses (after chronic bursts, black dot), in non-stimulated and ryanodine-treated synapses (no burst, green dot) and in burst-stimulated and ryanodine-treated synapses (after chronic bursts, green dot). Dots represent the mean ± Standard Deviation. (D) Relative change of contraction force during 2s-30Hz bursts of nerve stimulations in mouse soleus muscles (n=4), before and during exposure to a low dose (0.1 µM) of curare.

https://doi.org/10.7554/eLife.12190.015

In order to test whether synapses also converge if the initial Gsyn/Gin ratio is below the set point, we applied the same chronic activity in the continuous presence of low doses of curare (0.1–2 µM). At these concentrations, the bursts of evoked synaptic events are still composed of a majority of suprathreshold events, combining both potentiating and depressing processes. Figure 5A (red dots) shows that, again, the degree of plasticity (potentiation in this case) depends on the distance of the initial Gsyn/Gin ratio to the set point.

The plasticity orientation rule may then be expressed in a way showing its homeostatic character. We plotted the degrees of plasticity (relative change in the Gsyn/Gin ratio) in non-treated (Figure 5B, black dots) and curare treated (Figure 5B, red dots) synapses shown in Figure 5A, as a function of the difference between the initial Gsyn/Gin ratios and the mean Gsyn/Gin ratio after chronic activity (distance to the set point). In a perfect homeostatic process, with an attractor set point, the degree of plasticity can be expressed as a function of the distance to the set point:

(GsynGin)after(GsynGin)before=111+set point(GsynGin)beforeset point

The solid line in Figure 5B shows this theoretical relationship with a set point of 2.36, equal to the mean Gsyn/Gin ratio obtained after chronic activity. The overlap between the data points and the theoretical relationship shows the homeostatic function of the neuromuscular plasticity.

In order to place the potentiation we obtained with ryanodine (Figure 3C, 'ryanodine' bar) in regard of this homeostatic rule, we normalized the averaged synaptic conductances by the input conductance and added the data to the Figure 5A (green dots). The ryanodine-treated synapses did not converge towards a set point. The averaged Gsyn/Gin ratios after burst activity were above the set point, and the standard deviations from the mean were increased. These data suggest that the ryanodine receptors-dependent calcium signal participates to the stabilization of the synaptic efficacy at the set point.

In adult mouse neuromuscular synapses, an apparent Gsyn/Gin ratio can be extrapolated from the ePSPs. If we assume a linear passive leak, the Gsyn/Gin ratio can be expressed as:

GsynGin=80VpVp with Vp the peak membrane potential reached by the ePSP and 0 and -80 mV the reversal potentials of the synaptic and the leak currents respectively. These ratios are only apparent and underestimated given the effect of distance between the recording site and the neuromuscular junction. Figure 5C shows the apparent Gsyn/Gin ratio in non-stimulated and in burst-stimulated preparations, both for non-treated (black dots) and ryanodine-treated (green dots) preparations (same data as in Figure 1D and Figure 3D–E). These data in mouse (Figure 5C) confirmed the results seen in Xenopus (Figure 5A).

Finally, in mouse, we applied a continuous low dose of curare (0.1 µM) on soleus muscles, together with chronic burst stimulation of the nerve, in order to show that the homeostatic processes are also able to compensate for a decrease in the postsynaptic sensitivity to the neurotransmitter. At this curare concentration, the synaptic activity is still composed of a majority of suprathreshold events, combining both nicotinic and DICR calcium signals. We took advantage of the fact that the muscle contraction force integrates the synaptic efficacy over all the fibers of the muscle and all the synaptic events of the bursts, and as such is an indicator of even small changes in the synaptic gain. Figure 5D shows the 26% decrease in the contraction force due to low curare, followed by progressive recovery of the force back to a normal level in response to subsequent nerve burst stimulations. We could thus directly observe the functional effect of the synaptic homeostasis mechanism we describe.

Discussion

We have demonstrated in two evolutionarily distant vertebrate species that the skeletal muscle cell constantly senses the synaptic events and their ability to trigger postsynaptic action-potentials. Two distinct calcium signals report the pre- and postsynaptic activity. Synaptic events are detected through the nicotinic Ca2+ influx and trigger a positive retrograde feedback that targets presynaptic neurotransmitter release. Postsynaptic APs are detected through the sarcoplasmic reticulum DICR signal and trigger a negative retrograde feedback. In Xenopus, convergence of the synaptic efficacy —here captured by the ratio between the synaptic and the input conductances— towards an attractor set point located in the suprathreshold range of synaptic efficacy suggests that a dynamic balance between the potentiating and depressing synaptic processes ensures homeostasis of the neuromuscular transmission. This equilibrium automatically sets the neurotransmitter release to a level matching the postsynaptic excitability. Obviously, ACh release is not the only actor of synaptic efficacy, nicotinic receptors density and muscle input conductance being also under control. However, the advantages of the calcium sensors-based plasticity of presynaptic release are its dynamical aspect and its fast kinetics, confirming in vertebrates that rapid adjustment of the neurotransmitter release is a primary variable in the homeostatic control of neuromuscular synaptic efficacy (Plomp et al., 1992; Petersen et al., 1997; Davis et al., 1998; DiAntonio et al., 1999; Paradis et al., 2001; Frank et al., 2006; Davis and Müller, 2015).

This mechanism may primarily explain the ubiquity of the unity neuromuscular synaptic gain across different muscles and species, and is likely to be involved in several synaptic regulation processes that occur throughout the life of the neuromuscular synapse. In particular, it could be involved in the development and stabilization of forming synapses, in the heterosynaptic depression (Lo and Poo, 1991; 1994) which takes part in the important developmental phase of synaptic selection leading to the characteristic mono-innervation (Gouzé et al., 1983; Sanes and Lichtman, 1999; Witzemann, 2006), and it could counterbalance modulations of muscle excitability associated with growth and physical exercise. This homeostatic control might also be involved in the quantal content increase which partially compensates for the reduced number of postsynaptic receptors at myasthenic human end-plate (Cull-Candy et al., 1979; 1980), or in rat diaphragm muscle treated with the nicotinic antagonist α-bungarotoxin, mimicking myasthenia gravis syndrome (Plomp et al., 1992).

A novel plasticity orientation rule

When selectively triggered, the nicotinic Ca2+ influx and the DICR signal induce a form of LTP and LTD respectively. The causal links between the nicotinic Ca2+ influx and the synaptic events, and between the DICR signal and the muscle APs, allow the establishment of an orientation rule for neuromuscular synaptic plasticity: synaptic events-induced LTP versus muscle AP-induced LTD. This orientation rule finds its equilibrium in the suprathreshold range of the synaptic strength and provides the specific 'homeostatic' function to this plasticity. This neuromuscular plasticity differs from the other forms of long-term plasticity, and in particular should not be confounded with anti-Hebbian plasticity (Kullmann and Lamsa, 2008; Roberts and Leen, 2010). Anti-Hebbian plasticity underlies synaptic strength modulations while the present homeostatic mechanism maintains the stability of synaptic efficacy.

Modulation of synaptic transmission has been extensively studied in vitro at the Xenopus neuromuscular synapse by Poo group (Fu and Poo, 1991; Lo and Poo, 1991; 1994; Lohof et al., 1993; Lo et al., 1994; Dan et al., 1995; Cash et al., 1996; Wang and Poo, 1997; Wan and Poo, 1999). Our work, partly done on the same experimental model, suggests that a number of results obtained by Poo group might be interpreted in the more specific framework of homeostatic plasticity.

In vertebrates, neurotrophic factors such as brain-derived neurotrophic factor, neurotrophin 3 and neurotrophin 4, are trans-synaptic factors considered candidates to mediate presynaptic release modulations in many forms of synaptic plasticity (Schinder and Poo, 2000; Poo, 2001). Skeletal muscle synthesize and release neurotrophins in an activity-dependent manner (Funakoshi et al., 1995; Wang and Poo, 1997; Xie et al., 1997), and motor nerve terminals contain the receptors tyrosine kinase TrkB and C (Henderson et al., 1993; Koliatsos et al., 1993; Wong et al., 1993; Yan et al., 1993). At the Xenopus neuromuscular synapse in vitro, Poo group has shown that the exogenous application (Lohof et al., 1993; Wang and Poo, 1997) of these factors reproduces the synaptic LTP induced in the present work with subthreshold synaptic events or postsynaptic ryanodine treatment. Finally, upregulation of the ACh evoked release in α-bungarotoxin treated rat was markedly reduced by inhibition of the tyrosin kinase receptors of the neurotrophins (Plomp and Molenaar, 1996). Therefore, neurotrophic factors should be considered for future investigations to determine whether they can be retrograde factors mediating homeostatic plasticity at the neuromuscular synapse.

Postsynaptic calcium signaling

The diversity of the effects of the calcium signaling comes from the variety of its temporal patterns, amplitude, and location. The two antagonist calcium signals that we demonstrated here to be involved in the orientation of the neuromuscular plasticity exhibit different temporal patterns and amplitudes. The AP-associated DICR signal exhibits fast transient large calcium concentration elevations for each AP. The kinetics of the nicotinic calcium build-up is much slower and does not contain transients during repetitive nicotinic receptor stimulations. Its amplitude under voltage-clamp is low compare to the DICR signal (Figure 2F), and even lower under current-clamp given the reduced driving-force for the calcium ion due to the postsynaptic depolarization. This dichotomy between fast large transient and slow low calcium build-up is known to trigger selectively, via calmodulin, the CAMKII kinase and the calcineurin phosphatase respectively. This mechanism is considered in the central nervous system to implement Hebbian plasticity (Malenka et al., 1989; Malinow et al., 1989; Mulkey et al., 1993; 1994) by modulating the conductance and number of the postsynaptic receptors (Barria et al., 1997; Mammen et al., 1997; Beattie et al., 2000). The CAMKII signaling pathway was also shown to be required for normal 'presynaptic homeostasis' in Drosophila (Haghighi et al., 2003). Furthermore, Wan and Poo 1999 showed at the Xenopus synapse in vitro that induction of LTP and LTD can be blocked by pre-loading the muscle cell with peptide inhibitors of calcineurin and CAMKII respectively. As noted by Wan and Poo 1999, this situation is opposite to the central synapses, where calcineurin is associated to depression and CAMKII to potentiation. Our results suggest that the low-calcium nicotinic signal and the calcineurin activation have a possible causal link with the detection of the synaptic events and the synaptic potentiation, while the high-calcium DICR signal and the CAMKII activation are associated with the detection of the postsynaptic APs and the synaptic depression.

A push-pull mechanism for homeostatic plasticity

The convergence of the synaptic efficacy towards a set point (Figure 5) suggests that the potentiating and depressing synaptic processes balance each other in the suprathreshold range of the synaptic strength. These observations strongly suggest that the set point of the homeostatic plasticity is sustained by a push-pull mechanism. Our results, however, do not determine the stages where this push-pull mechanism could operate in the causal chain linking the evaluation of synaptic efficacy to the presynaptic modulation. In particular, our findings do not necessarily imply that two independent antagonist trans-synaptic retrograde factors balance their effects at the presynaptic level. The coexistence of potentiating and depressing trans-synaptic factors is a possibility among others. The increased and decreased secretion of a single positive retrograde factor —like for example endostatin in Drosophila (Wang et al., 2014) and neurotrophins in vertebrates— could also bi-directionally regulate neurotransmitter release. In this case, it could be hypothesized that a push-pull mechanism operates downstream of the calcium signaling, for example at the level of the calcineurin/CAMKII balance, ultimately determining the secretion level of a single factor. Therefore, the use of blue and red arrows in Figure 2 is intended as a schematic representation of the retrograde mechanisms that can bi-directionally change neurotransmitter release, without presuming the existence of two distinct retrograde factors.

Comparison with the Drosophila model

Most of what we know about synaptic homeostasis stems from experimental studies at the Drosophila larva neuromuscular junction. Given the numerous differences between Drosophila and vertebrates the comparison between these systems is not straightforward. Action potentials and the associated DICR signal in vertebrates being absent in Drosophila muscle cells (Hong and Ganetzky, 1994), the homeostatic mechanism we propose here cannot be directly transposed to Drosophila. Since no orthologue of the tyrosine kinase receptors are found in Drosophila (Frank, 2014), the neurotrophin hypothesis in vertebrates cannot either be applied to Drosophila. However, different actors could play similar roles in both systems. The calcium permeability of glutamate receptors in Drosophila could have a similar role than the calcium permeability of nicotinic receptors in vertebrates. The low-voltage activated calcium channels activated by the ePSPs and the 'calcium-induced calcium release' signal responsible for the excitation-contraction coupling in Drosophila (Peron et al., 2009) could be the orthologue of the DICR signal in vertebrates. Finally, endostatin in Drosophila (Wang et al., 2014) may play the role of neurotrophins in vertebrates.

In Drosophila, Frank et al. (2006) found that spontaneous glutamate release was sufficient to induce a rapid potentiation in the absence of nerve activity, while we had to use a 30Hz motor command to induce rapid potentiation in vertebrates. In order to establish the Figure 5A, cells were incubated in curare in absence of neurons spikes during 30–60 min before recording (red dots). Therefore, the spontaneous ACh release did not restore the normal Gsyn/Gin ratio (found in non-treated synapses) before chronic bursts were applied. A possible explanation of this difference between the two systems is the high frequency of miniatures in Drosophila (10–20 Hz) (Frank et al., 2006), while the average frequency of spontaneous release in the Xenopus cells culture was 100 fold lower. In vertebrates, evoked activity responsible for muscular tonus and voluntary motions represents most of the synaptic activity, and the 20–30 Hz frequencies we used in our experiments are in the normal range for a motor command (Gorassini et al., 2000). Therefore, the evoked activity in vertebrate is more likely able to rapidly mobilize the homeostatic machinery than the spontaneous miniatures.

NMJ as a model for homeostasis in the central nervous system?

In the central nervous system, homeostatic forms of synaptic plasticity have been proposed to restrain the mean level of neuronal activity within a physiological regime, and to maintain the stability of recurrent network activity challenged by associative plasticity. Because of the robustness of its synaptic transmission, the neuromuscular junction is considered as a model of homeostasis. However, the neuromuscular junction being the end effector of the motor network, its main function is the faithful translation of the motor command into muscle activity. The mean level of muscle activity is not involved in any recurrent network that requires stability, and therefore does not need to be controlled. While the synaptic strength is the adjustment variable for the control of the mean level of neuronal activity in the CNS, at the NMJ it is the efficacy of synaptic transmission itself that must be maintained constant for reliable relay of the motor command. Therefore, homeostatic plasticity could be regarded as a functionally different phenomenon at the NMJ than in the CNS. Nevertheless, the plasticity orientation rule described in the present work matches a synaptic 'relay' function. Despite differences in the nature of the activity sensors, other 'relay' synapses in the CNS, such as those involved in sensory inputs to the thalamus (Guido, 2008), might share with the NMJ a comparable homeostatic control based on pre- and post- synaptic activity detection.

Material and methods

Animal care followed the European Union regulations (OJ of EC L358/1 18 December 1986), and the European directive 2010/63/UE.

Primary cell culture

Request a detailed protocol

Myotomal and spinal tissues from 1-day-old Xenopus laevis embryos (stages 23 to 25) were mechanically dissociated using a Ca2+- and Mg2+-free medium of the following composition (in mM): 115 NaCl, 2.6 KCl, 0.4 EDTA, 10 HEPES (pH = 7.6). Cells were directly plated in a plastic recording chamber, and grown at 19°C for 12 hr prior to the experiments. The culture medium consisted of 50% Leibovitz L-15 medium (Gibco, Invitrogen Corp., Cergy-Pontoise, France), 1% fetal calf serum, 1% antibiotic mixture (ibid., final concentration: 100 units/mL penicillin G and 100 µg/mL streptomycin), and 48% physiological solution of the following composition (in mM): 113 NaCl, 2 KCl, 0.7 CaCl2, 5 HEPES (pH=7.8).

Patch-clamp recordings

Request a detailed protocol

Perforated patch-clamp recordings, to preserve the cell integrity, were performed at room temperature (20–22°C). Pipettes were made from borosilicate glass (Clark Electromedical Instruments, Reading, England) and pulled on a P-1000 puller (Sutter Instrument Company, Novato, CA, U.S.A.). Patch electrodes had a resistance of 2–3 MΩ when filled with internal physiological solution. Membrane currents and potential were recorded using Axopatch200B patch-clamp amplifiers (Axon Instruments, Union City, CA). Access resistances were compensated at 80%. Myocyte membrane currents were filtered with an integrated low-pass Bessel filter at 2 kHz. The filtered signals were digitized by a 12 bit A/D converter (Digidata 1200B, ibid.) and stored using pCLAMP 8 software (ibid.). Recordings were analyzed using the Origin 7 software (OriginLab Corp., Northampton, MA). Motor neurons were current-clamped through an amphotericin-perforated membrane patch (400 pA, 3 ms current step to induce the presynaptic AP); the intra-pipette solution had the following composition (in mM): 1 NaCl, 140 K-gluconate, 1 MgCl2, 10 HEPES (pH=7.2), and amphotericin-B at 300 µg/ml. Myocytes were either voltage- or current-clamped, through an amphotericin-perforated membrane patch, using the following intra-pipette composition (in mM): 1 NaCl, 20 KCl, 125 K-gluconate, 1 MgCl2, 10 HEPES (pH=7.2), and amphotericin-B at 300 µg/ml. The external solution had the following composition (in mM): 140 NaCl, 3 KCl, 1 MgCl2, 2 CaCl2, 10 HEPES (pH=7.4).

Iontophoretic ACh applications

Request a detailed protocol

Pipettes made with borosilicate glass had a resistance of 80–120 MOhm when filled with 0.5 to 1 M ACh-Cl. ACh+ efflux was induced by positive current steps, using a home-made constant current generator.

Fluorescence

Request a detailed protocol

Xenopus muscle cells were loaded with the calcium indicator in whole-cell patch-clamp configuration for 10 min with an intra-pipette medium of the following composition (in mM): 110 KCl, 1 NaCl, 2 Mg2+, 10 HEPES (pH 7.2), and 0.2 Fluo4-pentapotassium salt. The patch pipette was then removed and a perforated patch was performed as described above. Fluorescence intensity was quantified with an Olympus photomultiplier (forming part of the OSP system, Olympus, Japan), and the tension signal was digitized with the Digidata converter. After background fluorescence subtraction, signals were normalized according to the baseline fluorescence.

Nicotinic ionic permeability

Request a detailed protocol

The reversal potential of the nicotinic current induced by ionophoretic acetylcholine application was measured under voltage-clamp in solutions of various ionic compositions (Figure 2—figure supplement 1). The following equation, derived from the Goldman-Hodgkin-Katz flux equation(Goldman, 1943; Hodgkin and Katz, 1949), was used to calculate the permeability ratios:

Vrev=RTF[0.75×PK[K]o+0.75×PNa[Na]o+0.25×4PCa[Ca]o+0.25×4PMg[Mg]o0.75×PK[K]i+0.75×PNa×[Na]i+0.25×Mg[Mg]i]

In this equation, the terms in square brackets are ion concentrations, PS is the permeability of the S ion species, T is the absolute temperature, R and F are the gas and Faraday constants, and i and o are the intra- and extracellular compartments, respectively. Factors represent the ionic activity coefficients.

Dynamic-clamp

Request a detailed protocol

The dynamic-clamp technique (Sharp et al., 1993; Robinson and Kawai, 1993) was used to inject computer-generated conductances in Xenopus muscle cells. Dynamic-clamp experiments were run using the hybrid RT-NEURON environment (Le Franc et al., 2001; Sadoc et al., 2009), a modified version of NEURON (Hines and Carnevale, 1997) running under the Windows operating system (Microsoft Corp., Redmond, Washington), augmented with the capacity of simulating models in real time, synchronized with the intracellular recording. A PCI DSP board with 16 bit A/D-D/A converters (Innovative Integration, SimiValley) was used to input the membrane potential into the equations of the model, and to output the current to be injected into the cell with a time resolution of 0.1 ms. Passive K+ leak and K+ inward rectifying (Kir) conductances were injected into the muscle cell to simulate an increase in the input conductance (Figure 3—figure supplement 1). The passive K+ leak was expressed in the form: Ileak = gleak x (V-Eleak). Eleak was set to -80 mV, and gleak to 5–15 nS. The Kir conductance was expressed in the form: IKir =gmax x m x (V – EK)

where the maximal conductance gmax was set to 20–50 nS, and changes with time of the gating variable m were calculated by solving the differential equation:

dmdt=mmτm
m(t)=m(mm0)×tτm

where fitting voltage-clamp recordings of the real isolated Kir current of Xenopus muscle cells gave and m(v)=11+exp(0.074×(EKV))andτm=0.2ms

Measurement of Xenopus muscle cell contraction

Request a detailed protocol

The 'contraction' trace shown in Figure 2A represents a qualitative measurement of the contraction of a muscle cell in culture. In addition to the membrane potential recording pipette, a second patch pipette vertically approached the membrane cell until a slight increase in the pipette electrical resistance became visible. The pipette was then held in that position, and increase in the cell thickness accompanying contraction was monitored through further variations of the pipette resistance.

Adult mouse nerve-muscle preparations

Request a detailed protocol

3- to 5-month-old Swiss mice were anesthetized with isoflurane, and cervical dislocated. Dissections were performed within 15 min in an oxygenated Ringer solution of the following composition (in mM): 145 NaCl, 3 KCl, 2 CaCl2, 1 MgCl2, 10 HEPES (pH 7.4) and 11 glucose.

Floating electrode recordings

Request a detailed protocol

Intracellular recordings (Figure 1—figure supplement 1) were performed at 34°C, in the oxygenated Ringer solution defined above. Sharp pipettes were made from borosilicate glass (Clark Electromedical Instruments), pulled on a P-1000 puller (Sutter Instrument Company), and had a resistance of 40–60 MΩ when filled with a KCl 3 M solution. The filled pipette tip was cut at the limit of the pulled zone, and used as electrode. The chlorided end of a 10 cm long, 50 µm diameter, silver wire was introduced inside this electrode, and plugged by its opposite end to the headstage of an Axoclamp 2A amplifier (Axon Instruments), where it hung loosely above the muscle. A drop of mineral oil was added at the back of the intra-pipette medium to avoid evaporation. The pipette pendulum was vertically manipulated to enter the muscle cells, and its flexibility allowed stable membrane potential recordings in contracting muscles. Nerve APs were induced with 6 V, 30 µs voltage steps in a suction pipette. 2 µM of µ-conotoxin GIIIB (10 min) was used to isolate the ePSPs. Data acquisition and analysis were made as above.

Under µ-conotoxin the amplitudes of the recorded ePSPs depend on the distance between the recording site and the synaptic region. With the floating electrode method, the placement of the electrode is not precisely controlled. We limited the distance effect by using the FDB mouse muscle, where cells are shorter (300 µm) than the muscle length (1 cm). Therefore, blind placement of the electrode can be performed on this muscle, with a maximal 150 µm distance between the recording site and the end plate. The specific cells size and organization in FDB limit the distance effect in our recordings, but this effect is presumably responsible for an underestimation of the ePSPs amplitudes and for most of the variability between the recorded cells. Despite the underestimation of the ePSPs amplitudes, the recorded amplitudes were above the -63 mV AP threshold determined in muscle cells of rat fast and slow muscles (Wood and Slater, 1995).

Products

Request a detailed protocol

The salts composing the media used for electrophysiological recordings, EGTA used to obtain Ca2+-free media, and Amphotericin B used for perforated patch-clamp were obtained from Sigma-Aldrich (Sigma-Aldrich, Saint Quentin Fallavier, France). Ryanodine used in some experiments was 'Ryanodine fractions' from Latoxan (Latoxan, Valence, France). The µ-conotoxin GIIIB used to block mouse muscle action-potentials was obtained from Alomone Labs (Alomone Labs, Jerusalem, Israel). The Ca2+ dye 'Fluo4-pentapotassium salt' was obtained from Molecular Probes (Molecular Probes, Eugene, Oregon).

References

  1. 1
  2. 2
  3. 3
  4. 4
  5. 5
  6. 6
  7. 7
    Conus geographus toxins that discriminate between neuronal and muscle sodium channels
    1. LJ Cruz
    2. WR Gray
    3. BM Olivera
    4. RD Zeikus
    5. L Kerr
    6. D Yoshikami
    7. E Moczydlowski
    (1985)
    The Journal of Biological Chemistry 260:9280–9288.
  8. 8
  9. 9
  10. 10
  11. 11
  12. 12
  13. 13
    Glutamate receptor expression regulates quantal size and quantal content at the drosophila neuromuscular junction
    1. A DiAntonio
    2. SA Petersen
    3. M Heckmann
    4. CS Goodman
    (1999)
    Journal of Neuroscience 19:3023–3032.
  14. 14
  15. 15
  16. 16
  17. 17
    Ryanodine receptors of striated muscles: A complex channel capable of multiple interactions
    1. C Franzini-Armstrong
    2. F Protasi
    (1997)
    Physiological Reviews 77:699–729.
  18. 18
  19. 19
  20. 20
  21. 21
  22. 22
  23. 23
  24. 24
  25. 25
    Activity of hindlimb motor units during locomotion in the conscious rat
    1. M Gorassini
    2. T Eken
    3. DJ Bennett
    4. O Kiehn
    5. H Hultborn
    (2000)
    Journal of Neurophysiology 83:2002–2011.
  26. 26
  27. 27
  28. 28
  29. 29
  30. 30
  31. 31
  32. 32
  33. 33
    Spatial and temporal expression patterns of two sodium channel genes in drosophila
    1. CS Hong
    2. B Ganetzky
    (1994)
    Journal of Neuroscience 14:5160–5169.
  34. 34
  35. 35
  36. 36
  37. 37
    Nociceptive signal transfer through the dorsal horn network: Hybrid and dynamic clamp approaches using a real-time implementation of the neuron simulation environment
    1. Y Le Franc
    2. B Foutry
    3. F Nagy
    (2001)
    Society for Neuroscience Abstracts 927:18.
  38. 38
  39. 39
  40. 40
    Depression of developing neuromuscular synapses induced by repetitive postsynaptic depolarizations
    1. YJ Lo
    2. YC Lin
    3. DH Sanes
    4. MM Poo
    (1994)
    Journal of Neuroscience 14:4694–4704.
  41. 41
    Heterosynaptic suppression of developing neuromuscular synapses in culture
    1. YJ Lo
    2. MM Poo
    (1994)
    Journal of Neuroscience 14:4684–4693.
  42. 42
  43. 43
  44. 44
  45. 45
  46. 46
  47. 47
  48. 48
  49. 49
    The agrin hypothesis
    1. UJ McMahan
    (1990)
    Cold Spring Harbor Symposia on Quantitative Biology 55:407–418.
    https://doi.org/10.1101/sqb.1990.055.01.041
  50. 50
  51. 51
  52. 52
  53. 53
  54. 54
  55. 55
  56. 56
  57. 57
  58. 58
  59. 59
  60. 60
  61. 61
  62. 62
  63. 63
  64. 64
  65. 65
  66. 66
  67. 67
    Anti-hebbian spike-timing-dependent plasticity and adaptive sensory processing
    1. PD Roberts
    2. TK Leen
    (2010)
    Frontiers in Computational Neuroscience, 4, 10.3389/fncom.2010.00156.
  68. 68
  69. 69
  70. 70
  71. 71
  72. 72
    Dynamic clamp: Computer-generated conductances in real neurons
    1. AA Sharp
    2. MB O'Neil
    3. LF Abbott
    4. E Marder
    (1993)
    Journal of Neurophysiology 69:992–995.
  73. 73
  74. 74
    Culturing spinal neurons and muscle cells from Xenopus embryos
    1. N Tabti
    2. J Alder
    3. MM Poo
    (1998)
    In: G Banker, K Goslin, editors. Culturing Nerve Cells (Edition 2). Cambridge: MIT press. pp. 237–260.
  75. 75
  76. 76
  77. 77
  78. 78
  79. 79
  80. 80
  81. 81
  82. 82
  83. 83
  84. 84
  85. 85
    Activity-dependent expression of NT-3 in muscle cells in culture: Implications in the development of neuromuscular junctions
    1. K Xie
    2. T Wang
    3. P Olafsson
    4. K Mizuno
    5. B Lu
    (1997)
    Journal of Neuroscience 17:2947–2958.
  86. 86

Decision letter

  1. Sacha B Nelson
    Reviewing Editor; Brandeis University, United States

In the interests of transparency, eLife includes the editorial decision letter and accompanying author responses. A lightly edited version of the letter sent to the authors after peer review is shown, indicating the most substantive concerns; minor comments are not usually included.

Thank you for submitting your work entitled "A novel synaptic plasticity rule explains homeostasis of neuromuscular transmission" for consideration by eLife. Your article has been favorably evaluated by Eve Marder (Senior editor) and three reviewers, one of whom, Sacha Nelson, is a member of our Board of Reviewing Editors.

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

The following individuals involved in the review of your submission have agreed to reveal their identity: Benjamin Eaton and Alberto Cangiano (peer reviewers).

Summary:

The authors record from Xenopus neuromuscular junctions in culture and mouse ex vivo nerve-muscle preps and observe potentiation and depression which together implement a homeostatic learning rule for maintaining the gain of neuromuscular transmission close to 1. Both the potentiation and depression have been previously studied at the neuromuscular junction but here, by independently regulating pre and postsynaptic efficacy, the authors demonstrate a dual positive and negative feedback on presynaptic release mediated by calcium influx through nicotinic receptors and depolarization induced calcium release respectively.

Essential revisions:

1) The authors misstate the prior results of Wan and Poo who reported achieving LTP when the postsynaptic cell was either in voltage clamp or current clamp. The subsection “A stable synaptic gain” should entirely be rewritten since: 1) Wan and Poo (Science 1999) saw synaptic potentiation not only after presynaptic burst stimulation under myocyte Vclamp but also Iclamp, and 2) the statements of "…no apparent synaptic plasticity was visible. On the contrary…", are not clear. Also, the effects of nicotinic calcium signaling on presynaptic function presented here are seemingly in contrast to the previous observations (i.e. Dan and Poo, Science 1992) that found that increased nicotinic calcium produces synaptic depression and not LTP. The authors need to address these discrepancies.

2) The results presented on triggering LTP suggest that the sensitivity of the muscle to neurotransmitter is not what determines the increase in ACh release but rather the lack of the LTD mechanism induced by postsynaptic firing. This seems to be contrary to numerous studies from flies, mice, and humans supporting the idea that presynaptic compensation for changes in muscle sensitivity to neurotransmitter is one of the primary homeostatic mechanisms at the NMJ and driven by reduced AChR function or increasing K leak (see below). This point needs shoring up both with further discussion and with controls for the effects of curare incubation and/or changes in input conductance on ACh release in the absence of bursting. With respect to further discussion, the authors need to do a better job of explaining other differences between this data set and previous studies such as those of Paradis et al. 2001, Frank et al. 2006, Petersen et al. 1997; Cull-Candy et al. 1979 and 1980; Plomp et al. 1992 and Plomp and Molenaar 1996.

3) It is not clear whether the authors are suggesting that the positive and negative feedback mechanisms are dynamically balanced to achieve unity gain or gate each other to shut off plasticity once unity gain has been achieved. Given the fact that both forms of plasticity have previously been described and what is novel here is the interaction between the two it seems reasonable to insist on a better characterization of that interaction. This could be performed along one or both of the following dimensions:

A) Timing: Is the plasticity "gated" by SR calcium release or is there a temporal dependence to this mechanism? A tight temporal requirement would support that the gating mechanism is occurring within the muscle cell. Demonstrating a shared cellular localization for the convergence of these processes would strengthen the argument that LTD regulates/gates the LTP. Related to this point, is it surprising that the LTD is so stable in Figure 4E given that the ePSPs seem to all fail to reach threshold? Shouldn't this induce LTP?

B) Gradation: It is predicted that homeostatic regulation of PSP amplitudes would require a graded response to provide needed precision to the regulation (Davis, 2006). This should be demonstrated. Is either of the signals or their respective outputs graded? This could be revealed by investigating the freq dependence and/or the number of postsynaptic depolarizations required to generate LTD (Dan and Poo, Science 1992). Or is bulk unloading to the SR with thapsigargin sufficient to induce LTD? Overall, the authors need to strengthen their argument that this system is truly homeostatic. Similarly, is there a threshold level of nicotinic calcium required to induce LTP? Is there any freq dependence? Given previous observations from Drosophila and patients with myasthenia gravis, it might be expected that sub-blocking levels of curare would induce increased ACh release independent of bursting. Thus, an important control for the curare experiments is whether curare incubation in the absence of bursting or with chronic activity results in LTP.

One related concern is the safety factor for NMJ transmission – if homeostasis is ongoing wouldn't it tend to obliterate excess EPSP above that needed to evoke an AP?

Although it would be strongest to address both "axes" doing one or the other should be sufficient for publication (though the curare control experiment should be done in either case). This need not be carried out in both systems. Hence revision could involve a modest number of additional experiments which could be completed in two months or less.

https://doi.org/10.7554/eLife.12190.023

Author response

Essential revisions:

1) The authors misstate the prior results of Wan and Poo who reported achieving LTP when the postsynaptic cell was either in voltage clamp or current clamp. The subsection “A stable synaptic gain” should entirely be rewritten since: 1) Wan and Poo (Science 1999) saw synaptic potentiation not only after presynaptic burst stimulation under myocyte Vclamp but also Iclamp, and 2) the statements of "…no apparent synaptic plasticity was visible. On the contrary…", are not clear. Also, the effects of nicotinic calcium signaling on presynaptic function presented here are seemingly in contrast to the previous observations (i.e. Dan and Poo, Science 1992) that found that increased nicotinic calcium produces synaptic depression and not LTP. The authors need to address these discrepancies.1A) “Misstatement of the prior results of Wan & Poo”:

In the pioneer study by Wan & Poo (Science, 1999) the induction of LTP was indeed shown both under postsynaptic current-clamp (Figure 1A(i)) and voltage-clamp (Figure 1A(ii-v)). Nonetheless, it was not clear to us whether the plasticity orientation rule (Figure 1C in Wan & Poo 1999), dependent on the initial synaptic conductance, was obtained under post-synaptic voltage- or current-clamp, since this information was not provided in the main text nor in the figure legend. Since the plasticity we found under current-clamp (Author response image 1, blue dots) was different from that found by Wan & Poo 1999, we explicitly addressed the question whether this discrepancy was due to different clamping modes. We then performed the experiment under postsynaptic voltage-clamp, following the protocol of Figure 1A(ii) in Wan & Poo. Author response image 1 (black dots) shows the orientation rule obtained under voltage clamp, which perfectly matched the orientation rule found in Wan & Poo 1999 (Figure 1C).

Author response image 1
Effect of burst stimulation of the motor neuron under postsynaptic voltage- and current-clamp.

Black dots: same protocol as Wan & Poo 1999 (Figure 1A (ii)). Blue dots: same data as Author response image 4 (black dots), expressed as a function of the initial averaged ePSCs amplitude.

https://doi.org/10.7554/eLife.12190.016

In the single illustration of potentiation under postsynaptic current-clamp in Wan & Poo (Figure 1A(i)), the synaptic efficacy during the conditioning burst was mainly sub-threshold (1 action-potential for the first event followed by 4 sub-threshold events composed the post-synaptic response to the bursting presynaptic stimulation). Figure 1 of our manuscript shows that given the correlation between the synaptic and the muscle input conductances (Figure 1C), sub-threshold synaptic efficacy turned out to be rare and not representative of the initial conditions in our Xenopus cell culture. We then had to artificially reduce the ratio between the synaptic and the input conductances to obtain sub-threshold synaptic bursts that resulted systematically in potentiation, independently on the initial synaptic conductance (Author response image 2, red dots).

All these observations brought us to conclude, maybe wrongly, that the orientation rule of the Figure 1C in Wan & Poo 1999 was established under postsynaptic voltage-clamp. However, since we think that this point is not crucial for our argumentation, we removed all mentions of voltage-clamp from the text.

Author response image 2
Degree of plasticity (relative change in ePSC amplitude) as a function of the initial ePSC.

Potentiation obtained with a burst of subthreshold ePSPs (red dots) and depression obtained with direct triggering of the muscle APs (black dots). Same data than Figure 3C ('curare' bar) and Figure 4C ('direct AP' bar) expressed as a function of the initial averaged ePSCs amplitude.

https://doi.org/10.7554/eLife.12190.017

The statements "…no apparent synaptic plasticity was visible. On the contrary.…", are not clear.

Under postsynaptic Iclamp, the bursting stimulation of the motor neuron did not induce drastic change in the synaptic gain in Xenopus (Figure 1E) contrary to what was expected regarding the plasticity orientation rule obtained under post-synaptic Vclamp (Author response image 1, black dots). This discrepancy was the object of the statements in the subsection “A stable synaptic gain”. Because of the correlation with the input conductance, most of synapses were efficient in eliciting the muscle spikes, and according to homeostatic rules, a synapse able to trigger the muscle spikes (i.e. close to the set point of the homeostasis) should not exhibit strong plasticity. This is further confirmed by the fact that chronic bursting stimulation of the motor neuron under post-synaptic Iclamp in non-treated synapses induced a low depression and extremely rarely a potentiation of the synaptic currents (Author response image 1, blue dots).

We clarified the statement in the subsection “A stable synaptic gain”.

1B) “Also the effects of nicotinic calcium signaling on presynaptic function presented here are seemingly in contrast to the previous observations (i.e. Dan and Poo, Science 1992) that found that increased nicotinic calcium produces synaptic depression and not LTP.”:

Our study was more qualitative (nicotinic calcium versus DICR) than quantitative, and we estimate that our results do not contradict that “increased nicotinic calcium induces depression and not potentiation”. We think that our work is not in contrast with the idea that low calcium induces potentiation and high calcium depression, but points out how these two calcium regimes are physiologically achieved via different pathways.

Plasticity orientation rule under postsynaptic Vclamp supports the idea that “increased nicotinic calcium produces synaptic depression and not LTP”:

Even more than the data in Dan & Poo 1992, the orientation rule obtained under Vclamp may also suggest that “increased nicotinic calcium produces synaptic depression and not LTP”. Indeed, in Dan & Poo 1992, depression was obtained via ionophoretic ACh application. Since in the Xenopus cell culture the nicotinic receptors are not limited to the synaptic region, ionophoretic ACh application may mimic a competing synapse. This situation may be more related to synaptic competition and hetero-synaptic depression than to plasticity at mono-innervated muscle cell. Nicotinic calcium may have different effects when induced by another synapse. This idea is supported by the fact that Dan & Poo obtained depression with ionophoretic ACh alone and not when ACh was applied under postsynaptic Vclamp concomitantly with the pre-synaptic stimulation, while in this case nicotinic calcium build up should have been more important (Figure 1 of Dan & Poo 1992).

In contrast, under post-synaptic Vclamp at mono-innervated muscle cell, DICR is prevented (owing to the lack of depolarizations) and nicotinic calcium at the synapse is the only source of calcium. Since under Vclamp the driving force for Ca2+ is constant and the nicotinic calcium depends solely on the membrane permeability (i.e. the synaptic conductance), then nicotinic calcium increases linearly with the X-axis of Author response image 1. Low nicotinic calcium-induced LTP and high nicotinic calcium-induced LTD may thus be an attractive explanation for the orientation of plasticity under post-synaptic Vclamp, and is compatible with the demonstration by Wan & Poo 1999 that potentiation depends on calcineurin (selectively activated by low calcium) while depression depends on CAMKII (selectively activated by high calcium).

The orientation rule in Wan & Poo 1999 (Figure 1C), which matches our results under postsynaptic Vclamp (Author response image 1, black dots), leaded to another interpretation: the initial synaptic currents amplitude may reflect the degree of maturity of the synapse, and immature synapses tend to potentiate while mature ones tend to depress. But we showed that 1) the synaptic current amplitude reflects in fact the adaptation to the post-synaptic input conductance and that the synapses are functionally equivalent (they trigger spikes), and 2) all the synapses are capable of both potentiation and depression. The potentiation shown in Figure 3C (“curare” bar) and the depression shown in Figure 4C (“direct AP” bar) were independent of the initial synaptic currents (Author response image 2). Therefore, we believe that nicotinic calcium level-dependent plasticity is a more attractive explanation of the orientation of plasticity under post-synaptic Vclamp than a structural or functional capability of the synapses in plasticity.

Under postsynaptic Iclamp, the nicotinic calcium is low:

Are our data under post-synaptic Iclamp compatible with the view that low-calcium induces potentiation and high-calcium induces depression?

Under post-synaptic Iclamp, nicotinic calcium no longer depends solely on the synaptic conductance as in Vclamp but also on the variation of the membrane potential. We investigated the voltage dependency of the Ca2+ component of the nicotinic current in Xenopus myocytes (Author response image 3A) and we found the same Ca2+/V relationship as in Miledi et al. 1980 (J Physiol. 1980 Mar;300:197-212), i.e. the nicotinic Ca2+ exponentially decays with a potential constant of 30 mV. The nicotinic calcium build up is consequently strongly sensitive to the membrane potential. Thus for a given nicotinic conductance the nicotinic calcium is much lower under Iclamp than under Vclamp with a holding potential close to the resting potential.

In order to quantify the effect of the potential on the nicotinic calcium during a burst we did a conductance-based computer model based upon our characterization of the ionic currents of the Xenopus myocyte, including the voltage-dependency of the nicotinic calcium. For a given membrane surface, the increased permeability combined with the opposite voltage-dependency resulted in a bell-shape relationship between the nicotinic calcium and the synaptic conductance (Author response image 3B). As an illustration, simulations show that for a burst of 5 synaptic events at 20-30 Hz with a conductance producing a 1.5 nA synaptic current at a holding potential close to the resting potential, the use of Vclamp increases by 5 times the nicotinic calcium compared to Iclamp. In other words, under Iclamp the nicotinic calcium during this burst is equivalent of that provided under Vclamp by a much weaker synaptic conductance producing -300 pA of current. Therefore, the range of nicotinic calcium mobilized under Iclamp is globally in the range mobilized under Vclamp by low synaptic conductances which induce potentiation (Author response image 1, black dots). Finally, the AP-induced DICR, known to be largest calcium signal in skeletal muscle cells, might be more susceptible than the nicotinic calcium to reach under IClamp the high level of calcium necessary to activate CAMKII and to induce depression.

The corollary could be that under Vclamp, the nicotinic calcium artificially reaches the threshold inducing depression because of the maximum driving force maintained by the holding potential.

Author response image 3
Voltage-dependency of the nicotinic calcium.

(A) Nicotinic I/V relationship with iohophoretic ACh applications in media for which the Ca2+ ion only carries the nicotinic current. Na+ and Mg2+ were removed, K+ was adjusted for each holding potential in order to remain at the equilibrium, and Lysine hydrochloride was inversely adjusted to KCl in order to hold both the osmotic pressure and the ionophoretic ACh application. Dashed line shows the rectification predicted with the GHK free diffusion model. (B) Example for a given cell surface of simulation of the nicotinic calcium in V- and I-clamp. The model comprised a synapse, a linear passive K+ leak, an inward rectifying K+ leak, two voltage-gated K+ conductances and a voltage-gated Na+ conductance.

https://doi.org/10.7554/eLife.12190.018

In conclusion, our results under Iclamp seem to not be in contrast with the fact that “increased nicotinic calcium induced depression and not potentiation”. We thought interesting to discuss here the discrepancies between Iclamp and Vclamp, but the issue of Vclamp being not crucial to the paper, we chose to avoid a lengthy discussion on this point and we suppressed its mention from the text. We added to the Discussion of the revised version of the manuscript that the low level of nicotinic calcium under Iclamp and the high level of calcium reached by the DICR signal are consistent with the demonstration by Wan & Poo of the roles of calcineurin and CAMKII in plasticity.

2) The results presented on triggering LTP suggest that the sensitivity of the muscle to neurotransmitter is not what determines the increase in ACh release but rather the lack of the LTD mechanism induced by postsynaptic firing. This seems to be contrary to numerous studies from flies, mice, and humans supporting the idea that presynaptic compensation for changes in muscle sensitivity to neurotransmitter is one of the primary homeostatic mechanisms at the NMJ and driven by reduced AChR function or increasing K leak (see below). This point needs shoring up both with further discussion and with controls for the effects of curare incubation and/or changes in input conductance on ACh release in the absence of bursting. With respect to further discussion, the authors need to do a better job of explaining other differences between this data set and previous studies such as those of Paradis et al. 2001, Frank et al. 2006, Petersen et al. 1997; Cull-Candy et al. 1979 and 1980; Plomp et al. 1992 and Plomp and Molenaar 1996.

2A) Controls in absence of bursting:

The main concern raised in this point seems to be that we used evoked neurotransmitter release (in the form of trains of presynaptic stimulations called “bursts”) to induce potentiation, while in the Drosophila larva evoked release is not necessary to a rapid compensatory potentiation in response to the reduction of the post-synaptic sensitivity (Frank et al. 2006). If evoked release is not necessary for potentiation in Drosophila, synaptic activity is nonetheless required, since the authors propose that spontaneous release remaining in absence of motor neuron activity is responsible for this potentiation.

Following the reviewer suggestions, we performed two control experiments:

1) In our potentiation induction, it has to be noted that curare was transiently (1 min or less) applied to reduce the synaptic efficacy during the conditioning burst. In order to test that evoked bursts, and not spontaneous release, was responsible for potentiation, we transiently applied curare during 2 min in absence of pre-synaptic stimulation, and we did not obtain potentiation (relative change in ePSC of 0.89 ± 0.05, n=3). We have added these data to Figure 3C (“Curare No burst” bar).

2) We did another control with longer lasting incubation of curare. Given the large range of initial synaptic conductances (Gsyn) in the cell culture but the strong correlation (Figure 1C) between this conductance with the input conductance (Gin), we compared the Gsyn/Gin ratio in non-treated synapses with synapses incubated in curare after 30-60 min of exposure. The averaged Gsyn/Gin ratio in curare (Gsyn/Gin = 1.35 ± 0.32, n = 8) was significantly lower than the averaged Gsyn/Gin ratio in non-treated synapses (Gsyn/Gin = 3.78 ± 0.49, n = 10) with a significance level of 0.001. Since motor neurons do not spontaneously fire in the Xenopus cell culture, this result shows that within 60 min of curare incubation the spontaneous ACh release was unable to sufficiently promote the evoked release in order to restore a normal evoked Gsyn/Gin ratio.

2B) The frequency of spontaneous synaptic activity may explain some differences between Drosophila and vertebrates:

The 60 min control incubation of the above experiment does not presume the outcome of a longer curare exposure that could result in a full recovery of the synapse efficacy. In the Xenopus cell culture, motor neurons are “isolated” (not connected by afferences) and do not spontaneously fire. Despite the absence of motor neurons spikes, they connect the muscle cells and the mechanisms responsible for the evoked release establish during the 12-24h before the recordings. The establishment of a functional evoked release in absence of neuronal spikes, and moreover dependent on the muscle input conductance (Figure 1C), presumably implicates the spontaneous release. The difference between Drosophila and Xenopus in the speed of compensatory potentiation induced by spontaneous activity may simply be related to the strong difference between these systems in the frequency of the spontaneous activity. In Paradis et al. 2001, Frank et al. 2006 or Petersen et al. 1997, it seems that the frequency of spontaneous release in Drosophila larva is around 10-20 Hz, while it is 100 time less on average in our Xenopus cell culture.

2C) Evoked activity should also mobilize homeostatic mechanisms:

If spontaneous synaptic events are able to mobilize some synaptic plasticity mechanisms, it seems reasonable to envisage that evoked release does too. Moreover, homeostasis of the neuromuscular transmission adapts the evoked neurotransmitter release to the post-synaptic input conductance in order to ensure the reliability of the motor command transmission. Therefore, it seems useful for the homeostatic mechanisms that evoked release participates to the evaluation of its own efficacy and that this evaluation does not depend solely on the spontaneous release. The large literature produced by the Poo’s group, and the present work, suggest that in Xenopus cell culture bursts of evoked activity are able to trigger the plastic properties of the synapse more rapidly than spontaneous activity.

In Drosophila, Petersen et al. or Paradis et al. observed the outcome of the homeostasis but it does not exclude that evoked neurotransmitter release had participated to the homeostasis before its observation. Even if Frank et al. show that 10-20 Hz spontaneous activity is able to trigger potentiation in absence of evoked activity, this does not prove that evoked activity with the frequency of a normal motor command cannot do the same. A possible stabilizing role of synaptic depressing processes mobilized by evoked activity remains to be investigated but is not yet excluded in Drosophila.

2D) Evoked burst at 30 Hz is natural:

In vertebrate, evoked activity is responsible for the muscular tonus and voluntary motions, and represents the prominent part of synaptic activity. In vertebrates, the frequency range of a motor command is 10-60 Hz, and muscle activity being made of temporary motions, the motor command is made of bursts. Thus, the conditioning bursts we used in our experimental protocols closely mimic the typical firing rate of individual motor units, as recorded in fast and slow muscles in the conscious rat during unrestrained walking (Gorassini et al., 2000 Journal of Neurophysiology 83 (4): 2002–11). In addition, bursting activity with frequency of action potentials in the range of 10-60 Hz has also been observed in the developing spinal cord (L. T. Landmesser and M. J. O’Donovan, 1984, J. Physiol. London347, 189; M. J. O’Donovan et al., 1998, Ann. N.Y. Acad. Sci.860, 130).

2E) The results presented on triggering LTP suggest that the sensitivity of the muscle to neurotransmitter is not what determines the increase in ACh release but rather the lack of the LTD mechanism induced by postsynaptic firing.

This question is related to the point 3 of the essential revisions, concerning the type of interaction between the potentiating and the depressing synaptic mechanisms. We used the extreme cases of subthreshold synaptic events and ryanodine treated preparations to prevent DICR and to emphasize the potentiating role of the nicotinic calcium. But as discussed in the answer to point 3, and as shown in the new Results section of the revised version (Figure 5A–B), potentiation also occurs in a range of synaptic efficacy mixing nicotinic calcium with DICR signals. We show in this new section that potentiation is more related to how far from an “attractor Gsyn/Gin ratio” is the initial Gsyn/Gin than it is due to “the lack of depressing mechanism”.

One goal of our work was to show that suprathreshold synaptic activity mobilizes both potentiating and depressing mechanisms, and that homeostasis might be due to a balance between these antagonist mechanisms. Even if synaptic strength stabilization at a set point results from such balance, the “postsynaptic sensitivity to the neurotransmitter” – here assimilated to the ratio Gsyn/Gin– remains “one of the primary variables determining presynaptic compensation”.

2F) Differences between this data set and previous studies such as those of Paradis et al. 2001, Frank et al. 2006, Petersen et al. 1997; Cull-Candy et al. 1979 and 1980; Plomp et al. 1992 and Plomp and Molenaar 1996.:

Beyond the difference between vertebrates and Drosophila concerning the frequency of spontaneous synaptic activity, there are other parameters that make a direct comparison of these systems difficult. In particular, given the lack of muscle action potentials and DICR signal in the Drosophila larva, the mechanism we propose here cannot be directly transposed from vertebrate to Drosophila. However, it could well be that in Drosophila the evoked release mobilize a putative negative side of the homeostasis through actors other than DICR. The literature on Drosophila focuses mainly on the presynaptic expression of the homeostatic potentiation, and not on the postsynaptic mechanisms of evaluation of synaptic efficacy leading to the homeostasis. However, despite the absence of muscle spikes in Drosophila, evoked postsynaptic potentials should activate low-voltage-activated Ca2+ channels and the calcium-induced calcium release responsible for muscle contraction. It is tempting to speculate on the possible role of these calcium signals in the stabilization of the evoked release in Drosophila, and this remains to be investigated. We added to the Discussion the comparison between Drosophila and our data, including the differences in the role of the spontaneous release.

In the study by Cull-Candy et al. 1979 and 1980, the authors compared the neurotransmitter release in Myasthenia Gravis (MG) and control patients. They found that the neurotransmitter release was not responsible for the weakness characteristic of the MG disease (Cull-Candy et al. 1979), and that in contrast a five-fold increase in the quantal content of the evoked release partially compensated the reduced number of postsynaptic receptors (Cull-Candy et al. 1980). Therefore, this work is a post-observation of the outcome of a compensatory process which had occurred into the organism, and does not predict what kind of synaptic activity between spontaneous release and evoked motor command was responsible in the triggering of such compensation. Despite their weakness, MG patients’ movements are indeed due, as mentioned above, to bursts of evoked release in a frequency range containing the 30Hz we used in our experiments. Therefore, the increase in the quantal content we obtained in response to reduced Gsy/Gin and the mechanism we propose seem not in contrast with these observations.

A similar argument can be applied for the study by Plomp et al. 1992 and Plomp & Molenaar 1996. In this work, the author mimicked the MG syndrome in rat with chronic injections of α-bungarotoxin, and observed that an increase in the evoked release compensate for the reduced number of functional postsynaptic receptors, an observation that is not in contrast with our work. The time course of the compensation cannot be compared with our work since bungarotoxin was injected chronically every 48h. However, a difference with our work is the apparent long delay (3h minimum) before the observation of the compensation. This apparent delay might be linked to the mode of bungarotoxin administration, i.e. subcutaneous injection, which renders the time course of the appearance of the drug’s effect at the diaphragm neuromuscular junctions difficult to predict. We introduced these references, and these elements of discussion, to the Discussion section of the revised manuscript.

3) It is not clear whether the authors are suggesting that the positive and negative feedback mechanisms are dynamically balanced to achieve unity gain or gate each other to shut off plasticity once unity gain has been achieved. Given the fact that both forms of plasticity have previously been described and what is novel here is the interaction between the two it seems reasonable to insist on a better characterization of that interaction. This could be performed along one or both of the following dimensions: A) Timing: Is the plasticity "gated" by SR calcium release or is there a temporal dependence to this mechanism? A tight temporal requirement would support that the gating mechanism is occurring within the muscle cell. Demonstrating a shared cellular localization for the convergence of these processes would strengthen the argument that LTD regulates/gates the LTP. Related to this point, is it surprising that the LTD is so stable in Figure 4E given that the ePSPs seem to all fail to reach threshold? Shouldn't this induce LTP? B) Gradation: It is predicted that homeostatic regulation of PSP amplitudes would require a graded response to provide needed precision to the regulation (Davis, 2006). This should be demonstrated. Is either of the signals or their respective outputs graded? This could be revealed by investigating the freq dependence and/or the number of postsynaptic depolarizations required to generate LTD (Dan and Poo, Science 1992). Or is bulk unloading to the SR with thapsigargin sufficient to induce LTD? Overall, the authors need to strengthen their argument that this system is truly homeostatic. Similarly, is there a threshold level of nicotinic calcium required to induce LTP? Is there any freq dependence? Given previous observations from Drosophila and patients with myasthenia gravis, it might be expected that sub-blocking levels of curare would induce increased ACh release independent of bursting. Thus, an important control for the curare experiments is whether curare incubation in the absence of bursting or with chronic activity results in LTP.

One related concern is the safety factor for NMJ transmission – if homeostasis is ongoing wouldn't it tend to obliterate excess EPSP above that needed to evoke an AP? 3A) “Positive and negative feedback mechanisms are dynamically balanced to achieve unity gain or gate each other to shut off plasticity once unity gain has been achieved?”

Some elements for answering this question may already be found in the data of the paper, and they support the dynamical balance hypothesis. It should be noted that despite an apparent safety factor (slope in Figure 1C ≥ 3, Author response image 4), the synaptic gain in the Xenopus cell culture is not strictly 1 as it is in the mouse neuromuscular junction, but rather close to 0.7-0.8 (Figure 1E). This results from the variability of neurotransmitter release in this model (Figure 1A). Evoked post-synaptic activity is thus made of 70-80% of action potentials (some triggered by release largely exceeding the threshold) and 20-30% of sub-threshold ePSPs. The bursts composing the chronic activity in Figure 1E were made of these proportions of sub- and supra-threshold events. We reasoned that in the case of a “gating mechanism”, supra-threshold events would have been neutral in terms of plasticity (“shut off plasticity”), and thus the remaining 20% of sub-threshold events would have potentiated the synapse up to gain 1. Instead of that, the synaptic gain slightly decreased (Figure 1E) due to a slight depression of the averaged synaptic currents (0.2, subsection “A stable synaptic gain “of the initial manuscript). The absence of potentiation despite the presence of sub-threshold events is not in favor of “gating” hypothesis, and the plasticity outcome – i.e. depression – excludes it.

Gradation of depression:

To examine these issues in more details, we performed new experiments addressing the plasticity response to chronic bursting activity in non-treated Xenopus synapses (same protocol than in Figure 1E). The small depression observed after chronic activity was not homogenous among tested synapses (Author response image 4, black dots). Interestingly, the degree of depression of a given synapse seems dependent on how far its initial Gsyn/Gin ratio was from the averaged ratio of the population measured after chronic activity (Author response image 4A, inset, dotted line). Consequently, standard deviation from the mean diminished after chronic activity. We interpret this synaptic behavior as a convergence towards the set point of homeostasis, induced by the evoked bursting activity.

Author response image 4
Homeostatic control of the synaptic efficacy.

(A) In Xenopus, ratios between averaged synaptic conductance ('Gsyn', calculated from 30–40 ePSCs) and muscle cell input conductance ('Gin') before and after 20-30 min of chronic burst stimulation of the motor neuron (burst of 20–60 events at a 20–30 Hz frequency, every 30–40 s) under postsynaptic current-clamp in non-treated (black dots) and low curare-treated (red dots) synapses. Green dots represent the Gsyn/Gin ratio before and after 1–3 bursts of 5 presynaptic stimulations at 30 Hz in ryanodine loaded muscle cells (same data than in Figure 3C, ryanodine bar). Inset show the mean ± standard deviation of the Gsyn/Gin ratios in the three conditions. The dotted lines show the mean Gsyn/Gin ratio after chronic activity in non-treated synapses. (B) Degree of plasticity shown in A expressed as a function of the difference between the initial individual Gsyn/Gin ratio and the mean ratio after chronic burst activity ('Distance to the set point'), in non-treated (black dots) and curare-treated (red dots) synapses. The solid line shows the theoretical relationship between plasticity and the distance to a set point of 2.36, calculated as the mean Gsyn/Gin ratio after chronic activity in non-treated synapses.

https://doi.org/10.7554/eLife.12190.019

Gradation of potentiation:

In order to test whether synapses also converge after that the initial Gsyn/Gin ratio was decreased below the set point, we applied the same chronic activity in the continuous presence of low doses of curare (0.1–2 µM), as proposed by the reviewer (Author response image 4, red dots). In presence of curare, chronic burst stimulation induced convergence of synapses towards the same set point, with a potentiation magnitude also depending on how far the initial Gsyn/Gin ratio was from the averaged ratio after chronic activity.

“A truly homeostatic system”:

In a perfect homeostatic process, with an attractor set point, the degree of plasticity (relative change in the Gsyn/Gin ratio) can be expressed as a function of the distance to the set point (difference between the initial Gsyn/Gin ratio and the set point):

degreeofplasticity=111+setpointdistance

Author response image 4B shows this theoretical relationship (line) with a set point of 2.36 (the averaged Gsyn/Gin ratio obtained after chronic activity). In order to confront our data with a perfect theoretical homeostasis, we added to Author response image 4B the degrees of plasticity in non-treated (black dots) and curare treated (red dots) synapses shown in A, expressed as a function of the difference between the initial Gsyn/Gin ratio and the 2.36 set point.

In conclusion, gradation in both potentiation and depression could be obtained in curare-treated and non-treated synapses respectively, suggesting that this system “is truly homeostatic” and that potentiating and depressing mechanisms balance rather than gate each other.

Why is the strength of Xenopus synapses, in initial conditions (12-24 hours after cultivation start), correlated to the postsynaptic input conductance above the target set point reached after evoked activity? As discussed in the answer to point 2 of the questions of the reviewers, Xenopus motoneurons in culture do not spontaneously fire and the mechanisms responsible for the evoked ACh release take place under the influence of the spontaneous release, which has lower amplitudes (a minority reaches the threshold). The potentiation/depression balance mobilized by spontaneous release is in favor of potentiation, and this may explain that the evoked release is found in initial conditions above the normal set point.

The effects of ryanodine in Xenopus and mouse in regards to the set point of homeostasis:

In order to place the potentiation, we obtained with ryanodine treatment in regards to the homeostatic law described above, we normalized by Gin the synaptic conductances found in ryanodine-loaded Xenopus muscle cells in Figure 3C (“Ryanodine” bar), and we added these data to Author response image 4A (green dots).

In adult mouse neuromuscular synapses, an apparent Gsyn/Gin ratio can be estimated from the ePSPs in the following manner. If we assume a linear passive leak, the Gsyn/Gin ratio can be expressed as:

GsynGin=80VpVp with Vp the membrane potential reached by the ePSP and 0 and -80 mV the reversal potentials of the synaptic and the leak currents respectively. Author response image 5A shows the apparent Gsyn/Gin ratio in non-stimulated and in burst-stimulated preparations, both for non-treated and ryanodine-treated preparations. Contrary to Xenopus where the membrane of the muscle cells is isopotential, these ratios in mouse are only apparent since the recorded ePSP amplitude depends on the distance between the synapse and the recording site. This distance effect is limited in FDB muscle because the cells are “only” 300 µm long, but nonetheless the apparent Gsyn/Gin ratio is presumably lower than the actual ratio at the synapse location. For this technical reason, the distance effect might also participate to the variability in apparent ePSPs amplitudes.

Both in mouse and Xenopus, the ryanodine-treated synapses did not converge towards a set point. The averaged Gsyn/Gin after burst activity was higher than the set point, and the standard deviation from the mean was increased. These data suggest that the ryanodine receptors-dependent calcium signal participates to a dynamical balance stabilizing the synaptic efficacy at a set point.

Author response image 5
Plasticity in mouse.

(A) Apparent averaged Gsyn/Gin ratios calculated in mouse FDB muscles from the data of Figure 3E with equation X, in non-stimulated and non-treated synapses (no burst, black dot), in burst-stimulated and non-treated synapses (after chronic bursts, black dot), in non-stimulated and ryanodine-treated synapses (no burst, green dot) and in burst-stimulated and ryanodine-treated synapses (after chronic bursts, green dot). Dots represent the mean ± Standard Deviation. (B) Relative change of contraction force during 2s-30Hz bursts of nerve stimulations in mouse soleus muscles (n=4), before and during exposure of a low dose (0.1 µM) of curare.

https://doi.org/10.7554/eLife.12190.020

The goal of this work was to show that the physiological range of nicotinic calcium plays a potentiating role on the synaptic strength, that the AP-induced calcium release from the reticulum plays a depressing role, and that this depressing effect participates to stabilization of the neurotransmitter release in a supra-threshold range. However, important questions such as the precise roles of the two calcium signals in the balance, and the location(s) where the push-pull mechanism(s) acts require more experiments.

We have included the new data in the Results of the revised version of our manuscript under the title “Plasticity orientation rule is homeostatic”. A newly added figure contains the data presented above as well as a low curare experiment (0.1 µM) done on the mouse nerve-muscle preparation in order to show that the mouse neuromuscular synapse is also capable of compensatory potentiation under curare.

In this last data set, we took advantage of the fact that the muscle contraction force during a presynaptic burst integrates the synaptic efficacy over all the synapses and all the synaptic events. The Author response image 5B shows that burst stimulation of the nerve triggers a compensatory potentiation that results in the recovery of a normal contraction force.

3B) Timing between nicotinic calcium and DICR:

The timing approach is another interesting suggestion, but we think that this approach is more related to hetero-synaptic depression and synaptic selection than to homeostasis at a mono-innervated muscle cell. In order to show that the calcium release from reticulum through the ryanodine receptors induce synaptic depression we have directly triggered the muscle AP in absence of pre-synaptic stimulation. However, in the natural circumstances this situation occurs only in multi-innervated muscle cells, during development. In our work we did not investigate the possible role of the AP-associated calcium signal in heterosynaptic depression, a hypothesis which remains to be tested. It is tempting to hypothesize that if the global DICR signal (activated across the whole fiber) is associated with the release by all the muscle surface of a negative retrograde feedback factor, it could participate to the depression of the less active synapses in the case of a competition. Interestingly, Drosophila larva muscles, that do not possess the global DICR signal of the vertebrates, are multi-innervated.

3C) Is it surprising that the LTD is so stable in Figure 4E given that the ePSPs seem to all fail to reach threshold? Shouldn't this induce LTP?

Actually, it does, but slowly. Author response image 6 shows in a muscle, for which the nerve was relatively regularly stimulated at low rate, the slow increase in ePSPs amplitudes (r=0.23, p=0.07).

Author response image 6
In mouse, slow potentiation after depression.

ePSPs peak in individual FDB cells after depression induced by bursts of external stimulations in a Ca2+ free medium.

https://doi.org/10.7554/eLife.12190.021

As we discussed in our answer to point 2 of the reviewers, we attribute the apparent absence of potentiation (or slow potentiation) in Figure 4E to the low rate of stimulation used to measure the ePSPs. In the course of the 60 min test protocol, in Figure 4E and Author response image 6, each dot after the external stimulation correspond to ePSP peak in response to a discrete “test” stimulation of the nerve. ePSPs are sequentially recorded in different muscle fibers, resulting in a low rate stimulation of the nerve. The Figure 4—figure supplement 4 of the initial submission (Figure 4—figure supplement 3 in the revised version) was dedicated to show that rapid LTP is indeed inducible after depression as expected by the reviewer, if higher frequency of nerve stimulation is used (arrows, bursts of nerve stimulations at 20-30Hz). Similarly, in Xenopus, potentiation is obtained by motoneuron stimulation (1-3 bursts of 5 events at 30Hz) after depression induced by direct muscle AP (1-3 bursts of currents steps into the muscle cells at 30Hz) (not shown). For clarity, we removed the Figure 3—figure supplement 3 where we did not wait enough in Xenopus to restore the normal synaptic gain after depression.

3D) Dependence on Frequency and number of postsynaptic events

We did not attempt to perform such studies for the revised paper. Instead we have focused on the new experiments described above, that show the graded behavior of plasticity for both depression and potentiation.

3E) Given previous observations from Drosophila and patients with myasthenia gravis, it might be expected that sub-blocking levels of curare would induce increased ACh release independent of bursting.

This point is directly related to the point 2 of the essential revisions. Please see our detailed answer to point 2.

3F) Is there a threshold level of nicotinic calcium required to induce LTP?

The apparent threshold number and frequency of stimulations necessary to induce potentiation in Wan & Poo 1999 (Figure 3) could suggest that there is a threshold of nicotinic calcium build-up required to induce rapid potentiation. Related to this point, we detailed in the point 1 of the essential revisions that the nicotinic calcium is strongly dependent on the membrane potential. For plasticity induced under post-synaptic Vclamp, in contrast with a -80 mV holding potential, the temporary use of a -50 mV holding during the conditioning burst reveals a threshold synaptic conductance to induce a rapid potentiation (Author response image 7), suggesting that a threshold level of nicotinic calcium is required.

Author response image 7
Dependency of plasticity under postsynaptic Vclamp on the holding value during the conditioning burst.

Same protocol than Wan & Poo 1999 (Figure 1A (ii)) with different holding potentials during the conditioning burst.

https://doi.org/10.7554/eLife.12190.022

3G) One related concern is the safety factor for NMJ transmission – if homeostasis is ongoing wouldn't it tend to obliterate excess EPSP above that needed to evoke an AP?

Yes, in the case of a gating mechanism. But we argued in the text above that a dynamical balance is more likely than a gating. In the case of a dynamical balance, the mean level of synaptic conductance reaching the equilibrium is difficult to predict when we don’t know the location(s) and the nature(s) of the push-pull mechanism(s). Equilibrium is not necessarily reached at the spike threshold.

https://doi.org/10.7554/eLife.12190.024

Article and author information

Author details

  1. Gilles Ouanounou

    Unité de Neuroscience Information et Complexité, Centre National de la Recherche Scientifique, FRE 3693, Gif-sur-Yvette, France
    Contribution
    GO, Designed the research, Performed experiments, Analysis and interpretation of data, Wrote the paper
    For correspondence
    gilles.ouanounou@unic.cnrs-gif.fr
    Competing interests
    The authors declare that no competing interests exist.
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-5658-7005
  2. Gérard Baux

    Unité de Neuroscience Information et Complexité, Centre National de la Recherche Scientifique, FRE 3693, Gif-sur-Yvette, France
    Contribution
    GB, Contributed with critical discussion on the manuscript, Drafting or revising the article
    Competing interests
    The authors declare that no competing interests exist.
  3. Thierry Bal

    Unité de Neuroscience Information et Complexité, Centre National de la Recherche Scientifique, FRE 3693, Gif-sur-Yvette, France
    Contribution
    TB, Provided the dynamic-clamp technique, Contributed with critical discussion on the manuscript, Drafting or revising the article
    Competing interests
    The authors declare that no competing interests exist.

Funding

Centre National de la Recherche Scientifique

  • Gilles Ouanounou
  • Gérard Baux
  • Thierry Bal

Association Française contre les Myopathies

  • Gilles Ouanounou
  • Gérard Baux

Fondation pour la Recherche Médicale

  • Gilles Ouanounou
  • Thierry Bal

Neuropole de recherche francilien

  • Gilles Ouanounou
  • Thierry Bal

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We thank Jean-Pierre Changeux, David McCormick and Yves Frégnac for critical discussions about the manuscript, Zuzanna Piwkowska, Evan Harrell and Francesca Barbieri for proof reading, and Patrick Parra and Jean-Yves Tiercelin for shaping the mechanical parts of the experimental setups.

GO, GB and TB were supported by the 'Centre National de la Recherche Scientifique' (CNRS). GO and GB were supported by a grant from 'Association Française contre les Myopathies'. GO and TB were supported by grants from 'Neuropôle de Recherche Francilien' and 'Fondation pour la Recherche Médicale'.

Ethics

Animal experimentation: Animal care followed the European Union regulations (O.J. of E.C. L358/1 18 498 December 1986), and the European directive 2010/63/UE.

Reviewing Editor

  1. Sacha B Nelson, Brandeis University, United States

Publication history

  1. Received: October 9, 2015
  2. Accepted: April 7, 2016
  3. Version of Record published: May 3, 2016 (version 1)

Copyright

© 2016, Ouanounou et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 1,343
    Page views
  • 294
    Downloads
  • 17
    Citations

Article citation count generated by polling the highest count across the following sources: Scopus, Crossref, PubMed Central.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)

Further reading

    1. Cell Biology
    2. Chromosomes and Gene Expression
    Isabelle C Kos-Braun et al.
    Feature Article Updated

    Core facilities are an effective way of making expensive experimental equipment available to a large number of researchers, and are thus well placed to contribute to efforts to promote good research practices. Here we report the results of a survey that asked core facilities in Europe about their approaches to the promotion of good research practices, and about their interactions with users from the first contact to the publication of the results. Based on 253 responses we identified four ways that good research practices could be encouraged: (i) motivating users to follow the advice and procedures for best research practice; (ii) providing clear guidance on data-management practices; (iii) improving communication along the whole research process; and (iv) clearly defining the responsibilities of each party.

    1. Cell Biology
    2. Physics of Living Systems
    Andrea Serra-Marques et al.
    Research Article Updated

    Intracellular transport relies on multiple kinesins, but it is poorly understood which kinesins are present on particular cargos, what their contributions are and whether they act simultaneously on the same cargo. Here, we show that Rab6-positive secretory vesicles are transported from the Golgi apparatus to the cell periphery by kinesin-1 KIF5B and kinesin-3 KIF13B, which determine the location of secretion events. KIF5B plays a dominant role, whereas KIF13B helps Rab6 vesicles to reach freshly polymerized microtubule ends, to which KIF5B binds poorly, likely because its cofactors, MAP7-family proteins, are slow in populating these ends. Sub-pixel localization demonstrated that during microtubule plus-end directed transport, both kinesins localize to the vesicle front and can be engaged on the same vesicle. When vesicles reverse direction, KIF13B relocates to the middle of the vesicle, while KIF5B shifts to the back, suggesting that KIF5B but not KIF13B undergoes a tug-of-war with a minus-end directed motor.