Activity-dependent decline and recovery of synaptic transmission in central parts of surviving primary afferents after their peripheral cut in crayfish

Unlike vertebrates, in which axons do not survive when deprived of their nucleus (Koeppen, 2004), the peripheral parts of cut soma-deprived motor axons do not degenerate in crustaceans owing to the invading glial soma (Bittner and Baxter, 1991; Parnas et al., 1998). This distal axon survival allows for their functional reconnection to the outgrowing proximal stump of the motor axon (Hoy et al., 1967), as indicated by the regenerative sprouting response observed both in motor peripheral and sensory central cut ends (Kennedy and Bittner, 1974; Cooper, 1998). Moreover, during the first months, soma-deprived motor axons keep their capacity to conduct spikes and trigger transmitter release (Bittner, 1973). The same observations were made for sensory axons, which have their cell bodies located in peripheral sensory organs. Once cut, soma-deprived central axons maintain their spike conduction and synaptic release capacities. This is also accompanied by glial cell reorganization (Govind et al., 1992).

Peripheral motor axons of cut motor nerves continue to conduct spikes for at least a year after they are cut (Atwood et al., 1989), with a velocity that is not different from that of control axons (Parnas et al., 1991). However, even though such cut axons continue to elicit neurotransmitter release, the kinetics of synaptic release undergo deep changes across time. In particular, the decay phase of quantum events, which follows a single exponential decay in control preparations, is much slower in cut axons. It can no longer be fitted with a single exponential decay. Moreover, the duration of the synaptic event increases progressively after the section, owing to a longer decay phase. This longer decay phase is probably explained by post-synaptic mechanisms (Parnas et al., 1991). An increased delay was also observed in these preparations (Parnas et al., 1991).

Concerning sensory axons, their survival and degeneration after nerve cut was studied in crayfish tailfan sensory nerves (Govind et al., 1992). Unlike motor axons, in this model, 95% of sensory axons deprived of their cell body underwent a degenerative process (Govind et al., 1992). Central connections of remaining axons onto target neurons were deeply modified. Surviving sensory axons were invaded by glial cell nuclei (Govind et al., 1992), as was observed in motor axons.

In axons that have been cut from their cell body (i.e. peripheral stumps for motor axons, and central parts for sensory axons), the absence of spikes for long periods of time (up to 1 year) could lead to dramatic changes in synapse function. Indeed, activity-dependent changes in synaptic function have been shown in invertebrates (Lnenicka, 2020; Goel and Dickman, 2021) and vertebrates (Magee and Grienberger, 2020). In Hebbian rules of synaptic plasticity, when a presynaptic neuron is not involved in the activity of the post-synaptic neuron, its synaptic transmission decreases. If such a mechanism is present, this lack of activity could explain the synaptic changes observed in motor (Parnas et al., 1991) and sensory (Govind et al., 1992) axons deprived of their cell body and source of activity.

In this study, we addressed this question by studying the synaptic transmission in a well-documented sensorimotor circuit of the crayfish composed of the coxa-basipodite chordotonal organ (CBCO), which contains sensory neurons. The resistance reflex is ensured via monosynaptic connections from release-sensitive CBCO axons to depressor motoneurons (MNs) and from stretch-sensitive CBCO axons to levator MNs (El Manira et al., 1991a,b), and disynaptic inhibitory connections from release-sensitive CBCO axons to levator MNs and from stretch-sensitive CBCO axons to depressor MNs (Le Bon-Jego and Cattaert, 2002). After in vivo section of the sensory CBCO nerve, the sensorimotor system was dissected out at different time periods (from a few days to more than 6 months post lesion) to be studied in vitro. This allowed us to follow the evolution of monosynaptic excitatory synaptic transmission and disynaptic inhibitory transmission. In parallel, electron microscopy studies were carried out on the central part of the cut sensory nerve to confirm the survival of CBCO axons over months, and glial cell invasion. In a series of experiments, we show that all synaptic transmission is lost 3 months after peripheral cut of the CBCO nerve. To test whether this loss was due to the absence of sensory activity, we repeated these experiments but kept a sustained orthodromic spiking activity during the time period following the cut of the sensory nerve. This was done by placing two stimulating cuff electrodes on the central end of the cut nerve. We also show that, even in the absence of sensory activity for long periods of time in vivo, it is still possible to re-establish functional synaptic connections with MNs and with inhibitory interneurons (INs) of the disynaptic pathway.

Experiments were performed on adult crayfish [Procambarus clarkii (Girard 1852)] of either sex measuring approximately 12 cm. Animals were purchased from a commercial supplier (Chateau Garreau, France), and maintained in indoor aquaria at 15–18°C. Animals were fed once a week. After surgery, animals were kept in individual aquaria.

The sensory nerve of the primary afferent is very superficial. This renders possible both the section of the nerve (as previously described in Le Ray et al., 2005) and the implantation of extracellular electrodes in vivo on anesthetized animals. The procedure used for chronic recordings was adapted from the technique developed by Böhm (1996). Animals were anesthetized on ice, and immobilized (as previously described in Le Ray et al., 2005) for dissection.

A small piece of cuticle from the coxo-basipodite joint of the fourth walking leg was removed to access the CBCO sensory nerve and to section it. After the efficiency of the section was confirmed by the disappearance of the resistance reflex (Fig. 1Ai,ii), the hole in the cuticle was filled with a mixture of rosin and heated beeswax. Animals were then placed in individual aquaria and later used for recordings or electron microscopy.

Four thin monopolar wires traveling in a grounded cable fixed with wax on the back of the animals were used. Three of them were attached to 50 µm insulated wire electrodes to record muscular and nerve activities (Fig. 1), whereas a fourth electrode was placed under the carapace and used as a reference. Electrodes were isolated from the hemolymph with a flexible silicone elastomer Kwik-Cast (World Precision Instruments). This two-component very low viscosity silicone sealant was developed to embed peripheral nerves with electrodes for acute multi-fiber recordings. Each recording electrode was fixed separately on the leg cuticle with wax before it reached its attachment to the grounded cable and connection to homemade extracellular amplifiers. Amplified neurograms and electromyography signals were directed through a CED 1401 interface (Cambridge Electronic Design, Cambridge, UK) to a computer for storage and analysis. For the chronic stimulation of the sensory nerve, trains of 500 ms duration were applied at a frequency of 0.25 Hz. Each train was composed of 0.3 ms pulses at 20 Hz.

After various delays (from 1 to 180 days after CBCO nerve section) (see protocol in Fig. 1B), the ventral thoracic nerve cord was dissected out with the sensory and motor nerves to the fourth left leg. This in vitro preparation was used to study sensory–motor connections (El Manira et al., 1991a; Le Bon-Jego and Cattaert, 2002; Le Bon-Jego, 2004; Le Bon-Jego et al., 2006). It consists of the last three thoracic (T3–T5) ganglia dissected out along with all the nerves (motor and sensory) of the two proximal segments of the left fourth leg (Fig. 1). In control experiments, the chordotonal organ (CBCO), which monitors movements of the second joint (coxo-basipodite), was also dissected out and kept intact. For the operated animals, the remaining part of the sensory nerve without the CBCO was carefully dissected. The preparation was pinned down dorsal side up in a silicone elastomer covered Petri dish. The fourth ganglia were de-sheathed to improve the continuous superfusion of central neurons with oxygenated saline and to allow intracellular recordings of all depressor MNs. The saline was composed of (in mmol l⁻¹): 195 NaCl, 5 KCl, 13 CaCl₂ and 2 MgCl₂, buffered with 3 mmol l⁻¹N-2-hydroxyethylpiperazine-N′-ethanesulfonic acid (Hepes, Sigma-Aldrich) and pH adjusted to 7.65 at 15°C (optimal pH value to obtain the best survival state for in vitro preparation; Le Ray and Cattaert, 1997). In some experiments, in order to increase the spiking threshold of INs, we used a high divalent cation solution containing (in mmol l⁻¹): 34 CaCl₂ and 6.4 MgCl₂ with a sodium concentration reduced accordingly to preserve the osmolarity of the solution (157 NaCl). The use of this altered saline allowed us to identify the monosynaptic reflex responses (Berry and Pentreath, 1976).

Extracellular recordings from motor nerves innervating depressor and levator muscles and from the sensory nerve of the CBCO were performed using stainless steel pin electrodes contacting the nerves and insulated from the bath with Vaseline. Recorded signals were amplified by differential AC amplifiers (Grass, Quincy, MA, USA, gain of 10,000×). Intracellular recordings from depressor MNs were performed from their main neurite within the ganglion using glass micropipettes (Clark Electromedical Instruments, Reading, UK) filled with 3 mol l⁻¹ KCl (resistance 20–25 MΩ) connected to an Axoclamp 2B amplifier (Axon Instruments, Foster City, CA, USA) used in the current-clamp mode. Depressor MNs were identified using the following criteria: (1) spikes evoked by electrical stimulation of the depressor nerve were recorded by the intracellular microelectrode and (2) spikes evoked by a depolarizing current injected into the intracellularly recorded neuron were correlated one-to-one with the extracellular spikes recorded in the depressor nerve. Note that 12 depressor MNs exist (Le Ray and Cattaert, 1997). We aimed at recording these 12 depressor MNs in each in vitro experiment. However, owing to the difficulty of finding each depressor MN main neurite in the neuropile (blind search), we failed to record some of them in some experiments. An eight-channel stimulator (A.M.P.I., Jerusalem, Israel) was used for intracellular stimulation of MNs during the identification procedure, and for CBCO nerve stimulation. Data were digitized and stored onto a computer hard disk (sampling speed 10 KHz) through an appropriate interface (1401plus) and software (Spike2) from Cambridge Electronic Design.

The changes in sensory–motor connections after section of the CBCO sensory nerve were tested in vitro. The CBCO nerve (Fig. 4) was electrically stimulated and the evoked response in the depressor MN was recorded intracellularly (Fig. 4). Classically, this electrical stimulation protocol activates simultaneously the two components of the resistance reflex [monosynaptic excitation (excitatory post-synaptic potential, EPSP) and disynaptic inhibition (inhibitory post-synaptic potential, IPSP)] because release- and stretch-sensitive CBCO neurons are located in the same nerve (Fig. 4).

Data were analyzed using the Spike2 analysis software. Statistical analyses were performed with Prism (Graphpad Software, San Diego, CA, USA). Results are given as means±s.e.m. The significance of the effect of peripheral axotomy on the synaptic connection (number of MNs eliciting an EPSP or an IPSP) over time was assessed by one-way ANOVA followed by Tukey's multiple comparison tests (Fig. 4). Comparison between evolution of EPSPs and IPSPs over time was assessed by two-way ANOVA (Fig. 4). Changes in synaptic delays after a given period after axotomy were assessed by unpaired t-tests.

The CBCO sensory nerve from control and operated animals was collected and fixed overnight at 4°C in glutaraldehyde (0.1%). Samples were then rinsed in phosphate buffer containing sucrose, post-fixed at room temperature in osmium 1% for 45 min and stained with uranyl acetate (2%). Dehydration was obtained through an ascending series of ethanol baths. Nerves were embedded in araldite and resin, then polymerized at 60°C for 3 days. Ultrathin sections (70–80 nm) were collected on nickel grids contrasted with uranyl acetate and lead citrate. They were then observed with a transmission electron microscope (Hitachi H600). For some experiments, grids were immunostained for GABA following the protocol in Watson et al. (2005). Briefly, sections were treated at room temperature with 2% periodic acid (3 min) and in sodium metaperiodate at 50°C (3 min), then washed in Tris buffer (TB), pH 7.2. After 30 min in 5% normal goat serum diluted in TB, the grids were transferred for 2 h in polyclonal rabbit anti-GABA at 1:1000 (Sigma-Aldrich) in TB. After further washing, they were transferred for 1 h to 15 nm gold-labeled goat anti-rabbit antibody (British Biocell International, Cardiff, UK) diluted 1:20 in TB, pH 8.2. Finally, sections were counterstained with uranyle acetate and lead citrate. Preabsorption with GABA conjugated to bovine serum albumin with glutaraldehyde eliminated all labelling with GABA antibody.

The surface occupied by glial cells was calculated as follows. Using ImageJ, the contours of glial cells were manually drawn on the electron microscopy image. The area and perimeter of the contours were then calculated by ImageJ, using the scale bar, and the total image area was used to calculate the percentage of surface occupied by glial cells.

In this study, we modeled the interaction between a GABA synapse responsible for primary afferent depolarizations (PADs; responsible for presynaptic inhibition) and a disynaptic inhibitory circuit of reciprocal inhibition (Le Bon-Jego and Cattaert, 2002; Le Bon-Jego et al., 2004) (see Fig. 2). The GABA synapse producing PADs is located on a first-order branch in the region of the first branching point of a primary afferent (CBCO) sensory fiber in the ganglion neuropile (Cattaert et al., 1992, 2001; Cattaert and El Manira, 1999). The synapses from CBCO to the inhibitory interneuron of the reciprocal inhibition are located on the CBCO main axon (Fig. 2A,B). The effect of PADs was simulated using the NEURON 7.3 program (Hines and Carnevale, 1997). The temporal integration step was 20 µs.

Five neurons were simulated: a primary afferent fiber and terminal (in green in Figs 2C and 8), an MN, a primary afferent depolarization interneuron (PADI; inhibitory neuron producing PADs in the CBCO terminal; in blue in Figs 2C and 8) and a reciprocal inhibition interneuron (in red in Figs 2C and 8). Each neuron is made of a soma, an axon initial segment (AIS), an axon, and synaptic compartments (either for input or output). The two antagonistic MNs (depressor and levator) were more complex, with some passive sections (soma, thin neurite between soma and large neurite – neurite 1 and neurite 2) and some active sections (AIS and axon). Note that input synapses on the MN were placed onto neurite 1 (see Fig. 2C). The primary afferent conveyed spikes in the axon but not in the terminal branches. The dimensions and properties of all neuron compartments are given in Table 1.

Note that Na/K channels were present in CBCO branches, but their density did not allow active conduction of spikes.

All compartments had the same specific membrane resistance (R_m=3000 Ω cm²). All computations were carried out assuming a specific capacitance, C_m, of 1 µF cm⁻² and a specific axoplasmic resistance, R_a, of 100 Ω cm.

Active compartments were equipped with Na and K Hodgkin–Huxley channels. The equilibrium potential for Na⁺ ions was set at E_Na=+40 mV. The equilibrium potential for K⁺ ions was E_K=−90 mV. The density of Na and K channels in each compartment is given in Table 1.

The activation level of the postsynaptic channel, s, ranges from 0 to 1 when the synapse is ‘closed’ or ‘open’, respectively. The maximum conductance G_syn is fixed, but the value of the actual conductance, G_syns, varies with s; that is, when the synapse opens and closes. The kinetics of s are controlled by two parameters: GABA synapses: tau rise=1 ms; tau decay=5 ms; and excitatory synapses: tau rise=0.2 ms; tau decay=3.5 ms.

For GABA synapses, specifically PADs, G_Cl was fixed to 110 nS and E_Cl was fixed to −35 mV; for other GABA synapses, G_Cl was fixed to 0.11 μS and E_Cl was fixed to −70 mV. For excitatory synapses, G_Exc was fixed to 22 nS and E_Exc was fixed to 20 mV.

Six months after axotomy, we analyzed the ultrastructure of the sensory nerve. The axons of the central part of the sectioned sensory nerve did not show noticeable change (Fig. 3A,B). By contrast, glial cell populations underwent deep reorganization. Among all profiles in electron microscopy, glial cells were identified by their nuclear electron density profile as they have a higher electron density than other profiles. Each nucleus was counted as one cell. Their density increased from 1.37±0.25 cells 1000 µm^–2 (n=5 animals) in control to 4.38±1.20 cells 1000 µm^–2 (n=4 animals) 6 months after axotomy (Fig. 3H). Their mean individual surface area did not change (8.89±1.20 µm², n=12 cells in control to 8.62±0.60 µm², n=32 cells after axotomy, Fig. 3I). Consequently, the global percentage of surface occupied by glial cell increased from 1.21±0.28% (n=5 animals) in control to 3.41±0.76% (n=4 animals) after axotomy (Fig. 3J).

Glial cell proliferation was accompanied by a change in their anatomical disposition with respect to axons. In the control (Fig. 3C), glial cell nuclei were clearly separated from axons, and remained outside of the endoneurium. Six months after axotomy, glial cell nuclei migrated inside endoneuria, tending to enwrap axons (Fig. 3D). Moreover, the close contact between axons and invading glial cell nuclei presents anatomical features characterized by invagination (Fig. 3E) and possible membrane discontinuities (Fig. 3F,G) (we did not perform serial section analysis to confirm such discontinuities).

After cutting the peripheral sensory nerve from its sensory organ (CBCO), which contains the cell bodies, sensory axons did not convey any sensory information to the central nervous system, but displayed sparse bursting antidromic activity (see below) (Fig. 7).

To assess the evolution of central synapses from sensory CBCO axons onto target neurons (MNs and INs) after sectioning the CBCO nerve in vivo, dissections were performed after various periods (from a few weeks to >6 months post CBCO nerve section; Fig. 4B). In these in vitro preparations, we electrically stimulated the remaining central part of the sensory nerve. The electrical stimulation of the sensory nerve evoked mixed excitatory/inhibitory responses in intracellularly recorded MNs as it stimulated both the direct stretch reflex excitatory pathway (Cattaert et al., 1992) (Fig. 4Ai) and the inhibitory reciprocal innervation circuits (Le Bon-Jego and Cattaert, 2002) (see Fig. 2C for a schematic diagram of the circuit). In single experiments, all 12 depressor MNs controlling movements of the coxo-basipodite joint (Fig. 4Aii) were intracellularly recorded, allowing us to follow the evolution of identified sensorimotor synapses (Fig. 4Bi) over time. In the control condition, nine out of the 12 depressor MNs had monosynaptic EPSPs, while three of them showed only polysynaptic PSPs (Le Ray and Cattaert, 1997) (Fig. 4A). During the 3 weeks that followed the nerve section, monosynaptic EPSPs progressively declined, and less than 10% of MNs presented a monosynaptic EPSP (Fig. 4Bii,iii). By contrast, in 60% of MNs, disynaptic IPSPs could still be recorded up to 6 months after section (Fig. 4Bii,iii). After this delay, no PSPs were ever recorded from MNs (n=3 animals, n=29 MNs).

Surprisingly, even after the complete loss of synaptic transmission, 6 months after sensory nerve section, it was possible to restore synaptic function by applying repetitive electrical stimulation (0.2 Hz) of the sensory nerve for 60 min in the in vitro preparation (Fig. 5Ai). After this period of stimulation, each electrical stimulus elicited an IPSP in the intracellularly recorded MN (Fig. 5Aii,iii). In the following stimuli (i.e. at t=60 min, see Fig. 5Aiii), polysynaptic EPSPs were progressively recruited (see the progressive weakening of IPSP peak, orange arrow in Fig. 5Aii), and at t=70 min some monosynaptic EPSPs were also observed (see fast EPSP in Fig. 5Aii, yellow arrow and Fig. 5Aiii). In the experiment illustrated in Fig. 5Bi, synaptic transmission was recovered in each of the 11 recorded depressor MNs. Similar results were obtained in two other experiments with a total of 29 restorations of PSP out of the 30 intracellularly recorded depressor MNs (Fig. 5). However, the recovered monosynaptic responses were significantly delayed (6.07±0.33 ms instead of 4.01±0.42 ms in control, P<0.001; Fig. 5Bii), whereas the delay of inhibitory responses was unchanged (13.0±0.74 ms instead of 12.83±1.50 ms in control, P=0.91; Fig. 5Biii).

The loss of sensorimotor synaptic transmission described above was activity dependent. Indeed, when the sensory nerve was sectioned, chronic electrical stimulation (Fig. 6Ai) applied in vivo to the proximal end (Fig. 6Ai,Bi–iii) was sufficient to avoid the loss of sensorimotor synaptic transmission (Fig. 6C,D) that was observed after 45 days (see Fig. 4B). In the in vitro preparations issued from these in vivo experiments, the electrical stimulation of the proximal end of the sensory nerve evoked an EPSP in 71.96±3.22% of the recorded MNs (Fig. 6Di). This percentage was not significantly different from control preparations (74.24±0.75%), whereas in the absence of chronic stimulation, only 8.0±4.9% of MNs displayed a monosynaptic EPSP. A similar tendency was observed for IPSPs (Fig. 6Dii) (82.27±2.35% with chronic stimulation, 77.52±7.11% in control, and 46.67±13.32% without chronic stimulation), but those changes were not significant owing to the lesser extent of IPSP loss.

After peripheral section of the CBCO nerve, antidromic spikes were recorded in the proximal part of the nerve (connected to the central nervous system) (Fig. 7A). We previously showed that such antidromic spikes are produced by PAD elicited by depolarizing GABA synaptic events from PADI (Cattaert et al., 2001). This was also the case 10 days after CBCO nerve section in vivo (Fig. 7B), and in vitro preparations made 20 days after CBCO nerve section in vivo (Fig. 7C). Given the anatomical disposition of output synapses from CBCO onto GABA INs (Watson et al., 2005), it is possible that such synapses located on the main axon of a CBCO in the central nervous system (Fig. 2A,B) may be activated during PADs occurring slightly more distantly in a CBCO branch. This hypothesis would explain how, in the absence of sensory spikes, the reciprocal inhibitory pathway (Le Bon-Jego and Cattaert, 2002) could be activated by antidromic spikes (Figs 2C, 7), which would have preserved it, and would explain why it was easily restored even 9 months after the peripheral cut of the CBCO nerve. We tested this possibility by simulating the involved circuits (Figs 2C, 8A,B) using the NEURON 7.3 program (Hines and Carnevale, 1997), and by considering the spatial arrangement of output and input GABA synapses (Fig. 2D). This simulation confirmed that PADs that elicit antidromic spikes in the main CBCO axon are capable of recruiting output proximal synapses activating the reciprocal inhibitory pathway (Fig. 8A). Note that although an IPSP is evoked in the levator MN (Fig. 8B), the monosynaptic excitatory synapse to the depressor MN is not activated and no EPSP is produced in the depressor MN (Fig. 8B).

Our study outlines the absence of anatomical and functional degeneration in axons and their central synapses in neurons deprived of nuclei for long periods of time (>6 months). However, stimulation of the central part of the cut CBCO nerve produced fewer and fewer synaptic responses in MNs up to 6 months after section, at which time no response was observed. We showed that the failure of evoked synaptic activity 6 months after CBCO nerve section is the result of the lack of activity, and can be restored by repetitive electric stimulation of the CBCO nerve, even after 6 months. These findings indicate that the machineries for spike propagation and synaptic transmission persist within the central nervous system in axotomized sensory neurons deprived of their cell bodies. This absence of degeneration of axon functionality is accompanied by a deep reorganization of the glial cell population (proliferation and migration in endoneurium to get in close contact with the axons). Our study therefore confirms previous findings on the absence of anatomical and functional degeneration of sensory axons (Govind et al., 1992; Cooper, 1998) and motor axons (Parnas et al., 1998) deprived of nuclei in crustaceans. In these studies, proliferation and migration of glial cell nuclei in the axon tube were observed. Indeed, long survival of invertebrate axons deprived of their nucleus, and their capacity to conduct action potentials, has been reported in invertebrates (Hoy et al., 1967; Nordlander and Singer, 1972; Wine, 1973; Bittner and Johnson, 1974; Ballinger and Bittner, 1980; Bittner, 1988; Atwood et al., 1989; Blundon et al., 1990; Bittner and Baxter, 1991; Parnas et al., 1991; Sheller and Bittner, 1992) and vertebrates (Matsumoto and Scalia, 1981; Cancalon, 1982; Lubińska, 1982; Zottoli et al., 1987; Blundon et al., 1990).

Here, we were interested in the activity dependence of the synaptic function of nucleus-deprived sensory axons. After 9 months of nucleus deprivation, the stimulation of the peripheral part of the sensory nerve did not produce any response in post-synaptic MNs (Fig. 4). However, after 1 h of stimulation, post-synaptic responses were recorded in MNs, indicating that the machineries to conduct spikes and to activate functional synapses could still be activated. Why was the first hour unsuccessful? There are two possibilities: (1) a spike conduction blockade that was restored by electrical stimulation, or (2) a synaptic machinery blockade. Indeed, previous observations in crayfish showed that sensory nerve section did not immediately prevent nerve activity, induced by electrical stimulation, to activate the target networks (Govind et al., 1992). The same stimulation did not evoke any response 3 weeks after the lesion; however, because this was due to axonal degeneration, synaptic function was not explored. In contrast, decentralized motor axons controlling lobster deep abdominal extensors continue to conduct spikes and evoke post-synaptic responses even 1 year after nerve cut (Parnas et al., 1991).

Owing to the absence of the sensory structure, no sensory activity could be recorded in the sensory nerve (Fig. 1Aii). However, antidromic activity was recorded (Fig. 7). In freely moving animals, such antidromic activity is linked to the central control of sensory terminal activity via presynaptic inhibition produced by depolarizing GABA events in invertebrates (Cattaert et al., 1992, 2001; Cattaert and El Manira, 1999) and vertebrates (Dubuc et al., 1985; Vinay et al., 1999). It seems, however, that the antidromic activity fades over time, and it was rarely observed after 6 months of nucleus deprivation. Here, we have shown that maintenance of output synapses on primary afferents is dependent on spiking activity as chronic stimulation of the central end of the cut CBCO nerve prevented the loss of synaptic transmission (Fig. 6). Note that the loss of PSP in post-synaptic target neurons was not due to spike conduction problems as spikes could still be recorded in the main nerve trunk after electrical stimulation of the sensory nerve (data not shown). Although we did not explore the mechanisms of this activity dependence of synaptic function, our results indicate that such a control exists. Moreover, we hypothesize that differential effects of antidromic spiking activity on (1) excitatory synapses to MNs (monosynaptic resistance reflex circuit) and (2) disynaptic inhibitory connections to antagonist MNs (reciprocal inhibitory circuit) could explain why responses did not disappear at the same time (Fig. 4Biii). Owing to their anatomical arrangement, synapses from primary afferents onto MNs are located at the ending of sensory axons, in an area where antidromic spikes are not actively propagated (Cattaert et al., 1992; Cattaert and El Manira, 1999; Cattaert et al., 2001). Therefore, these synapses would not be maintained by antidromic activity. By contrast, synapses from primary afferents onto inhibitory INs of the reciprocal inhibitory circuit are located in the region of the first branch of the main sensory axons within the central nervous system (Fig. 2A–C), where PADs can elicit spikes. This hypothesis was validated with a simulation pointing out this difference in spiking activity due to PADs, in the vicinity of these two types of synapses (Fig. 8A,B). Antidromic activity was recorded in intact sensory nerves and the central part of cut sensory nerves. Although this antidromic activity tends to decrease over the months after lesion, the presence of spikes could explain the difference observed between EPSPs and IPSPs triggered by electrical stimulation of the central part of the CBCO nerve (Fig. 4Biii). Nevertheless, because two synapses are involved in the disynaptic inhibitory circuit, the question remains open as to what occurs to the synapse between the IN and the MN. If PADs are strong enough to activate the IN, then the synapse from the IN to the MN could follow the same evolution. If it is not the case, then it is also possible that these INs can be activated by other neurons, as was shown for the 1a INs in vertebrates (Jankowska and Hammar, 2013).

We hypothesize that the differential effects of antidromic spikes (triggered by PADs) on monosynaptic EPSPs and disynaptic IPSPs could be the reason for the order of reappearance of PSP during repetitive CBCO nerve stimulation (Fig. 5). Because the synapse to inhibitory INs had been regularly stimulated by antidromic spikes during the first months after nerve section, the degree of blocking machinery was less than the synapse to MNs that was not activated by antidromic spikes. This would imply that the process of synapse blocking is somewhat gradual. More work on the activity dependence of synapse maintenance in crustaceans is needed to understand the involved mechanisms and explain its gradual nature.

We are grateful to Anaelle Braine for language assistance on the manuscript.