Calcineurin and cytoskeleton in low-frequency depression


Address correspondence and reprint requests to Milton P. Charlton, Department of Physiology, Medical Sciences Building, Rm. 3308, University of Toronto, 1 King’s College Circle, Toronto, ON, Canada M5S 1A8. E-mail:


J. Neurochem. (2009) 109, 716–732.


Transmitter release at high probability phasic synapses of crayfish neuromuscular junctions depresses by over 50% in 60 min when stimulated at 0.2 Hz. Inhibition of the protein phosphatase calcineurin by intracellular pre-synaptic injection of autoinhibitory peptide inhibited low-frequency depression (LFD) and resulted in facilitation of transmitter release. Since this inhibitor had no major effects when injected into the post-synaptic cell, only pre-synaptic calcineurin activity is necessary for LFD. To examine changes in phosphoproteins during LFD we performed a phosphoproteomic screen on proteins extracted from motor axons and nerve terminals after LFD induction or treatment with various drugs that affect kinase and phosphatase activity. Proteins separated by PAGE were stained with phospho-specific/total protein ratio stains (Pro-Q Diamond/SYPRO Ruby) to identify protein bands for analysis by mass spectrometry. Phosphorylation of actin and tubulin decreased during LFD, but increased when calcineurin was blocked. Tubulin and phosphoactin immunoreactivity in pre-synaptic terminals were also reduced after LFD. The actin depolymerizing drugs cytochalasin and latrunculin and the microtubule stabilizer taxol inhibited LFD. Therefore, dephosphorylation of pre-synaptic actin and tubulin and consequent changes in the cytoskeleton may regulate LFD. LFD is unlike long-term depression found in mammalian synapses because the latter requires in most instances post-synaptic calcineurin activity.Thus, this simpler invertebrate synapse discloses a novel pre-synaptic depression mechanism.

Abbreviations used



excitatory post-synaptic potential




bisindolylmaleimide I


low-frequency depression


long-term depression


long-term facilitation


matrix-assisted laser desorption/ionization time-of-flight mass spectrometry


miniature EPSPs


neuromuscular junction


protein kinase A


protein kinase C


phorbol 12-myristate13-acetate


protein phosphatase 1


protein phosphatase 2A


protein phosphatase 2B


adenosine 3′,5′-cyclic monophosphorothioate Rp-isomer triethylammonium salt


sodium dodecyl sulphate

A fundamental property of synapses is their ability to modify the efficiency of synaptic transmission through various forms of synaptic plasticity such as potentiation or depression. These changes may be involved in learning, memory and development. Long-term depression (LTD) of synaptic transmission occurs at certain mammalian synapses stimulated at 0.5–5 Hz, can last tens of minutes to hours, and involves several mechanisms depending on the particular cells involved and their age (Malenka and Bear 2004). Although depression encompasses a variety of induction, expression and maintenance mechanisms that can be pre-synaptic, post-synaptic or a combination of both (Anwyl 2006), a common element is regulation by phosphorylation. The variability of LTD mechanisms is achieved through the kinases and phosphatases and their substrates, especially receptors and nerve terminal proteins involved in neurotransmission. Activation or inhibition of a variety of kinases and phosphatases can induce depression. For example, activation of protein kinase C (PKC) and PKG acting on metabotropic glutamate receptors induces depression in Purkinje cells (Crepel and Krupa 1988; Hartell 1994a,b). In contrast, in adult hippocampus in vivo, LTD involves a decrease in PKC activity that is mediated in part by dephosphorylation of the catalytic domain of PKC by protein phosphatases activated after LTD-inducing stimulation (Thiels et al. 2000). Activation of protein phosphatase 1 (PP1) and 2A activity can promote LTD in visual cortex (Kirkwood and Bear 1994) and pyramidal cells (Ajima and Ito 1995). Interestingly, both activation and inhibition of protein phosphatase 2B (PP2B, calcineurin) can produce LTD. In hippocampus (Mulkey et al. 1993; Lin et al. 2008), visual cortex (Kirkwood and Bear 1994) and amygdala (Lin et al. 2003) calcineurin activation promotes LTD. The inverse scenario is found in cerebellum where PP2B inhibition induces LTD (Belmeguenai and Hansel 2005). Inhibitors of PP1 block AMPAR and NMDAR hippocampal LTD but inhibition of PP2B blocks only AMPAR LTD (Morishita et al. 2005).

Post-synaptic cerebellar LTD mediated by AMPA receptors involves PP1, PP2B and receptor endocytosis. Post-synaptic LTD mediated by NMDA receptors requires PP1 but not PP2B and involves actin depolymerization (Morishita et al. 2005). Calcium-dependent actin depolymerization is involved in NMDAR LTD but not AMPAR LTD (Morishita et al. 2005). Actin may modulate NMDAR channel activity (Rosenmund and Westbrook 1993). Microtubules (MT) have also been implicated in NMDAR delivery in cortical pyramidal neurons (Yuen et al. 2005).

Pre-synaptic LTD of transmitter release involves metabotropic glutamate receptors (mGluR) and cannabinoid receptors (Rammes et al. 2003; Bellone et al. 2008; Lovinger 2008; Qiu and Knopfel 2009). Pre-synaptic LTD based on mGluR is age-dependent and, unlike post-synaptic LTD, may be inhibited by activity of PP2B (Li et al. 2002). However, pre-synaptic interneuron LTD requires activity of PP2B (Heifets et al. 2008).

The function of phosphatases as well as their substrates in pre-synaptic LTD is relatively less understood than in post-synaptic LTD. Therefore, it is worthwhile to search for new pre-synaptic functions of phosphatases in synaptic plasticity. We chose to study the mechanism of synaptic depression in crayfish neuromuscular junction (NMJ) synapses because this system allows simultaneous recordings from pre- and post-synaptic sites and permits direct intracellular injection of large molecules into large pre-synaptic axons (Hua and Charlton 1999). Moreover, at crayfish NMJs, different types of synapses on the same post-synaptic muscle cell exhibit extreme kinds of plasticity (Atwood and Karunanithi 2002). For instance, phasic high probability synapses in crayfish NMJs exhibit low-frequency (0.2 Hz) depression (LFD) that lasts for over 1 h (Silverman-Gavrila et al. 2005). Experiments with permeant inhibitors showed that LFD at crayfish NMJs is regulated by PP1, PP2A and PP2B and recovery requires kinase activity (Silverman-Gavrila et al. 2005). Since during LFD there is no change in post-synaptic sensitivity to transmitter quanta, we inferred that the activity of these kinases and phosphatases to regulate LFD is pre-synaptic.

This report shows that pre-synaptic activity of calcineurin is required for LFD. Therefore, it is likely that the phosphorylation status of some proteins changes during LFD. The identification of proteins whose phosphorylation changes during synaptic plasticity may help to unravel this mechanism and identify key proteins that determine synaptic strength. In the present study, we exploited the large pre-synaptic axons at crayfish NMJs as sources of protein to examine differential phosphorylation during LFD. A phosphoproteomic screen and immunocytochemistry identified pre-synaptic cytoskeletal proteins that are dephosphorylated during LFD.

Materials and methods


Intracellular recordings to assess evoked neurotransmitter release

To determine changes in evoked neurotransmitter release we measured the peak amplitude of the excitatory post-synaptic potentials (EPSPs) recorded intracellularly with sharp microelectrodes (3–15 MΩ, backfilled with 3 M KCl) from single muscle fibers of the crayfish (Procambarus clarkii) leg extensor muscle as described previously (Silverman-Gavrila et al. 2005). The extensor muscle is innervated by a single ‘tonic’ axon and a single ‘phasic’ axon. The excitation current required to excite the phasic axon is higher than that for the tonic axon and therefore both are stimulated when the phasic axon is stimulated. However, tonic synapses usually release few, if any, quanta at frequencies used here (Bradacs et al. 1997).

Data collection and analysis

Intracellular potentials detected by electrometer amplifiers (Warner Instruments, Model IE-201, Hamden, CT, USA) were low-pass filtered at 5 kHz, further amplified 10X (Warner Instruments, Model LPF202) and digitized by a PowerLab/4sp data acquisition system (AD Instruments, Round Rock, TX, USA) using Scope for Windows v.3.6.10 program (AD Instruments, Pty Ltd, Castle Hill, NSW, Australia).

Data are presented as mean and SE. Curve fitting to the time course of LFD was performed by SigmaPlot (Systat Software, Richmond, CA, USA) using non-linear regression. Data were transferred to SigmaPlot for producing graphs and SigmaStat 3.0 (SPSS, Chicago, IL, USA) was used for statistical analysis. Statistical significance was evaluated using the paired t-test.

Action potential recordings

Intracellular recordings of the action potential were made from secondary branches of the pre-synaptic phasic axon.

Pressure injection

Pulses of 20–60 psi, 100 ms were applied (Picospitzer II, General Valve Corporation, Fairfield, NJ, USA) to the lumen of a sharp microelectrode (resistance 20–30 mΩ) containing an equal volume mixture of 100 μM calcineurin autoinhibitory peptide (in water) and Texas Red dextran neutral 3 kDa (Invitrogen/Molecular Probes, Eugene, OR, USA) in 1 M KCl. The solution was sonicated prior to loading to reduce clogging. Calcineurin autoinhibitory peptide is a specific calcineurin inhibitor (2.93 kDa, Calbiochem, La Jolla, CA, USA) that corresponds to the C-terminal domain residues of the calmodulin-binding domain of calcineurin (H-DRIM-OH aa 457–482, ITSFEEAKGLDRINERMPPRRDAMP) (Perrino et al. 1995). The Texas Red dextran allowed visualization of successful injection and, since its molecular weight is similar to that of the peptide, we assumed that its diffusion was also similar.

Muscle fibers were filled with the fluorescent marker by diffusion within 5–10 min of injection (Fig. 1a). Axonal injection required 10–15 min and after diffusion for 5–10 min fluorescence could be detected throughout the axonal arbor to boutons. Experiments were conducted with a Nikon Optiphot upright epifluorescence microscope with an Hg lamp. Fluorescent (dichroic mirror 580 nm, excitation filter 510–560 nm, barrier filter 610 nm) and bright field images were obtained using a color CCD camera (CV-S3200, JAI Corporation, Japan) connected to an ATI All-In-Wonder (ATI, Toronto, Canada) frame grabber. Muscle cells under filled terminals were penetrated with a second microelectrode.

Figure 1.

 Post-synaptic calcineurin does not cause LFD. (a) Fluorescence micrograph of a muscle fiber injected with impermeant calcineurin autoinhibitory peptide and fluorescent tracer Texas Red dextran (3 kDa). Bar = 25 μm. (b) Bright field image of same fiber. (c). The decline in transmitter release elicited by stimulation at 0.2 Hz was not abolished after injection of the calcineurin autoinhibitory peptide into the muscle fiber from which EPSPs were recorded [open circles, τ1 = 3.6 min, τ2 = 130 min (n = 4)] compared with control conditions [black circles, τ1 = 4 min, τ2 = 105 min (n = 6)]. Thus, inhibition of post-synaptic calcineurin did not prevent LFD, but only slightly inhibited the depression. Measurements of EPSP amplitude were normalized to the first EPSP. Each point is the average of six consecutive EPSPs in four experiments (mean ± SE) for treatment and five experiments (mean ± SE) for control experiments. Biexponential decay curves are fitted to the treatment data (solid black line) and controls (white line). (d) Calcineurin autoinhibitory peptide pressureinjected into the muscle cell 10 min before stimulation at 0.2 Hz did not prevent development of LFD. However, subsequent application of FK-506 allowed recovery from depression. Each point is the average of six consecutive EPSPs in four experiments (mean ± SE).

Spontaneous miniature EPSPs

To determine if drugs act on pre-synaptic or post-synaptic sites we analyzed 130 mEPSPs caused by spontaneous release of individual quanta before and after treatment with 10 μM cytochalasin, 1 μM jasplakinolide, 10 μM taxol (paclitaxel), 10 μM nocodazole, 50 μM calcineurin autoinhibitory peptide, and 40 μM FK-506 in 4–6 experiments. mEPSPs in the muscle fiber were detected by a low-resistance (2–7 MΩ, 3 M KCl) microelectrode and recorded using Chart for Windows v. 4.2 program (AD Instruments, Castle Hill, NSW, Australia). In crayfish NMJs, the mini frequency is very low (5–10 minis per minute), therefore, we usually recorded for 30 min before treatment and 30 min after treatment. The first 130 mEPSP events well above the noise level before treatment and 130 mEPSPs after treatment were counted manually and the amplitudes and frequencies were determined automatically using Mini Analysis Program v. 6.01 (Synaptosoft, Decatur, GA, USA). The non-parametric Kolmogorov–Smirnov test was used to detect differences in distribution of mEPSP amplitudes. Distributions were considered different using a critical probability level of p < 0.05. The data for each experiment were normalized relative to baseline. All values are means ± SE. Student’s paired t-tests were used to determine difference in the amplitude or frequency before and after treatments. Numbers of experiments are indicated by n. Probability values (p) < 0.05 were considered to represent significant differences.

Drug application

Stock solutions (1–10 mM) of Bisindolylmaleimide I (GF 109203X); FK-506 (tacrolimus); cAMPS- Rp-isomer TEA-salt, adenosine 3′,5′-cyclic monophosphorothioate Rp-isomer, triethylammonium salt (Rp-cAMPS); (Calbiochem); phorbol 12-myristate13-acetate (PMA), staurosporine (Sigma-Aldrich, Oakville, ON, USA); permeant and impermeant calcineurin autoinhibitory peptide (Calbiochem); nocodazole (Sigma-Aldrich); cytochalasin D (Sigma-Aldrich); taxol-Paclitaxel (Fluka); jasplakinolide (Invitrogen/Molecular Probes, Eugene Oregon, USA); latrunculin B (Sigma) were prepared in dimethylsulfoxide (DMSO) or H2O and were dissolved before use in 2 mL saline and mixed thoroughly to a final concentration < 0.1–0.2% DMSO (v/v) in the static bath. Control experiments indicate that this concentration of DMSO does not alter LFD (data not shown). Drug concentrations are indicated in the text associated with each experiment and were chosen empirically based on dose-effect trials and on published concentrations used in crayfish for Rp-cAMPS, staurosporine, GF, PMA, FK-506 (see Silverman-Gavrila et al. 2005 and citations therein), cytochalasin D (Beaumont et al. 2002), jasplakinolide (Zhong and Zucker 2004), latrunculin (Zhong and Zucker 2004), taxol (Weaver and Viancour 1992); or in other invertebrates such as Drosophila (nocodazole and cytochalasin, Kuromi and Kidokoro 1998), and Lymnaea (taxol, Boer et al. 1995).

Cell permeant calcineurin auto-inhibitory peptide (11RCaN-AID, AcRRRRRRRRRRRGGGRMAPPRRDAMPSDA-NH2) (Calbiochem) is composed of calcineurin autoinhibitory domain AID fused to a poly-arginine protein transduction domain [11R (polyarginine)] and acts as an inhibitor of phosphatase activity (Terada et al. 2003).


To observe tubulin and actin distribution in pre-synaptic axon terminals and their modification during LFD or drug treatment we double stained for tubulin and actin, tubulin and synaptotagmin (pre-synaptic vesicle marker), actin or phosphoactin and syntaxin (pre-synaptic compartment marker). Dissected preparations of crayfish leg extensor muscles from the same animal were either stimulated at 0.2 Hz to undergo LFD or kept in saline (control), or cytoskeleton-affecting drugs for 1 h, then processed for imunostaining as described previously (Silverman-Gavrila et al. 2005). As primary antibodies we used monoclonal anti α tubulin antibody clone DM1A, purified mouse immunoglobulin (Sigma, Montreal (T6199)) diluted 1 : 1000; rabbit polyclonal against synaptotagmin I (DSYT2 epitope C- terminal aa 134-474 (Bellen laboratories, Howard Hughes Medical Institute, Baylor College of Medicine, Houston, TX, USA, highly cross-absorbed) diluted 1 : 500; mouse syntaxin-1 6D2 clone epitope: aa 2-190 (1 : 500, Dr Takahashi, University of Kitasato, Japan), phosphoactin [1 : 100, raised in rabbit against a highly conserved actin phospho-peptide CGYSFTTTPAEREIVR (T’jampens et al. 1999) encompassing Thr203 of Physarum actin (Vandekerckhove and Weber 1978)]; rabbit polyclonal anti-actin (Abcam, 1 : 100) in saline containing 0.1% Triton X-100 and as secondary antibodies Alexa Fluor 594 goat anti-rabbit IgG (H + L) and Alexa Fluor 488 goat anti-mouse IgG (H + L) highly cross-adsorbed (Invitrogen/Molecular Probes) diluted 1 : 1000 in saline plus 0.1% Triton X-100. Control preparations without the primary antibodies showed no staining with secondary antibodies (data not shown). Slides were imaged with a Leica DNLFS confocal laser scanning microscope running a TCS LS software (Leica Microsystems, Heidelberg, Germany) using 488 and 594 nm excitation lines and either a 63X NA 1.32 or 40X NA 1.25 objective lens. Images from different focal planes were stacked as layers and combined in projected images.

We examined five crayfish extensor muscle preparations stained for actin and tubulin, six for tubulin and synaptotagmin and two for each experimental condition using similar settings on the confocal microscope. The endings of the phasic axon often extend well beyond the tonic endings (Bradacs et al. 1997) and we examined at least 20 such areas where only phasic terminals were present for each preparation; the pattern predominant in at least 80% of the analyzed terminals was exemplified.

Western blotting

Antibodies were tested for specificity by western blot analysis of proteins extracted from crayfish ventral nerve cords, pre-synaptic axonal processes and neuromuscular junction as described previously (Silverman-Gavrila et al. 2005). Protein extracts were separated by SDS-PAGE (sodium dodecyl sulphate-polyacrylamide gel electrophoresis). Blots were probed with the primary antibodies: anti-actin (1 : 2000); anti-phosphoactin (1 : 100), anti-α tubulin (Sigma, 1 : 1000), synaptotagmin I (1 : 500), syntaxin (1 : 1000), and then with ECL goat anti-rabbit IgG secondary antibody conjugated to horseradish peroxidase (1 : 2000–5000) or sheep anti-mouse IgG secondary antibody, peroxidase-linked species-specific-whole antibody NA 931(1 : 2000–5000) (Amersham Pharmacia Biotech). As molecular weight markers we used pre-stained protein ladder (Fermentas) or Kaleidoscope Prestained Standards (BioRad, Richmond, CA, USA).


Protein sample preparation

After stimulation to induce LFD at phasic synapses or treatment for 20 min with regulators of protein phosphorylation (FK-506, Rp-cAMPS, PMA, GF, staurosporine, in saline) at concentrations used previously for pharmacological studies (Silverman-Gavrila et al. 2005) the preparations were treated with collagenase (type I, 1 mg/mL in saline, Calbiochem) for 30 min at 22°C (Kanai et al. 1998). Collagenase treatment detaches pre-synaptic terminals from muscle and leaves post-synaptic receptors intact (Robitaille et al. 1990; Wilkinson and Lunin 1994). Therefore, we assume that we do not have much post-synaptic contamination. While it was not possible to eliminate axonal protein from the samples, crayfish axons stimulated with low frequency admit Ca2+ (Saito and Ueno 1978) and might therefore have some reactions that are similar to pre-synaptic boutons. Pre-synaptic structures (phasic and tonic axons and boutons) were detached with very fine insect pins (minuten pins 0.1 mm, Fine Science Tools, CA, USA) from the extensor muscle, frozen in liquid nitrogen, thawed and homogenized twice in a lysis buffer (50 mM Tris–HCl, 2% (w/v) sodium dodecylsulphate, 10% (v/v) glycerol, 20 mM β glycerophosphate, 5 mM EDTA, 5 mM sodium fluoride, 5 mM orthovanadate, 10 mM dithiothreitol (DTT) and protease inhibitor cocktail (Roche Diagnostics, Laval, QC, Canada). Lysate was then boiled for 10 min, centrifuged to remove debris and the supernatant collected. Protein extracts were separated by SDS-PAGE on 15-well 15% gels (Bio-Rad Laboratories Hercules, CA, USA) as described previously (Silverman-Gavrila et al. 2005). Most of the boutons are formed by tertiary axon branches on the muscle surface and are likely to be included when the axon is stripped of the muscle.

Protein visualization

Proteins separated by SDS-PAGE were visualized by staining with Pro-Q Diamond (Invitrogen/Molecular Probes), a fluorescent dye specific for phosphoaminoacids. The gel was fixed in 50% methanol and 10% acetic acid for 60 min, washed in dH2O for 3 × 10 min and stained with Pro-Q Diamond for 90 min (as per manufacturer protocol) and destained for 3 × 30 min in destaining solution containing 20% acetonitrile and 50 mM sodium acetate, pH 4. As a molecular weight marker and as positive and negative control for detection of phosphorylated proteins we used PeppermintStick Phosphoprotein molecular weight standards (Invitrogen/Molecular Probes) that contain a mixture of phosphorylated and nonphosphorylated proteins. Separation by SDS-PAGE resolved the marker mixture into two phosphorylated and four non-phosphorylated bands. Stained gels were visualized with a TyphoonTM 9400 Variable Mode imager (GE Healthcare/Amersham Biosciences) with excitation at 532 nm and emission collected at 580 nm (±30 nm) using ImageQuant software (GE Healthcare/Amersham Biosciences). To optimize images of Pro-Q Diamond staining, the gray-scale was adjusted so that the phosphorylated proteins from the standard lane showed up as two dark bands and non-phosphorylated proteins were not detected. This ensured that all other bands imaged from the gel corresponded to the phosphorylated proteins.

The same gel was then washed twice in dH2O for 5 min per wash and re-stained overnight with a solution of SYPRO Ruby dye for total proteins. Then the gel was washed two times for 30 min in 10% methanol and 7% acetic acid, rinsed with dH2O and scanned at 473 nm and the emission collected at 610 nm (±30 nm). Images were optimized by focusing on the lane containing the standards and adjusting the gray-scale so that both phosphorylated and non-phosphorylated proteins showed up as six dark bands corresponding to total proteins.

To detect changes in the ratio of phosphorylated to total protein we measured the fluorescence from Pro-Q Diamond bound to phosphoprotein and the fluorescence from SYPRO Ruby bound to all the proteins and calculated the ratio of these measurements. By performing a ratiometric analysis of each band, it is possible to distinguish a large amount of minimally phosphorylated protein from a small amount of heavily phosphorylated protein. We used the program ImageJ ( to measure the intensity of the Pro-Q Diamond signal (P) and the SYPRO Ruby signal (T) from a rectangular area of interest along each lane of the gel. This provided both pixel intensity and position information along the length of a lane. The maximum values of the peaks corresponding to each band from which we subtracted the background intensity values gave the values used to calculate the phosphorylated/total ratios (P/T) for the standard proteins as well as for proteins of interest. It is critical to note that since each gel contained experimental and control protein lanes, all procedures and measurements were applied equally to experimental and control proteins. We cannot compare the intensity in gel A versus gel B as gel A is stained for phosphoproteins and gel B for total proteins. Comparisons cannot be made among lanes of the phosphostained gel (A) because the lanes reflect treatment with various phosphorylation regulators; moreover, equal loading was not maintained for each lane, which also prevents comparison among lanes in gel B stained for total proteins. Instead, the “actual” phosphorylation level is expressed as a ratio of phophorylation/total protein, obtained by ratioing the band intensity in gel A phosphostained with Pro-Q staining that reflects phosphoprotein levels to the intensity of the same band in gel B stained with SYPRO Ruby dye that quantitatively stains total proteins. The technique corrects for different loading between lanes and makes equal loading unnecessary. The lanes in A that appear qualitatively darker relative to their corresponding total protein lanes in B generally represent experiments in which phosphorylation was increased by drugs that increase kinase activity (lane 8) or that decrease phosphatase activity (lane 5).

Next, the gel was stained with Coomassie Brilliant Blue R-250 (BioRad Laboratories) to visualize the bands with a visible permanent dye. Bands of interest were then excised from the gel, digested and proteins identified by mass spectrometry.

2-D gel electrophoresis

Proteins were extracted from pre-synaptic axonal processes by freezing twice in liquid nitrogen and homogenizing. The sample was lysed with a lysis buffer [30 mM Tris pH 8.8, 7 M urea, 2 M thiourea, 3% CHAPS and 10 mM DTT and protease inhibitor mixture (Roche Diagnostics), added just before lysis]. The lysate was centrifuged for 10 min at 9660 g and the supernatant collected. Protein concentration was measured with a 2-D Quant kit (GE Healthcare/Amersham Bioscience). Next the sample was treated with DeStreak Rehydration Solution (GE Healthcare/Amersham Biosciences) to prevent streaking and buffered with IPG buffer solution pH 3–10 NL (nonlinear gradient) (GE Healthcare/Amersham Biosciences). Proteins were separated overnight in the first dimension by isoelectric focusing on a 13 cm, broad-range pH 3–10 IPG Immobiline DryStrip gel (GE Healthcare/Amersham Biosciences) in a Ettan IPGphor II apparatus (GE Healthcare/Amersham Biosciences). Then the strip was rehydrated and equilibrated in a buffer containing SDS, urea, TRIS pH 8.8, bromphenol blue, glycerol, DTT and iodoacetamide. The second dimension electrophoresis was performed using a lab cast 20 cm 10% SDS-PAGE gel. The gel was stained with Page Blue protein staining solution (Fermentas) and proteins visualized with a Kodak Image station 2000R. Spots of interest were excised from the gel, digested and proteins identified by mass spectrometry.

Mass spectrometry

In-gel trypsin digestion of proteins was performed as per manufacturer’s protocol using a Pierce In–gel Tryptic Digestion Kit (ThermoFisher Scientific, Rockford, IL, USA). Two samples were analyzed in duplicate at the Proteomic and Mass Spectrometry Centre (University of Toronto). The peptide mixture was separated by a Voyager–DESTR matrix-assisted laser desorption/ionization time-of-flight mass spectrometer (MALDI-TOF MS) equipped with a 337 nm nitrogen laser (Applied Biosystems). Peptide-mass fingerprinting chromatograms of MS/MS data showing spectra peak lists of m/z values sets were analyzed against all available proteins from database using MASCOT Search program (Matrix Science Ltd, London). Some search parameters included: enzyme: trypsin; fixed modification: carbamidomethyl; variable modification: oxidation; mass value: MH+; monoisotopic, protein mass: unrestricted. Peptide mass tolerance was set to ± 100–200 ppm (fraction expressed as parts per million); peptide charge state: 1+, maximum missed cleavages allowed: 1. Mass fragment spectra were compared with MSDB (comprehensive, non-identical protein database), NCBI (non-redundant) and Swiss Prot database merged into the UniProt database, The Universal Protein Resource (


Pre-synaptic calcineurin activity in LFD

LFD at phasic synapses involves phosphorylation-related changes in the probability of transmitter release regulated by protein kinase A (PKA), PKC, PP1 and 2A and calcineurin (Silverman-Gavrila et al. 2005). When calcineurin activity was inhibited by either the membrane-permeant calcineurin inhibitor FK-506 or permeant calcineurin auto-inhibitory peptide, both phases of LFD were eliminated (Silverman-Gavrila et al. 2005). When tested at a non-depressing frequency (0.0016 Hz), FK-506 and permeant calcineurin auto-inhibitory peptide caused a small increase in transmitter release that stabilized within 10–20 min showing that time-dependent effects are not a major factor in blockade of LFD. Subsequent stimulation at 0.2 Hz, which normally causes LFD, now caused facilitation (Silverman-Gavrila et al. 2005). Effects of the calcineurin blockers to increase transmitter release were at maximum before depression was tested; when depression stimuli were given, there was facilitation instead of depression, suggesting that when calcineurin is inactive, LFD is not expressed.

Since these permeant drugs could have acted on both pre-synaptic and post-synaptic cells we performed additional experiments to determine whether the site of calcineurin activity in LFD is pre- or post-synaptic. We pressure-injected membrane impermeant calcineurin auto-inhibitory peptide into post-synaptic muscle cells or pre-synaptic axons and assessed its effects on neurotransmitter release. We measured the amplitude of the intracellularly recorded excitatory post-synaptic potential (EPSP) evoked by stimulating the phasic axons at 0.2 Hz.

The membrane impermeant calcineurin autoinhibitory peptide (50 μM) and Texas Red-dextran (3 kDa) were pressure-injected into a muscle fiber (Fig. 1a and b) 10 min before starting stimulation at 0.2 Hz, but depression still occurred (τ1 = 3.6 min, τ2 = 130 min; n = 4) compared with control (τ1 = 4 min τ2 = 105 min; n = 6) (Fig1c). Moreover, LFD obtained after the autoinhibitory peptide was injected into the muscle cell was reversed with subsequent application of membrane-permeant calcineurin inhibitor FK-506 (40 μM) to the bath (n = 4) (Fig. 1d). This suggests that most of LFD is caused by the activity of pre-synaptic calcineurin and not by post-synaptic calcineurin because inhibition of the latter caused only a slight decrease of LFD.

To determine whether these drugs affect the kinetics of transmitter release from individual vesicles or the sensitivity of post-synaptic receptors we recorded mEPSPs which are the responses to spontaneously released single quanta of transmitter. We measured the amplitude and frequency for mEPSPs obtained before (usually 130) and after drug treatment (usually 130) for each of the five experiments. FK-506 did not affect the average mEPSP amplitude (0.13 ± 0.01 mV before vs. 0.13 ± 0.01 mV after, p = 0.78, n = 5) or the amplitude distribution (Kolmogorov–Smirnov test, p = 0.69 for the example experiment (Fig. 2a). The frequency (0.12 ± 0.03 Hz before vs. 0.15 ± 0.01 Hz after, p = 0.1, n = 5), rise to peak time (1.8 ± 0.17 ms vs. 1.7 ±0.25 ms, p = 0.7, n = 5) and decay to 50% time (2.2 ±0.07 ms vs. 1.9 ± 0.56 ms, p = 0.7, n = 5) of mEPSPs did not change with FK-506 (Fig. 2a inset). This is similar to the action of FK-506 on NMJs of the crayfish opener muscle where the amplitude of EPSPs was increased but there was no effect on the frequency of spontaneous transmitter release (Swain and Charlton, unpublished results). This probably means that spontaneous transmitter release and evoked transmitter release do not utilize exactly the same set of phosphoproteins dephosphorylated by calcineurin.

Figure 2.

 FK-506 and permeant calcineurin autoinhibitory peptide have no obvious post-synaptic effects. (a) Cumulative amplitude distribution of mEPSPs was not affected by FK-506 (Kolmogorov–Smirnov test, p = 0.69 for example experiment; n = 5). Inset. Averages of 102 mEPSPs recorded before and after treatment with FK-506 show little effect on rise to peak time (1.31 ± 0.05 ms vs. 1.32 ± 0.05 ms) and decay to 50% time (0.19 ± 0.07 ms vs. 0.18 ± 0.06 ms) for this experiment. (b) Cumulative amplitude distribution of mEPSPs was not affected by permeant calcineurin autoinhibitory peptide (Kolmogorov–Smirnov test p = 0.26 for example experiment; n = 5). Inset. Averages of 130 mEPSPs recorded before and after treatment with permeant calcineurin autoinhibitory peptide show little effect on mEPSP rise to peak time (1.67 ± 0.04 ms vs. 1.68 ± 0.04 ms) and decay to 50% time (1.96 ± 0.12 ms vs. 1.71 ± 0.11 ms for this experiment).

The membrane permeant calcineurin autoinhibitory peptide previously shown to block LFD (Silverman-Gavrila et al. 2005) did not affect the amplitude of spontaneous mEPSPs (0.16 ± 0.01 mV before versus 0.16 ± 0.01 mV after, p = 0.81, n = 6). The Kolmogorov–Smirnov test (p = 0.26 for example experiment) did not detect statistically significant changes between the amplitude distributions of mEPSPs before and after treatment with permeant calcineurin autoinhibitory peptide (Fig. 2b). Averages of 130 mEPSPs recorded before and after the treatment show that the mEPSP rise to peak time (1.76 ± 0.07 ms vs. 1.67 ±0.07 ms, p = 0.43, n = 6), and the decay to 50% time (2.87 ± 0.45 ms vs. 2.86 ± 0.60 ms, p = 0.98, n = 6) did not change (Fig. 2b inset). The frequency of spontaneous mEPSPs was also not significantly affected following application of this permeant peptide (0.15 ± 0.04 Hz before and 0.19 ± 0.07 Hz after, p = 0.65, n = 6). Therefore, it is unlikely that these drugs affect post-synaptic receptors or kinetics of individual release events.

When the impermeant autoinhibitory peptide was pressure-injected into the pre-synaptic axon prior to stimulation at 0.2 Hz (Fig. 3a) both phases of LFD were blocked (Fig. 3b and c) and facilitation occurred. In control preparations with pre-synaptic terminals injected with similar quantities of Texas Red or muscle cells injected with fluorescein-marker solution alone, LFD was no different from that recorded from uninjected preparations (data not shown). This indicates that activity of pre-synaptic calcineurin on unknown phosphosubstrates is required for LFD.

Figure 3.

 Pre-synaptic injection of impermeant calcineurin autoinhibitory peptide abolishes both phases of LFD. (a) Injection puffs of peptide with fluorescent dye marker Texas Red dextran (3 kDa) in main phasic axons (a) and diffusion in main (i–iii), secondary (ii, iii, v), tertiary and quaternary branches (iv, v) and phasic boutons (iv). In all figures bars = 20 μm. (b) Representative simultaneous recordings of pre-synaptic APs from axons injected with impermeant calcineurin autoinihibitory peptide and EPSPs from muscle fiber at 0 min (i) and 23 min (ii) of stimulation at 0.2 Hz show an increase in the EPSP amplitude with no effect on the action potential. (c) Pressure-injection of the calcineurin autoinhibitory peptide into phasic axons prior to stimulation prevented the usual decline in amplitude of EPSPs elicited at 0.2 Hz stimulation. Instead, there was a gradual increase of transmitter release. Measurements of EPSP amplitude are normalized to the first EPSP. Each point is the average of six consecutive EPSPs in four experiments (mean ± SE).

Dephosphorylation of actin and tubulin during LFD

In order to identify possible phosphoproteins involved in LFD, we employed a combined proteomic and electrophysiological approach by comparing phosphoprotein levels before (unstimulated controls) and after production of LFD. Proteins from pre-synaptic axonal processes extracted before and after LFD or treatments with various regulators of protein phosphorylation (FK-506, Rp-cAMPS, PMA, GF, staurosporine) were separated by SDS-PAGE. Phosphoproteins were detected by gel staining with Pro-Q Diamond and subsequently total proteins were stained with SYPRO Ruby (Fig. 4). Coomassie Brilliant Blue R-250 was used to visualize the bands in order to excise them from the gel. Next, we analyzed 10 bands whose phosphorylation level was altered and whose protein concentration was determined by visual inspection to be sufficient for further mass spectrometry analysis. The bands of interest were excised from the gel and proteins were trypsin digested and analyzed by MALDI-TOF MS. Three statistically significant Mascot protein scores were obtained leading to the MS identification of three proteins whose phosphorylation state changed on the basis of probability-based Mowse Score, protein profiles peptide mass matching and protein ID data base searching (Swiss Prot and MSDB). Relatively few crayfish protein sequences are known and therefore it is necessary to compare the MS results with those from other organisms. Using criteria given in the Materials and methods the phosphoproteins were identified as actin 1 from the shrimp (Penaeus monodon) 42 kDa, tubulin α chain neuron-specific isoform from marbled electric ray (Torpedo marmorata) 50.82 kDa and cGMP-phosphodiesterase δ subunit from Ciona intestinalis 17.6 kDa.

Figure 4.

 LFD stimulation and pharmacological manipulation alter protein phosphorylation in pre-synaptic processes. These changes are detected by analysis of the ratio between phosphorylated (a) versus total proteins (b). For each band this ratio is then compared between treatments. Five preparations were stimulated at 0.2 Hz for 60 min to elicit LFD or were treated for 20 min with regulators of protein phosphorylation FK-506 (40 μM) to block calcineurin, Rp-cAMPS (300 μM) to block PKA, PMA (100 μM) to stimulate PKC, GF (10 μM) to block PKC, staurosporine (10 μM) to block several kinases. Control preparations were not stimulated and were untreated. (a) Proteins isolated from crayfish pre-synaptic axonal processes were separated electrophoretically (PAGE) and stained with Pro-Q Diamond for phosphorylated proteins. (b) After washing out Pro-Q Diamond, the same gel was stained with SYPRO Ruby for total proteins. The fluorescence intensities of Pro-Q Diamond (P) and SYPRO Ruby (T) were recorded with a gel imager. Intensities were measured in ImageJ by placing a rectangular area of interest (black vertical bands) along the lanes and the resulting data were plotted versus distance (MW axis). (c) Shows fluorescence of Pro-Q Diamond, SYPRO Ruby and ratio Pro-Q/SYPRO Ruby versus distance for the unstimulated controls and (d) after LFD. The maximum values of the peaks corresponding to each band from which the background intensity values were subtracted gave the values used to calculate the P/T (phosphorylated/total) ratios for the standard proteins as well as for proteins of interest presented in Table e. White boxes (T1,2,3; P1,2,3) indicate bands that were excised from gel, analyzed and identified by mass spectrometry as tubulin (ratio P1/T1), actin (ratio P2/T2) and cGMP-phosphodiesterase (ratio P3/T3). (e) Effect of drugs and LFD on phosphorylation of actin and tubulin. Ratiometric measurements of phosphorylated/total protein intensities for bands from which tubulin (ratio P1/T1) and actin (ratio P2/T2) were identified by mass spectrometry in various experimental conditions.

Quantitative ratiometric analysis of phosphorylated versus total proteins showed that phosphorylation of actin and tubulin was decreased during LFD (Fig. 4). We wondered whether drugs that alter activity of kinases and phosphatases would also affect the phosphorylation state of actin and tubulin in unstimulated cells. We found that phosphotubulin levels were decreased by the kinase inhibitors GF, staurosporine and, Rp-cAMPS and by the PKC agonist PMA, but were increased by the phosphatase inhibitor FK-506. Phosphoactin levels were decreased by GF, staurosporine and Rp-cAMPS, but were increased by PMA and FK-506 (Fig. 4). The experiment was duplicated with similar results. Thus when calcineurin was inhibited, phosphorylation of both proteins increased, but when kinase activity was inhibited, phosphorylation decreased. Other protein bands also showed changes in P/T ratios with experimental manipulation, but the screening technique identified only a few proteins with high certainty.

Actin and tubulin are involved in LFD

To test the hypothesis that actin and tubulin are involved in LFD we applied drugs that affect the cytoskeleton and assessed their effects on LFD and on basal evoked transmitter release at the non-depressing frequency of 0.0016 Hz.

We used two drugs that reduce actin filaments by different mechanisms. Latrunculin sequesters monomeric G-actin, leading to the disassembly of actin filament, while cytochalasin D caps the actin filaments at the barbed ends to decrease actin polymerization. Disruption of actin microfilaments with latrunculin B (3 μM, n = 6) or cytochalasin D (10 μM, n = 5) inhibited LFD. Compared with 62% depression after 60 min in controls, a 30% depression was observed in latrunculin experiments and a 10% depression for cytochalasin treatments (Fig. 5a,b). Latrunculin also blocks the induction of long-term facilitation (LTF) at some crayfish NMJs (Zhong and Zucker 2004). To assay for non-stimulus-dependent drug effects we tested each drug with stimulation at 0.0016 Hz, which does not cause depression. Neither drug had major non-activity-dependent effects on transmitter release at 0.0016 Hz (Fig. 5a,b). The actin polymerization activator, jasplakinolide (1 μM) eliminated the fast phase, but there was only 57% depression after 60 min (n = 5) (Fig. 5c). Since jasplakinolide reduced transmitter release by about 20% at 0.0016 Hz (Fig. 5c′) (cf Zhong and Zucker 2004), it appears that this drug reduced LFD.

Figure 5.

 Cytoskeleton drugs affect LFD. To assess the importance of actin and tubulin, drugs that affect these proteins were applied before and during stimulation to evoke LFD (a–e). To assess nonstimulus-dependent effects of these drugs they were also applied during non-depressing stimulation at 0.0016 Hz (a′–e′). (a) Application of actin depolymerizing drug latrunculin (3 μM, white circles) 10 min before stimulation at 0.2 Hz inhibited LFD. Mean normalized EPSP amplitude in six drug experiments and five control experiments (dark circles, shown for comparison) (a′). When latrunculin was added after stimulation at 0.0016 Hz for 20 min, there was a slight increase in transmitter release. (b) Application of microfilament-disrupting drug cytochalasin (10 μM) 10 min before stimulation at 0.2 Hz inhibited LFD (n = 5) (b′). There was a small increase in transmitter release that reached a maximum within 10 min of cytochalasin treatment (0.0016 Hz) and then declined. (c) The actin polymerization activator jasplakinolide (1 μM) eliminated the fast phase of LFD (n = 5) and caused about 20% inhibition of basal transmission (c′). (d) Nocodazole, a microtubule-depolymerizing agent (10 μM), accelerated LFD (n = 5) and caused about 20% inhibition of basal transmission (d′). (e) Taxol (10 μM), a microtubule stabilizer, inhibited the second phase of depression (n = 5) but caused about 10% reduction in basal transmission (e′). Measurements of EPSP amplitude (a–e) were normalized to the first EPSP. Each point is the average of six consecutive EPSPs in five experiments mean ± SE for each treatment. (a′–e′) Panels show control experiments for non-activity-dependent effects of cytoskeleton regulators. After 20 min of stimulation at 0.0016 Hz, the drugs were added to the bath and transmitter release measured for additional 60–70 min in five experiments for each treatment. Bars indicate the presence of drugs in the bathing solution.

With the microtubule depolymerizer nocodazole (10 μM) LFD appeared greater than control (74% depression after 60 min, n = 5) (Fig. 5d), but since this drug caused about 20% depression at 0.0016 Hz (Fig. 5d′) there was no real effect on LFD. Taxol (10 μM), a microtubule aggregating agent that causes tubulin monomers to polymerize, inhibited the second phase of LFD (23% after 40 min) (n = 5) (Fig. 5e) and had no significant effect on transmitter release at 0.0016 Hz (Fig. 5e′).

To determine whether cytoskeletal drugs had post-synaptic effects we analyzed spontaneous mEPSPs. Data from representative experiments are presented in Fig. 6. For statistical analysis using paired t-test, mean mEPSP frequencies and amplitudes were normalized by dividing the treatment values by the control ones before the treatment. Cytochalasin (0.34 ± 0.26 Hz before vs. 0.99 ± 0.64 Hz after treatment, n = 5, p = 0.03) and taxol (0.11 ± 0.01 Hz before vs. 0.84 ± 0. 6 Hz after treatment, n = 6, p = 0.01) increased mEPSP frequency but jasplakinolide (0.33 ± 0.14 Hz before vs. 0.33 ± 0.16 Hz after treatment) and nocodazole (0.12 ± 0.03 before vs. 0.19 ± 0.05 treatment, n = 5) did not. None of these treatments had statistically significant effects on the mEPSP amplitude [(0.10 ± 0.01 mV before versus 0.11 ± 0.01 mV after jasplakinolide treatment, n = 5, Fig. 6a and b); (0.09 ± 0.01 mV before vs. 0.09 ± 0.01 mV after cytochalasin treatment, n = 5 Fig. 6c); (0.12 ± 0.01 mV versus 0.11 ± 0.02 mV after nocodazole treatment, n = 5, Fig. 6d); and (0.16 ± 0.04 mV vs. 0.13 ± 0.03 mV after taxol treatment, n = 6, Fig. 6e)] suggesting that they act pre-synaptically.

Figure 6.

 Jasplakinolide, cytochalasin, taxol, and nocodazole have no post-synaptic effects. (a) Representative average traces of 130 mEPSPs recorded before and 130 mEPSPs after jasplakinolide treatment shows that the mEPSP rise time and decay constant did not change. (b) Example of a cumulative fraction plot of the distribution of mEPSP amplitudes before and after jasplakinolide treatment. The Kolmogorov-Smirnov test indicates that there was no significant change in mEPSP amplitude distributions with jasplakinolide treatment (p = 0.96, n = 130). (c) Cytochalasin (10 μM) increased the mEPSP frequency without affecting the mEPSP amplitude distribution (Kolmogorov–Smirnov test, p = 0.82 for this experiment). Similar results were obtained in four additional experiments, suggesting that cytochalasin inhibits depression by acting on the pre-synaptic terminals. (d) When 10 μM nocodazole was applied, there was no change in mEPSP frequency (n = 6). The amplitude, time course of spontaneous mEPSPs and mEPSP amplitude distribution were also not affected by the presence of nocodazole (p = 0.41, Kolmogorov–Smirnov test for this individual experiment) suggesting that nocodazole accelerates the depression by acting pre-synaptically. (e) Treatment with taxol resulted in a significant increase in the rate of spontaneous transmitter release (n = 6), without altering mEPSP amplitude distribution (p = 0.82, Kolmogorov–Smirnov test for this individual experiment), suggesting that taxol inhibits depression by acting on the pre-synaptic terminals.

In control experiments, in which saline was exchanged with saline plus 0.2% DMSO there was no change in the amplitude or frequency of spontaneous transmitter release (data not shown). The frequency of mEPSPs was stable for 30 min. These results indicate that the muscle sensitivity to quantal glutamate release was not affected by treatments with anti-cytoskeletal drugs. It is likely that the drugs affect the pre-synaptic terminal and modulate evoked EPSPs by affecting transmitter release. LFD was decreased by tubulin stabilization and by disruption of actin.

Tubulin and actin in pre-synaptic terminals

We used double immunostaining for tubulin [Fig. 7a(ii)] and the vesicular marker synaptotagmin [Fig. 7a(i)] to image tubulin distribution relative to that of synaptic vesicles. In untreated and unstimulated crayfish nerve terminals tubulin immunoreactivity occurs under or beside synaptotagmin immunoreactivity at phasic boutons as seen in the merged image [Fig. 7a(iii)]. After application of taxol (10 μM) tubulin appeared to wrap around the vesicle clusters as loops [Fig. 7b(iii)]. During LFD there was a decrease of microtubule immunoreactivity around boutons [Fig. 7c(iii)], though tubulin staining was present in main axons (data not shown). A similar pattern was observed after treatment with nocodazole (10 μM), a microtubule-depolymerizing agent that caused the disruption of microtubules (Fig. 7d(iii)). Western blot analysis (Fig. 7e) showed that anti α-tubulin antibody recognized a single protein band of 51 kDa in both crayfish nerve cord (N) and pre-synaptic processes (P) and anti-synaptotagmin-1 detects a band of about 86 kDa in crayfish nerve cord (N) and pre-synaptic processes (P). Actin immunoreactivity at phasic boutons was imaged relative to tubulin [Fig. 8a(iii)]. There was also phosphoactin immunoreactivity at boutons that appeared to co-localize with syntaxin [Fig. 8b(iii)]. There was no immunoreactivity in preparations in which only the secondary antibodies were applied (data not shown). Phosphoactin immunoreactivity was reduced after LFD at phasic boutons compared with syntaxin [Fig. 8c(iii)].

Figure 7.

 Tubulin in crayfish phasic nerve terminals. (a) Representative double immunostaining confocal images showing the presence of tubulin (ii, green) pre-synaptic terminals and its distribution relative to synaptic vesicles stained with synaptotagmin (i, red). In control unstimulated and untreated phasic terminals tubulin runs under and beside synaptotagmin staining as seen in the overlay image (iii). (b) Microtubule stabilizer taxol (10 μM) that inhibits LFD caused wrapping around the vesicle clusters as loops (iii). (c) During LFD a flattening of vesicle cluster and disappearance of most of microtubules was observed. (d) A similar pattern was observed after treatment with nocodazole (10 μM), a microtubule-depolymerizing agent that caused the disruption of microtubules. Scale bars = 7.83 μm. (e) Western blot analysis for α tubulin and synaptotagmin I. Anti-α tubulin antibody recognized a single band of about 51 kDa protein and antisynaptotagmin I recognized a band of about 86 kDa in both crayfish nerve cord (N) and pre-synaptic structures (P). Thus, both antibodies had appropriate specificity.

Figure 8.

 Actin in crayfish phasic terminal. (a) Double immunostaining for actin [red, a(ii)] and tubulin [green, a(i)] showed immunoreactivity at phasic boutons [a(iii)]. (b) Double immunostaining for phosphoactin [red b(ii)] and pre-synaptic membrane marker syntaxin [green, b(i)] showed co-localization at control phasic boutons [b(iii)]. (c) Phosphoactin immunoreactivity [red, c(ii)] was reduced after LFD at phasic boutons [merged c(iii)] compared to syntaxin [green, c(i)]. Scale bars = 2.6 μm. (d) Anti-actin antibody detected a protein of approximately 41 kDa in crayfish pre-synaptic axonal processes (P) and nerve cord (N). (e) Anti-phosphoactin antibody recognized a protein band of about 43 kDa in both crayfish nerve cord (N) and pre-synaptic processes (P). (f) Anti-syntaxin antibody identified a band of about 38 kDa in nerve cord (N) and pre-synaptic processes (P). The western blots in d, e, and f demonstrate appropriate specificity of the antibodies. (g) Western blot of phosphoactin using proteins isolated from control unstimulated preparations and after LFD. Phosphoactin level declined by about 50% after LFD compared with tubulin used as an equal loading control protein. (h) 2D gel electrophoresis image of proteins isolated from crayfish pre-synaptic structures. Actin and phosphoactin could be isolated by 2D gel electrophoresis; and were later identified by mass spectrometry. Proteins isolated from crayfish pre-synaptic structures (axons and butons) were separated in the first dimension by isoelectric focusing using a broad range 13 cm, pH 3–10 strip gel followed by a second dimension electrophoresis on a 20 cm 10% SDS gel. The gel was stained with Page Blue protein staining solution (Fermentas) and proteins visualized with a Kodak Image station 2000R. The migration position of the molecular weight markers is indicated at the right in kDa and pI values at the bottom of the panel. Two spots (arrows) were excised from the gel, digested and proteins were identified by mass spectrometry. Mascot Search Matrix programs identified them as actin. The staining pattern may indicate two stages of actin phosphorylation, with the phosphorylated acidic actin towards the +end.

Treatment with cytochalasin disrupted actin cytoskeleton (data not shown). On Western blots the anti-actin antibody detected a band of approximately 41 kDa in crayfish pre-synaptic axonal processes (P) and nerve cord (N) (Fig. 8d); anti-phosphoactin antibody recognized a single protein band of about 43 kDa in both crayfish nerve cord (N) and pre-synaptic processes (P) (Fig. 8e), and anti-syntaxin antibody a band of about 38 kDa in pre-synaptic processes (P) and nerve cord (Fig. 8f). During LFD phosphoactin concentration decreased as shown by quantitative western blotting of phosphoactin versus tubulin used as equal loading protein (Fig. 8g). Western blots using an antibody for total actin showed no difference before and after LFD (not shown). On the 2D gel shown in Fig. 8h, two spots of similar MW but displaced on the pH axis were excised and shown by MS to contain actin. It is likely that these spots represent actin and phosphoactin.

In summary, the phosphorylation status of pre-synaptic proteins such as actin and tubulin is regulated during LFD.


LFD at crayfish phasic synapses occurs with minimal stimulation (0.2 Hz) as a biphasic long-lasting depression (Silverman-Gavrila et al. 2005). LFD is a frequency-dependent phenomenon likely caused by a decrease in transmitter release. These characteristics are shared with other LFDs such as LFD that occurs at the synapses between neonatal rat spinal Ia afferents and motoneurons (Lev-Tov and Pinco 1992), or at rat visual cortex (Akaneya et al. 2003).

In this study, we have begun to decifer the mechanism underlying LFD at crayfish synapses and found that it requires activity of pre-synaptic calcineurin and involves dephosphorylation of pre-synaptic cytoskeletal elements.

Phosphoproteomics of LFD

Only a few proteins whose phosphorylation state changed during LFD could be identified and, of these, we chose to study actin and tubulin because they were previously implicated in pre-synaptic function and could be manipulated by accepted pharmacological techniques. Protein databases contain relatively few Crustacean proteins, but future enrichment of these databases would allow us to identify more proteins involved in LFD.

Actin and tubulin in pre-synaptic terminals

Crayfish terminals contain actin and tubulin (Figs 7 and 8) as do other synapses (reviewed in Cingolani and Goda 2008; Hirokawa et al. 1989; Guo et al. 2005; Gordon-Weeks et al. 1982). Tubulin localization was detected in close proximity to synaptic vesicles labeled with anti-synaptotagmin antibody (Fig. 7a). The pre-synaptic terminals contain both actin and phosphoactin (Fig. 8a–e and g). Since the antibody for phosphoactin was made with a conserved phosphopeptide sequence (see Materials and methods) that is found in crayfish actin (Kang and Naya 1993; accession number P45521) and identifies the correct molecular weight on western blots, it is highly likely that the immunoreactivity to phosphoactin is specific. Phosphorylated actin and tubulin appear in mouse forebrain synaptosomes (Collins et al. 2005) and developing forebrains and midbrains (Ballif et al. 2004), but this is the first demonstration of phospho-actin in intact pre-synaptic terminals. The immunoreactivity of phospho-actin decreased during LFD (Fig. 8c) indicating stimulus-dependent phosphatase activity. Two spots separated on the pH axis of 2D gels were identified by MS as actin (Fig. 8h) and probably correspond to actin and phospho-actin (Gu et al. 2003).

LFD and cytoskeleton dephosphorylation

During LFD phosphorylation of actin and tubulin was reduced by 30–40% as seen by ratio staining (Fig. 4e) and by about 50% for actin as seen by western blotting (Fig. 8g). The stimulus-dependent changes at synapses may have been underestimated because the protein sample was not exclusively from phasic pre-synaptic boutons but included phasic axons and the unstimulated tonic axons and boutons. The reduction in phosphorylation of actin and tubulin caused by kinase inhibitors (Fig. 4e) in the absence of stimulation shows that there is ongoing phosphorylation and dephosphorylation. The increase in actin and tubulin phosphorylation caused by the calcineurin inhibitor alone (Fig. 4e) indicates that calcineurin has a strong effect on phosphorylation of cytoskeletal elements.

Inhibition of calcineurin causes an increase in phosphorylation of actin and tubulin, an increase in transmitter release, and blocks LFD. This suggests that control of actin and tubulin phosphorylation by calcineurin is involved in LFD.

LFD and cytoskeleton polymerization

The consequences of actin phosphorylation on polymerization depend on the sites phosphorylated. For instance, phosphorylation by PKC promotes actin polymerization (Ohta et al. 1987; Reiss et al. 1996), inhibits disassembly and protects F-actin assembly (Banan et al. 2004a), while actin phosphorylated by PKA has weaker polymerizability than the unphosphorylated form suggesting that various protein kinases phosphorylate actin differently and induce opposite effects on its polymerizability (Ohta et al. 1987). Phosphorylation of actin at Thr203, the epitope detected by our antibody, results in loss of its ability to polymerize into actin filaments (Furuhashi 2002).

Tubulin polymerizability is also affected by the site of phosphorylation (Wandosell et al. 1987). α and β-tubulin phosphorylated at the C terminal by Ca2+/CaM-dependent kinase fails to polymerize and this is reversed by phosphatase treatment (Wandosell et al. 1986). Tyrosine phosphorylation of α-tubulin inhibits its ability to polymerize into MT (Ley et al. 1994). In some instances, phosphorylation may inhibit incorporation of tubulin into MT, while dephosphorylation promotes MT depolymerization (Banan et al. 2004b). Brain β-tubulin phosphorylated at serine by casein kinase after MT assembly retains its ability to polymerize; phosphorylated tubulin is mainly present in the assembled MT protein fraction (Serrano et al. 1987; Diaz-Nido et al. 1990).

Since actin and tubulin dephosphorylation could affect their polymerization, we examined the effects of drugs that influence actin and tubulin polymerization to determine whether they affect LFD. Changes in neurotransmitter release following pharmacological disruption of the actin cytoskeleton (Wang et al. 1996; Kim and Lisman 1999; Cole et al. 2000; Beaumont et al. 2002; Sankaranarayanan et al. 2003; Richards et al. 2004; Zhong and Zucker 2004) or microtubules (Bouron 1997; Ruiz-Canada et al. 2004; Trotta et al. 2004) have been used to determine the function of tubulin and actin cytoskeleton in pre-synaptic plasticity (reviewed by Doussau and Augustine (2000); Dillon and Goda (2005), Cingolani and Goda (2008). We found that cytoskeletal drugs affected LFD.

Both latrunculin and cytochalasin cause actin depolymerization and both blocked LFD to different extents. This is in contrast to the effects at snake motor synapse where latrunculin disrupts actin and partially blocks transmitter release (Cole et al. 2000) and frog NMJ where it reduces vesicle mobilization and exocytosis (Richards et al. 2004). Thus, our data contradict the model in which actin promotes vesicle delivery to the active zone, by providing tracks that guide the movement of synaptic vesicles towards the readily releasable pool (RRP) via actin-based molecular motors (Prekeris and Terrian 1997; Evans et al. 1998) as at snake (Cole et al. 2000) and Drosophila NMJs (Kuromi and Kidokoro 1998; Delgado et al. 2000) where N-ethylmaleimide sensitive factor participates in establishing and maintaining pre-synaptic actin (Nunes et al. 2006). Our results are in agreement with the facilitatory effect of actin depolymerization by latrunculin A on neurotransmitter release at central synapses where actin may restrain fusion of synaptic vesicles at the releasing site (Morales et al. 2000); partial disassembly of actin filaments may facilitate vesicle transport within the terminal and reassembly is necessary to limit that movement (Bernstein et al. 1998). We speculate that at crayfish synapses, cytochalasin might inhibit the second phase of LFD by increasing vesicle mobilization from the reserve pool (RP) to RRP (Wang et al. 1996; Ashton and Ushkaryov 2005) by disrupting a barrier of actin (Kuromi and Kidokoro 1998) for vesicle trafficking between RP and RRP and docking sites, possibly involving binding the vesicles via synapsin (Doussau and Augustine 2000). Perhaps in LFD, actin forms a barrier between RP and RRP holding the vesicles trapped in the RP during the second phase of LFD. Alternatively, actin may promote vesicle endocytosis to the RP as at lamprey giant reticulospinal synapse (Brodin and Shupliakov 2006) or frog NMJ (Richards et al. 2004).

Cytochalasin has minimal effect on basal transmission, yet almost completely blocked LFD. During stimulation, Ca2+ activation of actin regulatory proteins (Glenney and Weber 1981) and dephosphorylation of actin and might make actin more susceptible to cytochalasin. Depending on the preparation, cytochalasin can decrease or increase high frequency depression (Wang et al. 1996; Kuromi and Kidokoro 1998).

We expected that additional actin stabilization with jasplakinolide that stabilizes polymerized actin and shifts the balance towards assembly would accentuate the depression. Instead, jasplakinolide eliminated the fast phase of LFD similarly to cytochalasin and apparently reduced LFD. This suggests that in the first phase of LFD, rather than an intact cytoskeleton, continuous actin remodeling, filament turnover or flux of a certain population of actin is important (Halpain 2003). It is possible that jasplakinolide stabilizes the actin barrier and subsequent stimulation cannot increase the stabilization. As Cingolani and Goda (2008) point out, the understanding of how actin functions in pre-synaptic terminals is incomplete, however, the delineation of the exact mechanism by which changes in actin might affect LFD is beyond the scope of this paper.

Both stimulation to produce LFD and treatment with nocodazole disrupted microtubules (Fig. 7c and d). It is unlikely that the disruption of microtubules is the sole cause of LFD because nocodazole only caused about 20% depression by itself. However, when microtubules were stabilized by taxol, the slow phase of LFD was prevented. Since the tubulin stabilizer taxol partially blocked LFD, it seems that some part of the LFD mechanism requires changes in microtubules. MTs may function as tracks for vesicle transport (Schroer 1992) to the RP. Changes in MT cytoskeleton during LFD or treatment with taxol or nocodazole [Fig. 7b(iii), c(iii) and d(iii)] might suggest the involvement of MT in structural plasticity of pre-synaptic terminals during LFD (Becker et al. 2008).

Calcineurin and cytoskeleton

While tubulin is a substrate for calcineurin (Goto et al. 1985; Li and Handschumacher 2002), phospho-actin is not. Perhaps, as in LTD (Mulkey and Malenka 1992; Mulkey et al. 1993), calcineurin activates serine/threonine PP1 activity (Mulkey et al. 1994) by dephosphorylation and inactivation of Inhibitor-1 which, in its phosphorylated form, inhibits PP1 (reviewed in Malenka and Bear 2004; Kirkwood and Bear 1994; Mulkey et al. 1993, 1994; Mulkey and Malenka 1992). Phospho-actin is a substrate for PP1 and PP2A (Fiorentini et al. 1996; Shtrahman et al. 2005); the PP2A inhibitor calyculin induces actin serine phosphorylation, depolymerization, filament redistribution and disassembly of F-actin in renal epithelial cells (Gu et al. 2003). The function of actin is also affected by several accessory proteins (Revenu et al. 2004) some of which such as cofilin (Wang et al. 2005), filamin (Garcia et al. 2006) and GAP-43 (Lautermilch and Spitzer 2000) are substrates of calcineurin. For instance, phospho-cofilin which is dephosphorylated by calcineurin can then sever actin (Wang et al. 2005).

Pre-synaptic calcineurin activity in LFD

After pre-synaptic injection of the impermeant calcineurin inhibitor peptide, there was facilitation of transmitter release at 0.2 Hz rather than depression (Fig. 3). The conversion from depression to facilitation that occurs with the blockade of pre-synaptic calcineurin probably indicates activity of pre-synaptic stimulus-dependent kinases which are unopposed by calcineurin. Indeed, the recovery from LFD requires kinase activity (Silverman-Gavrila et al. 2005). Since the impermeant calcineurin inhibitor peptide blocks LFD only when injected into the pre-synaptic cell, LFD at crayfish phasic synapses requires pre-synaptic calcineurin activity. While we have not ruled out all possible post-synaptic mechanisms that could contribute to LFD, post-synaptic sensitivity to transmitter is unchanged. LFD differs from LTDs that require post-synaptic inhibition of calcineurin such as cerebellar LTD (Belmeguenai and Hansel 2005) and hippocampal AMPAR LTD (Morishita et al. 2005) or post-synaptic inhibition of PP1 and 2A such as cerebellar LTD (Ajima and Ito 1995) and both forms of hippocampal LTD: AMPAR LTD and NMDAR LTD (Morishita et al. 2005). LFD is also distinct from the mammalian LTD that requires activation of calcineurin in NMDAR–dependent LTD in hippocampus (Mulkey et al. 1993; Lin et al. 2008), visual cortex (Kirkwood and Bear 1994) and amygdala (Lin et al. 2003) or activation of PP1 and 2A in hippocampus (Mulkey et al. 1993, 1994), visual cortex (Kirkwood and Bear 1994), and pyramidal cells (Ajima and Ito 1995) because the activation is by post-synaptic Ca2+ signals. Several forms of LTD result from a decline in transmitter release (Choi and Lovinger 1997; Margrie et al. 1998; Li et al. 2004). However, LFD at crayfish phasic synapses differs also from pre-synaptic LTD based on mGluR because the latter is inhibited by activity of calcineurin (Li et al. 2002). Hippocampal endocannabinoid LTD that occurs at interneuron-pyramidal cell synapses (Heifets et al. 2008) requires only pre-synaptic calcineurin unlike LFD which also requires PP1 and PP2A (Silverman-Gavrila et al. 2005).

Since they seem to be evolved conservatively, many plasticity mechanisms discovered in invertebrates have been confirmed in the mammalian CNS. Future studies are needed to determine if the mechanism of LFD is more general.


Supported by a CIHR grants MGP-37773 and MOP-82827 to MPC and NSERC and HSF Postdoctoral fellowships to LBSG. Special thanks to Dr Gettemans, Ghent University for phosphoactin antibody and Dr Takahashi, University of Kitasato, Japan for syntaxin antibody. We thank Dr Lloyd Berger for technical assistance and Drs R. Collins (CIHR MOP12837) and A. Wilde (NCIC 15094) for material support of proteomics studies.