Abstract
Postsynaptic release of Ca2+ from intracellular stores is an important means of cellular signaling that mediates numerous forms of synaptic plasticity. Previous studies have identified a postsynaptic intracellular Ca2+ requirement for a form of short-term plasticity, post-tetanic potentiation (PTP) at sensory neuron (SN)–motor neuron synapses in Aplysia. Here, we show that postsynaptic IP3-mediated Ca2+ release in response to a presynaptic tetanus in an SN that induces PTP can confer transient plasticity onto a neighboring SN synapse receiving subthreshold activation. This heterosynaptic sharing of plasticity represents a dynamic, short-term synaptic enhancement of synaptic inputs onto a common postsynaptic target. Heterosynaptic sharing is blocked by postsynaptic disruption of Ca2+- and IP3-mediated signaling, and, conversely, it is mimicked by postsynaptic injection of nonhydrolyzable IP3, and by photolysis of caged IP3 in the MN. The molecular mechanism for heterosynaptic sharing involves metabotropic glutamate receptors and Homer-dependent interactions, indicating that Homer can facilitate the integration of Ca2+-dependent plasticity at neighboring postsynaptic sites and provides a postsynaptic mechanism for the spread of plasticity induced by presynaptic activation. Our results support a model in which postsynaptic summation of IP3 signals from suprathreshold and subthreshold inputs results in molecular coincidence detection that gives rise to a novel form of heterosynaptic plasticity.
Introduction
Synaptic plasticity exists in a wide range of temporal domains. Activity-dependent short-term plasticity, lasting on the order of seconds to minutes, includes paired-pulse facilitation, paired-pulse depression, and post-tetanic potentiation (PTP). Whereas long-lasting forms of plasticity, such as long-term potentiation (LTP) and long-term depression (LTD), are generally considered to provide a means of information storage in neurons, these short-term forms of plasticity are critical for dynamic, moment-to-moment adjustments of synaptic strength during online processing of neural information. At most synapses, PTP is typically thought to be mediated solely by presynaptic mechanisms (Zucker and Regehr, 2002). However, a major component of PTP at sensory neuron–motor neuron (SN-MN) synapses in Aplysia requires postsynaptic Ca2+ (Bao et al., 1997; Schaffhausen et al., 2001; Jin and Hawkins, 2003). The fact that postsynaptic Ca2+ contributes significantly to PTP at SN-MN synapses in Aplysia led us to propose that the induction of PTP at one synapse might transiently facilitate induction of plasticity at neighboring synapses onto the same MN via shared postsynaptic Ca2+-dependent mechanisms (Schaffhausen et al., 2001). Here, we show that heterosynaptic sharing of PTP can indeed occur when an input sufficient for PTP induction (20 Hz activation of an SN) is paired with a subthreshold input to a neighboring SN (activated at 7 Hz). The mechanism for heterosynaptic sharing is dependent upon postsynaptic Ca2+- and Homer-mediated interactions. Heterosynaptic sharing is blocked by postsynaptic injection of BAPTA, as well as inhibitors of intracellular Ca2+ release, including heparin, thapsigargin, and caffeine. Moreover, heterosynaptic sharing is facilitated by postsynaptic injection of nonhydrolyzable IP3 and by photoactivated caged IP3. Finally, heterosynaptic sharing is blocked by disruption of Homer-mediated protein interactions following postsynaptic injection of a synthetic peptide containing the Homer binding sequence of the Aplysia IP3 receptor (IP3R).
These findings provide evidence for a novel form of short-term heterosynaptic plasticity that arises from postsynaptic interactions between presynaptic inputs, and indicates that activity-dependent spread of plasticity is not limited to the long-term time domains. Moreover, this unique form of plasticity provides a mechanism for dynamic, temporally restricted amplification of synaptic inputs onto a common postsynaptic target.
Materials and Methods
General methods.
Wild-caught Aplysia californica (120–250 g) were obtained commercially (Marinus). Animals were anesthetized by injection of isotonic MgCl2, and isolated ganglia were pinned and desheathed in dishes containing 1:1 MgCl2/artificial seawater (ASW) (containing, in mm: 460 NaCl, 55 MgCl2, 11 CaCl2 10 KCl, and 10 Tris, pH 7.6).
Protein quantification was performed using the BCA Protein Assay (Pierce). Western blots were performed by electrophoresis using the Invitrogen NuPAGE gel system. Blotting conditions were varied depending on primary antibody used. Blocking was performed in 3% Blotto in TBS with 0.1% Tween 20, and blots were probed overnight in primary antibody in blocking solution, followed by 1 h at room temperature in HRP-conjugated secondary antibody at 1:2500. Blots were developed using the ECL Western Blotting Analysis System by GE Healthcare.
Cloning.
For use in degenerate reverse transcriptase (RT)-PCR, whole RNA was prepared from pleural and pedal ganglia using Trizol Reagent (Invitrogen). Initial forward and reverse degenerate Homer primer sequences were CAYGTNTTYCANATHGAYCC and CCAYTGNCCRAAMTTYTG, respectively. Subsequent RT-PCR was performed using a separate degenerate forward primer (ATGGGNGARCARTTYATHTT) and a gene-specific reverse primer from within the initially identified sequence. PCR was then performed on Aplysia sensory neuron or CNS libraries [generously provided by Kelsey Martin (University of California, Los Angeles, CA) and Wayne Sossin (McGill University, Montreal, Canada), respectively] with different combinations of forward gene-specific primers and T3 or T7 vector primers. Typically, PCR products were cloned into the Invitrogen pCR2.1 sequencing vector and submitted for DNA sequencing (Laguna Scientific). This approach yielded a total of 353 aa. While a termination codon was not identified within the library screen, alignment of ApHomer with Drosophila and vertebrate Homers indicated that the sequence obtained likely represented >95% of the coding region of ApHomer. The translation start site was confirmed using 5′ rapid amplification of cDNA ends with the GeneRacer kit from Invitrogen, using pleural/pedal RNA as starting material.
Cloning of the IP3 receptor in Aplysia was performed by degenerate RT-PCR from pleural/pedal ganglia and PCR of an Aplysia CNS library using gene-specific and vector primers, as described for ApHomer. Initial degenerate forward and reverse primers were CCICCNAARAARTTYMG and AAWTTYTCYTGYTCNGCRTG. DNA sequence obtained from the initial fragment of Aplysia cDNA was used to make gene-specific primers, which were then coupled to vector primers for use in amplification from the Aplysia CNS library. For expression in bacteria, the first 258 aa of ApIP3R were cloned into the pENTR vector from Invitrogen. Following recombination into the pDEST15 glutathione S-transferase (GST) expression vector, recombinant protein was expressed in BL21(DE3)Star cells. Bacterial pellets were resuspended in lysis buffer (PBS + 0.5% Triton X-100, 10% glycerol, 10 mm imidizole) and sonicated. Soluble protein was recovered by centrifugation and coupled to glutathione agarose (Pierce). Constructs for the expression of recombinant ApHomer were prepared using the EcoRI and NotI sites in the pGEX4T-3 vector (GE Healthcare), containing either amino acids 1–335 or 163–335. Following expression, bacterial cell pellets were resuspended and lysed by sonication in a smiliar fashion. The accuracy of all constructs was confirmed by DNA sequencing (Laguna Scientific, Cogenics).
Tissue distribution and ApHomer expression.
Regulation of Homer expression was investigated using four separate methods: injury, in vivo 5-HT, KCl depolarization, and tailshock. To look at the effect of injury, the animal was anesthetized, and the pleural and pedal ganglia were immediately excised into Trizol for RNA preparation. One hour later, the other half was removed and processed identically. For in vivo 5-HT treatment, animals were placed in 6 L of ASW or 6 L of ASW containing 250 μm 5-HT for 1.5 h at 18°C. After 1.5 h, animals were either taken for tissue or moved to ASW. Tailshock was performed as previously described (Sharma et al., 2003). Five shocks were given to one side of the tail with a 10 min intershock interval. The unshocked side served as a control to that animal. Pleural and pedal ganglia were dissected into Trizol. In all cases, reverse transcription was performed using 3.0 μg of total RNA, and 1.0 μl of cDNA was used in the PCR. Cycle number was optimized for each primer set, to ensure amplification was kept within the linear range for comparison between control and experimental groups. In all cases, histone h4 was used as a negative control, and the immediate early gene c/ebp was used as a positive control. c/ebp has previously been demonstrated to be rapidly transcribed following injury and in vivo 5-HT exposure (Alberini et al., 1994). RT-PCR was also used to assess tissue distribution, similarly as described for investigation of experience-dependent expression of ApHomer.
Antibody preparation and detection of protein interactions.
Frozen rat brains were obtained from Pel Freez Biologicals. Metabotropic glutamate receptor 1 (mGluR1) antibody was from BD Biosciences, and IP3R antibody was from Affinity BioReagents. Recombinant ApHomer proteins were coupled to 0.1 ml of packed glutathione-agarose overnight at 4°C. As described by Tu et al. (1998), rat cerebellar extracts were prepared by sonication in PBS with 1% Triton X-100 (PBST) plus protease inhibitors, followed by centrifugation at 15,000 × g for 10 min. Glutathione beads were washed three times with 0.3 ml of PBST plus protease inhibitors, and 40 mg of cerebellar extract was allowed to bind for 4 h at 4°C. Beads were again washed once with PBS followed by two additional washes with PBST, and bound material was removed by boiling in 100 μl of SDS-PAGE sample buffer. Equal amounts of eluted material were separated by gel electrophoresis for Western blotting with antibodies against mGluR1a and type 1 IP3R.
In the case of recombinant GST-ApIP3R, soluble proteins were similarly coupled to glutathione-agarose overnight at 4°C. Ganglia from four to six animals were homogenized with a glass-Teflon, motor-driven homogenizer into 1.0–1.2 ml of PBST plus a 1:100 Sigma mammalian protease inhibitor mixture, 3 mm Na orthovanadate, and 30 mm NaF, and were centrifuged for 10 min at 10,000 × g. A small amount of supernatant was saved for use as input material and then was divided equally for addition to agarose-coupled recombinant proteins. To inhibit the Homer–IP3R interaction, a synthetic peptide corresponding to Aplysia IP3R amino acids 41–52 (GDLNNPPKKFRD; mutant peptide GDLNNPLKKRRD) was synthesized by Invitrogen. Peptide was added to Aplysia CNS extract before addition of the extract to glutathione-coupled recombinant protein.
Following binding on a rotator for 4 h at 4°C, unbound material was removed by three washes of PBST plus inhibitors. Bound material was eluted by addition of SDS-PAGE sample buffer and boiling for 5 min, followed by analysis by Western blot for ApHomer. A custom affinity-purified ApHomer antibody was prepared by Covance Research Products against the ApHomer peptide,-LQDSQRQLQDGRSSRDQEHRE (amino acids 324–344). The synthetic peptide was conjugated to KLH and used as immunogen for antibody generation in rabbits.
Electrophysiology.
For electrophysiology, pleural and pedal ganglia were desheathed in 1:1 isotonic MgCl2/ASW to expose the tail sensory and motor neuron clusters. MgCl2 was removed by perfusion of the preparation with ASW for at least 15 min. Intracellular recording was performed using glass microelectrodes containing 3 m KCl with electrode resistance of 8–12 MΩ. Motor neurons were hyperpolarized to approximately −70 mV. The resting potential and input resistance of the SN and MN were monitored throughout each experiment. Baseline EPSP amplitudes were determined by injection of current to elicit a single test spike in the SN in each of the two presynaptic SNs (∼200 ms apart). One minute later, a tetanizing stimulus (generated by individual current pulses [10 nA] at 20 or 7 Hz for 2 s) was applied to the SNs. A posttest EPSP was measured 10 s after the offset of the stimulus train. The 200 mm BAPTA (tetrapotassium salt, Invitrogen), 25 μm heparin (Calbiochem), and 125 μm IP3 were dissolved in 3 m KCl, 10 mm HEPES, and 3 mm Fast Green dye (Fisher Scientific). Drugs were injected into the MN with two to four pulses (20 ms, 20–40 psi) from a picospritzer (General Valve), and at least 20 min were allowed for drugs to take effect before pretest values were obtained. For caffeine experiments, caffeine (5 mm, Sigma) was dissolved in ASW and perfused over the ganglion through the experiment. Thapsigargin (5 μm, Sigma) was prepared in 0.1% DMSO and continually perfused as in the case of caffeine.
For mGluR1-blocking experiments, LY367385 (300 μm) and S-MCPG (500 μm) (Tocris Bioscience) were prepared in 110 mm NaOH and diluted 1:250 in ASW immediately before each experiment. In vehicle control experiments, 200 μl of 110 mm NaOH was added to 50 ml of ASW (1:250) and perfused in a similar manner as drug-treated preparations. Following removal of MgCl2 by perfusion, drug or vehicle was bath applied for at least 30 min, at a flow rate of 0.5–1 ml/min, before the commencement of recording.
Synthetic ApIP3R peptide (active and mutant) was prepared at a concentration of 1.0 mm in 0.5 m KOAc, 10 mm TrisCl, pH 7.5, and 15 mm Fast Green dye (∼1.0%). Peptides were injected into the MN by 28–32 pulses, 10–40 ms each, at 20 psi. The number and duration of pulses was determined depending upon the resistance of injection electrode and the size of the cell. Following injection, cells were continually perfused in ASW for at least 45 min before recording from MN. Monosynaptic EPSPs were measured and PTP performed as described above. Short-term facilitation induced by a single pulse of 5-HT was also assessed following three pretests taken at 15 min intervals. Ten minutes after the last pretest, 50 μm 5-HT was perfused for 5 min, and a single post-test was performed immediately.
Input resistance and excitability of the injected MN were monitored in all experiments, and any preparations that indicated possible cellular damage were discarded before onset of the stimulus train (∼10% of cases).
Cell culture.
Cell culture was performed based on a modified protocol as described previously (Gruenbaum and Carew, 1999; Lyles et al., 2006). Briefly, pleural and abdominal ganglia from 80 to 100 g of Aplysia were first incubated in 10 mg/ml Protease type IX (Sigma) at 34.5°C for 2 h and 15 min and 2 h and 10 min, respectively. The ganglia were then desheathed. SNs were isolated from the pleural ganglia, and LFS MNs were isolated from the abdominal ganglia. Identified SNs or MNs were then plated in culture dishes. The sensory neuron processes were carefully manipulated to contact the motor neurons. Culture dishes were precoated with poly-l-lysine (0.5 mg/ml). Culture medium consisted of 50% Aplysia hemolymph and 50% L-15 medium supplemented with salts (NaCl, 260 mm; CaCl2, 10 mm; KCl, 4.6 mm; MgSO4, 25 mm; MgCl2, 28 mm; NaHCO3, 2 mm; HEPES, 15 mm) and D-glucose (34.6 mm). Additionally, glutamine (2 mm), penicillin (100 U/ml), and streptomycin (100 U/ml) were also included in the culture medium. Cultured cells were left on the microscope stage overnight and then transferred to an 18°C incubator. After 4 d, Ca2+ imaging was performed on MNs, and synaptic strengths between the SN-MN cocultures were measured. Electrophysiological recordings were performed similarly to those in the intact CNS.
Photolysis of caged IP3 and calcium imaging in MN neurites.
Cultured MNs were injected with 1 mm caged IP3 (Invitrogen) and 10 mm cell-impermeable Fluo-4 Ca2+ indicator dye (Invitrogen) using a picospritzer, similar to the microinjection procedure described in the intact ganglia. To aid visualization of calcium release in MN neurites, cells were pretreated with a high-K+ (100 mm), high-calcium (55 mm), and low-magnesium (11 mm) ASW for 5 min (Malkinson and Spira, 2010). This conditioning depolarization loads intracellular calcium stores (Garaschuk et al., 1997; Hong and Ross, 2007). Imaging was accomplished using a custom-built microscope system based around an Olympus IX 50 microscope equipped with an Olympus 60× oil objective lens (numerical aperture, 1.45), using excitation by a solid-state 488 nm laser (Spectra Physics sapphire) and imaging emitted fluorescence at λ > 510 nm using a Cascade 650 electron-multiplied CCD camera (Roper Scientific).
Fluorescence signals are expressed as ratios (F/F0 or ΔF/ F0) of the fluorescence (F) at each pixel relative to the mean resting fluorescence (F0) before stimulation. IP3 was photoreleased from a caged precursor by delivering flashes of UV light (350–400 nm) focused uniformly throughout the field of view. Image data were streamed to computer memory and then stored on disc for offline analysis using the MetaMorph software package (Molecular Dynamics).
Photolysis of caged IP3 and electrophysiology in SN-MN coculture.
The electrophysiological studies using SN-MN cocultures were performed on a different experimental setup than that used to visualize Ca2+ transients in neurites (as described above). Following microinjection with caged IP3 and fluo-4, cultured cells were again subject to a conditioning depolarization by treatment with a high K+ (100 mm) ASW for 5 min to load intracellular Ca2+ stores and then subject to a 2 s UV flash. This conditioning depolarization pretreatment, followed by washout, had no effect on subsequent baseline synaptic transmission. Flashes of UV light were derived from a Zeiss Xite 120 metal halide light source via a UV filter set in a Zeiss Axiovert 200 inverted microscope (Leissring et al., 1999) and were delivered to the MN soma and proximal neurites.
EPSPs of sensory–motor synapses before and after (1) 20 Hz stimuli, (2) 7 Hz stimuli, (3) 7 Hz stimuli paired with the UV flash, (4) UV flash alone, and (5) no stimulation were then compared.
Data analysis and statistics.
Postsynaptic EPSP amplitudes were measured as the peak voltage of the EPSP. In the case of 5-HT-induced facilitation, baseline EPSP was determined based on the average of three pretests in ASW, and only EPSPs with pretests within 20% of the mean were used for further analysis. In the case of PTP, a single pretest was performed as described previously (Schaffhausen et al., 2001). All EPSP amplitudes are expressed as a percentage of the baseline mean. Two-tailed t tests were used for data analysis. All data are shown as ±SEM.
Results
PTP can be shared at SN-MN synapses
To directly test the hypothesis that convergent inputs might result in postsynaptic sharing of PTP, we first asked whether presynaptic activation of an SN at a level that is ineffective in producing PTP might gain the ability to induce PTP when paired with effective activation of a neighboring presynaptic SN. Intact ganglia (abdominal or pleural-pedal, in separate experiments) were excised, and neurons were impaled for intracellular recording. By eliciting a single presynaptic action potential and recording the evoked postsynaptic EPSP, two presynaptic SNs were identified that made synapses onto a common MN. At least 15 min after identification of convergent synapses, each SN was activated with a train (2 s in duration) of either 7 Hz (which produces no PTP) or 20 Hz (which induces PTP) (Fig. 1, inset). EPSPs were then measured 10 s after the offset of the trains. While 7 Hz alone produced no PTP, when paired with a 20 Hz train to the other SN, the 7 Hz train induced a robust enhancement of synaptic strength (Fig. 1A,B). Additional experiments with no activation of one SN and 20 Hz activation of the other SN revealed that nonactivated synapses do not passively inherit this enhancement (Fig. 1B), indicating that the 7 Hz train likely generates a subthreshold postsynaptic signal that summates with the signal initiated by the 20 Hz train. Thus, the temporal registry of trains from both cells is required for the “sharing” of plasticity between synapses.
Induction of PTP can be shared with a neighboring synapse receiving subthreshold activation. A, Inset, Experimental design: two presynaptic SNs are identified that synapse onto the same postsynaptic MN. The 20 and 7 Hz trains are delivered intracellularly to the two SNs, while postsynaptic EPSPs are recorded in the MN. Representative traces are shown from synapses receiving 7 Hz alone (A1, top left), 20 Hz alone (A1, bottom left), or 7 Hz paired with 20 Hz (A2, top and bottom, respectively). B, Histogram of results indicating that an SN receiving a 7 Hz train acquires the ability to induce heterosynaptic sharing, but only when paired with a 20 Hz train to a neighboring SN. Previously, we have shown that small synapses (<6.2 mV) exhibit an enhanced form of PTP, which is more robust than the magnitude of PTP observed across synapses of all sizes (∼160 vs 285%) (Schaffhausen et al., 2001). Here, we observed a similar result, in which 20 Hz PTP varied from ∼200%, when tested among small and large synapses alike, to ∼300%, when experiments were restricted to small synapses (compare Fig. 1B and Figs. 1C, 4A, and 7C). Interestingly, the magnitude of the shared 7 Hz PTP is consistently ∼200%, regardless of the degree of potentiation at the “donor” SN. *p < 0.05, n = 4–7 per group. C, Top, schematic diagram of experimental protocol. Pretests of both SNs are taken ∼100 ms apart, 1 min before the onset of the overlapping 7 and 20 Hz trains (each 2 s in duration). Post-test measurements are typically made 10 s after the offset of the trains (e.g., as in 2B). For time course experiments, a single post-test measurement was made for each synapse at 1 s, 1 min, or 10 min after the overlapping trains. As shown in C (bottom), while 20 Hz PTP remains significantly above baseline at 10 min, the shared 7 Hz PTP is no longer significantly above baseline by 1 min (n = 4–7 per group). D, Sharing of PTP requires that the 7 and 20 Hz trains must occur within a narrow time window.
We next examined the time course of heterosynaptic sharing and found that it decayed faster than PTP induced by the 20 Hz train (Fig. 1C). By 1 min after simultaneous activation, the EPSP associated with the 7 Hz train was no longer significantly different from baseline, while the corresponding 20 Hz EPSP was still significantly elevated at 10 min. Next, to determine the temporal constraints on induction of heterosynaptic sharing, we varied the time between 7 and 20 Hz activation of the two SNs. Sharing was observed when the onset of trains was separated by ≤3 s (Fig. 1D) and was independent of the order of activation, indicating that the two SNs need not be coactive but that their activation must occur within a restricted time window. The symmetrical and relatively broad (3 s) permissive time window for heterosynaptic sharing contrasts with previously described forms of associative plasticity such as spike timing-dependent plasticity and input timing-dependent plasticity, which have much more restricted temporal windows (in the range of milliseconds), and even for associative LTD in the cerebellum, in which pairing of inputs can occur on the order of hundreds of milliseconds (Sjöström and Nelson, 2002; Wang et al., 2003; Dan and Poo, 2004; Dudman et al., 2007; Sarkisov and Wang, 2008). This dissociation of temporal requirements suggests that heterosynaptic sharing represents a unique form of synaptic interaction whereby activity at one set of inputs can change the gain for local information flow at other synaptic sites.
Heterosynaptic sharing requires postsynaptic Ca2+
To examine the postsynaptic Ca2+ dependency of heterosynaptic sharing, we injected the Ca2+ chelator BAPTA into the MN, and 20 and 7 Hz trains of activation were delivered to two SNs. In contrast to the control condition, synaptic enhancement in the 7 Hz SN was entirely blocked (Fig. 2A). We and others have previously shown that NMDA receptor activation is not required for induction of PTP at the SN-MN synapse (Schaffhausen et al., 2001; Jin and Hawkins, 2003). Moreover, Jin and Hawkins (2003) have previously shown that PTP in SN-MN cultures is dependent on intracellular Ca2+ release, but not on voltage-gated Ca2+ influx or extracellular Ca2+ concentration. Thus, we next examined the contribution of Ca2+ from intracellular stores. Perfusion of caffeine (which causes Ca2+ store depletion via the ryanodine receptor) and thapsigargin (a sarco/endoplasmic reticulum Ca2+-ATPase pump antagonist), as well as postsynaptic injection of heparin (an IP3R antagonist), also resulted in a total block of heterosynaptic sharing (Fig. 2A). Interestingly, although injection of heparin (25 μm) completely blocked heterosynaptic sharing, this concentration had no effect on PTP. However, when we injected a higher concentration of heparin (100 μm) we observed a significant attenuation of PTP (Fig. 2A), which is consistent with previous observations by Jin and Hawkins (2003), who also observed a significant attenuation of PTP in cultured SN-MNs by injecting this concentration of heparin into the MNs. The remaining PTP in the heparin (100 μm) experiments (Fig. 2A) is likely due to the presynaptic contribution to PTP at these synapses (Schaffhausen et al., 2001).
Heterosynaptic sharing requires postsynaptic intracellular calcium. A, Degree of sharing (control, open bar, n = 7) is significantly blocked in the presence of Ca2+ inhibitors BAPTA (n = 8), caffeine (n = 5), heparin (n = 7), and thapsigargin (n = 5). While 25 μm heparin is sufficient to block sharing, that concentration is without effect on PTP (black bar, n = 6). In contrast, 100 μm heparin results in significant, attenuation of PTP (n = 4). *p < 0.05, within-group comparison. B, Injection of nondegradable IP3 into an MN has no effect on baseline synaptic transmission from an unstimulated SN (0 Hz, open bar), but when paired with a 7 Hz train delivered to the SN results in significant PTP (n = 7 per group).
While perfusion of thapsigargin and caffeine in the bath could affect both presynaptic and postsynaptic Ca2+, the block observed by these reagents, taken together with the block observed following injection of heparin into the MN, are consistent with a possible role of IP3-mediated postsynaptic plasticity. To directly test this idea, we injected nondegradable IP3 into the MN. In this case, a 7 Hz SN tetanus alone induced significant PTP (Fig. 2B). Collectively, these results suggest that effective (20 Hz) SN activation generates IP3-dependent release of Ca2+ from internal stores in the MN, which facilitates the induction of sharing at the neighboring (7 Hz) synapse.
Photolysis of caged IP3 induces Ca2+ transients in the MN neurites
As demonstrated above, nondegradable IP3–7 Hz pairing induces PTP (Fig. 2B). One potential mechanism that may account for this observation involves IP3-mediated Ca2+ release. To test this hypothesis, MNs were injected with Fluo-4 and caged IP3. As shown in Figure 3A,B, repetitive, localized Ca2+ transients in neurites were evoked by photorelease of IP3 following a UV flash. These observations are representative of results in four cells, obtained at 16 locations up to 500 μm away from the cell body. These localized Ca2+ transients likely arise from clusters of IP3Rs and are reminiscent of puff activity as reported in mammalian cells (Smith et al., 2009).
Photolysis of caged IP3 induces Ca 2+ release in MN neurites. A, Monochrome image of resting fluo-4 fluorescence in neurites from a motor neuron. A, Circled white areas indicate regions of interest from which intensity profiles were generated over time (shown at right). Traces depict local Ca2+ transients from each region evoked by photorelease of IP3 using a UV flash (indicated by the arrow). B, Pseudocolored Ca2+ images (F/F0) before (top left) and at different times after (1a, 2b, and 3c) photorelease of IP3. Images are taken from the same experiment as in A and were captured at times indicated by the corresponding letters by the traces. Arrows indicate the neuritic regions where Ca2+ transients were captured. Each image is an average of eight consecutive frames acquired at an exposure time of 200 ms.
Heterosynaptic sharing is mimicked by pairing 7 Hz activation and photolysis of caged IP3
To further elucidate the mechanisms of heterosynaptic sharing, we used the photolysis of caged IP3 to mimic the in situ generation of an IP3 signal induced by 20 Hz stimuli. We predicted that if IP3-mediated intracellular Ca2+ release is a critical component of the mechanism for sharing, then the transient Ca2+ release induced by the photoactivation of caged IP3 in the MN, when paired with a train of subthreshold 7 Hz tetanus in the SN, should mimic facilitation induced by heterosynaptic sharing. In SN-MN cocultures, MNs were injected with caged IP3 and the calcium indicator Fluo-4. A UV flash, which liberates free IP3 and induces calcium transients in the MN neurites (Fig. 3) was delivered to the MNs. We then examined the synaptic strengths at the SN-MN synapses before and after (1) a 20 Hz train, (2) a 7 Hz train, (3) a 7 Hz train paired with the photolysis of caged IP3, (4) photolysis alone, and (5) no stimulation (No Stim) (Fig. 4). We observed that the SN-MN synapses were significantly potentiated by 20 Hz stimuli alone (Fig. 4A,B). The significant potentiation exhibited following 20 Hz stimulation is all the more striking when compared with the No Stim control (Fig. 4B), which exhibited synaptic decrement typically observed in SN-MN cultured synapses (Eliot et al., 1994; Bao et al., 1997; Jin and Hawkins, 2003). Thus, the 20 Hz-induced potentiation is superimposed on a progressively decrementing baseline [a similar result was previously observed by Jin and Hawkins (2003)]. SN-MNs exposed to either 7 Hz alone or photolysis alone also exhibited synaptic decrement that was comparable to the No Stim condition (Fig. 4B). In contrast, the pairing of 7 Hz stimuli and photolysis of caged IP3 induced significant synaptic potentiation (Fig. 4A,B). The facilitation induced by pairing 7 Hz with IP3 photolysis was significantly greater than that induced by the 7 Hz stimuli alone and was not significantly different from the facilitation induced by the 20 Hz stimuli alone (Fig. 4B). These data support the hypothesis that the postsynaptic intracellular Ca2+ transients induced by IP3 receptor activation are critically involved in heterosynaptic sharing.
Sharing-like potentiation is induced by pairing 7 Hz activation with photolysis of caged IP3. SN-MN pairs are cocultured. A 20 or 7 Hz train is delivered extracellularly to the SN, while EPSPs are recorded in the MN. The MN is microinjected with caged IP3 and receives a UV flash (Fig. 3). A, Representative EPSPs are shown from SN-MN synapses before and after a 20 Hz train (top), a 7 Hz train alone (middle), or a 7 Hz train paired with a photolysis flash (bottom). B, Summary histograms indicating that 7 Hz activation paired with the photolysis of caged IP3 in the postsynaptic MN acquires the ability to induce potentiation comparable to a 20 Hz train. In contrast, 7 Hz activation alone or photolysis alone exhibits synaptic decrement comparable to unstimulated SN-MN pairs. n = 4–8 in each group.
Molecular mechanisms of heterosynaptic sharing require postsynaptic Homer-mediated protein interactions
We next explored the postsynaptic molecular mechanisms underlying sharing. Previous work by Jin and Hawkins (2003) indicated that mGluRs are required for PTP in cell culture. Thus, we examined the effects on monosynaptic PTP in pleural/pedal ganglia of two type I mGluR inhibitors, LY367385 and S-MCPG. While control synapses demonstrated robust PTP, both LY367385 and S-MCPG caused a significant attenuation of PTP, with no effect on baseline synaptic transmission (Fig. 5A,B). The remaining PTP observed in the presence of inhibitors is likely due to the presynaptic component of PTP at these synapses (Schaffhausen et al., 2001).
PTP in Aplysia requires type I mGluRs. A, PTP is significantly attenuated in the presence of mGluR inhibitors S-MCPG or LY367385 (ASW, n = 10 per group; S-MCPG/LY367385, n = 7 per group). *p < 0.05, within-group comparison. B, The blockers had no effect on synapse size (n = 7–10 per group).
Given the mGluR/Ca2+ dependency and IP3 involvement in sharing, an attractive molecular candidate for mediating PTP and heterosynaptic sharing was Homer, a well described protein in the vertebrate postsynaptic density. Homer proteins mediate a series of interactions with other proteins including mGluRs as well as IP3 receptors (Tu et al., 1998; Worley et al., 2007). Moreover, Homer proteins have been demonstrated both to couple mGluR signaling with intracellular Ca2+ release in vertebrate neurons (Tu et al., 1998) and to localize IP3Rs within dendritic subregions, allowing for synaptically localized IP3-mediated Ca2+ release (Sala et al., 2001, 2005). However, there was no available evidence to indicate that a Homer ortholog is expressed in Aplysia, or whether it might play a role in the spread of plasticity.
Thus, to directly explore the possible involvement of Homer in sharing, we first cloned Aplysia Homer (ApHomer) using a combination of degenerate RT-PCR and library screening (Fig. 6A,B). ApHomer shares the domain organization of other Homer proteins, including an N-terminal EVH1 domain and a C-terminal coiled-coil region. In vertebrates, Homer has been predominantly studied in the nervous system but is also expressed in tissues outside the CNS, as observed for ApHomer (Fig. 6C). In comparison, the gene for Aplysia MAGUK protein ApSAP is expressed predominantly within the CNS (Reissner et al., 2008). Because Homer was first identified as an activity-dependent immediate early gene (Brakeman et al., 1997; Kato et al., 1997), we used semiquantitative RT-PCR to assess message levels in response to injury, in vivo 5-HT, KCl depolarization, or tailshock (which induces behavioral sensitization). None of these treatments induced a significant change in the cDNA levels of histone H4 or Homer but did induce a substantial increase in a positive control, C/EBP (Fig. 6D). These findings thus indicate the lack of an activity-dependent splice variant of Homer and suggest that there is no change in the amount of the constitutively expressed splice form of Homer following experience. These results are consistent with findings in Drosophila, where there appears to be only one gene and one transcript for Homer (Diagana et al., 2002). However, we did identify an alternative splice variant at the extreme 5′ end of the mRNA, which would translate into proteins differing among the first six amino acids (supplemental Fig. 1, available at www.jneurosci.org as supplemental material).
Identification of a Homer ortholog expressed in Aplysia. A, Amino acid sequence of ApHomer is shown in alignment with Drosophila Homer and mouse Homer 1b. Identical amino acids are shaded in red. The EVH1 region is bracketed in blue. The cDNA sequence for ApHomer is available under GenBank accession number EU999995. B, Schematic comparison of Aplysia Homer with mouse and Drosophila Homers. Alignment within the EVH1 domain shows a particularly high degree of sequence identity among Homer family members. The percentages shown indicate the degree of sequence identity within that region between Aplysia and Drosophila Homer, or Aplysia and mouse Homer 1b. C, Semiquantitative RT-PCR indicates expression of Homer message in a variety of tissues. CDNA was prepared from buccal mass (lane 1), CNS (lane 2), heart (lane 3), muscle (lane 4), and sperm sac (lane 5). Lane 6 was amplified from water as a template. ApSAP was amplified as a CNS-enriched control, and histone H4 was amplified as a ubiquitously expressed control. D, Semiquantitative RT-PCR of RNA from pleural and pedal ganglia. Lane 1, Water control template; lane 2, cDNA prepared from RNA taken from an ASW-treated animal; compare with lane 3, cDNA prepared from an in vivo 5-HT-treated animal; lane 4, cDNA prepared from RNA taken immediately from an anesthetized animal; lane 5, cDNA prepared from the other half of the same animal, 1 h later. No change is observed in amplification of histone H4 or Homer, while C/EBP levels are robustly increased. Similar results were observed following tailshock and KCl (100 mm) depolarization (data not shown).
A prerequisite for our hypothesis that ApHomer has a mechanistic role in heterosynaptic sharing is that Homer-mediated interactions observed in vertebrates, particularly with mGluRs and IP3Rs, should be conserved in Aplysia. We found that recombinant ApHomer could successfully coprecipitate rat mGluR1α in a GST pull-down experiment (Fig. 7). However, a low signal-to-noise ratio result made it difficult to interpret pull-down results for IP3R (Fig. 7, bottom right). To directly examine a possible interaction between ApHomer and Aplysia IP3R, we determined the coding sequence for the initial 154 aa of ApIP3R. A cDNA for Aplysia IP3R has also been reported, matching the sequences we obtained (Cummins et al., 2007). The consensus sequence for Homer binding is PXXFr (Tu et al., 1998), and in rat IP3R it is found at amino acid positions 49–54. This sequence is 100% conserved between Aplysia and other vertebrates and invertebrates (Fig. 8A).
ApHomer coprecipitates rat mGluR 1α. Top, GST (1) or GST fusion proteins with either amino acids 163–335 (2) or the first 335 aa of ApHomer (3) were bacterially expressed and purified for a GST pull-down with rat cerebellar extract. Bottom, Western blot for mGluR1α and IP3R of pull-downs. Lane numbers represent the matching number beside constructs used as bait. In, input. Full-length GST–ApHomer coprecipitates mGluR1α (Western blot: left panel, upper portion, lane 3), whereas GST alone (lane 1) and a GST fusion protein containing the C-terminal portion of ApHomer (lane 2) do not. In contrast, an interaction between ApHomer and rat IP3R could not be conclusively demonstrated due to a nonspecific signal observed between all constructs and IP3R. Top, Left and right, shows Western blot for mGluR1 and IP3R, whereas bottom, left and right, shows the same blots stripped and reprobed with an anti-GST antibody. The presence of GST or GST fusion proteins is indicated by arrowheads at left.
Interaction between ApHomer and ApIP3R is required for PTP sharing. A, Alignment of the first 54 aa of ApIP3R with starfish, mouse, and Drosophila. Conserved amino acids are shaded. A synthetic peptide including the Homer-binding consensus sequence is boxed in blue. B, GST pull-down in which GST–ApIP3R was used to isolate ApHomer from Aplysia CNS extract. Lane 1, Input; lane 2, pull-down with GST as bait; lane 3, ApIP3R pull-down without any peptide included; lane 4, pull-down including low-concentration peptide; lane 5, pull-down including high-concentration peptide; lane 6, pull-down including high-concentration mutant peptide. Blots for ApHomer (top) were stripped and reprobed for GST (bottom). C, IP3R peptide (n = 6) or mutant peptide (n = 5) was injected into an MN before induction of PTP and sharing. The mutant peptide has no effect, whereas the blocking peptide completely prevents sharing. *p < 0.05, within-group comparison.
The first 256 aa of ApIP3R were expressed as a GST fusion protein for a pull-down assay, as described above. A custom antibody was generated against ApHomer amino acids 324–344, which resulted in immunoreactivity against multiple bands in Aplysia CNS extract (Fig. 8B, lane 1). However, Western blot following the GST pull-down assay indicated that, while the anti-ApHomer antibody recognizes a complex series of bands in CNS extract used as input, only a single band of ∼47 kDa is observed in samples eluted from beads coupled to GST-ApIP3R (Fig. 8B, lanes 3–6). This is in agreement with the ∼45–50 kDa size of mammalian and Drosophila Homers and the expected size of ApHomer. Moreover, inclusion of a 12 amino acid peptide matching the Homer recognition sequence within ApIP3R competed away this interaction in a concentration-dependent manner, whereas a mutant peptide had no effect (Fig. 8B).
The synthetic peptide used to inhibit the ApIP3R–ApHomer interaction in vitro now provided a powerful experimental tool to directly explore Homer-mediated IP3 activation in sharing. We pressure injected MNs with either the ApIP3R peptide or mutant peptide, in which two critical amino acids in the Homer-binding sequence were substituted. One hour postinjection, two SNs were activated with either 20 or 7 Hz to explore PTP and sharing (Fig. 8C). Following injection of the mutant peptide, 20 Hz activation induced significant PTP, and simultaneous 7 Hz activation induced significant heterosynaptic sharing. In contrast, following injection of the ApIP3R peptide into the MN, the 20 Hz PTP was reduced, indicating a partial block by the active peptide, which is consistent with our previous observation that 20 Hz PTP has both presynaptic and postsynaptic components (Fig. 5) (see also, Schaffhausen et al., 2001). Importantly, heterosynaptic sharing induced by simultaneous activation with 7 Hz was completely blocked. No effect was observed on baseline synaptic transmission. Finally, the attenuation of PTP and block of sharing by the peptide was restricted to short-term, activity-dependent plasticity at the SN-MN synapse. Short-term facilitation induced by a single pulse of 5-HT was unaffected by the peptide (supplemental Fig. 2, available at www.jneurosci.org as supplemental material).
Discussion
Our results elucidate a series of molecular steps that provide a mechanism through which short-term plasticity may be shared between two neighboring synapses. Collectively, our findings support the hypothesis that low-level (e.g., 7 Hz) activation of an SN alone leads to postsynaptic release of IP3, which is subthreshold for the induction of PTP, but can summate with IP3 liberated by high-level activation (e.g., 20 Hz) of another presynaptic input (Fig. 9). The idea that IP3 summation can result in molecular integration has been previously suggested by Parker and colleagues, who found that caged IP3 responses can summate at distances up to 20 μm (Parker, 1989; Parker and Miledi, 1989; Parker and Ivorra, 1992). SN-MN synapses in Aplysia commingle with intersynaptic distances in many cases well within this 20 μm range (Marinesco and Carew, 2002), supporting the idea that subthreshold IP3 released in the postsynaptic neuron by low-level activation may summate with additional IP3 from high-level input to generate a form of molecular coincidence detection that induces a sufficient Ca2+ signal to give rise to heterosynaptic sharing.
Molecular model for sharing of short-term synaptic plasticity. Model for molecular coincidence detection: dotted line represents IP3 threshold for intracellular Ca2+ release sufficient to induce PTP. Abscissa indicates Ca2+/IP3 spread within a neurite. A, 20 Hz delivered to an SN is sufficient for local threshold IP3 generation (blue), thus leading to intracellular Ca2+ release (green). B, However, 7 Hz alone does not achieve the IP3 threshold required for Ca2+ release. C, In contrast, when neighboring synapses are coactivated at 7 and 20 Hz within a narrow temporal window, summation of local IP3 signals surpass the threshold for Ca2+ release and thus give rise to potentiation of the 7 Hz SN synapse, resulting in sharing of synaptic enhancement.
In the long-term time domain (e.g., LTP and LTD), Ca2+-dependent heterosynaptic expansion of synaptic plasticity has been observed in other systems (Nishiyama et al., 2000; Royer and Pare, 2003). In addition, communication between proximal hippocampal CA1 synapses has been revealed by facilitation of LTP at a minimally activated synapse following LTP induction at a nearby synapse (Harvey and Svoboda, 2007). However, the heterosynaptic amplification of hippocampal LTP is not dependent on intracellular Ca2+, whereas the heterosynaptic sharing we observe requires a postsynaptic Ca2+ response coupled to Homer- and IP3R-dependent protein–protein interactions.
The current study raises two important questions regarding Ca2+ signaling. First, how might postsynaptic Ca2+ mediate the induction of PTP at strongly activated synapses, and second, how might postsynaptic Ca2+ contribute to heterosynaptic sharing at weakly activated synapses? Considering the first question, one candidate mechanism for the induction of PTP at strongly activated synaptic sites is CaMKII, since it is well established that Ca2+-dependent kinases can be critical mediators of synaptic plasticity. For example, CaMKII regulates the conductance of AMPA receptors during LTP by phosphorylation of GluR1 at Ser831 and also may indirectly mediate insertion of AMPA receptors into the membrane by phosphorylation of interacting molecules that govern membrane insertion (for reviews, see Bredt and Nicoll, 2003; Malenka, 2003; Boehm and Malinow, 2005). Consistent with this possibility, Jin and Hawkins (2003) have found that postsynaptic CaMKII is required for PTP in cultured SN-MN synapses in Aplysia, supporting the hypothesis that Ca2+-activated CaMKII may contribute to a postsynaptic mechanism for the induction of PTP we observe at strongly activated central synapses. Considering the second question, how might Ca2+ contribute to sharing? It is possible that the spread of IP3 from strongly activated synapses to neighboring weakly activated ones could induce Ca2+ release at the weakly activated synapses and thus engage the same CaMKII-mediated signaling events at those neighboring sites as are responsible for the induction of PTP at the “donor” synapse. However, we cannot exclude the possibility that the postsynaptic mechanisms contributing to the induction of PTP at strongly activated synapses could differ from those required for heterosynaptic sharing. What we can say, however, is that both classes of plasticity require postsynaptic Ca2+. Other Ca2+ targets beyond CaMKII may contribute both to the induction of PTP and to heterosynaptic sharing as well. For example, it has recently been shown that Ca2+-dependent Ras activation occurs rapidly within individual spines (Harvey et al., 2008) and is thus a candidate for PTP induction at the strongly activated site. Moreover, activated Ras can spread to neighboring spines (Harvey et al., 2008), making it also an attractive candidate for mediating signaling contributing to PTP sharing. Thus, a number of interactive Ca2+-dependent postsynaptic signaling events may contribute both to the initial induction of PTP as well the sharing of PTP that we observe.
Our data also indicate that ApHomer-mediated protein interactions are critical for heterosynaptic sharing, since the synthetic peptide that interrupts Homer–IP3R interactions completely blocks sharing. Interestingly, injection of the ApIP3R peptide into the MN also tends to reduce the level of PTP (although not to a level that reaches statistical significance), suggesting that, as in the case of heparin, blockade of the signaling pathway between Homer and intracellular Ca2+ release is more readily impaired in the case of heterosynaptic sharing compared with PTP. Another difference between heterosynaptic sharing and PTP is their time course. Sharing is significantly attenuated after 60 s, whereas PTP lasts >10 min. Thus, although not conclusive, the differential sensitivity of heterosynaptic sharing and PTP to both heparin concentration and the ApIP3R peptide, taken together with their differential time courses, suggests that these two forms of synaptic plasticity are not mechanistically identical.
What are the mechanisms by which disruption of Homer-mediated interactions might inhibit induction of PTP and heterosynaptic sharing? An obvious candidate is postsynaptic Ca2+ release at the activated synapse, localized by a protein scaffold linking type I mGluRs, Homer, and IP3Rs. However, since the conserved Homer binding sequence is present in additional Homer ligands, these interactions are also likely to be disrupted by the synthetic peptide as well. For example, a number of Homer ligands containing the PXXFr sequence have been identified, including the Ryanodine receptor Shank, Dynamin III, PI3 kinase enhancer, and transient receptor potential cation channels (Tu et al., 1998; Feng et al., 2002; Rong et al., 2003; Yuan et al., 2003; Lu et al., 2007). These interactions serve to regulate release of Ca2+ from intracellular stores, activation of PI3 kinase, maturation of dendritic spines, as well as the localization of synaptic AMPA receptors and modulation of synaptic strength (Sala et al., 2001; Rong et al., 2003; Sala et al., 2005; Lu et al., 2007). Although our results support a model in which mGluR–Homer–IP3R interactions mediate a local Ca2+ signal near activated synapses, which may travel to nearby synapses as part of the mechanism of sharing (Fig. 9), we cannot rule out other Homer-dependent mechanistic components to heterosynaptic sharing.
Finally, an intriguing question arises from the present work. What is the functional significance of heterosynaptic sharing? The SN-MN synapses examined in this study mediate the monosynaptic component of the tail-elicited tail withdrawal reflex in Aplysia, in which a tactile stimulus to the tail gives rise to the withdrawal of the tail from the site of stimulation. Because there is an approximate topographic arrangement of SN-mediated tactile input from the tail skin onto the tail MNs (Walters et al., 2004), the sharing of plasticity we observe could allow the animal to be transiently more responsive to a weak stimulus applied to a neighboring area on the tail where a strong stimulus recently occurred. This enhancement of sensitivity within the net receptive field of the reflex would be constrained to a restricted time window commensurate with the duration of the net plasticity induced at neighboring SN presynaptic inputs to the tail MNs. In this fashion, the functional tactile receptive field of the reflex would be dynamically and rapidly expanded following a focal strong stimulus to the tail, with a time course sufficient for the animal to make an integrated, adaptive behavioral response.
In principle, one could expand this general notion to any site in a neural network where there is convergent presynaptic input onto a common postsynaptic target. At these convergent sites, heterosynaptic sharing of short-term plasticity could provide a potential mechanism for the time-limited rapid and dynamic expansion of the functional “receptive field” of a postsynaptic neuron. From this perspective, inputs whose activation are below threshold for the induction of plasticity can, by virtue of temporal coincidence with other suprathreshold inputs, subsequently acquire plasticity and thus more effectively drive the postsynaptic target. In this fashion, the net enhanced synaptic input onto the postsynaptic neuron is dynamically expanded within a constrained time course dictated by the duration of the composite plasticity at those presynaptic sites. This general mechanism could allow for a novel form of integration in a neural network by providing a means of rapid and transient sharing of plasticity arising from temporally correlated inputs to a common postsynaptic target.
Footnotes
This work was supported by National Institutes of Health Grant R01 MH14-10183 and National Science Foundation Grant IOB-0444762 (T.J.C.). We thank Kelsey Martin and Wayne Sossin for kind gifts of Aplysia SN and CNS libraries. We thank Thomas Fischer, John Guzowski, Michael Sutton, Robert Zucker, and members of the Carew laboratory for constructive comments on an earlier version of the manuscript.
- Correspondence should be addressed to Thomas J. Carew at the above address. tcarew{at}uci.edu
