Abstract
The synapses formed by the mossy fiber (MF) axons of hippocampal dentate gyrus granule neurons onto CA3 pyramidal neurons exhibit an intriguing form of experience-dependent synaptic plasticity that is induced and expressed presynaptically. In contrast to most other CNS synapses, long-term potentiation (LTP) at the MF–CA3 synapse is readily induced even during blockade of postsynaptic glutamate receptors. Furthermore, blocking voltage-gated Ca2+ channels prevents MF-LTP, supporting an involvement of presynaptic Ca2+ signaling via voltage-gated Ca2+ channels in MF-LTP induction. We examined the contribution of activity in both dentate granule cell somata and MF terminals to MF-LTP. We found that the induction of stable MF-LTP requires tetanization-induced action potentials not only at MF boutons, but also at dentate granule cell somata. Similarly, blocking Ca2+ influx via voltage-gated Ca2+ channels only at the granule cell soma was sufficient to disrupt MF-LTP. Finally, blocking protein synthesis or blocking fast axonal transport mechanisms via disruption of axonal tubulin filaments resulted in decremental MF-LTP.
Collectively, these data suggest that—in addition to Ca2+ influx at the MF terminals—induction of MF synaptic plasticity requires action potential-dependent Ca2+ signaling at granule cell somata, protein synthesis, and fast axonal transport along MFs. A parsimonious interpretation of these results is that somatic activity triggers protein synthesis at the soma; newly synthesized proteins are then transported to MF terminals, where they contribute to the stabilization of MF-LTP. Finally, the present data imply that synaptic plasticity at the MF–CA3 synapse can be affected by local modulation of somatic and presynaptic Ca2+ channel activity.
Introduction
Plasticity at CNS synapses is thought to play a crucial role in the encoding and storage of information. While most synapses have the capacity to exhibit long-term increases (e.g., long-term potentiation; LTP) or decreases (e.g., long-term depression; LTD) in synaptic strength, the mechanisms underlying the induction and expression of these forms of plasticity are diverse. At many synapses, LTP induction requires the postsynaptic activation of NMDA-type glutamate receptors or Ca2+ channels during tetanic stimulation (Madison et al., 1991). The ensuing postsynaptic Ca2+ influx triggers various Ca2+-dependent signaling cascades that lead to the subsequent expression of LTP (Nicoll, 2003).
In other forms of LTP, induction and expression do not necessarily occur postsynaptically. For example, the induction of LTP at the mossy fiber (MF) synapse (MF-LTP) of the hippocampus is unaffected by the pharmacological blockade of postsynaptic NMDA and/or AMPA receptors [e.g., see Castillo et al. (1994), Nicoll and Malenka (1995), and Mellor and Nicoll (2001); for a review, see Nicoll and Schmitz (2005)]. In addition, MF-LTP persists even when postsynaptic Ca2+ signaling is interrupted by the introduction of high-affinity Ca2+ buffers into the postsynaptic CA3 neurons (Mellor and Nicoll, 2001) (but see Yeckel et al., 1999). MF-LTP is, however, disrupted or modified by the pharmacological blockade or the genetic deletion of the voltage-gated Ca2+ channel subtypes that are present in MF terminals (Castillo et al., 1994; Dietrich et al., 2003). Furthermore, MF-LTP is significantly reduced in mutant mice strains with deletions or modifications of components of the presynaptic neurotransmitter release machinery (Castillo et al., 1997; Castillo et al., 2002).
In general, it seems that postsynaptic signaling mechanisms are not important for the induction of MF-LTP, whereas presynaptic Ca2+ channels are. However, the aforementioned pharmacological and genetic manipulations of the MF presynaptic Ca2+ channels would have caused a general reduction of Ca2+ influx in the presynaptic cell, not just at the presynaptic terminals. Thus, it is unclear whether it is the Ca2+ increase in MF boutons that is crucial for LTP induction, or whether Ca2+ increases in other subcellular compartments of the presynaptic neuron are critical. We designed a series of experiments to selectively investigate the roles of dentate granule cell somata versus MF boutons in the induction and expression of MF-LTP. We found that, in addition to action potential firing at MF terminals, activity and concomitant Ca2+ influx at granule cell somata also play a critical role in MF-LTP induction. Furthermore, our data suggest that fast axonal transport links somatic induction mechanisms to the plasticity occurring at MF terminals.
Materials and Methods
Subjects.
Experiments were performed on male C57BL/6 mice (Charles River) that were 29–55 d old at the beginning of each experiment. Mice were group housed in wire-lidded plastic cages before each experiment and had continuous access to food and water under a 12:12 h light–dark cycle. All experimental procedures were approved by the University of Bonn Animal Care and Use Committee.
Hippocampal slice preparation and maintenance.
Hippocampal slice preparations were performed during the light phase of the light–dark cycle. Each mouse was perfused through the left ventricle of their heart with 5–15 ml of ice-cold carbogen (i.e., 95% O2, 5% CO2)-bubbled dissection fluid (40 mm NaCl, 2.5 mm KCl, 25 mm NaHCO3, 1.25 mm NaH2PH4, 0.5 mm CaCl2, 7 mm MgCl2, 25 mm d-glucose, and 150 mm sucrose; pH 7.4), under combined ketamine (Upjohn: 100 mg/kg, i.p.) and xylazine (Bayer: 80 mg/kg, i.p.) anesthesia. Next, the mouse was decapitated, its brain was removed and immediately immersed in ice-cold carbogen-bubbled dissection fluid. Finally, transverse 400 μm hippocampal slices were cut from the brain using a vibratome (either a VT 1000S, Leica, or a Microm HM650V) and directly transferred into a holding chamber that contained carbogen-bubbled room-temperature artificial CSF (ACSF; 125 mm NaCl, 3 mm KCl, 26 mm NaHCO3, 1.25 mm NaH2PH4, 2.5 mm CaCl2, 1.3 mm MgCl2, and 25 mm d-glucose; pH 7.4).
Stimulation and field EPSP recording.
Not less than 1 h after the preparation of the hippocampal slices, a single hippocampal slice was transferred into the recording chamber. Once in the recording chamber, the slices were continuously superfused (2–2.5 ml/min) with carbogen-bubbled ACSF. The temperature in the recording chamber was between 31° and 33°C. MFs were stimulated via a monopolar glass stimulation electrode filled with ACSF (∼3 MΩ) placed in the stratum lucidum of the CA3 region of the hippocampus. In some experiments, stimulation of MFs was performed by placing a bipolar stimulation electrode within the granule cell layer. Paired-pulse stimulations (40 ms apart, every 30 s) were delivered with a constant-current stimulus isolator (WPI A320; 0.1–0.3 mA, 100 μs). MF field EPSPs (MF-EPSPs) were recorded with an ACSF-filled borosilicate glass pipette (3 MΩ) placed within the CA3 stratum lucidum. Three criteria were used to ensure the MF origin of the recorded EPSPs. First, the EPSPs had to exhibit a paired-pulse facilitation of at least 250%. Second, the EPSPs had to display facilitation in response to 1 Hz stimulation of at least 450%. Third, bath application of (2S,2′R,3′R)-2-(2′,3′-dicarboxycyclopropyl) glycine (DCG-IV, 3 μm), which was performed at the end of all MF recordings, had to block the putative MF-EPSPs by at least 80%. A 25 μm concentration of d-AP5 was present in the bath ACSF for some of the experiments (i.e., for those experiments presented in Figs. 1 and 3). PP-EPSPs were recorded by placing stimulation and recording electrodes within the molecular layer of the dentate gyrus. Antidromic dentate granule cell population spikes were recorded by placing a recording electrode within the dentate granule cell layer.
Focal drug applications were administered with a Picospritzer (Intracel, Shepreth). The Picospritzer was used to apply brief pulses of air pressure to patch pipettes that were filled with an ACSF-based drug solution. Depending on the particular experiment, that drug solution included either 10 μm TTX, 5 mm Co2+, or an ACSF with NaCl replaced by 125 mm choline chloride. In some experiments, a 0 Ca2+-containing extracellular ACSF was applied directly via the field potential recording electrode via application of mild pressure (30–40 mbar).
MF tetanic stimulation consisted of 4 trains of 50 pulses at 100 Hz that were delivered 1 s apart and were of the same intensity as the MF baseline test stimuli. Analog data were digitized and stored for offline analysis (Digidata 1200A or 1322, pCLAMP 7 or 8, Molecular Devices). fEPSP peaks were normalized to baseline values and depicted as average ± SEM. The glutamate receptor ligands 2-methyl-6-(phenylethynyl)-pyridine (MPEP), 7-(hydroxyimino)-cyclopropa-[b]-chromen-carboxylate-ethylester (CPCCOEt), and d(−)-2-amino-5-phosphonopentanoate (d-AP5) were obtained from Tocris Bioscience. All other chemicals were purchased from Sigma.
Patch-clamp recording.
For patch-clamp experiments, mice were decapitated, the brain was rapidly removed into ice-cold ACSF and transverse hippocampal slices (300 μm) were made as described above. Cells were visualized using a Nikon Eclipse FN1 upright microscope equipped with infrared difference interference contrast optics and a water-immersion lens (×60, 1.0 NA, Nikon). Somatic whole-cell current-clamp recordings were made in conjunction with a Multiclamp1B amplifier (Molecular Devices). Data were filtered at 10 kHz and sampled at 50 kHz with a Digidata 1322A interface controlled by pClamp Software (Molecular Devices). Electrode resistance in the bath ranged from 2 to 3 MΩ, and series resistance ranged from 13 to 25 MΩ. The internal solution contained [in mm]: 130 potassium gluconate, 20 KCl, 10 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid, 0.16 ethylene glycol bis(2-aminoethylether)-N,N,N′,N′-tetraacetic acid, 2 Mg-ATP, 2 Na-ATP, 15 glucose (pH adjusted to 7.4 with KOH, 295 mOsmol). The extracellular ACSF was identical to that used in the mossy fiber LTP experiments (32°C). A liquid junction potential of 14.8 mV was determined for these solutions, and corrected such that the depicted potentials are 14.8 mV more hyperpolarized than the measured potentials.
The passive membrane properties were calculated from long (≥400 ms) subthreshold hyperpolarizing and depolarizing current pulses. Input resistance (Rin) was determined as the slope of the peak membrane potential deflections elicited by injecting −50 to +20 pA current pulses. From the same voltage responses as above, the membrane time constant (τ) was determined from a single exponential fit to the charging phase. Action potential properties were determined from the first action potential elicited by just suprathreshold long duration depolarizing current pulses. Action potential threshold was the voltage at which the first derivative of the spike waveform exceeded 15 V · s−1. Action potential duration was measured at half action potential amplitude. From the same voltage responses, the fast afterhyperpolarization was determined as the lowest membrane potential reached within 5 ms following the peak of the action potential. Afterdepolarization amplitude was determined as the peak membrane potential reached >5 ms following the peak of the action potential elicited by brief (4 ms) depolarizing current pulses.
Imaging experiments.
Imaging experiments were performed with a two-photon laser scanning microscope (Ultima, Prairie Technologies). Laser polarization was controlled with a λ/2 plate (Thorlabs). Second harmonic (SH) radiation at 485 nm was excited with a pulsed Ti:Sa laser (Chameleon, Coherent) set to 970 nm through a 20× water-immersion objective (NA 0.95, Olympus). SH radiation was collected in a forward direction with a condenser (NA 0.8, Olympus) in Köhler position and filtered with a 485/20 bandpass (Semrock). The power density at the sample plane was ∼5.5 MW/cm2 for a focal spot diameter of 650 nm and a laser power of 70 mW. Images were analyzed using ImageJ and Igor Pro (Wavemetrics).
Statistical analyses.
All statistical comparisons were calculated with the R statistical package (http://www.R-project.org) or GraphPad Prism (Statcon). Two sorts of analyses were performed. First, when a comparison was between the means of two groups with equal variance (as assessed with an F test), a Student's t test was used. Second, when the comparison was between the means of two groups with unequal variance as assessed with an F test, Welch's t tests were performed.
Results
Activation of postsynaptic CA3 pyramidal cells is not necessary for MF-LTP
Does MF-LTP require the activation of postsynaptic glutamate receptors? Many (Castillo et al., 1994; Nicoll and Malenka, 1995; Mellor and Nicoll, 2001), but not all (Yeckel et al., 1999), studies have found that it does not. Because this was a critical question for the bulk of our experiments, we first performed a series of experiments to verify that MF-LTP does not require activation of postsynaptic glutamate receptors under our experimental conditions.
Under control conditions, tetanization of the MFs induced LTP of the MF-EPSPs that was 176.1 ± 5.4% of the baseline response 2 h after tetanus (n = 5) (Fig. 1A). Next, we tested whether blocking AMPA-type glutamate receptors during LTP induction would diminish the magnitude of MF-LTP: We tetanized MFs during a bath application of the AMPA/kainate-type glutamate receptor antagonist CNQX (25 μm) (Fig. 1B). The bath application of CNQX almost completely blocked MF-EPSPs, and, following a 2 h washout period, a large potentiation (Fig. 1B,E) was observed in tetanized (n = 6) (filled circles in Fig. 1B) versus untetanized control (n = 4) (empty circles in Fig. 1B) slices that had received an equivalent bath application of CNQX (t(8) = 16.9, p = 8 × 10−8). It has been suggested that the activation of postsynaptic metabotropic glutamate receptors (mGluRs) is essential for the induction of MF-LTP (Yeckel et al., 1999). We therefore performed additional experiments in which we bath applied antagonists of mGluR1 and mGluR5 (10 μm MPEP and 100 μm CPCCOEt) together with 25 μm CNQX. Combined blockade of ionotropic glutamate receptors and mGluR1/5 receptors did not eliminate MF-LTP (n = 6) (filled circles in Fig. 1C) when compared to untetanized control slices (empty circles in Fig. 1C) (n = 3, t(7) = 10.3, p = 9 × 10−6). Thus, activation of postsynaptic AMPA/kainate-type glutamate receptors and mGluR1/5 receptors is not required for MF-LTP induction under our experimental conditions. Conversely, blocking voltage-gated Ca2+ channels during MF-LTP induction with Co2+ (2 mm) resulted in a complete loss of MF-LTP (n = 5) (filled circles in Fig. 1D) compared to untetanized controls (empty circles in Fig. 1D) (n = 5, t(8) = 1.5, p = 0.09).
The data from this first set of experiments support the view that the induction of MF-LTP requires presynaptic firing and concomitant Ca2+ influx via voltage-gated Ca2+ channels (Castillo et al., 1994; Nicoll and Malenka, 1995; Mellor and Nicoll, 2001). This first set of experiments relied on bath perfusion of pharmacological antagonists. Accordingly, these data do not allow one to discriminate the role of the soma versus axon terminals of dentate granule cells.
Action potential firing at granule cell somata contributes to MF-LTP induction
In view of that deficiency, we next performed a set of experiments that assessed the effects on MF-LTP of blocking action potentials at the granule cell somata, but not the MF terminals. To achieve such a selective blockade, we focally applied TTX to both the inferior and superior blades of the dentate gyrus using focal pressure-induced (“puff”) applications from two patch pipettes (pipettes containing 10 μm TTX and labeled with a lowercase d in Fig. 2A). During the puff application of TTX, antidromic population spikes were elicited with a stimulation electrode positioned in the CA3 region and recorded from the granule cell layer (pipettes labeled with a lowercase b and e, respectively, in Fig. 2A). After focal TTX application to the dentate gyrus, population spikes were significantly reduced in amplitude (to 17.7 ± 3.6% of baseline levels; n = 5) (open bar in Fig. 2B) relative to slices that did not receive TTX (t(9) = 18.8, p = 1.0 × 10−8).The magnitude of the MF-EPSPs was unaffected by the focal TTX applications, confirming that the focal TTX applications did not reduce the Na+ currents at MF terminals (97.4 ± 2.3% of baseline levels, n = 5) (open bar in Fig. 2C). The antidromic population spikes, as well as the MF-EPSPs, were stable in slices in which all pipettes were placed as above, but puff application was omitted (n = 6) (filled bars in Fig. 2B,C). These data demonstrate that focal TTX application can block action potentials at the granule cell somata while leaving action potentials and neurotransmitter release at the MF terminals unaffected.
When focal application of TTX was omitted, we obtained an initial post-tetanic potentiation (PTP) (filled bar in Fig. 2F), with a subsequent expression of MF-LTP that was comparable to control conditions in Figure 1A (189 ± 20% of baseline levels 1 h after tetanization; n = 6) (filled circles in Fig. 2D). In contrast, in those slices in which somatic Na+ channels and somatic spiking were blocked with the focal application of TTX, the level of MF-LTP 1 h after tetanization was significantly reduced (140.0 ± 11.0%; n = 5, t(9) = 2.0, p = 0.04) (empty circles in Fig. 2D; for comparison, see Fig. 2E), even though the post-tetanic potentiation was comparable (Fig. 2F).
Action potential firing at MF terminals is necessary for MF-LTP induction
We next asked whether action potential firing at MF axon terminals is required for LTP induction. Because the effects of puff applying TTX proved difficult to wash out, we used puff applications to the stratum lucidum of an ACSF-based solution in which NaCl had been replaced with choline chloride (lowercase c in Fig. 3A). MFs were stimulated with a stimulation electrode positioned in the dentate gyrus granule cell layer (lowercase a in Fig. 3A). Puff applications of Na+-free ACSF caused a significant reversible reduction of MF-EPSPs to 28.4 ± 5.3% of baseline levels (n = 4) (open bar in Fig. 3C) (t(7) = 13.4, p = 2 × 10−6). These applications did not affect EPSPs at the perforant path–granule cell synapse (91.2 ± 9.2%; n = 6) (open bar in Fig. 3B; the positions of the recording and stimulation electrodes are indicated with a lowercase e and d, respectively, in Fig. 3A), confirming that the application of the Na+-free ACSF was restricted to the CA3 region.
Puff applications of ACSF did not alter either MF-EPSPs (104.4 ± 2.8%; n = 5) (filled bar in Fig. 3C) or PP-EPSPs (99.6 ± 1.7; n = 6) (filled bar in Fig. 3B). Under control conditions (i.e., puff application of normal ACSF), stable MF-LTP could be induced by tetanization (157.3 ± 13.3%; n = 5) (filled circles in Fig. 3D; left filled bar in Fig. 3E). In contrast, MF-LTP induction was impaired when Na+-free ACSF was puff applied during tetanization (101.3 ± 6.2%; n = 4; t(7) = 3.5, p = 0.005) (filled squares in Fig. 3D; right filled bar in Fig. 3E). Control experiments in which Na+-free ACSF was applied without tetanization revealed that the effects of the focal application were completely reversible within 1 h (115 ± 21.1%, n = 4) (open squares in Fig. 3D; open bar in Fig. 3E).
Ca2+ influx at granule cell somata contributes to MF-LTP induction
Collectively, these results indicate that action potential firing at both granule cell somata and MF terminals is essential for MF-LTP induction. We next examined whether blocking action potential-dependent Ca2+ influx only at the granule cell somata would likewise affect MF-LTP induction. We puff applied 5 mm Co2+ to the granule cell layer via two patch pipettes during tetanization (pipettes labeled with a lowercase a in Fig. 4A). Puff application of Co2+ strongly and reversibly reduced PP-EPSPs (block to 27.0 ± 6.2%, and recovery to 95 ± 8% of baseline amplitude; n = 6; t(6.2) = 11.2, p = 1.0 × 10−5) (open bar in Fig. 4B), while leaving MF-EPSPs unaffected (104.6 ± 3.2% of baseline amplitude; n = 6) (open bar in Fig. 4C). In control experiments, placing all electrodes but omitting the puff application of Co2+ did not significantly alter PP- or MF-EPSPs (99.5 ± 1.8% and 101.0 ± 2.2%, respectively, n = 6 and 6) (filled bars in Fig. 4B,C). These data indicate that the focal application of Co2+ can effectively block Ca2+ channels in the dentate gyrus without significantly affecting Ca2+ channels in the CA3 region.
Under control conditions, tetanization induced a PTP (truncated in Fig. 4D, see filled bar in Fig. 4F), with a subsequent expression of MF-LTP 1 h after the tetanization (175.4 ± 18.3%; n = 6) (filled circles in Fig. 4D; filled bars in Fig. 4E). In slices in which Co2+ was focally applied to the granule cell layer during LTP induction, PTP levels were nearly identical (Fig. 4F), but the levels of MF-LTP 1 h after tetanization were significantly reduced (130.2 ± 7.7%; n = 5; t(9) = 2.1, p = 0.03) (empty circles in Fig. 4D; for a comparison, see Fig. 4E).
Importance of Ca2+ influx at MF terminals for LTP induction
We next examined whether blocking action potential-dependent Ca2+ influx only at the MF terminals would affect MF-LTP induction. We puff applied 5 mm Co2+ to the stratum lucidum of the CA3 region (pipette labeled with a lowercase a in Fig. 5A)—close to the MF-EPSP recording electrode (pipette labeled with a lowercase c in Fig. 5A). These focal applications of Co2+ potently reduced MF-EPSP amplitude to 41.9 ± 3.2% of baseline levels (n = 12) (Fig. 5B; see also Fig. 5C). PP-EPSPs were not affected by these focal Co2+ applications (99.2 ± 4.1%; n = 4) (Fig. 5B), demonstrating that the effects of the Co2+ applications were limited to the CA3 region. The reduction of MF-EPSPs following Co2+ puff application was not reversible. However, comparing slices tetanized in the presence of focal Co2+ (n = 7) to untetanized slices subjected to focal Co2+ applications (n = 5) revealed the absence of MF-LTP following synaptic applications of Co2+ [Fig. 5C; tetanized, 67.8 ± 8.3%, n = 7 (filled bar in Fig. 5E); untetanized, 67.6 ± 11.2%, n = 5 (open bar in Fig. 5E)]. In view of the limited reversibility of the focal Co2+ application, we locally reduced mossy fiber terminal Ca2+ influx by an alternative approach. We introduced nominally Ca2+ free extracellular solution directly into the MF recording electrode and ejected this solution locally by applying limited pressure (30–40 mbar) to the recording electrode. This form of application led to a profound reduction of MF-EPSPs to <10% (see Fig. 5D) that was not observed in control experiments in which the pipette contained normal ACSF (not shown). In these experiments, similar to the focal applications of Co2+, full recovery of MF-EPSPs could not be obtained. Nevertheless, also in these experiments, comparison of tetanized to untetanized slices did not reveal any evidence for MF potentiation (tetanized, 54.6 ± 8.4, n = 6, untetanized 54.7 ± 17.3, n = 6) (Fig. 5E). These results are consistent with the idea that MF-LTP does not occur in the absence of MF terminal Ca2+ influx. It should be noted, however, that these experiments also indicate that removing Ca2+ influx from mossy fiber terminals appears to have long-term depressing effects. Caution should therefore be exercised in the interpretation of this particular set of results.
Disrupting axonal microtubules or protein synthesis impairs the ability to sustain MF-LTP
The data so far indicate that action potential firing and Ca2+ influx into the somatic compartment of granule cells are required during tetanization for the induction of stable MF-LTP. This requirement is consistent with the notion that signaling between the granule cell soma and the MF terminal is necessary for stable MF-LTP. Accordingly, we next explored the contributions of fast axonal transport to MF-LTP by incubating hippocampal slices with the microtubule-depolymerizing agent nocodazole. To confirm that nocodazole disrupted the microtubule network, we measured the optical SH signal generated by MF-axons (Dombeck et al., 2003; Kwan et al., 2008) during a bath application of 25 μm nocodazole (Fig. 6A,B) with a two-photon microscope. During the 1 h applications of nocodazole, there was a strong reduction of the SH signal. To ensure that nocodazole application did not lead to dramatic changes in the properties of hippocampal neurons, we performed patch-clamp recordings from visually identified CA3 pyramidal neurons and assessed their firing properties (see Materials and Methods). Properties of CA3 cells, shown in Table 1, were closely comparable with those previously described in patch-clamp recordings (Brown and Randall, 2009), and for most parameters revealed no differences in action potential generation between slices preincubated with nocodazole and control slices. There was, however, a significant 5 mV hyperpolarizing shift in the action potential threshold (see Table 1 and supplemental Fig. 1, available at www.jneurosci.org as supplemental material).
Having confirmed the disruption of the microtubule network with nocodazole, we next recorded MF-EPSPs in slices that were incubated for at least 1 h with 25 μm nocodazole (n = 5) and untreated control slices (n = 5) (Fig. 6C,D). In all of these experiments, 25 μm d-APV was present during the tetanization. At time points >15 min following tetanization, MF-EPSP amplitudes in the nocodazole-treated slices declined, whereas in control slices, MF-LTP was stable after 1 h of recording (Fig. 6C). The amount of LTP 1 h after tetanization was significantly reduced in the presence of nocodazole (112.1 ± 9.3%) (open bar in Fig. 6E) relative to control (149.4 ± 5.6%, t(8) = 3.43, p = 0.005) (filled bar in Fig. 6E). At the same time, paired-pulse facilitation was unaltered by nocodazole exposure (25 μm, 274 ± 24 vs 277 ± 19% facilitation in control and nocodazole incubated slices, respectively, n = 8 for both groups). Baseline MF-EPSP size was also unaltered in a set of experiments in which we washed in 10 μm nocodazole for at least 30 min (MF-EPSP amplitude after application of nocodazole 107 ± 7% of baseline MF-EPSP amplitude). These data indicate that intact axonal transport via microtubules is required for the expression of stable MF-LTP, but not for short-term synaptic plasticity.
The most straightforward explanation for the present set of results is that somatic activity triggers somatic synthesis of proteins, which are then transported to MF terminals and contribute to MF-LTP stabilization. We therefore tested whether protein synthesis is required for stable MF-LTP by preincubation of hippocampal slices with the translation inhibitor emetine (25 μm, 60 min). In emetine-treated slices, MF-LTP was not maintained (MF-EPSPs after 1 h 110.9 ± 6.7% of baseline MF-EPSP, n = 6), in contrast to interleaved control slices in which stable MF-LTP was observed (185 ± 25% of baseline MF-EPSP, n = 5, p = 0.018) (Fig. 6D,E).
Discussion
We assessed the contributions of activity within the granule cell somata and the MF terminals to MF-LTP. There were three major findings. First, we showed that action potentials at both the dentate granule cell soma and at the MF boutons are necessary for MF-LTP. Second, we found that blocking Ca2+ channels only at the dentate granule cell soma is sufficient to disrupt MF-LTP. Finally, we demonstrated that the disruption of fast axonal transport mechanisms or protein synthesis can disrupt MF-LTP. Collectively, these data suggest that—in addition to Ca2+ influx into MF terminals—the induction of MF synaptic plasticity also requires action potential-dependent Ca2+ signaling at granule cell somata and fast axonal transport along MFs.
Tetanic stimulation of the MF pathway produces robust LTP of the MF–CA3 synapse. During MF tetanization, strong activation of postsynaptic NMDA-type receptors, AMPA-type receptors, or mGluRs undoubtedly occurs. However, blocking one or a combination of these receptor types does not reduce MF-LTP (Castillo et al., 1994; Nicoll and Malenka, 1995; Mellor and Nicoll, 2001; and the present study). In addition, introduction of high concentrations of the high-affinity Ca2+ buffer BAPTA into CA3 neurons does not block MF-LTP (Mellor and Nicoll, 2001) (but see Yeckel et al., 1999). This has led to the view that MF-LTP induction is dependent mainly on Ca2+ influx into the presynaptic granule cells. The results of the present set of experiments confirm that presynaptic voltage-gated Ca2+ channels mediate the Ca2+ influx required for MF-LTP induction: Blockade of ionotropic glutamate receptors and/or mGluRs during LTP induction failed to influence MF-LTP expression, whereas blockade of voltage-gated Ca2+ channels did reduce MF-LTP.
The dependence of MF-LTP on voltage-gated Ca2+ channels has triggered interest into which channel subtypes are involved. Prior studies have elaborated the contribution of specific subtypes of Ca2+ channels to synaptic transmission and LTP at MF terminals (Castillo et al., 1994; Dietrich et al., 2003). These investigations have assumed that the Ca2+ influx relevant for MF-LTP occurs at presynaptic terminals. However, it has remained unclear whether this is indeed the case. In addition to MF terminals, action potential-dependent Ca2+ entry at granule cell somata might also be necessary for the induction of stable MF-LTP. To address this issue, we first used focal drug applications to block action potential firing or Ca2+ influx at dentate granule cell somata but not at MF terminals during MF-LTP induction. These experiments unequivocally show that spiking and Ca2+ influx at dentate granule cell somata is required for the induction of stable MF-LTP. When Ca2+ influx or spiking at granule cell somata was prevented, tetanization caused a post-tetanic potentiation that was indistinguishable from the control situation. This potentiation subsequently decayed over 20–30 min. This suggests that MF terminals contain the machinery to generate a PTP and an early phase of LTP, but require the soma for maintenance of LTP over longer time periods. This idea was also put forward in a study demonstrating that MF-LTP is decremental if MFs are transected, or if transcription or translation blockers are applied (Calixto et al., 2003). It should be noted, however, that transection approaches are hard to interpret, because axonal transection can cause prolonged high-frequency activity in neurons, coupled with strong increases in intraneuronal Ca2+ (see e.g., Mandolesi et al., 2004; Gitler and Spira, 1998; Ziv and Spira, 1997). This issue is particularly relevant for mossy fiber LTP, because it can be induced by high-frequency presynaptic activity alone.
If activity at the somatic compartment of granule cells is important for the later stages of LTP, we reasoned that Ca2+ influx restricted to the soma might be capable of LTP induction. We tested this notion by selectively blocking either action potential firing or Ca2+ influx at MF terminals without affecting their somata within the dentate gyrus. These experiments revealed that MF-LTP is absent when action potential firing at MF terminals is blocked. The results of blocking Ca2+ influx only at mossy fiber terminals are consistent with a role of local Ca2+ influx into MF terminals for MF-LTP. However, because blocking Ca2+ influx had partially irreversible effects, the interpretation of these experiments is equivocal. Nevertheless, given that we show that action potentials at mossy fibers are required for stable MF-LTP, it seems likely that the action potential associated Ca2+ influx into MF terminals is also relevant for MF-LTP. Thus, stable MF-LTP requires activity-dependent Ca2+ influx through voltage-gated Ca2+ channels at the somatic compartment and likely also at the MF terminal.
Our results suggest that somatic activity is required to trigger protein synthesis at the soma. Newly synthesized proteins are then transported to MF terminals and contribute to MF-LTP stabilization. We therefore disrupted fast, microtubule-dependent transport by applying the microtubule-depolymerizing agent nocodazole. We verified the disaggregation of axonal microtubules by monitoring the SH radiation obtained when hippocampal slices were illuminated with a pulsed infrared laser. Microtubules generate a strong SH signal due to their intrinsic optical asymmetry (Dombeck et al., 2003; Kwan et al., 2008). Disrupting microtubules, and thereby axonal transport, impaired the ability of MFs to sustain LTP after tetanization. The inability to sustain LTP was similar to the observations after somatic application of Na+ or Ca2+ channel blockers, or inhibition of protein synthesis. It should be noted that microtubule depolymerization may also have additional contaminating effects, such as the small but significant hyperpolarizing shift in action potential threshold. This effect is unlikely to explain why MF-LTP is decremental after nocodazole preincubation. In addition, baseline mossy fiber synaptic transmission and short-term plasticity were unaltered (this study), and LTP at other synapses such as the Schaffer collateral–CA1 synapse has been shown to be unaffected by microtubule disruption (Vickers and Wyllie, 2007).
What is the current evidence supporting activity-dependent synthesis of presynaptic proteins and transport to presynaptic terminals? A number of studies have demonstrated that the abundance of presynaptic proteins can be modified by activity. For instance, LTP induction in the dentate gyrus in vivo causes upregulation of syntaxin1b and synapsin I mRNA and protein in the MF pathway, and an increase of glutamate release from MF terminals (Hicks et al., 1997; Helme-Guizon et al., 1998). In addition, an increase in synapsin, synaptotagmin, and synaptophysin levels has been observed in the dentate gyrus following perforant path LTP induction in vivo (Lynch et al., 1994). These studies suggest activity-dependent synthesis and perhaps transport of presynaptic proteins. In vitro studies of synaptogenesis have shown that components of the active zone are transported along the axons (Friedman et al., 2000; Roos and Kelly, 2000). This transport is thought to be necessary both for synaptogenesis and the maintenance of functional neurotransmitter release sites. Activity dependence of axonal transport has recently been shown for transport of active zone precursor proteins by a syntaxin–syntabulin–KIF5B complex along axonal microtubules (Cai et al., 2007). Additionally, the study by Cai et al. (2007) established the importance of axonal microtubules for transport of components of the presynaptic active zone.
Our data suggest that high-frequency activation of granule neurons induces plasticity-related processes both at MF terminals and at granule cell somata, and that both sets of processes are required for long-term changes in MF synaptic transmission. Induction of somatic protein synthesis and the transport of component proteins of the active zone to synaptic terminals may then promote plastic changes in presynaptic neurotransmitter release.
Footnotes
This work was supported by the Deutsche Forschungsgemeinschaft (SFB TR3), EPICURE, ERANET Neuron “EpiNet,” Marie Curie Actions, and the BONFOR program of the University of Bonn Medical Center.
- Correspondence should be addressed to Heinz Beck, Department of Epileptology, Laboratory of Experimental Epileptology, Sigmund-Freud-Strasse 25, 53127 Bonn, Germany. heinz.beck{at}ukb.uni-bonn.de