Abstract
Postsynaptic calcium (Ca2+) nonlinearities allow neuronal coincidence detection and site-specific plasticity. Whether such events exist in dendrites of interneurons and play a role in regulation of synaptic efficacy remains unknown. Here, we used a combination of whole-cell patch-clamp recordings and two-photon Ca2+ imaging to reveal Ca2+ nonlinearities associated with synaptic integration in dendrites of mouse hippocampal CA1 fast-spiking interneurons. Local stimulation of distal dendritic branches within stratum oriens/alveus elicited fast Ca2+ transients, which showed a steep sigmoidal relationship to stimulus intensity. Supralinear Ca2+ events required Ca2+ entry through AMPA receptors with a subsequent Ca2+ release from internal stores. To investigate the functional significance of supralinear Ca2+ signals, we examined activity-dependent fluctuations in transmission efficacy triggered by Ca2+ signals of different amplitudes at excitatory synapses of interneurons. Subthreshold theta-burst stimulation (TBS) produced small amplitude postsynaptic Ca2+ transients and triggered long-term potentiation. In contrast, the suprathreshold TBS, which was associated with the generation of supralinear Ca2+ events, triggered long-term depression. Blocking group I/II metabotropic glutamate receptors (mGluRs) during suprathreshold TBS resulted in a slight reduction of supralinear Ca2+ events and induction of short-term depression. In contrast, blocking internal stores and supralinear Ca2+ signals during suprathreshold TBS switched the direction of plasticity from depression back to potentiation. These data reveal a novel type of supralinear Ca2+ events at synapses lacking the GluA2 AMPA subtype of glutamate receptors and demonstrate a general mechanism by which Ca2+-permeable AMPA receptors, together with internal stores and mGluRs, control the direction of plasticity at interneuron excitatory synapses.
Introduction
Hebb's learning rule requires a tight temporal association of input and output at activated synapses. The backpropagating action potential (AP) informs the dendrites about the occurrence of neuronal output and provides the critical level of membrane depolarization necessary for the activation of voltage-dependent Ca2+ mechanisms and generation of supralinear Ca2+ signals (Magee and Johnston, 1997). However, in most inhibitory interneurons, AP backpropagation is largely attenuated with distance from the soma because of the nonuniform distribution of voltage-gated ion conductances (Goldberg et al., 2003a; Hu et al., 2010; Evstratova et al., 2011). As a result, backpropagating APs can generate significant Ca2+ elevations and, accordingly, control the induction of Hebbian forms of plasticity only at synapses located proximally. Local upregulation of AP-evoked Ca2+ transients by changes in synaptic activity can rescue the induction of Hebbian plasticity at interneuron excitatory synapses (Topolnik et al., 2009). Alternatively, the induction of associative plasticity at distal synapses may rely on local regenerative activity. Studies on pyramidal neurons revealed dendritic Ca2+ spikes initiated via activation of NMDA receptors (NMDARs) and voltage-gated Ca2+ channels (VGCCs) as a mechanism for coincidence detection at distal synapses (Regehr and Tank, 1990; Markram and Sakmann, 1994; Schiller et al., 1997; Golding et al., 2002; Losonczy and Magee, 2006; Remy and Spruston, 2007; Tsay et al., 2007). In addition, pyramidal neurons exhibit dendritic Ca2+ nonlinearities in the form of Ca2+ waves resulting from the activation of metabotropic glutamate receptors (mGluRs) and a subsequent Ca2+ release from internal stores (Wang et al., 2000; Nakamura et al., 2002; Watanabe et al., 2006). In interneurons, the mechanisms of dendritic regenerative activity remain largely unknown, although dendritic Ca2+ spikes have been reported in neocortical low threshold spiking cells and hippocampal CA1 stratum radiatum interneurons (Goldberg et al., 2004; Katona et al., 2011). In particular, it remains unclear how different patterns of synaptic activity can take part in the initiation of dendritic regenerative signals and how these events can regulate the efficacy of transmission at distal synapses.
Here, we used a combination of whole-cell patch-clamp recordings and two-photon Ca2+ imaging in the distal dendritic branches of mouse hippocampal CA1 fast-spiking (FS) cells to explore the mechanisms and functional relevance of Ca2+ nonlinearities evoked by the activation of glutamatergic synapses. In marked contrast to previously described mechanisms, synaptically evoked Ca2+ nonlinearities in FS interneurons involved primarily the activation of GluA2-lacking AMPA receptors with a subsequent Ca2+ release, were highly localized within individual dendritic branches and switched the direction of plasticity from long-term potentiation (LTP) to short-term depression (STD) or long-term depression (LTD), with the duration of depression depending on the additional recruitment of the group I/II mGluRs.
Materials and Methods
Slice preparation and patch-clamp recordings.
Hippocampal slices (300 μm) were prepared from CD1 mice of either sex (P13–P21) as described previously (Evstratova et al., 2011) in accordance with the animal welfare guidelines of the Université Laval. During experiments, slices were perfused with standard artificial CSF (ACSF) containing the following (in mm): 124 NaCl, 2.5 KCl, 1.25 NaH2PO4, 26 NaHCO3, 2 MgSO4, 2 CaCl2, and 10 glucose saturated with 95% O2 and 5% CO2, pH 7.4, at 30–33°C. In all experiments, GABAergic transmission was blocked by the GABAA and GABAB receptor antagonists gabazine (1 μm) and CGP55845 (2 μm), respectively. In some experiments, dl-AP5 (100 μm), 1-naphthyl acetyl spermine (NASPM; 3–60 μm), philanthotoxin 433 (PhTx, 1 μm), NBQX (10 μm), d-tubocurarine (dTC; 20 μm), nifedipine (10 μm), NNC 55-0396 (10 μm), Ni2+ (50 μm), CPA (30 μm), ryanodine (30 μm), SNX-482 (30 nm), or E4CPG (500 μm) were included in the ACSF and caffeine (1 mm) was puff applied (50 ms). As NASPM, PhTx, and nifedipine are use-dependent blockers (Koike et al., 1997; Tóth and McBain, 1998; Shen et al., 2000), the following procedure was applied when using these pharmacological agents: after an initial period of recording of control responses, the slice was incubated in a solution containing the blocker for 10 min. After this period, synaptic stimulation was resumed, and the current amplitude as well as the peak Ca2+ transient were blocked in a use-dependent manner during the next 20 min.
Whole-cell patch-clamp recordings were obtained from interneurons located in the CA1 stratum oriens adjacent to the stratum pyramidale, which were identified using a 40× water-immersion objective and infrared differential interference contrast (IR-DIC) microscopy. Recording pipettes (3–6 MΩ) were made from borosilicate glass capillaries (1B100F-4; World Precision Instruments). For whole-cell current-clamp recordings, the pipette solution contained the following (in mm): 130 KMeSO3, 2 MgCl2, 10 diNa-phosphocreatine, 10 HEPES, 2–4 ATP-Tris, 0.2 GTP-Tris, 0.2–0.3% biocytin, and Ca2+-insensitive (Alexa Fluor 594, 20 μm; Invitrogen) and Ca2+-sensitive [Fluo-5F, 300 μm or Oregon Green BAPTA 5N (OGB-5N), 500 μm; Invitrogen] dyes, pH 7.2–7.3, 275–290 mOsmol/L. In some experiments, spermine (100 μm) or QX-314 (2–5 mm) was included in the intracellular solution.
Two-photon Ca2+ imaging.
Dendritic Ca2+ imaging was performed using a TCS SP5 two-photon laser-scanning microscope (Leica Microsystems) based on a Ti-sapphire laser (Chameleon Ultra II; Coherent; >3 W, 140 fs pulses, 80 Hz repetition rate) tuned to 800 nm. A long-range water-immersion objective (40×; NA 0.8) was used to collect photons in the epifluorescence mode with external photomultiplier tubes. Neurons were filled with indicators via the patch electrode for 20–30 min before imaging. Red fluorescence from Alexa Fluor 594 was used to locate dendrites of interest. To measure Ca2+ signals, green and red fluorescence was collected during 500 Hz line scans along or across dendritic segments of 2–15 μm. Fluorescence changes were quantified as increases in green fluorescence from the baseline normalized to the red fluorescence: ΔG/R = (G − G0)/R or as changes in green fluorescence from the baseline ΔF/F = (F − F0)/F0.
Synaptic stimulation and plasticity induction.
To study the rules of synaptic integration in dendrites of interneurons, we took advantage of physiological synaptic stimulation using a local bipolar microelectrode. Although two-photon glutamate uncaging allows for the temporally precise activation of a defined number of synaptic inputs in pyramidal cells, it can produce undesirable effects (activation of perisynaptic and extrasynaptic receptors) in aspiny dendrites of interneurons. Local activation of distal synaptic inputs (80–200 μm from the soma) in the stratum oriens/alveus was achieved using a bipolar stimulating electrode made from the borosilicate theta-glass capillaries (BT-150–10; Sutter Instruments) and filled with ACSF containing 5–10 μm Alexa Fluor 594. The electrode was positioned 7-20 μm from the dendritic branch of interest under visual control (Figs. 1A, 7C) and connected to a constant current isolation unit (A360LA; World Precision Instruments). Dendritic Ca2+ transients and corresponding somatic responses were evoked by brief bursts of synaptic stimulation (three stimuli at 100 Hz, 0.1 ms each) of various intensities applied once every 30 s. The threshold stimulation was defined as the minimal intensity of stimulation necessary to evoke postsynaptic responses. The stimulus strength was then increased gradually with a step of 10–15% up to 200% of the threshold and both somatic and dendritic responses were recorded. In 31 of 40 cells examined, a 10–100% increase in stimulus intensity from the threshold resulted in a supralinear dendritic Ca2+ event (Fig. 2), which was associated with a generation of 2–4 APs in the soma.
In synaptic plasticity experiments, the pipette solution had the same composition as in Ca2+ imaging experiments, except for adding spermine (100 μm) and omitting the Ca2+ indicator. The plasticity induction protocol was applied within 8 min of whole-cell recording to avoid extensive dendritic washout. Similar to Ca2+ imaging experiments, the stimulating electrode was filled with ACSF containing 5 μm Alexa Fluor 594 and positioned 7–20 μm from the distal dendritic branch under visual guidance. Voltage-clamp recordings of baseline EPSCs were performed for 3 min at a subthreshold (for somatic AP initiation) level of stimulation. In one series of experiments (Fig. 8B), three theta-burst trains of stimuli (TBS: 3 pulses/100 Hz repeated eight times at 4 Hz for 2 s) of the same baseline intensity were applied at 30 s intervals to the cell recorded in current-clamp mode at –70 mV. This induction paradigm was defined as subthreshold TBS. Thereafter, voltage-clamp recordings of EPSCs were resumed for a further 20–30 min. To mimic the conditions necessary for the generation of dendritic Ca2+ nonlinearities, in the second series of experiments (Fig. 8C–E), the stimulation intensity during TBS was raised to 200% (to ensure the induction of supralinear Ca2+ events). Accordingly, such induction paradigm was defined as suprathreshold TBS. Thereafter, the stimulation intensity was reduced to the baseline level and voltage-clamp recordings of EPSCs were resumed.
To obtain the I–V relationship of EPSCs, cells were recorded in a voltage-clamp configuration using a Cs+-based internal solution containing spermine (100 μm) and QX-314 (2–5 mm) and EPSCs were evoked by single stimuli at different levels of membrane potential (from –80 to +60 mV with a step of 20 mV, liquid junction potential not corrected).
Data acquisition and analysis.
Data acquisition (filtered at 2–3 kHz, digitized at 10 kHz) was performed using Clampex 10.2 and analyzed using Clampfit 10.2 (Molecular Devices), Leica LAS, Igor Pro (WaveMetrics), and Excel (Microsoft). Series resistance was 10–25 MΩ. Pipette capacitance and series resistance (or bridge balance in current-clamp experiments) were compensated. Series resistance was verified every 5 min and readjusted if necessary. Cells with resting membrane potentials more positive than –50 mV, fluctuations in holding current, or unstable series resistance were discarded. In current-clamp experiments, interneurons were held at ∼–70 mV by injecting current (–20 to –50 pA), if necessary.
To measure ΔG/R peak amplitude, we averaged the values around the peak over a 10 ms time window. In summation experiments, both ΔG/R peak amplitude and the stimulus intensity were normalized to their threshold values (the minimal intensity necessary to evoke the minimal Ca2+ response). In the experiments illustrated in Figure 5, the Ca2+ nonlinearity threshold was determined as the minimal stimulus intensity necessary to evoke a supralinear Ca2+ event. In these experiments, subthreshold ΔG/R amplitude was determined as the ΔG/R peak amplitude associated with a subthreshold (for AP generation) somatic response. Spontaneous EPSCs (sEPSCs) were analyzed using the search events algorithm of Clampfit 10.2. Only those events that were not contaminated by ambiguous deflections at their rising or decaying phases, and decayed back to the baseline, were selected for the analysis of kinetics. All events were counted for frequency analysis. Typically, 200 events per cell were chosen. sEPSC amplitudes were measured at the peak of the waveform. The rise time was defined as the duration of 20–80% of the sEPSC peak amplitude. The decay phase of the sEPSCs was fitted with a single exponential. Summary data are shown as means ± SEM. Significance between groups was assessed using the paired t test, the Mann–Whitney test, or one-way ANOVA at p < 0.05.
Morphological and neurochemical identification of interneurons.
For anatomical reconstruction, neurons were filled with biocytin (Sigma) during whole-cell recordings. Slices with recorded cells were fixed overnight with 4% paraformaldehyde (PFA) at 4°C. To reveal biocytin, slices were permeabilized with 1.5–3% Triton X-100 and incubated overnight at 4°C with streptavidin-conjugated Alexa Fluor 546 or 488 (1:200; Jackson ImmunoResearch). Sections were mounted in Dako fluorescence medium and confocal images of biocytin-filled interneurons were obtained using a Leica TCS SP5 imaging system coupled with 543 nm He-Ne and 488 nm Argon lasers. Interneuron Z stacks were acquired using a 1 μm step. Anatomical reconstruction was performed using the Neurolucida 8.26.2 software (MBF Bioscience). The morphological identification of recorded interneurons was achieved by the analysis of their axonal arborization.
For neurochemical identification, slices with recorded neurons were fixed with 4% PFA and re-sectioned (thickness, 50 μm) using a Leica VT1000 vibratome (Leica Microsystems). Sections were permeabilized with 0.3% Triton X-100 in TBS and incubated in blocking solution containing 4% bovine serum albumin and 10% normal serum for 1 h. After this step, sections were incubated with streptavidin-conjugated Alexa Fluor 546 (1:200; Jackson ImmunoResearch) to reveal biocytin and a mouse monoclonal anti-parvalbumin antibody (P3088; Sigma) at 4°C for 24 h. The following day, slices were incubated with an anti-mouse secondary antibody (DyLight-647; Jackson ImmunoResearch) for 2–4 h, rinsed, and mounted on microscope slides. Confocal images of labeled processes were obtained using a Leica TCS SP5 imaging system and a 63× (NA 1.4) oil-immersion objective (Leica Microsystems).
Results
Postsynaptic Ca2+ mechanisms in distal dendrites of hippocampal CA1 FS interneurons
We performed whole-cell patch-clamp recordings from hippocampal CA1 FS interneurons tentatively identified based on their morphological properties in the IR-DIC image: large-diameter fusiform cell body located within the stratum oriens adjacent to the stratum pyramidale. Only cells that exhibited nonaccommodating FS firing pattern [with an average frequency of 207.7 ± 7.8 Hz (at 30–33°C) in response to a 1 s 1 nA current pulse; n = 84; Figure 1B] were included in this study. All recorded interneurons were filled with biocytin for post hoc anatomical and neurochemical identification (Fig. 1C). Among 84 FS cells recorded, 61 neurons were labeled successfully and reconstructed. Based on anatomical criteria, 25 recorded neurons were identified as basket cells (BCs; Fig. 1C, top left) and 30 neurons were identified as bistratified cells (BISs; Fig. 1C, top right). A subset of recorded cells was processed for parvalbumin (PV) immunoreactivity. As our typical Ca2+ imaging experiment lasted for at least 1 h, we were only able to detect PV expression in neuronal processes (distal dendrites and axonal branches) due to a rapid washout of cytoplasmic Ca2+ buffers from perisomatic compartments during whole-cell recordings (Müller et al., 2005). Nevertheless, seven of seven examined neurons revealed PV immunoreactivity in axonal branches (Fig. 1C, bottom). Consistent with previous findings (Sik et al., 1995; Buhl et al., 1996; Ali et al., 1998; Glickfeld and Scanziani, 2006), FS interneurons recorded here had a resting membrane potential of −63.6 ± 0.9 mV, input resistance of 99.3 ± 6.1 MΩ and a membrane time constant of 9.9 ± 0.5 ms (n = 55). Furthermore, as reported previously (Buhl et al., 1996), BCs exhibited a significantly lower input resistance and faster time constant when compared with BISs [BC: Rin = 84.4 ± 7.1 MΩ, τ = 8.4 ± 0.4 ms, n = 25; BISs: Rin = 114.2 ± 5.1 MΩ, τ = 11.5 ± 0.5 ms, n = 31; p < 0.05; Mann–Whitney test].
Because in many types of interneurons dendritic Ca2+ transients evoked by backpropagating APs (bAP-CaTs) decline with distance from the soma, we first examined the spatial profile of bAP-CaTs in CA1 FS interneurons. Cells were loaded with a medium-affinity Ca2+ indicator (Fluo-5F) and a Ca2+-insensitive morphological dye (Alexa Fluor 594) for 20–30 min (Fig. 1A). After the loading period, small trains of APs were initiated by somatic current injection (0.8–1 nA, 2 ms; 10 APs at 100 Hz) and corresponding bAP-CaTs were examined in dendritic shafts at different distances from the cell body (Fig. 1A, white lines). Consistent with previous findings in cortical FS interneurons (Goldberg et al., 2003a) and dentate gyrus BCs (Aponte et al., 2008), bAP-CaTs in CA1 FS interneurons were highly restricted within proximal dendritic sites (Fig. 1D).
To determine the sources of postsynaptic Ca2+ elevations that may operate in distal dendritic branches (80–200 μm from the soma), a theta-glass bipolar stimulating electrode was positioned at 7–20 μm from the branch of interest within the stratum oriens/alveus (Figs. 1A, 7C, 8A). Therefore, excitatory responses analyzed in this study were evoked primarily by the activation of local CA1 pyramidal cell (PC) input (Takács et al., 2012). First, to examine the presence of NMDARs at these synapses, EPSCs were evoked by single shocks (15–50 μA, 0.1 ms; Fig. 1E) in low Mg2+ (0.5 mm) ACSF. Surprisingly, in 6 of 11 interneurons, EPSCs evoked in such conditions showed a strong inward rectification [Rectification Index (RI) = 0.11 ± 0.04; Figure 1E, black traces and symbols], consistent with the major role of GluA2-lacking Ca2+-permeable AMPA receptors (CP-AMPARs) and with the absence of NMDAR component at these synapses. The remaining cells exhibited a mixed contribution from CP-AMPARs and NMDARs, as their EPSCs showed two components: one with a linear I–V relationship, which was sensitive to the NMDAR antagonist dl-AP5 (Fig. 1E, red traces and symbols), and the second, which exhibited a strong inward rectification (RI = 0.17 ± 0.02), with insensitivity to dl-AP5. Furthermore, in all cells examined, EPSCs were sensitive to NASPM (Fig. 1F), a synthetic analog of the Joro spider toxin, which is a selective inhibitor of CP-AMPARs (Koike et al., 1997). Importantly, CP-AMPAR-dominating synapses with variable contribution from NMDARs (range: 0–45%) were observed in both BCs and BISs. Moreover, when NMDARs were present, we found no significant difference in the ratio of the CP-AMPAR to NMDAR EPSC components between the two cell types (BCs: 1.7 ± 0.04, n = 4; BISs: 1.8 ± 0.06, n = 4, p > 0.05, Mann–Whitney test). Thus, for the rest of the study, the data obtained from BCs and BISs were pooled together.
To examine the relative contribution of NMDARs versus CP-AMPARs at these synapses under more physiological conditions, we studied their role in synaptically evoked Ca2+ elevations in normal ACSF (Mg2+, 2 mm). Interneurons were filled with a low-affinity Ca2+ indicator (OGB-5N) and a morphological dye Alexa Fluor 594, and postsynaptic Ca2+ transients were evoked by a brief burst of synaptic stimulation (three pulses at 100 Hz) at –70 mV (Fig. 1G). Under these conditions, local subthreshold (for somatic AP initiation) synaptic stimulation produced fast Ca2+ transients (decay time constant: 63.3 ± 11.2 ms, n = 6; Figs. 1G, 2A, 5A) that were highly localized within dendritic microdomains (6.0 ± 0.9 μm, n = 6; Figs. 1G, 7B), consistent with previous findings in aspiny dendrites of cortical FS interneurons (Goldberg et al., 2003b). Furthermore, the application of the NMDAR antagonist dl-AP5 slightly reduced the postsynaptic Ca2+ transients at some synapses and had no effect at others (Fig. 1G), which is in line with a variable contribution of NMDARs at these synapses. When put together, these results revealed a small but significant decrease of postsynaptic Ca2+ transients in the presence of dl-AP5 (to 77.1 ± 7.1% of control; n = 6; Fig. 1G), indicating that NMDARs make a minor contribution to postsynaptic Ca2+ influx at local CA1 synapses onto FS interneurons under physiological conditions. In contrast, postsynaptic Ca2+ transients exhibited a greater sensitivity to the CP-AMPAR inhibitor NASPM (decrease to 21.6 ± 9.9% of control; n = 7; Fig. 1G), which was also successful in blocking the residual DL-AP-5-resistant component of subthreshold postsynaptic Ca2+ transients (n = 6; Fig. 1G). Together, our data indicate that it is CP-AMPARs, not NMDARs, that provide the major contribution to postsynaptic Ca2+ signals at CA1 local collateral synapses on FS interneurons.
Dendritic Ca2+ nonlinearities in FS interneurons
Previous studies in CA1 pyramidal cells showed that burst stimulation of distal synapses elicits local dendritic Ca2+ spikes (Golding et al., 2002; Tsay et al., 2007). To determine whether any type of local regenerative activity can be initiated synaptically in dendrites of FS cells, we examined the summation of postsynaptic Ca2+ transients during synaptic stimulation of increasing intensity using a low-affinity Ca2+ dye (OGB-5N; Fig. 2). Similar to findings in pyramidal cells (Golding et al., 2002; Tsay et al., 2007), distal Ca2+ signals in interneurons exhibited a steep sigmoidal relationship to stimulus intensity (Fig. 2A,B), consistent with the generation of nonlinear regenerative responses. At low stimulus intensities, subthreshold for somatic AP generation, postsynaptic Ca2+ transients had a small amplitude (ΔG/R peak amplitude: 0.06 ± 0.004; Fig. 2A,B1). However, increasing the stimulus intensity by 10–100% was associated with a generation of supralinear Ca2+ signals that were almost five times higher than subthreshold postsynaptic Ca2+ transients (ΔG/R peak amplitude: 0.26 ± 0.02, n = 28; Fig. 2A,B2). Generation of supralinear Ca2+ signals was always associated with somatic spiking (2–4 APs; Fig. 2A,C,D). However, both the slope of somatic voltage responses (Fig. 2E) and the number of somatic APs (Fig. 2F) increased linearly, indicative of the gradual recruitment of an increasing number of synapses. Dendritic Ca2+ nonlinearities were induced in both BCs and BISs as well as in some other unidentified CA1 FS interneurons, which were not included in this study. Importantly, BCs and BISs exhibited a similar degree of summation of postsynaptic Ca2+ signals, pointing to a common mechanism for their initiation in different subtypes of FS interneurons.
Mechanisms of Ca2+ nonlinearities
To study the mechanisms of Ca2+ nonlinearities, we next examined their sensitivity to the NMDAR and AMPAR blockers. Consistent with a minor role of NMDARs at these synapses (Fig. 1), blocking NMDARs had only a small effect on the amplitude of Ca2+ nonlinearities (decrease to 77.3 ± 7.2% of control, n = 19, p < 0.05; t test; Fig. 3A,B). The presence of the NMDAR antagonist slightly decreased the degree of summation of Ca2+ transients but the summation remained supralinear (p < 0.05; n = 16, ANOVA; Fig. 3C,D), indicating that NMDARs contribute to supralinear Ca2+ signals but are not essential for their generation. In contrast, blocking CP-AMPARs resulted in a strong inhibition of these events (NASPM: decrease to 23.1 ± 9.6% of control, n = 10, p < 0.001, t test; PhTx: decrease to 19.5 ± 5.0% of control, n = 4, p < 0.001, t test; Fig. 3A,B). The effect of PhTx was not associated with blocking nicotinic acetylcholine receptors (nAChRs) as the nAChR antagonist dTC had no effect on supralinear Ca2+ transients (p = 0.32, t test, n = 3; Fig. 3B). Furthermore, summation of Ca2+ transients became linear in the presence of the CP-AMPAR antagonist NASPM (n = 7; Fig. 3C,D). The residual NASPM-resistant component of supralinear Ca2+ signals was blocked by the broad-spectrum AMPA/kainate receptor antagonist NBQX (Fig. 3A2,B), indicative of a partial blockade of CP-AMPARs by NASPM (Le Roux et al., 2013) or of an additional recruitment of kainate receptors (Cossart et al., 1998). In any case, this residual component's amplitude was small and was insufficient for the induction of supralinear Ca2+ signals (Fig. 3C,D). These results demonstrate that CP-AMPARs, not NMDARs, play a major role in synaptically evoked supralinear Ca2+ signals in distal dendrites of FS interneurons.
As CP-AMPARs could contribute to supralinear Ca2+ transients by providing postsynaptic depolarization necessary for the activation of VGCCs, we next examined the role of VGCCs in the generation of Ca2+ nonlinearities. Blocking L- or T-type VGCCs had no effect on the summation of postsynaptic Ca2+ transients. The L-type VGCC antagonist nifedipine (10 μm) reduced the amplitude of supralinear Ca2+ events (78.1 ± 5.1% of control, p = 0.053, n = 5, t test) but did not affect the summation of postsynaptic Ca2+ transients (Fig. 4A,B,F). No effects on the amplitude or summation of postsynaptic Ca2+ transients were found with T-type VGCC inhibitor NNC 55–0396 (10 μm; Fig. 4C,D,F) or nickel (Ni2+; 50 μm; data not shown), consistent with a postnatal downregulation of the T-type VGCC Cav3.1 in FS interneurons (Iftinca and Zamponi, 2009; Okaty et al., 2009). The use of the R-type VGCC antagonist SNX-482 (30 nm) resulted in a strong inhibition of somatic responses (n = 3; data not shown), preventing us from studying the specific role of these channels in Ca2+ nonlinearities. Furthermore, postsynaptic Ca2+ transients summated nonlinearly in the presence of the sodium channel blocker QX-314 (n = 3; Fig. 4E). These results indicate that CP-AMPARs do not act via activation of sodium and L- or T-type calcium channels but, instead, may provide a direct Ca2+ influx involved in the generation of Ca2+ nonlinearities.
To test this hypothesis, we next examined the calcium dependency of supralinear Ca2+ events (Fig. 5). We found that the threshold for initiation of supralinear Ca2+ signals was lower in cells that exhibited a higher subthreshold postsynaptic Ca2+ influx (Pearson correlation: r = −0.73, p = 0.0113; Fig. 5A–C, red traces and symbols), pointing to the Ca2+ sensitivity of supralinear Ca2+ signals. The latter was not a result of an additional contribution of NMDARs at some synapses, as higher NMDAR content had no effect on the amplitude of subthreshold Ca2+ influx (Pearson correlation: r = 0.29, p = 0.44) or on the threshold for initiation of supralinear Ca2+ signals (Pearson correlation: r = −0.38, p = 0.32; Fig. 5D,E). These data point to a specific role of CP-AMPAR-mediated Ca2+ influx in generation of Ca2+ nonlinearities. As supralinear Ca2+ signals were associated with a substantial somatic depolarization (Fig. 2), which could hamper CP-AMPAR-mediated Ca2+ influx in the presence of intracellular polyamine, we wondered whether excluding spermine from the intracellular solution had any effect on the generation of Ca2+ nonlinearities. Both the amplitude of supralinear Ca2+ events and the threshold for their initiation were not affected by omitting spermine from the intracellular solution (ΔG/R peak amplitude with spermine: 0.27 ± 0.01, n = 19; ΔG/R peak amplitude without spermine: 0.26 ± 0.03, n = 22; threshold with spermine: 45.6 ± 6.7 μA; threshold without spermine: 45.0 ± 10.9 μA, p > 0.05, Mann–Whitney test). These findings indicate that intracellular polyamines have no effect on dendritic Ca2+ nonlinearities, likely due to the insufficient membrane depolarization achieved in distal dendritic branches during initiation of supralinear Ca2+ events. In contrast, increasing the CP-AMPAR Ca2+ influx by membrane hyperpolarization during synaptic stimulation lowered the threshold for initiation of Ca2+ nonlinearities (stimulation threshold at –70 mV: 69.0 ± 4.1 μA; stimulation threshold at –90 mV: 43.3 ± 2.4 μA, n = 3). Collectively, these results indicate that synaptically generated supralinear Ca2+ events rely on Ca2+ influx via CP-AMPARs.
Previous evidence indicates that Ca2+-induced Ca2+ release (CICR) from intracellular stores can amplify the postsynaptic Ca2+ signal (Wang et al., 2000; Nakamura et al., 2002; Watanabe et al., 2006). To study whether this mechanism can operate in distal dendrites of FS interneurons, we examined the sensitivity of Ca2+ nonlinearities to pharmacological manipulations of Ca2+ release (Fig. 6). Our data showed that the summation of postsynaptic Ca2+ transients became linear in the presence of CPA, a specific blocker of smooth endoplasmic reticulum Ca2+/ATPases (SERCA; Fig. 6A,E), or ryanodine (Fig. 6B,E). Moreover, priming the intracellular stores by local puff application of caffeine (1 mm, 50 ms), the ryanodine receptor agonist, decreased the threshold for the initiation of supralinear Ca2+ transients (p < 0.05, n = 4, t test; Fig. 6C) without affecting their amplitude (p > 0.05, n = 4, t test; Fig. 6E). These data point to a major role of CICR in the generation of Ca2+ nonlinearities. As both NMDARs and L-type VGCCs made a significant contribution to supralinear Ca2+ signals (Figs. 3B, 4F) and together could provide Ca2+ influx necessary to induce Ca2+ release, we examined next whether blocking these two Ca2+ sources simultaneously may have a significant impact on the generation of Ca2+ nonlinearities. Our data showed that, compared with control recordings (Fig. 2B), the degree of summation of Ca2+ transients was reduced but the summation remained supralinear (n = 4, p < 0.05, ANOVA; Fig. 6D), suggesting that CICR initiated by the specific activation of CP-AMPARs is sufficient for the generation of Ca2+ nonlinearities. We did not observe the participation of group I/II mGluRs in supralinear Ca2+ signals induced by single bursts of synaptic stimulation because these events were completely blocked by a combination of dl-AP5, NASPM, and NBQX (n = 3; Fig. 3A2,B) and were not sensitive to the group I/II mGluR antagonist E4CPG (n = 5, p > 0.05, t test; Fig. 3B). Thus, intracellular Ca2+ release initiated by Ca2+ influx through CP-AMPARs is a primary mechanism of supralinear Ca2+ signals generated in FS interneurons.
Spatial profile of dendritic Ca2+ nonlinearities
The spatial extent of Ca2+ release events can vary from spatially restricted Ca2+ “sparks” to regenerative Ca2+ waves (Berridge, 1998; Callamaras et al., 1998; Watanabe et al., 2006; Manita and Ross, 2009; Miyazaki and Ross, 2013). To study the spatial profile of Ca2+ nonlinearities in dendritic branches of FS interneurons, we first performed fast line scan imaging along dendritic segments of 10–15 μm-length. Upon initiation, supralinear Ca2+ signals were restricted within small dendritic microdomains (6.0 ± 0.9 μm, n = 6; Fig. 7A) and remained localized within individual dendritic branches when a supralinear Ca2+ event was generated (11.2 ± 2.3 μm, n = 6; Fig. 7A,B). The SERCA pump inhibitor CPA removed the supralinear component of Ca2+ events but did not affect significantly the spatial extent of Ca2+ events (7.8 ± 0.8 μm, n = 6, p > 0.05, t test; Fig. 7A,B), consistent with a major role of Ca2+ release in local Ca2+ nonlinearities. To further examine the possible propagation of supralinear Ca2+ events, we performed line scan imaging across dendritic segments at different locations from the stimulation site (Fig. 7C, left). The focal plane was adjusted according to the z-position of the dendrite. The summation of postsynaptic Ca2+ transients was examined in each location by raising the stimulus intensity gradually up to 200% from the initial stimulus strength necessary to evoke the postsynaptic Ca2+ signal. Under these recording conditions, supralinear Ca2+ signals were also seen within individual dendritic branches (13.5 ± 2.1 μm; Fig. 7C, right). In addition, we found that supralinear Ca2+ events increased their amplitude with distance from the soma (Pearson correlation: r = 0.62, p = 0.0003; Fig. 7D) and were significantly higher in secondary and tertiary dendritic branches because of the smaller dendritic diameter (Pearson correlation: r = −0.53, p = 0.0024; Fig. 7E). This shows that in FS interneurons, supralinear Ca2+ events occur primarily within distal secondary and tertiary dendritic branches and remain localized.
Functional role of dendritic Ca2+ nonlinearities
At excitatory synapses of pyramidal neurons, supralinear Ca2+ events have been consistently associated with induction of LTP (Magee and Johnston, 1997; Koester and Sakmann, 1998; Golding et al., 2002; Dan and Poo, 2004). To test whether Ca2+ nonlinearities can regulate synaptic efficacy at excitatory synapses of FS interneurons, we examined how their generation at theta frequency, a natural pattern of hippocampal activity, affects the efficacy of transmission at distal excitatory synapses (Fig. 8). Interneurons were voltage clamped at –70 mV and EPSCs were evoked by local bipolar stimulation of distal inputs in the stratum oriens/alveus (100–250 μm from the soma; Fig. 8A). After a baseline recording, TBS (three trains at 4 Hz for 2 s with 30 s intervals) was applied to interneurons held in current-clamp mode, after which the voltage-clamp recordings of EPSCs were resumed. Consistent with previous findings (Lamsa et al., 2007; Oren et al., 2009; Nissen et al., 2010; Griguoli et al., 2013), low-intensity (subthreshold for somatic AP generation) TBS at rest (Vm = –70 mV) applied within 8 min of whole-cell recording produced LTP (EPSC amplitude, 264.3 ± 26.3% of control; Fig. 8B2; n = 7, p < 0.05, t test). Postsynaptic Ca2+ transients evoked by the same stimulation paradigm in a separate series of experiments had a small amplitude (ΔG/R peak amplitude: 0.07 ± 0.01; Fig. 8B1,F), consistent with the absence of supralinear Ca2+ events under basal conditions. Surprisingly, increasing the stimulation intensity during TBS (100% from the baseline level) to trigger dendritic Ca2+ nonlinearities (Fig. 8C1,F) induced LTD at these synapses (EPSC amplitude: 70.2 ± 13.2% of control; Fig. 8C2; n = 7, p < 0.05, t test). Ca2+ signals evoked by such suprathreshold TBS in a separate series of experiments had a large amplitude (ΔG/R peak amplitude: 0.28 ± 0.02, n = 5), consistent with the induction of supralinear Ca2+ events. Blocking internal Ca2+ release by CPA during suprathreshold TBS resulted in a significant decrease of supralinear Ca2+ events and in a switch in plasticity direction from LTD back to LTP (EPSC amplitude: 141.4 ± 4.6% of control; Fig. 8D,F; n = 5, p < 0.05, t test). To test whether the CPA application affects presynaptic Ca2+ release (Simkus and Stricker, 2002; Martín and Buño, 2003), which could contribute to plastic changes, we analyzed the properties of sEPSCs and evoked EPSCs (eEPSC) as well as the area under curve of TBS-evoked somatic responses in control and in the presence of CPA. Our data showed that the amplitude, frequency, and kinetics of sEPSCs were not affected by CPA application (amplitude control: 18.8 ± 0.4 pA; amplitude CPA: 17.7 ± 0.3 pA; frequency control: 5.1 ± 0.8 Hz; frequency CPA: 6.8 ± 1.7 Hz; rise control: 0.36 ± 0.01 ms; rise CPA: 0.35 ± 0.03 ms; decay control: 2.33 ± 0.09 ms; decay CPA: 2.25 ± 0.06 ms; n = 4, p < 0.05, Mann–Whitney test). Moreover, CPA had no effect on the paired-pulse ratio (PPR) of eEPSCs (PPR control: 1.85 ± 0.12, n = 7; PPR CPA: 1.86 ± 0.19, n = 5; p < 0.05, Mann–Whitney test). Furthermore, somatic voltage responses evoked by suprathreshold TBS were also not affected by CPA application (area under curve control: 1070 ± 236 mV × ms; area under curve CPA: 1315 ± 171 mV × ms, n = 5, p < 0.05, Mann–Whitney test). Together, these data indicate that, similar to other types of excitatory synapses (Carter et al., 2002), presynaptic Ca2+ stores do not contribute to transmitter release at CA1 synapses to FS interneurons.
As suprathreshold TBS could activate postsynaptic group I mGluRs, which also contribute to internal Ca2+ release, we examined the sensitivity of supralinear Ca2+ signals to the group I/II mGluR antagonist E4CPG (Fig. 8E1,F). Whereas E4CGP had no effect on the peak amplitude of supralinear Ca2+ events evoked by single bursts (Figs. 3B, 8E1; first burst in the train), it decreased significantly the area under the curve of the TBS-induced supralinear Ca2+ elevation (Fig. 8F; n = 5, p < 0.05, t test), indicating that mGluR-mediated postsynaptic Ca2+ release contributes to this signal. Accordingly, blocking group I/II mGluRs during suprathreshold TBS resulted in STD (EPSC amplitude 62.4 ± 13% of control; n = 4, p < 0.05, t test) with a complete recovery of EPSC amplitude within 15 min after induction (Fig. 8E2). Hence, supralinear Ca2+ events resulting from the Ca2+-induced Ca2+ release following the activation of CP-AMPARs switch the direction of synaptic plasticity from LTP to STD. Furthermore, additional recruitment of mGluRs, which takes place during repetitive theta-like activity, converts STD to a longer lasting depression.
Discussion
Here, we investigated the mechanisms and functional significance of supralinear Ca2+ signals generated by synchronous activation of excitatory synapses in distal dendritic branches of hippocampal CA1 FS interneurons. Our major findings are that supralinear Ca2+ signals (1) require Ca2+ influx through CP-AMPARs with a subsequent Ca2+ release; (2) remain localized within individual dendritic branches; and (3) switch the direction of synaptic plasticity from LTP to STD or LTD, depending on the additional recruitment of group I/II mGluRs. We conclude that, by linking together different signaling cascades, intracellular Ca2+ release in dendrites of FS interneurons provides a reliable and dynamically regulated source of Ca2+ that may signal coincident activation of synaptic inputs and, thus, represents an important mechanism in the regulation of synaptic strength.
Mechanisms of Ca2+ nonlinearities in FS interneurons
Our results provide new insights into mechanisms of regenerative activity in distal dendrites of interneurons. Consistent with previous findings in neocortical FS cells and hippocampal CA1 PV-positive interneurons (Goldberg et al., 2003b; Le Roux et al., 2013), we found that the GluA2-lacking CP-AMPARs provide a primary source of postsynaptic Ca2+ influx. CP-AMPARs were solely responsible for postsynaptic Ca2+ influx in 60% of synapses examined and mediated up to 80% of postsynaptic Ca2+ influx at rest. Accordingly, CP-AMPARs played a major role in the induction of Ca2+ release resulting in the generation of dendritic supralinear Ca2+ events. NMDARs and L-type VGCCs made a minor contribution to Ca2+ nonlinearities but were not sufficient for their generation. Therefore, unlike the mechanisms described previously (Schiller et al., 1997; Wang et al., 2000; Wei et al., 2001; Golding et al., 2002; Losonczy and Magee, 2006; Watanabe et al., 2006; Tsay et al., 2007), supralinear Ca2+ signals in FS interneurons do not rely on the activation of voltage-dependent Ca2+ sources but require instead Ca2+ mechanisms activated mainly at rest. This mechanism may predominate in interneurons because of the highly variable distributions of voltage-gated Ca2+ sources (e.g., NMDARs; Lei and McBain, 2002; Goldberg et al., 2003b; Nyíri et al., 2003; Le Roux et al., 2013). We did not find significant differences in the NMDAR contribution to synaptic responses between BCs and BISs, pointing to the lack of cell type-specific differences in NMDAR distribution within a population of FS interneurons. However, other excitatory projections that terminate within CA1 O/A could be recruited in our experiments (Somogyi and Klausberger, 2005; Takács et al., 2012), pointing to the synapse-specific variations in NMDAR distribution in FS interneurons. This question will require further examination using selective optogenetic targeting of specific excitatory projections. Alternatively, limited functional availability of voltage-dependent Ca2+ mechanisms in distal dendritic branches resulting from the failure of AP backpropagation and insufficient membrane depolarization can explain the minor role of these mechanisms in supralinear Ca2+ signals (Goldberg et al., 2003a; Aponte et al., 2008; Hu et al., 2010). Under these conditions, AMPAR-store coupling allows the coincidence detection of activated synapses and the site-specific regulation of synaptic strength depending on a number of synapses activated synchronously. Given the size of unitary excitatory responses at CA1 PC to FS interneuron synapses (∼2 mV; Lacaille et al., 1987; Ali et al., 1998), the size of EPSPs evoked during supralinear Ca2+ signals (∼10–15 mV) and the heavy innervation of FS cells (Gulyás et al., 1999), we estimate that at least ∼5–7 synapses are required to generate a nonlinear Ca2+ event in dendritic compartment of ∼5 μm (Fig. 7A,B). It should be pointed out that this measure is an approximation since the amplitude measurements during supralinear events were contaminated to some extent by sodium spikes riding on top of EPSPs. Further increase in the stimulation intensity was associated with recruitment of additional synapses and/or local spread of Ca2+ signals (Fig. 7C). However, no conclusion can be made regarding the actual number of synapses that may experience supralinear Ca2+ elevations under physiological conditions as the spatial extent of Ca2+ signals was likely overestimated due to the high mobility of synthetic Ca2+ indicators (Goldberg et al., 2003c).
Our data showed that ryanodine receptors (RyRs) mediate release of Ca2+ from internal stores in FS interneurons. RyR1 and, to a lesser degree, RyR2 are expressed in the CA1 hippocampal region (Hertle and Yeckel, 2007) and trigger fast and spatially localized Ca2+ “sparks” in dendrites of PCs (Manita and Ross, 2009; Miyazaki et al., 2012; Miyazaki and Ross, 2013), which appear similar to Ca2+ release events detected here. It should be pointed out that spontaneous Ca2+ release events have not been observed in our study, and their existence and degree of similarity with Ca2+ sparks described in cardiac myocytes (Cheng and Lederer, 2008) and hippocampal PCs (Manita and Ross, 2009; Miyazaki and Ross, 2013) need to be examined using a higher affinity Ca2+ indicator. Furthermore, we found that group I mGluR-mediated Ca2+ release makes an additional contribution to supralinear Ca2+ signals during repetitive suprathreshold TBS but not during single burst. The mGluR-mediated Ca2+ release likely involved the activation of mGluR5 (van Hooft et al., 2000) and RyRs (Topolnik et al., 2006, 2009), as PV-positive interneurons in this area do not express IP3 receptor 1 (Hertle and Yeckel, 2007) but exhibit RyR-mediated Ca2+ release following the activation of Gq-PLC signaling cascade (Lee et al., 2011). The absence of IP3 receptors, together with fast Ca2+ extrusion in PV-positive interneurons (Goldberg et al., 2003c), may explain why traveling Ca2+ waves were not observed in our study even following the group I mGluR activation during repetitive synaptic activity (Watanabe et al., 2006; Miyazaki and Ross, 2013).
Functional role of dendritic Ca2+ nonlinearities
Postsynaptic Ca2+ influx via CP-AMPARs has been associated with the induction of distinct forms of synaptic plasticity at excitatory synapses of hippocampal interneurons (Laezza et al., 1999; Lei and McBain, 2002; Lamsa et al., 2005, 2007; Oren et al., 2009; Sambandan et al., 2010). Activation of CP-AMPARs together with group I mGluRs or α7 nAChRs at a relatively hyperpolarized level of the postsynaptic membrane potential can induce the so-called “anti-Hebbian” LTP (Lamsa et al., 2007; Oren et al., 2009; Nissen et al., 2010; Le Duigou and Kullmann, 2011; Szabo et al., 2012; Griguoli et al., 2013; Le Roux et al., 2013). In addition, short- and long-term synaptic depression was observed at GluA2 AMPAR-lacking synapses following the activation of group I mGluRs (Le Duigou et al., 2011) or tetanic stimulation of excitatory inputs (Laezza et al., 1999; Lei and McBain, 2002). Together, these findings indicate that GluA2 AMPAR-lacking synapses are capable of regulating the efficacy of transmission in a bidirectional manner. It should be pointed out, however, that, as different hippocampal regions, cell and synapse types, as well as induction paradigms have been studied by different groups, the mechanisms underlying multiple forms of plasticity at GluA2 AMPAR-lacking interneuron synapses are still not clear. As postsynaptic Ca2+ signal is a cornerstone in plasticity induction, we focused on examining how activity-dependent Ca2+ fluctuations may control the induction of plasticity at CP-AMPAR-dominated synapses of interneurons. Our data highlights the coexistence of multiple forms of plasticity in the same synapse by providing direct evidence for the activity-dependent recruitment of additional Ca2+ mechanisms downstream of the CP-AMPAR Ca2+ influx. We demonstrate that, whereas small amplitude Ca2+ transients associated with subthreshold theta-like activity induce anti-Hebbian LTP, supralinear Ca2+ signals generated during suprathreshold theta-burst activity by the intracellular Ca2+ release induce LTD. The LTD induction required additional recruitment of group I/II mGluRs, as blocking these receptors converted LTD to STD. Thus, supralinear Ca2+ events generated by Ca2+ release following CP-AMPAR activation were responsible for STD. The expression site of different forms of synaptic plasticity observed here and the underlying signaling mechanisms remain to be determined. Previous findings showed that anti-Hebbian LTP is expressed presynaptically and, likely, involves the activation of a yet unknown retrograde messenger (Lamsa et al., 2007). Presynaptic expression of the group I mGluR-dependent LTD was also reported. Importantly, postsynaptic group I mGluR activation and endocannabinoid release mediated LTD in CA1 FS interneurons (Péterfi et al., 2012). When compared with PCs, a higher stimulation frequency was required for this LTD induction due to the lower level of diacylglycerol lipase-α in interneurons. As application of the mGluR antagonist decreased significantly TBS-evoked postsynaptic Ca2+ signals in our experiments, we assume that postsynaptic group I mGluRs were involved in suprathreshold LTD. Thus, it will be plausible to examine the role of these receptors and of endocannabinoids in the STD and LTD observed here.
In conclusion, our data shed light on the Ca2+ mechanisms of plasticity induction at GluA2 AMPAR-lacking synapses and revealed a novel metaplastic mechanism by which synaptic weight can be transiently adjusted depending on the number of synapses activated synchronously. Moreover, as excitatory synapses in many types of neurons throughout the CNS may incorporate GluA2-lacking AMPA receptors earlier during development or following synaptic inactivity (Kumar et al., 2002; Thiagarajan et al., 2005), this mechanism may be important for the detection of coincident synaptic inputs in the absence of voltage-dependent Ca2+ sources and may provide additional ways for associative learning.
Notes
Supplemental material for this article is available at www.neuronimaging.ca. This material has not been peer reviewed.
Footnotes
This work was supported by the Canadian Institute of Health Research, the Natural Sciences and Engineering Research Council of Canada (NSERC), and the Savoy Foundation. O.C. was supported by the NSERC PhD fellowship. L.T. is the recipient of the NSERC Faculty Award for Women. We thank Dimitry Kullmann and Marco Capogna for stimulating discussions and comments on the previous version of this manuscript, and Dimitry Topolnik for excellent technical assistance.
- Correspondence should be addressed to Lisa Topolnik, Axis of Cellular and Molecular Neuroscience, 2601 Ch. De La Canardière, CRULRG, Québec, PQ, G1J 2G3, Canada. Lisa.Topolnik{at}crulrg.ulaval.ca