Abstract
In cerebellar Purkinje neuron dendrites, the transient depolarization associated with a climbing fiber (CF) EPSP activates voltage-gated Ca2+ channels (VGCCs), voltage-gated K+ channels (VGKCs), and Ca2+-activated SK and BK K+ channels. The resulting membrane potential (Vm) and Ca2+ transients play a fundamental role in dendritic integration and synaptic plasticity of parallel fiber inputs. Here we report a detailed investigation of the kinetics of dendritic Ca2+ and K+ channels activated by CF-EPSPs, based on optical measurements of Vm and Ca2+ transients and on a single-compartment NEURON model reproducing experimental data. We first measured Vm and Ca2+ transients associated with CF-EPSPs at different initial Vm, and we analyzed the changes in the Ca2+ transients produced by the block of each individual VGCCs, of A-type VGKCs and of SK and BK channels. Then, we constructed a model that includes six active ion channels to accurately match experimental signals and extract the physiological kinetics of each channel. We found that two different sets of channels are selectively activated. When the dendrite is hyperpolarized, CF-EPSPs mainly activate T-type VGCCs, SK channels, and A-type VGKCs that limit the transient Vm ∼ <0 mV. In contrast, when the dendrite is depolarized, T-type VGCCs and A-type VGKCs are inactivated and CF-EPSPs activate P/Q-type VGCCs, high-voltage activated VGKCs, and BK channels, leading to Ca2+ spikes. Thus, the potentially activity-dependent regulation of A-type VGKCs, controlling the activation of this second set of channels, is likely to play a crucial role in signal integration and plasticity in Purkinje neuron dendrites.
SIGNIFICANCE STATEMENT The climbing fiber synaptic input transiently depolarizes the dendrite of cerebellar Purkinje neurons generating a signal that plays a fundamental role in dendritic integration. This signal is mediated by two types of Ca2+ channels and four types of K+ channels. Thus, understanding the kinetics of all of these channels is crucial for understanding PN function. To obtain this information, we used an innovative strategy that merges ultrafast optical membrane potential and Ca2+ measurements, pharmacological analysis, and computational modeling. We found that, according to the initial membrane potential, the climbing fiber depolarizing transient activates two distinct sets of channels. Moreover, A-type K+ channels limit the activation of P/Q-type Ca2+ channels and associated K+ channels, thus preventing the generation of Ca2+ spikes.
- calcium channels
- cerebellar Purkinje neuron
- climbing fiber
- neuron modeling
- neuronal dendrites
- potassium channels
Introduction
The climbing fiber (CF) synaptic input from the brainstem inferior olive to the cerebellar Purkinje neuron (PN) governs short-term (Brenowitz and Regehr, 2005) and long-term (Safo and Regehr, 2005) synaptic depression of parallel fiber (PF) synaptic inputs in the dendrites. While these important learning mechanisms eventually involve activation of metabotropic glutamate receptors at PF releasing sites and postsynaptic release of endocannabinoids (Marcaggi and Attwell, 2005), the CF signal is carried by a transient depolarization in the dendrite. The large CF-EPSP is generated in the cell body and in the proximal dendritic segment by glutamate release from hundreds of synaptic terminals (Silver et al., 1998). Then, the depolarization spreads in the dendritic arborization (Canepari and Vogt, 2008) where it activates voltage-gated Ca2+ channels (VGCCs) and voltage-gated K+ channels (VGKCs). The transient elevation of intracellular Ca2+ activates both SK K+ channels and BK K+ channels (Edgerton and Reinhart, 2003). Thus, the ensemble of activated Ca2+ and K+ channels, including VGKCs, shapes the waveform of the dendritic membrane potential (Vm) and of the Ca2+ transient (i.e., the two signals that the CF input transmits to the dendritic terminations) (Vogt and Canepari, 2010). A detailed characterization of the kinetics of each individual channel involved in the CF signal is therefore crucial for understanding how the CF transmits information to the dendritic arborization. To tackle this problem, a detailed compartmental model of PNs, including several ion channels, was proposed in the early 1990s for predicting the dendritic signals associated with synaptic responses (De Schutter and Bower, 1994). Since then, several computational models incorporating ion channel and morphology variations have been proposed to account for the emerging complex dendritic activity (Anwar et al., 2013, 2014). Yet, few, if any, of these predictions have been tested experimentally. Importantly, recent advancements in voltage-sensitive dye (VSD) imaging combined with Ca2+ imaging with a fast sampling rate, using low-affinity indicators to track the kinetics of VGCCs, allow the direct measurement of Vm and Ca2+ signals in the dendrites with a spatial resolution of a few microns (Jaafari et al., 2014, 2015; Jaafari and Canepari, 2016).
In this article, we report the first measurements of dendritic Vm and Ca2+ transients associated with the CF-EPSP at different initial Vm and the analysis of their correlation. Then, we report a detailed analysis of the changes in the CF-mediated dendritic Ca2+ transient produced by the local selective inhibition of various Ca2+ or K+ channels. Based on this rich experimental dataset, obtained from an ensemble of 59 cells, we built a NEURON model of a simplified PN dendritic compartment. The model is based on prior work (Anwar et al., 2012) and incorporates modified channel models to accurately reproduce the Vm and Ca2+ transients associated with the CF-EPSP at different initial Vm. The model included P/Q-type high-voltage activated (HVA) VGCCs, T-type low-voltage activated (LVA) VGCCs, A-type LVA-VGKCs, and Ca2+-activated K+ channels (both SK and BK). The consistency of this model was assessed by quantitatively comparing the experimental results of blocking different individual channels, with the elimination of respective channels from the model. The model also included a generic HVA-VGKC (HVAK), an immobile endogenous Ca2+ buffer (Canepari and Mammano, 1999), and the two Ca2+ binding proteins Calbindin-D28k and Parvalbumin highly expressed in PNs (Schmidt et al., 2003). The systematic and extensive feedback between experiments and model simulations allowed us unraveling the physiological kinetics of all dendritic Ca2+ and K+ channels activated by the CF-mediated Vm transient, under different initial Vm conditions, providing a realistic model for the ion channels involved. In particular, we found that two distinct sets of channels are activated when the initial resting Vm is either hyperpolarized or depolarized and that the separation between these two sets is governed by the activation of A-type VGKCs that is likely playing a crucial role in regulation and plasticity of PF inputs.
Materials and Methods
Slice preparation, electrophysiology, and pharmacology
Experiments were ethically performed in accordance with European Directives 2010/63/UE on the care, welfare, and treatment of animals. Procedures were reviewed by the ethics committee affiliated to the animal facility of the university (D3842110001). Cerebellar sagittal slices (250 μm thick) were prepared from 21 to 35 postnatal day old mice (C57BL6) following established procedures (Vogt et al., 2011a, b; Ait Ouares et al., 2016) with a VT1200 (Leica Microsystems) and incubated at 37°C for 45 min before use. The extracellular solution contained the following (in mM): 125 NaCl, 26 NaHCO3, 1 MgSO4, 3 KCl, 1 NaH2PO4, 2 CaCl2, and 20 glucose, bubbled with 95% O2 and 5% CO2. The intracellular solution contained the following (in mM): 125 KMeSO4, 5 KCl, 8 MgSO4, 5 Na2-ATP, 0.3 Tris-GTP, 12 Tris-phosphocreatine, 20 HEPES, adjusted to pH 7.35 with KOH. In combined Vm and Ca2+ imaging experiments, PNs were loaded with the VSD JPW1114 and with the Ca2+ indicator FuraFF (at 1 mM) using a previously described procedure (Vogt et al., 2011a). In experiments of Ca2+ imaging only, Oregon Green BAPTA-5N (OG5N) was added to the internal solution at 2 mM concentration. Patch-clamp recordings were made at 32°C-34°C using a Multiclamp amplifier 700A (Molecular Devices), and signals were acquired at 20 kHz using the A/D board of the CCD camera. The measured Vm was corrected for junction potential (−11 mV) as previously estimated (Canepari et al., 2010). CF-EPSPs were elicited by current pulses of 5–20 μA amplitude and 100 μs duration delivered by a pipette. Local block of the various channels was achieved by gentle pressure application of the extracellular solution containing the specific blocker at selected effective concentration, using a pipette of ∼2 μm diameter. Full names of chemicals used to block L-type, N-type, or T-type VGCCs were as follows: isradipine, 4-(2,1,3-benzoxadiazol-4-yl)-1,4-dihydro-2,6-dimethyl-3,5-pyridinecarboxylic acid methyl 1-methylethyl ester; PD173212, N-[[4-(1,1-dimethylethyl)phenyl]methyl-N-methyl-L-leucyl-N-(1,1-dimethylethyl)-O-phenylmethyl)-L-tyrosinamide;ML218, 3,5-dichloro-N-[[(1α,5α,6-exo,6α)-3-(3,3-dimethylbutyl)-3-azabicyclo[3.1.0]hex-6-yl]methyl]-benzamide-hydrochloride; and NNC550396,(1S,2S)-2-[2-[[3-(1H-benzimidazol-2-yl)propyl]methylamino]ethyl]-6-fluoro-1,2,3,4-tetrahydro-1-(1-methylethyl)-2-naphthalenyl-cyclopropanecarboxylate-dihydrochloride. The VSD and the Ca2+ indicators were from Invitrogen. ω-Agatoxin IVA, AmmTx3, iberiotoxin, and apamin were purchased from Smartox Biotechnology. All other chemicals were purchased from Tocris Bioscience, Hello Bio, or Sigma-Aldrich.
Optical signal recording and calibration
Sequential Vm and Ca2+ optical measurements (Canepari and Vogt, 2008; Canepari et al., 2008) were achieved by alternating excitation of FuraFF at 385 nm with an OptoLED (Cairn Research) and of the VSD at 532 nm using a 300 mW solid state laser (model MLL532, CNI). In experiments with Ca2+ imaging only, OG5N was excited at 470 nm with the OptoLED. Ca2+ fluorescence (either from FuraFF or from OG5N) and Vm fluorescence was recorded at 525 ± 25 nm and at >610 nm, respectively, using a NeuroCCD-SMQ camera (RedShirtimaging). Images, demagnified by ∼0.2× to visualize an area of ∼150 μm diameter, were acquired at 5 kHz with a resolution of 26 × 26 pixels. Electrical and optical signals associated with the CF-EPSP were recorded for 20 ms (100 frames) with the CF stimulation occurring 2 ms after the beginning of trials. Three or four trials, with 20 s between two consecutive trials, were obtained to assess the consistency of the signals. Fluorescence from these trials was averaged and corrected for bleaching using a filtered trial without signal. Vm and Ca2+ signals were initially expressed as fractional changes of fluorescence (ΔF/F0). To match experimental Ca2+ signals with NEURON simulations, ΔF/F0 was converted into Ca2+-bound-to-dye concentration ([Ca2+Dye]) as previously described (Ait Ouares et al., 2016). Briefly, [Ca2+Dye] = [Dye]TOT · ΔF/F0/σ, where [Dye]TOT is the total dye concentration and σ is the empirical dynamic range defined as the ΔF/F0 for saturating Ca2+ (Ait Ouares et al., 2016). Measurements of σ were obtained by saturating the indicators in the cytosol with a 30 s step of −500 mV (in voltage clamp) that makes the somatic membrane permeable to Ca2+. The value of σ obtained in this way for FuraFF was −0.9. VSD ΔF/F0 was calibrated into a Vm transient using an established procedure (Canepari and Vogt, 2008). Assuming that somatic hyperpolarizations from the resting Vm spread without attenuation through PN dendrites (Roth and Häusser, 2001), a prolonged hyperpolarizing step of 1 s was used in each experiment to convert the VSD ΔF/F0 signal into millivolts, as shown in the example of Figure 1A. Because, in this study, signals associated with the CF-EPSP were evoked at different initial Vm, depolarizing steps of 1 s were also used to estimate the dendritic Vm produced by somatic depolarization. In detail, as shown in the example of Figure 1A, the two largest depolarization steps lead to the same dendritic Vm, although corresponding to two different somatic Vm. This indicates that the dendrite reaches a maximal steady Vm regardless of the current injected into the soma above certain intensity. In our experiments, Vm recordings at depolarized states were performed by injecting currents above this intensity, in this way driving the initial dendritic Vm to the maximal value that was obtained by the calibration protocol.
Experimental design and statistical analysis
The effects of changing the initial Vm on the Vm and Ca2+ transients, or of blocking a Ca2+ or a K+ channel on the Ca2+ transients, were established by performing the paired Student's t test on the signal under the two different conditions. A change in the signal was considered significant when p < 0.005. The ROIs used for the statistical analysis of pharmacological tests were those adjacent to the pipette delivering the channel blocker.
Procedure of matching experimental transients with a neuron model
To deduce the kinetics of each channel involved in the dendritic electrical signal associated with the CF-EPSP, we developed an original approach based on matching optical data with a simplified model, implemented in NEURON (Hines and Carnevale, 1997) as depicted in Figure 1B. To build the model, we started from a published model designed to predict activation of Ca2+-activated K+ channels in PN dendrites (Anwar et al., 2012), available in the ModelDB database at https://senselab.med.yale.edu/ModelDB/ShowModel.cshtml?model=138382&file=/AnwarEtAl2010/cdp5.mod#tabs-1. From this model, consisting of a cylinder of 4 μm diameter and 20 μm length with standard passive membrane properties, several parameters and mechanisms were initially replaced with those from two earlier models (De Schutter and Bower, 1994; Solinas et al., 2007) also available in ModelDB database at (https://senselab.med.yale.edu/ModelDB/ShowModel.cshtml?model=7176#tabs-1) and at (https://senselab.med.yale.edu/ModelDB/ShowModel.cshtml?model=112685&file=/Golgi_cell/Golgi_SK2.mod#tabs-2). From this starting point, the kinetics of ion channels were changed to obtain traces from simulations that matched experimental Vm and Ca2+ optical measurements from ∼17 × 17 μm2 square regions at three different initial Vm: hyperpolarized (∼−80 mV), intermediate (∼−65 mV), and depolarized (∼−50 mV). The specific modifications implemented for each channel are described in detail in the paragraph below. After obtaining a satisfactory set of models for each channel in one cell, the same channel models were used to match our experimental data in three more cells, only by tuning the channel densities and the current input associated with the CF-EPSP.
Ion channel mathematical functions matching experimental data
The channel functions that matched experimental Vm and Ca2+ transients are reported below and are available in ModelDB at http://senselab.med.yale.edu/ModelDB/showModel.cshtml?model=244679. In detail, I is the current density (expressed in mA/cm2), ∏ is channel permeability to Ca2+ (expressed in cm/s), m∞, tm, mexp, and h∞, th are the voltage-dependent activation and inactivation parameters, respectively, z∞,zexp, and tz are the Ca2+-dependent activation parameters, and GHK is the Goldman-Hodgkin-Katz factor (expressing the current per unit permeability (Anwar et al., 2012). Voltage (V) and time are expressed in millivolts and milliseconds, respectively.
P/Q-type VGCC.
Starting from the formulation for P/Q-type VGCCs in Anwar et al. (2012), we multiplied the activation curve by a sigmoid function to account for the fact that we did not observe P/Q channel activation <−50 mV. We also reduced the activation time by 40% to reproduce the observed Ca2+ spiking rate at depolarized states.
T-type VGCC.
Starting from the formulation for T-type VGCCs in Anwar et al. (2012), we multiplied the activation curve by a sigmoid function to account for the observed activation at hyperpolarized and intermediate states. The activation time was decreased by 70%, whereas the inactivation time was doubled to fit the rise and decay of the experimental Vm trace at the hyperpolarized state as follows:
SK Ca2+-activated K+ channel.
We used the model in Solinas et al. (2007) with 95% of the SK channels coupled to T-type VGCCs to account for effect of blocking these channels observed exclusively at hyperpolarized states.
BK Ca2+-activated K+ channel.
Starting from the formulation in De Schutter and Bower (1994), we reduced the Ca2+-dependent activation time to half to account for the larger slow repolarization at depolarized states as follows:
A-type VGKC.
Starting from the formulation of A-type VGKCs channels in De Schutter and Bower (1994), the kinetic parameters were modified in line with modifications of T-type VGCCs to account for behaviors at hyperpolarized states. The density was corrected at intermediate states to account for partial inactivation as follows:
HVA-VGKC (HVAK).
Starting from the formulation of the “delayed rectifier channel” given in De Schutter and Bower (1994), the HVAK kinetic parameters were modified to account for the behaviors at depolarized states. Specifically, the activation curve was multiplied by a sigmoid function to track the occurrence of the first Ca2+ spike. Then the activation time was decreased by ∼95% to reproduce the number and shape of the observed Ca2+ spikes. Notably, this is the only channel for which the experimental pharmacological block was not available. See the following equations:
Model of the CF-associated current
The model of the current associated with a CF-EPSP, designed to mimic the shape of the current reported by Llano et al. (1991), was expressed by the following equation:
The parameters in this equation were tuned to obtain the match of experimental Vm and Ca2+ transients within the following ranges. IHOLD (holding current before CF-EPSP occurrence): between −0.03 and 0.04 mA/cm2. ISTANDING: between −0.4 and −0.1 mA/cm2. DELAY: between 2 and 2.4 ms. RISE: between 0.4 and 1.8 ms. DURATION: between 2.5 and 6.
Other fixed mechanisms and parameters of the neuron model
The following standard mechanism parameters in Figure 1B were used in the simulations.
Immobile buffer (concentration and Ca2+ reaction kinetic parameters) is as follows (Ait Ouares et al., 2016): concentration = 1 mM; KON = 570 μm−1 s−1; KOFF = 5.7 · 103 s−1.
Parvalbumin (concentration and Ca2+ reaction kinetic parameters for Ca2+ and for Mg2+), corrected from empirical values reported by Lee et al. (2000) to take into account the difference in temperature and radial diffusion: binding sites concentration (two per molecule) = 150 μm; KONCa2+ = 535 μm−1 s−1; KOFFCa2+ = 0.95 s−1; KONMg2+ = 4 μm−1 s−1; KOFFMg2+ = 25 s−1.
Calbindin D28-k (concentration and Ca2+ reaction kinetic parameters), corrected from empirical values reported by Nägerl et al. (2000) to take into account the difference in temperature and radial diffusion): fast binding site concentration (two per molecule) = 1.2 mM; KON = 217.5 μm−1 s−1; KOFF = 35.8 s−1; slow binding sites concentration (two per molecule) = 1.2 m; KON = 27.5 μm−1 s−1; KOFF = 2.6 s−1.
Ca2+ indicator (Ca2+ reaction kinetic parameters): KON = 570 μm−1 s−1; concentration FuraFF = 1 mM; KOFF(FuraFF) = 5.7 · 103 s−1; concentration OG5N = 2 mM; KOFF(OG5N) = 19.95 · 103 s−1.
Ca2+ extrusion equation was adopted from Destexhe et al. (1993), with kinetic parameters and density used in Anwar et al. (2012) as follows: KON = 3 · 10−3 μm−1 s−1; KOFF = 1.75 · 10−2 s−1; KEXT = 7.255 · 10−5 μm−1 s−1; density = 10−9 mol cm−2.
LEAK current: 0.002 mA/(mV · cm2) · (Vinit − Vrest), where Vinit is the initial Vm (in millivolts) and Vrest is the resting Vm (−65 mV).
Results
Dendritic depolarization and Ca2+ transients associated with the CF-EPSP
In the first series of experiments (N = 12 cells), we investigated the dendritic depolarization and the Ca2+ transients associated with the CF-EPSP by combining Vm and Ca2+ imaging as described in Materials and Methods. Within a recording field of ∼150 × 150 μm2 (26 × 26 pixels), we systematically averaged fluorescence over 1–3 dendritic regions of ∼17 × 17 μm2 (3 × 3 pixels) with sufficient signal-to-noise ratio to reliably measure ΔF/F0 signals with both indicators, at 5 kHz. In each cell and dendritic region, the Vm was determined from voltage-sensitive dye ΔF/F0 signals using the protocol described in Materials and Methods and illustrated in Figure 1A. In the cell reported in Figure 2A, the initial somatic Vm was set by current injection and a CF-EPSP was evoked at a state of hyperpolarization (hyp, blue trace) ∼−80 mV, at an intermediate Vm (int, green trace) close to the resting Vm, and at a state of depolarization (dep, red trace). These three conditions corresponded, in this particular cell, to initial dendritic Vm of −83, −64, and −54 mV, respectively. In the two analyzed regions (R1 and R2), the dendritic Vm transient associated with the CF-EPSP had a first peak occurring within 3 ms after the stimulation that ranged from −17 mV at the hyp state to 10 mV at the int state (Fig. 2B). A second sharp peak was observed at the dep state. Correlated with these Vm transients, the Ca2+ transient increased from the hyp state to the int state and exhibited two sharp peaks at the dep state that can be defined as spikes because they are characterized by a rapid rise and fall. This behavior was consistently observed in every cell investigated. The Vm and Ca2+ transients in R1 are shown again in Figure 2C to illustrate the quantitative analysis that was performed. For both dendritic Vm and Ca2+ transients, we measured the maximum (max) during the first 4 ms after the CF stimulation (first max) and between 4 and 14 ms after the CF stimulation (second max). The values for R1 are reported in Figure 2C, and the statistics (mean ± SD) for 19 regions in the 12 cells analyzed are reported in Figure 2D. At hyp states (with initial Vm between −87 and −74 mV), the first and second Vm maxima were −13 ± 8 and −38 ± 7 mV, respectively, whereas the Ca2+ transient (−ΔF/F0) max were 1.83 ± 0.55% and 3.14 ± 0.85%, respectively. At int states (with initial Vm between −68 and −61 mV), the first and second Vm maxima were 6 ± 7 and −21 ± 11 mV, respectively, whereas the Ca2+ transient max were 3.65 ± 1.27% and 3.65 ± 1.05%, respectively. Finally, at dep states (with initial Vm between −54 and −46 mV), the first and second Vm maxima were 9 ± 6 and −1 ± 9 mV, respectively, whereas the Ca2+ transients maxima were 5.36 ± 1.17% and 6.59 ± 1.42% respectively. The first max of both Vm and Ca2+ transients significantly increased from the hyp state to the int state (p < 0.005, paired t test), whereas the second max of both the Vm and Ca2+ transients significantly increased from the int state to the dep state. These results demonstrate that CF-EPSP-associated dendritic depolarization and Ca2+ influx increase with the initial Vm. Furthermore, dendritic spikes (typically two in the 18 ms after the stimulation) and correlated Ca2+ transients occur when the dendrite is depolarized. The dendritic depolarization is produced by the passive spread of the CF-EPSP from the soma and proximal dendrite and activates VGCCs and VGKCs. Ca2+-activated K+ channels contribute to the repolarization. To resolve the channels underlying the behaviors observed in the experiments reported above, we analyzed pharmacologically in detail the Ca2+ transient associated with the CF-EPSP.
Vm calibration protocol and illustration of NEURON model. A, Left, Fluorescence reconstruction of a PN with three ROIs (R1, R2, and R3). From the resting Vm (−67 mV), negative or positive current pulses of 1 s duration were delivered from the recording electrode. Right, Somatic Vm and the VSD-ΔF/F0 signals in R1-R3 associated with the current pulses. The VSD-ΔF/F0 signal in each region is converted into millivolts to quantify the Vm transient associated with the CF-EPSP, assuming that the resting Vm is uniform (int state) and that the hyperpolarizing pulse spreads into the dendrites without attenuation (hyp state). The protocol also allows determining the dendritic Vm associated with the strongest depolarizing pulse (dep state). B, A dendritic region of ∼17 × 17 μm2 is approximated with a cylinder of 4 μm diameter and 20 μm length in the NEURON model. The model contains P/Q-type and T-type Ca2+ channels; SK, BK, A-type K+ channels; and a generic HVAK. It includes four buffers: a fast immobile buffer, the Ca2+ indicator (either Fura-FF or OG5N), Parvalbumin, and Calbindin D-28k. It also includes Ca2+ extrusion and a LEAK channel.
Combined Vm and Ca2+ transients associated with the CF-EPSP. A, Bottom, Fluorescence reconstruction of a representative PN with two ROIs indicated (R1 and R2). Top, Somatic Vm associated with a CF-EPSP at three different initial Vm: hyperpolarized (hyp blue trace); intermediate (int green trace); and depolarized (dep red trace). B, Top, Dendritic Vm in R1 and R2 calibrated as illustrated in Figure 1A corresponding to the somatic CF-EPSPs in A. Bottom, Corresponding FuraFF ΔF/F0 signals. C, Analysis of the Vm and Ca2+ maxima (max) associated with signals in R1 reported in B; the first max of the Vm and Ca2+ transients is calculated within the first 4 ms after the CF stimulation; the second max of the Vm and Ca2+ transients is calculated between 4 and 14 ms after the CF stimulation. Blue traces represent the hyp state. Green traces represent the int state. Red traces represent the dep state. D, Mean ± SD for 19 regions in 12 cells analyzed as illustrated in C. The hyp states (blue columns) were with initial Vm between −87 mV and −74 mV. The int states (green columns) were with initial Vm between −68 mV and −61 mV. The dep states (red columns) were with initial Vm between −54 and −46 mV. *Significant increase in the max (p < 0.005, paired t test).
Dendritic Ca2+ channels activated by the CF-EPSP
In PNs, the dendritic Ca2+ transients associated with the CF-EPSP transient depolarization are mediated by VGCCs, in particular P/Q-type HVA-VGCCs (Usowitz et al., 1992) and T-type LVA-VGCCs (Isope et al., 2012). Thus, we investigated the changes in the Ca2+ transients from OG5N fluorescence produced by the selective block of one or more VGCCs. In the representative example of Figure 3A, the CF-EPSP was evoked at hyp (blue traces), int (green traces), and dep (red traces) states, and the associated OG5N ΔF/F0 signal was recorded in the control condition and after local application of the P/Q-type VGCC blocker ω-agatoxin-IVA (AgaIVA, 1 μm). Importantly, to assess the postsynaptic effect while excluding any possible presynaptic effect, the changes in the Ca2+ transient were analyzed and compared in the region next to the application pipette (R1) and in another region at >50 μm from the application pipette (R2). In R1 only, AgaIVA reduced the Ca2+ transient during the first few milliseconds after the CF-EPSP at the hyp and int states, and blocked the Ca2+ transient at the dep state. In the representative example of Figure 3B, the same protocol was used to assess the effects of local block of T-type VGCCs, using the inhibitors ML218 (ML, 5 μm) and NNC550396 (NNC, 30 μm). In this case, the blockers inhibited the late component of the Ca2+ transient at the hyp and int states but had no effect on the Ca2+ transient at the dep state. Finally, in the representative example of Figure 3C, local block of P/Q-type and T-type VGCCs together inhibited Ca2+ transient in all states. To rule out any additional component from other VGCC types, in the three representative examples of Figure 3D, we also tested the blockers isradipine (20 μm, L-type), PD173212 (5 μm, N-type), and SNX482 (1 μm, R-type). The Ca2+ transient remained unaffected under all states and cases tested. The results reported in Figure 3 were consistently observed in all cells tested with each VGCC blocker. To quantify these results, we measured again the first max (during the first 4 ms after the stimulation) and the second max (between 4 and 14 ms after the stimulation), and calculated the percentage of the two maxima in the presence of the VGCC blocker, with respect to the control condition, as illustrated in the examples of Figure 4. We tested the block of P/Q-type VGCCs (AgaIVA), of T-type VGCC (ML + NNC), or of P/Q-type and T-type VGCCs together (AgaIVA + ML + NNC) in N = 6 cells for each case. The block of P/Q channels significantly reduced the first max of the Ca2+ transient at all initial Vm states (p < 0.005, paired t test), whereas the second max was reduced only in the int and dep states. In contrast, the block of T channels significantly reduced the first and second maxima of the Ca2+ transient at hyp states and the second max only and int states, whereas neither max was changed at dep states. The block of P/Q and T channels together significantly reduced both maxima at all initial states. Finally, we tested the block of L-type VGCCs (Isr), of N-type VGCC (PD), or of R-type VGCCs (SNX) in N = 4 cells for each case. No changes in the two maxima were observed in any of the initial states. In summary, these results show the distinct kinetics of two components of Ca2+ influx, associated with the CF-EPSP and mediated by P/Q-type and T-type VGCCs, respectively. At hyp states, where the depolarization transient is typically <−10 mV (Fig. 2), P/Q-type VGCCs are weakly activated, but activation of these channels increases when the initial Vm becomes more positive and the transient depolarization larger. In contrast, T-type VGCCs are strongly activated at hyp states, but these channels inactivate as the initial Vm increases.
Dendritic Ca2+ channels activated by the CF-EPSP. A, Left, Fluorescence reconstruction of a PN with two ROIs indicated (R1 and R2): R1 is next to a pipette delivering 1 μm of the P/Q-type VGCC inhibitor AgaIVA; R2 is >50 μm from the application pipette. Right, Top, Somatic Vm associated with CF-EPSPs in control conditions and after local application of AgaIVA at three different initial Vm: hyp (blue trace); int (green trace); and dep (red trace). Right, Bottom, The corresponding OG5N ΔF/F0 signals. B, In another PN, same as A, but with the pipette delivering the T-type VGCC inhibitors ML (5 μm) and NNC (30 μm). C, In another PN, same as A, but with the pipette delivering both the P/Q VGCC inhibitor AgaIVA (1μm) and the T-type VGCC inhibitors ML (5 μm) and NNC (30 μm). D, In three other PNs, from a region next to a pipette delivering a VGCC blocker, the OG5N ΔF/F0 signals associated with CF-EPSPs in control conditions and after local application of 20 μm of the L-type VGCC inhibitor Isr, of 5 μm of the N-type VGCC inhibitor PD or 1 μm of the R-type VGCC inhibitor SNX at the three different initial Vm.
Quantitative analysis of dendritic Ca2+ channels activated by the CF-EPSP. Left, From two representative cells, OG5N ΔF/F0 signals associated with CF-EPSPs in control conditions at three different initial Vm: hyp (blue trace); int (green trace); and dep (red trace); superimposed (gray traces) are the OG5N ΔF/F0 signals after local application of either 1 μm of the P/Q-type VGCC inhibitor AgaIVA or of the T-type VGCC blockers ML (5 μm) and NNC (30 μm); the first max of the Ca2+ transient is calculated within the first 4 ms after the CF stimulation; the second max of the Ca2+ transient is calculated between 4 and 14 ms after the CF stimulation; the percentages from control ΔF/F0 maxima after application of the VGCC blockers are reported above or below the arrows. Right, Mean ± SD of the percentages from control ΔF/F0 maxima after application of the VGCC inhibitors AgaIVA (N = 6 cells), ML+NNC (N = 6 cells), AgaIVA+ML+NNC (N = 6 cells), Isr (N = 4 cells), PD (N = 4 cells), or SNX (N = 4 cells). Gray columns represent the statistics of the first max. White columns represent the statistics of the second max. *Significant change in the max (p < 0.005, paired t test).
Dendritic K+ channels activated by the CF-EPSP
Activation of dendritic P/Q-type VGCCs at int and dep states is correlated with larger depolarization transients associated with CF-EPSPs, possibly due to steady inactivation of A-type VGKCs that are expressed in PN dendrites (Otsu et al., 2014). To test this hypothesis, we investigated the change in the OG5N Ca2+ transient produced by local application of the A-type VGKC inhibitor AmmTx3 (Zoukimian et al., 2019). As shown in the representative example of Figure 5A, the block of A-type channels strongly enhanced the Ca2+ transient associated with the CF-EPSP both at the hyp and int states but did not produce any effect at the dep state. In contrast to the experiments with VGCC blockers, the effect of AmmTx3 was not observed exclusively in the area adjacent to the pipette delivering the toxin (data not shown). In N = 6 cells tested, AmmTx3 significantly enhanced the first and second maxima of the Ca2+ transient at hyp and int states, also at sites located at >50 μm from the pipette delivering the toxin, although it did not modify the Ca2+ transient at dep states, indicating that A-type VGKCs are fully inactivated at initial Vm > −55 mV (Fig. 5B). In contrast, A-type VGKCs are activated at more negative initial Vm, and this result suggests that their activation limits the activation of P/Q-type VGCCs.
Dendritic A-type VGKCs activated by the CF-EPSP. A, Left, Fluorescence reconstruction of a PN with an ROI indicated next to a pipette delivering 1 μm of the A-type VGKC inhibitor AmmTx3. Right, Top, Somatic Vm associated with CF-EPSPs in control conditions and after local application of AmmTx3 at three different initial Vm: hyp (blue trace); int (green trace); and dep (red trace). Right, Bottom, The corresponding OG5N ΔF/F0 signals. B, Left, From the cell in A, OG5N ΔF/F0 signals associated with CF-EPSPs in control conditions at the three different initial Vm; superimposed (gray traces) are the OG5N ΔF/F0 signals after local application of either 1 μm AmmTx3; the first max of the Ca2+ transient is calculated within the first 4 ms after the CF stimulation; the percentages from control ΔF/F0 maxima after application of the VGCC blockers are reported above the arrows. Right, Mean ± SD of the percentages from control ΔF/F0 maxima after application of AmmTx3 (N = 6 cells). Gray columns represent the statistics of the first max. White columns represent the statistics of the second max. *Significant change in the max (p < 0.005, paired t test).
Because Ca2+ transients activate Ca2+-activated K+ channels, we finally investigated the change in the OG5N Ca2+ transient produced by local application of the BK channel inhibitor iberiotoxin (1 μm; Fig. 6A) and of the SK channel inhibitor apamin (1 μm; Fig. 6B). Neither iberiotoxin nor apamin produced any change in the first and second maxima of the Ca2+ transient at hyp, inta̧nd dep states, a result observed in N = 5 cells tested with iberiotoxin and N = 6 cells tested with apamin (Fig. 6C). Nevertheless, in 5 of 5 cells where the block of BK channels was tested, the number of Ca2+ spikes increased from 2 to 3 at dep states. In addition, in 4 of 6 cells where the block of SK channels was tested, a slight but observable decrease in the decay of the OG5N ΔF/F0 signal was noticed (Fig. 6D). These two results indicate that both channels are activated by the CF-EPSP at different initial Vm states.
Dendritic Ca2+-activated K+ channels activated by the CF-EPSP. A, Left, Fluorescence reconstruction of a PN with an ROI indicated next to a pipette delivering 1 μm of the BK channel inhibitor iberiotoxin. Right, Top, Somatic Vm associated with CF-EPSPs in control conditions and after local application of iberiotoxin at three different initial Vm: hyp (blue trace); int (green trace); and dep (red trace). Right, Bottom, The corresponding OG5N ΔF/F0 signals. B, In another PN, same as A, but with the pipette delivering 1 μm of the SK channel inhibitor apamin. C, Mean ± SD of the percentages from control ΔF/F0 maxima after application of iberiotoxin (N = 5 cells) or apamin (N = 6 cells). Gray columns represent the statistics of the first max. White columns represent the statistics of the second max. D, Left, From the cell in A, OG5N ΔF/F0 signal associated with the CF-EPSP at dep state in control condition and after addition of iberiotoxin (gray trace). Right, From the cell in B, OG5N ΔF/F0 signal associated with the CF-EPSP at hyp state in control condition and after addition of apamin (gray trace).
Analysis of dendritic ion currents associated with the CF-EPSP by NEURON modeling
To extrapolate individual dendritic ion currents underlying Vm transients and Ca2+ signals associated with CF-EPSPs at different Vm initial states, we selected 4 dendritic sites from 4 different PNs within those analyzed in Figure 2. The criterion for the selection was that the somatic resting Vm and CF-EPSP at different initial Vm were stable for the entire duration of the recordings, to assume the same conditions for the Vm and Ca2+ recordings. In these 4 compartments, we matched the experimental Vm and Ca2+ transients with traces obtained by computer simulations of the NEURON model illustrated in Figure 1B, as described in Materials and Methods. In particular, we established a unique kinetic model for each of the six types of channels, to consistently reproduce the behaviors observed in all 4 cells. The model that matches Vm and Ca2+ transients is available in the ModelDB database at http://senselab.med.yale.edu/ModelDB/showModel.cshtml?model=244679. While the channel kinetics were the same in all simulations, the density of P/Q-type and T-type VGCCs, of A-type and HVAK VGKCs, and of BK and SK Ca2+-activated K+ channels was adjusted to match experimental data in each cell, as shown in Figure 7. To validate the consistency of the four variants of the model, we run computer simulations by modeling the replacement of 1 mM Fura-FF with 2 mM OG5N and analyzed Ca2+ ΔF/F0 modifications produced by the elimination of 90% of each channel, in this way mimicking the experiments in which individual channels were pharmacologically blocked. As an example, the results for the model variant of Cell 1 in Figure 7 are reported in Figure 8, showing that 90% reduction in the density of each channel qualitatively reproduced the experimental behavior observed after toxin or drug inhibition. To compare experiments and simulations, we calculated the percentage of the first max and the second max of the Ca2+ transient after elimination of 90% of each channel. The comparisons, reported in Table 1, indicate that the observed effects on Ca2+ signals in the experiments are all in line with the predictions from the simulations. In addition, in all 4 variants of the model, the 90% reduction of BK channels reproduced the appearance of a third Ca2+ spike that was also observed in experiments. In summary, matching experimental data with NEURON simulations generated a simplified yet biological plausible dendritic model that successfully reproduces the complexity of experimental signals.
NEURON model of 4 PN dendritic compartments reproducing Vm and Ca2+ transients associated with the CF-EPSP. Left, Experimental dendritic Vm and Ca2+ transients associated with CF-EPSPs from 4 selected cells at three different initial Vm: hyperpolarized (hyp, blue trace); intermediate (int, green trace); depolarized (dep, red trace). Right, Simulations of dendritic Vm and Ca2+ transients associated with CF-EPSPs reproducing experimental data (gray traces).
Simulations of block of P/Q-type VGCCs, T-type VGCCs, A-type VGKCs, BK and SK Ca2+-activated K+ channels from a NEURON model. Simulated Ca2+ transients (OG5N) associated with CF-EPSPs at three different initial Vm in control condition and after reduction of 90% of one individual channels from Cell 1 model variant reported in Figure 7. For each case of 90% channel reduction, traces under control conditions are reported in gray.
Comparison of the percentage of the control Ca2+ transient produced by the pharmacological block of a channel type, in the case of experimental data, with the percentage of the control Ca2+ transient produced by the elimination of 90% of the channel in the case of simulationsa
We then used the model variant of Cell 1 of Figure 7 to extrapolate the kinetics of each channel contributing to the CF-mediated signal under different conditions (Fig. 9). At hyp state (blue traces), the depolarization transient carried by the CF-EPSP activates a robust Ca2+ current, mediated by T-type channels, and a K+ current, mediated by A-type channels. A very small K+ current mediated by SK channels is elicited by the transient Ca2+ elevation. P/Q-type VGCCs and HVAK VGKSs are poorly activated that this state. As the initial Vm becomes more positive (int state, green traces), part of T-type and A-type channels become inactivated, reducing the associated currents and allowing more P/Q and HVAK channels to activate. Finally, at dep state (red traces), T-type and A-type channels are fully inactivated and the depolarization transient carried by the CF-EPSP activates a Ca2+ current, mediated by P/Q-type Ca2+ channels, and a K+ current, mediated by HVAK channels, which are responsible for multiple Ca2+ spikes. The number of spikes is limited by the K+ current mediated by BK channels. The strong activation of P/Q-type VGCCs and HVAK VGKCs is prevented at hyp states by the K+ current mediated by A-type VGKCs. Hence, when A-type channels are reduced by 90% (purple traces), activation of P/Q-type VGCCs is strongly enhanced, substantially increasing the dendritic Ca2+ transient. In summary, by extrapolating the kinetics of individual Ca2+ and K+ channels, we resolved their functional interaction establishing a specific role of A-type VGKCs in controlling activation of P/Q-type, HVAK, and BK channels.
Individual currents extracted from the NEURON model. Ca2+ currents of P/Q and T channels and K+ currents of A, BK, SK, and HVAK channels from the simulations of Cell 1 model variant reported in Figure 7. Simulations were at hyp (blue traces), int (green traces), and dep (red traces) states in control conditions, and at hyp (purple traces) after blocking 90% of A-type VGKCs, superimposed to currents in control conditions (gray traces).
Discussion
In this article, we report an original study of the dendritic Vm and Ca2+ transients associated with the CF-EPSP at different initial Vm. We explored experimentally, using combined Vm and Ca2+ imaging, the activation of five Ca2+ or K+ channels, and we propose a simplified single-compartment model that produces simulations of the Vm and Ca2+ transients that accurately match experimental data. Interestingly, we used 4 stable cells for this analysis, and the matching models were obtained using the same kinetic models of the six channels with relatively small variations of conductance densities. Using this strategy, which combines fast imaging techniques, pharmacological analysis and biophysical modeling, we unraveled the precise kinetics of dendritic Ca2+ and K+ channel activation in PNs during CF-EPSPs under different Vm conditions. In particular, we established a clear role for the A-type VGKC in limiting the membrane depolarization and the activation of P/Q-type VGCCs.
Two distinct sets of channels activated by the CF-EPSP
We found that two different sets of channels are selectively activated at different initial Vm. When the dendrite is hyperpolarized (Vm ∼ −80 mV), the transient depolarization produced by the CF-EPSP invading the dendritic branch activates T-type VGCCs (Isope et al., 2012) that enhance the distal dendritic Vm depolarization produced by the spread of the EPSP. The dendritic Vm is, however, capped ∼<10 mV by the K+ current via A-type VGKCs, limiting the opening of HVA Ca2+ and K+ channels. Under this condition, Ca2+ influx activates SK channels (Hosy et al., 2011) that regulate Vm repolarization and that appear selectively linked to T-type VGCCs, presumably by molecular coupling (Stocker, 2004). When, in contrast, the dendrite is depolarized (Vm ∼ −50 mV), T-type VGCCs and A-type VGKCs are inactivated and the CF-EPSP can drive the dendrite to more positive Vm values that activate first P/Q-type VGCCs (Usowicz et al., 1992) and then HVAKs leading to Ca2+ spikes (see the channels model available at http://senselab.med.yale.edu/ModelDB/showModel.cshtml?model=244679). In this case, Ca2+ influx activates BK channels (Rancz and Häusser, 2006) that fasten Vm repolarization limiting the number of Ca2+ spikes. HVAKs include Kv3.3, which is highly expressed in PNs (Goldman-Wohl et al., 1994), regulating dendritic Ca2+ spikes (Zagha et al., 2010; Veys et al., 2013). However, the lack of a selective channel blocker for this VGKC did not permit the experimental assessment of the kinetics of this channel, and we cannot exclude a contribution of other VGKCs with similar biophysical properties. The role of BK channels is to dampen the generation of Ca2+ spikes, which are typically only two when these channels are active. Finally, significant activation of both sets of channels occurs only at the intermediate initial Vm (∼−65 mV). In summary, The CF-evoked dendritic Ca2+ influx is mediated by two VGCCs that exhibit two different kinetics of activation and that are presumably associated with two distinct molecular pathways. In particular, P/Q-type VGCCs are believed to trigger endocannabinoid release and short-term synaptic depression (Rancz and Häusser, 2006). The scheme of activation of the two distinct sets of functionally coupled channels is illustrated in Figure 10.
Channel activation following CF-EPSPs at hyperpolarized and depolarized states. A, In control conditions, at hyp state, the CF-EPSP activates T-type channels, which activate SK channels, and A channels, which limit activation P/Q and HVAK channels; at dep state, the CF-EPSP activates P/Q-type channels, which activate BK channels, and HVAK channels, while T channels and A channels are inactivated. B, When A channels are blocked or inactivated, at hyp state, the CF-EPSP also activates P/Q-type channels, which activate BK channels, and HVAK channels.
A potential role of A-type VGKCs in synaptic plasticity
The scheme of Figure 10 also shows the result of blocking or inactivating A-type VGKCs. As these channels prevent the activation of the second set of channels when the dendrite is hyperpolarized, the modulation of this channel by synaptic transmission may provide a mechanism for triggering a CF nonlinear behavior playing a role in associative plasticity. Indeed, it has been shown that A-type VGKCs are modulated by Type 1 metabotropic glutamate receptors (Otsu et al., 2014), and this mechanism can play a role when the CF-EPSP is concomitant with PF activation. A-type VGKCs can also be rapidly inactivated by depolarization produced by EPSPs, driving the dendrite to a depolarized state. This mechanism occurs in the dendrites of CA1 hippocampal pyramidal neurons where inactivation of A-type VGKCs by Schaffer collateral EPSPs leads to boosting of backpropagating action potentials, a mechanism playing a role in Hebbian plasticity at these synapses (Magee and Johnston, 1997). While A-type VGKCs can act as a functional trigger of synaptic plasticity, these channels can be potentially the target of meta-plasticity mechanisms to regulate dendritic functions, in particular with respect to membrane excitability. In CA1 hippocampal pyramidal neurons, coupling between local dendritic spikes and the soma can be modified in a branch-specific manner through regulation of dendritic A-type K+ channels, a phenomenon that allows spatiotemporal correlation of synaptic inputs (Losonczy et al., 2008). In the cerebellum, this phenomenon occurs in the case of LTP of mossy fiber inputs to granule cells (Rizwan et al., 2016).
Relevance of our approach in the understanding of synergistic behaviors of ion channels
Voltage-gated and Ca2+-activated ion channels shape the integration of incoming inputs in dendritic compartments and determine the pattern of Vm and Ca2+ influx (Magee and Johnston, 2005). In particular, each ion channel contributes to the Vm transient that in turn regulates the state of the channel (open, close, or inactivated). This patterning of mutual interactions determines a global synergy with sets of distinct channels that are functionally coupled. Hence, the understanding of dendritic integration relies on the precise reconstruction of the kinetics of all principal channels underlying the response of a dendritic compartment to a given physiological input. This concept also applies to dendritic abnormal behaviors associated with channelopathies, such as the reported cases in Fragile X syndrome (Brager and Johnston, 2014; Zhang et al., 2014). Yet, the direct measurement of diverse ionic currents, mediated by different ions, is beyond available experimental techniques. In the last few years, we developed techniques to directly measure the kinetics of Ca2+ currents mediated by VGCCs in the dendrites of CA1 hippocampal pyramidal neurons (Jaafari et al., 2014, 2015; Jaafari and Canepari, 2016) and of PNs (Ait Ouares et al., 2016), starting from high-temporal resolution Ca2+ imaging. In the present work, we used an alternative strategy to extract, indirectly, all the Ca2+ and K+ currents underlying the CF-mediated signals, using the NEURON simulation environment (Hines and Carnevale, 1997) applied to the same type of recordings, combined with Vm imaging. This novel approach allowed us reconstructing the functional interaction among individual channels, demonstrating the role of A-type VGKCs in controlling the activation of P/Q-type VGCCs. Thus, it was possible, for the first time, to deduce the kinetics of several Ca2+ and K+ channels in parallel. It should be highlighted that a single compartment model or a multicompartment model with a few compartments is not a realistic neuronal model of a PN because they do not account for either the precise morphology of the cell or the mechanisms present in small protrusions, such as dendritic spines. Yet, our approach is based on the concept that fewer parameters allow better constraining their values when fitting with a set of experimental observations. A complex multicompartmental model based on thousands of compartments (De Schutter, 1998) can be used to predict general cell integrative behaviors leading to firing activity, but is not ideal for deducing the precise kinetics of dendritic channels from detailed experimental observations. Modeling a PN dendrite as a single compartment is practical because PN dendritic trees comprise only two VGCCs and no voltage-gated Na+ channels. It will be very interesting, however, to implement these optimized channel models into realistic PN models to predict the major physiological functions of these neurons. A single compartment approach would not be appropriate in dendrites of cortical and hippocampal pyramidal neurons because these dendrites are endowed with a more complex composition of voltage-gated channels and notably stronger action potential backpropagation. Precisely reconstructing dendritic or axonal compartments in other systems is possible, in principle, using the same approach by expanding the pharmacological analysis to all the channels involved in a signal. For this purpose, the database ModelDB for NEURON simulations already includes >1100 published models covering >130 research topics (McDougal et al., 2017), which can be used as a starting framework to produce more simplified models with realistic channel kinetics matching the complexity of Vm and Ca2+ imaging experiments at high temporal resolution. The application of this novel approach also concerns the study of alterations induced by mutated proteins associated with channelopathies.
Footnotes
This work was supported by Agence Nationale de la Recherche Grant ANR-14-CE17-0006 WaveFrontImag, Labex Ion Channels Science and Therapeutics Program Grant ANR-11-LABX-0015, and National Infrastructure France Life Imaging Noeud Grenoblois; and Federation pour la recherché sur le Cerveau Grant Espoir en tête, Rotary France. L.F. was also supported by European Union COST Action CA15124 (NEUBIAS). A.T. and P.P. were supported by European Research Council Starting Grant dEMORY GA 311435. A.T. was supported by Google EMEA Scholarship, Onassis Foundation MSc Scholarship, and Einstein Foundation Berlin PhD Fellowship.
The authors declare no competing financial interests.
- Correspondence should be addressed to Macro Canepari at marco.canepari{at}univ-grenoble-alpes.fr