Differential targeting and functional specialization of sodium channels in cultured cerebellar granule cells

Abstract

The ion channel dynamics that underlie the complex firing patterns of cerebellar granule (CG) cells are still largely unknown. Here, we have characterized the subcellular localization and functional properties of Na⁺ channels that regulate the excitability of CG cells in culture. As evidenced by RT-PCR and immunocytochemical analysis, morphologically differentiated CG cells expressed Nav1.2 and Nav1.6, though both subunits appeared to be differentially regulated. Nav1.2 was localized at most axon initial segments (AIS) of CG cells from 8 days in vitro DIV 8 to DIV 15. At DIV 8, Nav1.6 was found uniformly throughout somata, dendrites and axons with occasional clustering in a subset of AIS. Accumulation of Nav1.6 at most AIS was evident by DIV 13–14, suggesting it is developmentally regulated at AIS. The specific contribution of these differentially distributed Na⁺ channels has been assessed using a combination of methods that allowed discrimination between functionally compartmentalized Na⁺ currents. In agreement with immunolocalization, we found that fast activating–fully inactivating Na⁺ currents predominate at the AIS membrane and in the somatic plasma membrane.

Cerebellar granule (CG) cells integrate sensory information from somatosensory, vestibular, acoustic and visual origins via mossy fibres to regulate the activity of inhibitory interneurones and Purkinje cells (Ito, 1984).

CG cells respond repetitively to excitatory inputs conveyed by mossy fibres with complex discharge patterns. Electrical behaviour such as spike bursting, oscillations and resonance have been revealed by pharmacological manipulation (D'Angelo et al. 1998, 2001). More recently, sensory stimulation of the vibrissae in anaesthetized rats has been shown to generate bursts of action potentials in CG cells, with maximal frequencies as high as 200 Hz (Chadderton et al. 2004). Likewise, single extracellular stimulation of rat cerebellar parallel fibres unexpectedly triggers a doublet or a burst of action potentials in CG cells, a mechanism that may be involved in the induction of long-term depression at the parallel fibre–Purkinje cell synapse (Isope & Barbour, 2002; Isope et al. 2004) and that suggests that complex, as yet unresolved, ion channel dynamics underlie granule cell firing patterns.

An important step towards understanding the molecular determinants of the excitability of granule cells is to determine the properties of Na⁺ currents, as they control both subthreshold activity and action potential electrogenesis. Although substantial knowledge has been accumulated about voltage-dependent K⁺ and Ca²⁺ currents in granule cells (Cull-Candy et al. 1989; Pearson et al. 1995; Shibata et al. 2000; D'Angelo et al. 2001), Na⁺ currents have not been explored thoroughly. Among the 10 different subunits encoding Na⁺ channel subtypes (Goldin, 1999), in situ hybridization and immunodetection have demonstrated that CG cells of adult rodents express Nav1.2 and Nav1.6, and possibly Nav1.1 (Westenbroek et al. 1989; De Miera et al. 1997; Felts et al. 1997; Schaller & Caldwell, 2000). A major challenge then is to relate these different channel subunits to the Na⁺ currents observed in CG cells and to establish their distribution.

Functionally, Na⁺ channels recorded in mature granule cells are tetrodotoxin (TTX)-sensitive with fast activation/inactivation kinetics (Hockberger et al. 1987; Cull-Candy et al. 1989; D'Angelo et al. 1994; Stewart et al. 1995; Carlier et al. 2000). However, these conclusions are largely based on CG cell recordings in which the electronically remote regenerative currents were inadequately controlled, making it difficult to characterize the Na⁺ channel isoforms electrically. Moreover, based on current clamp recordings, D'Angelo et al. (1998) have suggested a role for a persistent Na⁺ current in sustaining subthreshold depolarizing potentials in CG cells, although direct voltage-clamp evidence for this current has yet to be presented.

Accordingly, our aim was to answer three main questions. First, which Na⁺ channel subunits are expressed in differentiated CG cells, and what is their specific subcellular location? Second, what are the properties of the Na⁺ currents in CG cells when these currents can be voltage clamped adequately? Third, are the different components of the Na⁺ current attributable to channel heterogeneity and/or subcellular location? We show that CG cells display different distributions of Nav1.2 and Nav1.6 isotypes in soma, AIS and dendrites, giving rise to functionally compartmentalized Na⁺ currents. Some of these results have appeared in abstract form (Osorio et al. 2004).

Methods

Culture of cerebellar granule cells

Animal use followed guidelines established by the European Animal Care and Use Committee (86/609/CEE). Cerebellar granule cells were cultured according to previously described procedures (Levi et al. 1984) with some modifications. Briefly, primary cultures were prepared from decapitated 7-day-old Wistar rats (P7). Immediately after death, cerebella were dissected out and treated with trypsin (0.02%) in Hank's balanced salt solution (HBSS, Sigma, St Louis, MO, USA) for 15 min at 37°C. The tissue was then washed several times in Neurobasal medium (Gibco-Invitrogen, Grand Island, NY, USA) supplemented with 10% fetal bovine serum and gently triturated using fire-polished Pasteur pipettes. The homogenate was centrifuged at 800 r.p.m. (50 g) for 2 min and the pellet was resuspended in plating medium that consisted of: Neurobasal medium supplemented with B27 (20 μl ml⁻¹, Gibco), glutamine (2 mm), penicillin (50 i.u. ml⁻¹), streptomycin (50 μg ml⁻¹) and KCl (24 mm). After a brief centrifugation at 800 r.p.m., cells were plated in Petri dishes coated with poly-l-lysine (0.05 mg ml⁻¹) at a density of 1 × 10⁵ cells cm⁻². Cells were incubated at 37°C in an atmosphere of 5% CO₂ humidity. After 72 h, 2 μm AraC (Sigma) was added to avoid glial cell proliferation. Half of the culture medium was changed every 4 days.

Immunocytochemistry and imaging

Granular cell cultures were fixed using 4% paraformaldehyde in phosphate-buffered saline (PBS). After several washes in PBS, non-specific binding was reduced by preincubating cells in blocking buffer containing 3% bovine serum albumin and 0.1% Triton X-100 in PBS. Primary antibodies were incubated in blocking buffer for 3 h at room temperature. Secondary antibodies were incubated in blocking buffer for 1 h at room temperature. Cells were then washed in PBS and mounted in Mowiol (Colbiochem, Merck, Darmstadt, Germany). Blocking controls for non-specific staining were performed by 1 h preincubation of the primary antibodies with a large molar excess of the corresponding immunizing peptides.

Immunostaining of HEK293 cells stably expressing the human Nav1.1 sodium (Clare et al. 2000; Mantegazza et al. 2005) channel was performed as described above.

Primary antibodies used and dilutions were: β-III tubulin (monoclonal SDL.3D10, Sigma) 1/1000; serine-rich domain of ankyrin-G node (Bouzidi et al. 2002) 1/500; MAP2 (monoclonal HM-2, Sigma) 1/400; PanNav (monoclonal K58/35, Sigma) 1/200; Nav1.1 (polyclonal ASC-001, Alomone, Jerusalem Israel) 1/100; Nav1.2 (polyclonal ASC-002, Alomone, and polyclonal 06-633, Upstate Biotechnology, Lake Placid, NY, USA) 1/100; Nav1.6 (polyclonal ASC-009, Alomone) 1/100 and ankyrin-B (monoclonal, Oncogene) 1/50. Alexa Fluor 488- and Alexa Fluor 546-conjugated goat secondary antibodies (1/200 to 1/400, Molecular Probes, Eugene, OR, USA) were used to detect rabbit polyclonal antibodies and mouse monoclonal antibodies, respectively.

Images were acquired using an optical Nikon microscope (Nikon, Tokyo, Japan) equipped with a digital Nikon Coolpix camera or a TCS SP2 laser-scanning confocal microscope (Leica Microsystems, Mannheim, Germany) and later exported into Photoshop (Adobe Systems, San Jose, CA, USA) for final processing. Images comparing peptide-blocked and unblocked antibody staining were acquired and digitally processed in an identical manner.

Reverse transcriptase-PCR

Total RNA was extracted from three cultures of CG cells at DIV 7 with TRI Reagent (Sigma) following the manufacturer's protocol. Reverse transcription reactions with primers for rat Nav1.2, Nav1.3, Nav1.6 and GAPDH (glyceraldehyde-3-phosphate dehydrogenase) were performed as described by Alessandri-Haber et al. (2002). PCR samples corresponding to an input of 0, 0.1 and 1 ng of reverse-transcribed cDNA were analysed on 1% agarose gels after 34 and 38 cycles of amplification.

Patch-clamp recordings

Whole-cell patch-clamp recordings were conducted at room temperature on CG cells kept 7–15 days in vitro (DIV 7–15). Patch pipettes were pulled from thick-walled borosilicate glass capillaries (Harvard Apparatus, Edenbridge, UK) and had a resistance of 8–12 MΩ. The small soma size of CG cells (< 10 μm) prevented stable recordings when using pipettes with lower resistance. All recordings were made using an Axopatch 200B amplifier (Axon Instruments, Foster City, CA, USA), low-pass filtered at 2 kHz and digitized at 25 kHz. Transient and leakage currents were digitally subtracted using a standard P/n protocol (n = 6), unless otherwise stated (e.g. voltage ramp protocol). The series resistance measured in the whole-cell mode was 12–20 MΩ and was compensated by 60–75%. The intracellular solution used to achieve a standard Na⁺ gradient consisted of (mm): 100 CsCl, 30 CsF, 10 NaCl, 1 MgCl₂, 0.5 CaCl₂, 10 Hepes, 10 EGTA, 4 Mg-ATP and 0.4 Na-GTP (Coste et al. 2004). The bathing solution contained the following (mm): 120 NaCl, 3 KCl, 2 CaCl₂, 1 MgCl₂, 11 glucose, 10 Hepes, 0.5 CdCl₂, 20 TEA-Cl and 1 4-aminopyridine (4-AP). In inverse Na⁺ gradient (Numann et al. 1991; Carlier et al. 2000), the intracellular solution consisted of (mm): 20 CsCl, 110 NaCl, 1 MgCl₂, 0.5 CaCl₂, 10 Hepes, 10 EGTA, 4 Mg-ATP and 0.4 Na-GTP and the extracellular solution was (mm): 210 sucrose, 10 NaCl, 3 KCl, 2 CaCl₂, 1 MgCl₂, 11 glucose, 10 Hepes, 0.5 CdCl₂, 20 TEA-Cl and 1 4-AP. pH was adjusted to 7.35 with either NaOH or CsOH (302–305 mosmol l⁻¹). In Na⁺-free external solution, the concentration of sucrose was increased to 220 mm. Input resistance and resting potential of CG cells were measured in current clamp mode with standard extracellular solution (without CdCl₂, 4-AP and TEA) and an intrapipette solution consisting of (mm): 125 potassium gluconate, 5 NaCl, 1 MgCl₂, 0.5 CaCl₂, 10 Hepes, 10 EGTA, 4 Mg-ATP and 0.4 Na-GTP.

Local application of TTX

Local application of TTX, Na⁺-free saline or kainate was achieved by pressure ejection from a small patch pipette (∼1 μm i.d.) positioned at ∼15 μm from the site of interest and directly in front of a large suction pipette (15–20 μm i.d.) (see Fig. 7A). Using small micropipettes and fast flow exchange rates, the maximum spread was estimated to be ∼5–10 μm in diameter as measured when sucrose was perfused in normal Na⁺. The fact that local application of TTX to proximal dendritic sites (7–15 μm away from the cell body) had little or no effect on the somatically recorded whole-cell current indicated that there was little, if any, diffusion of the applied solutions.

Figure 7. Somatic and AIS components of the Na⁺ current.

A, effects of local application of TTX to either the soma or the neurites on Na⁺ currents recorded in inverse Na⁺ gradient. Upper panel: illustration, approximately to scale, showing the perfusion system together with the somatic whole-cell electrode. The position of the perfusing micropipette in front of a large-bore suction pipette permitted rapid exchange of the external solution surrounding the site of interest. TTX (1 μm) was applied to the neurite (1) at the expected site of the AIS (10–15 μm away from the soma) and onto the cell body (2). Lower panel: outward Na⁺ currents evoked by depolarizing pulses to +20 mV in control, during application of TTX on the neurite (1) or the soma (2) and after washout of TTX. Perfusion of TTX on the soma caused 76% block of the Na⁺ current. Due to technical difficulties, small contaminating outward currents (arrowhead) were not subtracted from rapidly activating outward Na⁺ currents in this experiment. B, summary of the effects of TTX application on the soma and neurites in inverse Na⁺ gradient. Experimental configuration as in A, except that the cells (n = 7) were tested on the soma and on the two visually identified neurites. C, normal Na⁺ gradient. TTX was applied chronologically at the sites indicated schematically, following the sequence indicated (1 → 3). Na⁺ currents evoked at −20 mV (C: control; W1 and W2: washouts 1 and 2). Scale bars: 2 ms, 200 pA.

Data analysis

Data were analysed using pCLAMP 8.02 (Axon Instruments) and Prism 4.0 (GraphPad, San Diego, CA, USA) software. Under conditions of inverse Na⁺ gradients, Na⁺ currents were isolated from residual contaminating outward currents by subtracting TTX-insensitive currents (except in Fig. 7A).

Conductance–voltage curves were calculated from the peak current according to the equation G_Na=I_Na/(V–E_Na) where V is the test pulse potential and E_Na the reversal potential extrapolated from the current–voltage (I–V) curve. It should be noted that though E_Na (obtained by fitting I–V relationships in inverse Na⁺ gradient) was close to the theoretical E_Na, Na⁺ currents were typically isolated by subtraction of currents remaining in 1 μm TTX. The activation curve (G–V) was fitted using the Boltzmann function: G/G_max= 1/(1 + exp(V_0.5–V/k)), where G/G_max is the normalized conductance, V_0.5 is the potential of half-maximum activation and k is the steepness factor. Inactivation curves were constructed from normalized currents and fitted according to the Boltzmann function: I/I_max= 1/(1 + exp(V–V_0.5/k)), where V is the conditioning pulse potential and V_0.5 is the membrane potential at which half of the channels are inactivated. The time course of entry or recovery from inactivation was determined using the Chebyshev method to fit current traces with a single exponential equation: Y=Y_max(1 – exp(–t/τ)), where τ is the time constant and Y represents the fraction of I recovery (I_t/I_pre) or the normalized current (I/I_max). For the study of slow inactivation, which required long-lasting recordings, the measured I_Na was corrected for run-down (estimated periodically from a holding potential (V_h) of −100 mV). The concentration–inhibition curves for TTX were fitted with the Hill equation of the form Y=Y_max[TTX]^nH/(IC₅₀^nH+[TTX]^nH), where Y is the percentage inhibition (e.g. 100 ×I/I_{[TTX]= 0}), IC₅₀ the TTX concentration that produces half-maximal inhibition and n^H the Hill coefficient. Results are presented as mean ± s.e.m. and n represents the number of cells examined. Statistical analysis was performed using unpaired or paired t tests, P < 0.05 being considered significant.

Results

Identification and distribution of Na⁺ channels in morphologically differentiated granular cells

CG cells isolated at P7 were studied after 7–15 days in vitro, at a stage where they exhibit typical features of functionally mature cells, namely high input resistance (1 ± 0.1 GΩ), negative resting potential (−66 ± 2 mV, n = 7), overshooting action potentials and spontaneous synaptic inputs, often in the form of bursting patterns (Gallo et al. 1987; Hockberger et al. 1987; Galdzicki et al. 1991; Becherer et al. 1997).

From DIV 7 onwards, CG cells had elongated thin neurites and developed a bipolar or multipolar morphology as revealed by staining for the neurone-specific marker β-III tubulin (Fig. 1A). Axon initial segments were identified by the presence of ankyrin-G (Boiko et al. 2003), overlapping β-III tubulin-positive processes (Fig. 1A). Using low density cultures, it was possible to assign an individual AIS to a particular CG cell and to specify the characteristics of each AIS. Ankyrin-G labelling typically began 11.5 ± 1 μm away from the soma and extended to 16.6 ± 1 μm (n = 52) (Fig. 1). Superimposition of ankyrin-G and MAP2 staining showed that the two proteins did not colocalize. MAP2 labelling was restricted to the soma and to one or more processes, identified as dendrites, while in some instances staining was also observed in the axon hillock (Fig. 1B) as already reported in hippocampal neurones (Cáceres et al. 1986). The axonal nature of the neurites harbouring ankyrin-G-positive segments was verified by dual staining with the anti-MAP2 antibody and an antibody directed against an axon-specific epitope of ankyrin-B (data not shown).

Figure 1. Morphologically differentiated cerebellar granule (CG) cells in culture.

A, fluorescent photomicrographs of 3 representative CG cells (DIV 7) double-labelled with anti-β-III tubulin (red) and anti-ankyrin-G (green) antibodies. Arrowheads indicate ankyrin-G-positive axon initial segments (AIS). B, fluorescent photomicrographs of a CG cell double-labelled with anti-ankyrin-G (green) and anti-MAP2 (red) antibodies. The overlay is shown on the right panel. Arrowheads indicate AIS. Scale bars in A and B, 10 μm.

To localize Na⁺ channels, we used a monoclonal antibody (PanNav) that recognizes all α-subunit isoforms of neuronal Na⁺ channels. Double labelling with PanNav and ankyrin-G antibody revealed intense immunostaining for Nav channels that coincided with ankyrin-G labelling at the AIS, as shown by confocal analysis of the extensive neuritic network (Fig. 2A) as well as single neurones (Fig. 2B). Examination of adjacent confocal planes also revealed faint PanNav staining in CG cell bodies and along axons and dendrites (Fig. 2B). PanNav-labelled segments did not immediately border CG cell bodies, indicating that Na⁺ channels do not cluster at high density in the axon hillock and instead concentrate in the AIS (Fig. 2B). Pre-adsorption of the PanNav antibody led to virtually complete loss of staining in the AIS, leaving some background staining in the cell body (Fig. 2C).

Figure 2. Na⁺ channels concentrate at the axon initial segment (AIS) of CG cells.

A, confocal images of CG cells cultured for 8 days and double-labelled with anti-ankyrin-G (green) and PanNav (red) antibodies. Overlay shows that PanNav immunoreactivity colocalizes with ankyrin-G-immunoreactivity. Images are projections of 5 consecutive optical sections spanning 2.5 μm. Cells bodies are not clearly visible in these optical sections. Scale bar, 20 μm. Ba and b, details of PanNav staining in 2 different CG cells, double-stained as in A. Na⁺ channels accumulate in ankyrin-G-enriched AIS (arrowheads). Images are projection of 15 consecutive optical sections spanning 2.8 μm. Arrowheads indicate AIS. Scale bars, 10 μm. C, representative neurone double-labelled as in B when PanNav antibody was preincubated with the cognate peptide. Note that labelling is abolished at the AIS whereas a dim signal persists in the soma. Images are projections of 15 consecutive optical sections spanning 2.8 μm. Arrowheads indicate AIS. Scale bar, 10 μm.

Out of the 10 Nav1 isoforms, Nav1.1, Nav1.2 and Nav1.6 have been detected in developing and mature CG cells by in situ hybridization (De Miera et al. 1997; Felts et al. 1997). To test for the presence of these different Na⁺ channel isoforms, CG cells were double-labelled with specific Na⁺ channel α-subunit antibodies and PanNav, using PanNav immunoreactivity to mark AIS based on the results presented above. Staining with Nav1.1-specific antibody at DIV 8 produced a dim signal in the soma, which was not extinguished by preincubating the primary antibody with the immunizing peptide (Fig. 3A). Similar results were obtained at DIV 14, indicating that there was no specific staining for Nav1.1, though the Nav1.1 antibody specifically recognized a protein of 250 kDa in Western blots (data not shown) and labelled HEK cells stably expressing human Nav1.1 (Fig. 3A, inset). From DIV 7 to DIV 14, anti-Nav1.2 antibody (ASC-002) consistently labelled a short proximal segment of neurites that overlapped with PanNav immunoreactivity, indicative of an accumulation of Nav1.2 at the AIS. Out of 315 initial segments identified by PanNav staining at DIV 8, 77% were colocalized with bright Nav1.2 staining (Figs 3B and 4C). This percentage remained stable over time in culture, with 85% of co-labelled PanNav/Nav1.2 AIS at DIV 14 (see Fig. 4C). It should be noted that confocal through-focus series did not detect, or barely detected, specific Nav1.2 staining along distal axons and dendrites. This pattern of Nav1.2 staining with clustering at AIS was confirmed using a second Nav1.2 antibody (06-633, see Methods).

Figure 3. Nav1.2 immunofluorescence colocalizes with PanNav at initial segments.

CG cells cultured for 8 days were double-labelled with PanNav and anti-Nav1.1 antibodies (A) or PanNav and anti-Nav1.2 antibodies (ASC-002) (B). Right panels show immunoreactivity after competing with the respective cognate peptides. Images are projections of 15 consecutive optical sections spanning 2.8 μm. Arrowheads indicate AIS. Scale bars, 10 μm. Insets in A: confocal image of stably transfected HEK cells expressing human Nav1.1 and labelled with anti-Nav1.1 antibody with or without preincubation with the immunizing peptide. Images are projections of 30 consecutive optical sections spanning 6 μm. Scale bars, 20 μm.

Figure 4. Nav1.6 appears at AIS during maturation in culture.

CG cells were double-labelled with PanNav and anti-Nav1.6 antibodies at DIV 8 (A) or DIV 12 (B). Right columns show immunoreactivity for Nav1.6 when the primary antibody was preincubated with the immunizing peptide. At DIV 8, Nav1.6 staining was diffusely distributed and rarely concentrated at the AIS whereas accumulation was detected in most AIS at DIV 12. Images are projections of 15 consecutive optical sections spanning 2.8 μm. Scale bars, 10 μm. C, histogram showing the percentage of PanNav-immunoreactive AIS at DIV 8 or DIV 14, in which accumulation of Nav1.2 or Nav1.6 was seen. Note that Nav1.6 only concentrated in a subset of AIS by DIV 8.

From DIV 8 on, Nav1.6 labelling was seen in virtually all CG cells, but in contrast to Nav1.2 staining, immunoreactivity was diffusely distributed in the cell body and along the entire length of dendrites and axons (Fig. 4A), showing only in a few instances clustering at AIS. Out of 224 initial segments identified with PanNav labelling, only 28% showed accumulation of Nav1.6 by DIV 8, whereas Nav1.6 clustering was evident in most AIS by DIV 14, increasing the percentage of AIS co-labelled with PanNav to 72% (Fig. 4B and C). We were unable to determine whether Nav1.6 and Nav1.2 are spatially segregated within the AIS because of the lack of appropriate Na⁺ channel antibodies. Thus, these data suggest that the high density of Na⁺ channels at the AIS at early stages in culture stems from a preferential localization of Nav1.2 subunits, whereas by DIV 12–14 both Nav1.2 and Nav1.6 may coexist at the AIS.

In agreement with immunochemistry, reverse transcription-PCR from CG cells cultured for 7 days consistently identified mRNA transcripts for Nav1.2 and Nav1.6 subunits (Fig. 5). Because the Nav1.3 isoform is often present at early stages of development in multiple brain structures, we also tested for the expression of Nav1.3 mRNA in our culture system using specific primers for rNav1.3 (Alessandri-Haber et al. 2002). Nav1.3 transcripts were not detected using cDNA templates that typically produced amplicons for Nav1.2 and Nav1.6 (Fig. 5).

Figure 5. Both Nav1.2 and Nav1.6 transcripts are detected in CG cells.

RT-PCR experiments were performed on total RNA extracted from cultured granule cells at DIV 7. PCR products amplified from the indicated amounts of reverse-transcribed cDNA (ng) were analysed on 1% agarose gels. Primers for Nav isoforms and GAPDH are as in Alessandri-Haber et al. (2002). For each initial amount of cDNA, amplification was performed for 34 (first lane) and 38 (second lane) cycles. Position of the 0.5 and 0.3 kb markers are indicated on the right of each gel (arrowheads). Co-amplification of GAPDH mRNA provided an internal control for PCR efficiency.

Dissection of whole-cell Na⁺ current into somatic and AIS Na⁺ current components

Whole-cell Na⁺ currents in cultured CG cells were isolated using external Cd²⁺ (500 μm), 4-AP (1 mm) and TEA (20 mm) and Cs⁺-based intracellular solutions. Under these conditions, CG cells displayed a fast activating–inactivating inward current that was fully suppressed by isosmotically substituting external Na⁺ by impermeant molecules (see below) and by bath-applying TTX. The TTX dose–response curve gave an IC₅₀ of 13.2 ± 1 nm with a Hill coefficient of 0.8 ± 0.1 (data not shown).

Although cultured CG cells have small capacitance (4–7 pF) and are electrotonically compact, they possess long thin neurites, which make them not amenable to good time-clamp control. Thus, under regular or even reduced Na⁺ gradient, depolarizing voltage steps activated non-graded, rapidly activating Na⁺ currents, which manifested as late or all-or-none currents (left and middle panels in Fig. 6A). To gain better control of the transmembrane voltage, Na⁺ currents were examined using an inverse Na⁺ gradient (i.e. 110 mm[Na⁺]_i and 10 mm[Na⁺]_o), rendering the Na⁺ currents non-regenerative. Under these conditions, outward Na⁺ currents evoked by a standard activation protocol were smoothly graded with increasing depolarization and free of notches suggesting adequate control of the membrane potential (Fig. 6A, right panel). Outward Na⁺ currents were then isolated from outward contaminating currents by subtracting TTX-insensitive currents (Fig. 6B). Currents recorded in inverse Na⁺ gradient have rapid activation kinetics over the whole activation range and inactivate quickly and completely within a few milliseconds. Outward Na⁺ currents had TTX sensitivity indistinguishable from those recorded in standard Na⁺ and were fully blocked by 1 μm TTX (Fig. 6B). The current–voltage relationship of normalized TTX-sensitive currents plotted in Fig. 6C (n = 11) shows that currents activated at ∼−40 mV and gradually reached maximal activation at around +10 mV. The extrapolated reversal potential for Na⁺ ions was estimated to be −60 mV, which agrees well with the predicted Nernst potential for Na⁺ ions in inverse gradient (E_Na=−60.4 mV). It should be noted that I–V relationships of TTX-sensitive Na⁺ currents determined at DIV 7–8 or DIV 12–14 were indistinguishable, both in terms of voltage dependence and maximum peak currents (see below), indicating that redistribution of Nav1.6 subunits at DIV 7–14 was not correlated with detectable changes in current properties.

Figure 6. Voltage-clamp recording of Na⁺ currents in inverse Na⁺ gradient.

A, representative current traces elicited by depolarizing voltage steps in CG cells recorded either in normal Na⁺ gradient ([Na⁺]_o 120 mm and [Na⁺]_i 10 mm; left panel), in ‘symmetrical’ Na⁺ gradient ([Na⁺]_o 120 mm and [Na⁺]_i 110 mm; middle) and in inverse Na⁺ gradient ([Na⁺]_o 10 mm and [Na⁺]_i 110 mm; right panel). B, outward Na⁺ currents evoked in inverse Na⁺ gradient by voltage steps from −50 to +40 mV in the absence or presence of 1 μm TTX. TTX-sensitive current traces are shown in the right panel. C, normalized I–V relationship for TTX-sensitive currents in inverse Na⁺ gradient. Each data point is the mean ± s.e.m. of 11 CG cells. Dashed line shows the extrapolated equilibrium potential for Na⁺ (E_Na).

To determine the location of the Na⁺ channels recorded in inverse Na⁺ gradient, we applied TTX (1 μm) locally to either the soma or to the expected site of the AIS in cells cultured for 7–10 days. Local application of TTX was achieved by pressure ejection from a small patch pipette (∼1 μm i.d.), the tip of which was positioned at ∼15 μm from the site of interest and directly in front of a large suction pipette (15 μm i.d.) (Fig. 7A). This was exclusively done on CG cells that displayed a bipolar morphology with visually identified neurites comparable to those illustrated in Fig. 1. Local application of TTX at the expected site of the AIS or distally (up to 50 μm) had no effect on outward Na⁺ currents in 11 out of 14 neurites tested (2 neurites tested per cell; Fig. 7A). In the three remaining neurites, a partial block of 10, 32 and 52% was seen (Fig. 7B). In marked contrast, application of TTX onto the soma suppressed the Na⁺ current in all cells tested giving an average block of 79 ± 3% (n = 7; Fig. 7B). Total outward current block was never attained by local TTX application onto the soma whereas application of TTX via the general perfusion system fully inhibited Na⁺ currents. This indicates that ∼20% of the recorded current in inverse gradient actually resulted from channels located outside the CG cell bodies.We repeated the protocol of Fig. 7A by applying TTX locally to neuritic or somatic sites of the same cell in standard Na⁺ gradient (Fig. 7C). Application of TTX onto the soma blocked a lower fraction of the whole-cell Na⁺ current (42 ± 4%, n = 18), as expected from recruiting fast transient Na⁺ currents from poorly clamped portions of the neurites. Nevertheless, effects of TTX on the two neurites were not identical since application had either very small effects (2.1 ± 1%, range 0–10%, n = 11) or yielded substantial block (66 ± 6%, range 50–70%). Thus, when TTX could be applied successfully to both neurites of the same cell (n = 5), only one neurite was found to be clearly sensitive to TTX and identified a posteriori as the axon. An alternative approach was also used, which involved local application of a Na⁺-free external solution. This essentially replicated the results obtained with TTX since local application of Na⁺-free solution on the soma blocked on average 52 ± 4% of the whole Na⁺ current (n = 11). Likewise, puffing Na⁺-free solution on neurites at the expected site of the AIS produced either substantial (40 ± 5%) or very little (4.8 ± 2%) Na⁺ current block depending on the neurites.

Taken together, these data indicate that Na⁺ currents recorded in an inverse Na⁺ gradient result primarily from somatic Na⁺ channels, whereas in a standard Na⁺ gradient both somatic and axonal channels contribute to the whole-cell Na⁺ current.

Voltage dependence of activation and inactivation of somatic Na⁺ currents

We have analysed the biophysical properties of the somatic Na⁺ current to help assign them to their cloned counterparts. Whole-cell Na⁺ currents in inverse Na⁺ gradient were typically elicited by depolarizing steps from a holding potential of −80 mV preceded by a short prepulse to −100 mV to remove fast inactivation (Fig. 8A, left panel). The voltage dependence of activation was measured by plotting the normalized conductance against membrane potential and fitting the curve with the Boltzmann equation (Fig. 8B). The membrane potential at half-maximal activation (V_0.5) was −23.4 ± 0.4 mV and the slope (k) of the curve was estimated to be 7 ± 0.5 mV (n = 11; DIV 7–15). The voltage dependence of fast inactivation was further investigated by holding the cells at prepulse potentials between −110 and −35 mV before stepping to the test potential (typically +20 mV) (Fig. 8A, right panel). Relatively short preconditioning pulses of 100 ms were used in this protocol to prevent entry of Na⁺ channels into slow inactivation (see below). Data fitted to a single Boltzmann function gave a midpoint of fast inactivation of −61.8 ± 0.2 mV and a slope factor of 5.5 ± 0.2 mV (n = 11) (Fig. 8B). Note that when cells at DIV 7–9 or DIV 12–15 were analysed separately, V_0.5 values for activation and fast inactivation were not significantly different (activation: V_0.5=−21.5 ± 0.5 mV versus−24.5 ± 0.6 mV; fast inactivation: V_0.5=−61 ± 0.5 mV versus−62.3 ± 0.4 mV, respectively). The same holds for the Na⁺ current amplitude (280 ± 16 pA, n = 59 and 332 ± 27 pA, n = 48, respectively; P = 0.08; test step to +20 mV).

Figure 8. Voltage dependence of the somatic Na⁺ current.

A, left panel: Na⁺ current traces elicited by depolarizing voltage steps ranging from −60 to +25 mV in 5 mV increments. Right panel: fast inactivation of Na⁺ currents was obtained using a 100 ms conditioning voltage pulse immediately followed by a 20 ms test pulse to +20 mV. The conditioning voltage was varied from −100 to −40 mV. For clarity, only currents activated by the test pulse are shown. B, normalized peak currents (I/I_max, •) and relative conductance (G/G_max, ○) were plotted against membrane potential and fitted to single Boltzmann functions. Half-activation voltage for activation and fast inactivation are indicated. Each data point is the mean ± s.e.m. of 11 CG cells.

Repriming kinetics were measured at recovery voltages from −120 to −80 mV for increasing durations following a 20 ms inactivating prepulse at +20 mV (Fig. 9A). This conditioning prepulse was found to be sufficient to produce complete fast inactivation minimizing slow inactivation. Fractional Na⁺ current recovery at −120 mV was well fitted by a monoexponential function with a time constant of 2.5 ms, whereas at −100 and −80 mV recovery was best described with both fast and slow time constants (τ_fast of 3.4 and 7 ms and τ_slow of 3 and 2 s, respectively) (Fig. 9B).

Figure 9. Recovery from fast inactivation of somatic Na⁺ currents.

A, the time course of recovery from fast inactivation was studied using the voltage protocol shown in the inset. Na⁺ channels were completely inactivated by a 20 ms prepulse to +20 mV (V_pre) and then held at a recovery potential for different durations (Δt = 1–200 ms) followed by a test step to +20 mV (V_t). For the family of current traces shown, recovery was studied at −100 mV and Δt was 1−20 ms. The sweeps are arranged so that the currents evoked by V_t are gradually shifted rightward as Δt is lengthened by 1 ms between each sweep. B, fraction of I recovered (I_t/I_pre) plotted against the interpulse duration at different recovery potentials (−120 mV, ♦; −100 mV, ○; −80 mV, ▪). The interpulse duration was lengthened by 0.2, 1 or 10 ms between sweeps. The recovery subsequent to the initial delay can be fitted by a monoexponential function at −120 mV, which gave a time constant of 2.5 ms and by double exponential functions at −100 and −80 mV, which gave fast and slow time constants of 3.4 ms and 3 s (4% of I_total) and 7 ms and 2 s (14% of I_total), respectively. Note that because calculated slow time constants are longer that the actual duration of the voltage protocol, they should be considered as tentative. Each data point represents the mean ± s.e.m. of 11–12 cells.

To determine further the extent of the slow inactivation, entry into slow inactivated states was promoted by long depolarizations (30–60 s) followed by a 50 ms step to −100 mV, during which recovery from fast inactivation should be complete (> 10 times the time constants of recovery from fast inactivation) (Fig. 10A). The Na⁺ current evoked at +20 mV following these conditioning voltage pulses was thus inversely related to the number of channels accumulated in the slow inactivated state. The normalized available Na⁺ current was plotted against conditioning potential and could be fitted by a Boltzmann function with an average midpoint of −68.4 ± 1 mV and a slope factor of 6.1 ± 1 mV (Fig. 10B). The maximal steady-state fraction of Na⁺ channels that were slow inactivated after 60 s conditioning pulses was 90%, saturating at potentials positive to −40 mV (Fig. 10B). The entry into slow inactivated states was a mono-exponential process at potentials between −100 and −40 mV and showed mild voltage dependency over this range: average time constants were 17 ± 1 s at −70 mV, 13 ± 1 s at −60 mV, 11 ± 0.8 s at −50 mV and 9.3 ± 0.8 s at −40 mV (n = 3–9) (Fig. 10C). Recovery from slow inactivation was also mono-exponential but displayed no apparent voltage dependency with time constants of 10.3 ± 1.4 s at −70 mV, 8.7 ± 1.4 s at −60 mV, 9.5 ± 1.1 s at −50 mV and 11.4 ± 1.3 s at −40 mV (n = 3–6).

Figure 10. Slow inactivation of somatic Na⁺ current.

A, time course of slow inactivation. The CG cell was held at different conditioning potentials (V_h) to promote slow inactivation and stimulated by a test sequence comprising a 50 ms step to −100 mV (to allow recovery from fast inactivation) and a 20 ms test pulse to +20 mV. The stimulus waveform was applied at 0.2 Hz. Right inset: superimposed current traces collected at conditioning potentials of −50, −60, −70 and −100 mV at the times indicated. B, normalized Na⁺ currents are plotted as a function of conditioning potentials and fitted with a Boltzmann function, yielding a mean V_0.5 value of −68.4 mV and a mean slope factor of 6.1 mV (mean ± s.e.m. of 10–16 CG cells). C, kinetics of entry into (upper panel) and recovery from (lower panel) slow inactivation. Same protocol as in A. Upper panel: the development of slow inactivation was studied from the gradual reduction of the peak Na⁺ current by stepping the cell from a holding potential of −100 to either −70 mV (•) or −50 mV (○, inset). The continuous curves are single exponential fits to time courses, giving time constants of 15.5 and 10.2 s, respectively. Lower panel: the time course for recovery from slow inactivation at −100 mV from 60 s prepulse to −50 mV (○, inset) or −70 mV (•). Fitting gave time constants of 9.1 s and 8.9 s, respectively.

Thus, the slow rate of Na⁺ channel entry into slow inactivated states, together with the extremely rapid rate of recovery from fast inactivation would allow most Na⁺ channels to recover rapidly at very negative potentials. This accounted for the absence of use dependence (e.g. cumulative inactivation) we observed when stimulating CG cells with short depolarizations at frequencies ≥ 30 Hz (data not shown).

Detection of a persistent Na⁺ current

Based on current clamp studies, a persistent inward current (I_NaP) has been hypothesized in CG cells (D'Angelo et al. 1998). In a first series of experiments, we sought evidence for the presence of persistent Na⁺ currents under conditions of inverse Na⁺ gradients using slow depolarizing voltage ramps rising at a rate of 43 mV s⁻¹ (3 s duration). These slow ramps were selected because they usually allowed full inactivation of the fast-inactivating Na⁺ current described above. Figure 11A shows currents evoked by such protocols in a representative CG cell, both in control conditions and in the presence of TTX (1 μm). Isolation of TTX-sensitive Na⁺ currents by offline digital subtraction showed no persistent components, consistent with the lack of I_NaP using the pulse protocol (cf. Figs 6B and 8A).

Figure 11. Detection of a ‘persistent’ Na⁺ current in normal Na⁺ gradient.

A and B, currents elicited by a 3 s ramp depolarization from −100 to +30 mV (rising rate, 43 mV s⁻¹) in inverse Na⁺ (A) or in normal Na⁺ (B) gradients and in the absence or presence of TTX (1 μm, general superfusion). Signal averaged from 3 sweeps and plotted as a function of potential. Subtraction of the currents in the presence of TTX from those in controls (difference current) isolates the persistent inward current (I_NaP). Note that I_NaP is not detected in inverse Na⁺ gradient. C, 500 ms voltage step to −35 mV elicited a fast inactivating Na⁺ current followed by a ‘steady-state’ Na⁺ current (signal-averaged from 20 sweeps). Both components are blocked by bath-applied TTX. Inset: the current traces are expanded in amplitude to highlight the persistent component. D, conductance–potential relationship of I_NaP constructed from averaging I–V curves from 3 different cells. Data are fitted by a single Boltzmann function with a half-activation voltage of –41 mV and a slope factor of 4 mV. E, effect of local application of TTX to the soma (illustration) on transient (upper panel) and persistent (lower panel) Na⁺ currents recorded in normal Na⁺ gradient. Data shown are averaged current traces obtained in 5 different CG cells, in which transient and persistent components of the Na⁺ current could be isolated. Na⁺ currents were evoked by 20 ms voltage steps from −100 to −20 mV or using slow voltage ramps from −100 to +30 mV and subtracted from TTX-insensitive currents by bath-applying TTX at the end of the experiments. Note that the persistent currents seen using step or ramp protocols were unchanged by TTX application to the soma, whereas the transient Na⁺ current component was half-reduced.

Ramp protocols were also used in experiments under conditions of standard Na⁺ gradients. Figure 11B illustrates current traces from a CG cell obtained by slow voltage ramps prior to and upon superfusion of TTX. TTX subtraction revealed a persistent current with a threshold of activation at about −55 mV, a peak at −35 mV and a mean amplitude measured at the peak of the I–V relationship of −20.6 ± 3 pA (n = 13). This persistent current was detected in most CG cells from DIV 7 to DIV 14 with amplitudes ranging from −10 to −50 pA, irrespective of the culture stage (average amplitude at −30 mV of −21 ± 4 pA at DIV 7–9 and −20 ± 5 pA at DIV 12–14, n = 6–7, P = 0.8). I_NaP could also be evoked by delivering long-lasting depolarizing pulses of more than 500 ms (Fig. 11C). The G_NaP–V relationship derived from the I_NaP–V curve was well fitted by a single Boltzmann function, giving a half-activation potential of –41 mV and a slope factor of 4 mV (Fig. 11D).

An inhomogeneous voltage clamp is known to cause a potential gradient in remote electrotonic regions of neurones, which produces steady-state axial current flow into the soma, thereby increasing the persistent current measured somatically. To test this, we determined the reversal membrane potential of kainate (250 μm)-induced currents applied locally on the neurites at increasing distance from the soma (up to 80 μm, n = 4–9). No significant deviation of the reversal potential (range from −1.8 to +2.1 mV) was observed with distance from the cell body. Although these data argue that distal neurites are adequately clamped under steady-state conditions, they do not decisively prove that the persistent Na⁺ current is well clamped.

Because the persistent current was not observed under inverse Na⁺ gradient, where Na⁺ currents reflect primarily activity of somatic channels, we examined the issue of the localization of Na⁺ channels that generate I_NaP. Local application of TTX onto the soma did not cause significant block of I_NaP (−16.4 ± 2 pA and −13.8 ± 3 pA in control and upon TTX application, respectively; P = 0.1, paired t test; n = 5), whereas it did inhibit total transient Na⁺ current by 52 ± 5% in the same cells (Fig. 11E). In all these cells, TTX (1 μm) applied at the end of the experiment via the general perfusion system (i.e. superfusing both soma and neurites) fully abolished I_NaP suggesting that the persistent Na⁺ current results from channels mostly located in cell processes.

Discussion

In this study, we have characterized the properties of Na⁺ currents in cultured CG cells and correlated these data with the subcellular location of different Na⁺ channel isoforms.

Differential distribution of Na⁺ channel subunits in cultured CG cells

From day 7 onwards, CG cells in our culture system had developed polarity as reflected by the segregation of axonal- and dendritic-specific cytoskeletal markers to morphologically distinct compartments. The formation of AIS was demonstrated by clustering of ankyrin-G, a prerequisite for the correct targeting and retention of multiple proteins at the AIS (Zhou et al. 1998; Bennett & Chen, 2001; Jenkins & Bennett, 2001). Using dual staining with a PanNav antibody, we showed that Na⁺ channels accumulate at the AIS, showing colocalization with ankyrin-G domains. It should be noted that immunoreactivity for PanNav was clearly observed only at sites of relatively high Na⁺ channel density, namely the AIS and the cytoplasm of the cell body, where they may represent newly synthesized pools of channels. Hence, though the amount of functional channel protein is likely to be related to the amount of cytoplasmic channel protein, Na⁺ channel immunodetection should not be taken as reflecting channels inserted into the plasma membrane, but rather subcellular regions of intense synthesis/trafficking or with limited diffusion (e.g. the AIS).

In line with immunodetection, Nav1.2 and Nav1.6 mRNAs were detected in cerebellar cultures by RT-PCR. Though study of cultured cells may not be indicative of the in vivo condition, our results are generally consistent with previous immunohistochemistry and in situ hybridization studies on rat cerebellum (Westenbroek et al. 1989; Felts et al. 1997; Schaller & Caldwell, 2000), indicating that the expression of Nav subunits in cultured CG cells provides a reasonable qualitative match to that in situ. Although Nav1.2 and Nav1.6 are simultaneously expressed in cultured CG cells, they show some differences in their subcellular expression pattern. Thus, at DIV 7–8, Nav1.2 was concentrated at AIS and colocalized with ankyrin-G and PanNav, suggesting that the high density of Na⁺ channels at these sites is mainly due to Nav1.2 subunits. At this stage, Nav1.6 was not confined to AIS but diffusely distributed throughout the cell body, dendrites and axons; accumulation of Nav1.6 at most AIS was evident by DIV 12. The rearrangement of Nav1.6 at AIS does not seem to result from an overall increase in expression since the appearance of Nav1.6 clusters was not paralleled by significant changes in maximum current amplitudes (−605 ± 35 pA at DIV 7–10, n = 46 versus−570 ± 55 pA at DIV 12–15, n = 13; P = 0.6) but rather reflects the new ability of initial segments to sequester Nav1.6. This sequential distribution in which Nav1.2 precedes Nav1.6 at AIS resembles the developmental expression of these subunits at AIS of Purkinje neurones and retinal ganglion cells (Jenkins & Bennett, 2001; Boiko et al. 2003) and in nodes of Ranvier (Boiko et al. 2001; Kaplan et al. 2001). Moreover, Nav1.6 accumulation was not associated with a reduction in Nav1.2 as we did not notice any appreciable change in the number of Nav1.2 clusters; however, whether the two subunits are spatially segregated within the initial segment could not be directly tested due to the lack of appropriate Na⁺ channel antibodies.

The differential distribution of Nav1.2 and Nav1.6 subunits also implies that within the non-conserved structural domains of these two highly related proteins are signals that direct channels to their specific subcellular compartments. Ankyrin-G is clearly required for clustering of Na⁺ channels at AIS, as Na⁺ channels failed to accumulate at AIS of ankyrin-G knockout mice (Zhou et al. 1998). Recently, it has been shown that the cytoplasmic loop connecting domains II and III (AIS motif) is an important determinant conferring compartmentalization of Nav1.2 at the AIS in rat hippocampal neurones (Garrido et al. 2003). Given that the AIS motif identified in Nav1.2 is highly conserved in Nav1.6, our findings suggest that the AIS signal may be necessary but not sufficient to localize Nav1.6 at AIS. A corollary therefore is that signals for ankyrin-G-based targeting of Nav1.6 is likely to involve additional Nav1.6-interacting protein(s) downstream of ankyrin-G in the pathway to formation of the AIS specialized domain.

Properties of rapidly inactivating and persistent Na⁺ currents

The biophysical properties of the Na⁺ current could be satisfactorily investigated by using inverse Na⁺ gradients, rendering the currents non-regenerative. The biophysical signature of the rapidly inactivating TTX-sensitive Na⁺ current in inverse gradient was undistinguishable from those of Nav1.2 or Nav1.6 when coexpressed with auxiliary β subunits with normal Na⁺ gradient. With respect to the voltage dependence of activation and inactivation (fast and slow components), no major differences could be seen between the somatic Na⁺ current of CG cells and recombinant Nav1.2 or Nav1.6 or in acutely dissociated neurones (Li et al. 1992; West et al. 1992; Sarkar et al. 1995; Xie et al. 1995, 2001; Smith et al. 1998; Toib et al. 1998; Herzog et al. 2003). Likewise, the somatic current has fast repriming kinetics that ranged from 2 to 7 ms between −120 and −80 mV, which are comparable to those of Nav1.2 or Nav1.6 but are 2- to 10-fold faster than those of Nav1.3 and Nav1.7 (Cummins et al. 2001; Herzog et al. 2003). Taken together, our results suggest that the somatic Na⁺ current is dominated by fast gating channels, consistent with the contribution of Nav1.2 and/or Nav1.6 subunits, though we were unable to decipher the specific roles played by each subtype in the soma.

In addition to fast gating Na⁺ channels that play a critical role in transmitting high frequency action potentials, we observed that CG cells also exhibit a non-inactivating steady-state (‘persistent’) Na⁺ current. An important property of this current is its activation at voltages near the resting potential, suggesting a possible role in determining action potential threshold and in maintaining repetitive activity. Even though the window current of Na⁺ channels (see Fig. 8B) does appear to be capable of making a modest contribution to the persistent current at negative potentials, it is minute and its maximum amplitude would reach only 0.14% of the total Na⁺ conductance (as estimated from inverted Na⁺ currents), which is clearly incompatible with the size of the persistent current. Importantly, the persistent current was not observed in inverse Na⁺ gradient, where Na⁺ currents are thought to reflect activity of somatic channels, nor was it blocked by TTX applied onto the soma in normal Na⁺ gradient, suggesting that ‘persistent’ channels might be localized in the cell processes.

This interpretation, however, may be biased by poor neuritic voltage control, which can lead to the appearance of pseudo-persistent currents in the cell body (White et al. 1995). We have shown that CG cells are electronically compact at steady state (e.g. voltage can be controlled in remote neuritic segments), which makes them amenable to relatively ‘uniform’ space clamp. Naturally, the situation varies considerably under transient and steady-state conditions; that is injection of current from an eccentrically (somatic) placed patch electrode limits the rate at which the membrane capacitance of axons can be charged, thereby limiting the voltage clamp speed. This, along with high levels of ion channel expression in axons, results in poor control of fast Na⁺ currents in axons. In addition, the kainate experiments, although indicative of good voltage control of distal portions of neurites, do not provide evidence that remote neurites are clamped adequately while voltage-dependent Na⁺ currents are flowing, raising the possibility that steady-state axial current may be the prime source of I_NaP. Taken together, our present data therefore cannot determine whether I_NaP is caused by uncontrolled window Na⁺ currents occurring in unclamped portions of the neurones or by the activity of non-inactivating channels located in remote compartments (or both).

Previous studies have provided evidence for the preferential distribution of non-inactivating Na⁺ channels in the dendrites of hippocampal (Masukawa et al. 1991) and dorsal horn (Safronov et al. 1997) neurones. The existence of a persistent Na⁺ current of non-somatic origin has just been recently suggested in granular cells of rat cerebellar slices (Magistretti et al. 2004). The fact that Nav1.6 is commonly linked to the presence of persistent Na⁺ currents in a variety of neurones (Raman et al. 1997; Maurice et al. 2001), together with our finding that Nav1.6 is the main isoform expressed in dendrites of CG cells, support the proposal that Nav1.6 may contribute to the persistent currents. However, a better definition of the Na⁺ channel subtype(s) that generates I_NaP awaits further experiments.

In conclusion, our study establishes which Nav isoforms are expressed in CG cells and defines the subcellular distribution of each subtype. It also reveals the existence of functionally compartmentalized Na⁺ currents. These specialized Na⁺ channels, which include persistent and fast gating channels, are key determinants of CG cell excitability, making them reliable in terms of following sustained high-frequency stimulation and repetitive firing. Our model system therefore should prove useful for future studies addressing the molecular basis of Nav channel clustering as well as the functions of compartmentalized Na⁺ channels that shape the excitability of CG cells.