Expression and Biophysical Characterization of Voltage-Gated Sodium Channels in Axons and Growth Cones of the Regenerating Optic Nerve

Abstract

Purpose.: Successful regeneration and re-establishment of synaptic connections in the adult central nervous system is a complex process determined by both the exterior environment and the endogenous neural activity of the regenerating growth cones. The purpose of this study was to determine the expression and properties of voltage-gated sodium channels (Na_v) expressed by regenerating growth cones.

Methods.: Na_v channels were studied in an organotypic explant culture of the adult rat retina by immunocytochemistry and whole-cell, patch-clamp recordings.

Results.: Regenerating axons and growth cones, but not glial processes, expressed Na_v channels. Whole-cell, current-clamp recordings from growth cones displayed a high input resistance of 1.29 GΩ and a resting membrane potential of −69.0 mV. All growth cones responded to depolarizing voltage steps with fast, transient, inward currents mediated by Na⁺ ions, followed by slow, sustained outward K⁺ currents. Half-maximum activation clustered in two groups, suggesting the presence of at least two Na_v channel isoforms. Steady state inactivation and recovery from fast inactivation were characterized by a half-maximum value of −69.7 mV and by a time constant of 3.64 ms, respectively. Injection of depolarizing current steps larger than threshold (−29.3 mV) consistently induced a single action potential, whereas ganglion cell bodies responded to above-threshold stimulation with a series of fast, all-or-none action potentials.

Conclusions.: These experiments describe for the first time the biophysical properties of Na_v channels recorded from the growth cones of regenerating retinal ganglion cells and contrast their properties with those of adult retinal ganglion cells.

Regeneration of retinal ganglion cell axons after injury to the optic nerve depends critically on the environment through which the advancing growth cones pass. Interaction with extracellular cues is mediated by a variety of receptors expressed on the growth cone membrane that exert their effects by triggering intracellular signal transduction cascades. ^1,2 The increase in the concentration of intracellular Ca²⁺, mediated either by its release from internal stores or by influx through calcium-permeable ion channels from the extracellular side, plays a key role in growth cone motility and axon guidance. ³ In addition to Ca²⁺ channels, a role for voltage-gated K⁺ channels in the outgrowth of retinal ganglion cell axons has been reported. ⁴ However, the establishment of correct and functional synaptic connections in the target area is critically dependent on electrical activity, as represented by the generation and conduction of action potentials. In the visual system, Na⁺-dependent action potentials are crucial for the appropriate synaptic targeting of retinal ganglion cell axons, ⁵ whereas disruption of electrical activity with tetrodotoxin does not affect the process of axon extension and pathfinding. ^6,7  

Voltage-dependent ion channels are gated by changes in the transmembrane potential, and their biophysical and physiological properties determine the response of the growth cones to a variety of stimuli. Therefore, cation-permeable membrane channels are of fundamental importance in mediating and regulating the outgrowth and guidance of growth cones and in forming synapses in the target area. Although several studies have been performed in invertebrates, ^8,9 relatively little is known about the electrophysiological properties and ion fluxes across growth cone membranes in the mammalian nervous system. The presence of low- and high-voltage–activated Ca²⁺ channels has been described, with Ca²⁺ channels likely to be organized in local hot spots. ^10–13 Voltage-dependent Na⁺ (Na_v) channels are expressed in growth cones of neuroblastoma cells, ¹⁴ dorsal root precursor cells, ¹² and nerve growth factor–induced PC12 cells. ¹⁵ In the vertebrate visual system, expression of the voltage-gated K⁺ channel isoform Kv4.3 has been reported in Xenopus retinal ganglion cells, and blockade of this channel with 4-aminopyridine inhibits axonal growth. ⁴ Other isoforms of Kv channels have subsequently been detected with expression profiles, depending on the state of ganglion cell differentiation. ¹⁶ However, the properties of Na_v channels of growth cones in the regenerating visual system are largely unknown. In this study, we examined the presence and the biophysical characteristics of voltage-gated Na⁺ channels in the regenerating rat retina, and we compared the properties of action potentials generated in the growth cones with those triggered in retinal ganglion cell bodies. 

Methods

Optic Nerve Crush

Male and female adult Sprague-Dawley rats were obtained from the local colony at the university, where their care and maintenance conformed to the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. 

Sprague-Dawley rats (150–250 g body weight, 12–16 weeks after birth) were anesthetized by intraperitoneal injections of ketamine (60–80 mg/kg) and xylazine (10–15 mg/kg), and a 1- to 1.5-cm-long incision was made in the skin above the right orbit. The optic nerve was surgically exposed in the orbit under an operating microscope. The dura was opened longitudinally, and the nerve was crushed 1 mm behind the eye for 10 seconds with a jeweler's forceps while avoiding injury to the central artery. Optic nerve crush was confirmed by the appearance of clearing at the lesion site. The vascular integrity of the retina was verified by funduscopic examination. Optic nerve crush was followed by cataractogenic lens injury, which prevented traumatic ganglion cell death and promoted axonal regeneration. ¹⁷  

Organotypic Retinal Explants

Retinal explants were obtained from rats 4 days after the animals were subjected to optic nerve crush. The eyes were rinsed in fresh, sterile Hanks' balanced salt solution (HBSS) oxygenated with pure oxygen for 10 minutes before explantation, until saturation was reached, as monitored by colorization of the solution. Oxygenation of all solutions and culture medium was essential, as our preliminary experiments and previous experience with rat retinas ¹⁸ showed reproducible axonal growth only in an oxygen-rich atmosphere at pH 7.4. These conditions were established by using an incubator, which was manufactured to work with 5% CO₂ and 55% O₂ at 37°C (HeraCell 240; Kendro, Langenselbold, Germany). 

The retinas were dissected from the eye cup by removing all surrounding tissues, flattened, and subsequently cut into wedges centered on the optic nerve head. The explants were cultured with the ganglion cell side down on hydrophilic dishes (Petriperm; In Vitro Systems, Göttingen, Germany). The dishes were washed carefully with HBSS, coated overnight with poly-d-lysine at 37°C (PDL; 200 μg/mL in HBSS, 200–400 kDa) (Sigma, München, Germany), washed three times with HBSS, and coated with the second substrate for at least 1 hour at 37°C before they finally received the explants. The second substrate contained laminin-1 derived from EHS-sarcoma cells (20 μg/mL in HBSS; Boehringer Mannheim, Mannheim, Germany). The culture medium was prepared according to the chemically defined S-4 medium ¹⁸ and was routinely used without any neurotrophic factors, serum components, or other additives, except for the antibiotics penicillin and streptomycin. The medium was changed every third day after explantation. 

Explant cultures were also obtained from control animals that were not subjected to an optic nerve crush and targeted lens injury before explantation. In these animals, considerably fewer ganglion cells survived after 7 days in vitro, and their capacity to regenerate axons was significantly reduced. 

Preparation of Slices

Slices were obtained from the retinas of adult rats (12–16 weeks after birth) that were not subjected to optic nerve crush. Briefly, each retina was removed from the eye cup and cut in half. Each half was trimmed with a scalpel blade by removing peripheral tissue, and it was subsequently embedded in 2% agarose dissolved in Ames medium (Sigma), which had been equilibrated with a mixture of 95% O₂ and 5% CO₂ before use. After polymerization of the agarose, the embedded tissue was transferred to the stage of a vibratome (VT1200S; Leica Microsystems, Wetzlar, Germany) and cut into 200-μm vertical sections. The sections were collected and stored in equilibrated Ames medium at room temperature. 

Immunohistochemistry

Organotypic cultures were inspected by videomicroscopy before fixation. Moving growth cones indicated that the cultures were viable and healthy. Retinal explants were removed from the culture dishes, embedded in optimum temperature cutting compound (Tissue Tek; Sakura Finetek, Tokyo, Japan), and frozen in liquid nitrogen. They were used either as wholemounts to stain axons and growth cones or as cryosections (10–12 μm thick) mounted on gelatinized glass slides to stain the retinal cells. Explant wholemounts or sections were fixed in ice-cold methanol for 10 minutes, washed three times for 5 minutes each in PBS (pH 7.4), and blocked with 10% fetal calf serum (FCS) for 30 minutes. After removal of the blocking solution, the sections or explants were incubated with the primary antibody in a humidified chamber at 4°C overnight. After rinses in PBS three times for 5 minutes each, the sections or explants were incubated with the secondary antibody for 1 hour at room temperature. Sections and explants were rinsed in PBS three times for 5 minutes each and coverslipped with antifade Mowiol containing 10 mg/mL Hoechst 33258, to label the nuclei. Explants and sections were then viewed with a conventional fluorescence microscope equipped with the appropriate filters (Axiophot; Carl Zeiss Meditec, Göttingen, Germany) and documented digitally. Monoclonal anti-growth–associated protein-43 (GAP-43, clone GAP-7B10; Sigma) antibodies were diluted 1:500. Monoclonal antibodies against neuronal class III β-tubulin (Covance, Berkeley, CA) were diluted 1:400. Rabbit anti-glial fibrillary acidic protein (GFAP; Sigma) antibodies were diluted 1:1000. Polyclonal anti-pan Na_v (SP19; Alomone Laboratories, Jerusalem, Israel) antibodies were diluted 1:200. The secondary antibodies were conjugated to Cy2 or Cy3 for double-fluorescence labeling. Control slides were treated only with the secondary antibodies. Sections of control and experimental explants were stained simultaneously, to avoid variations in the immunohistochemical staining. 

Patch-Clamp Recordings

Explant cultures and vertical slices were placed in a recording chamber (Luigs and Neumann, Ratingen, Germany) on the stage of an upright microscope (Leica Microsystems). Growth cones were visually identified by using 40× and 63× water-immersion objectives equipped with differential interference contrast optics (Leica Microsystems). Whole-cell voltage-clamp and current-clamp recordings were performed with a patch-clamp amplifier (EPC9; Heka, Lambrecht, Germany). Current traces were monitored with a digital oscilloscope (Tektronix, Beaverton, OR) and directly stored to the hard disc of a computer. The data were acquired with a sample frequency of 10 kHz in voltage-clamp experiments and with a frequency of 20 kHz in current-clamp experiments (filtering frequency, 2.7 kHz). Leak currents were subtracted before sodium current measurements from all current traces by using a P/4 protocol. Patch pipettes were pulled from borosilicate glass (1.5-mm outer diameter, 1.275-mm inner diameter; Hilgenberg, Malsfeld, Germany) with a horizontal electrode puller (Sutter, Novato, CA). Electrodes with a resistance ranging from 7 to 10 MΩ were connected to the amplifier with an Ag/AgCl wire. The electrodes were coated as far as possible to their tips with a silicone elastomer (Sylgard; Dow Corning, Midland, MI) to reduce stray capacitance and optimize the fast and slow compensation procedures of the amplifier. The electrode holder combined with the headstage was mounted on a mechanical, remote-controlled device attached to a three-dimensional micromanipulator (Luigs and Neumann). In whole-cell experiments, the series resistance of the electrodes usually ranged between 8 and 15 MΩ and could be compensated for by up to 80%. In addition, the series resistance was carefully monitored in the time course of an experiment, and only those recordings with a stable series resistance were considered for analysis. Drugs were applied to the preparation in the extracellular bath solution by a pressure-driven application system (DAD-12 Superfusion System; ALA Scientific Instruments, Westbury, NY). 

Explanted retinal cultures were continuously superfused (0.5 mL/min) at room temperature with an extracellular solution containing (in mM): 137 NaCl, 5.4 KCl, 1.8 CaCl₂, 1 MgCl₂, 5 HEPES, and 10 glucose (pH 7.4). The intracellular solution for recordings of whole-cell currents contained (in mM): 140 KCl, 1 CaCl₂, 2 MgCl₂, 11 EGTA, and 10 HEPES (pH 7.2). When blockage of voltage-gated potassium channels was required, KCl was replaced by 120 mM CsCl and 20 mM tetraethylammonium chloride. All biochemicals were obtained from Sigma unless otherwise noted. A stock solution of tetrodotoxin (TTX; Alomone Laboratories) was prepared in distilled water and stored at −20°C. 

Results

Immunohistochemistry of Explants, Regenerating Axons, and Growth Cones

Outgrowth of ganglion cell axons started as early as 1 day after explantation of the retina. Processes originating from different ganglion cells crossed each other rather frequently, whereas individual processes rarely branched. A few of them stopped abruptly as blunt ends, but the majority extended growth cones of typical size and shape (see Figs. 1C, 3A). The neuron-specific class III β-tubulin subunit of microtubules is localized throughout axonal processes into the growth cones of regenerating neurons, ¹⁹ and growth-associated protein (GAP)-43 is a well-known marker of ganglion cell axons. ²⁰ Therefore, monoclonal antibodies directed against both proteins were used to identify axonal processes and growth cones of retinal ganglion cells in explant cultures. 

In contrast to the processes of retinal glial cells, axonal processes were intensely labeled with antibodies against class III β-tubulin and GAP-43. It appeared that both antigens were stained within explant sections (Figs. 1A, 1B). Class III β-tubulin labeling was positive within the ganglion cell layer and the inner plexiform layer (IPL), showing that the lamination of the retina was well-preserved in culture (Fig. 1A). GAP-43 was expressed within the ganglion cell layer and the axons in sections of explant cultures (Fig. 1B), thereby outlining growth cones and their filopodial processes (Fig. 1C). 

Figure 1.

 

Immunohistochemical detection of the neuron-specific markers class III β-tubulin and GAP-43. (A) Section through a retinal explant stained with class III β-tubulin antibodies, which in the retina specifically label the ganglion cell layer and ganglion cell dendrites in the IPL. Blue: nuclear counterstain. (B) Section through a retinal explant shows the GAP-43 staining of ganglion cells (orientation as in A). (C) GAP-43 staining of regenerating axons and growth cones indicates that the processes formed on the polylysine-laminin-1 substrate are ganglion cell axons. GCL, ganglion cell layer; INL, inner nuclear layer; ONL, outer nuclear layer.

To investigate whether Na_v channels are localized along regenerating axons and growth cones, we performed immunohistochemistry on retinal explant cultures. In double-labeling experiments, fibers negative for the glial marker GFAP were stained by anti-pan Na_v antibodies (Figs. 2A–C), indicating expression of Na_v channels in the neuronal processes. Likewise, double staining with axon-specific anti-GAP-43 and anti-pan Na_v showed that some axons and growth cones expressed Na_v channels (Figs. 2D–F). Double staining with anti-class III β-tubulin and anti-pan Na_v displayed strong expression of Na_v channels along the axons and the growth cones (Figs. 2G–I). 

Figure 2.

 

Immunohistochemical detection of voltage-gated Na channels. (A–C) Double staining of processes with an antibody against GFAP, a glia-specific marker (A, green) and a pan-Na channel antibody that labels axons (B, red). The merged image in (C) shows that processes negative for GFAP were labeled with Na channel antibodies. (D–F) Double staining with an antibody against GAP-43 (D) and a pan-Na channel antibody (E) shows that some axons and growth cones expressed Na channels. (F) Overlay of the red and green channels to show co-localization. (G–I) A growth cone double stained with class III β-tubulin (G) or pan-Na channel antibody (H) and a merged image (I) showing expression of channels along the axon and the growth cone.

Membrane Time Constant and Input Resistance

Explant cultures of the adult rat retina were kept in vitro for at least 4 days until commencement of the recordings, performed as shown in Figure 3A. Since growth cones are highly fragile structures, they tended to collapse when disturbed mechanically with an electrode, and stable whole-cell recordings were therefore extremely difficult to establish. Successful recordings were obtained from 27 growth cones. The resting membrane potential of the growth cones ranged from −65 to −73 mV (mean ± SD: −69.0 ± 2.3 mV, n = 9). Membrane time constants were determined in the current-clamp mode by hyperpolarizing the membrane with a current step (−20 pA, 400 ms) and fitting the trajectory of the charging curve after onset and offset of the current pulse with exponential functions (Figs. 3B, 3C). Current amplitude and duration were chosen to minimize activation of the hyperpolarization-activated cation current and to allow the membrane potential to reach steady state, respectively. The passive voltage response of the membrane after onset and termination of the current step was approximated with second-order exponential functions (Fig. 3C). Comparing fast and slow time constants from the respective fits did not reveal statistically significant differences (P < 0.01, paired t-test). The fast time constant τ₁ contributed 27% of the overall charging curve, whereas τ₂ contributed 73%. The values obtained from the fits are summarized in Table 1. 

Figure 3.

 

Membrane time constants of retinal growth cones. (A) Differential interference contrast image of a growth cone with a patch-clamp electrode attached. Axons crossed each other frequently (arrowhead) and showed occasional branching points (arrow). Scale bar, 20 μm. (B) Voltage response of a growth cone hyperpolarized by a negative current step (−20 pA, 400-ms duration). The stimulus protocol is indicated above the voltage trace. (C) The passive membrane responses induced by onset and offset of the hyperpolarizing current were fitted each with a second-order exponential function. Gray lines: the fits superimposed on the current trajectory. Fit parameters obtained for the response of this growth cone to onset and termination of the current pulse were τ₁ = 9.06 ms and τ₂ = 29.61 ms, and τ₁ = 8.47 ms and τ₂ = 28.27 ms, respectively.

Table 1.

 

Membrane Properties of Retinal Growth Cones

Resting membrane potential, mV	−69.0 ± 2.3
Input resistance, GΩ	1.29 ± 0.11
Time Constant Measurements	Segment 1	Segment 2
τ1, ms	6.98 ± 6.27	6.98 ± 2.91
τ2, ms	40.79 ± 22.27	35.19 ± 9.18
A1, mV	10.81 ± 7.97	−10.68 ± 3.29
A2, mV	29.93 ± 10.05	−27.93 ± 6.24

 Data are mean ± SD for 11 growth cones. Segments 1 and 2 represent the voltage response after onset and termination of the hyperpolarizing current step, respectively, which were fit separately with second-order exponential functions (Fig. 3C). A _i characterizes the coefficients derived fro the i ^th order exponential fit to the charging curves, reflecting the contribution of each time constant τ _i to the overall charging curve.

The input resistance of growth cone membranes was determined in the current-clamp mode by measuring the voltage responses to hyperpolarizing current steps (Fig. 4A). Growth cones responded almost linearly to current steps less than or equal to −80 pA, and they showed only a small decrease in apparent input resistance during the hyperpolarization (Fig. 4A). However, negative currents larger than or equal to −60 pA induced a significant amount of sag in the voltage response, indicating a decrease in input resistance due to the delayed opening of the ion channels (Fig. 4A). This behavior suggests the presence of the hyperpolarization-activated cation current I _h, which has been shown to be expressed in retinal ganglion cell bodies. ^21,22 In addition, the termination of larger hyperpolarizing current steps frequently triggered spikes in most growth cones, indicating a rebound depolarization arising from activation of I _h (data not shown). Thus, expression of I _h most likely is not restricted to the soma of ganglion cells but also extends to regenerating axons. The input resistance of growth cone membranes was high (mean ± SD: 1.29 ± 0.11 GΩ, R = 0.99,275, n = 6), as measured from the slope of the fit to the linear portion of the current-voltage relation (Fig. 4B). 

Figure 4.

 

Membrane input resistance of retinal growth cones. (A) Voltage response of a growth cone to hyperpolarizing current injections ranging from −20 to −120 pA (−20 pA decrement, 400-ms duration). The current stimulus is indicated above the voltage traces. (B) The input resistance was determined by plotting the mean change of the membrane voltage (ΔV) of all growth cones (± SD, n = 6) versus the injected hyperpolarizing current. Initial voltage changes were obtained by fitting the response to a first-order exponential function, whereas the steady state voltage response was determined manually. The regression line was fitted to the linear portion of the plot at small current pulses, and it did not differ significantly between initial and steady state values (shown is the fit to initial values).

Activation Properties of Voltage-Gated Na⁺ Currents

Voltage-gated Na⁺ and K⁺ currents were elicited by depolarizing the growth cone membrane in the voltage-clamp mode from a holding potential of −70 mV to voltages ranging from −60 to 80 mV (Fig. 5A). This protocol induced transient inward currents with highly variable amplitudes ranging between −164 and −1107 pA (mean ± SD: −394 ± 353 pA, n = 7) at a −20-mV step potential. These currents were completely abolished in the presence of 1 μM tetrodotoxin, indicating that they were mediated by Na_v channels (data not shown). With physiological intracellular K⁺ concentrations, we always observed sustained outward currents, which could be blocked by either 120 mM intracellular Cs⁺ or 40 mM extracellular tetraethylammonium, suggesting activation of voltage-gated K⁺ channels (data not shown). Their amplitudes were much less variable and measured, on average, 820 ± 94 pA (n = 6) at 80 mV. Voltage-gated Na⁺ channels were activated at approximately −50 mV and reached their peak amplitudes between −30 and −20 mV (Fig. 5B). At positive step potentials, the current-voltage relations of both the Na⁺ and K⁺ channels were approximately linear (Fig. 5B). 

Figure 5.

 

Voltage-gated Na⁺ and K⁺ currents. (A) Depolarization of a growth cone with voltage steps ranging from −60 to 80 mV (10-mV increments, 45-ms duration) induced fast transient inward Na⁺ currents followed by sustained outward K⁺ currents. (B) The current–voltage relation of peak Na⁺ and K⁺ currents shown in (A). (C) The dependence of Na⁺ current amplitude on the holding potential. Current traces were evoked by depolarizing the membrane from −70 and −90 mV to various step potentials. Shown is the step to −20 mV. (D) The current–voltage relation of the growth cone response shown in (C) over the entire voltage range. (B, D, dotted lines) fits obtained with the Goldman-Hodgkin-Katz equation. (E) Activation curves of seven growth cones showing variability of half-maximum voltage and slope. The normalized conductances (g/g _max) of individual growth cones were fitted with Boltzmann distributions (solid lines). (F) Histograms of half-maximum voltage of activation (E _h) and slope (k) of the Boltzmann function shown in (E). Bin sizes are 1.9 and 0.71 for E _h and k, respectively, and the number of classes is six for both parameters.

In 69% (4/13) of the growth cones, the peak Na⁺ current amplitudes depended on the holding potential. When the membrane was clamped at −90 mV, the average current response to a depolarizing voltage step to −20 mV was −444 ± 302 pA, compared with −204 ± 45 pA at −70-mV holding potential (Fig. 5C). These values were significantly different (P < 0.05, paired t-test). Their average maximum current density was 101.6 ± 42.7 pA/pF. A similar difference was observed over the entire voltage range, although it was most prominent at the nonlinear, declining phase of the current-voltage relation, when Na⁺ channels were activated. The Na⁺ currents of the remaining four growth cones were independent of the holding potential, but displayed larger current amplitudes (−861 ± 290 pA). The mean current density of these growth cones was −232.3 ± 56.1 pA/pF, which was significantly different from the current density of the first group (P < 0.01, t-test). Individual growth cones belonging to either group did not show any obvious morphologic differences. These results suggest expression of Na⁺ channel isoforms, which differ in their inactivation properties. The highly variable current amplitudes are likely to result from the observed differences in channel density and/or single-channel conductances. The current density of retinal ganglion cells was −455.7 ± 136.3 pA/pF (n = 7), and thus it was significantly higher than the average current density of the growth cones. 

Voltage-dependent activation of growth cone Na⁺ channels was determined by calculating the conductance at voltages ranging from −60 to 0 mV. Because of the high variability between individual growth cones, the conductances for each growth cone were normalized to the maximum value and plotted against voltage (Fig. 5E). Boltzmann distributions were fitted to each data set, resulting in activation curves that differed significantly between individual growth cones. Half-maximum activation (E _h) occurred between −33 and −43 mV, whereas the voltage dependence of activation (k), as reflected by the slope of the fit, ranged from 1.24 to 5.53. For both parameters, individual values were not distributed uniformly across all growth cones tested, but appeared to cluster in two groups (Fig. 5F). Na⁺ channels of seven growth cones were activated at an average potential of −42.1 ± 1.6 mV, whereas the remaining six were centered at a more depolarized mean level of −34.4 ± 0.9 mV. These results were statistically significant (P < 0.01, t-test). However, E _h did not correlate with k (χ² = 0.233). The average E _h differed significantly (P < 0.01, Mann-Whitney-U test) between growth cones (−38.5 ± 4.2 mV, n = 13) and ganglion cell bodies (−55.5 ± 5.4 mV, n = 12) recorded in a retinal slice preparation, whereas the mean slopes showed no such difference (P > 0.05, Mann-Whitney U-test). 

Steady State Inactivation and Kinetics of Fast Inactivation

We measured the steady state inactivation of retinal growth cone Na⁺ channels in a two-pulse protocol. From a holding potential of −70 mV, a conditioning pulse to potentials ranging from −100 to 10 mV in 10-mV increments was applied for 300 ms, followed by a test pulse to −10 mV (Fig. 6A). The full complement of Na⁺ channels was available when the membrane was hyperpolarized to values below its resting potential. When the membrane became more and more depolarized, an increasingly smaller fraction of Na⁺ channels was activated with the step to −10 mV (Fig. 6B). For analysis, current amplitudes of each growth cone were normalized to the peak value obtained at a potential of −100 mV and plotted against the voltage of the conditioning pulse (Fig. 6C). The data were highly variable among individual growth cones, and thus each data set was fitted with a Boltzmann distribution. Half-maximum inactivation ranged from −83.5 to −59.7 mV, with an average of −69.7 ± 8.8 mV (n = 11). At their resting membrane potential of −69 mV, on average, 47.8% of the Na⁺ channels were already inactivated, although variability was highest at this voltage. Likewise, the voltage dependence of inactivation, as reflected by the slope of the Boltzmann function, varied between 5.36 and 16.45 (mean: 8.73 ± 3.16, n = 11). In contrast to the activation parameters, E _h and k of inactivation were normally distributed (P > 0.05, Shapiro-Wilk test). 

Figure 6.

 

Steady state inactivation of Na_v channels. (A) Representative Na⁺ currents elicited by a two-step protocol. Growth cone membranes were clamped at voltages ranging from −100 to 10 mV (10-mV increment, 300-ms duration). Each conditioning voltage step was followed by a test pulse to −10 mV. (B) Inward Na⁺ currents induced by the voltage step to −10 mV at higher temporal resolution (A, inset). Numbers refer to the conditioning voltage step. (C) Steady state inactivation curves of eight growth cones with high variability of half-maximum voltage and slope. Currents were normalized to the current amplitude elicited from a conditioning potential of −100 mV (I/I _max) and fitted with a Boltzmann distribution (solid lines). (D) Plot of activation and inactivation curves of a single growth cone. Fit parameters are E _h = −42.1 mV, k = 1.77 for activation (Fig. 5E, □) and E _h = −65 mV, k = 9.94 for inactivation (C, ○).

In a direct comparison of activation and inactivation properties, Figure 6D shows the normalized Boltzmann functions of a representative growth cone. Activation was much steeper than inactivation, and the voltage range for a window current, indicative of a persistent Na⁺ current, was very narrow. For the growth cone shown, less than 10% of the total Na⁺ current was activated between −50 and −40 mV, although our experiments revealed no evidence of a persistent current (data not shown). 

We also measured the kinetics of fast inactivation by depolarizing growth cone membranes, similar to the experiment shown in Figure 5A. The time constants (τ) were determined by fitting the decay phases of the current traces with a single exponential function (Fig. 7A). When plotted against the different test potentials, mean τ decreased with increasing depolarization (Fig. 7B). The measured values ranged between 0.971 ± 0.447 ms at −30 mV and 0.636 ± 0.190 ms at 30 mV. The differences were statistically significant when the values were compared at −30 mV and 30 mV (P < 0.05). 

Figure 7.

 

Fast inactivation of Na_v channels. (A) The decay phase of Na⁺ currents was fitted with a single exponential function (top). Shown are current traces elicited by depolarizations to −20 and 20 mV from a holding potential of −70 mV. Lines: the fits. To better compare fast inactivation kinetics, currents were normalized to the maximum current value (bottom). (B) Plot of the mean ± SD (n = 8) of the time constants of fast inactivation against the step potential.

Recovery from Fast Inactivation

We studied the recovery of retinal growth cone Na⁺ channels from fast inactivation by using a two-pulse protocol. The cells were voltage clamped at −100 mV, to ensure that all Na⁺ channels were in an activatable state. The subsequent depolarization to −10 mV (50-ms duration) induced a large inward Na⁺ current that showed typical fast inactivation (Fig. 8A). This first depolarization was followed by a second voltage step of the same amplitude and duration, thereby eliciting a fraction of the original Na⁺ current. To measure the rate of recovery, we varied the time interval between the two pulses (Δt) between 2 and 21 ms in 1-ms increments. The fractional recovery of each growth cone was calculated by dividing the inward current amplitude obtained by the second pulse by the corresponding Na⁺ current of the first depolarization. The mean values of eight growth cones were plotted as a function of time between the successive voltage steps and fitted with a first-order exponential function (Fig. 8B). Growth cone Na⁺ channels recovered with a τ of 3.64 ± 0.79 ms (n = 13). Since the amplitude of the second pulse was consistently smaller than that of the first conditioning pulse (0.93 ± 0.08, Δt = 21 ms), recovery did not appear to be complete within the observed time frame. Therefore, we used a different experimental paradigm to study recovery over prolonged time intervals. To minimize the number of depolarizations, we incremented the time interval Δt exponentially from 1 to 257 ms. Figure 8C shows representative Na⁺ currents elicited with this protocol. Again, fractional recovery was determined by dividing inward currents induced by the second and the first pulses and plotting the mean against the time Δt between the voltage steps (Fig. 8D). Data were fitted with a first-order exponential function, resulting in a τ of 4.51 ± 2.34 ms (n = 13). This value was not significantly different from the τ obtained with the linear increment shown in Figure 8B (P < 0.01, paired t-test). The late time course of the recovery was better approximated with a second-order exponential function (τ₁ = 2.93 ± 1.77 ms, τ₂ = 34.26 ± 24.81 ms; n = 12), although values for the fast time constant did not differ significantly (P < 0.01, paired t-test). However, we cannot exclude the possibility that growth cone Na⁺ channels recovered with a slow time constant of 34 ms in addition to the fast time constant. 

Figure 8.

 

Recovery of Na_v channels from fast inactivation. (A) Representative Na⁺ currents obtained with the stimulus protocol shown above the current trace. The interval Δt between the depolarizations to −10 mV was incremented linearly according to the equation Δt = 2 ms + (i − 1), i = 1, …, 20. (B) Plot of recovered currents normalized to peak amplitudes (I/I _max) induced by the first step to −10 mV. Data points representing the mean ± SD of eight growth cones were fitted with a first-order exponential function. (C) Recovery of Na⁺ currents evoked by a nonlinear increment of the interval between successive depolarizations (Δt = 1 ms + 2⁽ⁱ⁻¹⁾, i = 1, …, 10). (D) Data points representing the mean ± SD (n = 8) of normalized currents were fitted with a first-order exponential function. Dotted line: the fit with a second-order exponential function provided a better approximation of the late phase of the recovery (χ² = 0.00857; R ² = 0.9997 compared with χ² = 0.11781; R ² = 0.99432 with a single exponential function). Fast time constants τ, as obtained from the respective first-order fits, are depicted in (B) and (D).

Generation of Action Potentials

As described, a single action potential after rebound depolarization was frequently observed in retinal growth cones. To further study the generation and the properties of growth cone action potentials, we made recordings in the current-clamp mode. We injected positive currents of increasing amplitudes to depolarize the membrane and thereby trigger the generation of action potentials. Current steps of 10 and 20 pA elicited passive membrane responses, as determined by the product of capacitance and input resistance, whereas current steps of 30 pA and larger always induced single action potentials (Figs. 9A, 9B). According to Ohm's law, a 30-pA current generates a voltage response of ∼39 mV based on the measured input resistance of retinal growth cones (1.29 GΩ). Thus, at a resting membrane potential of −69 mV, the growth cone membrane must be depolarized to −30 mV to induce an action potential. In fact, the threshold for action potential generation was determined to be −29.29 mV, which was very close to the theoretical value. 

Figure 9.

 

Properties of action potentials in growth cones and retinal ganglion cells. (A) Injection of increasing current amplitudes induced the generation of a single action potential in the growth cones of retinal ganglion cells. Shown are three current steps of 10, 20, and 30 pA (400-ms duration) and the corresponding voltage responses of a representative growth cone. (B) Series of voltage responses evoked by currents steps ranging from 10 to 100 pA (10 pA increment) plotted at high temporal resolution. Steps of 10 and 20 pA elicited only passive responses (bottom), whereas larger current amplitudes invariably evoked a single action potential. (C) Plot of action potential amplitude versus injected current. Data were fitted with a linear regression line (slope = 0.28; R = 0.99745). (D) Plot of time-to-peak versus injected current. Data were fitted with a first-order exponential function resulting in a 1/e value of 44.86 pA (R ² = 0.99322). (C, D) Data are the mean ± SD. (E) Voltage response of a ganglion cell body to a current step of 100 pA (100-ms duration). (F) At their resting membrane potential, retinal ganglion cells displayed spontaneous generation of action potentials.

The amplitude of the growth cone action potentials depended linearly on the injected current, increasing from 54.3 mV at 30 pA to 74.4 mV at 100 pA (Fig. 9C), with an average value of 64.68 ± 8.55 mV (n = 40). The action potentials appeared rather broad, displaying a mean width at threshold of 5.01 ± 0.88 ms (n = 40). In addition, the duration of the depolarization eventually leading to threshold became shorter with higher current amplitudes. The time from the threshold to the peak of the action potential was also shortened, but its contribution to the overall decrease was less significant (Fig. 9B). We measured the time-to-peak as the duration from the beginning of the current step to the peak of the action potential. The time-to-peak decreased from 21.2 to 6 ms at 30 pA and 100 pA, respectively. This dependence was nonlinear and could be well fit with a first-order exponential function. The time-to-peak was reduced to 37% (1/e) of its original duration by injecting a current of 44.9 pA (Fig. 9D). 

To compare the action potentials of growth cones with those of retinal ganglion cells, we also performed current-clamp recordings from identified ganglion cells in a vertical slice preparation of the adult rat retina. Even steps of 10 and 20 pA almost always induced at least one or two action potentials (data not shown). When larger positive currents were injected, all ganglion cells responded to the depolarization with a series of action potentials, characterized by an initial larger spike and moderate slowing of the frequency (Fig. 9E). The threshold of spike generation was significantly lower than that in growth cones. In addition, the ganglion cell spikes were faster and displayed larger overall amplitudes. The properties of growth cone and ganglion cell action potentials are summarized in Table 2. 

Table 2.

 

Properties of Action Potentials of Retinal Growth Cones and Ganglion Cells

	Growth Cones	Ganglion Cells
Threshold, mV	−29.29 ± 8.09	−50.39 ± 2.62
Width, ms	5.01 ± 0.88	2.88 ± 0.90
Amplitude, mV	64.68 ± 8.55	79.56 ± 8.73

 Data represent the mean ± SD of 40 events (growth cones) and 169 events (ganglion cells). The width of the action potentials was measured at threshold voltage. The amplitude was determined as the voltage difference between the resting membrane potential and the peak of the action potential. All values are significantly different between growth cones and ganglion cells (P < 0.01, t-test).

When recorded for prolonged time intervals at their endogenous resting membrane potentials, growth cones did not spontaneously generate action potentials (data not shown). In contrast, even at their rather hyperpolarized resting membrane potentials (−80.4 ± 2.8 mV, n = 7), retinal ganglion cells were capable of firing spikes, most often occurring in bursts of two to three action potentials (Fig. 9F). 

Discussion

In this study, we investigated the biophysical properties of voltage-dependent Na⁺ channels expressed in growth cones of regenerating ganglion cell axons. All growth cones responded to depolarization with fast transient inward currents, representing the influx of Na⁺ through voltage-gated ion channels. The current density was sufficient to trigger a single action potential. Of interest, Na_v channels expressed by retinal ganglion cell bodies displayed different electrophysiological properties, most notably reflected by the generation of a series of action potentials on depolarizing current steps. In cultured rat dorsal root ganglion neurons, only a single action potential could be evoked in most somas, whereas most their growth cones showed repetitive firing. ²³ The biophysical properties of individual action potentials recorded from the soma and growth cones, however, were very similar to those described in this study. In addition, growth cones in both systems displayed a high input resistance and similar time constants when their voltage transients were fitted with double-exponential functions. This close correspondence between growth cones of the somatosensory and visual system suggests that the membrane biophysics of growth cones is largely independent of the neural system. 

Na_v channels were located in the growth cones and along the regenerating axons, although the distribution did not appear to be homogeneous. These axons were not myelinated, and it is tempting to speculate that membrane specializations, as described previously in the nerve fiber layer of the rat retina, ²⁴ correspond to Na_v channels. 

The importance of Na_v channels in the correct targeting of projecting axons in the developing visual system has been described. ⁵ During development, spontaneous activity of retinal ganglion cells is the most likely source of activity-dependent signals. ²⁵ Correlated firing of retinal ganglion cells has further been implicated in the correct targeting of axons to eye-specific layers within the lateral geniculate nucleus. ²⁶ In the regenerating rat retina, however, we found no evidence of spontaneous activity in the growth cones, although ganglion cell bodies showed occasional action potentials. These were most likely induced by presynaptic release of excitatory neurotransmitter, which depolarized the ganglion cell membrane beyond threshold. A corresponding activity was not measured in the growth cones. These results could indicate that action potentials generated in the cell body are not actively conducted along the length of the axon in the regenerating retina. However, spontaneous activity of ganglion cell bodies was measured in an acute slice preparation, which differs from explant cultures in the time point of optic nerve lesioning and vertical sectioning. Therefore, it is possible that the apparent firing properties of retinal ganglion cells are at least partially determined by the preparation. In addition, growth cones are apparently not equipped with the ion channels necessary for the generation of spontaneous activity. ²⁷ Instead, depolarization of the growth cone membrane elicited a single action potential, which was characterized by its relatively small amplitude and long duration at threshold. It is therefore reasonable to assume that local signals, most likely released from the surrounding and potential target tissues, rather than propagating spikes from the cell body, trigger the generation of action potentials in the growth cones. A possible candidate is glutamate, since growth cones express ionotropic glutamate receptors, which is the topic of a different study. 

Although we cannot exclude the possibility that the single action potential generated in the growth cones is retrogradely conducted to the cell body, it appears to act as a local signal. Such a brief depolarization could lead to a time- and voltage-controlled influx of Ca²⁺ through voltage-dependent Ca²⁺ channels, which are also expressed locally in the growth cone membrane. Of note, all growth cones responded to depolarizing current steps with a single action potential of comparably small amplitude, but we never observed a burst or a series of spikes. Therefore, efficient mechanisms, as reflected in the biophysical properties of the Na_v channels, must have evolved to limit the number and amplitude of the action potentials. The amplitude of action potentials is largely determined by the number of Na⁺ channels available per area of membrane (current density), the conductance of individual channels, their inactivation properties, and the kinetics of repolarization. Of these parameters, the current density was significantly higher in ganglion cell bodies. The fact that only a single action potential is generated in the growth cones could be explained by the inactivation of Na_v channels in combination with a rather depolarized threshold of −29 mV. With 43% of Na⁺ channels already inactivated at the resting membrane potential, the number of available channels at even more depolarized values, as induced by positive current injections, is apparently not sufficient to trigger yet another action potential. Recordings from developing retinal ganglion cells (P7–P17) showed a very similar discharge pattern, which was caused by the slow inactivation kinetics of immature ganglion cells. ²⁸ In contrast, recovery from inactivation was extremely fast (4 ms) in regenerating growth cones; thus, we cannot exclude the contribution of A-type or Ca²⁺-dependent K⁺ channels in the persistent inhibition of growth cone action potentials. 

The first-spike latency is likely to carry information about the stimulus in various sensory modalities, especially in the auditory and visual system. ^29,30 In retinal growth cones, the time between the onset of depolarization and the peak of the action potential is exponentially dependent on the stimulus intensity. In addition, we observed a linear relationship between action potential amplitude and stimulus intensity. It is therefore possible that both parameters serve as potential sources of information about the stimulus. Growth cones are small structures, and thus they are likely to be isopotential during whole-cell experiments. However, current could flow between the growth cone and the axon, if the action potential is not generated in the growth cone itself, but somewhere along the axon. Axons of explant cultures are up to several hundred micrometers in length, which is long enough to change the properties of the action potential, because of the nonuniformity in the voltage distribution. ³¹ Saxitoxin- and tetrodotoxin-binding studies have directly shown the presence of Na_v channels in growth cones. ^32,33 In addition, voltage-gated Ca²⁺ and K⁺ channels have been demonstrated immunocytochemically. ^4,16,34 Our study showed the presence of Na_v channels along regenerating axons and in the terminal region of growth cones. These results indicate that all major voltage-gated ion channels are expressed in growth cones, suggesting that the Na currents recorded in the growth cones are at least partially mediated by Na_v channels present in these structures. 

Activation and inactivation curves of Na_v channels showed a rather broad variability between individual growth cones. Thus, half-maximum activation ranged between −34.4 and −41.3 mV, whereas the midpoint of inactivation was even more diverse (61.6 to −83.5 mV). The sample size was too small to correlate these two parameters and make an educated guess as to the nature of the isoforms expressed. However, the biophysical parameters mentioned correspond most closely to the Na⁺ channel α subunits Na_v1.1, Na_v1.6, and Na_v1.7. ^35,36 In the mammalian retina, the Na⁺ channel α subunits Na_v1.1, Na_v1.2, and Na_v1.6 have been demonstrated recently. ³⁷ However, the presence of Na_v1.7 has not yet been reported in the mammalian retina. Of interest, the isoform Na_v1.2, predominantly localized to unmyelinated and premyelinated axons, ³⁸ has different activation and inactivation properties, ^39,40 suggesting that it is not expressed in growth cones. In addition, the different characteristics of Na⁺ currents measured in retinal ganglion cell bodies indicate the differential expression of Na_v channel α subunit isoforms in the two compartments. A detailed immunocytochemical analysis of the subcellular expression of potential Na_v channel subunits will be the focus of a different study. It is also possible that differences in intramembrane electric fields in these two cellular regions contribute to the observed differences in Na_v channel behavior. ¹⁴ However, an electrophysiological and ultrastructural comparison between neuronal growth cone and cell body membranes performed in the invertebrate Helisoma revealed no such differences. ⁹  

Our data provide for the first time a detailed description of the biophysical properties of voltage-gated Na⁺ channels expressed in retinal growth cones of the regenerating rat retina. It is important to determine in further studies whether Na_v channels of growth cones act as a local signal transducer to mediate a precise connection with appropriate target tissue. The coordination of Na_v channels with voltage-gated Ca²⁺ channels and ionotropic glutamate receptors in shaping the electrical response of growth cones to endogenous target-derived signals is of particular interest. 

Footnotes

 Supported by Deutsche Forschungsgemeinschaft Grants FE 464/3-1 and FE 464/3-2.

Footnotes

 Disclosure: A. Feigenspan, None; K. Dedek, None; K. Schlich, None; R. Weiler, None; S. Thanos, None

The authors thank Helen Seyfried and Susanne Wallenstein for excellent technical assistance and Jennifer Trümpler for reading and improving the manuscript.