|Home | About | Journals | Submit | Contact Us | Français|
Nerve cells integrate and encode complex synaptic inputs into action potential outputs through a process termed intrinsic excitability. Here we report the essential contribution of fibroblast growth factor homologous factors (FHFs), a family of voltage-gated sodium channel binding proteins, to this process. In mouse cerebellar slice recordings, wild-type and Fhf1−/− granule neurons generate sustained trains of action potentials up to high frequencies (~60 Hz), but Fhf4−/− neurons typically fire for only 100 milliseconds, and Fhf1−/−Fhf4−/− neurons often fire only once. Additionally, the voltage threshold for spike generation is 9 mV higher in Fhf1−/−Fhf4−/− neurons compared to wild-type cells. The severity of ataxia and motor weakness in mutant mice parallels the degree of intrinsic excitability deficits in mutant neurons. While density, distribution, isotype, and activation of sodium channels in Fhf1−/−Fhf4−/− neurons are similar to those of wild-type cells, channels in Fhf1−/−Fhf4−/− neurons undergo inactivation at more negative membrane potential, inactivate more rapidly, and are slower to recover from the inactivated state. Altered sodium channel physiology is sufficient to explain excitability deficits, as tested in a granule cell computer model. These findings provide a physiological understanding for spinocerebellar ataxia syndrome associated with human Fhf4 mutation and suggest a broad role for FHFs in the control of excitability throughout the central nervous system.
Neurons integrate and encode postsynaptic currents into outputs of action potentials for efficient information transmission along axons. The relationship between current input and action potential output, termed intrinsic excitability, is dependent upon the subcellular distribution and biophysical properties of voltage-gated sodium channels. Resting sodium channels concentrated at the axon initial segment sense threshold depolarization in the neural soma and open rapidly to allow sodium influx and the rise in membrane potential representing the upstroke of the action potential. Fast inactivation of sodium channels along with gated potassium efflux mediates the rapid downstroke repolarization and after-hyperpolarization of the action potential. Thereafter, several physiological events, including the recovery of sodium channels from the inactivated state, determine if and when the axon initial segment can initiate the next action potential.
While the pore-forming alpha subunit of a voltage-gated sodium channel is inherently capable of undergoing voltage-dependent activation and inactivation (Caterall, 2000), protein-protein interactions and post-translational modifications modulate the distribution and response characteristics of the channel. Sodium channel beta subunits facilitate transfer of the alpha subunit from transport vesicles to cell surface membrane, and also modulate the voltage dependence of channel activation, inactivation, and resurgent currents (Chen et al., 2002; Grieco et al., 2005; Isom et al., 1995; Patton et al., 1994; Spampanato et al., 2004). Site-specific phosphorylation of alpha subunits by protein kinase A, protein kinase C, p38 stress kinase, or Fyn tyrosine kinase provides a means for signal transduction to modulate sodium channel response characteristics (Ahern et al., 2005; Cantrell and Catterall, 2001; Wittmack et al., 2005). Recently, fibroblast growth factor homologous factors (FHFs) have been identified as another class of alpha subunit-binding proteins (Liu et al., 2001; Liu et al., 2003; Wittmack et al., 2004).
FHFs comprise a family of vertebrate proteins related in sequence and structure to fibroblast growth factors (FGFs) (Goldfarb, 2005). While FHFs are sometimes referred to by FGF nomenclature (FHF1 = FGF12, FHF2 = FGF13, FHF3 = FGF11, FHF4 = FGF14), FHFs and prototypic FGFs have unrelated functions. FGFs bind to the extracellular domain of cell surface receptor tyrosine kinases (Ornitz and Itoh, 2001), whereas FHFs are expressed in excitable cells as intracellular proteins that bind to either the protein kinase scaffold protein islet brain-2 (Olsen et al., 2003; Schoorlemmer and Goldfarb, 2001) or to the C-terminal cytoplasmic tails of sodium channel alpha subunits (Liu et al., 2001; Liu et al., 2003; Wittmack et al., 2004). Expression of the four FHF genes begins during embryogenesis in postmitotic neurons and persists through adulthood, with individual neurons expressing distinct repertoires of FHF transcripts (Hartung et al., 1997; Smallwood et al., 1996; Wang et al., 2000). Consistent with their binding to voltage-gated sodium channels, FHFs localize to nodes of Ranvier along myelinated axons (Wittmack et al., 2004). Several studies have documented effects of transfected FHFs on the function of transfected or endogenous sodium channels in immortalized neuron-like cells and in cultured hippocampal neurons. FHF expression could change sodium channel surface density and alter the voltage dependence for channel activation and inactivation (Liu et al., 2003; Lou et al., 2005; Rush et al., 2006; Wittmack et al., 2004), although substantial quantitative and qualitative differences were seen in different experimental systems.
Insights into FHF neurological function have come from mutations of the Fhf4 gene in mice and man. Inheritance of a missense mutation in the human Fhf4 (Fgf14) gene causes early onset spinocerebellar ataxia (van Swieten et al., 2003). Similarly, mice rendered null for Fhf4 by gene targeting display cerebellar ataxia and dyskinesia (Wang et al., 2002) as well as learning deficits (Wozniak et al., 2007). However, a biological function for FHFs at the cellular level and a relationship of such function to known FHF binding targets has not been confirmed.
This paper presents an analysis of FHF function through whole cell patch clamp recordings of granule neurons in brain slices and in cultures prepared from Fhf mutant versus wild-type mice. Animals with loss-of-function mutations in both Fhf1 and Fhf4 genes display several neurological deficits, including a severe cerebellar ataxia. While cerebellar architecture is normal in these animals, cerebellar granule cells show severely impaired intrinsic excitability, characterized by failure of mutant neurons to generate repetitive action potentials and elevated voltage threshold for spiking in response to inwardly injected current. Underlying the excitability deficits is the altered physiological responses of voltage-gated sodium channels. Sodium channels in mutant neurons inactivate at more negative membrane potential, and the channels undergo both a faster rate of inactivation and a slower recovery from inactivation. The modulation of sodium channel physiology by FHFs is crucial for the excitability of granule neurons in response to depolarization, as assessed by computer modeling.
A null allele for the Fhf1 gene was produced through standard gene targeting technology (Experimental Procedures and Figure 1). Resulting Fhf1−/− animals were viable, fertile, and performed similarly to wild-type mice in several motor tasks. First, Fhf1−/− animals showed a normal ability to retain their grip on a flat metal grid rotated to vertical (90°) or inverted (180 °) positions (see videos in Supplemental Fig. 1). Second, Fhf1−/− mice could maintain themselves on a narrow ledge for times comparable to wild type mice (Figure 2A). Third, in a footprint assay of walking gait, Fhf1−/− mice properly superimposed their hind paws onto sites where their forepaws were previously placed (Figure 2B). Lastly, Fhf1−/− and wild-type mice performed similarly in a quantitative measurement of forelimb grip strength (data not shown). In contrast, Fhf4−/− mice on the same strain background (C57Bl/6 x 129/Svev hybrid backcrossed two generations to 129/Svev) could not maintain their grip on an inverted grid (Supplemental Figure 1), showed shorter ledge retention time (Figure 2A), and often failed to position their hind paws properly during walking (Figure 2C, E). These deficiencies are consistent with ataxia and motor weakness, as had been previously shown for Fhf4−/− mice on a predominantly C57Bl/6 background (Wang et al., 2002).
Fhf1 and Fhf4 genes are normally expressed in partially overlapping sets of neurons in the central nervous system, including cerebellar granule neurons (Smallwood et al., 1996; Wang et al., 2000). We reasoned that mice lacking functional alleles for both of these genes could display accentuated and novel behavioral deficits in comparison to Fhf4−/− animals. Indeed, Fhf1−/−Fhf4−/− mice showed more severe ataxia and motor weakness. They could not retain themselves on a narrow ledge for any measurable time (Figure 2A), they displayed a highly abnormal gait, often placing their hind paws in front of their previously positioned forepaws (Figure 2D, E), they rapidly stumbled off a metal grid tilted to only a 45-degree diagonal (Supplemental Figure 1), and could not hold on to a rod with any measurable force when pulled backward by the tail (data not shown). Additionally, Fhf1−/−Fhf4−/− mice showed increased locomotor activity when disturbed or placed in a new environment, a phenotype not observed in mice bearing even one functional allele of either Fhf gene. This behavior could be quantitated by counting revolutions that mice performed when placed in a small circular container (Supplemental Figure 2). Overall, these experiments revealed functional requirements for FHF1 when FHF4 was absent.
Functional deficits in Fhf1−/−Fhf4−/− mice were apparent at weaning and did not noticeably progress in severity as animals aged. Furthermore, mutant animals never died of natural causes past weaning over the maximum one-year time span that they were maintained. Histological evaluation of cerebellum in a six month-old mutant mouse was unremarkable; the internal granule cell layer and Purkinje cells were well formed, and Luxol Fast blue staining of myelin in fiber tracts beneath the cortex was apparent (Figure 3A, B). Confocal immunofluorescence showed that axon initial segments of granule neurons in the cerebellum of an Fhf1−/−Fhf4−/− mouse possess sodium channel Nav1.6 (Figure 3C–H), the channel which only appears at an advanced stage of granule neuron maturation (Osorio et al., 2005). Cerebellar granule neurons from Fhf1−/−Fhf4−/− pups maintained in culture for 30 days were indistinguishable from wild-type cultured cells in terms of both gross morphology under phase contrast microscopy (Figure 3I, J) and the presence of a well-formed axon initial segment bearing voltage-gated sodium channels (Figure 3K, L) that colocalize with ankyrin G (Figure 3M, N). In aggregate, these data argue against a developmental deficit or degenerative process in Fhf mutant animals.
Native complexes of FHF2 in association with voltage-gated sodium channels have been detected in extracts from whole rat brain (Wittmack et al., 2004). In addition, we can detect complexes of FHF1 and sodium channels in extracts from mouse cerebellum (Figure 4A). To investigate whether FHF modulation of cerebellar granule neuron intrinsic excitability underlies ataxia of Fhf mutant mice, we first tested whether FHF proteins colocalizes with sodium channels at the axon initial segment of granule neurons. Since suitable antibodies for immunostaining endogenous FHF1 or FHF4 are not available, cultured granule neurons were transfected with expression vectors for FHF1a or FHF1b fused to green fluorescent protein (GFP). As shown in Figure 4B–G, FHF1a-GFP and FHF1b-GFP are detected in neuron soma and in the axon initial segment together with voltage-gated sodium channels. In order to confirm a similar localization for native FHF protein, we probed untransfected cultured hippocampal neurons with antibodies specific for FHF2, sodium channels, and ankyrin-G, which demarcates the initial segment (Kordeli et al., 1995). Strong FHF2 immunoreactivity was detectable in the neural soma and in the axon initial segment (Figure 3H–M). Immunostaining of axon initial segments with the FHF2 antibodies was suppressed by preincubation with the immunizing FHF2 peptide, but not with an unrelated peptide (data not shown), demonstrating the specificity of the staining. In aggregate, these findings show that a broad repertoire of FHF isoforms colocalize with sodium channels at the axon initial segment and are good candidates to serve as modulators of intrinsic excitability.
The intrinsic excitability of cerebellar granule neurons in juvenile (20–30 day-old) rats and mice has been studied extensively. In response to sufficient sustained inward current injection through a patch pipette, these cells respond with a continuous train of action potentials at frequencies up to 50–150 Hz (Chadderton et al., 2004; D’Angelo et al., 1998; Gall et al., 2003). Intrinsic excitability of granule cells in adult (3–12 month old) mouse brain slices was examined here by blind whole cell patch clamp in current clamp mode, applying inward current of various amplitudes for the duration of 800 milliseconds. Suitable brain slice preparations were only achieved through the application of reagents to the tissue dissection and incubation solutions preventing N-methyl-D-aspartate receptor transmission and rapid neuronal cell death (see Experimental Procedures). Patched cells were confirmed as granule neurons by their measured membrane capacitance (Cm) of 2.5–6.0 pF, lack of spontaneous firing, and their generation of inward (sodium) and outward (potassium) currents following depolarization in voltage clamp (see below).
Above a minimum amplitude of injected current, all wild-type adult mouse granule neurons generated action potentials that increased in frequency with increased current amplitude. Maximum frequencies averaged 60 Hz (Figure 5A–C), an example of which is shown in Figure 5D. The depolarization voltage threshold for spike generation was −49.6 +/− 4.9 mV, and the amount of injected current needed to achieve threshold voltage varied among cells, reflecting their varying input resistance (Rin = 1.53 +/− 0.7 GΩ) (Table 1A). Input resistance variability is in part the consequence of patch seal weakening upon whole cell break-in. The wild-type neurons also varied in their current-to-spike relationship; some cells showed graded excitability over a 2.5-fold range in injected current amplitude (Figure 5C, solid lines), while other cells gave closer to an all-or-nothing firing response (Figure 5C, dashed lines). The type of current-to-spike profile did not correlate with variations in Rin, Cm, or magnitude of voltage-gated currents, and may reflect subtle differences in channel functionality or subcellular distribution among cells.
Granule neurons in Fhf1−/− cerebella (n=4) showed normal excitability, in terms of both maximum spike frequency and voltage threshold for spike induction (Figure 5A, B, E and Table 1A). By contrast, no granule neurons in Fhf4−/− or Fhf1−/−Fhf4−/− cerebella generated sustained repetitive firing. Fhf4−/− cells (n=4) generated an average maximum of ~8 spikes, with most of these action potentials occurring within the first 100 milliseconds of current injection (Figure 5A, B, F and Table 1A). Fhf1−/−Fhf4−/− cells (n=7) were more seriously impaired, averaging only two spikes maximum, with four of the seven recorded cells (57%) generating only a single early action potential (Figure 5A, B, G and Table 1A). In addition, Fhf1−/−Fhf4−/− neurons displayed a significantly higher voltage threshold for spike generation (−40.6 +/− 7.8 mV) than their wild-type counterparts (9 mV shift) and required greater injected current to depolarize to threshold voltage (Table 1A). These findings show a correlation between the magnitude of granule cell excitability deficits and degree of ataxia among Fhf mutant genotypes.
Mutant granule cells displayed a normal repertoire of voltage-gated currents. Upon voltage-clamped depolarization from −80 to −20 mV, wild-type and Fhf1−/−Fhf4−/− cells generated virtually the same amplitude of fast activating/inactivating inward current, which is predominantly a sodium current (Table 1A). Further depolarization revealed similar amplitudes and kinetics of this inward current, followed by inactivating and non-inactivating outward potassium currents (Figure 5H, I, and Table 1A). The variable input resistance recorded among cells is insufficient to account for the consistent deficit in repetitive firing and altered voltage spike threshold in mutant cells. For example, the wild-type and Fhf1−/−Fhf4−/− cells shown in Figure 5H, I had similar voltage-gated currents, Rin (0.81 vs. 0.66 GΩ), and Cm (3.6 vs. 4.0 pF), but still differed in repetitive firing (28 vs. 1 spike maxima during current injection) and voltage threshold for spiking (−48 vs. −40 mV).
The colocalization and physical interaction of FHFs with sodium channels suggested that intrinsic excitability deficits in Fhf mutant granule neurons could reflect altered sodium channel physiology. Sodium channel response characteristics were monitored by voltage-clamp protocols using patch pipette and bath solutions that inhibit active currents other than sodium (calcium, potassium, and chloride) (Magistretti et al., 2006). Initial analyses were performed on cultured granule neurons (16–30 days in vitro), some of which afforded excellent voltage clamp of the axonal sodium conductance.
The voltage dependence of sodium channel activation was assayed by step depolarizations from a holding potential of −95 mV. Current traces and derived plots from representative wild-type and Fhf1−/−Fhf4−/− cells are shown in Figure 6A,B, and the averaged data for all cells are provided in Figure 6E and Table 1B. Fhf mutation induces a small, but insignificant, leftward shift in the voltage dependence of sodium channel activation (wild-type −33.3 +/− 4.3 mV; Fhf1−/−Fhf4−/− −36.5 +/− 7.5 mV; P > 0.2). Peak fast sodium current upon depolarization and resurgent current upon repolarization were variable among cells in both groups, and Fhf mutation did not cause detectable changes in current densities (Table 1B).
The voltage-dependence of sodium channel fast inactivation was analyzed by measuring peak sodium current induced at −15 mV subsequent to a 125 millisecond preconditioning voltage command. Current traces and derived plots from representative wild-type and Fhf1−/−Fhf4−/− cells are shown in Figure 6C,D, and the averaged data for all cells are provided in Figure 6E and Table 1B. Fhf mutation induced a highly significant, 13–14 mV hyperpolarizing shift in the voltage dependence of channel inactivation (wild-type −59.1 +/− 4.8 mV; Fhf1−/−Fhf4−/− −72.8 +/− 4.3 mV; P < 0.00003). The shift in voltage-dependent inactivation reduces the so-called window current in Fhf1−/−Fhf4−/− cells (Figure 6E), suggesting that mutant cells would have reduced persistent sodium currents. Indeed, sodium currents persisting 25 milliseconds after depolarization to −35 mV were substantially diminished among Fhf1−/−Fhf4−/− cells (wild-type −26 +/− 7 pA; Fhf1−/−Fhf4−/− −12 +/− 6 pA; P < 0.0003)(Figure 6F and Table 1B).
The hyperpolarizing shift in voltage-dependent fast inactivation of sodium channels might be expected to accelerate the rate of channel inactivation in Fhf1−/−Fhf4−/− cells. The rates of channel inactivation were calculated from the decay profile of sodium current amplitude following channel activation. As shown in Figure 7A and Table 1B, the inactivation rates of sodium channels at −15 or −5 mV were significantly faster in Fhf1−/−Fhf4−/− cells. Consequently, the sodium current peak is narrower in width for mutant cells (see example in Figure 7B).
Nav1.6 sodium channels that predominate at the axon initial segment are known to undergo rapid recovery from inactivation following restoration of strong negative membrane potential (Herzog et al., 2003), and one isoform of FHF2 (FHF2a) has been reported to slow recovery (Rush et al., 2006). We have analyzed rates of sodium channel recovery in wild-type and Fhf1−/−Fhf4−/− granule neurons by subjecting cells to two 10 millisecond intervals of depolarization to −15 mV separated by a −85 mV recovery period of variable test duration. The relative amplitude of the second induced sodium current following the recovery period indicated the fraction of channels recovered. Sodium channels in wild-type cells (example in Figure 7C) recovered from inactivation (τrecovery = 3.80 +/− 0.80 msec) significantly faster than did channels in Fhf1−/−Fhf4−/−cells (τrecovery = 5.88 +/− 0.88 msec; p < 0.001)(Figure 7E and Table 1B, example in Figure 7D).
A similar analysis of sodium channel physiology was conducted on granule neurons in adult cerebellum slice preparations. Attempts to analyze the voltage dependence of sodium channel activation were complicated by clear signs of clamp escape (Supplemental Figure 3A, B), probably due to more remote localization of sodium channels in the axon initial segments of fully mature cells. Nevertheless, some channel parameters could be reliably recorded. Voltage dependent fast inactivation displayed a −14 mV hyperpolarizing shift (Supplemental Figure 3C, D and Table 1C) and a slowed rate of recovery from inactivation (Supplemenntal Figure 3E, F and Table 1C) in Fhf1−/− Fhf4−/− cells, comparable to data recorded from cultured neurons.
The altered inactivation of sodium channels in Fhf1−/−Fhf4−/− cells seemed sufficient to account for the substantial rise in spike voltage threshold and for the failure of such cells to fire action potentials repetitively. Pipette-delivered current, acting similarly to native excitatory postsynaptic current, depolarizes cells with a time constant (Cm × Rin) on the order of 4–6 milliseconds, such that sodium channels are vulnerable to inactivation during initial current injection. In mutant cells, sodium channel inactivation is more severe, and greater current would be required to depolarize the neuron faster and further to reach a threshold voltage for spiking. Following a first spike, several factors would combine to block repetitive firing in the mutant cell. The greater pipette current required for the first spike would oppose potassium currents to raise the afterhyperpolarization potential and, together with the negative shift in voltage dependence of channel inactivation, would strongly disfavor channel recovery. Furthermore, the slower rate of recovery would further reduce channel availability, thereby suppressing subsequent action potentials.
In order to substantiate a functional relationship between altered channel physiology and impaired excitability, a granule neuron computer model was employed (Figure 8) (D’Angelo et al., 2001). Sodium channel behavior was modeled as previously described (Khaliq et al., 2003; Magistretti et al., 2006; Raman and Bean, 2001), with inactivation governed by voltage-independent drain of channels into inactivated states (see Experimental Procedures). By enhancing the entry and retention of channels in the inactivated states, the model reproduced the most evident changes in sodium currents observed in Fhf1−/−Fhf4−/− mice. In particular, appropriate changes in kinetic constants accelerated current decay kinetics reducing τdecay from 0.6 to 0.3 msec at −10 mV (Figure 8A), shifted steady-state inactivation by10 mV to the left (Figure 8B), reduced the persistent sodium current (Figure 8A), and slowed recovery from inactivation (Figure 8B). Since inactivation limits the number of channels reaching the open state without modifying the voltage-dependent processes leading to channel opening, the Fhf1−/−Fhf4−/− model did not show appreciable shift in the voltage-dependence of sodium channel activation (Figure 8B). However, since drain into the inactivated states was increased, the maximum peak conductance in the Fhf1−/−Fhf4−/− model was reduced from 900 pS to 500 pS. The reduced peak sodium conductance in the mutant model is consistent with recording data from slice preparations (Table 1C), although at odds with similar peak currents recorded in physiological buffers (Table 1A) or of sodium currents recorded from cultured granule cells (Table 1B). This discrepancy leaves open the possibility that sodium channel density in mutant neurons is somewhat increased reflecting development of compensatory mechanisms.
The model was next used to test the effects of sodium channel modification on granule cell excitability. Indeed, the changes in sodium currents were sufficient to depress spike firing in current-clamp simulations, even after normalizing the maximum sodium conductance (Figure 8C). Whereas the wild-type model fired repetitively when 10 pA current injection was simulated, the Fhf1−/−Fhf4−/− model cell did not fire at all. With increased injected current in the simulation, the Fhf1−/−Fhf4−/− model cell could generate only a first rapid-onset action potential, albeit at a higher voltage threshold (Figure 8C inset), analogous to the excitability deficits of recorded mutant neurons.
Loss of mouse Fhf4 function prevents prolonged repetitive firing of cerebellar granule neurons and induces ataxia, while the additional loss of Fhf1 function exacerbates these deficits. Granule neurons relay and refine high frequency sensory and cerebral inputs transmitted by mossy fibers and distribute this information through their parallel fibers to Purkinje cell synapses (Eccles et al., 1967). In contrast, climbing fibers in the cerebellum exert strong low frequency inputs onto Purkinje cells. Therefore, impaired granule neuron excitability in Fhf mutant animals is likely to prevent high-speed information processing in the cerebellum, resulting in ataxia. Furthermore, impaired high frequency excitation of granule cells in Fhf mutant animals is anticipated to interfere with plasticity of granule cell/Purkinje cell synapses, which is normally driven by burst firing of the presynaptic granule cell (Casado et al., 2002). These studies do not rule out additional requirements for FHFs in the cerebellar circuitry mediated through potential FHF expression in other types of cerebellar neurons. Our findings also suggest that impaired neuronal excitability underlies a type of human spinocerebellar ataxia traced to an inherited missense mutation in the Fhf4 gene (van Swieten et al., 2003).
The impaired excitability of Fhf mutant granule neurons can be explained by altered inactivation and recovery of voltage-gated sodium channels, although a role for FHFs in the physiology of other classes of ion channels cannot be excluded. While sodium channel distribution, isotype, abundance, and voltage dependence of activation are not substantially affected by Fhf1 and Fhf4 mutations, sodium channels in mutant granule cells inactivate at more negative membrane potential, inactivate faster at any specific membrane potential, and are slower to recover from fast inactivation than are channels in wild-type cells. Consistent with these findings, ectopic expression of FHF4a or FHF4b proteins in immortalized Neuro2A cultured cells or in primary hippocampal neurons induces depolarizing shifts in the voltage-dependence of sodium channel inactivation (Lou et al., 2005), and transfection of FHF2 proteins into an immortalized cell line derived from sensory neurons also induces depolarizing shifts in sodium channel inactivation (Rush et al., 2006; Wittmack et al., 2004). Since FHFs bind to sodium channel alpha subunits, the above findings readily suggest that sodium channels are physiologically modified through the binding of FHFs. However, ectopic FHF expression studies have failed to document the role of FHFs in accelerating sodium channel recovery from fast inactivation (Rush et al., 2006; Wittmack et al., 2004). Therefore, FHF modulation of sodium channels in vivo appears to be molecularly complex and as yet not fully recapitulated in ectopic gain-of-function culture systems.
An understanding of the mechanism underlying FHF control of sodium channel inactivation shall require further investigation. Mathematical modeling suggests that alterations in the rate constants governing the transition from closed and open states into the inactivated states is sufficient to cause the electrophysiological changes observed in Fhf1−/−Fhf4−/− mice. Sodium channel inactivation is governed by interaction between an inactivation “particle” housed in the III–IV cytoplasmic loop and the inner pore of the channel (Patton et al., 1993; Vassilev et al., 1988; West et al., 1992). The channel C-terminal tail also contributes to inactivation (An et al., 1998; Glaaser et al., 2006; Mantegazza et al., 2001), potentially through contacts between the tail and the III–IV loop (Motoike et al., 2004). Since the sodium channel tail is the documented site of FHF binding (Liu et al., 2001; Liu et al., 2003; Wittmack et al., 2004), FHFs may act through modulation of interactions among channel alpha subunit inactivation domains or through FHF-mediated recruitment or displacement of other channel binding proteins. In this regard, it is intriguing that inactivation of sodium channel Nav1.1 is modulated by beta subunit interaction with the Nav1.1 C-terminal tail (Spampanato et al., 2004).
While the Fhf4 gene plays a greater role in granule neuron excitability than does Fhf1, it is not clear whether FHF4 proteins are more efficient than FHF1 as modulators of sodium channels or, alternatively, if FHF4 proteins are expressed at higher levels in these cells. The broader question is to what degree excitability of different neurons is dependent on the specific Fhf genes and protein isoforms expressed in these cells as opposed to the mere aggregate concentration of FHF. Resolving the question of FHF protein specificity awaits development of quantitative assays for FHF interaction with sodium channels and, more importantly, for FHF modulation of channel inactivation.
The behavioral deficits in Fhf1−/−Fhf4−/− mice extend beyond ataxia to anomalies in neuromuscular and locomotor functions. The association of other FHFs with sodium channels at axon initial segments and at nodes of Ranvier, together with the widespread expression of Fhf genes throughout the nervous system (Smallwood et al, 1996; Hartung et al, 1997; Wang et al, 2000), suggest a broad role for the FHF family of proteins in the control of intrinsic excitability and other sodium channel-dependent processes.
The mouse Fhf1 gene was isolated on a bacterial artificial chromosome by screening a genomic library with a mouse Fhf1 cDNA (Hartung et al., 1997) hybridization probe. The gene targeting vector consisted of a 13.1 kilobase pair (kbp) region of the Fhf1 gene in which a central 1.0 kbp segment containing coding exon 2 was replaced with a PGKneo G418 selection cassette (see Suppl. Fig. 1). R1 (129/Sv, agouti/agouti) embryonic stem (ES) cell G418-resistant clones transformed with the targeting cassette were screened by Southern blot hybridization to identify clones bearing a disrupted Fhf1 allele (targeting frequency approximately 2% of clones) (see Suppl. Fig. 1). Fhf1+/− ES cells were injected into C57Bl/6 blastocysts, which were reimplanted into foster mother hosts. Male chimeras were bred with C57Bl/6 females, and agouti progeny were screened by Southern blotting to identify Fhf1+/− mice (see Figure 1). Since exon 2 is common to the two splice isoforms of Fhf1 mRNA (Hartung et al., 1997; Schoorlemmer and Goldfarb, 2001) and encodes part of the β-trefoil fold required for FHF binding to target proteins (Liu et al., 2001; Liu et al., 2003; Olsen et al., 2003), the disrupted Fhf1 allele is presumed to be functionally null.
Fhf4−/− mice (from a mixed C57Bl/6 and 129/Sv background)(Wang et al., 2002) were crossed with Fhf1+/− animals, and mutant alleles in Fhf1+/−Fhf4+/− progeny were propagated through two backcross generations onto a 129/Svev background before intercrossing to yield desired genotypes for analysis, including wild-type, Fhf1−/−, Fhf4−/−, and Fhf1−/−Fhf4−/−. Thereafter, wild-type and Fhf1−/− mice were maintained as fertile colonies. For Fhf4−/− mice, which bred poorly, and Fhf1−/−Fhf4−/−, which could not breed at all, desired animals were continually generated through heterozygote mating. Genotyping was performed by Southern blotting or PCR (Figure 1)(Wang et al., 2002).
The ledge test monitored the length of time (up to a maximum of 1 minute) an animal could maintain its balance on a high (35 cm), narrow (8 mm) rubber ledge, as described elsewhere (Schaefer et al., 2000).
The footprint assay was performed as described previously (Wang et al., 2002), except that hind paws and forepaws were painted different colors to allow measurement of the polarity of paw placement disparities. For each animal the paw placement disparity in the forward direction was measured and averaged for six consecutive steps.
The grid grip test visually monitored the ability of a mouse to maintain itself on a wire bar cage lid held to 45-degree, perpendicular or inverted positions. The behaviors for each genotype tested are illustrated in the videos in Supplemental Figure 1; all mice within each genotype group (n > 3) performed indistinguishably.
For the activity test, each mouse was placed in a clean 1-liter glass beaker, and monitored during a one-minute period for how many 360-degree turns it executed. After a trial acclamation day, the results from the second day were recorded. This assay was routinely performed in favor of the standard open field laser beam-monitored activity assay, since, in the latter, Fhf1−/−Fhf4−/− mice were hyperactive for short periods of time followed by prolonged immobility, perhaps reflecting fatigue accompanying weakness in this genotype.
Individual mouse cerebella (postnatal day 10) were separately teased free from surrounding meninges, treated with trypsin and DNaseI, triturated, and each cell preparation plated in Dulbecco’s modified Eagle’s (DME) medium + horse (5%) and fetal calf (5%) sera onto four 12 millimeter diameter poly-D-lysine-coated glass coverslips. The medium was changed to serum-free after 48 hours, and cells were maintained for 18 days prior to DNA transfection and for between 16–30 days prior to electrophysiological recording. Granule neurons were identified by their round, small diameter soma and their small membrane capacitance measured in whole cell patch clamp (see below). Granule neurons were transfected with pEGFP-N1 vectors (Clontech) directing expression of mouse FHF1a or human FHF1b (Lou et al., 2005; Schoorlemmer and Goldfarb, 2001) fused to GFP using Lipofectamine 2000 (Invitrogen), and processed for immunofluorescence 24 hours later.
Hippocampi dissected from embryonic day 18 rats were digested with 7 Units/ml papain (Sigma Chemical) + 0.17 mg/ml L-cysteine in Hank’s Balanced Salt solution (HBSS), then incubated with HBSS + 1 mg/ml bovine serum albumin + 1 mg/ml trypsin inhibitor followed by trituration and plating onto poly-D-lysine-coated glass coverslips (12 mm diameter) at density of four coverslips per hippocampus in DME medium containing 5% fetal calf serum. Cells were switched to serum-free DME medium after 24 hours. Neurons were processed for immunofluorescence after 12 days in culture.
Sagittal vibratome sections (0.25 mm thickness) of adult mouse cerebellar vermis were fixed in phosphate buffered saline + 4% paraformaldehyde on ice for 1 hour, then cryosectioned at 15 micron thickness and stained with a combination of cresyl violet, eosin Y, and Luxol Fast Blue. Sections were otherwise analyzed by immunofluorescence and confocal microscopy. Rat hippocampal neurons (12 days in vitro) and mouse transfected (20 days in vitro) and untransfected (30 days in vitro) cerebellar neurons were fixed as above and analyzed by immunofluorescence. Primary antibodies utilized were mouse monoclonal anti-pan-sodium channel alpha subunit (Sigma Chemical), rabbit anti-Nav1.6 (Boiko et al., 2001), rabbit anti-glutamate transporter EAAC1 (Alpha Diagnostics), rabbit anti-FHF2 (Schoorlemmer and Goldfarb, 2001), mouse monoclonal anti-ankyrin G (Santa Cruz Biotechnology), rabbit anti-microtuble associated protein-2 (MAP2, Chemicon) and mouse monoclonal anti-green fluorescent protein (Chemicon). Secondary reagents were biotinylated anti-rabbit IgG, anti-mouse IgG, and anti-mouse IgG2a (Jackson Immunoresearch), and fluorescent reagents (Molecular Probes): anti-rabbit IgG-ALEXA 488, anti-mouse IgG1-ALEXA 594, Streptavidin-ALEXA 488, and Streptavidin-ALEXA 594. Cell nuclei were visualized with DAPI. Digital images were taken on a Nikon Axiophot microscope or on a Leica confocal microscope. Consecutive z-axis images were stacked to allow visualization of axons running oblique to the imaging plane.
Adult mice (3–12 months old) were anesthetized by halothane inhalation, decapitated, and their cerebellar vermis chilled with ice-cold modified Kreb’s solution: 120 mM NaCl – 3 mM KCl – 1.2 mM KH2PO4 – 26 mM NaHCO3 – 3 mM glucose – 2 mM Na pyruvate – 3 mM myo-inositol – 1.2 mM MgSO4 – 2 mM CaCl2 - 0.4 mM Na ascorbate – 0.5 mM kynurenic acid – pH 7.2. Vibratome saggital sections (200–250 micron thickness) were allowed to recover for at least thirty minutes in the same solution at room temperature continually bubbled with oxygen/carbon dioxide (95%/5%), and were then placed in the unheated recording chamber on the stage of a Nikon EF600 microscope and continuously perfused with oxygenated solutions during recordings. Alternatively, coverslips containing cerebellar granule neurons cultured for 16–30 days in vitro (see above) were directly transferred into the recording chamber.
Recordings were conducted using Axopatch 200A or 200B amplifiers interfaced to pCLAMP command/record software through a Digidata 1322 analog/digital converter (Axon Instruments/Molecular Devices). Patch pipettes were pulled to 1.0–1.5 micron tip diameter using borosilicate glass tubing (1.5 mm O.D./0.97 mm I.D.) in P-87 or P-97 pipette pullers (Sutter Instruments). For all current injection experiments and for examination of all currents in voltage clamp, pipettes were filled with 126 mM K gluconate – 4 mM NaCl – 5 mM KOH-buffered HEPES – 15 mM glucose – 1 mM MgSO4 – 0.15 mM BAPTA – 0.05 mM CaCl2 – 3 mM Mg ATP – 0.1 mM Na GTP – pH 7.2 and the tissue slice was perfused with modified Kreb’s solution. For measurement of sodium currents, the pipette solution consisted of 104 mM CsF – 50 mM tetraethylammonium (TEA) Cl – 10 mM CsOH-buffered HEPES – 5 mM glucose – 2 mM MgCl2 – 10 mM EGTA – 2 mM Na ATP – 0.2 mM Na GTP – pH 7.2, and the tissues or cultured cells were perfused with 74 mM NaCl – 3 mM KCl - 26 mM NaHCO3 – 3 mM CsCl – 5 mM glucose – 25 mM TEA-Cl - 2 mM MgCl2 – 10 mM NaOH-buffered HEPES – 5 mM 4-amino-pyridine – 4 mM BaCl2 – 0.2 mM CdCl2 – pH 7.2, as previously described (Magistretti et al., 2006), except that the perfusion solution was supplemented with 0.5 mM kynurenic acid, 2 mM Na pyruvate, 3 mM myo-inositol, and 0.4 mM Na ascorbate to limit neuronal death. Pipette resistances were 8–12 MΩ when using pipette and bath solutions for all currents or were 4–8 MΩ when using sodium-only solutions. Junctional potentials were manually offset before patching. Sealing onto granule neurons in the granule layer of slices was achieved without visual aid, while sealing onto cultured neurons was achieved while viewing cells at 400X magnification. Whole cell access was initiated after seal resistance exceeded 8 GΩ. Recordings were filtered at 5 kHz and digitally sampled at 10 or 20 kHz. Membrane capacitance (Cm) was calculated from measurement of area under current transients induced by 10 mV voltage clamp steps (−85 to −75 mV or −70 to −80 mV). Series resistance was calculated based upon the measured time constant for fast decay of the capacitative current transient using the formula Rseries = τfast / Cm, and the frequency of voltage clamp was calculated using the formula fVC = 1 / (6.28 × τfast ). Input resistance was calculated from change in leak current using formula Rin = ΔVm/ΔIleak. Measured membrane voltages from current clamp recordings were corrected by −10 mV to offset the liquid junction potential (LJP) (Vm = Vrecord + LJP), which was measured empirically as −10.1 +/− 0.3 mV using the agar-bridge reference electrode method (Neher, 1992). For recordings in sodium current-only solutions, a −5 mV offset was applied to convert Vcommand to Vmembrane (Vm = Vc + LJP), reflecting an empirically measured LJP of −5.2 +/− 0.4 mV.
A patched cell was first subjected to a multisweep voltage clamp protocol in which membrane voltage (Vm) was stepped from −80 mV to test voltages (−70 + 10N mV, N = 1 to 8) for 30 milliseconds and current was recorded. Passive current transients and leak currents were subtracted during data acquisition using a hyperpolarizing prepulse protocol (P/6). The cell was then placed under current clamp control, Vm was adjusted to −80 mV through manual current injection, and the cell was subjected to 800 millsecond intervals of inward current injection of different amplitudes while recording Vm. The amplitude of current injection was increased until action potential frequency plateaued or diminished. The cell was again analyzed by voltage clamp to confirm that active currents had not substantially degraded; current injection data from a cell was only included in analyses if voltage-activated currents had diminished less than 20% from values preceding current clamp. Threshold voltage for spike generation (Vthreshold) was determined from recorded voltage traces as the inflection point of accelerating spike upstroke. Threshold current for spike generation (Ithreshold) was the minimum depolarizing current needed to elicit at least one action potential. In all genotypes, Vthreshold reflected the amount of passive depolarization generated by Ithreshold, i.e., Vrest − Vthreshold = Ithreshold × Rin. Current injection excitability data are only reported for neurons in brain slice preparations; wild-type cultured granule neurons showed a deficiency in A-type potassium current resulting in reduction of spike afterhyperpolarizations and inconsistent repetitive firing.
To analyze voltage dependence of sodium channel activation, cells were subjected to an 18-sweep protocol in which Vm was stepped from −85 mV to test voltages (−85 + 5N mV, N = 1 to 18) for 30 msec. Peak sodium current was plotted against Vm, and maximum sodium conductance computed from the linear portion of the plot (−10 to +5 mV range), enabling a plot of percentage open channels versus Vm. The data were directly fit to a Boltzmann equation with the formula
where f (Vm) is the fraction of open channels, V1/2(act) is the voltage of half-maximal activation, and k is the slope factor. For recordings of neurons in adult brain slices, the plot of peak conductance vs. Vm showed a two-phase transition (e.g., Supplemental Figure 5A, B) indicative of clamp escape and precluding analysis of voltage-dependent activation.
The rate constant τfast for fast inactivation of sodium channels in cultured neurons was calculated by fitting the decay from 90% to 10% peak sodium conductance to a two-component exponential decay using the formula
τfast was only accepted for cells in which Ifast > 0.75 Itotal.
Following depolarization from −85 mV to −35 mV, average residual sodium current measured 25–30 milliseconds after onset of depolarization was taken as the persistent current.
To analyze voltage dependence of sodium channel inactivation, cells were subjected to an 18-sweep protocol in which Vm was stepped from −95 mV to preconditioning test voltages (−95 + 5N mV, N = 0 to 17) for 125 msec followed by further depolarization to −15 mV. The peak sodium current at −15 mV measures the fraction of channels still available. The plot of percentage channels available versus Vm was fit to the Boltzmann equation with the formula
where V1/2(inact) is the voltage of half inactivation and k is the slope factor.
To analyze the recovery rate of sodium channels from inactivation, cells were subjected to a 16-sweep protocol in which Vm for each sweep was stepped from −85 to −15 mV for 10 msec, returned to −85 mV for 1.5N msec (N = 1 to 16), then depolarized again to −15 mV. The second peak sodium current for each sweep was expressed as a percentage of maximal channel recovery and plotted versus recovery time (t). For each cell, τrecovery was measured by fitting data to the formula:
Computer simulations of granule cell electrical activity were performed in the NEURON environment (Hines and Carnevale, 1997) by employing a granule cell single-compartment model described in detail elsewhere (D’Angelo et al., 2001). The sodium current was based on the 13-state, allosteric sodium channel model of Raman & Bean (2001) in the modified form described by Khaliq et al. 2003 (Scheme 1) and subsequently implemented in granule cells by Magistretti et al. (2006): The values of the kinetic parameters in Scheme 1 were: α = 353.9·exp(V/kα) ms−1, β = 1.272·exp(−V/kβ) ms−1, kα = kβ = 13.9 mV, n1 = 5.42, n2 = 3.28, n3 = 1.83, n4 = 0.74; γ = 150 ms−1, δ = 40 ms−1, ε = 1.75 ms−1, ζ = 0.0201·exp(−V/kζ) ms−1, kζ = 25.0 mV; Con = 0.005 ms−1, Coff = 0.5 ms−1, Oon = 0.75 ms−1, Ooff = 0.005 ms−1, a = (Oon/Con)1/4, b = (Ooff/Coff)1/4. Sodium current was calculated as: INa = f·Na · (V − VNa), where f is the fraction of channels in the open (O) state as returned by numerical resolution of Scheme 1 for time and voltage; VNa, the Na+ equilibrium potential, was +87.4 mV; and Na, the maximal Na+ conductance, was 13.0 mS·cm−2. To reproduce results obtained in Fhf1−/−Fhf4−/− mice, Con and Oon were increased by 10 times, while Coff was decreased to 40% of its original value. In particular, changing Con and Oon shifted steady-state inactivation, while changing Coff caused the shift in recovery from inactivation. For current clamp simulations, the density of sodium channels in the mutant model was doubled, bringing peak sodium conductance up to the level achieved in the wild-type model (as seen in actual recorded cells). In the models, temperature was set at 20 °C.
We thank Dr. Jacopo Magistretti for invaluable discussion and for sharing research data prior to publication. We further thank Christy Pardee, Stacey Rubin, and Pasqualina Farisello for their technical assistance. This research was funded by PHS grant R01-NS39906 (M.G), CUNY equipment grant (M.G.), and EU-SENSOPAC and CARIPLO grants (E.D.’A.)
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.