Address correspondence and reprint requests to Dr. H.S. White at Anticonvulsant Drug Development Program, Department of Pharmacology and Toxicology, University of Utah, 20 S 2030 E, Room 408, Salt Lake City, UT 84112, U.S.A. E-mail: email@example.com
Summary: Purpose: Mutations in the genes that encode subunits of the M-type K+ channel (KCNQ2/KCNQ3) and nicotinic acetylcholine receptor (CHRNA4) cause epilepsy in humans. The purpose of this study was to examine the effects of the Szt1 mutation, which not only deletes most of the C-terminus of mouse Kcnq2, but also renders the Chnra4 and Arfgap-1 genes hemizygous, on seizure susceptibility and sensitivity to drugs that target the M-type K+ channel.
Methods: The proconvulsant effects of the M-channel blocker linopirdine (LPD) and anticonvulsant effects of the M-channel enhancer retigabine (RGB) were assessed by electroconvulsive threshold (ECT) testing in C57BL/6J-Szt1/+ (Szt1) and littermate control C57BL/6J+/+ (B6) mice. The effects of the Szt1 mutation on minimal clonic, minimal tonic hindlimb extension, and partial psychomotor seizures were evaluated by varying stimulation intensity and frequency.
Results:Szt1 mouse seizure thresholds were significantly reduced relative to B6 littermates in the minimal clonic, minimal tonic hindlimb extension, and partial psychomotor seizure models. Mice were injected with LPD and RGB and subjected to ECT testing. In the minimal clonic seizure model, Szt1 mice were significantly more sensitive to LPD than were B6 mice [median effective dose (ED50) = 3.4 ± 1.1 mg/kg and 7.6 ± 1.0 mg/kg, respectively]; in the partial psychomotor seizure model, Szt1 mice were significantly less sensitive to RGB than were B6 mice (ED50= 11.6 ± 1.4 mg/kg and 3.4 ± 1.3 mg/kg, respectively).
Conclusions: These results suggest that the Szt1 mutation alters baseline seizure susceptibility and pharmacosensitivity in a naturally occurring mouse model.
Many causes for epilepsy have been identified, but in the majority of cases, the seizure etiology remains unresolved. As a result, many forms of epilepsy are described as “idiopathic.” Advances in epilepsy research have identified a strong genetic component to many of these idiopathic epilepsies, including the rare pediatric disorders benign familial neonatal convulsions (BFNC) and autosomal dominant nocturnal frontal lobe epilepsy (ADNFLE). BFNC, a generalized epilepsy, is caused by mutations in the KCNQ2 and KCNQ3 genes that underlie the neuronal M current [IK(M)] (1,2); and ADNFLE, a partial epilepsy, is caused by mutations in the CHRNA4 gene, which encodes the α4 subunit of the nicotinic acetylcholine (nACh) receptor (3). BFNC is characterized by seizures that start between postnatal day 2 and month 4 and continue for several weeks to months, after which point, they spontaneously resolve (4). BFNC is considered benign, partly because patients exhibit normal psychomotor skills and learning ability; consequently, the prognosis is favorable. Patients are, however, at a 15-times greater risk for developing epilepsy as adults (5). ADNFLE is characterized by brief, violent, partial seizures of frontal lobe origin that occur during light sleep. The average age at onset is ∼10 years, and seizures often either remit or at least improve after puberty; however, as is the case with BFNC, patients have a higher incidence of epilepsy in adulthood (6). Thus these patients might be described as having long-term increased seizure susceptibility.
Potassium ion flux through the voltage-gated M-type K+ channel generates the neuronal M current [IK(M)], which plays a critical role in setting and maintaining the resting membrane potential of many neurons, and thus helps establish baseline neuroexcitability (7). In animal models, agents that decrease IK(M) amplitude are proconvulsant (8), whereas IK(M) enhancers are anticonvulsant (9,10); therefore a putative link exists between decreased IK(M) function and increased seizure susceptibility. In Xenopus oocytes, it has been shown that IK(M) amplitude is diminished by the inclusion of KCNQ2 or KCNQ3 M-channel subunits containing BFNC-causing mutations (11). It is therefore implied that mutations in the KCNQ2 and KCNQ3 genes precipitate IK(M) hypofunctionality in BFNC patients, the consequence of which is increased seizure susceptibility.
Presynaptic nACh receptors containing the α4-subunit subtype also play a critical role in modulating neurotransmitter release (12,13) and neuroexcitability (14,15). In animal models, agents that increase conductance through the nACh receptor (e.g., nicotine) are proconvulsant (16), and agents that decrease nACh receptor activity (e.g., scopolamine, caramiphen) are anticonvulsant (17). An oocyte expression study of one ADNFLE-causing CHRNA4 mutation suggests that such mutations increase the potency but decrease the efficacy of ACh (18). Thus it is not clear how CHRNA4 mutations directly increase neuroexcitability in ADNFLE.
Previously, in efforts to characterize better the roles of the KCNQ2 and CHNRA4 genes in epilepsy, separate groups have generated mice with targeted deletions of the Kcnq2 and Chrna4 genes. Kcnq2-null mice die shortly after birth, and hemizygous mice exhibit decreased transcript expression and increased susceptibility to pentylenetetrazol (PTZ)-induced seizures (19). Chrna4-null mice are viable and exhibit increased sensitivity to PTZ-induced seizures as well as elevated anxiety-like behavior (20). In related work, Yang et al. (21) identified a 300-kb spontaneous deletion mutation, Szt1, which deletes the region of the Kcnq2 gene that encodes the majority of the KCNQ2 C-terminus, as well as the Chrna4 (nACh-receptor α4 subunit) and Arfgap1[guanosine triphosphatase (GTPase)-activating protein that inactivates adenosine diphosphate (ADP)-ribosylation factor 1] genes. Mice heterozygous for the Szt1 mutation were identified in a subpopulation of stock C57BL/6J (B6) mice as the result of a large-scale electroconvulsive threshold (ECT) screen (21). Mice that are heterozygous for this deletion mutation exhibit a decreased seizure threshold.
With the Szt1 deletion encompassing this unique combination of epilepsy-related genes, the Szt1 mouse presents a model in which to study polygenic epilepsy in a naturally occurring system. ECT testing was chosen as a standardized acute seizure-threshold test to examine the effects of this mutation on seizure susceptibility and pharmacosensitivity in Szt1 mice. The aim of this study was to (a) examine further the effects of the Szt1 mutation on seizure thresholds in several electroconvulsive paradigms, and (b) characterize differences in pharmacosensitivity to pro- and anticonvulsant agents that act at the M-channel.
MATERIALS AND METHODS
Szt1 and B6 mice
Eight- to 12-week-old coisogenic male and female C57BL/6J-Szt1/+ (Szt1) mice (15–25 g) and their C57BL/6J-B6+/+ (B6) littermates were obtained from a research colony at the Jackson Laboratory (Bar Harbor, Maine) and used for all ECT testing. Animals were allowed free access to food and water and were housed in a temperature- and light-controlled (12 h on/12 h off) environment. All animal care and experimental manipulations were approved by the Institutional Animal Care and Use Committee (IUCAC) of the University of Utah and are in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals.
For all behavioral experiments involving drug injection, linopirdine (LPD) and retigabine (RGB) suspensions were made in 0.5% methyl cellulose (MC) and sonicated for 12 min before i.p. injection. Drug solutions were freshly prepared on the day of each experiment. All chemicals were purchased from Sigma (ST. Louis, MO, U.S.A.) unless otherwise noted. RGB was generously supplied by VIATRIS (Frankfurt, Germany).
Baseline seizure thresholds
Baseline convulsive current (CC) curves were constructed for mice of both genotypes. A drop of tetracaine (0.5%) was administered to each eye just before testing. Three different stimulation protocols were used in an effort to differentiate the effects of IK(M) modulators on limbic (6 Hz), forebrain (minimal clonic), and hindbrain (minimal tonic hindlimb extension) seizure thresholds. Partial psychomotor seizure testing was conducted with a Grass S48 stimulator (6-Hz, 0.2-ms rectangular pulse width, 3-s duration, varying current intensities). The phenotype for these seizures is rhythmic face movements, forelimb clonus, dorsal neck flexion, rearing and falling, and/or transient gait wobbliness/ataxia (22). Minimal clonic and minimal tonic hindlimb extension (THE) seizure testing (60-Hz, 0.2-ms sinusoidal current pulse, varying current intensities) was conducted with a stimulator previously described (23). Minimal clonic seizures are characterized by rhythmic face and forelimb clonus, rearing and falling, and ventral neck flexion. Minimal THE seizures are characterized by a tonic–clonic flexion–extension sequence that starts with tonic forelimb extension, followed by hindlimb flexion, and terminates in full THE (180 degrees to the torso) (24–26).
The previously described staircase estimation procedure (27) was used to generate population seizure thresholds for each mouse genotype. Using these data, full CC curves were generated and statistical significance was determined using Probit analysis in the statistical program MINITAB (State College, PA, U.S.A.). CC1-99 values also were calculated for each genotype and seizure type. Separate CC curves were constructed for male and female mice.
Behavioral effects of M-current modulation
After establishing baseline thresholds for each seizure phenotype in both mouse genotypes, the effects of two compounds that directly modulate the M-channel (LPD, a blocker, and RGB, an enhancer) were tested. LPD and RGB were prepared in 0.5% MC on the day of testing, such that 0.01 ml/g of body weight was injected. All drugs were administered by intraperitoneal (i.p.) injection. Mice were injected with either LPD or RGB (10 mg/kg) and tested for partial psychomotor or minimal clonic seizures 15 min after injection (the previously determined time-to-peak effect for both drugs). Control mice received an equivalent volume of 0.5% MC. Mice received either LPD or RGB injections and were stimulated at the previously established CC10 and CC90 values, respectively. A subsequent study was conducted to test whether the proconvulsant effects of LPD and the anticonvulsant effects of RGB could be mitigated when the two drugs were administered concurrently. For this study, both drugs were co-injected (10 mg/kg each), and B6 and Szt1 mice were stimulated at their respective CC50 values. Results obtained from this combination study were compared with those obtained from a study in which LPD and RGB were studied separately in both seizure tests at the CC50 value for each genotype. Groups of nine to 15 mice were used for this set of experiments.
Linopirdine and retigabine dose–response curves
Based on results from the initial LPD and RGB studies (see Table 2), full dose–response curves were constructed for RGB in the partial psychomotor model and for LPD in the minimal clonic model. For these experiments, RGB doses of 1.25, 2.5, 5.0, 10.0, 12.25, 15.0, and 20.0 mg/kg and LPD doses of 1.25, 2.5, 5.0, 7.5, 10.0, 15.0, and 20.0 mg/kg were used. Each successively lower dose was made by serial dilution. Groups of only four to six mice were used for most doses because of the scarcity of age-matched Szt1 and littermate control B6 mice available from the Jackson laboratory research colony.
Table 2. Genotype- and seizure phenotype–dependent changes in sensitivity to M current modulators
Stim. intensity treatment
Minimal clonic seizure
Partial psychomotor seizure
Data are reported as number of mice seizing/number of mice tested. Note that for groups treated with vehicle control methyl cellulose (MC), the fraction of mice seizing approximates each predetermined CC value.
RGB, retigabine; LPD, linopirdine.
aMice treated with RGB were tested at the previously determined CC90 value for each seizure type
bp < 0.05 (within-genotype drug effect, Fisher's Exact test).
cMice treated with LPD were tested at the CC10 value for each seizure type
dMice treated with LPD/RGB were tested at the CC50 value for each seizure type
ep < 0.05 (between-genotypes drug effect, Mantel–Haenszel χ2 test).
f,gThe effects of LPD and RGB were tested independently at the CC50 values for both seizure types. Data included in this table were acquired from female mice. Some data from male mice show similar trends in pharmacoresponsiveness (data not shown).
The effect of high doses of LPD and/or RGB (≥10 mg/kg total drug) on motor function was assessed by the rotarod test (28,29). Rotarod testing was conducted immediately before ECT testing by placing the mouse on a 1-inch diameter rod, textured with small ridges to aid in gripping, rotating at 6 rpm. Motor function was considered impaired if the mouse fell off the rotarod 3 times during a 1-min testing period.
For baseline seizure threshold estimates, seizure incidence was determined at several different stimulus intensities according to the staircase estimation procedure (27). Convulsive current (CC) curves were then constructed from these data by Probit analysis, and CC1-99 values were calculated using Minitab 13 (State College, PA, U.S.A.). CC curves for Szt1 mice were compared with those of wild-type B6 mice for each seizure type, and seizure thresholds were considered significantly different at p < 0.05.
The effect of drug treatment on seizure incidence at a given CC value was determined for each genotype. Fisher's Exact test was used for within-genotype analysis of drug versus MC effects; one-tailed tests were used for the effects of LPD and RGB alone, and two-tailed tests were used for the effects of LPD/RGB coapplication. The two-tailed Mantel–Haenszel χ2 test was used for between-genotype drug-effect comparisons. Significance was determined at p < 0.05.
RGB and LPD dose–response curves for B6 and Szt1 mice were constructed using Probit analysis and compared for genotype-dependent shifts in dose–response curves. Dose–response curves obtained in B6 and Szt1 mice were considered significantly different at p < 0.05.
As reported previously (21), the Szt1 mutation (when expressed on a B6 background) significantly decreases minimal clonic and minimal THE seizure thresholds. In the present study, CC curves were constructed for partial psychomotor seizures (Fig. 1A), as well as minimal clonic (Fig. 1B) and minimal THE extension seizures (Fig. 1C). The data presented here not only confirm, but also extend, the findings of Yang et al. by showing that Szt1 mouse seizure thresholds are significantly lower than B6 mouse thresholds for all three seizure phenotypes. Calculated CC50 values and corresponding 95% confidence intervals (CI95) are summarized in Table 1.
Table 1. Convulsive current 50 (CC50) values and corresponding 95% confidence intervals (CI95) for B6 and Szt1 mice in three ECT testing paradigms
CC50 (CI95) values (mA)
Szt1 mice exhibit significantly reduced seizure thresholds relative to B6 mice in all three seizure types tested.
ap < 0.05.
Female B6 and Szt1 mice display lowered seizure thresholds relative to male B6 and Szt1 mice
Previous studies demonstrated that the seizure threshold of female B6 mice is lower than that of their male counterparts (30,31). Consistent with these findings, the seizure thresholds of both female B6 and Szt1 mice were significantly decreased relative to male B6 and Szt1 mice in the minimal clonic and partial psychomotor seizure models. For example, CC50 values obtained in the partial psychomotor seizure test were as follows: B6 male, 22.4 ± 0.7; B6 female, 19.5 ± 0.7; Szt1 male, 19.6 ± 0.5; Szt1 female, 17.7 ± 0.5. Furthermore, Szt1 decreased the partial psychomotor seizure CC50 values of both male and female mice by 2.8 mA.
Szt1 mice exhibit altered responsiveness to drugs acting at the M-channel
Minimal clonic seizures
Consistent with its known anticonvulsant profile, RGB (10 mg/kg) significantly decreased seizure susceptibility to ECT testing in both Szt1 and littermate control mice (Table 2). When tested at the CC90, nine of 10 B6 mice treated with MC displayed a minimal clonic seizure versus four of 10 treated with RGB; eight of 10 Szt1 mice treated with MC seized versus three of 10 treated with RGB.
Conversely, the IK(M) blocker LPD (10 mg/kg) increased seizure susceptibility in both groups. When tested at the CC10, two of 15 B6 mice treated with MC versus six of 10 treated with LPD displayed a minimal clonic seizure, whereas one of 10 Szt1 mice treated with MC seized versus 19 of 20 treated with LPD. Interestingly, a between-strains statistical comparison of these data revealed that LPD was significantly more proconvulsant in Szt1 mice than in B6 mice (Mantel–Haenszel χ2 test, p < 0.05). Furthermore, not only did a higher proportion of LPD-treated Szt1 mice display minimal clonic seizures compared with B6 (19 of 20 vs. six of 10, respectively), but also 12 of the 19 Szt1 mice that displayed a minimal clonic seizure progressed directly to the more severe minimal THE seizure. This seizure spread was not observed in any of the LPD-treated B6 mice that exhibited minimal clonic seizures.
When mice were co-injected with LPD/RGB (10 mg/kg each) and subjected to ECT testing, another notable genotype-specific difference arose (Table 2): whereas co-injected B6 mice showed no difference in seizure incidence compared with vehicle control (12 of 25 vs. 12 of 22, respectively), co-injected Szt1 mice seized at a significantly higher rate compared with vehicle control (13 of 15 vs. seven of 15, respectively). The seizure rate observed in co-injected Szt1 mice was comparable to that observed when Szt1 mice were treated with LPD only (13 of 15 vs. 19 of 20, respectively).
Partial psychomotor seizures
With a protocol identical to the one outlined in the previous section, the effects of LPD, RGB, and LPD/RGB were examined in the partial psychomotor seizure model (Table 2). When tested at the CC90, RGB (10 mg/kg) significantly decreased the seizure incidence in B6 mice (nine of 11 for MC vs. two of 11 for RGB) but had no effect in Szt1 mice (seven of nine for MC vs. seven of nine for RGB). When tested at the CC10, LPD increased the seizure incidence in both B6 (one of 10 for MC vs. six of 10 for LPD) and Szt1 mice (one of 10 for MC vs. seven of 10 for LPD). Unlike the results obtained in the minimal clonic seizure test, no notable genotype-dependent difference in LPD sensitivity was noted in the psychomotor seizure test.
When mice were co-injected with LPD/RGB and subjected to ECT testing, seizure incidence was decreased in B6 mice (six of 11 for MC vs. two of 11 for LPD/RGB), but slightly increased in Szt1 mice (five of nine in MC vs. seven of nine in LPD/RGB). Although neither change was significant (Fisher's Exact test), the Mantel–Haenszel χ2 statistic revealed that LPD/RGB co-injection was significantly more proconvulsant in Szt1 mice than in B6 mice. Taken together, the data obtained in the minimal clonic and partial psychomotor seizure tests reveal both genotype-dependent and seizure type–dependent differences in pharmacosensitivity.
Several strains of mice display a propensity for respiratory arrest after THE seizures, and death may follow if the mice are not artificially resuscitated (32). Although B6 mice rarely underwent respiratory arrest after minimal THE extension seizures, Szt1 mice usually did (data not shown), and this condition was always lethal if mice were not resuscitated. With successive breedings, it became virtually impossible to resuscitate Szt1 mice after minimal THE seizures. The observed high mortality rate in response to THE seizures, compounded with the general paucity of Szt1 and littermate control B6 mice, precluded pharmacology experiments in the minimal THE seizure model.
LPD and RGB ED50 values are significantly shifted in Szt1 mice
The seizure type–dependent differences in pharmacosensitivity presented in Table 2 indicated the need for complete LPD and RGB dose–response curves. LPD curves were constructed to test the hypothesis that in the minimal clonic seizure model, LPD is a more potent proconvulsant in Szt1 mice (Fig. 2). Similarly, RGB curves were constructed to test the hypothesis that in the partial psychomotor seizure model, RGB is a less potent anticonvulsant in Szt1 mice (Fig. 3). The resulting data support this hypothesis and demonstrate that in the minimal clonic seizure model, Szt1 mice are more sensitive to the proconvulsant effect of LPD than are B6 mice (ED50s: 3.4 ± 1.1 mg/kg vs. 7.6 ± 1.0 mg/kg, respectively). Likewise, in the partial psychomotor seizure model, Szt1 mice are less sensitive to the anticonvulsant effects of RGB than are B6 mice (ED50s: 11.6 ± 1.4 mg/kg vs. 3.4 ± 1.3 mg/kg, respectively). Motor impairment, as determined by the rotarod test, was observed in only three of nine Szt1 mice in response to 20 mg/kg RGB injection. None of the B6 mice tested at the highest RGB dose (i.e., 12.5 mg/kg) exhibited impairment.
Previously it was demonstrated that the Szt1 mutation lowers the seizure threshold of B6 mice (21). The experiments presented here were designed to characterize further the seizure threshold shift induced by the Szt1 mutation and assess changes in seizure susceptibility after treatment with proconvulsant and anticonvulsant drugs. The results obtained from this study support the hypothesis that a deletion mutation affecting a combination of genes implicated in epilepsy alters both seizure susceptibility and pharmacosensitivity. Specifically, mice with a mutation in Kcnq2 and hemizygous deletion of Chrna4 and Arfgap1 do display reduced seizure thresholds and altered sensitivity to the M-channel–modifying drugs LPD and RGB. In addition, the changes in drug sensitivity are dependent on the seizure type elicited. The results presented here demonstrate that Szt1 mice are more sensitive to the proconvulsant properties of LPD in the minimal clonic model, and less sensitive to the anticonvulsant properties of RGB in the partial psychomotor model, than are B6 mice.
The seizure type–dependent differences in responsiveness to M-channel modulators may be explained in part by the fact that the partial psychomotor and minimal clonic seizure types result from activation of different brain areas. For example, c-fos expression studies have demonstrated that the stimulus required to elicit psychomotor seizures (at threshold levels) strongly activates the neocortex (22), whereas the stimulus required to induce minimal clonic seizures activates both forebrain and midbrain structures. Kcnq2 and Chrna4 transcripts are abundant in several brain regions that are relevant to epileptogenesis, including the cerebral cortex, thalamus, and hippocampus (20,33). However, it has not yet been determined if and where KCNQ2 subunits carrying the Szt1 mutation (and thereby Chrna4 hemizygosity) are expressed. Further experiments would prove useful in this regard.
The results from the LPD/RGB co-injection experiments are particularly interesting, as they suggest that Szt1 mice are more susceptible to seizure spread from forebrain/limbic areas to the hindbrain. For example, in Szt1 mice, the effects of LPD prevail over those of RGB in the minimal clonic seizure model; although a higher fraction of LPD/RGB co-injected Szt1 mice seized vs. MC-injected Szt1 mice, none progressed to minimal THE seizures. Recall that when Szt1 mice were treated with LPD alone and tested for minimal clonic seizures, 12 of the 19 that seized progressed directly to minimal THE seizures. The results from the LPD/RGB coadministration studies suggest that in B6 mice, the anticonvulsant effects of RGB attenuate the proconvulsant effects of LPD, whereas in Szt1 mice, the effects of LPD predominate over those of RGB. The stimuli required to elicit minimal clonic and minimal THE seizures are similar, except that minimal THE seizures require greater current intensity. This finding suggests that although RGB does not protect against minimal clonic seizures in Szt1 mice in the presence of LPD, it may prevent seizure spread from limbic/forebrain structures to the hindbrain.
Considering that KCNQ2 C-terminal deletion mutations are among the mutations identified in BFNC patients (34), it is not surprising that mice carrying the Szt1 mutation exhibit decreased seizure thresholds. Although CHRNA4 hemizygosity is not implicated in ADNFLE patients, previous mouse studies suggest that Chrna4 hemizygosity most likely contributes to the reduced seizure threshold of the Szt1 mouse. In contrast to Kcnq2 and Chrna4, the consequences of a decrease in Arfgap1 expression have not been characterized. Whereas it has been shown that in the presence of GTP, Arfgap1 promotes vesicle formation and cargo sorting in the Golgi complex (35), the role of Arfgap1 in regulating synaptic vesicle formation has not been established.
Whereas we can conclude that the Szt1 mutation decreases baseline seizure threshold, we cannot determine whether all of the affected genes contribute to the change in seizure susceptibility. It is noteworthy, however, that the Jackson Laboratory has identified two new mouse genotypes that carry only Kcnq2 point mutations and exhibit decreases in seizure threshold that are at least as severe as those reported in the Szt1 mouse (Yang and Frankel, unpublished results). These preliminary results suggest that Kcnq2 mutations alone may be sufficient to reduce electroshock seizure thresholds. The results from this study serve to illustrate the Szt1 mouse's altered sensitivity to IK(M) modifying drugs, but the altered pharmacosensitivity observed in this mouse is by no means limited to M-channel–modifying drugs. Further experiments to establish the effects of compounds acting at the nACh receptor and compounds that target other epilepsy-related receptors and ion channels would prove useful in establishing the Szt1 mouse as a model of increased seizure susceptibility. Although spontaneous seizures have not yet been reported, the Szt1 mouse does provide a naturally occurring hyperexcitable model system with which to investigate several issues pertinent to epilepsy research, including epileptogenesis, seizure susceptibility, the “second hit” hypothesis, and disease-modification strategies (36). Finally, electrophysiology experiments, in either expression systems or in vitro preparations, could be used to differentiate the possible roles of the individual gene mutations/deletions in this hyperexcitable mouse.
Acknowledgment: We thank Drs. Harold Wolf and Misty Smith-Yockman for encouragement and critical reading, as well as Barbara Beyer for genotyping and Carolyne Dunbar for maintaining the mouse breeding colony at the Jackson Laboratory. This work was supported by NS-40246 (W.N.F. and H.S.W.), NS31348 (W.N.F.), Primary Children's Medical Center Foundation 51001142 (K.S.W.), and a TJL postdoctoral fellowship (Y.Y.).