Journal of Neurophysiology

Seizures and Reduced Life Span in Mice Lacking the Potassium Channel Subunit Kv1.2, but Hypoexcitability and Enlarged Kv1 Currents in Auditory Neurons

Helen M. Brew, Joshua X. Gittelman, Robert S. Silverstein, Timothy D. Hanks, Vas P. Demas, Linda C. Robinson, Carol A. Robbins, Jennifer McKee-Johnson, Shing Yan Chiu, Albee Messing, Bruce L. Tempel

Abstract

Genes Kcna1 and Kcna2 code for the voltage-dependent potassium channel subunits Kv1.1 and Kv1.2, which are coexpressed in large axons and commonly present within the same tetramers. Both contribute to the low-voltage–activated potassium current IKv1, which powerfully limits excitability and facilitates temporally precise transmission of information, e.g., in auditory neurons of the medial nucleus of the trapezoid body (MNTB). Kcna1-null mice lacking Kv1.1 exhibited seizure susceptibility and hyperexcitability in axons and MNTB neurons, which also had reduced IKv1. To explore whether a lack of Kv1.2 would cause a similar phenotype, we created and characterized Kcna2-null mice (−/−). The −/− mice exhibited increased seizure susceptibility compared with their +/+ and +/− littermates, as early as P14. The mRNA for Kv1.1 and Kv1.2 increased strongly in +/+ brain stems between P7 and P14, suggesting the increasing importance of these subunits for limiting excitability. Surprisingly, MNTB neurons in brain stem slices from −/− and +/− mice were hypoexcitable despite their Kcna2 deficit, and voltage-clamped −/− MNTB neurons had enlarged IKv1. This contrasts strikingly with the Kcna1-null MNTB phenotype. Toxin block experiments on MNTB neurons suggested Kv1.2 was present in every +/+ Kv1 channel, about 60% of +/− Kv1 channels, and no −/− Kv1 channels. Kv1 channels lacking Kv1.2 activated at abnormally negative potentials, which may explain why MNTB neurons with larger proportions of such channels had larger IKv1. If channel voltage dependence is determined by how many Kv1.2 subunits each contains, neurons might be able to fine-tune their excitability by adjusting the Kv1.1:Kv1.2 balance rather than altering Kv1 channel density.

INTRODUCTION

Neuronal information processing is shaped by voltage-dependent potassium (Kv) channel tetramers, whose properties depend partly on which four subunits they contain. The subfamily of mouse genes Kcna1Kcna8 codes eight Kv1 subunit types, Kv1.1–Kv1.8, potentially leading to high functional diversity in Kv1 channels (Lock et al. 1994). Alternatively, if there is functional redundancy, it may occur between subunit types Kv1.1 and Kv1.2 because they are coded by the two most closely related Kcna genes, are present at many of the same CNS locations, and either type forms channels with rapid activation and slow inactivation when expressed in oocytes (Hopkins et al. 1994; Lock et al. 1994; Stuhmer et al. 1989; Wang et al. 1993, 1994). Kv1.1 and Kv1.2 are also the two most abundant Kv1 subunit types and are commonly in the same tetramers; Kv1.1:Kv1.2 and Kv1.1:Kv1.4 are common combinations detected by coimmunoprecipitation from mammalian brains along with several different combinations of Kv1.1, Kv1.2, Kv1.3, Kv1.4, and Kv1.6 (Coleman et al. 1999; Rhodes et al. 1997; Scott et al. 1994; Shamotienko et al. 1997; Wang et al. 1993, 1999). This makes it hard to discern whether Kv1.1 and Kv1.2 possess distinct functional roles.

One approach to dissecting out separate functions for individual Kv1 subunit types is genetic deletion. For example, evidence for an axonal role of Kv1.1 was provided by Kcna1-null mice in which several axonal and axon terminal sites exhibited hyperexcitability, thought to underlie their seizure susceptibility, cold-swim–induced myoclonus, and hyperalgesia (Clark and Tempel 1998; Smart et al. 1998; Zhang et al. 1999; Zhou et al. 1998). Such evidence is difficult to obtain in direct recordings from the sites where Kv1.1 and Kv1.2 are likely most strongly expressed, at the juxtaparanodes of large-diameter axons, because they are beneath myelin, which also impedes toxin access and makes their physiological role controversial (David et al. 1995; Wang et al. 1993). In the present study we created Kcna2-null (−/−) mice lacking Kv1.2 and compared them with control +/− and +/+ littermates, as well as previous results from Kcna1-nulls. The −/− mice had reduced life spans and exhibited spontaneous generalized seizures. To explore developmental processes pertaining to the postnatal day 15 (P15) onset of −/− seizures, we performed quantitative polymerase chain reaction assays (qPCR) on +/+ brain stems that showed Kcna2 mRNA was strongly upregulated from P7 on.

Kv1 currents have been characterized at some neuronal somata using direct recordings and toxins that block channels containing at least one sensitive subunit type. For example, channels containing at least one Kv1.2 subunit are blocked by tityustoxin-Kα (TsTx), a component of Tityus serrulatus scorpion venom (Hopkins 1998; Werkman et al. 1993), and dendrotoxin-I (DTX) from black mamba snake venom (Strydom 1976) blocks channels containing at least one of the DTX-binding subunits Kv1.1, Kv1.2, and Kv1.6 (Robertson et al. 1996). This probably means DTX blocks all Kv1 channels native to neurons, especially in auditory neurons of the medial nucleus of the trapezoid body (MNTB) where Kv1.1, Kv1.2, and Kv1.6 are the predominant Kv1 subunit types (Dodson et al. 2002; Fonseca et al. 1998). Applying DTX to murine MNTB neurons and other auditory neurons strongly expressing Kv1.1 and Kv1.2 has shown that Kv1 channels underlie a rapidly activating low-voltage–activated potassium current Ikl (Adamson et al. 2002; Bal and Oertel 2001; Brew et al. 2003; Grigg et al. 2000; Wang et al. 1994). Proposed functions of auditory Ikl are to powerfully repolarize large synchronized excitatory postsynaptic potentials, reduce membrane time constants, minimize temporal summation, facilitate temporally precise transmission across synapses, and preserve phase locking to sound peaks (Brew and Forsythe 1995; Manis and Marx 1991; Oertel 1983; Trussell 1999).

The homogeneity of the MNTB (90% principal neurons) makes it a prime location for combining the above-cited approaches: direct recordings, toxin applications, and genetic deletion. For example, MNTB neurons in brain stem slices from Kcna1-null mice had 30% reduced Ikl amplitudes and were hyperexcitable [they fired more action potentials (APs) and had smaller threshold currents than those of controls; Brew et al. 2003]. In analogous recordings described below, Kcna2-null MNTB neurons were hypoexcitable and possessed enlarged Ikl, exactly opposite to the changes in Kcna1-null MNTB. To distinguish subcomponents of Ikl due to Kv1 channels with or without Kv1.2 subunits, TsTx and DTX were applied. One unexpected finding was that all Kv1 channels in +/+ mouse MNTB neurons contained Kv1.2, in marked contrast to rat MNTB neurons in which only a subset contained Kv1.2 (Dodson et al. 2002). An analysis of voltage dependence suggests Kv1.2-free channels produce larger currents because they activate at more negative potentials than Kv1.2-containing channels, whose voltage dependence may also depend on the number of Kv1.2 subunits they contain. Thus neurons may adjust their balance between Kv1.2 and Kv1.1 expression to fine-tune their excitability for specialized information-processing tasks. Some MNTB results were previously described in abstracts (Brew et al. 2000, 2001).

METHODS

Generation of Kcna2-null mice

A genomic phage clone containing the Kcna2 gene was isolated from a 129/Sv mouse liver library (Stratagene, La Jolla, CA). For the targeting construct, the neomycin resistance cassette (provided by Dr. R. Behringer, The University of Texas M. D. Anderson Cancer Center, Smithville, TX) was inserted between a Kcna2 5′ Xba I–Eag I and 3′ Nco I–Xba I genomic fragment (see Fig. 1A). The thymidine kinase cassette (provided by Dr. R. Behringer) was cloned into the 3′ portion of the targeting construct. The targeting vector was linearized and electroporated into AB-1 embryonic stem (ES) cells (gift of Dr. A. Bradley, Wellcome Trust Sanger Institute, Cambridge, UK). The ES cells were then subjected to positive–negative selection for 8 days in 300 μM G418 and 200 nM fialuridine. Doubly resistant clones were expanded and analyzed by Southern blot analysis.

FIG. 1.

Generation and verification of Kcna2-null mice. A: diagram of the targeting strategy used to remove the Kcna2 open reading frame (ORF). Homologous recombination of the Kcna2 chromosomal locus (top) with the targeting vector construct (middle) generated a targeted locus (bottom) in which the neomycin resistance cassette (Neo) has replaced the Kcna2 ORF. B: Southern blot analysis of genomic DNA from offspring of a heterozygote intercross. An Eco RV digest probed with the 5′ (Xba I–Xba I) fragment detected an 8.5-kilobase (kb) wildtype allele and an 11.3-kb targeted allele. In this litter there were 2 wildtype (+/+) mice, 3 heterozygous (+/−) mice, and one Kcna2-null (−/−) mouse. C: Western blot analysis of Kv1.2 in protein isolated from whole brains. Kv1.2 protein was detected in +/+, reduced in +/−, and not detected in −/− brains. The broad band of Kv1.2 staining probably reflects variable levels of glycosylation. Blots were reprobed with anti-β-actin to control for loading. D: quantitative polymerase chain reaction (qPCR) expression of mRNA for Kcna2, Kcna1, and Kcna6 in whole brains from mice aged postnatal day 14 (P14). For each qPCR experimental run, the expression of the 3 Kcna genes was measured and normalized relative to the geometric mean of the expression levels for 3 reference genes: β-actin; γ-actin; and succinate dehydrogenase complex, subunit A (Table S1 and methods). Mean relative expression levels from 1 to 4 qPCR runs were calculated for each mouse and each gene (only 2 mice had only a single qPCR run per gene) before averaging across mice, and plotting the mean and SE. Kcna1 and Kcna6 mRNA expression were similar across genotypes. Kcna2 mRNA expression was approximately halved in +/− brain compared with +/+ controls, and Kcna2 mRNA was absent from −/− brain (the calculated −/− expression level and error bar were both smaller than the thickness of the x-axis).

Southern blot analyses with 5′ (Xba I–Xba I) and 3′ (Xba I–Sac I) probes flanking the targeted region were used to confirm that a homologous recombination event had occurred in the genomic DNA. The Eco RV restriction endonuclease site in the Kcna2 open reading frame (ORF) was removed following homologous recombination, allowing the 5′ and 3′ probes to detect fragments of increased size in the targeted allele (Fig. 1B shows sizes detected by the 5′ probe). Positive clones were expanded and injected into C57BL/6J blastocysts (The University of Cincinnati and Children's Hospital DNA Core Facility), and the chimeric mice generated from these blastocysts were bred to test for germline transmission of the mutant allele.

Breeding and genotyping

The mutation was transferred into C3HeB/FeJ mice (backcross generations N5–20) to create Kcna2-null mice (referred to here as −/− mice) and their wildtype (+/+) and heterozygous (+/−) littermates. A chimeric founder was also crossed to C57BL/6J mice, establishing a hybrid line and providing mutant mice referred to as −/− (B6/129) in this text. Breeding of the +/− mice took place in AAALAC-approved specific-pathogen–free facilities. Offspring were genotyped using tail clips taken from mice aged P6 or older and ear-punched for later identification on the day of experiments. When necessary, mice were killed by CO2 exposure, followed by decapitation. All animal protocols were reviewed and approved by the University of Washington IACUC or the University of Wisconsin IACUC.

DNA was isolated from the tail and then PCR-amplified using two primer sets (supplemental Table S1)1 : one to match the ORF of Kcna2, indicating the presence of an intact Kcna2 gene; the other to match the neomycin resistance cassette, indicating a Kcna2-null chromosome. The size of the DNA fragments generated told us which mice in a litter were wildtypes, heterozygotes, and Kcna2-nulls. (For further details, see http://depts.washington.edu/tempelab/Protocols/KCNA2.html.)

Western blot analysis of Kv1.2 protein expression

Total protein was isolated from whole brains and enriched for membrane-associated proteins. Brains were homogenized in 320 mM sucrose with protease inhibitors and cellular debris was pelleted by centrifugation at 1,000 × g. The supernatant was centrifuged at 120K × g for 1 h and the resulting pellet resuspended in 10 mM Tris (pH 7.4), 150 mM NaCl, and 1% Triton X-100 with protease inhibitors. Unsolubilized membrane proteins were removed by centrifugation at 120K × g for 1 h and the supernatant was saved. All steps were done on ice or at 4°C. Protein concentrations were determined using the BCA Protein Assay (Pierce, Rockford, IL).

Protein samples from each animal (20 μg) were denatured at 50°C for 10 min in loading buffer containing 100 mM dithiothreitol and separated by 8% SDS–PAGE. Protein was then electroblotted to a nitrocellulose membrane (BioRad, Hercules, CA). Membranes were blocked in phosphate-buffered saline Tween-20 with 10% nonfat milk and probed with an anti-Kv1.2 monoclonal antibody (1:5,000; Upstate, Waltham, MA). Anti-mouse horseradish peroxidase–conjugated secondary antibodies (1:1,000) were used for detection with enhanced chemiluminescent reagents (Amersham Biosciences, Piscataway, NJ). Blots were exposed to film for the amount of time necessary to obtain a clear image. Blots were then reprobed with anti-β-actin monoclonal antibody (1:1,000; Abcam, Cambridge, MA) and reprocessed for chemiluminescence detection.

Quantitative real-time PCR

qPCR techniques used here were similar to recently published procedures (Duncan et al. 2006; Silverstein and Tempel 2006), including the normalization of Kcna expression levels to the geometric mean of the three most stable internal reference genes from a panel of ten candidate reference genes, as described in Vandesompele et al. (2002). To measure Kcna gene expression versus genotype, +/+, +/−, and −/− mice were killed at age P14, and the whole brain was homogenized in 6 ml of Trizol (Invitrogen, Carlsbad, CA) and purified. To measure changes in brain stem Kcna expression during development C3HeB/FeJ control mice were killed at various ages (P1–P29). The brain was placed in RNAlater (Qiagen), after which the brain stem was dissected away and frozen at −80°C, for subsequent purification. Total RNA yield was measured spectrophotometrically and RNA integrity was verified by gel electrophoresis.

cDNA was reverse transcribed from 1 μg of total RNA using 50 pM of Random Hexamers (Amersham Biosciences) and 1 μl Powerscript Reverse Transcriptase (Clontech/BD Biosciences, Franklin Lakes, NJ) as detailed in Duncan et al. (2006). For each transcript of interest, qPCR primer pairs were designed using Primer3 software (Rozen and Skaletsky 2000) and following design criteria of: amplicon size 50–150 bp, primer Tm of 60–64°C, primer %GC of 35–65%, and complementarity between and within primers minimized (see supplemental Table S1 for sequences).

qPCR using about 5 ng starting total RNA was measured using SybR Green Supermix (BioRad). Cycling parameters were as follows: 95°C × 3 m (activate enzyme), 40 repeats of 95°C × 30 s, 60°C × 30 s (amplification), 95°C × 1 m, 55°C × 1 m (premelt curve), 90 repeats of 10 s each starting at 55°C and incrementing 0.5°C per step (melt curve). Real-time fluorescence was measured using the iCycler I/Q Module (BioRad). In each experiment, all samples were divided into two replicate reactions (same biological sample and reaction mixture), and the average threshold cycle number for the two replicates was used as a single data point. For each primer set melt curves were inspected and efficiencies determined as described in Duncan et al. (2006).

Kcna expression levels were normalized to the geometric mean of multiple internal reference genes, as previously described (Vandesompele et al. 2002). For all of the samples of a given data set, the qPCR threshold cycles were determined for a panel of ten candidate reference genes; the three most stable reference genes were used, as ranked by geNORM software (http://medgen31.ugent.be/jvdesomp/genorm/). These were β-actin, γ-actin, and succinate dehydrogenase complex, subunit A for the cross-genotype study (Fig. 1D); and β-actin, γ-actin, and hydroxymethylbilane synthase for the developmental study (Fig. 3). The primer sequences used for the three Kcna genes and these four reference genes are shown in supplemental Table S1.

The expression levels for potassium channel genes Kcna1, Kcna2, or Kcna6 were normalized as kcnai/r, where kcnai is the copy number for the relevant Kcna gene (i = 1, 2, or 6) and r is the geometric mean of the copy numbers for the three reference genes. The copy number for each gene was its “base of amplification” value B, raised to the power of its threshold cycle number. Final reported values of kcnai/r were based on multiple independent qPCR runs from multiple animals (details in legends of Figs. 1 and 3).

Motor tests and seizure testing

We tested gross motor behavior and susceptibility to flurothyl-evoked seizures in 28 mice aged P14 (three litters). Mice were weighed before the motor tests, which included 1) lowering the mice toward a cage top, to see whether they could grasp the rungs with their forelimbs, and testing the grip strength by pulling gently upward on their tail, 2) letting the mice climb vertically upward on cage rungs, and 3) letting them balance on a 1-cm-diameter aluminum rod held horizontally about 10 cm above the cage floor.

Later on the same day, these mice were tested for seizure susceptibility. After placing a mouse into a Plexiglas chamber of volume 10.7 liters, we applied the volatile convulsant flurothyl (Sigma–Aldrich, St Louis, MO) by dripping it in liquid form onto a filter paper at 20 μl/min. The chamber was cleaned and aerated between subjects. Two human observers noted the latencies from the first drip of flurothyl to the onset of various seizure-related behaviors, which included “flagpole” tail dorsiflexion (Straub tail), unilateral myoclonic jerks, tremors, and bilateral forelimb clonus. Also noted was the latency to a fully generalized seizure, which consisted of 2 to 5 s of running-bouncing seizure (RBS) followed by full tonic extension. At this point, the mouse was rapidly removed into fresh air and given abdominal massage to restore breathing, before being placed in a cage for observation.

Because Kcna1-null mice display a dramatic tremor when forced to swim in cold water (Zhou et al. 1998) we performed similar swim tests on Kcna2-null mice [in this case the −/− (B6/129) mice]. A tank, 18 × 29 cm (width × length), was filled with water to a depth of 7 cm. Mice were placed in the middle of the tank to swim. The water temperature was 17°C. The swim time was 2 min. After swimming, the mice were placed on a dry platform at room temperature for observation of abnormal motor behavior.

MNTB electrophysiology

BRAIN STEM SLICE PREPARATION AND RECORDING TECHNIQUES.

The preparation of brain stem slices containing MNTB neurons was carried out using previously published techniques (e.g., Brew et al. 2003). In brief, mice aged P9–P16 were killed by brief CO2 exposure, followed by decapitation and brain dissection. A tail clip was collected for confirmation of prior genotyping (see previous text). Dissection and slicing took place in ice-cold sucrose-based solution containing (in mM): 250 sucrose; 2.5 KCl; 26 NaHCO3; 1.25 NaH2PO4; 2 CaCl2; 1 MgCl2; 10 glucose; 2 Na pyruvate; 0.5 Na ascorbate; and 3 myo-inositol (pH was maintained at 7.4 by gassing with 95% O2-5% CO2 mixture). After separation of the forebrain by a transverse cut through the colliculi, the cut rostral surface of the brain stem was glued to the chamber of a vibrating slicer (Vibratome Series 1000; Technical Products International, St. Louis, MO). Slices of 150 μm each were cut, up to five of which contained MNTB. Slices were kept for 1 h at 37°C in an incubation chamber filled with artificial cerebrospinal fluid (ACSF) and gassed with 95% O2-5% CO2 mixture. The ACSF was identical to the sucrose-based solution described earlier except that it contained 125 mM sodium chloride instead of the 250 mM sucrose. The incubation chamber was allowed to cool to room temperature and slices were maintained there for up to 8 h before use. Slices were then placed in a recording chamber, also perfused with gassed ACSF at room temperature (22–25°C), sited on a stable recording platform built around a microscope (on an X–Y stage) fitted with differential interference contrast optics (Axioskop or Axioskop FS2, Zeiss, www.zeiss.com).

Recording pipettes were borosilicate glass (Vertical Pipette Puller 700B; David Kopf Instruments, Tujunga, CA) and filled with a solution containing (in mM): K gluconate 97.5, KCl 32.5, EGTA 5, HEPES 10, MgCl2 1 (adjusted to pH 7.2 using ∼14 mM KOH). For voltage-clamp recordings, pipettes were coated with Sylgard (Dow-Corning, Midland, MI) to reduce their capacitance. Pipettes were connected to the patch-clamp amplifier (Axopatch 200; Axon Instruments, Foster City, CA) by the amplifier headstage mounted on a micromanipulator [Narishige (Tokyo, Japan) or EXFO (Burleigh, Victor, NY)] bolted to the recording platform. Pipette resistance was 3–6 MΩ before gigaohm seal formation, measured from the responses to −5-mV voltage-clamp steps. Although recording locations were not measured, attempts were made to record from a variety of locations throughout the MNTB, to try to fully represent the properties of all MNTB neurons, and also to minimize bias from any variations in Kv1-based potassium current amplitudes across the tonotopic axis (e.g., as reported in rat MNTB; Brew and Forsythe 2005). The access resistance was monitored frequently throughout recordings and typically rose to about 10 MΩ on achieving the whole cell recording configuration (range 5–18 MΩ). Recordings were discontinued if the pipette access resistance went to >20 MΩ. There was a −7-mV liquid junction potential that is included in all the subsequently given membrane potential values.

Soon after the start of voltage-clamp recordings, the slice was perfused with ACSFV (a low-calcium version of the ACSF) containing 0.5 mM CaCl2, 2.5 mM MgCl2, and 0.5–1 μM tetrodotoxin to minimize sodium currents, calcium currents, and synaptic activity. The pipette access resistance was compensated using the series resistance compensation circuitry of the patch-clamp amplifier (“correction” dial at 85% and “prediction” dial at 85%).

Chemicals were from one source (Sigma–Aldrich) except sucrose and NaH2PO4 (J.T. Baker, Phillipsburg, NJ). Tetrodotoxin (TTX), dendrotoxin-I (DTX), and tityustoxin-Kα (TsTx) were from another source (Alomone Labs, Jerusalem, Israel). TsTx (100 nM) or DTX (100 nM) plus TsTx (100 nM) was added to the ACSFV and applied by perfusion. Because DTX block is almost irreversible in slices, for recordings following previous toxin applications, we used a fresh slice, replaced the perfusion tubing and reservoirs, and either replaced the recording chamber or washed it with dilute HCl for 60 min.

STIMULUS GENERATION AND DATA ACQUISITION.

The software for stimulus generation and data acquisition [either Synapse (Synergy, Bethesda, MD) or Axograph (Axon Instruments)] ran on a Macintosh computer (7100/80AV or G3) and sent command sequences to the patch-clamp amplifier by an AD/DA interface (ITC-16; Instrutech, Port Washington, NY). This interface also digitized signals on two channels from the patch-clamp amplifier so that they could be digitally recorded by the software. During current-clamp experiments, the two channels recorded were the current-clamp pulse amplitude and the resulting pipette potentials, each recorded at a digitization rate of 10 kHz (amplifier filtering 5 kHz). During voltage-clamp experiments, the two channels recorded were the voltage-clamp pulse amplitude and resulting pipette currents, each recorded at a digitization rate of 5 kHz (amplifier filtering 2 kHz).

Current-clamp pulses were applied at 1-s intervals, were of 180-ms duration, and increased in 10-pA increments from 0 to 200 pA, or from −100 to 200 pA. This is the same protocol as that previously used for the study of MNTB neurons in Kcna1-null mice, to allow straightforward comparisons between the data sets (e.g., supplemental Table S2 and Table 1 of Brew et al. 2003).

To measure current–voltage (IV) relations in voltage-clamped MNTB neurons, a standard IV protocol was used (also used previously for the Kcna1-null MNTB study; Brew et al. 2003). Voltage-clamp command test pulses were applied at approximately 1.3-s intervals, were of 180-ms duration, and increased in 10-mV increments from −40 to +90 mV. These test pulses were applied over a continual background potential of −60 mV, set using the amplifier, and the −7-mV liquid junction potential. This −67-mV holding potential was necessary to inactivate the large A-currents present in murine MNTB neurons (see Fig. 7A and supplemental methods). The resulting patch pipette potentials were −107 to +23 mV. Each pulse sequence included an initial 150-ms pulse to −107 mV (intended to remove any potassium channel inactivation that had occurred during previous test pulses), an 870-ms prepotential of −67 mV (to inactivate A-type potassium currents), the 180-ms test pulse, a 50-ms postpotential at −37 mV, and a subsequent step back to −67 mV for 50 ms. The 245-ms section recorded included the final 15 ms of the prepotential, the 180-ms test pulse, and the 50-ms postpotential.

Two alternative protocols differed from the standard protocol described earlier only during the 180-ms test pulses section. One was a “high-voltage–resolution” IV protocol, designed to increase the accuracy of the measurements of voltage dependence, by using smaller test pulse increments of 5 mV within a smaller range of test potentials, −107 to −22 mV (this protocol and the standard IV protocol were both used for analyses of voltage dependence in Fig. 12). The other was a “toxin-monitoring” protocol, which did not contribute to any data analyses. It consisted of repeated test pulses to −47 mV to facilitate visualization of the rapid decreases in current amplitudes caused by toxin perfusion (as shown in Fig. 9A). The frequent monitoring allowed experimenters to judge when current amplitudes had become stable (typically 1–3 min after the start of recordings or a switch of perfusion solutions), which prompted the collection of IV data (by performing several runs of the IV protocols).

DATA ANALYSIS.

Current-clamp data and voltage-clamp data from MNTB neurons were analyzed, fitted, and graphed using macros within the acquisition software and additional software (Kaleidagraph; Synergy Software, Reading, PA).

Current-clamp data for each neuron were measured from a single run of the current-pulse protocol before any averaging across runs or between neurons was performed. All membrane potentials presented include the addition of the −7-mV junction potential and a voltage correction for current flow across the pipette tip access resistance, both carried out off-line. The sustained potential response to a particular current pulse amplitude was the average potential during a 10-ms time window toward the end of the pulse (100 data points, 160 to 170 ms after current-pulse initiation). The resting potential was taken to be the sustained potential response to the 0-pA current pulse. Sustained potential responses were also used for calculations of input resistance (in supplemental Table S2).

In voltage-clamped neurons, the sustained outward current amplitudes (I) were measured from each test potential of the standard IV protocol or the “high-voltage–resolution” IV protocol as follows. First the absolute current amplitude was measured at the end of each 180-ms test pulse, as the average of the 50 data points between 160 and 170 ms after test pulse initiation. The current amplitudes at membrane potentials between −97 and −77 mV represented background “leak current” and were used to calculate the linearly extrapolated current at other test potentials, which was then subtracted from the absolute currents to yield the sustained outward current amplitudes, I. For each neuron and in each perfusion solution, these values of I were also averaged across two to three runs of the respective IV protocol. To obtain a measure of that neuron's current through Kv1 channels, the sustained IKv1, these average I values in the presence of both TsTx and DTX, were subtracted from the I values measured in ACSFV alone.

The membrane potentials referred to in the subsequent voltage-clamp results and figures are the pipette potentials with a −7-mV correction for the liquid junction potential, but do not include any voltage correction for current flow across the uncompensated 15% of the pipette tip access resistance (range 1–3 MΩ). This omission facilitated the averaging of data collected at the same pipette test potentials and is justified because the estimated corrections were all small in the potential range relevant to Kv1 channels (e.g., at −37 mV, the mean amplitude of the sustained current was +760 pA before leak subtraction, resulting in an average correction of −1.5 mV).

To fit the apparent activation time course of an MNTB neuron's IKv1, the current records were averaged from two to three runs of the standard IV protocol in ACSFV and two to three runs in the presence of both TsTx and DTX; the latter were subtracted from the former (which removed capacitative current transients). The resulting currents were well fitted by curves increasing with a single-exponential time course at test potentials −47 and −57 mV. Fits were performed on a time period of 20 ms (i.e., 100 data points) beginning soon after the start of the test pulse (1.2–21.1 ms). Because data were available from a larger number of neurons in ACSFV alone, the apparent activation time course was also fitted from each neuron's averaged IV responses in ACSFV, which required a nonstandard method of subtracting the leak and capacitative transient currents, as follows. First, the record containing the −77-mV pulse was subtracted from that containing the −87-mV pulse, which yielded the estimated passive “leak plus capacitance” current response to a single −10-mV pulse (the −67-mV pulse could not be used because some sustained outward current including IKv1 was usually present). Next, this passive “leak plus capacitance” estimate was multiplied by the relevant n before adding it to the current responses to test pulses differing by n increments of 10 mV from the −67-mV holding potential. Finally the current fits were performed (as described earlier for IKv1).

DATA SETS AND DATA EXCLUSION CRITERIA.

The current-clamp data are from a total of 68 MNTB neurons and 27 mice (26 from 15 −/− mice, 23 from 12 +/+ mice, and 19 from 10 +/− mice). The voltage-clamp data are from another 29 MNTB neurons and 21 mice (11 from 8 +/+ mice, 8 from 7 +/− mice, and 10 from 6 −/− mice). The recordings from neurons of each genotype had similar mean pipette access resistances and capacitances, as read off the amplifier dials, and were from mice of similar ages (see supplemental Table S2).

Current-clamped neurons were excluded if the resting potential was smaller than −57 mV (including the −7-mV liquid junction potential). Voltage-clamped neurons were excluded if a holding current larger than −100 pA was required to clamp at the background holding potential of −67 mV. We made exceptions to this criterion for two voltage-clamped neurons to which we subsequently successfully applied both toxins (one +/+ neuron and one −/− neuron requiring −160 and −190 pA, respectively). On average, the holding current was near zero for the remaining +/+ and −/− voltage-clamped neurons (+/+, −2 ± 24 pA, n = 10; −/−, −5 ± 18 pA, n = 9) and slightly larger for +/− neurons (−50 ± 11 pA, n = 8).

FITTING OF CONDUCTANCE DATA.

The conductance (G) and its toxin-sensitive component (GKv1) were calculated at each test potential (V) by dividing sustained current amplitudes (I or IKv1) by the driving force (V minus −80 mV, the approximate reversal potential for DTX-sensitive channels; Brew and Forsythe 1995; Stansfeld and Feltz 1988). The calculated G values were plotted versus V (−107 through −37 mV) and fitted with a single Boltzmann function G = Gmax/{1 + exp[(VVhalf)/−k]} with variable parameters Gmax, the maximum conductance, Vhalf, the half-activation voltage, and k, the slope factor. The fitting used a least-squares curve-fitting method (the Pearson's R general curve fit within Kaleidagraph software, which finds the minimum value of χ2, the sum of the squared residuals, using partial derivatives according to the Levenburg–-Marquardt algorithm, described in Press et al. 1992). In Fig. 12, 56 fits and data are shown normalized to their fitted Gmax values, to facilitate visual comparisons, but the subsequent description of goodness-of-fit is based on the fits before normalization.

The 56 fits shown in Fig. 12 appeared good to the naked eye, had small mean errors returned with each fitted parameter (Gmax, 0.9 nS; Vhalf, 1.6 mV; and k, 0.8 mV), and values of R2 > 0.93 (53 fits had R2 > 0.98). Of these, 46 fits were judged “very good” because the returned errors were <2 units for all three parameters (shown by solid lines in Fig. 12). Fits were judged too poor to be included if one or more parameters were returned with estimated errors >10 units, but in most cases the same neuron's current responses from an alternative protocol returned a good fit. Thus only one neuron was excluded from a plot because of a poor fit (a −/− neuron excluded from Fig. 12A). If both protocols returned very good fits, those from the high-voltage–resolution protocol were preferred for inclusion in Fig. 12 and statistical comparisons. Also shown are 10 “less-good” fits, for which at least one parameter was returned with an error estimate >2 units but <10 units (dashed lines in Fig. 12). The majority of these “less-good” fits came from +/+ neurons (7 of 10). This was probably partly because the conductance was not near its maximum value at −37 mV in +/+ neurons, whereas at that potential the +/− and −/− conductances had reached or closely approached their maxima (Fig. 12A; also see results). The results and discussion therefore focus on the +/− and −/− results, and only tentative conclusions are made from the +/+ data.

Statistical evaluation

Tests were unpaired two-tailed Student's t-tests assuming equal variance in each sample, except as noted for paired comparisons or comparisons involving appreciably nonnormal distributions, for which we used Mann–Whitney U tests. All tests were performed using StatView (SAS Institute, Cary, NC) or Kaleidagraph. All averaged values are expressed in text and figures as mean ± SE.

RESULTS

Verification of Kcna2 knockout and Kcna expression in +/+, +/−, and −/− brains

Our targeting strategy successfully knocked out the entire open reading frame of the Kcna2 gene (Fig. 1, A and B). The lack of Kv1.2 protein in the brains of Kcna2-null (−/−) mice was confirmed by Western blot (Fig. 1C). By qPCR there was also no Kcna2 mRNA present in brains from −/− mice aged P14, and Kcna2 mRNA expression was approximately halved in +/− compared with +/+ littermates' brains (Fig. 1D). The Western blot is suggestive of a reduced amount of Kv1.2 protein in +/− brain relative to +/+ brain (Fig. 1C). These results verify that both Kcna2 mRNA and Kv1.2 protein are absent in the −/− mouse brain.

To test whether other Kcna genes had altered expression as a result of the absence of Kcna2 in −/− mice, the qPCR experiments also measured Kcna1 and Kcna6 mRNA expression, which were found to be very similar between +/+, +/−, and −/− brains (Fig. 1D). These two genes code for Kv1.1 and Kv1.6 subunits, which are known to be expressed along with Kv1.2 in both rat and mouse MNTB neurons (Brew et al. 2003; Dodson et al. 2002; Fonseca et al. 1998). Because the expression of each Kcna gene was normalized to the same three reference genes, these data suggest that Kcna1 mRNA may be expressed at a higher level than either Kcna2 or Kcna6 in +/+ brain. The brains of P14 −/− mice did not exhibit any signs of compensatory changes or greater variability in mRNA expression for Kcna1 or Kcna6 because both their mean expression and variability were similar to those of +/+ and +/− brains (Fig. 1D).

Life span and gross motor behavior in Kcna2-null mice

The −/− mutation did not cause embryonic lethality, judging by the Mendelian proportions of each genotype in 14 litters consisting of 88 mice, none of which was subjected to any experimental testing (+/+, n = 21, 23.8%; +/−, n = 45, 51%; and −/−, n = 22, 25%). However, the −/− mice had a severely reduced life span, averaging 17 ± 0.2 postnatal days (Fig. 2 A, range P16–P19). Their littermate +/+ and +/− mice had normal life spans. In litters removed from the specific-pathogen–free facility for experimental testing, some −/− mice had even shorter life spans (range P14–P19). The mean life span was also 17 ± 2 postnatal days in eight −/− (B6/129) mice, although the range was much larger [one −/− (B6/129) mouse died at P6, the rest at P14–P25] perhaps because the mutation was in a genetic background with seizure-resistant (B6) and relatively seizure sensitive (129) alleles sorting in the mixed background (Frankel et al. 2001). Nonetheless, the data suggest the null mutation was penetrant in more than one genetic background. Adult +/− mice were good breeders, suggesting that they flourish despite the substantially reduced expression of Kcna2 mRNA and Kv1.2 protein demonstrated at P14 (Fig. 1, C and D).

FIG. 2.

Survival curves and seizure susceptibility in +/+, +/−, and −/− mice. A: percentage of mice surviving through each postnatal age (e.g., if they died at age P16, they survived through P15). All of the 22 −/− mice died at age P19 or younger, whereas none of the 21 +/+ mice or the 45 +/− mice died during the first 30 days. In total, 88 mice were from 14 litters, with roughly Mendelian proportions of each genotype and balanced gender proportions (45 males and 43 females). Line through the points was drawn by eye. B: mean latencies to the first seizure-related behavior (left bars) and running-bouncing seizure (RBS, middle bars) for 6 +/+ mice, 15 +/− mice, and 7 −/− mice, following the first drip of flurothyl onto the filter paper in the exposure chamber. Mean time between the first seizure-related behavior and the RBS is also shown (right bars). Significant differences between genotypes are denoted by single or double asterisks (P < 0.05 or P < 0.005, respectively; Mann–Whitney U tests). −/− mice differed significantly from +/+ and +/− mice on all 3 measures.

The −/− mice appeared physically normal during their first 2 wk of life. At P14 the majority of −/− mice had their eyes open and exhibited age-appropriate motor behavior (walking, running, rearing, grooming, exploration). We tested gross motor behavior in three litters of 28 mice aged P14 (see methods). All the mice passed motor tests 1–3, i.e., could grasp and climb vertically on cage rungs and balance on the horizontal rod. The −/− mice weighed about 1 g less (7.5 ± 0.17 g, n = 7) than their +/+ littermates (8.9 ± 0.20 g, n = 6, P < 0.05) and their +/− littermates (8.3 ± 0.19 g, n = 15, P < 0.05).

“Spontaneous” seizures in Kcna2-null mice

Many −/− mice were observed undergoing apparently spontaneous episodes that began with a sudden explosive onset of wild running and jumping, followed after 5–10 s by full tonic extension (TE). This sequence is highly reminiscent of (RBS), leading to TE, which is a typical feature of a fully generalized seizure in rats and mice (Gale 1992). When −/− mice were observed recovering from TE, they often exhibited myoclonic jerks and tremors, followed by 5–20 min of relative immobility before recovery of normal motor behavior, reminiscent of a postictal phase. The fatality rate of observed RBS/TE events was high, about 50%, and was probably due to the cessation of movement and breathing that accompanied TE. It is likely that all the −/− mice eventually experienced fatal RBS/TE because all those that died unobserved were found with their limbs and bodies fully extended. Thus the RBS/TE episodes in −/− mice are probably seizures with generalized onset, and the proximal cause of the reduced −/− life spans was probably fatal apnea occurring during TE.

The exact frequencies, fatality rates, and ages of occurrence of all RBS/TE events could not be recorded, although some conclusions can be drawn from the following quantitative estimates, as remembered by two human observers (L.R. and H.B.) who were with litters including −/− mice for ≤60 min per day of routine mouse care or experimental preparations. At least 50 RBS/TE episodes were observed in about 100 −/− mice aged P15–P19. This is more than would be expected if observed and unobserved RBS/TE events occurred at the same rate and had the same fatality rate (50%), in which case a typical −/− mouse would undergo only one or two RBS/TE events in its lifetime, and the expected number of events in 100 −/− mice would be 200, only eight of which should coincide with the daily hour of observation. Both observers reported that many RBS/TE events began when the observer moved the cage slightly, perhaps suggesting they were evoked by sounds or accelerations or stress. The fatality rate did not appear to increase with age because RBS/TE was fatal in the two youngest −/− mice ever observed undergoing RBS/TE, aged P14, and nonfatal RBS/TE events were observed in several −/− mice at ages P17 and P18, and in a −/− (B6/129) mouse aged P17, which survived but died later the same night. RBS/TE events were very rare and about 50% fatal during several days of quiet extended observation of five −/− mice in a cage with a foster mother, and were initiated from sleep or normal awake behavior without any provoking stimuli apparent to the human observer. Thus RBS/TE events in −/− mice were probably infrequent (less than one per day between P15 and P19), 50% fatal, and some observed RBS/TE events may have been evoked by unknown stimuli.

Susceptibility to seizure induction of Kcna2-null mice

To test whether the −/− CNS was abnormally susceptible to evoked seizures, mice aged P14 were tested for their latencies to flurothyl-induced seizures (see methods; same 7 −/− mice, 15 +/− mice, and 6 +/+ mice as in the gross motor testing described earlier). At this age, it was likely that none of the −/− mice had yet experienced an RBS/TE generalized seizure event. All mice behaved normally for a few minutes after the first drip of flurothyl into the exposure chamber, then exhibited seizure-related behaviors (see methods), eventually culminating in RBS followed by TE. The mean latency to occurrence of the first seizure-related behavior was shortest in −/− mice, intermediate in +/− mice, and longest in +/+ mice (Fig. 2B, left). The mean RBS latency for −/− mice was 40% shorter than that in +/+ or +/− mice (Fig. 2B, middle). The −/− mice progressed very rapidly from their first seizure-related behavior to RBS, taking an average of 23 s, whereas their +/+ and +/− littermates displayed ongoing seizure-related behaviors for about 2 min before progressing to RBS (Fig. 2B, right). There was no overlap in the range of latencies to flurothyl-induced seizures aged P14 in the seven −/− mice (182–251 s) and their +/− and +/+ littermates (311–465 and 292–455 s, respectively). Abdominal massage restored breathing for 18 of these 28 mice, including three −/− mice. The shorter latencies to RBS in −/− mice suggest there was network hyperexcitability present in the −/− CNS at P14.

Although the seizure profile of −/− mice (described earlier) is distinct from that of Kcna1-null mice (partial seizures from age P21 onward) the Kcna1-null mice did also have reduced seizure latencies in response to flurothyl at ages as young as P10 (Rho et al. 1999). Because Kcna1-null mice showed abnormal tremor after a cold swim, we performed similar cold-swim tests on five −/− (B6/129) mice (see methods). However, unlike the Kcna1-null mice, the Kcna2-null mice did not exhibit any signs of hyperexcitability (body tremors) after a cold swim.

Could sound be one of the stimuli able to induce RBS/TE in −/− mice? This question arises because RBS/TE can be induced audiogenically in certain strains of rats and mice that are either genetically susceptible or have been made susceptible by early partial deafening (see discussion). Typically, loud sounds of 100–130 dB lead after 2–20 s to RBS, closely followed by clonic seizures and/or TE (Ross and Coleman 2000). In preliminary tests, 70 of 71 applications of an octave band stimulus 8–16 kHz (112 dB SPL, 20-s duration) failed to induce RBS or TE in six −/− mice tested at ages P14–P18 (data not shown).

Developmental expression of Kcna mRNA in +/+ brain stem

To explore whether the second and third postnatal weeks might be an especially important period in the development of CNS Kv1 channels, the qPCR technique was used to measure the developmental time course of Kcna2 gene mRNA expression in the brain stems of 14 +/+ mice, two mice at each of seven ages tested (Fig. 3A). The expression level increased 10-fold between P1 and P29 for Kcna2 mRNA (Fig. 3A). This increase in +/+ Kcna2 expression occurs at approximately the same age as the onset of seizures in −/− mice (see discussion).

FIG. 3.

Developmental expression in brain stem of mRNA for Kcna2, Kcna1, and Kcna6. A: expression of Kcna2 mRNA in +/+ brain stems from 14 mice killed at ages P1–P29. Expression is given relative to the geometric mean of the 3 control genes β-actin, γ-actin, and hydroxymethylbilane synthase (supplemental Table S1 and methods). At each age, the expression level is plotted for 2 mice, as the mean of 3–5 qPCR measurements (SE also plotted for each mouse using either narrow or wide error bars). Line is a Boltzmann fit to the 14 mean values plotted. B: relative mRNA expression as in A, but shown as the average of both mice at each age, for Kcna2 (triangles), Kcna1 (inverted triangles), and Kcna6 (diamonds). At each age, Kcna1 and Kcna6 expression are shown as the mean and SE for 4 qPCR measurements, 2 from each mouse. Kcna2 expression was the mean and SE from 6–9 qPCR measurements. Each line is a Boltzmann fit to the 7 mean values for that gene.

The developmental time course of Kcna1 and Kcna6 gene mRNA expression was also measured using the same brain stem tissue samples (Fig. 3B). The mRNA expression increased approximately thirtyfold for Kcna1 but only twofold for Kcna6 (Fig. 3B). Kcna1 mRNA and Kcna2 mRNA reached their half-maximal expression at P12 and P11, respectively (values from fits to Boltzmann functions; Fig. 3B). A comparison of the relative expression of these three Kcna genes is also of interest. The data suggest that neonate brain stems had higher expression of Kcna6 mRNA than Kcna1 and Kcna2 mRNA, whereas the opposite was true for juvenile brain stems (Fig. 3B). Also, Kcna1 and Kcna2 mRNA were present at similar copy numbers in neonatal mice, whereas in juvenile mice Kcna1 mRNA contributed more than twice as many copies as Kcna2 mRNA (Fig. 3B). If this reflects the ratios of Kv1 subunit proteins produced, it suggests that the order of expression strength in neonatal brain stem may be Kv1.6 ≫ Kv1.2 > Kv1.1, whereas this order is reversed during maturation, becoming Kv1.1 > Kv1.2 ≫ Kv1.6.

Kcna2-null MNTB neurons were hypoexcitable

One obvious way in which a lack of Kv1.2 might cause network hyperexcitability in the −/− CNS would be by reduced potassium currents causing hyperexcitability in individual neurons. Although it is probably unlikely that MNTB neurons within the brain stem auditory system play a role in seizure susceptibility in either Kcna1-null or Kcna2-null mice, MNTB neurons are a useful model for study because even small reductions in their IKv1 have large effects on excitability (Brew and Forsythe 1995; Brew et al. 2003).

To find out whether −/− MNTB neurons had abnormal excitability, we recorded and analyzed the responses to current pulses (180-ms duration, −100 to 200 or 0 to 200 pA, in 10-pA increments) of 63 MNTB neurons in brain stem slices from mice aged P9–P16 (9 +/+ mice, 14 +/− mice, and 15 −/− mice). The responses of three example MNTB neurons, each typical of their genotype, are shown in Fig. 4A. The −/− neuron fired the smallest numbers of APs, e.g., only a single initial AP at the start of a 200-pA pulse contrasting with three APs generated by the +/+ neuron (Fig. 4A, top traces). The −/− neuron also had the highest threshold current amplitude (defined as the smallest current pulse amplitude that generated at least one AP) of 160 pA, contrasting with 80 pA for the +/+ neuron and 130 pA in the +/− neuron. Overall, the threshold current amplitudes were significantly smaller in the 21 +/+ MNTB neurons than those in the 25 −/− neurons (P < 0.0001) or the 19 +/− neurons (P < 0.005, Fig. 4B). The +/− neurons’ threshold current amplitudes were distinctly intermediate between +/+ and −/− neurons because they were also significantly smaller than those of the −/− neurons (P < 0.05).

FIG. 4.

−/− Medial nucleus of the trapezoid body (MNTB) neurons are less excitable by current pulses than +/+ neurons. Threshold currents and action potential (AP) numbers were measured from +/+, +/−, and −/− MNTB neurons subjected to 180-ms-duration current pulses −100 to 200 pA, applied at 1-s intervals, incrementing by 10 pA. A: example responses of typical +/+, +/−, and −/− MNTB neurons to the current-pulse protocol, shown here at 50-pA intervals. Linked arrows on the top traces mark the time during which sustained membrane potential was measured (used for Fig. 5 and for calculating the input resistances and resting potentials as described in results and supplemental Table S2). BD: summary data on threshold currents and AP numbers from 21 +/+ neurons (filled circles), 17 +/− neurons (triangles), and 25 −/− neurons (open circles). For each neuron, data came from a single run of the current-clamp protocol. Significant differences between the genotypes are shown by single, double, or triple symbols (P < 0.05, P < 0.01, and P < 0.005, respectively). B: threshold current amplitudes required to generate at least one AP in the MNTB neurons of each genotype. Threshold current amplitudes were larger in −/− neurons and +/− neurons than in +/+ neurons. C: mean number of APs generated at each current-pulse amplitude. Error bars show the SE. −/− and +/− neurons fired significantly fewer APs than the +/+ neurons. Significant differences between +/+ and −/− values are shown by small asterisks, between +/+ and +/− values by large asterisks, and between +/− and −/− values by circumflexes. D: mean numbers of APs during the 5 largest current steps tested (160–200 pA). This value was termed the APN(160–200) and it was largest in +/+ neurons, smallest in −/− neurons, and intermediate in +/− neurons.

Both the −/− and the +/− neurons fired fewer APs than +/+ neurons, for every current-pulse amplitude ≥80 pA (Fig. 4C, small and large asterisks show significant differences). There were only three current-pulse amplitudes at which +/− neurons fired significantly larger numbers of APs than −/− neurons (Fig. 4C, circumflexes). The genotypic differences in AP numbers were also present when data were lumped across all the current amplitudes because the +/+ AP numbers differed significantly from −/− AP numbers or +/− AP numbers (repeated-measures ANOVA, each at P < 0.001). The small overall difference between the AP numbers of −/− and +/− MNTB neurons was also significant (repeated-measures ANOVA, P < 0.05).

To more accurately represent the excitability of each MNTB neuron, and compare the distribution of excitability within each genotype, the AP numbers were averaged across the five largest current pulses tested, 160–200 pA, and this average was termed the APN(160–200). The APN(160–200) values were significantly larger in +/+ neurons than those in −/− neurons or +/− neurons (Fig. 4D).

In Fig. 4 and the statistical comparisons related earlier, five of an original total of 68 recordings were excluded because they fired so many APs that they were statistical outliers (supplemental Fig. S1A, arrows) and their tonic firing also disallowed their inclusion in Fig. 5 and supplemental Table S2 (see following text). These recordings were probably not from MNTB principal neurons and their exclusion made very little difference to any of the statistically significant differences between the genotypes (see legend to Fig. S1).

FIG. 5.

Sustained potentials were more depolarized in +/+ than in −/− or +/− MNTB neurons. A: mean sustained potential responses to current pulses, measured in artificial cerebrospinal fluid (ASCF) for 21 +/+ MNTB neurons and 13 −/− neurons. Depolarizing current pulses led to larger depolarizations in the +/+ neurons than in the −/− neurons. (Same current-pulse protocol as in Fig. 4; same +/+ and −/− neurons as in Fig. 4C and Table S2, except excluding data from 12 −/− neurons to which only depolarizing pulses were applied. Mean responses of these 12 neurons were identical to within 1 mV of the 13 −/− neurons shown.) Sustained potential was defined as the mean of 600 data points toward the end of the current pulse, during the 30-ms period marked by linked arrows in Fig. 4A. Sustained potentials were corrected for current flow across the pipette access resistance. SE is shown by error bars, sometimes obscured by the size of the symbols. Significant differences between genotypes are denoted by single and double asterisks (P < 0.05 and P < 0.005, respectively). B: as in A, but comparing 17 +/− neurons with the 21 +/+ MNTB neurons. Depolarizing current pulses led to larger depolarizations in the +/+ neurons than in the +/− neurons.

There was only slight overlap between the +/+ and −/− APN(160–200) values (Fig. 4D), which suggests almost all of the −/− MNTB neurons had abnormally low excitability, rather than some −/− MNTB neurons having excitability similar to that of +/+ neurons, having somehow compensated for their Kcna2 deficit. In principle, variable levels of compensation for the Kcna2 deficit in +/− and −/− neurons could lead to greater within-genotype variability than that for +/+ neurons, but in fact they exhibited similar variability in threshold currents and smaller variability in AP numbers (e.g., Fig. 4, BD). There was also no sign that +/− neurons and −/− neurons underwent compensation in the age range studied (P9–P16) because they were not more excitable in slices from mice aged P12–P16 than from mice aged P9–P11 (see supplemental Table S2). The significant differences between +/+ neurons and the hypoexcitable −/− and +/− neurons remained after this subdivision into two age groups (Table S2).

These data show that −/− MNTB neurons and +/− MNTB neurons were hypoexcitable compared with their +/+ counterparts. This is especially surprising because it was previously shown that Kcna1-null MNTB neurons were hyperexcitable and had the reduced Ikl amplitudes expected because of their Kv1.1 deficit (Brew et al. 2003). In marked contrast, the deficits of Kv1.2 subunits in −/− and +/− MNTB neurons were associated with hypoexcitability, suggesting their Ikl amplitudes were enlarged, not reduced.

Altered resting membrane properties and I–V relations in −/− MNTB neurons

An enlarged Ikl would be expected to cause slightly less depolarized steady-state IV relations in the range −70 to −40 mV, as well as enlarged conductance (reduced membrane resistance). Analysis showed that −/− MNTB neurons did have less depolarized sustained membrane potentials than those of +/+ MNTB neurons, for all depolarizing current-pulse amplitudes tested, as well as shallower IV relations around rest (Fig. 5A; same neurons and pulse protocol as in Fig. 4; linked arrows in Fig. 4A show the time window of sustained potential measurement). The +/− neurons' sustained potentials were similar to those of +/+ neurons, although there were small significant differences in potential at a few current amplitudes (Fig. 5B). The abnormal sustained IV relations in −/− MNTB neurons support the idea that their sustained Ikl amplitude is enlarged compared with +/+ neurons.

Although the mean resting potentials were slightly larger in the −/− and +/− neurons (−66 and −67 mV) than those in the +/+ neurons (−65 mV) the differences were not significant (supplemental Table S2, resting potential was defined as the sustained membrane potential during the 0-pA current pulse). Table S2 compares the MNTB neurons of each genotype across a range of excitability-related membrane parameters. The mean resting input resistance was about 90 MΩ in −/− neurons, significantly lower than the 130–140 MΩ in +/+ neurons (Table S2; see legend for values and details of both methods used to quantify input resistance). The mean resting input resistance of +/− neurons was similar to that of +/+ neurons (Table S2). These data do not support the idea that there had been variable levels of compensation for the Kcna2 deficit in +/− and −/− neurons because the SDs for each parameter were generally similar in the MNTB neurons of each genotype (Table S2). As expected for mouse strains repeatedly backcrossed into the same inbred C3HeB/FeJ background, the distributions of AP numbers, threshold currents, and other membrane properties for the +/+ MNTB neurons reported here were very similar to those for MNTB neurons from the control littermates of Kcna1-nulls (compare Figs. 4, 5, and supplemental S1 and Table S2 in the present study with Figs. 1 and 5 and Table 1 of Brew et al. 2003).

Action potentials and other properties were similar in +/+ and −/− MNTB neurons

Aside from the above-described differences that could be ascribed to potassium currents, MNTB neurons had similar properties irrespective of genotype. For example, the −/− and +/− sodium channels were probably functioning normally because MNTB neurons of all three genotypes had a similar initial AP during a 200-pA pulse, with an approximate mean latency of 3 ms, mean half-width of 0.9 ms, and a rapid rising phase beginning at −45 mV (supplemental Table S2). MNTB neurons of each genotype were probably of similar size because they had similar capacitances (Table S2). Also, the size of MNTB and its principal neuron somata appeared similar in brain stem slices of each genotype viewed in our recording chamber (data not shown). The pipette access resistances and ages were similar for MNTB neurons of each genotype (Table S2). Thus the genotypic differences in threshold currents, AP numbers, and IV relations described earlier could be produced solely by differences in potassium currents.

Dendrotoxin-I had greater effects on −/− MNTB neurons and abolished genotypic differences

Next, we measured the effects of dendrotoxin (DTX; see introduction) on threshold current amplitudes, AP numbers, and sustained potentials in five +/+ and four −/− MNTB neurons, to test whether −/− hypoexcitability could be caused solely by an enlarged IKv1 (the subcomponent of Ikl clearly attributable to Kv1-type channels because of its DTX sensitivity). If so, DTX should have greater effects on the excitability of −/− neurons than +/+ neurons and there should be no differences in excitability between +/+ and −/− neurons when DTX is present. The typical effect of DTX on MNTB neuron firing is a reduction in the threshold current for an AP and a conversion from phasic to tonic firing (e.g., in rat MNTB; Brew and Forsythe 1995). This phasic firing during a prolonged current pulse was shown earlier for two typical example +/+ and −/− MNTB neurons in control ACSF (Fig. 4A) and both converted to tonic firing throughout current pulses after 100 nM DTX was applied (Fig. 6, A and B). In terms of AP numbers, these +/+ and −/− MNTB neurons fired three APs and one AP, respectively, in ACSF at the example pulse amplitude of 200 pA, and in DTX this increased to 20 and 16 APs. The DTX application caused a much larger reduction in threshold current amplitude for the −/− neuron than for the +/+ neuron (140 vs. 40 pA). Similar results were obtained in all five +/+ neurons and four −/− neurons which had threshold current amplitudes in control ACSF averaging 100 and 170 pA, respectively, each typical for its genotype, and differing significantly from each other (Fig. 6D, left; compare with Fig. 4B). After DTX application, the +/+ neurons and −/− neurons had statistically indistinguishable threshold current amplitudes of about 40 pA (Fig. 6D, center) and fired similarly large numbers of APs (Fig. 6C, triangles). The mean reduction in threshold current amplitudes caused by DTX was approximately twice the size in −/− neurons as that in +/+ neurons (Fig. 6D, right).

FIG. 6.

Dendrotoxin-I (DTX) has greater effects on −/− MNTB neuron excitability than on +/+ MNTB neuron excitability. Responses of MNTB neurons to −100- through 200-pA pulses (protocol as in Fig. 4) were measured in control ACSF, then in DTX (ACSF containing 100 nM DTX). For CF the data from each neuron were averaged from 2–3 runs of the protocol, before averaging across 5 +/+ neurons or 4 −/− neurons. A: responses of an example +/+ MNTB neuron in DTX. (Same neuron as in Fig. 4A, recorded several minutes previously in control ACSF.) B: responses of an example −/− neuron in DTX (control data from same neuron shown in Fig. 4A). C: mean numbers of APs generated at each current-pulse amplitude for +/+ and −/− neurons in ACSF and in DTX. In DTX the AP numbers were similar across genotypes. D: mean threshold current amplitudes in control ACSF and in DTX for +/+ neurons (black bars) and −/− neurons (white bars). Before DTX application, the −/− threshold currents were significantly larger than those in +/+ neurons (left bars) but DTX abolished the differences, reducing the threshold currents to about 40 pA for each genotype (middle bars). The magnitude of the DTX-induced decrease in threshold current was significantly larger for the −/− neurons (right bars). Single and double asterisks denote significant differences (at P < 0.05 and P < 0.001, respectively). E: mean sustained potentials (measured as in Fig. 5) from 5 +/+ neurons and 4 −/− neurons in response to current pulses in ACSF (circles) and in DTX (triangles); −/− neurons (open symbols) had stronger current–voltage (IV) rectification in ACSF than the +/+ neurons (filled symbols) and the −/− rectification arose at a more negative potential. In DTX, the rectification was abolished and the responses became linear in each individual MNTB neuron (data not shown). DTX responses are shown only up to the current amplitude where all neurons had measurable sustained potentials (rather than spiking tonically as in most examples of A and B). Extrapolated linear fits were very similar for the mean DTX responses of −/− neurons (open triangles, solid line) and +/+ neurons (filled triangles, solid line, shown slightly lengthened to distinguish it from −/− fit), suggesting that in the presence of DTX, neurons of each genotype have similar background leak conductance.

Along similar lines, if different amplitudes of IKv1 accounted for the different sustained IV relations of +/+ and −/− MNTB neurons, then DTX should cause larger shifts in sustained potentials in −/− neurons, and the IV relations of +/+ and −/− neurons should become similar in DTX. Consistent with this, the mean sustained IV relations were initially quite different for the five +/+ and four −/− neurons in ACSF (Fig. 6E, filled and open circles) but became very similar after DTX was applied (Fig. 6E, triangles, and extrapolated lines). In DTX, sustained potentials were not attained for pulse amplitudes larger than the threshold current (∼30 pA) because they typically caused tonic firing (e.g., Fig. 6, A and B). However, the measurable sustained IV relations were highly linear in every neuron (data not shown); thus the fitted lines represent the data slightly beyond the average data shown by the triangles, whose range was restricted by neurons with very low threshold currents of 0 and 10 pA (Fig. 6E). Up to about the 30-pA current-pulse amplitude, the depolarizations of the sustained potentials caused by DTX were approximately twice as large in −/− neurons as those in +/+ neurons (Fig. 6E; compare vertical distances between the circle symbols and the lines fitted through triangles). The mean depolarization caused by DTX relative to the resting potential (the sustained potential during the 0-pA pulse) was more than twice as large in the four −/− neurons (4.8 ± 0.90 mV) as in the five +/+ neurons (1.5 ± 0.47 mV, P < 0.05; this included only one potential that was an extrapolated estimate, for one +/+ neuron that fired tonically at 0 pA in DTX).

These data suggest that IKv1 is larger in −/− than in +/+ MNTB neurons, and that this could underlie all their differences in excitability and membrane properties. The DTX sensitivity of the −/− Kv1 channels shows they must each contain at least one DTX-sensitive subunit, i.e., Kv1.1 or Kv1.6, given that Kv1.2 subunits are absent (Fig. 1C).

Larger Ikl in voltage-clamped −/− MNTB neurons than in +/+ neurons

To measure and compare Ikl amplitudes and properties, we carried out voltage-clamp recordings from MNTB neurons in brain stem slices from +/+, +/−, or −/− mice. After obtaining the whole cell recording configuration for an MNTB neuron, the slice was perfused with ACSFV containing tetrodotoxin, low calcium and high magnesium to block sodium and calcium channels, and the currents were recorded in response to test pulses −107 to +23 mV, after a −67-mV prepotential (see methods). Typical example +/+ and −/− MNTB neurons had current responses that appeared qualitatively similar in the −107- to −37-mV potential range (Fig. 7A; note that after test pulses more negative than −67 mV, the A-currents can be very large during the −37-mV postpotential). The amplitudes of sustained outward current were slightly larger in the −/− neuron than those in the +/+ neurons within the −67- to −37-mV range within which Kv1 channels are expected to dominate the outward current (Fig. 7A; bars show time window for measurements of sustained current). In 11 +/+ and 10 −/− MNTB neurons the mean amplitudes of sustained outward currents were similar and statistically indistinguishable at all test potentials except −67 and −57 mV (Fig. 7B; asterisks denote P < 0.05, Mann–Whitney U tests; see methods for details of leak current subtraction). The data points within the IKv1-relevant range of potentials were replotted to show the larger mean amplitudes of sustained current in −/− MNTB neurons than those in +/+ neurons (Fig. 7C; also including data from eight +/− MNTB neurons). At the −67-mV test potential, the mean amplitudes of sustained current were almost twice as large in −/− neurons (45 pA) as in those in +/+ MNTB neurons (26 pA). This is highly consistent with the above-cited current-clamp data showing that −/− neuron hypoexcitability could be due to enlarged IKv1 (Figs. 4, 5, and 6).

FIG. 7.

Outward K+ currents in +/+, +/−, and −/− MNTB neurons. A: qualitatively similar currents recorded from 2 MNTB neurons, each typical of their genotypes (+/+, left; −/−, right), in response to voltage-clamp test pulses −107 through −37 mV, from within the standard IV protocol (test pulses were applied in 10-mV increments between −107 and +23 mV, at 1.3-s intervals, each preceded by a −67-mV prepotential and followed by a −37-mV postpotential; see methods). Slices were perfused with ACSFV (a low-calcium version of the ACSF; see methods). A large transient A-type potassium current was visible during the −37-mV postpotential if the test pulse was −67 mV or more negative, but its inactivation by the −67-mV prepotential meant it was absent or very small during test pulses. Bars show the times of measurement of sustained currents for B and C. B: sustained current amplitudes from 11 +/+ neurons (8 mice) and 10 −/− neurons (6 mice) at test potentials −107 through +23 mV. For each run of the standard IV protocol, the sustained current amplitudes were measured toward the end of each test pulse (during time marked by bars in A) and leak currents were subtracted (see methods). Then, for each neuron, the sustained current amplitudes were averaged across 2–3 runs of the IV protocol, before averaging across the neurons within each genotype. Asterisks indicate significant differences between +/+ and −/− neurons (P < 0.05, Mann–Whitney U test). C: boxed region from B, plotted on an expanded scale to highlight the potential range where there were significant differences between genotypes (asterisks as in B). Also shown are the sustained currents from 8 +/− neurons (7 mice). There were no significant differences between the +/+ and +/− currents. Circumflexes (hats) indicate significant differences between +/− and −/− current amplitudes (P < 0.05, Mann–Whitney U tests).

The eight +/− MNTB neurons had sustained current amplitudes which were significantly smaller than those of the −/− neurons at the test potentials −67 and −57 mV but statistically indistinguishable from the +/+ neurons (Fig. 7C, P < 0.05, Mann–Whitney U tests). The similar current amplitudes in +/+ and +/− neurons were somewhat surprising given that +/− MNTB neurons had low excitability much more like that of −/− MNTB neurons than that of +/+ MNTB neurons (see Fig. 4). However, the +/− current amplitudes appear closer to those of −/− neurons after the exclusion of a few neurons with current amplitudes that were outliers for their genotypes (see Figs. 8 and 11).

FIG. 8.

Time course of low-voltage–activated potassium current Ikl in +/+, +/−, and −/− MNTB neurons. A: current responses of individual +/+, +/−, and −/− neurons during the first 10 ms of −57-mV pulses from the standard IV protocol (same neurons as in Fig. 7C). To facilitate visual comparisons of amplitude and kinetics, we subtracted capacitative and leak currents (see methods) and also offset the baseline current at −67 mV to zero. Each current trace increased in amplitude with an approximately single-exponential time course. Arrows show neurons with current amplitudes that were outliers for their genotype (as defined later in Fig. 11A). B: mean time constants of the current increases shown in A at −57 mV (left bars, same neurons as in A) or at −47 mV (right bars, at this potential one −/− neuron could not be included because it yielded a poor fit; it was the −/− neuron with tiny current arrowed in A). Time constants were determined by fitting a single-exponential curve to the current from each neuron before averaging across neurons (see methods for fitting details).

No other genotypic differences apparent

As described earlier for the current-clamp MNTB data, apart from any differences in potassium currents the 29 voltage-clamped MNTB neurons of Fig. 7 had similar properties irrespective of genotype. The background “leak” resistances were statistically indistinguishable between genotypes in the potential range used for leak subtraction (−97 to −77 mV; see methods: +/+, 186 ± 14 MΩ, n = 11; +/−, 228 ± 20 MΩ, n = 8; −/−, 244 ± 32 MΩ, n = 10). The pipette access resistances and neuronal capacitances noted from the compensation dials on our amplifier were also similar (+/+, 10.7 ± 1.1 MΩ; +/−, 11.8 ± 1.1 MΩ; −/−, 12.7 ± 1.1 MΩ: +/+, 13.2 ± 0.9 pF; +/−, 14.4 ± 0.7 pF; −/−, 13.0 ± 0.4 pF). The postnatal ages of the mice used were similar across genotypes (+/+, 12.4 days, n = 11; +/−, 12.3 days, n = 8; −/−, 12.6 days, n = 10). There were no statistically significant correlations between any of the electrophysiological parameters we measured and the age of the mouse, at least within the small age range tested.

Activation time course of Ikl after the −67-mV prepotential

Different activation kinetics could in principle contribute to excitability differences between +/+, +/−, and −/− MNTB neurons. However, the outward currents developed with very similar time courses in all MNTB neurons, attaining their maximal amplitudes within 10 ms of the step to −57 mV (Fig. 8A; same neurons as in Fig. 7C; arrows show outlier current amplitudes further addressed in Fig. 11). The current traces from all these neurons were fitted well by single exponentials and gave almost identical mean time constants of about 3 ms for the neurons of each genotype (Fig. 8B, left bars). Fits to the currents at −47 mV showed slightly more rapid current increase than at −57 mV, as would be expected, with time constants of about 2 ms in each genotype (Fig. 8B, right bars). Very similar time constants were obtained from fits to the IKv1 current component obtained by subtraction (e.g., 3 ms at −57 mV, data not shown; currents in the presence of Kv1-selective toxins were subtracted from those recorded a few minutes previously in ACSFV; see methods). This suggests that when depolarized from −67 mV (similar to their resting potentials, supplemental Table S2) the potassium currents in MNTB neurons of each genotype increased with a similar time course, implying that amplitude differences alone may underlie genotypic differences in excitability. However, subtle kinetic differences cannot be ruled out and might be more easily measurable under conditions where IKv1 activation was slow and sigmoidal, as occurred after holding potentials of −80 mV or more negative in Rothman and Manis (2003). Also, inactivation properties may contribute to the apparent activation time courses in Fig. 8, although inactivation at −67 mV is probably <10% (Brew et al. 2003).

Differential sensitivities to subunit-specific toxins of Ikl in +/+, +/−, and −/− neurons

To characterize the subunit contents of Kv1 channels in the MNTB neurons of each genotype, their currents were recorded while applying tityustoxin (TsTx), which blocks Kv1 channels containing at least one Kv1.2 subunit, and DTX, which probably blocks all MNTB neuron Kv1 channels (see introduction). Currents were recorded from MNTB neurons, initially during perfusion with control ACSFV alone, then during perfusion of ACSFV plus 100 nM TsTx, and finally during perfusion of ACSFV containing both 100 nM TsTx and 100 nM DTX. Effects of the toxins on sustained current amplitudes occurred within 1–3 min of the start of perfusion, as shown at −47 mV for typical example neurons of each genotype (Fig. 9A, top row). The application of TsTx to +/+ or +/− neurons caused decreased sustained current amplitudes within 1 min, reaching a new equilibrium level within 3 min (e.g., Fig. 9A, left, middle). TsTx had no effect on the currents in −/− neurons, as expected because of their lack of Kv1.2 subunits (e.g., Fig. 9A, right). The current amplitudes in +/+ neurons did not decrease further when the solution containing both TsTx and DTX was applied, presumably because the TsTx solution had already blocked all the DTX-sensitive channels (e.g., Fig. 9A, left). The additional application of DTX to −/− and +/− neurons when TsTx was already present substantially decreased their sustained currents (e.g., Fig. 9A, middle, right).

FIG. 9.

Effects of tityustoxin-Kα (TsTx) and DTX in +/+, +/−, and −/− MNTB neurons. A: sustained current amplitudes in response to −47-mV test pulses for an example MNTB neuron of each genotype (+/+, left; +/−, middle; −/−, right). Currents were measured first in control ACSFV, then while the perfusion solution was switched to ACSFV containing 100 nM TsTx, and finally to ACSFV containing both 100 nM TsTx and 100 nM DTX-I (times of toxin applications shown by black bars). In the +/+ neuron, TsTx led to substantial current block but the additional presence of DTX caused no further block (left). In the +/− neuron, TsTx led to substantial current block and the addition of DTX also led to substantial current block (middle). In the −/− neuron, TsTx had no effect but DTX led to substantial current block. Here leak current was not subtracted because most data points came from the toxin-monitoring protocol and only 9–12 data points came from the IV protocols that allow leak current subtraction (see methods). For each neuron in each solution, 3 runs of the standard IV protocol described in Fig. 7 were applied at times when the currents appeared approximately stable (e.g., shown by the letters c, t, and d in each panel, which also denote when the data in B were collected). For some neurons the “high-resolution” IV protocol was also applied for 3 runs per solution (see methods). B: current responses during −47-mV pulses of the same example neurons as in A (directly above) at the approximate times shown by letters c, t, or d in A, i.e., in ACSFV, TsTx, or (TsTx + DTX). Each trace is the average response from 2–3 runs of the standard IV protocol. In the +/+ neuron, the current block appeared identical in the TsTx solution and the (TsTx + DTX) solution (left). In the +/− neuron, the current amplitude in TsTx fell about halfway between the control amplitude and that in the solution containing both toxins (middle). In the −/− neuron, the current appeared identical in the control ACSFV or in the TsTx solution, and only the (TsTx + DTX) solution led to substantial current block (right). These characteristic effects of toxins on the current amplitudes of neurons of each genotype were as evident at the beginning of test pulses as at the end, when sustained current amplitudes were measured (see bars in Fig. 7A). C: voltage dependence of the mean sustained current amplitudes in +/+, +/−, and −/− neurons, in control ACSFV (c, solid lines), ACSFV plus TsTx (t, dashed lines), and ACSFV plus TsTx and DTX (d, dotted lines). For each neuron, the currents from each run of the standard IV protocol were subjected to leak subtraction (see methods) before averaging the amplitudes from 2–3 runs per solution. Inserts show mean current amplitudes for all the test pulses −107 through +23 mV, and the larger panels show the same data plotted on an expanded scale to highlight effects of the toxins in the range of potentials relevant to Kv1 channels. In the +/+ neurons, the current amplitudes at each potential were similar in TsTx or (TsTx + DTX). In the +/− neurons, TsTx led to substantial current block and (TsTx + DTX) caused even greater block. In the −/− neurons, TsTx had no effect at any potential and DTX led to substantial current block.

The subsequent analyses of toxin block are all based on the sustained current amplitudes measured at the ends of test pulses, but very similar results would have been obtained from currents measured early during the test pulse because the toxin effects were approximately uniform throughout 180-ms test pulses (Fig. 9B, same neurons as directly above in A). When both toxins were present, the toxin-insensitive currents were revealed and were similar for neurons of each of the three genotypes, usually including a small initial A-current (mostly inactivated by the −67-mV prepotential) and a slowly activating outward current (Fig. 9B, traces marked d). Note that the only substantial effect of any of the toxins on the baseline outward current amplitudes at −67 mV before the test pulse was in the −/− neuron, in which the approximately 50-pA reduction probably reflected block of Kv1 channels open at −67 mV (Fig. 9B, right).

The voltage dependence of the sustained current amplitudes in the presence of the control, TsTx, or TsTx + DTX solutions was calculated for 21 MNTB neurons and averaged within each genotype (Fig. 9C, leak subtraction and averaging across two to three runs of the standard IV protocol were as in Fig. 7, B and C; see methods). At the most positive potentials, neither TsTx nor the combined toxins significantly affected mean current amplitudes (Fig. 9C, insets) probably because Kv3-type high-voltage–activated currents are very large (Brew and Forsythe 1995; Wang et al. 1998). In the potential range relevant to Kv1 channels, at test potentials −67 through −37 mV, TsTx produced substantial current block in the seven +/+ and eight +/− MNTB neurons, but had no effect in six −/− MNTB neurons (Fig. 9C, dashed lines). The addition of DTX had no further blocking effect in the +/+ MNTB neurons but produced substantial current block in the +/− and −/− neurons (Fig. 9C, dotted lines). Throughout this potential range, TsTx block appeared to account for 100% of the total toxin block in +/+ neurons, in contrast with 0% in −/− neurons and about 50% in +/− neurons (Fig. 9C, and see following text for further analyses).

Some of these MNTB neurons exhibited toxin-induced current increases at positive potentials (e.g., see +/+ inset panel in Fig. 9C). The probable reason for this and details of how this also limits the reliability of the calculated IKv1 values at more negative potentials are described further in supplemental methods. This is relevant when considering the fits to the calculated GKv1 and the parameters for voltage dependence in Figs. 12 and 13 because the same mechanism could slightly distort some of the calculated GKv1 values (see supplemental methods).

Next, the amplitudes of current blocked at −47 mV by TsTx and by subsequent addition of DTX are presented for individual MNTB neurons (Fig. 10A, dark gray bars and light gray bars, respectively; same 21 neurons as in Fig. 9C). The mean amplitude of the current reduction caused by TsTx was 185 ± 29 pA in the seven +/+ neurons and only a little smaller at 153 ± 19 pA in the eight +/− neurons, but only 9 pA in the six −/− neurons. The consecutive addition of DTX reduced current amplitudes by only 4 pA in the +/+ neurons, 129 pA in the +/− neurons, and 287 pA in the −/− neurons. The total current blocked by the combined TsTx + DTX solution compared with the control ACSFV solution is the IKv1 component (represented by the total heights of the bars in Fig. 10A, unless one subcomponent was “negative”; i.e., when a toxin application was associated with an increased current amplitude). The mean amplitude of IKv1 at −47 mV was 188 ± 29 pA in +/+ neurons, significantly smaller than the 282 ± 30 pA in +/− neurons (P < 0.05). Although the IKv1 was even larger in −/− neurons, at 297 ± 61 pA, this was not significantly larger than that in either +/− or +/+ neurons, probably because one −/− neuron had extremely small IKv1 (shown at far right in Figs. 10, 11, and S2). The relatively tiny current shifts that occurred when TsTx was applied to −/− neurons were likely due to experimental noise; the shifts were not statistically significant, averaged 9 pA, and were not correlated with the IKv1 amplitude for that neuron (Fig. 10A; also see Fig. 9A for sample noise/drift). The apparent effects of the addition of DTX to TsTx-treated +/+ neurons were also small and not significant (averaging 4 pA, and the largest value was 33 pA) and probably attributable to noise.

FIG. 10.

Proportional sensitivities to TsTx and DTX in MNTB neurons at −47 mV. A: amplitudes of sustained current blocked at −47 mV for each of 21 MNTB neurons by TsTx (dark gray bars, control current minus current in TsTx) and by subsequent addition of DTX (light gray bars, TsTx current minus current in TsTx + DTX solution). If both subcomponents are positive, then the heights of the combined light and dark gray bars reflect the total current blocked by both toxins, or Kv1-based K+ currents (IKv1; also shown by the vertical extent of the black bars in Fig. S2A). Only very small current shifts were ever associated with DTX addition in +/+ neurons (light gray, left) or TsTx application in −/− neurons (dark gray, right). B: proportion of the total toxin block at −47 mV that is accounted for by TsTx in each of the neurons shown in A. TsTx-sensitive current (control current minus current in TsTx solution) was divided by IKv1 (control current minus current in TsTx + DTX solution). At −47 mV, the average TsTx percentage block relative to TsTx + DTX block was 98% for +/+ neurons, 56% for +/− neurons, and 1.2% for −/− neurons.

FIG. 11.

Outward K+ current amplitudes in individual MNTB neurons of each genotype. A: bar plot of the sustained current amplitudes in ACSFV for 29 MNTB neurons at −47 mV. Amplitudes were approximately normally distributed for each genotype, with the exception of 2 +/+ neurons, one +/− neuron, and one −/− neuron with amplitudes that were outliers for their genotype (asterisks; outliers defined as being >1.5-fold the interquartile range above the upper quartile or below the lower quartile, both at this test potential and at one or both of the test potentials −57 or −67 mV). B: IV relations plotted as in Fig. 7C, but excluding 4 outlier neurons (the neurons with outlier values of sustained current, asterisked in A). Small asterisks indicate significant differences between +/+ and −/− neurons, circumflexes indicate significant differences between +/− and −/− neurons, and large asterisks indicate significant differences between +/+ and +/− neurons. All symbols indicate significance levels of P < 0.005 (2-tailed unpaired t-test assuming equal variance).

To estimate the percentage of Kv1 channels that contain Kv1.2 in each MNTB neuron, we calculated the percentages of current blocked by TsTx relative to the total toxin block when DTX was also present. At −47 mV, the average percentage block accounted for by TsTx was almost 100% for +/+ neurons, 56% for +/− neurons, and about 0% for −/− neurons (Fig. 10B, from data of A). The percentages of TsTx block were similar when averaged across the four test potentials −67 through −37 mV (+/+, 98%; +/−, 49%; −/−, −1%). The absence of Kv1.2 protein in the −/− brain obviously makes it reasonable to assume that the true percentage of Kv1 channels containing Kv1.2 is 0% in the −/− neurons. However, although the +/+ data suggest that 100% of the +/+ Kv1 channels contain Kv1.2, this conclusion implies that all of the percentage variability is due to noise, but we cannot rule out the alternative idea that the four +/+ neurons with TsTx block percentages clustering in the 80–90% range (Fig. 10B) actually reflect reality and have 10–20% of their Kv1 channels lacking Kv1.2. Similarly, although the 35 to 93% range for +/− neurons’ TsTx block percentages at −47 mV may be due to noise, these values might reflect actual differences in each +/− neuron's proportion of Kv1 channels containing Kv1.2 (Fig. 10B; also see Fig. 13).

Typical current amplitudes and proportions of IKv1 for MNTB neurons of each genotype

MNTB neurons within each genotype exhibited some variability in their sustained current amplitudes in the control ACSFV solution (e.g., at −47 mV; Fig. 11A). This variability may reflect real differences between recordings because the current amplitudes at −47, −57, and −67 mV were highly correlated (data not shown). Of the 29 neurons shown, four had current amplitudes that were outliers for their genotypes at two or three of these test potentials (asterisks in Fig. 11A; see legend for statistical definition). Without these four outliers, the MNTB neurons within each genotype had approximately normally distributed current amplitudes, and there was very little overlap in current amplitudes between +/+ and −/− neurons (Fig. 11A). This is consistent with the above-cited current-clamp excitability data, which showed little overlap between +/+ and −/− neurons in the numbers of APs fired (Fig. 4D). Thus both voltage-clamp and current-clamp data suggest that the majority of −/− MNTB neurons were abnormal, rather than having compensated for their lack of Kv1.2 subunits by up- or downregulation of other genes or proteins.

To compare current amplitudes for more representative populations of MNTB neurons of each genotype, the IV data from Fig. 7C were replotted after excluding the outliers (Fig. 11B, excludes the four neurons asterisked in A). After the exclusion, the −/− current amplitudes remained significantly larger than those in +/+ neurons at all three test potentials, and larger than those in +/− neurons at −57 and −67 mV (Fig. 11B, P < 0.005). The +/− currents had intermediate amplitudes similar to those of +/+ neurons at −67 mV, but closer to −/− values at −47 mV (Fig. 11B), which might relate to the earlier findings that +/− MNTB neurons had properties similar to those of +/+ neurons when at their resting potentials, but more like their −/− counterparts at the slightly depolarized membrane potentials crucial to determining excitability (supplemental Table S2; e.g., resting input resistances, AP numbers). Although the exclusion of outliers makes statistical comparisons speculative, the current amplitudes were significantly larger in +/− neurons than those in +/+ neurons at −67, −57, and −47 mV (Fig. 11B, small asterisks, P < 0.005) and remained significantly different when all eight +/− neurons were included (P < 0.05). Abnormally large +/− currents also seem a probable explanation for the significantly lower excitability in +/− MNTB neurons than in +/+ neurons, shown earlier (Fig. 4).

In the majority of MNTB neurons of each genotype, most of the sustained Ikl amplitude at −47 mV was accounted for by the toxin-sensitive component IKv1 (70–80% for +/− and −/− MNTB neurons, and 60–70% for +/+ neurons; Fig. S2). This was after excluding three neurons with atypically large proportions of non-Kv1 current, all of which were also outliers in terms of their Ikl amplitudes (Fig. 11A, asterisks). This helps to justify the inclusion in the analyses of voltage dependence of four +/+ and four −/− MNTB neurons to which toxins were not applied because we can assume that most of their total Ikl was IKv1. Except for the three excluded neurons, the toxin-insensitive non-Kv1 currents had similar amplitudes irrespective of genotype, averaging about 100 pA at −47 mV (supplemental Fig. S2). This is consistent with the idea that the genotypic differences in sustained current amplitudes are attributable to different amplitudes of IKv1.

The finding of increasing current amplitudes between +/+, +/−, and −/− MNTB neurons suggests an inverse dependence on their respective gene dosages of Kcna2 (2, 1, and 0) and possibly on the proportions of their Kv1 channels that include Kv1.2 (roughly 100, 50, and 0%). In principle, individual +/− neurons with differential expression of Kv1.2 subunits might also exhibit an inverse dependence between current amplitudes and their proportion of Kv1.2-based current, but no such correlation is evident (compare middle panels of Figs. 10 and 11A). Possible differences between +/− neurons are explored further in the analyses of Figs. 12 and 13, which address whether the enlarged potassium current amplitudes in +/− and −/− neurons might be due to abnormally negatively activating Kv1 channels.

FIG. 12.

Voltage-dependence curves for +/+, +/−, and −/− Gkl and GKv1 conductances. AC: symbols show G/Gmax values from individual MNTB neurons plotted against test potentials −87 through −37 mV (see methods for conductance calculations and fitting methods). Conductances were calculated from the current responses obtained using the standard IV protocol (or the high resolution IV protocol if available) after leak subtraction and averaging across 2–3 runs. Fits of Boltzmann functions to these data are shown as solid lines, or dashed lines for slightly “less good” fits (see methods). Gray bars are visual aids showing the range −57 to −48 mV. A: symbols show Gkl/Gmax values plotted against test potential for 9 +/+ neurons (top), 8 +/− neurons (middle), and 8 −/− neurons (bottom, same neurons as in Fig. 7C but excluding three neurons asterisked in Fig. S2A, and one −/− neuron with a poor fit). Half-activation voltage (Vhalf) values for the +/+ neurons and −/− neurons fell in distinct ranges, whereas the +/− neurons had intermediate Vhalf values (approximately within the gray bar). B: GKv1/Gmax values from individual MNTB neurons plotted against test potentials for 17 MNTB neurons (same neurons as in Figs. 9, 10, but excluding 1 +/+ neuron and 2 −/− neurons yielding poor fits, and one +/− neuron for which the DTX-sensitive component was too small to be well fitted for C). Mean Vhalf values and their ranges and the genotypic differences between them were generally similar to those from A, although the mean Vhalf values were a little more negative than those in A. C: fits as in A and B, but to the distinct components of +/− GKv1 either TsTx-sensitive or TsTx-insensitive but DTX-sensitive. GKv1/Gmax values of TsTx-sensitive components became half-activated at less-negative potentials than those of the TsTx-insensitive components, but the Vhalf values for these 2 components did not fall neatly into the +/+ and −/− ranges, respectively.

FIG. 13.

Relationship between Vhalf and Kv1.2 subunit content. A: Vhalf values from the fits to total Gkl shown in Fig. 12A, plotted against the Kcna2 gene dosage for each neuron. Regression line through all 27 points shows a significant correlation (P < 0.01). B: Vhalf values from the fits to GKv1 shown in Fig. 12B, plotted against the estimated proportion of channels containing at least one Kv1.2 subunit. In the case of +/+ and −/− neurons, we estimated these proportions were 0 and 100%, whereas for +/− neurons the estimate was taken as the percentage of IKv1 blocked by TsTx at −47 mV. Here the regression line is through only the 7 +/− points, and shows a significant correlation (P < 0.05). This supports the idea that +/− Kv1 channel Vhalf values are determined by the proportion of Kv1.2-containing channels. Clusters of −/− and +/+ Vhalf values falling on the regression line suggest that most +/− neurons have channel populations with Vhalf values distinct from those of the other 2 genotypes, and that the 2 more negative +/+ Vhalf values may be due to distorted data (see results and supplemental methods). C: as in B, but including only TsTx-sensitive channels. Vhalf values from fits to +/+ GKv1 and TsTx-sensitive +/− GKv1 (shown in Fig. 12C) are plotted against the estimated proportion of channels containing at least one Kv1.2 subunit. Again the regression line is through only the 7 +/− points, and shows a significant correlation (P < 0.05). This suggests that +/− Kv1 channels including Kv1.2 can have different Vhalf values, probably depending on the number of Kv1.2 subunits each channel contains.

−/− GKv1 activated at more negative potentials than +/− GKv1

To explore whether the genotypic differences in current amplitudes were partly based on differences in voltage dependence, fits were made to the total conductance G (Fig. 12 A) and, where available, its explicitly toxin-sensitive component GKv1 (Fig. 12B, calculated from sustained Ikl or IKv1, respectively, as described in methods). For each neuron, the G and GKv1 values were first plotted against test potential and fitted by a Boltzmann function GKv1 = Gmax/{1 + exp[(VVhalf)/−k]} with fitted parameters as follows: Gmax, the maximum conductance; Vhalf, the half-activation voltage; and k, the slope factor. Note that the necessary use of the −67-mV prepotential (see supplemental methods) means that these curves could represent a mix of activation and inactivation for each channel in the population, although previous results suggest inactivation is probably <10% for the +/+ neurons (Brew et al. 2003). The data and fit for each neuron are shown normalized to the Gmax value for the fit (Fig. 12). Of the 56 fits shown, 46 were very good, using the fitting criteria described in methods (Fig. 12, solid lines). Of the ten “less-good” fits, seven came from +/+ neurons (Fig. 12, dashed lines). The focus of the subsequent analysis is therefore on the results from −/− and +/− neurons, although +/+ data are also presented. In each panel, the gray shaded region covers the same 9-mV range between −48 and −57 mV to facilitate visualization of the differences in conductance curves between neurons and between genotypes (Fig. 12). For example, in all eight +/− neurons the fits to G yielded Vhalf values between −48 and −55 mV, within the range shown by the gray bar (Fig. 12A, middle, 0.5 on the y-axes corresponds to half-activation at Vhalf on the x-axes). In contrast, six of the eight −/− neurons had Vhalf more negative than −57 mV, to the left of the gray bar (Fig. 12A, bottom). The mean Vhalf value from +/− neurons was 5 mV less negative than that in −/− neurons (P < 0.001). The Vhalf values returned from eight of nine fits to +/+ G data were less negative than −48 mV, i.e., less negative than any of the fits to +/− or −/− G, and reaching Vhalf at potentials beyond the right edge of the gray bar; only one +/+ Vhalf value of −49 mV fell within the gray bar (Fig. 12A, top). The mean +/+ Vhalf value was 7 mV less negative than the mean for +/− neurons, and 12 mV less negative than that for −/− neurons (P < 0.0001).

The mean Vhalf values for the G from MNTB neurons of each genotype remained similar when we excluded the five “less-good” fits (Fig. 12A, dashed lines, described earlier and in methods) because the remaining five +/+, eight +/−, and seven −/− neurons had mean Vhalf values of −45.0, −51.1, and 57.2 mV, respectively, and the differences between genotypes remained significant at P < 0.005 or better. The values of Gmax associated with the fits to G were very similar for the MNTB neurons of each genotype (+/+, 17.2 ± 1.2 nS; +/−, 18.4 ± 1.5 nS; −/−, 19.0 ± 1.4 nS). This suggests that MNTB neurons of each genotype may possess the same numbers of Kv1 channels and that the genotypic differences in current amplitudes may be solely due to differences in the voltage dependence of their Kv1 channels (possibly involving differences in both activation and inactivation).

Overall, the voltage dependence of G was steepest in −/− neurons, intermediate in +/− neurons, and shallowest in +/+ neurons (Fig. 12A). Although the genotypic differences in k values were significant, the differences were small and almost all the fits had k values between 5 and 10 mV (averaged data not shown). If the differences are real, it might mean that individual +/+ Kv1 channels activated less steeply than individual −/− Kv1 channels and/or might reflect a greater spread of Vhalf values among +/+ Kv1 channels.

The same ordering of voltage dependence between the genotypes was also found in the fits to GKv1, for which Vhalf values were 7 mV more negative in −/− neurons than in +/− neurons (P < 0.005, Fig. 12B). The Vhalf values from the fits to −/− GKv1 ranged between −58 and −62 mV, exhibiting hardly any overlap with the −48- to −57-mV range for +/− GKv1 (Fig. 12B). The fact that the fits yielded broadly similar Vhalf values whether calculated from G or GKv1 suggests that G did largely reflect the voltage dependence of Kv1 channels. The fits to +/+ GKv1 could be regarded as consistent with the fits to G if we focus on the four fits with Vhalf values less negative than −48 mV (Fig. 12B, top); the other two fits yielded very negative Vhalf values, probably distorted in the negative direction by toxin-induced space-clamp effects (see supplementary methods). If Kv1 channels do activate at less-negative potentials in +/+ MNTB neurons than in their +/− and −/− counterparts, this would itself contribute to making the +/+ G and GKv1 data harder to fit because they would not yet be approaching their maximum values at −37 mV, whereas the +/− and −/− G and GKv1 data show clear signs of approaching or reaching their maxima (see Fig. 12, A and B).

Both sets of fits, to G and GKv1 (Fig. 12, A and B), suggested that Vhalf is about 6 mV more negative for −/− than that for +/− Kv1 channels. The actual Vhalf values may be close to the −60 and −53 mV returned from the GKv1 fits because the G fits were probably shifted slightly to the right by the toxin-insensitive conductance component, which, when fitted separately, gave a less-steep voltage dependence and substantially more positive Vhalf than any of the fits shown (data not shown). Consistent with this, simulations in which we subtracted a typical amount of toxin-insensitive conductance (calculated from the neurons of supplemental Fig. S2B) from the G data (as in Fig. 12A) yielded +/− and −/− fits with Vhalf values 3–5 mV more negative than before the subtraction, and slightly steeper. The same subtraction from the +/+ G data yielded +/+ Vhalf values 7 mV more negative than before the subtraction, of −51 mV, presumably because of the larger proportional contribution to +/+ G of toxin-insensitive channels. From this, our overall best estimates of the mean Vhalf for +/+, +/−, and −/− MNTB neuron Kv1 channel populations are −51, −55, and −60 mV. Thus the Vhalf may be only 9 mV more negative for −/− than that for +/+ Kv1 channels, rather than the 12 mV suggested by the G fits. These results on voltage dependence should ideally be confirmed in future experiments when selective blockers become available for the various types of DTX-insensitive channels also present, which would allow improved isolation of the Kv1 channel currents and would probably reduce the effects of any DTX-induced space-clamp alterations (see supplemental methods).

The results in Fig. 12, A and B suggest that Kv1 channel populations activate at more negative potentials in neurons with smaller gene dosages of Kcna2, shown earlier to have smaller proportions of Kv1.2 subunits (Figs. 10 and 11). One simple explanation for this would be that Kv1 channels without Kv1.2 subunits activate at more negative potentials than those containing Kv1.2. If so, this might be detectable between the two distinct GKv1 components in +/− neurons, based on either Kv1.2-free or Kv1.2-containing channels. For the same seven +/− neurons as in Fig. 12B, the fits to the GKv1 component based on TsTx-sensitive currents activated at 4-mV less-negative potentials (Fig. 12C, top) than for the GKv1 component based on TsTx-insensitive but DTX-sensitive currents (Fig. 12C, bottom, P < 0.05). Note that the latter values for presumed Kv1.2-free +/− Kv1 channels are not as negative as those for −/− GKv1, possibly because the TsTx block was not fully completed when DTX was added. Nonetheless, this provides additional support for the idea that Kv1 channels without Kv1.2 subunits activated at more negative potentials than those containing Kv1.2.

G and GKv1 activate more negatively in +/− neurons with smaller percentages of TsTx block

The Vhalf values for G were highly correlated with the number of Kcna2 genes present (0, 1, or 2) in the 25 neurons of all three genotypes (Fig. 13A, P < 0.01). The relative gene dosage is probably a rough indicator of the relative proportions of Kv1.2 subunits present, but in +/− neurons a better indicator may be the proportion of IKv1 that was blocked by TsTx at −47 mV (Fig. 10). For the seven +/− neurons for which we obtained good fits to GKv1 (in Fig. 12B) the Vhalf values were significantly correlated with this TsTx block proportion (Fig. 13B, P < 0.05). Also plotted for comparison are the Vhalf values from +/+ and −/− neuron GKv1, assuming that their TsTx block proportions were 0 and 100%. These data show that the variations in Vhalf between +/− MNTB neurons are correlated with the proportion of their IKv1 that is due to Kv1.2-containing channels, which is likely to be larger when Kv1.2 subunits make up a larger proportion of the total pool of Kv1 subunits.

So far, these data could be explained by assuming only two possible Vhalf values for Kv1 channels: either Kv1.2-free channels with Vhalf of −60 mV or Kv1.2-containing channels with Vhalf of, say, −50 mV. If so, then the intermediate Vhalf of the Kv1 channel population in +/− MNTB neurons would be produced by a mix of these two channel types, although the Vhalf value should be the same for all Kv1.2-containing channels. However, the Vhalf values from the TsTx-sensitive component of +/− GKv1 (from Fig. 12C) varied and were themselves significantly correlated with the TsTx block proportion (Fig. 13C, P < 0.05). This suggests that different Vhalf values may be possible among Kv1.2-containing channels, perhaps correlating with their possession of different numbers of Kv1.2 subunits. For example, most +/− channels containing Kv1.2 may include only a single Kv1.2 subunit and have a more negative Vhalf than that of channels containing two or more Kv1.2 subunits, which predominate in +/+ neurons.

DISCUSSION

We generated −/− mice lacking the Kcna2 open reading frame and Kv1.2 subunit proteins. Their primary overt phenotype consisted of increased seizure susceptibility (at P14), seizures (from P15), and reduced life span. The +/− mice were overtly normal, despite the halved expression of Kcna2 mRNA in +/− brain. At the individual neuron level, −/− auditory MNTB neurons were hypoexcitable and had abnormally large Kv1-based K+ currents (IKv1) probably because their Kv1.2-free channels activated at substantially more negative potentials than the +/+ channels, all of which contained Kv1.2. Corresponding variations in Kv1.1:Kv1.2 stoichiometry between different neuronal types may represent one mechanism of fine-tuning potassium currents for particular information-processing tasks.

Severe brain stem seizures probably cause the −/− reduced life span

Wild running/bouncing and tonic extension (TE) involve the whole body and indicate the −/− mice had generalized, brain stem seizures, rather than partial, forebrain seizures. Studies using genetically epilepsy-prone rats (GEPRs) showed wild running and TE depend on brain stem structures and occur even when forebrain connections are severed, whereas partial seizures involve only parts of the forebrain and movements of parts of the body (Browning et al. 1999; Faingold 1999; Gale 1992). Although TE is usually not fatal in GEPRs, TE is commonly a fatal endpoint of mouse seizures (e.g., Rho et al. 1999). TE and apnea during seizures probably caused the reduced life span in −/− mice because their fatalities always coincided with TE. If so, this suggests a possible way to prolong the −/− life span because TE with apnea became nonfatal for mice placed in 100% oxygen just before audiogenic seizure induction (Venit et al. 2004).

If there were brain stem sites where Kv1.2 subunits dominated Kv1 channels, with no Kv1.1 expression, such sites would probably be hyperexcitable in −/− mice. However, Kv1.1 and Kv1.2 are coexpressed in all the same neurons of the mouse brain and brain stem, except in one neuron type in the olfactory bulb, which expresses Kv1.2 without Kv1.1, and in the hippocampus, which expresses Kv1.1 much more widely than Kv1.2 (Grigg et al. 2000; Veh et al. 1995; Wang et al. 1993, 1994). Although partial seizures exhibited by mice lacking Kv1.1 probably initiate in the hippocampus, where CA3 neurons are hyperexcitable (Lopantsev et al. 2003; Smart et al. 1998), it seems unlikely that −/− seizures would initiate in the olfactory bulb. Instead the −/− brain stem may be hyperexcitable at sites where Kv1.2 subunits dominate channels in normal mouse brain, even with some Kv1.1 present (see following text). Possibly, respiratory or cardiovascular abnormalities could also contribute to −/− seizures because arterial smooth muscle expresses Kv1.2 with Kv1.5 but not Kv1.1 (Albarwani et al. 2003; Thorneloe et al. 2001; Xu et al. 2000). However, intrinsic neural network hyperexcitability was apparently present in all −/− mice tested at P14, which had reduced latencies to flurothyl-induced seizures, and rapid progression to generalized seizures (Fig. 2B).

Kv1 subunit expression and the phenotypes of mice with Kv1 deficits

Kv1.2 may be more important than other Kv1 subunit types because −/− mice were more severely affected than mice with deficits of other Kv1 subunits. Kcna1-null mice lacking Kv1.1 exhibited milder partial seizures and lived approximately twice as long as −/− mice (Smart et al. 1998), whereas Kcna4-null mice had normal life spans and only occasional seizures (London et al. 1998). Mice with deficits in Kv1.3 or Kv1.5 had normal life spans and no reported seizures (Archer et al. 2001; Xu et al. 2003). The relative severity of each phenotype may reflect the relative abundance of each subunit type in normal mouse brain. Using antibody binding to DTX-sensitive channels from bovine brain, the order of subunit prevalence was Kv1.2 > 1.1 ≫ 1.6 > 1.4 (Scott et al. 1994). Kv1.3, Kv1.5, and Kv1.7 are weakly expressed or absent in neurons (Kalman et al. 1998; Trimmer and Rhodes 2004).

These relative abundances may alter during development. Our qPCR data suggest that the dominance of Kv1.1 and Kv1.2 develops in the brain stem between P1 and P29, when their mRNA expression increased ten- and fourfold, respectively, whereas mRNA for Kv1.6 subunits merely doubled (Fig. 3B). Thus Kv1.6 subunits may play a relatively important role in early postnatal development, but be less important in adults. During the second and third postnatal weeks, the upregulation of Kv1.1 and Kv1.2 is probably important for myelination of axons because both subunits are strongly coexpressed in large myelinated axons in adult mice (Wang et al. 1993), although the role of a transient strong embryonic expression of Kv1.1 is unknown (Hallows and Tempel 1998). The qPCR data also showed an increasing ratio of Kcna1 to Kcna2 mRNA during development (Fig. 3B). This ratio also trended upward with age in the rat cochlear nucleus (Bortone et al. 2006). The MNTB neuron results showing Kv1 channel voltage dependence varied with their Kv1.2 content (Figs. 10 and 12; also see following text) suggest this changing ratio might result in Kv1 channels that activate at more negative potentials, as the nervous system develops.

The usual time course of Kv1.1 or Kv1.2 subunit upregulation may have relevance to the age of −/− seizure onset. For example, large axons may be able to function with or without Kv1.2 during the first two postnatal weeks, after which Kv1 subunit production in −/− mice would fall behind requirements, causing reduced amplitudes of axonal IKv1 and seizures. Alternatively, the timing of −/− seizure onset may be determined by when IKv1 becomes enlarged due to Kv1.1-dominated channels with abnormally negative activation. However, an early juvenile onset is common for seizures of many different etiologies, perhaps related to increases in neuronal maximum firing rates (Swann and Hablitz 2000). For example, mice with sodium channel abnormalities developed spontaneous seizures and died aged P13–P26 (Chen et al. 2004) and there was a P16–P18 onset of seizure susceptibility in GEPRs (Reigel et al. 1986, 1989).

Hyperexcitability due to abnormal IKv1 in axons, synapses, or inhibitory neurons?

Possibly, Kv1.2-dominated channels populate axons by preferential targeting, although three subunit types Kv1.1, Kv1.2, and Kv1.6 are targeted to axons via an interaction with Kvbeta2 (Gu et al. 2003). If so, −/− axons or synapses may have strongly reduced Kv1 channel density, leading to network hyperexcitability and seizures. In Kcna1-null axons, it is thought that reduced IKv1 contributes to the hyperexcitability induced by cold temperatures at the final heminode before neuromuscular synaptic terminals, thought to account for the cold-swim–induced tremor in Kcna1-null mice, and in Kvbeta2-null mice (McCormack et al. 2002; Zhou et al. 1998). Kv1 conductance may be especially susceptible to cold temperatures, as shown in octopus neurons of the mouse cochlear nucleus (Cao and Oertel 2005). However, there was no cold-swim–induced tremor in −/− (B6/129) mice (described earlier), arguing against abnormally small IKv1 in −/− axons, unless their Kv1.2-free channels are relatively cold insensitive.

Preferential targeting of Kv1.2 to synaptic terminals was first suggested by localization discrepancies between Kv1.2 protein and mRNA within the hippocampus (Veh et al. 1995). A hippocampal lesioning study suggested Kv1.2 was in the synaptic terminals of entorrhinal afferents, whereas Kv1.1 and Kv1.4 were found together along axons (Monaghan et al. 2001). Kv1.2 subunits dominated the Kv1 currents at the giant synaptic terminals of cochlear nucleus bushy cells, whereas Kv1.1 subunits dominated their somatic current (Dodson et al. 2003). Future direct recordings from these synapses in −/− mice could reveal reduced IKv1.

Alternatively, an enlarged −/− IKv1 (as found in MNTB) could lead to seizures if it weakened inhibition more than excitation. A weakened glycinergic inhibitory output of −/− MNTB neurons seems unlikely to lead to seizures, but might interfere with sound localization using the cue of interaural level difference, calculated by lateral superior olive (LSO) neurons which integrate ipsilateral excitatory signals with contralateral inhibitory signals, arriving via the MNTB inverting relay (also see section on −/− auditory system). Neocortical pyramidal neurons had increased inhibitory postsynaptic current (IPSC) frequencies in Kcna1-null mice, whereas excitatory postsynaptic current (EPSC) frequencies were normal (van Brederode et al. 2001). The powerful effects of Kv1 channels on excitability at GABAergic cerebellar basket cell synapses were shown when DTX led to an increased IPSC frequency onto Purkinje neurons, as was also found in mice with Kv1.1 deficits, although the basket cell somata fired at normal rates (Herson et al. 2003; Southan and Robertson 1998; Zhang et al. 1999). Future recordings showing abnormal IPSC frequencies at these synapses in −/− CNS would point to whether the synapses have reduced or increased IKv1.

Hypoexcitability of −/− MNTB neurons is due to enlarged IKv1

MNTB neurons in brain stem slices from −/− mice were intrinsically less excitable than +/+ MNTB neurons. Consistent with this, IKv1 was larger in voltage-clamped −/− MNTB neurons than in the +/+ neurons. This contrasted significantly with results from Kcna1-null MNTB neurons, lacking the closely related subunit Kv1.1, which had reduced potassium currents and were hyperexcitable (Brew et al. 2003). Interestingly, the excitability of Kcna1-heterozygote MNTB neurons was identical to that of control counterparts, whereas here we showed Kcna2-heterozygote MNTB neurons had excitability and IKv1 properties intermediate between their −/− and +/+ littermates (compare Table S2 with Table 1 in Brew et al. 2003).

The excitability differences between −/− and +/+ MNTB neurons were attributable solely to differences in their Kv1 channels because genotypic differences were abolished by DTX and other electrophysiological properties such as capacitance and action potential shapes were similar across genotypes. The enlarged −/− IKv1 could also explain why DTX caused larger depolarizations from rest in −/− neurons and larger reductions in their threshold currents (Fig. 6 and related text).

Implications of toxin sensitivities for Kv1 channel stoichiometries in MNTB neurons

Both Kv1.1 and Kv1.2 are probably present in almost all Kv1 channels in murine MNTB neurons, as follows. Both are strongly expressed in rat and mouse MNTB, with some Kv1.6 (Brew et al. 2003; Dodson et al. 2002; Fonseca et al. 1998; Wang et al. 1994). The Kv1.1-selective blocker dendrotoxin-K (DTX-K; Robertson et al. 1996) blocked >70% of the total sustained Ikl at −47 mV in MNTB neurons of the same mouse strain used here, C3HeB/FeJ (Gittelman and Tempel 2006). This is consistent with the 50–70% block of Ikl, expected if IKv1 was fully blocked because Ikl was blocked by 50–60% by DTX in Brew et al. (2003) and by 66% by TsTx + DTX (Fig. S2B). This suggests Kv1.1 is present in all the Kv1 channels. Kv1.2 subunits were in 80–100% of +/+ MNTB neuron Kv1 channels because when TsTx was already present, no additional current block was caused by DTX (Figs. 9 and 10).

The 80–100% of Kv1 channels that contain Kv1.2 in murine MNTB neurons is a substantially larger percentage than the 50–60% in rat MNTB neurons. All the rat MNTB neuron Kv1 channels included Kv1.1 because either DTX or DTX-K blocked 90% of the total sustained current at −47 mV, but only 50% was TsTx sensitive and therefore due to Kv1.2-containing channels (Brew and Forsythe 1995; Dodson et al. 2002). Rat MNTB neurons also lack the large DTX-insensitive A-currents found in murine MNTB neurons (Fig. 8 and equivalents in Brew and Forsythe 1995; Brew et al. 2003). These different potassium currents may reflect species differences in information processing by MNTB. For example, the rat medial superior olive (MSO) receives substantial inhibitory input from MNTB (Smith et al. 2000) whose phase locking is thought to improve MSO neurons' encoding of interaural time difference, a useful cue for localizing low-frequency sound sources, although mice hear only at higher frequencies (more suited to interaural level comparisons by LSO) and their MSO is correspondingly vestigial. Perhaps very negatively activating Kv1 channels facilitate the preservation of phase-locked information across the giant synapse onto rat MNTB neurons, by ensuring consistent resetting of the membrane between every synaptic event, whereas potassium channels with less-extreme properties may be suited to the accurate temporal representation of the sound envelope to be used in the murine LSO level comparison (possibly by increasing membrane resistance at rest and maximum firing rates). Note that when Kcna2 expression was reduced in the +/− MNTB neurons with only one Kcna2 gene, they became very similar to rat MNTB neurons in terms of their percentage of Kv1 channels containing Kv1.2 because, in both cases, TsTx block was 50–60% of the DTX current block amplitude.

Next, we explore the possible stoichiometries of Kv1 channels in +/+ and +/− MNTB neurons, in terms of their Kv1.2 subunits. Because the +/− neurons each possess a single Kcna2 gene, they probably have half the usual +/+ copy numbers of Kcna2 mRNA (as in whole brain; Fig. 1D) and of Kv1.2 subunit production, accounting for a smaller proportion of their Kv1 current being TsTx sensitive. This probably also means that the +/− Kv1.2-containing channels would have fewer Kv1.2 subunits per channel, on average, than the Kv1.2-containing channels in +/+ MNTB neurons. The channel stoichiometries can be estimated if we assume that subunits assemble randomly into tetramers (binomial assembly of Kv1.2 and non-Kv1.2 subunits) and that stoichiometry does not affect the surface expression of channels. The mean proportional TsTx sensitivity in +/− MNTB neurons can be approximately simulated using binomial assembly if the Kv1 subunit population consists of 20% Kv1.2 subunits, resulting in 59% of Kv1 channels containing at least one Kv1.2 subunit. This 59% constitutes 41% containing a single Kv1.2 subunit and 18% containing two or more Kv1.2 subunits. If +/+ MNTB neurons produced the same numbers of non-Kv1.2 subunits as +/− neurons, and twice the numbers of Kv1.2 subunits, their overall proportion of Kv1.2 subunits would be 33%. Using binomial assembly, this would result in 80% of the Kv1 channels containing Kv1.2, not dissimilar to the values found for +/+ Kv1 channel TsTx sensitivity (80–90%, excluding three values >100%; Fig. 10). Of these simulated +/+ channels, 40% contain a single Kv1.2 subunit and another 40% contain two or more Kv1.2 subunits. Thus we estimate that more than two thirds of the Kv1.2-containing channels in +/− neurons contain only one Kv1.2 subunit (xxx2), whereas at least half of the Kv1.2-containing channels in +/+ neurons possess two or more Kv1.2 subunits (xx22 or x222).

Stoichiometry and differences in Vhalf between MNTB neuron Kv1 channel populations

The results suggest that there are at least three different Vhalf values for MNTB neuron Kv1 channels containing zero, one, or two Kv1.2 subunits, as follows. Our best estimates of Vhalf were −60 mV for −/− Kv1 channels (Kv1.2-free) and −51 mV for +/+ Kv1 channels (≥80% Kv1.2-containing channels). In principle, different proportions of Kv1 channels having one or the other of these two Vhalf values could produce the mean Vhalf of −55 mV for +/− Kv1 channels, and explain the significant correlation between Vhalf and TsTx-percentage block in individual +/− MNTB neurons (Fig. 13B). However, if there were also distinct Vhalf values for Kv1 channels containing one or two Kv1.2 subunits (xxx2 vs. xx22) this could explain why the Vhalf values from the subpopulation of +/− channels that contained Kv1.2 were also significantly correlated with the TsTx-percentage block (Fig. 13C). In oocytes, four distinct Vhalf values have been demonstrated for channels formed from Kv1.1 and Kv1.2 subunits (stoichiometries 1111, 1122, 1222, and 2222 had Vhalf of −31, −27, −23, and −16 mV, respectively; Akhtar et al. 2002). An unknown modulating factor may cause Kv1 channel Vhalf values to be about 20 mV more negative in auditory neurons than in oocytes (Trussell 1999). Our MNTB neuron results suggest that any such modulating factor does not overwhelm Vhalf differences between channels containing different numbers of Kv1.2 subunits.

The 10-mV negative shift of −/− Kv1 channels’ Vhalf relative to +/+ channels suggests they may be dominated by Kv1.1 subunits, or even be Kv1.1 homomers, because the published literature suggests Kv1.1 subunits would be the only Kv1 subunit type likely to exert a strong negative effect on Vhalf when substituting for Kv1.2 subunits. In oocyte expression systems, the average Vhalf values for each of Kv1.1 homomers, Kv1.2 homomers, and Kv1.6 homomers were −33, −26, and −17 mV and other Kv1 homomer types had Vhalf values of −26 to +4 mV (Akhtar et al. 2002; Grissmer et al. 1994; Grupe et al. 1990; Hopkins et al. 1994; Kalman et al. 1998; Lang et al. 2000; Stuhmer et al. 1989; Swanson et al. 1990). One caveat to this is that in some mammalian cell lines, Kv1.2 homomers had much more positive Vhalf values (Grissmer et al. 1994; Werkman et al. 1992). Preliminary experiments showed −/− Ikl had high sensitivity to block by external tetraethylammonium (TEA, Kd <1 mM) in contrast to +/+ Ikl (>5 mM; JX Gittelman and HM Brew, unpublished data). The Kd of expressed Kv1.1 homomers was also <1 mM (0.3–0.6 mM), contrasting with homomers of Kv1.6 (4 and 7 mM) or homomers of Kv1.2, Kv1.3, Kv1.4, or Kv1.5 (>10 mM; Grissmer et al. 1994; Grupe et al. 1990; Stuhmer et al. 1989; Swanson et al. 1990). Thus −/− Kv1 channels are probably Kv1.1 homomers or have 111x stoichiometry, given that TEA block is additive (Heginbotham and MacKinnon 1992).

Kv1.1-dominated channels, Kvbeta2, and surface expression

A strong surface expression of Kv1.1 homomers in −/− CNS would be particularly interesting both because Kv1.1 homomers have not been detected in mammalian brains (Coleman et al. 1999; Shamotienko et al. 1997; Wang et al. 1999) and because Kv1.1 subunits possess a strong ER retention motif (Shi et al. 1996). Although this retention motif is also present in Kv1.2 and Kv1.6 subunits, Kv1.1 subunits exhibit the greatest retention in expression systems (95%) and can apparently dominate retention in heteromers with one or two subunits of Kv1.2, or Kv1.4, which lacks the motif (Manganas and Trimmer 2000). Possibly MNTB neurons produce so much Kv1.1 and Kv1.2 that surface IKv1 is substantial despite very strong ER retention, accounting for Kv1.1 and Kv1.2 antibody staining being so strongly intracellular (Brew et al. 2003). Surface Kv1 channel density might instead be limited by the availability of the accessory protein Kvbeta2, present in MNTB neurons, which facilitates surface expression (Fonseca et al. 1998; Manganas and Trimmer 2000). However, mice lacking Kvbeta2 had only occasional seizures, lived for many months, and exhibited cold-swim–induced tremor, all consistent with a mild version of the Kcna1-null phenotype (McCormack et al. 2002; Smart et al. 1998; Zhou et al. 1998).

Kv1 channel voltage dependence in MNTB neurons and elsewhere

Some of the variation in reported Vhalf values for GKv1 between neuronal types may arise from differences in the Kv1.1:Kv1.2 balance of their Kv1 channel populations. In auditory neurons in rodent brain slices, GKv1 had Vhalf between −50 and −45 mV and toxin experiments suggested all the underlying channels contained Kv1.1, whereas a little over 50% contained Kv1.2: the percentages of TsTx-sensitive IKv1 were 65% in murine octopus neurons, 50% in rat MNTB, and 70% in rat bushy neurons (Bal and Oertel 2001; Dodson et al. 2002, 2003). If there were a gradient of Kv1.1:Kv1.2 expression across the tonotopic axis of rat MNTB, it could underlie the gradients of current amplitudes and Vhalf, which increased significantly from −46 mV laterally to −50 mV medially (Brew and Forsythe 2005). Tonotopic gradients of the Kv1.1:Kv1.2 expression ratio have been detected along the mouse spiral ganglion and within chick nucleus magnocellularis (NM), the avian equivalent of bushy neurons, along with gradients of firing properties indicating larger IKv1 where the ratio was largest, in regions with high best frequency (Adamson et al. 2002; Fukui and Ohmori 2004). Acutely isolated murine spiral ganglion neurons and chick NM neurons yielded the most negative reported Vhalf values for auditory GKv1, of −62 and −58 mV (Mo et al. 2002; Rathouz and Trussell 1998), perhaps suggestive of strongly Kv1.1-dominated channels.

DTX-sensitive conductances with less-negative Vhalf values of −27, −22, and −5 mV occur in striatal neurons (Shen et al. 2004) and pyramidal cells in neocortex and amygdala (Bekkers and Delaney 2001; Faber and Sah 2004). This may be because Kv1.1 is scarce or may mean these Kv1 channels lack whatever modulating factor causes the very negative Vhalf values for GKv1 in auditory neuron somata. At the giant synapse onto rat MNTB neurons, all the Kv1 channels contained Kv1.2 but only 37% of IKv1 was through Kv1.1-containing channels (Dodson et al. 2003). With binomial assembly this balance could be produced if the Kv1.1:Kv1.2 subunit ratio was 10:90, with most channels Kv1.2 homomers and 1222 channels. Based on our above-cited MNTB neuron results, we predict that this Kv1.2-dominated channel population would have Vhalf values less negative than any of the values from auditory neuron somata.

Although the populations of Kv1 channels in auditory neurons probably include channels with distinct Vhalf values, the voltage dependences of these channel populations were quite steep, with slope factors between 5 and 10 mV in the present study, and 6 and 10 mV in previous studies (Bal and Oertel 2001; Brew and Forsythe 1995; Dodson et al. 2002; Mo et al. 2002; Rothman and Manis 2003a). Similar values were found for homogeneous populations of expressed homomers of Kv1.1, Kv1.2, or Kv1.6, and some Kv1.1:Kv1.2 heteromers (Akhtar et al. 2002; Bosma et al. 1993; Grissmer et al. 1994; Grupe et al. 1990; Hopkins et al. 1994; Stuhmer et al. 1989; Swanson et al. 1990). The steep voltage dependence means quite small differences in Vhalf between +/+ and −/− MNTB neuron G and GKv1 are associated with large changes in channel open probabilities, sufficient to account for the genotypic differences in IKv1 amplitudes at particular membrane potentials (calculations not shown, but see Fig. 10). By analogy, a positive shift in Vhalf of a few millivolts could explain why IKv1 was reduced by 41% in MNTB neurons from Kcna1-null mice (Brew et al. 2003).

If differences in IKv1 are produced solely by changes in channel open probability, MNTB neuron somata may possess the same numbers and densities of Kv1 channels, irrespective of genotype (also with identical single-channel conductance). In support of this, the mean Gmax was similar in +/+, +/−, and −/− MNTB neurons (see results). It is reasonable to assume that +/+ MNTB neurons produce a larger total number of Kv1 subunit proteins than +/− neurons, which in turn produce larger numbers than −/− neurons because of the different numbers of Kv1.2 subunits produced, although our data argue against the idea that +/+ MNTB neurons have the highest Kv1 channel densities, at least at their somata. Thus the factors determining Kv1 channel density may be relatively independent of Kv1 subunit production.

Does the absence of Kv1.2 affect information processing by the −/− auditory system?

Enlarged IKv1 is likely in −/− spiral ganglion (SG) neurons whose axons are the auditory nerve fibers, and in their synaptic targets the −/− cochlear nucleus bushy cells (which send synapses to MNTB neurons) because in normal mice they exhibit MNTB-like firing, IKv1, and strong Kv1.1 and Kv1.2 expression (Adamson et al. 2002; Brew et al. 2003; Manis and Marx 1991; Mo et al. 2002; Wang et al. 1994). This enlarged IKv1 would predict reduced activity in both excitatory (SG and bushy cells) and inhibitory (MNTB projections) −/− auditory pathways, but −/− auditory responses in vivo are harder to predict because they would depend on the balance between the two (e.g., at LSO) and there could also be compensatory synaptic changes. Results from Kcna1-null mice also suggest auditory system abnormalities went beyond reduced IKv1 in SG cells, bushy cells, and MNTB neurons (Brew et al. 2003) because, in vivo, spontaneous firing rates and auditory thresholds were normal in bushy and MNTB neurons, although they exhibited increased jitter to auditory stimulation, but their maximum firing rates were actually reduced, perhaps suggesting compensatory strengthening of inhibition (Kopp-Scheinpflug et al. 2003).

Preliminary behavioral data show that −/− mice do not share the auditory abnormalities found in Kcna1-null mice, which had impaired ability to detect rapid sound offsets and changes in the azimuthal location of sounds, using reflex modification audiometry (Allen et al. 2003; Ison et al. 2002) because −/− mice aged P12 to P18 performed at least as well as controls (Brew et al. 2005). This suggests that firing patterns are approximately normal in the −/− auditory nerve. One interesting possibility is that the −/− auditory nervous system might have better than normal temporal resolution on some auditory processing tasks, if enlarged IKv1 leads to decreased time constants for membrane potential changes.

The known link between reduced firing in auditory afferents and audiogenic seizure susceptibility (ASS) may have some relevance to the −/− phenotype. Studies showing that experimentally-induced early hearing impairments led to ASS in normal mice (e.g., Chen and Fuller 1976) led to the idea that rodent ASS is caused by “reduction of neural activity in the auditory pathways from deafness during development” (Ross and Coleman 2000). In GEPRs, audiogenic seizures initiate in the inferior colliculus, where there is weakened γ-aminobutyric acid circuitry, thought to occur as compensation for the reduced activity in excitatory auditory inputs caused by their hearing impairment (Faingold 2002; Faingold et al. 1986). Although enlarged IKv1 is likely in −/− excitatory auditory afferents and could lead to reduced activity, we were not able to induce seizures audiogenically in young −/− mice. Note that GEPRs also exhibited reduced latencies to flurothyl-induced seizures, as well as heightened susceptibility to seizures induced by other chemical, electrical, or sensory stimuli (Franck et al. 1989; Reigel et al. 1986). However, the genes and chromosomes thus far linked to ASS in mice have not been Kcna genes or the chromosomes containing them (Lock et al. 1994; Misawa et al. 2002; Neumann and Collins 1992; Skradski et al. 2001).

Functionally distinct Kv1 channels and neuronal information processing

The range of Vhalf values possible for different Kv1.1/Kv1.2 stoichiometries suggests different neurons could tune their channel properties for different roles. For example, Kv1 channels are found in auditory pathways that rapidly relay temporally precise signals from the cochlea to the brain stem, by large axons with large synapses, and their main role is to powerfully and rapidly reset the membrane potential between synaptic events (Oertel 1983; Trussell 1999). (This may be analogous to a role for Kv1 channels in axonal conduction, with large depolarizing sodium currents at nodes of Ranvier substituting for synaptic inward currents.) Very negatively activating Kv1.1-dominated channels may be particularly well suited to this role. In contrast, membrane resetting between synaptic events would be disastrous in most other neurons, where instead temporal summation is crucial and less negatively activating Kv1 channels can shape onset latencies and firing rates as they inactivate during prolonged responses (e.g., Bekkers and Delaney 2001; Faber and Sah 2004; Shen et al. 2004; Storm 1988). Consistent with this, neurons within low-frequency parts of chick NM had the highest Kv1.2:Kv1.1 expression ratio and the smallest threshold currents and were also most likely to perform temporal summation because they received several smaller synapses, rather than a single large synapse from one auditory nerve fiber (Fukui and Ohmori 2004).

If the Kv1 channel density was the same in +/+, +/−, and −/− MNTB neurons, even though production of Kv1 subunits by the latter was probably reduced by deficits in Kv1.2 subunit numbers, this suggests neurons can adjust their potassium currents by fine-tuning channel properties rather than altering channel densities. This idea could be tested by generating mice with different Kcna gene configurations and recording from their MNTB neurons. For example, GKv1 should have the same amplitude and properties in MNTB neurons with two Kcna2 genes and four Kcna1 genes as the +/− neurons described here.

GRANTS

This work was supported by National Institute on Deafness and Other Communication Disorders Grants DC-003805, DC-02739, and P30-DC-04661. S. Y. Chiu was supported by National Multiple Sclerosis Society Grant RG-3058. During some of this work, H. M. Brew was supported by a Royal Society University Research Fellowship. Graduate training grants supported J. Gittelman (T32-GM-07108) and R. S. Silverstein (T32-DC-05361). T. Hanks was supported by a predoctoral fellowship from the Howard Hughes Medical Insititute.

Acknowledgments

We thank D. Perkel, J. Ison, and P. Allen for helpful comments.

Footnotes

  • 1 The online version of this article contains supplemental data.

  • The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

REFERENCES

View Abstract