Contrasting molecular composition and channel properties of AMPA receptors on chick auditory and brainstem motor neurons

Abstract

1. Neurons in the brainstem auditory pathway exhibit a number of specializations for transmitting signals reliably at high rates, notably synaptic AMPA receptors with very rapid kinetics. Previous work has not revealed a common structural pattern shared by the AMPA receptors of auditory neurons that could account for their distinct functional properties.

2. We have used whole-cell patch-clamp recordings, mRNA analysis, immunofluorescence, Western blots and agonist-evoked cobalt uptake to compare AMPA receptors on the first-, second- and third-order neurons in the chick ascending auditory pathway with those on brainstem motor neurons of the glossopharyngeal/vagal nucleus, which have been shown to have very slow desensitization kinetics.

3. The results indicate that the AMPA receptors of the cochlear ganglion, nucleus magnocellularis and nucleus laminaris share a number of structural and functional properties that distinguish them from the AMPA receptors of brainstem motor neurons, namely a lower relative abundance of glutamate receptor (GluR)2 transcript and much lower levels of GluR2 immunoreactivity, higher relative levels of GluR3 flop and GluR4 flop, lower relative abundance of the C-terminal splice variants GluR4c and 4d, less R/G editing of GluR2 and 3, greater permeability to calcium, predominantly inwardly rectifying I–V relationships, and greater susceptibility to block by Joro spider toxin.

4. We conclude that the AMPA receptors of auditory neurons acquire rapid kinetics from their high content of GluR3 flop and GluR4 flop subunits and their high permeability to Ca²⁺ from selective post-transcriptional suppression of GluR2 expression.

Ionotropic glutamate receptors (GluRs) of the AMPA subtype are assembled from four protein subunits, termed GluR1-4 or A-D. Native AMPA receptors are assembled from a variety of subunit, splice variant and mRNA editing combinations that result in diverse functional properties (Borges & Dingledine, 1998). Although some types of neuron are reported to show great cell-to-cell variation in AMPA receptor function (Angulo et al. 1997; Washburn et al. 1997), other cell classes are reported to display a predominant functional type of AMPA receptor. Rat CA3 pyramidal neurons, for example, express slowly desensitizing receptors with low permeability to calcium, and the AMPA receptors of Bergmann glial cells desensitize rapidly and show relatively high calcium permeability (Geiger et al. 1995). Since AMPA receptors mediate most rapid synaptic transmission in the CNS (Collingridge & Lester, 1989), understanding how different classes of neuron develop and maintain characteristic information-processing functions will require detailed analysis of the properties of their AMPA receptors.

Auditory neurons show a number of striking morphological and functional specializations that can be related to their roles in hearing (Trussell, 1999). The AMPA receptors of auditory neurons in birds and mammals also exhibit specializations. Several types of auditory neuron have AMPA receptors with unusually high permeability to divalent cations (Otis et al. 1995; Zhou et al. 1995; Caicedo et al. 1998) and very rapid desensitization rates – almost fivefold faster than the AMPA receptors of brainstem motor neurons, for example (Raman et al. 1994). The molecular bases for these AMPA receptor specializations are poorly understood. A few studies have analysed AMPA receptor subunit expression within certain auditory centres using in situ hybridization (Hunter et al. 1993; Sato et al. 1993; Niedzelski & Wenthold, 1995), mRNA analysis (Niedzelski & Wenthold, 1995) or immunohistochemistry (Petralia & Wenthold, 1992; Petralia et al. 1996, 1997; Levin et al. 1997; Wang et al. 1998; Caicedo & Eybalin, 1999). Geiger et al. (1995) included one type of auditory neuron in their study of the correlation of AMPA receptor functional properties with GluR mRNA profiles. Each of these studies has provided a partial characterization of AMPA receptor structure and there are discrepancies in their results which make it unclear whether there is a common structural pattern shared by the AMPA receptors of auditory neurons. To address this issue, we used mRNA analysis, whole-cell patch-clamp recordings, immunofluorescence, Western blots and agonist-evoked cobalt uptake to compare the molecular and functional properties of AMPA receptors in the first three neural centres of the chick auditory pathway – the cochlear ganglion (CG), nucleus magnocellularis (NM) and nucleus laminaris (NL) – with those of motor neurons in the glossopharyngeal/vagal nucleus (NIX/X), which Raman et al. (1994) have shown to have very slow desensitization kinetics.

METHODS

Chickens (Gallus domesticus) at embryonic day (E)17 (Hamburger-Hamilton stage 43) were decapitated, unless otherwise stated. Tissue from the CG, NM, NL and NIX/X was immediately obtained by dissection from the skull (CG) or from 300 μm-thick vibratome sections of brainstem and then used in the various analyses described below. All experiments were carried out in accordance with the guidelines for animal experimentation of the University of Utah's Institutional Animal Care and Use Committee.

cDNA synthesis

RNA was extracted as previously described (Lee et al. 1998) using a modification of the guanidinium isothiocyanate method (TRIZOL, Gibco BRL). The purity and concentration of RNA were assessed by spectrophotometry. For reverse transcription, this RNA was used in a reverse transcription reaction with Superscript II, a modified Moloney murine leukaemia virus reverse transcriptase (Gibco BRL), and an oligo-dT primer to produce cDNA for use as templates in PCR.

Analysis of relative abundance of AMPA receptor subunits

PCR of a homologous region of GluR1-4 was performed using PCR Supermix (Gibco BRL) and the pan-AMPA degenerate primers listed below, according to the manufacturer's protocol.

The primers amplified a 550 base pair (bp) fragment of GluR1 (bp 1853-2403), a 549 bp fragment of GluR2 (bp 2040-2589), a 553 bp fragment of GluR3 (bp 1950-2503) and a 539 bp fragment of GluR4 (bp 1841-2380). The cycling parameters were 91°C, 30 s; 53°C, 30 s; 72°C, 1 min, for 35 cycles. The PCR cycle was preceded by an initial denaturation for 2 min at 94°C and followed by a final extension of 5 min at 72°C. Equal aliquots of the pan-AMPA PCR product were then enzymatically digested using subunit-specific restriction endonucleases (Pvu II cuts GluRI at bp 2120, Bam H I cuts GluR2 at bp 2493, Eco R I cuts GluR3 at bp 2161 and Pst I cuts GluR4 at bp 2139). The resulting fragments were resolved on agarose gels. After being photographed under UV illumination, each gel was scanned using a Molecular Dynamics Fluorimager 575 and the resulting data were analysed with ImageQuant software (Molecular Dynamics, Sunny Vale, CA, USA). In each case, the fluorescence of the cut bands was expressed as a percentage of the total to yield the relative abundance of the subunit in question.

Analysis of relative abundance of GluR4 C-terminal splice variants

Three C-terminal splice variants of GluR4 (GluR4c, 4d and 4s) originally described by Ravindranathan et al. (1997) were studied by the following methods. PCR of the GluR4 region that contains the flip/flop sites and the C-terminal splice sites was performed using PCR Supermix and the pan-GluR4 primers listed in Table 1, following the manufacturer's protocol. These primers amplified a 719 bp fragment of GluR4 (bp 1892-2611), an 832 bp fragment of GluR4c (bp 1892-2724), a 903 bp fragment of GluR4d (bp 1892-2795) and a 470 bp fragment of GluR4s (bp 1892-2362). The cycling parameters were 91°C, 30 s; 48°C, 30 s; 72°C, 1 min, for 35 cycles. The PCR cycle was preceded by an initial denaturation for 2 min at 94°C and followed by a final extension of 5 min at 72°C. Samples were run on a 2 % agarose gel at 65 mV for 8 h, photographed under UV illumination, and the relative abundance of the individual C-terminal splice variants was quantified on the fluorimager. The identity of each band was confirmed by sequencing using ABI PRISM fluorescent dye terminator cycle sequencing.

Table 1.

PCR primers and conditions for AMPA receptor subunits and C-terminal splice variants

Receptor	Primer sequence	Size	Amplified region	PCR conditions*
GluR1	5′ TGG TGG TTC TTC ACC CTG ATC AT 3′	707 bp	1904–2611	48 °C
	5′ TAT GGC TTC ATT GAT GGA TTG C 3′
GluR2	5′ GGT GTC TCC CAT TGA AAG TG 3′	690 bp	2155–2845	55 °C
	5′ CCA TAT ACG TTG TAA CCT TCC 3′
GluR3	5′ GAT GAA CGA GTA CAT CGA GC 3′	522 bp	2313–2835	55 °C
	5′ CAC ATT CTC TGC TGC TTC TC 3′
GluR4	5′ TGG TGG TTC TTC ACC CTG ATC AT 3′	719 bp	1892–2611	48 °C
	5′ ACT CCC AGT GAT GGA TAA CCT G 3′
GluR4c	5′ GGT GGT TCT TCA CCC TGA TCA 3′	758 bp	1893–2653	59 °C
	5′ CGG TTC CAT ATA CGT TGT AAC C 3′
GluR4d	5′ GGT GGT TCT TCA CCC TGA TCA 3′	787 bp	1893–2682	59 °C
	5′ CTC TGT GTC TGC TCT GTA CTG 3′

^*Annealing temperature.

Single-cell mRNA profiling

The relative abundance of mRNAs for GluR1-4 in NM neurons was studied by a modification of the single-cell mRNA profiling method of Eberwine et al. (1992). Acutely dissociated NM neurons were obtained following enzymatic digestion and trituration, as described previously (Raman et al. 1994). Individual NM neurons were identified and carefully aspirated under visual control into a pipette containing 2 μl of a reverse transcription (RT) mixture: 50 mM KCl, 10 mM Tris-HCl, 5 mM MgCl₂, 1 mM each dNTP, 50 ng of primer (5′ aaa cga cgg cca gtg aat tgt aat acg act cac tat agg cgc t₁₅ 3′), 10 U RNase inhibitor and 2 U reverse transcriptase. The contents of the pipette were then transferred to a tube containing 20 μl of the reverse transcription cocktail and incubated at 42°C for 1 h. Following this, the synthesized single-stranded cDNA was made double stranded with the addition of a second-strand solution containing 100 mM KCl, 20 mM Tris-HCl, 5 mM MgCl₂, 10 mM ammonium sulfate, 1 mM ATP, 5 mg BSA, 1 U T4 DNA ligase, 1 U RNase H and 20 U DNA polymerase I (final volume, 80 μl) to the RT reaction. The entire mixture was incubated for 1 h at 12°C then for 1 h at 22°C. The double-stranded cDNA was blunt-ended through the addition of 10 U T4 DNA polymerase and 10 U Klenow (37°C, 1 h), then purified by ethanol precipitation and drop dialysis. Antisense RNA (aRNA; see Results) was synthesized using 4 μl of transcription buffer containing 10 mM DTT, 0.5 mM each of rATP, rUTP and rGTP, and 20 U RNAse inhibitor, with 5 μl cDNA template, 2.1 μl radiolabelled CTP (800 mCi mM⁻¹; Amersham, Arlington Heights, IL, USA) and 2000 U T7 RNA polymerase in a total volume of 20 μl. The reaction was incubated at 37°C for 4 h and then precipitated with ethanol. This aRNA was then used to probe slot blots containing equimolar amounts of cDNAs for rat GluR1-4 (a gift of Dr James Boulter at the Salk Institute), which share > 90 % sequence homology with chick GluR1-4 (Paperna et al. 1996), as well as cDNAs for chick actin and rat glial fibrillary acidic protein (GFAP). Cross-hybridization controls showed that under the hybridization conditions used, there was no detectable cross-hybridization between AMPA receptors (data not shown). The relative levels of AMPA receptor subunit mRNA expression were then expressed as a fraction of the total AMPA receptor mRNA.

Analysis of flip/flop isoform expression

PCR of the flip/flop region of individual AMPA receptor subunits and the GluR4 C-terminal splice variants was performed using PCR Supermix as described previously by Ravindranathan et al. (1996, 1997). The specific primers and PCR conditions are summarized in Table 1. The cycling parameters were as follows: 91°C denaturation for 30 s and 72°C extension for 60 s for 35 cycles. The PCR cycle was preceded by an initial denaturation for 2 min at 94°C and followed by a final extension of 5 min at 74°C. Restriction enzymes required to distinguish flip and flop domains of the four AMPA receptor subunits and C-terminal splice variants were identified by analysing restriction sites present in cloned chick AMPA receptor and C-terminal flip/flop sequences (Ravindranathan et al. 1996, 1997). Hsp 92 I (Promega) cuts GluR1 flip at bp 2371, GluR2 flip at bp 2558, GluR4 flip at bp 2358, GluR4c flip at bp 2358 and GluR4d flip at bp 2358. Hpa I (Promega) cuts GluR2 flop at bp 2517, GluR3 flop at bp 2431, GluR4 flop at bp 2317, GluR4c flop at bp 2317 and GluR4d flop at bp 2317. Tsp 509 I (New England Biolabs) cuts GluR1 flop at bp 2383. Msp I (Promega) cuts GluR3 flip at bp 2526. Restriction enzyme digests of PCR products were performed under standard conditions and the resulting fragments were resolved on agarose gels, photographed under UV illumination, and the relative ratio of flip and flop isoforms quantified on the fluorimager.

Electrophysiology

Neurons from the CG, NM and NIX/X were dissociated by enzymatic digestion and trituration as previously described (Raman et al. 1994). These neurons were then plated on polylysine-coated 35 mm dishes in minimal essential medium (MEM) and allowed to adhere for 1 h. Just prior to recording, the MEM was removed from the Petri dishes and replaced with a bathing solution containing 142 mM NaCl, 5 mM KCl, 2 mM CaCl₂, 2 mM MgCl₂, 10 mM Hepes and 10 mM glucose buffered to pH 7.3 with NaOH. In some experiments, cells were perfused with a solution in which NaCl was replaced with the impermeant cation N-methylglucamine and CaCl₂ concentration was raised to 10 mM. Pipettes were pulled on a two-stage Sutter puller and the tips polished on a Narishige microforge to yield electrodes with a resistance of 3–5 MΩ in the external solution; they were filled with an internal solution containing 145 mM CsCl, 4 mM MgCl₂, 10 mM Hepes and 10 mM EGTA. Spermine (30 μM) was added to the internal solution to prevent washout of rectification (Donevan & Rogawski, 1995; Koh et al. 1995). Neurons were voltage clamped at -60 mV unless otherwise indicated. The electrophysiological properties of the AMPA receptors on these neurons were then ascertained using agonists (kainate and glutamate) and antagonists (Joro spider toxin (JSTX) and GYKI53655). Drug solutions were applied using a multi-barrelled rapid perfusion system described previously (Donevan & Rogawski, 1993), in which the solution exchange time constant at the bare tip of a recording electrode was less than 2 ms. One barrel contained bathing solution while the other barrels contained agonists and drugs, either alone or in combination. The constant-field equation was used to estimate the relative permeability of the AMPA receptor responses to Ca²⁺, as described previously (Iino et al. 1990):

(1)

where V_rev is the reversal potential in the Ca²⁺ solution, P_Ca and P_Cs are the permeability coefficients for Ca²⁺ and Cs⁺, and F, R and T have their usual meanings. V_rev measurements in the high-calcium solution were not corrected for a -1.9 mV liquid junction potential.

Analysis of Q/R editing

The relative amounts of Q/R-edited and -unedited GluR2 transcripts in the CG, NM, NL and NIX/X were determined in restriction digest assays, based on the method of Akbarian et al. (1995) and Lee et al. (1998). RT-PCR with one ³²P-labelled primer was used to amplify a 331 bp fragment of GluR2 (bp 1582-1913) containing transmembrane domain II and the Q/R editing site from CG, NM, NL and NIX/X cDNA. The primer sequences are given below.

Forward: 5′ cct cag aag tcc aag cca gga gtg 3′

Reverse: 5′ cag gaa ggc agc taa gtt agc cg 3′

PCR products derived from unedited transcripts contained two sites for the Bbv I restriction endonuclease (GCAGC(N)₈), giving rise to three restriction products of 248, 63 and 20 bp. Edited transcripts have lost one of these sites, resulting in two restriction digest products of 311 and 20 bp. The RT-PCR products were digested with Bbv I and the resulting fragments resolved by SDS-PAGE. The ratio of edited to unedited GluR2 within a sample was determined by scintillation radiometry (ImageQuant) following exposure of the gel to a phosphor screen.

Analysis of R/G editing

Using cDNA from the NIX/X, CG, NM and NL as templates, the R/G editing regions of GluR2, 3 and 4 were amplified by PCR using PCR Supermix and the individual AMPA subunit-specific primers and conditions outlined in Table 1. GluR2 and 3 PCR products were purified according to the procedure we have previously described (Lee et al. 1998). GluR4 PCR products were purified using the QIAquick protocol (Qiagen) as per the manufacturer's instructions. The products were then sequenced using ABI PRISM fluorescent dye terminator cycle sequencing and the original PCR subunit-specific primers. The R/G site was identified on the chromatogram (GluR2 at amino acid 743, GluR3 at amino acid 747 and GluR4 at amino acid 744) and the editing level in the amplified fragment was ascertained by measuring the peak height of the G signal (edited) versus the total signal (A + G) at the editing site (Lee et al. 1998).

Agonist-evoked cobalt uptake

Whole dissected cochlear ganglia and brainstems with the cerebellum removed were processed for kainate-evoked cobalt accumulation by the method described in detail by Zhou et al. (1995). After the tissue had been allowed to rest in oxygenated chick Tyrode solution for 2 h, it was placed in oxygenated cobalt uptake buffer (140 mM NaCl, 4 mM KCl, 3 mM CaCl₂, 10 mM Hepes, 10 mM dextrose and 5 mM CoCl₂, pH 7.3) for 20 min with or without (control) added test substance – 100 μM kainic acid. The tissue was rinsed in uptake buffer followed by a rinse in uptake buffer to which 2 mM EDTA was added. This was followed by a wash in uptake buffer before the tissue was developed for 10 min in uptake buffer containing 1.2 % ammonium sulfide. Following double rinses in uptake buffer, the tissue was fixed in 4 % paraformaldehyde in 0.1 M phosphate buffer for 20 min, rinsed twice in distilled water, and immersed for 45 min at 50°C (all other incubations were at room temperature) in silver enhancement solution containing 1 mM AgNO₃, 292 mM sucrose, 15.5 mM hydroquinone and 42 mM citric acid. Cryostat sections were then cut at 15–18 μm intervals and mounted on slides to visualize cobalt uptake. Images captured on film with a photomicroscope were digitized with a scanner (LeafScan45) and the final figures were prepared with Adobe Photoshop software and printed on a high-resolution digital printer.

GluR2 immunochemistry

GluR2 immunoreactivity in the cochlear ganglion and brainstem was studied by immunofluorescence and Western blotting using a monoclonal antibody directed against an N-terminal peptide sequence of the rat GluR2 subunit (MAB 397, Chemicon, Temecula, CA, USA). For immunofluorescence, E17 chick embryos deeply anaesthetized with sodium pentobarbital (25 mg i.p.) were perfused transcardially with a wash (0.9 % NaCl, 1000 i.u. heparin, 1.0 g NaNO₂ in 500 ml) followed immediately by 4 % paraformaldehyde in 0.1 M phosphate buffer. The brainstem was dissected out and placed in fixative for 3 h and then stored at 4°C in 30 % sucrose and 0.1 M phosphate buffer. The brainstem was embedded and frozen in OCT (VWR, Salt Lake City, UT, USA). Coronal sections 18 μm thick containing NM, NL and NIX were cut and placed on microscope slides. The portion of the skull containing the basilar papilla and cochlear ganglion was blocked and decalcified by immersion for 1 week at 4°C in fixative containing 5 % EDTA. This tissue was then cut in the coronal plane at 18 μm intervals on a freezing microtome and the sections were thaw-mounted onto gelatinized slides. Sections of brainstem or inner ear on slides were blocked for 20 min with 1 % normal goat serum (Jackson Immunoresearch, West Grove, PA, USA) and 0.4 % Triton X-100 (Sigma) in phosphate-buffered saline. The tissue was then incubated overnight at 4°C with anti-GluR2 antibody diluted 1:500 in the blocking buffer. Primary antibody incubation was continued at room temperature for 1–3 h and was followed by 3 × 10 min washes in PBS. Each immunofluorescence experiment included some sections that had not been exposed to the primary antibody (control). Alexa 488-labelled secondary antibody (Molecular Probes) was diluted 1:800 in 0.4 % normal goat serum and PBS. The tissue was incubated with the secondary antibody for 90 min followed by 3 × 10 min washes in PBS. Slides were mounted with Fluorsave (Calbiochem) and stored in the dark at 4°C. Slides were read and digital images captured using an Olympus Fluoview confocal microscopy system. Final figures were prepared with Photoshop software and printed on a high-resolution digital printer.

For Western blot analysis of GluR2 expression, NM, NL and NIX were dissected from 300 μm coronal vibratome sections, placed into harvesting buffer (50 mM Trizma buffer (pH 7.4) containing 1 mM EDTA, 1 % Triton X-100, 7 μg ml⁻¹ chymostatin, 7 μg ml⁻¹ leupeptin, 5 μg ml⁻¹ aprotinin, 7 μg ml⁻¹ antipain, 7 μg ml⁻¹ pepstatin and 350 μg ml⁻¹ phenylmethylsulfonyl fluoride (PMSF)), homogenized and sonicated; whole cerebellum obtained from an adult male rat deeply anaesthetized with halothane and killed by decapitation was processed by the same method for use as a positive control. Connective tissue was removed from preparations by centrifugation and the protein concentration of samples was determined using a BCA assay (BioRad). Fifty micrograms of protein were resolved via 6.5 % SDS-PAGE at 100 V for 2 h at room temperature before transfer onto Immobilon-P membrane (Fisher) at 100 V for 2 h at 4°C. Blots were probed overnight at 4°C with the monoclonal antibody against GluR2 (Chemicon MAB 397) diluted 1:300. Bound primary antibodies were detected as described by Amersham using their enhanced chemiluminescence (ECL) protocol.

Statistical analyses

Descriptive and inferential statistical analyses were made with StatView 5.0 software (SAS Institute, Cary, NC, USA). For comparison of the relative abundance of GluR1-4 in NM as determined by the RT-PCR and single-neuron aRNA methods, the non-parametric Kolmogorov-Smirnov test was used. For other comparisons in the mRNA analysis, a factorial analysis of variance for each dependent variable was performed across the four neural structures studied, with Scheffé‘s method used for posthoc comparisons. For electrophysiological results, Student's unpaired t tests were used. For all inferential comparisons, α = 0.01.

RESULTS

Relative abundance of AMPA receptor subunit mRNAs in auditory and motor neurons

Degenerate primers based on sequences common to chick GluR1-4 were used to amplify AMPA receptor cDNAs from different brain regions, under the assumption that the relative abundance of each subunit would be maintained (Lee et al. 1998). Subunit-specific restriction endonucleases were then used to estimate the relative abundance of each AMPA receptor subunit mRNA. The relative abundance calculated using Fluorimager 575 and ImageQuant software was confirmed by scanning autoradiograms on a gel documentation system and subsequent analysis with BioRad MultiAnalyst software. Data were obtained from eight independent RNA preparations. As shown in Fig. 1A, the relative abundance patterns for the three auditory structures shared many similarities and differed from the pattern seen in the glossopharyngeal/vagal motor nucleus. In the CG, NM and NL, GluR1 transcripts accounted for 10–12 % of the total, whereas in NIX/X the mean (±s.e.m.) level was 14.5 ± 0.6 %. Analysis of variance showed a very highly significant main effect of tissue source on GluR1 transcript level (F_3,27 = 10.4, P < 0.0001) and posthoc comparisons revealed significant differences between NIX/X and both CG and NM; there were no reliable differences among the three auditory structures or between the NIX/X and NL (P = 0.09). In the three auditory structures, GluR2 transcripts accounted for 32–36 % of the total, whereas in NIX/X the level was 50.6 ± 2.3 %. Statistical analysis showed a very highly significant main effect of tissue source on GluR2 transcript level (F_3,27 = 47.4, P < 0.0001) and there were reliable differences between NIX/X and each of the three auditory structures but no differences among the CG, NM and NL. In the CG, NM and NL, GluR3 accounted for 26–35 % of the total AMPA receptor transcripts but in NIX/X the level was 16.4 ± 1.9 %. Analysis of variance showed a very highly significant effect of tissue source on GluR3 transcript level (F_3,27 = 39.8, P < 0.0001) and posthoc comparisons revealed that NIX/X differed reliably from all three auditory structures; GluR3 levels in the CG and NL also differed reliably but there were no other significant differences between the auditory structures. In the three auditory structures, GluR4 transcripts accounted for 23–26 % of the total, whereas the NIX/X showed a level of 18.6 ± 1.4 %. Statistical analysis showed a significant main effect of tissue source on GluR4 level (F_3,27 = 6.0, P = 0.0028); there was a reliable difference in GluR4 expression between NIX/X and NL and a marginally reliable difference between NIX/X and NM (P = 0.016) but there were no other significant differences.

Figure 1. Relative abundance of mRNAs for GluR1-4 in auditory and motor structures.

A, gels and bar graphs showing the relative abundance of AMPA receptors in the glossopharyngeal/vagal nucleus (Nucleus IX/X), the cochlear ganglion, the nucleus magnocellularis and the nucleus laminaris. In each case, the lane in the gel labelled U indicates the undigested PCR product generated by amplification with pan-AMPA primers and the respective cDNAs as described in Methods. Lanes labelled 1, 2, 3 and 4 contain equal amounts of the respective PCR products digested with subunit-specific restriction endonucleases. Arrows indicate the 600 bp mark. Bar graphs showing the mean (+s.e.m.) relative abundance of AMPA receptor message given as a percentage of the total AMPA receptor transcript expression are shown below each gel. Note the higher relative abundance of GluR2 transcripts in the glossopharyngeal/vagal motor nucleus. B, the relative abundance of AMPA receptor transcripts was also determined by single-neuron mRNA profiling of five acutely dissociated NM neurons. In the blot on the left, the profile of a single NM neuron is shown; positive signals are seen for GluR1-4 and β actin (+) but not for GFAP (-), an astrocyte-specific marker; the absence of the glial marker confirms that this profile is from a neuron uncontaminated by glial RNA. The relative abundance of GluR1-4 was determined separately for five neurons and the results are summarized in the bar graph as mean (+s.e.m.) percentages of the total AMPA receptor transcript expression accounted for by each subunit. Note that the RT-PCR/restriction enzyme (A) and aRNA (B) methods show similar patterns and rank orders of relative abundance for the four subunits.

Since the relative levels of expression of AMPA receptor subunits were ascertained using mRNA obtained from whole ganglia or nuclei dissected from 300 μm-thick sections, the possibility of glial contamination confounding the results must be considered. To confirm directly that data from whole nuclei or ganglia reflect AMPA receptor abundance in neurons, we used a non-PCR based single-cell mRNA profiling method to analyse the relative abundance of GluR1-4 in a control experiment with acutely isolated NM neurons (Fig. 1B). This ‘aRNA’ technique results in a linear amplification of all polyadenylated mRNAs in a single cell (Eberwine et al. 1992). An advantage of this method over the PCR/restriction enzyme approach is that the possibility of unequal cutting by restriction enzymes is eliminated. The relative abundance data from five single NM neurons (Fig. 1B) showed the same rank order for GluR1-4 as that obtained by RT-PCR of the whole NM from eight independent samples (Fig. 1A) and there was no reliable difference in the relative abundance patterns generated by the two methods (P = 0.063). These findings support the conclusion that data obtained with the PCR/restriction enzyme method from dissected tissue samples accurately reflect neuronal mRNA levels; accordingly, all of the mRNA analyses were made with this method.

Relative abundance of C-terminal isoforms of GluR4 in auditory and motor structures

Ravindranathan et al. (1997) described three new C-terminal splice variants of the chick GluR4 subunit (Fig. 2A). GluR4c, which is homologous to the rat GluR4c cloned from rat cerebellum by Gallo et al. (1992), differs from GluR4 by having a 113 bp insert spliced in downstream of the flip/flop splice site. GluR4d contains a 184 bp insert spliced into the same site as GluR4c. Both of these novel isoforms are expressed in flip and flop variants. GluR4s is a shortened transcript that lacks the nominal fourth transmembrane domain and flip/flop domain and shares a common C-terminal region with GluR4. Thus, counting the flip and flop isoforms of GluR4 and the three novel variants, seven distinct GluR4 transcripts are expressed in chick brain. All of these transcripts are included in the GluR4 relative abundance data illustrated in Fig. 1. Because we have previously shown (Ravindranathan et al. 1997) that the novel GluR4 isoforms are expressed in cerebellar granule cells at aggregate relative abundances comparable to that of GluR4, we thought it important to compare expression of these transcripts in auditory and motor structures. As shown in Fig. 2B, all three auditory structures expressed GluR4c RNA at levels about 25–30 % that of GluR4 whereas NIX/X expressed GluR4c at a relative abundance 55 % that of GluR4. Analysis of variance revealed a very highly significant effect of tissue source on the relative abundance of GluR4c (F_3,8 = 65.2, P < 0.0001) and posthoc comparisons showed that NIX/X differed significantly from all three auditory structures, none of which differed from any other. Similarly, the three auditory structures expressed GluR4d at levels 20–25 % of GluR4 whereas NIX/X expressed GluR4d message at 40 % that of GluR4. There was a highly significant effect of tissue source on expression of GluR4d mRNA (F_3,8 = 21.3, P = 0.0004) and, as with GluR4c, NIX/X differed reliably from the auditory structures, none of which differed significantly from any other. There were no significant differences in expression of GluR4s.

Figure 2. Relative abundance of the C-terminal splice variants of GluR4.

A, summary diagram of novel GluR4 isoforms described in chick brain by Ravindranathan et al. (1997). The diagram shows the splice sites and the sizes of the spliced regions. Splicing at the flip/flop and 4c sites can give rise to a total of seven variants of GluR4. B, gels and bar graphs showing the ratio of the C-terminal splice variant mRNAs to GluR4 mRNA in the glossopharyngeal/vagal nucleus (Nucleus IX/X), the cochlear ganglion, the nucleus magnocellularis and the nucleus laminaris. Arrows indicate the 600 bp mark. Bar graphs show means (+s.e.m.). Note the increased ratio of the C-terminal splice variants in the glossopharyngeal/vagal nucleus.

Auditory neurons show strong kainate-evoked cobalt uptake and low GluR2 immunoreactivity

Agonist-induced cobalt uptake can be used to assess divalent cation permeability through non-NMDA receptor channels. The entry of cobalt ions into neurons mirrors Ca²⁺ influx under physiological conditions and the cobalt is easily visualized as an insoluble cobalt sulfide precipitate (Pruss et al. 1991). We have previously shown that kainate-evoked cobalt accumulation in NM and NL neurons occurs exclusively through AMPA receptors (Zhou et al. 1995). In the present experiments, we examined cobalt uptake in four dissected cochlear ganglia and in four brainstems exposed to kainic acid. As illustrated in Fig. 3, cobalt accumulation evoked by kainic acid was absent in the NIX/X, present in the majority of neurons of the cochlear ganglion, and strong in the NM and NL. Although most neurons in the cochlear ganglion showed heavy cobalt accumulation, some did not (Fig. 3B). Immunofluorescence studies on six animals using a specific monoclonal antibody showed strong GluR2 immunoreactivity in NIX/X and above-background but lower expression in the CG, NM and NL. Western blots (including the representative example illustrated in Fig. 3I) were used to compare GluR2 immunoreactivity in two to three samples of adult rat cerebellum, the NIX/X, CG, NM and NL. These data demonstrate that, despite substantial levels of GluR2 transcript expression, most auditory neurons express very little GluR2 protein and show strong kainate-evoked cobalt uptake. In contrast, NIX/X neurons express GluR2 at high levels and show no kainate-evoked cobalt uptake.

Figure 3. Kainate-evoked cobalt uptake and GluR2 immunoreactivity.

Cobalt accumulation evoked by kainic acid was absent in neurons of the NIX/X (arrows in A), present in many neurons of the CG (B), and strong in all neurons of the NM (C and D) and NL (arrows in D). Although many neurons in the CG showed cobalt accumulation, some did not (arrow in B). Immunofluorescence with a specific monoclonal antibody showed substantial GluR2 immunoreactivity in NIX/X (E) and above-background but lower expression in the CG (F), NM (G) and NL (H); the sections in E, G (including inset) and H come from a single animal and were processed and photographed under identical conditions. The inset in G is from a section through NM stained only with the Alexa 488-labelled secondary antibody and illustrates the low level of background staining seen in the immunofluorescence experiments. I, a representative example of a Western blot comparing GluR2 immunoreactivity in samples of adult rat cerebellum (RCB), the NIX/X, CG, NM and NL. These data demonstrate that, despite substantial levels of GluR2 transcript expression, most auditory neurons express very little GluR2 protein and show strong kainate-evoked cobalt uptake. In contrast, NIX/X neurons express GluR2 at high levels and show no kainate-evoked cobalt uptake. Calibration bars: 100 μm in A and D, 25 μm in C, and 50 μm in B, E, F, G and H.

GluR2 subunits in both auditory and motor neuron AMPA receptors are fully edited at the Q/R site

Editing of GluR2 RNA at the Q/R site has an important influence on the calcium permeability of AMPA receptors containing a GluR2 subunit (Sommer et al. 1991). While it has been reported that GluR2 subunits from late embryonic stages onward are completely edited (Monyer et al. 1991), the possibility has not been excluded that some populations of mature cells may lack complete GluR2 editing. Also, the realization that expression of even relatively small amounts of unedited GluR2 can affect synaptic function (e.g. Feldmeyer et al. 1999) makes it important to assess Q/R editing directly in cell populations of interest. By means of a restriction enzyme-based editing assay, we found that GluR2 was essentially completely edited in the whole brain and in NM at E17 (Table 2). The CG, NL and NIX/X showed very small amounts of unedited GluR2, amounting to less than 0.5 % of the total GluR2 transcript (Table 2). Thus, all of the GluR2 transcripts analysed in the present study were essentially fully edited at the Q/R site.

Table 2.

mRNA editing of AMPA receptor subunits

Q/R editing*
GluR2
WB	0 ± 0
NIX/X	0.3 ± 0.1
CG	0.4 ± 0.1
NM	0 ± 0
NL	0.3 ± 0.1

	R/G editing†
	GluR2	GluR3	GluR4
NIX/X	100 ± 0	100 ± 0	100 ± 0
CG	62.1 ± 1.2	71.4 ± 0	100 ± 0
NM	61.4 ± 2.9	64.0 ± 3.1	100 ± 0
NL	70.6 ± 3.2	84.0 ± 9.3	89.7 ± 6.0

^*Percentage of GluR2 transcript with Q at position 586 (means ±s.e.m.).

^†Percentage of transcript with G at position 743 (GluR2), 747 (GluR3), or 744 (GluR4) (means ±s.e.m.). WB, whole brain; NIX/X, nucleus of the glossopharngeal/vagal nerves; CG, cochlear ganglion; NM, nucleus magnocellularis; NL, nucleus laminaris.

R/G editing of GluR2 and 3 differs between auditory and motor neurons

Since increased RNA editing at the R/G site of GluR2-4 has been reported to increase the speed of recovery from receptor desensitization (Lomeli et al. 1994), it was of interest to compare the degree of such editing in the three auditory structures and NIX/X, which exhibit very different AMPA receptor desensitization kinetics (Raman et al. 1994). As shown in Table 2, GluR4 was edited completely in NIX/X, CG and NM and 90 % in NL; there was no significant variation in GluR4 editing across these structures (F_3,11 = 0.9, P = 0.47). In contrast, transcripts for GluR2 and 3 were completely edited in NIX/X but only about 60–80 % edited in the auditory structures (Table 2). There was a significant overall difference for GluR2 (F_3,11 = 61.0, P < 0.0001) and posthoc comparisons showed that NIX/X differed reliably from all three auditory structures, none of which differed from the others. For GluR3, there was a significant main effect (F_3,11 = 9.3, P = 0.002) but posthoc comparisons showed that only the NIX/X versus NM difference was statistically significant.

AMPA receptors on auditory and motor neurons show distinct patterns of rectification, calcium permeability and block by polyamine toxin

Since the kainate-evoked cobalt uptake data suggested that auditory, but not motor, neurons express calcium-permeable AMPA receptors, we examined the functional properties of AMPA receptors in these cell populations directly by making whole-cell recordings from acutely dissociated CG, NM and NIX/X neurons. Specifically, we compared rectification properties, calcium permeability, and the sensitivity of the receptors to JSTX, one of the polyamine spider toxins that have been shown to block calcium-permeable, but not calcium-impermeable, AMPA receptors (Iino et al. 1996). The concentration of JSTX used (5 μM) was chosen based on previous reports of the concentrations needed to produce near-maximal block of calcium-permeable AMPA receptors in dissociated neurons (Washburn & Dingledine, 1996; Washburn et al. 1997).

Cochlear ganglion neurons

Cochlear ganglion neurons could be separated into two groups based on the functional properties of the AMPA receptor responses of the individual neurons. Most CG neurons (14/22) showed inwardly rectifying current responses to 100 μM kainate. The rectification index (RI) was calculated as the ratio of the kainate current at +40 mV over the kainate current at -60 mV, and was 0.31 for the inwardly rectifying neuron shown in Fig. 4A (mean, 0.31 ± 0.02). As expected of cells showing such an inwardly rectifying profile, the kainate response showed significant inward current when the normal buffer was replaced with one lacking Na⁺ and containing a high concentration of Ca²⁺. The P_Ca/P_Cs ratio was estimated using the constant field equation (eqn (1) in Methods) and was 0.95 for the neuron in Fig. 4A (mean, 0.98 ± 0.07; n = 4). Finally, as expected of AMPA receptors containing low levels of GluR2, the kainate response of this neuron was almost completely blocked by 5 μM JSTX (91.5 %; mean, 88.6 ± 1.2 %; n = 8), a relatively selective antagonist of GluR2-lacking, calcium-permeable AMPA receptors. The functional properties of the AMPA receptors of the remaining population of CG neurons were significantly different (P < 0.001), as indicated by the asterisks in Fig. 4C. These AMPA receptor responses (Fig. 4B) were linear to slightly outwardly rectifying (RI, 0.86 ± 0.02; n = 8), showed lower permeability to calcium (P_Ca/P_Cs, 0.42 ± 0.02; n = 4) and were less sensitive to block by JSTX (percentage block, 25.5 ± 3.1 %; n = 5). These properties are consistent with the expression of GluR2-containing AMPA receptors.

Figure 4. Electrophysiology of CG neurons.

Left, the traces in A and B are from different neurons and show responses to 100 μM kainate at various holding potentials between -80 and +60 mV. Middle, the current-voltage relationship for these two neurons is shown in normal high-Na⁺-containing buffer (•) and one in which the Na⁺ was replaced with the impermeant ion N-methylglucamine and the Ca²⁺ concentration was raised to 10 mM (^). The arrows indicate the reversal potential in the Na⁺-lacking buffer. Right, the traces show block of the 100 μM kainate current (holding potential, -80 mV) by 5 μM JSTX. Recovery from block was accelerated by depolarizing the membrane to +60 mV, consistent with a voltage-dependent mechanism of block. The data from CG neurons showing inwardly rectifying (IR) and linear to outwardly rectifying (OR) responses to kainate are summarized in C. The rectification index was calculated as the ratio of the kainate-evoked currents at +40 mV to that at -60 mV. Calcium permeability was determined according to the constant-field equation (eqn (1) in Methods), without correction for a -1.9 mV liquid junction potential. JSTX block was measured as the ratio of the kainate-evoked current in the absence of JSTX (I_con) and following steady-state block by JSTX (I_JSTX). Asterisks indicate statistically significant differences (P < 0.001).

Nucleus magnocellularis neurons

As illustrated in Fig. 5A, NM neurons showed inwardly rectifying responses to kainate (RI, 0.49 ± 0.04; n = 12). Moreover, when the bathing medium was switched to the Na⁺-free, high-Ca²⁺ solution, the cell passed significant inward calcium current. The P_Ca/P_Cs ratio for the NM neuron shown in Fig. 5A was 1.7 (mean, 1.33 ± 0.07; n = 15). As expected of neurons expressing inwardly rectifying, calcium-permeable AMPA receptors, the kainate-evoked current response of this NM neuron was almost competely blocked by 5 μM JSTX (mean, 92.5 ± 1.2 %; n = 9). Thus, in NM neurons almost all of the kainate-evoked inward current appears to flow through JSTX-sensitive, Ca²⁺-permeable AMPA receptors.

Figure 5. Electrophysiology of NM and NIX/X neurons.

Left, the traces in A and B are from NM and NIX/X neurons, respectively, and show responses to 100 μM kainate at various holding potentials between -80 and +60 mV. Middle, the current-voltage relationship for these two neurons is shown in normal high-Na⁺-containing buffer (•) and one in which the Na⁺ was replaced with the impermeant ion N-methylglucamine and the Ca²⁺ concentration was raised to 10 mM (^). The arrows indicate the reversal potential in the Na⁺-lacking buffer. Right, the traces show block of the 100 μM kainate current (holding potential, -80 mV) by 5 μM JSTX. Recovery from block was accelerated by depolarizing the membrane to +60 mV, consistent with a voltage-dependent mechanism of block. Data from NM and NIX/X neurons are summarized in C. The rectification index was calculated as the ratio of the kainate-evoked currents at +40 mV to that at -60 mV. Calcium permeability was determined according to the constant-field equation (eqn (1) in Methods), without correction for a -1.9 mV liquid junction potential. JSTX block was measured as the ratio of the kainate-evoked current in the absence of JSTX (I_con) and following steady-state block by JSTX (I_JSTX). Asterisks indicate statistically significant differences (P < 0.001).

Glossopharyngeal/vagal motor neurons

As shown in Fig. 5B, the functional responses of NIX/X neurons were similar to those of the outwardly rectifying population of CG cells illustrated in Fig. 4B. These neurons showed linear to outwardly rectifying responses to kainate (RI, 0.86 ± 0.4; n = 15), were relatively impermeable to calcium (P_Ca/P_Cs, 0.31 ± 0.02; n = 10) and showed minimal block by 5 μM JSTX (13.2 ± 0.3 %, n = 9). Asterisks in Fig. 5C indicate differences between NM and NIX/X neurons that were statistically significant (P < 0.001).

Thus, the AMPA receptors on most auditory neurons exhibit inwardly rectifying I–V relationships, high calcium permeability, and essentially complete block of kainate-evoked currents by Joro spider toxin. In contrast, the motor neurons have outwardly rectifying I–V curves, low calcium permeability and very low sensitivity to block by JSTX.

Auditory neurons express mainly the flop isoforms of GluR2-4

Since alternative splicing strongly affects the kinetics of glutamate receptors, we examined the relative amounts of the flip and flop isoforms of GluR1-4 in the cell poplulations under study using isoform-specific restriction endonucleases. As shown in Fig. 6, the flip/flop relative abundance patterns for the three auditory structures shared many similarities for GluR2-4 and differed from the pattern seen in the glossopharyngeal/vagal motor nucleus for GluR2-4. No such pattern was seen for GluR1. In the CG, NM and NL, GluR2 flop transcripts accounted for 67–79 % of the total, whereas in the NIX/X the mean (±s.e.m.) level was 47.2 ± 3.7 %. Analysis of variance showed a very highly significant main effect of tissue source on GluR2 flop transcript level (F_3,20 = 12.9, P < 0.0001) and posthoc comparisons revealed significant differences between NIX/X and both NM (P = 0.0001) and NL (P = 0.0016) and a marginally significant difference between NIX/X and CG (P = 0.0151). There were no reliable differences among the three auditory structures. In the CG, NM and NL, GluR3 flop transcripts accounted for 55–70 % of the total, whereas in the NIX/X the mean (±s.e.m.) level was 32.3 ± 3.2 %. Analysis of variance showed a very highly significant main effect of tissue source on GluR3 flop transcript level (F_3,20 = 22.8, P < 0.0001) and posthoc comparisons revealed significant differences between NIX/X and all auditory structures (CG, P = 0.0015; NM, P < 0.0001; and NL, P = 0.0002). There were no reliable differences among the three auditory structures. In the CG, NM and NL, GluR4 flop transcripts accounted for 73–81 % of the total, whereas in the NIX/X the mean (±s.e.m.) level was 42.3 ± 3.9 %. Analysis of variance showed a very highly significant main effect of tissue source on GluR4 flop transcript level (F_3,20 = 30.5, P < 0.0001) and posthoc comparisons revealed significant differences between NIX/X and all auditory structures (CG, P < 0.0001; NM, P < 0.0001; and NL, P < 0.0001). There were no reliable differences among the three auditory structures.

Figure 6. Relative abundance of flip and flop isoforms of GluR1-4.

Gels and bar graphs showing the relative abundance of the flip and flop splice variants of GluR1-4. In each case, the lane in the gel labelled U indicates the undigested PCR product generated by using subunit-specific primers and the respective cDNAs as described in Methods. Lanes labelled I (flip) and O (flop) contain equal amounts of the respective PCR products digested with splice-specific restriction endonucleases. Arrows indicate the 600 bp mark. Bar graphs showing means (+s.e.m.) for the relative abundance of AMPA receptor isoform message in each of the cell populations are shown below each gel; flip is indicated in grey and flop in black. Note the predominance of flop isoforms of GluR2-4 in the auditory structures versus the motor nucleus.

As illustrated in Fig. 7, GluR4d was expressed mainly as the flip isoform in both auditory and motor structures whereas GluR4c transcripts were expressed mostly in the flop isoform in auditory structures but about equally as flip and flop isoforms in NIX/X. Analysis of variance showed that whereas there was no significant effect of tissue source on GluR4d abundance (P = 0.197), there was a highly significant effect on the relative abundance of GluR4c flop (F_3,12 = 15.0, P < 0.0002). Posthoc comparisons showed that NIX/X levels were reliably lower than those in each of the three auditory structures, none of which differed from any other. Although it is known that homomeric rat GluR4c receptors expressed in Xenopus oocytes respond to kainate with inward currents (Gallo et al. 1992), nothing is known about the influence of any of the C-terminal splice variants on AMPA receptor desensitization kinetics, rectification or calcium permeability, or even whether these isoforms produce functional receptor subunits in CNS cells. The greater relative abundance of these splice variants in the NIX/X, however, suggests that they could contribute to the distinctive functional properties of the motor neurons.

Figure 7. Relative abundance of flip and flop isoforms of GluR4c and 4d.

Gels and bar graphs showing the relative abundance of the flip and flop splice variants of GluR4c and 4d. In each case, the lane labelled U indicates the undigested PCR product generated by using subunit-specific primers and the respective cDNAs as described in Methods. Lanes labelled I (flip) and O (flop) contain equal amounts of the respective PCR products digested with splice-specific restriction endonucleases. Arrows indicate the 600 bp mark. Bar graphs showing means (+s.e.m.) for the relative abundance of flip (grey bars) and flop (black bars) message in each of the cell populations are shown below each gel. Note that the flip isoform of GluR4d predominates in all structures but that about three-quarters of GluRc transcripts in auditory structures (but only about half in the NIX/X) are expressed as the flop variant.

Figure 8 provides, for each structure examined, a summary of the relative abundance of GluR1-4 and the proportions of each subunit expressed in the flip and flop isoforms. As shown in this figure, GluR1 and 2 accounted for about two-thirds of the AMPA receptor transcripts of NIX/X neurons but less than half of the total in the three auditory structures; although the flop isoforms of GluR1 and 2 were the majority in all of the structures examined, they were about 50 % more plentiful in the auditory neurons. More than half of the AMPA receptor transcripts in the auditory neurons were GluR3 and 4 and a substantial majority of these were in the flop isoform; less than half of the GluR4 transcripts had C-terminal splices. In contrast, GluR3 and 4 accounted for only about one-third of the AMPA receptor mRNA in NIX/X neurons, the majority of these transcripts were in the flip isoform, and nearly two-thirds of the GluR4 transcripts had C-terminal splices.

Figure 8. Molecular composition of AMPA receptors on auditory and motor neurons.

Summary figures showing, for each of the structures examined, the proportion of the total AMPA receptor transcripts accounted for by the flip and flop isoforms of GluR1-4 and the C-terminal splice variants of GluR4.

In summary, the results indicate that the AMPA receptors of the first three neuron types in the ascending auditory pathway share a number of structural and functional properties that distinguish them from the AMPA receptors of brainstem motor neurons, viz. a lower relative abundance of GluR2 transcript and much lower levels of GluR2 immunoreactivity, higher relative levels of GluR3 flop and GluR4 flop, lower relative abundance of the C-terminal splice variants GluR4c and 4d, less R/G editing of GluR2 and 3, greater permeability to calcium, predominantly inwardly rectifying I–V relationships, and greater susceptibility to block by Joro spider toxin, in addition to the faster desensitization kinetics reported by others (Raman et al. 1994).

DISCUSSION

Comparison with previous studies

Glossopharyngeal/vagal nucleus

To our knowledge, this is the first quantitative analysis of AMPA receptor structure in the NIX/X or its mammalian homologue, the nucleus ambiguus. It is also thought to be the only characterization of AMPA receptor function in this nucleus other than the finding by Raman et al. (1994) of very slow desensitization kinetics in chick glossopharyngeal neurons. Previous studies have provided some qualitative information about AMPA receptor subunit distribution in the nucleus ambiguus. Ambalavanar et al. (1998) used immunohistochemistry to study AMPA receptor subunit expression in cat nucleus ambiguus. These authors found neurons in the nucleus with dense GluR1 immunoreactivity and moderate and comparable levels of staining with antibodies against GluR2/3 and GluR4. The in situ hybridization data of Sato et al. (1993) showed low levels of GluR1 transcript, high levels of GluR2 and moderate levels of GluR3 and 4 in the nucleus ambiguus of rat; these results agree well with our relative abundance data for GluR1-4 in chick NIX/X.

Cochlear ganglion

Previous studies have reported that excitatory synaptic transmission from hair cell to CG neuron in guinea-pigs is mediated by AMPA receptors (Ruel et al. 1999) with linear current-voltage relationships (Nakagawa et al. 1991) and that these neurons show kainic acid-evoked cobalt uptake (Simmons et al. 1994). Raman et al. (1994) included embryonic chick CG neurons in their survey of AMPA receptor desensitization kinetics using outside-out membrane patch recordings of responses to rapidly applied glutamate. These authors found a mean desensitization time constant of 2.8 ms for CG neurons, comparable to the very low time constants found in NM (1.8 ms) and NL (2.3 ms) neurons and distinctly faster than those in non-auditory neurons. We have found that there are two populations of CG neurons with functionally distinct AMPA receptors and that the more common type shows inward rectification, high divalent cation permeability and high JSTX block.

Niedzielski & Wenthold (1995) used RT-PCR, in situ hybridization and immunohistochemistry to study AMPA receptor subunits in rat cochlear ganglion neurons. These authors found significant and comparable levels of expression of message for GluR2 and 3, with a predominance of GluR2 flip and GluR3 flop isoforms. GluR2 message expression was significantly greater in small cochlear ganglion neurons than in larger ones but all ganglion neurons expressed both message and immunoreactivity for GluR4 and GluR2/3/4c but not GluR1. We also found low GluR1 expression in CG neurons, roughly equivalent levels of expression of mRNA for GluR2-4, and a predominance of the flop isoform of GluR3; unlike these authors, we also found that GluR2 was mainly present in the flop variant. The finding of Niedzielski & Wenthold (1995) that rat cochlear ganglion neurons show distinct populations with respect to GluR2 message level is consistent with our finding of two distinct populations of CG neurons in chick with regard to rectification index, cobalt permeability and GluR2 immunofluorescence. It is possible that these two populations represent Type 1 and Type 2 ganglion neurons, which are well characterized in mammals but have not yet been identified clearly in birds (Fischer et al. 1994). A recent study used immunohistochemistry and Western blotting to characterize expression of AMPA receptor subunits in adult pigeon CG (Reng et al. 1999); these authors found low expression of GluR1 and significant immunoreactivity with antibodies that detect GluR4 and GluR2/3/4c. Since our results with a GluR2-specific antibody showed that GluR2 expression was very low in the CG, the available immunochemical data support the conclusion that AMPA receptors in avian CG neurons are composed mainly of GluR3 and GluR4.

Central auditory nuclei

There is good agreement among previous in situ and immunohistochemistry studies that GluR1 expression in central auditory neurons of rat and barn owl generally is low (Petralia & Wenthold, 1992; Sato et al. 1993; Levin et al. 1997; Caicedo & Eybalin, 1999); this is consistent with the low GluR1 transcript levels observed in the present study. Regarding GluR2, there is good agreement among previous studies that mRNA levels are moderate in central auditory neurons (Sato et al. 1993) but disagreement as to whether GluR2 immunoreactivity is very low (Petralia et al. 1997; Wang et al. 1998) or moderate (Caicedo & Eybalin, 1999); our results show moderate GluR2 transcript levels in NM and NL but very low expression of GluR2 immunoreactivity in tissue sections and immunoblots. For GluR3, there is reasonable agreement among previous studies that message levels are moderate to high in anteroventral cochlear nucleus (AVCN; Sato et al. 1993) and low in medial nucleus of the trapezoid body (MNTB; Sato et al. 1993; Geiger et al. 1995); we found moderate message levels in NM and NL neurons. The cross-reactivity with GluR2 and 4c of the antibody most frequently used to stain for GluR3 complicates detection of this subunit but there is general agreement in previous studies between inferred protein levels for GluR3 and message levels in rat VCN and MNTB and avian NM and NL (Levin et al. 1997). For GluR4, there are wide discrepancies among previous studies regarding message levels in VCN and MNTB and the correspondence of message and immunoreactivity levels in each structure (Petralia et al. 1992; Sato et al. 1993; Geiger et al. 1995; Caicedo & Eybalin, 1999). The largest difference is for the MNTB, where Sato et al. (1993) found low GluR4 message levels and Caicedo & Eybalin (1999) found low GluR4 immunoreactivity but Geiger et al. (1995) found the highest relative abundance of GluR4 message (55 %) of any of the nine cell types they examined. We observed moderate levels of GluR4 message in CG, NM and NL, which correlates well with the immunochemical results of Reng et al. (1999) and Levin et al. (1997) for these neurons. We found that about 80 % of the GluR4 transcripts in the three auditory structures were expressed in the flop isoform, a result that can be compared with the 100 % value reported for MNTB neurons by Geiger et al. (1995).

Structural correlates of AMPA receptor kinetics in avian auditory and motor neurons

Very rapid synaptic currents are thought to be the most important of several morphological and functional specializations used by neurons in the auditory pathway to maintain the relative timing of action potentials as they are passed through consecutive synaptic levels. These fast EPSCs depend upon the very rapid deactivation and desensitization kinetics characteristic of the AMPA receptors of auditory neurons (Trussell, 1999). The AMPA receptors of chick auditory neurons have been shown to exhibit faster desensitization and deactivation kinetics than any non-auditory neurons, the slowest of which were the glossopharyngeal motor neurons (Raman et al. 1994). A goal of the present study was to learn whether the distinct functional properties of AMPA receptors in auditory and non-auditory neurons could be correlated with structural differences in the receptors. The AMPA receptors of the three rapidly desensitizing auditory neuron types studied here have remarkably similar structural profiles that differ in some respects from what might be predicted from previous studies. It has sometimes been assumed that since the AMPA receptors of auditory neurons show the most rapid desensitization kinetics of any neurons yet studied (Raman et al. 1994; Geiger et al. 1995) the receptor must be composed largely of the receptor subunit showing the most rapid desensitization kinetics when expressed homomerically in Xenopus oocytes (Mosbacher et al. 1994), namely GluR4 flop (see Geiger et al. 1995). The present results show clearly that GluR1 message expression is low in three types of auditory neuron and that transcripts for GluR2-4 are expressed at comparable levels, with substantial and statistically significant proportions of GluR2 and 3 mRNA expressed in the flip isoform. Expression of GluR1 and 2 proteins in these avian neurons is quite low and expression of GluR3 and 4 is higher and comparable. It appears that there is good agreement between the relative levels of expression of GluR1, 3 and 4 transcripts and immunoreactivity for these subunits in avian auditory neurons. The AMPA receptors of CG, NM and NL neurons in birds therefore appear to be composed largely of GluR3 and GluR4, with perhaps a slight predominance of the former subunit. Based on the results of the present study, about one-third of the GluR3 subunits are present in the more slowly desensitizing flip isoform, as are about one-fifth of the GluR4 subunits.

R/G editing (conversion of an AGA codon for arginine into a GGA codon for glycine by mRNA editing) at a location just before the flip/flop splicing site leads to faster recovery from desensitization and, in most cases, slower desensitization (Lomeli et al. 1994). We found that GluR4 is edited completely in NIX/X, CG and NM and 90 % in NL. In contrast, transcripts for GluR2 and 3 were completely edited in NIX/X but only about 60–80 % edited in the auditory structures. Homomeric GluR4 flop receptors with complete R/G editing desensitize 43 % faster than the unedited version (Lomeli et al. 1994) so that the complete editing of GluR4 present in auditory neurons (in concert with the substantial relative abundance of GluR4 flop) could be expected to contribute to their unusually rapid desensitization. As shown above, NIX/X motor neurons express a significantly higher proportion of their GluR4 mRNA in the form of the C-terminal splice variants GluR4c and GluR4d than do auditory neurons. It has been reported that rat GluR4c forms functional homomeric receptors when expressed in Xenopus oocytes (Gallo et al. 1992). It is unknown, however, whether the C-terminal isoforms are incorporated into functional receptors in neurons and, if so, what influence they have on functional properties important for neuronal signalling.

Structural correlates of calcium permeability in avian auditory and motor neuron AMPA receptors

The calcium permeability of NM neurons measured in the present study was nearly identical to that determined by Otis et al. (1995) in similar studies on NM, where P_Ca/P_Cs was 1.2 when calcium concentrations were not corrected for activity (versus 1.3 in the present study). All neurons in the NM, and most in the CG, express AMPA receptors with high calcium permeability, inwardly rectifying I–V relationships and high susceptibility to JSTX block, and these auditory neurons (and NL) also express moderate levels of GluR2 message but only very low GluR2 immunoreactivity. In contrast, NIX/X motor neurons, which are strongly GluR2 immunoreactive and express a high relative abundance of GluR2 transcripts, exhibit AMPA receptors with low calcium permeability, outwardly rectifying I–V curves and low sensitivity to JSTX block. Calcium permeability of AMPA receptor channels is thought to depend on the relative abundance of GluR2 in its Q/R-edited form (Jonas et al. 1994; Washburn et al. 1997). We examined the editing of GluR2 transcripts at the Q/R site, since a lack of editing would create calcium-permeable receptors containing GluR2. No significant differences were seen in the editing status of GluR2 in CG, NM, NL or NIX/X, since all were essentially completely edited. The major observed difference between the auditory centres and the motor nucleus was the higher relative abundance of GluR2 message in the latter cells and the very much higher level of GluR2 immunoreactivity. Given the small amount of GluR2 expressed in the auditory neurons and the complexity of the influence of GluR2 on channel function (Washburn et al. 1997), these differences would appear to be sufficient to explain the observed differences in rectification and calcium permeability between motor and auditory neurons. Although there is (as discussed above) generally good correspondence between mRNA levels and immunoreactivity for GluR1, 3 and 4 in avian auditory neurons, in the present study we found a large discrepancy between the relative abundance of GluR2 mRNA in the CG, NM and NL and the generally very low expression of GluR2 immunoreactivity in these cells. This suggests that there is selective post-transcriptional suppression of GluR2 protein expression in auditory neurons comparable to that reported for NMDA receptors in cultured rat PC12 cells (Sucher et al. 1993) and some rat brain GluR2 transcripts expressed in Xenopus oocytes and cell-free reticulocyte lysates (Myers et al. 1999).