GABAergic neurons participate in the brain’s responseto birdsong auditory stimulation

Raphael Pinaud,1,2 Tarciso A. F. Velho,1,2 Jin K. Jeong,1 Liisa A. Tremere,1 Ricardo M. Leao,2,3

Henrique von Gersdorff2 and Claudio V. Mello1

1Neurological Sciences Institute, Oregon Health & Science University, Portland, OR 94006, USA2The Vollum Institute, OHSU, Portland, OR, USA3Department of Physiology, FMRP, University of Sao Paulo, Ribeirao Preto, SP, Brazil

Keywords: auditory, avian, GABA, GAD, NCM, plasticity, songbird, ZENK

Abstract

Birdsong is a learned vocal behaviour that requires intact hearing for its development in juveniles and for its maintenance duringadulthood. However, the functional organization of the brain circuits involved in the perceptual processing of song has remainedobscure. Here we provide evidence that GABAergic mechanisms are an important component of these circuits and participate in theauditory processing of birdsong. We first cloned a zebra finch homologue of the gene encoding the 65-kDa isoform of glutamic aciddecarboxylase (zGAD-65), a specific GABAergic marker, and conducted an expression analysis by in situ hybridization to identifyGABAergic cells and to map their distribution throughout auditory telencephalic areas. The results showed that field L2, thecaudomedial nidopallium (NCM) and the caudomedial mesopallium (CMM) contain a high number of GABAergic cells. Using patch-clamp brain slice recordings, we found abundant GABAergic mIPSCs in NCM. Pharmacological antagonism of mIPSCs inducedlarge EPSC bursts, suggesting that tonic inhibition helps to stabilize NCM against runaway excitation via activation of GABA-Areceptors. Next, using double fluorescence in situ hybridization and double immunocytochemical labelling, we demonstrated thatlarge numbers of GABAergic cells in NCM and CMM show inducible expression of the transcriptional regulator ZENK in response tosong auditory stimulation. These data provide direct evidence that GABAergic neurons in auditory brain regions are activated by songstimulation. Altogether, our results suggest that GABAergic mechanisms participate in auditory processing and perception, and mightcontribute to the memorization of birdsong.

Introduction

The study of learned vocalizations in oscines has contributedextensively to our understanding of the neuronal basis of vocalcommunication and vocal learning. The learning and production ofsong and of certain calls depend on the integrity of a set ofinterconnected telencephalic nuclei known as the song control system(see reviews in Brenowitz, 1997). Both the acquisition of learned songand the maintenance of song structure after learning require auditoryfeedback and, thus, depend on intact hearing (Konishi, 1965; Nordeen& Nordeen, 1992; Leonardo & Konishi, 1999). In addition, theauditory processing of birdsong is required for song perception anddiscrimination, and is essential for behaviours such as territorialdefense and mate selection (Catchpole & Slater, 1995; Kroodsma &Miller, 1996).The caudomedial mesopallium (CMM; formerly known as the

caudomedial hyperstriatum ventrale, or CMHV) and the caudomedialnidopallium (NCM; formerly known as the caudomedial neostriatum;we use here the new avian brain nomenclature described in Reineret al., 2004) have been implicated in the auditory processing,discrimination and perception of birdsong. These structures receivedirect or indirect input from the primary telencephalic auditory zone,

field L, are reciprocally connected and project mainly to other centralauditory areas (Vates et al., 1996). It has been postulated that,altogether, these telencephalic auditory areas comprise the avianhomologue of the mammalian primary auditory cortex. According tothis view, thalamo-recipient field L2 would correspond to the granularlayer while NCM and CMM would correspond to supragranular layersof the auditory cortex (Karten & Shimizu, 1989; Wild et al., 1993;Mello et al., 1998; Reiner et al., 2004). NCM and CMM show evokedelectrophysiological responses to song auditory stimulation that havelonger latency and are more selective towards complex stimuli thanthose in field L (Chew et al., 1995; Sen et al., 2001; Gentner &Margoliash, 2003). In NCM, the electrophysiological responses tosong show a rapid decrease, or habituation, upon repeated presenta-tions of the same song (Chew et al., 1995). This ‘habituation’ is song-specific, as a full response can be reinstated upon presentation of anovel song, suggesting a possible contributing mechanism for auditorydiscrimination of birdsong.Importantly, NCM and CMM show robust induction of the

immediate early gene zenk (we use zenk for the gene and ZENK forthe protein) in response to auditory stimulation with conspecific song(Mello et al., 1992; Mello & Clayton, 1994). Expression of zenk (alsoknown as zif-268, egr-1, NGFI-A, and krox-24) is activity-dependentand is, thus, useful to identify cells that respond to specific stimuli orbehavioural contexts (reviewed in Chaudhuri, 1997). The expressionof zenk has revealed brain areas involved in the response to song

Correspondence: Dr Claudio V. Mello, as above.E-mail: melloc@ohsu.edu

Received 24 March 2004, revised 23 June 2004, accepted 1 July 2004

European Journal of Neuroscience, Vol. 20, pp. 1318–1330, 2004 ª Federation of European Neuroscience Societies

doi:10.1111/j.1460-9568.2004.03585.x

auditory stimulation (including NCM and CMM; reviewed in Mello,2002b) or in the motor control of song production (Jarvis &Nottebohm, 1997; Jarvis & Mello, 2000; Jarvis et al., 2000) inawake, unrestrained animals, without interference with naturalbehaviours (Jarvis et al., 1997). zenk expression analysis has alsobeen invaluable to study the functional organization of song-respon-sive areas. For example, the zenk response in NCM is higher forconspecific song than for other stimuli, and it shows a song-specificdecrease, or habituation, upon repeated presentations of the same song(Mello et al., 1995). In canaries, zenk expression patterns in NCMresulting from stimulation with song syllables have a topographicalorganization based on frequency and tuned to features present innatural vocalizations (Ribeiro et al., 1998). Regional variations of thezenk response in NCM that correlate with acoustic properties of thesong stimulus have also been described in starlings (Gentner et al.,2001). These observations indicate that the zenk response in NCMreflects the acoustic properties of the song stimulus, and that NCMparticipates in song auditory processing and discrimination. zenkexpression in NCM may also be linked with aspects of song auditorymemory. For example, ZENK levels in adult NCM correlate withthe degree to which the bird copied the song used as a stimulus duringthe vocal learning period (Bolhuis et al., 2000, 2001). In addition, theinitiation of zenk inducibility by song in NCM correlates with thebeginning of the sensory acquisition period for song in zebra finches(Jin & Clayton, 1997).

GABAergic mechanisms play a pivotal role in the processing ofauditory stimuli, particularly in species that depend heavily on theauditory system for normal behaviour (Fujita & Konishi, 1991; Yanget al., 1992; Suga et al., 1997; Zheng & Knudsen, 1999; Chen & Jen,2000; Zheng & Knudsen, 2001). The GABAergic system has also been

implicated in mediating some aspects of experience-dependentplasticity in a number of sensory modalities (Zheng & Knudsen,1999; Tremere et al., 2001b; Morales et al., 2002). Here we usedmolecular, cellular and electrophysiological approaches to study thecontribution of GABAergic mechanisms to birdsong auditory process-ing. We found a high incidence of GABAergic cells and a prevalence ofGABAergic synapses in telencephalic auditory regions. Moreover, wefound that a large proportion of GABAergic cells participate in thebrain’s response to song stimulation. We conclude that GABAergicmechanisms participate in the physiology of brain areas that contributeto song auditory processing, perception and possibly song memoryformation.

Materials and methods

Animals

We used 30 adult zebra finches (Taeniopygia guttata; 14 females, 16males). Animals were purchased from a breeder and maintained in ourlocal aviary. Experimental protocols utilized in this study wereapproved by OHSU’s Institutional Animal Care and Use Committeeand are in accordance with NIH guidelines.

zGAD65 cloning

GenBank GAD65 sequences from several species were alignedand a pair of primers (forward: ccttcacatctgaacacagtc; reverse:accagaagcagacgtttgtgtg) designed to amplify a fragment spanningexons 3–9 of the mouse cDNA sequence (Fig. 1A) from a zebra finchbrain cDNA library. This library has been screened successfully for

Fig. 1. zGAD65 cloning. (A) Schematic diagram of the mouse GAD65 cDNA. Numbers indicate exons and arrows indicate the locations of the primers designedto amplify the zebra finch homologue (zGAD65). (B) Alignment of the predicted amino acid sequences of zGAD65 with those of the chicken (gGAD) and mouse(mGAD) homologues. Boxed regions indicate residue identity. (C) Northern blot of total telencephalic RNA from adult zebra finch probed with zGAD65. Arrowsindicate two bands identified by the probe (4.1 and 4.4 Kb in size).

GABAergic response to birdsong in auditory areas 1319

ª 2004 Federation of European Neuroscience Societies, European Journal of Neuroscience, 20, 1318–1330

several genes expressed in the brain of zebra finches (Holzenbergeret al., 1997; Denisenko-Nehrbass et al., 2000). We performed 30 PCRcycles, each consisting of 95 �C for 45 s, 44.2 �C for 1 min and 72 �Cfor 45 s, using Taq polymerase and standard PCR buffer. The resultingband was cloned in Bluescript and its identity was confirmed bysequencing.

Northern blot analysis

Total RNA from the telencephalon of an adult male zebra finch wasextracted according to Chomczynski & Sacchi (1987). Ten microgramsof total RNAwere run on aMOPS–formaldehyde gel and blotted onto anylon filter according to standard protocols (Sambrook et al., 1989).The filter was hybridized to a 33P-labelled antisense riboprobe forzGAD65 using established protocols, as described in Clayton et al.(1988), with modifications (detailed in Mello et al., 1997).

Song stimulation

Birds were taken from the aviary and placed overnight individually insmall soundproof boxes (� 76 · 31 · 28 cm), under a 12-hlight : 12-h dark cycle (lights on at 07.00 h). The day followingisolation, stimulated birds (n ¼ 25) were presented for 30 min withplaybacks of a medley of novel conspecific songs (70 dB mean SPL),while control birds were not stimulated. The stimulation paradigmused is known to induce a robust induction of the zenk gene inauditory areas of the zebra finch brain (Mello et al., 1992; Mello &Clayton, 1994). Unstimulated birds (n ¼ 5) were used to ascertain theeffectiveness of song stimulation and were not included in thequantifications. Birds used for in situ hybridization were killed,immediately following the end of stimulation, while birds used forimmunocytochemistry (ICC) were maintained in the soundproof boxesfor an additional hour and then killed.

Tissue preparation

For in situ hybridization, birds (n ¼ 19 song-stimulated and 3unstimulated) were killed by decapitation and brains were quicklydissected, frozen in a dry-ice and propanol bath in embedding medium(Tissue-Tek; Sakura Finetek, Torrance, CA), and cut parasagittally ona cryostat at 10 lm. For ICC, birds (n ¼ 5 song-stimulated and 3unstimulated) were killed with an overdose of Nembutal and perfusedtranscardially with 20 mL of 0.1 m phosphate buffer (PB; pH ¼ 7.4)followed by 60 mL of a 1% paraformaldehyde and 2% glutaraldehydein PB solution. Brains were then dissected and cryoprotected byimmersion in a 30% sucrose solution at 4 �C until equilibrated. Brainswere then frozen as for in situ hybridization, sectioned parasagitally orcoronally on a cryostat at 16 lm, and thaw-mounted on SuperfrostPlus slides (Fischer Scientific, Pittsburgh, PA, USA). Sections werethen dried and stored at )80 �C until further processing.

Labeling of zGAD65 riboprobes

Bluescript plasmids containing zGAD65 and zenk were isolated frombacteria using Qiagen miniprep kit (Qiagen Inc., Valencia, CA, USA),linearized with the appropriate restriction enzymes and purified usingQiagen PCR purification kit (Qiagen Inc., Valencia, CA, USA).33P-labelled sense and antisense riboprobes were generated andpurified as previously described in detail (Mello et al., 1997), and usedfor Northern blot and in situ hybridization analysis.

Radioactive in situ hybridization

We used a previously described protocol (detailed in Mello et al.,1997). Briefly, glass-mounted sections were fixed in a 3%paraformaldehyde solution for 5 min, rinsed in 0.1 m PBS anddehydrated in a standard series of alcohols. Sections were thenacetylated for 10 min in a solution containing 1.35% triethanolam-ine and 0.25% acetic anhydride in water, rinsed three times with2 · SSPE (sodium phosphate, sodium chloride and EDTA), dehy-drated in the alcohol series and air-dried. After addition ofhybridization solution containing the zGAD65 probe (16 lL persection of solution containing formamide 50%, 2 · SSPE, 2 lg ⁄ lLtRNA, 1 lg ⁄ lL BSA, 1 lg ⁄ lL poly A in DEPC-treated water and5 · 105 cpm of labelled probe), sections were coverslipped, sealedby immersion in mineral oil and incubated overnight at 65 �C.Sections were then rinsed twice in chloroform, decoverslipped in2 · SSPE and washed by incubating sequentially for 1 h at roomtemperature (RT) in 2 · SSPE, 1.5 h at 65 �C in 2 · SSPEcontaining 50% formamide, and twice for 30 min at 65 �C in0.1 · SSPE. Sections were then dehydrated in a series of alcoholsand analysed by phosphorimager autoradiography and ⁄ or processedfor emulsion autoradiography, followed by Cresyl Violet counter-staining. Radioactive in situ hybridization was used to investigatethe overall brain distribution of zGAD65 at low-power magnifica-tion (phosphorimager autoradiograms) and for an assessment of celltypes based on cytoarchitectonics (emulsion combined with Nisslcounterstaining).

Fluorescence in situ hybridization (FISH)

Sense and antisense zGAD65 probes were labelled using digoxigenin(DIG)-labelled uridine triphosphate (UTP) (as in Mello et al., 1997),adding a DIG nucleotide labelling mix (Roche Diagnostics Corp.) toour standard probe labelling buffer. The resulting products werepurified in G-50 columns and evaluated for size and yield by agarosegel electrophoresis. One microlitre of the final purified reaction (addedto the 16 lL of hybridization solution) was used per section.Hybridization and washing steps were as above for radioactivein situ hybridization, except that after the final wash sections wereprocessed for detection of the DIG-labelled probes using tyramidesignal amplification (TSA). Briefly, sections were incubated in 0.3%hydrogen peroxide in TNT buffer (0.1 m Tris-HCl, pH ¼ 7.2–7.4;5 m NaCl, and 0.05% Triton-X 100 in DEPC-treated water) for10 min to inactivate endogenous peroxidase. Sections were thenwashed in TNT buffer (3 · 5 min) and incubated for 30 min in ablocking solution (TNB) consisting of TNT buffer containing2 mg ⁄mL of bovine serum albumin. Next, sections were incubatedfor 2 h at RTwith a peroxidase-conjugated anti-DIG antibody (1 : 100dilution in TNB buffer; Roche Diagnostics Corp., Indianapolis, IN,USA), followed by washes (3 · 5 min) in TNT buffer. The sectionswere then incubated for 2 h in a solution containing tyramideconjugated with a fluorophore (Alexa 488 or Alexa 594; MolecularProbes, Eugene, OR, USA). The dilution used was 1 : 100 inamplification buffer provided by the manufacturer. Sections were thenwashed (3 · 5 min) in TNT buffer, counterstained in Hoechst, washedin TNT (3 · 5 min) and coverslipped with aquamount. SlowFadeLight Antifade kit (Molecular Probes, Eugene, OR) was used insections analysed by confocal microscopy. To control for thespecificity of the label, the same procedure was performed withomission of the anti-DIG antibody. FISH was used for a highresolution definition of cellular morphology in the analysis ofzGAD65 expression.

1320 R. Pinaud et al.

ª 2004 Federation of European Neuroscience Societies, European Journal of Neuroscience, 20, 1318–1330

Double-fluorescence in situ hybridization (dFISH)

dFISH was used for double-labelling analysis of zGAD65 and zenkexpression. First, sections were hybridized simultaneously with theantisense probes for zGAD65 (DIG-labelled) and for zenk (biotin-labelled), as described above for FISH. After hybridization and washes,sections were developed for zGAD65 labelling, as described above forFISH, using a peroxidase-conjugated anti-DIG antibody followed byincubation for 2 h in TNTcontaining tyramide coupled with Alexa 594(1 : 500 dilution in TNT). Sections were then incubated in 0.3%hydrogen peroxide for 10 min to inactivate any remaining peroxidase.The biotinylated-zenk probe was then detected by first incubating thesections for 2 h with a peroxidase-conjugated antibiotin antibody(Vector Laboratories, Burlingame, CA, USA; 1 : 500 dilution in TNT).Sections were then incubated for 2 h in Alexa 488-coupled tyramide(1 : 500 diluted in TNT), washed (3 · 5 min) in TNT, counterstainedwith Hoechst and washed further (3 · 5 min) in TNT. Sections were

then coverslipped with antifade solution. We also performed the entireprocedure above with biotin-labelled zGAD65 and DIG-labelled zenkprobes, and with the reverse combination of tyramide reagents. Tocontrol for the effectiveness of the peroxidase inactivation betweenlabels, we performed additional dFISH reactions omitting the anti-DIGantibody, in which case signal could only be observed in the appropriatefilter. We also ran additional controls omitting the antibiotin to controlfor the specificity of the biotin labelling.

Double-label immunocytochemistry (dICC)

Sections were removed from the )80 �C freezer and air-dried at RT for30 min. Slides were then immersed in PB for 30 min, followed byincubation for 2 h at RT in a blocking buffer (BB) that consisted of 0.5%albumin and 0.3% Triton X-100 in 0.1 m PB. Next, sections wereincubated overnight at 4 �Cwith a rabbit anti-egr-1 polyclonal antibody

Fig. 2. Radioactive in situ hybridization of zGAD65 in the zebra finch brain. (A) Schematics of a medial parasagittal section depicting regions within the auditorytelencephalon. (B) Phosphorimager autoradiogram of a section at the level represented in A, hybridized with zGAD65 antisense riboprobe. Inset depictsautoradiogram of sense probe hybridization. (C) Darkfield view of emulsion-dipped section (same level as in A) after hybridization with zGAD65, depictinglabelled cells in NCM, field L2 and CMM. (D–F) Brightfield views of emulsion-dipped sections showing zGAD65-positive cells in auditory areas: (D), NCM;(E), field L2; (F), CMM; arrowheads indicate cells with high expression levels of zGAD65. (G) Detailed view of emulsion-dipped section showing differentialaccumulation of silver grains over cells in NCM; arrowhead depicts a high-expression zGAD65-positive cell, arrow indicates a lower-expression zGAD65-labelledcell. (H) View of emulsion-dipped section showing zGAD65 expression in cerebellar Purkinje cells; ml, molecular layer; pcl, Purkinje cell layer; gcl, granule celllayer. (I) Detailed view of zGAD65-positive neurons in the pallidum. Scale bars, 4 mm (A,B), 100 lm (C), 50 lm (D-F), 10 lm (G), 25 lm (H,I).

GABAergic response to birdsong in auditory areas 1321

ª 2004 Federation of European Neuroscience Societies, European Journal of Neuroscience, 20, 1318–1330

(Santa Cruz Biotechnology, Santa Cruz, CA, USA) that recognizesZENK protein in songbird brain tissue (Mello & Ribeiro, 1998), in ahumidified chamber, followed by a 2-h incubation with a goat antirabbitbiotinylated antibody (Vector; 1 : 200 dilution in BB) at RT. Sectionswere then incubated for 2 h at RTwith ABC (Vector). Finally, tissue wasincubated in a filtered solution containing 0.03% diaminobenzidine(DAB), 0.15% Nickel sulphate and 0.001% hydrogen peroxide in PB.After 8–15 min, the reaction was stopped by immersion in PB. Each ofthe steps above was separated bywashes (3 · 10 min each) in PB. After

signal was developed in the DAB incubation step, sections were washed(3 · 10 min) in PB and incubated in a 0.3% hydrogen peroxide solutionfor 20 min to inactivate any peroxidase activity. Sections were thenwashed in PB, followed by another series of incubations using the anti-GABA antibody (Chemicon International, Temecula, CA; 1 : 200dilution in BB) essentially as detailed above for ZENK ICC, except thatnickel sulphate was not added to the DAB solution. The issue ofspecificity of the GABA antibody for songbird brain sections has beenaddressed in Grisham & Arnold (1994). The use of fluorescence ICC

Fig. 4. Spontaneous miniature postsynaptic events are GABAergic and provide tonic inhibition to NCM neurons. (A) Whole-cell patch-clamp recordings ofspontaneous events in an NCM neuron of a slice perfused with normal aCSF + TTX (top trace) and after the addition of bicuculline (bottom trace). Holding potential,)70 mV. (B) Expanded view of a representative mIPSC recorded in an NCM neuron. The red line shows a single exponential fit of the decay time, demonstratingthat these spontaneous miniature events are GABAergic. (C) Whole-cell patch-clamp recordings depicting examples of bicuculline-evoked postsynaptic currents inNCM neurons. The section was perfused with TTX-free aCSF; recordings from two representative neurons are shown.

Fig. 3. Fluorescence in situ hybridization (FISH) for zGAD65. (A) Low-power photomicrograph of parasagittal section depicts zGAD65 expression in NCMusing DIG labelling. (B) Same field as in A, with superimposed image depicting Hoechst counterstaining for cell nuclei; the high cell density area in the upper rightcorresponds to the ventricular zone. (C) High-power photomicrograph showing detailed view of zGAD65-positive cells in NCM; arrowheads indicate examples ofstrongly labelled cells with large soma size, arrows indicate cells with less intense labelling and smaller soma size. Scale bars, 50 lm (A,B), 25 lm (C).

1322 R. Pinaud et al.

ª 2004 Federation of European Neuroscience Societies, European Journal of Neuroscience, 20, 1318–1330

was not possible due to the high tissue auto-fluorescence when using theglutaraldehyde perfusion required for GABA fixation. Sections werethen dehydrated in a standard series of alcohols, delipidized in xyleneand coverslipped with Krystalon (EM Science, Gibbstown, NJ, USA).

Cell countsWe used a Nikon Eclipse E-600 microscope equipped with a motorizedstage drive and the Lucivid system, and conducted the mapping analysiswithNeurolucida software (Microbrightfield Inc., Colchester,VT,USA)

Fig. 5. dFISH for zGAD65 and zenk reveals song-responsive inhibitory neurons in NCM. (A) Top two rows, representative examples of neurons that coexpresszGAD65 and zenk mRNAs in NCM of zebra finches stimulated with conspecific song. Third row, representative example of zGAD65-positive cells that do notexpress zenk; section is the same as in top rows. Fourth row, representative example of a zenk-positive cell that is negative for zGAD65 from the same section as inupper rows. All images were obtained with confocal microscopy. (B) Schematics depicting the areas in rostral and caudal NCM that were mapped for double-labelled cells (black boxes on left image; not to scale). These maps are generated in Neurolucida software and allow for quantification and analysis of topographicaldistribution of zenk (red circles)- and zGAD65 (green crosses)-positive cells. Each mapped field is 100 · 100 lm in size. Scale bar, 5 lm (A).

GABAergic response to birdsong in auditory areas 1323

ª 2004 Federation of European Neuroscience Societies, European Journal of Neuroscience, 20, 1318–1330

installed in a PC. To estimate local densitites of labelled cells, a regulargrid of 100 · 100-lmsquares was superimposed on sections containingNCM after processing for FISH, dFISH or Nissl staining. For FISH anddFISH, five such fields over rostral and caudal NCM each were countedper bird for cells labelled for zGAD65, zenk, or both mRNAs. Hoechstwas used as a counterstaining method to identify and count all neuronswithin the same fields. In Hoechst-stained sections viewed under theproper fluorescence filter, neuronal cells typically exhibit lightly,nonhomogeneously stained nuclei, often with prominent nucleoli,whereas glial cells usually have strongly, homogeneously stained nuclei;individual cells, even within clusters, can be readily identified. Therelative density of zGAD65-labelled cells was calculated for each bird asthe percentage of such cells relative to the neuronal cell estimates basedonHoechst. The average results across fields were averaged across birds(n ¼ 3 birds per staining method). For dFISH, zenk-positive only ordouble-labelled cells were mapped and counted as above by alternatingthe appropriate fluorescence filters in each grid area examined. Therelative density of double-labelled cells was estimated as the percentageof such cells relative to the number of zenk-labelled cells in each field.Average results across fields were then averaged across birds. Ourestimates of neuronal density in NCM using the Hoechst and Nisslmethods yielded very similar results. Because all our estimates were ofrelative percentages of cells over sampled fields rather than total numberof cells within each class, we did not use stereological methods (asdiscussed in Coggeshall & Lekan, 1996; Saper, 1996). The dICCmaterial was not amenable to quantification because a large number ofcells displayed weak nuclear or cytoplasmic staining and thus double-labelled cells within the low intensity range of staining could not bereliably identified. In addition, cells with very high GABA content wereoften saturated in our preparations, thus possibly occluding the ZENKICC signal.

Imaging and photomicrographs

Photomicrographs were acquired with a Digital camera (DVC, Austin,TX, USA) coupled to the microscope. We used Neurolucida softwareto acquire digital images and Adobe Photoshop to assemble photo-micrograph plates.

Whole-cell patch-clamp electrophysiology

Animals were decapitated after overnight isolation in a soundproofbox. Brains were quickly dissected and placed in artificial cerebro-spinal fluid (aCSF) solution for 10–20 min at 4 �C, and 300-lm-thickparasagittal sections (0.2 mm lateral) containing NCM and adjacent

areas (field L and CMM) were obtained on a vibratome (LeicaMicrosystems, Heerbrugg, Switzerland). We also prepared slices thatcontained NCM only by dissecting this structure from the surroundingtissue. Whole-cell patch-clamp recordings were performed in normalaCSF at RT. aCSF solution consisted of (in mm): NaCl, 125; KCl, 2.5;NaHCO3, 25; NaH2PO4, 1.25; glucose, 25; CaCl2, 2; MgCl2, 1;ascorbic acid, 0.4; sodium pyruvate, 2; myo-inositol, 3; pH 7.4 whenbubbled with 95%CO2 ⁄ 5%O2. Soft, thin-walled glass pipettes (WPI,Sarasota, FL, USA) were pulled using a Narishige vertical puller(Narishige, East Meadow, NY, USA). Resistance of the pipettes in thebath was in the 3–7 MW range. The internal solution consisted of(in mm): Cs-methanesulphonate, 90; CsCl, 20; MgCl2, 1; Na2-phophocreatine, 5; HEPES, 40; TEA-Cl, 10; EGTA, 0.2; ATP-Mg, 2;GTP, 0.2; pH ¼ 7.3 with CsOH, 310 mOSM. To isolate spontaneousGABAergic postsynaptic currents, slices were perfused with 0.5 lmtetrodotoxin (TTX) and glutamate receptor antagonists NBQX (5 lm)and dL-AP5 (50 lm). The holding potential was )70 or )80 mV.Bicuculline concentration used was 50 lm. Data were collected usinga double EPC-9 patch-clamp amplifier and HEKA software (HEKA,Germany). Signals were sampled between 10 and 50 lsec and filteredat 2.5 kHz. All off-line analysis including the determination of thedecay kinetics for the GABAergic miniature inhibitory postsynapticcurrents (mIPSCs) was conducted with IgorPro software (Wavemet-rics, Lake Oswego, OR, USA).

Results

Isolation of zGAD65

To clone the zebra finch homologue of the 65-kDa isoform of glutamicacid decarboxylase, we identified conserved domains in the gene byaligning the human, mouse and chicken sequences, and designed PCRprimers to amplify a fragment of � 700 bp, extending from exon 3 toexon 9 of mouse GAD65 (Fig. 1A). The predicted amino acidsequence of the resulting fragment spanned residues 276–507 of themouse gene (nucleotides 824–1522; GenBank accession no.BC018380; Fig. 1B). The homologies with GAD65 of other specieswere 85% (mouse) and 93% (chicken) at the nucleotide level, and 95%(mouse) and 98% (chicken) at the amino acid level (Fig. 1B). Thehomologies with GAD67, the gene encoding the 67-kDa GADisoform, were lower (respectively 76% and 78% at the amino acidlevel with mouse and chicken GAD67). We named our clone zGAD65(GenBank accession no. AY364313).Northern blot analysis of total RNA from zebra finch telencephalon

with an antisense zGAD65 riboprobe revealed a doublet (Fig. 1C),whereas no band was detected with the sense probe (not shown).

Fig. 7. Schematic circuit diagrams indicating putative roles of inhibitory cells in NCM. (A) Models illustrating how zenk expression in excitatory and inhibitoryneurons in NCM may be determined by the excitatory drive from external inputs. In this and all other panels, excitatory neurons are indicated in grey, inhibitoryneurons are indicated in red, zenk-expressing or -activated neurons are indicated by a black dot and the NCM border is indicated by a grey horizontal line.(B) Models illustrating how zenk expression in excitatory and inhibitory neurons in NCM may be determined by inhibitory drive from external inputs to NCM.(C) Model illustrating how functional domains within NCMmay be sharpened by GABAergic cells. (D) Proposed mechanism for experience-dependent plasticityin NCM mediated by inhibitory neurons. Under normal conditions (top panel), GABAergic interneurons would participate in local inhibitory processes withinactivated domains (shadowed region). Under conditions of reorganizational pressure (bottom panel), GABAergic cells would mediate disinhibition of domains thatare not directly activated by auditory input (expanded shadowed region). The latter domains would become functionally coupled to the activated domain. Theinduction of plasticity-related activity-dependent genes (asterisk) might result in long-lasting circuit reorganization.

Fig. 6. Double immunocytochemistry reveals co-localization of ZENK and GABA in NCM of song-stimulated birds. Shown are representative examples ofdifferent cell types from the same histological section. (A) Cell double-labelled for ZENK (black nuclear staining) and GABA (brown cytoplasmic staining).Arrows depict GABAergic cells that do not express ZENK (cytoplasmic staining only). (B) Large GABAergic cell that does not express ZENK (cytoplasmicstaining only). (C) Another example of a GABAergic cell (cytoplasmic staining) that expresses ZENK (nuclear staining). (D) Example of non-GABAergic ZENK-positive cell (nuclear staining only); arrow depicts a small GABAergic cell that does not express ZENK. Scale bar, 25 lm.

1324 R. Pinaud et al.

ª 2004 Federation of European Neuroscience Societies, European Journal of Neuroscience, 20, 1318–1330

Fig. 6.

Fig. 7.

GABAergic response to birdsong in auditory areas 1325

ª 2004 Federation of European Neuroscience Societies, European Journal of Neuroscience, 20, 1318–1330

Although the significance of the doublet is unclear, the bands (4.1 and4.4 Kb) are close to the expected size based on the chicken and mousesequences. It seems unlikely that our probe cross-reacts with theGAD67 isoform, because that transcript is � 2Kb shorter in mice(Ahman et al., 1996).

zGAD65 expression

In situ hybridizationwith an antisense riboprobe revealed the anatomicaldistribution of zGAD65 expression in zebra finches (Fig. 2). In general,the results were in close accordance with the known distributions ofGAD65 in mammals and nonsongbird avian species (Veenman &Reiner, 1994; Bolam et al., 2000). For instance, zGAD65 expressionwas high in the striatum (Fig. 2B) within the basal ganglia. Our probealso labelled well-known GABAergic cell populations such as cerebel-lar Purkinje cells (Fig. 2H) and large neurons in the globus pallidus(Fig. 2I). High zGAD65 expression was also found throughout pallialregions of the telencephalon (Fig. 2B), including the caudomedialauditory areas field L2 (where expression was particularly high), NCMand CMM. The present study focuses on the latter three areas; a moredetailed analysis will be presented elsewhere. Importantly, no signal wasdetected with a sense probe (Fig. 2B, inset).Emulsion autoradiography analysis revealed numerous zGAD65-

expressing cells in NCM, field L2 and CMM (Fig. 2C–F). Based on aqualitative assessment (density of emulsion grains per cell), there appearto exist two types of zGAD65-positive neurons: cells with very high(Fig. 2G; arrowhead) and cells with moderate-to-low (Fig. 2G; arrow)zGAD65 expression. The distribution of these two kinds of cells differsacross areas (Fig. 2D–F), with a higher density of the high expressioncell type occurring in field L2 (Fig. 2E, arrowheads), and the low-to-moderate expression type being most numerous in NCM (Fig. 2D). Theexpression level of zGAD65 in the high-expression cell type (Fig. 2G,arrowhead) is similar to that in Purkinje and pallidal cells (Fig. 2H and I).No apparent qualitative differences in zGAD65 expression in telence-phalic auditory areas were observed when comparing males withfemales or song-stimulated with -unstimulated birds (not shown).We also used FISH with DIG-labelled probes to study the

expression of zGAD65 in field L2, NCM and CMM. Due to highspatial resolution, this method (with Hoechst counterstaining fornuclei) facilitates the identification of individual labelled neurons,particularly within cell clusters (Fig. 3B). Numerous zGAD65-positiveneurons were seen in NCM (Fig. 3A), field L2 and CMM (not shown),whereas no signal was observed when the anti-DIG antibody wasomitted. Our estimates of the density of GABAergic cells, expressedas the percentage of zGAD65-labelled cells relative to the totalnumber of neurons per area (means ± SEM, n ¼ 3 birds), were39.7 ± 0.2 and 38.2 ± 1.4 in caudal and rostral NCM, respectively,35.8 ± 1.1 in field L2 and 43.3 ± 0.9 in CMM, demonstrating the highincidence of GABAergic cells in telencephalic auditory areas. Whilemost cells had similar expression levels of zGAD65 based on aqualitative assessment of FISH sections, a small number of cells werevery strongly labelled (Fig. 3C; compare arrows and arrowheads).Cells of this latter type tended to be large and were more abundant infield L2 (not shown) than in NCM and CMM.

NCM contains active GABAergic synapses

To investigate the presence of active GABAergic synapses, we usedwhole-cell patch-clamp electrophysiology in slices, focusing on NCM(n ¼ 12 cells from seven birds). To isolate spontaneous GABAergicpostsynaptic currents, slices were perfused with TTX and glutamate

receptor antagonists. Under these conditions, we observed a number ofminiature inhibitory postsynaptic events in all recorded cells (Fig. 4A,top panel). The current magnitude of these events ranged between 8and 40 pA. Furthermore, these GABAergic currents reverted around)39 mV, which is in line with the predicted reversal potential forchloride. Importantly, application of bicuculline methiodide (BMI), aGABA-A receptor antagonist, blocked the occurrence of all miniatureevents (Fig. 4A, bottom panel), indicating the existence of activeGABAergic transmission in NCM mediated by GABA-A receptors.The kinetics of these mIPSCs were further analysed and presented anaverage decay time constant where s¼ 6.921 ms (n ¼ 4 animals, 20mIPSCs per cell) (Fig. 4B). This time course agrees with previouslydocumented time courses for GABA-A mediated mIPSCs in otherpreparations, in which the decay time constant can range from 4 to11 ms (Salin & Prince, 1996; Hollrigel & Soltesz, 1997; Smith et al.,2000). In TTX-free preparations, we also observed, upon BMIapplication, the appearance of bursts of large and asynchronouspostsynaptic currents of putative excitatory origin (examples inFig. 4C). These bursts were observed in slices containing NCM andareas that provide input to NCM (field L and CMM), as well as inslices containing NCM only (not shown). Finally, the postsynapticcurrents observed in the presence of BMI displayed a reversalpotential of 0 mV, which suggests that these events are excitatory innature. Altogether, our findings indicate that excitatory NCM neuronsare tonically inhibited by local GABAergic networks.

Inhibitory neurons in auditory processing areas aresong-responsive

To test whether some of the numerous GABAergic neurons in NCMand CMM might directly respond to birdsong stimulation, we usedsong-induced zenk expression to reveal song-responsive neurons inthese areas (field L2 was not included because it does not show song-induced zenk expression). We developed a dFISH procedure to detectzenk and zGAD65 mRNAs in the same brain sections from birdsstimulated with conspecific birdsong. We found that several zGAD65-positive cells are also labelled for zenk mRNA and, thus, are song-responsive (Fig. 5A, top two rows). Quantification of dFISH revealedthat 42.2% ± 2.2 (mean ± SEM, n ¼ 3 song-stimulated birds) ofzenk-expressing cells in NCM were GABAergic, the remaining 57.8%presumably consisting of excitatory neurons. From the overall zenk-expressing population in CMM, 33.8% ± 1.0 coexpressed zGAD65(mean ± SEM, n ¼ 3 song-stimulated birds). We also found aconsiderable number of zenk-negative GABAergic cells (Fig. 5A,third row), representing either GABAergic cells that are not song-responsive or cells that respond to acoustic features that were absent inthe song stimuli used. The three cell types above were foundthroughout NCM and CMM, without any obvious topographicaldistribution (Fig. 5B). As expected, zenk-expressing cells in unstim-ulated control birds were very few or absent and were not counted.To independently test whether GABAergic neurons in NCM are

song-responsive, we also developed a double-labelling ICC proce-dure for GABA and for ZENK protein. This procedure, due to thedifferential subcellular localization of the antigens, revealed double-labelled cells with a dark-blue nuclear staining (ZENK) and a browncytoplasmic staining (GABA). Several GABAergic neurons in NCM(Fig. 6A and C) and CMM (not shown) of song-stimulated birdsexpressed ZENK protein in response to song auditory stimulation,confirming the existence of song-responsive GABAergic cells inthese areas. In addition, several ZENK protein-expressing cells werenegative for GABA and most probably represent excitatory song-

1326 R. Pinaud et al.

ª 2004 Federation of European Neuroscience Societies, European Journal of Neuroscience, 20, 1318–1330

responsive neurons (Fig. 6D). Finally, several GABAergic cells didnot express ZENK protein [Fig. 6A (arrows) and B]. In unstimulatedbirds, ZENK-labelled cells were very few or absent and thus werenot counted. These findings provide direct evidence that GABAergiccells participate in the physiological response to song auditorystimulation.

Discussion

We have demonstrated a high incidence of GABAergic cells and aprevalence of active GABAergic synapses in auditory areas of thezebra finch telencephalon, and that a large proportion of these cellsrespond to birdsong stimulation. We discuss below the relevance ofour findings with respect to birdsong auditory processing andexperience-dependent plasticity.

zGAD65 cloning

To identify and characterize GABAergic cells that might be involvedin the auditory processing of birdsong, we cloned a songbirdhomologue of GAD65 and generated probes for in situ hybridization.The two known glutamic acid decarboxylase isoforms, GAD65 andGAD67, are encoded by distinct genes and synthesize GABAefficiently. However, whereas GAD67 appears to preferentiallysynthesize cytoplasmic GABA for cellular respiration, GAD65 seemsto catalyse GABA synthesis for vesicular release (Soghomonian &Martin, 1998). In addition, the two genes show differential expressionacross brain regions (Soghomonian & Martin, 1998) that correlateswith electrophysiological properties: while GAD67 is mainlyexpressed in neurons that fire tonically, GAD65 appears topreferentially locate to neurons whose activation relies on synapticinput (Soghomonian & Martin, 1998). Furthermore, while GAD65 ismostly targeted to the cellular membrane and nerve endings, GAD67has a broader intracellular distribution (Ruppert et al., 1993;Soghomonian & Martin, 1998). We focused on GAD65 to revealGABAergic cells that are more probably involved in the synapticresponse to birdsong. Our evidence demonstrates that we have isolatedthe zebra finch GAD65 homologue and that our probe identifiesGABAergic cells.

zGAD65 in telencephalic auditory processing areas

The caudomedial telencephalon is a bulging lobule that contains thethalamorecipient zone field L and the adjacent NCM and CMM, all ofwhich are part of the central auditory pathway, and display robustelectrophysiological and inducible gene responses to auditorystimulation with birdsong (Mello, 2002a; Mello, 2002b). We foundthat GABAergic cells are a major constituent of these areas. Given thatzGAD65 is more tightly linked with the synaptic pool of GABA, ourconservative estimate may represent a more accurate assessment ofGABAergic neurons directly involved in the synaptic response tobirdsong.

We observed heterogeneity in the size and relative labelling ofGABAergic cells, suggesting the existence of two subpopulations withdifferential distribution. It is possible that the smaller and larger cellscorrespond to local interneurons and projection neurons, respectively.This hypothesis is suggested by previous studies of song controlnucleus area X, where the fewer, large, GABAergic cells areprojection neurons that constitute the basal ganglia–thalamic projec-tion of the anterior forebrain pathway required for song learning (Luo

& Perkel, 1999). Testing this hypothesis will require carefulquantitative analysis of zGAD65 expression.It is also interesting that the overall proportion of GABAergic

neurons we observed in auditory pallial areas is high in comparisonwith primary sensory cortical areas in mammals, where the estimatesrange from 25 to 30% of neuronal cells for both visual andsomatosensory cortices (Jones, 1993). In mammals, such cells arehighly concentrated in the thalamorecipient cortical layer 4, as well asin layers 2 and 3 (Gabbott & Somogyi, 1986). It has been argued thatthe avian brain contains neuronal populations that are equivalent tothose that form cortical circuits in mammals. In addition, basic aspectsof their circuitry organization, such as thalamo-telencephalic recipro-city, are preserved. However, this organization is nuclear rather thanlayered (Karten & Shimizu, 1989). If NCM, CMM and field L are thefunctional equivalents of supragranular and granular layers of themammalian auditory cortex (Vates et al., 1996; Mello et al., 1998),then sampling of inhibitory neurons in these nuclei may over-representthe general inhibitory influence in auditory areas.

Active GABAergic synapses are prevalent in NCM

Our patch-clamp recordings showed that active GABAergic synapsesare prevalent in NCM slices, indicating a prominent role ofGABAergic transmission in this region. The high frequency of mIPSCoccurrence in NCM neurons suggests that these cells may be undertonic inhibition, possibly as a means to prevent runaway excitation.Consistent with this hypothesis, the pharmacological blockade ofGABA-A receptors led to the firing of bursts of postsynaptic currentsin NCM cells. Because inhibitory synapses could be detected in bothslices containing NCM, CMM and field L, as well as in morerestrictive slices containing only NCM, the tonic inhibition of NCM islikely to be exerted by local inhibitory interneurons. It is also possible,though, that at least some GABAergic neurons in field L provideinhibitory input to target areas such as NCM and CMM. Theseputative inhibitory projection neurons would, upon their activation,shut down GABAergic interneurons within NCM, relieving otherNCM neurons from tonic inhibition. Consistent with this model, anumber of NCM cells fire action potentials spontaneously in slices(not shown). In experiments where NCM was physically isolated, westill observed large spontaneous postsynaptic currents that correspon-ded to action potential-like events. Overall, our electrophysiologicalexperiments implicate GABAergic cells in the normal physiology ofauditory processing areas of the songbird brain.

Inhibitory neurons in auditory processing areas aresong-responsive

The use of zenk expression analysis by in situ hybridization or ICC toidentify song-responsive cells within auditory structures is well-documented (Mello et al., 1992; Mello & Ribeiro, 1998). zenkexpression in auditory and vocal control pathways depends on theprevious activation of neuronal cells by birdsong stimulation orsinging behaviour (Jarvis & Nottebohm, 1997; Mello & Ribeiro,1998). The expression of zenk in a given neuronal cell can be taken asclear indication of that neuron’s activation by the stimulus presented(Chaudhuri, 1997; Mello, 2002a; Mello, 2002b). Our double-labellingdata demonstrate that GABAergic cells in NCM and CMM expresszenk mRNA and protein in response to birdsong auditory stimulation,and that these cells represent a large proportion of the overall zenk-expressing cell population. Thus, GABAergic cells in telencephalic

GABAergic response to birdsong in auditory areas 1327

ª 2004 Federation of European Neuroscience Societies, European Journal of Neuroscience, 20, 1318–1330

auditory regions are activated by birdsong stimulation and thusparticipate in the auditory processing of song. Conversely, the severalnon-GABAergic zenk-expressing cells are most probably excitatorysong-responsive neurons. Although caution is required in interpretingzenk-negative cells (for example, some specific regions do not showsong-induced zenk expression (see Mello & Clayton, 1995) andneurons in NCM and CMM may exhibit habituation to repeated songstimulation (Chew et al., 1995), zenk-negative GABAergic and non-GABAergic cells most probably were not activated by the stimulation.Several potential scenarios could account for the zenk expression

patterns observed (see models in Fig. 7A and B). For example,excitatory inputs activated by song could act directly on NCMneurons. Alternatively, song could act through a release of the localinhibitory tone. This possibility requires song activation of localinhibitory interneurons within NCM microcircuitry. Our present datashow that GABAergic cells are indeed activated by song stimulation,and could be involved in local disinhibition. NCM anatomicallyoccupies a position that corresponds to supragranular layers of themammalian auditory cortex. In our models (Fig. 7A), the excitatoryinput to NCM would, thus, be equivalent to drive to the supragranularlayers of the mammalian sensory cortex from either granular orsupragranular layers (the latter corresponding to cortico-corticalinteractions). Alternatively, NCM could receive inhibitory inputs(Fig. 7B) that might then release tonic inhibition, by acting on localGABAergic interneurons, within NCM (Fig. 7B). This type ofinhibitory interaction would be similar to some forms of lateralinhibitory connections described in the mammalian sensory cortex.Overall, the concepts diagrammed in our models could apply, inprinciple, to any area containing inhibitory interneurons, regardless ofa layered (mammals) vs. nonlayered (birds, reptiles) organization.The neurochemical identity of zenk-expressing cells had not been

previously studied. Our data demonstrates that zenk is induced bybirdsong in GABAergic and non-GABAergic cells. Thus, the sameactivity-dependent gene can be induced by song in cells that playopposite roles in functional circuitry organization. Given that zenkencodes a transcriptional regulator (Christy & Nathans, 1989),identifying ZENK protein targets will be essential for establishingthe significance of zenk induction. It will be important, however, toconsider neuronal subtypes separately, as zenk targets and theirregulation (up- or down-) may differ in excitatory and inhibitory cells.In vitro studies have shown that ZENK regulates neuronal genes suchas synapsins I and II (Thiel et al., 1994; Petersohn et al., 1995). Wehave preliminary evidence that synapsin II is regulated by birdsongstimulation in NCM (T.A.F. Velho, R. Pinaud & C.V. Mello,unpublished observations), but whether this induction is zenk-dependent or occurs in different neuronal types remains to bedetermined.

Roles for inhibition in auditory processing, plasticity and learning

GABAergic inhibition plays a significant role in shaping the responseproperties of sensory neurons. For instance, antagonism of the GABA-A receptor leads to expansions of the classical receptive fields of cellsin the visual and somatosensory cortices (Sillito, 1977; Kyriazi et al.,1998; Tremere et al., 2001a). In the auditory system, the sharpness andspecificity of receptive fields are also dependent on GABAergicmechanisms. A clear example is the demonstration that the GABA-Areceptor is involved in the sharpening of auditory tuning curves ofneurons that participate in echolocation in the mustached bat (Yanget al., 1992; Suga et al., 1997; Chen & Jen, 2000). GABA-A receptorshave also been implicated in the formation of exquisitely fine auditory

spatial maps and in the generation of interaural time differenceselectivity in mesencephalic neurons in the barn owl (Fujita &Konishi, 1991; Zheng & Knudsen, 1999, 2001). In the chickentelencephalon, local application of bicuculline to regions postsynapticto the thalamorecipient field L2 enlarges isointensity-response areas(Muller & Scheich, 1988). By analogy, it is possible that GABAergiccells also mediate functional sharpening of receptive field properties inNCM neurons (see model in Fig. 7C).It has also been proposed that inhibition is pivotal in generating

conditions that enable experience-dependent plasticity in a number ofsensory systems (Gierdalski et al., 2001; Tremere et al., 2001b;Meredith et al., 2003). By inference, GABAergic mechanisms mightalso play a role in plasticity-related phenomena in a song-processingarea such as NCM. For example, the electrophysiological responses ofNCM neurons decrease (habituate) in consequence to repeated songpresentations (Chew et al., 1995). This habituation is rapid (withinseconds) and could be mediated by a fast-acting mechanism involvingGABA and ⁄ or noradrenaline (the latter suggested by Ribeiro & Mello,2000).GABAergic mechanisms have also been implicated in activity-

dependent functional reorganization of brain networks throughdisinhibition. Under conditions of reorganizational pressure (e.g. digitamputation, retinal lesions), GABAergic interneurons within activatedbrain domains may promote, through lateral inhibition, the shuttingdown of inhibitory interneurons within nonactivated domains(Tremere et al., 2001b; Tremere et al., 2003). Local GABAergicneurons may also mediate the functional properties and some forms ofplastic changes in NCM. Preliminary evidence suggests that the NCMof canaries undergoes reorganization of its frequency-dependentrepresentation of song syllables after prolonged exposure to fre-quency-shifted songs (S. Ribeiro and C.V. Mello, unpublishedobservations). Auditory input from birdsong stimulation wouldnormally activate only NCM cells that are anatomically positionedto receive that information (Fig. 7D, shadowed area in top panel).Under conditions of reorganizational pressure (e.g. frequency-shiftedsongs), GABAergic interneurons within the activated domain wouldinhibit GABAergic neurons within nonactivated domains. Cells innonactivated domains would, thus, be activated in the absence ofdirect auditory input through disinhibition, becoming functionallycoupled to the activated domain (Fig. 7D, extended shadowed area inbottom panel). Long-lasting physical changes in neuronal circuitorganization might also ensue, mediated by the induction of activity-dependent genes associated with phenomena such as synapticpotentiation, neurite sprouting and changes in neuronal excitability(Fig. 7D, asterisk in bottom panel). In mammals, this type of inhibition(GABAergic to GABAergic cell) is prominent in primary sensorycortical areas, and a probable anatomical substrate has been demon-strated (Winfield et al., 1981; McGuire et al., 1991). Experimentaltesting of this model in songbirds should provide significant insightsinto the functional role of the GABAergic system in the perceptionand memorization of song.The involvement of GABAergic transmission in some forms of

learning has been previously demonstrated in birds. For example, themarked plastic changes in receptive field properties of auditoryneurons that occur in the barn owl upon manipulations of the visualenvironment depend on an intact GABAergic system (Zheng &Knudsen, 1999; Knudsen et al., 2000). In addition, our data areconsistent with the previous finding that an activity-dependent gene,namely c-fos, can be induced in activated GABAergic cells in thecontext of visual imprinting in chicks (Ambalavanar et al., 1999). Thatphenomenon was observed in the intermediate medial mesopallium(IMM, formerly known as IMHV), an area required for visual

1328 R. Pinaud et al.

ª 2004 Federation of European Neuroscience Societies, European Journal of Neuroscience, 20, 1318–1330

imprinting. The induction of c-fos in the IMM has been associatedwith learning in this paradigm, rather than with sensory processing orlocomotor activity per se (McCabe & Horn, 1994). Therefore, c-fosactivation in GABAergic cells indicates that inhibitory neurons play aprominent role in visual imprinting (Ambalavanar et al., 1999; for arecent review see Horn, 2004). One of the auditory areas in the presentstudy, the CMM, is located in the same telencephalic subdivision asthe IMM and is involved in the auditory processing of vocalizations invarious avian species (Bonke et al., 1979; Scheich et al., 1979; Muller& Leppelsack, 1985; Capsius & Leppelsack, 1999; Sen et al., 2001;Gentner & Margoliash, 2003). Interestingly, significant changes insong selectivity occur in CMM neurons as a result of exposure to songin the context of song recognition learning (Gentner & Margoliash,2003). Thus, it is possible that the song-responsive GABAergicneurons we identified in the CMM based on zenk induction play a rolein neuronal selectivity changes associated with song perceptuallearning.

Together with our current findings, the data discussed aboveindicate a prevalent participation of GABAergic neurons in theprocessing of behaviourally relevant sensory stimuli in birds(Ambalavanar et al., 1999; McCabe et al., 2001). A similarinvolvement may also occur in mammals, where GABAergic neuronsplay a major role in cortical physiology. Further studies in mammals,birds and other vertebrates should help determine whether theinvolvement of GABAergic mechanisms in brain function, as itrelates to behaviour, has been evolutionarily conserved.

Acknowledgements

We would like to thank AeSoon Bensen for technical support and RowlandTaylor for allowing us to use equipment in his lab. Raphael Pinaud is an N.L.Tartar Research Fellow. This work was supported by NIH grant DC02853 andUnited States Public Health Service, National Center for Research Resources,Grant RR016858.

Abbreviations

aCSF, artificial cerebrospinal fluid; CMM, caudomedial mesopallium; DAB,diaminobenzidine; dFISH, double-fluorescence in situ hybridization; dICC,double-label immunocytochemistry; DIG, digoxigenin; FISH, fluorescencein situ hybridization; ICC, immunocytochemistry; mIPSC, miniature inhibitorypostsynaptic current; NCM, caudomedial nidopallium; PB, phosphate buffer;RT, room temperature; SSPE, sodium phosphate, sodium chloride and EDTA;TTX, tetrodotoxin.

References

Ahman, A.K., Wagberg, F. & Mattsson, M.O. (1996) Two glutamatedecarboxylase forms corresponding to the mammalian GAD65 andGAD67 are expressed during development of the chick telencephalon. Eur.J. Neurosci., 8, 2111–2117.

Ambalavanar, R., McCabe, B.J., Potter, K.N. & Horn, G. (1999) Learningrelated fos-like immunoreactivity in the chick brain: time-course andco-localization with GABA and parvalbumin. Neuroscience, 93, 1515–1524.

Bolam, J.P., Hanley, J.J., Booth, P.A. & Bevan, M.D. (2000) Synapticorganisation of the basal ganglia. J. Anat., 196, 527–542.

Bolhuis, J.J., Hetebrij, E., Den Boer-Visser, A.M., De Groot, J.H. & Zijlstra,G.G. (2001) Localized immediate early gene expression related to thestrength of song learning in socially reared zebra finches. Eur. J. Neurosci.,13, 2165–2170.

Bolhuis, J.J., Zijlstra, G.G., den Boer-Visser, A.M. & Van Der Zee, E.A. (2000)Localized neuronal activation in the zebra finch brain is related to thestrength of song learning. Proc. Natl Acad. Sci. USA, 97, 2282–2285.

Bonke, B.A., Bonke, D. & Scheich, H. (1979) Connectivity of the auditoryforebrain nuclei in the guinea fowl (Numida meleagris). Cell Tissue Res.,200, 101–121.

Brenowitz, E.A. (1997) Comparative approaches to the avian song system.J. Neurobiol., 33, 517–531.

Capsius, B. & Leppelsack, H.-J. (1999) Response patterns and their relation-ship to frequency analysis in auditory forebrain centers of a songbird. HearRes., 136, 91–99.

Catchpole, C.K. & Slater, P.J.B. (1995) Bird Song: Biological Themes andVariations. Cambridge University Press, Cambridge, UK.

Chaudhuri, A. (1997) Neural activity mapping with inducible transcriptionfactors. [Review.] Neuroreport, 8, iii–vii.

Chen, Q.C. & Jen, P.H. (2000) Bicuculline application affects dischargepatterns, rate-intensity functions, and frequency tuning characteristics of batauditory cortical neurons. Hear Res., 150, 161–174.

Chew, S.J., Mello, C., Nottebohm, F., Jarvis, E. & Vicario, D.S. (1995)Decrements in auditory responses to a repeated conspecific song are long-lasting and require two periods of protein synthesis in the songbird forebrain.Proc. Natl Acad. Sci. USA, 92, 3406–3410.

Chomczynski, P. & Sacchi, N. (1987) Single-step method of RNA isolation byacid guanidinium thiocyanate-phenol-chloroform extraction. Anal. Biochem.,162, 156–159.

Christy, B. & Nathans, D. (1989) DNA binding site of the growth factor-inducible protein Zif268. Proc. Natl Acad. Sci. USA, 86, 8737–8741.

Clayton, D.F., Huecas, M.E., Sinclair-Thompson, E.Y., Nastiuk, K.L. &Nottebohm, F. (1988) Probes for rare mRNAs reveal distributed cell subsetsin canary brain. Neuron, 1, 249–261.

Coggeshall, R.E. & Lekan, H.A. (1996) Methods for determining numbers ofcells and synapses: a case for more uniform standards of review. J. Comp.Neurol., 364, 6–15.

Denisenko-Nehrbass, N.I., Jarvis, E., Scharff, C., Nottebohm, F. & Mello, C.V.(2000) Site-specific retinoic acid production in the brain of adult songbirds.Neuron, 27, 359–370.

Fujita, I. & Konishi, M. (1991) The role of GABAergic inhibition in processingof interaural time difference in the owl’s auditory system. J. Neurosci., 11,722–739.

Gabbott, P.L. & Somogyi, P. (1986) Quantitative distribution of GABA-immunoreactive neurons in the visual cortex (area 17) of the cat. Exp. BrainRes., 61, 323–331.

Gentner, T.Q., Hulse, S.H., Duffy, D. & Ball, G.F. (2001) Response biases inauditory forebrain regions of female songbirds following exposure tosexually relevant variation in male song. J. Neurobiol., 46, 48–58.

Gentner, T.Q. & Margoliash, D. (2003) Neuronal populations and single cellsrepresenting learned auditory objects. Nature, 424, 669–674.

Gierdalski, M., Jablonska, B., Siucinska, E., Lech, M., Skibinska, A. & Kossut,M. (2001) Rapid regulation of GAD67 mRNA and protein level in corticalneurons after sensory learning. Cereb. Cortex, 11, 806–815.

Grisham, W. & Arnold, A.P. (1994) Distribution of GABA-like immunor-eactivity in the song system of the zebra finch. Brain Res., 651, 115–122.

Hollrigel, G.S. & Soltesz, I. (1997) Slow kinetics of miniature IPSCs duringearly postnatal development in granule cells of the dentate gyrus.J. Neurosci., 17, 5119–5128.

Holzenberger, M., Jarvis, E.D., Chong, C., Grossman, M., Nottebohm, F. &Scharff, C. (1997) Selective expression of insulin-like growth factor II in thesongbird brain. J. Neurosci., 17, 6974–6987.

Horn, G. (2004) Pathways of the past: the imprint of memory. Nature Rev.Neurosci., 5, 108–120.

Jarvis, E.D. & Mello, C.V. (2000) Molecular mapping of brain areas involvedin parrot vocal communication. J. Comp. Neurol., 419, 1–31.

Jarvis, E.D. & Nottebohm, F. (1997) Motor-driven gene expression. Proc. NatlAcad. Sci. USA, 94, 4097–4102.

Jarvis, E.D., Ribeiro, S., da Silva, M.L., Ventura, D., Vielliard, J. & Mello, C.V.(2000) Behaviourally driven gene expression reveals song nuclei inhummingbird brain. Nature, 406, 628–632.

Jarvis, E.D., Schwabl, H., Ribeiro, S. & Mello, C.V. (1997) Brain generegulation by territorial singing behavior in freely ranging songbirds.Neuroreport, 8, 2073–2077.

Jin, H. & Clayton, D.F. (1997) Localized changes in immediate-early generegulation during sensory and motor learning in zebra finches. Neuron, 19,1049–1059.

Jones, E.G. (1993) GABAergic neurons and their role in cortical plasticity inprimates. Cereb. Cortex, 3, 361–372.

Karten, H.J. & Shimizu, T. (1989) The origins of neocortex: connections andlamination as distinct events in evolution. J. Cogn. Neurosci., 1, 291–301.

Knudsen, E.I., Zheng, W. & DeBello, W.M. (2000) Traces of learning in theauditory localization pathway. Proc. Natl Acad. Sci. USA, 97, 11815–11820.

Konishi, M. (1965) The role of auditory feedback in the control of vocalizationin the white-crowned sparrow. Z. Tierpsychol, 22, 770–783.

GABAergic response to birdsong in auditory areas 1329

ª 2004 Federation of European Neuroscience Societies, European Journal of Neuroscience, 20, 1318–1330

Kroodsma, D.E. & Miller, E.H. (1996) Ecology and Evolution of AcousticCommunication in Birds. Cornell University Press, Ithaca, NY.

Kyriazi, H., Carvell, G.E., Brumberg, J.C. & Simons, D.J. (1998) Laminardifferences in bicuculline methiodide’s effects on cortical neurons in the ratwhisker ⁄ barrel system. Somatosens. Mot. Res., 15, 146–156.

Leonardo, A. & Konishi, M. (1999) Decrystallization of adult birdsong byperturbation of auditory feedback. Nature, 399, 466–470.

Luo, M. & Perkel, D.J. (1999) Long-range GABAergic projection in a circuitessential for vocal learning. J. Comp. Neurol., 403, 68–84.

McCabe, B.J. & Horn, G. (1994) Learning-related changes in Fos-likeimmunoreactivity in the chick forebrain after imprinting. Proc. Natl Acad.Sci. USA, 91, 11417–11421.

McCabe, B.J., Horn, G. & Kendrick, K.M. (2001) GABA, taurine and learning:release of amino acids from slices of chick brain following filial imprinting.Neuroscience, 105, 317–324.

McGuire, B.A., Gilbert, C.D., Rivlin, P.K. & Wiesel, T.N. (1991) Targets ofhorizontal connections in macaque primary visual cortex. J. Comp. Neurol.,305, 370–392.

Mello, C. (2002a) Immediate early gene (IEG) expression mapping of vocalcommunication areas in the avian brain. In Kaczmarek, L. & Robertson,H.A. (eds), Immediate Early Genes and Inducible Transcription Factors inMapping of the Central Nervous System Function and Dysfunction, 1st edn.,Elsevier Science B.V., Amsterdam, pp. 59–101.

Mello, C.V. (2002b) Mapping vocal communication pathways in birds withinducible gene expression. J. Comp. Physiol. A Neuroethol. Sens. Neural.Behav. Physiol., 188, 943–959.

Mello, C.V. & Clayton, D.F. (1994) Song-induced ZENK gene expression inauditory pathways of songbird brain and its relation to the song controlsystem. J. Neurosci., 14, 6652–6666.

Mello, C.V. & Clayton, D.F. (1995) Differential induction of the ZENK gene inthe avian forebrain and song control circuit after metrazole-induceddepolarization. J. Neurobiol., 26, 145–161.

Mello, C.V., Jarvis, E.D., Denisenko, N. & Rivas, M. (1997) Isolation of song-regulated genes in the brain of songbirds. Meth. Mol. Biol., 85, 205–217.

Mello, C., Nottebohm, F. & Clayton, D. (1995) Repeated exposure to onesong leads to a rapid and persistent decline in an immediate early gene’sresponse to that song in zebra finch telencephalon. J. Neurosci., 15,6919–6925.

Mello, C.V. & Ribeiro, S. (1998) ZENK protein regulation by song in the brainof songbirds. J. Comp. Neurol., 393, 426–438.

Mello, C.V., Vates, G.E., Okuhata, S. & Nottebohm, F. (1998) Descendingauditory pathways in the adult male zebra finch (Taeniopygia guttata).J. Comp. Neurol., 395, 137–160.

Mello, C.V., Vicario, D.S. & Clayton, D.F. (1992) Song presentation inducesgene expression in the songbird forebrain. Proc. Natl Acad. Sci. USA, 89,6818–6822.

Meredith, R.M., Floyer-Lea, A.M. & Paulsen, O. (2003) Maturation of long-term potentiation induction rules in rodent hippocampus: role of GABAergicinhibition. J. Neurosci., 23, 11142–11146.

Morales, B., Choi, S.Y. & Kirkwood, A. (2002) Dark rearing alters thedevelopment of GABAergic transmission in visual cortex. J. Neurosci., 22,8084–8090.

Muller, C.M. & Leppelsack, H.J. (1985) Feature extraction and tonotopicorganization in the avian auditory forebrain. Exp. Brain Res., 59, 587–599.

Muller, C.M. & Scheich, H. (1988) Contribution of GABAergic inhibition tothe response characteristics of auditory units in the avian forebrain.J. Neurophysiol., 59, 1673–1689.

Nordeen, K.W. & Nordeen, E.J. (1992) Auditory feedback is necessary for themaintenance of stereotyped song in adult zebra finches. Behav. Neural Biol.,57, 58–66.

Petersohn, D., Schoch, S., Brinkmann, D.R. & Thiel, G. (1995) The humansynapsin II gene promoter. Possible role for the transcription factorzif268 ⁄ egr-1, polyoma enhancer activator 3, and AP2. J. Biol. Chem.,270, 24361–24369.

Reiner, A., Perkel, D.J., Bruce, L.L., Butler, A.B., Csillag, A., Kuenzel, W.J.,Medina, L., Paxinos, G., Shimizu, T., Striedter, G.F., Wild, M., Ball, G.F.,Durand, S.E., Gunturkun, O., Lee, D.W., Mello, C.V., Powers, A., White,S.A., Hough, G., Kubikova, L., Smulders, T.V., Wada, K., Dugas-Ford, J.,Husband, S., Yamamoto, K.YuJ., Siang, C. & Jarvis, E.D. (2004) Revised

nomenclature for avian telencephalon and some related brainstem nuclei.J. Comp. Neurol., 473, 377–414.

Ribeiro, S., Cecchi, G.A., Magnasco, M.O. & Mello, C.V. (1998) Toward asong code: evidence for a syllabic representation in the canary brain. Neuron,21, 359–371.

Ribeiro, S. & Mello, C.V. (2000) Gene expression and synaptic plasticity in theauditory forebrain of songbirds. Learn. Mem., 7, 235–243.

Ruppert, C., Sandrasagra, A., Anton, B., Evans, C., Schweitzer, E.S. & Tobin,A.J. (1993) Rat-1 fibroblasts engineered with GAD65 and GAD67 cDNAs inretroviral vectors produce and release GABA. J. Neurochem., 61, 768–771.

Salin, P.A. & Prince, D.A. (1996) Spontaneous GABAA receptor-mediatedinhibitory currents in adult rat somatosensory cortex. J. Neurophysiol., 75,1573–1588.

Sambrook, J., Fritsch, E.F. & Maniatis, T. (1989) Molecular Cloning: aLaboratory Manual, 2nd edn. Cold Spring Harbor Laboratory, Cold SpringHarbor, NY.

Saper, C.B. (1996) Any way you cut it: a new journal policy for the use ofunbiased counting methods. J. Comp. Neurol., 364, 5.

Scheich, H., Bonke, B.A., Bonke, D. & Langner, G. (1979) Functionalorganization of some auditory nuclei in the guinea fowl demonstrated by the2-deoxyglucose technique. Cell Tissue Res., 204, 17–27.

Sen, K., Theunissen, F.E. & Doupe, A.J. (2001) Feature analysis of naturalsounds in the songbird auditory forebrain. J. Neurophysiol., 86, 1445–1458.

Sillito, A.M. (1977) Inhibitory processes underlying the directional specificityof simple, complex and hypercomplex cells in the cat’s visual cortex.J. Physiol. (Lond.), 271, 699–720.

Smith, A.J., Owens, S. & Forsythe, I.D. (2000) Characterisation of inhibitoryand excitatory postsynaptic currents of the rat medial superior olive.J. Physiol. (Lond.), 529, 681–698.

Soghomonian, J.J. & Martin, D.L. (1998) Two isoforms of glutamatedecarboxylase: why? Trends Pharmacol. Sci., 19, 500–505.

Suga, N., Zhang, Y. & Yan, J. (1997) Sharpening of frequency tuning byinhibition in the thalamic auditory nucleus of the mustached bat.J. Neurophysiol., 77, 2098–2114.

Thiel, G., Schoch, S. & Petersohn, D. (1994) Regulation of synapsin I geneexpression by the zinc finger transcription factor zif268 ⁄ egr-1. J. Biol.Chem., 269, 15294–15301.

Tremere, L., Hicks, T.P. & Rasmusson, D.D. (2001a) Expansion of receptivefields in raccoon somatosensory cortex in vivo by GABA (A) receptorantagonism: implications for cortical reorganization. Exp. Brain Res., 136,447–455.

Tremere, L., Hicks, T.P. & Rasmusson, D.D. (2001b) Role of inhibition incortical reorganization of the adult raccoon revealed by microiontophoreticblockade of GABA (A) receptors. J. Neurophysiol., 86, 94–103.

Tremere, L.A., Pinaud, R. & De Weerd, P. (2003) Contributions of inhibitorymechanisms to perceptual completion and cortical reorganization. In Pessoa,L. & De Weerd, P. (eds), Filling-in: From Perceptual Completion to CorticalReorganization. Oxford University Press, Oxford, pp. 295–322.

Vates, G.E., Broome, B.M., Mello, C.V. & Nottebohm, F. (1996) Auditorypathways of caudal telencephalon and their relation to the song system ofadult male zebra finches. J. Comp. Neurol., 366, 613–642.

Veenman, C.L. & Reiner, A. (1994) The distribution of GABA-containingperikarya, fibers, and terminals in the forebrain and midbrain of pigeons,with particular reference to the basal ganglia and its projection targets.J. Comp. Neurol., 339, 209–250.

Wild, J.M., Karten, H.J. & Frost, B.J. (1993) Connections of the auditoryforebrain in the pigeon (Columba livia) J. Comp. Neurol., 337, 32–62.

Winfield, D.A., Brooke, R.N., Sloper, J.J. & Powell, T.P. (1981) A combinedGolgi-electron microscopic study of the synapses made by the proximal axonand recurrent collaterals of a pyramidal cell in the somatic sensory cortex ofthe monkey. Neuroscience, 6, 1217–1230.

Yang, L., Pollak, G.D. & Resler, C. (1992) GABAergic circuits sharpen tuningcurves and modify response properties in the mustache bat inferiorcolliculus. J. Neurophysiol., 68, 1760–1774.

Zheng, W. & Knudsen, E.I. (1999) Functional selection of adaptive auditoryspace map by GABAA-mediated inhibition. Science, 284, 962–965.

Zheng, W. & Knudsen, E.I. (2001) Gabaergic inhibition antagonizes adaptiveadjustment of the owl’s auditory space map during the initial phase ofplasticity. J. Neurosci., 21, 4356–4365.

1330 R. Pinaud et al.

ª 2004 Federation of European Neuroscience Societies, European Journal of Neuroscience, 20, 1318–1330