Contents lists available at ScienceDirect

Behavioural Brain Research

journal homepage: www.elsevier.com/locate/bbr

Research report

Social context-dependent singing alters molecular markers of dopaminergicand glutamatergic signaling in finch basal ganglia Area XLisa Y. Soa,b, Stephanie J. Mungerb, Julie E. Millera,b,c,⁎

a Program in Neuroscience, University of Arizona, Gould-Simpson Building, Tucson, AZ, 85721, United StatesbDepartment of Neuroscience, University of Arizona, Gould-Simpson Building, Tucson, AZ, 85721, United Statesc Department of Speech, Language, and Hearing Sciences, University of Arizona, Speech, Language, and Hearing Sciences Building, Tucson, AZ, 85721, United States

A R T I C L E I N F O

Keywords:DopamineZebra finchSongbirdGlutamateBasal ganglia

A B S T R A C T

Dopamine (DA) is an important neuromodulator of motor control across species. In zebra finches, DA levels varyin song nucleus Area X depending upon social context. DA levels are high and song output is less variable when amale finch sings to a female (female directed, FD) compared to when he is singing by himself (undirected, UD).DA modulates glutamatergic input onto cortico-striatal synapses in Area X via N-methyl-D-aspartate (NMDA) andDA receptor mechanisms, but the relationship to UD vs. FD song output is unclear. Here, we investigate theexpression of molecular markers of dopaminergic and glutamatergic synaptic transmission (tyrosine hydroxylase– TH, alpha-synuclein – α-syn) and plasticity (NMDA 2B receptor – GRIN2B) following singing (UD vs. FD) andnon-singing states to understand the molecular mechanisms driving differences in song output. We identifiedrelationships between protein levels for these biomarkers in Area X based on singing state and the amount ofsong, measured as the number of motifs and time spent singing. UD song amount drove increases in TH, α-syn,and NMDA 2B receptor protein levels. By contrast, the amount of FD song did not alter TH and NMDA 2Breceptor expression. Levels of α-syn showed differential expression patterns based on UD vs. FD song, consistentwith its role in modulating synaptic transmission. We propose a molecular pathway model to explain how socialcontext and amount of song are important drivers of molecular changes required for synaptic transmission andplasticity.

1. Introduction

Dopamine (DA) is an important neuromodulator of motor control,motivation and reward-based behaviors across mammalian and avianspecies. Insight into the role of DA modulation in neural mechanismsfor human vocal motor control has been obtained from studyingsongbirds [1].

In male zebra finches (Taenopygia guttata) [2] and European star-lings (Sturnus vulgaris) [3], levels of DA in the brain vary dependingupon the social context in which the bird sings. When the male zebrafinch sings to a female, known as female-directed (FD) song, DA levelsmeasured via high-performance liquid chromatography (HPLC) arehigher in vocal control region Area X compared to when the male ispracticing his song alone, known as undirected (UD) song [2]. Neuronsfound in Area X, a song-dedicated nucleus in the finch basal ganglia,receive dopaminergic input from the substantia nigra (SN) and ventraltegmental area (VTA) as in mammals (Fig. 1) [4,5]. Pharmacologicalapproaches in adult zebra finches have shown that reduction of pre-

synaptic DA input or antagonism of post-synaptic DA receptors in AreaX can abolish social context-dependent song differences. Injection of theneurotoxin 6-hydroxydopamine (6-OHDA) into Area X depletes DAnerve terminals and results in UD song resembling the more stereotypedFD song [6]. The application of a D1 receptor antagonist into Area Xcauses FD song to show more variable pitch changes similar to that ofUD song by altering neuronal firing patterns [7,8]. In a related species,Bengalese finch (Lonchura striata domestica), treatment with 6-OHDA inArea X interferes with pitch changes as part of reinforcement-drivenvocal plasticity during UD song [9]. The molecular mechanismsthrough which DA modulates these aspects of UD vs. FD adult songbehavior are not well identified.

Based on in vitro slice electrophysiology experiments, activation ofD1-like receptors reduced NMDA and non-NMDA receptor-mediatedexcitatory post-synaptic currents in Area X medium spiny neurons(MSNs) [10]. DA application can also directly increase pallidal (PAL)neuron firing, driving thalamic inhibition and result in a less variablesong output associated with FD song [7]. How changes in the amount of

https://doi.org/10.1016/j.bbr.2018.12.004Received 28 August 2018; Received in revised form 28 November 2018; Accepted 1 December 2018

⁎ Corresponding author at: Department of Neuroscience, University of Arizona, 1040 E. 4th St. GS 423, Tucson, AZ, 85721, United States.E-mail address: juliemiller@email.arizona.edu (J.E. Miller).

Behavioural Brain Research 360 (2019) 103–112

Available online 03 December 20180166-4328/ © 2018 Elsevier B.V. All rights reserved.

T

DA available modify synaptic transmission and plasticity mechanismsat these cortico-striatal synapses in Area X to support UD vs. FD songoutput requires investigation. One strategy is to identify moleculartargets of DA modulation that are differentially expressed based onsocial context-dependent singing.

Social context has an impact on mRNA and protein expression levelsin song nuclei, including Area X. ZENK, an immediate early gene, showsincreased mRNA and protein levels in Area X following 30–45min ofUD but not FD song [11,12]. FoxP2, a speech-related gene and tran-scription factor, shows decreased mRNA expression in Area X followingtwo hours of UD but not FD song, but at the protein level, singing inboth contexts drive protein levels down in comparison to non-singers[13,14]. Intriguingly, FoxP2 mRNA levels significantly decrease withhigher numbers of UD song motifs over a two-hour period [15, Fig. S5].By contrast, there is a weak association between FoxP2 mRNA andamount of FD song [13, Fig. 4], but this relationship is absent at theprotein level [14, Fig. 5]. Social context and the amount of song aretherefore two determinants that can affect FoxP2 expression levels.Both ZENK and FoxP2 are examples of social context-dependent dif-ferences in gene expression but the underlying molecular circuitry isnot well identified.

We propose that molecular mechanisms of synaptic transmissionand plasticity mediate social context-dependent song output associatedwith differences in DA levels in Area X (e.g. low in UD; high in FD). As astarting point, we have identified three potentially important molecularmarkers in Area X. We selected two related pre-synaptic molecularmarkers (tyrosine hydroxylase – TH, alpha-synuclein – α-syn) and onepost-synaptic marker (NMDA 2B receptor – GRIN2B) based on priorwork implicating them in synaptic transmission and plasticity me-chanisms [15].

TH is an enzyme required for DA biosynthesis. TH is found interminals of dopaminergic neurons that project from the SN/VTA toArea X [16,17]. TH is commonly used as a marker to detect changes inDA signal in finch Area X and rat basal ganglia that are correlated toaltered vocalizations including parkinsonian-like output [3,6,18,19]. Asa biomarker of DA synthesis, we hypothesize that total TH protein le-vels in Area X will mimic the low vs. high levels of DA present duringUD vs. FD singing [2] but that TH will also be influenced by how muchthe bird sings.

In mammalian models, the pre-synaptic protein α-syn normallyregulates TH enzymatic activity in dopaminergic terminals, synaptic

vesicle function and DA release into the synaptic cleft, and helps trafficDA active transporter to the membrane surface [20]. In rodents, virally-driven overexpression of α-syn leads to reductions in DA release due todopaminergic terminal loss in the basal ganglia as well as changes invocalizations [21,22]. In zebra finches, α-syn expression is enriched atsynapses in song control nuclei during critical developmental periodsassociated with vocal learning although its role in synaptic plasticityand adult song behavior is unclear [23]. Here, we hypothesize that α-syn contributes to DA signaling in Area X differentially depending onthe social context.

The gene GRIN2B encodes a NMDA 2B glutamate receptor. NMDA2B receptors initiate synaptic plasticity events in cortical and basalganglia brain regions. In a previous study, Miller and colleagues [15,Fig. S6] showed that GRIN2B is a hub gene in a synaptic plasticitynetwork in Area X activated by two hours of UD singing. As the birdsings, GRIN2BmRNA expression increases in Area X and is correlated toa more variable UD song. In adult finches, virally-driven overexpressionof GRIN2B expression in cortical song nucleus LMAN causes FD song torevert back to its juvenile state of increased variability with morestuttering [24]. However, the effects of social context on GRIN2B pro-tein levels in Area X are unknown including the relationship betweenGRIN2B mRNA/protein levels in Area X and FD song.

Using these molecular markers of synaptic transmission (TH, α-syn)and plasticity (GRIN2B), we first determined whether protein levelswere regulated by activity-dependent mechanisms (singing vs. notsinging). Given the large range of protein values detected within thesinging groups, we hypothesized that the amount of song/time spentsinging resulted in differences in protein levels for TH, α-syn, andGRIN2B in Area X. Our findings supported our hypotheses indicating astrong relationship between protein levels and social context-dependentsinging. TH and GRIN2B protein levels in Area X increased with theamount of UD but not FD singing whereas α-syn protein levels in-creased with the amount of UD song but fluctuated depending on thetime spent singing FD song.

2. Materials and methods

2.1. Animal model and experimental design

All animal use was approved by the Institutional Animal Care andUse Committee at the University of Arizona. Non-breeding adult male

Fig. 1. Schematic of the song circuitry. Individual brainregions are denoted by circles/ovals. The anterior forebrainpathway is comprised of a cortico-basal-thalamic-cortico loop(LMAN, Area X, DLM). Area X receives dopaminergic inputfrom SN/VTA and glutamatergic inputs from cortical nucleiHVC and LMAN. DLM – dorsal lateral nucleus of thalamus;HVC – higher vocal center; LMAN – lateral magnocellularnucleus of anterior nidopallium; RA – robust nucleus of ar-copallium; SN – substantia nigra; VP – ventral pallidum; VTA– ventral tegmental area. [4].

L.Y. So et al. Behavioural Brain Research 360 (2019) 103–112

104

zebra finches (n= 42) between 120 and 300 days post-hatch were usedin this study. Male finches were moved to individual sound attenuationchambers and acclimated for three days under a 13.5:10.5 h light:darkcycle during which UD song was continuously recorded using SongAnalysis Pro [25]. After acclimation, birds were collected for one offour behavioral groups starting at lights-on in the morning (7:30 am)and following the time points established by Miller et al. [14]. Twogroups of non-singers were collected at lights-on (0 HR NS) or twohours after lights-on (2 HR NS). For the 2 HR NS condition, an ex-perimenter observed the bird and if the bird attempted to sing, whichhappened infrequently, they tapped on the cage to prevent the birdfrom singing. Our prior publication showed that neither cage tappingnor the presence of the investigator significantly alters serum corti-costerone levels between NS, undirected (UD) or female-directed (FD)behavioral groups [14]. Birds collected as 2 HR NS occasionally com-pleted motifs on an average of 4.2 ± 2.3 motifs within the two-hourperiod. The two singing groups sang either two hours of undirected (2HR UD) or female directed (2 HR FD) song. We excluded birds that sangless than 90 motifs in a two-hour period because previous work showedthis was the minimum number of motifs needed to elicit changes inprotein expression [14]. Any bird that was quiet in the last 30min ofthe two-hour singing period was excluded from collection given that30min of quiet after singing can reduce ZENK expression back tobaseline levels [11]. FD song was elicited by the presentation of femalezebra finches over the two-hour period [14]. For 2 HR FD birds, livevideo streaming and investigator observations were used to ensure thatthe bird sang only FD and not UD song. At the completion of the non-singing or singing assignments, finches were euthanized using anoverdose with isoflurane in order to collect the brain tissue.

2.2. Song analysis

The two hours of singing prior to collection was used for songanalysis. Motifs were identified as repeated sequences of syllables se-parated by silent periods. The numbers of motifs were manuallycounted throughout the two-hour span. Time spent singing was calcu-lated by averaging the motif length of 20 randomly selected motifswithin the two-hour period for each bird and then multiplying by thetotal number of motifs sung in the two hours. This random samplingapproach has been successfully employed in previous publications asbeing representative of the total time spent singing [14,15]. To confirmthe validity of this random sampling approach in the current study, wecompared data obtained using this method to a ‘summed method’ in thesame bird where the length of every motif was measured and addedtogether over the two-hour recording period to obtain time spentsinging. No statistically significant difference was detected between thetwo methods when comparing n= 3 birds per UD and FD groups(paired t-test, p= 0.215).

2.3. Antibody characterization

Tyrosine Hydroxylase (TH)-This antibody for TH (Millipore,#AB152, RRID: AB_390204, Table 1) was used to detect dopaminergicpositive terminals in Area X and has been previously validated by Milleret al. [6], detecting a protein band ∼55 kD in finches.

Alpha-synuclein (α-syn)-This antibody to α-syn (Proteintech,#10842-1-AP, RRID: AB_2192672, Table 1) was chosen due to 86%homology to zebra finch α-syn (∼16 kD, NCBI accession number NP_001041718). To validate this antibody, a preadsorption control wasperformed with the original antigen via Western blot (Fig. 2A).

Glutamate ionotropic receptor NMDA subunit 2B (GRIN2B, NR2B)-This antibody was chosen due to 95% homology to zebra finch GRIN2B(∼180 kD, NCBI accession number XP_002195885). Because the pep-tide was not available to do a preadsorption control, we validated a

commercially available GRIN2B primary antibody made against the N-terminus of the mouse peptide (Millipore, #06-600, RRID: AB_310193,Table 1) and compared it to another commercially available antibodymade against the C-terminus of human GRIN2B (Proteintech, #21920-1-AP, RRID: AB_11232223, Table 1) in Western blots to detect proteinbands at similar molecular weights (Fig. 2B). Mouse basal ganglia(MBG) and finch nidopallium (NP, non-basal ganglia region) were usedas positive controls alongside Area X. Both antibodies detected proteinbands at the predicted molecular weight. For all subsequent blots,Millipore’s GRIN2B antibody was used to detect protein levels.

Glyceraldehyde-3-phosphate dehydrogenase (GAPDH)-The anti-body for GAPDH (Millipore, #MAB374, RRID: AB_2107445, Table 1)was used to normalize for equal protein loading across the lanes andwas previously validated by Miller et al. [14] as detecting a proteinband of ∼35 kD in finch Area X. Over the course of the study, we de-tected a weakening of the antibody signal despite testing different lots.Therefore, we also used an antibody to GAPDH from Proteintech(Proteintech, #10494-1-AP, RRID: AB2263076, Table 1) that also de-tects a protein band at the same molecular weight in Area X and yieldedthe same results.

2.4. Brain sample preparation, western blotting, and quantification

Directly following overdose with isoflurane, brains were dissectedout and flash frozen with liquid nitrogen. Brains were cryosectioneduntil Area X was visualized and bilateral tissue biopsies of Area X wereobtained by punching 1mm deep using a syringe attached to a in-tramedic luer-stub 20-gauge adapter (catalog #427564, BD MedicalTechnology). Post-hoc Nissl staining of 40 μm coronal brain sectionswas performed to validate accuracy of Area X targeting [14]. Tissuebiopsies obtained from Area X were homogenized and isolated for totalprotein using modified RIPA lysis buffer, and concentrations were de-termined on a Bio-Rad RC DC assay as in Miller et al. [14]. 15 μg ofprotein lysate was run at 100 V using a 10% polyacrylamide-SDS gelthen transferred to 0.2 μm PVDF membranes for the Western blottingprocedure as in Miller et al. [14]. Primary antibodies were incubated onseparate portions of the blot to prevent cross-reactivity (refer toFig. 2C). Primary antibodies incubated overnight at 4 °C were TH(Millipore, #AB152, 1:500, RRID: AB_390204), α-syn (Proteintech,#10842-1-AP, 1:100-1:250, RRID: AB_2192672), GRIN2B (Millipore,#06-600, 1:100-1:500, RRID: AB_310193), and GAPDH (Millipore,#MAB374, 1:5000, RRID: AB_2107445; Proteintech, #10494-1-AP,1:250, RRID: AB_2263076) [14,15]. After TBST washes, the blots wereincubated at room temperature for two hours in secondary HRP-con-jugated rabbit antibody (GE Healthcare-Amersham, #NA934, RRID:AB_772206, Table 1) at concentrations 1:2000 (TH, GRIN2B, GAPDH)and 1:1000 (α-syn), and secondary HRP-conjugated mouse antibody(GE Healthcare-Amersham, #NA931, RRID: AB_772210, Table 1) atconcentration 1:1000 (GAPDH). Blots were developed using chemilu-minescence and imaged using a Bio-Rad system. Quantifications wereperformed on Quantity One (Bio-Rad) by an experimenter blind to thebehavioral states. A rectangular band was drawn surrounding the signalof interest (raw value) and the same-sized band was placed just aboveor below the band in the same lane to obtain the “background”. Thecorrected value was calculated by subtracting the background from theraw value. Corrected values were obtained for TH, α-syn, GRIN2B, andGAPDH. Corrected values for TH, α-syn, and GRIN2B were divided bythe corrected GAPDH value in the corresponding lane to get normalizedvalues to control for equal protein loading. GAPDH was used as thenormalization control because it is not behaviorally regulated and therewas no difference in expression of GAPDH between behavioral groups(Kruskal-Wallis, p= 0.34), replicating previous findings [14,15]. Theprotein signals of interest/GAPDH values were then divided by theaverage of the two lanes of the control condition, 0 HR NS, within the

L.Y. So et al. Behavioural Brain Research 360 (2019) 103–112

105

same blot for normalization (set= 1). The 0 HR NS birds are not ex-perimentally manipulated and provide a good control without anysinging. This normalization step enables inter-blot comparisons for atotal of 5 separate blots. Each western blot consisted of two differentbirds per behavioral state. For protein levels from Western blots, therewere 10 birds per behavioral group for TH and α-syn and eight birds perbehavioral group for GRIN2B (Fig. 3). For song analysis, the totalnumbers of birds are as follows: TH and α-syn (n=10 for number ofUD/FD motifs, n= 9 for time spent singing UD, n=8 for time spentsinging FD); GRIN2B (n=8 for number of UD/FD motifs and timespent singing UD, n= 7 for time spent singing FD). There were fewerbirds in the time spent singing analysis group due to experimenter error(missing song files).

2.5. Statistical analysis

For protein level comparison between the four behavioral groups,the means of each group were compared via the Kruskal-Wallis test(VassarStats, vassarstats.net) because the data did not fit a normaldistribution. No data points were deemed outliers in the graphical plotsusing exclusion criteria based on technical issues (e.g. Western blotrunning conditions or imaging issues) [26]. Therefore, all data pointswere determined to be experimentally valid data points and were notexcluded from analysis.

Regression analyses using the OriginLab graphing program was usedto determine if TH, α-syn, and GRIN2B protein levels varied as a

Table 1Primary and secondary antibodies. All primary and secondary antibodies used in the study and relevant information are listed in the table.

Antibody Reference RRID Antigen Concentrations

Primary antibodyAnti-tyrosine hydroxylase antibody (rabbit polyclonal) AB152 (Millipore) AB_390204 Denatured tyrosine hydroxylase from rat

pheochromocytoma (denatured by sodiumdodecyl sulfate)

1:500

Anti-NR2B (GRIN2B) antibody (rabbit polyclonal) 06-600 (Millipore) AB_310193 a.a. 1437-1456 of mature mouse NR2B or1463-1482 of NR2B precursor; signal isa.a. residues 1-26 [KFNGSSNGHVYEKLSSIESDV]. This sequence isidentical to a.a. 1437-1456 of mature ratNR2B and a.a. 1465-1484 of human NR3,containing a N-terminal lysine

1:100-1:500

GRIN2B (NR2B) antibody (rabbit polyclonal) 21920-1-AP(Proteintech)

AB_11232223 human GRIN2B-GST fusion proteinYLDQFRTKENSPHWEHVDLTDIYKERSDD-FKRDSVSGGGPCTNRSHIKHGTGDKHGV-VSGVPAPWEKNLTNVEWEDRSGGNFCRS-CPSKLHNYSTTVTGQNSGRQACIRCEACK-KAGNLYDISEDNSLQELDQPAAPVAVTSN-ASTTKYPQSPTNSKAQKKNRNKLRRQHSY-DTFVDLQKEEAALAPRSVSLKDKGRFMDG-SPYAHMFEMSAGESTFANNKSSVPTAGH-HHHNNPGGGYMLSKSLYPDRVTQNPFIP-TFGDDQCLLHGSKSYFFRQPTVAGASKAR-PDFRALVTNKPVVSALHGAVPARFQKDIC-IGNQSNPCVPNNKNPRAFNGSSNGHVYE-KLSSIESDV (1133-1484 aa encoded byBC113620)

1:500

Alpha-synuclein antibody (rabbit polyclonal) 10842-1-AP(Proteintech)

AB_2192672 human SNCA-GST fusion proteinMDVFMKGLSKAKEGVVAAAEKTKQGVAE-AAGKTKEGVLYVGSKTKEGVVHGVATVA-EKTKEQVTNVGGAVVTGVTAVAQKTVEG-AGSIAAATGFVKKDQLGKNEEGAPQEGIL-EDMPVDPDNEAYEMPSEEGYQDYEPEA (1- 140 aa encoded by BC013293)

1:100-1:250

GAPDH antibody (rabbit polyclonal) 10494-1-AP(Proteintech)

AB_2263076 human GAPDH-GST fusion proteinMGKVKVGVNGFGRIGRLVTRAAFNSGKV-DIVAINDPFIDLNYMVYMFQYDSTHGKFH-GTVKAENGKLVINGNPITIFQERDPSKIKW-GDAGAEYVVESTGVFTTMEKAGAHLQGG-AKRVIISAPSADAPMFVMGVNHEKYDNSL-KIISNASCTTNCLAPLAKVIHDNFGIVEGL-MTTVHAITATQKTVDGPSGKLWRDGRGA-LQNIIPASTGAAKAVGKVIPELNGKLTGM-AFRVPTANVSVVDLTCRLEKPAKYDDIKK-VVKQASEGPLKGILGYTEHQVVSSDFNSD-THSSTFDAGAGIALNDHFVKLISWYDNEF-GYSNRVVDLMAHMASKE (1 - 335 aaencoded by BC004109)

1:250

Anti-glyceraldehyde-3-phosphate dehydrogenase(GAPDH) antibody, clone 6C5 (mouse,monoclonal)

MAB374 (Millipore) AB_2107445 Glyceraldehyde-3-phosphatedehydrogenase from rabbit muscle

1:5000

Secondary antibodyAmersham ECL Rabbit IgG, HRP-linked whole Ab NA934 (GE

Healthcare-Amersham

AB_772206 1:1000-1:2000

Amersham ECL Mouse IgG, HRP-linked whole Ab NA931 (GEHealthcare-Amersham

AB_772210 1:1000

L.Y. So et al. Behavioural Brain Research 360 (2019) 103–112

106

function of the number of motifs or time spent singing with R2 valuesreported as in previous publications [11,13,14]. Scatterplots were madein the OriginLab graphing program, and the optimal fit (either linear orcurvilinear) is shown for each data set [26,27]. R2 values for both linearand curvilinear fit were obtained from Origin and confirmed by IBMSPSS Statistical software package in consultation with our universitystatistician Mark Borgstrom. Each graph was determined to be linear orcurvilinear based on the significant F change value determined by IBMSPSS Statistical software package. If the significant F change value wasp < 0.05 for either the linear or curvilinear model, the correlatingmodel was used. If neither model had a significant F change value ofp < 0.05, the linear model was used. Statistics for the linear and cur-vilinear (quadratic) plots were done using IBM SPSS Statistical softwarepackage where x= number of motifs/time spent singing and y=protein levels. The ANOVA F and p-values are reported based on x2 (x-squared) for the plots.

3. Results

3.1. Antibody validation

The polyclonal antibody against α-syn protein detected bands ofsimilar molecular weights across finch Area X, nidopallium (NP), andmouse basal ganglia (MBG) tissue (Fig. 2A). Incubation of the α-synantibody with 30 times excess of its peptide (Proteintech, #ag1285)resulted in the absence of protein bands (Fig. 2A), confirming specifi-city along with a similar molecular weight band detected in MBG. Todetect and validate GRIN2B protein in finch Area X, we used com-mercially available antibodies raised against either the mouse N-term-inal peptide sequence (Millipore) or the human C-terminal peptide se-quence (Proteintech) (Fig. 2B) and tested them in two different finchbrain regions (Area X, NP) and in MBG tissue. Both antibodies detectedGRIN2B protein bands at similar molecular weights (∼180 kD) in finchand MBG tissue.

3.2. Gene expression of non-singers vs. singers in Area X

In this study, we examined the total protein expression of TH, α-syn,and GRIN2B via Western blots from tissue punches of Area X followingfour behavioral states: 0 HR NS, 2 HR NS, 2 HR UD, and 2 HR FD(Fig. 2C) in which the 0 HR NS was used as the control condition giventhe lack of behavioral manipulation [6,14]. There were no significantdifferences between groups for TH, α-syn, and GRIN2B (Fig. 3, see le-gend for statistics). Like Miller et al. [14], we pursued the sources of therange in protein values observed within groups for the singers.

3.3. Rise in TH levels in Area X coincides with UD song amount but has norelationship to FD song

The number of motifs and time spent singing were plotted againstTH protein levels in Area X (Fig. 4). With UD song, TH levels sig-nificantly increased with the number of motifs (Fig. 4A) and trended inthe same direction with time spent singing (Fig. 4B). However, neitherthe number of motifs (Fig. 4C) nor time spent singing FD song (Fig. 4D)had an effect on TH levels.

3.4. Rise in α-syn levels in Area X coincides with number of UD motifs; α-syn levels follow a U-shaped curve in response to time spent singing FD song

With UD song, α-syn levels in Area X significantly increased withthe number of motifs (Fig. 5A) with an exponentially stronger increasewith a higher number of motifs sung. Time spent UD singing had nosignificant effect on α-syn levels (Fig. 5B). The α-syn levels plottedagainst time spent singing FD song follow a U-shaped curve whereprotein levels decrease in birds that spend<350 s singing, but as thebird exceeds around 350 s, α-syn levels rise (Fig. 5D). However, thenumber of FD motifs (Fig. 5C) had no effect on α-syn levels.

Fig. 2. Validation blots for α-syn and GRIN2B antibodies and a representative western blot with four different behavioral states.A: Western blot with α-syn protein detected at ∼16 kD in Area X, nidopallium (NP), and mouse basal ganglia (MBG). 15 μg of protein loaded per lane. †Preadsorptioncontrol: incubation of the α-syn antibody with 30 times excess of its peptide showed that the bands disappeared, confirming specificity of the antibody to α-syn. B:Western blot with GRIN2B protein detected at ∼180 kD in Area X, NP, and MBG. 15 μg of protein loaded per lane. ‡Antibody 1 for GRIN2B (Millipore, #06-600) and§antibody 2 for GRIN2B (Proteintech, #21920-1-AP) both detect protein bands at similar molecular weights, confirming specificity of the antibody to GRIN2B. C:Representative blot of 15 μg of tissue lysate per lane from bilateral tissue biopsies of Area X. Two birds were run per behavioral group. Blot shows protein levels of TH(∼55 kD, Miller et al. 2015), GRIN2B (∼180 kD), GAPDH (∼35 kD), and α-syn (∼16 kD); finch molecular weights obtained from: www.ncbi.nlm.nih.gov/protein.Protein signals in each lane were normalized to GAPDH, the loading control. Dotted lines denote where the blots were cut prior to antibody incubation to avoid cross-reactivity between antibodies, weak signals, and to minimize antibody usage.

L.Y. So et al. Behavioural Brain Research 360 (2019) 103–112

107

Fig. 3. Behavioral regulation of Area X TH,α-syn, and GRIN2B protein levels.A: Bar graph of the mean TH protein levels bybehavioral condition with standard error bars(n=10 birds per group; 0 HR NS: 1 ± 0.04, 2HR NS: 1.65 ± 0.49, 2 HR UD: 1.19 ± 0.27,2 HR FD: 1.5 ± 0.40). Each point representsan individual bird from five different blots(n=10/group). There was no statisticallysignificant difference between the four groups(Kruskal-Wallis test, p= 0.76). B: Mean α-synprotein levels by behavioral condition withstandard error bars (n= 10 birds per group; 0HR NS: 1 ± 0.06, 2 HR NS: 2.36 ± 0.76, 2HR UD: 1.82 ± 0.65, 2 HR FD: 1.88 ± 0.53).Each point represents an individual bird fromfive different blots (n= 10/group). There wasno statistically significant difference betweenthe four groups (Kruskal-Wallis test, p= 0.24).C: Mean GRIN2B protein levels by behavioralcondition with standard error bars (n=8 birdsper group; 0 HR NS: 1 ± 0.03, 2 HR NS:2.1 ± 0.92, 2 HR UD: 1.58 ± 0.29, 2 HR FD:2.68 ± 1.31). Each point represents an in-dividual bird from four different blots (n=8).There was no statistically significant differencebetween the four groups (Kruskal-Wallis test,p= 0.45).

Fig. 4. Changes in Area X TH protein levelswith UD and FD song.TH protein levels of UD singers and FD singersplotted against the number of motifs sangwithin the two-hour period (A, C) or time spentsinging (B, D). TH protein levels were depen-dent upon the number of motifs sung duringUD (n= 10, ANOVA, F= 18, p=0.002) butnot FD (n=10, ANOVA, F= 0.11, p=0.746).Both time spent singing UD (n=9, ANOVA,F=3.68, p=0.097) and FD (n= 8, ANOVA,F=0.02, p=0.887) did not have a significanteffect on TH levels. Each point represents anindividual bird. Gray lines denote curvilinearregression curve for A and linear regressioncurves for B, C, and D drawn as best fits to thedata based on the significant F change value.

L.Y. So et al. Behavioural Brain Research 360 (2019) 103–112

108

3.5. GRIN2B levels in Area X increase with the amount of UD song but notFD song

With UD song, GRIN2B levels significantly increased with thenumber of motifs (Fig. 6A) and time spent singing (Fig. 6B). Comparedto UD song, GRIN2B levels with FD song showed a trend for decreasedprotein expression with increased time spent singing (Fig. 6D), butthese trends were not statistically significant. The number of FD motifs(Fig. 6C) had no effect on GRIN2B levels.

4. Discussion

Here, we characterized the effects of social context and how muchthe bird sang on the expression of molecular markers of synaptictransmission and plasticity in zebra finch Area X. Our Western blotanalyses did not show differences in the mean protein levels of TH, α-syn, and GRIN2B between social context but revealed strong differencesdependent on the social context when comparing the relationship be-tween protein levels and how much the bird sang (amount of song andtime spent singing). The lack of overall mean differences in proteinlevels between UD vs. FD singers has also been reported for the tran-scription factor FoxP2; furthermore, the number of motifs sung duringtwo hours of UD but not FD behavior alters FoxP2 protein expression[14]. In our study, levels of TH, α-syn, and GRIN2B in Area X increasedwith more UD singing but did not show these same patterns with FDsong. With more FD song, TH levels did not change, α-syn levels fluc-tuated depending on how much the bird sang, and GRIN2B levelstrended downwards.

4.1. Social context and amount of song

TH protein levels in Area X increased with more UD song but

remained unchanged with increased FD song. Heimovics and Riters [3]looked at FD song for a shorter time period, 30–45min, and also did notfind changes in TH levels using immunocytochemistry with how muchtime the non-breeding male birds sang FD [3]. In our current study,birds that sang more UD song motifs had higher TH levels. Sasaki et al.[2] found there was no relationship between DA levels and number ofUD or FD song motifs. However, our data had a broader range of singersover a longer time course (2 h vs. 30min) dedicated to one state onlywhereas Sasaki et al. [2] alternated 30min of UD with FD song. Fur-thermore, we took a different methodological approach to Sasaki et al.[2] by measuring changes in TH, the enzyme for DA biosynthesis, usingWestern blotting following the song collection period vs. in vivo mi-crodialysis.

As a pre-synaptic protein involved in vesicle-mediated release ofneurotransmitter, fluctuating levels of α-syn modulate DA, glutamate,and norepinephrine release [28–30]. Given that singing amount in bothsocial contexts drove α-syn levels, fluctuating levels of α-syn may play amodulatory role in synaptic transmission and plasticity as the birdsings.

Previously, GRIN2B mRNA levels were shown to increase with UDsong and with the amount of song; however, the authors did not look atthe relationship with FD song [15]. Our results show that GRIN2Bprotein levels increase with more UD song but tends to decrease withFD song suggesting that synaptic plasticity mechanisms may be differ-entially activated based on social context.

4.2. Molecular pathway model for UD song

Based on our findings in the current study, genetic and electro-physiological evidence from published work [2,7,10–12], we propose amolecular pathway model for circuit-level changes that occur over atwo-hour period to drive differences in UD vs. FD song output. A model

Fig. 5. Changes in Area X α-syn protein le-vels with UD and FD song.α-syn protein levels of UD singers and FDsingers plotted against the number of motifssang within the two-hour period (A, C) or timespent singing (B, D). α-syn protein levels weredependent upon the number of motifs sungduring UD (n= 10, ANOVA, F= 16.81,p= 0.002) but not FD (n=10, ANOVA,F=0.038, p= 0.85). α-syn levels were de-pendent upon time spent singing FD (n= 8,ANOVA, F=6.29, p= 0.043) but not UD(n=9, ANOVA, F= 1.41, p= 0.274). Eachpoint represents an individual bird. Gray linesdenote curvilinear regression curves for A andD and linear regression curves for B and Cdrawn as best fits to the data based on thesignificant F change value.

L.Y. So et al. Behavioural Brain Research 360 (2019) 103–112

109

for the UD song circuit is depicted in Fig. 7A–B highlighting the roles ofArea X, DLM, and LMAN in driving more acoustic variability during UDsong [31]. Acoustic variability has been defined as greater pitchchanges from rendition to rendition of UD song compared to FD song[31,32]. Hilliard, Miller et al. [15] previously showed birds that sangthe most number of motifs of UD song in a two-hour period had greateracoustic variability measured as the change in mean Wiener entropy.Dopaminergic and glutamatergic input converge on medium spinyneurons (MSNs), the site of cortico-striatal synaptic plasticity. After30–45min of UD song (black pathway, Fig. 7A), low DA levels exist inArea X [2]. MSNs inhibit pallidal (PAL) neurons so that DLM providesexcitatory drive to song nucleus LMAN. LMAN drives song nucleus RAto promote more variability in pitch, etc. in the bird’s song, as part ofhis trial and error learning [33]. Based on findings from our study, wepropose the blue molecular pathway (Fig. 7B) that becomes activatedwith increased UD singing. As the bird continues to sing over the two-hour period (blue pathway), α-syn facilitates increased glutamate re-lease onto MSNs requiring up-regulation of NMDA 2B (GRIN2B) re-ceptors. With increased NMDA 2B receptor levels, MSNs excitabilityand inhibition over PAL neurons increase to support more UD songrendition-to-rendition variability [15]. However, by two hours from thestart of singing, TH levels begin to rise in Area X as a potential me-chanism to counter too much acoustic variability, leading to increasedDA synthesis.

4.3. Molecular pathway model for FD song

The traditional circuit that supports low acoustic variability in FDsong (in comparison to UD song), is represented by the black coloredpathway in Fig. 7C. With 30–45min of FD song (black pathway), thereare high levels of DA from SN/VTA onto Area X, which depress gluta-matergic input from LMAN onto Area X MSNs (Fig. 7C) [10].

Consequently, PAL neurons are disinhibited and subsequent strongDLM inhibition leads to decreased excitability between LMAN and RAwith a less variable, more stereotyped song output. Based on ourfindings, we propose a new molecular pathway in blue that becomesactivated (Fig. 7D). As the bird sings to the female (blue pathway,Fig. 7D), high levels of DA with FD song and fluctuating α-syn levelsmediate decreases in GRIN2B protein expression in MSNs, decreasingtheir excitability and supporting PAL neuron inhibition of DLM. In thismanner, low acoustic variability from one song rendition to the nextwould be preserved but has not been determined. TH levels during FDsong do not change in response to the number of motifs or time spentsinging because of endogenously high DA levels.

4.4. Future directions

Our study identified differential protein expression patterns forthree molecular markers in Area X based on the amount of singing inUD vs. FD song states. We found differences in total protein levelssuggestive of modifications in dopaminergic and glutamatergic synaptictransmission in Area X by social context. To verify that MSNs are thecellular target of glutamatergic-dependent changes in synaptic plasti-city in response to social context and song amount, it will be necessaryto measure anatomical changes at cortico-MSN synapses. We wouldpredict that two hours of UD song will lead to increased numbers ofdendritic spines and glutamatergic synapses that persist to maintain theflexibility of the song circuitry needed for vocal motor exploration. Bycontrast, FD song would be associated with decreased activation ofsynaptic plasticity mechanisms to promote stability in the song circuitryand therefore, maintain a more stereotyped song output. A direct test ofthe role of GRIN2B in differential activation of synaptic plasticity me-chanisms in UD vs. FD song would involve using a viral vector tooverexpress/knockdown its expression [24]. Based on our model

Fig. 6. Differential expression of Area XGRIN2B protein levels in response to UDand FD song.GRIN2B protein levels of UD singers and FDsingers plotted against the number of motifssang within the two-hour period (A, C) or timespent singing (B, D). GRIN2B protein levelswere dependent upon the number of motifssung during UD (n=8, ANOVA, F= 13.18,p= 0.011) but not FD (n=8, ANOVA,F=1.13, p=0.329). In addition, protein le-vels were dependent upon the time spentsinging UD song (n= 8, ANOVA, F= 12.22,p= 0.013) but not FD song (n=7, ANOVA,F=5.21, p=0.071). Although the time spentsinging FD song did not have a significant ef-fect on GRIN2B levels, a downward trend wasevident. Each point represents an individualbird. Gray lines denote linear regression curvesdrawn as best fits to the data based on thesignificant F change value.

L.Y. So et al. Behavioural Brain Research 360 (2019) 103–112

110

(Fig. 7), virally-driven increases in GRIN2B expression should make FDsong more variable, abolishing social context-dependent differences.

The role of DA modulation of glutamatergic pathways in Area Xduring UD vs. FD song is not known but TH clearly plays a role.Depletion of pre-synaptic TH levels by the neurotoxin 6-OHDA resultsin a less variable UD song output [6]. Therefore, whether the reducedvariability is driven by changes at the Area X glutamatergic synapsesrequires investigation. Alternatively, optogenetic activation/inhibitionof the dopaminergic input of VTA neurons onto Area X can be used as atool to drive changes in the glutamatergic plasticity circuit and evaluatedisruption of UD vs. FD song output [34].

Acknowledgments

JEM and LYS conceived of and designed the experiments for thisstudy. LYS and SJM executed the experiments and the analyses. LYSwrote the manuscript with input from all authors. We thank Dr. MarkBorgstrom and Dr. Elena Plante for statistical consultation and theUniversity of Arizona Animal Care. Declarations of interest: none. Thiswork was supported by the University of Arizona startup funds to J. E.Miller.

References

[1] K. Simonyan, B. Horwitz, E.D. Jarvis, Dopamine regulation of human speech andbird song: a critical review, Brain Lang. 122 (3) (2012) 142–150.

[2] A. Sasaki, T.D. Sotnikova, R.R. Gainetdinov, E.D. Jarvis, Social context-dependentsinging-regulated dopamine, J. Neurosci. 26 (35) (2006) 9010–9014.

[3] S.A. Heimovics, L.V. Riters, Evidence that dopamine within motivation and songcontrol regions regulates birdsong context-dependently, Physiol. Behav. 95 (2008)258–266.

[4] S.D. Gale, A.L. Person, D.J. Perkel, A novel basal ganglia pathway forms a looplinking a vocal learning circuit with its dopaminergic input, J. Comp. Neurol. 508(5) (2008) 824–839.

[5] S.D. Gale, D.J. Perkel, A basal ganglia pathway drives selective auditory responsesin songbird dopaminergic neurons via disinhibition, J. Neurosci. 30 (3) (2010)1027–1037.

[6] J.E. Miller, G.W. Hafzalla, Z.D. Burkett, C.M. Fox, S.A. White, Reduced vocalvariability in a zebra finch model of dopamine depletion: implications for Parkinsondisease, Physiol. Rep. 3 (11) (2015) 1–16.

[7] A. Leblois, B.J. Wendel, D.J. Perkel, Striatal dopamine modulates basal gangliaoutput and regulates social context-dependent behavioral variability through D1receptors, J. Neurosci. 30 (16) (2010) 5730–5743.

[8] A. Leblois, D.J. Perkel, Striatal dopamine modulates song spectral but not temporalfeatures through D1 receptors, Eur. J. Neurosci. 35 (11) (2012) 1771–1781.

[9] L.A. Hoffmann, V. Saravanan, A.N. Wood, L. He, S.J. Sober, Dopaminergic con-tributions to vocal learning, J. Neurosci. 36 (7) (2016) 2176–2189.

[10] L. Ding, D.J. Perkel, M.A. Farries, Presynaptic depression of glutamatergic synaptictransmission by D1-like dopamine receptor activation in the avian basal ganglia, J.

Fig. 7. Molecular pathway models for UD and FD song.A: Black lines represent the circuit activity after 30–45min of UD song. B: Blue lines denote circuit activity at the end of two hours of UD song in relation to howmuch the bird sang. A bird that sang more over the two hours of UD song had increased expression of TH, α-syn, and GRIN2B (up arrows) potentially leading to moreacoustically variable song. The increased levels of TH at two hours may be a self-regulatory check to prevent too much variability in song beyond this time point. C:Black lines represent the circuit activity after 30–45min of FD song. D: Blue lines denote circuit activity at the end of two hours of FD song in relation to how muchthe bird sang. A bird that sang more over the two hours of FD song had the same levels of TH denoted as ∼, fluctuating levels of α-syn (up and down arrow), andreduced levels of GRIN2B (down arrow). The rendition-to-rendition variability of FD song compared to UD song after two hours is unknown (denoted as ?). DLM –dorsal lateral nucleus of thalamus; LMAN – lateral magnocellular nucleus of anterior nidopallium; MSN – medium spiny neurons; PAL – pallidal neurons; RA – robustnucleus of arcopallium; SN – substantia nigra; VTA – ventral tegmental area.

L.Y. So et al. Behavioural Brain Research 360 (2019) 103–112

111

Neurosci. 23 (14) (2003) 6085–6086.[11] E.D. Jarvis, C. Scharff, M.R. Grossman, J.A. Ramos, F. Nottebohm, For whom the

bird sings: context-dependent gene expression, Neuron 21 (4) (1998) 775–788.[12] C.V. Mello, S. Ribeiro, ZENK protein regulation by song in the brain of songbirds, J.

Comp. Neurol. 393 (4) (1998) 426–438.[13] I. Teramitsu, S.A. White, FoxP2 regulation during undirected singing in adult

songbirds, J. Neurosci. 26 (28) (2006) 7390–7394.[14] J.E. Miller, E. Spiteri, M.C. Condro, R.T. Dosumu-Johnson, D.H. Geschwind,

S.A. White, Birdsong decreases protein levels of FoxP2, a molecule required forhuman speech, J. Neurophysiol. 100 (4) (2008) 2015–2025.

[15] A.T. Hilliard, J.E. Miller, E.R. Fraley, S. Horvath, S.A. White, Molecular micro-circuitry underlies functional specification in a basal ganglia circuit dedicated tovocal learning, Neuron 73 (3) (2012) 537–552.

[16] S.W. Bottjer, The distribution of tyrosine hydroxylase immunoreactivity in thebrains of male and female zebra finches, J. Neurobiol. 24 (1) (1993) 51–69.

[17] J.A. Soha, T. Shimizu, A.J. Doupe, Development of the catecholaminergic in-nervation of the song system of the male zebra finch, J. Neurobiol. 29 (4) (1996)473–489.

[18] M.R. Ciucci, S.T. Ma, C. Fox, J.R. Kane, L.O. Ramig, T. Schallert, Qualitativechanges in ultrasonic vocalization in rats after unilateral dopamine depletion orhaloperidol: a preliminary study, Behav. Brain Res. 182 (2) (2007) 284–289.

[19] L.M. Grant, D.G. Barnett, E.J. Doll, G. Leverson, M. Ciucci, Relationships among ratultrasonic vocalizations, behavioral measures of striatal dopamine loss, and striataltyrosine hydroxylase immunoreactivity at acute and chronic time points followingunilateral 6-hydroxydopamine-induced dopamine depletion, Behav. Brain Res. 291(2015) 361–371.

[20] L.L. Venda, S.J. Cragg, V.L. Buchman, R. Wade-Martins, α-Synuclein and dopamineat the crossroads of Parkinson’s disease, Trends Neurosci. 33 (12) (2010) 559–568.

[21] M. Decressac, B. Mattsson, A. Björklund, Comparison of the behavioural and his-tological characteristics of the 6-OHDA and α-synuclein rat models of Parkinson’sdisease, Exp. Neurol. 325 (1) (2012) 306–315.

[22] S.E. Gombash, F.P. Manfredsson, C.J. Kemp, N.C. Kuhn, S.M. Fleming, A.E. Egan,L.M. Grant, M.R. Ciucci, J.P. MacKeigan, C.E. Sortwell, Morphological and beha-vioral impact of AAV2/5-mediated overexpression of human wildtype alpha-

synuclein in the rat nigrostriatal system, PLoS One 8 (11) (2013) 1–15.[23] J.M. George, H. Jin, W.S. Woods, D.F. Clayton, Characterization of a novel protein

regulated during the critical period for song learning in the zebra finch, Neuron 15(2) (1995) 361–372.

[24] M. Chakraborty, L.F. Chen, E.E. Fridel, M.E. Klein, R.A. Senft, A. Sarkar, E.D. Jarvis,Overexpression of human NR2B receptor subunit in LMAN causes stuttering andsong sequence changes in adult zebra finches, Sci. Rep. 7 (1) (2017) 1–18.

[25] O. Tchernichovski, F. Nottebohm, C.E. Ho, B. Pesaran, P.P. Mitra, A procedural foran automated measurement of song similarity, Anim. Behav. 59 (6) (2000)1167–1176.

[26] S.W. Scheff, Fundamental Statistical Principles for the Neurobiologist A SurvivalGuide, Academic Press, Cambridge, MA, 2016.

[27] J.L. Myers, A.D. Well, Testing for curvilinearity in regression, in: D. Riegert (Ed.),Research Design and Statistical Analysis, 2nd ed., Lawrence Erlbaum Associates,Mahwah, NJ, 2003, pp. 598–601.

[28] L. Yavich, P. Jäkälä, H. Tanila, Abnormal compartmentalization of norepinephrinein mouse dentate gyrus in alpha-synuclein knockout and A40P transgenic mice, J.Neurochem. 99 (3) (2006) 724–732.

[29] I. Gureviciene, K. Gurevicius, H. Tanila, Role of alpha-synuclein in synaptic gluta-mate release, Neurobiol. Dis. 28 (1) (2007) 83–89.

[30] J.B. Watson, A. Hatami, H. David, E. Masliah, K. Roberts, C.E. Evans, M.S. Levine,Alterations in corticostriatal synaptic plasticity in mice overexpressing humanalpha-synuclein, Neuroscience 159 (2) (2009) 501–513.

[31] S.C. Woolley, M.H. Kao, Variability in action: contributions of a songbird cortical-basal ganglia circuit to vocal motor learning and control, Neuroscience 296 (2015)39–47.

[32] M.H. Kao, A.J. Doupe, M.S. Brainard, Contributions of an avian basal ganglia-forebrain circuit to real-time modulation of song, Nature 433 (7026) (2005)638–643.

[33] S.J. Sober, M.J. Wohlgemuth, M.S. Brainard, Central contributions to acousticvariation in birdsong, J. Neurosci. 28 (41) (2008) 10370–10379.

[34] L. Xiao, G. Chattree, F.G. Oscos, M. Cao, M.J. Wanat, T.F. Roberts, A basal gangliacircuit sufficient to guide birdsong learning, Neuron 98 (1) (2018) 208–221.

L.Y. So et al. Behavioural Brain Research 360 (2019) 103–112

112