Dynamic transcriptome landscape in the song nucleus HVC between juvenile and adult zebra finches

Abstract Male juvenile zebra finches learn to sing by imitating songs of adult males early in life. The development of the song control circuit and song learning and maturation are highly intertwined processes, involving gene expression, neurogenesis, circuit formation, synaptic modification, and sensory‐motor learning. To better understand the genetic and genomic mechanisms underlying these events, we used RNA‐Seq to examine genome‐wide transcriptomes in the song control nucleus HVC of male juvenile (45 d) and adult (100 d) zebra finches. We report that gene groups related to axon guidance, RNA processing, lipid metabolism, and mitochondrial functions show enriched expression in juvenile HVC compared to the rest of the brain. As juveniles mature into adulthood, massive gene expression changes occur. Expression of genes related to amino acid metabolism, cell cycle, and mitochondrial function is reduced, accompanied by increased and enriched expression of genes with synaptic functions, including genes related to G‐protein signaling, neurotransmitter receptors, transport of small molecules, and potassium channels. Unexpectedly, a group of genes with immune system functions is also developmentally regulated, suggesting potential roles in the development and functions of HVC. These data will serve as a rich resource for investigations into the development and function of a neural circuit that controls vocal behavior.

Song behavior is controlled by a group of interconnected brain nuclei commonly referred to as the song system ( Figure 1A). The song system consists of two distinct pathways: the motor pathway, which controls song production, and the anterior forebrain pathway (AFP), which is necessary for song learning. 7-10 HVC is a cortical nucleus at the junction of these two pathways. [11][12][13] Neurophysiological, lesion, and imaging studies suggest that HVC encodes the spectral features and temporal patterns of a song. 10,[14][15][16][17] It is also known that before 45 days of age, when juveniles sing a poorly structured subsong, HVC is not required for singing, and singing is controlled by the anterior forebrain nucleus lMAN. After 45 d, song control gradually transfers from lMAN to HVC. 16 The process of shifting functional dominance is accompanied by structural development in HVC. HVC begins to establish functional synapses with its downstream nucleus RA at about 35 d, 18 and the volume of HVC and the number of neurons in HVC continue to increase 19,20 in parallel with changes in electrophysiological firing properties of HVC neurons. 5 In recent years, applying molecular and genomics tools such as in situ hybridization and cDNA microarrays to birdsong research, progress has been made in understanding gene expression programs relevant to gender, age, brain regions, learning experiences, and song behavior. [21][22][23][24][25][26][27][28][29][30] However, the dynamic transcriptional landscapes in HVC that enable its structural and functional changes as juveniles mature to adulthood remain enigmatic. In this study, we profiled transcriptomes in the HVC of male zebra finches using RNA-Seq; we focused on two age groups: 45 d, when motor control of song begins to transfer to HVC, and juveniles sing a highly variable song; and 100 d, when song learning is complete, and the now adult zebra finches sing a mature adult song. 3,4,16 Using a combination of bioinformatics analysis and experimental validation, we identified gene repertoires that define molecular signatures of HVC at 45 d and 100 d. We also describe the dynamic changes in gene expression that occur in HVC as zebra finches mature. These results provide unique insights into the transcriptional landscape underlying the development and functional maturation of the neural circuit for vocal communication.

| Profiling transcriptomes in HVC by RNA-Seq
To survey gene expression and transcriptome changes in HVC during song development, we collected brain tissues of 45 d and 100 d male zebra finches obtained from our breeding colony. Juveniles were reared with their parents in breeding cages until the collection time.
The newly adult birds were separated from their parents at around 60 to 70 d, and subsequently kept in group cages until brain collection. It is known that moment-to-moment sensory-motor experiences, such as singing or hearing songs, induce changes in gene expression in song related brain regions. 31,32 To obtain a gene expression profile in HVC at basal levels that reflects developmental stages, we monitored the birds in the morning for one hour prior to brain collection to verify that they did not sing and did not hear songs. We dissected out HVC tissue from sagittal brain sections under a dissecting microscope. We isolated total RNAs, and pooled RNA samples from HVC tissues of four animals for making each cDNA library (see the methods section for details). To identify genes with enriched expression in HVC relative to the rest of the brain, we also made cDNA libraries from whole brain tissue (WB) of male zebra finches at 45 d and 100 d.
We sequenced these cDNA libraries using the Illumina GAII platform, which produced over 30 million raw sequence reads of 79 nt per library.
F I G U R E 1 Comparison of gene expression in the HVC and whole brain tissue. A, A schematic diagram showing the song control circuit with major song nuclei and the connections among them. This study focused on the HVC, highlighted in green. B, Left, Venn diagram showing the number of genes differentially expressed in the HVC relative to the whole brain tissue at 45 d and 100 d. Right, Venn diagram showing the number of genes with enriched expression in the HVC relative to the whole brain tissue at 45 d and 100 d (q < 0.1). DE, differentially expressed. WB, whole brain. C, Images of in situ hybridization on sagittal brain sections showing enriched expression of ISG12, LBD3, LY6E2, and THBS4 in the HVC of 45 d male zebra finches. Scale bar, 1.5 mm After filtering and trimming adaptor sequences, we obtained a total of 150 million high quality sequence reads from all libraries combined. We mapped these reads to the zebra finch genome assembly (3.2.4/taeGut1) using the Eland (Illumina) software package. Typically, we obtained 25-30 million high quality sequence reads for each library; among them, 50% to 60% were mapped to the zebra finch genome. About 15% to 25% of the mapped reads were mapped to exons of annotated genes, while the remaining reads were mapped to intronic, intergenic regions, and/or regions without annotation. Altogether, mapped genes represent about 50% of the total number of annotated zebra finch genes. The general characteristics of library samples, sequencing, and mapping results, including the read counts of each libraries, their mapping rates to the genome, and the number of genes covered by each sequenced sample, are summarized in Table S1.

| RNA-Seq analysis reveals distinct gene expression patterns in juvenile and adult HVC
We first compared gene expression profiles in HVC with those in the whole brain samples, which allowed us to identify a large number of genes with enriched expression in the HVC of juvenile and adult finches. Using a cutoff of q < 0.1 (FDR-adjusted p-value), at 45 d, 1577 genes were differentially expressed in HVC relative to the whole-brain sample, and at 100 d, 811 genes were differentially expressed. Among these, less than half, 659 and 365 genes, respectively, showed enriched expression in HVC in juveniles and adults ( Figure 1B and Data S1 and Data S2). These HVC gene expression patterns, distinct from the average gene expression in the whole brain tissue, define the transcriptional programs in the HVC at 45 d or 100 d. The larger number of genes with enriched expression in HVC at 45 d indicates higher transcriptional activities at 45 d, and suggests that gene expression at 45 d is not synchronized with the rest of the brain. The asynchronicity and transcriptional activity in HVC are gradually reduced as birds mature to adulthood, suggesting that chronological age plays a determinative role in regulation of gene expression in HVC.
The enriched expression of many genes in adult HVC, especially those expressed at high levels (eg, ALDH1A2, CCNB, CADPS2, CRHBP, GLAR2, MUSTN1, NTS, PVBL, RELN, etc.), have been reported previously using various experimental platforms including differential display, cDNA microarray, and/or in situ hybridization, 24,25,29,33,34 supporting the robustness of the present study. We focused on a few  Figure 1C). These results further show the suitability of our dissection method for isolating HVC-specific tissue. The in situ hybridization results also revealed gene expression patterns in areas in addition to HVC. For example, on a sagittal brain section, ISG12, LDB3, and THBS4 also showed enriched expression in RA, a song nucleus acting downstream from HVC, which, together with HVC, controls song related motor activities. These expression patterns suggest that gene expression in the song control circuit might be functionally segregated. In an analysis of cDNA microarray gene expression data in several song-related brain regions, Lovell et al observed that each song-related region has a distinct gene expression pattern compared to its adjacent region, and their data also suggest that HVC and RA share a large number of co-expressed genes. 25 Together, these data hint at the ontogeny and evolutionary history of the song control circuit.
To gain more insight into the biological functions of the differentially expressed genes, we performed REACTOME pathway enrich-  To verify the RNA-Seq gene expression results, we performed quantitative real-time PCR (RT-qPCR) for a handful genes that met with the q < 0.1 cutoff to assess their expression levels in a separate set of HVC samples. This group of genes included CD74, CD200, CHRM4, DLK1, IFRD1, IL16, ISG12, LY6E-2, PHF15, RGS10-2, and RASGRP1. All these genes showed significant expression changes between 45 d and 100 d in HVC ( Figure 3C). In the RNA-Seq experiment, expression levels of these genes were represented with read counts ranging from 15 to several thousand, and fold-changes from 1.7-fold to 13-fold.
Despite these wide ranges and different analysis platforms, regression analysis indicated a high level correlation between the RNA-Seq and RT-qPCR results (R 2 = 0.8318, Figure 3B). We further performed RT-qPCR to test eight differentially expressed genes that did not meet with the stringent q < 0.1 cutoff, but had p < 0.01 in the RNA-Seq experiment. This second group, which included CAMK2B-1, CRIPT, FAM19A1, GABRB1-2, GAD2, NEUROD6, RAPGEF1, and UBE2A, also showed significant expression changes between 45 d and 100 d HVC by RT-qPCR, and the RNA-Seq and RT-qPCR results showed a good correlation (R 2 = 0.69, Figure 3B,C. For all RT-qPCR experiments, n = 4-9 animals per age group were used. The RT-qPCR data are also shown in Data S5). Together, these results provide an independent validation of our RNA-Seq approach to identifying differentially expressed genes in HVC during song development.
To examine the functions of the developmentally regulated genes, they were sorted into groups according to their functional annotations.
The relative distribution of these groups is summarized in Figure 4A.
These groups include genes with functions related to cytoskeleton and microtubule, extracellular matrix and cell adhesion, G protein-coupled neuromodulator receptors, immune system function, mitochondrial function, RNA processing, and signal transduction ( Figure 4B). The cell adhesion and extracellular matrix proteins have important roles in neuronal migration and maturation, while the cytoskeletal and microtubule proteins represent essential structural components for establishing and modifying neuronal dendritic and synaptic morphology. Within these groups, some genes were upregulated and others downregulated between the two ages. Among the signal transduction genes, FZD10 is known to promote sensory neuron development in Xenopus 35 and was observed here as upregulated at 45 d. In contrast, genes encoding proteins for synaptic functions, including G-protein coupled neuromodulator receptors, were mostly upregulated at 100 d compared to 45 d. These patterns are consistent with and reflect the highly orchestrated processes of circuit maturation and synaptic reorganization, particularly, the transition from synaptogenesis at 45 d to mature synaptic functions as zebra finches mature into adulthood. The downregulation of genes related to mitochondrial function may indicate the shifting of energy demand during this process. Dysfunction of many of these genes has been associated with a wide range of neurodevelopmental disorders (Table 1).
Unexpectedly, a group of genes known for their functions in the immune system was found as developmentally regulated in HVC (Table 2). This group includes genes encoding the major histocompatibility complex class II invariant chain1 (CD74), the immune inhibitory molecule CD200, the brain-specific chemokine (FAM19A1), interferon-related developmental regulator 11 (IFRD1), interleukin IL16, interferon inducible protein ISG12, and Lymphocyte antigen complex 6 (LY6E-2). Expression changes of these genes were validated using RT-qPCR (see Figure 3C,D). In addition, at least some of them (eg, ISG12 and LY6E-2) exhibited enriched expression specifically in HVC and other song related brain regions ( Figure 1C). Together, these observations argue against the possibility that their expression patterns were due to pathogen-activated immune responses, although that cannot be entirely excluded.
We performed REACTOME analysis of genes displaying developmentally regulated expression in HVC. The most highly enriched F I G U R E 2 REACTOME enrichment analysis of genes in the HVC compared with whole brain at 45 d and 100 d. A, Enriched REACTOME terms among genes that show enriched expression in 45 d or 100 d HVC compared to whole brain tissues. B, Interconnections between enriched terms. In both A and B, p-values are color-coded; grey represents the absence of enrichment term among genes with higher expression in 45 d HVC is metabolism of amino acids and derivatives. This group includes many genes F I G U R E 3 Validation of gene expression changes using RT-qPCR. A, RNA-Seq analysis identified 113 differentially expressed genes in HVC as birds matured from 45 d to 100 d; 70 genes increased expression, and 43 genes decreased expression (q < 0.1). B, Regression analysis of RNA-Seq and RT-qPCR results showing correlations between gene expression changes. Note the black line and the dots represent genes whose expression changes met the q < 0.1 cutoff in RNA-Seq experiment. The blue line and the triangles represent genes whose expression changes did not meet the q < 0.1, but met the p < 0.01 cutoff in RNA-Seq. C, Validation by RT-qPCR of expression changes of specific genes that met the q < 0.1 cutoff in RNA-Seq experiment (black dots in B). D, Validation by RT-qPCR of expression changes of genes that did not meet the q < 0.1 cutoff, but had p < 0.01 in RNA-Seq (blue triangles in B). In both C and D, n = 4-9 animals per age group. *p < 0.05, **p < 0.01, ***p < 0.001 While a large portion of the genome is active in HVC during its development, we noted that a sizable fraction of the high-quality sequence reads (40%) failed to map to the zebra finch genome. Of the reads mapped to the genome, a large fraction mapped to regions without annotation. The incomplete assembly and annotation of the zebra finch genome may have contributed to these observations. However, in recent years, as the sensitivity of detection techniques improves, mounting evidence has indicated that, in addition to known genes, a large portion of mammalian genomes is transcriptionally active. This  16 Many genes pertaining to extracellular adhesion and Slit-ROBO signaling, typically acting via ligand-receptor interactions, have important roles in neuron migration, axon guidance, cell-cell and cellmatrix recognition, and establishing synaptic connections. 45 Members T A B L E 1 Genes developmentally regulated in HVC that have been implicated in nervous system related disorders  include autosomal recessive non-syndromic intellectual disability. 54,55 UBE2A encodes a member of the E2 ubiquitin-conjugating enzyme family, which catalyzes the covalent attachment of ubiquitin to substrate proteins. Disorders associated with UBE2A include mental retardation, seizures, poor speech, and aggressive behavior. [56][57][58] UPF3A encodes the regulator of nonsense mediated mRNA decay (NMD), a process cells use to eliminate faulty mRNAs. UPF3A and its paralog UPF3B antagonistically regulate NMD, and both have been implicated in ASDs, intellectual disability, and schizophrenia. 59

| cDNA library preparation and sequencing
Total RNAs from HVC of four birds per age group were pooled and prepared into one sequencing library using a NuGEN Ovation RNAseq V1 kit following the manufacturer's instructions. An additional step of S1 nuclease treatment of cDNA was used as described. 71

| RT-qPCR
RT-qPCR was performed as described previously. 75 Briefly, total RNA was isolated from HVC tissue using Trizol reagent (Invitrogen), and quantified using a Nanodrop spectrophotometer. Reverse transcription was performed using 50 ng of total RNA using an iScript Reverse Transcription Supermix kit (Bio-Rad) following the manufacturer's instructions. qPCR was performed using the iQ SYBR Green Supermix (Bio-Rad) following the manufacturer's instructions. GAPDH was used as a reference gene after determining that its expression did not change during development. Relative gene expression levels were determined using the comparative Ct (2 −ΔΔCt ) method after normalizing to GAPDH. 76 N = 4-9 animals per age group were used in the RT-qPCR experiment. For all samples, RT-qPCR was performed in triplicate twice. Results shown in Figure 3 and Data S5 are from one set of experiment. Dissociation curve analysis was performed to confirm a single peak PCR product for each gene, indicating the specificity of PCR reactions. All primers were obtained from IDT (Integrated DNA Technology); their sequences are listed in Data S7.

| In situ hybridization
In situ hybridization was performed as described previously. 34 Briefly, fresh frozen zebra finch brains were cut into 10 μm sagittal sections and kept at −80 C until use. Brain sections were fixed in 4% PFA for 10 minutes followed by acetylation for 10 minutes. PCR amplified cDNA fragments containing the probe sequences (200-300 bp, sequence-verified) were cloned into pBluescript plasmid vectors, and sequences were verified. 33 P-labeled ribo-probes were made by in vitro transcription using T7 or T3 RNA polymerase (PerkinElmer Kit).
For hybridization, 10 6 cpm probes in 40 μL hybridization buffer were added to each brain section and hybridized at 65 C overnight. After hybridization, brain sections were washed two times, 30 minutes each, at 65 C with wash solution (50% formamide, 1xSSC, and 0.1% Tween 20), followed by washing 2 times with 0.2 X SSC at 65 C. Slides were exposed to X-films for 1-7 days, depending on signal intensity. Experimental design, data analysis, manuscript preparation, and project management.

CONFLICT OF INTEREST
All authors declare no conflict of interest.

DATA AVAILABILITY STATEMENT
The