ZEBrA - Zebra finch Expression Brain Atlas: a resource for comparative molecular neuroanatomy and brain evolution studies

Peter V Lovell; Morgan Wirthlin; Taylor Kaser; Alexa A Buckner; Julia B Carleton; Brian R Snider; Anne K McHugh; Alexander Tolpygo; Partha P Mitra; Claudio V Mello

doi:10.1002/cne.24879

. Author manuscript; available in PMC: 2021 Aug 1.

Published in final edited form as: J Comp Neurol. 2020 Feb 19;528(12):2099–2131. doi: 10.1002/cne.24879

ZEBrA - Zebra finch Expression Brain Atlas: a resource for comparative molecular neuroanatomy and brain evolution studies

Peter V Lovell ¹, Morgan Wirthlin ^1,⁴, Taylor Kaser ¹, Alexa A Buckner ¹, Julia B Carleton ¹, Brian R Snider ², Anne K McHugh ¹, Alexander Tolpygo ³, Partha P Mitra ³, Claudio V Mello ^1,^*

PMCID: PMC8219259 NIHMSID: NIHMS1556781 PMID: 32037563

Abstract

An in-depth understanding of the genetics and evolution of brain function and behavior requires detailed mapping of gene expression in functional brain circuits across major vertebrate clades. Here we present the Zebra finch Expression Brain Atlas (ZEBrA; www.zebrafinchatlas.org, RRID: SCR_012988), a web-based resource that maps the expression of genes linked to a broad range of functions onto the brain of zebra finches. ZEBrA is a first of its kind gene expression brain atlas for a bird species, and a first for any sauropsid. ZEBrA’s >3,200 high-resolution digital images of in situ hybridized sections for ~650 genes (as of June, 2019) are presented in alignment with an annotated histological atlas and can be browsed down to cellular resolution. An extensive relational database connects expression patterns to information about gene function, mouse expression patterns and phenotypes, and gene involvement in human diseases and communication disorders. By enabling brain-wide gene expression assessments in a bird, ZEBrA provides important substrates for comparative neuroanatomy and molecular brain evolution studies. ZEBrA also provides unique opportunities for linking genetic pathways to vocal learning and motor control circuits, as well as for novel insights into the molecular basis of sex steroids actions, brain dimorphisms, reproductive and social behaviors, sleep function, and adult neurogenesis, among other fundamental themes.

Keywords: molecular signature, gene expression, brain organization, functional circuits, zebra finch, song system, brain atlas

Graphical Abstract

graphic file with name nihms-1556781-f0017.jpg

Introduction

A key step towards understanding the genetic basis of brain function and behavior is to map the expression of genes onto functional circuits and cells. The importance of brain gene expression mapping is well illustrated by the Allen Institute’s Mouse Brain Atlas, a resource that has allowed for the examination of >20,000 brain transcripts in a representative mammalian model organism (Hochheiser & Yanowitz, 2007; Lein et al., 2007; Sunkin et al., 2013). This atlas has made numerous contributions to our understanding of the molecular organization of the brain of mammals, including defining molecular signatures of major structures (e.g. cortical layers, hippocampus, amygdala), providing insights into developmental changes in gene expression, and facilitating discoveries linking genetic pathways to circuits that underlie specific functions and/or behaviors (e.g. Acquaah-Mensah & Taylor, 2016; Belgard et al., 2011; Bohland et al., 2010; Hawrylycz et al., 2010; Henry & Hohmann, 2012; Jones et al., 2009; Lein et al., 2017; Li et al., 2017; Sunkin et al., 2013; Thompson et al., 2014; Thompson et al., 2008).

A comparable resource has been missing up until now for birds. Considering that birds have been broadly used as experimental model organisms and have contributed important insights into brain function and behavior, the lack of such a resource constitutes a major gap in neurobiology. For example, birds are increasingly recognized as being capable of sophisticated learned behaviors (e.g. tool use, episodic memory), with corvids (e.g. crows, ravens, jays) and parrots possessing cognitive abilities that rival or even surpass those of mammals, including some non-human primates (e.g. Clayton & Emery, 2015; Gunturkun & Bugnyar, 2016; Jarvis et al., 2005; Pepperberg, 2006). Several avian groups (songbirds, parrots and hummingbirds) have also evolved vocal learning, a trait that enables animals to learn vocalizations by imitation, as extensively studied in zebra finches (Taeniopygia guttata). This trait evolved convergently in these avian vocal learners and in humans (Figure 1a, left, red asterisks), where it provides an important basis for speech and language acquisition, but is largely absent or rudimentary in most non-human mammals, including rodents and non-human primates (Jarvis, 2004, 2019; Petkov & Jarvis, 2012; Wirthlin et al., 2019). Birds also have highly developed visual systems and high visual acuity (e.g. Bennett et al., 1996; Gaffney & Hodos, 2003; Hodos et al., 2003; Martin, 2014; Shimizu et al., 2010), and seminal studies in species like chicken and guinea fowl, and hearing specialists like owls, have been at the forefront of discoveries on central auditory processing and the neurobiological basis of sound localization (Grothe et al., 2004; Hong & Sanchez, 2018; Knudsen & Konishi, 1978; Konishi, 2003; Langner, 1981; Overholt et al., 1992; Scheich & Bonke, 1981; Smith & Rubel, 1979; Theurich et al., 1984; Walton et al., 2017). Many bird species also have complex social behaviors and thus have been excellent models for understanding the neuronal basis of sociality (Goodson & Kabelik, 2009; Kelly & Goodson, 2014). Chicken and quail are also broadly used for development studies, and have contributed numerous insights into early brain patterning, axonal guidance and neurotrophins, among others (e.g. Fraser et al., 1990; Hamburger & Hamilton, 1951; Keynes & Cook, 2018; Keynes & Stern, 1984; Lance-Jones & Landmesser, 1980).

Figure 1. — (a) Simplified phylogeny and brain evolution in amniotes. Left: Schematic diagram depicting phylogenetic relationships across major amniote lineages; branches containing bird lineages are highlighted in blue. Only select major extant groups are depicted; dashed line indicates dinosaur lineage extinction. Red asterisks indicate species/groups that possess the trait of vocal learning. Right: Schematic diagrams of representative vertebrate brains highlighting the relative sizes of pallial (orange) and subpallial (pink) areas in representative synapsids and sauropsids. (b) Schematic generic representation of the ‘nuclear’ vs. ‘layered’ hypothesis of pallial organization in avian (top) and mammalian (bottom) brains (after Karten, 1991). Color coding indicates analogous (and possibly homologous) elements between lineages. (c) Schematic diagrams of specialized pathways for birdsong (top) and speech (bottom). Top: Vocal control circuitry of the adult male zebra finch, depicted in the parasagittal plane; some structures have been omitted for clarity. The system consists of a direct motor pathway for song production (in black) that includes connections from HVC to RA, and from RA to brainstem nuclei involved in vocal-motor and respiratory control (nXIIts, RAm), and an anterior forebrain pathway for song learning (in white) containing a cortico-striato-thalamo-cortical loop with projections from pallial LMAN to striatal Area X, from Area X to thalamic DLM, from DLM to LMAN, and from LMAN back to Area X. Additional connections are shown in grey. Bottom: Proposed vocal control circuitry in humans consisting of a direct motor pathway for speech production (in black), and striatal brain regions (Cd, Pt) that have been proposed to be involved in speech learning (in white), adapted from Pfenning et al. 2014. The descending projections from RA and dLMC to brainstem vocal nuclei are thought to be convergently evolved in songbirds and humans. Abbreviations: Am, nucleus ambiguus; Cd, Caudate nucleus of the striatum; dLMC, dorsal part of the laryngeal motor cortex; DLM, medial part of the dorsolateral nucleus of the anterior thalamus; GP, Globus pallidus; HVC, HVC proper name; LMAN, lateral magnocellular nucleus of the anterior nidopallium; nXIIts, tracheosyringeal portion of the nucleus of the hypoglossal nerve (XII); PAG, periaqueductal grey; Pt, Putamen; RA, robust nucleus of the arcopallium; RAm, nucleus retroambiguus medialis; X, Area X.

Notably, while birds possess highly developed cognitive and sensorimotor skills and share important brain features with mammals, their brain organization differs significantly from that of mammals. Birds belong to sauropsids (Figure 1a, left), a large group of amniotes that includes turtles, lizards, snakes, and crocodiles. While the telencephalon of birds and mammals is largely composed of pallium (Figure 1a right, orange area in brain diagrams) (Cobos et al., 2001; Fernandez et al., 1998; Jarvis et al., 2005; Puelles et al., 2000; Redies et al., 2001; Reiner et al., 1998), the avian pallium has expanded considerably relative to the basal ganglia when compared to non-avian sauropsids, similar to how the cortex is expanded in primates compared to more basal mammals (Figure 1a, right; pallium in orange, basal ganglia in pink across species). Intriguingly, bird brains contain more pallial neurons than equivalently-sized mammalian brains, and parrot and corvid brains contain on average twice as many neurons as comparably-sized primate brains (Olkowicz et al., 2016). Birds have a largely non-layered telencephalon (Jarvis et al., 2005; Karten, 2013), and their pallium has a more ‘nuclear’ organization compared to the layered cortex of mammals (Figure 1b). Nonetheless, birds have cortical-like circuitry elements such as distinct thalamo-recipient zones (e.g. auditory field L, visual entopallium) and nuclei that originate descending pallial projections (Mello et al., 1998; Wang et al., 2010; Wild, 1993; Zeier & Karten, 1971). These differences and similarities have raised questions about avian homologs of mammalian pallial structures like neocortex, claustrum, and amygdala (e.g., see Jarvis et al., 2005; Karten, 1997, 2013; Medina et al., 2004; Puelles et al., 1999; Reiner et al., 2004), as well as how the computational needs for sensory, motor, and integrative processing are met by brains with very different layouts (Guy & Staiger, 2017). Whether genetic mechanisms subserving layered vs. non-layered circuits are shared between mammals and birds, and more generally sauropsids, remains unclear despite previous efforts (e.g. Belgard et al., 2013; Dugas-Ford et al., 2012; Jarvis et al., 2013; Montiel et al., 2016; Puelles, 2017; Puelles et al., 2016; Suzuki & Hirata, 2014; Tosches et al., 2018), in part due to limited expression data in an avian species.

Birds also offer excellent opportunities for mapping molecular pathways onto circuits that subserve specific behaviors, largely facilitated by their unique brain features - discrete nuclei with distinct cytoarchitectonics. One of the best examples is the zebra finch, the premier model organism for studying the neurobiology of vocal learning (Fee & Scharff, 2010; Goodson et al., 2009; Mello, 2014; Zann, 1996; Zeigler & Marler, 2004, 2008). This songbird species shares with humans vocal learning features like a critical period, the presence of early babbling-like immature vocalizations, and the ability to modulate vocalizations during sensorimotor learning by reference to auditory feedback (Bolhuis et al., 2010; Doupe & Kuhl, 1999). While some mammals such as bats, marine mammals and elephants (Janik, 2014; Knornschild, 2014; Poole et al., 2005; Prat et al., 2015) may have evolved some of these features, finches are much more amenable to experimentation, and benefit from resources such as a high quality genome assembly and extensive cDNA/EST databases (Korlach et al., 2017; Li et al., 2007; Replogle et al., 2008; Wada et al., 2006; Warren et al., 2010). Most importantly, the finch vocal circuitry has been extensively characterized, and is comprised of: (a) a direct pathway containing discrete cortical pre-motor and motor areas and brainstem vocal and respiratory nuclei (Figure 1c, top, in black) involved in vocal-motor control (Nottebohm & Arnold, 1976; Nottebohm et al., 1982); and (b) an anterior forebrain pathway consisting of a cortico-striato-thalamo-cortical loop (Figure 1c, top, in white) that plays key roles in vocal learning and plasticity (Bottjer & Arnold, 1984; Brainard & Doupe, 2000a, 2000b; Kao et al., 2005; Scharff & Nottebohm, 1991; Sohrabji et al., 1990; Thompson & Johnson, 2007). This system bears remarkable analogies to human vocal areas (Figure 1c, bottom), particularly when considering the descending motor pathway. Knowledge of the organization and physiological properties of finch vocal nuclei has contributed key insights into the roles of cortex and basal ganglia in vocal learning and vocal-motor representation (e.g. Aronov et al., 2008; Dutar et al., 1998; Hahnloser et al., 2002; Kubke et al., 2005; Leonardo & Fee, 2005; Mooney, 2000; Mooney & Prather, 2005; Perkel et al., 2002; Rosen & Mooney, 2003) (see also chapters in Zeigler & Marler, 2004, 2008). Studies in the zebra finch have also been instrumental in elucidating the role of FOXP2, the first gene to be linked to inheritable speech disorders in humans (Burkett et al., 2018; Haesler et al., 2007; Haesler et al., 2004; Teramitsu et al., 2004; Wohlgemuth et al., 2014). Zebra finches have also been widely used to study themes such as the endocrine modulation of reproductive and social behaviors (Goodson & Kabelik, 2009; Kelly & Goodson, 2014), sex steroid actions and brain dimorphisms (Nottebohm & Arnold, 1976; Pinaud et al., 2006; Williams, 1985), the role of sleep in learning (Brawn et al., 2008; Dave & Margoliash, 2000; Hahnloser et al., 2006), how cortical and basal ganglia circuits contribute to motor coding and rhythms (Budzillo et al., 2017; Fee et al., 2004; Gadagkar et al., 2016; Goldberg & Fee, 2010; Hahnloser et al., 2002; Hisey et al., 2018; Leonardo & Fee, 2005; Roberts et al., 2012), adult neurogenesis and neuronal replacement (Alvarez-Buylla et al., 1990), behavioral modulation of gene expression (Drnevich et al., 2012; Hara et al., 2012; Hilliard et al., 2012a; Hilliard et al., 2012b; Li et al., 2007; Stevenson et al., 2012; Thompson et al., 2012; Wada et al., 2006; Whitney et al., 2014), and the role of endocannabinoids (Soderstrom & Johnson, 2003; Soderstrom & Tian, 2004) and ethanol (Olson et al., 2014) in addiction. However, the genetic basis of these functions has remained largely unexplored.

Previous studies have identified specializations of the finch vocal circuitry through focused microarray and in situ hybridization screenings (Li et al., 2007; Lovell et al., 2008; Lovell et al., 2018a; Wada et al., 2004), but there have been no concerted efforts to expand in situ datasets, to study brain-wide gene expression, to present data under a unifying platform, or to generate tools for correlational and comparative analyses. To address these gaps and further our knowledge about the genetic basis of brain function and behavior, we have mapped a large number of brain transcripts in zebra finches and documented the differential expression of genes in vocal nuclei and broader areas. Here we describe the Zebra finch gene Expression Brain Atlas (ZEBrA), an open access website (www.zebrafinchatlas.org, RRID: SCR_012988) that contains >3,200 high-resolution digital in situ hybridization images for ~650 brain transcripts, presented in registration with an annotated histological reference atlas. This atlas is the first of its kind for an avian species, and more generally the first for a sauropsid. To facilitate comparative neuroanatomy, genomics and bioinformatics, ZEBrA includes extensive metadata on gene involvement in speech and language function, neurological and psychiatric diseases and phenotypes, brain expression in mouse, regulatory roles in brain development and function, and links to songbird-related discoveries and to key mammalian databases (Allen Mouse Brain Atlas, OMIM, MGI). Among studies that have already benefited from ZEBrA, we note findings on avian brain organization that give insights into cortical and amygdalar evolution (Jarvis et al., 2013; Mello et al., 2019), as well as the discoveries of convergent specializations of vocal areas shared by vocal learning birds and humans (Pfenning et al., 2014), compensatory changes in response to a lineage-specific loss of a major neuromodulatory gene (Lovell et al., 2015; Lovell et al., 2014), and vocal circuitry specializations related to neuronal excitability regulation (Friedrich et al., 2019; Lovell et al., 2018a). ZEBrA is thus a unique resource that enables exploring brain gene expression patterns in birds, facilitates comparative analyses between birds and other vertebrate lineages, and enables linking molecular pathways to a range of fundamental questions in neurobiology and brain evolution.

Methods

Gene Selection for Portals

The strategy and inclusion criteria used to build the six ZEBrA portals are detailed below. While the criteria are independent across portals, every gene in ZEBrA has been systematically examined to verify if it meets the criteria for inclusion in each portal. As new genes are added to ZEBrA, they are cross-referenced to the primary list of gene entries and/or external databases specific to each portal as described below, so that the contents of each portal/subportal can be updated.

(1). Markers of the Song System.

To build this portal, we examined all relevant in situ hybridization images of adult male zebra finches available in ZEBrA to identify genes that are differentially up- or down-regulated in one or more song nuclei compared to adjacent brain areas, based on visual inspection of images. For some nuclei, we further subcategorized genes that showed sparse cellular expression (e.g. sparse cells in HVC, area X, or RA) and/or were expressed within a restricted subdomain within a nucleus (e.g. paraHVC). For LMAN, a so-called shell region can be distinguished molecularly from a core domain or central nucleus. This LMAN shell has distinct connectivity (Bottjer et al., 2000; Bottjer et al., 1989; Johnson et al., 1995) and has been considered by some a non-vocal motor area (e.g. Feenders et al., 2008), or an auditory area (e.g. Bottjer et al., 2000; Johnson et al., 1995), while other recent studies have suggested a role in vocal learning (Achiro et al., 2017; Bottjer & Altenau, 2010). We use the term LMAN shell as an neuroanatomical description without any implications of function or possible analogy with the shell of the oval nucleus of the anterior nidopallium (NAo), which is considered the parrot equivalent to songbird LMAN. With regards to DLM, a large number of ZEBrA patterns show differential expression in a discrete the dorsal thalamic nucleus that receives projections from striatal Area X, projects to LMAN (Johnson et al., 1995; Luo et al., 2001), and exhibits singing-related up-regulated of DUSP1 (Horita et al., 2012), suggesting a prominent role in vocal learning. This nucleus has been termed DLM in most studies by several labs and has been sometimes referred to as aDLM (after Horita et al., 2012; Kubikova et al., 2010; Wada et al., 2004). Some ZEBrA genes show differential expression in a shell-like region that seems to circle DLM as defined above, and/or in a larger that likely encompasses larger and/or additional nuclei of the dorsal-medial thalamus. It is possible that in some cases this may correspond to the dorsal thalamic region that projects to the LMAN shell (Johnson et al., 1995) and/or that has been labeled DLM to differentiate it from the aDLM (Horita et al., 2012; Kubikova et al., 2010; Wada et al., 2004). A clear definition of this complex region requires efforts beyond the scope of the present paper, but we have taken note of these cases by placing an Unknown Thalamus label over it whenever relevant. A complete list of song nuclei and respective subdomains is presented in SuppInfo:STable1.xlsx. The majority of genes in this portal were identified as candidate markers of song nuclei HVC, RA, or Area X, based on differential microarray screenings (Li et al., 2007; Lovell et al., 2008; Lovell et al., 2018a; Pfenning et al., 2014; Wada et al., 2004). However, every gene in ZEBrA was also evaluated for differential expression in the song system, and has been included in this portal as applicable.

(2). Speech and Language:

To build this portal, we consulted our custom Human Disorder classification tree, which was based on the Online Mendelian Inheritance in Man (OMIM; Dec., 2017) catalog of human genes and genetic disorders. Construction of this custom classification tree is described below under (3) Diseases and Phenotypes. The part of this classification tree that is represented in this portal contains the full set of human genes that were classified as having an association with speech, language, communication, and/or an autism-related disorder. In addition, we consulted the Simons Foundation Autism Research Initiative (SFARI; Q2 Release Oct. 2017) website and identified genes listed in that website at ‘High Confidence’, ‘Strong Candidate’, or ‘Suggestive Evidence’ levels of linkage to autism spectrum disorders. We also included a number of genes that have been highlighted in the literature as being involved (or theoretically involved) in disorders like specific language impairment, dyslexia, childhood dyspraxia, stuttering, and psychiatric disorders with strong language components like schizophrenia. We also included genes that have been linked to speech and language through their shared expression in human and avian vocal circuits (Pfenning et al., 2014). A complete list of speech and language disorders, highlighted genes, and their parent level organizing themes is presented in SuppInfo:STable1.xlsx. The collection of genes in this portal represents an exhaustive effort to list all candidate genes with functions related to speech and language, thus it includes human genes that have not been identified in the genome of zebra finch and/or of birds in general, or for which in situ probes are not currently available.

(3). Diseases and Phenotypes:

To build the Human Diseases subportal, we downloaded OMIM’s Synopsis of the Human Gene Map file (morbidmap.txt; Dec., 2017), which contains a complete listing of genetic disorders and gene associations presented in OMIM. As of December 2017, this corresponded to 11,705 genes, which are associated with a total of 6,668 unique human disorders. We then cross-referenced all genes in ZEBrA with this human gene list. Because OMIM does not provide a hierarchical classification of disorder types, we next constructed a custom Human Disorder classification tree containing 53 unique disorder themes or subthemes that are represented in the ZEBrA database. Specifically, we classified genes in ZEBrA as being associated with ‘Neurological/Psychiatric Disorders’ (n=29 themes), of which we highlight a subset where the associated genes are more highly represented in ZEBrA (n=6 themes: epilepsy, mental retardation, neurodegeneration, ataxia, neuropathy, addiction), or with ‘Other Disorders’ (n=22 themes) which have no obvious neurological and/or psychiatric association. We note that to build this classification each OMIM entry related to a given gene in ZEBrA was determined to belong to one or multiple themes/sub-themes based on the presence of related keywords in the disorder description (e.g. epilepsy, deafness) or an interpretation of the description itself. A complete list diseases, disorders, and their parent level thematic classifications is presented in SuppInfo:STable1.xlsx. To build the Mouse Phenotypes subportal we first downloaded the file: MGI_GenePheno.rpt (Dec. 2018) from the Mouse Genome Informatics (MGI) database. This file contains coded descriptions of specific brain and behavioral phenotypes obtained from gene deletion, knock-down, and/or overexpression studies in experimental mouse models. As of December 2018, this consisted of 3,212 genes related to 1,347 unique phenotypes. Next, we cross-referenced the full set of ZEBrA genes with the genes in this Mouse Phenotype database. Using the MGI Mammalian Phenotype Vocabulary classification system as a starting point, we next constructed a Mouse Phenotype classification tree based on genes that are represented in the ZEBrA database. This classification tree is organized around major phenotypic themes of ‘Abnormal Behavior’ or ‘Brain Morphology Or Physiology’, with a total of 648 phenotypic sub-themes. A complete list of phenotypes and their parent level classifications is presented in SuppInfo:STable1.xlsx. We note that the specific entries in the OMIM and MGI databases for the corresponding genes in ZEBrA can be accessed through the linkouts in this portal.

(4). Comparative Neuroanatomy:

To build the Songbird Brain Markers subportal, we systematically examined in situ hybridization images for every gene in ZEBrA, and identified subsets that are differentially up- or down-regulated in various general avian brain areas (other than song control nuclei) compared to adjacent brain areas or subdivisions. A complete list of anatomical structures and their parent level organization within the Songbird Brain Marker subportal is presented in SuppInfo:STable1.xlsx. We note that for n. Ovoidalis (Ov) and for the Entopallium (E) we present markers of both core and shell regions. For the dorsolateral corticoid area (CDL) we present markers of both the entire region and a small and cytoarchitectonically distinct nucleus within the CDL that we have named the core of CDL (CDLco), following the convention used in the chicken stereotaxic atlas (Puelles, 2007). We note that this core nucleus has also been referred to as the dorsal nucleus of the hyperpallium (DNH) by researchers studying night vision and magnetoreception in finches and other birds (e.g. Heyers et al., 2007; Mouritsen et al., 2005; Mouritsen et al., 2016). However, we believe there is currently not conclusive evidence placing this nucleus within the hyperpallium. To build the Mouse Brain Markers subportal, we first constructed a database of molecular markers of mouse brain structures comprised of lists of genes that have been identified by the Allen Mouse Brain Atlas as differential markers of specific cortical layers, basal ganglia, hippocampal and amygdalar subdivisions, as well as a specific thalamic and neuromodulatory cell groups and/or nuclei, among others. For most structures, we retrieved pre-populated list of genes provided by the “Browse by Differential Expression” tool located on the ISH Data page at mouse.brain-map.org. We used default search settings and initially retrieved the 100 top and 100 bottom genes with the largest fold-changes. We next verified the retrieved lists by examination of the Mouse Brain Atlas patterns for genes with fold-change close to 1, and adjusted our cutoff further if genes in this range were identified as non-differential for the structure under consideration. For structures of interest not listed in “ Browse by Differential Expression” tool, we retrieved pre-populated markers sets in the “Fine Structures Search”. In such cases, the lists typically contained less than 50 markers and all markers were retrieved. Next, we cross-listed every gene in ZEBrA against the retrieved mouse brain structure gene lists, and populated the subportal tree accordingly. A complete list of anatomical structures along with their parent level organization within the Mouse Brain Marker subportal is presented in SuppInfo:STable1.xlsx.

(5). Gene Function:

To build this portal, we first adapted the general classification scheme (based on the Gene Families Index of the Human Genome Nomenclature Consortium - HGNC), modifying it to focus on molecular functions of relevance to brain circuits and the regulation of behavior. This includes several aspects of brain development (e.g. regulators of neurite extension, axonal guidance and fasciculation, cell:cell and cell:matrix adhesion, synapse formation, among others) and physiology (neurotransmitter and neuromodulatory receptors, peptides and respective receptors, synaptic function modulators, among others). We also included categories like intracellular signaling and trafficking, gene expression regulation, immune function, myelin-related, and genes that affect neuronal excitability. Other more general categories relate to various aspects of metabolism and gene function. A complete list of gene functions along with their parent level of organization is presented in SuppInfo:STable1.xlsx. We next categorized ZEBrA genes based on this classification scheme. In many cases this was straightforward, as several functional categories are tightly linked to the names of specific genes and/or gene families. In other cases this gene categorization required evaluating Gene Summaries provided by NCBI:Gene, GeneCards.org, and HGNC, or conducting literature searches in NCBI:Pubmed. Genes known to have multiple functions have been listed under multiple categories within this portal.

(6). Songbird Discoveries:

For the ‘Highlighted ZEBrA-Based Discoveries’ subportal, we identified songbird studies that have benefitted from the gene expression patterns available in ZEBrA to derive novel insights into comparative and physiological questions. For the ‘Songbird Studies’ subportal, we identified fundamental themes to which research in zebra finches and other songbirds has contributed novel and basic insights. This is just our own sampling of interesting or important studies in songbirds, and this expanding portal can easily accommodate expansions as more studies are published that reveal functional links for genes in the ZEBrA database, and/or as the ZEBrA database itself expands. Genes within these subportals are arranged first by major theme, and then by a reference in the literature that provides the evidence for the gene inclusion within a given theme. A complete list of themes and their parent level is presented in SuppInfo:STable1.xlsx.

Identification of Zebra Finch Orthologs and Gene Model Curation

All genes presented in ZEBrA, including those not within a portal, have been individually and systematically vetted to verify their identity and orthology to mammalian genes using a combination of genomic alignments followed by synteny analysis (verification of flanking genes) in available high quality avian genomes (zebra finch, chicken), well assembled/curated mammalian genomes (human, mouse), and when necessary some out-groups (platypus, anole lizard, alligator, xenopus frog). Our curation pipeline consisted of the following steps. We first identified a candidate locus for each gene of interest in the zebra finch genome (taeGut3.2.4) by searching for either an Ensembl (v73.1-v94) or NCBI’s RefSeq gene model with the gene name of interest. We next examined the alignment of the human ortholog to the zebra finch genome by BLAT (UCSC genome browser; Kent, 2002). At this step we examined every locus in the genome with a significant alignment (i.e. BLAT score > 30) to distinguish the putative ortholog from hits to paralogs or related gene family members that may be similar in sequence. From this set of alignments we identified the orthologous locus by verifying which locus in the zebra finch genome with a significant BLAT alignment had the same local synteny (i.e. flanking genes) as the correct locus in the human genome. In cases where synteny was partial due to chromosomal rearrangements or regions where the genome is incomplete, verification was extended to other taxa related to finch and humans (e.g. chicken, mouse, platypus) or to representative species of more ancestral lineages or outgroups to birds and mammals (e.g. alligator, lizard, xenopus, zebra fish, as needed). Finally, we confirmed each locus by reciprocal best BLAT (or BLAST) alignments. The combination of reciprocal best alignments supported by conserved synteny provides the best evidence that a given locus in zebra finch is orthologous to that in humans and mouse. As an example, the medium weight neurofilament (NEFM) gene has nearly identical syntenies on zebra finch chr8 and human chr22 (Figure 2a). However, BLAT alignments of NEFM predictions from finch and human show the highest reciprocal alignment and nucleotide % identity scores to the correct corresponding loci (finch on human BLAT score = 1154, % identity = 87%; human on finch BLAT score = 1085, % identity = 88%; cross-species alignments shown in Figure 2b–c). No similarly high scoring secondary alignments (e.g. paralogs) exist, confirming that the finch and human NEFM loci are orthologous. We note that applying this strategy to this and other molecular studies in finches has resulted in the annotation of a large collection of unannotated Ensembl models, as well as the curation of numerous others that were either incorrectly annotated, or whose orthology could be unambiguously established (Lovell et al., 2013; Lovell et al., 2018b; Lovell & Mello, 2018; Wirthlin et al., 2014). We also note that for cases where a finch locus did not have a predicted model, we provide the chromosomal location where the finch ortholog was identified. For several genes with expression patterns presented in ZEBrA no clear orthologs could be identified in other species. This includes possible novel genes that may be unique to zebra finches or to broader groups like finches, songbirds, or even birds in general. These genes are currently named UNKNOWN in ZEBrA, a terminology that may change as these genes become identified, characterized, and/or named through research efforts.

Figure 2. — (a) Schematic depiction showing how conserved synteny establishes the one-to-one homology of zebra finch and human NEFM. (b-c) Screenshots from UCSC’s genome browser showing how NEFM gene predictions from zebra finch (Ensemble model ENSTGUT00000004667) and human (Refseq NM_005382.2) align to the correct loci with the highest reciprocal alignment scores. (b) Human NEFM (NM_005382.2) aligns to a region of the zebra finch genome containing NEFM (ENSTGUT00000004677.1) and NEFL (ENSGUT00000004664.1). The alignment of human NEFM has a higher scoring alignment, and more completely covers (i.e. higher exon coverage) the NEFM locus, than the NEFL locus (BLAT scores = 1085 vs. 25, respectively). (c) Zebra finch NEFM (ENSTGUT00000004677.1) aligns to a region of the human genome that contains NEFM and NEFL. The alignment of zebra finch NEFM model has a higher scoring alignment, and more complete coverage for the NEFM locus than for NEFL locus (BLAT scores = 1154 vs. 65, respectively).

Clone and cDNA template Selection

Unless otherwise indicated, all probes were derived from the zebra finch brain cDNA collection generated through the Songbird Neurogenomics Consortium (ESTIMA collection, all ESTs have been deposited in GenBank; details in Replogle et al., 2008). For most genes in ZEBrA we selected a single cDNA clone that is fully or partially overlapping and/or contiguous with the predicted Ensembl gene model based on a series of EST reads, or the corresponding model in chicken (via cross-BLAT alignment). Probes generated from clones of this type are labeled ‘Overlaps gene’ in the Gene Info page (example in Figure 3a). In some cases, we selected clones that were non-overlapping, but found to align within 2 kb of the 3’-end of a gene model. Given the close proximity of these reads to the gene model, evidence of polyadenylation (i.e. poly-A tail), and correct orientation to the +/− DNA strand we are reasonably confident that these cDNAs correspond to the 3’-end of the gene of interest. These clones are labeled ‘Downstream of gene (0–2 kb)’ (e.g. see HTR2A locus in Figure 3b). In other cases, the clones met the other criteria above but were located further downstream, and thus are labeled ‘Downstream of gene (2–10 kb)’ (e.g. see KCNS3 locus in Figure 3c). In some rare cases the only available clone was non-overlapping and separated from the 3’-end of the gene by a gap (or in some instances a few gaps) in the genomic sequence. Since the size of these gaps is not known, we could only tentatively assign these clones to a given gene. Such clones are labeled ‘Separated from gene by a gap’ (e.g. see NPY6R locus in Figure 3d). Clones of the latter two types were only used when no higher confidence alternatives were available. We also identified cases with evidence of alternative splicing, resulting in different transcripts with unique exon arrangements. These are labeled ‘Transcript variant’ (see PNLDC1 locus in Figure 3e). Whenever possible we selected clones consisting primarily of 3’-UTR sequence, since compared to coding sequence these regions are less likely to contain conserved domains present in related members of gene families, although we note that several clones contain some stretches of coding sequence. To confirm clone specificity, we systematically performed BLAT alignments of the selected ESTs to the zebra finch genome. The majority of ESTs aligned unambiguously with high scores to a single locus. In a small number of cases an EST aligned to secondary loci, but the secondary alignment scores were typically <50%, indicating a high likelihood of probe specificity. ZEBrA also includes some probes derived from clones that are not associated with a model in the genome (e.g. novel genes, or known orthologs not predicted by Ensembl or RefSeq). In a few cases where no ESTIMA clones were available, we generated template by RT-PCR cloning.

Figure 3. — (a) Example of overlapping probe type: schematic depiction of genomic region for the zebra finch FOXP2 gene (based on UCSC’s genome browser) shows that clone FE716855 (Genbank ID shaded in red) was selected because it overlaps with coding and non-coding (3’-UTR) portions of the FOXP2 Ensembl Gene model (ENSTGUG00000005315.1). The EST for this clone also aligns unambiguously to this locus. Therefore, the Probe Location in the FOXP2 Gene Info page is listed as ‘Overlaps gene’. We note that the FOXP2 gene and related clones are on the minus strand (indicated by arrowheads in the gene model and selected clone alignments). (b-d) Examples of non-overlapping probes according to clone location relative to the gene. In (d), ‘Gap’ refers to a gap present in the genome assembly (taegut1). (e) Example of apparent spliced variants, where only one variant (CK312297) was analyzed. In all panels, Ensembl models are in red, clones selected for *in situs* are in blue, and orientation of gene and related clones are indicated by the arrowheads in the model and/or alternative variant clone. For each gene in ZEBrA the chromosomal location of the selected clone is listed under Probe Location in the Gene Info Page

Probe preparation and in situ hybridization

The steps involved in conducting in situ hybridization on zebra finch brain sections, including the preparation of cDNA templates, synthesis of DIG-labeled riboprobes, hybridization and post-hybridization washes, immunohistological detection of DIG probes, chromagen development, and section imaging and processing are described in detail in Carleton et al., 2014 (see also Lovell et al., 2013; Lovell & Mello, 2011). An overview of the basic pipeline we use to generate data for ZEBrA is presented in Figure 4.

Figure 4. — Schematic representation of ZEBrA’s *in situ* hybridization pipeline. Serial cryostat brains sections encompassing most major brain structures are hybridized at close intervals (e.g. L1–8, in blue) for genes that exhibit highly differential patterns, or at select levels (e.g. L4/L6, in vermillion) for non-differential genes, alignment to reference sections from the histological atlas, and ZEBrA web publishing.

Animal and tissue preparation

All birds used in the preparation of ZEBrA were adult male zebra finches (Taeniopygia guttata) obtained from our own breeding colony or purchased from local breeders. All birds were first isolated overnight (12:12-hour light-dark cycle) in custom-built acoustic isolation chambers to reduce non-specific auditory stimulation. On the following morning (~9:00 AM) birds were monitored briefly (~15 min) to confirm they were awake, not singing and not stimulated by song, and then sacrificed by decapitation. Brains were quickly dissected out, split along the midline, and each hemisphere placed in a plastic mold, covered with Tissue-Tek OCT compound (Sakura-Finetek; Torrance, CA), and frozen in a dry ice/isopropanol bath.

ZEBrA in situ Pipeline

Brains were sectioned on a cryostat (10 μm; Leica CM1850), and two consecutive sections mounted per slide to allow for a replication of hybridization in adjacent sections. To minimize the number of zebra finches required for the project, we used both hemispheres interchangeably and did not systematically compare signal across hemispheres. However, the hemisphere used was coded in metadata that is linked to each slide and hybridization experiment, and can be made available to researchers interested in asymmetric brain gene expression.

For every probe we first run an initial hybridization utilizing our standard protocol, on a pair of test slides at 65°C (hybridization and post-hybridization wash temperatures). Following a 3-day incubation in chromogen (BCIP/NBT), slides were washed, coverslipped, and individually evaluated under brightfield to assess both signal strength and background labeling. With some few exceptions, labeling was determined to be strong when the deposition of chromogen was distributed in a characteristic donut-shaped pattern surrounding a central nucleus devoid of signal. Background was deemed low when the tissue overall, including somata and neuropil, as well as the majority of cells within major fiber fields and/or ependymal cells of the lateral ventricle were devoid of labeling. This initial test allowed us to determine best hybridization conditions and provided an initial assessment of brain distribution patterns. If necessary, before each probe was hybridized to a final set of brain sections to be imaged, optimization steps were run to establish hybridization stringency conditions that gave the highest possible signal and lowest background. For instance, if signal was undetectable or deemed too low for imaging in the first test, the hybridization was repeated at 63°C (lower stringency). If the signal in the first test was deemed strong, but the background was also high, the hybridization was repeated at 67° or 69°C to raise the hybridization stringency. In rare cases where a hybridization at 69° C did not reduce background labeling, an additional RNAse H step was applied after the post-hybridization washes. Genes that showed no detectable signal after several attempts to hybridize at lower stringency are presented without images, and labeled “No Signal Detected”. Some notable exceptions to the pipeline above consist of genes that are actually expressed in the aforementioned control structures. These often represent markers of astroglia, oligodendrocytes, and/or ependymal or pial cells, however in such cases the labeling in these control areas resisted and/or became more distinct under higher stringency conditions. In a few other cases strong expression that resisted high stringency conditions occurred in both the somata/neuropil and control structures, representing ubiquitously expressed genes. We also note a few cases where the labeling pattern had a nuclear localization that resisted high stringency conditions.

Genes that exhibited highly differential expression patterns in identified brain nuclei, divisions, and/or cellular fields were subsequently hybridized to a final series. This consisted of sections at 4–8 levels spaced ~0.2 – 4.0 mm from midline (represented schematically in Figure 4, levels L1-L8 in sky blue), encompassing most of the major brain subdivisions, cellular fields, fiber tracts, and nuclei, including the major telencephalic song nuclei (i.e. HVC, RA, LMAN, Area X) and auditory areas (Field L and adjacent parts of the caudal nidopallium and mesopallium). In contrast, genes with relatively non-differential or flat patterns of expression were hybridized to sections at 2–3 levels spaced between ~1.0 and 2.4 mm from midline, but still containing all major telencephalic song nuclei and auditory areas (represented schematically in Figure 4, levels L4 and L6 in vermillion). Since the various steps above were performed in different brains, we typically examined brain expression for each probe in 2–4 birds, providing replication for the expression pattern presented in ZEBrA. We note that the ZEBrA pipeline incorporated a number of quality control measures as well, including: (1) the use of gel electrophoresis to confirm cDNA template quantity and quality, (2) the inclusion of positive control slides hybridized with GAD2, and negative (i.e. no probe) control slides with each hybridization, (3) visual monitoring of the chromogen precipitation reaction, and (4) the custom selection of individual brain sections for imaging. Further variables and issues that are critical for generating high quality in situ hybridization data during high-throughput efforts like ZEBrA are discussed in Carleton et al (2015).

Image Preparation

Completed slides were imaged with a Nanozoomer 2.0HT digital slide scanner (Olympus; Center Valley, PA) in the Mitra lab at Cold Spring Harbor Laboratory and more recently with a Leica Aperio AT2 scanner at the Knight BioLibrary at OHSU. Images were converted from their native Nanozoomer NDPI file into compressed JPEG format. We next used Photoshop CS5 (Adobe; San Jose, CA) to adjust the contrast, brightness, and color balance of each image, and also to correct for any artifacts introduced during slide processing (e.g. scratches, chromagen precipitate). Additional color-balancing was performed across each image series to ensure that background levels were similar across brain sections. Such manipulations were kept minimal and restricted to image regions not containing labeled cells, in order to avoid any misrepresentation of results. Each image was then rotated to approximately match one of the eighteen levels selected from the series of line drawings derived from the histological atlas of the zebra finch brain (Karten et al., 2013). Finally, each image was tiled using Zoomify (Aptos, CA), which converts the large image format into a series of smaller tilesets that can be reconfigured and viewed like a mosaic. Final image series were then uploaded to the ZEBrA webserver at www.zebrafinchatlas.org (RRID: SCR_012988).

Neuroanatomical Marker Annotation

Once posted to the ZEBrA website, each gene series underwent an additional curation step to qualitatively assess whether the gene could be considered a molecular marker for a given structure or region. Using a combination of annotated brain drawings, information derived from other brain atlases (e.g. Pigeon, Karten & Hodos, 1967; Canary, Stokes et al., 1974; Chick, Kuenzel, 1988; Puelles, 2007; Dove, den Boer-Visser, 2004; Zebra finch, Nixdorf-Bergweiler B.E. & Bischof, 2007), and expression data from genes that define known cell populations (e.g. glutamatergic or GABAergic cells), we scored any structure that showed differential mRNA expression as being upregulated (+) or downregulated (−) relative to the surrounding brain tissue. These results were used to populate a Neuroanatomical Marker table subset that is queried when exploring the Markers of the Song System and Bird Brain Markers portals. Each structure that showed differential gene expression was assigned an appropriate anatomical label. These labels can be viewed when the “Labels ON” button is selected in the navigation cluster on the lower right of the In Situ Image Browser.

Results

The ZEBrA database and website

The Zebra finch Expression Brain Atlas (a.k.a. ZEBrA; www.zebrafinchatlas.org, RRID: SCR_012988) is a public access in situ hybridization database that systematically documents the brain distribution of a large collection of genes in an avian species (Figure 5). It consists of high resolution digital images presented in alignment with an annotated histological brain atlas, together with genomics and bioinformatics resources. While an expanding database, ZEBrA currently contains brain expression data for ~650 genes in adult male zebra finches (as of June, 2019), covering a range of gene families and pathways of relevance for brain function, development, and behavior. The data were generated via our in situ pipeline (Figure 4), using a high-throughput non-radioactive protocol optimized for low background and high sensitivity (Carleton et al., 2014). As our main goal is to determine the constitutive expression of genes, the data derive from awake non-singing and not song-stimulated male zebra finches.

Figure 5. — Image of the homepage of the ZEBrA website (www.zebrafinchatlas.org, RRID: SCR_012988), an online public repository of high-resolution *in situ* hybridization images for a large collection of genes that are expressed in the brain of the adult male zebra finch.

The Reference Atlas and In Situ Images

The Histological Reference Atlas

(http://www.zebrafinchatlas.org/gene_display/histological-atlas) consists of images and corresponding drawings from eighteen brain levels covering the main auditory and vocal areas, as well as major subdivisions and structures of the avian forebrain, brainstem, and cerebellum. The images are derived from a set of alternating serial parasagittal sections that were Nissl-stained to visualize cytoarchitectonic features and landmarks, or myelin-stained to visualize fiber tracts (Karten et al., 2013). The complete set of images in ZEBrA can be visualized in the Histological Atlas Browser by clicking on the Nissl image icon located on the left side of the Image Browser or homepage (Figure 6a). The section icon on the lower left indicates the size, position and type of the main image shown in the browser (blue rectangle) relative to the whole section, helped by the scale bar above the section icon. A cluster of clickable and self-explanatory buttons on the lower right of the browser provides options for navigation. When “Turn labels ON” is active, anatomical structure abbreviations or dots (for small structures) appear over the image (e.g., Figure 6b); the full name for each structure is viewed by hovering over that label. When labels are visible, a hyperlink beneath the image provides access to the Nissl image for that level at the repository website for the images (brainarchitecture.org; Karten et al., 2013), where other parasagittal levels for this brain can also be examined. Images of the alternating Nissl- and myelin-stained sections are also included (e.g., Figure 6c–d), and a colorized version indicating major brain divisions and subdivisions (e.g., Figure 6e) can be viewed by clicking the respective icons to the left within the browser window.

Figure 6. — (a) View of the Browser depicting an example of a drawing from one selected level of the Histological Atlas. The full set of parasagittal levels currently available is shown as thumbnails at the bottom of the browser. The thumbnail icons on the left allow users to toggle across each of the four images types. The position of the section being viewed relative to the midline is illustrated by the orange line on the dorsal-view schematic drawing of the brain, in the top right corner of the bowser. The distance (in mm) of the current section from the midline, as well as the section level in the series, are indicated beneath that schematic. The navigation buttons on the lower right provide options for panning (arrows) or zooming (+ and −) an image, resetting the image zoom level (diagonally opposing arrows), toggling across the four image types for a selected level (circular arrows), requesting an image from ZEBrA (camera symbol), turning anatomical labels ON/OFF, and moving to the next or previous image in the series (>> or <<). The Pan function can also be applied by clicking and dragging the main image in any desired direction, and all other navigation functions can be activated by the appropriate keyboard stroke, as described under the “Navigation Help” tab in the upper-left corner of the browser. The size and position of the field of view shown in the browser is indicated by the blue rectangle over the icon on the bottom left, and the image scale is indicated above that icon. (b) View of the same example drawing as in (a), with Turn Labels button activated (ON position). (c-e) Views of the Nissl- and myelin-stained sections, and of the colorized drawing depicting the: pallium (red), sub-pallium (pink), diencephalon (green), midbrain (yellow), pons (dark blue) and cerebellum (light blue).

The In Situ Image Browser

The In Situ Image browser is ZEBrA’s platform for visualizing and navigating high-resolution in situ hybridization images. The navigation features are similar to those of the Histological Atlas Browser, except that drawings cannot be panned or zoomed, and the corresponding Nissl and myelin series are not presented. For example, an in situ image showing ADAM23 reveals that this gene is highly enriched in nucleus LMAN (Figure 7a). This image can be zoomed by double-clicking the image or using the navigation tools on the lower right to allow browsing down to a cellular level of resolution (Figure 7b–d). Each image is presented with a drawing from the Histological Reference Atlas that represents the closest reference level for that image and provides guidance for interpreting gene expression patterns. Toggling between the in situ image and the corresponding reference drawing (e.g. Figure 7e for ADAM23, at the level shown in Figure 7a) aids in data interpretation by displaying which major brain divisions, subdivisions, nuclei, and cytoarchitectonic features such as fiber tracts and laminae are present at the level of the imaged section. An additional colorized schematic providing information about the major brain divisions and subdivision present in the section is presented to the left of the drawing (see inset in Figure 7e, expanded view in Figure 7f). Since the in situ images and the reference drawings are from different brains, the latter cannot provide exact matches for the in situ images, and are meant instead as guides to the approximate brain level and structures in each in situ image. If further anatomical details are needed, the Nissl image icon on the left provides direct access to the full Histological Atlas Browser, which opens as a separate window. Multiple gene series can be opened in dedicated windows, thus by design images from different genes can be viewed simultaneously and in parallel as adjacent windows on the user’s screen.

Figure 7. — (a) View of the Browser depicting an *in situ* image from an example gene (ADAM23), with Turn Labels button activated. The set of icons at the bottom of the browser indicate the levels of the *in situ* images available for this gene, and the highlighted icon indicates the level of the enlarged image at the center of the browser. The position of the section being viewed relative to the midline is illustrated by the orange line on the dorsal-view schematic drawing of the brain, in the top right corner of the bowser. The distance (in mm) of the current section from the midline, as well as the section level in the series, are indicated beneath that schematic. The navigation buttons on the lower right provide options for panning (arrows) or zooming (+ and −) an image, resetting the image zoom level (diagonally opposing arrows), toggling between the *in situ* image and the corresponding drawing from the histological reference atlas (circular arrows; note that the atlas drawing cannot be zoomed from this browser), requesting an image from ZEBrA (camera symbol), turning anatomical labels ON/OFF, and moving to the next or previous image in the series (>> or <<). The Pan function can also be applied by clicking and dragging the main image in any desired direction, and all other navigation functions can be activated by the appropriate keyboard stroke, as described under the “Navigation Help” tab in the upper-left corner of the browser. The size and position of the field of view shown at the center of the browser is indicated by the blue rectangle over the icon on the bottom left, and the image scale is indicated above that icon. The icon above the scale provides access to the histological atlas browser, which opens in another window. (b-d) Enlarged views of the areas indicated by the red boxes in (a-c), corresponding to the core region of song nucleus LMAN (LMANco). Scale bars: (b): 100 μm; (c): 50 μm, (d): 25 μm. (e) View of the reference drawing (with labels ON) associated with the *in situ* image in (a) depicts specific brain structures present at the level of that *in situ* image. (f) View of the corresponding colorized drawing depicts the position of general brain subdivisions present at that level.

Each gene has been annotated according to its ability to define specific brain regions and nuclei (e.g., ADAM23 is highly enriched in the core region of LMAN, Figure 7a). Only structures for which a given gene is considered a marker (up or down differential expression compared to the surrounding brain areas) are labeled in that gene’s image browser. These marker annotations are also accessible collectively, with the complete anatomical gene marker sets presented in the ‘Markers of the Song System’ portal and/or in the ‘Bird Brain Markers’ sub-portal within the ‘Comparative Neuroanatomy’ portal (discussed below). Overall, the In Situ Image Browser, integrated with the Histological Reference Atlas, allows for detailed explorations of gene expression images documenting molecular specializations of broad avian brain subdivisions, and/or of nuclei that comprise specific avian brain circuits, with emphasis on the vocal control and learning circuitry.

The Gene Info Page

Every gene in ZEBrA is presented along with a Gene Info Page. This page provides probe and genomic information, as well as links to various metadata. Specifically, the ‘Gene/Probe Information’ tab (e.g., FOXP2 Gene Info Page in Figure 8a) includes gene symbols and descriptions, NCBI and Ensembl entries to finch and human orthologs, and the chromosomal location of genes and probes. Aided by these resources, the orthology of each gene in ZEBrA has been confirmed by reciprocal top alignments and synteny verification in comparison with humans, and whenever necessary non-avian out-groups (see Methods for a description of gene orthology verification, and NEFM example in Figure 2). The probe location entry in the ‘Gene/Probe Information’ tab reflects the confidence level for each selected probe. Confidence is highest for clones that overlap the gene model or located within a short distance of that model, but decreases for clones that are far away or separated from the model by a gap (see Methods for full description of probe types in Figure 3). Transcript cases representing apparent splice variants are indicated when relevant. Probes in the low confidence categories were included only when no higher confidence alternatives were present.

The ‘Gene Attributes’ tab provides counts and lists of all attributes for a given gene (e.g., GRIN2B in Figure 8b). These attributes link genes to their inclusion in the six major thematic portals that are presented in ZEBrA, and provide valuable information about a gene’s association with specific functions, diseases or phenotypes, and marked expression in song circuitry and/or brain areas in both birds and mammals. These data, in turn, facilitate discovering new relationships between gene expression patterns and functional attributes. A description of the content and organization of these portals, and linking of genes to attributes is discussed below. Of note, on the Gene Info page each attribute is associated with a hyperlink that provides access to the full list of genes that share that attribute. This page thus allows systematic explorations of how genes relate to each other by sharing attributes across portals.

Navigating the ZEBrA website and main resources

The ZEBrA website is organized into six thematically organized portals that were designed to facilitate website navigation and user access to images and related resources. These portals are readily accessible from the homepage beneath a short summary of ZEBrA contents (Figure 5). The navigation tools to explore these portals and other features consist of self-explanatory clickable icons and buttons, and when necessary pop-up menus with user instructions. For example, mousing over portal images and clicking the ‘GO’ button provides access to thematic drop-down menus with clickable arrowheads, which in turn provide access to alphabetical gene lists (e.g. Figure 9a). Within a given list, the ‘magnifying glass’ icon provides access to the In Situ Image Browser, the gene name provides access to the ‘Gene info’ Page, the ‘Attributes’ column informs on associations with themes or sub-themes in all portals (e.g. Figure 9b; details below), and the column on the right provides linkouts to databases or references, or informs the direction of differential regulation. Some lists contain multiple entries for the same gene, reflecting multiple entries and linkouts in the respective human or rodent databases.

Figure 9. — (a) View of the organization and layout of a typical list of genes associated with a specific ZEBrA portal theme. In this example, a subset of genes associated with the Neurological/Psychiatric Disorder Epilepsy within the Human Diseases sub-portal is shown. Each gene entry includes a magnifying glass icon under the View *In Situ* column (which provides access to *in situ* images in the Image browser), and entries for gene name (which provides access to the Gene Info page), Gene Description, Attributes, and Linkouts (which in this case provide access to an entry in OMIM). (b-c) Views of example pop-up menus showing Attributes associated with the gene RELN that are derived from the Disease and Phenotypes (b) and Speech and Language (c) portals. The association of RELN with the OMIM Attribute “Epilepsy, familial temporal lobe, 7” provides the reasoning behind the classification of this gene under the Human Disease theme of Epilepsy.

Through buttons on the left sidebar of the homepage (Figure 5), or the top navigation bar in each portal, users can access alphabetical gene lists, tools for gene searches by name or description, and the histological reference atlas. All portals can be accessed from within any portal through a drop-down menu on the right side of the navigation bar. The menu bar on the top right provides information on the construction and usage of ZEBrA’s database and website.

ZEBrA portals

The contents of each portal are described below; the current gene numbers in each portal are presented in Table 1. The majority of genes in ZEBrA are part of at least one portal, and are typically presented in multiple portals. A small number of genes are not in any portal but can be accessed through the A to Z index or by gene name search.

Table 1.

Numbers of genes presented in major themes or categories within the six ZEBrA portals.

# Genes	Portal/Theme/Category
Portal 1: Markers of the Song System

184	HVC
140	RA
122	MMAN / LMAN
61	Area X
78	DLM
13	DM
14	Nif
Portal 2: Speech and Language

52	Speech / Language Disorders
57	Autism (13 OMIM entries; 53 SFARI entries)
41	FOXP2 targets
36	Language Readiness Hypothesis
57	Shared Markers Between Vocal Learners
Portal 3: Diseases and Phenotypes

212	Human Diseases (326 OMIM entries)
262	Mouse Phenotypes (511 MGI phenotypes)
Portal 4: Comparative Neuroanatomy

451	Bird Brain Markers
222	Mouse Brain Markers (Mouse Brain Atlas entries)
Portal 5: Gene Function

681	Gene Function
Portal 6: Songbird Discoveries

80	Highlighted ZEBrA-based Discoveries
641	Songbird Studies

Open in a new tab

Abbreviations: DLM, medial part of the dorsolateral nucleus of the anterior thalamus; DM, medio-dorsal nucleus of the intercollicular complex; HVC, HVC proper name; LMAN, lateral magnocellular nucleus of the anterior nidopallium; MGI: Mouse Genome Informatics; NIf: nucleus interfacialis; OMIM: Online Mendelian Inheritance in Man; RA, robust nucleus of the arcopallium; SFARI, Simons Foundation Autism Research Initiative; X, Area X.

(1). Markers of the Song System:

This portal consists of genes that are differentially up-or down-regulated in at least one song control nucleus of adult male zebra finches compared to adjacent brain areas, based on visual assessment of in situ images. Included are markers for most nuclei in both major pathways of the song system (examples in Figure 10). We note the broad diversity of patterns, density of labeling, and contrasts with adjacent areas. Most of these genes were initially identified as candidate song nuclei markers through microarray screenings (Lovell et al., 2008; Lovell et al., 2018a; Pfenning et al., 2014), and in many cases provided stringent and objective criteria for distinguishing true vs. false positives and negatives in these screenings (Lovell et al., 2018a). Other subsets were initially included in other ZEBrA portals but later identified as markers of song system nuclei. Overall, they represent highly validated sets of molecular specializations that are likely linked to the unique anatomical, neurochemical and physiological properties of the song system. Song nuclei markers display broad ranges of labeling patterns at the cellular level (examples in Figure 11). While some markers are strongly expressed in high proportions of cells within various song nuclei (e.g. in Figure 11: CAMTA1 in HVC, CACNB4 in Area X), others label distinctly sparse populations of different sizes and shapes (e.g. in Figure 11: GAD2, CADPS2, UTS2B, and UGT8 in HVC; CALB2 and PLS3 in RA, etc.), suggesting that they might correspond to discrete and distinct cell types within these nuclei. Overall, genes in this portal constitute molecular specializations of song nuclei, but they also serve as markers that can help to define nuclear boundaries, allowing for greater precision in studies of nucleus size, shape, and cell type composition. Furthermore, while the differential expression in song nuclei has been previously reported for some markers (e.g. Hilliard et al., 2012b; Kato & Okanoya, 2010; Kubikova et al., 2010; Lovell et al., 2013; Lovell et al., 2008; Lovell et al., 2018a; Pfenning et al., 2014; Wada et al., 2004; Whitney et al., 2014; Wirthlin et al., 2014; see also Songbird Discoveries portal), ZEBrA provides extensive documentation of the patterns across all song nuclei and throughout the brain.

Figure 10. — Upper left: Schematic drawing of a parasagittal view of the adult male zebra finch brain depicts major brain subdivisions and the relative location of major song control nuclei; nuclei of the direct vocal-motor pathway are in black, nuclei of the anterior forebrain pathway are in white, colored dashed rectangles indicate the areas shown in the *in situ* hybridization images. Other panels: Examples of *in situ* hybridization images of positive or negative markers (up- or down-regulated expression compared to adjacent areas) of zebra finch song nuclei HVC (blue), RA (vermillion), LMAN (yellow), Area X (orange), DLM (reddish purple), DM (green) and NIf (sky blue). We note that various markers of LMAN specifically label the core (LMANco), shell (LMANsh), or both core and shell regions of LMAN. DM: mediodorsal nucleus of the intercollicular complex; NIf: nucleus interfacialis; N: nidopallium; St: Striatum; †: Unknown region of the dorsal thalamus surrounding nucleus DLM; ‡: Unknown region of the dorsal thalamus likely corresponding to the DLM defined by Wada et al., 2004. For other abbreviations, see Figure 1 legend. Scale bars for LMAN, DLM, and Area X = 200 μm; for all other nuclei = 100 μm.

Figure 11. — Upper left: Schematic drawing of a parasagittal view of the adult male zebra finch brain depicts major brain subdivisions and the relative location of major song control nuclei; nuclei of the direct vocal-motor pathway are in black, nuclei of the anterior forebrain pathway are in white, colored dashed rectangles indicate the areas shown in the *in situ* hybridization images. Other panels: Examples of *in situ* hybridization images of markers that show high density of labeling (left panels for each nucleus) vs sparse labeling (other panels) for song nuclei HVC (blue), RA (vermillion), LMAN (yellow), NIf (sky blue), Area X (orange), and DLM (reddish purple). For abbreviations, see legends of Figures 1 and 11. Scale bars: 50 μm.