The Songbird Neurogenomics (SoNG) Initiative: Community-based tools and strategies for study of brain gene function and evolution

1471-2164-9-131 1471-2164 Research article The Songbird Neurogenomics (SoNG) Initiative: Community-based tools and strategies for study of brain gene function and evolution Replogle Kirstin replogle@igb.uiuc.edu Arnold P Arthur arnold@lifesci.ucla.edu Ball F Gregory gball@jhu.edu Band Mark markband@uiuc.edu Bensch Staffan staffan.bensch@zooekol.lu.se Brenowitz A Eliot eliotb@u.washington.edu Dong Shu shudong@uiuc.edu Drnevich Jenny drnevich@uiuc.edu Ferris Margaret mferri3@uiuc.edu George M Julia j-george@uiuc.edu Gong George ligong@uiuc.edu Hasselquist Dennis dennis.hasselquist@zooekol.lu.se Hernandez G Alvaro aghernan@uiuc.edu Kim Ryan wrkim@express.cites.uiuc.edu Lewin A Harris h-lewin@uiuc.edu Liu Lei leiliu@scbit.org Lovell V Peter lovellp@ohsu.edu Mello V Claudio melloc@ohsu.edu Naurin Sara Sara.Naurin@zooekol.lu.se Rodriguez-Zas Sandra rodrgzzs@uiuc.edu Thimmapuram Jyothi jyothit@uiuc.edu Wade Juli wadej@msu.edu Clayton F David dclayton@uiuc.edu

Cell & Developmental Biology, Univ. of Illinois, Urbana, IL, USA

Physiological Sci., UCLA, Los Angeles, CA, USA

Psychological & Brain Sci., Johns Hopkins Univ., Baltimore, MD, USA

Psychology, Biology, and Bloedel Hearing Research Center, Univ. of Washington, Seattle, WA, USA

Mol. & Integrative Physiology, Univ. of Illinois, Urbana, IL, USA

Neurological Sci. Inst., Oregon Hlth. Sci. Univ., Beaverton, OR, USA

Psychology, Zoology & Neuroscience, Michigan State Univ., East Lansing, MI, USA

W.M. Keck Center for Comparative & Functional Genomics, Univ. of Illinois, Urbana, IL, USA

Institute for Genomic Biology, Univ. of Illinois, Urbana, IL, USA

Animal Sciences, Univ. of Illinois, Urbana, IL, USA

Animal Ecology, Lund University, S-223 62 Lund, Sweden

Neuroscience Program, Univ. of Illinois, Urbana, IL, USA

BMC Genomics 1471-2164 2008 9 1 131 http://www.biomedcentral.com/1471-2164/9/131 18366674 10.1186/1471-2164-9-131 15 10 2007 18 3 2008 18 3 2008 2008 Replogle et al; licensee BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Abstract

Background

Songbirds hold great promise for biomedical, environmental and evolutionary research. A complete draft sequence of the zebra finch genome is imminent, yet a need remains for application of genomic resources within a research community traditionally focused on ethology and neurobiological methods. In response, we developed a core set of genomic tools and a novel collaborative strategy to probe gene expression in diverse songbird species and natural contexts.

Results

We end-sequenced cDNAs from zebra finch brain and incorporated additional sequences from community sources into a database of 86,784 high quality reads. These assembled into 31,658 non-redundant contigs and singletons, which we annotated via BLAST search of chicken and human databases. The results are publicly available in the ESTIMA:Songbird database. We produced a spotted cDNA microarray with 20,160 addresses representing 17,214 non-redundant products of an estimated 11,500–15,000 genes, validating it by analysis of immediate-early gene (zenk) gene activation following song exposure and by demonstrating effective cross hybridization to genomic DNAs of other songbird species in the Passerida Parvorder. Our assembly was also used in the design of the "Lund-zfa" Affymetrix array representing ~22,000 non-redundant sequences. When the two arrays were hybridized to cDNAs from the same set of male and female zebra finch brain samples, both arrays detected a common set of regulated transcripts with a Pearson correlation coefficient of 0.895. To stimulate use of these resources by the songbird research community and to maintain consistent technical standards, we devised a "Community Collaboration" mechanism whereby individual birdsong researchers develop experiments and provide tissues, but a single individual in the community is responsible for all RNA extractions, labelling and microarray hybridizations.

Conclusion

Immediately, these results set the foundation for a coordinated set of 25 planned experiments by 16 research groups probing fundamental links between genome, brain, evolution and behavior in songbirds. Energetic application of genomic resources to research using songbirds should help illuminate how complex neural and behavioral traits emerge and evolve.

Background

Songbirds offer unique opportunities for studying the links between genome, brain structure, neurophysiology, behavior and evolution. They are one of the most successful vertebrate radiations, diversifying into more than 4000 different species over the past ~65 million years 12 and displaying a great range of physical, behavioral and social adaptations 34. Many songbird species are readily observed in their natural habitats and several species have been domesticated (e.g., zebra finch, canary). All songbirds share a highly specialized neural system for learned vocal communication. Indeed, songbirds are one of the few animal groups capable of complex vocal learning and stand alone as accessible experimental models relevant to human speech 56. Neurobiological research using songbirds has consistently generated new insights that were later found to be true for other vertebrates; examples include definitive evidence for sexual differentiation of brain circuits, seasonal changes in brain anatomy, and lifelong neuronal replacement 7.

Against these opportunities, the songbird as a research model also presents challenges. Although the zebra finch has emerged as a primary focus, many other songbird species are studied especially in natural contexts. It will be important to develop tools and reagents that allow study not just of one species (the zebra finch) but of many, so as to exploit the diversity of vocal behavior shown by songbirds. Some of these studies will be comparative in nature (e.g., how is the genome different in species with lifelong vocal learning, compared to species that learn song only once?). Other research objectives may focus on a single wild-caught species to address questions of gene regulation and variation in a particular niche, environment or behavioral paradigm. A related challenge is that researchers with expertise in field biology, neuroscience, physiology or behavioral analyses may not necessarily have equal expertise in molecular genetics, and vice versa.

In response to these needs and opportunities, in 2002 we inaugurated the Songbird Neurogenomics (SoNG) Initiative 89. Our goals were to leverage rapid advances in genomic technology, so as to bring new resources to songbird research and stimulate collaborative approaches that would integrate field researchers, physiologists and molecular biologists. Here we describe the following accomplishments, which form the foundations of the ongoing SoNG Initiative: 1) three generations in the development of an evolving catalog of gene sequences expressed in the zebra finch brain; 2) production and validation of DNA microarrays useful for study of diverse songbird species; 3) organization of an efficient community-based mechanism for stimulating use of these resources.

Results

High-Throughput Expressed Sequence Analysis and Annotation

Over the course of five years, we generated three sequential assemblies of expressed sequence information, each one incorporating more data drawn from both our own primary sequencing efforts and from other zebra finch research groups working in parallel (Table 1). For efficient generation, annotation and presentation of expressed sequence information, we made use of the ESTIMA software interface 10 and a production pipeline that had been refined in development of resources for cattle 11, honeybees 12 and other emerging research organisms 13. In the first phase of the project, we applied high-throughput DNA sequencing methodologies to generate a collection of Expressed Sequence Tags (ESTs) from a normalized zebra finch brain cDNA library. The resulting sequence information was assembled (ESTIMA:Songbird Build 1), made publicly available via the internet 9 in 2004, and used to build the cDNA array described later in this report. In the second phase of the project, we incorporated sequence information obtained from a sister project 14. The sequences were re-assembled (ESTIMA:Songbird Build 2, replacing Build 1 on the SoNG Initiative website in 2005) and used in fabrication of an Affymetrix array as described 15. In the third phase, we incorporated additional data from subtractive sequencing of our original cDNA library and data from another sister project 16 (ESTIMA Build 3, replacing Build 2 in October 2007).

Table 1

Sequence sources and statistics for the three sequential ESTIMA:Songbird assemblies

ESTIMA:Songbird

BUILD 1

BUILD 2

BUILD 3

Sequence Source

Raw reads

After filtering

20 K cDNA microarray

Lund/Affy microarray

SB01

2,880

1,840

√

SB02

18,720

16,360

√

SB03 (a)

19,104

17,032

√

SB03 (b)

9,120

5,717

√

Duke chromatograms

36,469

23,225

√

Duke from NCBI

17,312

16,442

√

SB05

11,328

10,056

√

SB06

11,040

9,760

√

Rockefeller

9,845

9,764

√

Raw reads

40,704

86,293

99,349

# SEQS INTO PTA

35,232

58,457

86,784

Pure singlets

10,123

11,823

13,908

Clustered singlets

1,900

2,457

3,662

Total singlets

12,023

14,280

17,570

Contigs

5,855

8,348

14,088

Total unique sequences

17,878

22,628

31,658

Total redundant sequences

17,354

35,829

55,126

Redundancy

49%

61%

64%

ESTIMA:Songbird Build 1

For the primary EST assembly used in production of the SoNG 20 K array (below), cDNA clones were 5'-end-sequenced using high-throughput methods. 35,232 filtered high-quality ESTs were clustered and assembled into 17,878 non-redundant sequences (Table 1), comprising 10,120 pure singlets, 5,855 contigs and an additional 1,903 clustered singlets (i.e., some sequence relationships in initial cluster analysis but could not be assembled into unbroken contigs; possible splice variants). In subsequent EST builds and BLAST analyses (below), 2900 of the Build 1 sequences were absorbed into larger contigs or aligned against redundant gene targets in other species, and 3409 failed to align anywhere (including the whole genome assembly of the chicken). Thus we estimate Build 1 and the corresponding SoNG 20 K microarray represent products coming from between 11,569 and 14,978 unique genes, with the uncertainty depending on the origin of the 3409 ESTs that cannot yet be aligned with a known orthologous genomic region.

ESTIMA:Songbird Build 2

The ESTIMA:Songbird database was designed to be a dynamic repository for zebra finch cDNA sequences in an evolving collection. A complementary initiative centered at Duke University emphasized the isolation of full-length cDNA clones from libraries representing a diversity of zebra finch brain states 14. The Duke investigators provided us with sequencing chromatograms from their project, and we combined this with our SB01-02-03 sequence data to produce a single assembly representing 58,457 filtered high-quality reads and 22,628 assembled unique sequences (Table 1). This assembly was used as the basis for design of a short-oligo Affymetrix array in a parallel project at Lund University 15 (see below). To estimate gene representation empirically, we performed a test BLAST search for 61 orthologs of specific interest across a range of functional categories (Table 2). Forty-one (~67%) were present in Build 2, including representatives of each of the functional categories of interest. However, numerous gaps in coverage remain, as shown for example by absence of several important steroid receptors and channel proteins in the assembly.

Table 2

ESTIMA:Songbird Build 2, presence ("yes") or absence ("no") of specific sequences of interest assessed by direct BLAST with available avian or mammalian sequences against 58,211 sequences in the ESTIMA database as of June 2005.

Yes

Steroid Receptors

Extracellular Signaling

GABA transmission

Androgen Receptor

•

BDNF

√

GABA-A alpha 1

√

estrogen receptor alpha

•

cannabinoid receptor

•

GABA-A alpha 2

√

estrogen receptor beta

•

IGF-2

√

GABA-A alpha 4

•

estrogen receptor gamma

√

N-CAM

√

GABA-A alpha 5

√

RARalpha

√

GABA-A beta 3

√

RARbeta

√

Intracellular signaling

GABA-A delta

√

RARgamma

•

Calbindin

√

GABA-A gamma 2

√

RXRalpha

√

CAMKII

√

GABA-A gamma 3/4

√

RXRbeta

•

Canarygranin

√

GABA-B

√

RXRgamma

•

CRABPI

√

GABA-C rho-2B

•

Steroid Metabolism

CRABPII

√

GAD-65

√

Aromatase

√

CRBPI

•

GAD67

√

P450-CYP26

•

CRBPII

•

Potassium channels

zRalDH

√

GAP-43

√

Kcnma1

•

Immediate Early Genes

MAPKK

√

Kcnmb1

•

ARC

√

Narp

√

Kcnmb2

•

c-fos

√

parvalbumin

√

Kcnmb3

•

c-jun

√

PAX-6

√

Kcnmb4

•

zenk

√

SNAP-25

√

Kv3.1

√

Cytoskeleton

Synapsin I

•

Transporters

Beta Tubulin

√

synapsin IIb

√

VGAT

•

GFAP

√

VGLUT

•

NFM

√

VIAAT

√

vimentin

√

ESTIMA:Songbird Build 3

In a third build of the database, we incorporated two more rounds of subtractive sequencing plus additional sequences from colleagues at Rockefeller University 16. The combined total of high-quality filtered reads now reached a total of 86,784, which assembled into 31,658 non-redundant transcripts (Table 1). Having reached a marginal redundancy of 64%, we suspended further sequencing of this cDNA source. However, the frequency of singletons remains high (~55%) indicating considerable sequence diversity still remains unrepresented.

Annotation

We performed machine annotations of each of the sequential ESTIMA builds by BLAST sequence similarity searches against external databases, making the results publicly available through the ESTIMA:Songbird website 9. For Build 3, the BLAST analyses aligned 77% of the assembled unique sequences to putative orthologs in other vertebrates (Table 3A). For nomenclature, we used the International Protein Index (IPI) of the European Bioinformatics Institute (EBI) as a primary reference point 17. IPI is a database of cross references between the primary data sources, providing a minimally redundant yet maximally complete set of protein sequences and stable identifiers for featured species (one sequence per transcript). IPI also maps to a growing set of qualified Gene Ontology (GO) terms for its sequence entries. Searching against chicken IPI, we obtained alignments for 13,219 (42%) of our assembled Build 3 sequences (Table 3B).

Table 3

Annotation statistics for ESTIMA:Songbird Build 3

Build 3

Percent total

20 K array subset

Percent of array

Percent of Build 3 category

Unique sequences after assembly

31658

100%

17214

100%

54%

A) ALL HITS BY DATABASE

Gga Genome

21601

68%

12396

72%

57%

Chicken_TC

20208

64%

11715

68%

58%

Gga Unigene

17980

57%

10490

61%

58%

NCBI_Chicken_RNA

15904

50%

9110

53%

57%

Ensembl_Chicken_cdna_all

14609

46%

8348

48%

57%

Ensembl_Chicken_cdna_abinitio

13223

42%

7588

44%

57%

Chicken IPI

13219

42%

7898

46%

60%

Hs Unigene

12373

39%

7224

42%

58%

NCBI_Chicken_protein

7776

25%

4517

26%

58%

Hits against any database

24466

77%

13803

80%

56%

B) Hierarchy of CUSTOM ANNOTATION

1 Use IPI-annotation*

13219

42%

7553

44%

57%

2 in GGA_Unigene but not IPI

5614

18%

3364

20%

60%

3 in HS but not IPI or GGA_unigene

265

104

39%

Total number of "custom annotations"

19098

60%

11021

64%

58%

4 additional "conserved in chicken"

5368

17%

2784

16%

52%

Hits against any database

24466

77%

13805

80%

56%

5 remainder = "TGU-specific" (no hits)

7192

23%

3409

20%

47%

31658

100%

17214

100%

54%

*C) IPI Annotations

Number of sequences with IPI identifiers

13219

7898

(Unigene used for 345)

Number of unique IPI identifiers

8127

6035

All identifiers in IPI release 3.26

25500

Fraction of total IPI identifiers

32%

24%

In addition to the 13,219 sequences that aligned with IPI records, another 5614 found matches against the chicken Unigene database but not IPI, and a further 265 aligned against Human Unigene but neither Chicken Unigene nor IPI (Table 3C). In sum, we were able to assign "custom annotation" names for 19,098 sequences, or 60% of the Build 3 assembly. A further 17% showed significant alignment to the chicken genome at sites with no current functional gene annotation, and the remaining 23% have not yet been aligned to any putative orthologous genomic region.

Microarray Production and Validation

For gene expression studies, two different DNA microarrays were produced from the first two ESTIMA:Songbird builds (Table 1). After ESTIMA:Songbird Build 1, the sequenced cDNA inserts were PCR-amplified and spotted onto a glass microarray; for contigs, a single representative clone was spotted (Methods). In addition to these clones, we also added 36 previously characterized cDNA clones as controls on the array (Tables 4 and 5). The total number of addresses on each array = 20,215 (hence we refer to this as the

SoNG 20 K

array). An annotation file for this array based on the Songbird3 Assembly is presented as Additional File 1. From ESTIMA:Songbird Build 2, a separate short oligonucleotide array (Lund-zfa
array) was produced using sets of 25 mer "perfect match" probes designed by Affymetrix 15 (additional information about this array is provided in Methods). We performed three detailed analyses to assess the performance of these microarrays to validate their use for specific research objectives.

Table 4

Content of SoNG 20 K array

Unique sequences in SB1 assembly

17878

Unique in SB3 assembly

17214

Redundant in SB3 contigs

664

Controls and knowns

368

Spots that replicate SB clones

967

Total DNA spots on array

19213

Blanks and buffer spots

950

Total addresses on array

20160

Table 5

Previously Cloned Genes and Replicated Controls spotted on 20 K Array

Seq ID (Gal file)

Provided by

Genbank Accession

Notes

Art Arnold

Androgen receptor

ARC

Claudio Mello

beta actin

SB02028A1F09.f1

CK311416

Replicated control

CAB

Keck¹

BE190670

Replicated control (soybean)

ER alpha

Art Arnold

Estrogen Receptor alpha

ER beta

Art Arnold

Estrogen Recetor beta

GAPDH

SB02023A1H08.f1

CK304281

Replicated control

Histone H3

SB02008B2G06.f1

CK315888

Replicated control

KA1

K. Wada

AB107130

Wada et al.²

KA2

K. Wada

AB107131

Wada et al.²

MSG

Keck¹

AJ239127

Replicated control (soybean)

RAR alpha

Claudio Mello

AY714582

Retinoic Acid Receptor

RAR beta

Claudio Mello

AY714583

Retinoic Acid Receptor

RAR gamma

Claudio Mello

AY714584

Retinoic Acid Receptor

RUB

Keck¹

AI495218

Replicated control (soybean)

zenk

David Clayton

EF050732

Replicated control (canary "e12" clone)

zf GluR1

K. Wada²

AB042749

AMPA subunit, Wada et al.²

zf GluR2

K. Wada²

AB042750

Wada et al.²

zf GluR3

K. Wada²

AB042751

Wada et al.²

zf GluR4

K. Wada²

AB042752

Wada et al.²

zf GluR5

K. Wada²

AB107127

Kainate, Wada et al.²

zf GluR6

K. Wada²

AB107128

Wada et al.²

zf GluR7

K. Wada²

AB107129

Wada et al.²

zf mGluR1

K. Wada²

AB042753

Metabotropic, Wada et al.²

zf mGluR2

K. Wada²

AB042754

Wada et al.²

zf mGluR3

K. Wada²

AB107132

Wada et al.²

zf mGluR4

K. Wada²

AB042755

Wada et al.²

zf mGluR5

K. Wada²

AB107133

Wada et al.²

zf mGluR8

K. Wada²

AB107134

Wada et al.²

zf NR1

K. Wada²

AB042756

NMDA, Wada et al.²

zf NR2A

K. Wada²

AB042757

Wada et al.²

zf NR2B

K. Wada²

AB107125

Wada et al.²

zf NR2C

K. Wada²

AB042758

Wada et al.²

zf NR2D

K. Wada²

AB042759

Wada et al.²

zf NR3A

K. Wada²

AB107126

Wada et al.²

zf ZENK

Claudio Mello

ZENK cDNA clone ZZF23

¹contributed to Keck Center by Lila Vodkin ²Wada K et al, JCN 476:44-64 (2004)

Additional File 1

SoNG20K. Annotation of the SoNG 20 K array. standard GAL file format describing each spot on the 20 K microarray, 20215 rows × 6 columns. Rows 1–55 describe block organization. Rows 56-20215 describe the 20160 features (spots) on the array (see Table 4). Columns 1–3 define feature position. Column four gives the unique identifier for the feature. Column five gives the "custom annotation" name for that spotted DNA, based on the Build 3 annotation (described in Tables 3). Column six identifies the source of the information used for that custom annotation.

Click here for file

ZENK Hybridization Analysis (SoNG 20 K array)

One of the most-studied examples of differential gene expression in the songbird brain is the induction of the immediate early gene zenk in the auditory lobule (AL) of the telencephalon following presentation of song playbacks 1819202122232425262728. We used this phenomenon as the basis for a primary test of the 20 K microarray. The original canary cDNA clone for study of zenk expression in both canaries and zebra finches 20 was spotted redundantly on the array as a replicated internal control. Additionally, a near-full-length cDNA from the zebra finch was spotted once on each array as one of the "known" genes. Fortuitously, three other cDNAs representing distinct parts of the transcript were also present on the array, as they did not form contigs in our EST sequence assembly (Fig. 1). Thus the microarray contains 5 different probes spanning the zenk mRNA and also representing the sequences of two different songbird species.

Figure 1

Schematic alignment of zenk sequences on array

Schematic alignment of zenk sequences on array. Five partial zenk cDNAs are spotted on the 20 K microarray. Here they are aligned against the derived contig ("mRNA") and its predicted protein coding sequence ("CDS"). Canary e12 cDNA [Genbank: EF050732] is the original songbird zenk sequence [20]. Zebra finch ZZF23 cDNA [Genbank: EF052676] was isolated using the canary sequence to probe a zebra finch cDNA library (C.V. Mello et al., unpublished). The three clones beginning in "SB..." are three independent cDNAs that did not form contigs and were spotted independently on the array; sequence extent is based on raw EST sequence. Scale bar in kb from putative 5' end of mRNA.

For this first experiment we used 6 microarrays and a simple direct-comparison design, where each array was hybridized with two probes from a different pair of zebra finches. For each pair, one bird heard a 30 minute song stimulus and the other bird heard only silence. They were immediately euthanized and total RNA was extracted from the dissected ALs. We analyzed the hybridization data using a two-stage parametric linear modeling approach (Methods). Data for the five zenk-hybridizing probes are summarized in Table 6.

Table 6

zenk-hybridizing sequences on the 20 K microarray (ordered as in Figure 1).

Description

Fold Increase ^a

Raw p ^b

FDR ^c

Mean intensity ^d

Canary e12 cDNA

Replicated control

2.0

1.5E-59

0.00

3661

Zebra finch cDNA

Single-spotted control

2.1

2.1E-03

0.16

3602

SB02047B2F12.f1

5'-most EST

1.2

1.0E-01

0.35

9005

SB03008A1H11.f1

Central EST

1.7

3.6E-04

0.15