JARO 03: 140–148 (2001) DOI: 10.1007/s101620020029 An Inner Ear Gene Expression Database ZHENG-YI CHEN,1,2 AND DAVID P. COREY3–5 1Neurology Service, Massachusetts General Hospital, Boston, MA 02114, USA 2Department of Neurology, Harvard Medical School, Boston, MA 02115, USA 3Neurosurgery Service, Massachusetts General Hospital, Boston, MA 02114, USA 4Department of Neurobiology, Harvard Medical School, Boston, MA 02114, USA 5Howard Hughes Medical Institute, Chevy Chase, MD 20815-6789, USA Received: 30 May 2001; Accepted: 10 August 2001; Online publication: 18 October 2001 ABSTRACT INTRODUCTION Microarray technology has provided an unprece- The development and the function of the inner ear are products of the genes expressed, the timing of their dented opportunity to study gene expression profiles expression, and the functional pathways in which they at a whole-genome level. As a first step toward a com- operate. Knowing the pattern of gene expression in prehensive understanding of inner ear gene expres- the inner ear will help in understanding development sion, mouse cochleas were examined at two devel- and function. Knowledge of genes expressed, in combi- opmental stages (P2 and P32) using GeneChip nation with their chromosomal locations, will also assist oligonucleotide arrays. A large number of genes and in identifying genes for hereditary hearing loss. Tradi- ESTs (Ͼ10,000) were found to be expressed in the tional approaches, such as RT-PCR (including differen- cochlea. Expression profiles derived from duplicate tial display), cDNA library screening, and sequencing samples at the same developmental stages showed gen- of cDNAs from inner ear libraries, have had some suc- eral agreement and indicated the reproducibility of cess in identifying inner ear genes (Gong et al. 1996; the assay. The expression of many known hair-cell Heller et al. 1998; Skvorak et al. 1999). There are, genes was detected in the whole-cochlea samples, dem- however, limitations to those approaches. RT-PCR and onstrating the relatively high sensitivity of the assay. cDNA library screening can deal with only very limited Genes highly expressed only at P2 or P32 were also numbers of genes. Sequencing of cDNA libraries from identified and their expression patterns correlate with inner ear has identified thousands of genes, many of their functions in the cochlea. A web-based database them novel. In order to recognize low-copy-number with external links was set up for public access, which genes with random sequencing of clones, however, an should facilitate the discovery of genes important in impractically large number of genes have to be the development and function of the inner ear and sequenced. For instance, about two-thirds of the should aid the identification of additional deafness ϳ20,000 genes expressed by a typical cell type are genes. expressed at low abundance, i.e., at Ͻ5 copies per cell Keywords: cochlea, utricle, microarray, genomic, Wnt (Zhang et al. 1997). If 50,000 clones from a cDNA library were sequenced at random, we calculate that more than half of these low abundance transcripts (a third of genes expressed) would be missed. The SAGE (serial analysis of gene expression) method to profile expression (Velculescu et al. 1995; 2000) relies on the sequencing of a large number of clones of concate- Correspondence to: Zheng-Yi Chen, Ph.D. • WEL415 • Neurology nated, unique sequence tags derived from multiple Department • Massachusetts General Hospital • Boston, MA 02114. Telephone: (617) 724-8599; fax: (617) 726-5256; email: cDNAs. While SAGE analysis does not depend on previ- [email protected] ous knowledge of the cDNA sequences (therefore new 140 CHEN AND COREY: Inner Ear Gene Expression Database 141 genes can be discovered) and is very quantitative, it is the cochlea. The cochleas were removed into MEM also laborious and requires the preparation, screening, (Invitrogen, Carlsbad, CA) buffered with 10 mM and sequencing of thousands of clones for each sample. HEPES to pH 7.4. P32 cochleas were transferred into Microarray technology, with which tens of thousands chilled RNAlater solution (Ambion, Austin, TX) and of genes can be monitored simultaneously, offers an stored at 4ЊC. P2 cochleas were frozen in liquid nitrogen alternative and efficient approach to studying gene and stored at Ϫ80ЊC. A minimum of 1 ␮goftotal RNA expression in the inner ear. There are different forms was required to produce sufficient cRNA for GeneChip of microarray technology, but all use the same principle: hybridization (Z-Y Chen, unpublished data). For this to immobilize targets (cDNAs, oligonucleotides, etc.) study, an average of ten mouse cochleas were collected on the surface of a solid support, to hybridize a sample for each of four samples, which produced an average to the targets, and to record all the hybridization signals of 1.5 ␮gtotal RNA. simultaneously. Since many targets (ranging from The procedures for RNA extraction and cDNA and 10,000 cDNAs to 400,000 oligonucleotides) can be cRNA synthesis followed the methods recommended immobilized on one surface, comprehensive expres- by Affymetrix. Total RNA was extracted using the sion patterns can be detected in a given sample through RNeasy RNA isolation kit (Qiagen, Valencia, CA). Tis- several rounds of hybridization (Lander 1999). The sues were homogenized in 700 ␮LofRLT solution with success of the human and mouse genome projects has ␤-mercaptoethanol. An equal volume of 70% ethanol produced an enormous amount of sequence data. was added to the homogenized lysate, mixed well, and Chips commercially available have sequences represent- spun through an RNeasy mini spin column. The col- ing 60,000 human genes and EST clusters, a large frac- umn was washed once with RW1 solution and then tion of the genome. transferred to a new collection tube. To avoid DNA We have used oligonucleotide arrays (GeneChip, contamination, 10 ␮L DNase I with 70 ␮LRDDbuffer Affymetrix, Santa Clara, CA) to profile gene expression (RNase-free DNase set, Qiagen) was mixed and added in the mouse cochlea. We chose the GeneChip system to the column for 15 min. The column was washed with because it is commercially available and covers the high- RW1, followed by wash with 500 ␮LofRPE solution. est number of genes. Importantly, the data obtained Diethyl pyrocarbonate-treated H O(50␮L) was used to can be compared with data from other resources using 2 elute RNA. Oligonucleotide assays require high-quality the same platform. We report here gene expression in RNA for cDNA synthesis. The total RNA was further the mouse cochlea assayed with the GeneChip technol- purified with 3M NaAc and ethanol in the presence ogy. These data are available as a searchable database of glycogen. that can be downloaded from the web. First- and second-strand cDNA synthesis used the T7- (dT)24 (designed by Affymetrix and synthesized and METHOD HPLC purified by Genset, La Jolla, CA) and Gibco/ BRL SuperScript Choice system (Invitrogen, Carlsbad, CA), following the manufacturer’s manual. To improve The oligonucleotide array set Mu30K (comprising the efficiency of cDNA synthesis from total RNA, synthe- Mu11K plus Mu19K) was used for the study. This set Њ covers ϳ13,000 known genes and ϳ21,000 EST clusters. sis was done at 42 C. Double-stranded cDNA was then Because of the overlap between some of the ESTs and purified with an equal-volume phenol/chloroform/ known genes, many genes or ESTs are represented isoamyl alcohol (25:24:1) mix and separated by spin- more than once on the chip. The total number of ning in a tube containing Phase Lock Gels (PLG; unique genes covered by the set is ϳ22,000. For simplic- Eppendorf-5 Prime, Inc. Boulder, CO). The aqueous- ity, we will treat each entry on the chip (a total of cDNA-containing phase was purified with 7.5 M NH4Ac 33,945 known genes or EST clusters) as one gene in and ethanol in the presence of glycogen. our analysis. The Enzo Bioarray High Yield RNA Transcript Label- Two developmental stages were used for the study, ing Kit (Enzo Diagnostics, Farmingdale, NY), was used postnatal day 2 (P2) and postnatal day 32 (P32). Since for in vitro transcription to make cRNA. The T7 pro- the inner ear is not fully functional at P2 but is mature moter linked to the (dT)24 used for cDNA synthesis at P32, the comparison may reveal genes important in served as the binding site for T7 RNA polymerase for the maturation process. The CBA/CaJ mouse strain in vitro transcription. Following the kit instructions, we was used because it has good hearing (based on ABR routinely obtained ϳ11 ␮gofcRNA from ϳ1.5 ␮g testing) even at 2 years old (Zheng et al. 1999). P32 total RNA. The cRNA was then purified with Qiagen’s mice were anesthetized with isoflurane; P2 mice were RNeasy kit before it was fragmented in 5 ϫ RNA frag- anesthetized with CO2.Mice were sacrificed by decapita- mentation solution [40 mM Tris-acetate (pH 8.1), 100 tion and the skull was hemisected, allowing a medial mM KOAc, and 30 mM MgOAc] at 94ЊCfor35min. Gel approach to the temporal bone that offers access to electrophoresis was run with ds cDNA, unfragmented 142 CHEN AND COREY: Inner Ear Gene Expression Database cRNA, and fragmented cRNA to confirm that no obvi- TABLE 1 ous degradation of cRNA occurred during the prepara- tion and that the cRNA was fragmented to the optimal Detection of hair-cell genes in the whole-cochlea samples size of 35–200 bases for hybridization. Hybridization Genes P2aa P2ba P32ab P32bb and scanning were conducted in a core facility, using Myosin-I␤ ϩϩ ϩϩ the protocol provided by Affymetrix. The final amount ϩϩ Ϫϩ ϳ ␮ Myosin-VI of the cRNA from each sample was 10 g, which is Myosin-VIIa Ϫϩ/ϪϪ Ϫ only enough to hybridize one chip.