Development and Characterization of a Normalized Canine Retinal cDNA Library for Genomic and Expression Studies

Abstract

purpose. Identification of causative mutations for retinal blinding disorders is often limited by restricted understanding of gene expression and underlying molecular mechanisms that trigger degenerative processes. This study was conducted to develop a catalog of canine retina-expressed genes that would provide a unique tool to investigate normal and altered function in the adult retina. Because of the conserved syntenies between the dog and human, this approach would identify new potential disease candidate genes for both species.

methods. A canine normalized retinal cDNA library was produced and analyzed by using a modified PhredPhrap algorithm. Computerized annotation provided gene homology and chromosomal location for individual clones and contigs in a Web-accessible database.

results. From 6316 cDNA clones, 3980 retinal expressed sequence tags (ESTs) were derived. Homology to the canine genome draft sequence was found for more than 99% of all ESTs, but only for 32% when compared with annotated canine cDNAs. Functional analysis suggests an enrichment of this library for genes involved with eye function and development, chaperone, or ribosomal functions when compared with mouse and human National Center for Biotechnology Information (NCBI) RefSeq entries.

conclusions. A combination of annotation approaches with ongoing mapping and expression studies provide functional data covering at least 27% to 30% of the currently proposed canine catalog of genes expressed in the retina. This is an essential first step toward establishing an integrated network for gene identification and expression patterns suitable for functional genetics, comparative genomics and evolutionary analysis of genes and gene families with respect to the developmental and degenerative processes of the retina.

The distinctive population structure of the domestic dog has been recognized as a unique resource for investigation of the genetic basis of morphologic, behavioral, and disease traits. ¹ To date, we recognize more than 400 hereditary disorders in the purebred dog (http://omia.angis.org.au/ provided in the public domain by the Australian National Genomic Information Service, University of Sydney, NSW, Sydney, Australia). ^{2 3} Of the approximately 30 diseases for which the underlying gene mutation has been identified, 7 are related to forms of inherited retinal degenerations, ^{4 5 6 7 8 9 10 11} the predominant group being progressive retinal atrophy (PRA). PRA comprises a group of hereditary blinding diseases with broadly similar clinical features thought to be present in at least 46 different breeds. ¹² The disease is postulated to be the conditional homologue to human retinitis pigmentosa (RP). In most cases, the disease is characterized by photoreceptor degeneration, initially affecting rods and then cones. Although the age of onset and progression rate varies between different dog breeds, PRA typically advances to complete blindness. It is assumed that, as in other species, this common pathway is apoptotic. ^{13 14} However, little is known about the exact molecular mechanisms that trigger degenerative processes in different subsets of these disorders in dogs or humans. Thus, the causative mutations and pathogenesis have still gone unidentified for many forms of PRA and RP (http://www.sph.uth.tmc.edu/RetNet/ provided in the public domain by the University of Texas Houston Health Science Center, Houston, TX). 

The recent release of a 7.6x canine genome sequence draft (http://www.genome.gov/11007358) ¹⁵ in addition to the original published 1.5x coverage of the canine genome, ¹⁶ has greatly improved the available resources for canine studies. These developments drive rapid identification of new disease genes in Canidae, and also yield detailed insight into disease and functional pathways common to mammals. The steadily decreasing number of estimated genes per genome ¹⁷ suggests that many of the observed phenotypic variations are driven by modification at the RNA or protein level. Although comparative genome projects contribute information about the structure and variation of genes, the need for expression maps becomes more and more apparent. The collection of expressed sequence tags (ESTs) from a specific tissue or cell type provides an expression profile including information about the level and complexity of gene expression in that tissue or cell type. ESTs are an essential resource for functional genomic studies, particularly for the construction of cDNA microarrays, that enable comparison of gene expression levels altered with disease, aging, and development. ¹⁸ Furthermore, EST analysis, together with genomic sequence information, has been a powerful tool for discovering novel and uniquely expressed genes. ^{19 20 21 22} Although these projects are well on the way for some species, ^{23 24} cDNA library characterization in the dog has just started in brain tissue, ²⁵ and only a few retinal EST clones have been identified. ²⁶  

To enhance our knowledge of cellular pathways related to retinal development, function, and disease, we have established an integrated database of canine retina-expressed genes that is Web accessible. Initially, six canine cDNA libraries were developed to augment our understanding of retinal and photoreceptor-specific gene expression using a well-defined canine disease model, progressive rod–cone degeneration (prcd). ^{27 28} We have demonstrated normal visual pigments with this disorder, ²⁹ and the mutant photoreceptors and their surrounding insoluble matrix develop and function normally before undergoing a series of topographically specified changes of the rod photoreceptor that progressively affect the outer segment (OS), inner segment, and nuclei. ^{27 30 31 32 33} Cone disease and degeneration occur later, once the rod disease is well established. The hallmark abnormalities in the disease are a marked decrease in rod outer segment (ROS) renewal that begins in structurally normal rods. ^{30 31 32} In miniature poodles (MPs) this abnormality can be measured as early as 9 to 13 weeks, and morphologic alterations in ROS structure are typically observed soon thereafter. To track changes due to the original disease rather than subsequent apoptosis, we have currently focused on the development of the retina and comparison between normal and diseased status at 16 weeks. At this time, the retinal structural abnormalities are limited to ROS disorganization, and there is no degeneration. 

Herein, we describe the development of these libraries and the characterization of the first normalized canine retinal cDNA library, containing approximately 4000 nonredundant gene clusters. The organization of a canine retinal EST database will provide detailed information and access to the actual clones for future use in expression and comparative genomic studies based on chromosomal location and sequence comparisons. The combination of these approaches will provide new potential candidates for the investigation of retinal function and disease for both dogs and humans. 

Methods

cDNA Library Construction

A 16-week-old beagle, homozygous normal for known retinal disorders, and four age-matched, poodle-derived, cross-bred dogs, homozygous affected with prcd, were used after euthanatization with a barbiturate overdose. Because of the unique breeding strategy, these dogs are considered to have similar genetic backgrounds; thus, the main genetic variation is contributed by the inheritance of prcd. The procedure complied strictly with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research, and received Institutional Animal Care and Use Committee (IACUC) approval. Retinas from normal and affected dogs were collected within 1 to 2 minutes of death, and brain from the normal beagle within 5 to 10 minutes. Brain tissue was rapidly separated into frontal, occipital, and temporal lobes. Retinas were gently isolated from the eye cup with forceps, and all tissues were flash frozen in liquid nitrogen. A total of six libraries were constructed. Three primary libraries were obtained: one from normal brain (DP01; combining equal amounts of RNA from the three identified regions), and one each from normal (DR01), and prcd-affected (DP02) retinas. Subtraction libraries were constructed by subtracting the normal brain from the normal retina (DP05), and normal and prcd-affected retina libraries from each other (DP03 and DP04, respectively). 

Total RNA was isolated using a total RNA isolation kit (RNeasy; Qiagen, Valencia, CA) followed by treatment with DNase (1 unit RQ1 DNase; Promega, Madison, WI). Poly(A)+ RNA was purified with the mRNA kit (Oligotex; Qiagen). Double-stranded cDNA was synthesized (Superscript Choice System kit; Invitrogen, Carlsbad, CA), and oligo d(T) primers that included tag sequences specific for the libraries (Table 1) . The cDNAs were directionally cloned into the EcoRI- and NotI-digested pGEM-Zf11(+) vector (Promega). The cDNA inserts are flanked by an adapter sequence at the 5′-end (5-AATTCCATTGTGTTGG-3′), and by the 3′-end tag sequences. The libraries were subsequently normalized and subtracted as described in Bonaldo et al. ³⁴  

DNA Sequencing and Sequence Analysis

Subtracted and normalized libraries were transformed into DH12S cells (Invitrogen) and plated on Luria-Bertani (LB) agar containing ampicillin at a final concentration of 100 μg/mL. Individual clones were chosen and grown in medium (CircleGrow; Qbiogene, Irvine, CA) containing ampicillin at the same concentration. Each clone was stored at a final volume of 40 μL in the same medium with 8% glycerol on a 384-well plate at −80°C. Plasmid DNA was extracted and sequenced (model 3700 sequencer; Applied Biosystems, Inc. [ABI], Foster City, CA) using T7 primer (5′-TAATACGACTCACTATAGGG-3′). Base-calling was performed with PHRED as part of the PhredPhrap software package. ^{35 36} Vector sequences were trimmed using cross-match. Low-quality bases (quality score < 20) were trimmed from both ends of sequences using Qualtrim (produced at the Keck Center for Comparative and Functional Genomics, University of Illinois, Urbana-Champaign, IL), and subsequently screened for a continuous stretch of bases not exceeding more than 20% of low-quality bases. Those ESTs having a length of more than 200 bp after both vector and quality trimming were considered “high-quality” ESTs. After removing vector and adaptor sequences, repeat sequences were identified and then masked by the RepeatMasker program with mammalian repeat sequences as a reference. The masked sequences were further screened for bacterial contamination, RNA, viral DNA, rRNA, and mitochondrial DNA, by using BLASTN (.ncbi.nlm.nih.gov/blastn/ provided in the public domain by the National Center for Biotechnology Information, Bethesda, MD). The sequences that passed the screening were considered to be “clean” EST sequences and were used for further analysis of assembly with the PHRAP program. Using default settings in PHRAP (minimum overlap, 30 bp, and 75% identity in the overlap region), we generated consensus sequences for all the contigs. All contigs and singletons were made up into the tentative unigene set of these ESTs. 

Annotation and Database Construction

To identify homology to known genes, BLAST searches were performed against the NCBI RefSeq sequences (http://www.ncbi.nih.gov/RefSeq/) using e-value cutoffs of E ≤ 1e ^{− 5}, E ≤ 1e ^{− 3}, and E ≤ 1e ^{− 1}, respectively, for BLASTN and BLASTP search. Each clone was annotated with a title based on the best BLAST results of the clone sequence and the respective contig sequence. Homology at a medium significance level (1e ^{− 3} < E < 1e ^{− 5}) was labeled “similar to,” whereas low significance (1e ^{− 1} < E < 1e ^{− 3}) is indicated by “has low similarity to.” BLASTN (E ≤ 1e ^{− 50}) against canine cDNA and genomic draft sequence provided genomic location of the EST. The resultant chromosomal location has been indicated by the base pair position on the respective chromosome corresponding to the latest annotated draft sequence. An SQL (structured query language) relational database was constructed to store the computational analysis together with all other experimental data for each clone. The EST database can be accessed through a web interface (URL: www.bioinformatics.upenn.edu/canine_retinalESTs/ provided in the public domain by the University of Pennsylvania, Philadelphia, PA). 

Results

We produced three normalized canine cDNA libraries needed for detailed functional studies of retinal diseases. Normal retina (DR01) and brain (DP01) libraries were obtained from a dog not affected with inherited retinal degeneration, and a prcd (DP02) library was obtained from prcd-affected retinas. Both retinal libraries were subtracted from each other in either direction. The brain library was subtracted from the normal retinal library to enrich for genes specific to retina (DP05). For each of the six final libraries, an initial 192 clones were isolated for sequencing, to assess quality (Table 1) . Ten thousand clones were chosen and sequenced for the normalized normal retinal library (DR01) to initiate a database for canine retinal genes. Although the other five libraries are stored for future investigation and all clones are available through the generated database, data presented herein were obtained from analysis of the normalized normal retinal library unless otherwise indicated. 

A total of 10,368 retinal EST clones from the DR01 library were sequenced and analyzed through a modified phredPhrap algorithm. Of these, 3549 did not pass quality checks, 332 were identified as containing contamination, another 166 were excluded based on short or no inserts, and one cluster of 5 sequences represented mitochondrial DNA. The result was 6335 clean, high-quality genomic EST sequences with an average read length of 404 bp (Table 2) . Automated analysis identified nine potential recombinant clones. Manual revision, however, confirmed only one of these clones to be a true recombinant (DR010017A10H07), a recombination rate of less than 0.02%. Subsequent clustering reduced the number of unigene sequences to a total of 3980 suggested genes, of which the majority represents individual clones (singlets). More than half of the 861 contigs were created by alignment of only two clone sequences, whereas the most abundant cluster (GNGT1) consists of 60 individual clones. The 47 contigs represented by at least 10 EST clones are listed with their deduced gene homology (Table 3) . Although 17 of the 30 most abundant retinal ESTs in humans deposited at the NEIBank ²³ are also present in the canine retinal library (rhodopsin, elongation factor 1-α, ferritin [heavy], transferrin, S-antigen, tubulin β2, ribosomal protein P0, glutamine synthetase, ribosomal protein S3A, α-transducin, NRL, ROM1, vimentin, recoverin, ribosomal protein L13a, elongation factor-1-γ, and unc119), only one of them appears within these most abundant clusters (Table 3 , gene 20: retinal S-antigen). 

Gene homology was inferred by BLASTN of EST and contig sequences against the RefSeq human and mouse library. Cutoff for interspecies comparison was chosen at E < 1e⁻⁵, which allowed successful annotation of 2680 (67%) unigenes of 3980 suggested genes, corresponding to 4658 (74%) individual ESTs. However, annotation strictly based on EST sequences only recognized homology for 4554 EST sequences (72%). By dropping the confidence level to 1e⁻⁵< E < 1e⁻³, we identified another 139 unigene sequences (170 EST sequences). These were annotated to be “similar to” the aligned gene. This class produced 184 hits based on EST sequences. The excess of ESTs, compared with the number corresponding to unigene sequences, belonged to sequences for which the annotation had already been achieved at the higher confidence level (E < 1e⁻⁵), when using the respective unigene sequences. Six hundred eighteen ESTs were found to have “low similarity to” aligned genes (1e⁻³< E < 1e⁻¹). Of these, 457 unigenes assigned to this group correspond to 552 individual ESTs. Overall, 5356 (85%) ESTs were annotated on an individual basis, and 5380 total annotations were achieved using 3276 (82%) unigene sequences. Annotations based on individual EST or unigene sequences differed for approximately 9% of all clones. In any case, best BLAST hits are displayed in the database for both. 

By restricting the BLAST search to cDNA entries annotated for Canis familiaris at E < 1e⁻⁵⁰, only 32% of all unigenes (1264) that represented 41% of individual clones (2590) were identified as homologues to previously annotated canine genes. Using the same criteria against the current canine genome draft, we identified 6284 of the canine ESTs, represented in 3951 of the 3980 unigene sequences, as being present in the canine genome. Of the remaining 32 sequences (28 unigene sequences), 21 had a putative gene homology (17 unigene sequences), whereas 11 putative individual EST sequences, each representing an individual unigene, did not align to any annotated sequence. Of the 2819 unique library sequences that were assigned a title line (E < 1e⁻³), one of nine functional groups (metabolism, signal transduction, cell cycle, cell structure, apoptosis, immune response, ribosome, eye function and development, and chaperone) could be inferred for 2152 clones based on the corresponding Gene Ontology (GO) database annotation (76%). ³⁷ For comparison, 35,989 of 56,940 currently deposited RefSeq mouse or human sequences (63%) have a GO identifier. A direct comparison of the distribution of the different sequence pools with respect to putative gene function suggests a more than two-fold enrichment of our library for genes related to chaperone function, eye function and development, and ribosomal genes in comparison to the RefSeq annotation (Fig. 1) . 

As part of developing the cDNA libraries for functional genomic studies, we have characterized a small number of clones from the other five normalized libraries: brain library (DP01), a prcd-affected retina library (DP02), three subtraction libraries corresponding to normal and prcd-affected retina subtracted from each other (DP03 and DP04), and brain subtracted from the normal retinal library to enrich for retinal genes (DP05). The number of unigene sequences present in each of the libraries, and overlapping with other libraries, is shown in Figure 2 . About half of the sequences identified from the two subtraction libraries (DP04 and DP05) that are based on the normal retinal library were also present in the original normalized retinal EST library (49% and 46%, respectively). In contrast, the two prcd-affected retina-based libraries, DP02 and DP03, showed lower representation in this pool (38% and 35%, respectively). Similar results were found for normalized brain EST sequences (DP01) present in the normalized retinal expression pool (30%). As the number of clones sequenced for these other five libraries was small, it is likely that the representation profiles will change as each library is characterized further. 

Sequence information of all ESTs has been submitted to GenBank (Accession numbers DT535933 through DT543037; provided in the public domain by the National Center for Biotechnology Information, Bethesda, MD). As well, complete information on each clone, annotation, and subsequent expression data are managed in a server-based database. The complete database can be accessed through any Web server (www.bioinformatics.upenn.edu/canine_retinalESTs/), and offers search functions such as BLAST and queries based on clone identifiers, annotation, or chromosomal location. The database will be continuously updated with respect to new annotations and results relevant to the location or expression of the clones. 

Discussion

With the completion of the human genome ¹⁷ and draft sequences for several model organisms, it is now important to understand the function and interaction of genes. To catalog and study genes relevant to retinal function, currently available data are being collected to generate the mammalian catalog of expressed genes. ³⁸ Although this study indicates that as many as 15,000 genes are expressed in the retina and RPE, the results represent a compilation from various sources with widely different supporting data. Furthermore, of 13,037 genes classified as high quality, less than 10% originate from nonhuman species, primarily mouse and cattle, with only 65 genes reported for the dog. To this end, we have produced the first normalized canine retinal cDNA library to initiate the complete annotation of canine retinal genes, and facilitate identification of those genes relevant to retinal disease research and comparative genomics. The almost 4000 unique sequences in our current database cover ∼27% to 30% of the proposed retinal catalog. Estimation of the total number of genes expressed in the retina may include genes expressed only in RPE. Although these genes may significantly affect retinal function, some (e.g., RPE65 ^{39 40 41 42} ) are not present in our retinal library. This suggests that our database is indeed specific for genes expressed in the canine retina and therefore may represent a higher percentage of the overall genes expressed only in this tissue. 

Normalization of the libraries has produced 38% redundancy within the investigated clones, which is within the range found in other studies, ⁴³ and has lowered the abundance of highly expressed retinal genes such as rhodopsin, which is represented by only nine clones. Consequently, almost 80% of all unique library entries are supported by a single clone, and less than 5% of all assemblies merged more than 10 clones (Table 2) . Through this process we also confirmed one recombinant clone. Overall, the data confirm the successful normalization and high quality of the library. The completion of the canine genome sequence draft has allowed us to locate more than 99% of the retinal sequences within the canine genome. The canine genome may be enriched for repetitive elements ¹⁶ that occur not only in coding sequences, but also can contribute significantly to developmental functions. ⁴⁴ Although coded by the RepeatMasker program (http://www.repeatmasker.org/ provided in the public domain by the Institute for Systems Biology, Seattle, WA) before the BLASTN search, clones containing repetitive elements were not removed from the library, and surrounding sequences may contribute to difficulties in unique alignment against the canine genome sequence draft. Alternatively, errors in the draft assembly, gene duplications, and gene families sharing common sequence motifs would also be reflected in alignments of one clone to more than one genomic region. These problems are common to automated annotations and cannot be overcome easily. We are currently integrating several clones, not readily mapped with analytical and comparative genomics, into the canine/hamster RHDF₅₀₀₀ panel. ⁴⁵ This process will allow us to understand better the dynamics of the canine genome, improve future mapping methods, and contribute to resolve discrepancies between the sequence and RH map in the dog. 

With only 32% of the canine retinal cDNA sequences representing annotated canine cDNA entries, the library significantly increases experimental support for retinal genes not yet cloned in the dog. At the same time, annotation is achieved through comparative genomics, which, in some cases, generates several problems as individual genes or splice variants might not be present in different species, or not yet identified. Furthermore, if EST clones are located in the 3′ untranslated region (UTR) of a gene, homology between species is often not sufficient for alignment. 

We have analyzed several clones, particularly those not yielding intron–exon boundaries, finding that the library is indeed slightly biased toward the 3′ UTR. For at least two clones, chosen due to their position in the reported prcd disease interval, ^{28 46} we confirmed that the corresponding genes are present in retinal mRNA pools, and the clones are not contributed through genomic contamination. In addition, these clones have been found critical for the respective disorders and are currently under investigation (data not shown). An even more prominent issue with cDNA libraries is the clustering of nonoverlapping sequences into genes. To date, no algorithm is ready to automate this procedure, and currently all nonoverlapping sequences are listed independently, even if assigned to the same gene (eg, Table 3 ; no. 8 and 26). This problem becomes even more evident for contigs that have no or poorly supported title line annotation. For example, based on chromosomal location, the unidentified unspliced cluster (Table 3 , no. 32) is likely part of the 3′UTR of the contig identified to be similar to MALAT1 (Table 3 ; no. 2), for which homology also is not very well supported. We are currently establishing an upgraded relational database to improve annotation in the future through a combination of title line annotation and mapping data with expression profiles. 

A previous study based on human NEIBank EST entries lists the 30 most abundant retinal genes in humans, ²³ 17 of which were present in our canine retinal library. However, because of successful normalization, only one of these (retinal S-antigen) is listed within the top 47 genes in the present database. Relevance of the library to retinal function has been further assessed by comparison of distribution of unique entries into nine distinct functional GO classes between our library and all annotated RefSeq sequences of mouse and human (Fig. 1) . In addition to an overrepresentation of genes associated with eye function and development, the relative increase of genes related to chaperone and ribosomal functions is in concordance with molecular mechanisms in the retina. Even within the 47 most abundant genes of the retinal library (Table 3) , 5 have expression only in the retina, and 12 have ribosomal function. Seven genes in this group play an important role in mitochondrial function, a reflection of the high-energy metabolism requirements of retina, and another seven are associated with DNA or RNA processing mechanisms. Surprisingly, two of the highly abundant genes in the library do not have a coding function. One of these, RNU47, is a well-described small nuclear RNA, but little is known about the second common observed cluster (homologue to MALAT1). This noncoding RNA transcript has been reported in Macaca mulatta to modulate proliferation of retinal neuroprogenitor cells in primate experimental myopia (Tkatchenko AV, Walsh PA, Tkatchenko TV, Gustincich S, Raviola E, unpublished data, 2005; GenBank accession number DQ148151). Of these 47 genes, only 8 have been cloned previously in the dog. Although most have been predicted from genomic data, both of the noncoding genes are unidentified in the dog, thus supporting the necessity for accurate records of retina-expressed genes to elucidate molecular functions. 

Assessment of the small subset of clones available for the remaining five normalized and subtracted libraries revealed substantial amount of overlap (30.6% to 49.3%) with the normalized retinal EST library. However, no bias toward any particular functional class has been observed. Subtraction of cDNA libraries largely eliminates the effect of differential gene expression. Subtraction libraries, therefore, should be highly enriched for transcripts that are significantly reduced or missing from the subtracted population. We are in the process of investigating the function of the clones enriched in the respective subtraction libraries to determine whether these might reflect chemical or molecular pathways that are specific to the photoreceptors or changed with the onset of prcd. To facilitate the generation of an integrated retinal gene library, clones characterized in this study have been used as a driver for subtraction from the normalized retinal library, reducing the redundancy of the library to 7% with an estimated 20% overlap to the original library. This process will allow us to add missing genes to the database on a continuous basis. At the same time, genes known to be critical to retinal function, but missing from the current library, are cloned and manually added to complete a canine retinal gene catalog. 

The database can be searched through a Web browser (www.bioinformatics.upenn.edu/canine_retinalESTs/ provided in the public domain by the University of Pennsylvania, Philadelphia, PA) using clone ID, key words included in title line annotation, chromosomal location, or BLASTN. Individual clone data sheets contain the current results, as well as information on RH mapping and cDNA microarray expression providing hyperlinks to respective results, if clones were included in these projects. The combination of annotation, location, and expression data of clones provides an easy tool to select new splice variants or novel candidates based on their characteristics. Clones are readily available to other investigators on request. Beyond the obvious advantage of providing the database for canine retinal disease studies, the library also will contribute further to refining synteny of the canine genome with other species of interest to retinal research and comparative genomics. We consider these results an important first step toward an integrated network for gene identification and expression patterns relevant to developmental and degenerative processes of the retina, and will continue to update this information and interlink our data with other existing tools. 

 Supported by National Eye Institute Grants EY06855 and EY13132, The Foundation Fighting Blindness, The Morris Animal Foundation/The Seeing Eye, Inc., The Laura J. Niles Foundation, The Van Sloun Fund for Canine Genetic Research, and Pfizer, Inc.

 Submitted for publication November 15, 2005; revised February 10, 2006; accepted April 17, 2006.

 Disclosure: B. Zangerl, None; Q. Sun, None; J. Pillardy, None; J.L. Johnson, None; P.A. Schweitzer, None; A.G. Hernandez, None; L. Liu, None; G.M. Acland, None; G.D. Aguirre, None

 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

 

Table 1.

 

Identification and Number of Clones Obtained for Each of Six Normalized EST Libraries

Library	Library Identifier	5′Tag	Clean Sequences (n)	UniGene Sequences (n)	Redundancy (%)
Normal retina	DR01	18(A)TGCGT	6316	3980	37.00
prcd retina	DP02	18(A)TCGGA	111	110	0.90
Normal brain	DP01	18(A)TCGGT	174	170	2.30
Normal retina-normal brain	DP05	18(A)TGCGT	154	149	3.25
Normal retina-prcd retina	DP04	18(A)TGCGT	152	148	2.63
prcd retina-normal retina	DP03	18(A)TCGGA	149	148	0.67
Totals			7077	4410	37.70

 Redundancy is determined by the number of the UniGene sequence set obtained for each library with respect to the total number of clean sequences. With the exception of the normal retina library (DR01) only a small number of ESTs were analyzed.

Table 2.

 

Sequence Analysis of the Normalized Retinal Library (DR01) Clones Using PhredPhrap

	Total	% of Original Clones
Clones sequenced	10,368	100.0
Failed sequences	3,549	34.2
Contaminated with E. coli/viral DNA	332	3.2
Clones containing empty vector or insert <200 bp	166	1.6
Mitochondrial DNA	5	0.05
Successful sequences	6,316	60.9
Singlets	3,119	30.1
Contigs
All	861	8.3
2 Clones	(474)
3 Clones	(158)
4 Clones	(68)
5 Clones	(46)
6–10 Clones	(77)
>10 Clones	(38)
UniGene sequence set	3,980	38.4

Table 3.

 

Genes Represented in the Normalized Retinal Library DR01 by 10 or More Clones

	Putative Gene/Clone Identifier	Gene Symbol	Number of Clones Clustering
1	Guanine nucleotide-binding protein (G protein), gamma transducing activity polypeptide 1	GNGT1	60
2	Metastasis-associated lung adenocarcinoma transcript	MALAT1	33
3	Ribosomal protein L23	RPL17/L23	32
4	Peripherin protein	RDS	30
5	Cytochrome c oxidase subunit VIIa	COX7A2	30
6	Tumor protein, translationally controlled 1	TPT1	25
7	Ribosomal protein S13	RPS13	22
8	Cytochrome c oxidase subunit VIc	COX6C	22
9	Thymosin beta 4	TMSB4X	20
10	Cold-inducible RNA binding protein	CIRBP	19
11	Guanine nucleotide binding protein (G protein), beta polypeptide 3	GNB3	19
13	Ribosomal protein S8	RPS8	19
12	NADH dehydrogenase (ubiquinone) 1 alpha subcomplex, 4	NDUFA4	17
14	Myosin, light peptide 6, alkali	MYL6	17
15	Nonmetastatic cell 1 protein	NME1	17
16	Ribosomal protein S17	RPS17	15
17	H3 histone, family 3B	H3F3B	15
18	Ribosomal protein L27a	RPL27a	15
19	Reticulon 4 (RTN4), transcript variant 5	RTN4	15
20	Retinal S-antigen	SAG	14
21	Phosducin	PDC	14
22	Translocase of inner mitochondrial membrane 8 homolog B	TIMM8B	14
23	AP20 region protein, transcript variant E	APRG1	14
24	Chromobox homolog 3	CBX3	14
25	Cu/Zn superoxide dismutase	SOD1	14
26	Cytochrome c oxidase subunit VIc	COX6C	13
27	Urate oxidase	UOX	13
28	Heterogeneous nuclear ribonucleoprotein A1	HNRPA1	13
29	Hypothetical protein HSPC016	HSPC016	13
30	Ubiquitin-like/S30 ribosomal fusion protein	FAU	12
31	Guanine nucleotide binding protein	GNB1	12
32	Unspliced cluster 1		12
33	ATP synthase, H+ transporting, mitochondrial F0 complex, subunit c	ATPG1	12
34	Ribosomal protein L36a	RPL36A	12
35	Ribosomal protein S12	RPS12	12
36	Ribosomal protein L12	RPL12	11
37	Ribosomal protein S7	RPS7	11
38	Ribosomal protein L22	RPL22	11
39	Translocase of outer mitochondrial membrane 7 homologue	TOMM7	11
40	Ribosomal protein S14	RPS14	10
41	Solute carrier family 24 (sodium/potassium/calcium exchanger), member 1	SLC24A1	10
42	ATP synthase, H+ transporting, mitochondrial F0 complex, subunit f, isoform 2	ATP5J2	10
43	Topoisomerase (DNA) II beta	TOP2B	10
44	Small nuclear RNA U47	RNU47	10
45	Splicing factor 3b, subunit 1	SF3B1	10
46	Dickkopf homologue 3	DKK3	10
47	Hypothetical gene supported by AL449243	LOC91689	10

Figure 1.

 

Comparison of distribution of sequences in nine functional classes based on GO annotation between the normalized canine retinal EST library (dark gray) and the RefSeq database (light gray). There is enrichment (*) of the EST library for genes related to ribosomal function (2.43-fold), eye function and development (2.39-fold), and chaperone function (2.08-fold). Note differences in scales.

Figure 2.

 

Overlap in the number of unique sequences present in five normalized canine EST libraries representing transcripts expressed in retina (DP02–DP05) and brain (DP01) and the normalized canine retinal EST library (DR01). Clones present in libraries not adjacent to each other are illustrated in circles outside or within the DR01 pool. Highest similarity to the DR01 library is observed with the libraries resulting from a subtraction of brain (DP05) or prcd retina (DP04) from the same clone pool, whereas the normalized brain library (DP01) shows the lowest amount of unigene sequences shared with the retina based libraries.