|Molecular Vision 2007;
Received 23 April 2007 | Accepted 20 November 2007 | Published 28 November 2007
A comparative gene expression profile of the whole eye from human, mouse, and guinea pig
Jun Yu,3 Jia
(The first two authors contributed equally to this publication)
1School of Ophthalmology and Optometry, Eye Hospital, Wenzhou Medical College, Wenzhou, Zhejiang, China; 2State Key Laboratory Cultivation Base and Key Laboratory of Vision Science, Ministry of Health and Zhejiang Provincial Key Laboratory of Ophthalmology and Optometry, Wenzhou, Zhejiang, China; 3Key Laboratory of Genome Sciences and Information, Beijing Institute of Genomics, Chinese Academy of Sciences, Beijing, China
Correspondence to: Jia Qu, School of Ophthalmology and Optometry, Eye Hospital, Wenzhou Medical College, 270 Xueyuan Road, Wenzhou, Zhejiang, 325003, China; Phone: 0086-577-88824116; FAX: 0086-577-88824115; Email: firstname.lastname@example.org
Purpose: To characterize and compare gene expression patterns of the whole eyeball among human, mouse, and guinea pig based on expressed sequence tags (ESTs).
Methods: Approximately 10,000 clones were sequenced from the 5'-end for each cDNA library made from mRNAs isolated from whole eyeballs of human, mouse, and guinea pig. ESTs were assembled and analyzed based on standard methods.
Results: We acquired 31,464 high-quality ESTs including 9,949 for the human, 11,495 for the mouse, and 10,020 for the guinea pig cDNA libraries, which were clustered into 12,253 unigenes for all three species. After removal of non-mRNA contaminations, we were able to match 96%, 97%, and 63% of the human, mouse, and guinea pig unigenes to sequences in the nonredundant library in GenBank, respectively. The high-abundance and medium-abundance genes in each library correlate with the anatomic structure and physiologic function of the eye in the three species. The large contribution from the lens in the mouse was related to the abundance of crystallin. Differentially expressed genes were observed among three libraries. Some of them appeared species-specific.
Conclusions: According to their gene expression patterns, guinea pig and human eyes are more similar compared to those of the mouse, making the guinea pig a promising animal model for eye research.
An appropriate animal model is an essential element in biomedical research. Several mammals have been chosen as models for eye research including mouse, rat, rabbit, guinea pig, dog, and monkey. Guinea pigs (cavies) are small mammals that have been used as an animal model to study pathogenesis and genetics for cataracts, retinal diseases, visual impairment in neonates, and other ocular complications of systemic disorders [1-3]. Furthermore, cavies have been used to study refractive changes associated with defocus of the eye and have shown responses similar to chickens, tree shrews, and primates in dimensional and refractive development of the eye [4-8]. Recently, cavies were also used for the study of form-deprivation myopia where moderate myopia (-5.8 Diopters) could be induced in two weeks by form deprivation starting at birth . Additionally, cavies have advantages over tree shrews, rabbits, cats, and primates as experimental animals because they are not only easier to rear and to breed but also because they grow quickly (developmentally mature at the age of five months), are more cooperative, and are cost-effective .
Given the significant physiologic and pathological differences between the eyes of human and animal models, selecting the right animal model for the study of certain diseases seems crucial to the acquisition of comparable and accurate results. Different eyes show great differences in anatomic structures. For example, despite the obvious similarities in the eyes of human, guinea pig, and mouse, there are gross quantitative differences in the eye structure among these three species. The human eye is almost eight times the diameter of the mouse eye, but the lens of a human eye is only twice as thick as that of the mouse, which fills most of the posterior segment. In addition, the anterior segment is relatively small in the human eye as compared to the anterior segment of the guinea pig and mouse (Figure 1).
Aside from the recently acquired genomic sequences for rat and mouse as well as ample ESTs (expressed sequence tags) for mouse, data for other mammalian models are still far from adequate. Although the NEIBank Database Project collects EST data for medical research on eyes from different species, mostly from mouse and rat, the current sum is still insufficient to meet the increasing demand . Most of the gene expression profiles are generated from mouse cDNA libraries, and mouse is an ideal animal for genetic studies. However, there are pitfalls in using mice as the model for eye research. For instance, the mouse has eyes that are too small for the experiments involving surgical procedures and molecular studies involving RNA and protein isolations. Our study focused on acquiring a large amount of ESTs from eyeballs from human, mouse, and guinea pig to compare gene expression profiles of the three mammals, which may provide clues in selecting a promising animal model for eye research on myopia and other ocular disorders.
The Ethics Committee of Wenzhou Medical College in China approved this study for both human subjects and animals. This study complied with the tenets of the Declaration of Helsinki for Research Involving Human Tissue and the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. Two adult human eyeballs were obtained from two humans (males aged 28 and 30) from the Wenzhou Eye Bank. The donors' eyes were free of ocular disease. Sixty eyeballs from the B6 mouse and 20 eyeballs from the Peruvian guinea pig had also been collected and supplied by the Wenzhou Medical College.
cDNA library construction and sequencing
Total RNA was extracted from eye tissues with Trizol agent (Invitrogen, Carlsbad, CA), and mRNAs were isolated by using the PolyATtractTM mRNA isolation system (Promega, Madison, WI) according to the manufacturer's instructions. Three nonnormalized cDNA libraries were constructed by using the Superscript II system and the directional pBluescript® II XR vector (Stratagene, La Jolla, CA) exploiting the EcoRI and XhoI restriction sites. The quality of the cDNA libraries was assessed by using colony polymerase chain reaction (PCR) and sequencing for insert size and E. coli contamination, respectively. Sequencing templates were prepared from plasmids in E. coli DH10B (Invitrogen) and sequenced on MegaBase® 1000 sequencers with a universal primer from a single end.
Expressed sequence tags assembly, annotation, and bioinformatic analysis
A Phred-Phrap-Consed package [11,12] with default parameters was used to determine sequence qualities (Q20 or 99% accuracy) and to assemble the sequences. A crossmatch program (University of Washington, Seattle, WA) was used to trim off vector sequences, and sequences longer than 100 bp were saved in FASTA format for sequence comparisons. The human and mouse ESTs were assembled according to supervised clustering strategy that utilizes known cDNA sequences in the public databases. The guinea pig ESTs were assembled independently. The assembled contigs and singlets were manually examined. Assembled ESTs or unigenes were annotated by comparing them to NCBI nonredundant (nr) or dbESTs (NEIBank) databases with BLASTn (e-value cutoff 1e-10) or BLASTx (e-value cutoff 1e-5). Functional classes were assigned according to GO (Gene Ontology) mapping provided by the Uniprot database. KEGG analysis was based on the comparative results between our unigenes and the updated KEGG database.
The IDEG6 program [13,14] was used to determine differentially expressed genes in multiple tag sampling experiments. Differentially expressed genes from three libraries were detected based on six different statistics tests, and finally, the Χ2 test was used to analyze differential gene expression after comparing all six arithmetically (AC.1-2, Fisher.1-2, Chi2x2.1-2, GT, R, Χ). A gene is considered differentially expressed when different from a random distribution model at a level of p<0.05.
Validation by reverse-transcriptase polymerase chain reaction
Total RNAs were extracted from eyes of the three species with Trizol (Invitrogen), and DNA contaminations were removed with RNase-free DNase I. An amount of 500 ng total RNAs was used for reverse transcription. The first strand cDNA of the three species was synthesized with poly (T) oligo nucleotides and superscript II reverse polymerase (Invitrogen). The primer pairs used in the validating experiment were detailed in Table 1. Specific primers were designed based on our EST data with average product size of 180-200 bp and S-antigen gene with product length about 500 bp. The reaction mixture was first denatured at 95 °C for 3 min. Cycling was performed at 94 °C (15 s), 55 °C (20 s), and 72 °C (30 s) for 30 cycles followed by an extension time of 10 min at 72 °C. An equal volume (3 μl) of all PCR products was electrophoresed on 2% agarose gels containing 1 μg/ml ethidium bromide and then photographed under ultraviolet illumination. The sizes of the amplified products were estimated by Marker DL2000.
Summary of the cDNA libraries
A total number of 31,464 high-quality ESTs were obtained including 9,949, 11,495, and 10,020 from humans, mice, and guinea pigs in the three respective cDNA libraries. These ESTs were assembled into 12,253 unigenes in which 4,315 were for humans, 3,070 for mice, and 4,868 for guinea pigs (Table 2). Mitochondrion-encoded genes (4.4%-6.4%) and rRNA sequences (2.4%-3.9%) were removed from the most abundant ESTs analyzed. The percentage of unigenes matching sequences from the nonredundant library in GenBank was 96% in humans, 97% in mice, and 63% in guinea pigs. Only a subset of the unigenes matched ESTs in the NEIBank collection, 77%, 83%, and 63% from libraries of human, mouse, and guinea pigs, respectively. The rest, collectively 27% of total unigenes, were not found in ESTs of NEIBank. The ESTs were also annotated according to the Uniprot database. This combined procedure increased the annotation rate for the unigenes (99.4% annotated for humans, 97.3% for mice, and 64.5% for guinea pigs). Since the sampling for the libraries may not be deep enough for random transcriptomes, the coverage was unable to be assessed.
Approximately 6.1% of the total unigenes have more than five clones, and 26% have more than one clone, indicating that the sampling is adequate for a qualitative test (discovering new genes) but inadequate for a quantitative test for transcriptomes. According to GO analysis, the major annotated groups could be classified into four categories with a percentage of unigenes: cell (48%), binding (51%), cellular process (71%), and physiologic process (70%; Figure 2). Among the three libraries established by the samples, the fractions of the genes in each category are similar across species, suggesting tissue-specificity of the unigenes.
Highly-expressed genes in the cDNA libraries
Crystallin and mitochondrial genome-encoded genes were among the highly expressed genes in all three libraries. There were trends in the results that suggested certain species-specificity. Mitochondrial genes were found to be most abundant in the human library whereas crystallin genes were enriched in the mouse library as compared to both human and guinea pig. We have listed the medium- and high-abundance genes in Table 3, excluding crystallin genes, rRNAs, and mitochondrion genome-encoded genes. The top ranked genes were found to be similar among the three libraries; they may have different rankings but are at the same magnitude. The high-abundance genes shared by the three libraries such as rhodopsin, S-antigen, transthyretin, and bestrophin are generally characteristics of the eye. The highly abundant hemoglobin transcripts may reflect a more enriched vascular system in the eye.
The abundant genes were categorized into 10 major functional groups based on the GO categories after exclusion of crystallin, rRNA, and mitochondrial genes. The top five functional categories include physiologic process, cellular process, cell, binding, and catalytic activity. Humans and guinea pigs shared six major categories with similar numbers of genes. In contrast, humans and mice shared only three major categories with similar numbers of genes. The third level GO categories were consistent with these results and provided more information on cellular functions and subcellular locations. The major functional groups based on these categories included intracellular, organelle, ion binding, protein binding, cell communication, localization, metabolism, cellular physiologic process, and organismal physiologic process. Furthermore, major functional categories with the number of genes were more similar between humans and guinea pigs than between humans and mice. Human eyes had a higher proportion of genes than the other two species in the categories related to metabolism and intracellular activities. Examination based on the KEGG database showed that highly-abundant genes covered most of the metabolic pathways associated with amino acids, purine, glucose, and proteins associated with oxidative phosphorylation and ATP synthesis.
Differentially-expressed genes of the three cDNA libraries
The pair-wise comparison (BLASTx with an e-value cutoff of 1e-5) showed that the pair-wise shared genes between each of the two libraries were approximately 20%-31% of the total. The human library shared more unigenes with the guinea pig library than with the mouse library. There were approximately 10% of the unique genes in each library. One thousand and ninety-seven genes were found unique to the human library but did not match any unigenes in either the mouse or the guinea pig library. There were 1,036 and 703 genes found that were unique to the guinea pig and mouse libraries, respectively. As majority of these "candidate" unigenes were primarily expressed in medium- or low-abundance (Table 4), their uniqueness in each library was likely due to bias during sampling. The GO analysis showed that major functional categories of eyeball genes included cell, organelle, binding, catalytic activity, cellular process, and physiologic process.
Between the human and guinea pig libraries, the Χ2 test found 131 and 54 differentially expressed genes at p values of <0.05 and <0.01, respectively. Similarly, 162 and 101 genes were found between mouse and human and 127 genes between mouse and guinea pig at p<0.01. The KEGG analysis showed that these differentially-expressed genes were related to signaling pathways including the MAPK, calcium, and TGF-β signals as well as the pathways related to cell cycle, apoptosis, circadian rhythm, and ubiquitin-mediated proteolysis. When referring to common metabolic pathways like citrate cycle and glycolysis, we found that the three species have enzymes differentially expressed within the same pathway
Comparison between the newly acquired expressed sequence tags data and the NEIBank collection
It was found that 8,945 (73%) of 12,253 unigenes from the three libraries matched the ESTs in the NEIBank. The fractions of hits to each tissue-specific library in the NEIBank are 54% for the retina, 10% for the lens, and 8% for the iris. The minor hits showed 76 unigenes in the trabecular tissue and 12 unigenes in the cornea. Overall, 3,333 (75%) unigenes from the human library matched the ESTs in the NEIBank collection, and over 85% of our mouse unigenes collection matched the public counterpart in the NEIBank. Since the data for guinea pigs in the NEIBank is limited (6,617 ESTs), the present study had only 3,079 unigenes matching the public unigenes. Among these matched genes, guinea pigs shared 1,073 ESTs with humans and 1,264 ESTs with mice. Therefore, 3,308 unigenes from three libraries found in this study enriched the data in the NEIBank in novel sequence information and expression profiles.
Expression pattern validated with reverse-transcriptase polymerase chain reaction
The expression pattern of nine differentially expressed genes was further validated with reverse-trancriptase polymerase chain reaction (RT-PCR) assay. Most of them exhibited changes that were consistent with the result of EST analysis (Figure 3) except rhodopsin with no significant differences among the three that observed 95, 64, and 146 ESTs from the libraries of human, mouse, and guinea pig, respectively. The Grifin gene failed to be amplified during the RT-PCR assay because it had a relatively low-abundance (usually less than five ESTs).
In the current study, we generated 31,464 high-quality ESTs that were clustered into 12,253 unigenes from three cDNA libraries made from entire eyeballs of humans, mice, and guinea pigs. This timely collection not only contributed novel sequences to the public databases but also helped us to determine which mammalian model most resembles human eyes at the transcriptome level. The transcriptomic profiles also provided molecular clues to differences among the three species in anatomy and physiology. Based on GO and KEGG analysis, we have classified most of the unigenes into different functional categories or distinct pathways. There are great differences in the physiologic structure of the eyeballs as well as visual behavior and circadian systems between human and rodent. This is consistent with the large number of differentially expressed genes that show species-specific patterns, which were found through comparisons of unigenes from each of the two libraries.
Phototransduction specific genes
Phototransduction specific genes were located primarily in the retina and were found in medium and high abundance in all three libraries. These genes included the opsin family, RAN binding protein, S-arrestin, recoverin, phosducin, and retinal G protein, which were observed to be differentially expressed. Rhodopsin was most abundant in the opsin family in all three libraries with occurrences of 95 clones in the human library, 64 clones in the mouse library, and 146 clones in the guinea pig library. This gene encodes rhodopsin pigment in the retinal rods that mediate vision in dim light. Variations in cone photopigment genes enable the cone photoreptors to discriminate colors . We found not only variations in amino acid sequences but also differentially expressed patterns that imply great differences in color vision formation among the three species. Phosducin is homologous in solubility and primary amino acid structure to phosducin localized in the pineal gland. It has been shown to bind the beta and gamma subunits of the retinal G-protein transducin and regulate the light-driven translocation of G protein transducin from the outer segments of rod photoreceptors to other compartments within the rod cell [16,17]. We observed phosducin expressed at a significantly higher level in mouse than both human and guinea pig. Recoverin, another retina-specific gene, mediates the recovery of dark current after photoactivation in the retina . It was found to be expressed more than five-fold higher in human and guinea pig than in mouse. Since the mouse is the only nocturnal animal among the three, this special expression pattern might correspond to the nocturnal behavior of the mouse. Photoperiod is the main environment influence on the circadian system of all living creatures, and the eye is the most important sensor responding to photoperiod stimuli [19-21]. Differential expression patterns of phototransduction specific genes among the three species were most likely to correspond to great differences in vision formation and the circadian system between human, a diurnal animal, and rodent, a nocturnal animal.
Correlation of gene expression patterns with anatomic structures
The eyes of the three species have differences in anatomic structure, especially in the lens. The lens fills most of the mouse eye, less so in the guinea pig, and only about 10% of the axial length of the human eye. Since we used the whole eyeball for the cDNA libraries, the gene expression patterns should correlate to differences in their anatomic structures. The expression pattern of lens-specific genes was observed to correlate with the specific anatomic differences. For instance, the crystallin genes were found to be richest in the tissue of the lens, and the three libraries have significantly different ratios reflecting the different proportions of lens in the eyeballs. Crystallin gene members numbered 16 in the mouse library. Except for γ-crystallins, other members were all found to have more than 45 ESTs sampled; the most is αA-crystallin with 300 copies. Only 13 members were detected in the guinea pig library. Except for αA-crystallin and γA-crystallin, all other members have much fewer copies in human and guinea pig as compared to what is found in the mouse; only nine members were observed in the human library and βB2-crystallin is the most redundant, having 13 copies. GRIFIN, a major lens specific member of the galectin family, seemed to be expressed in the lens of both mouse and guinea pig, but none of this cDNA was found in the human library. Until now, there is still no evidence to indicate that this gene is expressed in the human lens. The GRIFIN gene is detected in the human genome, but it appears to have been silenced during long-term evolution . Another lens-enriched gene, glutamine synthetase, was also found differentially expressed with more clones observed in the mouse and guinea pig libraries than in the human library.
Disease-related genes among the three species
Some notable disease-related genes were also found to be differentially expressed. Bestrophin, a chloride channel protein, is indicated for Best disease. Based on the comparison with the EST database from NEIBank, this gene was localized mainly in the libraries of RPE/choriod and has been reported to be a marker for RPE in human but is expressed in higher abundance in the testis than the eye . The alternative splicing of this gene also showed an age-related defect . Bestrophin was found most likely to have a species-specific expression pattern with 78 copies in the human library, which was significantly higher than what is found in the mouse and guinea pig libraries, which have 33 and 28 copies, respectively. We also observed a disease-related channel protein family that includes three differentially expressed genes, ApoD, ApoJ, and ApoE. ApoD and ApoJ were both found to be expressed at higher levels in human and guinea pig than in mouse, and ApoE were more abundant in the guinea pig library than in the other two libraries. Prostaglandin D2 synthase, another RPE/choroid-specific gene, was found to be expressed at a much higher level in mouse than in human and guinea pig. Decorin, a component of connective tissue, which binds to type I collagen fibrils and plays a role in matrix assembly, is altered by abnormalities in the structure or secretion of types I and III collagen. It has also been reported to be a myopia-related gene . It was expressed at a higher level in human than in mouse and guinea pig in our data set, which seems to be another example of major species differences in the use of a gene that is important in the function of the human eye tissue.
Another alternative animal model
A good experimental model for biomedical research should be easy to rear and surgically operate on. Therefore, the guinea pig appears to be a better model for ophthalmologic research because the anatomic structure of a guinea pig eyeball is much closer to that of the human and it is much easier for surgical procedures. When referring to gene expression profiles, we found that human eyes are closer to those of the guinea pig than of the mouse. According to the high-abundance genes in the three libraries, more mitochondria genes and relatively fewer crystallin genes were found in the list of high-abundance genes in human and guinea pig as opposed to the highly expressed crystallin genes in the mouse library, which is consistent with the different anatomic structure of the eye from the three species. Human and guinea pig also have similar transcriptional profiles for medium- and high-abundance genes, which were also observed most notably from the result of GO analysis. When comparing the expression profiles pair-wise, we found human eyes shared more unigenes with guinea pig eyes than with mouse eyes, and fewer differentially expressed genes were observed between human and guinea pig based on Χ2 tests. All of the present results support guinea pig as a promising animal model for experimental ophthalmologic research.
This study was sponsored by National Natural Science Foundation of China (30500549 and 30600227), Zhejiang Provincial Natural Science Foundation of China (Y204396, R205739, and Z205735), and Zhejiang Science & Technology Bureau (2003C23005; wkj2005-2-048-02). The authors thank Dr. Frank Thorn from the New England College of Optometry for reading an earlier version of the manuscript.
1. Han J, Little M, David LL, Giblin FJ, Schey KL. Sequence and peptide map of guinea pig aquaporin 0. Mol Vis 2004; 10:215-22 <http://www.molvis.org/molvis/v10/a27/>.
2. Bantseev V, Oriowo OM, Giblin FJ, Leverenz VR, Trevithick JR, Sivak JG. Effect of hyperbaric oxygen on guinea pig lens optical quality and on the refractive state of the eye. Exp Eye Res 2004; 78:925-31.
3. Lei B. The ERG of guinea pig (Cavis porcellus): comparison with I-type monkey and E-type rat. Doc Ophthalmol 2003; 106:243-9.
4. Lu F, Zhou X, Zhao H, Wang R, Jia D, Jiang L, Xie R, Qu J. Axial myopia induced by a monocularly-deprived facemask in guinea pigs: A non-invasive and effective model. Exp Eye Res 2006; 82:628-36.
5. Zhou X, Lu F, Xie R, Jiang L, Wen J, Li Y, Shi J, He T, Qu J. Recovery from axial myopia induced by a monocularly deprived facemask in adolescent (7-week-old) guinea pigs. Vision Res 2007; 47:1103-11.
6. McFadden SA, Howlett MH, Mertz JR. Retinoic acid signals the direction of ocular elongation in the guinea pig eye. Vision Res 2004; 44:643-53.
7. Howlett MH, McFadden SA. Form-deprivation myopia in the guinea pig (Cavia porcellus). Vision Res 2006; 46:267-83.
8. Howlett MH, McFadden SA. Emmetropization and schematic eye models in developing pigmented guinea pigs. Vision Res 2007; 47:1178-90.
9. Woerpel RW, Rosskopf WJ Jr. Avian-exotic animal care guides. Goleta (CA): American Veterinary publication; 1988.
10. Wistow G. The NEIBank project for ocular genomics: data-mining gene expression in human and rodent eye tissues. Prog Retin Eye Res 2006; 25:43-77.
11. Ewing B, Green P. Base-calling of automated sequencer traces using phred. II. Error probabilities. Genome Res 1998; 8:186-94.
12. Gordon D, Abajian C, Green P. Consed: a graphical tool for sequence finishing. Genome Res 1998; 8:195-202.
13. Romualdi C, Bortoluzzi S, D'Alessi F, Danieli GA. IDEG6: a web tool for detection of differentially expressed genes in multiple tag sampling experiments. Physiol Genomics 2003; 12:159-62.
14. Romualdi C, Bortoluzzi S, Danieli GA. Detecting differentially expressed genes in multiple tag sampling experiments: comparative evaluation of statistical tests. Hum Mol Genet 2001; 10:2133-41.
15. Sasaki J, Phillips BJ, Chen X, Van Eps N, Tsai AL, Hubbell WL, Spudich JL. Different dark conformations function in color-sensitive photosignaling by the sensory rhodopsin I-HtrI complex. Biophys J 2007; 92:4045-53.
16. Sokolov M, Strissel KJ, Leskov IB, Michaud NA, Govardovskii VI, Arshavsky VY. Phosducin facilitates light-driven transducin translocation in rod photoreceptors. Evidence from the phosducin knockout mouse. J Biol Chem 2004; 279:19149-56.
17. Brown BM, Carlson BL, Zhu X, Lolley RN, Craft CM. Light-driven translocation of the protein phosphatase 2A complex regulates light/dark dephosphorylation of phosducin and rhodopsin. Biochemistry 2002; 41:13526-38.
18. Murakami A, Yajima T, Inana G. Isolation of human retinal genes: recoverin cDNA and gene. Biochem Biophys Res Commun 1992; 187:234-44.
19. Kelliher P, Connor TJ, Harkin A, Sanchez C, Kelly JP, Leonard BE. Varying responses to the rat forced-swim test under diurnal and nocturnal conditions. Physiol Behav 2000 Jun 1-15; 69:531-9.
20. Cheng MY, Bittman EL, Hattar S, Zhou QY. Regulation of prokineticin 2 expression by light and the circadian clock. BMC Neurosci 2005; 6:17.
21. Fuller CA, Ishihama LM, Murakami DM. The regulation of rat activity following exposure to hyperdynamic fields. Physiologist 1993; 36:S121-2.
22. Ogden AT, Nunes I, Ko K, Wu S, Hines CS, Wang AF, Hegde RS, Lang RA. GRIFIN, a novel lens-specific protein related to the galectin family. J Biol Chem 1998; 273:28889-96.
23. Ida H, Boylan SA, Weigel AL, Smit-McBride Z, Chao A, Gao J, Buchoff P, Wistow G, Hjelmeland LM. EST analysis of mouse retina and RPE/choroid cDNA libraries. Mol Vis 2004; 10:439-44 <http://www.molvis.org/molvis/v10/a55/>.
24. Allikmets R, Seddon JM, Bernstein PS, Hutchinson A, Atkinson A, Sharma S, Gerrard B, Li W, Metzker ML, Wadelius C, Caskey CT, Dean M, Petrukhin K. Evaluation of the Best disease gene in patients with age-related macular degeneration and other maculopathies. Hum Genet 1999; 104:449-53.
25. Rada JA, Nickla DL, Troilo D. Decreased proteoglycan synthesis associated with form deprivation myopia in mature primate eyes. Invest Ophthalmol Vis Sci 2000; 41:2050-8.