|Molecular Vision 2006;
Received 29 November 2005 | Accepted 18 April 2006 | Published 8 May 2006
Gene expression and discovery during lens regeneration in mouse: regulation of epithelial to mesenchymal transition and lens differentiation
Mario Medvedovic,1 Craig R. Tomlinson,1 Mindy K.
Call,2 Matthew Grogg,2
1Genomics and Microarray Laboratory, Department of Environmental Health, Center for Environmental Genetics, University of Cincinnati Medical School, Cincinnati, OH; 2Laboratory of Molecular Biology, Department of Biology, University of Dayton, Dayton, OH
Correspondence to: Panagiotis A. Tsonis, Laboratory of Molecular Biology, Department of Biology, University of Dayton, Dayton, OH, 45469-2320; Phone: (937) 229-2579; FAX: (937) 229-2021; email: firstname.lastname@example.org
Purpose: It has been shown that after extracapsular lens removal by anterior capsulotomy in the mouse, the lens can be regenerated. However, as the capsular bag is filled with fibers, epithelial to mesenchymal transition (EMT), an event which is common after cataract surgery as well, takes place during early stages. This study, using a unique mouse model, was undertaken to identify novel regulators and networks in order to more clearly understand secondary cataracts at the molecular level.
Methods: We examined global gene expression via microarray analysis of mouse lens regeneration after extracapsular surgery. Gene expression at different times after surgery was correlated with the processes of EMT, which is seen in the initial stages of regeneration, and lens fiber differentiation, which occurs later.
Results: Several notable patterns were observed from the gene clustering data. It was obvious from the analysis that initially there is a response to injury, extensive matrix remodeling, and severe downregulation of genes encoding lens structural proteins. The patterns returned gradually to normal three weeks after surgery. New genes were identified from the clustering results that might be potential regulators of EMT and lens differentiation.
Conclusions: With this approach, we demonstrated the utility of a mouse model to study secondary cataracts at the molecular level. Extension of these studies in mice with known mutations affecting EMT or lens differentiation should allow the identification of the crucial molecular players that could lead to better treatments of secondary cataracts.
Traditionally, the newt has been hailed as the most powerful animal model for lens regeneration [1,2]. True enough, adult newts always replace their lens following removal. Lens regeneration in newts is achieved by transdifferentiation of the pigment epithelial cells from the dorsal iris. Other amphibia, such as frogs, are capable of lens regeneration by transdifferentiation of the cornea, but only during a short window of time before metamorphosis . The situation in higher vertebrates, especially in mammals, is very different. Lens regeneration has been shown in rabbits, but only if the lens capsule is left behind . Obviously, some lens epithelial cells remain attached to the lens capsule and they differentiate to lens fibers to "regenerate" a lens, which nevertheless is not perfect. Some similar, but limited, observations have been seen in cats . The studies with rabbits suggest that while lens regeneration does not follow the same traditional road of transdifferentiation as in newts, regeneration can nevertheless occur by differentiation of lens epithelial cells remaining on the capsule. Rabbits (or cats), however, are not favorable mammalian animal models for approaching the problem of lens regeneration with the frontline technology of molecular biology and, therefore, extensive studies at the molecular level are hindered. We reported previously that when the lens is removed in adult mice, leaving the capsule in the eye cavity, lens fibers rapidly differentiate from the adherent lens epithelial cells and fill the capsule within a few weeks, and is not limited to Soemmerring's ring only . Such "regeneration" of the lens is quite remarkable and has been reported in mouse and rats by others as well [7,8]. Importantly, epithelial to mesenchymal transition (EMT) has been observed at the initial stages in mice undergoing regeneration, indicating that the process undergoes an initial phase of repair and of lens differentiation.
EMT is a prominent process after cataract surgery. During modern cataract surgery, extracapsular lens removal allows the synthetic lens to be placed on the remaining capsule. However, adherent lens epithelial cells (LECs) still persist in most cases. LECs tend to transdifferentiate to mesenchymal cells, and this process leads to the so-called posterior capsule opacification (PCO), which requires expensive laser treatment, even though such procedures have been considerably reduced recently because of the state of the art instruments and the techniques applied . The most common experimental procedures to study EMT in LECs are either treating LECs in vitro or to injure lenses in vivo. Both rats and mice are commonly used for this assay. The in vivo procedure is usually performed by injury of the anterior subcapsular region with a needle. Such injury leads to cataract-related changes involving EMT. Culture of capsular bags from humans is another experimental system where proliferation and transdifferentiation of LECs can be studied . In these experiments, the capsular bags from deceased donors can be placed in culture, and the degree and development of EMT can be assessed in vitro. Such studies have provided important information about the factors involved, but are limited for genetic manipulations. TGF-β is considered an important factor in the initiation of EMT [11-13].
The ability of mice to regenerate the lens under the outlined conditions provides us with a valuable animal model system to study basic biology of EMT at the molecular level and to identify targets that eventually could lead to further understanding of the medical complications and the design of effective treatments. The availability of mutant mice and microarray analysis enables us to profile genomic activity during EMT, which is otherwise impossible to carry out. Furthermore, such studies as the ones presented here might elucidate factors that will improve the quality of the regenerated lens, information that will be important in future applications when the lens is accidentally damaged. In the present study, we have examined gene expression during different stages of lens regeneration in mice via microarray hybridization and analysis. Our results provide unique and interesting insights in gene regulation during EMT and lens differentiation.
C57BL mice (six to eight months old) were anesthetized with either intraperitoneal or subcutaneous injections of ketamine (87 mg/kg) and xylazine (13 mg/kg). Mice were also subcutaneously given the analgesic buprenorphine (2 mg/kg) preemptively. Pupils were dilated with the use of 1% tropicamide and 2.5% phenylephrine hydrochloride. A corneal incision was made, and anterior capsulotomy was performed. The lens was then removed by pushing the eye cavity with forceps. The anterior chamber was filled with sodium hyaluronate. In our hands, this procedure effectively removes the whole lens (we have examined removed lenses and lentectomized eyes by histology) and leaves the capsule behind, eventually with LECs. Figure 1 shows the morphology of the capsule one day after lens removal.
Microarray hybridization methods were used to obtain global gene expression profiles from intact and regenerating eyes after extracapsular lens removal in C57BL mice eight weeks of age. We examined four times, time 0 at the time of lens removal and 1, 2, and 3 weeks after surgery. In our previous publication  we presented a histological study of the early stages of these events. During week one we observed fiber differentiation and EMT. Week two was basically marked by increased fiber differentiation and a lower degree of EMT. By week three, EMT was virtually absent. Since the goal of this study was to identify genes that affect EMT and lens fiber differentiation, these times were sufficient because both EMT and fiber differentiation occur rapidly after surgery and follow a particular course.
A 70-mer oligonucleotide library from Operon Technologies, Inc. (Huntsville, AL) representing 24,878 known mouse genes (e.g., genes involved with cell signaling, apoptosis, cell proliferation, etc.; including most, if not all, of the available known genes involved in regeneration) were used for the microarray experiments.
As depicted in Figure 2, each regenerating tissue was directly compared to the corresponding intact tissue. RNA from control tissues was obtained from four independent animals and hybridized with RNA obtained from the same number of independent animals at week one and three animals at weeks two and three after the surgery. Biological variation was accounted for by including the multiple biological replicates per experimental condition. Our statistical model does indeed unequivocally factor out the gene-specific dye effect from the estimates of differential expression. This is achieved by fitting a linear statistical model with a "dye" effect, as described in the data normalization and analysis section of the methods, to each gene separately. This approach has been demonstrated to work well in unbalanced situations such as the week two and three comparisons.
Isolation of tissues and total RNA
Total RNA was isolated by standard methods using the Nucleospin RNA purification kit from BD Bioscience (San Jose, CA). We analyzed the quality of mRNA using an Agilent Bioanalyzer 2100 and NanoDrop 1000.
For each hybridization experiment (a microarray slide), total RNA from two single animals were used. Approximately 10 μg of total RNA was used for each Cy-3 or Cy-5 labeling procedure. cDNA target was synthesized using an indirect labeling method, in which aminoallyl-dUTP (7:3 ratio of aa-dUTP:TTP) was incorporated in the cDNA via an oligo(dT) primed reaction by reverse transcriptase (Superscript III; Invitrogen, Carlsbad, CA). The cDNA was decorated with Cy-3 and Cy-5 (Cy DyeTM Post-Labeling Reactive Dye Packs; Amersham, GE Healthcare, Piscataway, NJ) following the accompanying instructions. When necessary, the RNA was amplified using the Amino Allyl MessageAmpTM kit from Ambion (Austin, TX), which in our hands produced approximately 50-120 μg of amplified RNA (aRNA) with the incorporated amino allyl nucleotides starting from one μg of total RNA (10 μg of each aRNA was used per slide). The aRNA is an accurate representation of the original total cellular RNA .
The mouse 70-mer oligonucleotides were suspended in 3X SSC at 30 μM and printed at 22 °C and 65% relative humidity on aminosilane-coated slides (VSA-25C; Cel Associates, Inc. Pearland, TX) using a high-speed robotic OmniGrid machine (GeneMachines; San Carlos, CA) with Stealth SMP3 pins (Telechem, Sunnyvale, CA) [15,16]. The microarray slides were placed in prehybridization buffer (5X SSC, 0.1% SDS, and 1% BSA) and incubated at 48 °C for 45-60 min. The slides were washed twice in deionized water and used immediately for hybridization (2X hybridization buffer: 50% formamide, 10X SSC, and 0.2% SDS). The Cy-3 and Cy-5 labeled targets were suspended in nine μl water and heated at 95 °C for 3 min. The following were added to each tube of labeled target to inhibit nonspecific hybridization: eight μl of 1 mg/ml COT1-DNA (Roche Diagnostics, Basel, Switzerland), two μl of 10 mg/ml poly(A)-DNA (Sigma, St. Louis, MO), and two μl of 4 mg/ml yeast tRNA (Sigma). Next, 21 μl of 2X hybridization buffer preheated to 48 °C was added to the target mixture, mixed well, and centrifuged. The labeled target was applied to the prehybridized microarray slides, covered with a 22x60 mm glass cover slip, and placed in a sealed hybridization chamber (Corning, Acton, MA). The sealed chamber was placed in a 48 °C water bath and incubated for 40-60 h. For the posthybridization washes, the coverslips were removed in 1X SSC, 0.1% SDS, and 0.1 mM DTT at 48 °C, and the slides were agitated for 15 min. The microarray slides were transferred to a staining dish containing 0.1X SSC, 0.1% SDS, and 0.1 mM DTT at 48 °C and agitated for 5 min. The aforedescribed wash was repeated two more times. The slides were then washed two times in 0.1X SSC and 0.1 mM DTT at room temperature and agitated for 5 min. The slides were then spin dried .
Scanning and data generation
Imaging was carried out using a GenePix 4000A and GenePix 4000B (Axon Instruments, Union City, CA) with GenePixPro 5.0 software. Images were captured in JPEG and TIFF files, and the DNA probes were measured by the adaptive circle segmentation method. Information extraction for a given spot was calculated using the median value for the signal pixels minus the median value for the background pixels to produce a gene set data file for all the DNA spots. The Cy-3 and Cy-5 fluorescence signal intensities were normalized by adjusting total fluorescence levels.
Data normalization and analysis
The data representing raw spot intensities generated by GenePix® Pro version 5.0 was analyzed to identify differentially expressed genes. Data normalization was performed in three steps for each microarray separately . First, channel-specific local background intensities were subtracted from the median intensity of each channel (Cy3 and Cy5). Second, background adjusted intensities were log-transformed and the differences (R) and averages (A) of log-transformed values were calculated as R=log2(X1/X2) and A=[log2(X1 X2)]/2, where X1 and X2 denote the Cy5 and Cy3 intensities after subtracting local backgrounds, respectively. Third, data centering was performed by fitting the array-specific local regression model of R as a function of A. The difference between the observed logarithmic ratio and the corresponding fitted value represented the normalized log-transformed gene expression ratio. Normalized log-intensities for the two channels were then calculated by adding half of the normalized ratio to A for the Cy5 channel and subtracting half of the normalized ratio from A for the Cy3 channel. A statistical analysis was performed for each gene separately by fitting the following mixed effects linear model. Yijk=μ+Ai+Ck+Tj+εijk, where Yijk corresponds to the normalized log-intensity on the ith array (i=1,...,10), at the jth time (j=1,2,3), and labeled with the kth dye (1 for Cy5 and 2 for Cy3). μ is the overall mean log-intensity, Ai is the effect of the ith array, Tj is the effect of the jth time, and Ck is the effect of the kth dye. Assumptions about model parameters were the same as described in the literature  with array effects assumed to be random and treatment and dye effects assumed to be fixed. Statistical significance of differential expression between RNA samples at each time after the treatment, after adjusting for array and dye effects, was assessed by calculating p values and applying False Discovery Rates (FDR) multiple hypotheses testing [19,20]. Data normalization and statistical analyses were performed using SAS statistical software package (SAS Institute Inc., Cary, NC).
Clustering was performed with the Bayesian infinite mixture (BIM) model-based clustering for replicated microarray data [21,22] using replicated normalized logarithmic ratios from each microarray. BIM model-based clustering allowed for the fitting of the statistical mixture model without knowing the number of clusters in the data . The statistical model was fitted using the Gibbs sampler, and hierarchical clustering was produced by treating pair-wise posterior probabilities as the similarity measure and applying the traditional average-linkage principle. The clustering results were displayed using the TreeView program .
Clusters of co-expressed gene identified by the cluster analysis were correlated with functional groupings defined by Gene Ontologies (GO) . Clusters of genes with significantly over-represented genes from specific GO categories were identified using the EASE software . Statistical significance of over-representation of genes from a cluster in any given GO category was assessed using the Fisher's exact test with the Benjamini-Hochberg adjustment for multiple hypothesis testing . A GO category was considered to be significantly associated with a cluster if it contained more than one gene from the cluster and the adjusted Fisher's exact p value was less than 0.1.
Quantitative real-time polymerase chain reaction
RNA was isolated from intact eyes and eyes undergoing lens regeneration using Tri Reagent® (Molecular Research Center, Inc., Cincinnati, OH) according to the manufacturer's instructions. RNA (0.75 μg) was used to synthesize cDNA using the iScriptTM cDNA Synthesis Kit (BioRad). All Real-Time PCRs were performed using the iCyclerTM (BioRad). For each Real-Time PCR reaction run in triplicate, 2 μl of cDNA, 800 nM primers, and iQTM SYBR® Green Supermix (BioRad) were used. The data were analyzed using the Pfaffl method .
Results & Discussion
Labeled target representing mouse mRNA from lens was used to hybridize to arrayed 70-mer probes representing nearly 25,000 mouse genes. Gene expression profiles of regenerating lens (one, two, and three weeks post-lentectomy) were compared to the expression profiles of intact lens (Figure 1). We identified the genes that were significantly differentially expressed during the regeneration process. In all, we identified 2,094 genes that showed regulation during regeneration (FDR<0.05 in at least one comparison). Six clusters of co-expressed genes defining distinct patterns of expression were significantly correlated with at least one GO category (FDR<0.1 and more than one gene from a cluster was a member of a given GO category).
A general pattern emerged indicating that during the first week post-lentectomy, there was an increase in RNA levels of genes involved in tissue repair, inflammation, and reorganization of the cytoskeleton and the extracellular matrix (Figure 3). On the other hand, there was a significant decrease in RNA levels of genes encoding lens structural proteins, such as crystallins and other lens-fiber specific markers. As differentiation and growth of the lens ensued, some of the differentially expressed genes gradually returned to control levels of expression. The profile of the crystallins indicated that their synthesis followed the normal developmental program. At the same time, we observed that some genes never reached control levels. Another interesting and novel discovery from the clustering analysis was that RNA levels decreased for genes involved in transcription and protein synthesis and may be a key early event.
The overall pattern clearly follows two different biological processes that take place after the extracapsular operation. In the initial stages, there is EMT and considerable remodeling of the extracellular matrix. At later stages, a lens differentiation program takes over due to regeneration of lens fibers. This observation is also clear from a list of the top 50 differentially expressed genes that showed the greatest increase in mRNA levels across the different times after surgery relative to control lens (Table 1). Thrombospondin-1 (TSP-1) precursor showed the greatest fold-change increase. TSP-1 is a glycoprotein involved in the activation of TGF-β, which is considered to be the main inducing factor of EMT . TSP-1 has been shown to accumulate during PCO and decline during fiber differentiation . Other highly upregulated genes encode proteins that are involved in matrix remodeling, such as procollagen, TIMP-1, cathepsin, tenascin C, proteinases, and leucine-rich repeat containing protein (Table 1). Among the 50 genes that showed the greatest decrease in mRNA levels in the regenerating lens relative to the control (Table 2) are genes that encode structural proteins of differentiated lens fibers. The list includes several crystallins, phakinin, beaded filament structural protein, lens fiber membrane intrinsic protein, and lens fiber major intrinsic protein. Also, several regulatory genes, such as the homeo box NKX-2.2, the Kruppel factor 7, the cAMP responsive element binding protein, and NFAT are clustered with the lens fiber-specific ones (Table 2). In Table 1 and Table 2, the time with the highest (or lowest) regulation is highlighted in red. This was designed to help the reader to identify with a glance the times and the genes showing the most regulation. Interestingly, it becomes obvious that at week 2 we have the most severe regulation, positive or negative. These findings may eventually allow us to identify specific gene regulation programs involved in the distinct processes of EMT and fiber differentiation that take place during the process of mouse lens regeneration. The only genes that also coincide with cataract loci are the crystallin genes.
We further examined five general patterns of expression identified by correlating the clusters formed by the cluster analysis of gene expression profiles and functional clusters based on GO categories.
Weak uniform increase in RNA levels
In this group, the clustered genes showed a general pattern of a relatively slight increase in RNA levels throughout the regeneration process. In Table 3, the genes are divided according to main GO category, biological process, cellular component, and molecular function. Table 4 presents a general feature of this subgroup in that it contains genes involved in defense, response to injury, and extracellular matrix metabolism and also includes TGF-β and TGF-β-binding proteins, which are known mediators of EMT.
Strong uniform increase in RNA levels
As in the previous group, the genes in this cluster showed a general increase in RNA levels but more pronounced. These genes are involved in immune response, adhesion and remodeling, and processes that mediate injury and re-building of tissues after damage. Thrombospondins and disintegrins are included in the list (Table 5, Table 6).
Strong delayed increase in RNA levels
The mRNA levels of the genes in this group showed a sharp increase at week 2. Most of these genes are involved in cytoskeletal organization and negative regulation of transcription (Table 7, Table 8).
Weak early decrease in RNA levels
The main characteristic of this group was that the genes are involved in nucleic acid biosynthesis and ribosomal function. This result suggests that during the early events of repair, there is a general inhibition of transcriptional and translational events (Table 9, Table 10).
Strong uniform decrease in RNA levels
The genes in this cluster are involved in sensory organ development, perception of light, and the structural components of the lens (Table 11, Table 12). mRNA levels for the crystallins and other structural proteins of the lens, such as phakinin, are severely decreased indicating that lens fiber differentiation is not at its final stages during the repair process. Naturally, the drop in RNA levels of some genes in this group becomes less severe with the later stages of lens fiber differentiation (3 weeks post-lentectomy).
Verification of expression by quantitative real-time polymerase chain reaction
We selected ten genes to verify their expression by quantitative real-time polymerase chain reaction (QPCR). The selected genes showed different patterns of expression in the microarray experiments. TIMP1 showed a strong uniform increase in RNA levels, lysozyme showed strong increase at week 2, ceruloplasmin showed a weak uniform increase and γB-crystallin showed strong uniform decrease in RNA levels. Others showed not much variation and had lower levels. Ratios observed in microarray experiments for low expressed genes are most likely more variable for overall poorly expressed genes than for highly expressed genes. However, since we are using the statistical significance as the main criteria for identifying differentially expressed genes, such higher variability will be accounted for. That is, genes with higher variability in observed ratios will have lower statistical significance than genes with low variability ratios. Therefore, the statistical significance of low-expressed genes has been implicitly adjusted in our analysis and the statistically significant genes have equal chance of being false positives regardless of the overall level of expression. Nevertheless, we also decided to test such genes. As seen in Figure 4, expression of these genes as examined by QPCR was in excellent agreement with the microarray data. The housekeeping gene ATP synthase, epsilon subunit was used as the reference. This gene was found to have no differential expression in our microarray analysis and showed no differential expression in the QPCR experiments as well.
The mouse model for lens regeneration that we have described previously  is a valuable one because both EMT and lens fiber differentiation take place. Specifically, while the capsular bag is filled gradually with fibers, EMT is seen during the early stages and diminishes later. This has led us to utilize this model and examine global gene expression in order to associate clustered genes with both processes and identify new genes and networks. The availability of mutant mice will supplement these studies. By extensive genomic studies with mice lacking genes involved in EMT or lens fiber differentiation, the patterns of gene expression reported in this study could be sorted out in order to identify the role and regulation of known and novel genes involved in these processes. Extension of these studies, therefore, will lead to the establishment of databases and will provide indispensable and long-sought animal models for approaching PCO at the genetic level. At the same time, these studies will complement databases related to ocular bioinformatics [28-34].
This research was supported in part by NEI grant EY10540 to PAT.
1. Del Rio-Tsonis K, Tsonis PA. Eye regeneration at the molecular age. Dev Dyn 2003; 226:211-24.
2. Tsonis PA, Del Rio-Tsonis K. Lens and retina regeneration: transdifferentiation, stem cells and clinical applications. Exp Eye Res 2004; 78:161-72.
3. Freeman G. Lens regeneration from the cornea in Xenopus laevis. J Exp Zool 1963; 154:39-65.
4. Gwon AE, Gruber LJ, Mundwiler KE. A histologic study of lens regeneration in aphakic rabbits. Invest Ophthalmol Vis Sci 1990; 31:540-7.
5. Gwon A, Gruber LJ, Mantras C. Restoring lens capsule integrity enhances lens regeneration in New Zealand albino rabbits and cats. J Cataract Refract Surg 1993; 19:735-46.
6. Call MK, Grogg MW, Del Rio-Tsonis K, Tsonis PA. Lens regeneration in mice: implications in cataracts. Exp Eye Res 2004; 78:297-9.
7. Lois N, Dawson R, McKinnon AD, Forrester JV. A new model of posterior capsule opacification in rodents. Invest Ophthalmol Vis Sci 2003; 44:3450-7.
8. Lois N, Taylor J, McKinnon AD, Forrester JV. Posterior capsule opacification in mice. Arch Ophthalmol 2005; 123:71-7.
9. Ibaraki N. A brighter future for cataract surgery. Nat Med 1997; 3:958-60.
10. Duncan G, Wormstone IM, Liu CS, Marcantonio JM, Davies PD. Thapsigargin-coated intraocular lenses inhibit human lens cell growth. Nat Med 1997; 3:1026-8.
11. Lee EH, Joo CK. Role of transforming growth factor-beta in transdifferentiation and fibrosis of lens epithelial cells. Invest Ophthalmol Vis Sci 1999; 40:2025-32.
12. Lovicu FJ, Schulz MW, Hales AM, Vincent LN, Overbeek PA, Chamberlain CG, McAvoy JW. TGFbeta induces morphological and molecular changes similar to human anterior subcapsular cataract. Br J Ophthalmol 2002; 86:220-6.
13. Wormstone IM, Tamiya S, Anderson I, Duncan G. TGF-beta2-induced matrix modification and cell transdifferentiation in the human lens capsular bag. Invest Ophthalmol Vis Sci 2002; 43:2301-8.
14. Iscove NN, Barbara M, Gu M, Gibson M, Modi C, Winegarden N. Representation is faithfully preserved in global cDNA amplified exponentially from sub-picogram quantities of mRNA. Nat Biotechnol 2002; 20:940-3.
15. Guo J, Sartor M, Karyala S, Medvedovic M, Kann S, Puga A, Ryan P, Tomlinson CR. Expression of genes in the TGF-beta signaling pathway is significantly deregulated in smooth muscle cells from aorta of aryl hydrocarbon receptor knockout mice. Toxicol Appl Pharmacol 2004; 194:79-89.
16. Karyala S, Guo J, Sartor M, Medvedovic M, Kann S, Puga A, Ryan P, Tomlinson CR. Different global gene expression profiles in benzo[a]pyrene- and dioxin-treated vascular smooth muscle cells of AHR-knockout and wild-type mice. Cardiovasc Toxicol 2004; 4:47-73.
17. Sartor M, Schwanekamp J, Halbleib D, Mohamed I, Karyala S, Medvedovic M, Tomlinson CR. Microarray results improve significantly as hybridization approaches equilibrium. Biotechniques 2004; 36:790-6.
18. Wolfinger RD, Gibson G, Wolfinger ED, Bennett L, Hamadeh H, Bushel P, Afshari C, Paules RS. Assessing gene significance from cDNA microarray expression data via mixed models. J Comput Biol 2001; 8:625-37.
19. Benjamini Y, Hochberg Y. Controlling the false discovery rate: a practical and powerful approach to multiple testing. Journal of the Royal Statistical Society. Series B (Statistical Methodology) 1995; 57:289-300.
20. Reiner A, Yekutieli D, Benjamini Y. Identifying differentially expressed genes using false discovery rate controlling procedures. Bioinformatics 2003; 19:368-75.
21. Medvedovic M, Sivaganesan S. Bayesian infinite mixture model based clustering of gene expression profiles. Bioinformatics 2002; 18:1194-206.
22. Medvedovic M, Yeung KY, Bumgarner RE. Bayesian mixture model based clustering of replicated microarray data. Bioinformatics 2004; 20:1222-32.
23. Eisen MB, Spellman PT, Brown PO, Botstein D. Cluster analysis and display of genome-wide expression patterns. Proc Natl Acad Sci U S A 1998; 95:14863-8.
24. Ashburner M, Ball CA, Blake JA, Botstein D, Butler H, Cherry JM, Davis AP, Dolinski K, Dwight SS, Eppig JT, Harris MA, Hill DP, Issel-Tarver L, Kasarskis A, Lewis S, Matese JC, Richardson JE, Ringwald M, Rubin GM, Sherlock G. Gene ontology: tool for the unification of biology. The Gene Ontology Consortium. Nat Genet 2000; 25:25-9.
25. Hosack DA, Dennis G Jr, Sherman BT, Lane HC, Lempicki RA. Identifying biological themes within lists of genes with EASE. Genome Biol 2003; 4:R70.
26. Pfaffl MW. A new mathematical model for relative quantification in real-time RT-PCR. Nucleic Acids Res 2001; 29:e45.
27. Saika S, Miyamoto T, Ishida I, Barbour WK, Ohnishi Y, Ooshima A. Accumulation of thrombospondin-1 in post-operative capsular fibrosis and its down-regulation in lens cells during lens fiber formation. Exp Eye Res 2004; 79:147-56.
28. Wistow G. A project for ocular bioinformatics: NEIBank. Mol Vis 2002; 8:161-3 <http://www.molvis.org/molvis/v8/a22/>.
29. Wistow G, Bernstein SL, Wyatt MK, Behal A, Touchman JW, Bouffard G, Smith D, Peterson K. Expressed sequence tag analysis of adult human lens for the NEIBank Project: over 2000 non-redundant transcripts, novel genes and splice variants. Mol Vis 2002; 8:171-84 <http://www.molvis.org/molvis/v8/a24/>.
30. Wistow G, Bernstein SL, Ray S, Wyatt MK, Behal A, Touchman JW, Bouffard G, Smith D, Peterson K. Expressed sequence tag analysis of adult human iris for the NEIBank Project: steroid-response factors and similarities with retinal pigment epithelium. Mol Vis 2002; 8:185-95 <http://www.molvis.org/molvis/v8/a25/>.
31. Chauhan BK, Reed NA, Zhang W, Duncan MK, Kilimann MW, Cvekl A. Identification of genes downstream of Pax6 in the mouse lens using cDNA microarrays. J Biol Chem 2002; 277:11539-48.
32. Wride MA, Mansergh FC, Adams S, Everitt R, Minnema SE, Rancourt DE, Evans MJ. Expression profiling and gene discovery in the mouse lens. Mol Vis 2003; 9:360-96 <http://www.molvis.org/molvis/v9/a50/>.
33. Ahmed F, Torrado M, Zinovieva RD, Senatorov VV, Wistow G, Tomarev SI. Gene expression profile of the rat eye iridocorneal angle: NEIBank expressed sequence tag analysis. Invest Ophthalmol Vis Sci 2004; 45:3081-90.
34. Hawse JR, Hejtmancik JF, Huang Q, Sheets NL, Hosack DA, Lempicki RA, Horwitz J, Kantorow M. Identification and functional clustering of global gene expression differences between human age-related cataract and clear lenses. Mol Vis 2003; 9:515-37 <http://www.molvis.org/molvis/v9/a65/>.
35. Dobbin K, Shih JH, Simon R. Statistical design of reverse dye microarrays. Bioinformatics 2003; 19:803-10.