Identification of Genes Downstream of Pax6 in the Mouse Lens Using cDNA Microarrays* 210

Pax6 is a transcription factor that regulates the development of the visual, olfactory, and central nervous systems, pituitary, and pancreas. Pax6 is required for induction, growth, and maintenance of the lens; however, few direct Pax6 target genes are known. This study was designed to identify batteries of differentially expressed genes in three related systems: 8-week old Pax6 heterozygous lenses, 8-week old Pax6 heterozygous eyes, and transgenic lenses overexpressing PAX6(5a), using high throughput cDNA microarrays containing about 9700 genes. Initially, we obtained almost 400 differentially expressed genes in lenses from mice heterozygous for a Pax6 deletion, suggesting that Pax6 haploinsufficiency causes global changes in the lens transcriptome. Comparisons between the three sets of analyses revealed that paralemmin, molybdopterin synthase sulfurylase,Tel6 oncogene (ETV6), a cleavage-specific factor (Cpsf1) and tangerin A were abnormally expressed in all three experimental models. Semiquantitative reverse transcription (RT)-PCR analysis confirmed that all five of these genes were differentially expressed in Pax-6 heterozygous and Pax6(5a) transgenic lenses. Western blotting and immunohistochemistry demonstrated that paralemmin is found at high levels in the adult lens and confirmed its down-regulation in the Pax6(5a)-transgenic lenses. Collectively, our data provide insights into the genetic programs regulated by Pax6 in the lens.

Pax6 is among the most widely studied transcription factors because of its participation in the organogenesis of the eye, brain, head, and pancreas (1)(2)(3)(4). The essential role of Pax6 in early eye induction is conserved throughout the evolution of multicellular animals with ectopic expression able to direct conversion of wing imaginal disks to eyes in Drosophila mela-nogaster (2) and head ectoderm to lenses in Xenopus laevis (5). In the vertebrate eye, Pax6 is required for lens placode formation, growth of the lens (6), correct placement of a single retina in the eye (7), formation of the iris, maintenance of the corneal epithelium, and fate of retinal progenitor cells (8).
The diverse functions of Pax6 appear to originate from both the complex regulatory mechanisms controlling the tissue-specific transcription and splicing of the Pax6 mRNA as well as its ability to participate in multiple molecular interactions. A prevailing form of Pax6 in mouse embryos contains two DNAbinding domains, the paired domain and homeodomain (HD), 1 which can interact both independently and cooperatively with DNA, whereas the C terminus comprises the transcriptional activation domain (9,10). The paired domain contains two subdomains, PAI and RED, each of them capable of binding independently to DNA (9). A splice variant, Pax6(5a), has an additional 14 amino acids inserted into the PAI subdomain. This results in its recognition of only a subset of Pax6 binding sites (11)(12)(13)(14). Recent evidence suggests that Pax6 function is further modulated by interactions of its homeodomain with a diverse set of proteins, including the homeodomain-containing proteins Six3, Prox1, and Lhx2 (15) and the transcription factors TFIID and pRb (16). Pax6 also physically interacts with c-Maf/Maf A (17) and MitF (microphthalmia) (18), two important transcription factors controlling lens differentiation, and retinal development, respectively.
While Pax6 is clearly a central player in many developmental processes, relatively few genes have been shown to be directly regulated by Pax6. In Drosophila, Pax6/ey directly regulates the transcription of rhodopsins (19) and sine oculis (20). In vertebrates, Pax6 directly affects expression of Pax2 in the developing optic cup and stalk (21). Genetic evidence suggests that the genes for the eye development regulators Eya1 and -2 (22), Sox-2 (7), and c-Maf (23) are also direct targets. In addition to these developmental regulators, Pax6 can directly regulate the insulin, glucagon, and somatostatin genes expressed in the pancreas (22); L1-CAM expression in the brain (25); keratin K12 (26) and gelatinase B (27) expression in the cornea; and ␣A-, ␣B-, ␦1-, ␤B1and -crystallin expression in the lens (28 -30). Although the mechanism of Pax6 function has not been studied in detail in many of these cases, it appears that it can function both as a transcriptional activator and repressor (30) in in vitro assays.
cDNA microarray technology has been developed to decipher the complex genetic networks altered in response to environmental insults and disease (31)(32)(33). Here, this technology is used to study Pax6 function by determining which genes are affected by both Pax6 haploinsufficiency in the eye (1, 6 -8) and Pax6(5a) overexpression (14) in the lens. Our studies demonstrate the usefulness of microarray analysis for the analysis of gene expression in pathological conditions and give some insight into the function of Pax6 in the mature lens.

EXPERIMENTAL PROCEDURES
Mice-NMRI mice heterozygous for a Pax6 knockout/lacZ knock-in allele were generously provided by Dr. Peter Gruss (Max-Planck-Institute of Biophysical Chemistry, Gottingen, Germany) (34), while wild type NMRI mice were obtained from Charles River Laboratories (L'Arbresle, France). FVB/N mice overexpressing Pax6(5a) in lens fiber cells under the control of the mouse ␣A-crystallin promoter and wild type strain matched controls were described previously (14).
RNA Preparation-Lenses and eyes from which lenses were surgically removed were isolated from 8-week-old Pax6 heterozygous and wild type mice and stored in RNAlater (Ambion, Woodlands, TX) until RNA isolations were performed using the Totally RNA kit (Ambion). The genotype of Pax6 heterozygous lenses was confirmed by assaying the expression of Pax6 and lacZ using RT-PCR using primers designed to amplify Pax6 (5Ј-TTT AAC CAA GGG CGG TGA GCA G-3Ј and 5Ј-TCT CGG ATT TCC CAA GCA AAG ATG-3Ј) and lacZ mRNAs (5Ј-GTC AGG TCA TGG ATG AGC AG-3Ј and 5Ј-CAC TAC GCG TAC TGT GAG C-3Ј) employing the One Step RT-PCR system (Invitrogen). The initial RT step was conducted at 50°C for 30 min, and amplifications were conducted at the annealing temperature of 58°C.
Lenses were isolated from 3-and 8-week mice overexpressing PAX6(5a) in lens fiber cells and strain-matched controls as described (14), and RNA was immediately prepared using the SV Total RNA Isolation System (Promega, Madison, WI).
Microarray Procedures-cDNAs were generated using 2-5 g of total RNA and indirectly labeled with Cy3-and Cy5-specific dendrimers, employing the 3DNA detection system from Genisphere, Inc. (Montvale, NJ) (35) according to the manufacturer's protocol. Glass slide microarrays containing about 9700 mouse sequence verified genes were described elsewhere (36). The hybridizations were performed at 50°C, with three subsequent washes of the slides performed in 2ϫ SSC, 0.2% SDS; 2ϫ SSC; and 0.2ϫ SSC buffers. The chips were scanned using the GenePix 4000A scanner (Axon Instruments, Union City, CA), and primary data were analyzed using the Genepix 3.02 software. Each experiment was conducted in triplicate. Control self-hybridizations were performed using wild type RNA to determine S.D. values that were used to determine the eventual cut-off values.
Quality Control, Data Analysis, and Statistics-Primary data were flagged using four default parameters set in the Genepix 3.0 program. Intensity data for both channels were normalized by the widely used global intensity normalization method (37). The intensity of each spot in each channel was adjusted by subtracting the local background from the observed intensity (IЈij ϭ Iij Ϫ Bij, where IЈij, Iij, and Bij are the adjusted intensity, observed intensity, and background for the jth gene (j ϭ 1, 2, . . . n) in the ith channel (i ϭ 1, 2), respectively), and then subjected to log transformation (ln(IЈij)). The overall intensity for each channel was calculated by taking the power of the average of the log of the adjusted intensity for all genes (I Ch1 ϭ e ln(IЈ1j)/n , and I Ch2 ϭ e ln( IЈ 2 j)/n, where n is the number of genes, and I Ch1 and I Ch2 are the overall intensity for channel 1 and channel 2, respectively). The intensities for both channels were therefore balanced by multiplying the adjusted intensity of each spot in channel 2 by the ratio of the overall intensity in channel 1 over that in channel 2 (r ϭ I Ch1 /I Ch2 ). Means and S.D. values were calculated for those genes with no more than one flagged data point. For normalized data tables, see the Supplemental Material. Genes were classified into 12 functional groups (38) using annotations from the Swissprot data base (available on the World Wide Web at ca.expasy.org/sprot). The tangerin A was identified from EST AA217475 using the Gencarta data base (Compugen Ltd., Tel Aviv, Israel).
RT-PCR-All transcripts studied were reverse-transcribed and amplified using the One Step RT-PCR system (Invitrogen). The initial reverse transcriptase step was conducted at 50°C for 30 min. The annealing temperatures used for individual experiments are indicated in Table I. All amplifications shown here were performed at 29 cycles. All primers used in this study (Table I) were designed to cross intronexon boundaries and were tested in the absence of reverse transcriptase. Control reactions were performed initially to ensure linearity of amplification over concentrations of total RNA ranging from 5 to 100 ng.
Western Blotting-Lenses were dissected from 6-week-old Pax6(5a) transgenic and wild type litter mates and homogenized in radioimmune precipitation buffer (1ϫ phosphate-buffered saline, 1% Nonidet P-40, 0.5% sodium deoxycholate, 0.1% SDS, 0.575 mM phenylmethylsulfonyl fluoride, 45 g/ml aprotinin, 1 mM sodium orthovanadate). Supernatants were collected following two spins at 10,000 ϫ g. Protein concentrations were immediately determined using Bio-Rad DC protein assay (Bio-Rad), and 56 g of protein were loaded on each lane of a 10% discontinuous SDS-PAGE gel. The protein was transferred to nitrocellulose and incubated with a 1:2000 dilution of anti-paralemmin rabbit crude serum (39). Bound antibodies were detected with 1:2000 horseradish peroxidase-linked anti-rabbit IgG (New England BioLabs). Blots were developed using LumiGLO (Cell Signaling Technologies, Beverly, MA) and exposed to film.
Immunohistochemical Labeling and Preparation-Mouse eyes were enucleated and embedded in Tissue Freezing Medium (Triangle Biomedical Sciences, Durham, NC), and 16-m frozen sections were prepared. Sections were then fixed in 1:1 acetone/methanol for 10 min at Ϫ20°C and blocked with 1% bovine serum albumin/phosphate-buffered saline for 1 h at RT. Paralemmin and preimmune paralemmin primary antibodies (39) were prepared in 1% bovine serum albumin/phosphatebuffered saline at dilutions of 1:150. The bound primary antibodies were visualized following incubation with anti-rabbit IgG conjugated with Alexa Fluor 568 (1:50 dilution in 1% bovine serum albumin/  (40).

Genes Abnormally Expressed in Pax6
Heterozygous Lenses-Initially, lens RNA obtained from 8-week-old NMRI mice was labeled with either Cy5-or Cy3-labeled dendrimers (35) and self-hybridized to cDNA microarrays containing about 9700 sequence-verified genes (36) to determine the S.D. value of the hybridization ratios (Cy5/Cy3). Since the S.D. obtained from three independent hybridizations was 0.28 -0.31, expression ratios more than 1.60 for up-regulated and less than 0.63 for down-regulated genes are statistically significant, since they represent values that differ by two S.D. values. Genes differentially expressed in Pax6 heterozygous lenses as compared with normal lenses were determined by labeling cDNAs from the two samples with Cy3-and Cy5-specific dendrimers before simultaneously probing onto the cDNA microarrays described above. A representative scatter plot from one experiment showing the distribution of hybridization signals generated using GeneSpring 3.02 software (Silicon Genetics, San Carlos, CA) is shown in Fig. 1A. To demonstrate the reproducibility of data using the 3DNA labeling technology (35), we randomly selected six genes (GenBank TM accession numbers AA387340, AA120030, AA445775, AA238399, AA260490, and AA000249), and their ratios of expression from triplicate microarrays are given in Fig. 1B. This is the first report, to our knowledge, using the 3DNA detection system (Genisphere) and poly-Llysine-coated slides used to print the microarrays, allowing one to work with 2-5 g of total RNA without any mRNA amplification step (33). In addition, the standard deviation of the control experiment, 0.28 -0.31, was comparable with direct incorporation methods employing Cy3Ј and Cy5Ј modified UTP, which typically yielded values between 0.19 and 0.21. 2 Normalized data tables can be obtained on the World Wide Web at www.aecom.yu.edu/thecvekllab. Some of the data were flagged due to the unacceptable signal intensity above background intensity (i.e. if the signal intensity above background intensity was less than 100, then the spot would be flagged). This resulted in the identification of more than 400 differentially expressed genes; the vast majority of them were down-regulated, consistent with Pax6 roles as an activator of transcription. From these data, three lists of genes were generated. The first list, shown in Table II, includes genes with known functions classified into 12 subcategories (38), flagged no more than once, and expressed in Pax6 heterozygous lenses at reduced levels up to a factor of 0.63 and up-regulated genes by a factor of at least 1.6. When a single flag was found, we included the data if mean and median values were similar. The second list, shown in Table III, includes known genes that could not be classified into one of the 11 functional categories. The third list, shown in Table IV, includes ESTs showing strong and moderate similarities with genes deposited in public data bases and contained no more than one flagged value, as described above. Two genes down-regulated in Pax6 heterozygous lenses and relevant to known lens biology are homeodomain-containing transcription factor Pitx3 and structural ␤A4-crystallin (41).
Genes Abnormally Expressed in Pax6 Heterozygous Lenses, Eyes, and PAX6(5a) Transgenic Lenses to Identify Most Commonly Differentially Expressed Genes-Since such a large number of genes were differentially expressed, it was difficult to select which genes were appropriate for further analysis. Thus, a parallel series of microarray hybridizations were conducted using RNAs obtained from normal and Pax6 heterozygous eyes that have had the lenses surgically removed. This analysis 2 B. K. Chauhan and A. Cvekl, unpublished data.
FIG. 1. Gene expression monitored with the use of the 3DNA labeling system. A, a representative experiment displayed as a scatter plot obtained using mouse lens cDNAs from normal (labeled with Cy-3 dendrimers) and Pax6 heterozygous (labeled with Cy-5 dendrimers). Horizontal and vertical axes represent Cy-3-and Cy-5-generated signals in logarithmic scale shown in green and red, respectively. Shown are cut-off points (indicated by horizontal and vertical dashed lines) for genes producing signals below 100. Genes used for subsequent analysis are located in the upper right quadrant, labeled by a gray arrow. The upper and lower boundaries (blue lines) represent a 1.6-fold difference in the expression between Pax6 heterozygous and normal lenses, and the inner slope (blue line) indicates a ratio of 1.0 as described under "Experimental Procedures." The yellow signal represents unchanged genes. B, a diagrammatic representation of six randomly selected genes annotated with their GenBank TM accession numbers (see "Experimental Procedures"), from the pool of unchanged, down-and up-regulated genes. The horizontal axis represents three independent microarrays, labeled 1, 2, and 3; and the vertical axis represents the ratio of relative Cy-5/Cy-3 intensities for each independent hybridization. A disintegrin and metalloproteinase domain 9 (meltrin ␥) AA210306 Ϫ3.3 Cappa1 Capping Secreted modular calcium-binding protein 1 AA000223 Ϫ2 Thrombospondin 4 AA003452 Ϫ1.9 Tubb5 Tubulin, ␤ 5 W16254 Ϫ1. Paired-like homeodomain transcription factor 3 AA062140 Ϫ3.9 revealed that 79 genes were differentially expressed in Pax6 heterozygous eye, with 14 of these genes being common with the Pax6 heterozygous lens (Fig. 2). Since important genes expressed predominantly in the lens could have been missed in this analysis, a final set of microarray hybridizations were also performed using wild type FVB/N lenses and transgenic lenses overexpressing the PAX6(5a) splice variant in lens fiber cells (14) in the same mouse strain, and five genes were differentially expressed. These genes are paralemmin (39), molybdopterin synthase sulfurylase (MOCS3), 3 the Tel6 oncogene (ETV6) (43), a cleavage-specific factor (Cpsf1), and tangerin A 4 encoding a large protein found in brain cDNA libraries. In the group of nine down-regulated genes between Pax6 heterozygous lenses and eyes, the majority are ESTs without any apparent function. The exception is a mouse EST highly similar to human p20-CGGBP, a (5Ј-CGG-3Ј) n -binding transcription factor (45). In contrast, the group of five genes down-regulated in Pax6 heterozygous and 5a-transgenic lenses contains mostly known genes. One of these genes is ubiquitin-conjugating enzyme 7 (Ubc7p) spotted and identified twice on the microarray in this group, confirming the reproducibility of our experiments. Based on these findings, we decided to pursue confirmations on paralemmin, MOCS3, Tel6 oncogene (ETV6), a cleavage-specific factor (Cpsf1), and tangerin A at the mRNA level and for paralemmin at the protein level as well.

RT-PCR Analysis of Differentially Expressed Genes-
The differential expressions of mRNAs encoding paralemmin, MOCS3, Tel6 oncogene, a cleavage-specific factor, and tangerin A were confirmed using semiquantitative RT-PCR. As a control, the genes encoding argininosuccinate synthetase and uridine monophosphate kinase and encoding ubiquitin-conjugating enzyme 4 and ubiquinone biosynthesis gene coq7/clk1 were selected from the pool of unchanged genes obtained from the microarray experiments with Pax6 heterozygous and PAX6(5a) transgenic lenses, respectively. As expected, no difference in expression of these genes was found (Figs. 3B and 4B) using the same amounts of RNA of the compared samples. Using these conditions, we also found reduced expression of Pax6 mRNA in Pax6-heterozygous lenses (Fig. 3B). Using a set of specific primers, we then showed that expression of paralemmin, MOCS3, Tel6 oncogene, a cleavage-specific factor (Cpsf1), and tangerin A was indeed reduced both in the Pax6 heterozygous lenses (Fig. 3C) and in transgenic lenses overexpressing PAX6(5a) (Fig. 4C).
Paralemmin Protein Levels Are Decreased in the Pax6(5a) Transgenic Lens but Not the Pax6 Heterozygous Lenses-Immunohistochemical labeling was performed to confirm the differential gene expression of paralemmin at the protein level. In normal lenses, paralemmin protein is detected at the cell membrane in all lens cells (Fig. 5). In Pax6(5a) transgenic lenses, its expression was decreased throughout the fiber mass while appearing normal in the germinative and transition zone (Fig.  5). In lenses from Pax6 heterozygous mice, its expression appeared to be slightly increased in the lens epithelium as well as primary and secondary fibers (Fig. 5). Western blot confirmed reduced level of paralemmin in the extracts of 5a-transgenic lenses (Fig. 5).
Candidate Pax6 Binding Sites in Paralemmin, MOCO3, Pitx3, and ␤A4-Crystallin Genes-To examine the possibility that Pax6 directly binds to regulatory regions of paralemmin, MOCO3, Pitx3, and ␤A4-crystallin, we inspected the putative promoter regions of these genes for potential Pax6 binding sites. Selection of these genes is explained under "Discussion." For the search, we used a consensus binding sequence for the Pax6 paired domain, P6CON (46). Since Pax6 binds a diverse spectrum of target sequences, prediction of binding sites was limited to 1 kb of genomic DNA fragments 5Ј from the end of the longest known transcripts. The sites found using the "Findpatterns" algorithm from the GCG package (Oxford Molecular Group, Campbell, CA) are given in Fig. 6. To demonstrate sequence heterogeneity in known Pax6 binding sites, five well characterized Pax6 binding sites exhibiting 3-6 mismatches from P6CON are shown for comparison. Collectively for paralemmin, MOCO3, Pitx3/PITX3, and ␤A4-crystallin, we found nine Pax6 candidate binding sites with 3-5 mismatches from the 20-bp-long P6CON sequence. Additional experiments are required to test the direct regulation of paralemmin, Moco3, ␤A4-crystallin, and Pitx3 genes by Pax6. DISCUSSION A large body of information about Pax6 has been accumulated over the last decade since its discovery in 1991 (1-4, 28, 29). Pax6 is essential for normal vertebrate visual system development and has been implicated as a regulatory gene for other regulatory genes and various structural genes. To date, genes directly regulated by Pax6 were either identified by educated guesses (14, 19 -30) or from in situ studies of gene expression in Pax6 homozygous embryos (47)(48)(49). Multiple functions of Pax6 during the organogenesis of the eye, brain, and pancreas indicate that the majority of genes regulated by Pax6 remain to be discovered.
Pax6 heterozygous mouse lenses are smaller than normal and develop anterior subcapsular cataracts, which appear to be due to epithelial/mesenchymal transition of lens epithelial cells (50,51). The abnormal lens size appears to result from a 3 The GenBank TM accession number for mocs3 is AF102544. 4 The GenBank TM accession number for tangerin A is AF305087.  reduction in the number of cells incorporated into the lens vesicle, perhaps due to decreased cell division in the lens placode. However, cell division rates appear to be normal after lens formation (6). In contrast, the cataracts appear to develop due to impaired responses of the lens epithelium to the eye environment (52), which may explain why Pax6 ϩ/Ϫ epithelial cells in Pax6 heterozygote/wild type chimeras undergo preferential lens fiber cell differentiation (53). While the phenotypic alterations in the Pax6 heterozygous lens are typically mild, the gene expression profiling described here suggests that the transcriptome is globally disrupted. Since the current estimations of genes in of the mouse and human genome are between 40,000 to 70,000, and the cDNA microarrays used here contain over 9000 independent genes, it is possible to conservatively predict that at least 2000 genes are differentially expressed in the Pax6 heterozygous lenses.
The genes discovered to be differentially expressed in the Pax6 heterozygous lenses are likely to fall into three classes. First are genes whose expression is directly controlled by Pax6 whose reduced expression is the result of a reduction in Pax6 expression (Fig. 3). The genes in the second group are reduced, since their expression is controlled by transcription factors (such as c-Maf, Eya1 and -2, or Sox2) whose expression is controlled by Pax6 (7,22,23). The third group of genes are those differentially expressed as an indirect result of the phenotypic alterations seen in Pax6 heterozygous lenses such as the presence of myofibroblasts (51,52) and elevated levels of total tissue calcium (51).
Since the gene profiling described here was performed to discover both direct Pax6 target genes and those whose expression is controlled by Pax6-dependent pathways in the lens, the set of genes differentially expressed in the Pax6 heterozygous lens was compared with those whose expression was altered in the remainder of the eye. While this observation reduced the number of genes to be analyzed to 19, this left the concern that important Pax6-dependent genes expressed predominantly in the lens would be improperly sidelined. Thus, expression profiling was also performed on lenses from transgenic mice overexpressing the Pax6(5a) splice form in fiber cells, which normally contain reduced amounts of Pax6 compared with the lens epithelium (54). While a remarkably small number of genes were found to be differentially expressed in Pax6(5a) lenses in light of the drastic alterations in lens fiber cell morphology

FIG. 2. Venn diagram showing total numbers of genes downregulated in Pax6 heterozygous lenses (SEL), eyes minus lenses (SEE), and Pax6(5a)-transgenic lenses (5a).
Five genes in the intersection of SEL/SEE/5a (white), nine genes from SEL/SEE (yellow), and five genes from SEL/5a (purple) are displayed. caused by this ectopic expression (14), it was remarkable that 13 of the 27 genes were also differentially expressed in either Pax6 heterozygous lenses or eyes, whereas five were differentially expressed in all three data sets. While the Pax6(5a) transgenic mice express additional Pax6(5a) protein in the lens, and the Pax6 heterozygous lenses express reduced levels of both Pax6 and Pax6(5a), it is interesting to note that additional copies of the entire Pax6 locus caused ocular and lens abnormalities similar to the Pax6 haploinsufficiency (55). In addition, many reporter genes activated by low concentrations of Pax6 in transient transfections are repressed as Pax6 con-centrations increase (28). Thus, it is possible that overexpression of Pax6(5a) in the lens could cause repression of a small group of genes that are otherwise activated by Pax6/Pax6(5a) in normal lenses. We focused on these five genes, and using semiquantitative RT-PCR analysis, we showed that expression of paralemmin, MOCS3, Tel6 oncogene, a cleavage-specific factor (Cpsf1), and tangerin A was indeed reduced both in the Pax6 heterozygous lenses (Fig. 3C) and in transgenic lenses overexpressing PAX6(5a) (Fig. 4C). We also found significantly reduced paralemmin in 5a-transgenic but not in Pax6 heterozygous lenses (Fig. 5). We do not know the reason for this unex-  2 and 4) that were used for each assay. Note that the levels of argininosuccinate synthetase and uridine monophosphate kinase are unchanged, while Pax6 expression is reduced between Pax6 ϩ/Ϫ and wild type lenses. C, differential expression of paralemmin, MOCS3, Tel6 oncogene, tangerin A, and Cspf1 in Pax6 heterozygous (Pax6 ϩ/Ϫ) and normal (Wild-type) lenses. Lanes 1-4 and triangles are described above. The sizes of specific PCR products are given in parentheses; the 100-bp DNA ladder is shown on the left.

FIG. 4. Confirmation of microarray results obtained from wild type (WT) and Pax6(5a) transgenic (5a) lenses using semiquantitative RT-PCR.
A, a representative gel showing specific amplification of ubiquitin-conjugating enzyme 4 in the presence (ϩRT) and absence (ϪRT) of reverse transcriptase. B, experiment showing similar expression levels of two unchanged genes, ECE4 and ubiquinone biosynthesis gene coq7/clk1, and increased expression level of Pax6(5a) in Pax6(5a)-transgenic lenses. C, differential expression of paralemmin, tangerin A, MOCS3, Tel6, and Cspf1 in 5a-transgenic (5a) and normal (Wild-type) lenses. Lanes 1 and 2 represent total RNA isolated from wild type lenses, and lanes 3 and 4 represent total RNA isolated from Pax6 heterozygous lenses. The triangles indicate the two amounts of RNA (50 ng in lanes 1 and 3 and 17 ng in lanes 2 and 4) that were used for each assay. Sizes of specific PCR products are given in parentheses; the 100-bp DNA ladder is shown on the left. pected discrepancy between RNA and protein data; however, several factors could contribute to this finding. First, there may be differences in paralemmin protein turnover between these lenses. Second, Pax6 haploinsufficiency may cause reduced transcription of paralemmin but increased translation and accumulation of the protein. For example, little Dach mRNA is detected in lens fibers by in situ hybridization (56,57), but protein accumulation is easily detected by immunohistochemistry (58). Finally, it is possible that paralemmin mRNA levels are abundant in the lens epithelium and activated by Pax6, but lower levels of paralemmin transcription normally occur in fibers that are further repressed by high levels of Pax6(5a). These possibilities are not mutually exclusive, since a combination of differential RNA abundance and differential translation in epithelial and fiber cells may best explain these data.
Our finding of paralemmin expression in the lens raises the possibility of its specific role in the shape of lens fiber cells. Paralemmin is a putative new morphoregulatory protein highly expressed in the forebrain and cerebellum, two prominent tissues expressing Pax6 (59), but also less abundantly in many other tissues (39). It has been proposed that paralemmin plays a specific role in the plasma membrane architecture of neurons (39). The known expression pattern of Pax6 and paralemmin combined with the presence of candidate Pax6 binding site in its putative promoter region raises the possibility that the Pax6 directly regulates its expression. Future studies will be aimed at addressing roles of paralemmin in lens morphology and its transcriptional control.
MOCS3 catalyzes one of the final steps in the formation of the organic complex of molybdenum (Moco) (60). Moco is a cofactor essential to the function of three enzymes: sulfite oxi-dase, xanthine dehydrogenase, and aldehyde oxidase. Deficiency of Moco results in neurological abnormalities, mental retardation, and, in some cases, dislocated lenses (61, 62). Comparably, dislocated lenses are also found in some aniridia patients (63), suggesting indeed the possibility that MOCS3 is downstream of PAX6. Our data show that the expression levels of MOCS3 are reduced, but not abolished, in the lens and indicate the presence of several Pax6 binding sites (Fig. 6). Hence, the data raise the possibility that Pax6 may act as a modulatory transcription factor fine tuning expression of these genes in the lens, eye, and possibly other tissues in mammalian brain. However, the expression pattern of MOCS3 in the mammalian models is not known due to the lack of antibodies.
We also confirmed reduced expression of three remaining genes from the pool of five genes down-regulated in all three models studied (see Fig. 2): Tel6 oncogene, a cleavage-specific factor (Cpsf1), and tangerin A in both Pax6 heterozygous lenses and PAX6(6a)-transgenic lenses. Tel6/ETV6 encodes a widely expressed transcriptional repressor that is often rearranged in human leukemias (43). Tangerin A is a novel large protein with a calponin homology domain. 4 The calponin homology domain is present in many actin binding cytoskeletal and signal-transducing proteins (64). Interestingly, there are more genes with reduced expression in Pax6 heterozygous lenses (i.e. mACF7, Cappa1, Dsn, and Pfn2) that also bind actin and are shown in Table II (group Cellular organization). In the absence of relevant information about these genes and/or specific reagents, we did not study them further; nevertheless, their confirmed abnormal expression supports the reliability of the cDNA microarray technology.
While not differentially expressed in Pax6(5a) lenses or Pax6 heterozygous eyes, two genes differentially expressed in Pax6 heterozygous lenses, ␤A4-crystallin and Pitx3, require special attention. The known expression of ␤A4-crystallin and of Pitx3 in the lens, supported by phenotypes of mouse and human Pitx3 mutants, makes these two genes prime candidate direct targets for Pax6-mediated regulation of transcription. Pitx3, a homeobox-containing gene, is required for normal lens devel-FIG. 6. Candidate Pax6 binding sites in 1-kb genomic regions of human paralemmin, MOCS3, ␤A4-crystallin, and PITX3 and mouse Pitx3. Genomic fragments of 1 kb upstream from the longest EST found in public data bases were analyzed for the presence of Pax6 binding sites using a consensus sequence, P6CON (5Ј-ANNTTCACGC-WTSANTKMNY-3Ј) (56). P6CON is divided into two regions. The 5Јhalf of the binding site is recognized by PAI, ␤-turn, and linker to RED. 3Ј-half of P6CON is recognized by RED (42). Five well known Pax6 binding sites from the chicken ␦1-, mouse ␣Aand ␣B-, and guinea pig -crystallins (28) and mouse L1 CAM (25) are shown for comparison. Conserved (shown in red) and nonconserved (shown in black) nucleotides were as follows: W represents A/T; S represents G/C; K represents G/T; M represents A/C; and Y represents T/C; and N represents A/C/ G/T. The number of mismatches (n) between the P6CON and the individual sequences is given.

FIG. 5.
Paralemmin is down-regulated at the protein level in the Pax6(5a) lens. A, Western blot of paralemmin expression in 6-week Pax6(5a) transgenic and wild type mouse lens. Note that paralemmin protein levels are significantly reduced in the lenses of Pax6(5a) transgenic mice. B, paralemmin expression in a 4-week-old wild type lens. Note that high levels of paralemmin protein are associated with fiber cell membranes. Magnification was ϫ200. C, paralemmin expression in a 4-week-old Pax6(5a) transgenic lens. Note that the paralemmin staining seen in lens fibers is greatly reduced. Magnification was ϫ200. D, paralemmin expression in a 3-week-old Pax6 heterozygous lens. Magnification was ϫ200. E, central lens epithelium and fiber cells of a 4-week-old wild type mouse. Magnification was ϫ630. F, central lens epithelium and fiber cells of a 4-week-old Pax6(5a) transgenic mouse. Magnification was ϫ630. G, central lens epithelium and fiber cells of a 3-week-old Pax6 heterozygous mouse. Magnification was ϫ630. H, interface between the zone of denucleation and nucleated lens fibers in a 4-week-old wild type mouse. Magnification was ϫ630. I, central dysgenic fibers of a 4-week-old Pax6(5a) transgenic mouse. Magnification was ϫ630. J, 3-week-old wild type lens stained with preimmune serum. Magnification was ϫ630. wt, wild type; 5a-t, Pax6(5a) transgenic; e, lens epithelium; f, fiber cells; sf, secondary fiber cells; t, transition zone; zd, zone of denucleation. Red, paralemmin; green, DNA opment (65,66), and mutations in PITX3 cause congenital cataract (67). The expression pattern of the Pitx3 gene in the mouse lens supports the possibility that Pax6 might directly control its expression (48), and we have confirmed reduced expression of Pitx3 in Pax6 heterozygous lenses by RT PCR analysis as described elsewhere. 5 Pitx3 expression in lens (from E11) found in the lens vesicle and later in the entire lens with the highest expression in the anterior epithelium and lens equator (65). Thus, the developmental patterns of Pax6 and Pitx3 spatially overlap with Pitx3 expression following Pax6 (59,65). ␤A4-crystallin appears to be highly expressed only in the lens. In the absence of published data, we searched human and mouse TIGR expression data bases (available on the World Wide Web at www.tigr.org/tdb) and found that human ␤A4crystallin was predominantly expressed in the lens (22 clones) and that mouse ␤A4-crystallin clones were found mainly in total embryos (11 clones) but not in individual tissue-specific libraries. Only two clones were found in human retina and one clone in human placenta-and mouse skin and epidermis-derived libraries, respectively. Two putative Pax6 binding sites were found in the human ␤A4-crystallin 5Ј-flanking region, and two evolutionary conserved Pax6 binding sites were found in human and mouse 5Ј-flanking sequences of PITX3/Pitx3 genes.
Several genes found in the present study are good candidates to explain abnormal lens development in Pax6 heterozygous lenses (6) and properties of chimeric lenses from wild type and Pax6 ϩ/Ϫ cells (53). The selective exclusion of Pax6 heterozygous lens cells from the E16.5 lens was probably due to the aberrant expression of cell adhesion and extracellular matrix proteins (53). Our list of differentially expressed genes in 8-week-old Pax6 heterozygous lenses includes several membrane-associated proteins (i.e. Adam9, cadherin 3, Img/LIG-1, paralemmin/Palm, Spnb3, and Thbs4 (cellular organization genes in Table II) and a receptor-like tyrosine kinase Ddr1 (signal transduction genes in Table II). Some of these proteins may also play critical roles during the lens embryogenesis. Our data also show reduced level of expression of cyclin E2 (69,70). Similarly, they raise the possibility that reduced cellular proliferation of Pax6 heterozygous lens precursor cells might be linked to the reduced level of cyclin E2 expression during the formation of the lens pit (6). We also found up-regulation of plasma membrane calcium-transporting ATPase (PMCA1/ ATP2B1; EC 3.6.3.8). This ATPase is an integral membranebound protein that transports Ca 2ϩ out of the cell and is expressed in lens epithelium (71). Previously, we have shown increased levels of intracellular Ca 2ϩ in Pax6 heterozygous lenses (Sey ϽDeyϾ /ϩ) (51). These findings raise the possibility that PMCA1 up-regulation is a secondary response of the lens to deal with the elevated Ca 2ϩ . Future experiments to probe the molecular aspects of lens induction, growth, and physiology will be facilitated by this pool of genes. Finally, some of the genes identified in this study (i.e. paralemmin, plexin 2, receptor-like tyrosine kinase (Ddr1), nuclear FMRP-interacting protein (Nufip), and synaptotagmin 4) may exhibit abnormal expression in Pax6 heterozygous and homozygous brains, which may contribute to psychiatric disorders linked to the Pax6 haploinsufficiency (68,72) and PAX6 promoter polymorphism (44).
In conclusion, our data demonstrate that Pax6 haploinsufficiency results in the reduced expression of a large number of genes in the lens, while the remainder of the eye was somewhat less affected. However, comparisons between gene expression alterations in mice lacking one copy of Pax6 and expressing ectopic Pax6(5a) in the lens have revealed a population of 10 genes for future study that have the potential to be direct or indirect Pax6 targets. Our data base will serve as an important gateway for future systemic analysis of genes acting in the same genetic pathway as Pax6, not only in the lens but in other tissues expressing Pax6. In the long term, such differential gene expression information will be useful in understanding how Pax6 mutations result in the diverse eye and brain diseases observed in humans.