Identification of Pax6-Dependent Gene Regulatory Networks in the Mouse Lens

Lineage-specific DNA-binding transcription factors regulate development by activating and repressing particular set of genes required for the acquisition of a specific cell type. Pax6 is a paired domain and homeodomain-containing transcription factor essential for development of central nervous, olfactory and visual systems, as well as endocrine pancreas. Haploinsufficiency of Pax6 results in perturbed lens development and homeostasis. Loss-of-function of Pax6 is incompatible with lens lineage formation and results in abnormal telencephalic development. Using DNA microarrays, we have identified 559 genes expressed differentially between 1-day old mouse Pax6 heterozygous and wild type lenses. Of these, 178 (31.8%) were similarly increased and decreased in Pax6 homozygous embryonic telencephalon [Holm PC, Mader MT, Haubst N, Wizenmann A, Sigvardsson M, Götz M (2007) Loss- and gain-of-function analyses reveals targets of Pax6 in the developing mouse telencephalon. Mol Cell Neurosci 34: 99–119]. In contrast, 381 (68.2%) genes were differently regulated between the lens and embryonic telencephalon. Differential expression of nine genes implicated in lens development and homeostasis: Cspg2, Igfbp5, Mab21l2, Nrf2f, Olfm3, Spag5, Spock1, Spon1 and Tgfb2, was confirmed by quantitative RT-PCR, with five of these genes: Cspg2, Mab21l2, Olfm3, Spag5 and Tgfb2, identified as candidate direct Pax6 target genes by quantitative chromatin immunoprecipitation (qChIP). In Mab21l2 and Tgfb2 promoter regions, twelve putative individual Pax6-binding sites were tested by electrophoretic mobility shift assays (EMSAs) with recombinant Pax6 proteins. This led to the identification of two and three sites in the respective Mab21l2 and Tgfb2 promoter regions identified by qChIPs. Collectively, the present studies represent an integrative genome-wide approach to identify downstream networks controlled by Pax6 that control mouse lens and forebrain development.


Introduction
Embryonic organ development is contingent on complex coordinated interactions of multiple transcription factors that regulate the expression of selected downstream target genes. The Paired (Pax) family of genes has been shown to control development of many organs, such as brain, ear, eye, kidney, muscle, pancreas and thyroid [1,2]. Nine mammalian Pax genes, Pax1 to Pax9, encode specific DNA-binding transcription factors that recognize DNA via their N-terminal paired domains, PDs. Among them, Pax6 is essential for eye, brain, olfactory and pancreas development [3][4][5][6]. Gaining insight into the genes regulated directly by Pax6 is fundamental in deciphering its role in the genetic regulatory networks governing embryonic development.
During early stages of visual system formation, Pax6 is required for the establishment of lens progenitor cells [7] and for multipotency of retinal progenitor cells [8]. In later stages of eye development, Pax6 plays a number of complex functions during anterior segment [9], lacrimal gland [10] and neuroretina [11] development. The formation of lens progenitor cells appears to require co-expression of at least three genes, Pax6, Six3 and Sox2, in the pre-placodal region of the mouse embryo. Expression of these genes is linked to FGF/MAPK, BMP4 and BMP7 signaling pathways. During the growth and invagination of lens placode, Pax6 controls expression of c-Maf, Foxe3, Mab21l1, N-cadherin, Prox1 [7] and one or more components of retinoic acid (RA) signaling, such as Raldh3/Aldh1a3 [12,13]. In the differentiating lens fiber cells, Pax6 regulates expression of a number of crystallin genes [7,14] and the a5b1 integrin complex [15,16]. In addition to these specific targets, we predict that Pax6 may be engaged in transcriptional regulation of additional cohorts of genes during lens development that can be identified by genome-scale studies.
Two recent high throughput studies, focused on the function of Pax6 in mouse telecephalon, have shown a number of novel genes whose expression is controlled by Pax6. A DNA microarray study of Pax6 2/2 embryonic dorsal (cortex) and ventral (ganglionic eminence, GE) telencephalon, E12 and E15, identified a number of novel Pax6-regulated genes [17]. High-throughput in situ hybridization analysis of hundreds of co-expressed genes in midgestation mouse embryo (E14.5) resulted in the prediction of 30 genes regulated by Pax6 in the embryonic cortex, with , one half considered as putative direct target genes [18]. In Drosophila, DNA microarray expression studies identified batteries of genes regulated by eyeless (ey), a fly homologue of Pax6 [19,20].
In this study, we have analyzed differential gene expression in newborn mouse Pax6 heterozygous and wild type lens and compared this to expression profiles in embryonic forebrain from Pax6 homozygous embryos [17]. We show that approximately 1/3 of differentially expressed transcripts regulated by Pax6 are shared between these model tissues. In contrast, the majority (2/3) of the transcriptional profile represented by the set of 559 transcripts is differently regulated by Pax6 in the respective different cell types. Therefore, our data suggest that specific cellular environments promote common and distinct functions of Pax6. We identified Cspg2, Mab21l2, Olfm3, Spag5, and Tgfb2, as five novel putative direct Pax6-target genes in mouse lens. Finally, we characterized two Pax6-binding sites in the Mab21l2 regulatory region and three Pax6-binding sites in the Tgfb2 promoter.

Results
Identification of 559 genes differentially expressed in Pax6 +/2 compared to wild type lens To identify genes regulated by Pax6 in lens, we compared newborn lenses from Pax6 +/2 to wild type mice (Fig. 1). Pax6 heterozygous lenses (Fig. 1B) are smaller compared to the wild type lens (Fig. 1A), exhibit subtle morphological defects such as abnormal lens fiber cells [21], and often develop a persistent corneal-lenticular stalk (Fig. 1B). In contrast to the nearly homogenous expression of Pax6 in wild type lens epithelium, expression of Pax6 protein in Pax6 heterozygous lenses shows variable staining patterns in the individual cells (see Fig. 1B, and Fig. S1). In addition, the amount of Pax6 protein in the bow region is much smaller, i.e. below the detection limit of the assay, in Pax6 heterozygous (Fig. 1B) lenses compared to the wild type lens (Fig. 1A). Interestingly, these differences between expression patterns of Pax6 in wild type and Pax6 +/2 eyes are not as pronounced in the adjacent presumptive iris/ciliary body (compare Fig. 1A with 1B). Individual newborn mouse lens provides sufficient amount of total RNA for expression profiling. In contrast, hundreds of lenses are required to obtain lens chromatin for ChIPs [22].
To assess differential gene expression in newborn mouse Pax6 heterozygous lenses compared to wild type lenses, we performed DNA microarray hybridizations using the Affymetrix Mouse Genome 430 2.0 Arrays. Three biological replicate experiments were performed and analyzed as described in Methods. Initially, we found 591 differentially expressed transcripts in newborn Pax6 +/2 lenses which represented 559 differentially expressed genes (see Supporting information) from a total number of over 22,000 mouse genes represented on the array. The lens transcriptome was represented by 7,009 (,32%) genes with a median of RMA-normalized raw signal intensities above 100, and 1,844 (,8%) genes with signal intensities between 50-100, respectively. The boxplot shown in Fig. 2A illustrates the variability of gene expression between the two samples studied and reproducibility of each biological replicate. An increased variability of mRNA abundance in Pax6 +/2 lens originates most likely from variable phenotypes of individual mutant lenses [21,23]. Reduced expression of Pax6 transcripts in Pax6 +/2 lenses measured using both the microarrays and by qRT-PCR is shown in Fig. 2B. Within the 559 differentially expressed transcripts, we found 385 genes with reduced, and 174 with increased expression in mouse lens, respectively. This result is consistent with Pax6's role as both a transcriptional activator [22] and repressor [24][25][26] at the molecular level.

Downstream pathways sensitive to Pax6 haploinsufficiency
The 559 genes were organized according to Gene Ontology (GO) categories of biological process, molecular function and cellular compartment (Fig. 2C). This analysis suggests that the majority of genes downstream of Pax6 in mouse lens are involved in signal transduction, cell differentiation, transcriptional regulation, nervous system development and cell adhesion. For example, the GO cellular component categories ''extracellular space'' and ''extracellular matrix'' are occupied by 51 and 11 genes (Fig. 2C), respectively, representing 45% and 11% of all genes assigned to these categories by the GO system [27]. These two specific categories included six genes with established roles in lens biology, Cspg2, Igfbp5, Olfm3, Spock1, Spon1 and Tgfb2 (see Table 1).
A subcategory ''Eye development'' within the GO Biological process (Fig. 2C) yielded four genes, Mab21l2, NeuroD1, Nfil3, and Tgfb2. The expression of Mab21l2 [28], a gene encoding a critical evolutionary conserved regulatory protein [29], is essential for lens development [28]. Thus, we added Mab21l2, a gene up-regulated in Pax6+/2 lenses, to Table 1. To generate a more representative list of genes for subsequent studies, we added Nr2f2 and Spag5, to the genes listed in Table 1. A nuclear orphan zinc fingercontaining receptor Nrf2f2 (COUP-TF2) functions in mouse forebrain development [30], and its transcripts are upregulated in Pax6 heterozygous lenses. Downregulation of mitotic spindleasscociated protein p126, Spag5 [31], classified in the GO category Cell cycle (see Fig. 2C), was found in Pax6 heterozygous lenses. A recent study has shown that overexpression of Pax6 causes defects during mitosis [32].

Validation of microarray results by quantitative RT-PCR
Differential expression of nine genes, Cspg2, Igfbp5, Mab21l2, Nrf2f, Olfm3, Spag5, Spock1, Spon1 and Tgfb2 (Table 1), in Pax6 heterozygous lens was validated using quantitative RT-PCR. This analysis was carried out with cDNAs prepared from independent pools of RNAs, see Methods. The results (see Fig. 3) showed downregulation of Igfbp5, Spag5, Spock1, Spon1 and Tgfb2, and upregulation of Cspg2, Mab21l2, Nrf2f, and Olfm3 transcripts in Pax6 heterozygous lenses. As controls, expression of B2M, Hprt and Sdha was found virtually unchanged in total RNA samples prepared from wild type and Pax6 heterozygous lenses (data not shown).
To further validate the microarray results [33], we performed qRT-PCR analysis of expression of 15 additional genes. Three of these genes, Serpinb6b, Stmn2 and Sultx1, showed more than 2fold increase/decrease of their expression in Pax6 heterozygous lenses. Seven genes (Ctsh, Aldh1a3, Zw10, Wdhd1, Kif22, Rdm1 and Cdh11) showed 1.24 to 1.45-fold up-regulation. Finally, five genes (Rock1, Dnase2b, Gaa, Acvr1b and Camk1d) showed 1.21 to 1.73-fold down-regulation in this system. Analysis of expression of these genes by qRT-PCR confirmed their deregulation in Pax6 heterozygous lenses (Figs. S2 and S3). In summary, using additional biological replicates, we positively validated differential expression of 25 transcripts of the 559 differentially expressed genes in Pax6 +/2 lens.

Identification of genes commonly regulated by Pax6 in mouse lens and brain
Both lens and brain are of the ectodermal germ layer origin. Specific roles of Pax6 were established in the development of both tissues [7,34]. Taking advantage of expression data from mouse Pax6 Sey/Sey embryonic telencephalon [17], we next compared the 559 lens gene list (see above, Supporting information) with differentially expressed genes in E12 and E15 dorsal (cortex, Ctx), and E12 and E15 ventral (the ganglionic eminence, GE) telencephalon [17] as described in Methods. The results identified 178 genes, representing 31.8% of genes identified in Pax6 heterozygous lens, similarly deregulated in various compartments of the mouse telencephalon. Among those genes, kinesin family member 1B (Kif1b), a monomeric motor for anterograde transport of mitochondria [35], was downregulated in each of the five tissues (lens, E12 Ctx, E15 Ctx, E12 GE and E15 GE) examined (Fig. 4). In addition, expression of nine genes (Rdm1, Melk, Ctsh, Crlf3, Bdh, Gaa, Pygl, Celsr1 and Camk1d; see Fig. 4) was changed in lens and in three regions of the embryonic mouse telencephalon (Fig. 4). Similarly, from those 178 genes described above, 41 genes were changed in the lens and in two compartments/stages of the embryonic telencephalon. Finally, 127 genes were changed in the abnormal mouse lens and in one stage/region of the telencephalon. A representative list of 19 genes from this group, organized into several functional subcategories, is shown in Fig. S4. Identification of a relatively high fraction of genes commonly regulated by Pax6 between two distinct mouse embryonic tissues suggests that Pax6 participates in similar regulatory events during both lens and forebrain development.
As lens cells and cortical neurons represent distinct cell types, it is not surprising that the majority of 381 (68.2%) Pax6-differentially expressed genes in lens show ''opposite'' changes in their expression in Pax6 homozygous telencephalon [17]. The ''opposite'' changes represent contrasting up-and down-regulation of an individual gene or no-change in one tissue combined with either up-or downregulation in the other system. A list of 31 genes up-/down-regulated in lens and in all four compartments of embryonic telencephalon is shown in Fig. 5. The function of Epha3, Necab2, NeuroD1, and Pygb (Fig. 5C), and their significance for telencephalic development is given in the Discussion.

Identification of Pax6's presence in lens chromatin of five genes and identification of Pax6-binding sites in Mab21l2 and Tgfb2 loci
The list of differentially expressed genes in lens and telencephalon (see Supporting information) contains both direct and indirect Pax6 target genes. In a separate report (Y.Y., and A.C., unpublished data), we assessed whether Cspg2, Igfbp5, Mab21l2, Nrf2f, Olfm3, Spag5, Sparc1, Spon1 and Tgfb2 genes (Table 1) are Pax6-direct targets by chromatin immunoprecipitation coupled to DNA microarray analysis (ChIP-on-chip) of chromatin obtained from newborn mouse lens. Within this group, the arrays identified presence of Pax6 in five genes, Cspg2, Mab21l2, Olfm3, Spag5, and Tgfb2 (see Figs. S5,S6,S7,S8, and S9, respectively). To confirm Pax6 in these five loci, qChIP assays were performed using primers corresponding to the ''peak'' regions and surrounding regions in Table 1. A summary of genes regulated by Pax6 and studied as its candidate direct targets.

Gene
Function of the protein related to lens biology Reference(s)

Spock1
Sparc/osteonectin, cwcv and kazal-like domain proteoglycan 1 (testican). Mutations in a related gene, Sparc (osteonectin), cause abnormal lens development [90] Spon1 F-spondin, a secreted extracellular matrix protein, highly expressed in the embryonic lens [91] Tgfb2 TGF-b2, Tgfb2 2/2 mouse embryos form corneal-lenticular stalks found in Pax6 +/2 embryos [45] doi:10.1371/journal.pone.0004159.t001 the same locus. The results (see Fig. 6) confirmed presence of Pax6 in genomic regions predicted by the ChIP-on-chip data. Direct binding of Pax6 in Mab21l2 and Tgfb2 genes in lens chromatin suggests two novel regulatory mechanisms for embryonic eye development (see Discussion). Therefore, we wanted to localize the putative Pax6-binding sites within the regions occupied by Pax6 in lens chromatin. Using the 20 base pair Pax6 PD ''consensus'' sequence [36] and a 20 bp Pax6-specific derivative of the 17 base pair PHO (PD/HD) Drosophila paired ''consensus'' [37] described in Methods and in Fig. S10, we identified five putative Pax6-binding sites corresponding to the ''peak'' regions identified by ChIP-on-chip in Mab21l2 and seven Pax6-binding sites in Tgfb2 loci, respectively. Relative expression levels of Cspg2, Igfbp5, Mab21l2, Nr2f2, Olfm3, Spag5, Sparc, Spon1 and Tgfb2 transcripts in wild type (WT, shown in black) and Pax6 +/2 (het, shown in grey) lenses were determined using qRT-PCR as described in Methods. Transcripts encoded by B2M, Hprt and Sdha, were tested as internal references [86], and all were found unchanged between the wild type and Pax6 +/2 lenses. The data are expressed relative to the unchanged expression level of B2M transcripts. doi:10.1371/journal.pone.0004159.g003 Based on these 12 predicted binding sites, we prepared 11 probes for EMSAs (see Fig. S10 and Table S4). As candidate sites 11 and 12 in the Tgfb2 Pax6-binding region were close to each other, a single probe was used. The P6CON probe was used to determine the optimal concentration range of GST-Pax6 recombinant proteins, PD and PD/HD, needed to detect specific protein-DNA complexes as we described earlier [15,38]. Five of eleven probes tested (probe 1, 5, 6, 8 and 11/12, Fig. S10) generated specific protein-DNA complexes that were reduced in the presence of an excess of P6CON cold oligonucleotide competitor, as shown in Fig. 7. Each probe was incubated with similar amounts of recombinant GST-Pax6 proteins; however, the exposure times ranged from 10 to 21 hours, compared to the 2.5 hour exposure needed to visualize Pax6-binding to the ''optimal'' P6 CON probe. These results identified at least one Pax6-   binding site from major ChIP-on-chip ''peaks'' and showed that the binding affinities of the natural Pax6-binding sites were lower compared to the P6CON. In addition, two probes, site 1 and 5, showed increased affinity towards Pax6-PD/HD compared to PD alone, showing that the internal HD modulates the DNA-binding properties of the PD as described for structurally similar Pax3 proteins [39]. The identification of six ''false'' positives further underscores the need to improve prediction of Pax6-binding sites in regulatory regions of its candidate direct targets [15,18,20,40] Discussion A major question in developmental biology is to elucidate the function of lineage-specific DNA-binding transcription factors encoded by ''stage-selector'' genes, such as Gata1, HNF4a, MyoD, Mitf, Nr2e3, Nrl, Oct4, Pdx1, PU.1, Pax5/BSAP, Pax6, Runx1, Runx2 and Sox9, in various cell types. Here, we examined the function of Pax6 in mouse lens development by identifying batteries of genes differentially expressed in Pax6 heterozygous lens; Pax6 homozygous embryos (Sey/Sey) do not form any lens [41]. No lens progenitors are formed in the pre-placodal region surrounding the neural plate of the Pax6 2/2 mouse embryo [42], thus precluding such analysis. As a significant portion of these ''stage-selector'' genes are expressed in multiple developing lineages, (e.g. Pax6 is expressed in surface ectoderm giving rise to the lens and cornea, in the optic vesicle from which the retina forms, in the neural plate and neuronal progenitor cells, in the olfactory epithelium, in the anterior pituitary primordium and endocrine pancreas [43]), we wanted to determine if there is any overlap between Pax6's functions in the lens and in another tissue, the embryonic telencephalon.
Lens-and telencephalon-specific developmental programs regulated by Pax6 Using the Pax6 lens haploinsufficiency experimental model, we identified 559 genes differentially expressed between wild type and Pax6 heterozygous newborn mouse lens. We then took advantage of similar analysis in telencephalon microdissected into the cortex and ganglionic eminence (GE) in Sey/Sey mouse embryos [17] for data comparison. We found that approximately 1/3 of these genes differently expressed in Pax6 haploinsufficient lens were similarly up-or down-co-regulated in four samples dissected from mouse E12 and E15 telencephalon. These 178 genes represent a diverse group in terms of established and/or putative functions of their encoded proteins.
Within this group, we found that a locus encoding TGF-b2, which was positively regulated by Pax6 (Fig. 3), was occupied by Pax6 in lens chromatin (Fig. 6). In addition, using in vitro assays, we identified at least three Pax6-binding sites (Fig. 7) in the Tgfb2 promoter region (27.5 to +2.5 kb) from two ''peaks'' identified by ChIP-on-chip followed by qChIP confirmations. Furthermore, the expression patterns of Pax6 and TGF-b2 overlap in the mouse lens [44]. Both Tgfb2 2/2 and Pax6 +/2 mouse embryos show similar defects in their eyes; specifically, the lens does not separate from the cornea [21,23,45]. Loss-of-function of the TGF-b receptor, Tgfbr2, in mouse E15 and E18 embryos resulted in perturbed expression of Pax6 in the embryonic retina while lens was not analyzed [46]. A recent study has shown repression of Pax6 promoter via TGF-b signaling [47]. Thus, present data combined with a genetic link between Pax6 and Tgfb2 [44][45][46] suggest a regulatory feedback that may directly participate in fine tuning of Pax6 expression in differentiating lens fiber cells [43] and during cortical neurogenesis [48][49][50]. Approximately 2/3 of genes studied here were not co-regulated in lens and telencephalon. This analysis suggests that Pax6 actively promotes expression of specific genes in one tissue, i.e. embryonic cortex, while simultaneously repressing their expression in a different cell type, i.e. lens. Thus, it appears that dual functions of Pax6 as a stage-selector gene are to ''unfold'' a specific developmental program and suppress this specific program in different cellular contexts. For example, transcripts encoding Cspg2, Mab21l2, Olfm3 and Nr2f2 were up-regulated in Pax6 heterozygous lenses (Fig. 3) while downregulated in mouse Pax6 2/2 E12 cortex [17].

Novel insights into genetic networks regulated by Pax6
The ability of an individual regulatory protein to activate and suppress mutually exclusive developmental programs could be a much broader property of the ''stage-selector'' class of regulatory genes, supported by RNA profiling studies of transcription factors HNF4a in liver, small intestine and fetal colon [57,58], Crx in embryonic E10.5 retina and adult brain [59], and Runx2 in bone and tooth development [60][61][62][63]. During hematopoiesis, PU.1 appears to both activate the myeloid differentiation while suppressing the erythroid program [64]. Pax5/BSAP is essential for B-cell development and suppression of alternative cell fates [65]. A recent ChIP-on-chip study of Pax5/BSAP in B-cell development identified binding of this factor during B-cell development [66] consistent with Pax5/BSAP acting as both a transcriptional activator (56%) and a transcriptional repressor (44% of genes), respectively [67]. In retinal progenitor cells, Pax6 also plays dual roles as it is required for both multipotency of these cells and suppression of the premature activation of the photoreceptor-specific differentiation program [68].
Here we show that a complex analysis of differentially expressed genes in lens compared to four related telencephalic samples provide evidence for dual activities of Pax6 both to promote and suppress mutually exclusive developmental programs, during lens and forebrain development. Nevertheless, our data show that there is still a significant fraction of genes that appear to be similarly coregulated by Pax6 in both systems. Thus, this dual function may be a general property of many other lineage-specific DNA-binding transcription factors.
Although we assume that ''stage-selector genes'' initiate distinct developmental programs through their activities as transcriptional activators and/or repressors, it is not well known which initial targets have to be activated or repressed to achieve formation of committed cell progenitors. Previous studies have suggested that Pax6 directly or indirectly regulates expression of DNA-binding transcription factors Six3, Sox2, Pitx3, Prox1, Sox1, and c-Maf [7], Sox11 [69] as well as transcriptional co-activators Eya1, Eya2 [70] and a co-repressor Dach1 [71] during early stages of lens development, i.e. lens placode and lens vesicle formation. The present study did not identify any novel gene that could be definitively linked to this critical stage of lens development as this would require more direct analysis of gene expression during the formation of lens placode.
Interestingly, up-regulation of DNA-binding transcription factor Myb was found in lens and E15 cortex/GE (Fig. 4). In neuroretinal development, Myb positively regulates the Pax6 promoter [72]. Thus, it is possible that up-regulation of Myb in Pax6 heterozygous lens can be used as a compensatory mechanism to increase Pax6 expression to correct the haploinsufficiency effect.
In addition, our data identified perturbed expression of a number of genes that participate in a large number of signal transduction pathways (see Fig. 2C and Fig. S2). Pax6 expression is regulated via TGFb/BMP [73] and FGF signaling [74]. In addition, it has been shown that expression of a reporter gene activated by multiple copies of retinoic acid-responsive elements (RAREs) was reduced in Pax6 Sey/+ embryos [13]. Loss of Raldh3 expression in the surface ectoderm of rat Pax6 homozygous embryos (rSey/rSey) suggest this enzyme as Pax6 regulated gene [12]. In the present study we also found perturbed expression of Raldh3/Aldh1a3 in Pax6 heterozygous lens (see Fig. S3); however in the opposite direction. Retinoic acid (RA) signaling is required for lens placode invagination and separation of the lens vesicle from the surface ectoderm [7]. The present data suggest that Pax6 modulates expression of various previously unknown genes that participate in FGF, RA, TGFb/BMP, and Wnt signaling pathways, both in lens and embryonic forebrain.
Specifically, we found that Mab21l2 is upregulated in Pax6 heterozygous lens (Fig. 3), and we identified promoter-proximal regions occupied by Pax6 in lens chromatin (Fig. 6). In addition, we identified two Pax6-binding sites in Mab21l2 locus (Fig. 7). Although Mab21l2 is weakly expressed in lens, it is highly expressed in the optic vesicle/cup [28]. Expression of Mab21l2 in the optic vesicle is required for the formation of lens placode; however, its expression is not reduced in optic vesicles of Sey/Sey mouse embryos [75]. Thus, the present data suggest that Mab21l2 expression is directly suppressed by Pax6 in the lens but not regulated in the optic cup. At the molecular level, Mab21l2 may inhibit function of BMP4 signaling via direct binding to Smad1 [76]. BMP4 signaling is essential for expression of Pax6 in the lens placode [73]. Collectively, the available data suggest that in order to serve such unique roles in embryonic development, Pax6 both selectively responds to upstream extracellular cues, and modulates signal transduction pathways that function in various stages of lens development by positively (negatively) regulating expression of TGF-b2 (Mab21l2) as well as many other genes (see Fig. 5 and Fig.  S4), respectively.
Within the group of 15 genes used for additional validations, Dnase2b was examined (Fig. S3). DNase IIb is an enzyme critical for lens terminal differentiation [77,78]. Reduced expression of Dnase2b transcripts was also observed in an earlier microarray study of the lens-specific knock out of AP-2a gene [79]. Thus, it appears that Dnase2b is genetically downstream of two important regulators of lens development, Pax6 and AP-2a [7].
Our previous efforts to identify genes regulated by Pax6 in lens used much larger quantities of total RNA (,5 mg/sample) obtained from 6-week old lenses [80]. In addition, the experiments were preformed as technical replicates, hybridized using dual-color in house-produced cDNA microarrays, and analyzed using an abandoned fold-change approach. In contrast, in the current study, we used biological replicates, common Affymetrix platform and consensual statistical methods of data analysis [33]. Additional experiments using biological replicates and materials prepared from 6-week lenses would be necessary to achieve a direct comparison of differential gene expression in 1-day and 6-week old Pax6 heterozygous lenses.

Integration and comparison of multiple approaches to study Pax6-dependent networks
A summary of multiple strategies to identify top candidate genes regulated by Pax6 is shown in Fig. 8. The left arm of the diagram illustrates a procedure to identify a relatively small number of genes that deserve further attention to understand their function and regulation during lens development. The right arm of the analysis shows a parallel approach that considers both shared and opposite (suppression-activation) Pax6's functions in embryonic lens and cortical development. We also included data from a recent study in which 82 Pax6-regulated genes emerged from high-throughput in situ hybridization studies using an atlas of ,1,000 spatial gene expression patterns of the midgestation mouse embryo [18]. Of these 82, eight genes, i.e. Cdh8, Fgf14, Gabrg2, Neurod1, Neurod6, Pde1c, Tmem2 and Wnt2b, were found common between 381 differentially expressed genes identified here with opposite trends responding to the loss of Pax6 function in newborn mouse lens and embryonic cortex. At present, the endpoints of the RNA expression profiling/ChIP approach (present study) and in situ hybridization studies [18], are experimental validations of candidate Pax6-binding sites using EMSAs. This is a necessary but not a sufficient step to further proof that Pax6 is indeed a direct regulator of expression of the gene of interest in a specific cell type.
One of the major limitations of genome-wide studies of embryonic development is the limited availability of adequate biological materials. Tissue and organ development depend on formation of multiple cell progenitors of different embryonic origins and their mutual interactions in a complex threedimensional landscape mediated by short-, mid-and long-range signaling [81]. The major advantage of studying the role of Pax6 during lens development is that the ocular lens is comprised of cells of a single embryonic origin making it relatively easy to isolate pure tissue to obtain RNA and chromatin for genome-scale studies. A development of alternate technologies to process and analyze the ChIP studies, such as ChIP-seq [82] will allow identification of those genomic regions such as distal 59-and 39enhnacers occupied by Pax6 that are not included in the ChIP-onchip promoter arrays. Future studies will be aimed to probe the molecular mechanism of Pax6 using a selected group of putative target genes identified in this study.

Immunohistochemistry
Eyes were obtained from Pax6 lacZ/+ mice [6] and normal littermates at postnatal day (P) 1. The genotypes were determined by PCR analysis on DNA extracted from the tail. For PCR analysis the following primers were used: sense 59-CCGGCCGCTTGGG-TGGAG-39, antisense 59-CGG TCCGCCACACCCAGC-39. After enucleation, the eyes were fixed for 24 h in 4% paraformaldehyde, washed extensively in phosphate buffered saline (PBS), incubated in ascending (10%, 20%, 30%) concentrations of sucrose/ PBS for 8 h to overnight at 4uC and shock frozen in tissue freezing medium (DiaTec, Hallstadt, Germany). Sections of 10-12 mm were cut at 230uC, washed three times in PBS (5 min each), and blocked with 2% BSA in PBS (45 min at room temperature). The primary Pax6 antibody (Eurogentec, Seraing, Belgium) was diluted (1:50) in blocking solution and incubated at 4uC overnight. After three washes in PBS (5 min each), the secondary antibody, diluted in blocking solution, was applied for 1 h (Alexa 488, goat anti rabbit, 1:1000, Invitrogen, Karlsruhe, Germany). After three washes with PBS, counterstaining was performed with DAPI and the sections were embedded with fluorescent mounting medium (Dako, Hamburg, Germany).

DNA microarray hybridizations
Total RNA was isolated from individual wild type (NMRI background) and Pax6 heterozygous (Pax6 lacZ/+ ) 1-day old lenses (2 lenses from the same mouse per sample) using the RNeasy MiniElute Kit (Qiagen, Valencia, CA) according to manufacturer's instructions. RNA quality was determined with an Agilent 2100 Bioanalyzer and cDNAs were then generated with the Ovation TM Biotin RNA amplification and Labeling System (Nugen, San Carlos, CA) using 50 ng of RNA according to the manufacturer's protocol. Three biological replicates were subsequently hybridized on Mouse Genome 430 2.0 Arrays (Affymetrix, Santa Clara, CA). Animal husbandry and experiments were conducted in accordance with the approved protocol of the Albert Einstein College of Medicine Animal Institute Committee and the ARVO Statement for the use of animals in Ophthalmic and Vision Research.  [17], (**) represent data from [18]. In situ hybridization data, ISH; Pavlidis Template Matching, PTM; shared responses, S; suppression-activation, SA. doi:10.1371/journal.pone.0004159.g008 Bioinformatic tools and statistical filtering of RNA microarray results Genes/mRNAs differentially regulated between wild-type and heterozygous (Pax6 +/2 ) P1 lens were identified using triplicate sets of Robust multichip average (RMA)-normalized Affymetrix CEL files [83] by Student's T-test (p,0.10) and by Pavlidis Template Matching (PTM, p,0.05) [84]. Primary data from this study were deposited in the NCBI Gene Expression Omnibus database under accession number GSE13244. The R-based extension to GeneSpring 7.2 (Agilent Technologies, Santa Clara, CA) was used to create a boxplot representation of 559 Pax6 target genes in Fig. 1A, to generate a fivenumber summary including the smallest observation, lower quartile, median, upper quartile, largest observation, and indicates outlier observations. The comparison of mRNA profiles between the lens and the telencephalon was performed using PTM (p,0.05) of the Multi-experiment Viewer of the TIGR TM4 Analysis package [84]. The GO and KEGG pathway functional annotations were performed using the FatiGO+ tool of the Babelomics suite [27]. The ChIP-on-chip data (Figs. S5,S6,S7,S8, and S9) were analyzed through Model-based Analysis for Tiling arrays, MAT [85] and Integrated Genome Browser (Affymetrix).

Quantitative RT-PCR (qRT-PCR)
Relative expression levels of ten genes encoding Cspg2, Igfbp5, Mab21l2, Nr2f2, Olfm3, Pax6, Spag5, Spock1, Spon1 and Tgfb2 in WT and Pax6 +/2 lenses were determined using qRT-PCR (see Table S3 for oligonucleotides). For data normalization, expression of three reference genes, B2M, HPRT and SDMA was examined. Total RNA was isolated using TrizolH Reagent (Invitrogen, Carlsbad, CA), according to the manufacturer's instruction and digested with DNase I (Promega, Madison, WI). cDNA was subsequently generated with oligo(dT 20 ) primers (Invitrogen) and Superscript TM III Reverse Transcriptase (Invitrogen), according to the manufacturer's protocol. The cDNA was diluted 10 times and qRT-PCR was conducted using an Applied Biosystems (ABI, Foster City, CA) 7900HT fast Real-Time PCR system with Power SYBRH Green PCR master mix (ABI). qRT-PCR was conducted with the primers shown in Table S2. The primers were designed using Primer3 and cross-checked by NCBI BLAST. Transcripts encoding B2M, SDHA, and HPRT [86] genes were used for normalization of expression levels in Pax6 heterozygous lenses. As no significant changes of expression of B2M, SDHA and HPRT were found, the final data were expressed relative to the expression level of B2M. RNAs prepared from three biological replicates were analyzed as shown in Figs. 2 and 3. Different RNA preparations were used to evaluate expression levels of 15 genes shown in Figs. S2 and S3 (see Table S3 for oligonucleotides).

Chromatin Immunoprecipitations (ChIPs)
For the ChIP-on-chip studies, the ''standard'' assays using chromatin prepared from 20 lenses [22] were scaled up 5-times to proceed with the chromatin obtained from 100 lenses (CD-1 mouse, Charles River Laboratories, Wilmington, MA). Three biological replicates were performed and analyzed. The complete analysis will be published elsewhere. For quantitative ChIPs, 40 microdissected P1 lenses were crosslinked in freshly prepared 1% formaldehyde for 15 minutes at room temperature. The crosslinking was stopped by 0.125 M glycine. The lenses were lysed and homogenized on ice followed by sonication using Bioruptor (Diagenode, Sparta, NJ) to 200-500 bp fragments. Chromatin was further cleared using Protein A and G beads (Sigma, St. Louis, MO) and immunoprecipitated with 5 mg anti-Pax6 antibody (H-295X, Santa Cruz Biotechnology, Santa Cruz, CA) in a total volume of 1 ml. After three washes, crosslinking was reversed and enriched chromatin was eluted into 250 ml H 2 O using the QIAquick Spin Gel Purification Kit (Qiagen).
The PCR primers were designed using Primer3 and their specificity was checked using BLAST. Default parameters were used (GC% no more than 60%, 18-22 bp length and Tm = 60uC), and PCR products were limited to 80-100 bp. The primers are given in Table S2. Quantitative PCR was conducted using a 3-step protocol consisting of 40 cycles of denaturation at 95uC for 30 seconds, annealing at 60uC for 30 seconds and extension at 72uC for 30 seconds with ABI 7900HT equipment in a total volume of 8 ml.
Raw data were calculated and analyzed using SDS2.1 software (ABI). C t values of series dilution of input samples (0.05%, 0.2% and 1% input) were used to generate standard curves. Immunoprecipitation data were referred to the standard curve and normalized to relative input units. Every sample was tested in triplicate per individual 384-well microplate and repeated as three independent biological experiments.
Eleven double stranded oligonucleotide probes (Table S4) were labeled and tested in EMSAs. As sites 11 and 12 were next to each other, a single oligonucleotide, site 11/12, was examined. Briefly, recombinant Pax6 GST-PD and GST-PD/HD proteins were expressed in E.coli (BL21 DE3), incubated with 0.5-1 ng of the 59end labeled oligonucleotide in the presence of 2 mg of poly[d(I2C)] (Pharmacia, Piscataway, NJ) at room temperature for 10 mins. The optimal amounts of proteins were determined using the P6CON probe as we described elsewhere [15]. Individual oligonucleotide probes were incubated with identical amounts of Pax6 GST-PD and GST-PD/HD proteins and specific protein-DNA complexes were resolved by 5% PAGE in 0.5xTBE buffer. In some reactions, 50 ng of cold oligonucleotides were used as specific competitors. Figure S1 Immunofluorescence detection of Pax6 in lens epithelium. Panels (A-C) are newborn Pax6 WT lenses, panels (D-F) are Pax6 heterozygous lenses. White arrows in (F) demonstrate cells expressing higher levels of Pax6 in the epithelium than those cells marked by the red arrows. Abbreviations: epithelium; e, fiber cells; f. Scale bar = 20 mm. and Pax6 +/2 (het, shown in grey) lenses were determined using qRT-PCR as described in Methods and in legend to Fig. 3. Found at: doi:10.1371/journal.pone.0004159.s002 (0.62 MB TIF) Figure S3 Verification of microarray results of down-regulated transcripts by qRT-PCR. Relative expression levels of Acvr1b, Dnase2b, Sultx1, Rock1, Camk1d and Gaa transcripts in wild type (WT, shown in black) and Pax6 +/2 (het, shown in grey) lenses were determined using qRT-PCR as described in Methods and in legend to Fig. 3. Found at: doi:10.1371/journal.pone.0004159.s003 (0.96 MB TIF) Figure S4 Genes regulated by Pax6 in lens and one region of embryonic telencephalon. A representative list of 43 genes regulated by Pax6 in lens. 19 of these genes shown here (from the total number of 127) were also differentially expressed in a single region of the developing telencephalon. These genes were grouped into six categories: Chromatin regulation, Mitosis, and Signaling (FGF, RA, TGFb/BMP, and Wnt). Found at: doi:10.1371/journal.pone.0004159.s009 (6.20 MB TIF) Figure S10 A list of putative Pax6-binding sites in Mab21l2 and Tgfb2 loci. A) P6CON, PHO and P6PHD ''consensus'' sequences. B) Alignment with twelve predicted Pax6-binding sites (site 1 to 12). These sites are grouped as ''active'' and ''inactive'' sites. Conserved nucleotide (upper case letters), non-conserved nucleotides (lower case letters). Total number of missmatches (n) between the examined site and the ''consensus'' sequence and orientation (ori) of the respective site in the promoter (forward, +; reverse, 2) is also given. Found at: doi:10.1371/journal.pone.0004159.s010 (0.69 MB TIF)

Supporting Information
Table S1 A ''559'' master gene list of differentially expressed transcripts in Pax6 heterozygous lens. The transcripts are organized alphabetically into two group, upregulated (yellow) and downregulated (blue), respectively. Average fold change (FC) in Pax6 +/2 lens is shown. Although a number of transcripts show relatively small fold changes, they were included as they both passed the statistical criteria, and, may represent genes whose expression is severely deregulated in a hypothetical Pax6 2/2 lens. Found at: doi:10.1371/journal.pone.0004159.s011 (0.10 MB XLS)   Table S4 A list of putative Pax6-binding sites in Mab21l2 and Tgfb2 loci. A) P6CON, PHO and P6PHD ''consensus'' sequences. B) Alignment with twelve predicted Pax6-binding sites (site 1 to 12). These sites are grouped as ''active'' and ''inactive'' sites. Conserved nucleotide (upper case letters), non-conserved nucleotides (lower case letters). Total number of missmatches (n) between the examined site and the ''consensus'' sequence and orientation (ori) of the respective site in the promoter (forward, +; reverse, 2) is also given. Found at: doi:10.1371/journal.pone.0004159.s014 (0.03 MB DOC)