Critical Role of Klf5 in Regulating Gene Expression during Post-Eyelid Opening Maturation of Mouse Corneas

Background Klf5 plays an important role in maturation and maintenance of the mouse ocular surface. Here, we quantify WT and Klf5-conditional null (Klf5CN) corneal gene expression, identify Klf5-target genes and compare them with the previously identified Klf4-target genes to understand the molecular basis for non-redundant functions of Klf4 and Klf5 in the cornea. Methodology/Principal Findings Postnatal day-11 (PN11) and PN56 WT and Klf5CN corneal transcriptomes were quantified by microarrays to compare gene expression in maturing WT corneas, identify Klf5-target genes, and compare corneal Klf4- and Klf5-target genes. Whole-mount corneal immunofluorescent staining was employed to examine CD45+ cell influx and neovascularization. Effect of Klf5 on expression of desmosomal components was studied by immunofluorescent staining and transient co-transfection assays. Expression of 714 and 753 genes was increased, and 299 and 210 genes decreased in PN11 and PN56 Klf5CN corneas, respectively, with 366 concordant increases and 72 concordant decreases. PN56 Klf5CN corneas shared 241 increases and 98 decreases with those previously described in Klf4CN corneas. Xenobiotic metabolism related pathways were enriched among genes decreased in Klf5CN corneas. Expression of angiogenesis and immune response-related genes was elevated, consistent with neovascularization and CD45+ cell influx in Klf5CN corneas. Expression of 1574 genes was increased and 1915 genes decreased in WT PN56 compared with PN11 corneas. Expression of ECM-associated genes decreased, while that of solute carrier family members increased in WT PN56 compared with PN11 corneas. Dsg1a, Dsg1b and Dsp were down-regulated in Klf5CN corneas and their corresponding promoter activities were stimulated by Klf5 in transient co-transfection assays. Conclusions/Significance Differences between PN11 and PN56 corneal Klf5-target genes reveal dynamic changes in functions of Klf5 during corneal maturation. Klf5 contributes to corneal epithelial homeostasis by regulating the expression of desmosomal components. Klf4- and Klf5-target genes are largely distinct, consistent with their non-redundant roles in the mouse cornea.

Klf4 and Klf5 exert tissue-dependent and non-redundant influences on the mouse ocular surface in spite of possessing an identical DNA-binding domain. Conditional disruption of Klf4 in the developing mouse ocular surface resulted in numerous defects including corneal epithelial fragility, stromal edema, altered stromal collagen fibril organization, endothelial vacuolation and loss of mucin producing conjunctival goblet cells [21,22,32]. Similar conditional disruption of Klf5 also resulted in multiple defects including translucent cornea, abnormal eyelids with malformed meibomian glands and a conjunctiva devoid of goblet cells [31]. Microarray comparison of WT and Klf4CN corneal and conjunctival transcriptomes identified significant differences in Klf4-target genes in these adjacent tissues, suggesting tissuedependent functions for Klf4 [21,33].
Here, we test the hypothesis that the basis for non-redundant functions of structurally related Klf4 and Klf5 lies in their distinct target genes in the mouse cornea. As most of the Klf5CN ocular surface defects appeared in post-eyelid opening stages [31], we identified the corneal Klf5-target genes before eyelid opening at PN11 and in young adults at PN56. This study design also enabled us to examine the changes in gene expression accompanying WT corneal maturation between PN11 and PN56. We report that Klf5 regulates a wide array of genes associated with a diverse spectrum of functions such as cell adhesion, barrier function, maintenance of hydration, and xenobiotic metabolism. We also show that the corneal Klf5-and Klf4-target genes are largely distinct, consistent with their non-redundant roles in the mouse cornea. Furthermore, we identified significant differences in Klf5-target genes between PN11 and PN56, revealing dynamic changes in Klf5 functions in the maturing cornea.

Microarray analysis and validation of results
We compared the WT and Klf5CN corneal transcriptomes in immature PN11 corneas just before eyelid opening and in young adult PN56 corneas to identify the changes in gene expression associated with post-eyelid opening Klf5CN corneal phenotype [31]. We also compared the PN56 Klf5-target genes with those reported previously for Klf4 [21] to determine the extent of overlap between Klf4-and Klf5-target genes. Scatter plots of the WT vs. Klf5CN comparisons at PN11 (Fig. 1A) and PN56 (Fig. 1B), and the PN11 WT vs. PN56 WT comparison (Fig. 1C) show overall distribution of the panels measured by these microarrays. A large number of genes with distinct or overlapping expression were identified between (i) corneal Klf5-target genes at PN11 and PN56 (Fig. 1D), (ii) corneal Klf4- [21] and Klf5-target genes at PN56 (Fig. 1E), and (iii) the genes modulated during WT corneal maturation compared with the Klf5-target genes at PN11 and PN56 (Fig. 1F). Microarray results were validated by QPCR comparison of selected genes whose expression was increased, decreased or relatively unaffected in PN11 or PN56 Klf5CN  (Fig. 2). There was a general conformity between the microarray and QPCR results, indicating that the microarray results accurately represent the changes in Klf5CN corneal gene expression at these two stages (Fig. 2).
Changes in Klf5-target genes during corneal maturation Table 1 gives a complete breakdown of the changes in 21,815 unique characterized genes represented on the microarray. Corneal ablation of Klf5 resulted in decreased expression of 299 and 210 genes at PN11 and PN56 respectively (with 72 concordant decreases), and increased expression of 714 and 753 genes (with 366 concordant increases; Fig. 1D; Table 1). About 41% of the genes modulated in Klf5CN corneas were also modulated during WT corneal maturation between PN11 and PN56, compared with only 15% for those unaffected by disruption of Klf5 (Table 1), providing evidence for the important role of Klf5 in regulating post-eyelid opening corneal maturation. The top 50 most affected genes in PN11 and PN56 Klf5CN corneas are listed in Tables 2, 3, 4, and 5 (See Tables S2, S3, S4, S5 for the complete list).

Changes in gene expression during WT corneal maturation
Comparison of the WT corneal transcriptomes at PN11 and PN56 revealed that the expression of 1574 genes decreased and 1915 genes increased by more than 2-fold between PN11 and PN56 (Fig. 1C). The 50 most affected genes in the WT PN56 compared with PN11 corneas are listed in Tables 6-7 (See Tables  S6 and S7 for the complete list). Transcripts encoding different collagens and other major extracellular matrix (ECM)-related proteins were significantly decreased between PN11 and PN56, suggesting that most of the ingredients for stromal ECM are produced before or around eyelid opening (Tables 7 and 8). Similarly, expression of Adam family proteinases and other MMPs that play significant roles in remodeling ECM [34] also was sharply decreased between PN11 and PN56 (Table 9), providing further evidence that most of the stromal ECM is in place by eyelid opening stage and little stromal remodeling occurs in the adult cornea [35].
Expression of several cell-junctional complexes and late markers of stratified squamous epithelial cells increased significantly between PN11 and PN56 (shown in bold in Table 6), when much of the corneal epithelial stratification occurs. In addition, expression of several members of the solute carrier family was significantly elevated between PN11 and PN56 (Table 10), reflecting the elevated need for solute transport in metabolically active adult corneas. While specific corneal functions of many of these solute carrier family members are not known, it is noteworthy that mutations in SLC4A11 and SLC16A12 are associated with congenital hereditary endothelial dystrophy (CHED) [36] and microcornea [37], respectively. Another important change that takes place between PN11 and PN56 corneas is the increased expression of several oxidative stress related genes including ceruloplasmin, an antioxidant enzyme upregulated in different neurodegenerative disorders including glaucoma [38,39], Arachidonate lipoxygenase-12 and -15, which promote epithelial wound healing and host defense [40], carbonic anhydrase-2, -12, and -13, overexpressed in human glaucoma [41][42][43], and calcium binding proteins S100A8 and A9 (Table 6), suggesting an increase in oxidative stress in the adult compared with the PN11 corneas.

Differences in PN56 corneal Klf4-and Klf5-target genes
Comparison of the PN56 corneal Klf5-target genes with those described previously for Klf4 [21] identified 260 common targets (204 increased and 56 decreased; Fig. 1E; Tables S8 and S9), with many more modulated exclusively in the Klf4CN (270 increased and 349 decreased) or Klf5CN (512 increased and 109 decreased) corneas. Most of the common increases in Klf4CN and Klf5CN corneas are associated with immune response, reflecting enhanced inflammatory conditions in those corneas. Regulation of largely distinct sets of target genes by Klf4 and Klf5 is consistent with their non-redundant functions in the mouse cornea [22,31].

Elevated immune response in Klf5CN corneas
Canonical pathway analysis of the aggregate Klf5-target genes identified 26 significantly (p,0.001) enriched pathways, predominantly associated with immune function (Table S10). Expression of most of the genes associated with these pathways was increased, increasing the likelihood that these pathways represent indirect response to disruption of Klf5. To overcome this limitation, we repeated pathway analyses selectively for the genes with decreased expression in Klf5CN corneas. Several xenobiotic stress responserelated pathways were predominantly enriched in genes with decreased expression upon disruption of Klf5, suggesting that Klf5 plays a role in xenobiotic stress response in the cornea (Table 11).
Expression of 107 of 368 genes (29.1%) containing ''immun'' or ''inflamm'' in the GO Biological Process notation was increased, while expression of only seven of these genes decreased in PN11 or PN56 Klf5CN corneas (Table S11). Increased expression of the immune response related genes, together with the previously reported hypercellularity of the Klf5CN corneal stroma [31] suggested a robust increase in immune-response in the Klf5CN corneas. Immunofluorescent staining of corneal flat mounts with anti-CD45 antibody demonstrated increased influx of CD45+ cells distributed throughout the Klf5CN compared with the WT corneal stroma sparsely populated with CD45+ cells (Fig. 3).

Klf5CN corneal neovascularization (CNV)
Whole-mount corneal immunofluorescent staining with anti-CD31 and anti-Lyve1 antibody revealed that the enhanced inflammatory environment in Klf5CN corneas is accompanied by extensive CNV (Fig. 4). Klf5CN CNV was apparent as early as PN21, when the Lyve1+ lymph vessels were much more pronounced and penetrated deeper into the central cornea, unlike the CD31+ blood vessels that remained in the peripheral region without reaching the central cornea (Fig. 4). By PN56, CNV was observed throughout the Klf5CN cornea, with blood vessels overtaking lymph vessels, which appeared to have regressed (Fig. 4). Examination of the XY-stack of confocal images revealed that CD31+ blood vessels are present in the anterior of the Klf5CN corneal stroma, unlike the Lyve1+ lymph vessels located in the posterior (Fig. 4).     Desmosomes are essential for corneal epithelial homeostasis [44][45][46]. Previously, we reported that Klf4 contributes to the formation and maintenance of corneal epithelial permeability barrier by regulating the expression of desmosomal components [47]. Microarray data presented here revealed that desmosomal components Dsg1a and Dsg1b are decreased in Klf5CN corneas (Table 4, Tables S2 and S4). Consistent with the microarray data, immunofluorescence revealed a sharp decrease in the epithelial expression of desmogleins, and a moderate decrease in desmoplakin in the Klf5CN corneas (Fig. 5A). Next, we tested the effect of Klf4 and/or Klf5 on Dsg1a, Dsg1b and Dsp promoter activities by transient co-transfection assays in NCTC cells using the previously described reporter vectors [47]. Dsg1a, Dsg1b, and Dsp promoter activities were stimulated by 7.5-, 6.5-and 8.7-fold, 5.8-, 9.9-and 10.8-fold, and 9.6-, 3.5-and 9.6-fold, respectively, when cotransfected with Klf4, Klf5, or both (Fig. 5B). Relative to Klf4, Klf5 had a comparable effect on Dsg1a, stronger stimulatory effect on Dsg1b and weaker stimulatory effect on Dsp promoter activities (Fig. 5B). Co-transfection with both Klf4 and Klf5 did not result in an additive or synergistic stimulation of these promoter activities (Fig. 5B), suggesting that Klf4 and Klf5 function through the same ciselements within the Dsg1a, Dsg1b, and Dsp promoters [47]. Taken together with our previous report [47], these results demonstrate that one of the ways by which Klf4 and Klf5 contribute to corneal epithelial homeostasis is by regulating the expression of desmosomal components Dsg1a, Dsg1b and Dsp.

Influence of Klf4 and Klf5 on gene regulatory networks in the cornea
In order to determine the influence of Klf4 and Klf5 on gene regulatory networks in the cornea, we examined the expression of other transcription factors in PN11 and PN56 WT and Klf5CN corneas and compared them with the previous results from PN56 Klf4CN corneas [21]. Comparative analysis of the transcription factors decreased in more than one dataset (i.e., (a) PN56 Klf4CN vs. WT, (b) PN56 Klf5CN vs. WT, (c) PN11 Klf5CN vs. WT, and (d) PN56 WT vs. PN11 WT) identified Pax6, Bnc1, Cux1, Tox and Satb1 as common targets of Klf4 and Klf5 that were also modulated during corneal maturation (Table 12). Pathway analysis of the affected transcription factors revealed distinct networks predominantly associated with development and tissue homeostasis ( Figures S2 and S3). The differences in these associated networks in spite of the five common transcription factor targets for Klf4 and Klf5 are consistent with their nonredundant functions in the mouse cornea. Among the common transcription factor targets of Klf4 and Klf5, while Pax6 and Bnc1 are known to regulate corneal epithelial homeostasis [5,14,[48][49][50][51][52], corneal functions of Cux1, Tox and Satb1 are not yet known. Increased expression of Cux1, which suppresses collagen synthesis [53], is consistent with the decreased expression of collagens during WT corneal maturation. The transcription factors whose   Ccl9 chemokine (C-C motif) ligand 9 5.93 9.06 8.74 Genes whose expression is also increased in the PN11 Klf5CN corneas are shown in bold. doi:10.1371/journal.pone.0044771.t005 Table 6. Top 50 genes whose expression is most increased during post-eyelid opening WT corneal maturation between PN11 and PN56.    Table 12) are known to be upregulated during inflammation [54][55][56][57], raising a possibility that their increased expression reflects pro-inflammatory conditions in these corneas and may not be due to direct derepression in the absence of Klf4 or Klf5. In order to identify if a regulatory relationship exists between different Klfs, we examined the effect of disruption of Klf5 on expression of other Klfs in the cornea. While transcripts for Klfs 1, 14 and 15 were essentially absent in the cornea, those for Klfs 2, 4, 6, 7, 10, 11, 12, 16 and 17 were present in all samples but showed no robust changes. During WT corneal maturation, the expression of Klfs 3, 5 and 6 was increased, while that of Klfs 2, 12 and 13 was decreased in PN56 compared with the PN11 corneas. Klf4 and Klf5 remained unaffected in Klf5CN and Klf4CN corneas, respectively, suggesting a lack of regulatory relationship between these two Klfs highly expressed in the cornea. The closely related Klf9 and Klf13 [27] were both increased in both Klf4CN and Klf5CN corneas, suggesting that Klf9 and Klf13 show similar compensatory changes whether Klf5 or Klf4 is ablated. Whether this represents a true regulatory relationship among Klfs 4, 5, 9 and 13 in the mouse cornea remains to be established.

Discussion
Previously, we demonstrated that disruption of Klf5 resulted in defective maturation of the mouse cornea in post-eyelid opening stages [31]. Here, we have employed microarray analysis to obtain a comprehensive view of the changes in corneal gene expression upon deletion of Klf5 in immature corneas around eyelid opening (PN11) and in adult (PN56) corneas. Our findings reveal the molecular basis of the wide ranging influence of Klf5 on corneal homeostasis and identify candidate target genes of Klf5 in the mouse cornea. In addition, the design of our study allowed us to identify the changes in gene expression between PN11 and PN56 WT corneas.
Earlier reports of Klf5-target gene profiling used chromatin immunoprecipitation followed by microarray (ChIP-Array) in embryonic stem cells [58], or microarray analysis following selective disruption of Klf5 in the mouse lung [59] or bladder urothelium [60]. Our results are largely consistent with those findings and identify additional Klf5-target genes such as Pax6 and Dsg1a, which play critical roles in the cornea. Differences in Klf5target genes in the cornea (this study), ESCs [58], lung [59] and bladder urothelium [60] provide evidence for tissue-dependent A striking feature of our results is the large number of genes whose expression is influenced by the absence of Klf5 in the mouse cornea. By comparison with similar studies of transcription factors such as FoxP2 [61], Sox2 [62], Myb [63] and Bcl11b [64], we predict that Klf5 is likely to directly regulate only a fraction of those genes whose expression is modulated in the Klf5CN corneas. The remaining genes are expected to be indirect targets of Klf5, through other transcription factors such as Pax6 [48][49][50]65], whose expression is reduced in Klf5CN corneas (Table 12). Alternatively, they may represent physiological responses to the phenotype brought about by disruption of Klf5. For example, a large number of genes upregulated in Klf5CN corneas are immune response related, consistent with the massive infiltration of CD45+ cells (Fig. 3), and may not be directly regulated by Klf5. Though it is likely that the Klf5CN corneal neovascularization and influx of CD45+ cells indirectly contributed to increased expression of many of the immune response related genes, a similar effect may not be extended to a majority of the negatively regulated genes, which are more likely to represent true targets for Klf5.  Klf4 and Klf5 are both abundantly expressed in the mouse cornea [29], where they play critical, non-redundant roles [22,31]. In order to understand the molecular basis for their corneal functions, it is necessary to identify and distinguish their target genes. Comparison of the Klf5CN corneal gene expression profile (this study) with that of Klf4CN corneas [21] revealed that roughly 2/3 of the corneal Klf4-and Klf5-target genes are Klf4-or Klf5specific, with the rest being common targets. Canonical pathway analysis of genes exclusively modulated by Klf4 yielded ''human embryonic stem cell pluripotency'' as the most significantly enriched (p,10 25 ) pathway, in agreement with the importance of KLF4 in inducing pluripotency [66]. In contrast, ''hepatic fibrosis/hepatic stellate cell activation'' was the most significantly enriched (p,10 212 ) pathway among the genes exclusively modulated by Klf5 (Table S10), indicative of a general fibrotic response such as that observed in cultured human keratocytes exposed to TGF-b [67]. Molecular basis for the ability of Klf4 and Klf5 to regulate distinct sets of target genes in spite of possessing identical DNA-binding domain remains to be understood. Genome-wide identification of the nucleotide sequence of Klf4and Klf5-bound cis-elements by chromatin immunoprecipitation followed by large scale sequencing (ChIP-Seq) is necessary to better understand target site selection by Klf4 and Klf5. Being environmentally exposed, the cornea is frequently exposed to xenobiotic stress. Also, when the eyelids are closed during sleep, the avascular cornea is subjected to almost 75% drop in oxygen partial pressure [68,69]. Thus, hypoxic and xenobiotic response pathways are expected to play an active role in corneal homeostasis. The important role of inhibitory PAS domain protein (IPAS) -a hypoxia repressor protein-in maintaining corneal avascularity [70,71] further supports this contention. Moreover, corneal crystallin genes are induced by hypoxia or xenobiotics, further implicating hypoxic and xenobiotic stress in corneal gene expression [16,72,73]. Our data demonstrated significant enrichment of xenobiotic metabolism-related pathways among genes whose expression is decreased in Klf5CN corneas (Table 11). Thus, we suggest that Klf5 serves an important role in detoxification of the environmentally exposed avascular cornea, by supporting the expression of xenobiotic metabolism related genes.
Important changes take place in the mouse cornea as it matures following eyelid opening around PN12 [8,74]. This study identified the changes in corneal gene expression between PN11 and PN56, revealing the molecular events underlying post-eyelid opening corneal maturation. Specifically, we demonstrate that during WT corneal maturation between PN11 and PN56, transcripts encoding (i) ECM components are sharply decreased, (ii) epithelial barrier-related proteins are sharply increased, and (iii) members of the solute carrier family proteins are elevated, consistent with (a) the active formation of stromal ECM around eyelid opening, with little remodeling taking place in adult corneal stroma, (b) rapid stratification of squamous epithelium in posteyelid opening stages and (c) elevated demand for solute transport in the metabolically active adult cornea, respectively.
We compared the present data with a previous analysis of changes in corneal gene expression associated with post-eyelid opening maturation [75], which used Affymetrix MG74Av2 chips targeting a subset (8,666) of the 21,815 unique characterized genes examined here. Applying the present selection rules to their data [75] yielded 442 genes differentially expressed between their immature (PN10) and adult (PN49 to PN56) groups. Though there were differences between the two datasets (which could be attributed to several factors including differences in the mouse Table 9. Expression of metalloproteinases and other genes associated with matrix remodeling between PN11 and PN56. strains used, the nature of microarray chips used, and the parameters employed in filtering and analyzing the microarray data), our study confirmed 202 (45.7%) of these genes as differentially expressed between PN11 and PN56 corneas, with 36 concordant increases and 149 concordant decreases (7.0 and 5.1 times the number expected by contingency analysis, respectively, with a x 2 test yielding p = 6.8610 2161 ). Thus, the current study confirmed and expanded our knowledge of the changes in gene expression associated with post-eyelid opening corneal maturation [29,75]. In summary, this report identifies dynamic changes in gene expression accompanying post-eyelid opening corneal maturation, and the role of Klf5 in this process. Our results show that Klf5 contributes to maturation and maintenance of cornea by regulating the expression of subsets of genes involved in specific functions such as cell-cell adhesion, epithelial barrier formation, maintenance of proper level of hydration and xenobiotic metabolism. The changes in Klf5CN corneal gene expression are consistent with the elevated immune response and CNV. These results also revealed significant differences between Klf4-and Klf5-target genes, consistent with their non-redundant roles in the mouse cornea. Taken together with our previous report [22], the present studies establish Klf5 as another important node in the

Ethics Statement
Mice used in these studies were maintained in accordance with the guidelines set forth by the Institutional Animal Care and Use Committee (IACUC) of the University of Pittsburgh, and the ARVO statement on the use of animals in ophthalmic and vision research. All procedures performed on mice reported in this study were approved by the University of Pittsburgh IACUC.

Conditional disruption of Klf5
Klf5CN mice were generated on a mixed background by mating Klf5 loxP/loxP , Le-Cre/mice with Klf5 loxP/loxP mice as described before [31]. This mating scheme generated equal numbers of Klf5 loxP/loxP , Le-Cre/-(Klf5CN) and Klf5 loxP/loxP (control) offspring. Genomic DNA isolated from tail clippings was assayed for the presence of the Klf5-LoxP and Le-Cre transgenes by PCR using specific primers.
Isolation of total RNA, quality control, labeling and microarray analysis WT and Klf5CN littermates (4 each at PN56 and 3 each at PN11) were used for comparison of corneal gene expression by microarrays. All corneas used in these studies were microdissected from freshly harvested eyeballs under a surgical microscope. Corneas were dissected using a pair of fine scissors (RS-5611 Vannas Curved Spring Scissors; Roboz Surgical Company, Germany) around the limbus, ensuring that they are free of contamination from iris, ciliary body and/or trabecular meshwork. Two dissected corneas from each mouse were pooled for isolation of total RNA using the RNeasy Mini kit (Qiagen, Germantown, MD). The quality and integrity of the isolated total RNA was confirmed using an Agilent Bioanalyzer ( Figure S1); 1.0 mg sample RNAs were subsequently amplified and labeled using a 39 IVT Express Kit (Affymetrix Inc., Santa Clara, CA) and hybridized to Affymetrix MG 430 2 chips [33]. Utilization of the same amount of total RNA (1.0 mg) from WT and Klf5CN corneas for labeling and hybridization ensured that the smaller size of Klf5CN corneas did not skew the microarray results.
The raw data obtained from microarray analysis were processed using Affymetrix GeneChip Operating Software (GCOS v 1.4) using software defaults, to assess the presence or absence of each transcript target sequence, its expression level, and then to make all relevant pair-wise statistical comparisons among samples. Expression levels were scaled to a target value of 150 using the software default (2% trimmed mean). Prior to scaling, mean microarray expression levels were 449638 (mean 6 SD, n = 13 with one outlier of 255), and target sequences were detected (called Present) for 57.8%62.5% (mean 6 SD, n = 14) of panels on the chip. After scaling, three redundant panels for the housekeeping gene GAPDH reported coefficients of variation of 12.3%, 10.9% and 13.1% respectively (n = 13 for each panel since each had one outlier value).

Validation of microarray results by real time QPCR
Applied Biosystems (ABI: Foster City, CA) was the source of the reagents, equipment and software for quantitative real time RT-PCR assays (QPCR). Total RNA isolated from pooled corneas of 10 WT or Klf5CN mice was quantified and the concentration adjusted with RNase-free water to 100 ng/ml. cDNA was generated using an ABI High Capacity cDNA Archive Kit and real-time QPCR assays were performed in a ABI StepOnePlus thermocycler using GAPDH as endogenous control for SYBR green assays and Pcx (pyruvate carboxylase) as endogenous control for FAM assays; the results were analyzed using ABI StepOnePlus software. Nucleotide sequence of different primers used is given in Table S1.

Immunofluorescent staining of corneal whole mounts
Corneas were dissected, flattened by three radial incisions, washed 3 times for 15 minutes each in PBS with 4% FBS and blocked in FC block (BD Pharmingen, San Jose, CA) for 20 minutes prior to incubation with FITC-conjugated anti-CD45 antibody, or FITC-conjugated anti-Lyve1 antibody and PE-conjugated anti-CD31 antibody (BD Pharmingen) overnight at 4uC. Corneas were then washed three times each for 30 minutes in PBS/4% FBS, fixed in 1% Paraformaldehyde for 2 hours at 4uC, rinsed 3 times again for 30 minutes each in PBS/4% FBS

Immunofluorescence
Paraformaldehyde fixed, paraffin-embedded sections were deparaffinized with xylene, blocked with 10% goat serum in PBS for 1 h at room temperature (RT) in a humidified chamber, washed twice with PBS for 5 minutes each, incubated overnight with a 1:50 dilution of anti-Dsg (recognizes both Dsg1a and Dsg1b) or anti-Dsp primary antibody raised in rabbit (Santa Cruz Biotechnology, Santa Cruz, CA) at 4uC, washed thrice with PBS for 5 minutes each, incubated with secondary antibody (Alexa-Fluor 546-coupled goat anti-rabbit IgG antibody; Molecular Probes, Carlsbad, CA) at a 1:300 dilution for 1 h at RT, washed twice with PBS for 5 minutes each, incubated with DAPI for 10 minutes, mounted with Aqua Polymount (Polysciences, Inc), and observed with a fluorescence microscope. All images presented within each composite figure were acquired under identical settings and processed in a similar manner using Adobe Photoshop and Illustrator (Adobe, Mountain View, CA).

Reporter Vectors, Cell Culture, and Transient Transfection Assays
Different reporter vectors where approximately 2 kb Dsg1a, Dsg1b and Dsp promoter fragments drive the expression of luciferase reporter gene were described earlier [47]. Full-length Klf4 in pCI-Klf4 and full length Klf5 in pCMV-Sport-Klf5 was transiently expressed using the CMV promoter. Human skin keratinocyte NCTC cells were grown in Dulbecco's modified Eagle's medium(DMEM) supplemented with 10% fetal bovine serum, penicillin, and streptomycin at 37uC in a humidified chamber containing 5% CO 2 in air. NCTC cells in 12-well plates were co-transfected with 0.5 mg pDsg1a-Luc, pDsg1b-Luc or pDsp-Luc, along with 20 ng pRL-SV40 (for normalization of transfection efficiency) and 0.5 mg pCI (control), or 0.25 mg each of pCI and pCI-Klf4, or pCI and pCMV-Sport6-Klf5, or pCI-Klf4 and pCMV-Sport6-Klf5 using 2.5 mL transfection reagent (Expressfect; Denville Scientific). After 2 days, cells were washed with PBS and lysed with 200 mL passive lysis buffer (Promega). 50 mL of lysate was analyzed using a dual-luciferase assay kit (Promega) and a microplate luminometer (Synergy-II; Biotek Instruments, Winooski, VT). The measurement was integrated over 10 seconds, with a delay of 2 seconds. Firefly luciferase activity normalized for transfection efficiency using the SV40 promoter-driven Renilla luciferase activity, were used to obtain mean promoter activities. Fold-activation was determined by dividing mean promoter activity by the promoter activity with only pCI. Error bars indicate Standard Error of Mean (SEM). Results presented are representative of three independent experiments.         (*) = developmental effect, (**) = Klf5CN effect in PN11 cornea also, (***) = both. All changes in the PN11 Klf5CN are increases and all developmental changes are decreases, except those shaded grey. (XLSX)