Serum response factor: positive and negative regulation of an epithelial gene expression network in the destrin mutant cornea

Abstract

Increased angiogenesis, inflammation, and proliferation are hallmarks of diseased tissues, and in vivo models of these disease phenotypes can provide insight into disease pathology. Dstn^corn1 mice, deficient for the actin depolymerizing factor destrin (DSTN), display an increase of serum response factor (SRF) that results in epithelial hyperproliferation, inflammation, and neovascularization in the cornea. Previous work demonstrated that conditional ablation of Srf from the corneal epithelium of Dstn^corn1 mice returns the cornea to a wild-type (WT) like state. This result implicated SRF as a major regulator of genes that contributes to abnormal phenotypes in Dstn^corn1 cornea. The purpose of this study is to identify gene networks that are affected by increased expression of Srf in the Dstn^corn1 cornea. Microarray analysis led to characterization of gene expression changes that occur when conditional knockout of Srf rescues mutant phenotypes in the cornea of Dstn^corn1 mice. Comparison of gene expression values from WT, Dstn^corn1 mutant, and Dstn^corn1 rescued cornea identified >400 differentially expressed genes that are downstream from SRF. Srf ablation had a significant effect on genes associated with epithelial cell-cell junctions and regulation of actin dynamics. The majority of genes affected by SRF are downregulated in the Dstn^corn1 mutant cornea, suggesting that increased SRF negatively affects transcription of SRF gene targets. ChIP-seq analysis on Dstn^corn1 mutant and WT tissue revealed that, despite being present in higher abundance, SRF binding is significantly decreased in the Dstn^corn1 mutant cornea. This study uses a unique model combining genetic and genomic approaches to identify genes that are regulated by SRF. These findings expand current understanding of the role of SRF in both normal and abnormal tissue homeostasis.

neovascularization, inflammation, and cell hyperproliferation are phenotypes that distinguish diseased tissues from a normal, wild-type (WT) state. Identification of genes and molecules responsible for these physiological processes contribute to a better understanding of the mechanisms underlying tumorigenesis and other pathologically abnormal tissues.

The corneal epithelium, a transparent and tightly regulated tissue located at the front of the eye, serves as a unique model to investigate epithelial tissue abnormalities in vivo. In the cornea of mice lacking the gene for the protein destrin (DSTN), termed Dstn^corn1, DSTN deficiency leads to accumulation of F-actin, epithelial cell hyperproliferation, inflammation, and neovascularization (11, 12, 31, 34, 37). Analysis of gene expression differences between Dstn^corn1 and WT cornea confirmed transcript-level changes in actin dynamics and the cytoskeleton, and also hinted at a role for serum response factor (SRF) in the development of abnormal phenotypes (34). Subsequent work led to the discovery that conditional knockout of Srf in the corneal epithelium rescues the neovascularization, inflammation, and epithelial hyperproliferation caused by the Dstn^corn1 mutation, returning the tissue to a state that closely resembles that of a WT cornea (36).

SRF is a member of the MADS (MCM1 in Saccharomyces cerevisiae, AGAMOUS in Arabidopsis thaliana, DEFICIENS in Antihirrhinum majus, and SRF in Homo sapiens) box family of transcription factors and is expressed ubiquitously in many organisms (30). SRF derives its name from the ability to bind to a serum response element (SRE) and influence gene expression. SRF is required for proper activation of many immediate early growth-response genes; mice that lack Srf die early in embryonic development (24). Tissue-specific promoters have allowed for the manipulation of SRF expression in the corneal epithelium as well as the epidermis, heart, and muscle tissues, to name a few (13, 15, 17, 35, 44a). These studies highlight the essential role of SRF in the control of gene expression within and among tissues throughout the body of the organism. Improper or misregulated expression of SRF has been implicated in numerous cancers (25, 38, 47), and SRF activity has been found to be modulated by actin dynamics by several in vitro studies (1, 20, 21). This study uses a unique model to identify how F-actin-driven increased SRF transcriptional activity leads to increased proliferation, recruitment of inflammatory molecules, and angiogenesis in vivo.

The purpose of this study is to identify genes and proteins that are part of the transcriptional network controlled by SRF in the Dstn^corn1 cornea. Building upon our knowledge of gene expression differences in Dstn^corn1 cornea compared with WT, we can use our system of Srf ablation in the corneal epithelium to determine which genes act downstream from SRF to influence Dstn^corn1 phenotypes. We also continue our analysis of differences between Dstn^corn1 mutant and WT tissue by determining where SRF is bound. Our system allows us to identify both known and novel in vivo targets of SRF regulation in the corneal epithelium. By using Dstn^corn1 mice to explore the intersection of genomic analysis and genetic manipulation, we utilize this unique model as a powerful tool to identify genes and gene interactions that are functionally important in epithelial tissues.

MATERIALS AND METHODS

Mouse husbandry.

A.BY-H2^b H2-T18^b/SnJ-Dstn^corn1/J (A.BY Dstn^corn1), B6.129-Srf^tm1Rmn/J (Srf^f/f), and FVB.Cg-Tg (tetO-cre) 1Jaw/J (TetOcre⁺) mice were obtained from the Jackson Laboratory (Bar Harbor, ME) and bred and maintained in an animal facility at the University of Wisconsin-Madison. Krt12^rtTA/rtTA mice were generated as described previously (4). Dstn^corn1 mutant (Dstn^corn1/corn1; Srf^f/f; Krt12^rtTA/wt; TetOCre⁻), Dstn^corn1 rescued (Dstn^corn1/corn1; Srf^f/f; Krt12^rtTA/wt; TetOCre⁺), and WT (Dstn^corn1/wt; Srf^f/f; Krt12^rtTA/wt) mice were generated and treated as described previously (36). All mouse procedures were performed in accordance with protocols approved by the Animal Care and Use Committee at the University of Wisconsin, Madison, and conform to the Association for Research in Vision and Ophthalmology statement for the use of animals in ophthalmic and vision research.

Doxycycline administration.

Doxycycline (Dox) was administered either through diet (1.0 g/1 kg Dox, BioServ CN S3949) or via intraperitoneal injection, as described previously (36). All mice were killed at postnatal day (P)58.

RNA isolation from cornea.

RNA isolation was performed as described previously (34). For each biological replicate, 10 corneas (5 mice) were pooled for microarray analysis and also used for quantitative real-time PCR (qPCR).

Microarray analysis.

Microarray experiments were designed to comply with minimum information about microarray experiment (MIAME) guidelines (3). Microarray analysis for each genotype was performed with three derived RNA samples (10 cornea/5 mice) for each sample, each being hybridized to one Affymetrix MG 2.0 ST array (Affymetrix, Santa Clara, CA) according to the manufacturer's instructions. In brief, cRNA was synthesized using the Ambion WT Expression kit (Ambion). Single-stranded (ss) cDNA was synthesized and labeled using the Affymetrix WT Terminal Labeling kit. We hybridized 1 μl (∼114 ng) of purified, fragmented, labeled ss cDNA to an array at 45°C for 16 h in an Affymetrix 640 hybridization oven. The posthybridization process was performed in an Affymetrix 450 fluidic station according to the manufacturer's instructions. All gene chips were scanned on an Affymetrix GC3000 G7 scanner, and data were extracted from scanned images using Affymetrix Command Console. All technical microarray and qPCR procedures were carried out at the Gene Expression Center in the University of Wisconsin-Madison Biotechnology Center.

Microarray data analysis.

Microarray gene expression analysis was performed using Partek Genomic Suite software, version 6.6 Copyright 2013 (Partek, St. Louis, MO). The data were imported in .CEL file formatted and normalized using the robust multi-array average method. Using principle components analysis, we were able to identify that the major effect influencing expression value was sample group. One-way ANOVA was performed, and gene lists were created using a P value with a false discovery rate < 0.01. Functional annotation utilized the Database for Annotation, Visualization, and Integrated Discovery Functional Annotation Tool (http://david.abcc.ncifcrf.gov/) with Mus musculus as the background. To find enriched Gene Ontology (GO) terms, the default settings were used except that the P value cutoff was set to 0.01. The complete microarray data set generated for this study and discussed in this article has been uploaded to the National Center for Biotechnology Information Gene Expression Omnibus (NCBI GEO; http://www.ncbi.nlm.nih.gov/geo/) and can be accessed with GEO series accession number GSE49688.

qPCR analysis.

We treated the isolated RNA with Turbo DNA-free (Ambion). We converted 100 ng of total RNA to cDNA per 20 μl reaction using Superscript III first strand synthesis system (Invitrogen) according to the manufacturer's instructions. Amplification was performed using a Roche 480 Lightcycler system. Each cDNA reaction was performed in triplicate. The primer sequences are listed in Table 1. Thermal cycle conditions were as follows: activation step for 10 min at 95°C, 45 cycles of denaturation step for 10 s at 95°C, an annealing step for 10 s at 60°C, and an extension step for 5 s at 72°C. Fluorescence was measured at the beginning of each extension step for each cycle. To check the specificity of each primer pair, the predicted PCR amplicon melting temperature was confirmed by a dissociation curve analysis. PCR products were denatured at 95°C for 5 s and gradually reannealed, bringing the temperature to 65°C and measuring fluorescence with 8 acquisitions per 1°C. Expected amplicon size was confirmed by electrophoresis on an agarose gel (2%) with the same product using traditional PCR (Platinum Taq, Invitrogen). All qPCR reactions were performed by the University of Wisconsin Gene Expression Center.

Table 1.

Primers used for qPCR reactions listed in this study

Gene Name	Accession ID	Forward Primer	Reverse Primer	cDNA Product Size
Dstn	NM_019771	GAGTTCAGGTTGCGGA-GAAG	TGCACTGAGACAGAAAATGACAG	116
Srf	NM_134156	CTGCCTCAACTCGCCAGAC	TCAGATTCCGACACCTGGTAG	111
Cold1	NM_145575	CGTCCGCAATATCAAGAGCA	GTTCCGCTTGCCAGATACAT	208
Cobl	NM_172496	AAGCCAAATACGTTGATTGGGT	GGTTAGGCCAGGTTTAACTCTTT	90
Fmnl1	NM_019679	CTGCTGAGCCAGTATGACAATG	CGGTATCCAGGTAGCTCTTCA	118
Spnb3	NM_021287	TTCAACTCCTATCGCACTGTGG	TCTCGCTCATGTTCGGCTTTC	191
Krt12	NM_010661	CCCAGCTTGAGACCCTCAC	GCCTGGAAACTT-GGAGTTCT	76
Fmr1	NM_008031	TTGCGACAAATTGGAGCTAGT	CCCATTCCTTGACCATCATC	89
Mier	NM_001039081	GCCTTGAGAAGACTGAGATTTAATG	CTTCAGCCCTTGCTCAAAAT	96
Cnot8	ENSMUST000C0108843	CTGCGGCACTTGTAGAAAACA	TGAGGACAATCTCACGGATCTTC	93

qPCR data analysis.

Qbase Plus software (version 2.5, Dev build 201303141230 http://www.biogazelle.com/qbaseplus) was used to calculate the amplification efficiency for each primer set and generate the relative expression value. The amplification efficiency for each primer set is listed in Table 2. Relative expression values for the genes of interest were normalized to the mean of the relative expression values of two reference genes (Fmr1 and Mier1). One-way ANOVA was performed to compare the mutant, WT, and rescued expression values. Three reference genes were selected based on our microarray data (Fmr1, Mier1, Cnot8). After analysis of the relative expression values of these genes by geNorm, the gene with the highest M value (a stability measure) was eliminated. This was repeated until elimination of the gene with the highest M value had no significant effect on the calculated normalization factor. GeNorm determined that Fmr1 and Mier1 were the best normalization genes.

Table 2.

Mean efficiency values for the qPCR primers used in this study

Gene Symbol	Mean Efficiency, %	SE
Dstn	0.831	0.023
Srf	0.874	0.035
Cald1	0.686	0.051
Cobl	0.786	0.02
Fmnl1	0.798	0.037
Spnb3	0.812	0.041
Krt12	0.892	0.019
Fmr1	0.804	0.051
Mier	0.932	0.057
Cnot8	0.863	0.069

Chromatin immunoprecipitation.

Mouse cornea from 25 Dstn^corn1 mutant mice and 28 WT mice were dissected, pooled, and flash-frozen in liquid nitrogen. Further processing and chromatin immunoprecipitation sequencing (ChIP-seq) analysis were performed at Active Motif (Carlsbad, CA). For each genotype, cornea were fixed in PBS and 1% formaldehyde, cut into smaller pieces with a razor blade, and incubated at room temperature for 15 min. Fixation was halted by addition of 0.125 M glycine. The tissue was then treated with TissueTearer, spun down, and washed twice in PBS. Chromatin was isolated from each sample by the addition of lysis buffer and disruption with a Dounce homogenizer. Lysates were sonicated with a microtip to shear the DNA to an average length of 300–500 bp. Lysates were cleared by centrifugation and stored at −80°C. Genomic DNA (input) was prepared by treatment of aliquots with RNase, proteinase K, and heat for decross-linking, followed by phenol-chloroform extraction and ethanol precipitation. Purified DNA was quantified on a NanoDrop spectrophotometer.

For each chromatin immunoprecipitation (ChIP) reaction, 30 μg of chromatin was precleared with protein A agarose beads (Invitrogen). ChIP reactions were prepared using precleared chromatin and an SRF antibody (sc-335, lot #I2712, Santa Cruz Biotechnology) and incubated overnight at 4°C. Protein A agarose beads were added and the incubation continued for another 3 h. Immune complexes were washed, eluted from the beads via SDS buffer, and treated with RNase and proteinase K. Cross-links were reversed by incubation overnight at 65°C, and ChIP DNA was purified by phenol-chloroform extraction and ethanol purification.

The quality of the ChIP enrichment was assayed by qPCR using primers against candidate SRF sites located near genes Cyr61, Actc1, and Cald1. The resulting signals were normalized for primer efficiency via qPCR for each primer pair using input DNA.

ChIP-sequencing (Illumina).

ChIP and input DNAs were prepared for amplification by converting overhangs into phosphorylated blunt ends and adding an adenine to the 3′-ends. Illumina genomic adapters were ligated, and the sample was size-fractionated (200–300 bp) on an agarose gel. After a final PCR amplification step (18 cycles), the resulting DNA libraries were quantified and sequenced on HiSeq 2000.

ChIP-seq analysis.

Sequences (50 nt reads, single end) were aligned to the mouse genome (mm10) using the BWA algorithm (14). Aligns were extended in silico at their 3′-ends to a length of 150 bp, which is the average genomic fragment length in the size-selected library, and assigned to 32 nt bins along the genome. The resulting histograms (genomic “signal maps”) were stored in BAR and bigWig files. Peak locations were determined using the MACS algorithm (v1.4.2) with a cutoff of P value = 1e-10 (46). Signal and peak locations were used as input data to Active Motif's proprietary analysis programs, which creates excel tables containing detailed information on sample comparison, peak metrics, peak location, and gene annotations. The “gene margin” was defined as being located within the interval containing 10 Kb upstream from and downstream from a gene as well as the genomic sequence. The complete ChIP-seq data set generated for this study and discussed in this manuscript has been uploaded to the NCBI GEO (http://www.ncbi.nlm.nih.gov/geo/) and can be accessed with GEO series accession number GSE54209.

To determine if there were differences in the binding motif of peaks that were affected by the Dstn^corn1 mutation, we generated motif logos using Multiple EM (Estimation Maximization) for Motif Elicitation (MEME, http://www.meme.nbcr.net/meme).

Immunohistochemistry.

For immunohistochemistry (IHC) on frozen sections and whole cornea, tissues were prepared and IHC was carried out as described previously (36, 37). All mice for this study were sampled at P58. The primary and secondary antibodies along with the dilutions used are listed in Table 3.

Table 3.

Antibodies, their sources, and dilutions used for this study

			Dilution
Primary Antibody	Source	Catalog #	WM	FS	Secondary Antibody	Source	Catalog #
ZO-1	Invitrogen	61-7300	1:100		anti-rabbit Cy3	Jackson Immuno Research	711-165-152
Caldesmon	Santa Cruz	SC-271222		1:100	anti-mouse AF488	Molecular Probes	A21206
CD146/MCAM	abcam	ab75769		1:100	anti-rabbit Cy3	Jackson Immuno Research	711-165-152
DAPI	Sigma	D9542	1:1000	1:1000

Imaging.

Images acquired on sections and whole mount cornea were captured either on an Eclipse E600 microscope (Nikon, Tokyo, Japan) using a SPOT camera (Spot Diagnostics, Sterling Heights, MI) or a Zeiss 510 confocal laser scanning system and Axio Imager microscope using LSM 510 software (release 4.2) (Carl Zeiss MicroImaging, Thornwood, NY).

Fluorescein staining and microscopy.

Mice were killed, and 50 μl of 0.25% fluorescein (diluted in molecular grade water from 10%, AK-Fluor; Akron, Lake Forest, IL) was applied to the eye. A UV flashlight (UV 365 1W, TK566, Tank007) was used to illuminate the fluorescence under a Zeiss Stemi SV11 microscope camera hooked up to a Canon EOS Rebel T4i camera. EOS utility software was used to acquire images of fluorescein staining the cornea.

RESULTS

Identification of genes affected by conditional ablation of Srf.

In this study, we examined the effect of Srf genetic ablation on the corneal epithelium of Dstn^corn1 mice. We performed microarray analysis with Dstn^corn1 mutant, Dstn^corn1 rescued, and WT corneas isolated from mice at P58 after 30 days of Dox induction. We identified 3,442 genes that were differentially expressed when comparing Dstn^corn1 mutant vs. WT (group A), 421 when comparing Dstn^corn1 mutant vs. Dstn^corn1 rescued (group B), and 266 when comparing Dstn^corn1 rescued vs. WT (group C). The 403 genes that are contained in both group A and group B represent genes affected by Srf ablation that are responsible for phenotypic rescue of Dstn^corn1 phenotypes (Fig. 1A).

Fig. 1.

The proportion of upregulated, downregulated, and serum response factor (SRF)-bound genes in group A and the intersection of groups A and B. A: Venn diagram showing the number of genes affected when comparing Dstn^corn1 mutant vs. wild-type (WT) (group A) together with the genes affected when comparing Dstn^corn1 mutant vs. Dstn^corn1 rescued (group B) cornea. The intersection of these 2 groups represents the genes that are significantly affected both by the Dstn^corn1 mutation and genetic ablation of Srf from the Dstn^corn1 corneal epithelium. B: the proportion of upregulated (pink and red) and downregulated (light green and bright green) genes in the set of genes that are differentially expressed between Dstn^corn1 and WT (group A). Patterned sections represent genes that contain are bound by SRF based on chromatin immunoprecipitation sequencing (ChIP-seq) analysis. C: the proportion of upregulated (red) and downregulated (green) genes in the set of genes that is the intersection of groups A and B. Patterned sections represent genes that are bound by SRF based on ChIP-seq analysis.

Identification of differentially expressed genes bound by SRF in Dstn^corn1 mutant vs. WT cornea.

Of the 3,442 genes that were differentially expressed when comparing Dstn^corn1 mutant and WT cornea (group A), we found that 1,550 (45%) were upregulated and 1,892 (55%) were downregulated in Dstn^corn1 mutant cornea (Fig. 1B). Since we previously identified the transcription factor SRF as a major contributor to Dstn^corn1 mutant phenotypes, we then identified which differentially expressed genes were bound by SRF (Fig. 1B). We found that 2,425 (73%) of the genes found to be differentially expressed in the microarray analysis were bound by SRF in either Dstn^corn1 mutant or WT tissue as indicated by ChIP-seq (Fig. 1B). Of the 2,425 genes bound, 1,078 (44% of genes bound) were genes that were upregulated in Dstn^corn1 cornea compared with WT and 1,347 (56% of genes bound) were genes that were downregulated in Dstn^corn1 cornea compared with WT.

Genes affected by Srf ablation and direct targets of SRF in the Dstn^corn1 mutant cornea.

To determine which genes were significantly affected by genetic ablation of Srf from the Dstn^corn1 corneal epithelium, we identified genes that are differentially expressed in Dstn^corn1 rescued compared with Dstn^corn1 mutant cornea (group B). Of the 421 genes, 403 are shared with group A (Fig. 1A). Of this set of 403 genes, 105 (26%) were upregulated and 298 (74%) were downregulated in Dstn^corn1 mutant compared with Dstn^corn1 rescued cornea (Fig. 1C). Since this comparison represents the genes that are affected in Dstn^corn1 rescued mice and show abnormal expression in Dstn^corn1 mutant cornea, this group of 403 genes represents the group of genes that is both affected by Srf knockout and required for the persistence of Dstn^corn1 mutant phenotypes. Genes that are bound by SRF in Dstn^corn1 mutant or WT tissues represent direct targets of SRF whose expression likely changes due to increased expression of Srf. We found that 311 (78%) of the 403 genes were bound by SRF (Fig. 1C). Of these genes bound, 79 (25% of genes bound) were upregulated in Dstn^corn1 mutant tissue compared with WT, and 232 (75% of genes bound) were downregulated in Dstn^corn1 mutant tissue compared with WT.

Clustering analysis, GO term enrichment of genes affected by knockout of Srf.

The effect of Srf ablation on Dstn^corn1 cornea can be further visualized when hierarchical unsupervised clustering is performed on the 403 genes shared between groups A and B. A dendrogram clustering the nine arrays indicates that the arrays are successfully grouped by genotype (Fig. 2). Additionally, clustering based on expression patterns highlights at least two distinct clusters: genes that are upregulated in Dstn^corn1 mutant compared with Dstn^corn1 rescued and WT (Fig. 2, red dendrogram) and genes that are downregulated in Dstn^corn1 mutant compared with Dstn^corn1 rescued and WT (Fig. 2, blue dendrogram).

Fig. 2.

Heat map displaying hierarchical clustering of the 403 genes that make up the intersection of groups A and B. These genes are the most affected by Srf ablation in the corneal epithelium of Dstn^corn1 mice. Each row corresponds to a gene, and each column to a single array. Relative expression levels are indicated by colors shown in the scale at the bottom. Gene expression values were normalized by a standardization that shifted genes to a mean of 0 and scaled to a standard deviation of 1. The dendrogram at the top shows the clustering of arrays. The dendrogram at left demonstrates that genes can be split into 2 groups: genes that are upregulated in Dstn^corn1 mutant compared with rescued and WT cornea as red lines, and genes that are downregulated in Dstn^corn1 mutant compared with rescued and WT cornea as blue lines.

GO term analysis of upregulated genes within this group of 403 genes yielded particular enrichment for “brown fat cell differentiation,” “sarcomere,” and “actin binding” (Table 4). Since SRF is a known regulator of the actin cytoskeleton, this result further demonstrates the role of SRF in abnormal actin dynamics in Dstn^corn1 cornea (32). Genes responsible for these enrichments are listed in Supplemental Table S1.¹

Table 4.

GO term enrichment for the set of genes affected by Srf ablation that are upregulated in Dstn^corn1 mutant cornea compared with WT

GO Term	Genes, n	Fold Enrichment	P Value
Biological process
Brown fat cell differentiation	3	23.04352742	0.007210445
Cellular component
Sarcomere	4	9.472727273	0.008285928
Molecular function
Actin binding	7	5.209229391	0.002010319

GO, Gene Ontology; Srf, serum response factor; WT, wild type.

GO term analysis of downregulated genes within this group of 403 genes yielded particular enrichment for terms specific to epithelial tissue (Table 5). Enrichment for the terms “epidermis development” and “ectoderm development” underscore a lack of proper epithelial tissue formation in Dstn^corn1 mutant cornea. Enrichment for the terms “apical junction complex,” “apicolateral plasma membrane,” and “occluding junction” and “tight junction,” which relate to junctions between epithelial cells, are all enriched for by genes that are downregulated in Dstn^corn1 mutant cornea compared with rescued and WT. Genes responsible for enrichments are listed in Supplemental Table S2. Downregulation for epithelial-specific genes in Dstn^corn1 mutant cornea further emphasizes that gene expression changes taking place in Dstn^corn1 cornea during rescue influence the identity of cells and the structural rearrangement of the corneal epithelium. The terms “cell-cell junction” and “plasma membrane,” while not epithelial specific, point to further changes that occur that return the entire cornea to a WT-like state.

Table 5.

GO term enrichment for the set of genes affected by Srf ablation that are downregulated in Dstn^corn1 mutant cornea compared with WT

GO Term	Genes, n	Fold Enrichment	P Value
Biological process
Epidermis development	6	4.941090909	0.007201051
Ectoderm development	6	4.643882433	0.009287635
Cellular component
Apical junction complex	8	7.503150315	8.96E-05
Apicolateral plasma membrane	8	7.357458076	1.01E-04
Occluding junction	6	7.4784689	0.001191422
Tight junction	6	7.4784689	0.001191422
Cell-cell junction	8	4.380451918	0.002278533
Plasma membrane	44	1.434273916	0.007593248
Molecular function
None

Analysis of epithelial junction defects caused by the Dstn^corn1 mutation.

To assess the physiological impact of abnormal gene expression we identified in the Dstn^corn1 corneal epithelium, we performed histological and functional analysis of epithelial junctions in Dstn^corn1 mutant, Dstn^corn1 rescued, and WT corneal epithelium. Whole cornea staining for the tight junction marker ZO-1, also known as TJP1, shows an abnormal junction pattern in the Dstn^corn1 corneal epithelium that returns to a WT-like pattern in rescued cornea. Our analysis of WT and Dstn^corn1 rescued whole cornea show flat, squamous cells joined by tight junctions. Dstn^corn1 mutant cornea contain round, loosely packed cells whose junctions are not clearly outlined (Fig. 3A). E-cadherin, a component of adherens junctions, can be seen throughout the layers of the Dstn^corn1 mutant, Dstn^corn1 rescued, and WT corneal epithelium and does not appear to be affected by the Dstn^corn1 mutant or Srf ablation (Fig. 3B). A functional test for tight junction function reveals that the Dstn^corn1 corneal epithelium lacks an impermeable WT epithelial junction network. Fluorescein staining highlights the loss of impermeability that develops in Dstn^corn1 mutant cornea compared with WT (Fig. 3C).

Fig. 3.

Junction defects compromise proper structure and function in Dstn^corn1 mutant cornea. A: ZO-1/TJP-1, a tight junction marker, highlights the restructuring of the epithelium that occurs as a result of rescue of Dstn^corn1 cornea. Whole mount cornea ZO-1 staining clearly shows junctions between flat squamous epithelial cells in Dstn^corn1 rescued (middle) and WT cornea (right). In the Dstn^corn1 mutant cornea (left), ZO-1 is not concentrated in the junctions but also observed in the cytoplasm of round, loosely packed epithelial cells. Scale bar, 10 μm. B: E-cadherin, a component of adherens junctions, is highlighted in the epithelium of Dstn^corn1 mutant, Dstn^corn1 rescued, and WT cornea. Staining is consistent between the cells of each genotypes. Scale bar, 10 μm. C: gross anatomical imaging of fluorescein staining green highlights permeable areas of the Dstn^corn1 mutant cornea (left). WT cornea (right) are impermeable and do not demonstrate any fluorescein staining. Fluorescein also highlights tears surrounding both eyes.

qPCR analysis for differentially expressed genes.

We performed qPCR on a select group of genes to further confirm our microarray analysis results. Data for the gene Dstn confirm that it is not expressed in mutant or rescued tissues (Fig. 4A). The results also confirm that Srf expression has decreased in the cornea of Dstn^corn1 rescued mice compared with Dstn^corn1 mutant mice (Fig. 4B). Figure 4 also shows that both up- and downregulated genes were consistently measured as such by both microarray and qPCR analyses.

Fig. 4.

Quantitative real-time PCR (qPCR) confirms trends observed during microarray analysis. ANOVA was performed on gene expression values for the 3 sample groups (***P < 0.001, ****P < 0.0001). A: qPCR demonstrates that expression of Dstn is detectable in WT cornea but not in Dstn^corn1 mutant or rescued cornea, which lack the coding sequence for DSTN. B: Srf expression is significantly reduced following ablation of the gene from the cornea of Dstn^corn1 mice. C: qPCR analysis is consistent with results from microarray analysis. Cald1, Cobl, Fmnl1, and Spnb3 are genes selected from the 403 genes that are affected by Srf ablation. Krt12 is a marker of corneal epithelial cell identity. ANOVA was performed on gene expression values for the 3 sample groups.

Global SRF binding is decreased in Dstn^corn1 mutant compared with WT cornea.

To determine patterns of SRF binding, we performed ChIP-seq analysis on Dstn^corn1 mutant and WT tissues. Visualization of binding to Srf revealed that SRF binds to its coding sequence in Dstn^corn1 mutant and WT tissue (Fig. 5A), consistent with Srf being a direct target of itself. Dstn is not bound in the Dstn^corn1 mutant cornea since its genomic sequence is absent (Fig. 5B), while it is bound in WT. Global analysis of peak values in Dstn^corn1 compared with WT revealed that there is an average twofold decrease of SRF binding in the Dstn^corn1 mutant cornea (Fig. 5C). We generated motif logos for the most extreme 100 peaks in the following sets: peaks whose signal decreased in the Dstn^corn1 mutant compared with WT (Fig. 5D, green box), peaks whose signal was the same in Dstn^corn1 mutant and WT (Fig. 5D, black box), and peaks whose signal increased in Dstn^corn1 mutant compared with WT (Fig. 5D, red box).

Fig. 5.

Overall SRF binding decreases in the Dstn^corn1 mutant compared with WT cornea. A: SRF ChIP binding pattern for the Srf gene. Most peaks show a decrease in Dstn^corn1 mutant (purple) SRF binding compared with WT (blue) cornea. B: SRF ChIP binding pattern for the Dstn gene, as a control. The Dstn locus is absent in the Dstn^corn1 mutant (purple) and no SRF ChIP signal is expected or observed. C: histogram of SRF peaks suggesting that there is an average 2-fold decrease in SRF binding in the Dstn^corn1 mutant cornea compared with WT on a global scale (n = 31,812 peaks). The SRF peaks can be grouped as a decrease in binding (green), increase in binding (red), or no change in binding (black) in the Dstn^corn1 mutant cornea compared with WT. D: SRF motifs logos generated for a select set of peaks for the groups described in C. For the groups of peaks that showed decreased or increased binding in Dstn^corn1, the most extreme 100 peaks were used for the generations. For the group of peaks with no change, the 100 most midrange values were used for generation of the motif. Notably, the motif logo generated for the select peaks in the group of peaks with increased SRF binding in Dstn^corn1 mutant cornea show the highest similarity to a CArG box. E: SRF peaks can also be grouped by presence of absence in the 2 genotypes: present only in the WT (dark blue), present only in the Dstn^corn1 mutant (purple), and present in both the Dstn^corn1 mutant and WT (light blue). F: the log ratio of ChIP-seq signal between Dstn^corn1 mutant and WT for the 3 groups described in E. The average log ratios are significantly different P value = 2 × 10⁻¹⁶. G: SRF motifs logos generated for a select set of peaks for the groups described in E. For each group, a random group of 100 genes was selected for generation of each motif logo. Notably, the motif logo generated for the peaks that are bound by SRF in Dstn^corn1 mutant and not WT show the highest similarity to a CArG box.

We also grouped peaks based on whether they were present in both Dstn^corn1 and WT, or exclusive to either genotype. This grouping shows that, while almost half of the peaks are present only in the WT cornea, only 2% are unique to the Dstn^corn1 mutant cornea (Fig. 5E). Even for the regions bound in both Dstn^corn1 mutant and WT tissue, there is significantly decreased average signal in the mutant compared with WT cornea (Fig. 5F). We also generated motif logos for a random group of 100 peaks in the following sets: peaks common to both Dstn^corn1 mutant and WT (Fig. 5G, light blue box), peaks only bound in WT (Fig. 5G, dark blue box), and peaks only bound in Dstn^corn1 mutant (Fig. 5G, purple box).

Data from the motif logos suggest that peaks whose signals are increased in Dstn^corn1 or are present in Dstn^corn1 and not WT are most similar to the CArG [CC(A/T₆)GG] box, a sequence contained in the SRE known to be targeted for SRF binding (2).

Analysis of genes bound by SRF in Dstn^corn1 mutant and WT cornea.

From our ChIP-seq data we further identified genes that displayed SRF binding in the gene margin. Our analysis indicates that SRF was bound to the gene margin of 16,698 genes; 12,157 (73%) genes were bound in both Dstn^corn1 mutant and WT tissues. We found that 4,367 (26%) genes were bound by SRF in WT tissue but not bound in Dstn^corn1 mutant tissue. Conversely, 174 (1%) genes were bound in Dstn^corn1 mutant tissue but not WT (Fig. 6A). To investigate whether SRF demonstrated a different pattern of binding positions for genes bound only in Dstn^corn1 mutant or only in WT, we compared binding events between these two groups. Our analysis did not indicate that SRF binding positions were changed if a gene experienced SRF binding only in WT or only in Dstn^corn1 mutant tissue (Fig. 6B). These data, together with our microarray results, suggest that the Dstn^corn1 mutant cornea experiences less SRF binding, despite demonstrating a higher abundance of the protein.

Fig. 6.

Analysis of SRF-bound genes and gene binding events in WT and Dstn^corn1 mutant cornea. A: 16,698 genes can be grouped by presence or absence of SRF-binding in Dstn^corn1 mutant and WT: present only in the WT (dark blue), present only in the Dstn^corn1 mutant (purple), and present in both the mutant and WT (light blue). B: by comparing the location of binding events in genes that are bound by SRF only in the Dstn^corn1 mutant (purple) or WT (dark blue), we can see that SRF binding unique to WT or Dstn^corn1 mutant tissues does not appear to affect the location of binding events relative to the transcriptional start site (TSS), indicated by a black vertical line.

Since SRF is a known regulator of miRNAs (16, 26, 27, 45, 48), we investigated whether our ChIP-seq analysis indicated that gene margins of miRNAs were bound by SRF. The analysis revealed SRF binding to 265 miRNA gene margins; 174 (66%) demonstrated SRF binding in both Dstn^corn1 mutant and WT tissues. We found that 88 (33%) miRNA margins were bound by SRF in WT tissue but not bound in Dstn^corn1 mutant tissue. Conversely, 3 (1%) miRNA margins were bound by SRF in Dstn^corn1 mutant tissue but not WT (Fig. 7).

Fig. 7.

Analysis of SRF bound miRNAs in WT and Dstn^corn1 mutant cornea. SRF binding to the gene margin of 265 miRNAs can be grouped by the presence or absence of SRF binding in Dstn^corn1 mutant and WT: present only in the WT (dark blue), present only in the Dstn^corn1 mutant (purple), and present in both the Dstn^corn1 mutant and WT (light blue).

In vivo analysis of SRF targets caldesmon and melanoma cell adhesion molecule.

We sought to further confirm our microarray and ChIP seq results by characterizing changes in downstream targets of SRF at the protein level in vivo.

Based on our analyses we found that caldesmon (Cald1), a previously confirmed SRF target and regulator of actin dynamics, is bound by SRF (Fig. 8A) and upregulated in Dstn^corn1 mutant tissues (Fig. 4C). IHC for CALD1 correlates with our observation that Cald1 expression is upregulated in Dstn^corn1 cornea and returns to near-WT levels upon ablation of Srf from the corneal epithelium of Dstn^corn1 mice (Fig. 8B). CALD1 is restricted to the basal cell layer in both WT and Dstn^corn1 rescued tissues but highlights all layers throughout the Dstn^corn1 mutant epithelium. CALD1 appears to be concentrated in the basal portion of individual WT corneal epithelial cells. This pattern of localization is not observed in Dstn^corn1 corneal epithelium.

Fig. 8.

In vivo analysis of select SRF targets. A: SRF ChIP binding pattern for the Cald1 gene in WT (blue) and Dstn^corn1 mutant (purple) cornea. B: a stain for CALD1 highlights differences in cell polarity between Dstn^corn1 mutant, rescued, and WT corneal epithelia. CALD1 is localized to the basal side of corneal epithelial cells in WT corneal epithelium (right). In Dstn^corn1 rescued mice, CALD1 highlights the cytoplasm of entire cells in the basal layer of the corneal epithelium (middle). In the Dstn^corn1 mutant corneal epithelium, CALD1 is present in all layers and throughout the cells of the abnormal corneal epithelium (left), suggesting a defect in cell polarity. Scale bar, 10 μm. C: SRF ChIP binding pattern for the Mcam gene in WT (blue) and Dstn^corn1 mutant (purple) cornea. D: an antibody against MCAM/CD146/MUC18 highlights the nuclear envelope and shows increased expression in Dstn^corn1 mutant cornea. MCAM is localized to the nuclear envelope of WT (right) and Dstn^corn1 rescued (middle) corneal epithelial cells. The level of MCAM in Dstn^corn1 mutant corneal epithelium (left) is increased compared with WT and Dstn^corn1 rescued tissues. Scale bar, 10 μm.

Based on our analyses we found that melanoma cell adhesion molecule (Mcam) is bound by SRF, identifying it as a new potential direct target of SRF regulation. Mcam is not only bound by SRF (Fig. 8C) but, based on our microarray data, is also highly upregulated in the Dstn^corn1 cornea. IHC analysis confirms upregulation of MCAM in the Dstn^corn1 corneal epithelium and also reveals that MCAM is localized to the nuclear envelope of mutant, rescued, and WT corneal epithelial cells (Fig. 8D).

DISCUSSION

Dstn^corn1: an in vivo model of epithelial tissue abnormalities.

Previous work on the Dstn^corn1 cornea established in vivo links between abnormal actin dynamics and SRF and also implicated SRF as a major factor in the formation of neovascularization, chronic inflammation, and epithelial hyperproliferation (34, 36, 37). This study utilizes a conditional knockout of Srf to continue the construction of a molecular network that underlies these disease phenotypes. These results shed more light on the role of SRF in both the Dstn^corn1 mutant and WT cornea.

Conditional ablation of Srf from the cornea of Dstn^corn1 mice leads to gene expression changes.

Our results demonstrate that genetic ablation of Srf in the corneal epithelium of Dstn^corn1 mice via a tetracycline inducible system leads to gene expression changes that contribute to the rescue of Dstn^corn1 phenotypes. We observed 3,442 differentially expressed genes when comparing Dstn^corn1 mutant and WT cornea. However, only 421 genes were shown to be differentially expressed when comparing Dstn^corn1 mutant and Dstn^corn1 rescued cornea (Fig. 1). The 403 genes that are common between these two groups are SRF targets that are chronically misregulated and functionally important for the persistence of Dstn^corn1 mutant phenotypes.

Our data suggest that a small subset of the total number of genes whose expression is changed as a result of the Dstn^corn1 mutation are primarily affected by SRF and allow abnormal phenotypes to persist. These results indicate that SRF acts as effector to influence the gene expression of a specific set of target genes in vivo. We hypothesize that SRF acts via these targets to activate gene networks that contribute to neovascularization, epithelial hyperproliferation, and inflammation Dstn^corn1 cornea.

Rescue of Dstn^corn1 phenotypes affects epithelial cell-cell junctions.

Among the 403 genes affected by Srf ablation, genes downregulated in Dstn^corn1 mutant cornea compared with rescued and WT cornea are enriched for GO terms that influence epithelial cell-cell junctions (Table 5, Supplemental Table S2). Enrichment for these terms suggests that these genes are regulated by SRF and play a particularly important role in the development and maintenance of the corneal epithelium.

Tight junctions throughout the WT epithelium contribute to the impermeability of the membrane. Our results suggest that epithelial junction components are mislocalized in the Dstn^corn1 mutant cornea and do not contribute to the formation of a functional epithelial cell junction network. Our IHC data also indicate an increase in localization of ZO-1 to the cytoplasm in Dstn^corn1 mutant compared with Dstn^corn1 rescued and WT tissues (Fig. 3A). Cytoplasmic localization of ZO-1 has been highlighted as a symptom of inflammation (6). Inflammation that occurs in the Dstn^corn1 cornea may drive the migration of ZO-1 from the cell membrane to the cytoplasm. E-cadherin, a component of adherens junctions, does not appear to change between Dstn^corn1 mutant, Dstn^corn1 rescued, and WT corneal epithelia, suggesting that the defect may be specific to certain junction molecules (Fig. 3B).

Using fluorescein staining to test for corneal epithelial permeability, we are also able to show that the loss of epithelial junction gene expression leads to functional defects in Dstn^corn1 mutant cornea. Due to loss of junctions, the Dstn^corn1 corneal epithelium demonstrates a clear loss of impermeability that is required for the cornea to serve as a functional environmental barrier (Fig. 3C).

Previous studies in endothelial cells, WT epithelium, and stem cells have demonstrated that genetic ablation of Srf in WT tissue leads to defects in junctions between cells (10, 29, 35). In our system, however, Srf ablation from the corneal epithelium corrects defects in epithelial cell-cell junctions. These opposing effects of Srf ablation likely result from the fact that previous studies knocked out SRF in WT tissue, while our model knocks out Srf in the Dstn^corn1 mutant tissue. The Dstn^corn1 mutation has a known effect on the balance of actin and cytoskeletal dynamics, which may significantly change the effect of Srf ablation on cell-cell junctions. In addition, the correction of epithelial junction defects may be corneal epithelial cell-specific. SRF is not present in high abundance and may not have very significant effects in the WT corneal epithelial tissue (36). In the Dstn^corn1 mutant, SRF is abnormally upregulated, leading to corneal epithelial defects that include cell junction defects. Downregulation of SRF in the Dstn^corn1 mutant, therefore, may have different effects than in WT tissues where SRF has significant roles in the establishment and maintenance of cell-cell junctions.

Our model demonstrates that the establishment of epithelial junctions is tightly regulated by both the balance of actin dynamics and SRF. These data confirm that proper regulation of actin dynamics and limited SRF-driven transcription play an essential role in the establishment of functional junction complexes in the corneal epithelium. In addition, these data suggest that the proper establishment of epithelial junctions in the corneal epithelium may play a significant role in establishment of an avascular cornea with limited inflammation and controlled proliferation.

Paradox of increased SRF protein level and decreased expression of SRF target genes.

SRF is thought to activate its targets, and our data, together with our previous work, identified approximately equal numbers of up- and downregulated genes in Dstn^corn1 mutant cornea compared with WT (Fig. 1B) (34). By using Dstn^corn1 rescued cornea, whose cells lack Srf due to conditional ablation, we hoped to identify genes that are directly affected, that is, are transcriptionally activated by, SRF. Since Srf is upregulated in the Dstn^corn1 mutant cornea, we expected to see that genes differentially expressed between Dstn^corn1 mutant compared with WT and Dstn^corn1 rescued cornea would contain more upregulated genes. On the contrary, 298 (74%) of the genes affected by Srf ablation, including 232 direct SRF targets, are downregulated in Dstn^corn1 mutant cornea compared with WT (Fig. 1B). Despite upregulation of Srf, there appears to be more downregulation of SRF target genes in the Dstn^corn1 mutant cornea.

On the basis of this observation, we speculate that there must be a molecular pathway through which SRF negatively regulates gene expression. ChIP-seq analysis enabled us to determine patterns of and differences in SRF binding in Dstn^corn1 mutant and WT cornea. Intriguingly, we found that 26% of the 16,698 genes identified in our ChIP-seq analysis are bound by SRF in WT cornea but lack SRF binding in the Dstn^corn1 mutant cornea (Fig. 6A). The set of genes that are bound by SRF in Dstn^corn1 mutant cornea but not WT make up only 1% of the genes identified in the analysis (Fig. 6A). Further analysis of the SRF binding motif showed that, while SRF binding to degenerate CArG boxes appears to be lost or decreased, binding to the consensus sequence [CC(A/T₆)GG] is increased in Dstn^corn1 cornea. Mutations that disrupt a consensus, or true, CArG box and instead degenerate to a CArG-like sequence are known to dramatically alter the binding kinetics and subsequent transcriptional activity of SRF (2, 9, 23, 32). Our findings indicate that the Dstn^corn1 cornea displays loss of or decreased SRF binding to CArG-like sequences and increased binding to consensus, or “true” CArG boxes (Fig. 5, D and G).

Another possible route of negative transcriptional regulation is provided through generation of microRNAs (miRNAs) by SRF, which is well established in development and adult tissues (16, 26, 27, 45, 48). Increased generation of miRNAs in the Dstn^corn1 could be responsible for negative regulation of genes in Dstn^corn1 cornea. However, our ChIP-seq data indicate that trends of miRNA regulation parallel that of gene regulation in the Dstn^corn1 mutant cornea (Fig. 7). SRF binding is absent for one-third of identified miRNAs in Dstn^corn1 mutant compared with WT, suggesting that, for miRNAs as well as genes, there is an significant loss of SRF binding and decreased SRF-driven transcription in Dstn^corn1 mutant tissue. Nonetheless, it is still possible that a small number of miRNAs that are bound by SRF in Dstn^corn1 but not in WT cornea could participate in the negative regulation of genes to some degree.

Together, our microarray and ChIP-seq results indicate that downregulation of genes in the Dstn^corn1 corneal epithelium occurs due to decreased binding of SRF, despite the protein being present in higher abundance (12, 36). Our findings suggest a novel role of SRF in the large-scale negative regulation of gene expression. Although further analysis is necessary, one possible mechanism for decreased SRF binding in Dstn^corn1 mutant cornea is that other factors that inhibit the binding of SRF may also be misregulated. For example, YY1 is known to compete with SRF for binding to SREs (18). Increased expression of such SRF inhibitors in the Dstn^corn1 cornea compared with WT cornea could result in decreased SRF binding and global changes in the transcription of genes in the Dstn^corn1 mutant cornea. Considering that the effect of SRF binding appears to be dependent on the sequences of the binding motifs (Fig. 5, D and G), binding of such SRF inhibitors may also be sequence dependent.

Dstn^corn1 phenotypes mimic abnormalities in epithelial cancers and are corrected by ablation of Srf from the corneal epithelium.

The formation of new blood vessels, chronic inflammation, and epithelial hyperproliferation are symptoms of tumorigenesis and other cancerous and diseased tissues. Genetic mutations in mutant cells allow these abnormal tissues to proliferate at accelerated rates with nutrients and resources rushed to the site by inflammation and supplied by neovascularization. Increased expression of SRF has been shown to play a role in the formation and metastasis of many different types of cancer (5, 19, 28, 42). In fact, SRF consistently correlates with more aggressive tumors and metastasis (5, 43). Many of the gene expression changes we observed in Dstn^corn1 mice compared with WT, including increased expression of SRF, are similar to those in metastatic epithelial cancers (28). Tumor-specific splice variants of Cald1 have been identified in studies of colon, bladder, and prostate cancers (8, 28a, 33). Mcam was so named because of its high expression in melanoma tissues and can be used to diagnose different types of tumors undergoing a transition from an epithelial state (22, 38, 40, 41). Interestingly, we have also found that defects in cell polarity and identity, also characteristic of cancer tissues, are prevalent throughout the Dstn^corn1 mutant corneal epithelia (36, 44).

Our study further demonstrates the essential role SRF plays in the development of abnormal neovascularization, proliferation, and inflammation originating from the epithelial tissue. We have revealed the potential to control these processes in abnormal epithelial tissue via genetic ablation of Srf. The results of this study highlights differences in SRF regulation in the corneal epithelia of mutant and WT tissues, and a detailed network of genes will enhance the resolution of molecular pathways underlying epithelial abnormalities in the Dstn^corn1 mutant cornea.

GRANTS

This work was supported National Institutes of Health (NIH) Grant R01EY-016108 and Core Grant P30HD-03352 to the Waisman Center. Support for S. V. Kawakami-Schulz was partially provided by the NIH predoctoral training program in genetics (NIH 5T32GM-07133).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the author(s).

AUTHOR CONTRIBUTIONS

Author contributions: S.V.K.-S., A.M.V., S.G.S., and S.I. performed experiments; S.V.K.-S., S.G.S., E.J., A.I., and S.I. analyzed data; S.V.K.-S., A.M.V., S.G.S., E.J., W.W.-Y.K., A.I., and S.I. interpreted results of experiments; S.V.K.-S., S.G.S., and E.J. prepared figures; S.V.K.-S., A.I., and S.I. drafted manuscript; S.V.K.-S., A.M.V., S.G.S., E.J., W.W.-Y.K., A.I., and S.I. edited and revised manuscript; S.V.K.-S., A.M.V., S.G.S., E.J., W.W.-Y.K., A.I., and S.I. approved final version of manuscript; A.M.V., W.W.-Y.K., A.I., and S.I. conception and design of research.