Genetic background-dependent role of Egr1 for eyelid development

Ocular Anomalies in Egr1^−/− BALB/c but not Egr1^−/− C57BL/6 Mice.

Egr1^−/− mice on the C57BL/6 background were backcrossed to the BALB/c background. Although no abnormalities were observed in Egr1^+/− and Egr1^−/− C57BL/6 mice (Fig. 1A) or in Egr1^+/+ and Egr1^+/− BALB/c mice after five generations of backcrossing (Fig. 1B), substantial abnormalities were observed at the F5 Egr1^−/− BALB/c generation, including cryptophthalmos (narrow palpebral fissure), microphthalmia (small eye size) and, in some mice, opaque cornea (Fig. 1C). Compared with normal mice (Fig. 1D), histological analysis revealed anophthalmia (absence of one or both eyes) (Fig. 1E), severe microphthalmia (Fig. 1F), retinal dysplasia (Fig. 1G), cataracts (Fig. 1 H and I), keratitis (Fig. 1 H–J), corneal neovascularization (Fig. 1 H–J), and/or calcification (Fig. 1H and Table 1). Interestingly, the severity of defects often varied substantially between the two eyes in the same mouse, indicating variable penetrance of the phenotype even within a single animal.

These results suggested that Egr1 deficiency conferred these ocular abnormalities on the BALB/c but not the C57BL/6 background, but conceivably a de novo mutation in another gene might have occurred spontaneously during backcrossing. We thus performed a second independent backcross of Egr1^−/− C57BL/6 mice to the BALB/c background, and by the F4 generation the Egr1^−/− but not Egr1^+/− or Egr1^+/+ mice again showed ocular anomalies similar to those of the first backcross, including anophthalmia (Fig. S1A), severe microphthalmia (Fig. S1B), retinal folds (Fig. S1C), keratitis/corneal neovascularization (Fig. S1D), and cataracts (Fig. S1D and Table 1). These results confirmed that Egr1 mutation was indeed responsible for the eye defects on the BALB/c background, underscoring the importance of the genetic background and the possibility of contributions from strain-specific modifier gene(s).

Abnormal Eyelid Formation in Egr1^−/− BALB/c Embryos and Ocular Abnormalities After Birth.

To determine how early the ocular abnormalities occurred, we next studied 1- to 4-d-old mice. As expected, eyes from Egr1^+/+ and Egr1^+/− BALB/c mice were normal (Fig. 2A); however, histological analysis showed mild corneal neovascularization and keratitis abnormalities in Egr1^−/− BALB/c mice at postnatal days 1 and 2 (Fig. 2B and Table S1). In one case, an eye was severely malformed (Fig. 2C). At postnatal days 3 and 4, Egr1^−/− BALB/c mice exhibit a greater severity, including microphthalmia, keratitis, and corneal neovascularization as well as corneal melting and perforation (Fig. 2 D and E), vitreous hemorrhage (Fig. 2 E and F), inflammation (Fig. 2 E), and retinal detachment (Fig. 2 E and F) or only residual remnant retina (Fig. 2G) (abnormalities are summarized in Table S1). In addition to these abnormalities, a striking finding was that although all Egr1^+/+ and Egr1^+/− BALB/c mice had normal palpebral fissures and thus had normally closed eyelids (Fig. 2 H and I), 13 of 16 Egr1^−/− BALB/c mice had no evidence of eyelid formation and closure at postnatal days 1–4 (Fig. 2J and Table S1). Importantly, when eyelids were normally closed in Egr1^−/− BALB/c mice, the ocular structure was either normal or had only mild anomalies, compared with mice with unformed eyelids (Table S1), suggesting a causal relationship between the failure of normal eyelid closure and ocular damage.

In mice, eyelid formation is not evident at day E13.5, but the epidermis of both upper and lower eyelids grows rapidly and merges to cover the cornea at approximately day E15.5 over the next 24 h (15–17). To identify when defective eyelid formation occurred, Egr1^+/− and Egr1^−/− BALB/c embryos were analyzed. At day E16.5, 4 Egr1^+/− BALB/c embryos showed epidermis growth, with four of eight embryos having completely merged eyelids and the other four having one eyelid merged (Fig. 3A, Right) and the other still only partially merged (Fig. 3A, Left, and Table S2). In contrast, all seven Egr1^−/− embryos had defective eyelid formation; fusion was completely defective in four of the seven cases and was partial in the other three (Fig. 3B and Table S2). In all E17.5 and E18.5 Egr1^+/− BALB/c embryos, both eyelids were completely merged (Fig. 3 C and E), in marked contrast to the defective eyelid formation in all Egr1^−/− BALB/c embryos (Fig. 3 D and F and Table S2). Despite deformed eyelid formation, the internal ocular structure was relatively normal in all Egr1^−/− BALB/c embryos, indicating that the ocular anomalies observed postnatally and in adults, as shown in Fig. 1, occurred later than day E18.5.

Egr1 Is Expressed in the Eyelid During Embryogenesis.

Because of the defective eyelid formation, we next investigated whether the Egr1 gene is expressed in the eyelid during embryogenesis. Serial sections of E13.5, E15.5, and E17.5 embryos from C57BL/6 WT mice were stained by immunohistochemistry using an antiserum to EGR1 in the presence (Fig. 4 A, C, and E) or absence (Fig. 4 B, D, and F) of an EGR1-blocking peptide. Although no clear signal was seen in the presence of the blocking peptide, staining of eyes from E13.5 mice demonstrated robust nuclear staining of a subpopulation of mesenchymal cells in the developing eyelids, orbit, and periocular tissues as well as in the epithelia of the developing lens and conjunctival sac (Fig. 4B). Strong nuclear EGR1 staining was absent from the neuro-ectoderm–derived presumptive retinal pigment epithelium (RPE) and neural retina (Fig. 4B). By E15.5, EGR1 staining was prominent in the surface epithelium covering the now-closed eyelids as well as at the base of the developing cilia (eyelashes) (Fig. 4D). At day E17.5, staining was most prominent at the base of hair follicles and in scattered mesenchymal cells (Fig. 4F). Similar results were observed when sections from E15.5 embryos from BALB/c WT mice were stained with an antiserum to EGR1 in the presence (Fig. 4G) or absence (Fig. 4H) of an EGR1-blocking peptide. Prominent EGR1 staining was detected at the base of the developing cilia at the leading edge of the eyelid (Fig. 4H). Thus, Egr1 is indeed expressed in the developing eye, including in the eyelid. Given the expression of EGR1 in the eyelid and the defective eyelid development in Egr1-deficient BALB/c background mice, it will be interesting in the future to perform microdissection and subsequent RNA-sequencing (RNA-seq) of eyelid regions in BALB/c WT and KO embryonic mice; this investigation might provide more information related to the genes that are important for eyelid development.

Mutations of EGR1 Were Not Found in Patients with Coloboma and Multiple Congenital Anomaly Syndrome.

Given the abnormalities that we observed in Egr1-deficient BALB/c mice, we wondered if EGR1 deficiency might be associated with congenital ocular abnormalities in humans as well. “Congenital ocular coloboma” refers to defects in which normal tissue in or around the eye is absent from birth and has been reported in 3–11% of blind children worldwide. Typical coloboma—a defect in the inferonasal quadrant of the iris, retina/RPE/choroid and/or optic nerve—is caused by failure of the ventral optic fissure to close during the fifth week of human gestation (18). Rarely, coloboma and posterior staphyloma are reported in isolated cases (19, 20). To evaluate whether EGR1 might be responsible for a subset of patients with coloboma, 67 unrelated patients with this disease were analyzed, but no mutations were identified in the EGR1 coding regions or adjacent intronic regions, indicating that mutations in EGR1 do not explain the occurrence of coloboma in these patients. We also screened 88 multiple congenital anomaly (MCA) patients available to us for EGR1 mutations and detected a heterozygous variant involving nt1600 (C to T), P443L, in one patient. Although this variant has not been recorded in the Ensembl database, heterozygosity suggests that this change is unlikely to be responsible for the defects in this patient.

Identification of Differentially Expressed Genes in Egr1^−/− BALB/c Mice.

A common abnormality in the Egr1^−/− BALB/c mice was corneal neovascularization. Previously, it was reported that the proinflammatory cytokine IL-1β can induce neovascularization in both C57BL/6 and BALB/c mice (21). We thus measured ocular Il1b mRNA levels and found the levels were increased in adult Egr1^−/− BALB/c mice but not in Egr1^+/+ BALB/c, Egr1^+/+ C57BL/6, or Egr1^−/− C57BL/6 mice (Fig. 5A). Similarly, Il1b mRNA was increased in the eyes of newborn Egr1^−/− BALB/c mice but not in Egr1^+/+ BALB/c or Egr1^+/− BALB/c mice (Fig. 5B), suggesting that IL-1β may be involved in the development of ocular anomalies in Egr1^−/− BALB/c mice. We therefore backcrossed C57BL/6 mice homozygous for the Il1r^tm1Imx-targeted mutation, which prevents responsiveness to IL-1, to BALB/c for five generations and then crossed these mice to F5 Egr1^−/− BALB/c mice to generate double-KO mice. As expected, most (six of seven) Egr1^−/− BALB/c mutant mice had defects in one or both eyes, but all four Il1r^tm1ImxEgr1^−/− BALB/c double-KO mice also had defects related to one or both eyes (Table S3), indicating that disease still progressed even when IL-1β could not function, as is consistent with contributions by other cytokines or growth factors.

To identify other genes that potentially contribute to the ocular phenotype, RNA-seq was performed with RNA from eyes from adult Egr1^−/− BALB/c mice (see Dataset S1A for the complete RNA-seq data). RNA-seq results confirmed higher Il1b expression in Egr1^−/− BALB/c mice than in WT mice. In addition, there was increased expression of cytoskeleton genes, including the Krt1, Krt16, Lor, Sprr family, and Lce family genes, which are involved in the process of cornification in which an epidermal barrier is formed in stratified squamous epithelial tissue. We also identified increased expression of several kallikrein genes that are involved in the regulation of blood pressure and activation of inflammation. Interestingly, the expression of γ-crystallin genes, Cryga and Cryge, was significantly lower in Egr1^−/− BALB/c mice. Crystallins are the main structural proteins of the vertebrate ocular lens, and their appropriate arrangement is critical for lens transparency, with mutations in the Cryg genes leading mainly to lens opacification and cataracts (22, 23).

Because we found that internal ocular anomalies occurred after birth, we also performed RNA-seq analyses using eyes from newborn Egr1^−/− BALB/c mice, revealing increased expression of some of the genes identified in adult mice (see Dataset S1B for the complete RNA-seq data, comparing gene expression in “open” vs. “closed” eyes). In addition to cytoskeleton genes related to cornification, many more keratin genes are highly expressed in Egr1^−/− BALB/c mice, suggesting keratinization at an early age. In one Egr1^−/− BALB/c mouse, the expression of many crystallin and lens protein genes was strikingly decreased, possibly explaining why Egr1^−/− BALB/c mice often develop cataracts.

Because in some Egr1^−/− BALB/c mice, one eye had normally formed eyelids and a relatively normal ocular structure, whereas the other eye had unmerged eyelids and malformed ocular structure, we performed RNA-seq using RNA from each eye of three 3-d-old Egr1^−/− mice with unmerged eyelids and from Egr1^+/− littermate controls. A multidimensional scaling (MDS) plot shows that gene-expression profiles in eyes from Egr1^−/− mice with unmerged eyelids (eyes that appeared open) were very different from those in eyes from either Egr1^+/− or Egr1^−/− mice in which the eyelids formed normally (Fig. 5C and Dataset S2 for the complete RNA-seq data), suggesting that normally closed eyelids prevented ocular anomalies. To identify genomewide targets of EGR1-regulated genes that are important for eye development, differentiation, and function, we analyzed differential gene expression in eyes from mice without or with eyelid closure. We identified 854 and 37 genes that were significantly up- or down-regulated [fold change (FC) >2, P < 0.05], respectively, in eyes with defective eyelid closure (Fig. 5D). We identified the top 50 most differentially expressed genes, displayed by heat map (Fig. 5E; sorted by P value) and volcano plot (Fig. 5F). Very interestingly, Sprr1a (24), Krt16 (25), and Il1b, which previously had been shown to be involved in the development of ocular disease, were highly up-regulated in eyes without eyelids (Fig. 5G). Our data overall suggest that the differences in gene expression are secondary to the failure of eyelid formation rather than directly caused by the absence of Egr1.

The Role of the Tyr Gene in the Ocular Abnormalities.

A major difference between C57BL/6 (black) and BALB/c (albino) mice is eye and skin pigmentation. Interestingly, the enzyme tyrosinase, which catalyzes the first two steps of melanin biosynthesis, was previously identified as a modifier of the trabecular meshwork phenotype in a mouse model of human primary congenital glaucoma in which the presence of the homozygous Tyr^c-2J allele conferred much more severe developmental defects to Cyp1b^−/− mice (26). Because homozygous Tyr^c-2J mice are on a C57BL/6 background but are phenotypically indistinguishable from BALB/c mice, which harbor the mutant Tyr^c allele, we investigated whether the Tyr gene might contribute to the eyelid abnormalities we observed by crossing Egr1^−/− C57BL/6 mice to Tyr^c-2J homozygous mice. Strikingly, although all pigmented Egr1^−/− C57BL/6 mice had normal eyes (Fig. 6A), 7 of 12 albino Tyr^c-2J homozygous Egr1^−/− C57BL/6 mice developed ocular anomalies. These abnormalities tended to be milder than those in the BALB/c background, with keratitis, corneal neovascularization (Fig. 6B), cataracts, anterior synechiae, retinal dysplasia, anterior segment angle defect, and persistent hyperplastic primary vitreous (PHPV) (Fig. 6 C and D and Table S4), and two albino Egr1^−/− C57BL/6 mice exhibited full-thickness corneal defects with invasion of epithelium (Fig. 6C) as well as staphyloma (Fig. 6D). We did not observe staphyloma in any Egr1^−/− BALB/c mice. Thus, although the Tyr^c-2J homozygous mutation increased ocular abnormalities in Egr1^−/− C57BL/6 mice, the phenotype was less severe than and was somewhat different from that observed in Egr1^−/− BALB/c mice, indicating that additional modifier gene(s) may contribute to the ocular phenotype of Egr1^−/− BALB/c mice.

Mice.

All animal experiments used protocols approved by the National Heart, Lung, and Blood Institute (NHLBI) Animal Use and Care Committee and followed NIH guidelines for humane animal use. Egr1^−/− C57BL/6 mice (43) were from the Jackson Laboratory (Stock no. 012924) and genotyping was performed by PCR with primer sequences from the Jackson Laboratory. The mice were backcrossed to BALB/c mice from the Jackson Laboratory for five generations in the initial backcross or for four generations in the second backcross. Homozygous Tyr^c-2J mice (stock number 58; Jackson Laboratory) were crossed to Egr1^−/− C57BL/6 mice. Mice homozygous for the Il1r^tm1Imx-targeted mutation (stock number 3245; Jackson Laboratory) were backcrossed to BALB/c mice for five generations and then were crossed with F5 Egr1^−/− BALB/c mice to generate double-KO mice. Adult mice were examined between 6 and 21 wk of age.

Ocular Examination.

Eyelids, conjunctivae, corneas, anterior chambers, irises, lenses, and retinas of embryonic, neonatal, and adult mice were examined under a slit-lamp biomicroscope (Space Coast Laser, Inc.) and/or dissecting microscope.

Histology and Immunohistochemistry.

Eyes including the eyelids from neonatal and adult mice were enucleated and fixed in 4% glutaraldehyde for 30 min, in 10% formalin for >24 h, embedded in methacrylate, and sectioned via a vertical pupillary–optic nerve plane. Mouse embryos were decapitated, and whole heads were fixed in 10% formalin for >24 h, embedded in methacrylate, and sectioned in the coronal plane through the pupillary–optic nerve axis. Specimens that contained bone (osseous tissue) were deparaffinized in xylenes and rehydrated through graded ethanols. Each eye was cut and stained with H&E. For EGR1 staining, C57BL/6 mouse embryos were fixed in 10% buffered formalin, embedded in paraffin, and serial sections (5 μm thick) were cut for histology and immunohistochemistry. Endogenous peroxidase was blocked with 3% H₂O₂. Sections then were subjected to antigen retrieval using a citrate buffer (BioGenex) for 10 min at 100 °C. Following a 5% normal goat serum block, three sets of slides were incubated overnight at 4 °C with (i) Egr1 Ab (rabbit mAb) (no. 4153; Cell Signaling Technology) at 1:100, (ii) Egr1 Ab/blocking peptide (no. 1015; Cell Signaling Technology) mix (1:2 by volume), and (iii) rabbit monoclonal IgG isotype control reagent (no. 3900; Cell Signaling Technology). Slides were rinsed, and biotinylated goat anti-rabbit IgG (Vector Labs) that had been diluted 1:100 in TBS-Tween was applied for 30 min. After rinsing in TBS-Tween, Avidin Biotin Complex (Peroxidase ABC kit; Vector Labs) was applied to the slides for 30 min. Localization was visualized by incubation in diaminobenzidine (DAB) substrate for 5 min. Sections were counterstained with H&E, and then coverslips were added.

RNA Isolation and RT-PCR Analysis.

Eyes were enucleated and placed into Eppendorf tubes filled with RNAlater (Thermo Fisher), and total RNA was extracted using the RNeasy Mini Kit (Qiagen). The first strand was reverse-transcribed using the Omniscript Reverse Transcription Kit (Qiagen) and random primers. Real-time quantitative PCR was performed on a 7900HT Real-time PCR system (Applied Biosystems) using real-time primers and TaqMan probes for Il1b from Applied Biosystems. Expression was normalized to that of Rpl7.

RNA-Seq Analysis.

Poly(A)⁺ RNA was selected twice from total RNA using Dynal oligo magnetic beads (Invitrogen). Double-stranded cDNA was synthesized using SuperScript (Invitrogen) and fragmented using Bioruptor (Diagenode). Adaptors (Illumina) were ligated to the dsDNA using T4 DNA ligase (New England Biolabs) after end repair using the End-It Kit (Epicentre) and dA addition. Libraries were size-selected for 250- to 400-bp fragments, purified using 2% E-Gel (Invitrogen), and amplified for 18 cycles by PCR with PE 1.0 and PE 2.0 primers (Illumina) and Phusion High Fidelity PCR Master Mix (New England Biolabs). PCR products were sequenced on a Solexa 2G Genome Analyzer (Illumina). DNA sequence reads were mapped to the mouse genome [RefSeq gene database using the mm8 revision from the University of California, Santa Cruz (UCSC) genome browser]; 24,769 genes were used for RNA-seq analysis. RNA-seq data have been deposited in the Gene Expression Omnibus (National Center for Biotechnology Information).

DNA Sequencing of the Human EGR1 Gene.

The EGR1 gene was amplified by PCR with the following primers: for exon 1, forward primer 5′-CCGGTCCTGCCATATTAGG and reverse primer 5′-GGCAAGGCAAGTCTTACTGG; for exons 2–3, forward primer 5′-GCCACTGGTGCGGGTCCAGGA and reverse primer 5′-CATGGCTGTTTCAGGCAGCTG. PCR conditions were 95 °C for 5 min, followed by 35 cycles of 95 °C for 30 s, T °C for 30 s (where T = 58 °C for exon 1 and 68 °C for exons 2–3), and 72 °C for 1 min, followed by 72 °C for 7 min. The PCR products were gel-purified (Qiagen) and sequenced directly using the following primers: for exon 1, 5′-TCCCTCTAACACATGACTCTG; for exons 2–3, 5′-TTGGATGGCACTGCGCGCTC, 5′-GCCAGGAGCGATGAACGCAAG, 5′-GAACAACGAGAAGGTGCTGGT, 5′-GCACCCAGCAGCCTTCGCTAA, 5′-TTAGCGAAGGCTGCTGGGTGC, and 5′-GCCTCTTGCGTTCATCGCTCCT.