Introduction

FOXC1 encodes a transcription factor that binds regulatory DNA elements of its direct targets via a winged helix forkhead domain [1]. Together with FOXQ1 and FOXF2, FOXC1 is part of a conserved block of FOX transcription factor genes on human chromosome 6. FOXC1 is involved in the development of the anterior segment of the eye and other organs [2]. The initial FOXC1-related phenotype was identified as Axenfeld–Rieger Syndrome Type III (ARS, OMIM #602482), which involves posterior embryotoxon, iris hypoplasia, irido-corneal adhesions and ~ 50% chance to develop glaucoma [3]. Later, FOXC1 mutations were discovered in patients with aniridia, Peters anomaly and primary congenital glaucoma (PCG) [4,5,6]. Patients with mutations in FOXC1 often have additional non-ocular anomalies, such as heart defects, craniofacial dysmorphisms, hearing loss, skeletal anomalies (hip dysplasia or scoliosis), feeding issues, dental enamel hypoplasia, hypotonia/delay, and white matter lesions in the brain [7,8,9,10,11]. Mutations in FOXC1 explain a high proportion of cases affected with ARS and related disorders; however, there is still a considerable number of patients with an unknown genetic cause [8].

The expression of developmental genes is finely controlled by their regulatory elements which are often evolutionarily conserved [12]. Mutations in regulatory elements have been implicated in eye and other developmental disorders [4, 13,14,15,16]. Moreover, genome-wide association studies (GWAS) indicated that the majority of disease-associated loci lie in noncoding regions of the genome [17]. Specific to FOXC1, a GWAS study discovered a SNP at ~ 61.7 kb 5′ of FOXC1 (rs2745572[A]) that was significantly associated with primary open-angle glaucoma (POAG) and vertical cup-to-disk ratio, an important endophenotype for glaucoma [18, 19]. However, the regulatory elements of FOXC1 are still unknown.

Zebrafish has proved to be a robust animal model for the study of genes involved in embryonic development [12, 20]. There are two genes orthologous to human FOXC1 in zebrafish, foxc1a on chromosome 2 and foxc1b on chromosome 20 [21]. These two genes show high conservation at the protein level with 66% (Foxc1a) and 55% (Foxc1b) identity to human FOXC1 [21]. Moreover, the human and zebrafish genes are located within blocks of conserved synteny. Both foxc1 genes are expressed in developing zebrafish embryos in overlapping but distinct patterns [22]. Studies of knockout lines for foxc1a and foxc1b identified embryonic lethality, altered somitogenesis, cardiac anomalies/heart edema, and facial cartilage defects for the foxc1a knockout (KO), and no visible phenotype for the foxc1b KO [23,24,25,26]. Our group characterized the eye phenotype of foxc1a KO (mw711) and foxc1 double-KO lines and showed that both are similarly affected with major ocular defects overlapping ARS [26].

In order to discover and characterize regulatory elements of FOXC1, we performed various analyses to identify candidate regulatory regions and then used CRISPR-Cas9 technology to delete those regions in zebrafish followed by in vivo evaluations of the resultant lines. In total, five human noncoding regions corresponding to seven regions in zebrafish were examined and found to have variable effects on the expression of foxc1 or surrounding genes.

Results

Identification of candidate regulatory elements in the FOXC1 genomic region

To identify candidate regulatory elements, we performed a multispecies comparison of FOXC1/foxc1 genomic sequences, with a focus on regions conserved between human and zebrafish. The examined area included genomic sequences of human chromosome 6 starting at ~ 1.6-Mb upstream of FOXC1 and ending ~ 1-Mb downstream of this gene, which corresponds to ~ 213 kb and ~ 148 kb upstream and ~ 262 kb and ~ 142 kb downstream of zebrafish foxc1a and foxc1b, respectively. In humans, the studied region encompasses DUSP22, IRF4, EXOC2, HUS1B, FOXQ1, FOXF2, FOXC1, GMDS, 7 pseudogenes, 21 long noncoding RNA genes (18 uncharacterized), 5 long intergenic non-protein coding RNAs genes and 1 microRNA gene. In zebrafish, the chromosome 2 region contains dusp22b, irf4a, foxq1a, foxf2a, and foxc1a but not exoc2, hus1b and gmds, while the chromosome 20 region contains irf4b, exoc2, foxq1b, foxf2b, foxc1b, and gmds but not dusp22 or hus1b; there was no evidence for the presence of pseudogenes, noncoding, or microRNA genes on any of the zebrafish chromosomes.

This analysis identified four conserved regions in humans, one distant upstream and three downstream of FOXC1 (Fig. 1A), that corresponded to five zebrafish regions, two distant upstream of foxc1a or foxc1b and three downstream of foxc1a. The conserved element upstream of FOXC1, named Conserved Element Upstream 1 (CEU1), is located 221 kb upstream of FOXC1 (1.57 kb upstream of FOXF2 and 73 kb downstream of FOXQ1). Sequences with high homology to CEU1 were identified on both zebrafish chromosomes 2 (CEU1a) and 20 (CEU1b), upstream of FOXF2 orthologs foxf2a and foxf2b, respectively (Fig. 1A). The zebrafish CEU1a element is located 519-bp upstream of foxf2a and spans 78-bp with 78% identity with humans. The zebrafish CEU1b element is located 565-bp upstream of foxf2b and spans 160-bp with 69% identity with humans (Table 1; Additional file 1: Fig. S1). For the three conserved elements downstream of human FOXC1, the first one, named Conserved Element Downstream 1 (CED1), is situated in the intergenic region between FOXC1 and GMDS at 2.9 kb from FOXC1 (Fig. 1A), while the other two elements, CED2 and CED3, are located between exons 7 and 8 of GMDS (NM_001500.4) at 194.4 kb and 290 kb from FOXC1, respectively (Fig. 1A). All three downstream elements had homologous sequences on zebrafish chromosome 2 (foxc1a) but not chromosome 20 (foxc1b): in zebrafish, the CED1, CED2 and CED3 regions are positioned 2.5, 109.3 and 151 kb downstream of foxc1a and span 99-bp (76% identity), 67-bp (84% identity) and 140-bp (81% identity), respectively (Table 1; Additional file 1: Fig. S1). Review of the Encyclopedia of DNA Elements (ENCODE) [27] showed that all four human conserved regions, CEU1 and CED1-3, co-localize with predicted cis-regulatory elements (Table 1 and Additional file 1: Fig. S1). Also, CED1 overlaps with a previously predicted (using EnhancerFinder) enhancer region in the FOXC1-GMDS block [28].

Fig. 1
figure 1

Schematic of human FOXC1, zebrafish foxc1a and foxc1b loci, and developed zebrafish lines. A Schematic drawing of human chromosome 6 aligned with zebrafish chromosomes 2 and 20 showing the positions of the identified conserved elements (filled orange boxes labeled at the top). Exons are indicated with labeled black boxes, while intergenic and intronic regions are shown with dotted lines. Position of rs2745572[A], an SNP associated with primary open-angle glaucoma, is shown in blue on human chromosome 6 and corresponding intergenic regions on zebrafish chromosomes 2 and 20 are marked with blue rectangles. B Schematic drawing showing generated lines and corresponding genomic deletions (sequence gaps are indicated)

Full size image
Table 1 Summary of conserved elements in human FOXC1 and zebrafish foxc1a/b genomic regions
Full size table

Literature review additionally revealed a SNP, rs2745572[A], associated with primary open-angle glaucoma (POAG) and increased vertical cup-to-disk ratio (a glaucoma endophenotype) [18, 19], located in the intergenic region between FOXC1 and FOXF2 at ~ 61.7 kb upstream of FOXC1 and ~ 152.6 kb downstream of FOXF2. BLAST-based comparisons of the human intergenic sequence to corresponding genomic regions between zebrafish foxc1a/b and foxf2a/b failed to identify any conserved elements in these regions.

However, the reported association possibly indicates the presence of a regulatory sequence(s) in the intergenic region between FOXC1 and FOXF2.

Generation of deletion lines in zebrafish

A CRISPR-Cas9 genome editing system was utilized to generate various zebrafish lines carrying deletions of the identified candidate regulatory regions. For the distant upstream conserved element of FOXC1/foxc1a/foxc1b, two different deletion lines were generated: the first line, named foxc1a∆CEU1a, carries a 1226-bp deletion encompassing CEU1a located upstream of foxf2a, while the second line, foxc1b∆CEU1b, carries a 779-bp deletion including CEU1b located upstream of foxf2b (Fig. 1B; Additional file 1: Table S1, Fig. S2A, F). In order to test upstream sequences that are positionally homologous to the human region containing POAG-associated SNP rs2745572 (but lacking any clear sequence conservation), additional lines foxc1a∆UPa and foxc1b∆UPb, were generated carrying deletions of the entire intergenic regions between foxf2a and foxc1a (20,671-bp) or foxf2b and foxc1b (7726-bp), respectively, excluding the ~ 3 kb fragments immediately upstream of foxc1a or b, to ensure retainment of all promoter sequences (Fig. 1B; Additional file 1: Table S1, Fig. S2B, G).

For the conserved elements downstream of FOXC1/foxc1a, three different deletion lines were generated: the first line, foxc1a∆CED1, carries a 69,072 kb deletion containing CED1 only; the second line, foxc1a∆CED2−3, carries an 82,715 kb deletion including CED2 and CED3; while the third line, foxc1a∆CED1−3, carries a 151,989-bp deletion encompassing all three conserved elements (CED1, CED2 and CED3) downstream of foxc1a; (Fig. 1B; Additional file 1: Table S1, Fig. S2C–E).

Analysis of lines carrying deletions of upstream regions of foxc1a/b

Careful examination of foxc1a∆CEU1a and foxc1b∆CEU1b lines carrying deletions of distant upstream regions did not identify any consistent phenotype in homozygous or heterozygous embryos or adults. To further evaluate their possible regulatory function, qRT-PCR analysis of foxc1a, foxc1b, foxf2a/b and foxq1a/b transcripts in wild-type and mutant embryos was performed. This analysis identified various mild effects on foxc1a and foxc1b expression (Fig. 2A, B), while significant changes in expression levels of foxf2a/b and foxq1a/b were observed in corresponding homozygous embryos (Fig. 2C–F). These results indicate that the distant upstream conserved elements are primarily involved in the regulation of the nearby foxf2 and foxq1 genes rather than foxc1.

Fig. 2
figure 2

Changes in gene expression in mutants carrying deletions of upstream regions. A–F qRT-PCR relative expression of foxc1a (A) and foxc1b (B), foxf2a (C) and foxq1a (D) and foxf2b (E) and foxq1b (F) in 1-, 2-, 3- and 4-dpf wild-type, foxc1a∆CEU1a and/or foxc1b∆CEU1b homozygous zebrafish embryos (whole bodies). G, H qRT-PCR relative expression of foxc1a (G) and foxc1b (H) in 1-, 2-, 3- and 4-dpf wild-type, foxc1a∆UPa and foxc1b∆UPb homozygous embryos. β-actin (actb1) was used as the reference transcript in all experiments. *: p < 0.05; **: p < 0.01; ***: p < 0.001; ****: p < 0.0001

Full size image

Examinations of the foxc1a∆UPa and foxc1b∆UPb lines carrying deletions of the intergenic regions between the foxc1a and foxf2a or foxc1b and foxf2b (excluding the ~ 3 kb fragments immediately upstream of foxc1a/b) identified no visible phenotype in either homozygous or heterozygous animals. We next examined the expression levels of foxc1a and foxc1b in wild-type and mutant embryos at different developmental stages. A mild but statistically significant decrease in foxc1a or foxc1b levels at later stages of development was observed in the respective lines: to 0.92 and 0.8 at 3- and 4-dpf (days post-fertilization) for foxc1a in foxc1a∆UPa embryos (Fig. 2G) and to 0.88 at 4-dpf for foxc1b in foxc1b∆UPb mutants (Fig. 2H). These data suggest that each intergenic region may contain elements contributing to the proper expression of foxc1a and foxc1b at later stages of development.

Analysis of lines carrying deletions of downstream conserved regions of foxc1a

Examination of foxc1a∆CED1−3 homozygous embryos identified a completely penetrant ocular phenotype, while heterozygous animals appeared normal. The affected embryos showed an enlargement of the anterior chamber of the eye that was visible at 3-dpf and became more pronounced at later stages (Fig. 3D–F, M). Homozygous embryos also demonstrated reduced blood flow in the caudal region (Additional file 2: Video SV1) and juvenile lethality (100% of fish die by ~ 30-dpf). The fish that survived to 30-dpf showed general edema and variable anterior chamber defects (Fig. 4E–H). The majority of animals displayed an enlargement of the anterior chamber that was most pronounced in the nasal–dorsal part of the eye (Fig. 4G, H), along with deformed and irregularly shaped eyes (Additional file 1: Fig. S3).

Fig. 3
figure 3

Phenotypic analysis of zebrafish mutants carrying deletions of downstream elements. A–I Dorsal images of the head region of 3-, 4- and 6-dpf wild-type (WT) (A–C), foxc1a∆CED1−3 (D–F) and foxc1a∆CED2−3 (G–I) homozygous zebrafish embryos. Both mutant lines showed the enlargement of the anterior chamber of the eye that was first noticeable at 3-dpf and became more pronounced by 6-dpf (black arrows in D–I). J–L Lateral and dorsal views of the 3-dpf wild-type (J), foxc1a∆CED1−3 (K) and foxc1a∆CED2−3 (L) homozygous zebrafish embryos. Please note no obvious morphological changes (aside from ocular defects presented in A–I) in mutant embryos. M Comparison of the anterior chamber area in wild-type and mutant embryos at 3-, 4-, and 6-dpf. *: p < 0.05; **: p < 0.01; ***: p < 0.001; ****: p < 0.0001

Full size image
Fig. 4
figure 4

Developmental defects in juvenile and adult foxc1a∆CED1−3 and foxc1a∆CED2−3 mutants. A–L Lateral and dorsal whole body and head images of 30-dpf wild-type (A–D), foxc1a∆CED1−3 (E–H) and foxc1a∆CED2−3 (I–L) homozygous zebrafish embryos. Please note general swelling, including abdominal and heart edema (black arrowheads), in mutant embryos (E, I) as well as bilateral/unilateral enlargement of the anterior chamber of the eye, particularly in the dorso-nasal region (orange arrowheads in G–H, and K). M–P’ Ocular images of adult wild-type (M–P) and foxc1a∆CED2−3 mutants (M’–P’) showing bulging in the nasal part of the anterior chamber of the eye (orange arrowheads in N’–P’). Panels N, P, N’ and P’ show the regions outlined by white boxes in panels M, O, M’ and O’ at a higher magnification

Full size image

The foxc1a∆CED2−3 homozygous embryos carrying a deletion encompassing CED2 and CED3 but not CED1 demonstrated a similar fully penetrant embryonic phenotype: an enlargement of the anterior chamber of the eye at 3-dpf that became more pronounced at 4-dpf (Fig. 3G–I, M) and mildly reduced blood flow in the caudal region (Additional file 3: Video SV2); heterozygous embryos did not show any visible phenotype. However, most embryos recovered at later stages and survived to adulthood, thus showing a milder overall phenotype in comparison with the foxc1a∆CED1−3 line. At 30-dpf, only a small percentage (7%) of the juvenile animals displayed an enlargement of the anterior chamber, general edema and lethality, similar to foxc1a∆CED1−3 fish (Fig. 4I–L), while the majority of homozygotes (93%) appeared normal. However, examination of the surviving foxc1a∆CED2−3 homozygotes at later stages (7-month post-fertilization adults) identified visible ocular defects in 18.75% (3 out of 16) (Fig. 4M’–P’).

With respect to the foxc1a∆CED1 line, neither homozygous nor heterozygous embryos showed any visible phenotype and all embryos survived to adulthood, were fertile and bred normally.

In order to determine the specificity of the observed phenotypes to foxc1a, we generated compound heterozygous zebrafish carrying the foxc1a knockout allele mw711 ([26]; from here on referred to as foxc1aKO) with either ∆CED1-3 or ∆CED2-3 in trans (foxc1aKO/ΔCED1−3 or foxc1aKO/ΔCED2−3). These fish demonstrated similar enlargements of the anterior chamber as seen in the homozygous lines described above (Additional file 1: Fig. S4), thus supporting a role for the deleted regions in normal foxc1a function. Compound heterozygous animals carrying the mw711 and ∆CED1 alleles, foxc1aKO/ΔCED1, showed no visible phenotype.

Histological and marker analysis of affected embryos from foxc1a ∆CED1−3 and foxc1a ∆CED2−3 lines

To further evaluate the developing eye, hematoxylin–eosin (H&E)-stained histological head sections of foxc1a∆CED1−3 and foxc1a∆CED2−3 homozygous embryos at 6-dpf were examined. Consistent with the gross morphological observations, an enlargement of the anterior chamber of the eye was noticeable in 6-dpf homozygous embryos from both lines, with foxc1a∆CED1−3 embryos showing a more severe phenotype (Fig. 5A, A’; Additional file 1: Fig. S5A’). In addition to this, variable hypoplasia of the dorsal irido-corneal angle was observed, which again was more pronounced in foxc1a∆CED1−3 embryos. No visible defects in the retina or lens were detected in either line.

Fig. 5
figure 5

Histological analysis of ocular anomalies in foxc1a∆CED1−3 homozygous embryos. A, A’ H&E-stained transverse sections of the eye of 6-dpf wild-type and mutant embryos. B–C’ H&E-stained transverse sections through central (B and B’) and nasal (C and C’) eye regions of 30-dpf wild-type and mutant fish. Mutants show a marked enlargement of the anterior chamber and abnormal development of both dorsal and ventral annular ligaments (orange arrows, A’–C’); dislocation of lenses toward the back of the eye (B’); and deformed/misplaced scleral ossicles at the dorsal irido-corneal angle (black arrowhead, B’). D–E’ 20× magnifications of the dorsal irido-corneal angle (D and D’) and cornea (E and E’) showing details of the hypoplastic dorsal annular ligament (orange arrow, D’), and thin cornea at 30-dpf (orange arrowhead, E’). Transverse (F and F’) and coronal (G and G’) 40× magnifications of the ventral irido-corneal angle and canalicular network showing an apparent absence of the glycoprotein aggregates in the ventral annular ligament (orange arrow in F’), narrowing of the irido-corneal canal (blue arrow, F’), hyperplasia of the ventral iris stroma in this region (orange asterisks in F’ and G’) and thin cornea at 30-dpf (orange arrowhead in F’). H–K’ immunostaining of cornea sections of 30-dpf wild-type and mutant fish with anti-CKS (red) and anti-cdh2 (green), showing a thinner corneal stroma (I’) and a disorganized corneal epithelium (J’). AL, annular ligament; C, cornea; CaN, canalicular network; CC, ciliary canal; ce, corneal epithelium; cn, corneal endothelium; cs, corneal stroma; I, iris; IC, irido-corneal canal; Le, lens; ON, optic nerve; Os, scleral ossicle R, retina

Full size image

Examination of histological transverse and coronal head sections of 30-dpf foxc1a∆CED1−3 mutants identified a considerable enlargement of the anterior chamber with noticeable bulging in the dorsal–nasal area of the cornea being most frequently present (Fig. 5B’–C’; Additional file 4: Video SV3). Additional anomalies included a posteriorly displaced lens with highly reduced/absent vitreous space (Fig. 5B’); absent (4/8) or highly hypoplastic (4/8) dorsal annular ligament (Fig. 5B’, C’) and displaced or deformed scleral ossicles at the dorsal, nasal and temporal irido-corneal angle (Fig. 5B’–D’; Additional file 1: Fig. S5B’, C’; Additional file 4: Video SV3); a thinner cornea (Fig. 5E’); and notable defects in the ventral irido-corneal angle (Fig. 5F’, G’).

Normally, the annular ligament has a fibrous and porous meshwork appearance in aldehyde-fixed preparations (Fig. 5F) and the ‘pores’ were found to be non-membrane-bound aggregates of glycoprotein [29]. In foxc1a∆CED1−3 homozygous embryos, at the ventral annular ligament, a sharp reduction in ‘pores’ was observed, with an overall denser appearance of this tissue (Fig. 5F’). Most importantly, developmental defects in the aqueous humor drainage structure were detected in mutants (Fig. 5F’, G’). Drainage of aqueous humor in zebrafish occurs in a morphologically specialized structure called the canalicular network localized in the ventral irido-corneal angle. In wild-type adult fish, this structure consists of the irido-corneal canal and the ciliary canal that connect the anterior and posterior chambers, respectively, with the angular aqueous plexus where the aqueous humor is returned to the bloodstream [30] (Fig. 5F). The canalicular network is positioned between the iris and the annular ligament and is comprised of endothelial-lined openings of loosely organized juxtacanalicular connective cells; it is functionally analogous to the aqueous humor outflow system in mammals [30] (Fig. 5F). In foxc1a∆CED1−3 mutants, an underdeveloped canalicular network was observed (Fig. 5F’) with hyperplasia of the iris stroma detected in this region in some fish (3/7) (Fig. 5F’, G’). Therefore, the observed enlargement of the anterior chamber could be caused by an increase in the intraocular pressure due to impaired drainage of the aqueous humor.

To further study the cornea defects, 30-dpf foxc1a∆CED1−3 mutant sections were stained for N-cadherin (cdh2) that marks corneal epithelium and endothelium, and corneal keratan sulfate proteoglycan (CKS), which is a marker for corneal stroma [31]. Cornea epithelium at 30-dpf is composed of several layers of epithelial cells, an acellular and well-ordered stroma, and an endothelial monolayer (Fig. 5H–J). Mutant corneas showed thinner stroma and disorganized epithelial layer (Fig. 5I’, J’).

Since foxc1aKO embryos showed defects in the hyaloid vasculature [26], we studied this structure in the foxc1a∆CED1−3 homozygotes in a Tg(fli1a:EGFP) background. Tg(fli1a:EGFP) expresses eGFP under the control of the fli1a promoter, which is an early endothelial marker that allows monitoring of blood vessel formation [32]. During normal eye development at 3-dpf the superficial choroidal vasculature comprises three radial vessels, nasal (NRV), dorsal (DRV) and ventral (VRV), that project from the periphery of the optic cup toward the lens and are connected by a ring-shaped vessel named the superficial annular vessel (SAV); the same vessels continue to develop at 5-dpf and 8-dpf (Fig. 6A–C) [33]. Examination of 3- and 5-dpf foxc1a∆CED1−3 homozygotes revealed an enlarged SAV and disorganized NRV including irregular shape and/or bifurcation (Fig. 6A’ and B’). At 8-dpf, mutant eyes show a more diffuse fli1a signal exposing abnormal development of the superficial choroidal vasculature, a more pronounced enlargement and irregularity (particularly in the dorsal–nasal part) of the SAV (Fig. 6C’), and a misplaced (in all) and erroneously divided into daughter branches (in about half) NRV (9/20). The DRV can be detected in all mutants; however, it had not grown enough to connect to the SAV in most (16/20). Most remarkably, the VRV was not detectable in the ventral part of the eye of all but one mutant (19/20) (Fig. 6C’).

Fig. 6
figure 6

foxc1a∆CED1−3 mutant embryos display defects in the developing superficial choroidal vasculature. A–C’ Three-dimensional maximum intensity projection images of the ocular vasculature in live control (A–C) or foxc1a∆CED1−3 homozygous (A’–C’) embryos carrying fli1a:EGFP transgene at 3-, 5- and 8-dpf. Mutant embryos show abnormal development of the dorsal and nasal radial vessels (orange arrowheads in A’–C’), enlarged and deformed superficial annular vessel (orange arrows) and a highly disorganized vasculogenesis in the ventral part of the eye with no visible ventral radial vessel at 5- and 8-dpf (orange asterisks). DRV (dorsal radial), NRV (nasal radial), SAV (superficial annular), and VRV (ventral radial) blood vessels are indicated

Full size image

Excavation of the optic nerve head is a recognized clinical feature of glaucoma indicating likely death of retinal ganglion cells. Considering this, we examined the appearance of retinal ganglion cells in 30-dpf juvenile foxc1a∆CED1−3 homozygous mutants in a transgenic Tg(gap43:eGFP) background. The Tg(gap43:eGFP) line expresses eGFP under the promoter of gap43, an axon growth-associated gene, which is expressed during developmental or regenerative axon growth [34] and is a useful tool to monitor optic nerve damage and regeneration [35]. Since retinal axons are still growing at 30-dpf, a similar signal was observed in foxc1a∆CED1−3 and control siblings (Additional file 1: Fig. S6A–F). However, several structural differences in foxc1a∆CED1−3 mutant eyes in comparison with their normal siblings were observed: The distribution of axons was irregular, exposing thinner axon bundles and reduced branching in mutant eyes (Additional file 1: Fig. S6C–F); additionally, the head of the optic nerve appeared to be enlarged in many (5/7) and irregularly shaped (elongated instead of circular) in some (2/7) (Additional file 1: Fig. S6C, E).

Finally, since foxc1a is expressed in neural-crest (NC) derived periocular mesenchyme, we examined this cell population in embryos carrying the foxc1a∆CED1−3 allele in a Tg(foxd3:GFP)zf15 transgenic background (expressing GFP in migrating NC cells [22]). We observed no visible difference in intensity or distribution of GFP-positive cells between control (wild-type or heterozygous) and foxc1a∆CED1−3 homozygous embryos (Additional file 1: Fig. S7), suggesting no defects in migration of NC cells to the periocular mesenchyme in this mutant.

Analysis of gene expression in lines carrying deletions of downstream elements, foxc1a∆CED1−3 , foxc1a ∆CED2−3 and foxc1a ∆CED1

To determine the effect of the downstream deletions on the expression of foxc1a and foxc1b, qRT-PCR experiments were performed using wild-type and homozygous mutant whole embryo (1-, 2-, 3-, 4- and 6-dpf) and ocular (1-, 2- and 3-dpf) RNA samples. This analysis identified a significant decrease in the foxc1a transcript level at early stages in both whole embryos and eyes in foxc1a∆CED1−3 and foxc1a∆CED2−3 mutants (Fig. 7A, C). foxc1a∆CED1−3 homozygotes demonstrated a downregulation of foxc1a for both whole embryos (ranging from 0.44- to 0.61-fold) and mutant eyes (ranging from 0.52- to 0.59-fold) at 1–2-dpf and no difference at 3-dpf (Fig. 7A, C). Expression levels in foxc1a∆CED2−3 homozygotes were decreased at most stages in both embryos (ranging from 0.48- to 0.82-fold across all stages with the exception of 3-dpf) and mutant eyes (0.43- to 0.6-fold across all stages) (Fig. 7A, C). On the other hand, foxc1a∆CED1 homozygotes demonstrated a significant upregulation of foxc1a at all stages in both whole embryos (ranging from 1.67- to 2.64-fold) and eyes (1.45- to 2.09-fold) (Fig. 7A, C). These results strongly suggest that the downstream regions of foxc1a are involved in its transcriptional regulation.

Fig. 7
figure 7

Expression studies in zebrafish mutants carrying deletions of downstream regions of foxc1a. qRT-PCR relative expression of foxc1a (A, C), foxc1b (B, D) transcripts in 1–6-dpf whole bodies (A, B) and 1–3-dpf dissected eyes (C, D) of wild-type and mutant embryos; *: p < 0.05; **: p < 0.01; ***: p < 0.001; ****: p < 0.0001. E–M’ RNAscope in situ hybridization analysis of foxc1a (yellow) and foxc1b (magenta) expression in 1-, 2- and 3-dpf wild-type and foxc1a∆CED1−3 mutant embryos. Mutant embryos showed a visible reduction in ocular foxc1a expression at 1- and 2-dpf (orange arrows in E’ and H’) as well as in the branchial arches at 2-dpf (orange arrowhead in H’), while expression of foxc1b appeared normal (white asterisk; I’)

Full size image

In addition to foxc1a, expression of foxc1b was also affected in foxc1a∆CED1−3, foxc1a∆CED2−3, and foxc1a∆CED1 mutants. The foxc1a∆CED1−3 embryos showed an increase in the foxc1b level at all stages in whole embryos and eyes (Fig. 7B, D). The foxc1a∆CED2−3 embryos similarly demonstrated upregulation in foxc1b expression in whole embryos and eyes at some stages (Fig. 7B, D). The foxc1a∆CED1 embryos displayed no difference in foxc1b expression in mutant eyes at any stage, while mild effects (upregulation or downregulation) were detected in whole embryos at some stages (Fig. 7B, D).

Whole-mount in situ hybridization for foxc1a and foxc1b as well as antibody staining for Foxc1 in wild-type and foxc1a∆CED1−3 mutants at 1-, 2- and 3-dpf revealed weaker foxc1a/Foxc1 staining in the periocular mesenchyme and the branchial arches at 1- and 2-dpf (Fig. 7E’, H’; Additional file 1: Fig. S8A’–C’), consistent with the qPCR analysis.

Discussion

Despite the wide use of exome and genome sequencing, a large number of individuals with inherited disorders lack a genetic diagnosis [36, 37]. While there are still novel factors to be discovered, various studies highlight the importance of noncoding regions in human disease [17, 38,39,40]. One important class of functional sequences located in noncoding regions are regulatory elements that can be predicted based on evolutionary conservation, open chromatin state, the three-dimensional structure of chromatin, and other approaches [20, 27, 41]. Variants in these regions affect gene expression through alteration/removal of binding sites for transcription factors and/or disturbing the three-dimensional structure of chromatin [42, 43]. Consistent with this, many studies have shown that mutations in cis-regulatory elements of disease-associated genes can cause similar phenotypes to the ones reported for coding region variants [4, 13, 14, 16].

FOXC1 encodes a forkhead box transcription factor involved in vertebrate embryonic development. FOXC1 is located at 6p25.3 in a conserved cluster of FOX genes (FOXQ1, FOXF2 and FOXC1). Mutations in FOXC1 are responsible for several developmental disorders of the anterior segment of the eye [3,4,5, 7, 8, 10], while no human disease phenotypes are currently identified for FOXF2 or FOXQ1. Copy number variants represent an important class of FOXC1 pathogenic alleles and include both deletions and duplications of this gene [3, 4, 8, 44]. This highlights the importance of a precise dosage of FOXC1 for proper development. Accordingly, disruption of FOXC1 regulatory elements, or its upstream factors, is likely to result in disease; however, the mechanisms of FOXC1 regulation are currently unknown. In this manuscript, we present the first data on cis-regulatory elements of FOXC1 that have been studied in vivo.

We identified five conserved elements in the zebrafish foxc1a and foxc1b genomic environment corresponding to four human noncoding FOXC1 regions. Two of these elements are situated upstream of foxc1a or foxc1b and relate to the same remote upstream region of FOXC1, while the other three are located downstream of foxc1a/FOXC1. The conserved elements ranged from 67 to 160 bp in length and showed 76–84% identity between zebrafish and human. In terms of their position within the conserved block, the distant upstream elements reside 5’ of the FOXF2/FOXC1 or foxf2/foxc1 clusters, 27.9-/12.94 kb from foxc1a/b and 221.96 kb from FOXC1. The downstream elements are located at 2.5/2.9 kb to 151/290 kb downstream of foxc1a/FOXC1. In humans, one downstream element is situated in the intergenic region between FOXC1 and GMDS while the other two are located within an intronic region of GMDS. In zebrafish, all three downstream elements reside in the intergenic region between foxc1a and the neighboring gene mylk4a, since gmds is bordering foxc1b but not the foxc1a ortholog; interestingly, despite this downstream synteny for foxc1b, no conserved elements were identified in this region. foxc1a represents the main zebrafish ortholog of FOXC1 in terms of functional significance: foxc1aKO fish display strong developmental defects that recapitulate FOXC1 disease-associated features, while foxc1bKO fish do not show any visible phenotype [26]. The presence of conserved elements downstream of the foxc1a gene despite the lack of syntenic gmds further supports a possible role for these elements in the regulation of foxc1a. This is consistent with evolutionary studies suggesting that when genomic duplications occur, as in zebrafish, coding sequences of the extra bystander gene may be erased, while cis-regulatory modules for developmental genes remain conserved [12, 45].

Deletion of predicted cis-regulatory elements in animal models has become a powerful approach to exploring their role in gene regulation and disease [20, 41, 46, 47]. To investigate the possible function of the identified sequences, we generated a series of zebrafish lines carrying various deletions encompassing the identified candidate elements. The deletion of a 152 kb region comprising all three downstream regions (CED1-3) resulted in developmental defects in the anterior segment of the eye and juvenile lethality. Further dissection of this region revealed that deletion of an 82.7 kb fragment containing two of the three downstream conserved elements (CED2-3) produced a similar but milder/transient phenotype, while removal of only the first downstream element (CED1; 69.1 kb deletion) did not have any noticeable effect on zebrafish development or survival. In terms of expression changes, deletions of CED1-3 or CED2-3 resulted in significant downregulation of the foxc1a transcript in zebrafish embryos and developing eyes while removal of CED1 caused a detectable increase in the level of foxc1a transcript. In contrast, deletions of distant conserved elements upstream of either foxc1a (∆CEU1a) or foxc1b (∆CEU1b) did not produce a visible phenotype; zebrafish lines lacking the CEU1a or CEU1b elements had a minor change in foxc1a/b expression but showed a considerable alteration in the levels of foxf2a/b and foxq1a/b transcripts located in close proximity to those elements. A complementation test confirmed that the ∆CED1-3 and ∆CED2-3 noncoding deletions and foxc1a knockout allele mw711 [26] are allelic to each other and thus all affect foxc1a.

The phenotype observed in foxc1a∆CED1−3 and foxc1a∆CED2−3 mutants is consistent with features reported in foxc1aKO animals, including ocular abnormalities, vascular/blood flow defects, edema and lethality [26] but shows later onset and milder presentation. The later onset/milder phenotype in mutants with regulatory region deletions is likely related to the higher foxc1a level as these animals have at least 40% of normal foxc1a in comparison with its complete absence in foxc1aKO embryos. Similar phenomena were described for other transcription factors with dosage-dependent phenotypes [20, 41, 46, 47].

Drainage of the aqueous humor in humans takes place through a circular structure located in the irido-corneal angle that includes the trabecular meshwork (TM) and Schlemm’s canal [48]. The TM is a porous structure formed by several layers of connective tissue beams and collagenous elastic fibers. The TM represents the main pathway for drainage of aqueous humor out of the eye with a critical role in maintaining normal intraocular pressure; its function is to provide a pressure gradient resistance to the aqueous humor flow to the Schlemm’s canal [48]. Schlemm’s canal is considered a unique blood–lymphatic intermediate-type vessel that is originally formed by endothelial cells from the choroidal vein and acquires lymphatic characteristics later in development [49]. Unlike in humans, drainage of aqueous humor in zebrafish occurs in the ventral part of the irido-corneal angle only, through the canalicular network and the angular aqueous plexus (homologous structures to the human trabecular meshwork and Schlemm’s canal, respectively) [30]. The foxc1a∆CED1−3 mutants demonstrated defects in the ventral canalicular network, possibly related to the noted abnormalities in the development of the ventral superficial choroidal vasculature of the eye. Thus, the observed enlargement of the anterior chamber (indicating an increase in intraocular pressure) is likely due to an impairment of aqueous humor drainage through the malformed outflow structures in affected animals. High intraocular pressure accompanied by an enlargement of the ocular globe (known as buphthalmos) is a common manifestation of congenital glaucoma [50], one of the developmental phenotypes associated with mutations in FOXC1. The mechanism of this disorder is not fully known; however, human studies identified developmental defects in the drainage structures of patients with congenital glaucoma caused by CYP1B1 mutations [51]. No similar reports are available for FOXC1, but studies in mice demonstrated that Foxc1+/- [52] heterozygotes as well as animals homozygous for mutations in congenital glaucoma genes Cyp1b1 [53] and Angpt [54] have abnormally formed trabecular meshwork and/or Schlemm’s canal. Thus, the generated foxc1a mutants will serve as powerful models for studies of human developmental glaucoma.

Expression levels of the foxc1a transcript were altered in all downstream deletion lines implying a regulatory role for those regions. Although the deleted regions comprised large noncoding segments, it is plausible to assume that the identified conserved elements within the region are making the most important contribution to the observed regulatory effect. However, other factors, such as the presence of additional, yet-to-be-identified, regulatory DNA elements and/or noncoding RNA in this region as well as possible positional effects due to the changes in the architecture of topological associating domains (TADs) often associated with larger genomic deletions [55, 56], cannot be ruled out. Further studies including deletions of each downstream element separately could provide further information about their distinct functions.

Interestingly, deletions of foxf2a-foxc1a or foxf2b-foxc1b intergenic fragments positionally orthologous to the region containing the POAG-associated rs2745572[A] SNP [18] but lacking sequence conservation resulted in downregulation of foxc1a or foxc1b expression, respectively, indicating a possible role in transcriptional regulation. This is supported by the growing evidence that cis-regulatory regions may diverge in their primary sequences in different species while maintaining their functional (regulatory) role [57]. Another possibility is that these intergenic deletions affected distances between other regulatory elements and/or overall chromatin structure which was followed by a negative effect on foxc1a and foxc1b expression; however, the specific nature of the observed changes (limited to late developmental stages in both zebrafish lines) is consistent with the likely presence of distinct cis-regulatory element(s) in these regions. Further dissection of these regions in both human and zebrafish may provide additional insight into the location of regulatory sequences and their roles in FOXC1/foxc1 expression and POAG.

In summary, this study identified several regulatory regions that are critical for the normal expression of FOXC1/foxc1 in vertebrates. Specifically, we show that foxc1a and thus likely FOXC1 embryonic expression is governed by conserved elements located downstream of the gene and that deletions of these elements result in a range of phenotypes with variable severity in zebrafish. Further studies of these regions in human patients are likely to explain additional cases of Axenfeld–Rieger syndrome, aniridia, Peters anomaly, and glaucoma, and may possibly contribute to the extreme variability in phenotypes caused by FOXC1 heterozygous variants.

Materials and methods

Analysis of sequence conservation at the nucleotide level

Visual analysis of the Vertebrate Multiz Alignment & Conservation (100 Species) track at UCSC Genome Browser (http://genome.ucsc.edu) included a region from the start of human chromosome 6 (~ 1.6-Mb upstream FOXC1) until the start of MYLK (~ 1-Mb downstream FOXC1). Regions of high conservation from human to zebrafish (excluding coding regions) were selected and alignments were verified. Additionally, BLAST comparisons of FOXC1/foxc1a/foxc1b genomic regions were carried out manually: relevant human (hg38) or zebrafish (DanRer11) genome sequences were aligned with the RefSeq zebrafish or human genome, correspondingly, using BLASTN and the ‘somewhat similar’ alignment option (coding regions, UTRs and immediate promoters were disregarded).

Animal husbandry

Zebrafish (Danio Rerio) were raised and maintained under standard conditions as previously described [31]. The Tg(fli1a:eGFP), Tg(gap43:eGFP) and Tg(foxd3:GFP) lines were used to monitor blood vessel [32], optic nerve development [34] and neural-crest populations [58]. Developmental stages were determined by previously described morphological criteria [59]. All experiments were conducted in accordance with the guidelines established by the Institutional Animal Care and Use Committee at the Medical College of Wisconsin.

Generation of genomic deletions in zebrafish

Integrated DNA Technologies (IDT, Coralville, IA) custom Alt-R CRISPR-Cas9 guide RNA tool (https://us.idtdna.com/site/order/designtool/index/CRISPR_CUSTOM) was used to design sgRNAs. Two guides were designed for each deletion, one at each flank (Additional file 1: Table S2). Trans-activating CRISPR RNA (tracrRNA), Cas9 and sgRNAs were purchased from IDT. Each sgRNAs (7.5 μM) and tracrRNA (7.5 μM) were mixed, incubated for 5 min at 95ºC and cooled at room temperature. Cas9 protein (Alt-R® CRISPR-Cas9, IDT) and the pair of tracrRNA/sgRNA complexes were mixed to a final concentration of 0.5 μg/μL for Cas-9 and 1.5 μM for each complex and incubated for 10 min at 37 °C. Single-cell-stage embryos were microinjected with 9.2nL of Cas9/tracrRNA/sgRNA complex and 0.05% Phenol red (Sigma, St. Louis, MO) using the Nanoject II Injector (Drummond Scientific, Broomall, PA). Mosaic breeders were identified by analysis of their offspring via PCR amplification and sequencing with specific primers to identify mutant and wild-type alleles (Additional file 1: Table S3). Founder fish carrying deletions (Additional file 1: Table S1) were selected for further analysis and corresponding lines were established.

Morphologic analysis of embryos and adults

Zeiss SteREO Discovery V12 microscope (Carl Zeiss, Thornwood, NY) with either a 1.0X stereo objective or a 10X compound objective and Nikon SMZ-1500 with a 1.0X stereo objective were used for gross morphological observations and imaging. Fluorescent maximum intensity projection of z-stack images of transgenic line embryos was obtained using an AxioImager.Z1 microscope with an ApoTome attachment, an AxioCam 503 mono camera and ZEN pro software (Zeiss). Anterior chamber surfaces were measured using dorsal images and ImageJ 1.52 k [60]. Student’s t test was used to determine statistical significance.

Histological studies

Adult fish and embryos were immersed overnight in modified Davidson’s fixative (30% of a 37% solution of formaldehyde, 15% ethanol, 5% glacial acetic acid, and 50% distilled H2O) [61] and then transferred to 70% ethanol. Fixed samples were submitted to the Children’s Research Institute Histology Core at the Medical College of Wisconsin for paraffin sectioning and hematoxylin–eosin staining per standard protocols. NanoZoomer digital slide scanner was used to image the slides and NDP.view2 viewing software was used to visualize the images (Hamamatsu, Hamamatsu City, Japan).

RNAscope in situ hybridization and immunofluorescence

Whole embryos were fixed overnight in 4% paraformaldehyde and then transferred to 100% methanol. Embryos were hybridized with RNAscope probes for foxc1a (499611-C2), and foxc1b (584981-C3) (Advanced Cell Diagnostics, Newark, CA) using manufacturer protocols with minor modifications [31].

For immunohistochemistry, whole-mount embryos or paraffin sections were stained with DAPI (62247; Thermo Fisher, Waltham, MA) and various antibodies including human anti-FOXC1 (8758; Cell Signaling, Danvers, Ma), anti-cdh2 (GTX125962, GeneTex, Irvine, CA) and anti-CKS (MAB2022, Millipore, Burlington, MA) primary antibody, as well as donkey anti-rabbit Alexa Fluor 488 (A21206, Thermo Fisher) and donkey anti-mouse Alexa Fluor 568 (A10037, Thermo Fisher) secondary antibody, as previously described [31].

Quantitative RT-PCR transcript level analysis of wild-type and mutant

RNA extraction of whole embryos or dissected eyes was performed using Direct-zol RNA MiniPrep (Zymo Research, Irvine, CA); all samples were treated with DNase I (Invitrogen) prior to cDNA synthesis.

cDNA was synthesized using SuperScript III reverse transcriptase (Thermo Fisher). CFX384 Touch Real-Time PCR Detection Systems (BioRad, Hercules, CA), SYBR Green PCR Master Mix (Applied Biosystems) and transcript-specific primers (Additional file 1: Table S4) were used to analyze selected genes by real-time qPCR. β-actin (actb1) was used as the reference gene for the relative quantification of expression levels. All samples were run in triplicate to obtain average Cq values. Technical replicates that fell multiple standard deviations from the average were considered outliers and, in agreement with standard practice, removed from the analysis. Total fold changes and standard deviations were calculated as the average of three independent biological repeats via the 2−ΔΔCt method [62]. Student’s t test was used to determine statistical significance.