A Zebrafish Model of Axenfeld-Rieger Syndrome Reveals That pitx2 Regulation by Retinoic Acid Is Essential for Ocular and Craniofacial Development

Abstract

Purpose.: The homeobox transcription factor PITX2 is a known regulator of mammalian ocular development, and human PITX2 mutations are associated with Axenfeld-Rieger syndrome (ARS). However, the treatment of patients with ARS remains mostly supportive and palliative.

Methods.: The authors used molecular genetic, pharmacologic, and embryologic techniques to study the biology of ARS in a zebrafish model that uses transgenes to mark neural crest and muscle cells in the head.

Results.: The authors demonstrated in vivo that pitx2 is a key downstream target of retinoic acid (RA) in craniofacial development, and this pathway is required for coordinating neural crest, mesoderm, and ocular development. pitx2a knockdown using morpholino oligonucleotides disrupts jaw and pharyngeal arch formation and recapitulates ocular characteristics of ARS, including corneal and iris stroma maldevelopment. These phenotypes could be rescued with human PITX2A mRNA, demonstrating the specificity of the knockdown and evolutionary conservation of pitx2a function. Expression of the ARS dominant negative human PITX2A K50E allele also caused ARS-like phenotypes. Similarly, inhibition of RA synthesis in the developing eye (genetic or pharmacologic) disrupted craniofacial and ocular development, and human PITX2A mRNA partially rescued these defects.

Conclusions.: RA regulation of pitx2 is essential for coordinating interactions among neural crest, mesoderm, and developing eye. The marked evolutionary conservation of Pitx2 function in eye and craniofacial development makes zebrafish a potentially powerful model of ARS, amenable to in vivo experimentation and development of potential therapies.

Axenfeld-Rieger syndrome (ARS, OMIM 180500) is a spectrum of autosomal dominantly inherited malformations that predominantly affect the eye but are also associated with craniofacial dysmorphism and dental abnormalities, including maxillary hypoplasia and microdontia. ^1,2 Ocular findings in ARS are typically limited to the anterior segment of the eye and include posterior embryotoxon (anteriorization of the angle structure/Schwalbe's line), iris hypoplasia, and corectopia (maldevelopment of the iris with shifting of the pupil). Secondary glaucoma, caused by iris strands that bridge the iridocorneal angle and trabecular meshwork, can be refractory to treatment and lead to significant visual impairment. Specific gene mutations have been identified in approximately 60% of cases; one of the most commonly affected genes is the paired homeobox transcription factor PITX2 (see Refs. 3, 4 for review); OMIM 180500). Multiple types of point and chromosomal mutations, which include both gain-of-function and loss-of-function of the PITX2 gene, have been identified in patients with ARS. ^{5 –9}  

The PITX2 gene codes for as many as four mRNA transcripts (PITX2A-D). ¹⁰ In humans, PITX2A, B, C, and D have been identified, whereas in zebrafish only the pitx2a and pitx2c isoforms have been detected. ¹¹ Pitx2 is expressed in cranial neural crest (CNC) and mesoderm-derived cells, which form tissues that interact with the developing eye. ¹² The Pitx2 knockout mouse dies at embryonic day (E) 14.5 because of heart defects and displays severe ocular defects and loss of extraocular muscle (EOM). ¹³ Conditional knockout studies using mice demonstrated that Pitx2 in CNC is required for the ocular development and optic stalk formation. ¹² In mice, activation of muscle-specific transcription factors in the EOM is dependent on Pitx2, ^{14 –16} and Pitx2 is also required for pharyngeal arch development and subsequent jaw and dental formation. ^{17 –19} Thus, Pitx2 appears to be critical for craniofacial and ocular development, and the regulation of its expression may mediate interactions among the developing eye, CNC, and mesoderm. 

Retinoic acid (RA) is an essential morphogen that regulates craniofacial development in mammals and fish. ^{20 –22} In humans, both the disruption of RA synthesis (e.g., fetal alcohol syndrome) and in utero exposure to excessive retinoids (e.g., through medications such as isotretinoin) can result in craniofacial dysmorphism. ^{23 –26} Studies using mouse models reveal that RA acts as a paracrine signal that is produced by the developing eye and targets the periocular tissues during craniofacial development. Expression of retinaldehyde dehydrogenase (raldh) in the developing eye (retina, lens, and surface ectoderm) is tightly regulated both temporally and spatially. ^27,28 An additional gradient is formed through the expression of cyp26, a cytochrome P450 enzyme, in the central anterior-posterior axis of the retina. This creates an RA-free zone that separates the dorsal from the ventral retina. ^29,30 Thus, before the synthesis of visual pigment chromophore in the developing retina, both RA production and degradation are tightly regulated, supporting an early role for RA in ocular and craniofacial development. Indeed, mouse RA receptors RARβ and RARγ are expressed in the periocular mesenchyme and the developing craniofacial region, ^{31 –33} zebrafish rarα and rarγ are expressed in the periocular mesenchyme, ^34,35 and zebrafish rarα are also expressed in the EOMs. ³⁶ These studies suggest that RA acts as a paracrine signal with targets in periocular tissues. 

Congenital craniofacial abnormalities seen with pitx2 mutations and alterations in RA levels are caused predominantly by the disruption of neural crest development resulting in anomalies in the underlying bone, cartilage, connective tissue, and eyes. Numerous human disorders, including ARS, have been described as “neurocristopathies”. ³⁷ However, there is a wide range in severity of craniofacial defects among these conditions, and the expressivity likely reflects the stage at which the CNC is affected by the underlying genetic or environmental insult. Complex reciprocal interactions between the CNC and the surrounding ectoderm, endoderm, and mesoderm are required for formation of the tissues and the sensory organs of the head. ^{22,38 –44} RA and pitx2 are two critical regulators of this process, and studies in mouse demonstrate that Pitx2 is a downstream target of RA in the periocular mesenchyme. ⁴⁵  

In the present study, an in vivo experimental approach was used to probe the relationship between RA and PITX2 in ocular and craniofacial development. The results show that RA regulates pitx2 expression in the periocular mesenchyme and EOMs, and exogenously added human PITX2 mRNA can compensate for the inhibition of RA synthesis. Furthermore, human PITX2 mutations associated with ARS, ^{46 –49} and alterations in RA levels or pitx2 expression, can recapitulate ocular and craniofacial ARS phenotypes in zebrafish. Importantly, human PITX2 mRNA can rescue the ARS-like phenotypes caused by knockdown of the endogenous zebrafish pitx2 gene. We conclude that the roles of RA and PITX2 in ocular and craniofacial development reveal remarkable functional conservation between vertebrate classes. Such conservation can be used to probe the regulation of morphogenesis using a variety of approaches, including in vivo models, and to screen through phenotypic modifiers to identify potential therapies for ARS and other genetic disorders. 

Materials and Methods

Zebrafish Care, Mutants, and Transgenics

Zebrafish (Danio rerio) were raised in a laboratory breeding colony on a 14-hour light/10-hour dark cycle. Embryos were maintained at 28.5°C and staged as described. ⁵⁰ Embryo age is defined as hours postfertilization (hpf) or days postfertilization. Tg(sox10::EGFP) strain was the generous gift of Thomas Schilling. ^51,52 The Tg(α-actin::EGFP) strain was a generous gift of Simon Hughes. ⁵³ These transgenic strains were crossed into the Roy background, ⁵⁴ which was the generous gift of Rachel Wong, to decrease endogenous fluorescence. Phenylthiourea 0.003%, which inhibits pigmentation, was added to the media of embryos harvested for wholemount in situ hybridization. ⁵⁵ The protocols have met guidelines established by the University of Michigan Committee on the Use and Care of Animals and adhere to the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. 

In Situ Hybridization

In situ hybridization was performed as previously described ⁵⁶ using digoxigenin (DIG) or fluorescein-labeled RNA antisense probes, or both. RNA probes were in vitro transcribed with T7 or T3 polymerase using PCR product with a 3′ sequence of the appropriate promoter. For colorimetric reactions, probes were labeled with alkaline phosphatase-conjugated antibodies directed against DIG or fluorescein and were visualized with 4-nitroblue tetrazolium/5-bromo-4-chloro-3-indolyl phosphate (NBT/BCIP; Roche Molecular Biochemicals, Indianapolis, IN) or Fast Red (Roche). For fluorescence detection, probe haptens were labeled with peroxidase-conjugated antibodies, and the signal was amplified using tyramide Alexa fluorophores (Invitrogen, Carlsbad, CA) or tyramide Cy fluorophores (Perkin-Elmer, Waltham, MA). Embryos were cryoprotected and embedded in optimum cutting temperature compound for sectioning. 

Zebrafish Ocular Histology, Ocular Measurements, and Statistical Analysis

Zebrafish embryos were fixed in 2% paraformaldehyde/1.5% glutaraldehyde overnight at 4°C and then embedded in methyl acrylate. Blocks were sectioned at 5 μm and mounted on slides. Staining with Lee's stain or trichrome-like stain was performed using standard techniques. ^57,58 Permanent coverslips were placed using mounting medium (CytoSeal; Richard-Allan Scientific, Kalamazoo, MI). Sections were imaged with an inverted-light microscope (DM6000B; Leica Microsystems CMS GmbH, Wetzlar, Germany) and a color camera (DFC500; Leica Microsystems CMS GmbH). Images were processed using graphics editing software (Photoshop; Adobe Systems, Inc., San Jose, CA) and the Leica LAS application. 

Ocular measurements were obtained using ImageJ software (developed by Wayne Rasband, National Institutes of Health, Bethesda, MD; available at http://rsb.info.nih.gov/ij/index.html). For corneal thickness, in each eye a line radial to the central axis of the lens was drawn between the surfaces of the corneal epithelia and the lens epithelia and was measured in microns using three consecutive sections that encompassed the lens equator. The dorsal-ventral dimension of the eye was measured as the distance in microns between the dorsal and ventral retinal pigment epithelia at the equator of the eye in three consecutive sections that encompassed the thickest portion of the lens. The anterior-posterior dimension of the eye was measured as a line perpendicular to the central axis of the lens and was drawn between the surface of the corneal epithelium and the retinal pigment epithelium near the optic stalk in three consecutive sections that encompassed the lens equator. The average of the three consecutive sections for each eye was used for statistical analysis. Six to 10 eyes for each treatment or morpholino/mRNA group were measured. 

The Bartlett method was used to determine whether there was equality of SD between the eyes in each group. The Kolmogorov-Smirnov method was used to determine whether the measurements held to a Gaussian distribution. ANOVA with Tukey-Kramer multiple comparisons test was used for comparisons that satisfied the Bartlett and Kolmogorov-Smirnov methods. A Kruskal-Wallis test (nonparametric ANOVA) was used for comparisons that failed the Bartlett or Kolmogorov-Smirnov methods. Statistical significance was considered P < 0.05 and specific values are denoted in Tables 1 and 2. All statistical analyses were performed using a statistics program (InStat, version 3; GraphPad, La Jolla, CA). 

Table 1.

 

Ocular Measurements in 96-hpf Embryos Injected with Oligonucleotide Morpholinos and PITX2A mRNA

	Corneal Thickness (μm)	Dorsal-Ventral Dimension (μm)	Anteroposterior Dimension (μm)
(A) Control MO	5.5 ± 0.5	231.4 ± 5.6	171.4 ± 3.7
(B) pitx2a MO	7.6 ± 0.7 1 *	158.6 ± 12.4 1 *	145.3 ± 7.4 1 *
(C) PITX2A mRNA + control MO	8.2 ± 0.4 1 †	142.2 ± 17.3 1 †	125.7 ± 9.3 1 †
(D) PITX2A mRNA + pitx2a MO	5.4 ± 0.8 2 †, 3 ‡	246.4 ± 7.6 2 †, 3 †	170.9 ± 7.0 2 †, 3 †
(E) K50E PITX2A mRNA + control MO	6.1 ± 0.5 3 ‡	221.2 ± 4.0 3 ‡	165.2 ± 2.9 3 ‡
(F) K50E PITX2A mRNA + pitx2a MO	9.5 ± 2.1 1 †, 2 ‡, 4 †, 5 †	171.1 ± 19.9 4 *	167.5 ± 14.6 3 ‡
(G) raldh2/p53 MO	8.6 ± 1.2 1 †, 4 †, 5 †	142.9 ± 7.4 1 †, 4 †, 5 *	147.9 ± 3.3 1 *, 4 ‡
(H) PITX2A mRNA + raldh2/p53 MO	5.0 ± 1.1 2 *, 3 †, 6 †, 7 †	209.0 ± 10.3	166.4 ± 8.4 3 *
Test used	ANOVA	Kruskal-Wallis	Kruskal-Wallis

 Corneal thickness and eye dimensions (dorsal-ventral axis and anteroposterior axis) measured in microns in 96-hpf embryos injected with control MO (A), pitx2a MO (B), PITX2A mRNA and control MO (C), PITX2A and pitx2a MO (D), K50E PITX2A mRNA + control MO (E), K50E PITX2A mRNA + pitx2a MO (F), raldh2/p53 MO (G), or PITX2A mRNA + raldh2/p53 MO (H).

1  Compared with control MO.

2  Compared with pitx2a MO.

3  Compared with PITX2A mRNA + control MO.

4  Compared with PITX2A mRNA + pitx2a MO.

5  Compared with K50E PITX2A mRNA + control MO.

6  Compared with K50E PITX2A mRNA + pitx2a MO.

7  Compared with raldh2/p53 MO.

*   P < 0.01;

†   P < 0.001;

‡   P < 0.05.

Table 2.

 

Ocular Measurements in 96-hpf Embryos Treated with Exogenous RA and DEAB

	Corneal Thickness (μm)	Dorsal-Ventral Dimension (μm)	Anteroposterior Dimension (μm)
(A) 0.1% DMSO control	3.6 ± 0.8	231.4 ± 10.1	144.8 ± 10.2
(B) 100 nM RA	7.5 ± 1.0 1 *	184.0 ± 7.2 1 *	134.2 ± 7.5
(C) 10 μM DEAB	8.2 ± 0.8 1 *	195.3 ± 13.2 1 *	156.6 ± 8.3 2 ‡
(D) 1 nM RA + 10 μM DEAB	6.7 ± 1.6 1 *	217.6 ± 10.6 1 †, 2 *, 3 *	150.5 ± 5.6
(E) 10 nM RA + 10 μM DEAB	4.1 ± 0.8 2 *, 3 *, 4 *	214.9 ± 3.5 1 †, 2 *, 3 ‡	167.3 ± 8.7 2 *
Test used	ANOVA	ANOVA	Kruskal-Wallis

 Corneal thickness and eye dimensions (dorsal-ventral axis and anteroposterior axis) measured in microns in 96-hpf embryos treated with 0.1% DMSO (A), 100 nM RA at 28 hpf (B), 10 μM DEAB at 24 hpf (C), 1 nM RA at 28 hpf + 10 μM DEAB at 24 hpf (D), or 10nM RA at 28 hpf + 10 μM DEAB at 24 hpf (E).

1  Compared with 0.1% DMSO control.

2  Compared with 100 nM RA.

3  Compared with 10 μM DEAB.

4  Compared with 1 nM RA + 10 μM DEAB.

*   P < 0.001;

†   P < 0.05;

‡   P < 0.01.

Oligonucleotide Morpholinos

Antisense oligonucleotide morpholinos (MOs) were synthesized by Gene Tools, LLC (Cowallis, OR) and reconstituted in deionized water. MO sequences for raldh2, p53, and standard control (globin) were previously published. ^59,60 MO directed against raldh3, pitx2a, and pitx2c were custom designed according to the manufacturer's specification (sequences in Supplementary Table S1), and the same raldh3 (aldh1a3) morpholino sequence was recently published. ⁶¹ The concentration of MO for each gene that yielded consistent and reproducible phenotype was determined. MO sequences that were not lissamine-tagged were coinjected with Texas Red for fluorescent tracking; 1 nL MO (0.1–0.5 mM) was injected into the yolk of one- to two-cell–stage embryos. 

To test the functionality of the pitx2a and pitx2c MO (Supplementary Fig. S6), complementary sequences to the MO (sequences in Supplementary Table S1) were cloned, using the Gateway cloning system, into pdEST R4-R3 VECTOR II (final construct: CMV/SP6-pitx2a/c complementary sequence of MO-EGFP). mRNA of the complementary sequence was transcribed (mMessage mMachine kit; Ambion Biosystems, Austin, TX), resuspended at 100 to 200 ng/μL in nuclease-free water containing 0.1% phenol red, and injected into embryos with or without the pitx2a/c MO at the one-cell stage. Injection of mRNA containing the complementary sequence to the MO resulted in diffuse green fluorescent protein (GFP) expression (Supplementary Figs. S1B, S1G). Coinjection of pitx2a or pitx2c MO with the corresponding mRNA abrogated GFP expression, demonstrating the specificity of each MO (Supplementary Figs. S1D, S1I). To test the functionality of the raldh3 MO, PCR was used to amplify EGFP from pCS2-EGFP with the raldh3 morpholino binding sequence upstream and in frame with the EGFP start codon using the T3 3′ primer and the following 5′ primer containing the SP6 promoter: 5′- GATTTAGGTGACACTATATGCTATGGCACAGAACGGGACTATAGTGAGCAAGGGCGAGG. After 20 PCR cycles with this primer pair, the product was gel-isolated and used as a template for another PCR cycle using the SP6 and T3 primers. mRNA was synthesized (mMessage mMachine kit; Ambion Biosystems). A control EGFP mRNA was transcribed from a PCR primer pair of SP6 and T3 that lacked the morpholino sequence insert. The MO-EGFP mRNA was coinjected into embryos with either the lissamine-tagged raldh3-MO or an untagged control globin-MO (Supplementary Fig. S2). 

Embryos were analyzed with a combi-scope (M205FA; Leica) using bright-field imaging (DFC290; Leica) and fluorescent imaging (ORCA-ER; Hamamatsu, Hamamatsu, Japan) cameras. 

mRNA Synthesis and Microinjection

Human PITX2A mRNA (wild-type, T30P mutant, K50E mutant) was transcribed from the plasmid pCI vector tagged with HA (generous gifts of Michael Walter and Philip Gage). ⁶² Capped mRNA was synthesized (mMessage mMachine kit; Ambion Biosystems), and suspended at 100 to 200 ng/μL in nuclease-free water containing 0.05 mM Texas Red. mRNA (1 nL) was injected into one-cell stage embryos. 

Pharmacologic Treatment of Embryos

All-trans RA (Sigma, St. Louis, MO), diethylbenzaldehyde (DEAB; Sigma), Ro 41–5253 (Enzo Life Sciences, Plymouth Meeting, PA), TTNPB (Ro 13–7410; Enzo Life Sciences), and methoprene acid (GR-106; Enzo Life Sciences) were diluted in dimethyl sulfoxide (DMSO) at 1000× the final concentration. Pharmacologic agents were added to embryo media to their final concentrations (RA, 0.1 nM-1 μM; DEAB, 10–20 μM; Ro 41–52534, 1 μM; TTNPB, 0.01 μM; methoprene acid, 1 μM) at the indicated time. DMSO (0.1%) served as control. Exogenous treatment was initiated at 24 or 28 hpf, as indicated. Embryo media were changed every 24 hours with fresh pharmacologic agent until the embryos were harvested. 

Zebrafish Enucleation and Bead Experiments

AG 1-X8 ion-exchange beads (40–60 μm) were soaked in DMSO or 100 nM RA diluted in DMSO for 1 hour before implantation. ⁶³ Embryos from Tg(sox10::EGFP) and Tg(α-actin::EGFP) zebrafish were dechorionated and enucleated at 16 hpf using sharp tungsten needles and a dissecting microscope. Care was taken not to injure the cranium or the surrounding orbit during the procedure. A DMSO- or RA-soaked bead was placed in the orbit. The embryos were then returned to zebrafish growth medium and allowed to recover at 28.5°C. Embryos were imaged at 72 and 96 hpf and harvested in 4% paraformaldehyde. 

Results

Pitx2a Is Required for Ocular Development in Zebrafish Embryos

Given that haploinsufficiency of PITX2 in humans and mice results in the dysgenesis of the anterior segment of the eye, ^{3,4,49,64 –66} we first determined whether these findings could be recapitulated in zebrafish by injecting MO targeting pitx2a at the one- to two-cell stage (Figs. 1A–E). Indeed at 96 hpf, relative to controls, MO knockdown of pitx2a resulted in maldeveloped eyes that were significantly smaller in the dorsal-ventral (P < 0.01) and anterior-posterior (P < 0.01) axes (Table 1; compare Figs. 1A and 1P). The corneas in pitx2a MO knockdown embryos were significantly thicker (P < 0.01; Table 1, compare Figs. 1G and 1Q). The corneal epithelium in pitx2a MO knockdown embryos was disorganized and had a scalloped appearance, more similar to skin than to the smooth surface observed in control corneas. Using a trichrome-like stain, which better highlighted the corneal layers, it was evident that pitx2a MO embryos lacked a definitive corneal endothelium (compare Figs. 1G and 1Q). By 96 hpf, the dorsal iridocorneal angle of the control MO embryos contained CNC-derived iris stromal cells, including xanthophores, iridophores, and undifferentiated cells (Fig. 1R). The ventral iridocorneal angle, which has been previously described to be less cellular and slightly developmentally delayed compared with the dorsal angle, ⁶⁷ contained numerous undifferentiated cells in the control embryos (data not shown). After pitx2a knockdown, there was decreased cellularity in the dorsal (Fig. 1C) and ventral (data not shown) iridocorneal angles. Retinas in pitx2a MO lacked photoreceptor outer segments (Fig. 1E) but were otherwise similar to control (Fig. 1L). Thus, the knockdown of pitx2a expression in zebrafish embryos recapitulates features of the ocular phenotype in ARS. 

Figure 1.

 

Pitx2 regulates zebrafish ocular development. Embryos that were injected at the one-cell stage with antisense oligonucleotide MO directed against pitx2a MO (A–E), human wild-type PITX2A mRNA with control MO (F–J), human wild-type PITX2A mRNA with pitx2a MO (K–O), or control MO (P–T) were harvested at 96 hpf. Coronal plastic sections demonstrated that pitx2a MO knockdown resulted in small eyes and malformed jaw cartilage and muscles (A) compared with control embryos (P). Trichrome-like stain demonstrated thickened corneas with a scalloped epithelial (C Epi) surface and a lack of endothelial cells in pitx2a MO knockdown embryos (B) compared with the corneas in control embryos (Q). The dorsal iridocorneal angle of the pitx2a MO knockdown embryos (C) had fewer iris stromal cells, including Xn, Ir, and Un, which are seen in the control MO (R). The medial rectus (MR) in pitx2a MO knockdown on cross-section was thickened and contained disorganized myofibers (D) compared with the MR in the control MO (S). The developing retina in pitx2a MO knockdown embryos (E) showed mild disorganization of the inner nuclear layer (INL), poor demarcation of the outer plexiform layer (OPL), and lack of photoreceptor outer segment. Coinjection of human PITX2A mRNA with control MO similarly resulted in thickened corneas (G) with a scalloped appearance that lacked endothelial cells and dorsal iridocorneal angles (H) that contained fewer iris stromal cells. Injection of human PITX2A mRNA with control MO did not disrupt proper medial rectus formation (I) or retinal (J) development. Coinjection of human PITX2A mRNA with pitx2a MO restored corneal architecture (L), dorsal iridocorneal angle cellularity (M), myofiber organization in the medial rectus (N), and retinal development (O). SR, superior rectus; IR, inferior rectus; MR, medial rectus; C Epi, corneal epithelium; C S, corneal stroma; C End, corneal endothelium; Lens Epi, lens epithelium; Xn, xanthophore; Ir, iridophore; Un, undifferentiated cell; GCL, ganglion cell layer; IPL, inner plexiform layer; INL, inner nuclear layer; OPL, outer plexiform layer; ONL, outer nuclear layer.

Because both loss-of function and gain-of-function mutations are associated with ARS, we took advantage of the accessibility of zebrafish embryos and investigated whether microinjection of human PITX2A mRNA into one-cell stage embryos also disrupted ocular development. Injection of 1 nL of 75 ng/μL wild-type human PITX2A mRNA along with control MO resulted in eyes with thickened corneas (P < 0.001; Table 1, Fig. 1G). As observed in the pitx2a MO knockdown, the epithelium had a scalloped surface and the corneas lacked endothelium in embryos injected with human PITX2A mRNA and control MO. Dorsal (Fig. 1H) and ventral iridocorneal angles of the embryos injected with human PITX2A mRNA also had decreased cellular structure compared with control (Fig. 1R). In contrast to the MO knockdown, the retinas of embryos injected with human PITX2A mRNA appeared normal (Fig. 1J). 

We next determined whether the human allele of PITX2A mRNA could substitute for the zebrafish homolog and rescue the ocular phenotype in the zebrafish pitx2a MO knockdown. Coinjection of human PITX2A mRNA with the pitx2a MO completely restored ocular morphology (Fig. 1K), including the dimensions of the eye (Table 1), corneal architecture (Fig. 1L), angle structure (Fig. 1M), and retinal formation (Fig. 1O). Thus, there is functional homology between human and zebrafish forms of pitx2a in ocular development. 

The K50E mutation (with a substitution of lysine for glutamate) in PITX2A, which was identified in ARS patients, alters the conserved lysine in the DNA-binding domain, leading to a dominant negative effect by sequestration of cofactors and preventing wild-type PITX2A protein from activating target gene transcription. ^8,46 We synthesized human PITX2A mRNA containing the K50E mutation and microinjected the mRNA (1 nL of 75 ng/μL) along with control MO or the pitx2a MO into zebrafish embryos at the one-cell stage. Expression of the dominant negative K50E human PITX2A, when coinjected with control MO, did not significantly alter ocular size, anterior segment formation, or retinal development (Supplementary Figs. S3A–E) compared with embryos injected with control MO alone. On the other hand, expression of K50E human PITX2A, together with morpholino knockdown of the zebrafish pitx2a (Supplementary Fig. S3F–J), resulted in an additive phenotype, worse than pitx2a knockdown alone (Figs. 1A–E). In these embryos (Table 1) the corneas were significantly (P < 0.001) thickened (Table 1; Supplementary Fig. S3G) and lacked endothelium, whereas the dorsal (Supplementary Fig. S3H) and ventral (data not shown) iridocorneal angles were almost devoid of iris stromal cells. The retinas were disorganized and showed loss of cells (Supplementary Fig. S3J). 

Pitx2a Regulates CNC in Craniofacial Development in Zebrafish Embryos

Because ARS can be associated with craniofacial dysmorphism and Pitx2 mutations in mice cause pharyngeal arch and EOM defects, ^14,15,18,19 we next used an in vivo approach to investigate the role of pitx2 on CNC and muscle development using transgenic fish in which GFP is expressed in neural crest cells or differentiated muscle [Tg(sox10::EGFP) or Tg(α-actin::EGFP)]. ^{51 –53,68,69} Knockdown of the pitx2a isoform in Tg(sox10::EGFP) embryos demonstrated that at 72 hpf (data not shown) and 96 hpf, CNC-derived jaw cartilage was maldeveloped and the pharyngeal arch cartilage was absent (compare Figs. 2A and 2B). Similarly, knockdown of pitx2a in Tg(α-actin::EGFP) embryos resulted in abnormal jaw musculature and no pharyngeal arch musculature (compare Figs. 3A and 2B; compare Figs. 1A and 1P). In addition, knockdown of pitx2a was associated with EOM insertions that were placed closer together because of the smaller eye size (Fig. 3A). Cross-sections of the medial rectus in the pitx2a knockdown embryos (Fig. 1D) demonstrated disorganization of myofibers and thickening of the muscle. Knockdown of the pitx2c isoform, in contrast to pitx2a, showed no discernible craniofacial phenotypes (data not shown), consistent with the different roles of these isoforms in embryogenesis. ^70,71  

Figure 2.

 

pitx2 is required for zebrafish CNC development. Microinjection of MO directed against pitx2a (A) at the one- to two-cell stage into Tg(sox10::EGFP) embryos demonstrated at 96 hpf inhibition of neural crest-derived PA development and jaw malformation compared with control MO embryos (B). Coinjection of human PITX2A mRNA with pitx2a MO partially rescued jaw cartilage and PA formation (C), whereas injection of human PITX2A mRNA with control MO also suppressed PA and jaw development (D). Injection of the human dominant negative K50E mutant allele of PITX2A resulted in deformed jaw and PA formation when coinjected with control MO (F) and complete jaw and PA suppression when coinjected with pitx2a MO (E). Human PITX2A mRNA from the T30P autosomal recessive allele did not affect PA or jaw development when coinjected with control MO (H), nor did it rescue the defects when coinjected with pitx2a MO (G). PA, pharyngeal arch.

Figure 3.

 

pitx2a regulates craniofacial muscle development in zebrafish. Tg(α-actin::EGFP) embryos that were injected with pitx2a MO at the one- to two-cell stage demonstrated malformed PA and jaw muscle formation at 96 hpf (A) compared with control embryos (B). MO knockdown of pitx2a did not affect EOM organization. Coinjection of human PITX2A mRNA with pitx2a MO restored jaw and PA muscle development (C). Injection of human PITX2A mRNA with control MO caused mild deformities of jaw and PA musculature (D). PA, pharyngeal arch.

Given the functional equivalence of human PITX2A in ocular development, we tested the effect of injecting wild-type human PITX2A mRNA on CNC and muscle development. Overexpression of human PITX2A revealed malformation of jaw and pharyngeal arch cartilage (Fig. 2D) and minimal changes in jaw, pharyngeal arch, and EOM formation (Fig. 3D; see also Figs. 1F, 1I). 

Because the wild-type human PITX2A mRNA was able to substitute for the zebrafish homolog in ocular development, we next tested whether this was also true for CNC and craniofacial muscle development. Coinjection of pitx2a MO and human PITX2A mRNA resulted in restoration of jaw cartilage and muscle morphology and partial rescue of pharyngeal arch cartilage and muscle formation (Figs. 2C, 3C) compared with injection of pitx2a MO alone (Figs. 2A, 3A). 

Although injection of K50E mutant PITX2A mRNA had minimal effect on ocular development, we examined the effect of this dominant negative mutant on jaw and pharyngeal arch formation. Injection of K50E mutant human PITX2A mRNA resulted in malformed jaw cartilage and inhibition of pharyngeal arch formation (Fig. 2F). Coinjection of K50E mutant human PITX2A mRNA with pitx2a MO resulted in a more severe phenotype (Fig. 2E) than either the mutant mRNA (Fig. 2F) or pitx2a MO (Fig. 2A) alone. 

We next tested another ARS-related mutation in the PITX2 gene, the recessive T30P mutant (substituting threonine for proline). ^48,49 Unlike the K50E mutant, T30P does not prevent wild-type PITX2 from activating gene transcription, although it is transactivation defective. ^48,49,72,73 Microinjection of T30P mutant PITX2A mRNA may have a very mild effect on jaw or pharyngeal arch cartilage formation (Fig. 2H) at 96 hpf. Furthermore, injection of T30P mutant PITX2A mRNA did not rescue the jaw and pharyngeal arch defects caused by pitx2a knockdown (Fig. 2G). 

Taken together, these results demonstrate that in addition to ocular morphogenesis, pitx2 is required for the formation of CNC and mesoderm-derived structures in craniofacial development. Furthermore, these data demonstrate a remarkable level of functional conservation of pitx2 activity between fish and humans. 

Pitx2a Expression Is Regulated by RA in the Perioptic Mesenchyme and Developing Jaw

In mice, pitx2 expression in the perioptic mesenchyme is regulated by RA. ⁴⁵ We investigated whether this was also true in zebrafish. Colorimetric and fluorescent in situ hybridization of pitx2 between 24 hpf and 72 hpf revealed expression in the periocular mesenchyme, developing jaw, and brain (Figs. 4C, 4F; Supplementary Figs. S4C, S4F, S4I, [48 hpf]; data not shown). Furthermore, pitx2 was colocalized with the muscle-specific transcription factor myoD by double in situ hybridization, demonstrating the expression of pitx2 in the mesoderm, which gives rise to the EOM (Fig. 4G). 

Figure 4.

 

pitx2 expression in the periocular mesenchyme of the developing zebrafish is regulated by RA. Whole mount and section colorimetric in situ hybridization of 48 hpf wild-type embryos revealed expression of pitx2 in the periocular mesenchyme, developing jaw, PA, and pituitary in DMSO-treated embryos (C, F). Treatment of embryos with 1 μM RA, starting at 28 hpf, increased pitx2 expression in the orbit and jaw but decreased pitx2 expression in the developing pituitary (A, D). pitx2 expression was decreased throughout the periocular mesenchyme but increased in the pituitary with treatment with 20 μM DEAB at 24 hpf (B, E). Double fluorescent in situ hybridization (G) demonstrated colocalization of pitx2 (pink) expression in the periocular mesenchyme with the muscle-specific transcript myoD (green) in 48 hpf wild-type embryos.

We next determined whether alterations in RA levels affected pitx2 expression. RA signaling is known to regulate early developmental processes such as gastrulation, axis formation, and neural crest specification and migration. To avoid confounding effects from these earlier processes, ^74,75 we added exogenous all-trans RA or the pan-aldehyde dehydrogenase inhibitor diethylbenzaldehyde (DEAB, which inhibits RA synthesis) to the embryo media at 28 hpf or 24 hpf, respectively. Treatment with RA (1–2 μM; Figs. 4A, 4D; Supplementary Figs. S4A, S4D, S4G), upregulated pitx2 expression in the periocular mesenchyme and jaw. In contrast, exogenous RA decreased the expression of pitx2 in the brain. Inhibition of RA synthesis by treatment with DEAB (10–20 μM; Figs. 4B, 4E; Supplementary Figs. S4B, S4E, S4H) starting at 24 hpf decreased pitx2 expression in the periocular mesenchyme but appeared to increase expression in the brain. Thus, RA is a positive regulator of pitx2 expression in the periocular mesenchyme and jaw and a negative regulator of pitx2 expression in the brain, revealing the complexity of this regulatory cascade. 

RA Regulates Ocular Development

The regulation of pitx2 by RA led us to investigate whether exogenous treatment with RA or DEAB had effects similar to those of alterations in pitx2 expression on ocular and craniofacial development. Embryos were treated with exogenous RA or DEAB starting at 28 or 24 hpf, respectively, and then ocular development was examined at 96 hpf. Similar to knockdown of pitx2a expression, we found that treatment with exogenous RA (100 nM; Fig. 5A) starting at 28 hpf resulted in eyes with thickened corneas (P < 0.001) that were significantly smaller in the dorsal-ventral (P < 0.001) and anterior-posterior (P < 0.01) axes (Table 2) at 96 hpf. RA treatment impaired anterior segment development because there was no defined corneal endothelium (Fig. 5B) and there were fewer cells in the dorsal (Fig. 5C) and ventral (data not shown) iridocorneal angles. This was in contrast to DMSO-treated control embryos (Fig. 5K) in which the eyes showed well-defined corneal endothelia (Fig. 5L) and dorsal iridocorneal angles that contained iridophores, xanthophores, and undifferentiated cells (Fig. 5M). The inner nuclear layer of the retina in RA-treated embryos was also mildly disorganized, and the photoreceptor outer segment failed to form (Fig. 5E). These features have some overlap with the pitx2a knockdown ocular phenotypes. 

Figure 5.

 

Excessive RA and pharmacologic inhibition of RA synthesis alter ocular development in zebrafish. Coronal plastic sections of 96-hpf embryos treated with 100 nm RA (A–E) starting at 28 hpf demonstrated smaller, malformed eyes and complete lack of jaw formation (A) compared with 0.1% DMSO control (K). Trichrome-like stain of RA-treated embryos demonstrated lack of corneal endothelial cells (B) and decreased iris stromal cells in the dorsal iridocorneal angle (C) compared with DMSO-control treated embryos (L, M). High magnification of the medial orbit in RA-treated embryos (D) revealed no identifiable MR compared with a cross-section in control embryos (N). The developing retina in RA-treated embryos (E) demonstrated disorganization of the INL, poor demarcation of the OPL, and lack of photoreceptor outer segment formation compared with control embryos (O). Treatment with the pan-aldehyde dehydrogenase inhibitor DEAB (10 μM) starting at 24 hpf resulted in smaller eyes and mild disruption of jaw cartilage and muscle formation (F) compared with DMSO control (K). Trichrome-like stain of corneas of embryos treated with DEAB revealed a mildly thickened and scalloped epithelial (G, C Epi) surface but maintenance of stromal and endothelial layers. The dorsal iridocorneal angle of DEAB-treated embryos (H) had decreased differentiated and undifferentiated iris stromal cells compared with DMSO control (M). The MR in DEAB-treated embryos (I) in cross-section was mildly thickened compared with DMSO control (N). The retina in DEAB-treated embryos (J) demonstrated mild disorganization of the INL, poor demarcation of the OPL, and lack of photoreceptor outer segment formation compared with control embryos (O). MR, medial rectus; C Epi, corneal epithelium; C S, corneal stroma; C End, corneal endothelium; Lens Epi, lens epithelium; Xn, xanthophore; Ir, iridophore; Un, undifferentiated cell; GCL, ganglion cell layer; IPL, inner plexiform layer; INL, inner nuclear layer; OPL, outer plexiform layer; ONL, outer nuclear layer.

Inhibition of RA synthesis with DEAB also caused ocular maldevelopment. The eyes of embryos treated with 10 μM DEAB from 24 to 96 hpf (Fig. 5F, Table 2) were mildly but significantly (P < 0.001) smaller in the dorsal-ventral axis than they were in controls (Fig. 5K, Table 2). The corneas of DEAB-treated embryos, which were significantly thickened (P < 0.001; Table 2), had a scalloped epithelium but contained a defined endothelial layer (Fig. 5G). The dorsal (Fig. 5H) and ventral (data not shown) iridocorneal angles in DEAB-treated embryos were less cellular than in controls (Fig. 5M). As in the pitx2a knockdown embryos, the retinas in DEAB-treated embryos also showed ill-defined outer plexiform layer and lack of outer segment (Fig. 5J). Hence, inhibiting RA synthesis also recapitulates many of the pitx2a knockdown phenotypes. 

Given that both reduced and excess RA levels can induce eye defects, we tested whether exogenous RA can compensate for the inhibition of RA synthesis by DEAB. After treatment with 10 μM DEAB at 24 hpf, RA (1 nM or 10 nM) was added to the embryo media at 28 hpf. We found that 1 nM RA improved, but did not restore, the DEAB effect on corneal thickness and ocular size (Table 2). Treatment with 10 nM RA also improved the DEAB effect on ocular size and restored corneal architecture and thickness (Table 2), anterior segment formation, and retinal development (data not shown). 

Taken together, these data demonstrate that regulation of RA levels during the pharyngula, hatching, and early larval stages ⁵⁰ is required for proper ocular development, including eye size, anterior segment formation, and retinal development. Furthermore, the embryo is very sensitive to alterations in RA levels. 

Craniofacial Tissues Have Different Sensitivities to RA

Because deficiencies and overexposure to RA can lead to craniofacial anomalies, we next investigated the effects of exogenous treatment with RA or DEAB on craniofacial development. Increasing concentrations of RA (0.1 nM-1 μM) starting at 28 hpf disrupted CNC development, and 100 nM completely inhibited pharyngeal arch and jaw formation (Figs. 6A, 6D). RA at concentrations higher than 100 nM caused marked growth and developmental arrest and death by 72 hpf (data not shown). In contrast, 96 hpf control embryos had cartilage that demarcated five to six well-developed pharyngeal arches and a protruding jaw (Fig. 6C, 6F). Similarly, embryos treated with 100 nM RA starting at 28 hpf developed only primitive jaw musculature, and there was a complete lack of pharyngeal arch muscles and EOMs (Figs. 6G, 6J, see also Figs. 5A, 5D). Inhibition of RA synthesis by treatment with 10 μM DEAB starting at 24 hpf resulted in only two distinct pharyngeal arches and a jaw that was shortened and malformed (Figs. 6B, 6E, 6H). However, treatment with 10 μM DEAB largely did not affect the differentiation and organization of the EOMs (Fig. 6K), and the medial rectus showed relatively normal appearance of myofibers (Fig. 5I). 

Figure 6.

 

Excessive RA and pharmacologic inhibition of RA synthesis disrupt zebrafish craniofacial development. Tg(sox10::EGFP) embryos (A–F) treated with 100 nM RA starting at 28 hpf showed an almost complete lack of neural crest with no formation of PA or jaw (A, D). In Tg(α-actin::EGFP) embryos, 100 nM RA treatment showed absence of PA and extraocular musculature and only formation of primitive jaw muscle (G, J) compared with control DMSO-treated embryos (I, L). Tg(sox10::EGFP) and Tg(α-actin::EGFP) embryos treated with 10 μM DEAB starting at 24 hpf had mildly malformed jaws (B, H) and only two PAs (E, H) compared with the protruding jaws (C, I) and five to six well-developed PAs (F, I) seen in control DMSO-treated embryos. DEAB treatment alone (K) did not have an effect on EOM organization when compared with control DMSO (L). PA, pharyngeal arch; IR, inferior rectus; IO, inferior oblique; AM, anterior mandibulae.

Based on our findings that both reduced and excessive RA levels caused dysregulated craniofacial development, we tested whether exogenous RA could rescue pharmacologic inhibition of raldh. After treatment with DEAB at 24 hpf, the addition of different concentrations of exogenous RA (0.1 nM, 1 nM, 10 nM, 100 nM) at 28 hpf rescued craniofacial development by 96 hpf. However, the pharyngeal arches, jaws, and EOMs had different sensitivities to RA. After treatment with 10 μM DEAB, lower concentrations of exogenous RA (0.1–1 nM) improved jaw cartilage and muscle formation (Figs. 7C, 7D, 7K, 7L), whereas higher concentrations of RA disrupted jaw formation (Figs. 7A, 7B, 7I, 7J). In contrast, formation of the five pharyngeal arches occurred only with 10 nM RA (Figs. 7F, 7J). Lower concentrations of RA (Figs. 7G, 7H, 7K, 7L) did not abrogate the DEAB effect on the pharyngeal arches, whereas higher concentrations (Figs. 7E, 7I) had a teratogenic effect. EOMs were the least sensitive to alterations in RA because treatment with DEAB alone or with RA concentrations up to 1 nM (Figs. 7O, 7P) had minimal effect on EOM organization. Interestingly, treatment with 10 nM RA and 10 μM DEAB resulted in inferior oblique and rectus muscles that were attached to each other at the midline (Fig. 7N). 

Figure 7.

 

Zebrafish craniofacial structures have different sensitivities to RA. After treatment with 10 μM DEAB at 24 hpf, Tg(sox10::EGFP) (A–H) or Tg(α-actin::EGFP) (I–P) embryos were treated with increasing concentrations of RA (0.1–100 nM) at 28 hpf. Treatment with 0.1 nM and 1 nM RA rescued the DEAB-induced jaw cartilage (C, D) and muscle (K, L) defects but did not rescue PA structure (G, H, K, L). Treatment with 10 nM RA restored five PAs (F) but disrupted jaw cartilage (B) and muscle (J) development. Treatment with 100 nM RA with 10 μM DEAB inhibited jaw and PA cartilage (A, E) and muscle (I, M) development. After treatment with 10 μM DEAB, treatment with 0.1 nM and 1 nM RA did not affect EOM organization (O, P). Treatment with 10 nM RA with 10 μM DEAB resulted in EOMs that were thickened and disorganized as the IO muscles were connected to each other. IO, inferior oblique; PA, pharyngeal arch.

To determine whether the effects of exogenous RA or DEAB occurred through the activation of RARs, we tested whether treatment with RAR agonists or antagonists affected craniofacial development. Treatment with Ro 41–5253, a selective RARα antagonist at 24 hpf, recapitulated the effects of DEAB-mediated inhibition of RA synthesis (data not shown). Furthermore, treatment at 24 hpf with TTNPB (Ro 13–7410), a selective RAR agonist, mimicked the effects of exogenous RA. In contrast, treatment at 24 hpf with methoprene acid, an RXR agonist, did not alter CNC and EOM development (data not shown). Thus, in zebrafish the effects of RA on craniofacial formation are mediated by RAR. 

RA Is Produced by the Developing Eye during Zebrafish Embryogenesis

The effect of pharmacologic alterations in RA levels on ocular and craniofacial development after early events during embryogenesis led us to further investigate the source of RA production in zebrafish embryos. In situ hybridization in 36-hpf zebrafish embryos demonstrated that raldh2 was expressed predominantly in the anterodorsal retina (Figs. 8A, 8B), whereas raldh3 was expressed in the posteroventral retina (Figs. 8A, 8B). Raldh2 expression was also noted in the brain ventricle, but no other adjacent RA source was noted at this stage of development. 

Figure 8.

 

RA is produced by the developing eye, and human PITX2A mRNA rescues ocular defects caused by RA deficiency. Whole mount (A) and section (B) double in situ hybridization in 36-hpf wild-type embryos showed expression of raldh2 (pink, arrows) predominantly in the anterodorsal retina and raldh3 (purple, arrowhead) in the posteroventral retina in the developing eye. Raldh2 was also expressed in the brain around the ventricles. Coronal plastic sections of 96-hpf embryos injected at the one- to two-cell stage with antisense oligonucleotide MO directed against raldh2 and p53 MO (C–G) demonstrated small and malformed eyes and deformed jaw cartilage and muscles. High magnification and trichrome-like stain of raldh2/p53 MO knockdown embryos (D) revealed a thickened cornea with a scalloped epithelium (C Epi) and lack of endothelium (C End) and a dorsal iridocorneal angle (E) that contained fewer iris stromal cells. Cross-section of the MR in raldh2/p53 MO (F) demonstrated thickening of the muscle and disorganization of the myofibers (K). High magnification of the developing retina in raldh2/p53 MO knockdowns (G) demonstrated mild disorganization of the INL, poor demarcation of the OPL, and lack of photoreceptor outer segment formation. Coinjection with raldh2/p53 MO and human PITX2A mRNA at the one-cell stage revealed restoration of ocular and jaw development (H). Human PITX2A mRNA rescued corneal architecture (I), dorsal iridocorneal angle structure (J), MR muscle, and myofiber morphology (K), and retinal development (L) in raldh2/p53 MO knockdown embryos. SR, superior rectus; MR, medial rectus; C Epi, corneal epithelium; C S, corneal stroma; C End, corneal endothelium; Lens Epi, lens epithelium; Xn, xanthophore; Ir, iridophore; Un, undifferentiated cell; GCL, ganglion cell layer; IPL, inner plexiform layer; INL, inner nuclear layer; OPL, outer plexiform layer; ONL, outer nuclear layer.

Knockdown of raldh2, but Not raldh3, Disrupts Ocular and Craniofacial Development

We further investigated the role of RA on ocular and craniofacial development by using MO directed against raldh2 and raldh3. Injection of raldh2 MO into one- to two-cell stage embryos recapitulated the previously described ⁵⁹ neckless phenotype, which characteristically has anteroposterior axis truncation anterior to the somites, absence of pectoral fins, and midline defects. Similar results were obtained when we combine MO knockdown of raldh2 with MO knockdown of p53 to reduce apoptosis and nonspecific phenotypes. Knockdown of raldh2 and p53 resulted in maldeveloped eyes (Table 1, Fig. 8C) that were similar in appearance to those observed in the pitx2a knockdown embryos (Fig. 1A). These embryos had smaller malformed lenses that contained areas devoid of lens fibers (Figs. 8C, 8E asterisks). Corneas in raldh2/p53 MO knockdown embryos (Fig. 8D) were significantly thicker than in control (P < 0.001; Table 1) and had a scalloped epithelium and no endothelium. There was decreased cellularity in the dorsal (compare Figs. 8E and 1R) and ventral (data not shown) iridocorneal angles, and the retinas were mildly disorganized with an ill-defined outer plexiform layer and a lack of photoreceptor outer segment (Fig. 8G). 

We next used an in vivo approach to test the effects of raldh2 knockdown on craniofacial cartilage and muscle development. Knockdown of raldh2 and p53 in caused inhibition of the rostral wave of the CNC (data not shown) by 48 hpf, which by 96 hpf resulted in malformation of jaw cartilage and muscle and loss of pharyngeal arch formation (compare Figs. 9A, 9B and 9E, 9F; see also Fig. 8C). Because of their smaller eyes, the insertions of the EOM onto the globe were more closely spaced in the raldh2/p53 knockdown embryos (Fig. 9B), and the medial rectus had disorganized myofibers (Fig. 8F). 

Figure 9.

 

RA regulates zebrafish craniofacial CNC and muscle development. Microinjection of MO directed against raldh2 and p53 at the one- to two-cell stage into Tg(sox10::EGFP) embryos demonstrated inhibition of neural crest development (A) at 96 hpf. In the raldh2/p53 MO knockdown, there was absence of neural crest-derived jaw and PA formation compared with embryos injected with control MO (E). Knockdown of raldh2 and p53 in Tg(α-actin::EGFP) embryos revealed deformed jaw and PA musculature formation (B) compared with control embryos (F). The EOMs properly differentiated, but the insertions on the globe were spaced more closely together in the raldh2/p53 knockdown because of the smaller eye (B). Injection of human pitx2a mRNA in Tg(sox10::EGFP) and Tg(α-actin::EGFP) embryos demonstrated partial rescue of jaw and PA cartilage (C) and muscle (D) defects in raldh2/p53 MO. PA, pharyngeal arch; IR, inferior rectus; IO, inferior oblique; AM, anterior mandibulae.

In contrast to MO knockdown of raldh2, embryos injected with MO directed against raldh3 had no apparent jaw, pharyngeal arch, EOM, or ocular abnormalities and appeared morphologically similar to controls (data not shown). Coinjection of raldh2, raldh3, and p53 MO, however, resulted in global developmental arrest, with death by 96 hpf. 

Pitx2 Partially Mediates RA Regulation of Ocular and Neural Crest Development

We took advantage of our in vivo zebrafish model to assess the functional significance of the regulation of periocular pitx2a by RA during ocular and craniofacial development. We coinjected human PITX2A mRNA with raldh2/p53 MO at the one-cell stage and found that at 96 hpf, human PITX2A mRNA rescued the ocular phenotype of the raldh2/p53 MO (compare Figs. 8H and 8C), including the ocular dimensions (Table 1), corneal architecture (Fig. 8I), angle cellularity (Fig. 8J), and retinal formation (Fig. 8L). Furthermore, human PITX2A mRNA partially rescued the jaw and pharyngeal arch cartilage phenotypes of raldh2/p53 knockdown (compare Figs. 9A and 9C). Similarly, human PITX2A mRNA restored jaw and pharyngeal arch muscle formation in raldh2/p53 knockdown (compare Figs. 9B and Fig. 9D). 

We next sought to determine whether RA was not only necessary but sufficient for directing these events during embryogenesis. This was accomplished by replacing the developing eye with an ion-exchange bead soaked in RA. One eye was enucleated at 16 hpf; this was followed by implantation of a 40- to 60-μm ion exchange bead soaked in DMSO or 100 nM RA. Similar to previous reports, ^38,44 we found that enucleation inhibited CNC and EOM development. Implantation of an RA-soaked bead into an empty orbit did not rescue either CNC development or EOM organization (data not shown). 

These findings suggest that RA produced by the developing eye regulates pitx2 expression in the periocular mesenchyme and that both RA and pitx2 are key components in ocular and craniofacial development. RA alone, however, is not sufficient for properly organizing the orbit around the eye, demonstrating that additional signaling pathways are also required for craniofacial development. 

Discussion

Vertebrate craniofacial development involves complex signaling between the CNC and the surrounding tissues, and disruption of these signals during embryogenesis can result in congenital craniofacial and ocular anomalies such as those found in ARS. In the present study, we used zebrafish to investigate the function and regulation of pitx2, a paired-homeobox transcription factor that is commonly mutated in ARS. We first established that our zebrafish system could be used as a model for craniofacial and ocular development by recapitulating many of the characteristics of ARS by knocking down pitx2a expression. ^3,4,49,65,66 Moreover, by rescuing phenotypes using the human form of PITX2A mRNA, these experiments demonstrate remarkable functional homology of pitx2 between teleost fish and mammals and provide further evidence that similar signaling pathways are used during craniofacial and ocular development across vertebrate classes. The partial nature of the rescue may be related to the fact that injected PITX2A mRNA may be present in every cell in the embryo, which is, of course, abnormal in and of itself. We also expressed known disease-causing alleles of PITX2A and demonstrated their significance in vivo. ^{46 –49} The K50E mutant PITX2 gene product has an altered affinity for the DNA-binding site and can act in a dominant negative fashion by preventing wild-type PITX2 from activating gene transcription. ^72,73 Microinjection of the K50E mutant human PITX2A mRNA did not cause significant ocular abnormalities but did disrupt CNC development, possibly reflecting the dissimilar availability of transcriptional cofactors in the different cell populations. Furthermore, coinjection of the dominant negative mutant mRNA with pitx2a MO resulted in a worse ocular and neural crest phenotype than injection of either the mutant mRNA or the MO alone, uncovering in this model that graduated levels of haploinsufficiency correlate with phenotypic expressivity. These results recapitulated in zebrafish the sensitivity of the phenotypes to relative pitx2a expression levels in the context of this dominantly inherited mutant allele. ^46,47,49 On the other hand, the recessive T30P mutant PITX2 gene product lacks transactivation ability but does not interfere with wild-type PITX2 activity. ^{46 –49,72,73} The T30P mutant human PITX2A mRNA did not induce significant defects when injected alone and was incapable of rescuing the pitx2a knockdown phenotypes, maintaining its recessive character in our zebrafish model. 

Both gain-of-function and loss-of-function mutations in PITX2 are associated with ARS, ^{3 –8,49,65,66} and our results demonstrate the importance of tightly regulating pitx2a expression in both craniofacial and ocular development. In the eye, alterations in pitx2a expression resulted in impairment of corneal development and absence of neural crest-derived corneal endothelial and iris stromal cells. We found that in zebrafish embryos, alterations in pitx2a expression resulted in thicker corneas, whereas a study in humans and mice found that mutation in PITX2 caused thinned corneas in adulthood. ⁶⁴ This difference may be due to the fact that our studies examined the embryonic stage during which the corneal architecture is still being formed. The observed delay in retinal development may be secondary to a general delay in ocular development or may reveal a role for pitx2 in retinal development. Pitx2 has been reported to regulate optic stalk formation in a mouse model, ¹⁵ whereas a role for pitx2 in retinal development has not previously been described. Interestingly, alterations in pitx2a expression were associated with arrest of retinal development at 48 hpf; pitx2 expression around the eye peaks at 48 hpf. Further studies will be required to fully assess the role of pitx2 in retinal development. 

In our studies, alterations in pitx2a expression also caused abnormalities in jaw and pharyngeal arch cartilage and muscle formation. This correlates to studies in the mouse as well as the observation that patients with mutations in PITX2 more frequently have associated craniofacial dysmorphism compared with mutations in other ARS-causing genes such as FOXC1 and PAX6. ^{4,17 –19} Thus, our data are consistent with the causative role of pitx2 mutations in ARS and recapitulate many aspects of ARS. 

In mouse models, conditional knockout of Pitx2 disrupts development of the EOMs, and pitx2 has been shown to activate a cascade of muscle-specific transcription factors. ^{14 –16} In evaluating pitx2 expression patterns, we found that pitx2 colocalized with the muscle-specific transcript myoD in the periocular mesenchyme. However, despite its expression pattern and effect on jaw and pharyngeal muscles, altering pitx2a expression did not drastically affect development of the EOMs, although it did cause mild disorganization of the myofibers. It may be that other pitx2 isoforms or even pitx3 expression in EOM can compensate for the reduction in pitx2a expression. ⁷⁶ Further studies are required to evaluate the role of pitx2 in EOM formation and the way in which this process may differ from muscle development in other craniofacial regions in the zebrafish and myogenesis in mammalian models. 

Despite its link to ARS and the essential role of pitx2 in ocular and craniofacial development, little is known about the regulation of its expression. It has been shown that RA regulates pitx2 expression in the periocular mesenchyme in mice, and, indeed, the pitx2 promoter contains an RA response element. ⁴⁵ However, the in vivo significance of this is difficult to assess in mammalian models. By treating embryos with exogenous RA or the inhibitor DEAB at 24 to 28 hpf, we focused the treatments to late stages of embryogenesis with the goal of avoiding broad disruptions to early embryonic patterning. We found that in zebrafish embryos, RA regulated pitx2 expression not only in the periocular mesenchyme but also in the jaw and pharyngeal arches. Treatment of embryos with exogenous RA increased pitx2 expression in these regions, whereas inhibition of RA synthesis by DEAB decreased pitx2 expression. It is important to note that pitx2 expression in the brain had the opposite response to changes in RA levels. By adding the pharmacologic agents later in embryo development, we attempted to reduce their effects on embryo patterning, which is primarily completed by 24 hpf. Using the zebrafish model, we were able to demonstrate in vivo that pitx2 is a functional downstream target of RA in ocular and craniofacial development because injection of human PITX2A mRNA partially rescued the ocular and craniofacial morphogenesis defects caused by knockdown of raldh2. Thus, human PITX2A mRNA partially bypassed the RA requirement for proper development. The incomplete nature of the rescue is likely the result of additional RA-sensitive genes that are important in craniofacial morphogenesis, both known and unknown, and a reflection of the challenge of providing the precise amount of pitx2a that will lead to optimal rescue. 

RA is recognized as both an essential morphogen and a teratogen that regulates craniofacial development in vertebrates. Numerous studies have demonstrated complex local regulation of RA concentration by synthesis through raldh and degradation by cyp26. Previous studies in the mouse have shown that the developing ocular tissues express raldh2, raldh3, and raldh4, as well as cyp26, in a precise temporal-spatial manner, which creates an RA gradient that targets periocular tissues. ^{27 –30,77} We demonstrate that a similar process occurs in zebrafish development. Dehydrogenase enzymes expressed in the developing retina indicate the eye serves as a source of RA in the craniofacial region. Periocular, jaw, and pharyngeal arch structures displayed different sensitivities to RA that were consistent with the distance from the source of RA and the expression of degradation enzymes. The pharyngeal arches had the narrowest window of RA sensitivity, which correlates with the anatomic distance from the eye and the predominant expression of cyp26a1 in the posterior pharyngeal arches. EOMs, on the other hand, had high tolerance to variations in RA levels, whereas the jaw had intermediate tolerance to changes in RA levels. 

In addition to the sequence homology (Supplementary Fig. S5), the functional homology of human and zebrafish pitx2 underlies its critical role in craniofacial development and the importance of craniofacial and ocular structures in vertebrate evolution. Indeed, even blind and grossly eyeless fish have maintained primordial eye structures buried within the facial skeleton that may interact with signaling centers in generating proper craniofacial structures. ³⁹ The evolutionary conservation of RA and pitx2 function in the context of significant differences in craniofacial structures represents remarkable modularity of the gene networks that are responsible for vertebrate morphogenesis. Similarly, despite the significant differences between human and zebrafish anterior segments, the same genes and signals appear to play key roles in ocular development. This suggests that the zebrafish model of ARS may be useful for not only studying this genetic disorder but for using certain strengths of the zebrafish model, such as high throughput screening, to develop novel therapies. 

In summary, we report that, in zebrafish, pitx2 is a downstream target of RA and that together pitx2 and RA mediate signals among the developing eye, CNC, and surrounding tissues to regulate ocular and craniofacial development. Furthermore, zebrafish pitx2 and human PITX2 share significant functional homology, revealing a remarkable conservation of function between teleost fish and humans that may prove to be useful in generating novel therapies for human RA- and PITX2-related disorders. 

Supplementary Materials

Table st1, PDF - Table st1, PDF 

Figure sf01, TIF - Figure sf01, TIF 

Figure sf02, TIF - Figure sf02, TIF 

Figure sf03, TIF - Figure sf03, TIF 

Figure sf04, TIF - Figure sf04, TIF 

Figure sf05, TIF - Figure sf05, TIF 

Figure sf06, TIF - Figure sf06, TIF 

Footnotes

 Supported by a Research to Prevent Blindness Career Development Award (AK); the Alliance for Vision Research (AK); Postdoctoral Training Grant T32 EY013934 from the National Eye Institute (NEI) of the National Institutes of Health (NIH) (BLB); a Knights Templar Eye Foundation award (BLB); a Fight for Sight student award (DSK); Grants K08 EY018689 (AK) and R01 EY018132 (DG) from the NEI of the NIH; a Vision Research Core Award P30 EY007003 to the Department of Ophthalmology and Visual Sciences, Kellogg Eye Center, University of Michigan; the Zebrafish International Resource Center (supported by Grant P40 RR102546 from the NIH-NCRR); and the Helmut F. Stern Career Development Professorship in Ophthalmology and Visual Sciences (AK).

Footnotes

 Disclosure: B.L. Bohnsack, None; D.S. Kasprick, None; P.E. Kish, None; D. Goldman, None; A. Kahana, None

The authors thank Rose Elsaeidi, Donika Gallina, and Steven Grzegorski for technical assistance; Mitchell Gillett for tissue processing, sectioning, and staining of the methyl acrylate ocular sections; Phillip Gage and Peter Hitchcock for critical reading of the manuscript; and Michael Walter, Rachel Wong, and Phillip Gage for sharing reagents.