The suggestion that the iris smooth muscle is generated by the anterior rim of the optic cup neuroepithelium was made nearly a hundred years ago (Szili, 1901; Nussbaum, 1901; Lewis, 1903), and, since then, avian chimeric tissue studies have supported that view (Johnston et al., 1979; Nakano and Nakamura, 1985; Yamashita and Sohal, 1986). Little recent research has focused on the development of the iris smooth muscle and the genes that regulate its development have yet to been identified. The development of iris smooth muscle is interesting both in relation to eye development, and more generally, in how neural tissue generates smooth muscle.
Whereas in mammals the iris and ciliary muscles are smooth muscle type, in birds the ciliary muscle and much of the iris muscle are striated (Gabella and Clarke, 1983; Pilar et al., 1987), although the sphincter muscle probably remains partially smooth muscle in the adult. We know very little about the early events in the development of the iris smooth muscle in mammals, but studies nearly a hundred years ago described morphologically the development of the iris smooth muscle in chick (Lewis, 1903); the developing smooth muscle evaginates from the most anterior rim of the optic cup neuroepithelium and forms epithelial buds that project into the surrounding mesenchyme. Initially, distinct buds appear to be spaced around the pupil, and by some unknown mechanism, perhaps by fusion of the buds or generation of smooth muscle between the buds, a continuous ring of smooth muscle later surrounds the entire pupillary margin. Since the original observations of Lewis (1903), very little research has focused on the initial induction and differentiation of iris smooth muscle from the optic cup neuroepithelium; instead, most of the research has focused on the question of whether smooth muscle in birds transdifferentiates into striated muscle.
The only other vertebrate neural tissue that has been shown to generate smooth muscle is the neural crest (Le Lievre and Le Douarin, 1975). Neural crest cells are generated by the dorsal neural tube, they migrate away from the neural tube and give rise to many different cell types including the peripheral nervous system, melanocytes, and some smooth muscle (Le Douarin, 1982). The expression patterns of many genes that may regulate neural crest development have been described, and because smooth muscle is a derivative of neural crest, I determined whether some of the genes thought to regulate neural crest development are expressed at the right time and place to regulate the development of iris smooth muscle from the anterior optic cup neuroectoderm. First, I show that developing smooth muscle is labeled by the HNK1 antibody, an antibody that is a commonly used neural crest marker in chick (Bronner-Fraser, 1986), and then use HNK1 immunoreactivity to document the development of the iris smooth muscle. Second, I show that developing iris smooth muscle coexpresses two paired box transcription factors, Pax3 and Pax6. Third, I show that BMP4 and BMP are expressed by cells in the anterior optic cup and that the cells at the site of smooth muscle generation highly express both genes. BMP4 and BMP7 are expressed in the dorsal neural tube, and BMP4 and BMP7 can induce neural crest markers in chick neural plate explants, including HNK1 immunoreactivity and Pax3 expression (Liem et al., 1995).
Antibody Labeling of Developing Iris Smooth Muscle
I found that the HNK1 antibody, which labels neural crest cells in chick (Bronner-Fraser, 1986) and other cell types, labels developing iris smooth muscle cells (Fig. 1). HNK1 antibody labels cells in the anterior rim of the optic cup starting around embryonic day 7 (E7). Initially, some HNK+ cells that lie within the neuroepithelium and at the border between the pigmented and nonpigmented epithelium are labeled (Fig. 1, E7). At E8, more HNK1+ cells are present in the rim and by E9, many more cells are HNK1+ and the HNK1+ cells have evaginated from optic cup neuroepithelium to form an epithelial bud (Fig. 1, E9, black arrow). Most of the HNK1+ cells lie within the bud that has formed at the rim of the optic cup neuroepithelium, although some scattered process-bearing HNK1+ cells lie outside the bud (Fig. 1, E9, arrowheads). In sections of E10.5 eyes, most HNK1+ cells lie within the epithelial bud. More posterior, or distal, to the epithelial bud that forms at the most anterior tip, a cluster of cells are HNK1+ (Fig. 1, E10.5), these cells are either a secondary epithelial bud (Lewis, 1903) or a nerve bundle (see Fig. 4A, white arrow), the HNK1+ process-bearing cell is probably a Schwann cell (arrowhead). To identify smooth muscle cells, eyes were labeled with antibodies against α-smooth muscle actin (SMA) and calponin (Skalli et al., 1986; Gimona et al., 1990). The developing iris smooth muscle and cells surrounding large blood vessels (Fig. 2A, arrows) intensely label with anti-SMA in sections of E7, E8, and E11 chick eyes. In addition to the intense labeling of iris smooth muscle and blood vessels, other cells within the iris stroma are labeled less intensely with anti-SMA, and trabecular meshwork cells also label faintly with anti-SMA (Fig. 2A, E8, TM). Iris smooth muscle is also calponin immunoreactive (Fig. 2B). No labeling was observed when the tissue was incubated with only the secondary antibody (data not shown).
To assess the timing and pattern of the developing iris smooth around the pupillary margin whole-mounts of the anterior eye were labeled with HNK1 antibody at E7 and E8 (Fig. 3). In the E7 pupillary margin (Fig. 3A), many cells in the temporal–dorsal quadrant are HNK1+, whereas few cells in the nasal–ventral quadrant are HNK1+. HNK1 antibody labels cells in patches (Fig. 3B, arrows) that are larger in the temporal direction and smaller in the nasal direction. Whole-mount labeling reveals a pattern and gradient of developing iris smooth muscle and shows that sections through the pupillary region are not necessarily of an equivalent developmental stage. When tissue was incubated with only the secondary antibody, conjugated to alkaline phosphatase, no labeling was observed surrounding the pupillary margin (data not shown).
The presence of ciliary folds is correlated with the presence of HNK1+ cells (Fig. 3C). Thus, the ciliary folds are well developed in the temporal–dorsal quadrant where there is a high level of HNK1 antibody labeling. In contrast, the ciliary folds are not yet developed in the nasal–ventral quadrant where there is little HNK1 antibody labeling. At E8, both HNK1 antibody labeling and ciliary folds surround the entire circumference of the pupil (Fig. 3D, and data not shown).
Expression of Pax3 and Pax6
Because mutations in Pax6 cause eye defects, including the absence of irides and small eyes, in humans (Ton et al., 1991; Jordan et al., 1992; Hanson et al., 1994) and mice (Hill et al., 1991), I assessed the expression pattern of Pax6 in the iris smooth muscle. Sections of embryonic chick eyes were hybridized with Pax6 digoxigenin (DIG) -labeled antisense probe. Figure 4B shows the expression (blue reaction product) of Pax6 in E7, E10, and E13 anterior chick eyes; this expression pattern resembles that Pax3 (Fig. 4A). Other cells, in addition to iris smooth muscle, also express Pax6, including lens epithelial cells, the ciliary epithelium, the pigmented epithelium, and the inner layer of the iris, which has not yet pigmented (Fig. 4B), as well as cells in the neural retina (data not shown). In the section of E10 anterior eye, a putative secondary epithelial bud also expresses Pax6 (Fig. 4B, E10, arrow).
I also assessed whether Pax3 was expressed at the right time and place to be involved in iris smooth muscle development. Sections of embryonic chick eyes were hybridized with Pax3 DIG-labeled antisense probe. Figure 4A shows the expression (blue reaction product) of Pax3 in E7, E10, and E13 developing iris smooth muscle cells. Expression appears in iris smooth muscle cells that are also immunoreactive for HNK1, anti-SMA, and anti-calponin antibodies (Figs. 1, 2). The developing iris smooth muscle was the only tissue observed to express Pax3 in the embryonic chick eye from E5 to E13.
Expression of BMP4 and BMP7
Both BMP4 and BMP7 are expressed in the dorsal neural tube and recombinant BMP4 and BMP7 protein can induce markers of dorsal cell fate and neural crest, including Pax3 expression and HNK1 immunoreactivity, in neural plate explant assays (Liem et al., 1995). To assess the role of members of the BMP superfamily, including BMP2, BMP4, BMP6, and BMP7, in regulating HNK1 and Pax3 expression in developing iris smooth muscle, sections of embryonic chick eyes were hybridized with DIG-labeled antisense probes for the BMP transcripts. Figure 5A shows that, at E7, BMP4 is expressed in cells lying at the most anterior tip of the optic cup neuroepithelium (Fig. 5A, E7, arrow). Ciliary epithelial cells (Fig. 5A, E7, bracket) and cells at the most anterior edge of the iris stroma also express BMP4 (Fig. 5A, E10, arrowheads). At E10, BMP4 is highly expressed by the cells at the very tip of the optic cup neuroepithelium (Fig. 5A, E10, arrow) and by ciliary epithelial cells (Fig. 5A, E10, double arrows) and by cells at the anterior edge of the iris stroma (Fig. 5A, E10 arrowheads). BMP4 expression in E13 is similar to E10, with peak expression in cells at the very tip of the optic cup neuroepithelium (Fig. 5A, E13, arrow).
BMP7 expression (Fig. 5B) is more widespread than BMP4 expression in the anterior chick eye. At E7, the highest level of BMP7 expression is by cells lying at the most anterior tip of the optic cup neuroepithelium (also the site of highest BMP4 expression, see Fig. 5A), BMP7 is also expressed by ciliary epithelial cells, by developing iris smooth muscle cells, and by mesenchymal cells within the iris stroma. Expression of BMP7 in the E9 and E13 anterior eye is similar to E7 BMP7 expression (Fig. 5B).
Weak expression of BMP6 was observed in the ciliary epithelium at E7 and E9 (data not shown). No expression of BMP2 was detected in the anterior eye at any ages examined (E5 to E13, data not shown).
In the past several years, many of the mechanisms and genes that govern various aspects of skeletal muscle development have been identified. In contrast, very little is known about the mechanisms of smooth muscle development and very few genes have been identified that are involved in its induction and differentiation. Most smooth muscle is of mesodermal origin but some is of ectodermal origin; some neural ectoderm (rhombencephalic), by means of a neural crest intermediary, generates smooth muscle of the aortic arches (Le Lievre and Le Douarin, 1975). The iris smooth muscle is the other instance of neural ectoderm generating smooth muscle. As an initial step toward understanding the mechanisms and genes that regulate the development of the iris smooth muscle, I examined the spatial and temporal development of iris smooth muscle and the expression of several genes, including some genes that have been suggested to play a role in neural crest development, which is the other neural tissue that generates smooth muscle (Le Douarin, 1982).
Whole-mount HNK1 antibody labeling of anterior E7 chick eyes reconfirms that the iris smooth muscle initially develops as discrete patches or epithelial buds (Fig. 3B), an observation that was made previously by examination of histological sections (Lewis, 1903). Several interesting questions remain; how is the spacing of the buds regulated, and how does the iris smooth muscle transforms from a pattern of discrete buds to a continuous ring of muscle around the pupil? Do the buds fuse or does new muscle form between the buds? It is noteworthy that iris smooth muscle development, as judged by HNK1 (Fig. 1) and SMA (data not shown) antibody labeling, is not evident until the mesenchyme in the anterior eye extends to the very tip of the optic cup neuroepithelium (see Fig. 1, E6 and E7). Whole-mount HNK1 labeling further reveals that there is a gradient of smooth muscle development around the pupil: the most developed muscle is in the dorsal–temporal quadrant, and the least developed in the ventral–nasal quadrant in the E7 eye (Fig. 3A). The pattern of anti-calponin labeling suggests that smooth muscle cells more distal within the epithelial bud may be more differentiated than proximal cells because distal cells label more intensely (Fig. 2B).
Whole-mount labeling of anterior E7 chick eyes with the HNK1 antibody also shows that the development of the iris smooth muscle correlates in time and position with the development of the ciliary folds (Fig. 3C). Folding of the ciliary epithelium is thought to arise from differential division of ciliary epithelial cells (and overlying pigmented epithelial cells) relative to posterior retinal cells (Bard and Ross, 1982; Reichman and Beebe, 1992). The observation that the smooth muscle development is correlated in time and position with the folding of the ciliary epithelium raises several interesting possibilities, one of which is that iris smooth muscle is required for the proper formation of the ciliary folds. Perhaps the iris smooth muscle acts like a purse string that fixes the diameter of the pupil, while the ciliary epithelium is growing in surface area and intraocular pressure is increasing, thus causing the folding of the ciliary epithelium.
Humans that are heterozygous for mutations in PAX6 manifest aniridia and Peters' anomaly, both of which are characterized by multiple anterior eye defects, including an absence of irides, small eyes, cataract, and glaucoma. It is noteworthy that, although Pax6 is also expressed in the developing posterior eye in both the neural retina and pigmented retinal epithelium (Grindley et al., 1995), these tissues develop relatively normally in Pax6 heterozygous mice, although in humans foveal hypoplasia has been observed in some cases (Azuma et al., 1996). In Pax6 heterozygous mutants the iris is hypoplastic (Singh et al., 2002; Baulmann et al., 2002) and the iris sphincter muscle (smooth muscle) is much reduced (Baulmann et al., 2002). The expression pattern of Pax6 in the developing iris smooth muscle has not been previously reported. In this study, I show that Pax6 is expressed in both layers of the developing iris epithelium and in the developing iris smooth muscle. It will be interesting to determine whether the iris smooth muscle and ciliary folds develop normally in Pax6 heterozygous mutant mice, in Pax6−/+:Pax3−/− and Pax6−/+:Pax3−/− mice.
I examined the expression of members of the BMP superfamily in the region that generates the iris smooth muscle because of their previously identified roles in the development of the dorsal neural tube and neural crest (Liem et al., 1995). It has been shown that treatment of neural plate explants with BMP4 and BMP7 induces expression of Pax3 and HNK1 immunoreactivity (Liem et al., 1995), both of which I show are expressed by developing iris smooth muscle. Furthermore, treatment of neural crest with TGFβ superfamily members BMP2 and TGFβ1 instructively promotes smooth muscle fate in clonal cultures of neural crest (Shah et al., 1996), and in cultures of embryonic chick iris, another member of the TGFβ superfamily, Activin, promotes smooth muscle development (Link and Nishi, 1998). In this study, I show that both BMP4 and BMP7 are expressed at an appropriate time and place to regulate iris smooth muscle development (Fig. 5); both genes are highly expressed by cells at the very tip of the rim of the optic cup neuroepithelium, which is the site where iris smooth muscle is generated. BMP2 is closely related to BMP4 (Hogan, 1996); therefore, perhaps BMP4 can also promote smooth muscle fate. The mesenchyme that surrounds the developing iris smooth muscle and the iris smooth muscle express BMP7. In addition to their putative role in iris smooth muscle development, BMP4 and BMP7 are expressed in a location to regulate lens and ciliary epithelial development.
Eye defects occur in mice deficient in BMP4 (Dunn et al., 1997) and BMP7 (Luo et al., 1995; Dudley et al., 1995; Jena et al., 1997). Homozygous BMP4 null mice die early embryonically, but BMP4 heterozygous mice survive and exhibit an increased frequency of microphthalmia and anophthalmia (Dunn et al., 1997). Homozygous null BMP7 mice have eye defects ranging from anophthalmic to eyes of normal size (Luo et al., 1995; Dudley et al., 1995; Jena et al., 1997). It is unknown whether the ciliary epithelium and iris smooth develop in the BMP4- and BMP7-deficient mice.
I have been unable to identify any crest-like cells that are generated intermediate between the optic neuroepithelium and the iris smooth muscle; for instance, I did not detect the expression of Slug, a transcription factor expressed by neural crest cells (Nieto et al., 1994), in cells in the rim of the optic cup neuroepithelium (data not shown). This is perhaps not surprising, as presumptive smooth muscle precursor cells do not appear to undergo an epithelial to mesenchymal transition like neural crest cells, rather, it seems smooth muscle is generated directly by the optic cup neuroepithelium.
I show the expression pattern of several genes in the anterior eye that have been implicated in the development of dorsal cell fate in the neural tube, including BMP4, BMP7, and Pax3 (Liem et al., 1995). Furthermore, the signaling molecule Sonic hedgehog (Shh) that has ventral neural tube patterning activity is expressed in the mouse neural retina (Jensen and Wallace, 1997) and is similarly expressed in chick neural retina (A.M.J., unpublished observations). The Shh expression extends to the ora serrata, which is the border between the neural retina and ciliary epithelium (A.M.J., unpublished observations). Given these similarities, I propose that the vertebrate eye may use some of the genes for patterning the anterior–posterior axis of the optic cup neuroepithelium as those used in patterning the dorsal–ventral axis of the neural tube, namely Sonic hedgehog and BMPs. Furthermore, I propose that similar genes and mechanisms may generate both neural crest and iris smooth muscle from neuroectoderm. For instance, the neural crest is generated at the neural folds, which is the border region between the non-neural ectoderm and the neural plate, and the iris smooth muscle is generated at the border between the pigmented non-neural epithelium and the unpigmented epithelium (Fig. 1). An interesting issue is whether the unpigmented epithelium at the anterior rim of the optic cup still retains the capacity to generate neural tissue.
In Situ Hybridization
Eyes from embryonic White Leghorn chickens were dissected in phosphate buffered saline (PBS), pierced with a needle, and fixed in 4% paraformaldehyde in PBS overnight at 4°C. Eyes were embedded in 1.5% agar with 5% sucrose and sunk in 30% sucrose in PBS for up to 1 week at 4°C. Cryostat sections (18 μm) were cut and transferred onto Vectabond (Vector Laboratories) -coated slides, air-dried for 2–6 hr at room temperature, and stored desiccated at −20°C.
In vitro transcription.
The following templates (all subcloned into pBluescript) were used to generate DIG-labeled antisense RNA probes: Pax3 linearized with BamHI, transcribed with T3 polymerase. Pax6 linearized with EcoRI, transcribed with T7 polymerase. BMP2 linearized with HindIII, transcribed with T3 polymerase. BMP4 linearized with BamHI, transcribed with T3 polymerase. BMP6 linearized with XbaI, transcribed with T7 polymerase. BMP7 linearized with XhoI, transcribed with T3 polymerase.
Labeling of RNA probes with DIG-UTP was performed according to the manufacturer's recommendations. Briefly, 1 μg of linearized template was transcribed in a total volume of 20 ml using T7 or T3 polymerase (Boehringer Mannheim) and DIG-UTP (Boehringer Mannheim) for 2 hr at 37°C. The reaction products were precipitated, resuspended in 100 μl water or hybridization buffer, and stored at −20°C.
In situ hybridization.
DIG-labeled RNA probes were diluted in hybridization buffer (50% formamide, 10% dextran sulfate, 1 mg/ml yeast RNA, 1× Denhardt's, and 1× salt) and denatured for 10 min at 7°C. Approximately 100 μl of diluted probe was placed on the sections and a coverslip placed on top. Sections were hybridized overnight at 65°C in a humidified box. The slides were washed twice in 50% formamide, 1× standard saline citrate, 0.1% Tween-20 (Tw) at 65°C for 30 min followed by two washes in MABT (100 mM maleic acid, 150 mM NaCl, pH 7.5, 0.1% Tw) for 30 min at room temperature. Sections were blocked for 2 to 4 hr at room temperature in MABT containing 20% sheep serum (Sigma) and 2% blocking reagent (Boehringer Mannheim). The blocking solution was then replaced with blocking solution containing a 1:1,500–3,000 dilution of alkaline phosphatase–conjugated Fab fragments of sheep anti-DIG antibodies (Boehringer Mannheim), a coverslip was placed over the antibody solution, and the slides were incubated overnight at 4°C in a humidified box. The slides were washed a minimum of 5 times in MABT for 20 min at room temperature, twice in staining buffer (100 mM NaCl, 50 mM MgCl2, 100 mM Tris pH 9.5 and 0.1% Tw) and incubated overnight in staining buffer plus 10% v/vw polyvinyl alcohol (molecular weight, 70,000–100,000; Sigma) containing 4.5 μl of nitroblue tetrazolium (NBT) and 3.5 μl of 5-bromo-4-chloro-3-indolyl phosphate (BCIP)/ml (Boehringer Mannheim) in the dark at room temperature. The slides were washed in distilled water, dehydrated to 100% ethanol, cleared with Xylene, and mounted in DPX (Fluka). No cross-reactivity between probes within a gene family was observed at the stringency used on sections of stage 13–17 or stage 25 chick embryos (data not shown).
Sections were prepared as described above. Sections were rehydrated in PBS, incubated in blocking solution (20% goat serum in 50 mM Tris buffer [pH 7.4], 10 mM lysine, 145 mM NaCl, and 1% bovine serum albumin) for 30 min, incubated overnight at 4°C in primary antibody diluted in blocking solution with 0.1% Tween (HNK1 supernatant, 1:20; anti-calponin mouse ascites, 1:500 [Sigma]; anti-SMA mouse ascites, 1:400 [Sigma]). The sections were washed multiple times in PBS plus 0.1% Tween (PBS-Tw) and then incubated in fluorescein-conjugated goat anti-mouse immunoglobulin (IgG+IgM; Jackson, diluted 1:100 in blocking solution plus 0.1% Tw). After washing in PBS-Tw, the coverslips were mounted onto the glass slides in 50% glycerol:50% Tris (pH 8.3) and examined in a Zeiss Axioskop fluorescence microscope.
E6, E7, and E8 chick eyes were dissected and the approximate anterior third of the eye was cut away from the posterior portion, the vitreous was carefully removed, and the anterior portion was fixed in 4% paraformaldehyde in PBS overnight at 4°C. The tissue was washed a minimum of two times for 30 min in 100mM Tris-HCl (pH 7.5), 150 mM NaCl, 0.1% Tw (TTBS) and then heated to 65°C for 30 min in TTBS (to neutralized endogenous alkaline phosphatases), allowed to cool to room temperature, and then incubated in 10% goat serum (GS)/TTBS plus 0.5% Triton X-100 (TTBST) for 4 hr at room temperature. The tissue was incubated overnight in HNK1 (supernatant diluted 1:100 in TTBST) at 4°C. The tissue was washed a minimum of five times in TTBST for 1 hr each wash, then incubated 2 hr in 10% GS/TTBST, then incubated overnight at 4°C with the secondary antibody (alkaline phosphatase–conjugated goat anti-mouse IgM; diluted 1:5,000 in 10% GS/TTBST). The tissue was washed three times, 90 min each, in TTBST, then washed 45 min in 100 mM Tris-HCl (pH 9.5), 150 mM NaCl, 25 mM MgCl2, 0.5% Triton X-100 (CT). Tissue was washed in 5% PVA/CT for 1 hr, in 10% PVA/CT for 1 hr, and then incubated in 10% PVA/CT containing 4.5 μl of NBT and 3.5 μl of BCIP/ml (Boehringer Mannheim) in the dark at room temperature for approximately 3 hr. The tissue was briefly washed twice in TTBST, washed overnight at 4°C in TTBST. The tissue was dehydrated through a methanol series, then washed in CMFT (0.8 g/L NaCl, 0.02 g/L KCl, 0.115 g/L Na2HPO4, 0.02 g/L KH2PO4, 0.02 g/L ethylenediaminetetraacetic acid, 0.1% Tw). The tissue was cleared in a graded series of glycerol-CMTF until 80% glycerol. All washes and incubations, except for the color reaction, were on a rocking platform.
Photos were taken with a Zeiss Axioskop or Zeiss Axioplan fluorescence equipped microscope. Photographic slides were scanned into Photoshop (Adobe).
I thank the following people for generously providing cDNA clones: Martyn Goulding (Pax3 and Pax6), Philippa Francis-West (BMP2 and BMP4), Brian Houston (BMP6 and BMP7), and Tom Jessell (slug). I thank Jim Weston and members of his laboratory for their help and advice. I also thank Kate Barald for her helpful comments on the manuscript. I am especially grateful to Monte Westerfield, in whose laboratory this study was conducted. A.M.J. was funded by a NRSA fellowship, and Monte Westerfield was funded by the NIH.