Fate Maps of Neural Crest and Mesoderm in the Mammalian Eye

Home , Cardiac neural crest, Eye development, Mesoderm, Optic cup (embryology), Optic stalk

Philip J. Gage,1,2 William Rhoades,1 Sandra K. Prucka,3 and Tord Hjalt4

PURPOSE. Structures derived from periocular mesenchyme arise mesenchyme is to provide multiple mature cell lineages that by complex interactions between neural crest and mesoderm. are necessary for normal ocular development and vision, in- Defects in development or function of structures derived from cluding the corneal endothelium and stroma, trabecular mesh- periocular mesenchyme result in debilitating vision loss, in- work, Schlemm’s canal, sclera, ciliary body muscles, iris cluding glaucoma. The determination of long-term fates for stroma, extraocular muscles, and ocular blood vessels. A sec- neural crest and mesoderm in mammals has been inhibited by ond essential function of the periocular mesenchyme during the lack of suitable marking systems. In the present study, the ocular development is to provide essential signals for pattern- first long-term fate maps are presented for neural crest and ing of ocular ectoderm primordia, including induction of lac- mesoderm in a mammalian eye. rimal glands from the surface ectoderm, specification of retinal METHODS. Complementary binary genetic approaches were pigmented epithelium from the optic cup, and differentiation used to mark indelibly the neural crest and mesoderm in the of the optic stalk from the neural ectoderm.1–3 Genetic or developing eye. Component one is a transgene expressing Cre acquired defects in development or function of the periocular recombinase under the control of an appropriate tissue-spe- mesenchyme result in ocular disease and vision loss, including cific promoter. The second component is the conditional Cre particularly glaucoma.4,5 reporter R26R, which is activated by the Cre recombinase Historically, the periocular mesenchyme was thought to expressed from the transgene. Lineage-marked cells were arise developmentally from the mesoderm. However, fate counterstained for expression of key transcription factors. maps developed in birds by using quail chick chimeras, vital dye labeling, or neural crest-specific antibodies have demon- RESULTS. The results established that fates of neural crest and strated that periocular mesenchyme actually receives initial mesoderm in mice were similar to but not identical with those 6–10 in birds. They also showed that five early transcription factor contributions from both neural crest and mesoderm. These genes are expressed in unique patterns in fate-marked neural studies further established that the corneal endothelium and crest and mesoderm during early ocular development. stroma, trabecular meshwork, most of the sclera, and the ciliary muscles in birds are derived solely from the neural crest. CONCLUSIONS. The data provide essential new information to- In contrast, all vascular endothelium, the caudal region of the ward understanding the complex interactions required for nor- sclera, Schlemm’s canal, and extraocular muscles are derived mal development and function of the mammalian eye. The from the mesoderm. These results have generally served well results also underscore the importance of confirming neural as the model for all vertebrates, including mammals. However, crest and mesoderm fates in a model mammalian system. The the documentation of important developmental differences complementary systems used in this study should be useful for between birds and mammals has become increasingly com- studying the respective cell fates in other organ systems. (In- mon.11,12 In the neural crest, these include differences in the vest Ophthalmol Vis Sci. 2005;46:4200–4208) DOI:10.1167/ timing of neural crest migration, the migratory pathways taken, iovs.05-0691 and their ultimate fates,11 making it essential to uncover any species differences that may exist in mammals to account he vertebrate eye is constructed during development from accurately for the embryonic origins of each mature structure. Tthree general embryonic precursor sources: neural ecto- Several key regulators of periocular mesenchyme development derm, surface ectoderm, and a loose array of cells termed the have recently been identified.2,13–17 Determining the neural periocular mesenchyme. A primary function of the periocular crest and mesoderm expression patterns, as well as the lineage- specific effects of genetic lesions in these genes, would significantly enhance our understanding of the normal mechanisms From the Departments of 1Ophthalmology and Visual Sciences, by which these genes function during ocular development. 2Cell and Developmental Biology, and 3Human Genetics, University of The major impediment to addressing any of these questions Michigan Medical School, Ann Arbor, Michigan; and the 4Department has been the lack of a reliable strategy for quantitative, long- of Cell and Molecular Biology, Lund University, Lund, Sweden. term labeling of neural crest and mesoderm in a mammalian Supported by a University of Michigan UROP Summer Biomedical eye. Fortunately, the use of two complementary Cre-lox based Research Fellowship (WR), the Glaucoma Research Foundation (PJG), approaches has recently provided the necessary breakthrough. National Eye Institute Grants EY014126 and EY07003 (PJG), Research A binary transgenic system has been developed that allows to Prevent Blindness (PJG), the Swedish Science Council (TH), and for indelible, permanent labeling of early presumptive neural National Institute of Child Health and Development Grant 34283 to 18 Sally Camper. crest and all subsequent progeny cells. The first component Submitted for publication June 2, 2005; revised July 1, 2005; of this system is a Wnt1-Cre transgene, which expresses Cre accepted August 30, 2005. recombinase under the control of the promoter for the Wnt1 Disclosure: P.J. Gage, None; W. Rhoades, None; S.K. Prucka, gene. Like Wnt1 itself, Cre expression from the transgene is None; T. Hjalt, None transient and limited to a 24-hour period in early presumptive The publication costs of this article were defrayed in part by page neural crest cells before their emigration from the dorsal neural charge payment. This article must therefore be marked “advertise- tube.18 The second component of the system is the conditional ment” in accordance with 18 U.S.C. §1734 solely to indicate this fact. Cre reporter allele, R26R.19 Cre-mediated activation of ␤-galac- Corresponding author: Philip J. Gage, Department of Ophthalmol- ␤ ogy and Visual Sciences, University of Michigan Medical School, 350 tosidase ( -gal) expression from the R26R reporter allele by Kellogg Eye Center, 1000 Wall Street, Ann Arbor, MI 48105; DNA recombination provides for indelible, long-term labeling [email protected]. of all cells derived from the neural crest.19 This system has

Downloaded from iovs.arvojournals.org on 10/01/2021 IOVS, November 2005, Vol. 46, No. 11 Ocular Mesenchyme Fates in Mammals 4201

been used to map neural crest fates in cardiovascular, cranio- mM potassium ferrocyanide, 0.1% X-gal in wash buffer), rinsed in wash facial, and skull vault development.20–22 buffer, embedded in JB-4 resin, and sectioned at 3 ␮m. Mounted The mouse glycoprotein hormone ␣-subunit gene (␣GSU)is sections were counterstained with nuclear fast red (Sigma-Aldrich, St. expressed initially in Rathke’s pouch, the oral-ectoderm–de- Louis, MO). rived primordium of the anterior pituitary gland, and subsequently in multiple lineages of the mature anterior pituitary Immunohistochemistry gland.23 ␣GSU expression has also been reported in mesoderm of the early ocular primordium and in the olfactory epithe- Embryos harvested for immunostaining were fixed for 2 to 4 hours lium.24 The promoter regulatory elements required to recapit- with 4% paraformaldehyde in PBS, washed and dehydrated, and em- ulate this expression pattern have been identified and used to bedded in paraffin. Mounted sections were deparaffinized and treated express Cre recombinase in transgenic mice.25 Mice doubly for antigen retrieval by boiling for 10 minutes in citrate buffer (pH 6.0). ␣ Immunostaining was performed according to standard methods. transgenic for GSU-Cre and a Cre-responsive reporter cassette ␤ exhibit indelible expression of ␤-gal in a pattern consistent Briefly, sections were incubated with antibodies directed against -ga- with expression of endogenous ␣GSU itself.25 ␣GSU-Cre has lactosidase (Eppendorf-5Prime, Boulder, CO), FOXC1 or -2 (Abcam, also been used successfully to generate a lineage-specific gene Inc., Cambridge, MA), myogenin (MYOG; clone F5D, developed by knockout of the transcription factor gene Dax1 in the anterior Woodring Wright and obtained from NICHD/Developmental Hybrid- 26,27 oma Studies Bank, Iowa City, IA), PITX1 (a gift from Jacques Drouin, pituitary gland. 28 29 In the current study, we used the complementary Wnt1- Montreal, Canada ) or PITX2 followed by biotinylated species-spe- Cre/R26R and ␣GSU-Cre/R26R labeling systems to establish cific secondary antibodies (Jackson ImmunoResearch, West Grove, for the first time the long-term fates of neural crest and meso- PA). Signals were detected using tyramide signal-amplification kits derm, respectively, in a model mammalian eye. The fates were (PerkinElmer, Boston, MA). similar to but not identical with those previously reported in birds, with the most significant differences occurring in the ESULTS anterior segment. We also demonstrated that five known or R potential transcriptional regulators of periocular mesenchyme Contributions of Cranial Neural Crest and have unique expression patterns in the neural crest and meso- Anterior Paraxial Mesoderm to the Early derm during early ocular development. Finally, the data also imply that the mechanism of activation for the homeobox gene Ocular Primordia Pitx2 in the neural crest and mesoderm is likely to be distinct. We crossed Wnt1-Cre males with R26R females to generate These findings have important implications for ocular develop- Wnt1-Cre;R26R progeny (neural-crest–marked) with indelibly ment and function and may provide a model for understanding labeled neural crest by virtue of Cre activation of ␤-gal reporter interactions between the neural crest and mesoderm in other expression. In parallel experiments, we crossed ␣Gsu-Cre and organ systems. R26R mice to generate analogous ␣Gsu-Cre;R26R mice (mesoderm-marked) with ␤-gal-labeled ocular mesoderm. In either system, the presence and location of lineage-marked ␤-galϩ MATERIALS AND METHODS cells was detected by incubation with X-gal, a chromogenic Animals and Isolation of Lineage-Marked Mice substrate of ␤-gal, or by immunohistochemistry with an anti-␤- gal antibody. Wnt1-Cre mice transmit a transgene consisting of a cDNA cassette X-gal staining of whole embryos from each class at E10.5 expressing Cre protein placed under the transcriptional control of the demonstrated the distinct labeling specificity of each binary 18 ␣ murine Wnt1 promoter. Gsu-Cre transgenic mice (TgN(Cga- labeling system. The heads of neural-crest–marked embryos cre)S3SAC, cat no. 004426; Jackson Laboratories, Bar Harbor, ME) were heavily invested with ␤-galϩ cells by this time point, transmit a construct including the cDNA for a nuclear-localized Cre particularly in the branchial arches and trigeminal ganglia (Fig. protein under the transcriptional control of the pituitary glycoprotein 1A). There was also heavy labeling of the midbrain neural ␣ 25 subunit promoter. R26R mice were purchased from Jackson Lab- ectoderm, as has been reported.18 In the trunk, there was oratories and transmit a ␤-galactosidase reporter allele that is geneti- 19 substantial staining of the cardiac outflow tract and the dorsal cally activated in vivo by Cre recombinase activity. All procedures root ganglia along either side of the neural tube in neural-crest– using mice were approved by the University of Michigan Committee marked embryos (Fig. 1A). In contrast, head staining in ␣Gsu- on Use and Care of Animals. All experiments were conducted in Cre;R26R embryos at E10.5 was limited to a wedge of mesen- accordance with the principles and procedures outlined in the NIH chyme located ventrally and caudally to the optic cup and the Guidelines for the Care and Use of Experimental Animals and in the oral ectoderm. These positions are analogous to those reported ARVO Statement for the Use of Animals in Ophthalmic and Vision in ␣Gsu-lacZ transgenic mice and in endogenous ␣Gsu expres- Research. sion (Fig. 1B).24 ␣Gsu-Cre;R26R embryos also showed heavy Timed pregnancies were produced by mating homozygous R26R staining of the somites, atria and ventricles of the heart, and gut females either with Wnt1-Cre transgenic males to produce progeny mesentery (Fig. 1B). In histologic sections, classic neural-crest– ␣ with marked neural crest or with Gsu-Cre transgenic males to pro- derived structures such as the trigeminal ganglia, meninges, duce progeny with marked mesoderm. The morning a plug was de- and presumptive calvarial vault were heavily stained in neural- tected was designated as embryonic day (E)0.5. Embryos were col- crest–marked embryos (Figs. 1C, 1E) but were unstained in lected by cesarian section after the mother was euthanatized. All mesoderm-marked embryos (Figs. 1D, 1F). The dorsal neural genetic loci were genotyped by using PCR-based methods with prim- 25 19 tube, which was labeled in neural-crest–marked embryos (Fig. ers for Cre and R26R. 1E), was unstained in mesoderm-marked embryos (Fig. 1F). ␤ Collectively, these data are consistent with the Wnt1-Cre;R26R Detection of -Galactosidase Activity and ␣Gsu-Cre;R26R marking systems labeling mesenchyme Enucleated whole eyes from 6-week-old mice were fixed for 1 to 2 derived from neural crest and mesoderm, respectively. hours at 4°C with 4% paraformaldehyde in phosphate-buffered saline We next examined each binary system for ␤-gal labeling in (PBS) and rinsed three times in wash buffer (0.1 M sodium phosphate the developing eye at the cellular level. Labeled neural crest

[pH 8.0], 2 mM MgCl2, 2% NP-40). Fixed eyes were stained overnight cells in Wnt1-Cre;R26R embryos surrounded the optic cup and at 37°C in standard staining solution (5 mM potassium ferricyanide, 5 stalk at E10.5 (Fig. 1G). The neural crest also occupied the

Downloaded from iovs.arvojournals.org on 10/01/2021 4202 Gage et al. IOVS, November 2005, Vol. 46, No. 11

unique morphology of these cells relative to surrounding mesenchyme, the previous mapping of cranial paraxial mesoderm to this location at this time point, and the absence of labeling in Wnt1-Cre;R26R embryos all argue that the labeled cells in the ocular primordia of ␣GsuCre;R26R embryos represent cranial paraxial mesoderm. Thus, the two lineage-specific ␤-gal labeling systems resulted in unique staining patterns in the early ocular primordia that were consistent with the previously reported arrangement of neural crest and mesoderm precursors in the chick.6–10 Occasional labeled mesoderm cells were also present within the presumptive cornea, in the hyaloid space, and adjacent to the optic cup and stalk in ␣GsuCre; R26R embryos at this early time point (Fig. 1H and data not shown), an observation that is consistent with previous reports that mesoderm cells have already begun to intermingle with neural crest by this time point.30,31 By E12.5, neural crest and mesoderm were extensively comingled in multiple ocular structures (data not shown and see Figs. 3, 4, and 5). Transcription Factor Expression in Marked Neural Crest and Mesoderm Molecular markers that are specifically expressed in neural crest or mesoderm precursors have not been described for the mammalian ocular primordia. Therefore, we examined the protein expression patterns of several transcription factors previously implicated in ocular development to determine their expression profiles between the two embryonic precursor pools. The homeodomain transcription factor PITX2 is associated with ocular defects in humans and is essential for ocular development in mice.2,13,32,33 PITX2 expression labeled the neural crest, beginning initially in the presumptive anterior segment and quickly extending to the periphery of the optic cup in Wnt1Cre;R26R embryos by E11.5 (Fig. 2A). In contrast, PITX2 expression had marked all ␤-gal-labeled mesoderm in ␣GsuCre;R26R embryos at E11.5, before these cells entered the ocular field (Fig. 2B). By E12.5, PITX2 expression had spread to all ocular neural crest (data not shown). The forkhead transcription factor FOXC1 is also essential for POM development in mice and humans.16,34 FOXC1 expression marked the neural-crest–derived cells in the anterior segment and surrounding the optic cup and stalk at E11.5 (Fig. 2C). FIGURE 1. Lineage-specific ␤-galactosidase expression in neural crest and mesoderm of Wnt1-Cre;R26R and ␣Gsu-Cre;R26R embryos. X-gal Colabeling was also identifiable in a small percentage of meso- staining of E10.5 neural crest (A) and mesoderm (B) labeled embryos derm cells at E11.5 (Fig. 2D). Therefore, PITX2 and FOXC1 reveals distinct ␤-gal expression patterns generated by two binary were each expressed in both the neural crest and mesoderm, labeling systems. Strong ␤-gal expression in the trigeminal ganglion of albeit to very different degrees in the mesoderm. The transcrip- neural crest (C) but not mesoderm (D) labeled embryos stained with tion factor FOXC2 is necessary in periocular mesenchyme for X-gal. The mesoderm-derived endothelial lining of the primary head normal anterior segment development in mice, but not in ␤ vein is labeled in mesoderm embryos (D, inset). Strong -gal expres- humans.17,35 In contrast to FOXC1, expression of FOXC2 was sion in the dorsal root ganglion of neural crest (E) but not mesoderm ૺૺ detectable only in neural-crest–labeled cells and was not (F) marked embryos stained with X-gal. The dorsal neural tube ( )is ␣ labeled in neural crest (E) but not mesoderm (F) marked embryos. present in mesoderm-labeled cells of Gsu-Cre;R26R embryos Complementary, nonoverlapping ␤-gal expression patterns detected at E11.5 (Figs. 2E, 2F). The homeodomain protein PITX1 is by immunohistochemistry in E10.5 ocular primordia of neural crest (G) related to PITX2, and the two genes are often coexpressed and and mesoderm (H) labeled embryos. X-gal-stained sections are coun- are functionally redundant in other tissues.36,37 However, mes- terstained with nuclear fast red to display all nuclei. TG, trigeminal enchymal PITX1 expression in the developing eye was limited ganglion; BA, branchial arches; DRG, dorsal root ganglia; C, cornea; to a subset of labeled mesoderm cells and was never observed CM, cephalic mesoderm; G, gut; H, heart; S, somites. in neural-crest–labeled cells at E11.5 (Figs. 2G, 2H). Consistent with a previous report, PITX1 expression was also detected in presumptive cornea, beginning immediately after separation of the early lens vesicle (Fig. 2H).38 Finally, we examined MYOG the lens vesicle from the surface ectoderm, and a few cells as a second marker for mesoderm-derived muscle precursors. appeared within the hyaloid space posterior to the lens. Mes- MYOG expression was limited to a subset of mesoderm-labeled enchymal cells located in a morphologically distinct wedge of cells in ␣GsuCre;R26R embryos and was not expressed in cells located ventrally and caudally to the optic cup and stalk neural-crest–labeled cells at E11.5 (Figs. 2I, 2J). Based on these did not label as neural crest in Wnt1-Cre;R26R embryos, sug- collective results, particularly the complete absence of FOXC2 gesting that these cells were of mesoderm origin (Fig. 1G). colabeling in mesoderm-marked embryos and the specificity of Indeed, the position of these cells corresponds well with that PITX1 and MYOG for subsets of mesoderm cells, we conclude of the previously reported rostral limit of cranial paraxial me- that, in the developing eye, the Wnt1Cre;R26R and ␣GsuCre; soderm.30,31 Cells in this morphologically distinct wedge are R26R systems specifically mark the neural crest and meso- prominently labeled in ␣GsuCre;R26R embryos (Fig. 1H). The derm, respectively.

Downloaded from iovs.arvojournals.org on 10/01/2021 IOVS, November 2005, Vol. 46, No. 11 Ocular Mesenchyme Fates in Mammals 4203

the fate of each embryonic precursor pool was determined based on cellular morphology. Connective fascia cells in perinatal and adult muscles were derived exclusively from the neural crest (Figs. 3C, 3E). In contrast, myotubules in the developing muscle (Fig. 3D) and myoﬁbers in the mature muscle (Fig. 3F) are derived exclusively from mesoderm. No ectopic labeling by either labeling system, such as neural crest labeling of muscle cells or mesoderm labeling of connective fascia cells, was ever observed in multiple speci- mens examined at any time point.

Fates in the Ocular Vasculature The hyaloid vasculature forms posterior to the lens as a transient, embryonic blood system that maintains the posterior lens and developing retina. Mesenchymal precursor cells were present within the hyaloid space by E11.5 and primitive blood vessels were present by E12.5 (Figs. 4A, 4B). Both neural crest and mesoderm-derived cells contributed to the early hyaloid vessels (Figs. 4A, 4B). Subsequently, endothelial cells lining the mature hyaloid blood vessels were derived solely from the mesoderm (Figs. 4D, 4F). Pericytes attaching themselves to the unstained endothelial tube were derived exclusively from the neural crest (Figs. 4C, 4E). Endothelial and smooth muscle cells of the hyaloid artery were derived from the mesoderm and neural crest, respectively (data not shown).

FIGURE 2. Coexpression of transcription factors and ␤-gal in fate- marked embryos. E11.5 Wnt1Cre;R26R and ␣Gsu-Cre;R26R embryos were immunostained for expression of ␤-gal (green) to display neural crest (A, C, E, G, I) and cephalic mesoderm (B, D, F, H, J), respectively. Sections were counterstained (red) to detect the expression of PITX2 (A, B), FOXC1 (C, D), FOXC2 (E, F), PITX1 (G, H), and MYOG (I, J). Labeling in (A) provides orientation for all panels. NC, neural crest; CM, cephalic mesoderm; C, cornea; L, lens; OC, optic cup; OS, optic stalk. FIGURE 3. Detection of ␤-gal by immunohistochemistry in developing and mature extraocular muscles. ␤-gal expression (green) in presump- Fates in the Extraocular Muscles tive extraocular muscles (PM) of E12.5 neural crest (A) and mesoderm ␤ The early primordia of extraocular muscles were apparent as (B) marked embryos. -gal expression labels connective fascia cells (arrowheads) in developing (C) and mature (E) extraocular muscles of mesenchymal condensations lateral to the optic cup and neural-crest–marked mice. ␤-gal expression labels myotubules and stalk by E12.5 (Figs. 3A, 3B). A small number of neural crest myofibers (arrows) in developing (D) and mature (F) extraocular and a significant number of mesoderm precursor cells are muscles of mesoderm marked mice. Adult micrographs (E, F) are already present within the presumptive muscle condensa- presented as the merge of red and green channels to enhance visual- tions at this early time point (Figs. 3A, 3B). Subsequently, ization of the myofibers.

Downloaded from iovs.arvojournals.org on 10/01/2021 4204 Gage et al. IOVS, November 2005, Vol. 46, No. 11

order of appearance, the corneal endothelium and stroma, ciliary muscles, ciliary blood vessels, anterior iris, and trabecular meshwork and Schlemm’s canal within the iridocorneal angle. Neural crest cells migrated into the presumptive anterior segment between the anterior face of the newly formed lens vesicle and the surface ectoderm by E10.5 (Fig. 1G). Mesoderm cells followed 12 to 24 hours later (data not shown). By E14.5, neural-crest–derived cells comprised most of the cells within the endothelium and stroma of the cornea, and the presumptive iridocorneal angle (Fig. 5A). Although significantly fewer in number, mesoderm-derived cells were also consistently present in the same structures (Fig. 5B). The presence and relative ratios of the two embryonic cell lineages remain unchanged as these structures proceeded through differentiation (Figs. 5C, 5D). To detect ␤-galϩ cells while preserving the integrity of mature anterior segment structures, we stained enucleated adult eyes in X-gal and then processed them for plastic-embedded sections. The relative contributions of neural crest and mesoderm after this processing was unchanged from that observed with fluorescent immunostaining on samples derived from earlier time points, indicating efficient penetration of the stain. Most cells in the endothelium and stroma layers of the mature cornea expressed ␤-gal in neural-crest–marked animals, establishing a neural crest origin (Fig. 5E). However, cells that did not express ␤-gal were also consistently present in both layers of neural-crest–marked embryos (Fig. 5E), indicating that not all cells in these layers derived from neural crest in mice. A comparable number of ␤-galϩ cells were consistently present in the mature corneal endothelium and stroma of mesoderm- marked mice (Fig. 5F). The number of ␤-galϩ cells in mesoderm-marked mice parallels the number of unlabeled cells in neural-crest–marked mice. Collectively, these data indicate that most of the corneal endothelial and stromal cells are derived from neural crest but additional cells in both layers are derived from mesoderm. This finding was unexpected, because only cells of neural crest origin have been reported to be present in the corneal endothelium and stroma of birds.6–10 Contributions of the neural crest and mesoderm to the limbal region and iridocorneal angle of the mature eye were complex and highly variable. Ciliary muscles were derived FIGURE 4. Neural crest and mesoderm fates in ocular blood vessels. Immunohistochemical detection of ␤-gal expression (green) in eyes from the neural crest but not the mesoderm (Figs. 5I–L). The from neural crest–and mesoderm-marked animals. Neural crest (A) and endothelial lining of Schlemm’s canal was derived from the mesoderm (B) precursor cells are present within the hyaloid space by mesoderm (Figs. 5K, 5L). Most cells within the trabecular E12.5. Pericytes (C, E, arrowheads) in mature hyaloid vessels are meshwork labeled as neural crest, but a measurable number derived from neural crest, whereas the endothelial lining (D, F, ar- did not (Figs. 5I, 5J). Similar to observations in the cornea, rows) of mature hyaloid is derived from mesoderm. Neural crest (G, labeling of some trabecular meshwork cells in mesoderm-la- arrowheads) and mesoderm (H, arrows) have analogous fates in the beled mice indicates that these cells derived from the meso- mature choroid vasculature. All micrographs presented as the merge of derm (Figs. 5K, 5L). We observed little to no labeling for neural red and green channels to facilitate visualization of the blood vessels crest in the iris stroma (Figs. 5M, 5O). In contrast, iris stroma and the location of red blood cells, which are autofluorescent (red/ orange). Hy, hyaloid blood vessels; L, lens; R, retina; PE, pigmented regularly labels as mesoderm (Figs. 5N, 5P). epithelium; S, sclera; RBC, red blood cells.

DISCUSSION The choriocapillary bed comprises a network of microves- sels that immediately overlie the retinal pigmented epithelial We used binary transgenic systems based on Cre/loxP technol- cells and provide vascular functions to the RPE and underlying ogy to label speciﬁcally the neural crest or mesoderm in eyes of photoreceptors of the retina. Fates of the embryonic precursor mouse as a model vertebrate. Early indelible labeling of the pools in the choroid were identical with those in the hyaloid. respective embryonic precursor cell lineages by the two com- Pericytes and other associated cells were derived from the plementary systems allowed their fates to be followed through- neural crest (Fig. 4G), whereas endothelial cells were derived out development and in mature mouse eyes. The data demon- from the mesoderm (Figs. 4H). Endothelial cells of blood ves- strate that neural crest and mesoderm fates in mammalian eyes sels within the ciliary body were similarly derived from the are similar to but not identical with those previously reported mesoderm (Fig. 5H). in birds.6–10 We also report the neural crest and mesoderm expression patterns of several key transcription factors during Fates in Anterior Segment Structures early ocular development. Collectively, these results have im- Multiple cell lineages within the anterior segment of the eye portant mechanistic implications for understanding the devel- were derived from mesenchyme, including, in approximate opment and function of structures derived from periocular

Downloaded from iovs.arvojournals.org on 10/01/2021 IOVS, November 2005, Vol. 46, No. 11 Ocular Mesenchyme Fates in Mammals 4205

FIGURE 5. Complex neural crest and mesoderm fates in the ocular anterior segment. Immunohistochemistry (green) and X-gal (blue) were used to detect ␤-gal expression in neural crest (NC) and mesoderm (M). Neural crest–and mesoderm-derived cells were both present in the developing anterior segment of E14.5 (A, B) and P1 (C, D) eyes. Neural crest and mesoderm contributions to the mature corneal endothelium and stroma (E, F), ciliary body (G, H), iridocorneal angle (I–L), and iris (M–P) of adult eyes. Note ␤-galϪ cells (∧) in corneal stroma and endothelium of neural crest-marked animals (E). Limbus (ૺ), Schlemm’s canal (open arrows), trabecular meshwork (closed arrows), iris stroma (arrowheads). X-gal stained eyes are counterstained with nuclear fast red to display all nuclei. S, corneal stroma; E, corneal endothelium; ICA, iridocorneal angle.

mesenchyme, and the roles mediated by specific transcription cial, and skull vault development further established that this is factors in these processes. a robust approach for quantitative, specific, and long-term fate mapping of neural crest in mice.20–22 Our results confirm and Characterization of the Cell-Labeling Systems extend these conclusions. The ability to accurately track cell fates throughout develop- Our data also identify the ␣Gsu-Cre;R26R combination as a ment and into adulthood depends on the fidelity and perma- similarly robust and specific system for labeling mesenchyme nence of the labeling system used.39 Neural crest and meso- populations derived from mesoderm. The cells marked in derm fates have been extensively characterized in birds by ␣Gsu-Cre;R26R embryos are distinct from those in comparable using quail chick chimeras, vital dye labeling, and antibody neural crest-labeled Wnt1-Cre;R26R embryos during early oc- staining for neural-crest–specific markers.6–10,40,41 However, ular development, when the two embryonic precursor popu- these labeling systems are not effective for determining partic- lations have not yet mixed. The location of cells labeled in ularly the long-term fates of these lineages in mammals.39 ␣Gsu-Cre;R26R embryos at E11.5 corresponds precisely with Activation of the R26R reporter construct by Cre-mediated the previously established position of cephalic paraxial meso- DNA recombination in the two systems used in the study derm.30,31 The expression patterns of the transcription factors ensured indelible ␤-gal labeling of the precursor cells in which FOXC2, PITX1, and MYOG also distinguish between the ␤-gal- Cre was originally expressed, as well as in all subsequent marked populations in Wnt1-Cre;R26R versus ␣Gsu-Cre;R26R progeny cells. The Wnt1-Cre;R26R labeling system was initially embryos. It is unlikely that there was a previously unrecog- shown to generate specific, quantitative labeling of the mid- nized population of neural crest cells that was not marked by brain neural ectoderm and all premigratory neural crest.18 ␤-gal activation in Wnt1-Cre;R26R embryos. Therefore, these Subsequent experiments examining cardiovascular, craniofa- results are consistent with the ␤-gal-marked cells in ␣Gsu-Cre;

Downloaded from iovs.arvojournals.org on 10/01/2021 4206 Gage et al. IOVS, November 2005, Vol. 46, No. 11

FIGURE 6. Neural crest and mesoderm fates in bird and mammalian eyes. Cross-sectional diagrams summa- rizing similarities and differences in contributions of neural crest (red) and mesoderm (blue) to adult chicken and mouse eyes. The major differences occur in the anterior segment. Chicken fates are synthesized from previously published reports.6–10

R26R being of mesoderm origin. Future identification of addi- mammals. However, a small population of cells within each of tional molecular markers is likely to refine our understanding these structures consistently did not express ␤-gal in neural- of these cells further. crest–labeled animals, strongly suggesting that certain cells in The absence of staining in well-established neural-crest– these structures have a different origin. These cells were derived tissues, such as the trigeminal ganglia in the head and present by E12.5 and persisted into adulthood. In contrast, a dorsal root ganglia in the trunk, provides further evidence that comparable number of cells in each of these structures consis- the ␣Gsu-Cre;R26R system marks mesoderm and not neural tently labeled positively for ␤-gal in mesoderm-marked animals crest elsewhere as well, suggesting that this combination may throughout the same time period, consistent with a mesoderm have further utility in other mesoderm-derived tissues else- origin of these cells. Although the possibility of ectopic expres- where in the body. However, further analysis of lineage-spe- sion is always a concern when using transgenic mice, the cific markers within individual organ systems would be neces- absence of ectopic labeling in any other neural-crest–derived sary to confirm this preliminary interpretation. Although the structures or the dorsal neural tube, from which the neural ϩ source of rare ␤-gal cells in lens and surface ectoderm is not crest arises, strongly argues that these cells were not neural clear, PITX1 is capable of transactivating the same ␣GSU pro- crest. This, together with the demonstration that not all cells in moter in cell culture and PITX1 expression has been reported the affected structures labeled as neural crest, leads us to in the lens placode and early lens vesicle (Fig. 2H and Ref. 38). conclude that a minority population of cells within the corneal Therefore, it may be that expression of PITX1 ectopically endothelium and stroma, the limbus, and the trabecular mesh- activates the ␣Gsu-Cre transgene at a low frequency in lens or work are derived from the mesoderm. The identification of the surface ectoderm from which it derives. additional lineage-specific markers that are expressed at later time points is essential for further confirmation of this inter- Ocular Fate of Neural Crest and Mesoderm in pretation and to determine the mature identity of these cells. Birds Versus Mammals The underlying reason(s) for the surprising presence of Our results indicate that there is general conservation of neural mesoderm cells in multiple anterior segment structures in crest and mesoderm fates in the eye between birds and mam- mice, where they had not been reported in birds, remains mals and that there is frequently a division of labor within unknown. One possibility is that these cells are also present in individual structures where the neural crest and mesoderm the chick but are not recognized because they represent a contribute different mature lineages (Fig. 6).38 For example, in relatively small minority of cells in the relevant structures. This extraocular muscles the muscle fibers themselves are derived seems unlikely, given that multiple laboratories have reported from the mesoderm, whereas connective fascia cells arise from equivalent findings using different labeling approaches.6–10 the neural crest. Similarly, the endothelial lining of ocular Therefore, we propose a model wherein the mature corneal blood vessels arises from the mesoderm, whereas associated endothelial and stromal layers, the mesenchymal layer of the smooth muscle and pericytes derive from the neural crest. limbus, and the trabecular meshwork in mammals each consist Differentiation of ciliary muscles from the neural crest is also of two mature cell lineages, one of which is derived from the conserved between birds and mammals. neural crest and the other from the mesoderm. The lineages Despite the general conservation of neural crest and meso- derived from the mesoderm may not be present in homologous derm fates in the eye between birds and mammals, several structures in birds. This addition of new cells would represent important differences are apparent (Fig. 6). Of note, these all an evolutionary advance from birds to mammals. Attractive occur within anterior segment structures. The endothelial and candidates for the identity of at least some of the mesoderm- stromal layers of the cornea, as well as cells of the trabecular derived cells are a complex mix of antigen-presenting cells that meshwork are reported to derive solely from the neural crest in serve as immune sentinels and are present in mammalian eyes, birds.6,7,9,10 Endothelial cells lining Schlemm’s canal are the including murine eyes. In the limbus and peripheral cornea, sole anterior segment lineage reported to derive from the these include dendritic cells and macrophages.42–44 A similar mesoderm in birds.40 Our results indicate that most cells population of cell lineages has recently been demonstrated in within the corneal endothelial and stroma layers and the tra- the central cornea of mice as well, but the cells in this location becular meshwork are also derived from the neural crest in are generally immature, lacking the typical dendritic cell mor-

Downloaded from iovs.arvojournals.org on 10/01/2021 IOVS, November 2005, Vol. 46, No. 11 Ocular Mesenchyme Fates in Mammals 4207

phology and do not express major histocompatibility complex precursor cell populations. Anterior epithelial cells of the lens (MHC) class II antigen in normal, uninflamed eyes.45 These vesicle act as a signaling center that is needed in multiple steps cells probably arise from bone marrow, which is consistent during development of the anterior segment, including induc- with a mesoderm origin. It is intriguing that there have been no tion of the corneal endothelium and stroma layers.49 Pitx2 published reports of dendritic or macrophage cells’ having expression is activated within neural crest cells immediately been identified in chick eyes, and highly related Langerhans after their arrival within the nascent anterior segment. To- cells are reportedly absent from chick eyes.46 Taken together, gether with the lack of corneal endothelium and stroma spec- these observations suggest that cells within the dendritic and ification in Pitx2-deficient mice, this suggests a model in which macrophage lineages may account for at least some of the Pitx2 itself is an essential early target of the cornea-inducing mesoderm-derived cells that we identified in the anterior seg- signal(s) expressed by the anterior lens epithelia. Activation of ment of murine eyes. Recently, lymphohematopoietic lineages Pitx2 in wild-type mice, presumably together with additional have been shown to modulate pathogenicity in a murine model essential genes, results in the initiation of a genetic cascade of pigmentary glaucoma.47 Although the significance of these leading to specification and differentiation of corneal endothe- findings remains to be established for human disease, immune lium and stroma. Pitx2-deficient animals are genetically unable surveillance cells in the form of dendritic and Langerhans cells to express PITX2 protein, effectively resulting in uncoupling of are also found in the anterior segment of human eyes. There- the required genetic cascade and agenesis of the cornea. Fu- fore, the presence of mesoderm-derived dendritic cells may ture extension of this pathway by identification of the up- contribute to the underlying etiology of glaucoma. stream inducing signal(s), additional genes coinduced with Pitx2 in the neural crest, and the downstream effector genes Implications of Transcription Factor will provide essential insights into our understanding of cor- Expression Patterns neal development. In contrast to the neural crest, activation of Pitx2 in the mesoderm occurs before interaction of these cells The ability to identify lineage-specific expression patterns of with the ocular primordia and therefore, presumably, by a key regulatory and functional genes in the developing eye is an distinct mechanism. important advance, because it provides useful insights into We have illustrated the use of complementary systems for potential mechanisms of gene function and interpretation of long-term monitoring of neural crest and mesoderm cell fates mutant phenotypes. Demonstration that the five transcription in the mammalian eye. The subtle but important differences in factors we examined were expressed in overlapping but dis- cell fates between mature chick and mouse eyes underscore tinct patterns in the neural crest and mesoderm at E11.5 is the importance of determining whether such variations exist significant because it suggests that initial molecular patterning elsewhere in the body where the neural crest and mesoderm of the mesenchymal precursor cells is probably already in place interact during development. Differential expression of tran- at this early time point, even though morphologic differences scription factors within the two embryonic lineages in wild- were not yet discernible. It is particularly striking that FOXC1 type mice provides important new insights into the potential and PITX2 had the most similar expression patterns, because mechanisms of gene function in normal and abnormal ocular heterozygous mutations in the human FOXC1 or PITX2 genes development. Determining the lineage-specific expression pro- both result in Axenfeld-Rieger syndrome, an autosomal domi- files of additional important genes that are expressed in peri- nant condition including anterior segment defects and a high ocular mesenchyme should provide similarly useful insights. risk for glaucoma.33,34 This supports the hypothesis that the Mutant mice are now available for the genes encoding each of two factors probably regulate common downstream targets in these transcription factors14,16,17 and introduction of the two cells in which they are coexpressed. Our current demonstra- labeling systems onto each mutant genetic background will tion of wider Pitx2 expression in the developing eye is con- allow identification of potential lineage-specific changes in the sistent with the demonstration that complete loss of Pitx2 behavior of embryonic cell types and/or their derivatives. Fi- function in mice results in a more severe ocular phenotype nally, it will be possible, by using the Wnt1-Cre and ␣Gsu-Cre than loss of Foxc1. Foxc2 heterozygous mice exhibit ocular transgenes, to generate neural-crest–and mesoderm-specific phenotypes very similar to the Foxc1 heterozygous pheno- knockouts of relevant genes in the mammalian eye. For genes type.17 This is consistent with the two genes’ being coex- having complex expression patterns and mutant phenotypes, pressed in neural crest. this will allow parsing of specific features of the phenotype to Pitx2 is essential to specify correctly the multiple lineages a requirement for gene function either in neural crest or me- derived from the periocular mesenchyme, including the cor- soderm, and to distinguish intrinsic versus extrinsic effects. neal endothelium and stroma and the extraocular muscles.2,13,15,32,48 Our current demonstration that these struc- Acknowledgments tures are chimeric, receiving contributions from both the The authors thank Andy McMahon and Sally Camper for the gifts of neural crest and mesoderm, raises the intriguing mechanistic Wnt1-Cre and aGsu-Cre transgenic mice, respectively; Mitchell Gillett question of whether the requirement for Pitx2 function in for expert technical assistance; David Reed for graphic illustration; each affected structure lies in the neural crest or the meso- Amanda Evans, Sally Camper, Peter Hitchcock, and Anand Swaroop for derm. For example, specification of extraocular muscles may critical readings of the manuscript; and Sally Camper for support require Pitx2 for an intrinsic function in mesoderm precursors during the early stages of the work. for them to initiate their myogenic program. Alternatively, Pitx2 function may be necessary in the neural crest for expres- References sion of an upstream signaling molecule(s) that acts extrinsically on the mesoderm precursor cells. Neural crest- and mesoderm- 1. Fuhrmann S, Levine EM, Reh TA. Extraocular mesenchyme patterns the optic vesicle during early eye development in the em- specific knockouts will ultimately be necessary to address this bryonic chick. Development. 2000;127:4599–4609. question in each relevant tissue. 2. Gage, PJ, Suh H, Camper SA. Dosage requirement of Pitx2 for The distinct timing and location of initial Pitx2 expression development of multiple organs. Development. 1999;126:4643– in neural crest versus mesoderm prompts intriguing new in- 4651. sights into potential gene function in the developing cornea 3. Govindarajan V, Ito M, Makarenkova HP, Lang RA, Overbeek PA. and provides strong evidence that the mechanism of gene Endogenous and ectopic gland induction by FGF-10. Dev Biol. activation is likely to be different between the two embryonic 2000;225:188–200.

Downloaded from iovs.arvojournals.org on 10/01/2021 4208 Gage et al. IOVS, November 2005, Vol. 46, No. 11

4. Alward WL. Axenfeld-Rieger syndrome in the age of molecular 27. Zhao L, Bakke M, Krimkevich Y, et al. Steroidogenic factor 1 (SF1) genetics. Am J Ophthalmol. 2000;130:107–115. is essential for pituitary gonadotrope function. Development 2001; 5. Lines MA, Kozlowski K, Walter MA. Molecular genetics of Axen- 128:147–154. feld-Rieger malformations. Hum Mol Genet. 2002;11:1177–1184. 28. Lanctot, C, Gauthier Y, Drouin J. Pituitary homeobox 1 (Ptx1) is 6. Johnston MC, Noden DM, Hazelton RD, Coulombre JL, Coulombre differentially expressed during pituitary development. Endocrinol- AJ. Origins of avian ocular and periocular tissues. Exp Eye Res. ogy. 1999;140:1416–1422. 1979;29:27–43. 29. Hjalt TA, Semina EV, Amendt BA, Murray JC. The Pitx2 protein in 7. Noden DM. Periocular mesenchyme: neural crest and mesodermal mouse development. Dev Dyn. 2000;218:195–200. interactions. In: Jacobiec FA, ed. Ocular Anatomy, Embryology, 30. Trainor PA, PP Tam, Cranial paraxial mesoderm and neural crest and Teratology. Hagerstown, MD: Harper & Row; 1982:97–119. cells of the mouse embryo: co-distribution in the craniofacial 8. Le Douarin NM. The Neural Crest. Cambridge, UK: Cambridge mesenchyme but distinct segregation in branchial arches. Devel- University Press; 1982. opment 1995;121:2569–2582. 9. Le Douarin NM. Migration and differentiation of neural crest cells. 31. Trainor PA, Tan SS, Tam PP. Cranial paraxial mesoderm: regional- In: Moscona A, Monroy A, eds. Current Topics in Development. isation of cell fate and impact on craniofacial development in New York: Academic Press; 1980. mouse embryos. Development. 1994. 120:2397–2408. 10. Le Lievre CS, Le Douarin NM. Mesenchymal derivatives of the 32. Lu MF, Pressman C, Dyer R, Johnson RL, Martin JF. Function of neural crest: analysis of chimaeric quail and chick embryos. J Rieger syndrome gene in left-right asymmetry and craniofacial Embryol Exp Morphol. 1975;34:125–154. development. Nature. 1999;401:276–278. 11. Kulesa P, Ellies DL, Trainor PA. Comparative analysis of neural 33. Semina EV, Reiter R, Leysens NJ, et al. Cloning and characteriza- crest cell death, migration, and function during vertebrate embry- tion of a novel bicoid-related homeobox transcription factor gene, ogenesis. Dev Dyn. 2004;229:14–29. RIEG, involved in Rieger syndrome. Nat Genet. 1996. 14:392–399. 12. Barald K, Kelly M. From placode to polarization: new tunes in 34. Mears AJ, Jordan T, Mirzayans F, et al. Mutations of the forkhead/ winged-helix gene, FKHL7, in patients with Axenfeld-Rieger anom- inner ear development. Development. In press. aly. Am J Hum Genet. 1998;63:1316–1328. 13. Kitamura K, Miura H, Miyagawa-Tomita S, et al. Mouse Pitx2 35. Finegold DN, Kimak MA, Lawrence EC, et al. Truncating mutations deficiency leads to anomalies of the ventral body wall, heart, extra- in FOXC2 cause multiple lymphedema syndromes. Hum Mol and periocular mesoderm and right pulmonary isomerism. Devel- Genet. 2001;10:1185–1189. opment. 1999;126:5749–5758. 36. Charles MA, Suh H, Hjalt TA, Drouin J, Camper SA, Gage PJ. PITX 14. Pressman CL, Chen H, Johnson RL. LMX1B, a LIM homeodomain genes are required for cell survival and Lhx3 activation. Mol class transcription factor, is necessary for normal development of Endocrinol. In press. multiple tissues in the anterior segment of the murine eye. Gene- 37. Marcil A, Dumontier E, Chamberland M, Camper SA, Drouin J. sis. 2000;26:15–25. Pitx1 and Pitx2 are required for development of hindlimb buds. 15. Lin CR, Kioussi C, O’Connell S, et al. Pitx2 regulates lung asym- Development. 2003;130:45–55. metry, cardiac positioning and pituitary and tooth morphogenesis. 38. Chang W, KhosrowShahian F, Chang R, Crawford MJ. xPitx1 plays Nature. 1999;401:279–282. a role in specifying cement gland and head during early Xenopus 16. Kidson SH, Kume T, Deng K, Winfrey V, Hogan BL. The forkhead/ development. Genesis. 2001;29:78–90. winged-helix gene, Mf1, is necessary for the normal development 39. Stern CD, Fraser SE. Tracing the lineage of tracing cell lineages. of the cornea and formation of the anterior chamber in the mouse Nat Cell Biol. 2001;3:E216–E218. eye. Dev Biol. 1999;211:306–322. 40. Couly GF, Coltey PM, Le Douarin NM. The developmental fate of 17. Smith RS, Zabaleta A, Kume T, et al. Haploinsufficiency of the the cephalic mesoderm in quail-chick chimeras. Development. transcription factors FOXC1 and FOXC2 results in aberrant ocular 1992;114:1–15. development. Hum Mol Genet. 2000;9:1021–1032. 41. Tucker GC, Aoyama H, Lipinski M, Tursz T, Thiery JP. Identical 18. Danielian PS, Muccino D, Rowitch DH, Michael SK, McMahon AP. reactivity of monoclonal antibodies HNK-1 and NC-1: conservation Modification of gene activity in mouse embryos in utero by a in vertebrates on cells derived from the neural primordium and on tamoxifen-inducible form of Cre recombinase. Curr Biol. 1998. some leukocytes. Cell Differ. 1984;14:223–230. 8:1323–1326. 42. Dana MR. Corneal antigen-presenting cells: diversity, plasticity, 19. Soriano, P. Generalized lacZ expression with the ROSA26 Cre and disguise: the Cogan lecture. Invest Ophthalmol Vis Sci. 2004; reporter strain. Nat Genet. 1999;21:70–71. 45:722–727. 20. Chai Y, Jiang X, Ito Y, et al. Fate of the mammalian cranial neural 43. Hamrah P, Zhang Q, Liu Y, Dana MR. Novel characterization of crest during tooth and mandibular morphogenesis. Development. MHC class II-negative population of resident corneal Langerhans 2000;127:1671–1679. cell-type dendritic cells. Invest Ophthalmol Vis Sci. 2002;43:639– 21. Jiang X, Rowitch DH, Soriano P, McMahon AP, Sucov HM. Fate of 646. the mammalian cardiac neural crest. Development. 2000;127: 44. Hamrah P, Huq SO, Liu Y, Zhang Q, Dana MR. Corneal immunity 1607–1616. is mediated by heterogeneous population of antigen-presenting 22. Jiang X, Iseki S, Maxson RE, Sucov HM, Morriss-Kay GM. Tissue cells. J Leukoc Biol. 2003;74:172–178. origins and interactions in the mammalian skull vault. Dev Biol 45. Hamrah P, Liu Y, Zhang Q, Dana MR. The corneal stroma is 2002;241:106–116. endowed with a significant number of resident dendritic cells. 23. Japon, MA, M Rubinstein, MJ Low, In situ hybridization analysis of Invest Ophthalmol Vis Sci. 2003;44:581–589. anterior pituitary hormone gene expression during fetal mouse 46. Rodrigues MM, Rowden G, Hackett J, Bakos I. Langerhans cells in development. J Histochem Cytochem. 1994;42:1117–1125. the normal conjunctiva and peripheral cornea of selected species. 24. Kendall SK, Gordon DF, Birkmeier TS, et al. Enhancer-mediated Invest Ophthalmol Vis Sci. 1981;21:759–765. high level expression of mouse pituitary glycoprotein hormone 47. Mo JS, Anderson MG, Gregory M, et al. By altering ocular immune alpha-subunit transgene in thyrotropes, gonadotropes, and devel- privilege, bone marrow-derived cells pathogenically contribute to oping pituitary gland. Mol Endocrinol. 1994;8:1420–1433. DBA/2J pigmentary glaucoma. J Exp Med. 2003;197:1335–1344. 25. Cushman LJ, Burrows HL, Seasholtz AF, Lewandoski M, Muzyczka 48. Suh H, Gage PJ, Drouin J, Camper SA. Pitx2 is required at multiple N, Camper SA. Cre-mediated recombination in the pituitary gland. stages of pituitary organogenesis: pituitary primordium formation Genesis. 2000;28:167–174. and cell specification. Development 2002;129:329–337. 26. Zhao L, Bakke M, Krimkevich Y, et al. Hypomorphic phenotype in 49. Beebe DC, Coats JM. The lens organizes the anterior segment: mice with pituitary-specific knockout of steroidogenic factor 1. specification of neural crest cell differentiation in the avian eye. Genesis. 2001;30:65–69. Dev Biol. 2000;220:424–431.

Downloaded from iovs.arvojournals.org on 10/01/2021