Volume 246, Issue 8 p. 598-609
Research Article
Free Access

Retinal pigment epithelium expansion around the neural retina occurs in two separate phases with distinct mechanisms

Paula Bernice Cechmanek

Paula Bernice Cechmanek

Department of Cell Biology and Anatomy, Hotchkiss Brain Institute, Alberta Children's Hospital Research Institute, University of Calgary, Calgary, Alberta, Canada

Search for more papers by this author
Sarah McFarlane

Corresponding Author

Sarah McFarlane

Department of Cell Biology and Anatomy, Hotchkiss Brain Institute, Alberta Children's Hospital Research Institute, University of Calgary, Calgary, Alberta, Canada

Correspondence to: Sarah McFarlane, 2164 Health Sciences Building, 3330 Hospital Drive, NW, Calgary, AB, T2N 4N1. E-mail: [email protected]Search for more papers by this author
First published: 27 May 2017
Citations: 22


Background: The retinal pigment epithelium (RPE) is a specialized monolayer of epithelial cells that forms a tight barrier surrounding the neural retina. RPE cells are indispensable for mature photoreceptor renewal and survival, yet how the initial RPE cell population expands around the neural retina during eye development is poorly understood. Results: Here we characterize the differentiation, proliferation, and movements of RPE progenitors in the Zebrafish embryo over the period of optic cup morphogenesis. RPE progenitors are present in the dorsomedial eye vesicle shortly after eye vesicle evagination. We define two separate phases that allow for full RPE expansion. The first phase involves a previously uncharacterized antero-wards expansion of the RPE progenitor domain in the inner eye vesicle leaflet, driven largely by an increase in cell number. During this phase, RPE progenitors start to express differentiation markers. In the second phase, the progenitor domain stretches in the dorsoventral and posterior axes, involving cell movements and shape changes, and coinciding with optic cup morphogenesis. Significantly, cell division is not required for RPE expansion. Conclusions: RPE development to produce the monolayer epithelium that covers the back of the neural retina occurs in two distinct phases driven by distinct mechanisms. Developmental Dynamics 246:598–609, 2017. © 2017 Wiley Periodicals, Inc.


The retinal pigment epithelium (RPE) is a monolayer of pigmented cells lining the back of the neural retina. RPE cells sit on the basement membrane (Bruch's membrane) adjacent to the vascular meshwork of the choroid, and their interactions with photoreceptors of the outer retina are essential for photoreceptor function, maintenance, and survival (Strauss, 2005). RPE cells maintain water balance and ion exchange between the choroid and neural retina, absorb stray light, provide protection from damaging free radicals, and phagocytose photoreceptor outer segments (Martinez-Morales et al., 2004; Strauss, 2005). Defects in any of these functions lead to progressive photoreceptor neurodegeneration and loss. RPE dysfunction participates in the etiology of diseases such as retinitis pigmentosa and age-related macular degeneration, where devastating vision loss occurs (Housset et al., 2013; Strauss, 2005). While much is known about RPE function in the adult retina, little is known about the early development of the RPE and how the epithelium comes to wrap around the back of the neural retina. Such knowledge could benefit RPE cell transplant and stem cell therapies for the treatment of retinal degenerative disorders (Carr et al., 2013; Westenskow et al., 2014).

The vertebrate RPE shares its embryonic origin with the neural retina, with both tissues arising from the anterior neural plate (Esteve and Bovalenta, 2006; Martinez-Morales et al., 2004; Zuber et al., 2003). A network of transcription factors specifies the early eye field, which splits to form two bilateral eye primordia. These primorida evaginate from the neural tube to form the optic vesicles. Through interactions of the eye vesicles with surrounding ectoderm, mesoderm, and the neural retina, the RPE and optic stalk domains are specified (Fuhrmann, 2010). The distal-most region of the eye vesicle forms the neural retina, the dorsomedial region gives rise to the RPE, and the ventromedial region becomes the optic stalk. In vertebrates, including mouse and chick, the first known RPE specification genes to be expressed are microphthalmia associated transcription factor (Mitf) and orthodenticle homeobox 2 (Otx2). These genes are initially expressed throughout the entire optic vesicle and then become restricted to the prospective RPE at the onset of visual system homeobox 2 (Vsx2) expression in the neural retina, although notably in Zebrafish, mitf does not appear to be expressed in the neural retina and is expressed only in the presumptive RPE (Fuhrmann, 2010; Lane and Lister, 2012; Nguyen and Arnheiter, 2000). The cells within the domain specified as RPE expand around the neural retina to become a flat epithelial monolayer (Kwan et al., 2012). In Zebrafish, visible RPE melanin-based pigmentation occurs after this period of RPE expansion (Zhang et al., 2014).

We know little about the cellular behaviors that accompany RPE expansion. The coupling of cell-specific fluorescent transgenic lines with the transparent nature of the embryos makes Zebrafish an excellent development model to study RPE development and morphogenesis. Importantly, the structure of the vertebrate eye and the genes that drive its development are conserved across species (Fadool and Dowling, 2008; Gross and Perkins, 2008). Previous work in Zebrafish described some aspects of the timing of the spread of a presumptive population of RPE cells around the back of the neural retina (Heermann et al., 2015; Kwan et al., 2012; Li et al., 2000a; Schmitt and Dowling, 1994). To what extent the RPE progenitor domain undergoes morphogenesis earlier in development, how morphogenesis relates to RPE differentiation, and how the RPE expands over the entire extent of the eye is unknown.

A comprehensive temporal and spatial understanding of early RPE development is lacking both in Zebrafish and in other species. Thus, we identified molecular markers in order to follow Zebrafish RPE development from the time that RPE cells are first specified in the dorsomedial eye vesicle shortly after eye vesicle evagination from the neural keel through to expression of visible pigment. We propose a two-phase model for RPE development. In the first phase, the RPE progenitor domain expands both in cell number and in the area it occupies within the anteroposterior axis of the inner leaflet of the eye vesicle, in what appears to be a largely proliferation-independent manner. During this phase, RPE progenitors start to express differentiation markers. A second phase of RPE development then takes place, where the cuboidal RPE cells of the inner eye vesicle elongate as the inner eye vesicle leaflet is stretched around the neural retina. This elongation of the RPE cells occurs alongside the movement of neural retina progenitors of the ventral inner eye vesicle around the distal rim of the optic cup (Kwan et al., 2012; Picker et al., 2009).


RPE Progenitors are Detected Shortly After the Eye Vesicles Evaginate

An early marker of RPE in Zebrafish is the transcription factor microphthalmia-associated transcription factor-a (mitfa) (Lister et al., 2001; Lister et al., 2011; Thisse and Thisse, 2004), which turns on in the middle of optic cup morphogenesis at the 16 somite stage (ss) (17 hours postfertilization [hpf]). The transcription factor EC (tfec) gene, however, appears to turn on in presumptive RPE in the dorsoposterior eye vesicle of 8-ss (13-hpf) embryos (Lister et al., 2011), shortly after eye vesicle evagination from the neural keel has occurred. tfec is expressed in presumptive and definitive Zebrafish RPE cells (Lister et al., 2011; Miesfeld et al., 2015) and Tfec in mouse controls pigmentation of the RPE (Rowan et al., 2004). We wanted to understand the early events in RPE development from the initial specification of RPE precursors (8 ss/13 hpf) to the completion of RPE morphogenesis (26 ss/24 hpf), when a single-cell-layer epithelium surrounds the neural retina. Thus, we identified a number of markers that appeared to be expressed in developing vertebrate RPE and followed their expression in Zebrafish over time. Note that the orientation of the eye changes with brain and eye morphogenesis, so that the early 8-ss dorsoventral and anteroposterior axes become the mature (>26 ss) nasotemporal and dorsoventral axes, respectively (Picker et al., 2009). For simplicity, we use the early axis designations in our descriptions (unless otherwise specified).

We collected Zebrafish embryos every two somite stages from 8 to 20 ss and at the 26 ss and processed them for RNA in situ hybridization with antisense riboprobes. In Zebrafish, the eye vesicles start to evaginate at the 4 ss (11 hpf), elongate over the next 4 hr, and begin to invaginate around the forming lens at the 16 ss (17 hpf) (Ivanovitch et al., 2013; Kwan et al., 2012). Optic cup morphogenesis and RPE expansion over the retinal surface is complete by the 26 ss (24 hpf). The first of our markers to be expressed in the presumptive RPE are the two transcription factors, tfec and basic helix-loop-helix e40 (bhlhe40; also known as dec1/sharp2/stra13), at the 8 ss (13 hpf) and the 12 ss (15 hpf), respectively (Fig. 1). In dorsal views of whole-mount embryos, tfec mRNA appears at the 8 ss in a small domain in the posterior eye vesicle (Fig. 1A) and, potentially, neural crest cells (Lister et al., 2011). To better understand spatial aspects of the expression of tfec and bhlhe40, we generated transverse sections through the eye vesicle (Fig. 2). In section, the tfec+ domain is located in the dorsomedial leaflet of the vesicle in a group of closely associated, cuboidal-shaped cells (Fig. 2A,A′). Beginning at the 12 ss (15 hpf), bhlhe40 mRNA is observed in a domain that overlaps with tfec (Figs. 1B,F, 2F,F′). It is important to note that by 24 hpf, both genes are expressed in the RPE that covers the eye (Fig. 1E,I).

Details are in the caption following the image
RPE progenitors are detected as early as the 8 ss and expand over the developing neural retina between 8 and 26 ss (13–26 hpf). Dorsal views of Zebrafish embryos from 8 to 26 ss processed for RNA in situ hybridization for tfec and bhlhe40 mRNA. A–E: tfec mRNA is first detected in the dorsoposterior eye vesicle at the 8 ss (13 hpf) in presumptive RPE progenitor cells (A) and expands anteriorly (B,C) until the 16 ss (17 hpf). After the 16 ss, tfec signal spreads laterally as RPE progenitors wrap around the neural retina toward the lens (D,E). tfec mRNA is also detected in neural crest cells. F–I: bhlhe40 mRNA is first detected at the 12 ss in presumptive RPE progenitor cells (F), and its expression closely follows that of tfec mRNA throughout eye morphogenesis. A, anterior; e, eye; ec,; eye cup; ev, eye vesicle; le, lens; nc, neural crest cells; nk, neural keel; os, optic stalk; P, posterior. Scale bar = 100 μm.
Details are in the caption following the image
RPE progenitors expand over the developing neural retina in two phases. Lateral whole-mount views (A–I) and transverse sections either through the posterior eye vesicle (A′) or approximately midway through the eye vesicle (B′–I′; see schematic) of RNA in situ hybridization for tfec (A–E) and bhlhe40 (F–I). A–E: tfec expression in the inner leaflet of the dorsoposterior eye vesicle (A,A′) spreads anteriorly (A–C) and ventrally (B′,C′) (12–16 ss), and then dorsally (asterisk in D′) and further ventrally (arrows in D′,E′) around the distal rim of the eye cup (16–26 ss). RPE progenitors adjacent to the lens at 24 hpf do not express tfec (bars in E,E′), but do express bhlhe40 (I). F–I: bhlhe40 is first detected in dorsal RPE progenitor cells (F,F′) (12 ss) and follows a similar temporal expression pattern to that of tfec (G–I). A, anterior; D, dorsal; ev, eye vesicle; il, inner leaflet; L, lateral; le, lens; M, medial; nc, neural crest cells; nk, neural keel; of, optic fissure; ol, outer leaflet; os, optic stalk; P, posterior; pnr, presumptive nasal retina; ptr, presumptive temporal retina; V, ventral; ve, ventricle; y, yolk. Scale bar = 50 μm.

RPE Expansion Occurs in Two Phases

To understand the expansion of the initial RPE domain, we followed tfec and bhlhe40 expression in the eye vesicle over the period of optic cup morphogenesis (12–26 ss) (Figs. 1, 2). Of note, by the 12 ss (15 hpf), the eye vesicles have evaginated and elongated, and a furrow has moved anteriorly to reduce the connection between the eye and the brain, forming the optic stalk. By this stage, the tfec expression domain has expanded anteriorly within the dorsomedial eye vesicle (Figs. 1B, 2B), and also to some degree ventrally within the inner (medial) vesicle leaflet (Fig. 2B′). Between the 12 and 16 ss (15–17 hpf), both tfec and bhlhe40 expression domains continue to expand anteriorly toward the optic stalk (Figs. 1C,G, 2C,G) and also ventrally, as seen in transverse sections (Fig. 2C′,G′). In sections, it is clear that as the domain expands, the RPE progenitors remain in a monolayer sheet-like epithelium. At the 16 ss (17 hpf), RPE cells are still cuboidal in shape (Fig. 2C′,G′) and occupy the dorsal medial eye vesicle leaflet over most of its anteroposterior extent (Figs. 1C,G, 2C,G).

At the 16–18 ss, eye invagination around the developing lens starts (Kwan et al., 2012). During invagination, the presumptive temporal neural retina, which sits in the medial leaflet just ventral to the RPE expression domain, moves around the distal rim of the optic cup and comes to sit adjacent to the presumptive nasal neural retina in the lateral eye vesicle leaflet (Picker et al., 2009). Recent work argues that a similar, though more limited, movement of dorsomedial progenitors occurs around the distal rim of the dorsal eye vesicle (Kwan et al., 2012; Heermann et al., 2015). Concurrent with this movement of progenitors in the inner vesicle, we find extensive expansion of the RPE marker expression domains (Figs. 1D,E,H,I and 2D,E,H,I). Between the 16 and 26 ss, the bhlhe40+ RPE domain expands gradually around the dorsal (asterisk in Fig. 2H′), posterior (arrow in 1H), and ventral (compare arrows in Fig. 2H′ and 2I′, and for tfec compare arrows in Fig. 2D′ and 2E′) aspects of the eye vesicle. Coupled to this change in the morphology of the RPE domain is a dramatic elongation of individual RPE cells (Fig. 3E). Ultimately, by 24 hpf the distal edges of the bhlhe40 expression domain come to rest adjacent to the lens (arrowheads in Fig. 2I′), with RPE covering the entire eye (Fig. 2I). Notably, while the tfec and bhlhe40 domains are initially (12–18 ss) for the most part congruent, as verified by double in situ hybridization for tfec and bhlhe40 at the 14 ss (data not shown), the domains diverge somewhat by the 18–26 ss: bhlhe40 in situ label is adjacent to the lens at the 26 ss (Fig. 2I, arrowheads in 2I′), whereas tfec signal is absent from the RPE that abuts the lens (lines in Fig. 2E,E′).

Details are in the caption following the image
Tracking of individual RPE cells over the period of RPE expansion. A: In vivo time-lapse imaging of eyes shows RPE progenitors maintain their neighbors during RPE expansion. Tol2-tfec:egfp plasmid and transposase mRNA was injected into embryos and the eyes confocal-imaged between the 10 and 26 ss in a lateral orientation. Shown is an example of an F0 Tg(tfec:egfp) embryo; maximal optical projection of four 4-μm confocal z-sections with individual EGFP-positive cells tracked between the 10 and 18 ss. B: Schematic of individual cell trajectories (gray line t = 14 hpf; black line t = 24 hpf). C: Cells were tracked until 24 hpf, when they displayed an elongated, RPE-like morphology, at the distal rim of the optic cup. D: The average speed of movement (μm/hr) at different developmental stages from 15 to 20 hpf (N.B. the 15-hpf time point indicates the average speed of movement in the 14- to 15-hpf period). Numbers of cells are indicated and were pooled from movies of four independent F0 embryos. Error bars are standard deviation. Cells in the most posterior region of the eye vesicle show significant ventral movement at the same developmental time as those in the more anterior eye vesicle, but move more slowly. E: In transverse microtome sections, RPE progenitors (tfec mRNA- labeled cells) initially exhibit (shown at 12 ss) a cuboidal shape. By 24 hpf, RPE progenitors have dramatically elongated. F: Number of RPE progenitors at various stages as quantified in plastic sections. Number of retinas examined above bars. G: RPE cells tracked over the second phase of RPE expansion. Shown is a lateral view of an eye from an F2 Tg(tfec:egfp) embryo imaged from the 14 ss to 25 hpf; maximal optical projection of four 4-μm confocal z-sections (also see Movie 1). H: Schematic of RPE cell positions over time from 16 to 25 hpf. D, dorsal; nr, neural retina; V, ventral; ve, ventricle. A,G: Scale bar = 50 μm.

In order to assess whether cell migration plays any role in the antero-wards expansion of the RPE domain from the 10 to 16 ss, we performed in vivo time-lapse imaging of embryos where enhanced green fluorescent protein (EGFP) is expressed under a 2.4-kb upstream regulatory region of the tfec gene (tfec:egfp) (Miesfeld et al., 2015) (Fig. 3; Movie 1). To ensure that we followed RPE cells, we chose the brightest EGFP-positive cells within the posterior dorsomedial developing eye vesicle. It is important to note that we tracked the cells until 21–25 hpf to verify that the cells ultimately took on an elongated cell morphology at the rim of the developing optic cup (Fig. 3B,C,H). Cells were followed that moved either dorsally (future nasal), ventrally (future temporal), or posteriorly (future dorsal), over the first (10–16 ss) (Fig. 3A,B) and second (16 ss/24 hpf) (Fig. 3G,H) phase of RPE expansion. Over the initial period of RPE development (10–16 ss), presumptive RPE progenitors show minimal or no net antero-wards movement (Fig. 3A,B, 14–16 ss in 3H). The cells do move gradually (6–12 μm/hr) in the ventral direction (Fig. 3D). At the time of eye invagination around the lens (16 ss), however, the presumptive RPE cells move rapidly in ventral (approximately 20 μm/hr, Fig. 3D; Movie 1), dorsal, and posterior directions toward the rim of the forming eye cup (Fig. 3G,H), at which point the speed of movement slows and the cells eventually stop (Fig. 3D,H). Of note, over the entire expansion period, cells located more posteriorly move more slowly than those situated more anteriorly (Fig. 3D), and presumptive RPE cells exhibit a simple morphology with little protrusive activity (Fig. 3C,G). Further, presumptive RPE cells that are initially neighbors remain in close proximity as they move dorsally or ventrally (Fig. 3A–C,G,H), and the movement happens simultaneously for cells at different locations along the anteroposterior axis of the eye vesicle. Future work will need to address the possibility that the RPE domain expands as a single sheet of epithelium.

RPE Progenitor Division Plays Little or No Role in the Antero-wards Expansion of the RPE Domain

Shortly after the appearance of our first marker (tfec) in RPE progenitors at the 8 ss, the RPE domain expands both anteriorly and, in a more modest fashion, ventrally. This initial expansion appears to be driven by an increase in the numbers of RPE cells (Fig. 3F), in that cell counts of RPE cells in transverse sections indicate more RPE progenitors at the 12 ss (15 hpf) than at the 10 ss (14 hpf) (10 ss, 37.7 ± 4.0 tfec+ cells/eye, n = 10 eyes, error is standard error of the mean; 12 ss, 65 ± 15.0 bhlhe40+ cells/eye, n = 2). Proliferation could underlie this increase in cell number. To determine whether the antero-wards expansion of the RPE progenitor domain is driven in part by cell division, we first performed immunohistochemistry using an anti-phosphohistone-H3 (pHH3) antibody to label cells that are actively undergoing mitosis on embryos processed for RNA in situ hybridization against tfec or bhlhe40. Because cell numbers continue to increase between the 12 and 18 ss, with more bhlhe40+ cells at 18 than 12 ss (12 ss, 65 ± 15 bhlhe40+ cells/eye, n = 2; 18 ss, 86.4 ± 6.3 bhlhe40+ cells/eye, n = 8), we assayed for mitotically active cells between the 8 and 18 ss. Transverse microtome sections were collected, and in a blinded fashion we quantified the percent of tfec+ (8–10 ss) or bhlhe40+ (12–18 ss) RPE cells that were pHH3+ (Fig. 4A,B). Note that bhlhe40 is a more robust marker but is not expressed at the 8–10 ss. Between the 8 and 18 ss, the vast majority of RPE progenitors are pHH3 negative (Fig. 4B). Indeed, at the 10 ss (14 hpf) (Fig. 4B), less than 2% of the tfec+ RPE progenitors are pHH3+, suggesting that the changes in the numbers of tfec+ RPE progenitors over the 1-hr period between the 10 and 12 ss are due to de novo specification of additional RPE progenitors rather than from cell proliferation. The proportion of mitotically active RPE progenitors increases marginally from the 12 to 18 ss (15–18 hpf), with the highest percent of mitotically active RPE progenitors present at the 16 ss (Fig. 4B).

Details are in the caption following the image
RPE progenitors divide minimally as the RPE expands around the eye. Immunohistochemistry was performed using an anti-pHH3 antibody to label cells that are actively undergoing mitosis on embryos processed for RNA in situ hybridization for either tfec (8–10 ss) or bhlhe40 (12–26 ss) mRNA, and transverse sections cut. A: An example of a pHH3+ RPE progenitor (arrowhead) from a section of a 14-ss (16-hpf) embryo. B: Percent of bhlhe40+ cells that are pHH3+ positive at different stages. Number of eyes quantified above bars. C,D: Transverse sections through the central (C; left eye) and most posterior (most dorsal, mature eye) (C, right eye; D) eye at 24 hpf. E: Quantitation of the average number of actively dividing pHH3+ RPE progenitors in 24-hpf embryos from sections at three different locations (A; central, B; posterior [future dorsal], C; most posterior) along the anteroposterior axis as indicated in the schematic (**P < 0.01 with respect to central domain; One Way ANOVA and Tukey post hoc test). A, anterior; br, brain; D, dorsal; e, eye; nr, neural retina; P, posterior; rpe, retinal pigment epithelium; V, ventral; ve, ventricle. Scale bar = 50 μm.

To determine if RPE cell proliferation occurs to any great extent once RPE morphogenesis is complete, we extended our pHH3 analysis to 24 hpf. At 24 hpf, individual RPE cells are elongated and their numbers cannot be assessed (Fig. 4C,D), so we counted the number of pHH3+ bhlhe40+ cells/retinal section (Fig. 4E). Interestingly, at this developmental stage, proliferation appears regionalized within the eye, with few (0.77 cells/retinal section) mitotically active RPE cells present in the central eye, and more than five times as many (4.1 cells/section, P < 0.01) pHH3+ bhlhe40+ cells in the RPE that covers the most posterior (mature dorsal) region of the eye. For comparison purposes, at the 16 ss, approximately 1.5 cells/section are pHH3+. These data are in agreement with a study on proliferation in the early Zebrafish eye vesicle, which showed that eye progenitors exhibit a prolonged cell cycle length (> 30 hr) that shortens to about 10 hr only after 24 hpf, coincident with the start of a period of extensive developmental eye growth (Li et al., 2000b).

To directly test the requirement for cell division of RPE progenitors in the full expansion of the RPE domain, we blocked cell division by treating embryos with aphidicolin (150 μM) and hydroxyurea (20 mM) (Ciruna et al., 2006; Girdler et al., 2013; Twak et al., 2007) from the 8 to 16 ss, when the RPE domain expands in the antero-wards direction, and from the 16 ss to 24 hpf, when the RPE stretches around the back of the neural retina. In both cases, at the end of the treatment period, we assessed the size of the RPE progenitor domain in control and inhibitor-treated embryos by bhlhe40 in situ hybridization. We confirmed by quantification of pHH3+ cells that aphidicolin+hydroxyurea–treated embryos at 16 ss and 24 hpf showed a decrease in mitotically active cells compared to controls (inhibitor-treated embryos showed an 85% decrease in total number of pHH3+ cells in the head and eyes compared to controls; data not shown). Interestingly, preventing cell division has little or no effect on RPE expansion (Fig. 5). The anteroposterior extent of the bhlhe40+ domain at the end of the first phase of RPE expansion (16 ss), expressed as a percent of the length of the eye vesicle, is indistinguishable between controls and those treated with cell division blockers (Fig. 5A,B). Further, embryos treated from 16 ss to 24 hpf also show no difference in the extent of RPE coverage of the eye, as quantified from images of lateral views of the optic cup, indicating that the stretching of the RPE domain around the neural retina to abut the lens occurs normally when cell division is inhibited (Fig. 5C,D). While previous studies had suggested that proliferation was not required for RPE morphogenesis (Miesfeld et al., 2015; Kwan et al., 2012; Li et al., 2000b), whether this was true for expansion of the RPE over the entire extent of the eye vesicle was unclear. Our data indicate that proliferation is not required for any component of the full expansion of the RPE.

Details are in the caption following the image
RPE progenitor division plays little or no role in the expansion of the RPE domain around the neural retina. Embryos were treated either from the 8–16 ss (A,B) or 16–26 ss (C,D) with either DMSO (control) or 150 μM aphidicolin plus 20 mM hydroxyurea. bhlhe40 mRNA in situ hybridization analysis was used to assay for normal RPE development. A: Dorsal whole-mount views of 16-ss control and treated embryos showing similar antero-wards expansion of the bhlhe40+ RPE domain. B: RPE length was quantified with respect to eye vesicle length at the 16 ss (shown in schematic), and no difference in RPE length was observed when cell division was blocked (N = 2; control n = 33; Aph+HU n = 32; P = 0.22). C: Lateral whole-mount views of 24-hpf embryo eyes show similar RPE coverage of the eye when cell division is blocked as compared to controls. D: RPE coverage was quantified with respect to eye vesicle size (shown in schematic), and no difference was observed between controls and inhibitor-treated groups (N = 2; control n = 14; Aph+HU n = 22; P = 0.90). A, anterior; ev, eye vesicle; le, lens; nt, neural tube; P, posterior. Scale bar = 100 μm.

RPE-specific Differentiation Markers are Evident in the Eye Vesicle as it Finishes Elongating

We next asked when differentiation markers first appear in RPE cells, given that pigmentation appears at approximately 22–24 hpf. We reasoned that genes involved in the melanin biosynthesis pathway would turn on prior to when black pigment is evident. We therefore generated antisense riboprobes for candidate genes involved in pigment biosynthesis and melanosome structure and performed RNA in situ hybridization on fixed embryos between 8 and 26 ss.

The earliest marker we have of RPE specification, tfec, turns on at the 8 ss, and bhlhe40 follows 2 hr later at the 12 ss (15 hpf). We find that other RPE-specific genes turn on shortly after bhlhe40 at staggered time points prior to the onset of pigmentation. mRNA for connexin 43 (cx43), a gene encoding a gap junction channel that is expressed by mammalian RPE progenitors (Kojima et al., 2008), appears at the 12 ss (Fig. 6A) and is also present in the developing lens from the 16 to 26 ss (Fig. 6B,C). At the 14 ss, several genes involved in producing pigment turn on in the dorsal-medial leaflet of the eye vesicle, including pre-melanosome protein-a (pmela), dopachrome tautomerase (dct), and tyrosinase-related protein 1b (tyrp1b), as well as the solute carrier 45a2 (slc45a2) (Higdon et al., 2013; Hoek et al., 2008) (Fig. 6E–H). We followed the expression of these markers until the completion of RPE morphogenesis at 24 hpf and found that their expression patterns all follow tightly that of bhlhe40, both temporally and spatially (data not shown). In contrast, as shown previously, an important transcription factor for RPE differentiation in mammals, mitfa, turns on at the 16 ss (Fig. 6D) (Lister et al., 2001). Surprisingly, the pigment associated genes, pre-melanosome protein-b (pmelb) and purine nucleoside phosphorylase-4a (pnp4a), are first expressed at the 18 ss in a highly restricted manner in the dorsomedial RPE, where tfec and bhlhe40 first turn on 3–4 hr earlier, and not in the majority of RPE cells as seen at this stage for bhlhe40 and tfec (compare Fig. 1D,H with Fig. 6I,J). Even at 24 hpf, pmelb and pnp4a are expressed in a scattered group of posterior (mature dorsal) and medial RPE, with few of the RPE cells on the lateral eye surface expressing these pigmentation genes (Fig. 6M–O). Interestingly, at 24 hpf, pmela is expressed in all melanocytes, whereas pmelb mRNA remains restricted to the RPE (Fig. 6K,L). The temporal regulation of gene expression described here is illustrated in Fig. 6P. Finally, black pigmentation begins around 22–24 hpf (Zhang et al., 2014) along the medial border of the eye and spreads laterally in a similar fashion as described above for tfec and bhlhe40 mRNA, but lags behind in time.

Details are in the caption following the image
Differentiation markers are evident in the eye vesicle shortly after RPE progenitors are specified. A,B,D–J: Dorsal whole-mount views of RNA in situ hybridization for RPE differentiation. A,B: cx43 mRNA is expressed in RPE progenitors and the lens beginning at the 12 ss (A) and through 20 ss (B). C: Microtome section of double-fluorescence in situ hybridization for cx43 (green) and slc45a2 (red), with the dorsoventral orientation of the mature eye. D: mitfa mRNA is present by the 16 ss (17 hpf). E–H: Several differentiation markers turn on at the 14 ss, including pmela, slc45a2, dct, and tyrp1b. I–O: pmelb (I,L,M), pnp4a (J,N,O), and pmela (K) mRNA expression from dorsal (I,J,M,N) and lateral (K,L) views of whole embryos and the eye (O). pmelb and pnp4a do not turn on until the 18 ss and only in a subset of the RPE cells (I,J), even at 24 hpf (M,N,O). P: Summary of the timing of the onset of differentiation gene expression. A, anterior; ev, eye vesicle; le, lens; nc, neural crest cells; nr, neural retina; P, posterior; rpe, retinal pigment epithelium. A: Scale bar = 100 μm. C,O: Scale bar = 80 μm. K,L: Scale bar = 200 μm.


In this article, we characterize the development and movement of Zebrafish RPE cells from their point of origin in the dorsal proximal eye vesicle over the surface of the developing optic cup. We provide new insight into this critical period of RPE development, revealing for the first time that genes that encode proteins involved in the differentiated function of the RPE, including pigmentation pathway genes and a solute carrier, turn on soon after RPE progenitors are specified, and prior to expansion around the distal edges of the optic cup. Additionally, we propose that RPE eye development can be separated into two temporally distinct phases (Fig. 7). The initial phase has not been well described and involves an apparent expansion of the RPE progenitor domain in the anteroposterior axis of the inner leaflet of the eye vesicle, which does not depend on cell proliferation but rather we propose on the de novo specification of RPE progenitors. In the second phase, the RPE progenitor domain is stretched around the back of the eye. This process does not begin until the RPE progenitor domain has expanded anteriorly to its full extent, and appears to occur alongside the movement of the neural retina around the distal rim of the forming optic cup.

Details are in the caption following the image
RPE progenitors cover the neural retina in two distinct phases. Schematic depicting the events of RPE development from a lateral view where RPE expansion occurs in two phases. The first phase, between the 8 and 16 ss, involves both an increase in cell number and in domain area in an antero-wards direction that we propose occurs primarily through de novo specification of RPE progenitors. In the second phase, the RPE stretches around the neural retina from the inner leaflet of the eye vesicle in all axes to abut the lens (16–26 ss), coincident with an elongation of RPE progenitors (intranuclear distance increases). A, anterior; D, dorsal; L, lateral; M, medial; os, optic stalk, P, posterior; rpe, retinal pigment epithelium; V, ventral.

In vertebrates, RPE cells arise in the dorsal proximal eye vesicle from a common pool of eye progenitors, likely through signals arising from extraocular tissues (Fuhrmann, 2010). In eye progenitors that become RPE precursors, neural retinal transcription factors such as Vsx2 are down-regulated and Otx2 is maintained. Our data indicate that shortly after specification of the RPE, the cells turn on tissue-specific genes as the RPE domain expands anteriorly. Interestingly, these genes turn on well prior to RPE cells becoming postmitotic (Li et al., 2000b) and displaying pigment. These genes are involved in processes that are related to the differentiated function of RPE cells, and yet are expressed in lineage-committed progenitors several hours earlier than required for pigment biosynthesis. The fact that Zebrafish RPE progenitors turn on differentiation genes well prior to the expression of mitf, a transcription factor thought to be key for the development and differentiation of mammalian RPE cells (Bharti et al., 2006; Bumsted and Barnstable, 2000), is unexpected. These data indicate that in Zebrafish, Mitf does not initiate RPE differentiation but may control the expression of genes that regulate later events in RPE development and/or function. In support, the two mitf homologues in Zebrafish are not required for RPE pigmentation or morphogenesis (Lane and Lister, 2012).

Out data argue that even early in their development, RPE cells are a heterogeneous population. For instance, even though tfec turns on earlier than bhlhe40 in RPE cells, it shows a more restricted expression pattern than bhlhe40 at 24 hpf and is absent from RPE cells that approach the lens. Further, pmelb and pnp4a are expressed at 24 hpf in a scattered population of RPE cells and are largely absent from RPE that covers the lateral surface of the optic cup. It is known that in mammals, a single RPE cell supports many rod photoreceptors, but that the ratio of RPE:cone photoreceptors is closer to one (Volland et al., 2015). Thus, it is possible that these variations in expression of early RPE markers reflect distinct functional requirements for the different RPE cells.

In vertebrates, the initial dorsal RPE domain spreads to cover the entire optic cup. How this spreading occurs is poorly understood. Our data argue that two separate events allow for a fully expanded RPE to form around the back of the retina. The first phase involves an anterior expansion from the 12 to 16 ss of the initial dorsoposteriorly located RPE progenitor domain. Cell elongation appears to play no role in this phase, in that tfec+ cells retain their cuboidal appearance between the 8 and 16 ss. Instead, we propose that the expansion results primarily from de novo specification of RPE progenitors, potentially as naïve eye progenitors come in to contact with inducing signals as the eye vesicle continues to evaginate from the 8 to 12 ss. While the number of RPE progenitors doubles between the 10 and 12 ss, our data indicate that antero-wards expansion is only minimally impacted by progenitor proliferation. First, we find few RPE progenitors are mitotically active (pHH3+) at these stages. In support, while all cells in the eye vesicle at these stages are proliferating, the cell cycle is prolonged (> 30 hr) and few cells actively divide (Li et al., 2000b). Further, expansion of the RPE still occurs when cell division is inhibited. While other reports have suggested that the RPE undergoes its normal expansion when cell division is inhibited from early (10.5 hpf) to late (24 hpf) eye morphogenesis (Kwan et al., 2012; Miesfeld et al., 2015), because we assessed RPE domain expansion at the 16 ss and not at the end of RPE morphogenesis (24 hpf) as in the previous studies, we could preclude the possibility that proliferation is important in the first phase of RPE expansion, but that compensatory mechanisms (e.g., cell size increases, decreased eye vesicle size) permitted sufficient antero-wards expansion of the RPE to allow it to subsequently stretch around the neural retina.

Whether RPE cells migrate in this early phase of RPE development is unclear. Kwan and colleagues tracked over time the nuclei of a small number of Zebrafish dorsomedial eye vesicle cells, which translocate in a minimal fashion either posteriorly or anteriorly between 12 and 14 hpf (6–10 ss) in a coordinated pinwheel movement alongside the rest of the eye vesicle (Kwan et al., 2012). This period is prior to the antero-wards expansion of the RPE domain we define and appears associated with evagination of the eye vesicle. Through our marker analysis, we identified that the RPE domain expands dramatically between 13 and 15 hpf (8–12 ss), and more minimally between 15 and 17 hpf (12–16 ss), but our time-lapse data do not indicate that significant migration of individual RPE cells occurs alongside this expansion (Fig. 7). In support, with the exception of the most distal region of the eye vesicle, RPE progenitor cells at these early stages appear to be organized in a single-layer epithelium and express at least one adhesive molecule, Cx43. It is interesting to note that our marker analysis suggests that the progenitors of the RPE domain remain constrained within the dorsal region of the inner eye vesicle leaflet over these few hours, suggesting an active mechanism prevents extensive dorsoventral expansion of the domain even as it expands anteriorly.

The second phase of RPE development is better understood in Zebrafish and involves the stretching of the RPE over the back of the outer eye vesicle leaflet (Heermann et al., 2015; Kwan et al., 2012; Li et al., 2000a; Schmitt and Dowling, 1994) (Fig. 7). Coincident with eye morphogenesis and movement of the ventromedial eye vesicle around the rim of the eye primordium and into the lateral leaflet, the cuboidal RPE cells are stretched. As a result, the lateral eye vesicle (neural retina) gains volume, while the medial eye vesicle loses volume (Kwan et al., 2012; Li et al., 2000a; Schmitt and Dowling, 1994). Previous descriptions of RPE morphogenesis have been restricted to either histological assessment of cells morphologically identified as RPE in sections (Li et al., 2000a; Schmitt and Dowling, 1994), or as described in a dorsal (future nasal) confocal section by time-lapse microscopy (Kwan et al., 2012). Thus, we contribute here a description of how RPE stretching occurs over the full anteroposterior extent of the eye vesicle. Our data indicate that stretching initiates at the 16 ss, only once the RPE progenitor domain has reached its full anterior extent following the first phase of RPE development. The time-lapse data suggest that RPE morphogenesis occurs simultaneously in the dorsal, ventral, and posterior directions along the full extent of the anteroposterior axis of the eye, though the dorsal stretching is more modest than that which occurs in the ventral direction, likely because the ventral tissue has to move over a greater distance. Indeed, at the 18 ss, the bhlhe40+ RPE is pulled up next to the lens dorsally, while it is still far from reaching the ventral (future temporal) lens.

Which cell is the driver in the coupled morphogenesis of the future temporal neural retina and the RPE is unclear. One possibility is that the flattening of RPE cells provides a propulsive force that pushes the future temporal eye progenitors. Alternatively, these temporal eye progenitors may pull out the sheet of RPE cells, elongating RPE cell shape as they do. Finally, one could also imagine a scenario where neither cell type is passive and their combined behaviors drive eye vesicle morphogenesis.

The molecular mechanisms that drive RPE expansion are unknown. We describe here the coordinated events of RPE specification, differentiation, and morphogenesis in Zebrafish, an excellent model organism for genetic and molecular analyses of protein function in a developmental process. This information, coupled with the use of RPE-specific fluorescent transgenic lines such as Tg(mitf:gfp), where GFP turns on in RPE cells by 18 hpf (Curran et al., 2009; Miesfeld and Link, 2014), or Tg(tfec:egfp), where EGFP is expressed in RPE progenitors as early as the 8 ss (13 hpf) (Miesfeld et al., 2015), will prove useful in the future for defining what goes wrong when genes important in RPE development and morphogenesis are disrupted.

Experimental Procedures

Zebrafish Husbandry

Tupfel Longfin (TL) (ZIRC, Eugene, Ore.) embryos were raised at 28.5°C in E3 medium supplemented with 0.25 mg/L methylene blue as described previously (Westerfield, 2000). Embryos were staged by number of somites up to the 18 ss, and by hpf thereafter (Kimmel et al., 1995). The University of Calgary Animal Care Committee approved the animal protocols.

Probe Synthesis

Digoxygenin (DIG)-labeled antisense riboprobes were synthesized as previously described (Thisse and Thisse, 2014), and fluorescein-labeled probes were generated similarly by substituting fluorescein for DIG-labeled nucleotides. DNA templates for each probe were amplified from 20-ss Zebrafish cDNA by using gene-specific primers (Table 1), and antisense riboprobes were transcribed using SP6 RNA polymerase at 40°C.

Table 1. Primers Used to Amplify cDNA for Antisense Riboprobe Synthesisa
Gene Primers (forward, reverse)
  • a SP6 RNA polymerase was used to transcribe antisense probes for RNA in situ hybridization analysis using the forward and reverse primers shown for each RPE-specific gene. Only the gene-specific sequence is shown for each reverse primer, which also contained the SP6 promoter sequence ATTTAGGTGACACTATAGA.

RNA In Situ Hybridization

Colorimetric and double-fluorescent RNA in situ hybridization was performed by using whole embryos as described previously (Tessmar-Raible et al., 2005; Thisse and Thisse, 2014). For embryos imaged at 24 hpf, 0.003% (w/v) 1-phenyl-2-thiourea was used to prevent pigment biosynthesis (between 18 ss and 24 hpf). Images were taken of whole-mount embryos in 3% methyl cellulose (Sigma) using a Stemi SV 11 microscope with an AxioCam HRc camera and AxioVision software. Some embryos were dehydrated and embedded in JB4 medium (Polysciences, Inc.) and 7-μm-thick transverse sections were cut using a Leica microtome. Sections were mounted with a coverslip in Aqua-Poly/Mount (Polysciences, Inc.) and imaged by using a Zeiss Axioplan2 microscope with an MRc camera and AxioVision software. Images were compiled and adjusted for brightness/contrast with Adobe Photoshop.


Immunostaining was performed on whole embryos after RNA in situ hybridization similar to methods described previously (Macdonald, 1999). Briefly, embryos were permeabilized in acetone at -20°C for 20 min, washed twice for 5 min in phosphate buffered saline plus 0.1% Tween20 (PBST), blocked for 1 hr in 10% normal sheep serum (NSS), and then incubated with primary antibody (pHH3; Millipore, 1:500) diluted in 10% NSS. Next, embryos were washed in 1% NSS diluted in PBST and then incubated with secondary antibody for 1 hr (Alexa Fluor 546; Fisher Scientific, 1:1000). Embryos were washed in PBST, dehydrated with an ethanol work-up, embedded in JB4 solution, and microtome-sectioned.

Pharmacological Inhibition of Proliferation

8-ss or 16-ss embryos were grown in their chorion in 0.3X Danieau's solution that contained either 4% dimethyl sulfoxide (DMSO) or 150 μM aphidicolin (Sigma) plus 20 mM hydroxyurea (Sigma). These embryos were fixed in 4% paraformaldehyde at the 16 ss and 24 hpf, respectively, and processed as whole-mounts either with an antibody against pHH3 to assess proliferation, or for RNA in situ hybridization with an antisense bhlhe40 riboprobe. To assess the degree of RPE expansion at the 16 ss, images of dorsal views of the embryos were taken, and the anteroposterior extent of the bhlhe40+ RPE domain was quantified as a percent of the length of the eye vesicle. To assess the degree of RPE expansion at 24 hpf, images of lateral views of the embryos were taken, and the area of the bhlhe40+ RPE domain measured and represented as a percent of the area of the lateral eye, including the lens. The areas of the bhlhe40+ domains were measured with the researcher blinded to experimental conditions. Data were assessed for statistical significance by using a two-tailed, unpaired Student's t-test with SigmaStat (3.0).

Cell Quantification

The number of tfec+ or bhlhe40+ cells in 8–18 ss (13–18 hpf) 7-μm plastic sections were counted and compared to cells also positive for pHH3 antibody label. Every second section (3–5 sections/eye) through the central (no optic stalk) to distal eye vesicle of 2–10 eyes was assessed. An RPE cell was considered to be mitotically active if the pHH3+ nuclei sat entirely within a cell whose borders were clearly defined by bhlhe40 in situ hybridization signal, and in central retinal sections was not radially aligned to the apical-basal orientation of the outer leaflet (neural retina) of the eye vesicle. Of note, similar percentages of pHH3+ RPE cells were present in the central and distal eye vesicles (data not shown). For each eye, the total number of cells double-positive for pHH3 and marker, divided by the total number of marker+ cells, gives the fraction of dividing RPE progenitors. At 24 hpf, it was not possible to count individual RPE cells because of their elongated shape, thus the number of pHH3+ cells/retinal section at different locations along the anteroposterior (future dorsoventral) axis of the optic cup were counted. P values were calculated using One Way ANOVA and Tukey post hoc tests with Prism 7.

In Vivo Confocal Time-lapse Microscopy

A Tol2 2.4-kb tfec:egfp plasmid (50 ng/μL) (kindly provided by the laboratory of Dr. Brian Link) (Miesfeld et al., 2015) was injected with transposase mRNA (100 ng/μL) into 1-cell-stage embryos and animals were raised and screened for germ-line transmission. One founder line was established, but ultimately the F2/F3 generations could not be used for imaging due to gene silencing issues. Therefore, additional analysis was done on F0 fish injected with the tfec:egfp construct and transposase mRNA (as above; N = 4 embryos and n = 13 cells). For in vivo imaging, 8-ss embryos were dechorionated and mounted (lateral orientation) in 0.8% low-melting-point agarose in 1X E3 medium, then covered with 1X E3 containing 0.16 mg/mL tricaine. A stage heater was used to keep embryos at 28.5°C. Images were acquired using a Zeiss LSM700 confocal microscope: 15–30 z-sections, 4-μm z-step, 5-min intervals, 20X objective, 10–12 hr. Movies were made from maximal projections of 3–4 sections of the medial eye vesicle that contained cells in the presumptive location of the RPE domain (based on our analysis of marker expression in Figs. 1 and 2). NIH ImageJ software was used to track the cells manually over time. Cells were tracked that were brightly EGFP-labeled and that ultimately, by 22–24 hpf, flattened in a plane radial to the axis of the eye epithelium at the distal rim of the eye vesicle; the expected location and shape of RPE cells.


The authors thank S. Childs for the use of her Zebrafish facility, and Drs. K. Atkinson-Leadbeater and J. Hocking for helpful comments on the article. P.B.C. was supported by a studentship from Natural Sciences and Engineering Research Council (NSERC) of Canada; S.M. was a Tier II Canada Research Chair in Developmental Neurobiology and an Alberta Innovates-Health Solutions scientist. The research was also supported by Foundation Fighting Blindness and a Lions Sight Centre Fund awards to S.M.