Individual Cell Migration Serves as the Driving Force for Optic Vesicle Evagination

See allHide authors and affiliations

Science  25 Aug 2006:
Vol. 313, Issue 5790, pp. 1130-1134
DOI: 10.1126/science.1127144


The cellular mechanisms underlying organ formation are largely unknown. We visualized early vertebrate eye morphogenesis at single-cell resolution by in vivo imaging in medaka (Oryzias latipes). Before optic vesicle evagination, retinal progenitor cells (RPCs) modulate their convergence in a fate-specific manner. Presumptive forebrain cells converge toward the midline, whereas medial RPCs remain stationary, predetermining the site of evagination. Subsequent optic vesicle evagination is driven by the active migration of individual RPCs. The analysis of mutants demonstrated that the retina-specific transcription factor rx3 determines the convergence and migration behaviors of RPCs. Hence, the migration of individual cells mediates essential steps of organ morphogenesis.

For over a century, the vertebrate eye has served as a paradigm to study vertebrate organogenesis (1). Fate mapping has revealed the origin of retinal progenitor cells (RPCs) and forebrain cells in teleosts (24); however, the cellular mechanisms underlying eye morphogenesis remain unclear. Eye development starts at late gastrula stages with the specification of RPCs within the eye field of the anterior neural plate (5). Evagination of the optic vesicles, the first step of eye morphogenesis, is visible after neurulation. We applied in vivo imaging techniques in medaka to reconstruct eye morphogenesis at single-cell resolution. RPCs were specifically marked by fluorescent tagging of the homeobox transcription factor rx3, a protein that is expressed in the RPCs from the gastrula to optic vesicle stages (stages 16 to 20) (6, 7). We generated rx3::green fluorescent protein (GFP)–transgenic embryos to visualize RPCs during optic vesicle morphogenesis. To follow the movement of all individual cells during optic vesicle evagination, we counterstained all nuclei of the embryo with histone H2B-monomeric red fluorescent protein (H2B-mRFP) (8, 9). These embryos were imaged at high spatial and temporal resolution by confocal in vivo time-lapse microscopy (Fig. 1, A and B, and movie S1) (10). At gastrula stages, RPCs (which are GFP-positive cells) are located within the eye field in the anterior neural plate (Fig. 1B at time 0:00). The eye field is surrounded at its anterior-lateral and posterior borders by tel- and diencephalic progenitor cells, respectively (2). In the course of neurulation, the eye field converges to the midline and remains as a single domain within the forebrain (Fig. 1B at 03:28). Three hours later, the optic vesicles have evaginated from the neural keel (Fig. 1B at 06:16). It has previously been assumed that the optic vesicles evaginate after the neural keel [the precursor of the central nervous system (Fig. 1A)] has formed. However, at the level of gross morphology, we find that the eye field remains wide and distinct from the neural keel during neurulation as the posterior neural axis narrows, indicating an earlier onset of eye morphogenesis (dashed white line in Fig. 1B at 03:28). Similar observations have been made in mouse and rat models, in which the optic pit appears during the elevation of the neural plate, thus before neural tube formation (11, 12).

Fig. 1.

Fate-specific modulation of convergence. (A) Schematics of the imaging setup. D, dorsal; V, ventral; A, anterior; P, posterior. (B and C) Confocal sections during optic vesicle morphogenesis in wild-type (wt) (B) and eyeless (el) (C). RPCs, green (rx3::GFP); nuclei, red (H2B-mRFP); anterior is at top. (B) Gastrula-stage eye field (time 00:00) condenses to a single domain in the forebrain (03:28), which is wider than the neural keel (dashed white line). Optic vesicles (ov) have evaginated from the neural keel (nk) (06:16). (C) The eye field in eyeless does not differ from the wild type [00:00, compare to (B)]. Subsequently, mutant RPCs converge, forming the neural keel (03:28 to 6:16). (D to F) 4D visualization of tracked cells. Forebrain cells, magenta spheres; RPCs, green spheres. (D) Anterior view, with dorsal at top. (E and F) Dorsal view, with anterior at top. In (D) and (E), stationary wild-type RPCs keep the eye field wide (red arrows, 0:52 to 4:12). Diencephalic cells of the lateral neural plate converge to the midline and form the neural keel (white arrows). Lateral RPCs (yellow arrows) move to the midline, change direction, and contribute to the optic vesicles (2:36 to 6:00). (F) eyeless mutant RPCs (yellow arrows) converge as forebrain cells (0:00 to 4:12, white arrows), forming the neural keel (6:00). Time is reported in hh:mm (h, hours; m, minutes).

To follow individual cell movements during optic vesicle evagination, we tracked more than 300 individual cells of the eye field and the surrounding forebrain in the acquired four-dimensional (4D) sequences by recording the approximate geometric center of the H2B-mRFP–labeled nuclei. The recorded data were visualized with the use of 3D rendering routines to reconstruct the movements that individual cells underwent during evagination (Fig. 1, D and E, and fig. S1, A to F) (10). This analysis revealed that, within the eye field, ventromedial RPCs do not converge and show no net movement toward the midline (red arrows in Fig. 1, D and E). These stationary RPCs keep the eye field wide during neurulation at the site of evagination [Fig. 1, D at 2:36 and E at 2:36 and 4:12 (red arrows); and movies S3 and S4). In contrast, RPCs of the lateral eye field and prospective prosencephalic cells migrate to the midline (yellow and white arrows, respectively, in Fig. 1, D and E). On approaching the midline, the RPCs of the lateral eye field show a biphasic behavior; first they dive ventrally and then migrate laterally into the evaginating optic vesicle [Fig. 1D from 2:36 to 4:12 (yellow arrows) and fig. S1, A to F, (cell 9)]. Therefore, optic vesicle evagination starts during neural keel formation, whereas the neural plate converges at this stage (Fig. 1, D and E). Ventromedial RPCs thus never pass through the neural keel before they evaginate. This precludes a simple evagination of RPCs from the preformed neural keel, as was previously assumed (13, 14).

To further study the mechanism of optic vesicle evagination, we analyzed the medaka rx3 mutant eyeless (15). The rx homeobox transcription factors are indispensable for vertebrate eye development, and the loss of rx function results in the absence of eyes (1520). Medaka and zebrafish (Danio rerio) rx3 mutants lack eyes because of a failure of optic vesicle formation despite the presence of RPCs (15, 18, 19). We performed confocal time-lapse analysis and single-cell tracking in the rx3 mutant background (Fig. 1, C and F; fig. S1, G to K; movies S2 and S5) (10). At late gastrula stages, the mutant eye field is indistinguishable from that of the wild type (Fig. 1, C and F at 0:00). However, at the onset of convergence, all RPCs move toward the midline as do the other forebrain cells (Fig. 1F from 0:00 to 4:12) to form a neural keel (Fig. 1F at 6:00). Initiation of optic vesicle evagination by stationary RPCs of the ventromedial eye field does not occur (Fig. 1, C at 3:28 and F at 2:36). Thus, a crucial first step in vertebrate eye morphogenesis is omitted in the rx3 mutant, resulting in the localization of RPCs within the neural tube (Fig. 1F at 6:00; Fig. 2, B and D) (21).

Fig. 2.

Cell shape [(A) to (D)] and mosaic analysis [(E) to (H)]. (A and B) Single confocal planes from rx3::mYFP embryos; anterior is at top. (C and D) α-tubulin (green) and α-GFP (red) stainings on transverse sections of rx3::GFP embryos; dorsal is at top. The dynamic morphology of wild-type RPCs during evagination is shown [(A), arrows and (C)]. (B) and (D) Epithelial organization of mutant lateral RPCs is visible (brackets) surrounding the round medial cells. (E to H) Mosaic analysis. Confocal sections of the 4D sequence (neurula to somitogenesis stages) are shown; anterior is at top. (E) to (F) Normal participation of wild-type cells (membrane-mRFP, red regions) transplanted into the wild-type eye field (rx3::mYFP, green regions). (G) Wild-type cells (red) in the eyeless host (green) at the lateral border of the eye field and at medial positions. (H) Cell-autonomous rescue of optic vesicle formation by wild-type cells. Asterisk, mutant tissue remaining in the forebrain; dashed white line, anterior border of the forebrain. Time is reported in hh:mm:ss (h, hours; m, minutes; s, seconds).

In the absence of rx3 function, mutant RPCs adopt a migratory behavior typical of forebrain cells, in that they converge to the midline to form a neural keel. To visualize the changes in cell shape during optic vesicle morphogenesis, we used a rx3::membrane yellow fluorescent protein (rx3::mYFP) transgenic line to address the correlation of the morphogenetic behavior of RPCs and their morphology (Fig. 2, A and B, and movie S6) and corroborated these findings by immunostaining for α-tubulin (Fig. 2, C and D) (10). Wild-type RPCs are highly motile and adopt various shapes during evagination, reflecting their dynamic movement (Fig. 2, A and C, and movie S6). In contrast, mutant lateral RPCs are columnar and form part of the epithelialized wall of the forebrain, surrounding medial rounded mutant RPCs that also fail to evaginate (Fig. 2, B and D, and movie S6). Thus, in the absence of rx3, lateral mutant RPCs adopt both the shape and morphogenetic behavior of neuroepithelial cells, resulting in the formation of a neural tube containing nonevaginated RPCs (15). This indicates that neuroepithelial morphology and migratory behavior are a default state and that rx3 sets apart a domain that moves and adopts a shape different from that of the forebrain.

The observed epithelialization of mutant RPCs could be the cause for the failure of evagination. However, single-cell tracking revealed that optic vesicle morphogenesis starts before neural tube formation. An alternative explanation is that an altered migratory potential or behavior of mutant RPCs induces the phenotype. We performed mosaic 4D analysis, allowing the simultaneous monitoring and comparison of wild-type and mutant cell behavior during evagination at high resolution. Wild-type cells from embryos ubiquitously expressing a membrane-tethered mRFP (8, 9) were transplanted at the blastula stage to the animal pole of an eyeless intercross expressing rx3::mYFP (10). Wild-type cells transplanted into a wild-type eye field contribute normally to the developing optic vesicles (Fig. 2, E and F). Individual cells located medially before evagination (Fig. 2E) migrate into the optic vesicles to integrate into the vesicular epithelia (Fig. 2F), contributing to their growth (movie S7).

In the mutant background, individual wild-type cells rescue optic vesicle evagination in a cell-autonomous fashion (Fig. 2, G and H, and movie S8). Small but otherwise normal optic vesicles are formed by wild-type cells, whereas mutant cells of the host are found exclusively in the forebrain keel (Fig. 2H, asterisk). 4D volume rendering shows the migratory behavior of wild-type cells in the mutant context (Fig. 3, A to C). At early neurula stages, wild-type RPCs within the mutant eye field are found either at the lateral border, because of the reduced convergence of ventromedial wild-type RPCs, or within the medial eye field (Figs. 2G and 3A). 4D rendering shows that, subsequently, wild-type cells from the medial part of the eye field actively penetrate the surrounding mutant tissue to move laterally, where they merge with the wild-type RPCs in the rescued optic vesicle (Fig. 3, A and B, asterisk, and movie S9). These clusters gradually merge into a single vesicle by posterior movements (Fig. 3C). In contrast, mutant cells form a neural keel and never contribute to the optic vesicles (Figs. 3C and 2H). Movements of wild-type cells in the mutant background can occur either as small groups of cells (Fig. 3, D to F, arrow, and movie S10) or as individual cells that migrate through the mutant tissue of the medial forebrain into the rescued optic vesicles (Fig. 3, G to I, asterisk, and movie S11). Previous transplantation experiments have shown that mutant cells in a wild-type background do not contribute to the forming optic vesicles (21). Thus, the observed phenotype in eyeless mutant embryos is due to the altered migratory behavior of mutant cells. Optic vesicle formation can be efficiently rescued by wild-type cells. This indicates that the cues directing this migration are not affected in the mutant background.

Fig. 3.

Individual cell movement. (A to C) 3D reconstruction of rescued evagination by wild-type cells (red) in the mutant host (green). One half is shown. Dashed blue line, midline (m). Anterior is at top left. Wild-type cells [(A), asterisk] move through mutant tissue (B), forming a small optic vesicle (C). (D to F) Maximum-intensity projection (MIP) showing wild-type cells (red) in the eyeless host (mutant cells not shown). A cluster of four medial wild-type cells [(D), arrow] migrates to the optic vesicle [(E) to (F)]. (G to I) Single cell (red, asterisk) migrating through eyeless mutant tissue (green).

To further address the migratory properties of wild-type RPCs, we transplanted wild-type cells labeled by rx3::mYFP to an unlabeled wild-type background (Fig. 4, A to F) (10). Individual RPCs are highly motile and extend lamellipodia (Fig. 4F, arrowhead) and filopodia (Fig. 4F, arrow) as they converge toward the midline (movie S14). Both lamellipodia and filopodia are characteristic features of actively migrating cells. Also, at the beginning of optic vesicle evagination, RPCs extend filopodia (Fig. 4B, solid arrowhead) as well as lamellipodia (Fig. 4B, open arrowhead) (movie S12) in the direction of their movement. Furthermore, migrating RPCs elongate and actively migrate into the forming optic vesicle (Fig. 4C). Finally, cells actively exchange neighbors during their migration. This demonstrates individual cell migration. A representative example is shown in Fig. 4 (D and E). Cell 1 first lies lateral to cell 2 (Fig. 4D, 0:00). Within 40 min, it has moved over cell 2 (Fig. 4E at 0:38, and movie S13). Thus, active single-cell migration of RPCs is a crucial feature of vertebrate eye morphogenesis, both during convergence of the eye field and subsequently during optic vesicle evagination.

Fig. 4.

Cell morphology during migration. Morphology of cells is shown during evagination [(A) to (E)] and convergence (F). An MIP of wild-type cells (rx3::mYFP, green) in unlabeled wild-type background is shown. (A to C) Cell extending a lamellipodium (solid arrowhead) and filopodia (open arrowheads). The line in (C) indicates the distance moved by the cell. (D and E) Cell (1) moving over a neighboring cell (2). Dashed line, lateral border of eye field. (F) Cell in the eye field extending lamellipodia (arrowheads) and filopodia (arrow). (G) Model of optic vesicle morphogenesis. RPCs, green; forebrain, red. Panels 1 to 3 show that modulated convergence of ventromedial RPCs (blocked green arrow), as opposed to forebrain cells (red arrows), results in the formation of a wide eye domain. RPCs of the lateral eye field move toward the midline (green arrows in panels 1 and 2) and subsequently toward the forming vesicles (green arrows in panel 3). Evagination is driven by single-cell migration (green arrows in panel 4) at all stages. Hypothetical midline signal (blue circle) or two different signals could direct both processes: the block of convergence and subsequent evagination.

Previous studies of optic vesicle formation used fate mapping and gross morphological analysis (24, 13, 22). Consequently, models were not available to describe the behavior of cells during the transition from the eye field to optic vesicles. Our time-lapse analysis using fate-labeled RPCs allowed us to resolve the behavior of individual cells in this complex process. We show that the migratory properties of RPCs differ from those of forebrain cells in an rx3-dependent manner. Single wild-type cells migrate through the surrounding tissue in eyeless mutants, thereby rescuing optic vesicle evagination. We propose a two-step process of optic vesicle formation (Fig. 4G). First, during the gastrula and early neurula stages, ventromedial RPCs show a modulated, slower convergence when compared to the neural plate (Fig. 4G, steps 1 and 2, blocked green arrow), resulting in a wide domain where lateral evagination of the optic vesicle is initiated (Fig. 4G, step 3). In addition, RPCs are prevented from forming the epithelialized neural keel structure of more posterior forebrain cells. Second, RPCs in the forebrain elongate mediolaterally and migrate actively from medial positions laterally, enlarging the volume of the growing optic vesicles (Fig. 4G, step 4). Our 4D analysis indicates that, in both steps, the coordinated migration of individual RPCs is a driving force.

The migration of RPCs could be driven by a neuroectodermal midline signal as proposed for the convergence and extension (CE) of the neural plate in Xenopus (23, 24). In that model, an unidentified signal polarizes and orients cells toward the midline and defines distinct morphogenetic areas within the posterior neural plate (23, 24). We hypothesize that, during optic vesicle morphogenesis, a neuroectodermal midline signal first promotes convergence movements that are modulated in a cell-autonomous fashion in rx3-positive cells. Second, the same or a second signal directs the RPCs during subsequent optic vesicle evagination. Examples of such signals acting on axon growth are known (25, 26). Alternatively, pioneering cells in the optic vesicle could provide attractive cues for the migratory RPCs. It will be interesting to examine whether the signaling mechanisms involved in CE and optic vesicle evagination build on the same molecular players.

Supporting Online Material

Materials and Methods

Fig. S1


Movies S1 to S14

References and Notes

View Abstract

Navigate This Article