Balanced Rac1 and RhoA activities regulate cell shape and drive invagination morphogenesis in epithelia

Abstract

Epithelial bending is a central feature of morphogenesis in animals. Here we show that mutual antagonism by the small Rho GTPases Rac1 and RhoA determines cell shape, tissue curvature, and invagination activity in the model epithelium of the developing mouse lens. The epithelial cells of the invaginating lens placode normally elongate and change from a cylindrical to an apically constricted, conical shape. RhoA mutant lens placode cells are both longer and less apically constricted than control cells, thereby reducing epithelial curvature and invagination. By contrast, Rac1 mutant lens placode cells are shorter and more apically restricted than controls, resulting in increased epithelial curvature and precocious lens vesicle closure. Quantification of RhoA- and Rac1-dependent pathway markers over the apical–basal axis of lens pit cells showed that in RhoA mutant epithelial cells there was a Rac1 pathway gain of function and vice versa. These findings suggest that mutual antagonism produces balanced activities of RhoA-generated apical constriction and Rac1-dependent cell elongation that controls cell shape and thus curvature of the invaginating epithelium. The ubiquity of the Rho family GTPases suggests that these mechanisms are likely to apply generally where epithelial morphogenesis occurs.

Actin remodeling is principally regulated by the highly conserved small Rho GTPase family (1, 2), which in the interconvertible GTP-bound active state, binds to specific effector proteins that determine their cellular function in signal transduction pathways (3). Early reports in cultured fibroblasts showed that overexpression of RhoA, Rac1, and Cdc42, the three most-studied members, induced the formation of contractile stress fibers, protrusive lamellipodia, and probing filopodia, respectively (2). These and later studies identified distinctions in the mechanisms by which these three small Rho GTPases remodel the actin cytoskeleton. For RhoA, activation at the rear of cells stimulates the downstream effector Rho-associated kinase (ROCK) (3), thereby up-regulating phosphomyosin activity on preexisting filamentous actin (F-actin) and enhancing formation of contractile actin (4). For activated Rac1 and Cdc42, each binds to a particular effector protein at the cell leading edge. Rac1 indirectly associates with WASP family verprolin-homologous protein (WAVE) (3) or activates the effector p21-activated kinase (PAK) that suppresses local formation of contractile actin (2), and Cdc42 binds directly to neural Wiskott-Aldrich syndrome protein (N-WASP) or WASP (3). These latter three effectors bind an actin nucleator (Arp2/3 or diaphanous-related formins) (5) and promote generation of distinct F-actin types; Y-branched F-actin networks for Rac1 and bundled F-actin for Cdc42. However, recent biosensor detection studies show that RhoA, together with Rac1 and Cdc42, are active with spatial and temporal distinctions at the leading edge of migrating cells (6), whereas myosin has an additional role in actin complex disassembly at the trailing edge (7). Interestingly, Rac1 and RhoA have also been shown in migrating cells to exhibit mutual antagonism (8–13), mediated by p190Rho–GTPase-activating protein (GAP) (10), the kinase sticky (13), or FilGAP (8).

The mechanisms for cell migration are coming into sharper focus (14, 15), in contrast to those regulating developmental morphogenesis. Epithelial invagination is a conserved type of morphogenesis found, for example, during ascidian gastrulation, Drosophila mesoderm invagination, mammalian inner ear and neural tube formation. Epithelial sheets first establish and maintain apicobasal polarity, chiefly by the Par complex (16) regulated by the Rho GTPases, to localize cytoskeletal components and signaling machinery necessary for invagination (17). The actual process of invagination requires cell shape changes that drive folding of a tissue sheet; apical constriction is believed to be important for driving invagination (17). Studies on Drosophila gastrulation have shown that the upstream factor RhoGEF1 stimulates Rho1, thereby activating Drok/ROCK and consequently nonmuscle myosin to similarly constrict epithelial cell apices (17). Mechanisms of apical constriction mediated by phosphorylated myosin have also been verified for vertebrate invagination (17). For maturing polarized epithelia, actomyosin complexes engage adherens junctions through α-catenin and eplin (18) and then circumferential actin belts replace radial actin arrays by an unknown mechanism (19). In Drosophila ventral furrow formation, these contractile actin belts ensure apical constriction of cells across the ventral furrow to initiate tissue bending. Recently, it has shown that the actomyosin belts display a pulsing behavior, rather than contracting simultaneously, to drive apical constriction (20).

Early studies of lens placode morphogenesis revealed apical constriction during invagination (21). Recently, a gene-trapped mouse with a disruption in Shroom3 (22) verified that lens pit cells undergo Shroom3-mediated apical constriction (23), and another study showed that Cdc42-dependent contractile filopodia, emanating from the basal lens pit, function as physical tethers to coordinate invagination of presumptive lens and retina (24). Here, we used conditional mouse genetics to assess RhoA and Rac1 roles in the developing mouse lens. This showed that RhoA controls apical cell width and that Rac1 controls cell length. By quantifying RhoA and Rac1-dependent pathway markers over the apical–basal axis of lens pit cells we showed that in RhoA mutant epithelial cells, there was a Rac1 pathway gain of function and vice versa. These findings suggest that mutual antagonism produces balanced activities of RhoA-generated apical constriction and Rac1-dependent cell elongation in controlling cell shape and thus curvature of the invaginating epithelium.

Results

Invaginating Lens Epithelial Cells Show Trailing and Leading Edge Markers of Migration.

Prompted by work on single migrating cells and by studies showing that GTPases regulate epithelial morphogenesis in the sea urchin (25), the fly (26–28), and mouse (29), we determined whether lens placode cells had a polarized distribution of molecules similar to single migrating cells. We labeled the embryonic day E10.5 lens pit for established leading edge and trailing edge markers. Arpc2 is a subunit of the Rac1 or Cdc42-dependent actin nucleator Arp2/3 (15). Cortactin is an Arpc2 interactor that enhances actin branch formation (30). Both Arpc2 and cortactin identify lamellipodia, whereas myosin IIB identifies the preceding lamellae at the leading edge of migrating cells. Labeling of the lens pit for these markers (Fig. S1 A–D) revealed that all three were found at the base of the epithelial cells. Although any one example does not reveal this finding, quantification and averaging of labeling intensity over a 5-μm line interval for many examples (Fig. S1 B, yellow dashed lines, and C) showed that as in single migrating cells (31), the peak intensity of myosin IIB labeling appeared displaced toward the cell center compared with Arpc2 and cortactin. This combination and spacing of markers suggest that the base of lens pit cells might be similar functionally to the protrusive edge of a single migrating cell. The prior demonstration that Cdc42-dependent filopodia project from the basal surface of lens pit epithelial cells (24) is also consistent with this hypothesis.

To determine whether lens pit epithelial cells might also have RhoA-dependent complexes (1), we labeled the E10.5 lens pit for phospho-myosin regulatory light chain (MRLC), myosin IIB, and F-actin. According to immunofluorescence visualization (Fig. S1 E, F, and H) and quantification of that labeling (Fig. S1G) all three markers were found at the apex of lens pit epithelial cells in overlapping peaks that were within 1 μm of the apical edge. Although phospho-MRLC was also found at the base of lens pit epithelial cells, the majority of the activity was at the apex (Fig. S1H). Collectively, these labeling studies indicated that the epithelial cells of the lens placode cells have a similar molecular signature as single migrating cells.

RhoA and Rac1 Lens Mutants Generate Opposing Invagination Phenotypes.

The earlier observations suggest that the mechanisms of invagination might be related to mechanisms of single cell migration. To test this hypothesis, we generated placode-specific Rac1 and RhoA somatic mutants in the mouse using conditional alleles (Fig. S2 A and C) and the Le-cre driver. Without effective antibodies to RhoA, we used PCR genotyping of DNA isolated from lens pits to show that the loxP sites were recombined in the RhoA^fl/fl; Le-cre embryos (Fig. S2B). Using antibodies to Rac1, we also showed that the Rac1^fl/fl; Le-cre mutants had lost Rac1 immunoreactivity in the lens pit (Fig. S2 D–I). These studies also show that Rac1 immunoreactivity in the lens pit cells has a membrane-association pattern (Fig. S2 F–K).

Somatic deletion of Rac1 or RhoA resulted in subtle but reproducible phenotypes in the shape of the lens pit at E10.5 (Fig. 1). In the RhoA mutant, the lens pit showed a more open shape (Fig. 1B) compared with control (Fig. 1A). By contrast, the Rac1 mutant showed a lens pit that was more closed (Fig. 1C). Using coordinate geometry, we combined many examples of curved lines that represented the apical surface of the lens pit to quantify the average shape in control and mutant examples. This confirmed that on average, lens pits of RhoA mutants showed less curvature, whereas those of the Rac1 mutants showed more (Fig. 1D). These changes in curvature showed that at E10.5 the shape of the lens pit was in delicate balance and could be influenced either positively or negatively in these mutants.

Fig. 1.

Rac1 and RhoA mutations have opposite effects on epithelial curvature. (A–C) The appearance of control (A), Le-cre; RhoA^fl/fl (B), and Le-cre; Rac1^fl/fl (C) lens pits at E10.5. (D) Curves representing the average shape of the apical surface of E10.5 lens pits from control (gray), Le-cre; RhoA^fl/fl (red), and Le-cre; Rac1^fl/fl (blue) and Le-cre; Rac1^fl/fl; RhoA^fl/fl (green). Although individual examples of any genotype show some variability in lens pit shape, coordinate geometry averaging shows that the RhoA mutant pit has less curvature, whereas that for Rac1 has more. Quantification in E10.5 lens pits (n = 10) of cell numbers.

RhoA family GTPases are known to control cell cycle progression in some contexts (32). To determine whether the Rac1 and RhoA conditional mutants showed any evidence of this, we quantified cell number in comparison with wild type and, as an additional control, used the previously characterized Le-cre; Cdc42^fl/fl conditional mutant (24) (Fig. S3E). This showed that a normal lens pit contains about 60 cells in section and that none of the mutants showed a significant change. We also did not observe any indication of the multinucleate cells that might result from RhoA regulation of cytokinesis (33). These distinct consequences of RhoA mutation are likely to result in part from differential expression of RhoA effectors.

We previously showed that the Le-cre; Cdc42^fl/fl conditional mutant fails to generate filopodia that connect presumptive lens and retina (24). These processes aid in morphogenesis by constraining interepithelial distance and pulling the lens pit down into the optic cup (24). To determine whether the Rac1 or RhoA conditional mutants had defects in the generation of these filopodia, we determined filopodial indices and measured interepithelial distances (Fig. S3). As shown previously, the Cdc42 mutant has almost no filopodia (Fig. S3B) and more than double the interepithelial distance (Fig. S3 F and G, aqua bar) compared with control (Fig. S3 A, F, and G, gray bar). The Rac1 mutant has no change in filopodial index and no change in interepithelial distance (Fig. S3 D, F, and G, blue bar). By contrast, the RhoA mutant has a reduction in the filopodial index that is about 66% of the control value (Fig. S3 C and F, red bar). This may not be surprising given that these processes contain functional contractile myosin complexes (24). However, this 34% reduction in filopodial index is not accompanied by a change in the interepithelial distance (Fig. S3G, red bar). This probably means that the RhoA mutant still has sufficient filopodia to constrain the interepithelial distance. Consistent with this argument, a focal adhesion kinase (FAK) conditional mutant has even fewer filopodia (an index of 0.43, 50% of the control value) but still has no change in the interepithelial distance (Fig. S3 F and G, pale cream). On the basis of this analysis, we cannot conclude that the number or length of filopodia in the RhoA mutant has any impact on invagination morphogenesis of the lens pit.

We generated RhoA, Rac1 double mutants with Le-cre and assessed cell pit shape, which revealed a lens pit shape very similar to the RhoA single mutant (Fig. 1D). Although it was tempting to conclude that this revealed an epistatic relationship between RhoA and Rac1, further investigation showed that the double mutant lens pit displayed dramatically reduced levels of F-actin (Fig. S4A), abnormal patterns of cell packing (Fig. 2D) and reduced levels of the adherens junction marker β-catenin (Fig. S4 E and I) similar to those observed in lens-specific β-catenin somatic mutants (34). This all suggested that there was a fundamental loss of the actin cytoskeleton in the double mutants, and that the change in cell pit shape was unlikely to reflect the normal genetic relationship between the two GTPases.

Fig. 2.

Rac1 and RhoA mutations have opposite effects on cell shape. (A–D) Cell profiles from the lens pits of control and somatic mutants were generated from β-catenin–labeled cryosections. The profiles were exported, measured for width and height, and combined to produce the average profiles in E. (E) Apical and basal cell dimensions are indicated (above and below the profile) relative to the control basal dimension. Similarly, the absolute dimensions in micrometers are indicated for the average control cell. Significance values for apical dimension: control (2.08 ± 0.14) to RhoA (2.74 ± 0.18), P < 0.001; control to Rac1 (1.72 ± 0.12), P = 0.043; RhoA to Rac1, P < 0.001. Significance values for cell length: control (26.1 ± 0.46) to RhoA (32.6 ± 0.37), P < 0.001; control to Rac1 (23.8 ± 0.44), P = 0.008; RhoA to Rac1, P < 0.001. Basal dimensions are not significantly different. This analysis shows that Le-cre; RhoA^fl/fl cells (red) were longer and less apically constricted and that Le-cre; Rac1^fl/fl cells (blue) were shorter and more apically constricted. Control cells occupied an angle of 6.0°, whereas Le-cre; RhoA^fl/fl and Le-cre; Rac1^fl/fl cells occupied angles of 4.0° and 7.4°, respectively.

RhoA and Rac1 Lens Mutants Display Contrasting Cell Shapes.

The changes in cell pit shapes described above suggested that the Rac1 and RhoA mutants would show changes in cell shape. To assess this proposal, we recorded cell shape profiles from the entire lens pit of control and mutant embryos (n = 5 lens pits, about 350 cell profiles) at E10.5 (Fig. 2 A–C), and then generated average cell shapes (Fig. 2E) using a coordinate system. In both RhoA and Rac1 mutants, the basal cell dimension was not significantly changed (Fig. 2E). However, in RhoA mutants, the ratio of the apical to control basal dimension was 0.57 compared with 0.44 in the control (Fig. 2E). This was a subtle, but statistically significant change (P < 0.001). Accompanying reduced apical constriction, the RhoA mutants showed an unexpected cell elongation with a height index of 1.25 (Fig. 2E, red, P < 0.001). These two shape changes mean that the average RhoA mutant cell occupies an angle of 4.0° compared with 6.0° in the control. This cell shape change is consistent with the documented change in the shape of the lens pit (Fig. 1 B and D), as cells that occupy a reduced angle will generate an epithelial structure with reduced curvature.

The averaging of cell profiles from the lens pits of Rac1 mutant embryos (Fig. 2C) revealed that they were shorter than wild type with a height index of 0.91 (Fig. 2E, blue, P = 0.008). Unexpectedly, they were also more apically constricted with an index of 0.36 (P = 0.043). With these shape changes, the angle occupied by a Rac1 mutant cell is 7.4° (compared with 6.0° in the control), meaning that when combined in an epithelium, they will generate tighter curvature than average, consistent with our observation (Fig. 1 C and D). Combined, this analysis of lens pit cell shape in the Rac1 and RhoA mutants suggested that each GTPase can regulate both the height and apical width of a cell.

RhoA and Rac1 Show Both Apical and Basal Antagonism for Pathway Markers.

To investigate possible mechanisms of integration of Rac1 and RhoA activities, we labeled control and mutant lens pits for molecular markers of activity in Rac1- and RhoA-dependent pathways. It is well established that RhoA generates contractile actin via a pathway requiring RhoA activation of ROCK, ROCK phosphorylation-mediated activation of MRLC (35, 36) and suppression of myosin light chain phosphatase (MLCP) (37). Thus, phosphorylated MRLC is an activity surrogate for this pathway. For Rac1-dependent pathways, we measured the labeling intensity for the c2 subunit of the actin nucleator Arp2/3 as this is required for the generation of protrusive actin in lamellipodia (15). We also measured the level of cortactin, an enhancer of actin branch formation (30). Marker labeling was quantified over apical–basal line intervals that were distributed radially around the curve of the subject lens pit. For each genotype the data were obtained from 17 line intervals on each of five E10.5 lens pits (Fig. S5). The data from the 85 line intervals were combined by normalization to nuclear labeling with Hoechst 33258. To allow comparison of intensity profiles for epithelia of the different mutants, we also performed length normalization according to established procedures (38).

Quantification of phalloidin labeling revealed that F-actin is found in two intensities located at the apex (the major intensity) and at the base of control cells (Fig. 3 A and B, gray line). The distribution of F-actin in the lens pits of single mutants was changed very little (Fig. 3B, red and blue profiles) with the exception that the Rac1 mutant showed a slight increase in the level of F-actin in the basal half (Fig. 3B, blue). These profiles show that there was no fundamentally aberrant change in the ability of RhoA or Rac1 mutant cells to assemble or distribute filamentous actin.

Fig. 3.

Mutual antagonism of RhoA and Rac1 in regulating the cytoskeletal machinery during lens pit invagination. (A, C, E, G, I, and K) Labeling for F-actin (A), myosin IIB (C), phospho-MRLC (E), Arpc2 (G), cortactin (I), and Rac1 (K) in control and Le-cre; RhoA^fl/fl and Le-cre; Rac1^fl/fl (except Rac1) lens pit cells at E10.5. (B, D, F, H, J, and L) Quantification of the labeling shown in (A, C, E, G, I, and K) for lens pits over an apical–basal line interval (n = 5 eyes, n = 85 line intervals). The gray line shows quantification of the indicated marker in control cells and the red and blue lines, the quantification in Le-cre; RhoA^fl/fl and Le-cre; Rac1^fl/fl lens pit cells, respectively. The red and blue arrowheads in A, C, E, G, I, and K (only red) indicate the apical and basal marker labeling, and where numbered, cross-reference a peak on the quantification profiles.

RhoA mutant mice showed substantially reduced myosin II at both the apex and the base (Fig. 3 C and D, red). At the cell apex, this change was consistent with a reduced level of phospho-MRLC in the RhoA mutant (Fig. 3 E and F, red profile, apex). Although the absence of RhoA reduced the level of basal myosin II, there was a very limited reduction in the level of basal phospho-MRLC (Fig. 3 E and F, red profile, base). These observations are consistent with the established model for RhoA function where ROCK activation by RhoA leads to the formation of phospho-MRLC (36, 37). They further suggest that a RhoA–phospho-MRLC pathway is active at both the cell apex and base. When combined with the observation that RhoA mutant lens pit cells are less apically constricted, these data suggests that RhoA-dependent formation of apical phospho-MRLC is required for the contractile actin that mediates apical constriction.

Rac1 somatic mutants show a myosin IIB labeling intensity profile that is very similar to the control but shows a slightly higher, broader peak at the apex and a slightly lower peak at the base (Fig. 3 C and D, blue profile). By contrast, the phospho-MRLC labeling profile in the Rac1 mutant shows dramatic increases right across the profile but most obviously at the apex and base (Fig. 3 E and F, blue profile). Because the quantification of myosin IIB and phospho-MRLC in the RhoA mutant show that their distribution and activity is RhoA dependent, the Rac1 mutant response reveals that normally, Rac1 suppresses RhoA-dependent formation of phospho-MRLC. The presence of increased phospho-MRLC in apical and basal peaks indicates that Rac1 suppression of RhoA occurs in both locations.

In the Rac1 mutant, the labeling intensity for Arpc2 is substantially reduced in the lower half of lens pit epithelial cells (Fig. 3 G and H, blue profile). This effect is most dramatic where the normally intense peak of basal labeling is absent (Fig. 3H, blue arrowhead 3). This is in complete contrast to the Arpc2 quantification in RhoA mutant cells (Fig. 3 G and H, red profile) where there were increases in Arpc2 labeling at both the base (Fig. 3 H, red arrowhead 2) and at the apex (a broad, but modestly increased peak, Fig. 3H, red arrowhead 1). Combined, these data indicate that the level of Arpc2 at the cell base is dependent on Rac1 and in addition, that RhoA normally suppresses the Rac1-dependent basal Arpc2. The distribution of the Arp2/3 enhancer cortactin is also changed in the Rac1 mutant and shows a reduced level from the base to a point just below the apex (Fig. 3J, blue arrowhead 3). Like Arpc2, the basal level of cortactin in the RhoA mutant is opposite to that of the Rac1 mutant and shows an increase (Fig. 3J, red arrowhead 2). Thus, according to two markers of the Rac1-Arpc2/3/cortactin pathway, RhoA normally suppresses the assembly of protrusive actin complexes. Finally, with the availability of validated Rac1 antibodies (Fig. S2 D–I), we examined the relationship between RhoA mutation and Rac1 distribution. This showed (Fig. 3 K and L) that when RhoA is mutated, Rac1 immunoreactivity is not significantly increased apically (Fig. 3L, red arrowhead 1) but is significantly increased basally (Fig. 3L, red arrowhead 2). This suggests that one mechanism for RhoA suppression of Rac1 is to suppress its abundance in the base of the cell.

Discussion

Present models for tissue invagination suggest that increased polymerization of cadherin- anchored apical contractile actin provides the driving force for curving epithelia (17). Here we present data showing that both RhoA and Rac1 modulate apical constriction and cell length, coordinated by mutual antagonism. In this model (Fig. 4), the balanced regulation of apical cell width by RhoA and cell length by Rac1 determines the angle occupied by an epithelial cell and in aggregate, the curvature of an epithelium.

Fig. 4.

Schematic describing the role of RhoA–Rac1 mutual antagonism in epithelial bending in the lens pit. From the current analysis, we can infer that both Rac1 and RhoA have dual functions. Rac1 has a role in elongating cells through Arpc2 and cortactin but also suppresses the production of phospho-MRLC and thus the generation of contractile actin. By contrast, RhoA is required for apical constriction through the production of phospho-MRLC and contractile actin, but also suppresses the basal Arpc2 and cortactin complexes and thus inhibits cell elongation. In this way, a balance between the activities of RhoA and Rac1 controls the apical width and cell length. In turn, the ratio of these two dimensions controls the angle formed by the cells and, in aggregate, the curvature of the epithelium.

Our initial observations on the open invagination phenotype for the RhoA mutants and the opposite for the Rac1 mutants (Fig. 1) led us to study their respective average cell shapes. We show that Rac1 and RhoA have opposite effects (Fig. 2). Rac1 promotes cell elongation and suppresses apical constriction, whereas RhoA promotes apical constriction and suppresses cell elongation. The resulting changes in the RhoA and Rac1 pathway marker levels (Fig. 3) are consistent with the phenotypic data and lead to a model (Fig. 4) describing how balanced activities of RhoA and Rac1 control cell shape. We suggest, on the basis of existing data (15) and current observations, that one activity of RhoA is to produce contractile actin. This is consistent with the loss of apical phospho-MRLC and reduced apical constriction observed in RhoA mutant lens pit epithelial cells. Given the reduced apical level of myosin IIB in the RhoA mutant, we cannot exclude the possibility that reduced apical phospho-MRLC might in part be a consequence of reduced apical mysoin IIB availability. The dependence of lens pit epithelial cell apical constriction on Rho kinase (39) suggests that the RhoA-ROCK pathway (35, 36) is also required.

Because Rac1 somatic mutants show increased apical and basal phospho-MRLC, they are equivalent to a RhoA gain of function and this can explain why Rac1 mutants show increased apical constriction. Currently, we do not have a good explanation for the absence of basal constriction in Rac1 mutants in which the level of basal phospho-MRLC is increased, but perhaps integrin-mediated anchoring to basal lamina confers rigidity. Similarly, we argue that the primary activity of Rac1 is to produce protrusive actin (Fig. 2). This is consistent with the reduced levels of Arpc2 and cortactin at the cell base and with the observation that Rac1 mutant cells are shorter than normal. Because RhoA mutant cells show increased basal levels of Arpc2 and cortactin, we argue that they are equivalent to a Rac1 gain of function and this can explain why RhoA mutant cells are longer than normal. The increased labeling signal intensity for Rac1 at the base of RhoA mutant cells indicates that that one mechanism, by which RhoA suppresses Rac1 activity, is to control its level.

When combined, these observations argue that during invagination of the lens pit epithelium, RhoA and Rac1 are mutually antagonistic and that a balance between their activities fine tunes the apical width and cell height. In turn, the ratio of these two dimensions determines the angle that an individual epithelial cell occupies and in aggregate, the curvature of an epithelium. Because there are many examples where epithelial curvature changes dramatically over just a few cell diameters, it is likely that the ratio of Rac1 to RhoA activity can also be controlled locally, presumably by the signaling pathways known to regulate these GTPases (15). With the ubiquity of the Rho family GTPases, these are likely to be general mechanisms regulating epithelial morphogenesis. The similarity of the mechanisms of cell migration in cultured cells (15) and those uncovered here for epithelial invagination, further suggests that these mechanisms will be generally applicable.

Complex Interplay of Rho GTPase Activities Regulates Invagination.

We have previously shown that another GTPase, Cdc42, has an important function in regulating epithelial invagination in the lens system. In this case, Cdc42-dependent filopodia project from the base of the lens pit, connect to the adjacent optic cup, and appear to serve as physical tethers that assist coordinated invagination. The loss of lens pit filopodia in a Cdc42 somatic mutant results in a relative increase in invagination distance (24). Combined with the current findings, this indicates that at least three different GTPase-dependent modulations of the actin cytoskeleton, each of which produces only a subtle consequence, function concurrently to generate the full process of lens morphogenesis. There is also the very strong possibility given recent data from a culture system (40) that the presumptive retina and presumptive lens interact in a dynamic way to refine the shape of the lens pit and optic cup.

There are interesting parallels between these mechanisms and those observed in the fly (17, 27, 41) and sea urchin (25) where RhoA (Rho1 in the fly) is required for apical constriction and epithelial invagination. Furthermore, in the sea urchin, epithelial invagination is assisted by contractile filopodia that extend from secondary mesenchymal cells across the blastocoel (42, 43). Studies so far have focused on the three most studied Rho GTPases in epithelial invagination and further work is required to unravel the role of other small Rho GTPases in this process.

The current study has identified the importance of the small Rho GTPases in the process of epithelial invagination. However, invagination still occurs in the absence of each of the three common GTPases (Fig. 1 B and C) (24). Other factors are possibly involved in the process of epithelial invagination, so that knocking down one of these activities produces a subtle phenotype. Adhesion as a factor required for invagination of the lens placode has been shown by conditionally knocking out the adherens junctions (AJ) component β-catenin in the lens ectoderm using the Le-cre driver (34). The phenotype was more dramatic by reduced adhesion in the lens ectoderm leading to increased apoptosis, but an attempt at invagination was still made. Similar adhesion-deficient phenotypes have been shown in Drosophila during gastrulation (44) and in mice during follicular morphogenesis (45). Polarity has been determined as a factor for invagination as seen by the aberrant invagination of crumb mutants during Drosophila tracheal placode invagination (46). One of the emerging factors involved in invagination is the extracellular matrix (ECM), where it has recently been shown that loss in ECM components leads to lens placode invagination defects (47). Lens invagination defects were previously seen in a heparan sulfate mutant (48) and another example of ECM playing a role in invagination has been documented for the sea urchin during gastrulation (49).

Materials and Methods

RhoAfl (SI Materials and Methods) and Rac1fl (50) conditional alleles were crossed with the a lens-specific cre driver, Le-cre, to generate conditional mutants. Immunofluorescence labeling (34) and imaging, via an Apotome-based Zeiss fluorescence microscope, was used for visualizing proteins of interest.

For quantitation, curvature analysis was performed by a combination of Matlab 7.1 and Mathematica 7.0 (SI Materials and Methods), average cell shapes were obtained from membrane-labeled β-catenin cryosections and Axiovision software to measure and average cellular dimensions, and signal intensity profiles were calculated using ImageJ software on immunolabeled images and Microsoft Excel processing (SI Materials and Methods).