The cell adhesion molecule F-cadherin is expressed in Xenopus embryos at boundaries that subdivide the neural tube into different regions, including one, the sulcus limitans, which partitions the caudal neural tube into a dorsal and ventral half (alar and basal plate, respectively). Here we examine the role of F-cadherin in positioning cells along the caudal neuraxis during neurulation. First, we show that ectopic expression of F-cadherin restricts passive cell mixing within the ectodermal epithelium. Second, we show that F-cadherin is first expressed at the sulcus limitans prior to the extensive cell movements that accompany neural tube formation, suggesting that it might serve to position cells at the sulcus limitans by counteracting their tendency to disperse during neurulation. We test this idea using an assay that measures changes in cell movements during neurulation in response to differential cell adhesion. Using this assay, we show that cells expressing F-cadherin localize preferentially to the sulcus limitans, but still disperse when located away from the sulcus limitans. In addition, inhibiting cadherin function prevents cells from localizing precisely at the sulcus limitans. These results indicate that positioning of cells at the sulcus limitans is mediated in part by the differential expression of F-cadherin.
During the development of the vertebrate central nervous system (CNS), the neuroepithelium of the neural tube is divided into distinct regions with different developmental fates. This process begins during neurulation when the CNS first forms as a neural tube with distinct dorsoventral (D-V) and anteroposterior (A-P) axes. One likely outcome of neural patterning is to endow neural cells with the ability to precisely position along the neuraxis, even during periods of tissue morphogenesis when extensive cell movements occur. However, the mechanisms by which cells are localized to particular regions within the developing CNS remain poorly understood.
One mechanism proposed to determine cell position within the neural tube is differential cell adhesion as mediated by members of the cadherin family. This family of cell adhesion molecules consists of integral membrane proteins with conserved domains that allow for protein-protein interactions on both sides of the plasma membrane. The cadherin extracellular domain binds homotypically, thus promoting adhesion between adjacent cells expressing the same cadherin type. Cell adhesion, moreover, requires interactions between the conserved intracellular domain and the cytoplasmic catenin proteins which link the cadherins to the actin-based cytoskeleton (Ozawa et al., 1989). Both interactions allow the cadherins to serve as the transmembrane linkage in a specialized cell-cell contact site, termed the adherens junction (Geiger and Ayalon, 1992). By linking the actin cytoskeleton of adjacent cells, the cadherins are particularly well positioned to regulate cell movements that underlie morphogenesis. In addition, the ability of the cadherins to mediate cell sorting in vitro suggests that the differential expression of cadherins could mediate the preferential association of cells, allowing cells to position themselves within the developing CNS (Takeichi, 1991, 1995). A number of cadherins have been identified that are expressed in developing neural tissue. One of these, N-cadherin, is expressed by a majority of the neuroepithelial cells of the developing neural tube (Hatta et al., 1987). Disruption of the N-cadherin gene by homologous recombination results in homozygous mutant mice with severe neural tube defects, suggesting a role for N-cadherin in the formation of neural tissue (Hynes, 1996). Other members of the cadherin family have been identified which are expressed in subregions of the neural tube, and thus could be involved in mediating differential adhesion among neuroepithelial cells. For example, E-cadherin, which is expressed extensively outside the nervous system, displays a restricted pattern of expression in developing brain (Redies and Takeichi, 1993; Shimamura and Takeichi, 1992). R-cadherin is also expressed in restricted subdomains of the developing brain that include neuromere boundaries and a subset of developing nuclei (Ganzler and Redies, 1995; Matsunami and Takeichi, 1995; Redies et al., 1993). In addition, several of the type II cadherins and protocadherins are also expressed in subregions of neural tissue (Nakagawa and Takeichi, 1995; Redies and Takeichi, 1996; Sano et al., 1993; Suzuki et al., 1991). Thus, the differential expression of cadherin family members during neural development lends further support to the model that these molecules form subdomains within neural tissue.
We have previously described F-cadherin as a type II cadherin that is expressed by a subpopulation of neuroepithelial cells in the developing neural tube of Xenopus embryos. F-cadherin-expressing cells often lie at boundaries that subdivide the neural tube into different regions along the neuraxis (Espeseth et al., 1995). One of these boundaries, termed the sulcus limitans, divides the caudal neural tube into a dorsal half, involved in sensory function (the alar plate), and a ventral half involved in motor function (the basal plate). F-cadherin expression at the sulcus limitans first appears at the open neural plate stage (see Fig. 2) when the neural anlage is initially patterned. During neurulation, the neural anlage undergoes a morphogenetic process, called convergent-extension (C-E), during which cells tend to disperse along the D-V neuraxis. Nonetheless, the stripe of cells expressing F-cadherin remains at the sulcus limitans, and even sharpens in width during neurulation (Espeseth et al., 1995). Thus, if F-cadherin prevents cells from dispersing during C-E, then this might explain how cells that express F-cadherin localize to the sulcus limitans during neural tube formation.
Here we test the role of F-cadherin in determining cell position along the D-V axis of the caudal neural tube. We first show that ectopic expression of F-cadherin restricts passive cell mixing within the ectoderm, creating a boundary between F-cadherin and non-F-cadherin-expressing cells. We then employ a transplantation assay to determine whether ectopic expression of F-cadherin can influence cell position during neural tube formation. Using this assay, we show that cells-expressing F-cadherin localize preferentially to the sulcus limitans, but still disperse when located away from the sulcus limitans. We also show that cells expressing an inhibitor of F-cadherin fail to localize precisely at the sulcus limitans. Finally, we show that cells disperse even further than normal along the D-V axis of the neural tube when they express a pan-inhibitor of cadherin function. These observations, supplemented with a computer model of cell dispersion during neurulation, suggest that cells require F-cadherin, along with other factors, to localize to the sulcus limitans during neural tube formation.
MATERIALS AND METHODS
In vitro RNA synthesis
A template for the in vitro synthesis of F-cadherin RNA was constructed by inserting the F-cadherin cDNA into the CS2 vector (Turner and Weintraub, 1994), and linearizing the vector DNA with NotI. A myc-tagged version of F-cadherin was constructed by fusing in-frame, 6 copies of the myc-epitope sequence at the carboxyl terminus (Turner and Weintraub, 1994). The CBR-MT construct has been described previously (Riehl et al., 1996). The cDNA encoding ΔF-cadherin was constructed by introducing the 6 copies of the myc epitope into the F-cadherin sequence 8 amino acids after the transmembrane domain, followed by a stop codon. The myc-tagged versions of F-cadherin, ΔF-cadherin and CBR were used to show that protein expression from injected RNAs persisted out to neurulae stages. Capped RNAs encoding F-cadherin and ΔF-cadherin were synthesized in vitro using standard techniques (Kintner, 1988).
Ectoderm mixing assay
In vitro transcribed mRNAs encoding β-galactosidase or myc-tagged forms of F-cadherin and ΔF-cadherin (0.25-1.0 ng/embryo) were injected into single blastomeres of 8-16 cell embryos. After embryos had completed gastrulation (st. 12.5), they were fixed in MEMFA, and either stained in whole-mount for β-galactosidase expression in 0.15% X-gal in 0.1 M ferricyanide, 0.1 M ferrocyanide, and 0.1 M sodiun phosphate, pH 6.3, or for the myc-epitope using a monoclonal antibody (Evan et al., 1985) and HRP immunohistochemistry. After staining, embryos with a patch of labeled cells within the ventral ectoderm were selected and mounted under a coverslip. Using camera-lucida, the edge of patches of labeled cells was drawn, as well as any labeled cells lying outside of the patch, or unlabeled cells lying within the patch of labeled cells. The number of isolated cells in a measured area was counted for at least five embryos injected with a given RNA, and then expressed as the average number of isolated cells per square millimeter of epithelium.
Transplantation assay and bin analysis
Synthetic nLacZ RNA (0.4 ng/embryo) or a mixture of nLacZ and experimental RNA (1.0 ng/embryo) was injected, in a volume of 10 nl, into the animal pole of all four blastomeres of donor embryos at the four cell stage (Coffman et al., 1993). Tissue (15-30 cells) was isolated from the ectoderm of donor embryos at stage 10, and placed into host embryos shortly after the formation of the blastopore lip (st. 10.5). In the case of alar plate transplants, donor cells were placed between 45 and 90 degrees from the center of the blastopore lip of the host, while for basal plate transplants, the donor cells were placed between 20-45 degrees. After host embryos had reached neurulae stages, they were fixed in MEMFA, reacted for β-galactosidase expression in 0.15% salmon-gal (Molecular Probes) or X-gal as described above and then double stained by whole-mount in situ hybridization (Harland, 1991) for expression of F-cadherin RNA (Espeseth et al., 1995). After postfixing in MEMFA, stained embryos were embedded in Paraplast, sectioned at 10 µm on a rotary microtome, and mounted on glass slides in Permount. Using camera lucida, a grid subdividing the neural tube was used to assign donor nuclei into each bin along the dorsal ventral axis (see Fig. 4). Embryos with 20-220 cells were included in the subsequent analysis of cell distributions. The distribution of transplanted cells was plotted for each embryo by assigning the cells to bins as described in Fig. 4. Average distributions were obtained by first assigning each transplant a peak location corresponding to the bin containing the largest number of transplanted cells, and then averaging the data from transplants with the same peak location.
In order to compare the spread of transplant distributions for the various kinds of transplants, we also looked at the full width at half maximum (FWHM) of transplant distributions in individual embryos; this was measured in number of bins and the values ranged between 1 and 6. We calculated the FWHM means and standard deviations for two groups of embryos for each kind of transplant: one group consisted of the embryos that had at least 30% of transplant cells at the sulcus limitans, the ‘sulcus limitans group’, and the other contained the rest of the embryos, the ‘away from the sulcus limitans group’. We carried out pairwise comparisons of these means for different kinds of transplants and tested the statistical significance of their differences using a t-test. This analysis was done for embryos with dorsal transplants; there were not enough embryos with ventral transplants to carry out meaningful statistical comparisons. Embryos with bimodal transplant cell distributions were excluded from the FWHM analysis. The results of the FWHM analysis are presented in Table 3.
Computer simulations were used to model the movements of transplanted cells in the neural plate during neurulation. The tissue is modeled as a 2-dimensional rectangular grid at the vertices of which the cells are placed. At the start of a simulation the tissue has 18(A-P)×56(D-V) cells, each placed on a single vertex; a 6×6 patch of those cells (corresponding to a transplant) is labeled differently from the rest (see Fig. 10 B,C); each cell has an adhesion value indicating the adhesion strength expressed by the cell (we assume a default adhesion for all cells).
More specifically, f takes a default value 1.0 for all kinds for cell-cell contacts, which would correspond to non-specific adhesive interactions and friction. For border-cell/border-cell interactions f takes the value 4.0 (= 3.0 plus the default value). For patch-cell/patch-cell interactions, depending on the condition being tested, f takes either the value 1.0 (= just the default value), which corresponds to the ‘patch non-adhesive to itself’ condition, or the value 2.0 (= 1.0 plus the default value), which corresponds to the ‘patch adhesive to itself’ condition. For patch-cell/border-cell interactions, again depending on the condition being tested, f takes either the value 1.0 (= just the default value), which corresponds to the ‘patch non-adhesive to border’ condition, or the value 4.0 (= 3.0 plus the default value), which corresponds to the ‘patch adhesive to border’ condition.
During 25% of the time, or when no moves have negative energy, a move is selected at random from all the possible moves. This process is repeated about 100 times during each diffusion-adhesion step at randomly chosen grid positions which uniformly cover the whole extent of the tissue.
After several rounds of such rearrangement and diffusion-adhesion steps the tissue has elongated (as the neural tube does) and shrunk in width to a final grid with 16 points in the D-V direction (see Fig. 10B,C); we run one hundred simulations under each condition and gather statistics about the final distribution of patch cells on the grid (see Fig. 10D). Such information can be ‘collapsed’ to one dimension to produce histograms of the distributions of patch cells in the D-V direction, as in Fig. 6 of the Results.
F-cadherin inhibits cell mixing in ventral ectoderm
The predicted structural features of F-cadherin suggest that it mediates homotypic adhesion, as reported for other members of the type II cadherin family (Nakagawa and Takeichi, 1995; Suzuki et al., 1991). If so, one would predict that embryonic cells that differentially express F-cadherin would stick together and not mix with cells not expressing F-cadherin. As a first test of this prediction, we ectopically expressed F-cadherin in the ectoderm of early Xenopus embryos. RNA encoding a myc-tagged form of F-cadherin was injected into one animal blastomere at the 8-16 cell stage, thus restricting the expression of the injected RNA to a subpopulation of ectodermal cells. The embryos were allowed to develop through gastrulation which normally causes passive cell mixing among ectodermal cells (Detrick et al., 1990). At stage 12.5, embryos were fixed and stained for the expression of the epitope-tagged F-cadherin in order to determine the degree of mixing between F-cadherin-expressing cells and non-expressing cells (Fig. 1). The results show that cells expressing F-cadherin mixed much less with non-expressing cells than did cells in control embryos injected with lacZ alone (Fig. 1D). In addition, staining with the myc antibody shows that F-cadherin is concentrated at the site of contact between F-cadherin-expressing cells, suggesting that F-cadherin mediates adhesion by homotypic binding (Fig. 1B). Thus, these results show that differential expression of F-cadherin can suppress cell mixing within the embryonic ectoderm.
We constructed an antimorphic form of F-cadherin, called ΔF-cadherin, which lacks the intracellular domain. Similarly deleted forms of other members of the cadherin family have been shown to be markedly deficient in adhesion in cell mixing assays, and even to act as dominant-negative mutants, presumably by binding extracellularly to like-cadherins, without interacting with cytoplasmic components required for adhesion (Kuhl et al., 1996; Lee and Gumbiner, 1995; Levine et al., 1994). As predicted, ΔF-cadherin did not significantly affect cell mixing when ectopically expressed in the ectoderm, suggesting that the F-cadherin cytoplasmic domain is required for its function (Fig. 1 C,D). To ask whether ΔF-cadherin can act as a dominant-negative mutant, we mixed F-cadherin and ΔF-cadherin mRNA in different ratios, and introduced these mixtures into the ectoderm. The results show that when F-cadherin and ΔF-cadherin are expressed together, ΔF-cadherin inhibits the adhesive activity of F-cadherin, suggesting that ΔF-cadherin can act as a dominant negative inhibitor of F-cadherin (Fig. 1D).
Ectopic expression of F-cadherin disrupts C-E
Expression of F-cadherin is first detected at stage 12.5 (open neural plate stage) in two lateral domains of the neural plate that correspond in position to the prospective sulcus limitans (Espeseth et al., 1995). At this stage, the width of the caudal neural plate along its prospective D-V axis is two-fold greater than its length along the A-P axis. The neural anlage then undergoes a remarkable transformation in shape by narrowing along the D-V axis and lengthening along the A-P axis, resulting in a neural tube which is ten-fold longer along the A-P axis, than along the D-V axis. This change in shape, called convergent-extension (C-E), is brought about in part by active cell intercalation, resulting in extensive cell mixing along both the D-V and A-P neuraxis (Keller et al., 1992). During neural tube formation, F-cadherin expression is detected on a coherent stripe of cells that lie at the future sulcus limitans (Fig. 2). Thus, these observations suggested that differential expression of F-cadherin might contribute to the localization of cells to the sulcus limitans during a period of active cell movement.
As a first test of whether differential expression of F-cadherin might affect cell movements during C-E, we ectopically expressed F-cadherin on one side of the embryo by RNA injections at the two-cell stage (Fig. 3). Embryos were left to develop until late neural plate stages, and the extent of C-E was then measured by staining the embryos with a probe for N-tubulin, which marks the formation of primary neurons.
Primary neurons first form at the open neural plate, in three longitudinal domains on both sides of the midline (Chitnis et al 1995). During neurulation, these domains undergo a change in shape and position that mirrors the overall change in shape of the neural plate during C-E. In embryos ectopically expressing F-cadherin, the stripes of primary neurons form properly but fail to converge and extend normally on the injected side (Fig. 3). This result indicates that ectopic expression of F-cadherin perturbs C-E, in line with previous observations showing thatperturbation in cadherin function can alter C-E in the mesoderm (Kuhl et al 1996; Lee and Gumbiner, 1995). In addition, we ectopically expressed ΔF-cadherin on one side of Xenopus embryos by RNA injection and assessed the effects of this on C-E by staining for the formation of primary neurons (Fig. 3; Table 1). In contrast to F-cadherin, ectopic expression of ΔF-cadherin did not appear to have a dramatic affect on C-E as measured by the pattern of primary neurons, suggesting as above that the cytoplasmic domain is required for F-cadherin function (Figs 1, 3; Table 1). In addition, examination of the ΔF-cadherin-expressing embryos revealed none of the lesions in neural or epidermal tissues that were produced when embryos were injected with similar truncated forms of N-cadherin and E-cadherin, respectively (data not shown; Levine et al., 1994). Thus, these results indicate that restricted expression of F-cadherin is required for C-E to occur normally. In addition, these results indicate that ΔF-cadherin does not give the same effects as similar truncated forms of E- and N-cadherin, suggesting that it not interfering with the function of other cadherins (Levine et al., 1994).
The observations described above suggest a model where the differential expression of F-cadherin affects cell movements during neurulation, thus helping to position cells at the sulcus limitans. In order to measure the effects of F-cadherin expression on cell movements during C-E, we developed an assay based on transplanting marked cells into host embryos, and then following these marked cells during neural tube formation (also see Guthrie et al., 1993). As diagrammed in Fig. 4, small groups of cells from nLacZ RNA-injected donor embryos were transplanted into the prospective neural plate of host embryos prior to the onset of C-E. At neural tube stages, when C-E is largely complete, the embryos were fixed, and stained for the expression of nlacZ and F-cadherin, allowing the distribution of the transplanted cells relative to the sulcus limitans to be determined. As shown in Fig. 4, grafted cells incorporate into the host neural tissue and intercalate during C-E, spreading over considerable distances and mixing extensively with host cells. To measure the extent of these movements, we scored the distribution of transplanted cells in each embryo by dividing the neural tube into a number of radial bins along the D-V axis and scoring the number of cells in each bin (Fig. 4F). We analyzed the overall distribution of the transplanted cells by pooling data from transplants with the same peak location, defined as the bin containing the largest percentage of donor cells, allowing us to generate an average distribution. Plots of these average distributions give one indication of how the transplanted cells distributed within each region of the neural tube during C-E. In addition, the full width at half maximum (FWHM) was analyzed for individual transplant distributions as described in the Materials and Methods.
Bin analysis of nLacZ-expressing control cells showed that cell movements during C-E varied in nature depending on the position of the transplanted cells along the D-V axis of the neural tube. In cases where the peak location occurred towards the middle of the alar plate (Fig. 5A), the transplanted cells showed a symmetrical distribution along the D-V axis of the neural tube. A different result was obtained when the peak location of the transplanted cells occurred at the sulcus limitans (Fig. 5B), which is marked by the expression of F-cadherin. In these cases, the transplanted cells showed an extremely skewed distribution, with most cells on the alar side and very few on the basal side. We also examined the distribution of cells that were transplanted medially, thus directing them into the basal plate of the host neural tube. As was the case with alar plate transplants, the cells in a transplant with a peak location within the middle of the basal plate were symmetrically distributed along the D-V axis (Fig. 5C), while those with a peak location at the sulcus limitans showed an extremely skewed distribution (Fig. 5D). Thus, these results show that cells disperse within the alar and basal plates during C-E but are restricted from crossing between the alar and basal plates. In addition, cells in both alar and basal plate transplants entered into the stripe of F-cadherin expressing cells, indicating that both alar and basal plate cells contribute to the position in the neural tube where F-cadherin is expressed.
Computer simulation of cell distributions during neural plate elongation
To examine how the movements of cells during neural plate elongation affect their final distribution along the D-V neuraxis, we developed a computer simulation in which cells move randomly within a tissue but are restricted in their movements at a boundary. In this model, which is described in detail in the Material and Methods, cells are placed on a grid that then elongates via a repeating succession of two steps; a step of random cell rearrangement on the grid and a step of cell movement subject to adhesive forces and opposing diffusion-like processes that tend to spread cells uniformly on the grid. Cells can move up or down and left or right within the interior of the grid, while cells on the left and right borders of the tissue move only to the right or left, respectively, with some bias for moving towards the middle of the D-V axis and outwards in the A-P axis. During each iteration, the grid converges along one axis and extends along the other axis.
When a small group of cells are placed away from the edge of the cell array and disperse during the simulation, their final symmetric distribution mirrors the experimental distribution of transplanted cells whose peak location occurred in the center of the alar or basal plate (compare Fig. 6A with Fig. 5A or C). It is interesting to note that cells spread out as in the experimental case, despite the movement bias mentioned above, which would tend to restrict the spread. In contrast, if a patch of cells is placed closer to the border of the cell array, their subsequent asymmetric distribution mirrors the experimental distribution of cells whose peak location occurs at the sulcus limitans (compare Fig. 6B with Fig. 5B or D). Thus, the experimental measurements of cell dispersion during neurulation are consistent with a model where cells, as they undergo cell rearrangements, move randomly within the alar or basal plates, but encounter a boundary to cell movements at the sulcus limitans.
Ectopic expression of F-cadherin in transplanted cells
The results from the transplantation assay suggest that during C-E, cells disperse along the D-V axis of the neural tube by cell intercalation even though cell movements appear to be restricted across the sulcus limitans, where F-cadherin is expressed. This observation suggests several potential roles for F-cadherin in terms of regulating cell movements during C-E. One possibility is that F-cadherin expression contributes to the restriction in cell movements between the alar and basal plate. Another possibility is that F-cadherin is not responsible for this boundary, but only for localizing cells at the sulcus limitans. If F-cadherin does indeed localize cells to the sulcus limitans, it might do so on its own by preventing cell dispersion during C-E, or else it might require other factors present at the sulcus limitans in order to influence cell movements. To examine these different possibilities, we asked how the ectopic expression of F-cadherin would affect the position of transplanted cells during C-E. Accordingly, donor cells taken from embryos injected with F-cadherin and nLacZ RNAs were transplanted into host embryos and subjected to the analysis described above. Control experiments indicated that transplanted cells continued to express exogenous F-cadherin protein at high levels out through tadpole stages (data not shown).
Ectopic expression of F-cadherin substantially changes the distribution of transplanted cells within the neural tube by several criteria. First, of all the F-cadherin transplants analyzed, over half gave a peak location at the sulcus limitans while only a quarter of the control transplants fell into this category (Table 2). Secondly, while control transplants included a number of examples where the peak location occurred in Bin 3, adjacent to the sulcus limitans, only one of the F-cadherin transplants gave this distribution (Table 2). Both of these observations suggest that F-cadherin-expressing cells were preferentially localized to the sulcus limitans. Indeed, of the total cells analyzed, 27% of the F-cadherin-expressing cells and 13% of the control cells were localized at the sulcus limitans. Thirdly, the distribution of cells in transplants with a peak location at the sulcus limitans showed less of a spread by FWHM analysis than similar transplants of nLacZ-expressing cells (Table 3; compare Fig. 7B with Fig. 5B). Finally, when cells expressing F-cadherin were transplanted into the middle of the alar plate, their subsequent distribution in the neural tube tended more often to be bimodal, with a second smaller peak occurring at the sulcus limitans (compare Fig. 7A with Fig. 5A). When all transplants into the alar plate are viewed individually, 26% (7/27) of the F-cadherin-expressing transplants, but only 5% (2/38) of the nLacZ-expressing transplants have bimodal distributions. Thus, all of these observations are strong indications that the differential expression of F-cadherin biases cell position to the sulcus limitans during neural tube formation.
F-cadherin appears to be necessary but not sufficient for localizing cells along the D-V axis of the neural tube
The results described above suggest that F-cadherin localizes cells to the sulcus limitans. However these results also suggest that F-cadherin does not mediate this localization by simply limiting the general dispersion of cells during C-E. Thus, in the transplants of F-cadherin-expressing cells whose peak location occurred in the middle of the alar plate, the spread of cells along the D-V axis was not statistically different from that of control cells expressing β-galactosidase as determined by FWHM analysis (Table 3; compare Fig. 7A with Fig. 5A). In addition, the A-P distribution of F-cadherin-expressing cells did not significantly differ from that of control cells, indicating that F-cadherin does not affect the overall extent of cell movements during C-E (data not shown). Both of these results indicate that F-cadherin does not have a general effect on cell dispersion, or restriction of cell movements during C-E.
To account for how F-cadherin might localize cells during C-E without generally affecting cell dispersion, we examined the effects of differential cell adhesion on the distributions of cells in our computer simulations. In the first case, we examined a situation where cells in the patch are adhesive both to themselves and to the border of the array. When an ‘adhesive’ patch of cells is positioned away from the edge of the array, the final distribution of these cells is narrower than that obtained for ‘nonadhesive’ cells, as expected if adhesion acts to reduce the spread of cells undergoing C-E (Fig. 6C). This outcome is in contrast to that observed experimentally where ectopic expression of F-cadherin does not appear to narrow the final distribution of cells along the D-V axis (Fig. 7A and FWHM analysis in Table 3). We then examined a situation using the model where the cells in the patch are not adhesive to themselves but only to the border of the array. When such cells are started away from the border of the array, they produce bimodal distributions with one peak at the edge of the array. These cells produce a bimodal distribution for start positions that would have led to symmetric distributions had the added adhesion not been there (Fig. 6E). This result mirrored that experimentally observed, i.e. increased incidence of bimodal distributions for F-cadherin-expressing transplants with a peak location at the middle of the alar plate (compare Fig. 6E to Fig. 7A). In addition, when a patch of cells, which is adhesive for the border, is started closer to the border of the array, the subsequent distribution is narrower, as also observed experimentally (compare Fig. 6F with Fig. 7C). Thus, the changes in cell position that are observed in F-cadherin-expressing, transplanted cells can be explained by a model where F-cadherin affects the adhesion of cells, and thus cell movements, only when these cells are at the sulcus limitans where F-cadherin is normally expressed.
ΔF-cadherin prevents cells from localizing precisely at the sulcus limitans
The results obtained above indicate that F-cadherin is one factor required for localizing cells to the sulcus limitans. However, F-cadherin may only contribute substantially to this localization when other factors at the sulcus limitans come into play. To examine this further, we next asked how the inhibition of F-cadherin function might affect cell movements depending on where cells lie along the D-V axis of the neural tube. Accordingly, we transplanted cells expressing ΔF-cadherin and determined their distribution using bin analysis (Fig. 8). Significantly, the distribution of cells in transplants expressing ΔF-cadherin differed from control cells for those whose peak location occurred at the sulcus limitans. In these cases, cells expressing ΔF-cadherin appeared to have crossed the boundary marked by F-cadherin expression, and this was true when transplants were made into either the basal or alar plate (Fig. 8). Thus, these results indicate that F-cadherin is necessary for cells to localize at the sulcus limitans during neurulation and that cells disperse from this location when F-cadherin function is inhibited. Importantly, among the transplants of ΔF-cadherin-expressing cells, those with peak locations away from the sulcus limitans had a symmetric D-V distribution that was not significantly different from that of control cells by FWHM analysis (Table 3, compare Fig. 8A to Fig. 5A). Therefore, ΔF-cadherin affects the movements of cells only at the sulcus limitans where F-cadherin is normally expressed.
Inhibition of cadherin function alters position along the D-V axis
The results described above indicate that F-cadherin is likely to be only one factor determining localization of F-cadherin-expressing cells to the sulcus limitans. To ask more generally whether cadherin-mediated adhesion contributes to this localization, we examined the effects of generally disrupting cadherin function on the positioning of transplanted cells. We have previously shown that constructs containing just the intracellular domain of N-cadherin are pan-inhibitors of endogenous cadherin function when expressed in Xenopus embryos (Kintner, 1992; Riehl et al., 1996). When expressed at high levels, these cytoplasmic constructs cause extensive cell dissociation, but can be titered to a level at which cell dissociation does not occur but adhesion seems to be affected as shown by a general inhibition of C-E (data not shown). Thus, we used one of these cytoplasmic inhibitors, called CBR, to inhibit cadherin-mediated adhesion in transplanted cells during C-E. Cells were taken from embryos injected with CBR RNA (5 pg/embryo), transplanted into host embryos and analyzed for their D-V distribution as above. The results show (Fig. 9) that cells expressing CBR behave significantly differently from control cells in terms of their distribution along the D-V axis of the neural tube: away from the sulcus limitans, the average spread of CBR transplant distributions in the FWHM analysis is larger than that of controls (Table 3). Also, cells expressing CBR tend to cross the sulcus limitans more frequently (Fig. 9B,D), much more than cells expressing ΔF-cadherin (Fig. 8B). Thus, while 38% of all CBR transplants (13/34) had more than 10% of the cells which crossed the sulcus limitans, only 22% of all ΔF-cadherin transplants (5/31), and 13% of all control transplants fell into this category (9/70). These results support the idea that cadherin function contributes to the positioning of cells along the D-V neuraxis during C-E, and that F-cadherin might be only one of several factors contributing to the localization of cells to the sulcus limitans.
During the patterning of the vertebrate CNS, cells acquire a regional identity that reflects their position along the neuraxis. One likely outcome of neural patterning is to endow some cells with the ability to localize within the developing CNS, even during periods of active cell movement such as those occurring during neurulation. The well-established properties of the cadherins are consistent with the idea that differential cell adhesion could be a factor in how cells determine their position within the CNS during neural morphogenesis. For example, in vitro studies have shown that cadherins can promote sorting in cell aggregation assays (Nose et al., 1988; Steinberg and Takeichi, 1994). In addition, dissociated embryonic tissue has been shown to reaggregate in a cadherin specific manner (Matsunami and Takeichi, 1995). The differential expression of cadherins by different tissues in early development and within tissues during regionalization has also suggested a role for cadherins in cell segregation in vivo (Choi and Gumbiner, 1989; Dahl et al., 1996; Ganzler and Redies, 1995; Hatta et al., 1987; Kimura et al., 1995; Kintner, 1992; Nakagawa and Takeichi, 1995; Redies, 1995). Finally, the ectopic expression of N-cadherin in Xenopus appears to create ectopic boundaries within tissues (Detrick et al., 1990; Fujimori and Takeichi, 1993). These observations have suggested that differential cadherin expression might contribute to the localization of neural cells to different regions, but direct experimental evidence for a such a role has been notably lacking.
F-cadherin expression marks a subdomain in the caudal neural tube
F-cadherin is expressed by a small population of cells that lie at the boundary between the alar and basal plate (Espeseth et al., 1995). A number of other genes show a pattern of expression that is restricted to this region in the hindbrain and spinal cord, including XASH-3 in Xenopus, Dbx in mouse, and the Dbx homologs hlx-1 in zebrafish and CHoxE in chick (Fjose et al., 1994; Lu et al., 1992; Rangini et al., 1991; Zimmerman et al., 1993). These observations suggest that the sulcus limitans represents a subdomain of the caudal neural tube, containing cells distinct from other dorsal and ventral cell types. The early expression of F-cadherin indicates that this subdomain might form as early as stage 13 when neurulation begins (Espeseth et al., 1995). In addition, the results of the transplantation assay indicate that both dorsal and ventral cells contribute to this subdomain. Thus, we suggest that cells at the sulcus limitans are derived from both the alar and basal plate, perhaps from cells that lie at the boundary between these two domains, and that these cells appear to form a distinct population during neural tube formation, as evident from the stripe of F-cadherin-expressing cells at this stage (Fig. 2). Moreover, in this model, the expression of F-cadherin by the sulcus limitans cells would presumably allow them to maintain their position in the neural tube during neurulation.
F-cadherin is necessary, but not sufficient, for the localization of cells into a stripe at the sulcus limitans
Using a transplantation assay, we examined the role of F-cadherin in positioning cells at the sulcus limitans during neural tube formation. Our results suggest strongly that differential adhesion is one factor required for cells to localizeat the sulcus limitans. When transplanted cells express F-cadherin, their localization to the sulcus limitans is markedly increased (Fig. 7). In addition, when F-cadherin function is inhibited in transplanted cells using ΔF-cadherin, cells fail to localize precisely to the sulcus limitans as is evident from the fact that they cross from the basal into alar plate, or vice versa (Fig. 8). Thus, these results suggest strongly that the differential expression of F-cadherin is required by cells to localize to the sulcus limitans during the extensive cell movements that occur during neurulation.
Ectopic expression of F-cadherin in the surface ectoderm produces a sharp boundary between expressing and nonexpressing F-cadherin cells (Fig. 1). Thus we first hypothesized that a similar boundary might form between F-cadherin-expressing and nonexpressing cells within the neural tube during neurulation. In this model, the initial broad domain of F-cadherin expression observed at the open neural plate stage would give rise to the subsequent narrow stripe of F-cadherin-expressing cells in the neural tube, simply because F-cadherin-expressing cells tend not to intercalate with non-expressing cells during C-E. This, however, does not appear to be the case. Cells ectopically expressing F-cadherin intercalate with non-expressing cells as well as control cells during C-E, as is evident from the spread of transplanted F-cadherin-expressing cells along both the D-V and A-P neuraxis (Fig. 7, data not shown). We suggest, therefore, that in terms of regulating cell movements during C-E, the differential expression of F-cadherin is a factor only when cells come in contact with the sulcus limitans where F-cadherin is normally expressed. Indeed, the cell distributions that we observed experimentally can be explained in a computer-simulated model by a case where cells do not have adhesion for themselves (which tends to narrow the final distribution) but instead by one where cells are only adhesive for the edge of the array (i.e. the sulcus limitans). This predicts that the differential expression of F-cadherin only influences cell movements at the sulcus limitans where additional factors come into play but is largely irrelevant elsewhere. Among the other factors that could regulate cell position at the sulcus limitans is the restriction in cell movements between the alar and basal plate that can be observed in the transplantation assay. This restriction appears to be co-incident with the F-cadherin-expressing cells and a further indication that this region of the neural tube may be specialized in several different ways. A simple model to explain these observations is that several cadherins work together to account for the final distribution of sulcus limitans cells. This model would account for the more dramatic dispersion of cells observed when cadherin function is more generally inhibited with CBR (Fig. 9).
Cell adhesion and neural patterning
We have shown that the differential expression of F-cadherin is one factor determining cell position along the D-V axis of the neural tube. This finding may have implications for the processes that pattern the caudal neural tube along its D-V axis. Recent studies suggest that ventral cell types in the neural tube are specified by sonic hedgehog or a related molecule which is secreted from the notochord, while dorsal cell fates appear to be specified by BMP-like molecules, perhaps provided dorsally by the nonneural ectoderm (Ericson et al., 1996; Liem et al., 1995; Tanabe et al., 1995). Moreover these signaling molecules appear to act as morphogens in that they specify different cells fates along the D-V axis as a function of their concentration (Roelink et al., 1995). If so, then factors such as F-cadherin will indirectly affect cell fate by determining the position of cells relative to the source of the dorsal and ventral morphogens. Thus, neural patterning may involve an interplay between the signals that initially specify cell fate according to position within the neural tube, and factors such as F-cadherin which determine position by mediating differential adhesion.
We wish to thank Roger Bradley and Sam Pfaff for their thoughtful comments on the manuscript and Ajay Chitnis for simulating discussions. We also wish to thank the anonymous reviewers for their detailed and helpful comments. C. K. acknowledges support from the NIH and the McKnight foundation.