A segmented body plan is prevalent throughout the animal kingdom. It is thought that during evolution, the modulation of segment shape, size, and number has contributed to the diversity of organisms. In vertebrates, segmentation involves the partitioning of the presomitic mesoderm (PSM) into numerous metameric units called somites. This process proceeds in an anterior-to-posterior sequence, with each bilaterally symmetric somite pair forming at a consistent time interval. Somite formation begins in the PSM where cells at the intersection of opposing caudorostral Wnt/fibroblast growth factor (Dubrulle and Pourquie,2004) and rostrocaudal retinoic acid gradients (Moreno and Kintner,2004) become competent to respond to segmentation cues involving the Notch pathway genes (Pourquie,2003). This process is then followed by the expression of genes associated with the separation of somites from the unsegmented PSM, such as protocadherin (PAPC), which is expressed in the anterior halves of prospective somites (Kim et al.,2000; Rhee et al.,2003); the Eph receptor, which is also expressed in the anterior half of the presumptive somite; and the Eph ligand, ephrinB2, which is expressed in the posterior half of the same presumptive somite (Durbin et al.,1998).
Although such studies provide a clearer understanding of the molecular pathways involved in somite formation, the cell behaviors driving this process remain elusive. Two notable exceptions to this are found in chick and zebrafish where somite formation is amenable to time-lapse analysis. In the chick, Kulesa and Fraser (2002) have shown that somite formation involves a complex pattern of cell behaviors referred to as the “ball-and-socket” mode of tissue separation in which the epithelializing somite moves apart from the PSM. In the zebrafish, Henry et al. (2000) showed that somite formation consists of the short-range movements of epithelializing presumptive border cells that form a boundary separating the newly formed somite from the PSM. Thus, although both chick and zebrafish form epithelial-type somites, the cell behaviors underlying this process are very different.
Somite formation in X. laevis is different from most vertebrates because it does not form an epithelial somite. In X. laevis, the paraxial mesoderm forms blocks of cells that undergo a 90° rotation to form myotome fibers that are aligned parallel to the notochord (Hamilton,1969; Youn and Malancinski,1981). Due to the high opacity of these tissues, visualization of somite segmentation and rotation in X. laevis has been challenging. To circumvent this problem, Wilson et al. (1989) used X. laevis explants in which the archenteron roof was removed, thereby exposing the ventral surface of the paraxial mesoderm. Using this approach, they showed that somite formation consists of the expansion of short discrete fissures that bisect the PSM to form a stable segmental boundary. Although this study demonstrated a subset of the cell behaviors that lead to the separation of the somite from the PSM, the aforementioned tissue opacity prevented them from observing detailed cell behaviors associated with somite rotation. In addition, their analysis was performed on explants that had typically formed only the first six somites, thus preventing a more comprehensive analysis of somite formation along the entire axis. A previous study using scanning electron micrography (SEM) of tissue fractures showed that individual cells within a somite bend during rotation and that cells positioned in the central region of the somite begin the rotation process (Youn and Malancinski,1981). While these studies have provided us with a foundation for understanding the cell behaviors underlying somite segmentation and rotation, many of the details remain elusive.
Given that the cell behaviors underlying somite segmentation and rotation occur in a three-dimensional context, we decided to use a confocal imaging approach that would allow us to visualize individual cell morphologies along the anteroposterior, dorsoventral, and mediolateral axes. This approach involves the use of a membrane-targeted green fluorescent protein (GFP; Kim et al.,1998) to enable the visualization of cell shapes during the segmentation and rotation of somites. Our data reveal that somitogenesis consists of the following four steps: (1) the mediolateral elongation of anterior PSM cells, (2) the formation of filopodial protrusions at the onset of rotation, (3) the bending of elongated cells around the dorsoventral axis, and (4) the alignment of myotome fibers parallel to the notochord. We find that these cell behaviors are consistent along the entire length of the anteroposterior axis, with the exception that somite segmentation and rotation are asynchronous at the anterior and become more synchronous toward the posterior end of the developing embryo. Lastly, we show that the overall pattern of somite morphogenesis follows a dorsal-to-ventral progression, with cells in the dorsal aspect of the somite completing alignment before cells in other regions of the same somite. Together, our results provide a more detailed view of X. laevis somitogenesis.
In X. laevis, somite formation begins at stage 17 and continues until stage 40 at which time 45 pairs of somites have formed (Nieuwkoop and Faber,1967). The process begins at the anterior end of the paraxial mesoderm with the formation of fissures that bisect the PSM to form a somite that is designated as “S0” (Pourquie and Tam,2001). Within the PSM are groupings of cells that are organized into prospective somites that are referred to as somitomeres S-I through S-III, which reflects their progressive distance posterior to S0 (Fig. 1A). As somite maturation progresses posteriorly, the newest forming somite (S0) will complete the necessary cell movements to become a mature somite (SI). Consequently, somitomere S-I becomes S0, as it is now positioned at the anterior end of the PSM. In this manner, somite maturation progresses posteriorly at precise, species-specific time intervals. In this study, we describe the cell shapes associated with different stages of somite maturation (S-III to SIII) in fixed and cleared embryos. We also compare the cell morphologies that are present during the formation of anterior somites, which occur between stages 17 and 20 (1 to 6 somites), to those that are present during the formation of more posterior somites, which occur between stages 24 and 28 (15 to 22 somites).
Cell Shapes Associated With Somite Segmentation and Rotation
To obtain a three-dimensional view of the cell shape changes occurring during somitogenesis, we injected embryos at the one-cell stage with an mRNA encoding a membrane-targeted GFP (GAP43-GFP). The embryos were then allowed to develop to stage 24, at which time they contain approximately 15 somites, and then were fixed and cleared for imaging. We collected stacks of confocal images along the dorsoventral and mediolateral axes to gain a complete view of the cell shapes associated with somite segmentation and rotation. Using this approach, we found that cells in the PSM are distinctly columnar as indicated by both their mediolateral elongation in the dorsal perspective (Fig. 1B; S-I to S-III), and their round cross-sections in a similarly staged embryo in the lateral perspective (Fig. 1C; S-I to S-III). In a more posterior region of the PSM (S-III), cells appear less elongated along the mediolateral axis than cells in the more anteriorly positioned region of the PSM (S-I; Fig. 1B). To confirm this observation, we measured the average length/width ratio (LWR) of five cells in S-I and S-III in eight different embryos (Fig. 1D). We found that cells positioned in S-I had an average LWR of 11.7, whereas cells positioned in S-III had an average LWR of 7. This analysis confirms the observation that, just before segmentation, cells in S-I become more elongated along the mediolateral axis.
Once segmentation is complete the somite undergoes a 90-degree rotation. Our analysis of GFP-labeled embryos shows that cells within the rotating somite adopt varying cell shapes (Fig. 1B,C; S0). This observation further supports the finding that individual cells, rather than the entire block of cells within a somite, undergo rotation (Youn and Malancinski,1981). Given the complexity of the cell shapes adopted during rotation, we developed an approach that allowed us to render three-dimensional images of individual cells in a rotating somite. This approach involved transplanting both rhodamine- and fluorescein-labeled mesoderm cells from the lateral marginal zone of stage 12 embryos (end of gastrulation) to the same region of an unlabeled host embryo (Fig. 2A). By allowing the grafted embryos to develop to the tail bud stage, a subset of the grafted cells could be captured in the process of rotation (Fig. 2B; S0). These grafted embryos were then fixed, cleared, and imaged with a confocal microscope. The Amira three-dimensional rendering software package was used to process the resultant z-stacks into a single image that could be analyzed along the dorsoventral, lateromedial, and anteroposterior axes. From a lateral perspective, we observe that cells positioned in the dorsal-most region of the rotating somite appear to complete rotation before cells in other regions of the somite (Fig. 2C; yellow cells in S0). However, from this perspective, it is difficult to determine whether the cells that have not completed rotation are elongated or round in shape. To accurately determine these cell shapes, the image was rotated to provide a ventral perspective, which shows that cells in the process of rotation remain elongated (Fig. 2D). In fact, cells positioned in the anterior region of the somite appear to bend in the posterior direction, whereas cells positioned in the posterior region of the same somite appear to bend in the anterior direction. The opposing direction of these bending cells suggests that somite rotation does not involve the unidirectional movement of cells. In addition, that cells bend during rotation reveals that the elongated cell shape established in the PSM is largely maintained during somite rotation.
Given our observation that cells bend in an anterior or posterior direction to reorient themselves, we hypothesized that this reorientation involved movement around the dorsoventral axis. If this hypothesis were correct, then we would expect to observe that cells stay in the same dorsoventral position during rotation. To test this hypothesis, two different approaches were used to generate a clonal, labeled population of paraxial mesoderm cells. One approach consisted of injecting RNA encoding GFP-GAP43 into the B1 blastomere of a 32-cell stage embryo, from which a subset of cells will give rise to GFP-labeled paraxial mesoderm cells (Moody,1987). The other approach involved transplanting rhodamine dextran-labeled cells from the lateral marginal zone region of a stage 12 embryo to the same region of a similar stage GFP-labeled host embryo (Dali et al.,2002). With each approach, we observed that, although labeled cells are dispersed along the anteroposterior axis, they are restricted to the same dorsoventral position. For example, Figure 3A,A′ depicts a population of GFP-GAP43–labeled cells derived from a B1 blastomere injection that is located exclusively within the central region of the paraxial mesoderm. Similarly, Figure 3B,B′ demonstrates that a transplanted population of rhodamine-labeled cells is restricted to the dorsal and ventral regions of the paraxial mesoderm. Thus, the distribution of these labeled populations of cells supports our hypothesis that somite rotation involves the movement of cells around the dorsoventral axis.
Next, we examined whether specific cell contact behavior is associated with somite rotation. We used the approaches described above to label a subset of paraxial mesoderm cells. This strategy was advantageous because we found that labeling a subset of cells rather than the entire embryo provided more contrast between cells, thus resulting in clearer images of filopodial protrusions. In general, we found that cells positioned within the PSM (S-I through S-III) appear to be in close juxtaposition with the notochord (Fig. 4A). The medial aspect of these PSM cells form flattened edges in contact with the notochord (Fig. 4B, red star), whereas the lateral aspect of these same cells appear less flattened and more dynamic (Fig. 4A, white star). Once rotation begins, the lateral ends of S0 cells form filopodial extensions that appear to be posteriorly directed toward the most recently formed segmental boundary (Fig. 4B, arrow), whereas the medial ends appear to maintain broad contacts as they move more anteriorly. Once reorientation is complete, cells replace their filopodial extensions with broad membrane contacts directed toward the intersomitic boundaries (Fig. 4A; SI). These observations were quantified by counting the number of cells with filopodial extensions in three regions: myotome (SI), rotating somite (S0), and the PSM (S-I). We found that labeled cells in the process of rotation were 4 times more likely to exhibit filopodial protrusions than cells located in the PSM or mature somites (Fig. 4C). These results support our observations that cell reorientation involves an increase in the number of filopodial protrusions.
Relative Timing of Cell Behaviors During Rotation and Segmentation Vary With the Axial Position of the Somite
Our initial observations of tail bud stage embryos revealed a specific sequence of cell shape changes that occur during the segmentation and rotation of mid-level trunk somites. However, it is known that anterior somites, also known as cranial somites, are somewhat smaller than later stage somites and that they degenerate by the onset of metamorphosis. In addition, it has recently been found in zebrafish that different molecular signaling processes govern the formation of anterior and posterior somites (Julich et al.,2005). Thus it is possible that different sets or subsets of cell behaviors are responsible for forming somites at the anterior and posterior ends of the X. laevis embryo. We, therefore, asked whether the behaviors that we observed are consistent along the entire anteroposterior axis. To address this possibility, we compared the cell shape changes associated with the formation of both anterior and posterior somites by performing both dorsal-to-ventral and lateral-to-medial confocal z-scans of over 200 GAP43-GFP–labeled, fixed, and cleared embryos between stages 18 (3 somites) and 28 (22 somites). We found that the overall sequence of cell shape changes underlying somite formation and rotation were similar between anterior and posterior somites (Fig. 5). These cell shape changes include the mediolateral elongation of cells in the PSM, the bending of individual cells, and the eventual alignment of cells parallel to the notochord. Interestingly, the presence of multiple somites in the process of rotation led us to a more careful examination of the timing of rotation. For example, in a 5-somite embryo, S0 cells have just initiated rotation (Fig. 5A), whereas S0 cells in a 22-somite embryo have almost completed the rotation process (Fig. 5C; S0, pink cells). A lateral view of the same 5-somite embryo clearly shows the presence of a boundary between S0 and S-I (Fig. 5B). In addition, the lateral views confirm the mediolateral alignment of the S0 cells in the 5-somite embryo (Fig. 5B) and the anteroposterior alignment of S0 cells in the 22-somite embryo (Fig. 5D). Another notable observation is that, during the formation of posterior somites, cells in S-I begin to bend, despite that they have not formally separated from the PSM. This finding suggests that, in comparison to anterior somites, posterior somites may expedite rotation by beginning the process at an earlier stage of somitogenesis (Fig. 5A,C; S-I, blue cells). Taken together, these results show that, although the overall sequence of cell shape changes is the same, temporal differences do exist between anterior and posterior somites.
There are several possible explanations that could account for the differences in the relative timing of somite segmentation and rotation between anterior and posterior somites. One possibility is that segmentation among anterior somites occurs at a faster pace then it does among posterior somites. If this were the case, we would expect to observe an apparent build up of anterior somites undergoing rotation in contrast to the posterior end where segmentation would be slower and more coincident with the rate of somite rotation. Support for this scenario is found in both zebrafish (Kimmel et al.,1995) and chick (Sara Venters, personal communication) in which anterior somites have been shown to form at a faster pace than posterior somites. Another possibility is that segmentation occurs at a constant pace, but somite rotation is slower at the anterior end. If this were the case, we would expect the same outcome described above.
To distinguish between these two possibilities we examined the rate of segmentation and somite rotation. Because X. laevis embryos are opaque, it was not possible to observe the formation of somites in live embryos. Instead, we performed a time-course analysis using synchronized group of GFP-labeled embryos that were fixed every 50 min starting at stage 18, at which time approximately three somites have formed. A 50-min interval was chosen because this is the rate of somite formation determined by Neuwkoop and Faber (1967). Fixed embryos were imaged with a confocal microscope, and both the total number of somites and the number of rotating somites was documented for each time interval. We reasoned that, if the rate of somite formation increases, we would observe more than one somite forming within each time interval during the experiment. However, our data revealed a linear relationship between the number of somites and developmental time, suggesting that somite formation occurs at a similar rate along the anteroposterior axis (Fig. 6A). With respect to somite rotation, we found that, at the earliest time point, embryos contained three somites, which were all engaged in rotation (Fig. 6B). As time progressed, this number increased slightly, with a peak during the formation of the seventh somite (t = 200). After this time point, the number of rotating somites gradually declines from four to two (Fig. 6B). These results indicate that somite rotation is a slower process at the anterior end of the axis in comparison to the posterior end. Thus the temporal differences observed between anterior and posterior somites appears to be due to a slower rotation process among anterior-forming somites and not a difference in the timing of segmentation.
Somite Segmentation and Rotation Proceeds in a Dorsal-to-Ventral Progression
Initial analysis of mid-trunk level somites revealed that cells positioned in the dorsal region of a rotating somite are the first to become aligned parallel with the notochord (Fig. 1C; S0, pink cells). This observation was confirmed by three-dimensional imaging of rotating cells in similar staged embryos (Fig. 2C, yellow cells). Given the differences observed between anterior and posterior somites, we next examined whether rotation occurs in a dorsal-to-ventral direction for both anterior and posterior somites. This was done by performing dorsal-to-ventral confocal z-scans of the paraxial mesoderm in over 200 GAP43-GFP–labeled, fixed, and cleared embryos between stages 18 (3 somites) and 28 (22 somites). We found that regardless of the axial position of the somite, cells in the dorsal region of the paraxial mesoderm (at the level of the neural tube) were consistently more advanced in somite morphogenesis than cells located in the ventral region of the paraxial mesoderm (at the level of the notochord; Fig. 7). Because somite rotation is a relatively slower process at the anterior end of the PSM, the dorsal-to-ventral progression in somite rotation is less obvious than at the posterior end. For example, in a five-somite embryo, cells in the dorsal region of SI have begun to rotate (Fig. 7A; SI, green cells), whereas cells positioned in the ventral region of the same somite, remain aligned in the mediolateral direction (Fig. 7B; SI, green cells). Moreover, cells positioned in the dorsal region of a forming somite (S0) appear to have an indentation suggestive of somitic furrow formation (Fig. 7A; S0, pink cells), while cells in the ventral region of the same somite remain associated with the PSM (Fig. 7B; S0, pink cells). During the formation of more posterior somites, the dorsal-to-ventral progression in somite rotation becomes more pronounced. For example, in a 19-somite embryo, cells in the dorsal region of somite S0 have almost completed rotation (Fig. 7C; S0), while ventral cells in the same somite remain oriented in the mediolateral direction (Fig. 7D, S0). To further quantify this dorsal-to-ventral trend in somite rotation, we measured the relative angle of rotation of S0 cells positioned in both the dorsal and ventral regions of the somite (Fig. 7E). We found that, within a given newly forming somite, regardless of its axial position, cells in the dorsal region are more advanced in rotation than cells positioned ventrally. In addition, the temporal difference between dorsally and ventrally located cells appears to grow with the formation of more posterior somites (Fig. 7F).
By using several imaging approaches, we have investigated the cell shape changes involved in somite morphogensis along the anteroposterior, dorsoventral, and lateromedial axes in X. laevis. These observations lead us to propose a four-step process for somite rotation (Fig. 8A) that extends previous models proposed by Youn and Malancinski (1981) and Keller (2000). We show that this process begins at the anterior end of the PSM (S-I) where cells become more elongated along the mediolateral axis (Step 1). Next, cells increase the number of filopodial projections directed toward the posterior end, where the most recent boundary has formed (Step 2). As cells begin to rotate, they maintain their elongated cell shape as they bend around the dorsoventral axis (Step 3). Finally, cells become aligned parallel to the notochord (Step 4). This sequence of cell behaviors is similar for somites forming along the entire anteroposterior axis, with the exception that the rate of somite rotation is slower during the formation of anterior somites (Fig. 8B). We also show that somite morphogenesis begins dorsally among cells at the level of the neural tube, and then progresses ventrally to cells located at the level of the notochord (Fig. 8B,C).
Somite Rotation in X. laevis
Somitogenesis in X. laevis provides an excellent model for studying tissue morphogenesis. It consists of the partitioning of the PSM into blocks of cells that then undergo a 90-degree rotation. Although this process was first described by Hamilton in 1969, the precise cell behaviors remained elusive. The first analysis of the cell shape changes associated with somitogenesis in X. laevis was done by Youn and Malancinski (1981), who used SEM to show that individual cells within the somite, rather than the entire somite in unison, undergo rotation. Our data and those of other researchers (Wilson et al.,1989), support this conclusion. In addition, Youn and Malancinski (1981) found that somite rotation begins with cells positioned in the central region of the somite, at the level of the notochord, followed by cells positioned in the dorsal and ventral regions of the somite. By observing the ventral aspect of the rotating somite, Youn and Malancinski, also observed that somite rotation occurs in a dorsal-to-ventral gradient. Our data supports Youn and Malancinski's observations, however, we find that the process initiates in the dorsal-most region of the somite, at the level of the neural tube, rather than at the level of the notochord (Fig. 7). One possible reason for the differences in our observation is that Youn and Malancinski examined mid-trunk level somites (9–15 somites) where differences between dorsal and ventral intrasomitic cell shapes are less obvious than in more posterior somites (18–20 somites; Fig. 7). In addition, unlike the tissue fractures that were needed to expose specific regions of the paraxial mesoderm for SEM analysis, our confocal z-scans provide a less disruptive technique for visualizing tissues in a three-dimensional context, thereby facilitating the detection of relative differences in cell shapes between neighboring cells.
Somitogenesis in X. laevis has previously been studied by means of time-lapse imaging of dorsal explants. Wilson et al. (1989) observed that cells parallel to the forming somitic boundary move in opposite directions, suggesting that this movement could be part of the somite rotation process. In fact, this observation led Keller (2000) to propose a model for somite rotation in which cells in the anterior region of a forming somite migrate laterally, while cells in the posterior region of the same somite migrate medially toward the notochord. In this manner, a rotating somite to the right of the notochord would appear to rotate in a clockwise direction while the parallel somite to the left of the notochord would appear to rotate in a counterclockwise direction. Our confocal images of posterior somites suggest a similar clockwise/counterclockwise rotation of the somite (Figs. 4A, 5C). However, our Amira three-dimensional image analysis suggests that this movement is achieved by cells bending in either the anterior or posterior direction rather than migrating (Fig. 2D). In this manner, it seems likely that cells positioned anteriorly in a recently segmented somite (S0) form the lateral edge of a rotated somite, whereas cells positioned posteriorly form the medial edge of the same somite. However, live time-lapse imaging is necessary to conclusively determine the precise movement of cells during rotation.
A recent study examining the expression pattern of slow myosin heavy chain (MyHC) during the formation of myotome fibers in X. laevis, led to another model for somite rotation (Grimaldi et al.,2004). This model proposes that PSM cells fated to form slow muscle fibers are initially positioned adjacent to the notochord. At the time of segmentation, these prospective slow muscle fiber cells migrate to the lateral edge of the somite and become the first cells to complete rotation. However, our observations do not support this model. First, we do not observe a subset of cells positioned adjacent to the notochord in the PSM. Instead, we find that virtually all the cells positioned in S-I of the PSM are in contact with the notochord (Fig. 1B). Second, we do not observe cells in the lateral region of the somite completing rotation first. Instead, we find that cells in the dorsal-most region of the developing somite completed rotation first (Figs. 1C, 2C, 7). Moreover, MyHC is not detected until stage 27/28 with the formation of the 20th somite. Given the late timing of this expression pattern, it seems unlikely that the presence of prospective slow muscle cells would play a role in somite rotation.
Dynamic Cell Contact Behaviors Associated With Somite Rotation
Somite rotation involves distinct changes in cell contact behavior. This process begins in the PSM with mediolaterally aligned cells that form broad contacts with the notochord–somite boundary and more dynamic contacts directed laterally (Fig. 4A). As cells begin to rotate, they increase the number of fine filopodial protrusions in both the lateral and posterior directions (Fig. 4C). Once the posterior segmental boundary has formed and rotation is complete, posteriorly directed protrusive behavior is replaced by broad cell contacts (Fig. 4A; see SI). Of interest, three-dimensional imaging of rotating cells reveals that cells bend in both the anterior and posterior directions to become aligned parallel to the notochord. Thus it seems likely that rotating cells would form filopodial protrusions in both the anterior and posterior directions during this process. However, confocal images of rotating cells show that the majority of the filopodia are posteriorly directed toward the newly forming somite boundary (Fig. 4B). One possibility is that anteriorly directed protrusions are quickly captured and transformed into broad cell contacts at the anterior somite boundary. Thus it becomes difficult to detect protrusive activity in this direction. Alternatively, it may be the case that the broad cell contact established by the medial edge of an unrotated cell becomes the anterior end of a rotated cell. In this manner, the broad cell contact is made with the anterior somite boundary in the absence of protrusive activity. Regardless of the precise mechanism, it is likely that the matrix filled somite boundary provides cues that polarize the protrusive activity and help orient cells along the anteroposterior axis.
Differences in Somite Morphogenesis Along the Anteroposterior Axis
We find that, although the cell behaviors underlying somite rotation are largely the same throughout the anteroposterior axis, somite rotation is slower among anterior somites. The cellular and molecular processes underlying the differences observed between anterior and posterior somites are not well understood. In X. laevis, it is known that the mesodermal tissue positioned next to the prospective notochord at the onset of gastrulation will form the first six somites, which includes four cranial and two anterior trunk somites (Shih and Keller,1992). However, these six somites will degenerate by the onset of metamorphosis (Neuwkoop and Faber,1967). Thus it may be that anterior somites are molecularly and structurally distinct from posterior somites.
Support for molecular differences between anterior and posterior somites comes from genetic analysis in zebrafish. Oates and Ha (2002) showed that, although the Delta/Notch signaling pathway is required for the formation of somites along the entire axis, the formation of anterior somites also requires the activity of her 7, a downstream target of the Notch signaling pathway. In addition, it has been shown that integrin α5 is required for the formation of anterior, but not posterior somites (Julich et al.,2005). How these molecular differences translate into the patterning and functional differences among somites remains unclear. While these sharp differences in the formation of anterior and posterior somites have been described in zebrafish, they have yet to be demonstrated in X. laevis. The fact that the changes in the rate of somite rotation seem to occur gradually along the X. laevis axis suggests to us that the formation of anterior and posterior somites is not differentially regulated by multiple signaling pathways.
A recent study provides evidence that there may be structural differences between anterior and posterior somites. Shook et al. (2004) showed that ingressing cells from the gastrocoele roof are found in somites from the mid-trunk region (approximately the 10th somite) to the end of the tail. They hypothesize that these ingressing cells may be involved in instructing the behaviors and fates of the somitic cells. Of interest, these ingressing cells are absent from the first nine somites. This finding suggests that the presence of these ingressing cells may influence the rate of somite cell rotation. If this were the case, then one would expect to see a change in the rate of somite rotation shortly after the formation of the 10th somite. In fact, we did observe that the rate of somite rotation gradually increased after the formation of the 10th somite (Fig. 6B; t = 350 min). Work by Minsuk and Keller (1997) showed that the presence of these ingressing cells is highly variable between embryos and along the anteroposterior axis, with a greater proportion of them present toward the posterior end of the axis. Given this variable population of ingressing cells, it may be the case that they are involved in the gradual increase in the rate of somite rotation observed along the anteroposterior axis. However, because all cells within a somite eventually complete rotation independent of the presence of these ingressing cells, it seems unlikely that they are the sole regulators of this process. Additional contributing factors, such as changes to the fibrillar matrix surrounding the paraxial mesoderm (Davidson et al.,2004), may play a role. We have shown that, during rotation, cells increase their numbers of protrusions (Fig. 4C). Perhaps changes in the fibrillar matrix influence the effectiveness of the protrusive cell behaviors involved in reorienting cells along the anteroposterior axis.
The molecular mechanisms that control the cell behaviors underlying somite segmentation and rotation remain largely unknown. Several cell adhesion molecules have been found to play a role in somite formation. For example, two protocadherin molecules, paraxial PAPC and protocadherin, in neural crest and somites (PCNS) have been shown to be involved in the maintenance of segmental boundaries. By using a dominant-negative form of PAPC (Kim et al.,2000) or a knockdown of PCNS (Rangarajan et al.,2006), investigators found that they could disrupt the maintenance of segmental boundaries, which in turn led to the disorganization of myotome fibers. A similar phenotype was also obtained by disrupting type 1 cadherin (Giacomello et al.,2002). It remains unclear whether this finding is due to a direct or indirect effect on somite rotation. One possibility is that cells can rotate normally, but by disrupting the segmental boundaries, they cannot organize themselves into parallel arrays. Alternatively, it could be that cadherin/protocadherin function is required for the directional movement of filopodia during rotation. Another possible mediator of this process is Xena, an actin regulatory molecule (Kragtorp and Miller,2006). Inhibition of Xena function led to enlarged and disorganized somites. However, closer inspection revealed that most cells within the enlarged somites were able to complete rotation. Thus it is likely that Xena works with additional gene products to instruct somite cell rotation.
Given that the first step in somite formation involves the mediolateral elongation of PSM cells, it seems likely that the planar cell polarity (PCP) pathway is involved in somitogenesis. Work by Wallingford et al. (2000) showed that Dishevelled (dsh), a downstream target of the PCP pathway, is localized to the mediolateral edges of mesodermal cells and is required for convergent extension cell movement during gastrulation in X. laevis. Thus dsh may be involved in the mediolateral elongation of cells as they mature in the PSM. Because somite rotation appears to consist of the bending of elongated cells, it may be that dsh maintains the cell's polarity during rotation. Alternatively, cells may limit or stop their expression of dsh during rotation and then later re-establish dsh expression once they become polarized along the anteroposterior axis. The role of dsh in somite formation and rotation remains unknown. However, disheveled-2–deficient mice show defects in somite segmentation (Hamblet et al.,2002), indicating that dsh is likely to play a role in X. laevis somitogenesis.
In this study, we have undertaken a more detailed analysis of the cell shape changes associated with somite segmentation and rotation. We propose a four-step process for somite morphogenesis that involves mediolateral cell elongation, increased filopodia activity, and bending of elongated cells to form an array of cells that are aligned parallel to the notochord. This coordinated set of cell behaviors appears to proceed more quickly during the formation of posterior somites, as evidenced by the fact that their segmentation and rotation appear to be more coincident than it does among anterior somites. The overall pattern of somite formation and rotation occurs in a dorsal-to-ventral progression. In the future, it will be interesting to determine whether the cell behaviors driving somite rotation are autonomous or nonautonomous. Moreover, it will be important to identify the genes that drive this morphological process and determine how they are differentially regulated along the anteroposterior axis.
Embryo Culture and Transplantations
Eggs were obtained by standard methods (Kay and Peng,1991) from adult X. laevis ovi-positive females. After in vitro fertilization, the eggs were dejellied in 2% cysteine and then cultured in 33% Modified Barth's Solution (MBS) at 15–23°C until they reached the desired stage.
To image a subpopulation of cells undergoing somitogenesis, we first injected donor embryos at the one-cell stage with a 70-kDa dextran conjugated to either rhodamine or fluorescein (Molecular Probes), and then cultured these embryos to stage 12. Approximately 10–40 labeled cells from the PSM were grafted to the corresponding region of unlabeled stage 12 host embryos. Grafted embryos were cultured to the early tail bud stage (stage 24), fixed in MEMFA, washed in 100% methanol for an hour and then transferred to benzyl benzoate/benzyl alcohol (2:1) solution for viewing on a confocal microscope.
GFP-Labeling of Embryos
To describe the changes in cell shapes associated with somite formation and rotation, we injected embryos at the one or two-cell stage with ∼600 pg of an in vitro transcribed RNA (from a NotI-linearized plasmid template pGAP43-GFP) encoding a membrane-tagged GFP protein (Moriyoshi et al.,1996; Kim et al.,1998). To examine a subpopulation of labeled embryos, we injected RNA encoding this GFP into the B1 blastomere at the 32-cell stage. Injected embryos were cultured to various stages, fixed, and stained with an anti-GFP antibody conjugated to AlexaFluor 647 (Molecular Probes).
Quantification of Cell Behaviors
To quantify the extent to which cells in the PSM are mediolaterally elongated, we used the SPOT software package to measure the length (mediolateral axis) and width (anteroposterior axis) of five cells in the S-I and S-III regions. Dorsoventral scans of the PSM region for eight different embryos were used for this analysis. The average LWR and standard error was then graphed using Microsoft Excel.
To determine the number of cells containing filopodia during somite formation, we used 18 embryos that were cultured to stage 28 and contained a subset of labeled cells in all three regions: SI, S0, and S-I. For each region, the average number of cells containing filopodia and the standard error were determined.
To measure the timing of segmentation and somite rotation, we first injected a large clutch of fertilized eggs with mRNA encoding GAP43-GFP. At the onset of gastrulation, we carefully examined the injected embryos for synchronous development. Any embryos that were either delayed in the formation of the dorsal blastopore lip or were more advanced in gastrulation were removed. In this manner, we maintained a highly synchronized clutch of developing embryos. Beginning at stage 18, 10 embryos were fixed every 50 min over the course of 17 hr. Fixed embryos were then processed with an anti-GFP antibody conjugated to AlexaFluor 647. Embryos were mounted as mentioned above, and dorsal confocal scans were performed to determine the number of formed and rotating somites in each embryo. The averages and standard error were determined for each time point and plotted using Excel.
The angle of rotation for somites was determined by comparing the angular movement of S0 cells to those in the PSM for three different axial regions: (1) anterior, consisting of embryos with 4 to 6 somites; (2) middle, consisting of embryos with 9 to 12 somites; and (3) posterior, consisting of embryos with 18 to 20 somites. All embryos were stained for GFP and imaged with a confocal microscope. The SPOT software (Diagnostic Instruments) was then used to measure the angle of movement of S0 cells relative to the starting position of cells in the PSM. For each axial region, measurements were made of five cells positioned in the dorsal region and five cells positioned in the ventral region of the same somite (S0) in five different embryos. Average angle of rotation and the standard error were determined for each axial region using Excel.
Confocal Imaging and Three-Dimensional Rendering of Embryos
Stained embryos were cleared in a benzyl benzoate/benzyl alcohol (2:1) solution and mounted on slides with handmade wells that were built up from layers of electrical tape. In this manner, we were able to image embryos from both the dorsoventral and mediolateral perspectives. Confocal laser scanning microscopy was performed using an Eclipse PhysioStation fluorescence microscope (E600FN, Nikon) with a Nikon PCM2000 confocal scan head. The system was controlled by means of the Nikon software package. Images were captured at 1-μm intervals through both ×20 and ×40 Nikon objectives. To facilitate the visualization of cell morphologies, the original confocal images were inverted in Adobe Photoshop such that cells were outlined in black. Individual cells were then pseudocolored in Adobe Illustrator.
Grafted embryos with labeled cells in the rotating somite were imaged on a Nikon confocal microscope using Simple PCI software version 3.5. Z-stacks of images were then rendered using the Amira software package (TGS Template Graphics). After three-dimensional rendering of these images, a subset of cells was pseudocolored. This subset allowed us to view the morphology and relative position of individual somite cells as they exist within a somite undergoing rotation.
We thank Ray Keller, Dave Daggett, and Sharon Minsuk for helpful discussions; Clarissa Henry for assistance with confocal imaging; TaiJuana Sylvester for cell transplantations; Dov Sinkoff-Lerman with Amira tracings; Ed Connor with statistical analysis; and Eddy DeRobertis for the GAP43-GFP plasmid.