Differential contributions of Mesp1 and Mesp2 to the epithelialization and rostro-caudal patterning of somites

Somitogenesis is not only an attractive example of metameric pattern formation but is also a good model system for the study of morphogenesis,particularly epithelial-mesenchymal interconversion in vertebrate embryos(Gossler and Hrabe de Angelis,1997; Pourquié,2001). The primitive streak, or tailbud mesenchyme, supplies the unsegmented paraxial mesoderm, known as presomitic mesoderm (PSM). Mesenchymal cells in the PSM undergo mesenchymal-epithelial conversion to form epithelial somites in a spatially and temporally coordinated manner. Somites then differentiate, in accordance with environmental cues from the surrounding tissues, into dorsal epithelial dermomyotome and ventral mesenchymal sclerotome (Borycki and Emerson,2000; Fan and Tessier Lavigne,1994). Hence, the series of events that occur during somitogenesis provide a valuable example of epithelial-mesenchymal conversion. The dermomyotome gives rise to both dermis and skeletal muscle, whereas the sclerotome forms cartilage and bone in both the vertebrae and the ribs. Each somite is subdivided into two compartments, the rostral (anterior) and caudal(posterior) halves. This rostro-caudal polarity appears to be established just prior to somite formation (Saga and Takeda, 2001).

Mesp1 and Mesp2 are closely related members of the basic helix-loop-helix(bHLH) family of transcription factors but share significant sequence homology only in their bHLH regions (Saga et al.,1996; Saga et al.,1997). During development of the mouse embryo, both Mesp1and Mesp2 are specifically expressed in the early mesoderm just after gastrulation and in the paraxial mesoderm during somitogenesis. Mesp1/Mesp2 double-null embryos show defects in early mesodermal migration and thus fail to form most of the embryonic mesoderm, leading to developmental arrest(Kitajima et al., 2000). Mesp1-null embryos exhibit defects in single heart tube formation, due to a delay in mesodermal migration, but survive to the somitogenesis stage(Saga et al., 1999),suggesting that there is some functional redundancy, i.e. compensatory functions of Mesp2 in early mesoderm. During somitogenesis, both Mesp1 and Mesp2 are expressed in the anterior PSM just prior to somite formation. Although we have shown that Mesp2, but not Mesp1, is essential for somite formation and the rostro-caudal patterning of somites(Saga et al., 1997), a possible functional redundancy between Mesp1 and Mesp2 has not yet been clearly established.

To further clarify the contributions of Mesp1 and Mesp2 to somitogenesis,analysis of Mesp1/Mesp2 double-null embryos is necessary, but because of the early mesodermal defects already described, these knockout embryos lack a paraxial mesoderm, which prevents any analysis of somitogenesis. We therefore adopted a strategy that utilized chimera analysis. As we have reported previously, the early embryonic lethality of a Mesp1/Mesp2 double knockout is rescued by the presence of wild-type cells in a chimeric embryo, but the double-null cells cannot contribute to the cardiac mesoderm(Kitajima et al., 2000). This analysis, however, focused only on early heart morphogenesis and did not investigate the behavior of Mesp1/Mesp2 double-null cells in somitogenesis. In this report, we focus upon somitogenesis and compare two types of chimeras using either Mesp1/Mesp2 double-null cells or Mesp2-null cells to investigate Mesp1 function during somitogenesis.

Another purpose of our chimera experiments was to elucidate the cell autonomy of Mesp functions. In the process of somite formation, mesenchymal cells in the PSM initially undergo epithelialization at the future segment boundary, independently of the already epithelialized dorsal or ventral margin of the PSM (Sato et al.,2002). Epithelial somite formation is disrupted in the Mesp2-null embryo, indicating that Mesp2 is required for epithelialization at the segment boundary. Although Mesp products are nuclear transcription factors and their primary functions must therefore be cell autonomous (transcriptional control of target genes), it is possible that the roles of Mesp2 in epithelialization are mediated by the non-cell autonomous effects of target genes. We therefore asked whether the defects in Mesp2-null cells during epithelialization could be rescued by the presence of surrounding wild-type cells. Additionally, we would expect to find that the role of Mesp2 in establishing rostro-caudal polarity is rescued in a similar way.

Our analysis suggests that Mesp1 and Mesp2 have redundant functions and are both cell-autonomously involved in the epithelialization of somitic mesoderm. In addition, our results highlight some non-cell autonomous effect of Mesp2-null and Mesp1/Mesp2-null cells.

As described previously (Kitajima et al., 2000), chimeric embryos were generated by aggregating 8-cell embryos of wild-type mice (ICR) with those of mutant mice that were genetically marked with the ROSA26 transgene(Zambrowicz et al., 1997). Mesp1/Mesp2 double-null embryos were generated by crossing wko-del(+/–) and Mesp1(+/–)/Mesp2(+/cre) mice as described previously (Kitajima et al.,2000). This strategy enables us to distinguish chimeric embryos derived from homozygous embryos, which have two different mutant alleles, from those derived from heterozygous embryos. Likewise, Mesp2-null embryos were generated by crossing P2v1(+/–) mice(Saga et al., 1997) and P2GFP (+/gfp) mice (Y.S. and S.K., unpublished) that were also labeled with the ROSA26 locus. The genotype of the chimeric embryos was determined by PCR using yolk sac DNA.

The chimeric embryos were fixed at 11 days postcoitum (dpc) and stained in X-gal solution for the detection of β-galactosidase activity, as described previously (Saga et al.,1999). For histology, samples stained by X-gal were postfixed with 4% paraformaldehyde, dehydrated in an ethanol series, embedded in plastic resin (Technovit 8100, Heraeus Kulzer) and sectioned at 3 μm. The methods used for gene expression analysis by in-situ hybridization of whole-mount samples and frozen sections and skeletal preparation by Alcian Blue/Alizarin Red staining were described previously(Saga et al., 1997; Takahashi et al., 2000). Probes for in-situ hybridization for Uncx4.1(Mansouri et al., 1997; Neidhardt et al., 1997), Delta-like 1 (Dll1)(Bettenhausen et al., 1995) and Paraxis (Burgess et al.,1995) were kindly provided by Drs Peter Gruss, Achim Gossler and Alan Rawls, respectively. A probe for EphA4(Nieto et al., 1992) was cloned by PCR. For detection of actin filaments, frozen sections were stained with AlexaFluor 488-conjugated phalloidin (Molecular Probes) according to the manufacturer's protocol.

During somitogenesis, both Mesp1 and Mesp2 are expressed in the anterior PSM just prior to somite formation and their expression domains overlap (Fig. 1A). Mesp1-null embryos form morphologically normal somites and show normal rostrocaudal patterning within each somite(Fig. 1B,E-H), indicating that Mesp1 is not essential for somitogenesis. By contrast, Mesp2 is essential for both the formation and rostro-caudal patterning of somites, as Mesp2-null embryos have no epithelial somites and lose rostral half properties, resulting in caudalization of the entire somitic mesoderm(Saga et al., 1997)(Fig. 1C,D).

Although somite formation and rostro-caudal patterning is disrupted in the Mesp2-null embryo, histological differentiation into dermomyotome and sclerotome is not affected. It is noteworthy that the Mesp2-null embryo still forms disorganized dermomyotomes without forming epithelial somites(Saga et al., 1997). As Mesp1 is expressed at normal levels in the PSM of Mesp2-null embryos(Fig. 1C,D), it is possible that Mesp1 functions to rescue some aspects of somitogenesis in the Mesp2-null embryo. In order to further clarify the contributions of both Mesp1 and Mesp2 during somitogenesis, we therefore generated chimeric embryos with either Mesp2-null cells or Mesp1/Mesp2 double-null cells and compared the behavior of mutant cells during somitogenesis (Fig. 2).

We first generated Mesp2-null chimeric embryos(Mesp2^–/– with Rosa26: wild) to analyze cell autonomy of Mesp2 function during somitogenesis. The control chimeric embryo (Mesp2^+/– with Rosa26:wild) showed normal somitogenesis and a random distribution of X-gal stained cells (Fig. 3A). The Mesp2-null chimeric embryos formed abnormal somites that exhibited incomplete segmentation (Fig. 3B), but histological differentiation of dermomyotome and sclerotome was observed. Within the incomplete somite, X-gal-stained Mesp2-null cells were mainly localized in the rostral and central regions, surrounded by wild-type cells at the dorsal, ventral and caudal sides (Fig. 3B). The surrounding wild-type cells, however, did not form an integrated epithelial sheet, but consisted of several epithelial cell clusters. Such trends were more obviously observed in other sections, where wild-type cells were found to form multiple small epithelial clusters(Fig.3C,D). Mesp2-null cells tended to be eliminated from the epithelial clusters, although they were partially integrated into these structures (blue arrows in Fig. 3C,D). Likewise, small numbers of Mesp2-null cells were found to contribute to the dermomyotome(Fig. 3E,F). Mesp2-null cells also appeared to form the major part of the sclerotome.

We have previously demonstrated that suppression by Mesp2 of the caudal genes Dll1 and Uncx4.1 in presumptive rostral half somites is a crucial event in the establishment of the rostrocaudal pattern of somites(Saga et al., 1997; Takahashi et al., 2000). As Mesp2-null embryos exhibit caudalization of somites, Mesp2-null cells are predicted to be unable to express rostral properties. Hence, Mesp2-null cells are expected to distribute to the caudal region of each somite where the rostrocaudal patterns are rescued by wild-type cells in a chimeric embryo. In this context, the localization of Mesp2-null cells at the rostral side was an unexpected finding. We interpret this to mean that the rostral location of Mesp2-null cells is due to a lack of epithelialization functions (see Discussion).

To examine rostro-caudal properties in Mesp2-null cells, located in the rostral side, we analyzed the expression of a caudal half marker gene, Uncx4.1 (Mansouri et al.,1997; Neidhardt et al.,1997). Analysis of adjacent sections revealed that lacZ-expressing Mesp2-null cells, localized at the rostral and central portion, ectopically expressed Uncx4.1(Fig. 4A-D). This strongly suggests that Mesp2-null cells cannot acquire rostral properties even if surrounded by wild-type cells, and that Mesp2 function is cell-autonomously required for the acquisition of rostral properties. We also observed that the small number of Mesp2-null cells distributed mostly to the caudal end of the dermomyotome (Fig. 3E,F) and that the expression pattern of Uncx4.1 was normal in the dermomyotome(Fig. 4E,F). In the sclerotome, lacZ-expressing Mesp2-null cells often distributed to the rostral side, where expression of Uncx4.1 was abnormally elevated(Fig. 4G,H). The vertebrae of the Mesp2-null chimeric fetus showed a partial fusion of the neural arches,which was reminiscent of Mesp2-hypomorphic fetuses(Fig. 4I,J)(Nomura-Kitabayashi et al.,2002). Fusion of proximal rib elements was also observed(Fig. 4K,L).

To address the question of whether Mesp1, in addition to Mesp2, exhibits any function during somitogenesis, we next generated Mesp1/Mesp2 double-null chimeric embryos and compared them with the Mesp2-null chimeric embryos described in the previous sections. We first performed whole-mount X-gal staining of embryos at 11 dpc. In the control chimeric embryo, the X-gal-stained Mesp1/Mesp2 double-heterozygous cells distributed randomly throughout the embryonic body, including the somite region(Fig. 5A,C). By contrast, the Mesp1/Mesp2 double-null chimeric embryo displayed a strikingly uneven pattern of cellular distribution in the somite region. The X-gal stained Mesp1/Mesp2 double-null cells were localized at the medial part of embryonic tail and were not observed in the lateral part of the somite region(Fig. 5B,D). Histological examination of parasagittal sections further revealed obvious differences in the cellular contribution to somite formation(Fig. 5E,F). In the control chimeric embryo, Mesp1/Mesp2 double-heterozygous cells distributed randomly throughout the different stages of somitogenesis (PSM, somite, dermomyotome and sclerotome: Fig. 5E). In the Mesp1/Mesp2 double-null chimeric embryo, neither the initial segment border nor epithelial somites were formed, but histologically distinguishable dermomyotome-like and sclerotome-like compartments were generated(Fig. 5F). In addition,Mesp1/Mesp2 double-null cells and wild-type cells were randomly mixed in the PSM, whereas the dermomyotome-like epithelium consisted exclusively of wild-type cells and the sclerotome-like compartment consisted mostly of Mesp1/Mesp2 double-null cells. This suggests that either Mesp1 or Mesp2 is cell-autonomously required for the formation of epithelial somite and dermomyotome. These results also indicate that PSM cells with different characteristics are rapidly sorted during somite formation.

Subsequent examination of transverse sections confirmed the elimination of Mesp1/Mesp2 double-null cells from dermomyotome(Fig. 5G,H). In the mature somite region, the wild-type dermomyotome-like epithelium was found to form the myotome (my) (Fig. 5I,J). Furthermore, the ventral part of this dermomyotome-like epithelium became mesenchymal and appeared to contribute to the dorsal sclerotome (dsc),implying that this initial dermomyotome-like epithelium actually corresponds to the epithelial somite exclusively composed of wild-type cells(Fig. 5I,J). Fluorescent phalloidin staining revealed that the apical localization of actin filaments is limited to the dorsal compartments, which are occupied by wild-type cells in the Mesp1/Mesp2 double-null chimeric embryo(Fig. 5K,L), indicating the Mesp1/Mesp2 double-null cells cannot undergo epithelialization.

It is known that the bHLH transcription factor paraxis (Tcf15 – Mouse Genome Informatics), is required for the epithelialization of somite and dermomyotome (Burgess et al.,1995; Burgess et al.,1996). Although Paraxis expression is not affected in Mesp2-null embryos (data not shown), it is possible that it is influenced by the loss of both Mesp1 and Mesp2. We therefore examined the expression patterns of Paraxis in our Mesp1/Mesp2 double-null chimeras. In wild-type embryos Paraxis is initially expressed throughout the entire somite region (in both the prospective dermomyotomal and sclerotomal regions) in the anteriormost PSM and newly forming somites, and then localizes in the dermomyotomes (Burgess et al.,1995). The dorsal dermomyotomal epithelium, composed of wild–type cells, strongly expressed Paraxis in the chimeric embryo (Fig. 6A,B). In addition, adjacent sections revealed that lacZ-expressing Mesp1/Mesp2 double-null cells expressed Paraxis in the medial sclerotomal compartment (Fig. 6A,B,brackets). This suggests that Paraxis expression in the future sclerotomal region is independent of Mesp factors. However, at present we cannot exclude the possibility that the maintenance of Paraxisexpression in the dermomyotome requires the functions of either Mesp1 or Mesp2.

To clarify the rostro-caudal properties of somites in our chimeric embryos,we examined the expression pattern of Uncx4.1. Control chimeric embryos exhibited a normal stripe pattern of Uncx4.1 expression throughout the segmented somite region(Fig. 7A). By contrast,Mesp1/Mesp2 double-null chimeric embryos exhibited continuous Uncx4.1expression in the ventral sclerotomal region(Fig. 7B). This continuity was observed in the entire sclerotome-like compartment of the newly formed somite region and in the ventral sclerotome in the mature somite region. The caudal localization of Uncx4.1 expression, however, was normal in the dermomyotome and the dorsal sclerotome, which consisted of wild-type cells(Fig. 5), even in Mesp1/Mesp2 double-null chimeras. This suggests that, like Mesp2-null cells, Mesp1/Mesp2 double-null cells are incapable of acquiring rostral properties. Since the mesoderm of Mesp1/Mesp2 double-null embryos lacks the expression of the major markers of paraxial mesoderm (Kitajima et al., 2000), and Mesp1/Mesp2 double-null cells do not exhibit histological features characteristic of epithelial somites in our current study, it is possible that Mesp1/Mesp2 double-null cells may lack paraxial mesoderm properties. However, the analysis of adjacent sections suggests that lacZ-expressing Mesp1/Mesp2 double-null cells themselves express Uncx4.1, a somite-specific marker(Fig. 7C,D), and they had also been found to have normal expression of Paraxis(Fig. 6A,B).

It is believed that the rostro-caudal pattern within somites and dermomyotomes is generated in the PSM and maintained in somites and dermomyotomes. We observed a normal rostro-caudal pattern in the dermomyotome(Fig. 7), although wild-type cells and Mesp1/Mesp2 double-null cells are mixed in the PSM(Fig. 5), of Mesp1/Mesp2 double-null chimeric embryos. As Mesp products are required for suppression of Dll1 in the anterior PSM, a normal Dll1 stripe pattern cannot be formed if Mesp1/Mesp2 double-null cells are randomly distributed in the anterior PSM. This is because 50% of cells cannot undergo suppression of Dll1 even in the future rostral half region. Therefore, our finding of a normal rostro-caudal pattern in the dermomyotome of double-null chimeras is surprising and raises the question of whether wild-type cells can be normally patterned in the presence of surrounding Mesp1/Mesp2 double-null cells. To determine how the rostro-caudal pattern in the dermomyotome is formed in the PSM, we examined the expression pattern of Dll1(Bettenhausen et al., 1995),the stripe expression profile of which is established in the anteriormost PSM via the function of Mesp2 (Takahashi et al., 2000). The lacZ-expressing Mesp1/Mesp2 double-null cells were subsequently found to be consistently localized in the sclerotome-like region, where Dll1 expression was abnormally expanded(Fig. 6C,D). In the dermomyotome-like region, however, Dll1 expression in the caudal half was normal. Intriguingly, strong Dll1 expression in the anteriormost PSM was suppressed in a rostrally adjoining cell population, which is mainly occupied by wild-type cells (Fig. 6C,D, arrows). This implies that wild-type cells and Mesp1/Mesp2 double-null cells rapidly segregate at S–1 to S0, after which the rostro-caudal pattern of Dll1 expression is formed in the partially segregated wild-type cell population but not in the randomly mixed cell population. In other words, the separation from Mesp1/Mesp2 double-null cells enabled normal rostro-caudal patterning of wild-type cells.

Mesp2-null fetuses display caudalized vertebrae with extensive fusion of the pedicles of neural arches and proximal elements of the ribs(Saga et al., 1997). The Mesp1/Mesp2 double-null chimeric fetuses also exhibited fusion of the pedicles of neural arches and the proximal ribs(Fig. 7E-H). Furthermore, the vertebrae of severe chimeric fetuses were indistinguishable from those of Mesp2-null fetuses. These observations indicate that Mesp1/Mesp2 double-null cells can differentiate into caudal sclerotome and possibly contribute to chondrogenesis.

Mesp1 and Mesp2 not only exhibit similar expression patterns but also share common bHLH domains as transcription factors. Previous studies using gene replacement experiments (Saga,1998) (Y.S. and S.K., unpublished) indicate that these genes can compensate for each other. However, the early lethality of double knockout mice hampered any further detailed analysis of somitogenesis. An obvious strategy to further elucidate the functions of Mesp1 and Mesp2 was, therefore,the generation of a conditional knockout allele for Mesp2 in Mesp1 disrupted cells in which the Cre gene is specifically activated in the paraxial mesoderm, which is now underway. Chimera analysis is also a powerful method as an alternative strategy. Comparisons of chimeras, composed of either Mesp2-null or Mesp1/Mesp2 double-null cells, made it possible to determine the contribution of Mesp1 to somitogenesis. Our results indicate that Mesp1 has redundant functions in the epithelialization of somitic mesoderm and additionally, by chimeric analysis, we were able to demonstrate the cell autonomy of Mesp1 and Mesp2 function during some critical steps of somitogenesis.

In Mesp1-null mice, epithelial somites with normal rostro-caudal polarity are generated, whereas Mesp2-null mice exhibit defects in both the generation of epithelial somites and the establishment of rostro-caudal polarity. Thus,it seems likely that Mesp2 function is both necessary and sufficient for somitogenesis. However, dermomyotome formation was observed, without normal segmentation, even in Mesp2-null mice. In view of the apparent redundant functions of Mesp1 and Mesp2 in somitogenesis, as demonstrated by our previous gene replacement study, it was possible that the Mesp1/Mesp2 double-null embryo would exhibit a much more severe phenotype in relation to somitogenesis. In our chimera analyses, both Mesp2-null and Mesp1/Mesp2 double-null cells exhibited complete caudalization of somitic mesoderm,indicating that Mesp1 function is not sufficient to rescue Mesp2 deficiency and restore rostro-caudal polarity. Likewise, both Mesp2-null and Mesp1/Mesp2 double-null cells were incapable of forming an initial segment boundary,showing that the contribution of Mesp1 is also minor during this process. By contrast, whereas Mesp1/Mesp2 double-null cells lacked any ability to epithelialize, Mesp2-null cells were occasionally integrated into epithelial somites and dermomyotome, indicating that the contribution of Mesp1 to epithelialization is significant and that Mesp1 can function in the absence of Mesp2 (Fig. 8). We therefore postulate that the epithelialization of dermomyotome, observed in Mesp2-null embryos, is dependent on Mesp1.

Conventional interpretations of the results of chimera analysis are generally based upon the regulative development of the vertebrate embryo and argue cell autonomy of specific gene functions in embryogenesis(Ciruna et al., 1997; Brown et al., 1999; Kitajima et al., 2000; Koizumi et al., 2001). Mesp1/Mesp2 double-null cells failed to form epithelial somites, even in the presence of surrounding wild-type cells. In addition, they were incapable of contributing to dermomyotome, where cell sorting occurs. This strongly suggests that Mesp factors are cell autonomously required for the epithelialization of somitic mesoderm. However, we also found striking non-cell autonomous effects of Mesp mutant cells on wild-type cell behaviors. That is, both types of Mesp mutant cell not only failed to undergo normal somitogenesis, but also inhibited the normal morphogenesis of wild-type cells. This implies that there are non-cell autonomous roles for Mesp factors in the establishment of the future somite boundary, as we will discuss further.

Initial epithelial somite formation is achieved by the mesenchymal-epithelial transition of cells located in the anterior PSM. A future somite boundary is established at a specific position in the PSM,followed by gap formation between the mesenchymal cell populations. Subsequently, cells located anterior to the boundary are epithelialized. This process is known to be mediated by an inductive signal from cells posterior to the boundary (Sato et al.,2002). Therefore, defects in epithelial somite formation can be explained in two principal ways: a lack of cellular ability to epithelialize(cell autonomous) and a lack of an inducing signal, which is produced in the anterior PSM by a mechanism mediated by Notch signaling (thus non-cell autonomous). In the case of chimeras of Mesp1/Mesp2 double-null cells, no local boundary formed by locally distributed wild-type cells was observed,i.e. even a gap between wild-type cells was never observed in the mixture of Mesp1/Mesp2 double-null cells and wild-type cells. It is likely, therefore,that the wild-type cell population can form a boundary only after separation from Mesp1/Mesp2 double-null cells (Fig. 8). By contrast, some local boundaries between epithelial wild-type cell clusters were occasionally observed in chimeras with Mesp2-null cells. Considering that there is functional redundancy between these transcription factors, it is possible that either Mesp1 or Mesp2 is necessary for the formation of a signaling center or source of the putative inductive signal. Hence, we cannot exclude the possibility that the lack of Mesp function may affect non-cell autonomous generation of the inductive signal in the anterior PSM.

Formation of epithelial somites requires paraxis, which is a transcription factor (Burgess et al., 1996; Nakaya et al., 2004). We observed that Mesp1/Mesp2 double-null cells at the medial sclerotomal region expressed Paraxis, indicating that Mesp factors are not absolutely required for Paraxis expression. Defects in epithelial somite formation in paraxis-null embryos, with normal Mesp2 expression(Johnson et al., 2001), and in Mesp2-null embryos, with normal Paraxis expression, imply that epithelial somite formation independently requires both gene functions.

The distribution of Mesp2-null cells in the Mesp2-null chimeric embryos may appear somewhat paradoxical, as they are localized at the rostral side in the incomplete somites but at the caudal side in the dermomyotome. Initial localization at the rostral and central region, however, is likely to be due to the relative lack of epithelialization functions. In mammalian and avian embryos, mesenchymal-to-epithelial conversion of the PSM commences from the rostral side of the future somite boundary, i.e. the caudal margin of the presumptive somite (Duband et al.,1987). Epithelialization then proceeds anteriorly in the dorsal and ventral faces and in such a process, Mesp2-null cells, which are less able to participate in epithelialization, may therefore be pushed to the central and rostral sides. Thus, the majority of the Mesp2-null cells localize to the central, prospective sclerotomal region and a small number of them are integrated in the future dermomyotomal region. The incomplete somites then undergo reorganization into dermomyotome and sclerotome, and small numbers of Mesp2-null cells in the dermomyotome may be sorted out to the caudal end. Therefore, the apparently complex distribution pattern of Mesp2-null cells is likely to reflect a combination of defects in epithelialization and rostro-caudal patterning. In the incomplete segments of Mesp2-null chimeric embryos, the Mesp2-null cells fail to acquire rostral properties even when localized at the rostral side. Moreover, in the dermomyotome, where rostro-caudal patterning is rescued, Mesp2-null cells are mostly localized in the caudal region. These observations suggested that the requirement of Mesp2 for the acquisition of rostral properties is cell autonomous. Similarly, it has been reported that presenilin 1 (Psen1) is required for acquisition of caudal half properties (Takahashi et al.,2000; Koizumi et al.,2001) and that Psen1-null cells cannot contribute to the caudal half of somites in chimeric embryos, showing cell autonomous roles for Psen1(Koizumi et al., 2001).

In a previous study, we have shown that the rostro-caudal patterning of somites is generated by complex cellular interactions involved in positive and negative feedback pathways of Dll1-Notch and Dll3-Notch signaling, and regulation by Mesp2 in the PSM (Takahashi et al., 2003). In chimeras with either Mesp2-null or Mesp1/Mesp2 double-null cells, the mutant cells were distributed evenly and did not show any sorting bias in a rostro-caudal direction in the PSM. Since both Mesp2-null and Mesp1/Mesp2 double-null cells have the ability to form caudal cells, it is likely that if wild-type cells could occupy the rostral part of future somite regions and have the ability to sort in the PSM, a normal rostro-caudal patterning would be generated. We did not observe this, however,and conclude that the presence of mutant cells lacking Mesp factors must have disrupted normal cellular interactions via Notch signaling. Thus these non-cell-autonomous effects of our mutant cells are strongly supportive of our previous contention that rostro-caudal pattering is generated by cellular interactions via Notch signaling.

We thank Mariko Ikumi, Seiko Shinzawa, Eriko Ikeno and Shinobu Watanabe for general technical assistance. This work was supported by Grants-in-Aid for Science Research on Priority Areas (B) and the Organized Research Combination System of the Ministry of Education, Culture, Sports, Science and Technology,Japan.