Redefining the role of ectoderm in somitogenesis: a player in the formation of the fibronectin matrix of presomitic mesoderm

Somites are transient segments of the paraxial mesoderm that are formed in a rostral-to-caudal progression during vertebrate embryogenesis. Each pair of somites, epithelial structures located symmetrically on either side of the neural tube, periodically detaches from the rostral end of the presomitic mesoderm (PSM), while new immature PSM tissue is added in the posterior part of the embryo. The strict temporal and spatial regulation of somitogenesis is of crucial developmental importance because segmental structures, such as the vertebrae, trunk skeletal muscles, peripheral spinal nerves and early blood vessels, develop according to the somite segmental pattern(Christ and Ordahl, 1995; Gossler and Hrabe de Angelis,1998).

The PSM can be divided in two regions that differ not only in terms of gene expression patterns, but also in the morphology of the PSM cells. In the caudal two-thirds of the PSM, high Fgf8 activity is believed to keep cells in a mesenchymal, undifferentiated state and oscillatory expression of segmentation clock genes occurs (Dubrulle et al., 2001; Freitas et al.,2005; Palmeirim et al.,1997). As PSM cells leave this caudal immature region, crossing what has been termed the determination front and enter the anterior third of the PSM, several changes take place. The cyclic expression of clock genes comes to a halt, retinoic acid signalling progressively replaces Fgf signalling (Diez del Corral et al.,2003) and somite anterior-posterior polarity is established(Saga and Takeda, 2001). Furthermore, the first signs of morphological somite formation occur as peripheral PSM cells undergo a mesenchymal-to-epithelial transition(Duband et al., 1987; Kulesa and Fraser, 2002). Finally, in the anterior-most region of the PSM, somite boundaries are specified and formed through the activation of transcription factors of the Mesp family (Sawada et al.,2000), most likely also involving Eph-ephrin(Barrios et al., 2003; Durbin et al., 2000) and Notch signalling (Sato et al.,2002).

It has been clearly demonstrated that the molecular segmentation of the anterior PSM is an intrinsic property of the PSM and does not require signalling from neighbouring tissues(Palmeirim et al., 1997). However, morphological somite formation does not occur in isolated cultured PSMs. For somites to form, ectoderm must be in contact with the PSM(Borycki et al., 2000; Borycki et al., 1998; Correia and Conlon, 2000; Palmeirim et al., 1998),suggesting that ectodermal signals, most likely Wnts(Borycki et al., 2000; Schmidt et al., 2004), are essential for morphological somite formation.

Interactions between PSM cells and the extracellular matrix molecule fibronectin have been implicated in somitogenesis(Duband et al., 1987; George et al., 1993; Lash et al., 1984; Lash and Ostrovsky, 1986; Ostrovsky et al., 1983). Fibronectin exists in plasma and cellular forms which, when assembled into a matrix, mediate a wide variety of cellular processes such as cell spreading,migration, proliferation, survival and differentiation(Pankov and Yamada, 2002). Fibronectin matrix assembly is a complex cell-dependent process that requires the engagement of fibronectin by cell surface integrins, usually theα5β1 integrin, and fibrillogenesis involving fibronectin-fibronectin binding (Mao and Schwarzbauer, 2005; Wierzbicka-Patynowski and Schwarzbauer,2003). Fibronectin 1 (Fn1)-null mouse embryos initiate gastrulation normally, but, although they have paraxial mesoderm, no morphologically distinguishable somites form(George et al., 1993; Georges-Labouesse et al.,1996). Furthermore, embryos null for both α5 and αV integrin subunits (Itga5 and Itgav) do not assemble a fibronectin matrix and fail to form somites(Yang et al., 1999). Together,these studies suggest that interactions between the fibronectin matrix and the PSM cells are crucial for normal somitogenesis.

The main objective of the present work was to address the contribution of the ectoderm and the fibronectin matrix in somitogenesis. We show that enzymatic treatments generally used to isolate PSM explants from the surrounding tissues destroy the fibronectin matrix surrounding the PSM. When these explants are cultured for 6 hours, no somites form. By contrast, we show for the first time that when collagenase is used, the endogenous fibronectin matrix remains intact and somites form even in the absence of all surrounding tissues. Moreover, addition of exogenous fibronectin to cultured pancreatin-isolated PSMs significantly improves their ability to form somites. Interestingly, Fn1 is primarily expressed in the ectoderm, whereas the PSM strongly expresses the fibronectin receptor Itga5, suggesting that ectoderm and PSM collaborate in constructing the fibronectin matrix present around the PSM. Finally, we demonstrate that inhibition of fibronectin matrix assembly in the presence of ectoderm abrogates somitogenesis. Based on these findings, we conclude that a fibronectin matrix is essential for somitogenesis and that a major role of ectoderm in this process is to provide the fibronectin protein for this matrix.

Fertilised chicken (Gallus gallus) eggs were incubated in a humidified atmosphere at 37.5°C until stage HH12-14(Hamburger and Hamilton,1992). Somite nomenclature follows Christ and Ordahl(Christ and Ordahl, 1995).

Embryos were fixed in 4% paraformaldehyde in phosphate buffer (77 mM Na₂HPO₄, 23 mM NaH₂PO₄, 0.12 mM CaCl₂) with 4% sucrose overnight at 4°C and processed for cryoembedding (Bajanca et al.,2004). Cryostat (Bright Clinicut 3020) sections (12 μm) were processed for immunohistochemistry or, if embryos had already been subjected to whole-mount in situ hybridisation, mounted in 80% glycerol in PBS and photographed.

Embryos were pinned down on resin-coated Petri dishes with PBS containing calcium and magnesium (PBS w/Ca²⁺Mg²⁺). PSM isolation,including removal of ectoderm and/or endoderm was achieved using tungsten needles, with or without the application of enzymes: pancreatin (4×,Gibco), an enzyme extract containing trypsin; dispase (2.4 U/ml, Roche), which cleaves fibronectin and collagen type IV(Stenn et al., 1989); and collagenase type II (125 U/ml, Sigma), which digests various collagens. None of these enzymes degraded laminin (data not shown). PSM isolation with pancreatin and dispase took 3-4 minutes, whereas collagenase isolation took about 8 minutes. Pancreatin was inactivated with fetal bovine serum (FBS,Gibco), whereas dispase and collagenase were washed away with PBS w/Ca²⁺Mg²⁺.

Embryo explants were positioned on 0.8 μm polycarbonate filters(Millipore) floating on M199 medium with 5% FBS and 10% chick serum(Palmeirim et al., 1997). In one set of experiments, the culture medium of one PSM was supplemented with 50-100 μg/ml of rat (Calbiochem) or bovine (Sigma) plasma fibronectin,whereas the contralateral PSM was incubated in control medium (medium only, or with BSA). Both plasma fibronectins produced the same results. In another set of experiments, posterior embryo explants were cultured with 50 or 100μg/ml of a 70 kDa fibronectin fragment (Sigma). Control embryos were incubated in medium with BSA.

A factorial analysis of variance (ANOVA) was used to test for the effect of the isolation method (pancreatin versus collagenase) and the culture method(PSM with or without endoderm or ectoderm) on the capacity of PSMs to form somites. The statistical significance of predicted specific differences was then analysed using contrast analysis. A paired Student's t-test was used to test for differences between explants grown with or without the fibronectin supplement. Differences between control embryo explants and explants cultured with the 70 kDa fibronectin fragment were tested with t-tests following a square-root transformation of the data. Statistical tests were computed with the STATISTICA 6 (StatSoft)programme.

Fibronectin immunohistochemistry was performed using a polyclonal antibody(Sigma, 1:400), or a monoclonal anti-cellular fibronectin antibody (clone FN-3e2; Sigma, 1:400) that recognises the EIIIA domain unique to cellular fibronectin (Barnes et al.,1995). Monoclonal anti-N-cadherin (cadherin 2) antibody (clone32;BD Transduction Laboratories, 1:100) was also used. Secondary antibodies were Alexa Fluor 488-, 546- or 568-conjugated anti-rabbit and anti-mouse IgG F(ab′)₂ fragments (Molecular Probes, 1:1000). For F-actin staining, Alexa Fluor 568-conjugated phalloidin was used (Molecular Probes,1:40) and TO-PRO3 (Molecular Probes, 1:500) incubated with secondary antibody and ribonuclease A (Calbiochem, 10 μg/ml) was used for nuclear staining.

Explants were fixed overnight at 4°C in 4% paraformaldehyde in PBS and washed in PBS. Both primary and secondary antibodies were incubated overnight in PBS containing 1% BSA and 0.1% Triton X-100 at room temperature, and explants were mounted in Vectashield (Vector Laboratories).

Cryosections were treated with 0.2% Triton X-100 in PBS for 20 minutes,washed in PBS and blocked for 30 minutes in 5% BSA in PBS. Incubation in primary antibody was overnight at 4°C, and incubation in secondary antibody was for 1 hour at room temperature. Sections were counterstained with 4′,6-diamidine-2-phenylidole-dihydrochloride (DAPI, 5 μg/ml in PBS,0.1% Triton X-100).

Images were acquired with an Olympus DP50 digital camera attached to an Olympus BX60 microscope equipped with epifluorescence and Nomarski optics, or with a Zeiss LSM 510 Meta confocal microscope. During confocal image acquisition, the detection parameters were adjusted to avoid under- or overexposed pixels. For fibronectin quantification, the average fluorescence intensity of six non-overlapping 75 μm-diameter regions of interest along each PSM was taken from the maximum z-projection of the seven most-superficial slices. The average fluorescence intensity of each PSM was normalised against its background fluorescence. ImageJ and Adobe Photoshop were used for image analysis and processing.

In situ hybridisation probes for Fn1, Itga5 and Itgavwere generated. The Fn1 probe was designed against a region present in all splice variants of the chick Fn1 mRNA(ffrench-Constant and Hynes,1988; Norton and Hynes,1987). Reverse transcription (RT) PCRs were used to isolate portions of chick Fn1, Itga5 and Itgav using the sense oligos 5′-CGTTCGTCTCACTGGCTACA-3′,5′-AGGTGCTGAGGGGGCAA-3′, 5′-TTCTCCACAGCAAACAGCC-3′ and the antisense oligos 5′-GGTCCTCTGGATGGGATTCT-3′,5′-CACGACGGTGAGCGAAG-3′, 5′-ATCCTCACCACAATCCAGCA-3′,respectively. The DNA fragments generated were cloned into the pCRII-TOPO vector (Invitrogen) and plasmid DNA was isolated. The constructs were confirmed by sequencing.

A plasmid carrying an insert of Paraxis (Tcf15) was kindly provided by C. Jouve (Developmental Biology Institute of Marseille,Marseille, France). The digoxigenin-labelled RNA probes were obtained from linearised plasmids, according to standard procedures adapted from Sambrook et al. (Sambrook et al.,1989).

Embryos and explants were fixed overnight at 4°C in 4% formaldehyde with 2 mM ethylene glycol-bis (β-amino-ethyl ether) tetra acetic acid(EGTA) in PBS, rinsed in PBT (PBS, 0.1% Tween 20), dehydrated in methanol and stored at -20°C. Whole-mount in situ hybridisation was performed according to Henrique et al. (Henrique et al.,1995), but BM Purple Substrate (Roche) was used as staining solution. Some embryos were cryosectioned as described above.

PSM and epithelial somites were isolated using pancreatin. Twenty samples of each of the following were collected for RT-PCR: (1) the four posterior-most somites; (2) the anterior third of the PSM; and (3) the posterior two-thirds of the PSM. Whole embryos were collected as positive control. Samples were processed for RNA extraction using the RNeasy Mini Kit with column DNA digestion (Qiagen). Total RNA was reverse-transcribed using the RevertAid First Strand cDNA Synthesis Kit (Fermentas). PCR was performed using the primers described above. Amplification using β-actin primers was used as positive control. Reactions from which reverse transcription (RNA control) or cDNA (water control) were omitted were used as negative controls.

Immunoreactivity for fibronectin was observed in embryo sections around the PSM and epithelial somites, as previously reported(Duband et al., 1987; Ostrovsky et al., 1983), and short strands of fibronectin immunoreactivity were distinguishable among PSM cells (Fig. 1A,B). Non-enzymatic removal of ectoderm from the anterior PSM in whole embryos revealed a network of fibrillar fibronectin-positive matrix surrounding the PSM (Fig. 1C). However, when PSMs were isolated with pancreatin or dispase, which are commonly used for this purpose, no immunoreactivity for fibronectin was detected(Fig. 1D,E). These data show that both pancreatin and dispase destroy the PSM fibronectin matrix within the time period necessary for PSM isolation. By contrast, we found that PSM isolation using collagenase preserved a fibrillar fibronectin matrix(Fig. 1F), suggesting that this is a better method for PSM isolation than pancreatin and dispase.

If fibronectin is important for somitogenesis, it is not surprising that PSMs isolated with pancreatin or dispase fail to form somites. However,ectoderm is able to support morphological somitogenesis even when explants are treated with these enzymes (Borycki et al.,1998; Correia and Conlon,2000; Palmeirim et al.,1998), whereas endoderm does not(Palmeirim et al., 1998). To evaluate the role of ectoderm, endoderm and fibronectin matrix in somite formation, pancreatin or collagenase were used to isolate three types of PSM explants: PSM without any adjacent tissues (PSM), PSM with the underlying endoderm (PSM+endoderm), or PSM with the overlying ectoderm (PSM+ectoderm). Two identical explants were isolated from each embryo. One was immediately fixed and immunolabelled for fibronectin, whereas the other was cultured for 6 hours (time to form 4 somites) and processed for fibronectin immunohistochemistry and F-actin staining(Fig. 2). The number of somites formed in each case was recorded (Fig. 3).

PSMs isolated from all surrounding tissues using pancreatin were negative for fibronectin immunolabelling (Fig. 2A) and generally did not present fully formed somites after 6 hours of culture (n=206; Fig. 2C, Fig. 3). Curiously, in pancreatin-treated PSM+ectoderm explants, fibronectin was observed between the PSM and the ectoderm(Fig. 2G) and these formed somites after 6 hours of culture (n=46; Fig. 2I, Fig. 3). By contrast, although some fibronectin is preserved between the PSM and the underlying endoderm(Fig. 2D), pancreatin-treated PSM+endoderm explants rarely formed somites (n=19; Fig. 2F, Fig. 3).

All PSM explants treated with collagenase (n=157) retained a fibrillar fibronectin matrix on their surface(Fig. 2J,M,P), and exhibited fully formed somites (Fig. 2L,O,R, Fig. 3). This was true even for the PSMs cultured in complete isolation from surrounding tissues (n=66).

We conclude that PSMs isolated with an enzymatic treatment that retains the fibronectin matrix are able to form somites in the absence of neighbouring tissues. Pancreatin-treated PSM+ ectoderm explants retain fibronectin matrix between the two tissues, which may facilitate the formation of somites. Curiously, somites rarely form in pancreatin-treated PSM+endoderm explants even though some fibronectin is retained between the two apposed tissues.

In order to uncover the relative contribution of ectoderm, endoderm and the fibronectin matrix in morphological somite formation, we analysed our quantitative data in more detail (Fig. 3). A comparison of PSM and PSM+endoderm explants isolated with pancreatin versus collagenase, indicates that the difference in the capacity to form somites is dependent both on the enzyme (P<0.0001) and the presence of endoderm (P=0.03). To determine whether this effect is mediated by endoderm itself, or by the fibronectin retained between the two apposed tissues, we compared the somite-forming capacity of collagenase-isolated PSM and PSM+endoderm explants. No significant difference(P=0.71) in the capacity to form somites was found between these two explant types. We conclude that endoderm does not by itself enhance somite formation in cultured PSMs and that the slight increase in somite-forming capacity of pancreatin-isolated PSM+endoderm explants must be due to the fibronectin matrix that is retained between the two tissues.

We then compared PSM and PSM+ectoderm explants isolated with pancreatin versus collagenase. We found that the difference in the capacity to form somites is both dependent on the enzyme (P<0.0001) and on the presence of ectoderm (P<0.0001). Unlike in the case of endoderm, a significant interaction between these two variables was detected(P<0.0001), indicating that they mutually enhance each other's effect. To determine whether ectoderm enhances somite formation by itself, we compared collagenase-isolated PSM and PSM+ectoderm explants and found that these differ significantly in their somite-forming capacity(P<0.0001). We conclude that, although isolation with collagenase enables PSMs to form somites, the additional presence of ectoderm enhances this capacity.

Pancreatin-treated PSM+ectoderm explants retain fibronectin matrix only between ectoderm and PSM, whereas collagenase-treated PSM+ectoderm explants retain an extensive fibronectin matrix all around. A comparison between these two situations did not reveal a significant difference between the number of somites formed (P=0.075), indicating that somite formation in PSMs in contact with ectoderm is not improved by the presence of a fibronectin matrix all around the PSM at the beginning of the culture period.

Altogether, these results demonstrate that whereas collagenase-isolated PSMs are clearly capable of forming morphological somites, the additional presence of ectoderm enhances this capacity further, whereas endoderm has no effect.

Paraxis expression in the anterior PSM is necessary for epithelial somite formation (Burgess et al.,1996) and it has been suggested that ectoderm is essential to induce its expression (Correia and Conlon,2000; Sosic et al.,1997). To determine if fibronectin and/or ectoderm influence Paraxis expression, pancreatin- or collagenase-isolated PSMs(Fig. 4B,D) and their contralateral control halves (Fig. 4A,C) were cultured for 6 hours and assayed for Paraxisexpression. Both pancreatin-isolated (n=11) and collagenase-isolated(n=11) PSMs showed Paraxis expression(Fig. 4B,D), identical to the contralateral control halves (Fig. 4A,C), although only collagenase-isolated PSMs formed somites(Fig. 4D). We conclude that Paraxis expression in the anterior PSM and newly formed somites is independent of all surrounding tissues and of an intact fibronectin matrix. Furthermore, Paraxis expression in pancreatin-isolated PSMs cultured for 6 hours is not sufficient to drive somite formation.

Pancreatin-isolated PSMs show a clear fibronectin-immunoreactivity after 6 hours of culture. This fibronectin matrix was recognised by a cellular fibronectin-specific antibody (Fig. 5A), indicating that it is at least partially PSM-derived and not only from the serum supplement(McKeown-Longo and Mosher,1983). However, practically no somites were formed in these PSM explants. Thus, the possibility arises that the amount of fibronectin(PSM-plus serum-derived) available to the fibronectin-stripped PSMs is insufficient to support somite formation. PSM isolation using pancreatin or dispase removes all the endogenous fibronectin(Fig. 1D,E, Fig. 2A) and after 2 hours of culture these PSMs showed only a faint immunoreactivity for fibronectin(Fig. 5B,E). Only after 4 hours of culture had a fibronectin matrix formed, which was similar in labelling intensity to the one observed after 6 hours(Fig. 5C-E).

To determine whether the amount of fibronectin available to the PSMs could be the limiting factor for their capacity to form somites, pancreatin-isolated PSM pairs (n=20) were cultured in fibronectin-supplemented medium versus control medium. The average number of somites formed by PSMs cultured with fibronectin for 6 hours was significantly higher (P=0.002) than that formed by the contralateral PSMs grown in control medium(Fig. 5F-J). Cells of PSMs cultured with fibronectin had aligned, elongated nuclei and apical enrichment of F-actin and N-cadherin, which was not observed in PSMs cultured in control medium (Fig. 6, compare B,E with A,D). In fact, cell morphology in PSMs cultured with fibronectin was very similar to that of collagenase-isolated PSMs(Fig. 6, compare B,E with C,F). We conclude that the culture of fibronectin-stripped PSMs in fibronectin-supplemented medium significantly improves their capacity to form somites and that their cells exhibit a morphology typical of somitic cells.

Our previous observations indicate that although isolated PSMs are able to produce and assemble a fibronectin matrix in culture, the amount of fibronectin available to them is insufficient to support somitogenesis. This led us to investigate where fibronectin is normally produced.

In situ hybridisation for chick Fn1 transcripts revealed strong expression in the ectoderm and in the posterior part of epithelial somites(Fig. 7A-D). Furthermore, Fn1 was expressed in the endoderm underlying anterior PSM and somites(Fig. 7B-D). Unexpectedly, Fn1 mRNA was not detected in the rostral PSM and only a faint signal was present in the caudal PSM (Fig. 7B,C). This was confirmed by removal of the ectoderm prior to fixation (Fig. 7E-I). Interestingly, RT-PCR analysis showed that both anterior and posterior PSM express Fn1 mRNA (Fig. 7J), which is in accordance with the cellular fibronectin immunoreactivity present on pancreatin-isolated PSMs after 6 hours of culture(Fig. 5A). Together, these data demonstrate that the major site of Fn1 transcription is the ectoderm overlying the PSM, whereas the PSM expresses only low levels of Fn1mRNA.

Our next step was to identify where the integrins known to mediate fibronectin fibrillogenesis (Yang et al.,1999) are expressed. The main integrin responsible for fibronectin fibrillogenesis, Itga5, was strongly expressed in the posterior two-thirds of the PSM (Fig. 7L,N), whereas Itgav transcripts were not detected in the PSM by in situ hybridisation (data not shown). Little or no Itga5mRNA expression was detected in the anterior-most PSM(Fig. 7L,M), but was again detected in epithelial somites (Fig. 7L). These data demonstrate that ectoderm is the major source of fibronectin, whereas the PSM expresses Itga5, the receptor necessary for its assembly into a fibrillar matrix.

We have shown that somite formation occurs in collagenase-isolated PSMs and in pancreatin-isolated PSMs cultured with fibronectin. Our results also demonstrated that although collagenase-isolated PSMs were able to form somites, the presence of ectoderm improved this capacity further. We now show that ectoderm is the major producer of fibronectin protein in the region of the PSM. The question arises, is the beneficial effect of ectoderm due to its production of fibronectin or does it provide some other essential signal to the PSM that contributes to somitogenesis?

To test this hypothesis, we cultured posterior embryo explants for 12 hours in medium containing a 70 kDa fragment of fibronectin known to block fibronectin fibrillogenesis (McKeown-Longo and Mosher, 1985). Thus, fibronectin matrix assembly is blocked in the presence of ectoderm. Endoderm was removed from one side of the embryo to improve the accessibility of the peptide, while the other side was kept intact. Control embryos formed the expected 7-8 somites(Fig. 8A,C,E). However, when explants were cultured in the presence of 50 μg/ml of the fibronectin peptide (Fig. 8B,E),somitogenesis was less efficient (this was only significant in the absence of endoderm; P=0.005). Strikingly, when embryos were cultured in the presence of 100 μg/ml of the peptide, somitogenesis was practically abolished, both in the presence (P=0.002) and absence(P=0.001) of endoderm (Fig. 8D,E). We conclude that fibronectin matrix assembly is absolutely necessary for morphological somite formation, even when ectoderm is left intact over the PSM.

The notion that fibronectin might be an important player in somitogenesis is not new. A fibronectin matrix outlines epithelial somites in both chick and mouse and its distribution pattern in the PSM shows a more complex matrix anteriorly than posteriorly, such that it has been suggested that this matrix is progressively laid down as the PSM matures(Lash et al., 1984; Lash and Ostrovsky, 1986; Ostrovsky et al., 1983; Ostrovsky et al., 1988). Experiments involving the addition of fibronectin to dissociated chick PSM cells caused their aggregation, suggesting that fibronectin might normally play a role in compacting and/or increasing the cell-cell adhesiveness of anterior PSM cells (Lash et al.,1984; Lash et al.,1987). In agreement with this, we show that adding exogenous fibronectin to pancreatin-isolated PSMs significantly improved their capacity to form somites.

In Fn1-null mouse embryos, paraxial mesoderm initially forms normally but does not generate any morphologically distinguishable somites(Georges-Labouesse et al.,1996). The Fn1-null phenotype thus represents one of the most severe somitogenesis phenotypes in a single mouse gene (reviewed by Pourquié, 2001). Moreover, a fibronectin matrix is required for normal somitogenesis and axial extension in Xenopus (Marsden and DeSimone, 2003) and both fn1 and itga5knock-down interfere with somite epithelialisation and boundary maintenance in zebrafish (Julich et al.,2005; Koshida et al.,2005).

Many authors have reported that PSMs cultured in the absence of all surrounding tissues do not form somites(Borycki et al., 1998; Borycki et al., 2000; Correia and Conlon, 2000; Lash et al., 1984; Lash and Ostrovsky, 1986; Lash and Yamada, 1986; Linker et al., 2003; Packard, 1976; Packard, 1980; Palmeirim et al., 1998). In all these studies, PSMs were isolated with pancreatin (which contains trypsin), trypsin or dispase. Here we demonstrate that 3-4 minutes of exposure to pancreatin or dispase is sufficient to remove all fibronectin immunoreactivity from isolated PSM, whereas collagenase-isolated PSMs retain their fibronectin matrix and form somites. We thus show for the first time that isolated PSMs can form somites in the absence of all surrounding tissues as long as the fibronectin matrix is preserved. We propose that the major cause for the reported failure in morphological somite formation in isolated PSMs is that the enzymes commonly used for their isolation degrade the fibronectin matrix.

Our results demonstrate that ectoderm strongly expresses Fn1,whereas the PSM expresses very little (see proposed model in Fig. 9). Cultured ectoderm assembles an extensive, basally located fibronectin matrix(Newgreen and Thiery, 1980),showing that ectodermal Fn1 expression results in the production of fibronectin, which is predominantly secreted basally towards the PSM. Furthermore, fibronectin-positive `dense bodies' appear to come off the basal surface of the ectoderm as fibronectin-positive extracellular material accumulates on the mesoderm (Sanders,1986). In fact, extracellular matrix material has been described as travelling freely between germ layers in the early chick embryo(Harrisson et al., 1985). Thus, we suggest that the bulk of the PSM fibronectin matrix originates from fibronectin protein produced and secreted by the ectoderm(Fig. 9).

The posterior two-thirds of the PSM strongly express Itga5. Immunoreactivity for the β1 integrin subunit has been shown to be enriched on the surface of the PSM and is much lower within the PSM(Duband et al., 1986; Krotoski et al., 1986). Given that fibronectin matrix accumulates almost exclusively on the surface of the PSM, we suggest that α5β1 protein is enriched on the outer surface of the PSM where it regulates fibronectin assembly(Fig. 9).

Our results also show that the amount of fibronectin available to the PSM is crucial for the formation of a fibronectin matrix that can support somitogenesis. As the PSM-derived fibronectin is not sufficient to support somitogenesis, the PSM appears to depend on an external source of fibronectin to construct its matrix. Our observations strongly suggest that this external source is ectoderm.

It is clear that Paraxis is necessary for somitogenesis(Burgess et al., 1996), and its long-term expression depends on the overlying ectoderm(Correia and Conlon, 2000; Schmidt et al., 2004; Sosic et al., 1997). However, Paraxis expression is maintained in the paraxial mesoderm for up to 10 hours without any interaction with ectoderm(Linker et al., 2005), which agrees with our observation of strong Paraxis expression in isolated PSMs cultured for 6 hours (Palmeirim et al., 1998) (this study). Thus, although ectoderm is essential to maintain Paraxis in the dermomyotome(Linker et al., 2005), it is dispensable for initial Paraxis expression in PSM and early somites.

Our results instead provide evidence of a previously unrecognised role of ectoderm in somitogenesis, which is the production of the bulk of fibronectin necessary for morphological somite formation. However, we also show that if enough fibronectin has assembled into a matrix in the anterior PSM, the ectoderm is no longer required for somite formation. How can these results be reconciled with the existing literature that advocates an essential signalling role of ectoderm in morphological somite formation? We suggest that when ectoderm is removed from the full extension of the chick PSM in ovo(Sosic et al., 1997), or from the caudal PSM (Schmidt et al.,2004), somites do not form because the PSM is deprived of the major source of fibronectin necessary to construct its matrix(Fig. 9). In fact, if ectoderm is replaced over the PSM after its manipulation, there is little effect on somitogenesis (Packard et al.,1993). Conversely, we propose that when ectoderm is physically separated from the anterior PSM, epithelial somites form(Linker et al., 2005) because the fibronectin matrix in the anterior PSM remains undamaged. Furthermore,when dispase-treated PSMs are cultured within a bag of tail ectoderm, somite formation is partially restored (Correia and Conlon, 2000). We show that pancreatin-treated PSMs significantly increased their ability to form somites when cultured in fibronectin-enriched medium. It is thus tempting to suggest that the ectoderm bags used by Correia and Conlon (Correia and Conlon, 2000) provided a rich source of fibronectin protein to the dispase-isolated PSMs, allowing them to assemble a fibronectin matrix and partially restore somitogenesis. We propose that for morphological somites to form, PSMs need a major external source of fibronectin - the overlying ectoderm - in order to build a fibronectin matrix able to support somite formation.

Although the role of Wnts in the segmentation clock and in the differentiation of epithelial somites is well documented (reviewed by Auhehla and Pourquié, 2006; Rida et al.,2004; Yusuf and Brand-Saberi,2006), their role in the epithelialisation of somites is much less clear. Wnt6, the only Wnt expressed in the ectoderm overlying the anterior PSM(Cauthen et al., 2001; Rodriguez-Niedenfuhr et al.,2003; Schubert et al.,2002), is the most likely candidate to play a role in somite formation. Wnt6-expressing cells placed in the caudal PSM after ectoderm removal rescue Paraxis, Pax3 and MyoD expression after 18 hours of culture (Schmidt et al.,2004). However, expression of these genes is independent of somite epithelialisation (Linker et al.,2005; Palmeirim et al.,1998; Burgess et al.,1996; Wilson-Rawls et al.,1999) (this study). Furthermore, our results show that somites form in collagenase-isolated PSMs in the absence of ectoderm and that somitogenesis is severely affected when fibronectin matrix assembly is inhibited in its presence. These data therefore strongly argue against a direct role of ectodermal Wnts on the epithelialisation of the anterior PSM. Although ectodermal canonical Wnt signalling cannot be ruled out as a player in morphological somite formation [some somite-like condensations form in dispase-isolated quail PSM explants cultured with LiCl(Borycki et al., 2000)], our experiments show that its role in somite epithelialisation can only be minor in comparison with that of the fibronectin matrix.

We have demonstrated that the PSM fibronectin matrix is essential for morphological somite formation. Fibronectin matrix assembly is a complex process (reviewed by Wierzbicka-Patynowski and Schwarzbauer, 2003; Mao and Schwarzbauer, 2005) and its strengthening and maintenance is dependent on the amount of free fibronectin available for continuous incorporation, even after an initial fibrillar matrix has formed. This might explain why collagenase-isolated PSM+ectoderm explants form significantly more somites than collagenase-isolated PSMs. Since ectoderm produces fibronectin during the 6-hour culture period, the matrix of PSMs cultured with ectoderm increases in strength, whereas PSMs isolated without ectoderm rely on the matrix they have at the time of isolation. The 70 kDa fibronectin fragment used in our experiments prevents the fibronectin-fibronectin binding essential for the formation of new fibrils and it also binds to and displaces fibronectin within young fibrils(McKeown-Longo and Mosher,1985), explaining its dramatic effect on somite formation. We propose that the presence of ectoderm promotes a continuous fibronectin fibrillogenesis that maintains and increases the strength of the PSM fibronectin matrix, essential to support somitogenesis.

Is the PSM fibronectin matrix solely providing a structural framework or could fibronectin also be playing an instructive role? Studies of Xenopus convergent extension have demonstrated that fibronectin affects cadherin-mediated cell-cell adhesion, which is important for morphogenetic movements (Marsden and DeSimone, 2003). Fibronectin signalling through α5β1 integrin upregulates N-cadherin expression in cultured myoblasts(Huttenlocher et al., 1998). Since N-cadherin has been implicated in somitogenesis(Horikawa et al., 1999; Linask et al., 1998; Radice et al., 1997) and the presence of fibronectin correlates with the cellular redistribution of N-cadherin and F-actin (this study), α5β1-mediated signalling might affect N-cadherin during somitogenesis. Integrin α5β1 and fibronectin have also been shown to regulate cell behaviour during convergent extension in Xenopus (Davidson et al., 2006). This is controlled by small GTPases of the Rho family(Fukata et al., 2003), which have been implicated in the regulation of the mesenchymal-to-epithelial transition of the anterior PSM (Nakaya et al., 2004).

In summary, we conclude that fibronectin is an essential player in somitogenesis and that the PSM fibronectin matrix is the product of a collaboration between the PSM and the overlying ectoderm. We propose that the continuous build-up of the fibronectin matrix is essential for somitogenesis and that the removal of this matrix or the inhibition of its assembly compromises morphological somite formation. It is also tempting to suggest that the increasing complexity of the fibronectin matrix might transmit specific signals, so that cell behaviour in the anterior PSM changes,promoting epithelialisation. Whether and how fibronectin collaborates with other factors to achieve this morphological transition into somites will be interesting issues to explore.

We thank Gabriel G. Martins and Fernanda Bajanca for helpful suggestions during the course of this work and for reading the manuscript; Jorge Palmeirim for statistical advice; Luís Marques for valuable help; and C. Jouve for cDNA. Financial support was from Fundação para a Ciência e a Tecnologia (FCT)/FEDER projects POCTI/BCI/40754/2001 and POCI/BIA-BCM/59201/2004 and the FP6/EU Network of Excellence `Cells into Organs' of which P.R., C.L., R.P.A., G.R., I.P. and S.T. are members. L.C. was supported by the European Social Fund contract 1/3.2/PRODEP/2001. R.P.A. was supported by FCT (SFRH/BPD/9432/2002).