Somites are an embryonic manifestation of the underlying metameric organization of the vertebrate body plan. Through changes in cell shape, adhesion, and position, mesodermal cells give rise to somites, regular, repeated structures along the anterior/posterior axis of the embryo, which then differentiate into skeletal muscle, bone, and portions of the dermis (Pourquie,2001). The periodic expression of key patterning genes establishes anterior/posterior identities and directs subsequent somite boundary formation (Aulehla and Herrmann,2004). The specification of somite boundaries must then be coordinated with equally precise physical changes in cell shape, cell-cell, and cell-matrix adhesion to form individual somites. While the molecular pathways involved in patterning somites have been studied in some detail, the mechanisms responsible for their morphological segmentation remain elusive.
Integrins are a diverse family of transmembrane receptors that mediate interactions with the extracellular matrix, including cell adhesion, migration, and signaling (Shimaoka et al.,2002). These receptors function as heterodimers consisting of alpha and beta subunits. The integrin α5β1 complex binds with high affinity to fibronectin and can assemble a fibronectin matrix in cell culture (Wu et al.,1993), and in vivo (Julich et al.,2005; Koshida et al.,2005; Yang et al.,1993,1999). There is ample evidence that integrin α5β1 plays a conserved role in somitogenesis. In mice, loss of integrin α5 is lethal at 10–11 days of gestation with embryos exhibiting posterior truncation and, in some genetic backgrounds, failure of epithelialization in the remaining anterior somites (Yang et al.,1993,1999). Integrin α5 and its binding partner, fibronectin, have also been implicated in the establishment of anterior, but not posterior, somite boundaries in zebrafish (Julich et al.,2005; Koshida et al.,2005), although knockdown of both fibronectin genes in zebrafish (fn1 and fn3) results in a more severe phenotype (Julich et al.,2005). Additionally, injection of an integrin β1 function blocking antibody into quail embryos disrupts somite morphology and adhesion to the notochord (Drake et al.,1992).
In all of these model systems, morphological formation of the somite boundary primarily involves a mesenchymal to epithelial transition, although the precise cell movements vary slightly between species (Stern and Keynes,1987; Ostrovsky et al.,1988; Stickney et al.,2000). In contrast, somite formation in Xenopus laevis is characterized by the orchestrated rotation of blocks of cells, a process that depends on a unique series of cell rearrangements and coordinated changes in cell motility, shape, and adhesion. At the onset of somitogenesis, blocks of mediolaterally elongated presomitic cells bend anteriorly in order to effect a 90° turn that brings them parallel to the long axis of the embryo. During somite turning, the distal pole of each presomitic cell exhibits filopodial extensions while the flattened medial pole moves anteriorly along the notochord (Afonin et al.,2006), suggesting an active migratory process. Both α5 and β1 integrin subunits are expressed in Xenopus somites (Gawantka et al.,1992; Joos et al.,1995), and overexpression of integrin α3 (Meng et al.,1997) or a dominant-negative form of integrin β1 (Marsden and DeSimone,2003) disrupts somite formation. However, it is unknown to what extent the role of integrin α5β1 is conserved in the unique method of somitogenesis occurring in Xenopus.
We investigated the role of integrin α5 in Xenopus somitogenesis through the use of a splice-blocking antisense morpholino. Targeted injection of the integrin α5 morpholino disrupted somite turning and subsequent boundary formation in the absence of changes in presomitic mesoderm or segmentation markers. Morpholino-treated embryos also exhibited a decrease in integrin β1 accumulation and disordered fibronectin matrix at early somite boundaries. As embryos matured, we observed incomplete recovery of myotome boundaries, which was more pronounced in posterior versus anterior somites. These results highlight the importance of integrin α5 in both cell turning and somite boundary formation in Xenopus laevis and indicate that the role of integrin α5 in somite formation is conserved among vertebrates.
The intgα5 MO Causes Severe Curvature Along the Anterior/Posterior Axis
The integrin α5 subunit is expressed in Xenopus presomitic mesoderm and persists in somites upon their formation (Joos et al.,1995). In order to examine the function of the integrin α5 subunit in Xenopus laevis somite development, we developed an integrin α5 antisense morpholino targeted to the splice donor site of exon 6 (intgα5 MO; Fig. 1A). To determine the efficacy of the intgα5 MO, 4-cell-stage embryos were injected bilaterally with 50 ng of the intgα5 MO or a control morpholino, and dorsal explants from stage-22 embryos were collected and analyzed by RTPCR to establish the presence of mis-spliced transcripts. In intgα5 MO injected embryos, a lower band representing a splice variant missing exon 6 was identified (Fig. 1B), while the higher band representing the normally spliced transcript was greatly reduced or absent. Sequence analysis of the smaller fragment confirmed mis-splicing of the integrin a5 mRNA, which results in a frameshift and subsequent premature stop codon. The mis-spliced mRNA resulting from intgα5 MO injection is predicted to produce a non-functional truncated protein lacking the transmembrane domain (Fig. 1C).
To determine the requirement for integrin α5 in somitogenesis, unilateral injections of intgα5 MO were targeted to the presumptive somite region at the 4-cell stage. Gross morphology of injected embryos was normal through cleavage, gastrulation, and early neurulation stages. Subsequent to the onset of somitogenesis (stage 20), intgα5 MO injected embryos began to show a curvature along the anterior/posterior axis with the concave side consistently corresponding to the side of injection (Fig. 2B). In contrast, curvature was rarely observed at these stages in embryos injected with the control MO, with no bias toward injected or uninjected sides (Fig. 2A). Axis curvature became increasingly pronounced as intgα5 MO treated embryos developed, and resulted in profoundly bowed embryos by the early tailbud stages (stage 34; Fig. 2D). Co-injection of 500 pg of Xenopus integrin α5 mRNA with the intgα5 MO greatly attenuated the curvature at early neurulation stages (not shown) and early tailbud stages (Fig. 2F; Table 1). Co-injection of the integrin α5 mRNA with a control MO had no effect on gross morphology of embryos (Fig. 2E; Table 1). In all experiments, GFP mRNA was co-injected with the morpholinos to confirm targeting. As expected, injections resulted in GFP-positive cells residing mostly in the somitic mesoderm, although occasionally GFP was also found in small numbers of cells in the adjacent notochord, neural tube, or epithelial cells (Fig. 3F).
Table 1. Rescue of intgα5 MO Curved Phenotype in Stage-35 Embryos by Integrin α5 mRNA Co-Injection
Left-directed curvature is greater than or equal to 90°.
Recent work has demonstrated that the integrin α5β1 complex is required for the orchestration of cell movements during gastrulation in Xenopus laevis (Davidson et al.,2006). In contrast, we did not observe any alterations in timing or morphology of embryonic development during gastrulation, and, in fact, observed normal fibronectin deposition in gastrulating embryos injected with the intgα5 MO (data not shown). However, we do not feel that these results are necessarily contradictory. Integrin α5 is maternally expressed, and both existing mRNA as well as protein would be unaffected by the splice-blocking intgα5 MO. In addition, our injection leads to a mosaic distribution of morpholino within the embryo (Fig. 4C,D). Thus, we hypothesize that the levels of integrin α5 from maternal stores and expression in cells lacking the morpholino are sufficient to carry the embryo through these early stages, but as levels fall during later stages defects in somitogenesis become apparent. Consistent with this hypothesis, targeting of the morpholino to the blastocoel roof (where fibronectin matrix crucial for gastrulation is assembled; Marsden and DeSimone,2001) did not affect gastrulation movements, but later produced a disruption in neurulation (data not shown). Therefore, we feel confident that the effects of the intgα5 MO on somitogenesis are specific to that process, and not due to prior disruption of gastrulation movements.
The intgα5 MO Does Not Disrupt Generation or Patterning of the Presomitic Mesoderm
The changes in gross morphology observed after treatment with the intgα5 MO could arise from a disruption of presomitic mesoderm generation, somite specification, or somite morphogenesis. We began our phenotypic analysis of intgα5 MO injected embryos by asking whether somite mesoderm was appropriately specified. Injected embryos collected at stages 19 and 25 were subjected to in situ hybridization with a probe for the muscle marker MyoD. At stage 19, no differences in pattern or level of MyoD staining were observed between intgα5 MO injected and uninjected sides of embryos (Fig. 3A,B; intgα5 MO, n = 10; Control MO, not shown, n = 5). At stage 25, the severe curvature of the intgα5 MO treated embryos prevented a clear comparison of MyoD staining patterns. However, there was no evidence of a discernable decrease in the level of MyoD staining in response to integrin α5 knockdown (Figure 3C,D; Control MO, n = 11; intgα5 MO, n = 7). Importantly, the intgα5 MO injected sides exhibited some evidence of segmentation (arrows in Fig. 3D), although the segments did appear substantially compressed. In addition, in situ hybridization staining for PAPC, a marker of segmental patterning (Kim et al.,2000), did not reveal any consistent disruptions of pattern or level of staining (data not shown).
Prior to somite turning, generation of presomitic mesoderm requires the condensation and elongation of paraxial mesoderm perpendicular to the notochord and neural tube (Wilson and Keller,1991). Immunofluorescence staining for β-catenin (Fig. 3E,F) or myosin heavy chain (MHC; Fig. 4B) revealed normal paraxial mesoderm morphology at intgα5 MO injection sites as compared to the uninjected side, indicating that neither convergent extension movements nor elongation of presomitic mesodermal cells were disrupted by integrin α5 knockdown (n = 10). The sum of these results pointed to a defect in somite morphogenesis, rather than presomitic mesoderm generation, segmentation, or muscle specification.
Knockdown of Integrin α5 Disrupts Formation of Somite Boundaries
Xenopus embryos utilize a pattern of movements to generate somites, which is distinct from other vertebrate embryos. At the onset of somitogenesis, blocks of cells from the presomitic mesoderm coordinately turn 90° around their long axes to take up a final position parallel to the notochord (Afonin et al.,2006; Fig. 4A,E). Confocal images of embryos stained for MHC revealed that intgα5 MO treated somitic cells were delayed in this turning process, leaving many cells perpendicular to the notochord when their counterparts on the uninjected side of the embryo had come to their final orientation parallel to the notochord (Fig. 4B,G). In more mature somites (further anterior), many intgα5 MO injected (GFP-positive) cells achieved a roughly parallel orientation, but the junctions between adjacent somites tended to remain poorly defined (Fig. 4B). This trend toward recovery continued as embryos matured and by late tailbud stages the majority of myocytes targeted with the intgα5 MO had successfully rotated and extended in an anterior/posterior direction (Fig. 4I). However, in the anterior myotome, myocyte terminations were highly random, suggesting that boundary formation did not fully recover. In the far posterior myotome, some boundaries formed and the classic chevron shape could be distinguished, although myotome morphology remained substantially disrupted. intgα5 MO injected embryos also consistently exhibited a foreshortening of the injected side that was apparent at both early neurula (not shown) and late tailbud stages (brackets in Fig. 4I).
Integrin α5β1 binds to fibronectin, and is required for fibronectin matrix formation (Wu et al.,1993). Intriguingly, the integrin β1 subunit was consistently absent from presumptive somite boundaries of late neurula stage embryos (Fig. 5A). This did not appear to be due to a general loss of integrin β1, as staining at cell membranes of somite cells was unaffected. Analysis of older embryos revealed that integrin β1 does eventually become enriched at somite boundaries, although, as observed with MHC staining, the boundaries themselves remain irregular (Fig. 5C).
We previously reported the accumulation of fibronectin at the anterior borders of somites during somite rotation (Kragtorp and Miller,2006), suggesting a role for fibronectin in this turning process. In normal embryos, strong fibronectin staining is observed at somite boundaries and between somites and the adjacent notochord. In embryos exposed to the intgα5 MO, fibronectin deposition at somite boundaries is disorganized and discontinuous (Fig. 6A). In addition, the somites are pulled away from the notochord and the intervening fibronectin matrix is fragmented. A cross-sectional view of stacked images through the somite boundaries reveals large holes in the fibronectin matrix lining the boundary, and a sparse fibronectin matrix connecting the somite to the notochord (Fig. 6B).
How groups of cells coordinate changes in cell shape, adhesion, and motility in order to form complex tissues is one of the central questions of developmental biology. The dynamic cell rearrangements that occur during somitogenesis make this an excellent model of tissue morphogenesis. In Xenopus laevis, cells must re-orient their long axes in a synchronized manner in order to create a morphologically distinct boundary between somites. We demonstrate here that loss of integrin α5 delays the process of rotation and the accumulation of integrin β1 at early somite boundaries, and results in persistently abnormal myotome boundaries. In addition, fibronectin matrix formation is disrupted at the boundaries between adjacent somites as well as between somites and surrounding tissues. Interestingly, these effects are less severe in the far posterior somites of the developing tail. Thus, integrin α5 is required for somite rotation and boundary formation in Xenopus, though its requirement appears to vary at different positions along the anterior/posterior axis.
The lack of observable effects of the intgα5 MO on events prior to somitogenesis highlights the usefulness of the splice morpholino strategy in avoiding defects associated with genes required in early developmental stages, and allowing examination of their function in later developmental events. As noted earlier, the maternal expression of Integrin α5 results in a pool of pre-existing mRNA and protein that is unaffected by the splice-blocking intgα5 MO. As this pool diminishes, generation of new mRNA is disrupted, resulting in the creation of nonfunctional protein. This method enabled us to directly examine the role of integrin α5 during somitogenesis, and will likely prove very useful in the investigation of later developmental roles of other Xenopus genes as well.
Integrin α5 in Somite Cell Turning
Analysis of cell turning behaviors during Xenopus somitogenesis reveals that cell movements are coordinated, but individualized (Afonin et al.,2006). Cells adjacent to the notochord are the first to show changes in orientation, and throughout somite formation cell shapes are variable, suggesting that each cell is following its own “best path” during rotation (Youn and Malacinski,1981; Wilson et al.,1989; Afonin et al.,2006). In addition, protrusive activity is consistently seen to increase during the turning process, indicating that cells are adopting a migratory phenotype. These observations raise the question of what surface somitic cells are using to create the traction necessary for rotation.
One possibility is that somitic cells crawl along other cells using cadherin-based adhesive contacts. Indeed, both paraxial protocadherin (PAPC) and protocadherin in neural crest and somites (PCNS) have been implicated in somite patterning in Xenopus (Kim et al.,2000; Rangarajan et al.,2006) and in mice (Rhee et al.,2003). In addition, disruption of cadherin I in Xenopus embryos results in disorganized somites, providing evidence that cell–cell interactions are important for the cell movements involved in somitogenesis (Giacomello et al.,2002). However, as this disruption is accompanied by a loss of MyoD staining, it is possible that the phenotype is secondary to altered somite specification (Giacomello et al.,2002).
Alternatively, somitic cells could move by migrating along the extracellular matrix. Immunostaining reveals that integrin β1 is present at cell borders and at the emerging somite boundary (Fig. 5). Furthermore, fibronectin matrix is laid down at the anterior boundaries of turning somites, and lines the junction between somites and surrounding tissues (Fig. 6; Kragtorp and Miller,2006). Finally, our data showing that somite rotation is substantially delayed in intgα5 MO injected embryos demonstrates that integrin α5β1 plays a critical role in regulating rotation.
It is probable that cells utilize a combination of cell–cell and cell–matrix interactions to generate the traction necessary for cell turning. The robust fibronectin matrix that surrounds developing somites could provide traction for cells at the periphery, while cell–cell interactions might play a larger role at the interior of the somite, where fibronectin matrix is comparatively sparse. In this model, interactions of peripheral cells with fibronectin would provide a large share of the traction necessary to drive cell rotation. Loss of this traction would result in the delayed cell rotation that we observe in intgα5 MO treated embryos, while the remaining forces from cell–cell interactions could account for the eventual completion of cell rotation.
Boundary Formation in Anterior and Posterior Somites
In addition to a delay in somite rotation, we also observe a delayed accumulation of both integrin β1 and fibronectin at developing somite boundaries in intgα5 MO treated embryos. As cells eventually attain their final, anterior/posterior orientation, some irregular boundary formation is observed in the form of disjointed integrin β1 and fibronectin accumulation. We previously observed that, in anterior somites, fibronectin accumulation at the presumptive somite boundary occasionally preceded somite turning (Kragtorp and Miller,2006). This would suggest that, in anterior somites, deposition the fibronectin matrix may serve an initial instructional role in boundary formation by creating a “track” for rotating cells to follow. We hypothesize that in the absence of this instructional cue, rotating cells lose their coordination and final cell position becomes more random. At the end of rotation, end-to-end cell contacts would determine the position of boundaries, leading to fragmented somite boundaries.
We report here that formation of posterior somites is less affected by the knockdown of integrin α5. This is reminiscent of reports of zebrafish integrin α5 mutants, in which anterior but not posterior somites are affected. When combined with other deficiencies, such as fibronectin or Delta D mutants, zebrafish somite boundary formation fails all along the anteiror/posterior axis, indicating at least one level of redundancy. Indeed, loss of fibronectin commonly results in more severe phenotypes than loss of a single integrin subunit (Georges-Labouesse et al.,1996; Yang et al.,1999; Julich et al.,2005), the activity of which could be taken over by other integrins. This prompts us to hypothesize that such a redundancy is responsible for the decreased requirement for integrin α5 activity in the posterior somites of Xenopus.
There are several possible candidates for this redundancy. In mouse embryonic cells, the activity of integrin α5β1 can be replaced by integrin αvβ1 (Yang and Hynes,1996). As integrin αv is expressed ubiquitously in the early Xenopus embryo (Joos et al.,1998), it is conceivable that this subunit contributes to the partial recovery we observe. In contrast to anterior somites, we do not observe the accumulation of fibronectin prior to somite turning in more posterior somites (Kragtorp and Miller,2006). Thus, it is also possible that interactions with other extracellular matrix molecules could provide some redundancy in the development of posterior somites. Laminin accumulates around somites and at somite boundaries (Krotoski and Bronner-Fraser,1990), and integrin α7β1, which binds laminin, is expressed in the myotomes of both chicken (Kil and Bronner-Fraser,1996) and mouse (Song et al.,1992,1993). The expression pattern of integrin α7 in Xenopus has not been investigated, however, and thus its potential role in somitogenesis remains unclear. Finally, it is possible that cadherin-based cell–cell interactions could provide a redundant mechanism for posterior somite boundary formation. In Xenopus, inhibition of cadherin I in somites resulted in tadpoles with a distinctive “kinked” tail (Giacomello et al.,2002), which could point to an increased requirement for cadherin activity in tail somites. A combinatorial approach, similar to that used in zebrafish, would be useful in Xenopus in order to create a clearer picture of the redundancies operating in the complex morphogenesis of somites.
Research conducted in mouse, quail, and zebrafish all point to a conserved role for integrin α5β1 in somite boundary formation, although the severity of the defects observed in the absence of integrin α5, integrin β1, or fibronectin, varies between species (Drake et al.,1992; Yang et al.,1993,1999; Koshida et al.,2005). These species all use epithelialization to delineate the somite boundary (Stern and Keynes,1987; Ostrovsky et al.,1988; Stickney et al.,2000), a trait not shared by Xenopus. That the requirement for integrin–fibronectin interactions in somitogenesis is conserved may indicate that their role is, at least in anterior somites, instructive: physical delineation of the somite boundary by fibronectin matrix deposition provides a cue or scaffold that assists presomitic cells in carrying out the morphogenetic program specific to that species. It appears that, despite evolving very different morphogenetic movements, each species has continued to use the same basic tools to accomplish the cell rearrangements and boundary specification necessary for somite formation.
The cDNA sequence for integrin α5 (Joos et al.,1995; GenBank accession no. U12683) was compared with the genomic sequences for Xenopus tropicalis, mouse, and human integrin α5 in order to identify sites of putative introns in Xenopus laevis integrin α5. Primers surrounding a likely site for intron 6 were created and used to amplify and sequence this intron from genomic DNA. A morpholino antisense sequence targeting the donor splice site prior to intron 6 was identified and obtained from Gene Tools (Philomath, OR): 5′ – ATATTTCTTACCTTGCCAGAAGTAG – 3′. Morpholino stocks were handled as recommended by Gene Tools.
Embryos and Microinjections
Xenopus laevis embryos were obtained by fertilization of eggs collected from females injected with human chorionic gonadotrophin (Sigma). Eggs were dejellied in 2% cysteine, cultured in 0.33× MMR (Sive,2000), and staged according to Nieuwkoop and Faber (1994). Embryos were injected in 4% Ficoll in 0.33× MMR. Morpholino stocks were diluted in sterile water to deliver 50 ng per injection. Injections were unilateral for assessment of gross morphology and immunofluorescence experiments, with the uninjected side serving as an internal control. Bilateral injections were used for mRNA collection. For rescue experiments capped mRNA was synthesized using the SP6 mMessage Machine kit (Ambion) and combined with morpholino stocks to achieve a final dose of 500 pg mRNA per injection.
For most immunofluorescence experiments, embryos were fixed in Dents fixative overnight at 4°C. For imaging of fibronectin, embryos were fixed in 2% trichloroacetic acid in PBS overnight at 4°C. Immunostaining was performed with the following antibodies: rabbit polyclonal anti-GFP (Santa Cruz Biotechnology); mouse monoclonals anti-fibronectin (Ramos and DeSimone,1996), anti-MHC (Developmental Studies Hybridoma Bank) and anti-integrin β1 (8C8; DSHB). Staining was visualized using Alexa568-, Alexa488-(Molecular Probes), Cy2-, or Cy5-conjugated (Jackson ImmunoResearch) secondary antibodies. For imaging of cross-sections, immunostained embryos were incubated overnight in PBST, and slices were cut with a surgical scalpel. All samples were dehydrated, cleared using Murray's clear (Sive,2000), and mounted in Sylgard (Dow Corning) wells. Images were captured using a Zeiss spinning disc microscope and merged images were produced using Adobe Photoshop.
In Situ Hybridization
In situ hybridization was carried out as described (Harland,1991). Digoxygenin-labeled MyoD (Hopwood et al.,1992) and PAPC (Kim et al.,2000) probes were synthesized using a MAXIScript kit (Ambion). Probes were detected by alkaline phosphatase-conjugated anti-digoxygenin Fab (Roche) using BM Purple substrate (Boehringer Mannheinm). Embryos were incubated for 1 hr in bleaching solution prior to imaging (Sive,2000).
RT-PCR was performed using total RNA isolated from dorsal explants of stage-22 embryos containing notochord, neural tube, and somitic tissue. cDNA was prepared using SuperScript III reverse transcriptase (Invitrogen). Forward and reverse primers used for RT-PCR were located in exons 4 and 7, respectively, of integrin α5: forward: 5′- TGATGCTGCTGGTCAAGGG – 3′, reverse: 5′ – ATGAAGCCGCCTGTCGTGTC – 3′.
We thank Richard Harland and Ralph Rupp for the gift of the MyoD in situ probe and Eddy DeRobertis for the gift of the PAPC in situ probe. We would especially like to thank Doug DeSimone for helpful discussions and advice on the manuscript as well as the kind gift of the fibronectin 4H2 antibody and the Xenopus laevis intgα5 plasmid. Funding was provided by NIH Kirschstein-NRSA postdoctoral fellowship AR051309-02 to K.A.K. and NSF grant 0315767 to J.R.M.