Segmental Assembly of Fibronectin Matrix Requires rap1b and integrin α5


  • Simone Lackner,

    1. Department of Molecular, Cellular and Developmental Biology, Yale University, New Haven, Connecticut
    Current affiliation:
    1. Vision to Action Laboratory, Champalimaud Neuroscience Programme at the Champalimaud Centre for the Unknown, Lisbon, Portugal
    Search for more papers by this author
    • Simone Lackner and Jamie Schwendinger-Schreck contributed equally to this article.

  • Jamie Schwendinger-Schreck,

    1. Department of Molecular, Cellular and Developmental Biology, Yale University, New Haven, Connecticut
    Search for more papers by this author
    • Simone Lackner and Jamie Schwendinger-Schreck contributed equally to this article.

  • Dörthe Jülich,

    1. Department of Molecular, Cellular and Developmental Biology, Yale University, New Haven, Connecticut
    Search for more papers by this author
  • Scott A. Holley

    Corresponding author
    • Department of Molecular, Cellular and Developmental Biology, Yale University, New Haven, Connecticut
    Search for more papers by this author

Correspondence to: Scott Holley, Department of Molecular, Cellular and Developmental Biology, Yale University, New Haven, CT 06511. E-mail:


Background: During segmentation of the zebrafish embryo, inside-out signaling activates Integrin α5, which is necessary for somite border morphogenesis. The direct activator of Integrin α5 during this process is unknown. One candidate is Rap1b, a small monomeric GTPase implicated in Integrin activation in the immune system. Results: Knockdown of rap1b, or overexpression of a dominant negative rap1b, causes a mild axis elongation defect in zebrafish. However, disruption of rap1b function in integrin α5−/− mutants results in a strong reduction in Fibronectin (FN) matrix assembly in the paraxial mesoderm and a failure in somite border morphogenesis along the entire anterior-posterior axis. Somite patterning appears unaffected, as her1 oscillations are maintained in single and double morphants/mutants, but somite polarity is gradually lost in itgα5−/−; rap1b MO embryos. Conclusions: In itgα5/ mutants, rap1b is required for proper somite border morphogenesis in zebrafish. The loss of somite borders is not a result of aberrant segmental patterning. Rather, somite boundary formation initiates but is not completed, due to the failure to assemble FN matrix along the nascent boundary. We propose a model in which Rap1b activates Integrin/Fibronectin receptors as part of an “inside-out” signaling pathway that promotes Integrin binding to FN, FN matrix assembly, and subsequent stabilization of morphological somite boundaries. Developmental Dynamics 242:122–131, 2013. © 2012 Wiley Periodicals, Inc.


The interaction between cells and the surrounding extracellular matrix (ECM) plays an important role in morphogenesis and cell migration. During embryonic development, the ECM is integral to various morphogenic movements, such as gastrulation, the migration of neural crest cells, and branching of the mammary epithelium (Rozario and DeSimone, 2010). The primary cell surface receptors mediating cell-ECM adhesion are the Integrins (Hynes, 2002).

Integrins are heterodimeric transmembrane proteins that consist of α and β subunits. Each subunit has a large extracellular domain, a single-pass transmembrane domain, and a short cytoplasmic tail. Integrins serve to physically link the extracellular matrix to the cytoskeleton, as well as respond to intracellular and extracellular cues. Integrin activation in response to intracellular cues is known as “inside-out” signaling, while transmission of signals from the ECM to the nucleus or cytoskeleton is called “outside-in” signaling (Hynes, 2002). Affinity for ligand depends on the conformation of Integrin dimers, which can be either bent and inactive with low affinity for ligand, or extended and active with a high affinity for ligand (Askari et al., 2009). Integrin α5 and αV are the α subunits of the primary receptors for Fibronectin (FN), a glycoprotein that forms linear and branched fibrous networks that surround cells and are necessary for embryonic development (Schwarzbauer and DeSimone, 2011).

Here, we focus on the role of Integrin-FN interactions during somite morphogenesis. Somitogenesis is the process by which vertebrate embryos become segmented along the anterior-posterior axis. Segments, or somites, bud off from the anterior of the presomitic mesoderm (PSM) concomitant with axis elongation. The PSM is made up of bilateral columns of mesenchymal cells on either side of the notochord. Somites form in pairs flanking the midline in an anterior-to-posterior sequence. The generation of somites requires three basic steps: establishing the segmental pattern within the mesoderm, generating anterior-posterior (AP) polarity within the nascent somite, and creating a morphological boundary via a mesenchymal-to-epithelial transition. The segmental pattern generated in the PSM is the result of interactions between the segmentation clock, which is largely made up of oscillating Notch pathway genes, and the wavefront, a gradient of Wnt and Fgf signaling that links posterior growth of the PSM to somite border formation (Pourquie, 2011; Oates et al., 2012). This process then sets up AP polarity in the nascent somites. The establishment of segment polarity culminates in opposing domains of ephA4 and ephrinB2a expression in the anterior and posterior halves of nascent somites, respectively (Barrios et al., 2003). EphA4 and EphrinB2 are a membrane-bound receptor and ligand pair that can signal bidirectionally (Pasquale, 2008). At the boundary between prospective somites, where EphA4 and EphrinB2a domains meet, an intersomitic boundary, or gap, forms (Watanabe and Takahashi, 2010). The juxtaposition of Eph- and Ephrin-expressing cells appears sufficient to promote border formation, as transplantation of ephA4-overexpressing cells in a background of ephrinB2-positive cells results in formation of ectopic morphological borders (Durbin et al., 2000; Barrios et al., 2003). Integrin α5 (Itgα5), which is necessary for morphogenesis of anterior somite boundaries in zebrafish, clusters at prospective somite boundaries, followed by FN matrix accumulation (Jülich et al., 2005; Koshida et al., 2005). Initiation of Itgα5 clustering is independent of ligand binding, but requires ephrinB2a, and reverse signaling through EphrinB2a is sufficient to promote clustering and FN matrix formation (Jülich et al., 2009). These results suggest that inside-out signaling initiates Integrin activation and FN matrix assembly during somitogenesis. Outside-in signaling likely also plays a role during segmentation. Outside-in signaling is initiated by Integrin binding to Fibronectin and results in increased cellular concentration of lipid second messengers, alterations to the actin cytoskeleton, and various changes in signaling pathways and gene expression (Legate et al., 2009). During zebrafish somite border formation, phosphorylated Focal Adhesion Kinase, an indicator of outside-in signaling, is localized to the somite boundary, as are focal adhesion proteins such as Paxillin (Henry et al., 2001; Crawford et al., 2003). As the boundary cells undergo a mesenchymal-to-epithelial transition, the resulting somite is made up of a core of mesenchymal cells surrounded by epithelial cells (Holley, 2007; Takahashi and Sato, 2008).

FN is required for somite boundary morphogenesis. Zebrafish have two fn genes, fn1 and fn1b, both of which are expressed in the paraxial mesoderm (Thisse et al., 2001; Trinh and Stainier, 2004; Jülich et al., 2005; Koshida et al., 2005). Mutation or knockdown of fn1 results in a weak border defect of anterior somites (Jülich et al., 2005; Koshida et al., 2005), while loss of fn1b leads to tail extension defects and disruption of posterior somite borders. Knockdown of both fn1 and fn1b exacerbates the axis extension defect and disrupts somite border formation along the entire AP axis (Jülich et al., 2005). Proper FN matrix assembly is also required for somitogenesis in chick (Rifes et al., 2007). In addition to impaired somite border formation, disruption of FN fibrillogenesis in cultured chick embryos perturbs epithelization of formed somites (Martins et al., 2009). In mouse, deletion of FN results in embryonic lethality, a truncated AP axis, and complete lack of somites (George et al., 1993). Loss of the primary FN receptor, Itgα5, partially recapitulates the mouse FN null phenotype, though border defects are restricted to posterior somites (Yang et al., 1993; Goh et al., 1997). In anterior somites, loss of Itgα5 is compensated for by ItgαV (Yang et al., 1999), though ItgαV null embryos do not have a somite phenotype (Bader et al., 1998). Itgα5 binds FN through the RGD motif. Mutation of RGD to an inactive RGE sequence phenocopies the Itgα5 mutant, though FN matrix assembly persists through the action of ItgαVβ3 (Takahashi et al., 2007).

While Eph/Ephrin signaling appears to be upstream of Integrin-mediated FN assembly, the direct activator remains unknown. One candidate for the inside-out activation of Integrins during somitogenesis is Rap1. Rap1 is a small monomeric GTPase protein that has been implicated in a number of vital cell processes, including changes in cell morphology (O'Keefe et al., 2012), promotion of VEGF signaling during angiogenesis (Lakshmikanthan et al., 2011), and phagocytosis by macrophage cells (Caron et al., 2000). Deletion of rap1 in Drosophila results in failure of the ventral furrow to close due to abnormal cell shape and cell migration (Asha et al., 1999). Rap1 may also play a role in cancer metastasis, as altered Rap1 activity affects rates of carcinoma cell invasion on Fibronectin or Collagen matrix in culture (Kim et al., 2012), and several studies have demonstrated a link between aberrant Rap1 levels and metastasis in multiple cancer types (Bailey et al., 2009; Ricono et al., 2009; McSherry et al., 2011; Huang et al., 2012). Rap1 can exist as either a GDP-bound, inactive, or GTP-bound, active form. GTPase-activating proteins (GAPs) and guanine exchange factors (GEFs) accelerate the transition between these two forms (Bourne et al., 1991). Rap1 responds to multiple different activating cues and utilizes various effectors, all of which function in a context-specific manner (Gloerich and Bos, 2011). Here, we examine the role Rap1 plays in Integrin activation during zebrafish somitogenesis. We use genetic analysis to probe the in vivo relationship between itgα5, ephrinB2a, and rap1b function in vertebrate segmentation.


Knockdown of the Small GTPase rap1b Results in a Mild Tail Elongation Defect But No Disruption of Somite Borders

Zebrafish have two homologs of rap1, rap1a and rap1b, with rap1b being the most abundant isoform during early zebrafish development (Tsai et al., 2007). In Danio rerio, reduction of rap1b results in impaired angiogenesis by 28 hr post fertilization (hpf) and reduced vascular integrity after 4 days (Gore et al., 2008; Lakshmikanthan et al., 2011). Knockdown of rap1b also leads to axis extension defects readily apparent in embryos 2–3 days old (Tsai et al., 2007). However, the effects of rap1b earlier in development have not been documented.

rap1b is expressed weakly throughout the embryo during multiple stages of development, with stronger expression in the notochord during early somitogenesis (Thisse et al., 2001; Gore et al., 2008). We cloned rap1b from zebrafish cDNA and confirmed its ubiquitous basal expression during gastrulation and early segmentation using in situ hybridization (data not shown). We then analyzed rap1b function by knocking down Rap1b protein levels with both translation-blocking (MO1) and splice-blocking (MO2) morpholinos. We performed a dose response to determine the lowest concentration of morpholino necessary to provide a strong, highly penetrant phenotype. Knockdown of rap1b resulted in a mild AP axis truncation and tail elongation defect. Embryos are slightly shorter than wild-type siblings at both the 12–15-somite stage (Fig. 1A,B) and 23–25-somite stage (data not shown), though the severity of the axis defect was variable. In addition to the elongation defect, a thickened layer of cells formed at the dorsal side of the tail bud, resulting in a pronounced bump. No effect on somite border formation was observed, though somite size and shape were mildly affected. The results of MO1 and MO2 injection were very similar, though the concentration of MO2 required to recapitulate the phenotype seen with MO1 was over twice that of MO1. For simplicity, only images for MO1 are shown. We also perturbed Rap1b function by injection of dominant-negative rap1b mRNA (rap1b DN). This construct was generated using PCR-mediated site-directed mutagenesis to alter conserved residue 17 from serine to asparagine (Fig. 2A) (Gabig et al., 1995). Gross morphology of embryos injected with rap1b DN was similar to that of rap1b morphants, though the bump over the tail bud was more pronounced (Fig. 1C). Gene specificity of the splice-blocking morpholino was verified using RT-PCR (Fig. 2B) and by sequencing the improperly spliced product. The altered splice-product lacks the second exon and is 69 bp smaller than wild-type rap1b cDNA, resulting in a deletion of 23 amino acids from the predicted protein. The second exon of rap1b contains the highly conserved sequence of the G-2 domain, which is involved in GTP binding (Bourne et al., 1991). We note that knockdown of rap1b by MO2 is not complete, as properly spliced mRNA product persists. Lastly, co-injection of rap1b MO1 and MO2 at reduced concentrations or co-injection of rap1b MO1 and rap1b DN mRNA recapitulated the axis extension and tail elongation phenotype of the single morphant without disruption of somite borders (Fig. 2C).

Figure 1.

Loss of rap1b synergizes with itgα5, resulting in a disruption of all somite borders and abrogation of Fibronectin matrix assembly. A–F: Lateral view DIC images showing clear somite borders for (A) wild-type, (B) rap1b MO1-injected (8 experiments, 83.7% affected, n = 341), and (C) rap1b DN mRNA-injected (3 experiments, 100% affected, n = 177) embryos. For synergy experiments, rap1b MO1 or rap1b DN was injected into the progeny of heterozygous itgα5+/ crosses. Therefore, 100% penetrance would result in approximately 25% of embryos exhibiting the phenotype. D: Anterior somite borders are disrupted in the itgα5−/− mutant, while (E) itgα5−/−; rap1b MO1 (9 experiments, 27.4% affected, n = 865) and (F) itgα5−/−; rap1b DN (6 experiments, 22.3% affected, n = 455) embryos lack somite borders along the entire AP axis. (G– L) Dorsal view and (G’–L’) higher magnification dorsal view of FN localization showing clear FN matrix surrounding somites in (G–G’) wild-type, (H–H’) rap1b MO1, and (I–I’) rap1b DN embryos. J,J’: FN matrix surrounds only posterior somites in itgα5−/− mutants, but few fibrils are seen within the paraxial mesoderm of either (K,K’) itgα5−/−; rap1b MO1 or (L,L’) itgα5−/−; rap1b DN embryos. For all images, anterior is to the left and embryos are at the 12–15-somite stage.

Figure 2.

Morpholino knockdown of rap1b is specific. A: Schematic of the rap1b gene showing exons in blue, introns in black, and the approximate position of morpholino binding sites in green. The S17N mutation generating dominant-negative rap1b is indicated with a red star. B: Agarose gel showing 100-bp ladder with 500- and 600-bp bands marked, PCR product of wild-type rap1b cDNA, and PCR product of wild-type and altered splice product generated via splice-blocking morpholino (MO2) for rap1b. As confirmed by sequencing, the altered splice-product lacks the second exon and is 69 bp smaller than the wild-type rap1b cDNA. This disruption of splicing results in deletion of 23 amino acids from the predicted protein. C: Lateral view of DIC image for 12–15-somite-stage embryo injected with both rap1b MO1 and rap1b DN mRNA (3 experiments). For rap1b MO1+MO2 injections, 71.4% were affected (3 experiments, n = 248) (data not shown). Anterior is left.

Rap1 has been shown to regulate Integrin β1, β2, and β3 affinity and clustering, likely by recruiting Talin to Integrin complexes (Boettner and Van Aelst, 2009). Previous work found that clustering of Itgα5 precedes FN matrix assembly along zebrafish somite boundaries (Jülich et al., 2009), and that formation of the FN matrix along anterior somite borders is dependent on itgα5 (Jülich et al., 2005; Koshida et al., 2005). Loss of both itgα5 and itgαV in mouse synergizes, generating a phenotype similar to FN null mutants (Yang et al., 1999). We analyzed FN matrix assembly along somite borders using immunohistochemistry, and found that FN matrix forms normally around all somites of both rap1b MO and rap1b DN-injected embryos (Fig. 1G–I).

Knockdown of rap1b Synergizes With Loss of itgα5

Loss of itgα5 results in specific disruption of anterior somite borders in zebrafish (Jülich et al., 2005; Koshida et al., 2005), suggesting redundancy of Integrin activity during posterior somite morphogenesis, possibly due to ItgαV. We were interested in whether Rap1b plays a role in Integrin activation during somitogenesis. Therefore, we perturbed rap1b function by either morpholino knockdown or injection of rap1b DN in integrin α5/ mutants. As the maternal contribution of itgα5 is maintained, we note that the somite phenotype of itgα5/ mutants is less severe than that of maternal-zygotic itgα5/ embryos (Jülich et al., 2005, 2009). Concomitant loss of itgα5 and rap1b resulted in a synergistic effect on somite border formation with only the most recently formed 2–4 somite borders being transiently visible. These embryos also exhibited a strong enhancement of the axis elongation defect (Fig. 1D–F). Formation of FN matrix surrounding the somites was completely disrupted in itgα5/; rap1b MO and itgα5/; rap1b DN embryos, with FN matrix apparent only surrounding the paraxial mesoderm and notochord (Fig. 1J–L). Though morphological boundaries were sometimes seen for the most recently formed somites, no segmental FN matrix was observed in itgα5/; rap1b MO or itgα5/; rap1b DN embryos.

Loss of somite borders in itgα5/; rap1b MO and itgα5/; rap1b DN Embryos Is Not Due to Failure to Establish the Segmental Pattern

Before morphological somite boundaries form, a segmental pre-pattern isestablished. Proper patterning depends on oscillations of hairy/enhancer of split-related (her) genes that are visible as stripes of mRNA expression in the unsegmented paraxial mesoderm (Holley et al., 2000). We examined her1 expression patterns in rap1b morphants and itgα5/; rap1b MO embryos, as well as in rap1b DN and itgα5/; rap1b DN embryos. The presence of 2–4 transient somite borders in the posterior of itgα5/; rap1b MO embryos suggests that somite border morphogenesis, rather than somite patterning, is disrupted after rap1b perturbation in itgα5/ mutants. In accordance with this hypothesis, oscillatory her1 expression is apparent in itgα5/; rap1b MO and itgα5/; rap1b DN fish, though her1 stripe boundaries are slightly blurred. Approximately half of itgα5/; rap1b DN embryos have clear her1 stripes, and half have compressed or blurred stripes, whereas the vast majority of itgα5/; rap1b MO embryos show clear stripes of her1 expression. Stripes of her1 are also present in rap1b MO and rap1b DN embryos. (Fig. 3A–E).

Figure 3.

Loss of somite borders is likely not due to patterning defects in the PSM. A–E: Double in situ hybridization for myoD (pink) and her1 (blue) for (A) wild-type, (B) rap1b MO1, (C) rap1b DN mRNA, (D) itgα5−/−; rap1b MO1, and (E) itgα5−/−; rap1b DN 12–15-somite-stage embryos. Stripes of her1 expression persist in all backgrounds. In contrast, segmental myoD expression is blurred in rap1b MO1 and rap1b DN embryos, and completely lost in itgα5−/−; rap1b MO or itgα5−/−; rap1b DN. F–H: Double in situ hybridization for myoD (pink) and mesp-b (blue) in (F) wild-type, (G) rap1b MO1, and (H) itgα5−/−; rap1b MO1 embryos at 12–15-somite stage. Stripes of mesp-b can be seen in both rap1b and itgα5−/−; rap1b morphants. I–K: In situ hybridization for ripply1 in (I) wild-type, (J) rap1b MO1, and (K) itgα5−/−; rap1b MO1 12–15-somite stage embryos. Segmental ripply1 expression is disrupted only in itgα5−/−; rap1b MO1 embryos. For all images, anterior is up.

Somite Polarity Is Gradually Lost in itgα5/; rap1b MO Embryos

We next examined the establishment of anterior-posterior polarity within each somite. Segment polarity is manifest in the segmental expression of genes such as mesp-b, which is present in the future anterior half of forming somites (Fig. 3F), and ripply1, expressed in the anterior PSM and anterior half of recently formed somites (Fig. 3I) (Sawada et al., 2000; Kawamura et al., 2005). The myogenic marker myoD is expressed in the posterior half of all formed somites (Fig. 3A, F) (Weinberg et al., 1996). EphrinB2a is primarily localized to the posterior border of each somite (Fig. 4A–A’’) (Zhang et al., 2008). We analyzed mesp-b, myoD, and ripply1 expression and EphrinB2a localization in both rap1b morphants and itgα5/; rap1b MO embryos.

Figure 4.

Localization of EphrinB2a suggests gradual loss of segment polarity after loss of both rap1b and itgα5 function. Dorsal views (A–C:) and higher magnification dorsal views (A’–C’’) of EphrinB2a protein localized to the posterior halves of somites in (A–A’’) wild-type and (B–B’’) rap1b MO1 embryos. Localized EphrinB2a is lost in anterior somites of (C–C’’) itgα5−/−; rap1b double morphants, but is transiently present in more recently formed posterior somites. Shown are 12–15-somite-stage zebrafish embryos. In A–C, anterior is up, while in A’–C’’ anterior is left.

Stripes of mesp-b expression persist in nascent somites of both rap1b and itgα5/; rap1b morphants (Fig. 3F–H). Consistent with our morphological analysis, embryos with disrupted rap1b function have segmental myoD expression, though the pattern is much more blurred than wild type. In contrast, segmental expression of myoD is completely lost in mature somites after perturbation of rap1b in the itgα5/ background. Instead, myoD expression is blotchy along the AP axis (Fig. 3A–H). In itgα5/ mutants, only anterior myoD expression is fused (Jülich et al., 2005; Koshida et al., 2005). Expression of ripply1 is segmental in rap1b morphants and greatly disrupted in itgα5/; rap1b MO embryos, so that only 2–3 blurry stripes can be seen in some embryos, and expression in nascent somites consists of a single domain of expression, rather than the distinct stripes seen in wild type embryos or single rap1b morphants (Fig. 3I–K).

We next examined localization of EphrinB2a. Again, segment polarity was retained in rap1b morphants (Fig. 4B). Polarized localization of EphrinB2a was observed in both wild type and rap1b morphants, clustered at the posterior boundaries of both anterior and posterior somites (Fig. 4A–B’’). In contrast, although some EphrinB2a appears localized to the presumptive somite borders in more posterior somites, this pattern is lost in older somites of itgα5/; rap1b MO embryos (Fig. 4C–C’’). The gradual loss of EphrinB2a localization suggests that, while forming somites are properly patterned, the AP polarity cannot be maintained after the failure of boundary morphogenesis.

No Genetic Interaction Is Observed Between rap1b and ephrinB2a

In zebrafish, Eph/Ephrin signaling can initiate morphological border formation (Barrios et al., 2003). ephrinB2a is also necessary for ligand-independent Itgα5 clustering, and EphrinB2a reverse signaling is sufficient to activate clustering of Itgα5 and subsequent FN matrix formation (Jülich et al., 2009). Further, knockdown of ephrinB2a eliminates detectable EphrinB2a protein but does not lead to a somite phenotype. However, knockdown of ephrinB2a in itgα5/ mutants disrupts morphogenesis of all somite borders (Koshida et al., 2005; Jülich et al., 2009). We repeated the knockdown of ephrinB2a in itgα5/ mutants, and in agreement with previously reported results, somite borders were disrupted along the entire AP axis in itgα5/; ephrinB2a MO embryos and no FN matrix was observed within the paraxial mesoderm (Fig. 5A, D).

Figure 5.

Somite morphogenesis is sensitive to genetic interaction between itgα5 and ephrinB2a, but not rap1b and ephrinB2a. A–C: Lateral view DIC images showing complete disruption of somite borders in (A) itgα5−/−; ephrinB2a MO embryos (7 experiments, 27.9% affected compared to an expected 25% for homozygotes, n = 605) and proper maintenance of somite borders in (B) ephrinB2a MO + rap1b MO1 (5 experiments, no synergy, n = 455) and (C) ephrinB2a MO + rap1b DN embryos (3 experiments, no synergy). Dorsal (D–F:) and higher magnification dorsal (D’–F’) views of Fibronectin localization for (D–D’) itgα5−/−; ephrinB2a MO, (E–E’) ephrinB2a MO + rap1b MO1, and (F–F’) ephrinB2a MO + rap1b DN embryos at the 12–15-somite stage. FN matrix formation is lost only in itgα5−/−; ephrinB2a MO embryos. For all images, anterior is left.

We were interested in whether rap1b serves as an effector of Eph/Ephrin signaling to activate Integrins at nascent somite borders, as the itgα5/; ephrinB2a MO phenotype is strikingly similar to the itgα5/; rap1b MO embryos described above. Therefore, we combined knockdown of ephrinB2a with perturbation of rap1b, either by morpholino or rap1b DN injection. Surprisingly, though the axis extension defect was exacerbated in double morphants, no synergistic effect on somite borders was observed (Fig. 5B,C). Indeed, FN accumulated around somites along the entire AP axis in a manner similar to that observed in wild type (Fig. 5E,F).


We report a genetic interaction between rap1b with itgα5 during somite border morphogenesis that implicates Rap1b in promoting FN matrix assembly. We knocked down Rap1b, a small monomeric GTPase, that has been shown to activate Integrins by affecting clustering and ligand affinity (Boettner and Van Aelst, 2009). In wild-type embryos, reduction of Rap1b led to a mild tail elongation defect, but did not perturb somite boundary formation or maintenance. In contrast, injection of rap1b morpholino or rap1b DN into itgα5/ mutants caused a disruption of somite borders along the entire AP axis. We propose that Rap1b activates Integrins during somitogenesis, leading to FN matrix assembly and stabilization of morphological boundaries.

We hypothesize that Rap1b activates multiple FN-binding Integrins. Though Itgα5 is only required in anterior somites, Itgα5-GFP clustering occurs at somite boundaries along the entire AP axis. Further, Itgα5-GFP is still evident at myotome borders at 24 hr post-fertilization (unpublished observations). The increase in severity of phenotype for itgα5/; rap1b MO embryos exceeds the phenotype of MZitgα5/ mutants, which have defects restricted to the anterior 〜8 somite boundaries. If Rap1b activated only Itgα5β1, the double itgα5/; rap1b morphant should more closely resemble the MZitgα5/ phenotype. Therefore, these data suggest that Rap1b is at least responsible for activation of an Integrin that compensates for loss of itgα5 in posterior somites, and likely also activates Itgα5 itself.

ItgαV is a good candidate to function in parallel with Itgα5 in posterior somites. Both α5 and αV form heterodimers with β Integrins that bind the RGD motif in FN (Yokoyama et al., 1999; García et al., 2002). ItgαV can also promote FN fibrillogenesis via binding the GNGRG motif in FN-I repeat 5 (Takahashi et al., 2007). Zebrafish itgαV mRNA is maternally deposited and is broadly expressed throughout gastrulation and in the notochord during segmentation (Ablooglu et al., 2007, 2010). Knockdown of itgαV perturbs formation of Kupfer's vesicle leading, to a defect in establishing left-right asymmetry in zebrafish (Ablooglu et al., 2010). Similarly, itgα5 is required for establishment of left-right asymmetry in both mice and zebrafish (Pulina et al., 2011). In mice, Itgα5 and αV cooperate to promote FN assembly in the mesoderm and endothelial cells (Yang et al., 1999; van der Flier et al., 2010). As in zebrafish, loss of Itgα5 in mice affects segmentation along a portion of the AP axis, though the effects are seen in posterior, rather than anterior, somites (Yang et al., 1993; Goh et al., 1997). Loss of both α5 and αV in mouse synergizes, leading to earlier lethality and reduced levels of FN matrix (Yang et al., 1999). We hypothesize that a similar redundancy operates in zebrafish, with itgα5 required for anterior somite borders, and both itgα5 and αV functioning in more posterior somites.

Rap1 has been shown to activate β1, β2, and β3 Integrins (Bos, 2005). The synergy seen between itgα5 and rap1b is likely due to Rap1b activating an Integrin that functions in parallel to itgα5. We also speculate that Rap1b activates Itgα5β1. However, if the latter hypothesis is true, what explains the lack of somite border defects in rap1b MO or rap1b DN-injected embryos? One possible explanation is that there is incomplete disruption of Rap1b function; however, the absence of a somite border phenotype in either rap1b MO1 + MO2 or rap1b MO1 + rap1b DN embryos would suggest this is not the case. Thus, we favor the explanation that there is further redundancy inthe system, such as Rap1b-independentactivation of Integrins along the nascent somite border.

Despite the loss of segment borders, segmental patterning was not affected in itgα5/; rap1b MO embryos, as her1 oscillations persisted and EphrinB2a was properly localized in nascent somites. Initial somite formation proceeded normally, as segmental mesp-b expression was present and morphological boundaries could be seen surrounding the most recently formed somites. However, the morphological borders were no longer visible in the more anterior paraxial mesoderm. In addition, EphrinB2a localization was gradually lost in mature somites, segmental expression of myoD was rapidly lost, and no FN matrix could be seen around either anterior or more recently formed somites. The gradual loss of morphological boundaries and mislocalization of EphrinB2a suggests that border formation is initiated but eventually fails. This phenotype is similar to that seen in itgα5/; fn MO or itgα5/; ephrinb2a MO embryos, where somite boundaries appear but fail to achieve stable mesenchymal-to-epithelial transition (Jülich et al., 2005; Koshida et al., 2005).

EphrinB2a reverse signaling promotes Itgα5 clustering and subsequent FN matrix assembly (Jülich et al., 2009), and we hypothesized that Rap1b might provide a link between Ephrin signaling and Integrin activation. Our analysis indeed suggests that Rap1b functions downstream of the establishment of the segmental pattern of EphrinB2a as this striped pattern is present in theabsence of rap1b function. Since loss of ephrinB2a synergizes with itgα5/ (Koshida et al., 2005; Jülich et al., 2009), we examined whether loss of rap1b causes an exacerbation of the ephrinB2a knockdown phenotype. However, we observed no synergy in ephrinB2a; rap1b double morphants. While knockdown of ephrinB2a does eliminate detectable EphrinB2a protein in injected embryos, loss of ephrinb2a alone does not cause a somite defect, suggesting that there is likely some redundancy with other ephrins expressed in nascent somites (Barrios et al., 2003; Jülich et al., 2009). Given that we also suspect some redundancy between rap1b and other activators of Integrins, single knockdown of both ephrinb2a and inhibition of rap1b should be considered hypomorphic conditions. As combining two hypomorphs within a “genetic pathway” is typically expected to produce a stronger phenotype, the lack of genetic interaction between rap1b and ephrinb2a suggests that Rap1b does not link Eph/Ephrin signaling to Integrin activation. However, based on the data presented here, one cannot firmly conclude that there is no connection between Eph/Ephrin signaling and Rap1b during zebrafish somite morphogenesis.


Zebrafish Strains

Zebrafish were maintained using standard protocols (Nüsslein-Volhard and Dahm, 2002) as approved by Yale IACUC. The wild-type strain used was Tübingen, and mutant alleles used were integrin α5: bfethl30 and bfetig453.

mRNA and Morpholino Microinjections

Injections were performed at the 1-cell stage according to standard protocols (Nüsslein-Volhard and Dahm, 2002). Full-length coding sequence of rap1b was isolated from cDNA of 12–15-somite-stage zebrafish embryos and cloned into pCS2+. rap1b DN mRNA lacks the rap1b MO1 target site and was generated using site-directed mutagenesis and transcribed in vitro (SP6 mMessage Machine Kit, Ambion, Austin, TX). Morpholinos targeting rap1b and ephrinB2a have been previously described (Koshida et al., 2005; Tsai et al., 2007). The sequence of these morpholinos are as follows: rap1b MO1: 5′ ACGCATTGTGCAGTGTGTCCGTTAA, rap1b MO2: 5′ CAATAGAAATGATGCAGAACTTGCC, and ephrinB2a MO: AATATCTCCACAAAGAGTCGCCCAT. The concentrations used and approximate amount injected are as follows: rap1b DN: 200 ng/μl (40 ng); rap1b MO1: 600 μM (120 μmol); rap1b MO2: 1.6 mM (320 μmol); ephrinB2a MO: 900 μM (180 μmol); except for rap1b MO1+MO2 experiments, where concentrations were 400 μM (80 μmol) and 800 μM (160 μmol), respectively.

In Situ Hybridization and Immunohistochemistry

In situ hybridization was performed using standard protocols. Riboprobes were generated from plasmid templates. Fibronectin and EphrinB2a localization was performed as previously described (Jülich et al., 2005; Zhang et al., 2008) using rabbit anti-human Fibronectin IgG (Sigma, St. Louis, MO) or goat anti-zebrafish EphrinB2a (R&D Systems, Minneapolis, MN). For in situ hybridization, myoD and her1 antisense probes were made as previously described (Müller et al., 1996; Weinberg et al., 1996). Antisense probes for mesp-b and ripply 1 were generated directly from RT-PCR products. Following the initial RT-PCR, T7 sites were incorporated for riboprobe synthesis in a second round of PCR (Zhang et al., 2008). Primer sequences are mesp-b (5′-ATGCAAACCTCAAGCAAGAAC-3′, 5′-TCTCCAGTAAGTCTGAGGAAC-3′, 5′-TCTCCAGTAAGTCTGAGGAACAATACGACTCACTATAG-3′) and ripply-1 (5′-ATGAATTCTGTGTGCTTTGCCA-3′, 5′-GTTGAAAGCTGTGAAGTGACT-3′, 5′-GTTGAAAGCTGTGAAGTGACTAATACGACTCACTATAG-3′). The first two primers for each gene were used in the initial qPCR while the first and third primers were used in the second round of PCR to add the T7 sequence. Three experiments were performed for her1/myoD analysis of rap1b DN and itgα5/; rap1b DN embryos. Four replicates were performed for rap1b MO1 embryos, and two replicates for itgα5/; rap1b MO1 embryos. Two replicates were performed for ripply1 analysis of all perturbations, except rap1b MO1 embryos, which had four replicates. Only rap1b MO1 and itgα5/; rap1b MO1 were analyzed for mesp-b expression, with three and two replicates, respectively. For immunohistochemistry showing FN localization: five replicates of itgα5/; rap1b MO1; three replicates of rap1b MO1; two replicates of rap1b DN, itgα5/; rap1b DN, itgα5/; ephrinB2a MO, ephrinB2a MO + rap1b MO1 and ephrinB2a MO + rap1b DN; and a single replicate of rap1b MO2 and itgα5/; rap1b MO2 perturbations were performed. Ephrin localization experiments were performed on three replicates of rap1b MO1; two replicates of itgα5/; rap1b MO1, itgα5/; ephrinB2a MO, and ephrinB2a MO+rap1b MO1; and a single replicate of itgα5/; rap1b DN. Representative images are shown.

Image Acquisition and Processing

In situ hybridization images were captured using a Zeiss (Thornwood, NY) Stemi SV6 dissecting scope and Leica FireCam DFC 300 software. Wide field DIC images were collected using a 20x objective on a Zeiss Axioskop mot 2 plus and Openlab (Perkin Elmer, Waltham, MA) software. Fluorescent confocal images were acquired using a 40× objective on a Zeiss LSM 510 and Zen2008 software. Images were processed using Adobe Photoshop CS3 and ImageJ. Figures were assembled using Adobe Illustrator CS3.


We thank Patrick McMillen for critical comments on the manuscript. This research was supported by a Research Scholar Grant from the American Cancer Society and a grant from the NSF (IOS-1051839) to S.A.H.