Extra-embryonic syndecan 2 regulates organ primordia migration and fibrillogenesis throughout the zebrafish embryo

The heart and gut develop from bilateral precursors located in the lateral plate mesoderm (Brand, 2003; McFadden and Olson, 2002; Roberts, 2000; Zaffran and Frasch, 2002). These precursors migrate from the lateral edges of the embryo to the midline where they fuse to form the primitive cardiac and gut tubes. These tubes subsequently loop and coil along the left-right axis to form a multichambered heart and organs of the digestive system(Horne-Badovinac et al., 2003; Stainier, 2001; Stainier, 2005). In zebrafish,cardiac precursors migrate between the endoderm and the extra-embryonic yolk syncytial layer (YSL) towards the midline(Fig. 1A,B) and fuse to form an intermediate structure termed the cardiac cone(Glickman and Yelon, 2002; Stainier, 2001; Trinh and Stainier, 2004; Yelon et al., 1999). This intermediate structure then elongates to form the primitive heart tube(Holtzman et al., 2007; Rohr et al., 2006). Disruption of cardiac primordia migration can lead to the formation of two separate hearts, a condition known as cardia bifida(Alexander et al., 1998; Chen et al., 1996; Stainier et al., 1996; Yelon, 2001). Although less well studied, a similar process appears to occur during gut morphogenesis(Sakaguchi et al., 2006).

Endoderm, primitive endocardium and extracellular matrix are required for normal cardiac primordia migration. Cardia bifida occurs in chick embryos after the disruption of regions destined to become the ventral gut or yolk sac(Rosenquist, 1970). Zebrafish mutants such as oep, fau, cas (sox32) and bon(Alexander et al., 1999; Dickmeis et al., 2001; Kikuchi et al., 2000) that lack endoderm or have severe endodermal defects also manifest cardia bifida. Zebrafish cloche mutants lack endocardium and have markedly abnormal hearts due to impaired cardiac cone and primitive heart tube formation(Holtzman et al., 2007). Defective function of sphingosine 1-phosphate, a lipid mediator implicated in cell-extracellular matrix interactions(Kawahara et al., 2009; Kupperman et al., 2000; Matsui et al., 2007; Osborne et al., 2008), or loss of fibronectin, a major constituent of the extracellular matrix, also lead to cardia bifida (Linask and Lash,1988; Sakaguchi et al.,2006; Trinh and Stainier,2004).

Initiation of fibronectin fibrillogenesis partly depends on interactions between the cell-binding domain of fibronectin and cell surface integrins(Akiyama et al., 1989; Dzamba et al., 1994; Fogerty et al., 1990; Sechler et al., 1997; Wu et al., 1993). Although less well understood, fibronectin heparin-binding domains also appear to interact with glycosaminoglycans on cell surface proteoglycans to augment the interaction of cells with fibronectin(Bultmann et al., 1998; Christopher et al., 1997; Hocking et al., 1994; McDonald et al., 1987; Schwarzbauer, 1991; Whiteford et al., 2007; Woods et al., 1988). Cells that lack glycosaminoglycan synthesis cannot adhere normally to fibronectin or establish a fibronectin matrix (Chung and Erickson, 1997; LeBaron et al., 1988). Dominant-negative syndecan 2 (Sdc2), a transmembrane heparin sulfate proteoglycan, blocks fibronectin fibrillogenesis in CHO cells(Klass et al., 2000) and in Xenopus animal cap ectoderm(Kramer and Yost, 2002). Based on these findings, Sdc2 may play a role in organ primordia migration by augmenting cellular adhesion of fibronectin and regulating fibrillar organization (Carey, 1997; Woods and Couchman, 1998).

Here we explore early roles of Sdc2 in zebrafish development. In early development, before the migration of cardiac and gut primordia, we find functional differences between embryonic and extra-embryonic Sdc2. Knockdown of sdc2 in the embryo alone does not alter heart or gut primordia migration, whereas knockdown of sdc2 exclusively in the extra-embryonic YSL results in defective heart and gut primordia migration,loss of fibronectin fibrillogenesis throughout the embryo, and deficient basement membrane laminin assembly adjacent to cardiomyocyte and gut endoderm precursors. These defects can be rescued by YSL expression of wild-type Sdc2 or YSL expression of Sdc2 in which the proteoglycan shedding site has been mutated, but not by expression of Sdc2 in the embryo. We propose a model in which organ primordia migration is dependent upon extracellular matrix assembly that requires extra-embryonic Sdc2 and begins at the interface between embryonic and extra-embryonic tissues.

Wild-type zebrafish (Danio rerio) embryos were obtained from crosses of Oregon AB fish. Zebrafish possessing green fluorescent hearts were derived from transgenic fish expressing a cardiac myosin light chain 2:green fluorescent protein fusion gene, Tg(cmlc2:GFP)(Huang et al., 2003). Endocardium was visualized using transgenic fish expressing a friend leukemia integration 1:green fluorescent protein fusion gene Tg(fli1:GFP) (Lawson and Weinstein, 2002). Embryos were maintained in embryo water between 24 and 28.5°C and staged according to age (hours post fertilization, hpf)and morphological criteria (Kimmel et al.,1995; Westerfield,1994).

Morpholinos designed against sdc2 were injected into the YSL at the dome stage (4.3 hpf). Either 16 ng/embryo of a splice-blocking morpholino,MO(SB) (5′-GTGATGCAGACGCTCACCTGATCCC-3′), or 8 ng/embryo of a translation-blocking morpholino, AUG(MO)(5′-ATCCAAAGGTTCCTCATAATTCCTC-3′), were injected. Embryos were co-injected with 1-2 ng/embryo of rhodamine-conjugated control morpholino(5′-CCTCTTACCTCAGTTACAATTTATA-3′). Using a dissecting microscope with a rhodamine filter cube, embryos with a uniform distribution of injected material within the yolk and YSL were selected for further analysis. Capped sense sdc2, GFP-sdc2, sdc2-hCD4JM and sdc2-ΔGAG mRNA was synthesized using the mMESSAGE mMACHINE in vitro transcription kit (Ambion) and purified as described previously (Rupp et al.,1994). For mRNA rescue experiments, between 400 and 600 pg/embryo of sdc2, sdc2-hCD4JM and sdc2-ΔGAG mRNA was injected into the YSL at the dome stage. The GFP-sdc2 construct was designed to detect proteolytic cleavage of the N-terminal ectodomain. In the sdc2-hCD4JM mRNA construct, the putative juxtamembrane cleavage sequence of Sdc2 (DPQDVQSENLFQRTE) was replaced with noncleavable sequence from human CD4 (VKVLPTWSTRVQPMA) (Wang et al., 2005). In the sdc2-ΔGAG construct,the N-terminal membrane localization sequence was retained but all GAG attachment serines were deleted. All morpholino and mRNA injections were repeated at least three times.

For RT-PCR of sdc2 cDNA, 50 wild-type and 50 sdc2MO-injected embryos were collected at the 21-somite stage and total RNA was isolated. RT-PCR reactions were performed using the SuperScript II RT-PCR Kit(Gibco BRL) with the sdc2-specific RT primer pair(5′-GACGACCTGTACCTGGAGGA-3′ and 5′-GCTCTGCTTTAGGCACGAAG-3′) to amplify the sdc2 cDNA fragments.

Digoxigenin-labeled antisense RNA probes were synthesized by in vitro transcription using T3, T7 and SP6 polymerases. cDNA templates used in this study included cmlc2 (Yelon et al., 1999), foxa3(Odenthal and Nusslein-Volhard,1998), sdc2, sox17(Alexander and Stainier, 1999),α ₅ integrin(Koshida et al., 2005),β ₁ integrin (itgb1b)(Mould et al., 2006) and fibronectin 1 (Trinh and Stainier, 2004). Whole-mount in situ hybridization was carried out as described previously (Thisse et al.,1993) using a Biolane HTI in situ machine (Huller and Huttner AG,Tübingen, Germany). Embryos were cleared in 70% glycerol in PBST and photographed with a Leica MZ12 stereoscope using a Dage-MTI DC330 CCD camera.

Embryos were fixed overnight in a 50:50 mixture of 8%paraformaldehyde:1× PBS at 4°C and were dehydrated stepwise into methanol for storage at -20°C. After stepwise rehydration, embryos were manually deyolked and blocked for 1 hour in PBDT (1% BSA, 1% DMSO, 0.1% Triton X-100 in PBS pH 7.3). The deyolked embryos were embedded in 4% low-melt agarose and were sectioned using a vibratome (Electron Microscopy Sciences,OTS-3000). Primary antibodies used in this study included: rabbit polyclonal anti-fibronectin antibody (1:200; Sigma, F3648), rabbit polyclonal anti-laminin (1:200; Sigma, L9393), rabbit polyclonal anti-aPKC (1:200; Santa Cruz Biotechnology, sc-216), mouse IgG anti-β-catenin (1:400; BD Transduction Laboratories, 610153), mouse IgG2 anti-GFP (1:300; Living Colors Monoclonal Antibody JL-8; Clontech, 632381). Alexa Fluor antibodies (Molecular Probes) were used as secondary antibodies at a concentration of 1:300. Sections were equilibrated in SlowFade Antifade buffer (Molecular Probes,2828) and mounted on microscope slides for imaging. Embryo sections were imaged using an Olympus FluoView scanning laser confocal microscope with a 60× objective. To ensure consistent adjustment of microscope power and gain settings, wild-type embryos were imaged in all experiments.

Protein lysates were made from 50 dechorionated embryos at the 21-somite stage (19.5 hpf) that were placed in 1 ml deyolking buffer (55 mM NaCl, 1.8 mM KCl, 1.25 mM NaHCO₃). Yolk sacs were disrupted by pipetting with a 200 μl tip. The embryos were shaken for 5 minutes at 300 rpm to dissolve the yolk. Cells were pelleted at 300 g for 30 seconds and the supernatant was discarded. Two additional wash steps were performed by adding 1 ml wash buffer (110 mM NaCl, 3.5 mM KCl, 2.7 mM CaCl₂, 10 mM TrisHCl pH 8.5), shaking and pelleting cells as before. The pelleted cells were dissolved in 2 μl 2× SDS sample buffer per embryo and incubated for 5 minutes at 95°C. Samples were loaded on an 8% SDS-polyacrylamide gel(ten embryos per lane) and transferred to a PVDF membrane. Immunoblotting was performed using rabbit polyclonal anti-fibronectin antibody (1:3000; Sigma,F3648) followed by HRP anti-rabbit IgG (1:6000, Cell Signaling, #7074). The membrane was developed using ECL Plus and chemifluorescence was measured using a Storm scanner (GE Healthcare, Piscataway, NJ, USA).

sdc2 is expressed ubiquitously during early development in zebrafish, including strong expression in extra-embryonic yolk syncytial layer(YSL) nuclei (Fig. 2A). During gastrulation, YSL nuclei accumulate within the cortical cytoplasm of the yolk as marginal blastomeres collapse onto the yolk cell(D'Amico and Cooper, 2001; Kimmel and Law, 1985). The YSL has been implicated in mesoderm induction(Chen and Kimelman, 2000; Ober and Schulte-Merker,1999), axial patterning via Wnt and Nodal cell signaling pathways(Bischof and Driever, 2004),and migration of cardiac primordia(Kawahara et al., 2009; Sakaguchi et al., 2006). Here,we explore the function of extra-embryonic YSL-expressed sdc2 by morpholino knockdown and compare this with the knockdown of sdc2 in embryonic cells.

Most morpholino-based gene knockdown studies in zebrafish involve injection at the 1-to-2-cell stage, allowing gene knockdown in all embryonic cells during early development (Nasevicius and Ekker, 2000). It is not widely appreciated that morpholinos injected at this stage, even when injected into the yolk cell, are predominantly swept into early cleavage blastomeres(Fig. 1C,E,G). Efficient morpholino uptake into embryonic cells is likely to be driven by microtubule-dependent cytoplasmic streaming that occurs during the first few cell cycles. Cytoplasmic bridges allow small molecules, including morpholinos,to pass from the yolk cell to marginal blastomeres until the sphere stage (4 hpf) (Amack and Yost, 2004). In order to target morpholinos exclusively to the yolk, embryos were injected slightly later, at the dome stage (4.3 hpf). Fluorescent morpholinos injected at the dome stage were detected by fluorescence in the yolk, yolk syncytial nuclei and posterior yolk tube, but not in embryonic cells(Fig. 1D,F,H). Hereafter, we refer to such embryos as yolk^MO embryos, to distinguish them from traditional morphant embryos injected at the 1-to-2-cell stage, in which morpholinos are mainly localized to embryonic tissues and are largely excluded from the yolk.

The yolk forms a functionally distinct compartment from the embryo. Despite ubiquitous expression of Sdc2 throughout early development, there were obvious phenotypic differences between traditional sdc2 morphants and those injected with the sdc2 morpholino at the dome stage(yolk^sdc2MO embryos). Yolk^sdc2MO embryos displayed a spectrum of cardiac and gut migration defects including cardia bifida and a bifid foregut, neither of which was observed in traditional morphants injected with either sdc2 SB(MO) (n=221) or AUG(MO) (n=369). Instead, left-right patterning defects including reversed heart and gut looping were seen in traditional sdc2 morphants (data not shown).

Using the yolk^MO technique described above, sdc2antisense morpholinos were injected into the yolk at the dome stage (4.3 hpf)to knock down sdc2 exclusively in the YSL. Co-injection of a rhodamine-tagged control morpholino allowed visualization of the injected material and selection of embryos with a uniform distribution of morpholino within the yolk and YSL, with no morpholino in embryonic cells. Dome-stage injections of either translation-blocking or splice-blocking morpholinos resulted in the same phenotype, suggesting specific targeting of sdc2mRNA by these morpholinos. Additional evidence supporting morpholino efficacy comes from RT-PCR analysis of sdc2 SB(MO)-injected embryos, in which sdc2 RNA retained intron 2 (929 bp), which is predicted to cause premature translation termination (see Fig. S1 in the supplementary material).

Early in development (18-somite stage), the patterning in yolk^sdc2MO embryos was fairly normal (compare Fig. 2B and Fig. 2C), indicating that YSL-derived Sdc2 is not necessary for early functions, including mesoderm induction, gastrulation, and anterior-posterior and dorsal-ventral patterning. Later in development (40 hpf), yolk^sdc2MO embryos were smaller,with a defective posterior yolk tube and a correspondingly larger yolk cell(compare Fig. 2G and Fig. 2H). yolk^sdc2MOembryos also appeared to have decreased pigmentation and impaired migration of pigmented cells to the region immediately dorsal to the yolk tube.

Evaluation of heart formation by in situ hybridization using the cardiac-specific marker cmlc2 (also known as myl7) revealed complete fusion of cardiac primordia in all wild-type control embryos by 20 hpf (22-somite stage, n=19). By contrast, over 67% (29/43) of yolk^sdc2MO embryos displayed defects in heart-field migration, 28%with incomplete heart-field fusion and 39% with persisting bilateral heart fields (Fig. 2D-F). Looping of the primitive heart tube is well underway by 40 hpf in wild-type embryos, but cardiac development was arrested in many yolk^sdc2MO embryos well before formation of the primitive heart tube. At 40 hpf, over 66% (179/270) of yolk^sdc2MO embryos displayed cardiac defects including impaired blood flow and pericardial edema, with pooling of blood around the heart field, and a spectrum of cardiac anomalies including a band-like distribution of myocardial cell precursors, cardia bifida, incomplete fusion of bilateral heart fields and impaired elongation of the cardiac cone to form the primitive heart tube (Fig. 2I). Using transgenic embryos, Tg(cmlc2:GFP)(Huang et al., 2003), we found that the aberrantly positioned cardiomyocytes were contractile and, in embryos with cardia bifida, two dysmorphic hearts beat independently (see Movie 1 in the supplementary material), demonstrating that other aspects of cardiac morphogenesis continue despite impaired cardiac primordia migration to the midline (Li et al., 2004).

To assess whether the phenotypes observed in yolk^sdc2MO embryos were specific to the function of Sdc2 and not the result of injecting morpholinos into the YSL, control experiments were performed in which 16 ng of either a rhodamine-conjugated control morpholino or a morpholino designed to knock down spadetail (spt, also known as tbx16)(Amack et al., 2007) was injected into each embryo at the dome stage. Yolk^sptMO embryos had normal cardiac primordia migration (n=26), as did yolk^contMO embryos (n=52). The only cardiac anomaly noted in these embryos was a 3-4% incidence of right-sided heart looping, which is similar to the rate of heart reversal in uninjected wild-type embryos. The normal cardiac development in yolk^sptMO embryos is striking when compared with traditional spt morphants, in which injection of only 3 ng of morpholino per embryo yields a high percentage of left-right cardiac patterning defects (Amack et al.,2007; Bisgrove et al.,2000). These results highlight the fact that the yolk forms a functionally distinct compartment from the embryo, that yolk expression of Spt is not necessary for early heart development, and that spt morpholino injected into the yolk at the dome stage does not inhibit known sptgene functions in embryonic cells, consistent with the observation that the yolk^MO injection does not label embryonic cells(Fig. 1H).

To assess whether the cardia bifida phenotypes seen in yolk^sdc2MO embryos are specific to the function of Sdc or are due in part to off-target effects of the morpholinos, rescue experiments were performed by co-injecting sdc2 MO(SB) and sdc2 mRNA into embryos at the dome stage. In the rescue experiments, the percentage of yolk^sdc2MO embryos with cardiac anomalies decreased from 66%(n=270) to 24% (n=358, P<0.01) and the cardiac defects tended towards the milder end of the spectrum, i.e. incomplete heart-field fusion and impaired primitive heart tube formation(Fig. 2I). The ability to rescue the severe cardia bifida phenotypes in yolk^sdc2MO embryos indicates that these phenotypes are attributable to the loss of extra-embryonic Sdc.

In addition to the cardiac migration defects seen in yolk^sdc2MOembryos, gut migration abnormalities were also observed. In wild-type embryos,posterior endoderm condenses into a single rod-like structure along the midline to form the foregut by 24 hpf(Field et al., 2003). Subsequently, there is asymmetric development of the liver and pancreas buds off the foregut (Ober et al.,2003). The foregut marker foxa3 was expressed in all yolk^sdc2MO embryos, indicating that gut primordia specification was normal. However, over 43% (25/58) of yolk^sdc2MO embryos had foregut abnormalities at 40 hpf, consistent with defects in gut primordia migration. Foregut anomalies included duplicated hepatic/pancreatic buds and bilateral condensation of the posterior endoderm, resulting in the formation of a two-sided foregut (Fig. 2J). The bifid gut phenotypes seen in yolk^sdc2MO embryos are specific to the function of extra-embryonic sdc2, as co-injection of sdc2 MO(SB) and sdc2 mRNA into the dome stage resulted in normal gut formation in 95% (58/61) of embryos, indicating a 10-fold rescue of gut development by the expression of sdc2 in the yolk.

These results indicate that migration of the embryonic heart and gut primordia depends specifically on the function of Sdc2 in the YSL. Strikingly,endogenously expressed sdc2 in embryonic cells, including migrating heart and gut primordia, cannot functionally compensate for the loss of Sdc2 in the neighboring extra-embryonic YSL. Therefore, cardiac and gut cell migration is regulated in a non-cell-autonomous manner by extra-embryonic Sdc2.

Since the endoderm and endocardium are required for normal cardiac primordia migration (Alexander et al.,1998; David and Rosa,2001; Holtzman et al.,2007), we examined yolk^sdc2MO embryos for abnormalities in these cell types. Expression of the early endodermal marker sox17was normal in yolk^sdc2MO embryos (n=20, data not shown). Endocardium development, as assessed by GFP expression, in yolk^sdc2MOTg(fli1:GFP) embryos was also normal(n=24 embryos) to the extent detectable with the transgene Tg(fli1:GFP). This suggests that cardiac and gut bifida phenotypes in yolk^sdc2MO embryos are not due to defects in endoderm or endocardium development.

Myocardial and foregut primordia form sheets of polarized epithelia that adhere to and migrate on key components within the extracellular matrix. A number of molecules provide crucial links between extracellular structures and the cytoskeleton to facilitate epithelial polarization and migration. Laminin is found within the basement membrane and required for the induction of cell polarity that leads to the organization of embryonic tissues(Li et al., 2003; Yarnitzky and Volk, 1995). In CHO cells, Sdc2 is required for assembly of laminin into a fibrillar matrix in vitro (Klass et al., 2000). To investigate whether the basement membrane and cardiomyocyte polarity are altered in yolk^sdc2MO embryos, transverse sections of 20- to 22-somite-stage embryos were immunostained for laminin and atypical protein kinase C (aPKC). In wild-type embryos, migrating cardiomyocytes and posterior endoderm had a strong apicolateral localization of aPKC(Fig. 3C,E), indicative of a well-organized polarized epithelium (Trinh and Stainier, 2004), and a clearly defined laminin matrix between the migrating cardiomyocytes and posterior endoderm(Fig. 3A). By contrast,cardiomyocytes in yolk^sdc2MO embryos had weak and disorganized aPKC staining, and posterior endoderm had little if any laminin staining (5/6 embryos; Fig. 3B,D,F). These results indicate that Sdc2 expressed in extra-embryonic YSL is required for cell polarity and laminin matrix formation in the basement membrane of cardiac and gut primordia.

Several lines of evidence support the requirement for fibronectin during cardiomyocyte precursor migration. Cardia bifida occurs in mutants with a disrupted fibronectin 1 gene and in mutants with impaired fibronectin-cardiomyocyte interactions(Matsui et al., 2007; Sakaguchi et al., 2006; Trinh and Stainier, 2004). Mtx1 (also known as Mxtx1), a YSL-specific transcription factor, controls fibronectin 1 transcription in a non-cell-autonomous manner and its knockdown causes cardia bifida (Sakaguchi et al., 2006). To investigate whether extra-embryonic sdc2 also impacts fibronectin fibrillogenesis, transverse sections of embryos at 20- to 22-somite stages were immunostained for fibronectin. Both uninjected controls and traditional sdc2 morphants displayed normal fibronectin fibrillogenesis (n=16 and n=4 embryos,respectively; Fig. 4B,C). By contrast, many yolk^sdc2MO embryos completely lacked fibronectin fibril formation, not just at the YSL-embryo interface, but throughout the entire embryo (8/15 embryos; Fig. 4D). Fibronectin fibrillogenesis was rescued by co-injection of sdc2 mRNA at the dome stage (17/20 embryos; Fig. 4E), whereas overexpression of sdc2 in embryonic cells did not rescue fibrillogenesis in yolk^sdc2MO embryos (7/10 embryos; Fig. 4F).

mRNA encoding fibronectin is expressed in embryonic cells but not in the YSL, and the protein is secreted as monomers into the extracellular matrix,where fibrillogenesis occurs (Trinh and Stainier, 2004). Multiple mechanisms might lead to decreased fibronectin fibril formation, including diminished fibronectin 1 mRNA expression, altered expression patterns of fibronectin 1, decreased levels of fibronectin monomer synthesis or secretion by embryonic cells, or loss of α₅β₁ integrin, the major integrin in zebrafish that plays a crucial role in the initiation of fibronectin fibrillogenesis (Koshida et al.,2005; Mould et al.,2006). These possible mechanisms were systematically assessed in yolk^sdc2MO embryos. Levels of fibronectin 1 mRNA were normal or slightly increased, with normal fibronectin 1 expression patterns (Fig. 5A,B). The accumulation of fibronectin monomers was normal. Although fibronectin monomer secretion cannot be assessed in whole embryos, fibronectin protein synthesis and accumulation were normal in yolk^sdc2MO embryos(Fig. 5C). As a negative control, yolk^mtx1MO embryos were analyzed and found to have a significant reduction in the accumulation of fibronectin monomers, as previously reported (Sakaguchi et al.,2006). Expression patterns of RNA encoding α₅ andβ ₁ integrin in wild-type and yolk^sdc2MO embryos were found to be similar (Fig. 5D-G) and, using real-time PCR, integrin mRNA levels (relative toβ actin, Bactin2) in yolk^sdc2MO embryos were found to be similar to wild-type embryos (0.7:1). Therefore, it appears that extra-embryonic Sdc2 is not required for fibronectin biosynthesis or for the expression of the major fibronectin receptorα ₅β₁ integrin, but is required for the polymerization of fibronectin monomers into complex fibrils throughout the embryo.

Sakaguchi et al. (Sakaguchi et al.,2006) described similar, although less severe, cardiac and foregut anomalies in zebrafish with knockdown of the YSL-specific transcription factor Mtx1. By an unknown mechanism, extra-embryonic Mtx1 controls fibronectin 1 transcription in a non-cell-autonomous manner. Despite the phenotypic similarities between yolk^sdc2MO and yolk^mtx1MO embryos,no apparent genetic interaction exists between Sdc2 and Mtx1. In contrast to mtx1 knockdown, sdc2 knockdown does not affect fibronectin 1 mRNA expression or fibronectin protein accumulation(Fig. 5) but instead it impairs fibronectin matrix assembly (Fig. 4). In addition, knockdown of extra-embryonic mtx1 does not alter sdc2 mRNA expression (data not shown). As a functional test, cardiac migration defects were seen in 50% of yolk^mtx1MOembryos (n=39) and the aberrant cardiac migration was not rescued by coexpression of sdc2 mRNA (65% with cardiac migration defects, n=54). Together, these results indicate that there are multiple genetic pathways in nonembryonic cells that control embryonic development and cardiac primordia migration, one of which involves induction of extracellular matrix formation at the YSL-embryo interface by extra-embryonic Sdc2.

The syndecan family of proteoglycans has a juxtamembrane cleavage site that allows release of the N-terminal ectodomain into the extracellular space in a process known as proteoglycan shedding(Fears et al., 2006). This process presumably allows the ectodomain to function at a distance from the cell in which it is synthesized. First, to detect whether the Sdc2 ectodomain moves a long distance from YSL into the regions in which fibrillogenesis is rescued by YSL-expressed sdc2, a construct was made with GFP tagged to the ectodomain of Sdc2, N-terminal to the juxtamembrane proteolytic cleavage site (Fig. 6A). With release of the ectodomain from the cell surface by `shedding', GFP would presumably move away from its cell of origin. Confocal microscopy images of embryos with GFP-sdc2 mRNA injected into the yolk at the dome stage revealed that GFP fluorescence was concentrated within and around the YSL(Fig. 6C), and was not detected deeply in embryonic tissues (arrow in Fig. 6C), where fibrillogenesis is altered by knockdown of sdc2 in extra-embryonic cells. At the level of resolution afforded by confocal microscopy, these results indicate that the GFP-tagged Sdc2 ectodomain expressed in the YSL did not move deeply into embryonic tissues. These experiments do not address whether endogenous Sdc2 is cleaved and moves a short distance from the source. Confinement of GFP-labeled Sdc2 near the YSL could be due to a lack of cleavage/shedding or to retention of the cleaved Sdc2 ectodomain at the cell surface by interaction of Sdc2 GAGs with the surrounding extracellular matrix. As a control to assess whether GFP-labeled Sdc2 could be detected in embryonic tissues, GFP-sdc2 mRNA was injected at the 1-to-2-cell stage and resulted in GFP expression throughout the embryo (arrow in Fig. 6B).

To functionally test whether cleavage/shedding of extra-embryonic Sdc2 is necessary for fibrillogenesis and organ primordia movement, we generated a noncleavable Sdc2 construct (sdc2-hCD4JM)(Fig. 6D), in which the juxtamembrane cleavage sequence was replaced with noncleavable sequence from human CD4 (Wang et al., 2005). Strikingly, expression of sdc2-hCD4JM in yolk^sdc2MOembryos was able to rescue heart migration and fibrillogenesis throughout the embryo (Fig. 6E), similar to wild-type sdc2 (Fig. 4E). These results indicate that shedding is not necessary for extra-embryonic Sdc2 function, and that Sdc2 tethered to the plasma membrane of the extra-embryonic YSL is capable of driving fibrillogenesis and cell migration throughout the embryo.

We propose a model in which extra-embryonic Sdc2 functions to drive extracellular matrix assembly and organ primordia movement within the embryo(Fig. 7). Sdc2 is expressed in both embryonic and extra-embryonic tissues, but the two forms of Sdc2 are functionally distinct. In syndecan 4, a homolog of Sdc2, the GAG chains interact with the heparin-binding domains of fibronectin(Carey, 1997; Klass et al., 2000; Whiteford et al., 2007). Hence, one exciting possibility is that specific GAG chain modifications on extra-embryonic Sdc2 (red in Fig. 7) drive fibrillogenesis, and these GAG chain modifications are distinct from those found on Sdc2 expressed in embryonic cells (blue in Fig. 7), which are incapable of driving fibrillogenesis, but perform other functions. In support of this hypothesis, cardiac migration defects are not rescued in yolk^sdc2MOembryos by co-injection of sdc2-ΔGAG, an Sdc2 construct lacking GAG attachment sites(Fig. 6F,G). These results demonstrate that Sdc2 GAG chains are crucial for cardiac primordia migration and suggest that GAG chain modifications vary from one embryonic compartment to another, leading to differences in proteoglycan function.

Sdc2 is a transmembrane heparan sulfate proteoglycan with heparan sulfate glycosaminoglycan (GAG) chains covalently bound to its ectodomain, and is expressed in both embryonic and extra-embryonic cells during zebrafish development. The roles of extra-embryonic cells, including the extra-embryonic yolk syncytial layer (YSL), in zebrafish are poorly understood. Here we present evidence indicating that extra-embryonic Sdc2 cell-autonomously directs fibronectin fibrillogenesis and laminin matrix assembly at the YSL-embryo interface, and non-cell-autonomously controls matrix formation, and cardiac and gut primordia movement throughout the embryo.

Given that sdc2 mRNA is expressed throughout the embryo, it is striking that neither endogenous embryonic Sdc2 nor overexpression of sdc2 mRNA throughout the embryo by injection of the mRNA at the single-cell stage can compensate for the loss of Sdc2 in extra-embryonic cells(see Fig. 4). Further evidence for a functional difference between embryonic and extra-embryonic Sdc2 comes from traditional morphant studies in which knockdown of sdc2 in embryonic cells alters angiogenesis (Chen et al., 2004) and left-right patterning (data not shown) but does not lead to a bifid heart or gut. Although it is possible that extra-embryonic Sdc2 modulates a signal in the YSL that induces secretion of an unknown signal out of the YSL into the embryo to mediate fibrillogenesis, cell polarization and cell migration in the embryo, or that extra-embryonic Sdc2 plays a role in yolk syncytial nuclei movement that somehow modulates cardiac primordia patterning (Carvalho et al.,2009), a more parsimonious explanation is that extra-embryonic Sdc2 directly controls fibrillogenesis at the YSL-embryo interface and subsequently controls cell migration. The ability of YSL-expressed Sdc2 to nucleate fibrillogenesis concurs with the role of syndecans in promoting fibrillogenesis in cell culture (Carey,1997; Klass et al.,2000; Woods and Couchman,1998). Although ectodomain shedding in extra-embryonic Sdc2 does not appear to be necessary for fibronectin polymerization, extra-embryonic GAG chains attached to Sdc2 appear to be required for organ primordia migration(Fig. 6). Overall, our results indicate that functional differences exist between embryonic and extra-embryonic Sdc2 and that these differences are at least in part driven by GAG chains on Sdc2. An intriguing possibility is it that the fine structural modifications of the GAG chains attached to Sdc2 in the extra-embryonic cells are distinct from the fine structural modifications of the GAG chains attached to Sdc2 in embryonic cells (Fig. 7).

The striking foregut anomalies observed in yolk^sdc2MO embryos are not observed in natter (fibronectin 1) mutants(Trinh and Stainier, 2004),suggesting that foregut migration is not dependent on fibronectin fibrillogenesis and that Sdc2 utilizes some other non-cell-autonomous mechanism to control gut epithelial migration. Members of the laminin family are required for basement membrane assembly. Basement membranes are specialized extracellular matrices that form sheets underlying cell layers and play a crucial role in cell polarization and tissue epithelialization(Li et al., 2003; Yarnitzky and Volk, 1995). In zebrafish, a basement membrane is located between the cardiac primordia and ventral foregut primordia. As assayed by laminin and aPKC immunoreactivity,this basement membrane is nearly absent and epithelial cell polarity is disrupted in yolk^sdc2MO embryos (see Fig. 3). Interestingly, similar basement membrane, cell polarity and gut migration defects occur in both yolk^sdc2MO and yolk^mtx1MO embryos, suggesting that posterior endoderm migration may depend on the formation of an organized basement membrane (Sakaguchi et al.,2006). Additional evidence supporting this hypothesis comes from zebrafish has (prkci) mutants that lack aPKC(Horne-Badovinac et al.,2001). A deficiency in aPKC results in impaired formation of the zonula adherens in the polarized epithelia of the digestive tract, and leads to gut malformations including duplication of the hepatic/pancreatic buds.

In summary, extra-embryonic Sdc2 appears to direct heart and gut primordia migration by initiating fibronectin and laminin matrix assembly at the yolk-embryo interface. Strikingly, fibrillogenesis spreads throughout the entire zebrafish embryo. Functional differences in GAG chain structure between embryonic and extra-embryonic Sdc2 might serve to regulate fibrillogenesis,cell migration and epithelial morphogenesis of multiple organ systems throughout the embryo.