|Home | About | Journals | Submit | Contact Us | Français|
The liver, pancreas and lungs are induced from endoderm progenitors by a series of dynamic growth factor signals from the mesoderm, but how the temporal-spatial activity of these signals is controlled is poorly understood. We have identified an extracellular regulatory loop required for robust BMP signaling in the Xenopus foregut. We show that BMP signaling is required to maintain foregut progenitors and induce expression of the secreted frizzled related protein Sizzled (Szl) and the extracellular metalloprotease Tolloid-like 1 (Tll1). Szl negatively regulates Tll activity to control deposition of a Fibronectin (FN) matrix between the mesoderm and endoderm, which is required to maintain BMP signaling. Foregut-specific Szl depletion results in a loss of the FN matrix and failure to maintain robust pSmad1 levels causing a loss of foregut gene expression and organ agenesis. These results have implications for BMP signaling in diverse contexts and the differentiation of foregut tissue from stem cells.
During embryogenesis the liver, ventral pancreas, and lungs are induced from progenitor cells in ventral foregut endoderm by a series of Wnt, Fibroblast Growth Factor (FGF) and Bone Morphogenetic Protein (BMP) signals from adjacent mesoderm (Zaret, 2008; Zorn and Wells, 2009). Recent evidence indicates that these pathways, in particular BMP, are highly dynamic with distinct biological effects on endoderm cells at different times in organogenesis (McLin et al., 2007; Wandzioch and Zaret, 2009). During initial endoderm patterning in the gastrula, BMPs promote hindgut development and inhibit foregut progenitor cell fate (Rankin et al., 2011; Tiso et al., 2002). However, hours later BMPs from the lateral plate and cardiac mesoderm segregate foregut progenitors, promoting hepatic and repressing pancreatic lineages (Chung et al., 2010; Chung et al., 2008; Rossi et al., 2001; Shin et al., 2007). Shortly after this, spatially distinct BMP activities then promote pancreatic development, organ bud growth and tracheoesophageal segregation (Rodriguez et al., 2010; Wandzioch and Zaret, 2009).
The molecular mechanisms controlling temporal-spatial BMP signaling dynamics in foregut organogenesis are poorly understood. One key means to regulate BMP signaling is through the extracellular control of ligand bioavailability, which is best understood in the context of Drosophila and Xenopus axial pattering (Umulis et al., 2010; Zakin and De Robertis, 2010). In Xenopus gastrulae BMP ligands interact with numerous secreted BMP-binding proteins including Chordin, Crossveinless-2 (Cv2), Twisted gastrulation (Tsg) and Olfactomedin-1, which can form multi-protein complexes (Inomata et al., 2008; Zakin and De Robertis, 2010). These BMP-binding complexes are regulated in part by secreted BMP1/Tolloid-like (Tll) family of metalloproteases (Hopkins et al., 2007), which cleave Chordin, thereby liberating BMP ligands in a spatially restricted manner (Blitz et al., 2000; Dale et al., 2002; Piccolo et al., 1997). Adding to this complexity, BMP ligands as well as many proteins in these BMP-binding complexes can interact with the ECM and although the mechanisms are not well understood, there is evidence that this can either restrict signaling by sequestering ligands or promote signaling by facilitating ligand diffusion and accumulation in the ECM followed by regulated release. (Muir and Greenspan, 2011; Ramirez and Rifkin, 2009; Zakin and De Robertis, 2010).
The extent to which such extracellular mechanisms can account for dynamic BMP signaling in foregut organogenesis is unknown. In this study we identified Sizzled (Szl), a secreted fizzled-related protein (Sfrp) (Salic et al., 1997), as a key regulator of BMP signaling in the Xenopus foregut. Previous studies in gastrulae have shown that Szl functions as a BMP feedback inhibitor during axial patterning where it inhibits Tll proteases to prevent Chordin cleavage, thus keeping BMP ligands sequestered (Collavin and Kirschner, 2003; Lee et al., 2006; Muraoka et al., 2006; Yabe et al., 2003). In contrast to this anti-BMP function we show that in the Xenopus foregut development Szl –Tll interactions exert a pro-BMP activity.
We demonstrate that prior to foregut organ specification, mesenchymal BMP signals are essential to maintain foregut progenitors and to induce szl and tolloid like 1 (tll1) expression. Szl in turn is required to inhibit Tll-proteases to maintain a robust positive BMP signaling loop in the foregut. Depletion of Szl from the Xenopus foregut results in: 1) disruption of the FN matrix between the bmp-expressing mesoderm and the foregut endoderm 2) failure to maintain pSmad1 levels in foregut progenitors causing 3) loss of BMP-responsive foregut gene expression, including reduced bmp2/4/7 ligand expression 4) Szl morphants also exhibit elevated apoptosis of foregut progenitors and 5) subsequent liver, pancreas and lung agenesis. Szl morphants can be rescued by knockdown of Bmp1 and Tll1 indicating that the Szl phenotype is due to inappropriately elevated protease activity. Our data suggest a Chordin-independent mechanism where a Szl-Tll feedback loop maintains BMP signaling in the foregut through modulation of the FN-rich ECM. Consistent with this FN knockdown phenocopies the loss of Szl and reduces BMP signaling. These results have broad implications for the extracellular regulation of BMP signaling and open the door to understanding the molecular mechanism controlling signaling dynamics during foregut organogenesis.
In Xenopus the liver and pancreatic lineages are specified by stage 30 as a result of earlier inductive signals from the precardiac lateral plate mesoderm, but exactly when these critical inductive event occur is unclear. In experiments separating the mesoderm and endoderm prior to organ specification, we found that foregut identity was labile and that mesoderm contact between stages 15 to 23 was required to maintain expression of the ventral foregut progenitor marker hhex (Figure 1). Because early target genes induced in foregut progenitors are largely unknown in any species we performed a microarray screen on foregut explants cultured either with or without the adjacent mesoderm. One hundred foregut explants were isolated at stage 15 (0 somite stage; ss). In half of the explants, the mesoderm was removed immediately; the remaining explants were left intact. Both groups were cultured until stage 23 (12 ss). The mesoderm was then removed from the “intact” group and gene expression in the endoderm tissue cultured with and without mesoderm was compared by microarray (Figure 1A).
We identified 85 transcripts induced more than 2-fold in foregut endoderm by adjacent mesoderm (Supplementary Table S1). Consistent with predictions from mouse explant studies, several of these were targets/components of the FGF and BMP pathways (Figure 1B). Sizzled, the most highly induced transcript encodes a secreted frizzled-related protein (Sfrp) family member, containing a motif homologous to the Wnt-binding domain of Frizzled receptors (Salic et al., 1997). However, previous studies in frog and fish gastrulae indicate that Szl does not act as a Wnt-antagonist, but rather as a BMP feedback inhibitor during axial patterning (Collavin and Kirschner, 2003; Lee et al., 2006; Muraoka et al., 2006; Salic et al., 1997; Yabe et al., 2003). In situ hybridization showed that in addition to its well-characterized expression in gastrula ventral mesoderm, szl is also transiently expressed in the foregut endoderm and lateral plate mesoderm during stages 15–28 (Figure 1C and Supplementary Figure S1) (Lee et al., 2006). In situ hybridization confirmed that like hhex, szl expression in foregut endoderm was dependent upon mesoderm (Figure 1C).
In Xenopus gastrulae szl transcription is induced by BMP4/7 (Reversade and De Robertis, 2005). After gastrulation, bmp4 and bmp7 are expressed throughout the ventral-lateral plate mesoderm, whereas bmp2 is localized to the precardiac mesendoderm (Figure 1D and Supplementary Figure S1), suggesting that BMPs might also induce szl expression in the foregut. To test this we injected an antisense bmp2-MO (Reversade et al., 2005) into the anterior mesendoderm, which resulted in a dramatic down regulation of szl and hhex foregut expression at stages 18–20 (Figure 1C). Injection of bmp4-MO, bmp7-MO or treatment with a BMP receptor inhibitor LDN193189 all caused a similar reduction in hhex and szl (data not shown). Importantly, the earlier expression of hhex and chordin in the gastrula organizer was unaltered by the anterior injection of the bmp2-MO, confirming that axial patterning was not disrupted (data not shown). We next injected recombinant BMP2 protein into the foregut cavity at stage 12–13, which expanded szl and hhex expression at stage 20 (Figure 1C). We conclude that high levels of BMP signaling are required to maintain foregut progenitors, induce szl expression, and regulate the size of the foregut domain.
To determine the function of Szl in foregut we microinjected a well-characterized translation blocking antisense szl morpholino oligo (szl-MO) (Collavin and Kirschner, 2003; Lee et al., 2006) into C1/D1 cells of 16-cell stage embryos, which contribute to the foregut. This enabled a foregut-specific loss-of-function and allowed us to avoid interfering with the earlier gastrula function of Szl in the ventral mesoderm. Co-injection of a fluorescent lineage label confirmed correct targeting to the foregut (Figure 2A–D).
Foregut-specific Szl knockdown resulted in a dramatic reduction in hhex expression at stage 20, loss of liver and thyroid (hhex), pancreas (pdx1) and lung (nkx2.1) markers at stage 35 and frequent foregut organ bud agenesis at stage 42 (Figure 2 E–T). Interestingly the remnant hhex expression at stage 20 was often restricted to the endoderm cells immediately adjacent to the bmp-expressing mesoderm (Figure 2F), suggesting that diffusion of BMPs into endoderm might be disrupted. Intestinal markers were not ectopically expressed in the foregut of Szl morphants (data not shown), indicating that they did not acquire a posterior fate. The early cardiac marker nkx2.5 was unchanged, but later heart differentiation (tnIc) was impaired.
To confirm that the phenotype was specific to loss of Szl we co-injected the szl-MO along with a szl expression plasmid in which the MO target sequence has been mutated (Collavin and Kirschner, 2003). The szl plasmid partially rescued foregut organ development, based on hhex expression at stage 32 and nkx2.1-expressing lung buds at stage 42 (Supplementary Figure S2). We also assayed foregut targeted szl-MO embryos at gastrulation and compared these to ventrally injected szl-MO embryos (Supplementary Figure S2). This confirmed that the foregut injections affect neither axial patterning nor initial hhex expression in the gastrula, whereas ventral injections increased BMP signaling in the gastrula and disrupted axial patterning as previously published (Lee et al., 2006). We conclude that Szl is required for Xenopus foregut organogenesis and that this function is distinct from its previously published role in axial patterning.
Sfrps were independently identified as a family of proteins that can regulate apoptosis (Melkonyan et al., 1997). We therefore evaluated whether the hypoplastic foregut in Szl morphants might be due to cell death. At stage 20, when we see loss of hhex expression, there are approximately 400 cells in the foreguts of both control and szl-MO embryos. Activated Caspase–3 immunostaining revealed that the apoptotic pathway was activated in 10–15% of Szl-depleted foregut cells at stage 20 compared to ~1% in controls (Figure 3A). This was partially rescued by co-injection of a szl expression plasmid, which alone also caused a modest increase in apoptosis, suggesting that survival of foregut progenitors requires precisely titrated Szl levels. Injection of the szl-MO throughout the embryo did not cause significant cell death in either the ectoderm or the hindgut (Figure 3B) ruling out general morpholino toxicity. Activated Caspase-3 was undetectable at stage 13 in either control of szl-MO embryos (data not shown). Phosphorylated-Histone H3 staining (PH3+) showed no significant differences in proliferation between controls and Szl morphants; the mean number PH3+ foregut cells at stages 15 and 20 were 42 ± 8.5 and 15.3 ± 9 in cont-MO with 40.7 ± 3.3 and 12.3±4 in the szl-MO, respectively.
To determine the extent to which cell death could account for the loss of foregut identity in Szl morphants we treated control and szl-MO injected embryos with caspase inhibitors. Using three different inhibitors that prevent cell death in Szl morphants, hhex expression in foregut progenitors and the liver bud was still down regulated (Figure 3C, D). This indicates that loss of foregut identity in Szl morphants is not due to apoptosis, although increased cell death likely contributes to the lack of organ buds later.
Sfrps can function as Wnt-antagonists and/or as modulators of Tll proteases (Bovolenta et al., 2008; Ploper et al., 2011). Although Szl is known to repress BMP signaling in the gastrula through the inhibition of Tll-chordinase activity, it was important to determine whether in the context of the foregut Szl might repress Wnt/β– catenin signaling because Wnt inhibition by the related factor Sfrp5 is critical for foregut organogenesis (Li et al., 2008). Analysis of a Tcf/β-catenin transcriptional reporter construct (TOP:flash) injected into foregut cells with or without the szl-MO showed that β-catenin activity was not increased in Szl depleted embryos relative to sibling controls (indeed Wnt signaling was reduced) (Supplementary Figure S3), suggesting that Szl does not act as a Wnt-antagonist in the foregut.
We next examined whether BMP signaling was altered in Szl-depleted foregut tissue using a BMP/Smad1 responsive transcriptional reporter (BRE:luc) (von Bubnoff et al., 2005). In control injection experiments depletion of Szl from the ventral mesoderm caused the predicted increase in BRE:luc activity at the gastrula stage. Surprisingly foregut injection of the szl-MO had the opposite effect and caused a progressive loss of BRE:luc activity between stages 18–23 (Supplementary Figure S3).
To verify that Szl was required to maintain BMP signaling in the foregut we examined C-terminal serine phosphorylation and activation of Smad1/5/8 (pSmad1). Immunostaining and western blot analyses of control embryos revealed that pSmad1 levels in the developing foregut were dynamic (Figure 4) peaking at stages 17–20 in a domain similar to that which expresses hhex, followed by a decline in pSmad1 levels between stages 21–24, mirroring szl’s temporal expression. This dynamic pSmad1 activity in the Xenopus foregut is similar to recent observation in the mouse (Wandzioch and Zaret, 2009). Western blot analysis of foregut explants from control-MO and szl-MO injected embryos showed that Szl morphants had a modest reduction in pSmad1 levels (Figure 5A), consistent with the BRE:luc reporter assay. These data indicate that in the foregut Szl exerts a pro-BMP activity, which was surprising considering its established anti-BMP role in the gastrula. We next examined whether the loss of pSmad1 was localized to specific sub-regions of the foregut (Figure 5B). Quantitation of nuclear pSmad1 confocal immunostaining in different cell populations revealed that at stage 19 Szl morphants had significantly reduced pSmad1 levels in the deep foregut endoderm located 4–9 cells from the bmp-expressing mesoderm (Figure 5C). However in endoderm cells immediately adjacent to the mesoderm pSmad1 levels were comparable between control and szl-MO embryos, which might explain the remnant hhex expression in the mesoderm-proximal endoderm in many Szl morphants (Figure 2F).
To determine whether this reduced pSmad1 could explain the Szl morphant phenotype we examined sibling embryos where just BMP2 was depleted (but not BMP4/7) or where all BMP signaling was blocked by treatment with the BMP-receptor inhibitor LDN193189. Bmp2-MO injected embryos also exhibited reduced nuclear pSmad1 levels in the deep foregut endoderm, with no change in the mesoderm-proximal region, very similar to the Szl morphants. In contrast LDN-treated embryos had reduced pSmad1 levels throughout the foregut (Figure 5B and C). This suggests that the reduced ligand concentration caused by BMP2 knockdown results in the deep foregut not receiving sufficient BMP to maintain robust pSmad1 levels. Importantly depletion of either BMP2 or Szl was sufficient to dramatically down regulate hhex expression (Figures 1 and and2)2) indicating that it requires robust pSmad1 activity.
In addition to hhex, we examined several other known Smad1-target genes including vent1/2 and szl itself (Karaulanov et al., 2004; von Bubnoff et al., 2005), as well as the expression of bmp ligands. These were all down regulated in Szl morphants (Figure 5D) and the reduced bmp2/4/7 transcripts were consistent with the decrease in pSmad1 levels. Injection of recombinant BMP2 protein into the closing blastocoel (which becomes the foregut) of szl-MO embryos at stage 12 was sufficient to rescue expression of hhex, szl, vent1/2 and bmp2/4/7 transcripts (Figure 5D), as well as rescue apoptosis (Figure 5E). These data indicate that sustained robust BMP/Smad1 activity is required to maintain foregut gene expression and to define the size of the foregut domain. The results also demonstrate that bmp2/4/7 gene expression in the foregut mesendoderm is positively regulated by BMP signaling and that Szl is required to maintain this positive BMP signaling loop.
The pro-BMP activity of Szl in the embryonic foregut is distinct from its anti-BMP role in the gastrula. However, we postulated that this pro-BMP activity might also be mediated by inhibition of Tll proteases. In situ analysis of tll-family genes revealed that tll1 (xlr) was expressed in the foregut endoderm at stages 15–20 in a pattern overlapping szl (Figure 6A, B) (Dale et al., 2002; Inomata et al., 2008). bmp1 (which encodes a Tll protease not a BMP ligand) was ubiquitously expressed at a low level throughout the embryo including the foregut and more strongly in the neural plate and notochord (Figure 6C), whereas tll2 (xolloid) transcripts were not detected (not shown). Treatment of embryos with the BMP-receptor inhibitor LDN193189 or combined bmp2/4/7-MO injection inhibited tll1 expression (Figure 6H; not shown). However foregut injections of szl-MO or bmp2-MO had very little if any impact on tll1 transcripts (Figure 6E–G). We note that although szl-MO embryos exhibit reduced (but not absent) bmp2/4/7 expression (Figure 5D) and have lower pSmad1 levels, (comparable to bmp2 knockdown), the nuclear pSmad1 levels in szl-MO and bmp2-MO foreguts are still significantly higher than those in LDN treated embryos where BMP signaling is more robustly inhibited (Figure 5B,C). This suggests that while the lower pSmad1 levels in szl-MO and bmp2-MO embryos are not sufficient to maintain hhex (Figures 1, ,22 and and5),5), they are sufficient to maintain tll1 expression (Figure 6E–H). Together the data suggest that tll1 expression in the foregut is positively regulated by a low level of BMP-signaling.
The co-expression of szl and tll1 in the foregut led us to postulate that Szl might normally function to restrict Tll protease activity, which would be inappropriately elevated in Szl morphants. To test this we co-injected tll1-MO and bmp1-MOs (Inomata et al., 2008) into the presumptive foregut cells with or without the szl-MO. Injection of the tll1/bmp1-MOs with a control-MO had no obvious impact on foregut gene expression. However co-injection of tll1/bmp1-MOs with the szl-MO rescued expression of hhex and vent1/2 at stage 20 (Figure 6I–P) and the liver marker for1 at stage 35 (Figure 6 Q–T). Co-injection of individual tll1-MO or bmp1-MOs only partial rescued Szl morphants (data not shown) suggesting that Szl represses both Tll1 and BMP1, which might have distinct substrates the foregut. We conclude that the loss of bmp-responsive foregut gene expression in Szl morphants was due to excessive Tll activity.
In the gastrula Szl represses BMP signaling by inhibiting Tll-mediated Chordin degradation. However the pro-BMP activity of Szl in the foregut is not consistent with reduced Chordin levels. Moreover, injection of Chordin protein into the foregut could not rescue the szl-MO phenotype (Supplemental Figure S4) and based on in situ hybridization chordin is not expressed in the foregut (Figure 6D). These data suggest that in the foregut Szl-Tll act via a Chordin-independent mechanism. We next considered the possibility that Szl-Tll might promote BMP signaling through regulation of the ECM. Tll proteases are known to regulate ECM deposition, particularly by proteolytic processing of collagens, and there are emerging reports that the ECM can influence BMP bioavailability (Bovolenta et al., 2008; Muir and Greenspan, 2011; Ramirez and Rifkin, 2009).
Immunostaining of various ECM proteins including; Collagens I–V, Fibrillin, Laminin and FN indicated that at stages 18–20 the foregut ECM is relatively simple (Supplementary Figure S5). Only FN, which is known to play an early role in initiating ECM assembly (Schwarzbauer and DeSimone, 2011), was detectable as faint pericellular matrix surrounding the foregut cells and in two prominent fibril layers; one between the endoderm and mesoderm (Figure 7; yellow arrows) and another between the mesoderm and ectoderm (Davidson et al., 2004). We could not detect Laminin, Collagen I–V and Fibrillin fibrils in the foregut until stages 28–32 (data not shown) well after the time when Szl-Tll act. Analysis of FN in Szl morphants revealed that the layer of fibrils between the foregut endoderm and mesoderm was absent or severely reduced (Figure 7A and and8A,8A, red arrows), whereas the FN matrix between the mesoderm and ectoderm was intact. Membrane localization of β1-integrin a major FN receptor was not obviously altered in Szl morphants. Western blot analysis of foregut explants showed very little if any degradation or reduction in total FN or β1-integrin levels, suggesting that Szl was required for fibrilogenesis rather than FN expression (Figure 7B).
We next tested whether reduced BMP signaling or elevated Tll activity could account for the loss of FN matrix in Szl morphants. Inhibition of BMP signaling either by LDN treatment or bmp2/4/7-MO injection did not disrupt the FN matrix (Figure 8A; data not shown), suggesting that reduced BMP signaling cannot account for the loss of the FN matrix in szl-MO embryos. However reducing Tll levels by co-injection of the tll1/bmp1-MOs along with the szl-MO, completely rescued the FN layer in the foregut, whereas injection of tll1/bmp1-MOs into control embryos resulted in an increase of disorganized FN matrix (Figure 8A). This indicates that Tll activity negatively regulates FN matrix deposition in the foregut. In addition to rescuing the FN matrix in Szl morphants, knockdown of Tll1/BMP1 also restored pSmad1 levels in the endoderm, rescued bmp2/4/7 gene expression and prevented apoptosis (Figure 8A–C). In short reducing Tll1/BMP1 levels rescued all aspects of the szl-MO phenotype, supporting the interpretation that Tll activity is inappropriately elevated in the absence of Szl. Consistent with this hypothesis, injection of recombinant BMP1 into the foregut cavity partially disrupted the FN matrix, reduced hhex levels and induced foregut cell apoptosis (data not shown). These data indicate that Szl-Tll interactions regulate FN matrix deposition and suggest that unrestrained Tll activity in the szl-MO foregut prevents the formation of FN fibrils, which correlates with the loss of BMP signaling.
To directly test the hypothesis that the FN matrix is required for BMP-signaling we injected well-characterized fibronectin morpholinos (fn-MO) (Davidson et al., 2006) into the C1/D1-cells of 16-cell stage embryos, which targets the foregut mesendoderm, but avoids the axial mesoderm and ectoderm (e.g. Figure 2B) where FN is known to regulate gastrulation. The fn-MO injected embryos gastrulated normally with the anterior mesendoderm migrating to the proper ventral foregut position. Immunostaining confirmed that the pericellular FN matrix in the foregut and the fibril layer between the endoderm and mesoderm was absent from the fn-MO foregut injected embryos, whereas the FN fibrils between the mesoderm and ectoderm were still present, albeit reduced (Figure 8A and Supplementary Figure S6). Depletion of FN from the foregut partially phenocopied Szl morphants with reduced bmp2/4/7 and hhex expression and significantly lower pSmad1 levels in the foregut (Figure 8). Later in development FN depleted embryos had reduced liver gene expression and cardia bifida (Supplementary Figure S6) similar to zebrafish fn mutants (Trinh and Stainier, 2004). Interestingly fn-MO embryos did not exhibit increased foregut apoptosis, nor did embryos where BMP/Smad1 signaling was either directly over-activated (Figures 5E) or blocked (Figure 8C). However recombinant BMP2 protein did rescue cell death in Szl morphants (Figure 5E). This suggests that Szl-Tll interactions are complex and likely promote foregut cell survival through parallel BMP-dependent and BMP-independent mechanisms.
We conclude Szl exerts a pro-BMP activity in the foregut by negatively regulating Tll proteases and that Szl-Tll interactions regulate FN matrix deposition in the foregut, which is required to maintain the positive BMP feedback loop essential for foregut organogenesis.
BMP signaling is highly dynamic during foregut organogenesis, but the molecular mechanisms controlling this are poorly understood. In this study we have identified a BMP regulatory loop that is essential for foregut organ development. We demonstrate that prior to organ specification in Xenopus, BMP signals from the mesoderm are essential to maintain foregut progenitors, and that the level of BMP signaling regulates the size of the foregut domain. Our data support a model where BMP signaling induces szl and tll1 expression in the foregut. Secreted Szl in turn negatively regulates extracellular Tll proteases and the balance between Szl-Tll activity controls proper FN matrix deposition in the foregut, which is required to maintain a positive BMP-signaling loop essential for foregut development. Foregut specific knockdown of Szl, BMP2 or FN all result in a failure to sustain robust pSmad1 levels in the deep endoderm causing a loss of BMP-dependent foregut gene expression including hhex, szl, and bmp2/4/7; thus breaking the positive BMP regulatory loop. Knockdown of Szl, but not BMP or FN, also results in increased foregut progenitor apoptosis, which is separable from loss of foregut identity and probably due to inappropriately elevated Tll protease activity. In summary Szl is required for foregut development, and it exerts a pro-BMP activity by restricting Tll protease activity and regulating the FN-rich ECM.
Our data adds to the growing body of evidence that the ECM plays an important role in regulating BMP signaling, and suggest that FN fibrils might act as a scaffold to assemble BMP-containing protein complexes (Muir and Greenspan, 2011; Ramirez and Rifkin, 2009). Isolated FN fragments containing the 12th – 14th type III repeats have recently been reported to bind BMP2 and BMP7 in vitro (Martino and Hubbell, 2010) although the in vivo relevance of this remains to be tested. FN can also bind to the Szl-related protein Sfrp2 in cultured mammary cells where it exerts an anti-apoptotic effect (Lee et al., 2003), and FN can physically interact with BMP1 enhancing its proteolytic activity (Huang et al., 2009; Muir and Greenspan, 2011). There is evidence that BMP1/Tll can also directly bind BMP ligands and these interactions can both inhibit Tll protease activity as well as prevent BMP ligand-receptor binding (Jasuja et al., 2007; Lee et al., 2009). Thus FN fibrils might form a scaffold essential to facilitate Szl-Tll-BMP interactions that regulate BMP signaling in the foregut.
BMPs can also bind to other ECM proteins and these interactions could be disrupted by loss of the FN matrix in Szl morphants, because FN fibrils are required to nucleate the assembly of more mature ECM (Ramirez and Rifkin, 2009; Schwarzbauer and DeSimone, 2011). For example, BMP2/4/7 prodomains can bind with high affinity to Fibrillin and fibrillin-2 mutant mice exhibit decreased BMP signaling (Nistala et al., 2010; Ramirez and Rifkin, 2009). In Drosophila BMPs can bind to collagen IV and depending on the cellular context this can either augment signaling and promote morphogen gradient formation or restrict the signaling range through ligand sequestration (Wang et al., 2008). Although we did not detect Collagen types I–V or Fibrillin fibrils in the Xenopus foregut it is possible that they are present below the level of immunostaining. In addition we did not examine all collagen isoforms, of which there are many, and therefore we cannot rule out the possibility that they are present and regulating BMP signaling. Other ECM associated proteoglycans including Syndecans, Biglycan and Tsukushi have also been shown to modulate BMP signaling is some contexts (Moreno et al., 2005; Ohta et al., 2004; Olivares et al., 2009), raising the possibility that they might also be involved in the Xenopus foregut. Finally the BMP-binding proteins Tsg and Cv2 also associate with the ECM (Zakin and De Robertis, 2010) and they are both expressed in the Xenopus foregut. Future studies will examine whether they participate in Szl-Tll-FN regulation of BMP signaling. Consistent with this possibility Tsg−/− mutant mice exhibit a reduction in hhex expressing foregut endoderm (Petryk et al., 2004). Together with our data this suggests that the FN-rich ECM is required for the movement and/or accumulation of BMP-containing complexes in the foregut endoderm to robustly activate BMP-receptor signaling over a threshold necessary to maintain foregut gene expression; loss of the FN matrix in Szl morphants would prevent the necessary level of Bmp signals required by foregut progenitors.
It will be important in the future to determine whether Sfrp and Tll proteins regulate BMP signaling dynamics in the mammalian foregut. Mammalian Sfrp2, which is related to Szl and expressed in the early mouse foregut, can either inhibit or enhance BMP1 activity depending on the context (He et al., 2010; Kobayashi et al., 2009; Lee et al., 2006). Analysis of compound sfrp2, sfrp1 and sfrp5 mutant mice suggest that they are redundantly involved in gut development (Matsuyama et al., 2009) and although this has been attributed to Wnt-repression it remains to be tested whether they regulate BMP signaling in this context. Bmp1−/− mutants are perinatal lethal with ventral body wall defects, whereas Tll1−/− mutants and Tll1−/−; Bmp1−/− double mutants die by e13.5 with cardiac defects and Tll2−/− mice are viable with a mild muscle hypoplasia (Lee, 2008; Muir and Greenspan, 2011). None of these phenotypes have been definitively linked to disrupted BMP activity. The absence of an early foregut defects in Tll1−/−; Bmp1−/− double mouse mutants is similar to the bmp1/tll1 knockdown that we observed. We postulate that the lack of a foregut defect in the bmp1/tll1 deficient mouse and xenopus embryos may be due to redundant Tll2. Alternatively it is possible that bmp1/tll1 are indeed dispensable for foregut development, whereas too much Tll activity, as In Szl-MO disrupts FN-dependant BMP signaling and foregut progenitor maintenance.
A growing number of Sfrps can modulate BMP signaling (Bovolenta et al., 2008; Ploper et al., 2011), however in all previous reports they exert an anti-BMP activity by inhibiting Tll-mediated Chordin degradation, thus preserving Chordin repression of BMP. In contrast we show that in the foregut Szl exerts a pro-BMP activity. Our results are not consistent with reduced Chordin levels but rather implicate Szl-Tll regulation of the FN-rich ECM as the primary mode of BMP modulation. Our data argue that the loss of FN matrix in the Szl morphants is due to unrestrained Tll activity. Tll-proteases have not previously been reported to regulate FN matrix deposition and in vitro studies indicate that FN is not a native BMP1 substrate (Wermter et al., 2007). Western analysis of Szl depleted foreguts did not reveal FN degradation products, suggesting that elevated Tll activity inhibited FN fibrilogenesis rather than promoting degradation. Indeed bmp1/tll1 knockdown resulted in increased disorganized FN matrix deposition. It is unclear how exuberant Tll protease activity negatively regulate the FN matrix as they generally promote ECM deposition by proteolytic maturation of various ECM related precursor proteins including, perlecan, laminin-332, lysl oxidase, probiglycan and the conversion of pro-collagens into mature collagen fibrils (Hopkins et al., 2007). A future goal will be to identify Tll substrates in the Xenopus foregut.
The fact that Sfrps can modulate both the BMP and Wnt pathways, either repressing or activating depending on the cellular context raises the possibility that they form a nexus for BMP-Wnt crosstalk. Consistent with this possibility Szl morphants also exhibit a reduction in Wnt signaling in the foregut. Moreover the related factor Sfrp5, acting at the same time in foregut development as Szl, is also required to maintain foregut progenitors, in this case by inhibiting posteriorizing Wnt signals. Future studies will address how Szl and Sfrp5 orchestrate BMP-Wnt crosstalk and whether regulation of the FN matrix can also impact Wnt signaling in the foregut.
In summary we have identified a previously unappreciated extracellular mechanism that regulates BMP signaling dynamics during foregut organogenesis. We show that the secreted frizzled related protein Szl and Tll proteases modulate BMP signaling through regulation of the FN matrix. These data may impact our understanding of ECM function and extracellular regulation of BMP signaling in diverse biological contexts.
Xenopus embryo manipulation, explant culture, and microinjections were performed as described previously (McLin et al., 2007). Embryos with clear dorsal-ventral pigmentation differences were selected for 16-cell stage injections targeting either the C1/D1 or C4/D4 cells that contribute to the foregut or hindgut, respectively. Protein injections into the closing blastocoel of the foregut were performed at stages 12, with either 40 nl of recombinant human BMP2 (5.8 μM; R&D Systems) in PBS + 0.1%BSA or PBS + 0.1%BSA as controls. Small molecule inhibitors were dissolved in DMSO and embryos were cultured from stages 11–20 with either 1% DMSO vehicle in 0.1XMBS or the following concentrations of inhibitors: 20 μM LDN193189 (Axon MedChem), 21 μM Apoptosis Inhibitor I (Calbiochem), 56 μM Apoptosis Inhibitor II (Calbiochem), or 40 nl of 1.2 μM InSolution™ Caspase Inhibitor I (Calbiochem) injected into the foregut.
All of the morpholino oligos used in this study have been previously published to generate specific loss-of-function phenotypes: szl-MO (10–20 ng) (Collavin and Kirschner, 2003; Lee et al., 2006), bmp2-MO (20 ng) (Reversade et al., 2005), tll1-MO (15 ng), bmp1-MO (15 ng) (Inomata et al., 2008), fn1-MO + fn2-MO (15ng each) (Davidson et al., 2006) whereas the standard negative control-MO (10–30 ng) had no effect. For rescue experiments, we injected a pCS2+wSzl-MT expression plasmid (200–500 pg) encoding a szl RNA with the MO target site mutated (Collavin and Kirschner, 2003).
In situ hybridizations, Immunohistochemistry and Western blots were performed as previously described (Li et al., 2008). Antibodies included anti-phospho-Smad1/5/8 antibody (1:500, Cell Signaling Technology), anti-total Smad1 (1:500, Invitrogen), mouse anti-tubulin (1:700, Sigma) anti-active Caspase 3 (1:250, BD Pharmingen), anti-Fibronectin 4H2, (Ramos and DeSimone, 1996)(1:500, a gift from Lance Davidson) and goat anti-Rabbit HRP (1:5000, Jackson IR) for westerns, and goat anti-rabbit-cy5 (1:300; Jackson IR) for immunostaining. Alexa-647 topro-3 DNA dye was used to counter stain nuclei. Quantitation of western blots was performed using Image-J.
To quantitate pSmad1 immunostaining 80-micron mid-sagittal confocal Z-projections were obtained from the foregut region of stage 18–20 embryos. The raw images were then analyzed with Amaris™ software. Nuclear topro-3 staining was used to define the nuclear volumes and surrounding cytoplasmic regions. The average pixel intensity in the pSmad1 channel was then measured in all foregut nuclei (~400/foregut) and adjacent cytoplasmic regions. The foregut was sub-divided into different sub-regions based on the distance in number of nuclei from the mesoderm (Figure 5B; white dash lines) and the mean nuclear/cytoplasmic ratio of pSmad1 intensity in each sub-region was determined from at least 4 embryos. All experiments were repeated 3 times with similar results. A mixed effects model was used to statistically assess mean nuclear/cytoplasmic pSmad1 intensity in each sub-regions and differences in least square mean estimates between conditions were evaluated with T-tests at α=0.05 using SAS (v9.3, SAS Institute, Cary, NC).
For each of three biological replicates, 100 stage 15 foregut explants were dissected from sibling embryos. In 50 explants the mesoderm was left intact and for the other half the mesoderm was removed. Both were cultured to stage 23, when the mesoderm was removed from the intact group. Total RNA (~20 μg) was extracted using Trizol (Invitrogen) and RNAeasy columns (Qiagen) and 1 μg of RNA was used to probe Affymetrix Xenopus laevis Genechips v1. GeneSpring 7.1 software (Silicon Genetics) was used for RMA preprocessing, data normalization, clustering and filtering, GEO accession number GSE38654.
We thank Drs. Collavin and Davidson for reagents. We are grateful to Jim Wells, Sang-Wook Cha, Lance Davidson, Robert Angerer, Lynne Angerer, David Kozlowski and members of the Endoderm club for helpful discussions. This project was supported by in part by a CCHMC Procter Scholarship and K08 HL105661 to APK, a T32 HD07463 award to ETS, PHS Grants P30 DK078392 and by DK070858 to AMZ.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.