Tailbud-derived Bmp4 drives proliferation and inhibits maturation of zebrafish chordamesoderm

In zebrafish, BMP signaling acts progressively during gastrulation to establish mesodermal cell identity. Mesoderm is specified early in gastrulation and contributes to tissues that can be roughly categorized as dorsal or ventral in nature. The specification of ventral fate involves Bmp2b/7 signaling and activation of downstream transcription factors such as SMAD1/5/8, vent1, ved and vox(Kimelman, 2006; Schier and Talbot, 2005; Yamamoto and Oelgeschlager,2004). SMAD1/5/8 activity on the presumptive ventral side of the blastula activates the expression of ventral genes, whereas vent1,ved and vox repress the expression of genes that confer dorsal identity such as the BMP antagonist chordin(Imai et al., 2001; Kawahara et al., 2000; Melby et al., 2000). Establishment of the DV axis is followed by a second phase of BMP signaling during mid to late gastrulation, when the interaction of BMP antagonists and agonists is thought to establish a BMP activity gradient(Kimelman, 2006; Schier and Talbot, 2005). According to this model, high, intermediate and low levels of BMP activity in the mesoderm specify ventral, intermediate and dorsal fate, respectively.

Low levels of BMP activity direct cells surrounding the dorsal organizer to become axial mesoderm, whereas their positions within this tissue influence their exposure to other activity gradients(Schier and Talbot, 2005; Stemple, 2005). For example,Nodal activity is crucial for the process of further specification of axial mesoderm. High levels of Nodal, received by cells deep in the organizer,specify cells to become prechordal plate(Gritsman et al., 2000; Saude et al., 2000). Low levels of Nodal activity found in the superficial organizer specify cells to become chordamesoderm, the antecedent of the notochord(Gritsman et al., 2000; Saude et al., 2000).

The molecular events surrounding the transition of chordamesoderm into mature notochord are not fully understood, although there are several defining features of the process (Stemple,2005). Differentiated notochord cells acquire large vacuoles that allow the tissue to provide structural support to the embryo. Coupled with this, genes expressed in chordamesoderm are extinguished as the tissue matures. These include the sonic hedgehog and indian hedgehog homologs, shha and ihhb, respectively, as well as the extracellular matrix (ECM) gene collagen 2a (col2a1a)(Currie and Ingham, 1996; Krauss et al., 1993; Yan et al., 1995). Mutagenesis screens and studies examining ECM members have begun to elucidate the chordamesodermal transition into mature notochord. Embryos lacking Laminin 1 subunits (bashful, grumpy and sleepy)(Parsons et al., 2002),Laminin α4/5 (Pollard et al.,2006) and Collagen15a1(Pagnon-Minot et al., 2008)continue to express chordamesodermal markers after expression has been extinguished in wild-type embryos, suggesting that notochord differentiation is impaired. However, given their nature, these proteins are unlikely to play instructive roles in notochord differentiation. The signal responsible for promoting notochord development has yet to be defined.

Although it is clear that initial specification of chordamesoderm is reliant upon the absence of BMP from the dorsal organizer, subsequent proliferation and differentiation of this tissue may depend upon it. Posterior to the notochord lies the chordoneural hinge (CNH), a stem cell pool that contributes to notochord, floor plate and tailbud mesoderm(Cambray and Wilson, 2002; Cambray and Wilson, 2007; Charrier et al., 1999; Davis and Kirschner, 2000; Kanki and Ho, 1997). Several agonists of BMP activity, including the BMP ligand ADMP, are expressed in axial territories during gastrulation and in the CNH during segmentation stages (Dickmeis et al., 2001; Lele et al., 2001). Thus, BMP signaling may play an important role in the development of axial mesoderm after specification of the DV axis. However, the role of BMPs in DV establishment masks their later functions. For example, manipulation of ADMP levels within the embryo causes DV patterning phenotypes that are difficult to resolve from tissue-specific defects (Lele et al., 2001; Willot et al.,2002).

In contrast to the known role of BMP activity in DV establishment, prior studies have not addressed the role of the BMP activity gradient in relation to axial mesoderm. In this study, we have identified both a requirement for BMP activity in the development of this tissue, as well as the signaling molecule responsible for notochord differentiation. We have been able to address these issues through the inactivation of two BMP antagonists, fstl1 and fstl2 (Dal-Pra et al., 2006), that act redundantly to inhibit BMP activity beginning at late gastrulation. Inactivation of these genes, as well as of Bmp4 and BMP signaling, reveals that Bmp4 promotes the proliferative capacity of notochord and CNH cells. In the absence of Bmp4, chordamesoderm fails to proliferate and the notochord differentiates prematurely. Our results illustrate the requirement of fstl1 and fstl2 in late gastrulation to maintain proper BMP activity levels, which are necessary for the development and timely differentiation of dorsal structures.

Tg(hsp70l:dnBmpr-GFP)^w30 (tBR) transgenic zebrafish were obtained from the Kimelman laboratory (University of Washington, Seattle). This transgenic line contains a truncated Type I BMP receptor containing GFP in place of the kinase domain under the control of a heat shock promoter (Pyati et al.,2005). tBR embryos were heat-shocked at 37°C for 1 hour at the time indicated according to Pyati et al.(Pyati et al., 2005). Where appropriate, wild-type embryos were heat-shocked under the same conditions to serve as controls.

The following probes were used: admp(Lele et al., 2001), bmp4 (Nikaido et al.,1997), col2a1a (Yan et al., 1995), eve1(Joly et al., 1993), flh (Talbot et al.,1995), fstl1 (Dal-Pra et al., 2006), fstl2(Dal-Pra et al., 2006), gsc (Schulte-Merker et al.,1994), ihhb (Currie and Ingham, 1996), ntl(Schulte-Merker et al., 1992), shha (Krauss et al.,1993) and spt(Griffin et al., 1998).

Full-length fstl1/2 cDNA was cloned into pCS2+. Capped mRNAs were transcribed using RNA polymerase in vitro transcription kits (mMESSAGE mMACHINE; Ambion). Approximately 100 pg of fstl1 or fstl2mRNA was injected into one- and two-cell stage embryos.

Translation-blocking morpholinos are overlap with 22 nucleotides at the 5′ end of the previously published MO sequence(Dal-Pra et al., 2006). fstl1 and fstl2 translation-blocking MOs were as follows: fstl1, 5′-GCAGCTGCATGGACAGAGTGAAAAC-3′; fstl2,5′-CACGGGTAAACACCGAAACATCATT-3′. fstl splice-blocking morpholinos were as follows: fstl1,5′-CAGACTTACCTTCACATTGTCCGTC-3′; fstl2,5′-AAATTAAAGCTCACCATCACAAGTC-3′. To ensure MO function, RNA was isolated from injected embryos and RT-PCR was performed using primers designed around intron-exon boundaries. Those sequences are as follows: fstl1,5′-TAATCATCCAGTCTGTGGCAGTAAT-3′ and 5′-CTTGGGCTGTTGATGAT-3′; fstl2,5′-GGTCTGCACCGCCATGACTTGT-3′ and 5′-ACACGGCGGGTCCACTCCTC-3′. For single morpholino injections, 8 ng of MO were injected into one- and two-cell stage embryos. For double morpholino injections, ∼6 ng of each MO was injected. The control morpholino sequence used was 5′-CCTCTTACCTCAGTTACAATTTATA-3′. Where appropriate, control MO was injected into embryos that were subsequently heat-shocked at 37°C for 1 hour. The bmp4 splice-blocking morpholino used has been previously characterized(Chocron et al., 2007).

Labeling with PSMAD1/5/8 was as previously described(Rentzsch et al., 2006). Labeling with Myf5 and F59 were as previously described(Hammond et al., 2007; Topczewska et al., 2001). The primary antibodies were against Myf5 (anti-Myf5 recognizes Myod1 in zebrafish;Santa Cruz, C-20) at 1:50, F59 (anti-MyHC; University of Iowa Developmental Studies Hybridoma Bank) at 1:10, 4D9 (anti-Engrailed; University of Iowa Developmental Studies Hybridoma Bank) at 1:20, Prox1 (Angiobio) at 1:500 and P-SMAD1/5/8 (Chemicon International) at 1:100. Alexa-conjugated Phalloidin(Molecular Probes) was used at a 1:50 dilution to label F-actin. Appropriate Alexa Fluor (Molecular Probes) secondary antibodies were used.

Protein extracts were prepared using standard procedures(Westerfield, 1994). Anti-P SMAD1/5/8 and Anti-P SMAD2 (Cell Signaling Technology) were used at a concentration of 1:1000.

Twelve-somite embryos were incubated with a solution of 425 μl BrdU labeling reagent (Roche Applied Science) and 75 μl DMSO for 45 minutes at 6°C. They were then washed in embryo medium for 30 minutes at 28.5°C and fixed with 4% PFA in PBS. BrdU incorporated cells were detected using Alexa Fluor 594 goat anti-mouse IgG at 1:500.

Surgical transplantation and extirpation experiments were performed as described (Saude et al.,2000). For ventral margin extirpation, ∼100 cells were removed by suction from the ventral margin of 80% epiboly embryos. For cell transplantation, donor embryos were injected with a 5% solution of fluorescein dextran (10K MW; Molecular Probes) at one- and two-cell stage embryos. Approximately 30 ventral margin cells taken from shield-stage donors were transplanted into the dorsal organizer of an equivalently staged host. tBR embryos were obtained from an incross of homozygous tBRparents. Both donor and host embryo were then heat-shocked after transplantation to maximize BMP signaling inhibition according to previous transplantation protocols (Pyati et al.,2005).

Published protocols were followed with slight modifications(Stern et al., 2005). Embryos were treated with a combination of 150 μM aphidicolin and 20 mM hydroxyurea in 4% DMSO between 80% epiboly and bud stage. Cell proliferation was assayed through BrdU incorporation at stages during and after application.

Human Bmp2/7 and zebrafish Bmp4 protein-coated beads (R&D Systems) were prepared by overnight incubation of 45 μm polystyrene beads (Polysciences)in a 500 μg/ml solution of recombinant Bmp2/7 or Bmp4 protein (R&D Systems) in PBS. Before implantation, beads were rinsed three times for 10 minutes in PBS. Control beads were loaded with 500 μg/ml BSA.

DAPI-positive cells were quantified and compared by one-way ANOVA, followed by a two-tailed, equal variance t-test.

To explore the significance of BMP activity after the onset of gastrulation, we further characterized the roles of two previously identified follistatin-like genes (fstl1 and fstl2)(Dal-Pra et al., 2006). These genes are structurally and functionally related to Follistatin (Fst), a known inhibitor of Bmp4 and ADMP (Dal-Pra et al., 2006; Dosch and Niehrs,2000; Fainsod et al.,1997; Iemura et al.,1998). Although fstl2 mRNA can be detected in cleavage-stage embryos, fstl1 expression is initiated during late gastrulation in dorsal structures. Unlike fst, which is expressed only in anterior paraxial mesoderm (Bauer et al., 1998), fstl1/2 share largely overlapping expression patterns throughout axial and paraxial mesoderm during late gastrulation and segmentation stages (see Fig. S1 in the supplementary material) (see also Dal-Pra et al.,2006). Fstl2 plays a minor role in BMP antagonism during DV axis specification, as depletion of Fstl2 in a Chordin/Noggin1-deficient embryo slightly increases the severity of the ventralization phenotype observed in Chordin/Noggin1 double morphants. Loss of Fstl2 in either Chordin or Noggin1 single morphant embryos does not enhance the ventralization phenotype(Dal-Pra et al., 2006). As shown by these authors, we observed no DV patterning defects when either Fstl protein was reduced by itself or in combination with each other. The most pronounced morphological defect observed in Fstl1/2 morphants was an enlarged undulating notochord that became less prominent as somitogenesis continued(e.g. see Fig. S2 in the supplementary material; Figs 3 and 6). In addition, we found an increase in the levels of molecular markers that indicated an increase in BMP activity (Figs 1 and 2).

The domain of the tailbud, which is dependant on BMP activity(Holley, 2006; Szeto and Kimelman, 2006), was increased in Fstl morphant embryos and enlarged to a greater degree when Fstl1 and Fstl2 were knocked down in concert. The expression domain of spadetail (spt), a T-box domain transcription factor required for trunk segmental identity(Griffin et al., 1998), was increased in Fstl1/2 morphant embryos (Fig. 1A,B; see Fig. S2 in the supplementary material). The domain of expression of BMP agonists, which are positively regulated in response to BMP activity, was expanded in the tailbud in a similar fashion(Yamamoto and Oelgeschlager,2004). This included the expression domains of bmp4 and the transcriptional target of BMP activity, eve1(Fig. 1E,F,M,N; see Fig. S2 in the supplementary material) (Pyati et al.,2005), as well as admp in the posterior notochord(Fig. 1Q,R). We also observed an increase in phosphorylated (P) SMAD 1/5/8 in the tailbud, which is activated in response to BMP signaling(Fig. 1I,J)(Yamamoto and Oelgeschlager,2004). Western blots against PSMAD1/5/8 indicated that BMP activity is increased in Fstl1/2 morphants beginning at late gastrulation. Prior to 80% epiboly, we observed no change in PSMAD1/5/8 levels (not shown). At the end of gastrulation, we observed an increase in BMP activity that was present at every stage assayed, through 18 somites(Fig. 1U). Follistatin also binds Activin ligands with high affinity, which initiates intracellular Nodal cascades (Harrington et al.,2006; Nakamura et al.,1991). We observed no change in PSMAD2 levels, a target of Nodal signaling (reviewed by Kitisin et al.,2007), in Fstl1/2 morphants at any stage assayed(Fig. 1U).

To assess morpholino specificity, we compared the phenotypes generated using non-overlapping translation- and splice-blocking morpholino sequences. The tailbud and notochord phenotypes observed using translation-blocking morpholinos were indistinguishable from those observed using splice-blocking morpholinos (see Fig. S2 in the supplementary material). In general, morphant phenotypes were reliable and were observed in greater than 90% of embryos(n>500). Furthermore, we attempted to rescue the tailbud phenotype observed in Fstl1/2 morphants. Injection of either fstl1 or fstl2 mRNA caused a reduction in the tailbud domain, consistent with a decrease in BMP activity and a role for Fstl1 and Fstl2 as BMP antagonists(Fig. 1C,G,K,O,S) (also data not shown). Co-injection of either fstl1 or fstl2 mRNA and fstl1/2 morpholinos was sufficient to rescue the Fstl1/2 double morphant phenotype, demonstrating the specificity of our morpholinos and a functional redundancy between Fstl1 and Fstl2(Fig. 1D,H,L,P,T) (also data not shown). As all of the morphant phenotypes were comparable with each other,the Fstl1/2 morphant phenotypes illustrated here are of embryos injected with fstl1/2 splice-blocking morpholinos.

The expression of fstl1/2 along the dorsal midline and the notochord defects observed in fstl1/2 morphants suggest a function in axial development. We observed no changes in the expression of axial markers in Fstl1/2 morphants until late gastrulation, consistent with the temporal rise in PSMAD1/5/8 levels. At 70% epiboly, goosecoid (gsc)expression in prechordal mesoderm(Schulte-Merker et al., 1994)and floating head (flh), no tail (ntl) and admp expression in chordamesoderm(Dickmeis et al., 2001; Schulte-Merker et al., 1992; Talbot et al., 1995) of Fstl1/2 morphants were comparable with controls(Fig. 2A,B,H,I,O,P,V,W). At 90%epiboly, the domain of prechordal mesoderm was unchanged in Fstl1/2 morphant embryos (Fig. 2C,D), but the domain of chordamesoderm was expanded anteriorly(Fig. 2J,K,Q,R,X,Y).

As changes in chordamesodermal gene expression in Fstl1/2 morphants appeared to be a result of increased BMP activity, we wished to determine the effects of lowering BMP activity on this tissue. We were able to conditionally inactivate BMP activity through the use of a transgenic zebrafish line carrying a truncated BMP-type I receptor fused to GFP under the control of a heat-shock promoter (abbreviated tBR)(Pyati et al., 2005). Heat shocking these embryos for 1 hour at 37°C blocks BMP effector expression for at least 2 hours and is capable of inducing severe dorsalization phenotypes depending on the timing of heat shock(Pyati et al., 2005). To determine the temporal effectiveness of this transgene, tBR embryos were heat-shocked during late gastrulation in hourly intervals and stained with PSMAD1/5/8 antibodies (see Fig. S3 in the supplementary material). Between 1 and 3 hours post heatshock, PSMAD1/5/8-positive cells were not detectable in tBR embryos, suggesting that heatshock attenuates BMP activity over this time range during early somitogenesis.

As we observed an expanse in chordamesoderm and an increase in PSMAD1/5/8 levels during late gastrulation in Fstl1/2 morphants, we heat-shocked tBR embryos at 80% epiboly to examine the consequence of blocking BMP activity on chordamesoderm. In these embryos, the domain of prechordal mesoderm was unchanged relative to heat-shocked controls(Fig. 2E), whereas chordamesoderm was absent anteriorly (Fig. 2L,S,Z). Knockdown of Bmp4 also resulted in a reduction of chordamesoderm (Fig. 2M,T,AA),but not of prechordal mesoderm (Fig. 2F). The reduced domain of chordamesoderm observed when Fstl1,Fstl2 and Bmp4 were knocked down together resembled that of Bmp4 morphant embryos (Fig. 2N,U,BB).

Using riboprobes against shha and ntl, we examined the effect of manipulating BMP activity on the size of the notochord and CNH. The size of the CNH in Fstl1/2 morphant embryos was expanded when compared with controls (Fig. 3A,B,K,L). Sections through the trunk between somites 5-10 revealed that the diameter of the notochord was also increased in these embryos; the mean number of cells populating the notochord was increased by 63% when compared with controls(Fig. 3F,G) (4.1 cells versus 6.5 cells, n=10; P<0.001).

In contrast to the expansion of the notochord and CNH observed in Fstl1/2 morphant embryos, attenuation of Bmp4 activity caused a reduction in both the diameter of the notochord as well as the size of the CNH. The axial phenotype observed when Bmp4 was knocked down largely resembled that of tBRembryos that were heat-shocked at 80% epiboly(Fig. 3C,E,M,O), with the mean number of cells populating the notochord decreased by ∼41%(Fig. 3H,J) (4.1 cells versus 2.4 cells and 2.9 cells, n=10; P<0.001). The difference in notochord cells of tBR embryos and Bmp4 morphants was not significant (P>0.05). In both tBR embryos heat-shocked at 80% epiboly and Bmp4 morphants, we observed the formation of ectopic tail structures. These observations are consistent with roles assigned to BMP activity in patterning ventroposterior tissues during late gastrulation(Pyati et al., 2005; Stickney et al., 2007).

To determine the temporal requirements of axial cells for BMP activity, we heat-shocked tBR embryos at hourly intervals beginning at 80%gastrulation, corresponding to the time at which increased PSMAD1/5/8 levels were observed in Fstl1/2 morphants. The size of the notochord and CNH in embryos heat-shocked at bud stage was comparable to those of embryos heat-shocked at 80% epiboly (not shown). Embryos heat-shocked at the three-somite stage yielded a notochord diameter and CNH domain that were indistinguishable from heat-shocked controls(Fig. 3D,I,N) (4.1 cells versus 4.3 cells, n=10; P>0.05).

We wished to address the underlying cause of the increase in chordamesoderm. Because chordamesoderm transfates to paraxial mesoderm in flh mutants (Halpern et al.,1995), we reasoned that an expanse in chordamesoderm may occur at the expense of paraxial mesoderm. However, the analysis of deltaC(dlC), myod1 and paraxial protocadherin(papc) expression revealed no changes in paraxial territory of Fstl1/2 morphants or in tBR embryos heat-shocked at 80% epiboly (see Fig. S7 in the supplementary material) (also data not shown).

Because our results did not suggest that fate assignment was disrupted in Fstl1/2 morphants, we addressed whether the increase in chordamesoderm was due to increased cell proliferation by assaying BrdU incorporation. In 14-somite controls, we observed PSMAD1/5/8 and BrdU antibodies co-labeling notochord cells (Fig. 4A-A″). In addition to labeling these cells in Fstl1/2 morphants, cells of the CNH were positive for PSMAD1/5/8 and BrdU, and axial cells that were BrdU-positive were also positive for PSMAD1/5/8 (Fig. 4B-B″). Axial cells of Fstl1/2 morphants that were BrdU positive were also found in positions more anterior than in controls (not shown). By contrast, heat-shocking tBR embryos at 80% epiboly or bud stage resulted in loss of BrdU incorporation in both endogenous and ectopic axial structures (Fig. 4C-C″) (also data not shown). We observed no changes in cell proliferation when we heat-shocked tBR embryos at the 3-somite stage(not shown). The presence of PSMAD1/5/8-positive axial cells in Bmp4 morphants(Fig. 4D-D″) indicates that axial cells receive signaling from multiple BMP ligands, and also that loss of Bmp4 does not affect signaling of other BMPs, consistent with previous reports (Stickney et al.,2007). Nevertheless, reduction of Bmp4 levels was sufficient to inhibit axial proliferation; cell proliferation in Bmp4 morphants resembled that of tBR embryos heat-shocked at 80% epiboly. This suggests that BMP activity, particularly Bmp4, acts prior to the 3-somite stage to promote chordamesodermal proliferation.

The ventral margin acts as the major secreting center of BMPs during gastrulation (Agathon et al.,2003; Schier and Talbot,2005; Yamamoto and Oelgeschlager, 2004) and is in close proximity to dorsal cells at the end of gastrulation. However, because BMP signaling initiates positive-feedback mechanisms in target cells(Yamamoto and Oelgeschlager,2004), an alternate source of BMP ligands might initiate from PSMAD1/5/8-positive chordamesoderm. To demonstrate that BMP signals originate from the ventral margin, we removed ∼100 cells from the ventral side of the embryo at 80% epiboly and examined the effects on chordamesoderm. Embryos lacking most of their ventral margin resembled Bmp4 morphants; the sizes of ventral fin and ventral somitic mesoderm were reduced (see Fig. S4A in the supplementary material). Both the domain of chordamesoderm as well as cell proliferation in this tissue was reduced, resembling that of Bmp4 morphants and tBR embryos heat-shocked at 80% epiboly (see Fig. S4B-E″ in the supplementary material). This suggests that chordamesoderm possesses a requirement for ventrally derived BMP ligands during late gastrulation.

In addition to notochord size, we wished to determine whether alteration of BMP activity affects the timing of notochord differentiation. In both tBR embryos heat-shocked at 80% epiboly and Bmp4 morphants, the expression of markers indicative of chordamesoderm was shifted such that they resembled expression patterns seen in more mature notochord. At the 9-somite stage, expression of ihhb, which can be detected throughout the entire notochord in control embryos, was maintained only in the caudal notochord of embryos lacking Bmp4 (Fig. 5A-C). Likewise, shha expression was lost from the notochord but maintained in floor plate and the diencephalon in 24 hpf tBR embryos, Bmp4 morphants and embryos lacking ventral margin(Fig. 5E-G,I-K; see Fig. S3G I the supplementary material). These results were confirmed by loss of ptc1 expression from the trunk and tail, where the expression pattern of 24 hpf embryos lacking Bmp4 resembled 36 hpf controls(Fig. 5M-O,DD; see Fig. S4H in the supplementary material). col2a1a expression was absent from the floor plate of 36 hpf tBR embryos heat-shocked at 80% epiboly and Bmp4 morphants, and more closely resembled the expression observed in 48 hpf controls (Fig. 5Q-S,GG).

Conversely, in Fstl1/2 morphants, chordamesodermal gene expression was prolonged when compared with controls, resembling that of embryos at younger developmental stages. Expression of ihhb, which is restricted to the caudal notochord in 18-somite stage embryos, was maintained in the rostral notochord of Fstl1/2 morphant embryos (Fig. 5U,V). We also observed shha expression in vacuolated notochord cells and ptc1 expression in axial and paraxial tissues of the trunk of 36 hpf Fstl1/2 morphants, where they are not expressed in controls (Fig. 5X,Y,AA,BB,DD,EE). col2a1a expression was maintained in the notochord of 48 hpf Fstl1/2 morphants(Fig. 5GG,HH), resembling expression of 24 hpf controls (Yan et al.,1995). The gene expression patterns observed in Fstl1/2 morphants are consistent with a delay in transitioning into differentiated notochord, as is observed in embryos lacking Col15a1(Pagnon-Minot et al., 2008)and Laminin1/4/5 (Parsons et al.,2002; Pollard et al.,2006).

As cells populating the notochord become vacuolated during differentiation,we examined the morphology of the notochord in embryos with altered BMP activity. In 24 hpf controls, notochord cells swell and become vacuolated in an anterior to posterior fashion (Fig. 6A,C,D). Although we did not observe any changes in anterior cell morphology at the level of the hind yolk extension in Fstl1/2 morphants, cells of the caudal notochord at the level of the cloaca were smaller than those observed in control embryos (Fig. 6B,E,F). Conversely, tBR embryos heat-shocked at 80%epiboly and Bmp4 morphants exhibited caudal notochord cells at the level of the cloaca that became vacuolated well before those observed in control embryos (Fig. 6G-J). Precocious vacuolation of notochord cells was also observed in embryos lacking ventral margin (see Fig. S4F in the supplementary material). This suggests that prolonged BMP activity delays the chordamesodermal transition into mature notochord. In the absence of Bmp4, the notochord differentiates prematurely.

Alterations in BMP activity levels influence both the proliferative properties of chordamesoderm, as well as the timing of its differentiation. To determine whether the proliferative capacity of chordamesoderm influences the timing of its maturation, we used small molecule inhibitors of the cell cycle(Stern et al., 2005) to block cell proliferation during the window in which BMP activity influences the proliferative capacity of chordamesoderm. We found that treating embryos from 80% epiboly to bud stage with a combination of hydroxyurea and aphidicolin effectively blocked cell proliferation during late gastrulation (see Fig. S5A,B in the supplementary material) (also data not shown). Removing the inhibitors at bud stage allowed proliferation to resume 2 hours after washing,although BrdU incorporating cells were not observed in chordamesoderm (see Fig. S5C-C″ in the supplementary material). The chordamesoderm defects observed in embryos treated with cell cycle inhibitors during late gastrulation resembled the defects observed in embryos lacking Bmp4. Expression of the chordamesodermal markers ihhb, shha and col2a1a resembled expression in Bmp4 morphants(Fig. 5D,H,L,P,T), and notochord cells at the level of the cloaca became vacuolated before those of controls (Fig. 6K,L).

In contrast to embryos with reduced BMP activity, chordamesoderm proliferation is increased and transition into mature notochord is delayed in Fstl1/2 morphants. By treating Fstl1/2 morphants with hydroxyurea and aphidicolin from 80% epiboly to bud stage, we were able to rescue the prolonged expression of chordamesoderm markers. Expression of ihhband shha was restricted to caudal notochord in these embryos,resembling comparably staged controls (Fig. 5W,Z,CC). Similarly, col2a1a expression was lost from the notochord and floor plate of 48 hpf Fstl1/2 morphants treated with cell cycle inhibitors, also resembling controls (Fig. 5II). This suggests that the proliferative state of chordamesoderm, which is established during late gastrulation, determines the timing of the chordamesoderm transition into mature notochord.

The presence of PSMAD1/5/8-positive cells(Fig. 4) suggests that chordamesoderm is directly influenced by BMP activity. To demonstrate that this tissue autonomously responds to BMP signaling, we removed ∼30 cells from the ventral margin of shield stage embryos and transplanted them into the dorsal organizer of an equivalently staged host(Fig. 6O). Upon placing wild-type cells into a wild-type host, these cells reliably populated the notochord, consistent with previous fate mapping and transplantation studies(Gritsman et al., 2000; Saude et al., 2000). Transplantation of ventral wild-type cells into tBR embryos heat-shocked after transplantation resulted in donor cell morphology resembling that of control embryos (94/132)(Fig. 6P,Q). Surprisingly, we found that these cells populated the entire notochord at the exclusion of endogenous tBR cells. Following the transplantation of ventral tBR cells into wild-type embryos, the tBR cells consistently failed to populate the notochord (0/151)(Fig. 6R,S). They did, however,populate adjacent paraxial mesoderm, contributing to myofibers and somitic cells. Thus, BMP activity biases cells in proximity of the dorsal organizer to become chordamesoderm.

To further explore the effects of Bmp4 activity on notochord fate,recombinant zebrafish Bmp4-soaked beads were implanted near the organizer of shield-stage embryos (Fig. 7A). This way, involuting cells would be exposed to Bmp4. Embryos exposed to ectopic Bmp4 exhibited an expanse of chordamesoderm at late gastrulation stages and an increase in notochord and CNH size in segmentation stage embryos when compared with embryos exposed to BSA-soaked beads(Fig. 7B,C,J,K). At 24 hpf, the notochord of these embryos was composed of a bilayer of cells rather than a monolayer observed in controls, and development of the floor plate was delayed(Fig. 7D,F,G,I). By 48 hpf, the floor plate became visible in Bmp4-treated embryos, although notochord cell morphology remained irregular (Fig. 7M,N). These cells did not fully vacuolate, and some cells remained arranged in a bilayer. tBR embryos exposed to Bmp4-soaked beads and heat-shocked at 80% epiboly did not exhibit enlarged notochords(Fig. 7E,H).

Because we observed that axial cells remained PSMAD1/5/8-positive when Bmp4 was knocked down, we wished to examine the effects of other BMPs on notochord development (see Fig. S6 in the supplementary material). Implantation of beads soaked with recombinant human Bmp2/7 did not noticeably alter the domain of chordamesoderm or notochord morphology when compared with controls.

Because delays in notochord maturation disrupt myotome patterning(Pagnon-Minot et al., 2008),we examined slow and medial fast muscle specification in Fstl1/2 morphant embryos. During early somitogenesis stages, Hedgehog (Hh) activity maintains myod1 expression in slow and medial fast muscle precursors(Barresi et al., 2000; Linker et al., 2003; Wolff et al., 2003). In embryos lacking Fstl1/2, myod1 expression was upregulated in the presumptive slow and medial fast territories of the somite (see Fig. S7A,B in the supplementary material). Immunostaining with antibodies against Myod1 confirmed a significant increase in Myod1-positive cells when compared with controls (see Fig. S7 in the supplementary material) (74.6 versus 120.4, n=5; P<0.001). Fstl1/2 morphant embryos also displayed an increase in horizontal myoseptum size, as evidenced by immunostaining with F59 antibodies that recognize mainly slow muscle (see Fig. S7G,H in the supplementary material) (Devoto et al.,1996). Likewise, the number of Engrailed-positive cells, which are induced in response to Hh activity (Wolff et al., 2003), was increased in Fstl1/2 morphants (see Fig. S7J,K,M,N in the supplementary material).

Attenuation of BMP activity at 80% epiboly had the converse effect on myofiber specification. Although myod1 expression was maintained in presumptive slow muscle cells in tBR embryos, a significantly smaller number of Myod1-positive cells were found in the presumptive medial fast muscle domain (see Fig. S7C,F in the supplementary material) (57.2 versus 74.6, n=5; P<0.001). Embryos failed to form an organized horizontal myoseptum, and exhibited a reduction in Engrailed-positive muscle pioneers and medial fast muscle fibers (see Fig. S7I,L,O in the supplementary material). The disruptions in myofiber specification suggest that the timing of notochord maturation influences Hh activity levels in adjacent paraxial mesoderm.

Members of the BMP family of signaling molecules are typically assigned roles in specification of ventral cell fate. Here, we describe a previously unappreciated function for BMP signaling in the promotion of dorsal fate. We demonstrate that Bmp4 patterns axial tissue by promoting chordamesoderm proliferation, ultimately influencing the size of the notochord and CNH. Two BMP antagonists, Fstl1 and Fstl2, modulate BMP activity to appropriate levels in chordamesoderm. In their absence, the chordamesoderm domain is expanded,and these cells do not transition into mature notochord in a timely manner. Conversely, in the absence of Bmp4 the chordamesodermal domain shrinks and notochord cells differentiate prematurely. As such, we demonstrate a requirement for BMP activity in the timing of notochord differentiation.

Models addressing BMP activity in embryogenesis have established the ventralizing role of BMP signaling in specifying DV fate(Kimelman, 2006; Schier and Talbot, 2005). Surprisingly, our study demonstrates a dynamic requirement of axial tissue for BMP activity during gastrulation. Following establishment of the DV axis,axial cells require tightly regulated levels of BMP activity for proper patterning of chordamesoderm and its precursor pool. Changes in axial requirements for BMP activity are consistent with two recent studies in zebrafish examining BMP and Nodal activity during zebrafish gastrulation. Tucker and colleagues (Tucker et al.,2008) demonstrated that BMP patterning of the DV axis occurs progressively in an anterior to posterior fashion as gastrulation proceeds. Hagos and Dougan (Hagos and Dougan,2007) demonstrated that cells responsive to Nodal signaling acquire their fate based upon the cumulative dose of Nodal signals the cell receives. This is a function of both the competence of the receiving cells as well as the strength of the signal. These studies suggest that cells possess different requirements for activity gradients as gastrulation proceeds. This is in agreement with the characteristic migration of two margins into proximity with the other during gastrulation: the ventral margin secreting BMP ligands and the dorsal organizer secreting BMP antagonists(Agathon et al., 2003). As the circumference of the margins decrease during epiboly, cells populating ventral and dorsal margins receive increasing amounts of factors secreted from the other margin. The question is, at what point does distance between the two margins influence the BMP activity gradient on the dorsal side of the gastrula?

Our results demonstrate that the ventral margin does indeed influence cells arising from the dorsal margin, beginning during the transition from mid- to late-gastrulation. Attenuation of total BMP activity at any developmental stage between 80% epiboly and the 3-somite stage arrests chordamesodermal proliferation, culminating in a reduction in the size of the notochord and CNH. Because attenuation of BMP activity after the 3-somite stage has no apparent effect on axial structures, we propose that BMP ligands emanating from the ventral organizer during late gastrulation signal presumptive axial cells in the dorsal margin.

It is possible, although unlikely, that axial cells require BMP activity prior to late gastrulation. Labeling embryos with PSMAD1/5/8 antibodies at hourly intervals has revealed that PSMAD1/5/8-positive cells are not observed in the dorsal gastrula until late gastrulation(Tucker et al., 2008). Therefore, the rise of PSMAD1/5/8 levels after 80% epiboly in Fstl1/2 morphants is consistent with a role in mediating BMP activity in axial tissue.

Nodal activity is crucial for both the induction of mesoderm and the specification of axial cell fate(Kimelman, 2006; Schier and Talbot, 2005; Stemple, 2005). The location of cells within the dorsal organizer dictates the level of Nodal activity that axial mesoderm receives. High levels of Nodal signaling specify deep gsc-expressing prechordal plate, whereas lower levels specify more superficial flh-expressing chordamesoderm(Gritsman et al., 2000; Saude et al., 2000). As Follistatin binds Activin ligands(Harrington et al., 2006; Nakamura et al., 1991), one possibility for the notochord phenotype observed in Fstl1/2 morphants is an alteration in Nodal activity. If this activity were significantly altered, one would expect a transfating between types of axial mesoderm, as has been observed in mutants lacking zygotic contribution of the Nodal co-receptor Oep(Gritsman et al., 2000). However, we observe no changes in the prechordal domain of Fstl1/2 morphants,as assayed by the expression of the Nodal-responsive gene gsc and PSMAD2 levels. Furthermore, the phenotypes observed in chordamesoderm and notochord of embryos with reduced BMP activity are complementary to those observed in Fstl1/2 morphants. Thus, the main role of Fstl1/2 is to modulate BMP activity.

Two models can explain the inability of prechordal cells to respond to BMP signaling: the location of these cells during gastrulation, or their competence to respond to Bmp4. As gastrulation proceeds, prechordal plate cells rapidly increase their distance from the ventral margin, distancing themselves from BMP-secreting cells. If prechordal cells were competent to respond to Bmp4, gsc expression would be altered in embryos receiving Bmp4-soaked beads. Therefore, a more plausible model is that prechordal plate cells are not competent to respond to Bmp4 during mid- to late-gastrulation,which would make them distinct from chordamesoderm precursors in the dorsal organizer.

Our results indicate that during late gastrulation through early somitogenesis, BMP signaling regulates the size of the chordamesodermal domain by establishing its proliferative state. In the absence of Bmp4,chordamesodermal cells prematurely transition into differentiated notochord. Conversely, chordamesodermal proliferation is increased in Fstl1/2 morphants and maturation of the notochord is delayed. BMP signaling may play a dual independent role to control the proliferative status of the CNH and chordamesoderm and to regulate the timely differentiation of notochord precursors. By inhibiting cell proliferation over the same window at which BMP activity influences the proliferative state of chordamesoderm, we were able to mimic the notochord phenotype observed in tBR embryos and Bmp4 morphants. Additionally, blocking cell proliferation inhibits the prolonged maintenance of chordamesodermal character observed in Fstl1/2 morphants. As heat-shocking tBR embryos after the 3-somite stage has no apparent effect on indicators of notochord differentiation, it is likely that BMP activity does not directly delay chordamesoderm maturation. Rather, the evidence presented suggests a single role for BMP signaling in controlling proliferation and favors a causal link between CNH proliferation and notochord maturation. We propose a model where the size of the axial precursor pool,established by Bmp4 and tightly controlled by two Fstl proteins, ultimately determines the maturation schedule of the notochord. Although we demonstrate the requirement of Fstl1 and Fstl2 in this process, both chordin(chd) and noggin1/2 (nog1/2) are also chordamesodermally expressed, with nog1/2 expression persisting in the CNH through notochord maturation(Furthauer et al., 1999; Miller-Bertoglio et al.,1997). Although their requirement in DV specification prevents simple morpholino-mediated inactivation(Dal-Pra et al., 2006; Hammerschmidt et al., 1996a; Hammerschmidt et al., 1996b; Schulte-Merker et al., 1997),we cannot rule out that either chd or nog1/2 might act synergistically with fstl1/2 in chordamesoderm development as they do in early DV patterning. Similar models for a role of Bmp4 signaling in balancing proliferation and differentiation have been proposed in other contexts. The influence Bmp4 exerts over hematopoietic stem cell (HSC)proliferation (Zhang and Li,2005) draws parallels with the effect of losing Bmp4 on the CNH. Bmp4 promotes the proliferative state of HSCs while delaying differentiation(Zhang et al., 2003). In the absence of Bmp4, HSCs exit the cell cycle and differentiate prematurely(Bhatia et al., 1999).

In addition to notochord contributions, cells from the CNH give rise to floor plate and tailbud mesoderm (Cambray and Wilson, 2002; Charrier et al., 1999; Davis and Kirschner, 2000). Our transplantation experiments suggest a particular requirement for BMP responsiveness in chordamesoderm:BMP-responsive cells preferentially populate the notochord, whereas cells unresponsive to BMP signaling contribute to paraxial and tailbud mesoderm. In addition to its role in proliferation and differentiation, differential levels of BMP activity appear to be crucial in relegating CNH cells to a particular fate.

Experiments by Agathon and colleagues(Agathon et al., 2003) have shown that the BMP-rich ventral margin in zebrafish posses tail organizer activity that contributes to non-axial derivatives in the tail, whereas the CNH, which is derived from the dorsal organizer, contributes to axial components. Their data, along with other studies(Beck et al., 2001; Gont et al., 1993; Pyati et al., 2005; Tucker and Slack, 1995)suggest that axial and non-axial tissues of the tailbud can develop independently under certain experimental conditions. However, as these two organizing centers move into close proximity at the end of gastrulation, they probably need to interact with each other to promote the coordinated growth,patterning and differentiation of the tailbud. Our studies have revealed a requirement for BMP signaling from the ventral margin on the dorsal organizer,providing a mechanism by which such coordination can be achieved and demonstrating previously unappreciated roles of BMP activity in the patterning and maturation of axial structures.

We thank M. Halpern for comments and criticisms on the manuscript, D. Kimelman and U. Pyati for the Tg(hsp70l:dnBmpr-GFP)^w30zebrafish line and M. Esterberg for assistance with statistics. The F59 and 4D9 antibodies developed by F. Stockdale and C. Goodman, respectively, were obtained from the Developmental Studies Hybridoma Bank developed under the auspices of the NICHD and maintained by The University of Iowa, Department of Biology, Iowa City, IA 52242. This work was supported by an NIH grant to A.F.(DC004701).