Anteroposterior and dorsoventral patterning are coordinated by an identical patterning clock

A bone morphogenetic protein (BMP) signaling gradient patterns cells along the embryonic dorsoventral (DV) axis (Little and Mullins, 2006). BMP signaling is abrogated dorsally by the secreted BMP antagonists Chordin, Noggin and Follistatin to generate low BMP signaling levels that allow neurectodermal tissues to develop. These and additional extracellular modulators generate the moderate and high levels of BMP signaling that specify lateral and ventral cell types, respectively (Langdon and Mullins, 2011).

In addition to DV patterning of axial tissues, the embryo is also patterned along the anteroposterior (AP) and left-right axes. Although patterning of each of these axes has been well studied, little is known about how patterning is coordinated between each of these axes. We previously showed that DV tissues are patterned temporally progressively from anterior to posterior (Tucker et al., 2008). It is known that AP tissues are also patterned temporally progressively from anterior to posterior (Gamse and Sive, 2000; Gamse and Sive, 2001; Stern et al., 2006). However, it is not known if the patterning of DV and AP tissues is coordinated. Here we investigate whether the temporal patterning of DV and AP tissues is coordinated by the same patterning clock or occurs independently.

AP patterning is mediated by fibroblast growth factor (FGF), Wnt and retinoic acid (RA) posteriorizing signals during gastrulation (Gamse and Sive, 2000; Maden, 2002; Schier and Talbot, 2005; Stern et al., 2006; Wilson and Houart, 2004). To allow anterior neural tissues to develop, FGF and Wnt signaling must be repressed while these signals actively specify more posterior neural tissue development (Erter et al., 2001; Houart et al., 2002; Kudoh et al., 2004; Kudoh et al., 2002; Ramel et al., 2005; Rhinn et al., 2005; Shimizu et al., 2005). RA signaling does not posteriorize the most anterior neurectodermal tissues, but acts to posteriorize more caudal neural tissue (Kudoh et al., 2002; Maves and Kimmel, 2005). In zebrafish, expression of FGF, Wnt and RA signaling components overlap in the marginal zone (Schier and Talbot, 2005). In addition, the expression of these AP patterning components in the ventral marginal zone overlaps with that of BMP signaling components, suggesting the potential interaction of DV and AP patterning pathways during gastrulation.

In Xenopus, a model has been proposed to coordinate DV and AP patterning in dorsal tissues through integration of FGF and Wnt signaling into the BMP signaling pathway. Following phosphorylation by the BMP receptor, Smad1 is subsequently phosphorylated by FGF/MAPK and Wnt/GSK3, which leads to its degradation (Fuentealba et al., 2007). Although degradation of phosphorylated (P) Smad1/5 is not evident in ventral regions along the AP axis of the zebrafish gastrula (Tucker et al., 2008), a mechanism independent of P-Smad1/5 degradation might integrate BMP signaling with MAPK and GSK3 on P-Smad1/5 to coordinate DV and AP temporal patterning.

Here we demonstrate that patterning of DV tissues along the AP axis is intimately coordinated with AP patterning during zebrafish gastrulation. When FGF, Wnt or RA signaling pathways were inhibited in combination with the temporal modulation of BMP signaling, we found that patterning of DV tissues by BMP signaling along the AP axis is coordinated with AP patterning by FGF, Wnt and RA. Furthermore, our results suggest that coordinated patterning is regulated in part by phosphorylation of the P-Smad1/5 linker region via FGF/MAPK. The coordinated patterning of AP and DV tissues allows a cell to adopt both DV and AP positional information simultaneously during gastrulation.

The mutant strain somitabun^dtc24 (sbn, smad5) (Mullins et al., 1996; Nguyen et al., 2000), which contains a missense mutation in the MH domain (Hild et al., 1999), was used. All progeny of sbn^dtc24 heterozygous females are mutant owing to the dominant maternal-effect nature of this mutation (Mullins et al., 1996). The transgenic lines Tg(hsp70:chordin) [Tg(hsp70:chd)] and Tg(hsp70:dkk1GFP) have been described (Stoick-Cooper et al., 2007; Tucker et al., 2008).

Zebrafish embryos were treated with 300 μM SU5402 (Mohammadi et al., 1997) (Calbiochem, TOCRIS) in E3 medium as described (Shimizu et al., 2005). RA (1 nM) and LiCl (300 mM) treatments were performed as described (Kudoh et al., 2002). DEAB (5 μM) treatment was performed as described (Maves and Kimmel, 2005). DMH1 (10 mM stock solution in DMSO; gift from C. Hong, Vanderbilt University) was diluted to 1 μM in E3 medium. As controls of DMH1, SU5402, DEAB and RA treatments, embryos were treated with equivalent concentrations of DMSO. Embryos retained their chorions during all treatments.

RNAs encoding synthetic human SMAD1 wild type (hSmad1WT) (250, 350 and 450 pg), SMAD1 mutated at the MAPK sites (hSmad1MM) (250, 350 and 450 pg) and SMAD1 mutated at the GSK3 sites (hSmad1GM) (350 pg) (Fuentealba et al., 2007) were synthesized using the mMessage mMachine Kit (Ambion). The RNAs were injected into one-cell stage embryos from an incross of homozygous Tg(hsp70:chd) fish or sbn^dtc24/+ fish. The sequences of smad5 morpholino 1 and morpholino 3 (smad5 MOs) are described (Schumacher et al., 2011). smad5 MO1 and MO3 (3 ng each) were injected into Tg(hsp70:chd) embryos at the one-cell stage. smad5 MOs and the above RNAs were injected using separate needles in double and single injections to control for the effectiveness of the MOs and RNAs.

In situ hybridization was performed as described using gbx1 (Rhinn et al., 2003), hoxb1a (Prince et al., 1998), hoxb1b (Alexandre et al., 1996), krox20 (Oxtoby and Jowett, 1993), no tail (Schulte-Merker et al., 1992), otx2 (Li et al., 1994), pax2.1 (pax2a) (Krauss et al., 1992) and six3 (Kobayashi et al., 1998) probes.

Anti-P-Smad1/5 and anti-actin western blots were carried out as described (Schumacher et al., 2011). Anti-total Smad5 western blot was performed using the same conditions. Primary antibodies used were P-Smad1/5 (1:500, Cell Signaling Technology) and total Smad5 (1:1500, GeneTex).

Immunostaining and confocal microscopy were performed as described (Tucker et al., 2008). Primary antibodies used were P-Smad1^MAPK [1:1250 (Fuentealba et al., 2007)], P-Smad1^GSK3 [1:1000 (Fuentealba et al., 2007)] and total Smad5 antibody (1:100, GeneTex).

We began by examining the temporal patterning of DV tissues along the AP axis by BMP signaling during zebrafish blastula and gastrula stages. We inhibited BMP signaling at a series of time points using the transgene Tg(hsp70:chd) to drive expression of the chordin BMP antagonist with a heat shock-inducible promoter (Tucker et al., 2008). This transgene can inhibit all BMP activity and P-Smad1/5 within 60 minutes after the initiation of a 1-hour heat shock (HS) (Tucker et al., 2008). Following the HS treatments, we then investigated how DV tissues were affected along the AP axis by examining the expression of genes that mark distinct neurectodermal regions along this axis.

We found that a 1-hour HS of Tg(hsp70:chd) embryos at the 1000-cell stage [mid-blastula, 3.0 hours postfertilization (hpf)] caused expression of the prospective forebrain marker six3 (Kobayashi et al., 1998) to expand fully to the ventral side (Fig. 1A,B), whereas HS at or after the high stage (3.3 hpf) showed normal localization of six3 expression (Fig. 1C-F). HS at sphere (4 hpf) and 40% epiboly (late blastula, 5 hpf) stages caused expression of the forebrain and mid-brain marker otx2 (Li et al., 1994; Mori et al., 1994) to expand to the ventral side, whereas HS at or after the 50% epiboly stage (5.3 hpf) led to normal otx2 expression (Fig. 1G-L) (Tucker et al., 2008). HS at sphere, 40% and 50% epiboly stages resulted in expansion of the rostral hindbrain marker gbx1 (Rhinn et al., 2003) fully to the ventral side (Fig. 1M-P), whereas expression was restricted to the dorsal side by HS at or after the shield stage (Fig. 1Q,R). Finally, expression of the caudal hindbrain marker hoxb1b (Alexandre et al., 1996) was expanded ventrally in embryos subject to HS at 40% epiboly, shield (6 hpf) and 65% epiboly (7 hpf) stages (Fig. 1S-V), whereas embryos with HS at or after 70% epiboly (7.5 hpf) displayed normal hoxb1b expression (Fig. 1W,X).

In Tg(hsp70:chd) embryos, P-Smad5 is greatly reduced 40 minutes after the start of HS and is absent by 60 minutes (Tucker et al., 2008). Thus, BMP signaling patterns each of the above AP tissue markers ∼60 minutes after initiation of HS. Taken together, neurectodermal tissues from prospective forebrain to caudal hindbrain are patterned progressively over time from late blastula to late gastrula stages.

Because the temporal patterning of AP tissues in vertebrates is known to occur progressively, we hypothesized that DV and AP patterning are coordinated by the same patterning clock. To test this hypothesis, we disrupted AP patterning by inhibiting or activating FGF, Wnt or RA signaling pathways and then determined how temporal patterning by BMP signaling was affected. If DV and AP patterning are coordinated, we would expect that alterations in AP patterning would effect similar alterations in the temporal patterning of DV tissues. For example, in anteriorized embryos, i.e. those showing a loss of posterior tissue with a concomitant expansion of anterior tissue (located in a more caudal position of the embryo; red dashed line in Fig. 2A, upper panels), the caudally expanded anterior tissue would be patterned with the normal anterior domain timing, rather than when the more caudal domain is normally patterned by BMP signaling (Fig. 2A, model 1). Conversely, if DV patterning instead proceeds by a distinct temporal patterning mechanism, then alterations in AP patterning will not affect the progressive temporal patterning of DV tissues by BMP signaling. For example, in anteriorized embryos, the caudally expanded anterior tissue would be patterned at a later time point when the more posterior tissue is normally patterned, i.e. the temporal progression of BMP patterning would not be altered by changes in AP patterning and would progress normally (Fig. 2A, model 2).

To examine whether DV patterning is coordinated with AP patterning, we inhibited FGF signaling, a posteriorizing signal, in combination with temporally inhibiting BMP signaling. FGF signaling was inhibited with SU5402, a specific FGF receptor inhibitor (Fürthauer et al., 2004; Mohammadi et al., 1997). Because FGF signaling also functions in DV patterning during blastula stages (Fürthauer et al., 2004; Maegawa et al., 2006), we treated embryos with SU5402 beginning at shield stage (early gastrula stage), which we found caused a specific AP patterning defect without altering DV patterning. Anteriorization was evident by examining the anterior neurectodermal marker otx2 in late gastrula. Inhibition of FGF signaling caused otx2 expression to expand posteriorly, but not ventrally (Fig. 2B,G) (Kudoh et al., 2004). Inhibition of BMP signaling by HS of Tg(hsp70:chd) embryos at sphere (4 hpf) and 40% epiboly (late blastula, 5 hpf) stages caused otx2 expression to expand ventrally (Fig. 2C,D) (Tucker et al., 2008). Embryos with HS at or after 50% epiboly (5.3 hpf) displayed normal otx2 expression (Fig. 2E,F) (Tucker et al., 2008).

We then examined whether BMP signaling patterns the caudally expanded anterior tissue when these anterior tissues are normally patterned or at a later time point corresponding to its new more caudal location. We treated Tg(hsp70:chd) embryos with SU5402 to anteriorize the tissue, followed by a series of HSs to determine when BMP signaling patterns the expanded anterior tissue. We found that the expanded anterior tissue (Fig. 2H-K) was patterned during the same temporal interval as the normally positioned anterior tissue (Fig. 2C-F). These results suggest that temporal DV patterning by BMP signaling along the AP axis is coordinated with AP patterning mediated by FGF signaling.

To examine whether DV patterning along the AP axis is coordinated with AP patterning by Wnt signaling, we inhibited Wnt signaling in combination with temporally inhibiting BMP signaling. Several chemical Wnt inhibitors that we tested had no effect on early embryos (data not shown); therefore, we used Tg(hsp70:dkk1GFP) (Stoick-Cooper et al., 2007) to inhibit Wnt signaling. Because Wnt signaling also functions in DV patterning before and after the mid-blastula stages (reviewed by Langdon and Mullins, 2011), Tg(hsp70:dkk1GFP) embryos were subject to HS at 50% epiboly (5.3 hpf) for 1 hour, which we found caused a specific AP patterning defect with minimal DV patterning effects. This HS condition led to the specific expansion of otx2 expression posteriorly (an anteriorization) without an expansion ventrally (compare Fig. 3B with 3G).

Since the Tg(hsp70:dkk1) line was used for inhibiting Wnt signaling, we could not also use the HS-driven Tg(hsp70:chd) to temporally inhibit BMP signaling. Therefore, to inhibit BMP signaling we used DMH1, a more selective inhibitor of the type I BMP receptors Alk2 and Alk3 (also known as Acvr1l and Bmpr1aa) than dorsomorphin (Hao et al., 2010). To determine the efficiency with which DMH1 inhibits BMP signaling, we examined P-Smad1/5 levels by western blot analysis at multiple time points after DMH1 treatment, similar to our previous analysis of the effectiveness of Tg(hsp70:chd) (Tucker et al., 2008). DMH1 treatment beginning at the 64-cell stage (2 hpf), 1000-cell stage (3 hpf), and high (3.5 hpf) and sphere (4 hpf) stages caused greatly reduced P-Smad1/5 by 40% epiboly (5 hpf), shield (6 hpf), 60% epiboly (6.5 hpf) and 65% epiboly (7 hpf), respectively (Fig. 3A). These results show that DMH1 significantly inhibits P-Smad1/5 levels ∼3 hours after treatment is initiated and that DMH1 can be used to temporally inhibit BMP signaling.

The temporal series of DMH1 treatments caused the same dorsalized phenotypes as the Tg(hsp70:chd) HS series, except with a 2-hour delay, consistent with the temporal effects of DMH1 treatment on P-Smad1/5 levels (3 hours to inhibit BMP signaling, Fig. 3A) compared with Tg(hsp70:chd) [<1 hour to inhibit BMP signaling after HS is initiated (Tucker et al., 2008)]. Treatment of embryos with DMH1 at the 64-cell stage, 1000-cell stage, high and sphere stages phenocopied the dorsalization caused by HS of Tg(hsp70:chd) embryos at sphere, 40% epiboly, 50% epiboly (5.3 hpf) and shield stages, respectively (compare Fig. 3C-F with Fig. 2C-F). Thus, DMH1 exhibits a 2-hour delay in its effectiveness compared with Tg(hsp70:chd), both in reducing P-Smad1/5 levels and in affecting DV patterning.

We then investigated how the temporal patterning of BMP signaling is affected when the embryo is posteriorized by Wnt signal inhibition. DMH1 treatment at the 64-cell and 1000-cell stages caused expansion of otx2 to the ventral side, whereas DMH1 treatment at or after high stage did not alter otx2 expression (Fig. 3C-F). Inhibition of Wnt signaling by a 1-hour HS at 50% epiboly of Tg(hsp70:dkk1) embryos caused otx2 to expand posteriorly (Fig. 3G). This posteriorly expanded otx2 expression domain (Fig. 3H-K) was patterned at the same temporal interval as the normally positioned otx2 domain (Fig. 3C-F).

To activate Wnt signaling and posteriorize the embryos, we treated them with LiCl, which blocks the function of the Wnt inhibitor GSK3 (Klein and Melton, 1996). LiCl treatment caused expression of hoxb1b, a marker of caudal hindbrain, to expand anteriorly (Fig. 3Q) (Kudoh et al., 2002). Tg(hsp70:chd) embryos that were subject to HS at shield (6 hpf) and 65% epiboly (7 hpf) stages displayed expansion of hoxb1b expression fully to the ventral side (Fig. 3M,N), whereas embryos with HS at or after 70% epiboly (7.5 hpf) displayed normal hoxb1b expression (Fig. 3O,P). We then investigated whether BMP signaling patterns the rostrally expanded hoxb1b tissue at the same time as the normal hoxb1b expression domain is patterned or at an earlier time point corresponding to its more rostral location. When we treated Tg(hsp70:chd) embryos with LiCl and performed a HS series, we found that the rostrally expanded hoxb1b domain was patterned during the same temporal interval as the normally positioned hoxb1b domain (Fig. 3, compare M-P and R-U). These data are consistent with the model that DV patterning along the AP axis is coordinated with AP patterning by Wnt signaling.

We examined patterning at a later stage to confirm these results. We analyzed six3 and krox20 (egr2 - Zebrafish Information Network) at the 6-somite stage as markers of anterior neurectoderm and rhombomeres 3 and 5, respectively (Fig. 3V). HS of Tg(hsp70:chd) embryos at sphere (4 hpf) and 40% epiboly (5 hpf) stages caused krox20 expression to expand fully to the ventral side (Fig. 3W,X). HS at 50% epiboly (5.3 hpf) also caused krox20 rhombomere 5 expression to fully expand, whereas the rhombomere 3 domain was restricted to the dorsal side (Fig. 3Y). HS at shield stage (6 hpf) now resulted in both rhombomeres 5 and 3 being restricted to the dorsal region (Fig. 3Z) (Tucker et al., 2008). Activation of Wnt signaling by LiCl caused krox20 expression to be shifted anteriorly and a loss of six3 expression at the 6-somite stage (Fig. 3AA) (Kim et al., 2002; van de Water et al., 2001). This anteriorly shifted posterior tissue (Fig. 3BB-EE) was patterned during the same temporal interval as the normally positioned posterior tissue (Fig. 3W-Z). In conclusion, these results confirm that DV patterning by BMP signaling along the AP axis is coordinated with AP patterning by Wnt signaling.

To examine whether DV patterning is coordinated with AP patterning mediated by RA signaling, we inhibited RA signaling, a posteriorizing factor. RA signaling was abrogated with 4-(diethylamino)-benzaldehyde (DEAB), a potent retinaldehyde dehydrogenase inhibitor (Russo et al., 1988). We examined expression of hoxb1a in rhombomere 4 at the 6-somite stage, which is expanded posteriorly by inhibition of RA (Fig. 4G-L) (Maves and Kimmel, 2005). Inhibition of BMP signaling by HS of Tg(hsp70:chd) embryos at sphere (4 hpf), 40% epiboly (5 hpf) and 50% epiboly (5.3 hpf) stages caused hoxb1a expression to expand ventrally (Fig. 4B-D). Embryos subject to HS at or after germ ring stage (5.7 hpf) showed normal expression of hoxb1a in the dorsal region (Fig. 4E,F). The caudally expanded rhombomere 4 tissue caused by DEAB treatment (Fig. 4H-L) was patterned with the same timing as the normally positioned rhombomere 4 (Fig. 4B-F).

To activate RA signaling and posteriorize the embryos, we treated them with RA. RA treatment caused expression of the caudal hindbrain marker hoxb1b to expand dramatically anteriorly (Fig. 4R) (Kudoh et al., 2002). In Tg(hsp70:chd) embryos with HS at shield and 65% epiboly (7 hpf) stages, hoxb1b expression was expanded fully to the ventral side (Fig. 4N,O), whereas embryos with HS at or after 70% epiboly displayed normal hoxb1b expression (Fig. 4P,Q).

We then investigated whether BMP signaling patterns the rostrally expanded hoxb1b tissue at the same temporal interval as the normal hoxb1b expression domain is patterned or at an earlier time point corresponding to its more rostral location. When we treated Tg(hsp70:chd) embryos with RA and performed a series of HSs, we found that the rostrally expanded hoxb1b domain was patterned during the same temporal interval as the normally positioned hoxb1b region (Fig. 4N-Q,S-V), rather than at an earlier interval corresponding to its new rostral location.

We also examined patterning at a later stage to confirm these results. We analyzed pax8 expression, a presumptive otic vesicle marker, at the 6-somite stage (Hans and Westerfield, 2007). HS of Tg(hsp70:chd) embryos at sphere stage caused a loss of pax8 expression (supplementary material Fig. S1B). HS at 40% epiboly (5 hpf) caused pax8 to be expressed in a small anterior ventral region (supplementary material Fig. S1C), whereas HS at 50% epiboly (5.3 hpf) resulted in expression that extended from the dorsal to ventral side (supplementary material Fig. S1D). Finally, HS at shield stage led to the normal dorsal localization of pax8 (supplementary material Fig. S1E). Activation of RA signaling causes the anterior limit of otic cells to extend rostrally to completely encircle the head (supplementary material Fig. S1F) (Hans et al., 2007; Hans and Westerfield, 2007). This rostrally expanded otic tissue was patterned during the same temporal interval as the normally positioned tissue (supplementary material Fig. S1C-E,H-J). These results indicate that the temporal patterning of DV tissues along the AP axis is coordinated with AP patterning by RA signaling.

Taking all of these results together, we conclude that DV patterning by BMP signaling along the AP axis is coordinated temporally with AP patterning mediated by FGF, Wnt and RA signaling.

A study in Xenopus showed that in dorsal tissues the linker region of Smad1 is sequentially phosphorylated by MAPK and GSK3, causing it to be degraded (Fuentealba et al., 2007). The degradation of P-Smad1 and loss of BMP signaling dorsally then allows neural tissue development and AP patterning by FGF and Wnt signaling. In zebrafish, the ventral P-Smad1/5 gradient appears to be stable along the AP axis until 75% epiboly, when prospective caudal hindbrain tissue is patterned (Fig. 5M-O) (Tucker et al., 2008). These data suggest that a degradation mechanism in ventral vegetal regions does not regulate the temporal function of BMP signaling in zebrafish. We tested whether a modified mechanism might be acting in which P-Smad1/5 function is inhibited by FGF/MAPK and Wnt/GSK3 through a non-degradation mechanism to temporally coordinate DV and AP patterning during gastrulation. We investigated the localization of GSK3-phosphorylated Smad1/5 (P-Smad1/5^GSK3) and MAPK-phosphorylated Smad1/5 (P-Smad1/5^MAPK) using antibodies specific for the phosphorylated forms of human SMAD1 (Fuentealba et al., 2007).

We tested the specificity of the human P-Smad1^MAPK and P-Smad1^GSK3 antibodies for zebrafish Smad5 and for their respective phosphorylation sites. The smad5 gene in zebrafish is expressed maternally and functions predominantly over smad1 for specification of ventral fates (Dick et al., 1999; Hild et al., 1999; Kramer et al., 2002). In the zebrafish Smad5 protein, the equivalent human MAPK and GSK3 phosphorylation sites in the SMAD1 linker region are present (supplementary material Fig. S2). In the zebrafish smad5 mutant somitabun^dtc24 (sbn^dtc24), we found that P-Smad1/5^MAPK and P-Smad1/5^GSK3 signals were greatly reduced or absent, similar to the BMP receptor phosphorylated form of Smad1/5, P-Smad1/5 (Fig. 5A,E,I). When we injected human SMAD1 RNA (hSmad1WT) into sbn^dtc24 zebrafish embryos, the P-Smad1^MAPK and P-Smad1^GSK3 signals were rescued and localized to ventral regions (Fig. 5B,F,J). Moreover, when we injected sbn^dtc24 mutant embryos with human SMAD1 mRNA that had mutations in the MAPK (hSmad1MM) or the GSK3 (hSmad1GM) phosphorylation sites (Fuentealba et al., 2007), we detected no P-Smad1^MAPK or P-Smad1^GSK3 immunostaining, respectively (Fig. 5G,K). From these results, we conclude that P-Smad1^MAPK and P-Smad1^GSK3 antibodies specifically detect their respective phosphorylation sites in the zebrafish Smad5 linker region. Furthermore, wild-type embryos treated with the FGF inhibitor SU5402 lacked P-Smad1/5^MAPK, indicating that P-Smad1/5^MAPK is specific for FGF signaling (compare Fig. 5H with 5R).

Although FGF and Wnt signaling are also present in dorsal marginal gastrula regions, we did not detect P-Smad1/5^MAPK or P-Smad1/5^GSK3 dorsally. To determine if Smad1/5 protein is present dorsally, we examined total Smad1/5 protein localization by whole-mount immunostaining. We found that Smad1/5 protein is present uniformly in both dorsal and ventral gastrula regions of the zebrafish embryo (Fig. 5C,D) and is absent from embryos injected with smad5 MOs (data not shown). Moreover, inhibition of FGF signaling did not change the total Smad1/5 level as determined by western blot analysis, indicating that FGF signaling does not affect the overall Smad5 level (Fig. 5L). These results are consistent with the previously reported requirement for BMP receptor phosphorylation of Smad1/5 prior to phosphorylation by MAPK or GSK3 (Fuentealba et al., 2007).

We then examined the localization of P-Smad1/5^MAPK and P-Smad1/5^GSK3 during gastrulation. P-Smad1/5^MAPK localized to ventral regions of the early gastrula, with high signal in the marginal region and lower levels animally (Fig. 5Q). P-Smad1/5^MAPK was maintained in a ventral vegetal region as gastrulation and epiboly progressed (Fig. 5R-T). By contrast, P-Smad1/5^GSK3 localized strongly to the entire ventral region of the early gastrula (Fig. 5U) and then shifted to an animal region during gastrulation (Fig. 5V-X). Because vegetal regions correspond to more posterior regions of the early embryo, these results suggest that FGF/MAPK, but not GSK3, phosphorylation of Smad5 is positioned in a region where it could temporally regulate BMP signaling along the AP axis.

To test whether the temporal patterning of DV tissues along the AP axis is modulated by the FGF/MAPK phosphorylation of the Smad1/5 linker, we examined the effect of mutating these phosphorylation sites. We depleted Tg(hsp70:chd) embryos of endogenous Smad5 by injecting smad5 MOs. We then injected these embryos with hSmad1WT RNA as a control or hSmad1MM mRNA, which has the serine/threonine phosphorylation sites mutated to alanine (Fuentealba et al., 2007). Injection of three doses of RNA into smad5-depleted embryos revealed that the hSmad1WT and hSmad1MM mRNAs have similar rescuing activity (supplementary material Fig. S3A). In addition, the Smad1 protein level in hSmad1WT- and hSmad1MAPK-injected embryos was similar based on western blot analysis (supplementary material Fig. S3B, noHS), suggesting that the hSmad1WT and hSmad1MM proteins have similar stability in the injected embryos. We then performed a series of HSs to investigate the temporal patterning of BMP signaling using Tg(hsp70:chd).

We investigated the temporal function of hSmad1MM in specifying DV tissues by analyzing the expression domain of hoxb1b, a prospective caudal hindbrain marker. We analyzed the strength of dorsalization by measuring the DV extent of hoxb1b expression relative to the overall DV embryo width. We classified the embryos as: ‘fully dorsalized’ if hoxb1b was expanded to 100% of the embryo width; ‘weakly dorsalized’ if expression was 66-99% of embryo width; ‘wild-type’ if expression was 50-66% of embryo width; and ‘ventralized’ for an expression domain that was less than 50% of the embryo width. Depletion of Smad5 by injection of smad5 MOs into Tg(hsp70:chd) embryos caused hoxb1b expression to expand fully around the ventral side independently of the HS treatments (Fig. 6B,E, second bar in each set). Injection of hSmad1WT or hSmad1MM mRNA into these embryos rescued hoxb1b expression to its normal expression domain (Fig. 6E, third and fourth bars of first set). Following a series of HSs of Tg(hsp70:chd), hSmad1WT-injected embryos showed the same temporal patterning of hoxb1b as wild-type embryos, with normal dorsally restricted hoxb1b expression by HS at and after the 70% epiboly stage (Fig. 6E, third bar in sets two to five; compare with Fig. 4N-Q). By contrast, injection of hSmad1MM mRNA caused 30-minute earlier specification of hoxb1b, as evident with HS at the 65% epiboly stage, compared with hSmad1WT-injected embryos (Fig. 6A-E). Injection of three different doses (250, 350 and 450 pg) of hSmad1WT and hSmad1MM also showed that only hSmad1MM leads to precocious patterning of the posterior ventral tissue, independent of the mRNA dose injected (supplementary material Fig. S4).

We similarly monitored dorsalization of anterior regions by examining expression of otx2. In addition, we examined no tail (ntl) expression in the notochord to allow us to precisely orient the embryos and determine the extent of dorsalization. We measured the DV width of the otx2 expression domain and compared it with the overall DV width of the embryo. The otx2 domain in wild-type embryos varied from 46 to 53% (n=8) of embryo width. In fully dorsalized embryos, the otx2 domain was 83-89% of embryo width (n=9; it is not 100% because the overall DV embryo width is measured at its widest point and the otx2 domain lies more anteriorly, where the DV width is narrower). We established four classes of phenotype: ‘fully dorsalized’, corresponding to an otx2 domain of 83-89% embryo width; ‘weakly dorsalized’, corresponding to 54-82% of embryo width; ‘wild-type’, corresponding to 46-53% of embryo width; and ‘ventralized’, corresponding to less than 46% embryo width.

Depletion of Smad5 by injection of smad5 MOs into Tg(hsp70:chd) embryos caused otx2 expression to expand fully to the ventral side independently of HS (Fig. 6G,J, second bar in each set). Injection of hSmad1WT or hSmad1MM mRNA into these embryos rescued otx2 expression to its normal domain (Fig. 6J, third and fourth bar in first set, respectively). Following a series of HSs of Tg(hsp70:chd) embryos, we found that hSmad1WT mRNA rescued otx2 expression to its normal dorsal region when subject to HS at and after the 50% epiboly stage, as observed in wild-type embryos (Fig. 6J, compare first and third bar in sets two to five; see also Fig. 1G-L). By contrast, injection of hSmad1MM caused 30-minute earlier specification of the otx2 domain, as evident with HS at the 40% epiboly stage, compared with hSmad1WT-injected embryos (Fig. 6H-J, compare the third and fourth bars in set three). These results suggest that FGF signaling through MAPK phosphorylates the linker of P-Smad1/5 in ventral vegetal regions of the embryo, inhibiting P-Smad1/5 activity. Thus, FGF/MAPK signaling in part controls the temporal patterning of DV tissues along the AP axis.

Here we show that there is an intimate coordination in the temporal patterning of DV tissues with AP patterning during zebrafish gastrulation. Altering AP patterning by activation or inhibition of FGF, Wnt or RA signaling pathways, in combination with temporal BMP inhibition, revealed that DV patterning by BMP signaling along the AP axis is coordinated with AP patterning (Fig. 7). A gradual shift of P-Smad1/5^MAPK in ventral vegetal regions during gastrulation suggests that FGF/MAPK could regulate the temporal patterning of BMP signaling along the AP axis (Fig. 7). We also showed that the temporal coordination of DV and AP patterning is in part mediated by FGF/MAPK (Fig. 6). Replacing Smad5 with hSmad1MM, which lacks the MAPK linker phosphorylation sites, caused precocious patterning of DV tissues. DV tissues in both anterior and posterior regions were patterned 30 minutes earlier by hSmad1MM. These results suggest that MAPK phosphorylation of the P-Smad1/5 linker normally inhibits BMP signaling by ∼30 minutes in both anterior and posterior regions. Although tissues are patterned precociously when Smad1/5 lacks the MAPK phosphorylation sites, the progressive patterning of DV tissues continues, suggesting that additional factors also regulate the temporal coordination of DV and AP patterning.

In addition to FGF/MAPK, a Wnt signaling component could be involved in coordinating DV and AP patterning. Our data indicate that DV patterning along the AP axis is coordinated with AP patterning by Wnt signaling. However, in contrast to P-Smad1/5^MAPK, P-Smad1/5^GSK3 was shifted to animal regions during gastrulation, suggesting that FGF/MAPK works independently of GSK3 to temporally regulate DV patterning through phosphorylation of the Smad1/5 linker region. However, the data do not exclude the possibility that the temporal coordination of DV and AP patterning is also mediated by Wnt signaling. It is possible that a Wnt signaling component other than GSK3 modulates DV temporal patterning along the AP axis. Further studies are needed to elucidate which component of the Wnt signaling pathway might temporally modulate BMP signaling during DV patterning.

In addition to the FGF and Wnt signaling components, RA signaling components could also be involved in coordinating DV and AP patterning. Nuclear RA receptor γ (RARγ) agonists decrease phosphorylation of Smad1/5/8 in mouse chondrocytic ADTC5 cells (Shimono et al., 2011). In mouse P19 (teratocarcinoma) cells and in the chick neural tube, RA promotes the degradation of P-Smad1 through MAPK activation (Sheng et al., 2010). Because P-Smad1/5 appears stable along the AP axis during zebrafish gastrulation (Fig. 5) (Tucker et al., 2008), another mechanism is expected to operate to coordinate patterning.

Based on our results, we propose that MAPK phosphorylation of the P-Smad1/5 linker reduces P-Smad1/5 function and thus acts in part to temporally regulate BMP signaling along the AP axis (Fig. 7). In Xenopus, the combined phosphorylation of the Smad1 linker by MAPK and GSK3 in dorsal tissues causes P-Smad1 degradation, thus blocking BMP signaling dorsally and allowing AP neural patterning (Fuentealba et al., 2007). In zebrafish, degradation of Smad5 is not evident in the marginal zone where FGF/MAPK activity exists and Smad5 levels appear unaffected when FGF signaling is inhibited (Fig. 5). Furthermore, in ventral tissues the P-Smad5 gradient appears stable along the AP axis and thus degradation of P-Smad1/5 is unlikely to regulate its temporal function along the AP axis during gastrulation (Fig. 5) (Tucker et al., 2008). We propose that, in ventral vegetal gastrula regions, FGF/MAPK in the absence of GSK3 negatively regulates P-Smad1/5 function in a distinct manner, possibly affecting its activity directly.

In Xenopus and zebrafish, AP tissues are temporally patterned progressively from anterior to posterior, as is evident in the earlier developmental expression of anterior compared with more posterior neurectodermal markers (Gamse and Sive, 2000; Gamse and Sive, 2001; Kudoh et al., 2002; Maves and Kimmel, 2005). The timing of specification also indicates temporally progressive AP patterning. Forebrain tissue is specified in the early gastrula (Grinblat et al., 1998), whereas the hindbrain is specified in mid-gastrula (Woo and Fraser, 1998). Here we showed that neurectodermal tissues are temporally patterned progressively from prospective forebrain to caudal hindbrain during late blastula to mid- to late gastrula stages (Fig. 1) (Tucker et al., 2008). Thus, AP and DV patterning are temporally coordinated, allowing cells to adopt both an AP and a DV identity simultaneously from anterior to posterior during blastula and gastrula stages.

Nodal is expressed in the marginal zone of the zebrafish gastrula (Erter et al., 1998; Rebagliati et al., 1998a; Rebagliati et al., 1998b; Sampath et al., 1998) and can posteriorize tissues dose dependently. Injection of antivin (lefty1 - Zebrafish Information Network) mRNA, which inhibits Nodal signaling activity, into zebrafish embryos causes a dose-dependent truncation of posterior tissues (Thisse et al., 2000). A genetic study in zebrafish has suggested that AP patterning by Nodal signaling is indirect, acting in part by regulating wnt8 expression (Erter et al., 2001).

In Xenopus, the transcription factor foxb1 was recently reported to coordinate DV and AP patterning of the ectoderm (Takebayashi-Suzuki et al., 2011). foxb1 is expressed in the posterior dorsal ectoderm of the Xenopus (Gamse and Sive, 2001) and zebrafish (Grinblat et al., 1998) gastrula and, when overexpressed in Xenopus, can negatively regulate BMP signaling (Takebayashi-Suzuki et al., 2011). Depletion of Foxb1 showed that it functions in AP patterning; however, a function in DV patterning was not found (Takebayashi-Suzuki et al., 2011). Thus, foxb1 regulates AP patterning but is unlikely to directly coordinate DV and AP patterning.

During gastrulation, the vertebrate body plan is established by evolutionarily conserved cell movements, such as epiboly and convergence and extension (C&E) movements (reviewed by Leptin, 2005; Montero and Heisenberg, 2004; Solnica-Krezel, 2005; Yin et al., 2009). Considering that BMP signaling regulates C&E movements of lateral mesodermal cells independently of DV patterning (von der Hardt et al., 2007) and that cranial DV tissues are patterned prior to the major C&E movements (Fig. 1) (Tucker et al., 2008), C&E cell movements are not expected to be related to the coordinated patterning of DV and AP tissues. However, we believe that epiboly movements play a key role in regulating the temporal patterning of DV tissues by moving the domain of FGF/MAPK signaling posteriorly during gastrulation. As epiboly proceeds during gastrulation, FGF/MAPK phosphorylation of the P-Smad1/5 linker becomes localized to progressively more posterior (vegetal) regions, allowing tissues to be temporally patterned and specified progressively in its wake (Fig. 7). Thus, this cell movement mechanism allows both AP and DV tissues to be temporally patterned progressively in a coordinate manner.

We thank Drs E. De Robertis and L. C. Fuentealba for the hSmad1WT and hSmad1MM constructs and P-Smad1/5^MAPK and P-Smad1/5^GSK3 antibodies; C. Hong for DMH1; M. Hibi and S. Maegawa for SU5402; R. T. Moon for Tg(hsp70:dkk1GFP); V. Prince for the hoxb1a plasmid; and J. Dutko, Y. Elkouby and Y. Langdon for comments on this manuscript.