Tsukushi Modulates Xnr2, FGF and BMP Signaling: Regulation of Xenopus Germ Layer Formation

Background Cell-cell communication is essential in tissue patterning. In early amphibian development, mesoderm is formed in the blastula-stage embryo through inductive interactions in which vegetal cells act on overlying equatorial cells. Members of the TGF-β family such as activin B, Vg1, derrière and Xenopus nodal-related proteins (Xnrs) are candidate mesoderm inducing factors, with further activity to induce endoderm of the vegetal region. TGF-β-like ligands, including BMP, are also responsible for patterning of germ layers. In addition, FGF signaling is essential for mesoderm formation whereas FGF signal inhibition has been implicated in endoderm induction. Clearly, several signaling pathways are coordinated to produce an appropriate developmental output; although intracellular crosstalk is known to integrate multiple pathways, relatively little is known about extracellular coordination. Methodology/Principal Findings Here, we show that Xenopus Tsukushi (X-TSK), a member of the secreted small leucine rich repeat proteoglycan (SLRP) family, is expressed in ectoderm, endoderm, and the organizer during early development. We have previously reported that X-TSK binds to and inhibits BMP signaling in cooperation with chordin. We now demonstrate two novel interactions: X-TSK binds to and inhibits signaling by FGF8b, in addition to binding to and enhancement of Xnr2 signaling. This signal integration by X-TSK at the extracellular level has an important role in germ layer formation and patterning. Vegetally localized X-TSK potentiates endoderm formation through coordination of BMP, FGF and Xnr2 signaling. In contrast, X-TSK inhibition of FGF-MAPK signaling blocks ventrolateral mesoderm formation, while BMP inhibition enhances organizer formation. These actions of X-TSK are reliant upon its expression in endoderm and dorsal mesoderm, with relative exclusion from ventrolateral mesoderm, in a pattern shaped by FGF signals. Conclusions/Significance Based on our observations, we propose a novel mechanism by which X-TSK refines the field of positional information by integration of multiple pathways in the extracellular space.


INTRODUCTION
During blastula stages of amphibian development, the embryo is organized into three distinct germ layers: ectoderm, mesoderm and endoderm, precursors of skin, connective tissue and gut respectively. Mesoderm is formed through inductive interactions by which cells of the vegetal region act upon overlying equatorial cells (Reviewed in [1]), an event relying upon pre-localized maternal determinants. Moreover, these overlying cells receive differential signals from specific vegetal areas, resulting in initial ventral-dorsal patterning of the embryo [2]. The vegetal cells themselves are destined to become endoderm, in a process that has not been as intensively studied, in comparison to mesoderm formation. Formation of the endoderm begins during early blastulation, where gene expression specific to this germ layer is reinforced by late blastula stages [3] and interestingly, the process of endoderm formation involves signals shared with the events of mesoderm induction and patterning.
Cell-cell communication is essential in germ layer formation and patterning. Studies on mesoderm formation have identified several mesoderm-inducing factors; Activin B [4], Vg1 [5], Derrière [6] and Xenopus nodal-related proteins (Xnrs), [7,8] members of the activin-like branch of the Transforming Growth Factor-b (TGF-b) family of signaling molecules. These proteins act as morphogens and demonstrate activity to induce dorsal mesoderm formation at higher concentrations [9,10]. A second, larger branch of the TGF-b family comprising of Bone Morphogenetic Proteins (BMPs) function to ventralize mesoderm, highlighting the importance of TGF-b signaling in both mesoderm formation and its subsequent pattering (Reviewed in [11]).
Interestingly, activin-like TGF-b signals are also essential components in the mechanism of endoderm induction. Vegetally localized T-box transcription factor (VegT) is essential for initiation of endoderm formation by activating expression of TGF-b-related Xnr family members and Derrière, in addition to endoderm-specific transcription factor Sox17 [12][13][14][15][16][17]. This occurs at the early blastula stage and functions to achieve a high level of activin-like TGF-b signaling within presumptive endoderm [13,18]. Sox17 expression activated early by VegT later relies upon Xnr signals for maintenance of its expression [19], and by late blastula stages, expression of endoderm specific transcription factors of the Mix and GATA families is induced (reviewed in [3]). Upon gastrulation, mesodermal and endodermal cell populations become distinct [20], yet is not known how a clear border is made between these segregated cell populations when common signaling pathways are employed for their formation and maintenance.
It is clear that morphogen gradients form part of the mechanism responsible for differential gene expression in the Xenopus embryo [21]. Additional signaling pathways may be accessory to differentiation between germ layers. For example, intact FGF signaling is essential for mesoderm induction [22,23], which is integrated with activin-like signaling [24,25]. In contrast to this, FGF signal inhibition has been implicated in endoderm formation [26]. Similarly, active BMP signaling is involved in ventral mesoderm formation, whereas BMP inhibition is linked to endoderm formation [26,27]. These germ-layer specific signal requirements are clarified in Figure 1, illustrating the point that in order to achieve precise germ layer specification, multiple signaling pathways initiated by extracellular molecules must be coordinated to produce an appropriate output. Intracellular crosstalk has been intensively studied in the context of signal integration; for example, activation of MAP kinase, downstream of FGF signaling, inhibits BMP through phosphorylation of the Smad1 linker domain [28]. There is now increasing evidence that individual extracellular regulators interact with multiple morphogens, such as Cerberus (Wnt, Xnrs, BMP), Coco (Wnt, Xnrs, BMP) and Follistatin (Activin, BMP) [29][30][31][32], suggesting that extracellular regulators play an important role in integration of multiple pathways.
We have previously reported that a short-range secreted protein, Tsukushi (TSK), interacts with and modulates activities of TGF-b family members, BMP and chick Vg1, in addition to the Notch ligand, Delta [33][34][35]. We aimed to study multiple signal regulation in early development, coordinated at the extracellular level by an individual type of molecule. In this study, we show that Xenopus Tsukushi (X-TSK) plays an important role in multiple signal integration, through binding and modulation of BMP, FGF and Xnr2. Zygotic X-TSK expression is activated in endoderm and the dorsal blastopore lip, with relative exclusion from ventrolateral marginal zone. Functional analysis shows that X-TSK potentiates endoderm formation whilst inhibiting ventrolateral mesoderm formation, through activation of Xnr2 signaling combined with inhibition of FGF and BMP signaling. Indeed, through FGFdependent localized expression of X-TSK and coordinated regulation of three distinct signaling pathways, TSK contributes to germ layer formation and patterning in Xenopus development.

X-TSK is expressed in ectoderm, dorsal mesoderm and endoderm
We previously demonstrated that TSK functions in primitive streak and Hensen's node formation through modulation of BMP and Vg1 in chick [33]. In Xenopus embryos, Chick TSK (C-TSK) induces dorsal mesoderm formation, site of the organizer, most likely through BMP inhibition. Although we have previously shown a role for TSK in neural crest formation in Xenopus through BMP and Notch signal modulation [34], we have yet to detail any possible multiple signal interactions during earlier stages of amphibian development.
In order to learn more about TSK function in early Xenopus development, X-TSK mRNA expression was examined spatially and temporally by in situ hybridization and RT-PCR. Whole mount in situ hybridization showed that X-TSK is localized to the animal hemisphere during late blastula and gastrula stages, as identified by purple staining (Figure 2A). At stage 10, X-TSK expression is also detected around the dorsal blastopore lip, relative to lower expression in the ventrolateral marginal zone. Furthermore, punctate staining within the fully formed blastopore is evident from stage 10.5, suggesting localization of X-TSK to the endoderm. Staining is not observed with sense X-TSK probe, confirming that X-TSK staining is specific.
As probe cannot penetrate into deep endodermal tissue by whole mount in situ hybridization, due to large size of the embryo and high yolk content [36], embryos were first sectioned, followed by in situ hybridization to examine TSK expression within the Selected signaling pathways involved in Xenopus mesoderm and endoderm formation. Activation of pathways indicated by '+++', inhibition of pathways indicated by '222'. Ectoderm = red, mesoderm = green, endoderm = blue. FGF signal activity is required for mesoderm formation in addition to activity of activin-like signaling (represented here by Xnr2). FGF and BMP signal inhibition with Xnr2 signal activation is involved in endoderm induction mechanisms. (B) Selected signaling pathways involved in Xenopus mesoderm patterning. Active BMP signaling produces mesoderm with ventral character, whereas inhibition of BMP signaling produces mesoderm of dorsal character. Also, Xnr2 expressed in the dorsal region has activity to induce dorsal mesoderm. doi:10.1371/journal.pone.0001004.g001 vegetal region. Figure 2B demonstrates that X-TSK mRNA is expressed in the pre mid-blastula transition (MBT, stage 7) embryo. This maternal expression is localized to the animal hemisphere, in addition to a small amount of vegetal expression in the blastocoel floor. At stage 10.5, zygotic expression of X-TSK is detected throughout the vegetal region and the dorsal blastopore lip, overlapping with expression of the endoderm-specific marker Sox17a [37] and with the organizer marker Goosecoid (Gsc), in the dorsal marginal zone [38] ( Figure 2D). In contrast to this, expression is diminished in the area where the pan-mesoderm marker, brachyury (Xbra), is expressed [39] ( Figure 2D). Later in gastrulation (stage 13), X-TSK is strongly expressed in the endodermal derived region and precaudal plate under the neuroectoderm ( Figure 2B). In order to demonstrate the specificity of staining for X-TSK expression, in situ hybridization with sense probe was again performed, resulting in minimal staining as shown in Figure 2B.
Temporally, X-TSK expression levels peak during germ layer formation and early gastrulation, as demonstrated by semiquantitative RT-PCR in whole embryos ( Figure 2C). These data showing the specific expression of X-TSK in ectoderm, endoderm and dorsal mesoderm during early stages of Xenopus development suggest a function for X-TSK in germ layer formation and patterning. Loss-of-function: X-TSK is a component of endoderm formation and mesoderm patterning In order to determine the importance of X-TSK in Xenopus germ layer formation and patterning we depleted X-TSK with antisense morpholino oligonucleotide (XMO) targeted to prospective endoderm or mesoderm. We have previously shown that this MO specifically depletes X-TSK protein levels [34,40], where morphants were analyzed from stage 15, and exhibited a reduction in neural tissue formation, although earlier stages have not previously been analyzed.
As we found X-TSK to be expressed throughout the vegetal region from gastrula stages, with a small amount of expression in the pre-gastrula embryo, we targeted XMO to prospective endoderm, followed by in situ hybridization analysis of endoderm markers Sox17a and GATA4 [37,41]. In addition to performing this analysis in the whole embryo, we again subjected sectioned embryos to in situ hybridization in order to access the deep endodermal tissue. XMO was co-injected with b-Galactosidase RNA to facilitate identification of the targeted area, as observed from the region of light blue staining. As part of these loss-offunction approaches, we also injected control MO (CMO) to rule out non-specific effects of MO on development. Sox17a and GATA4 are expressed throughout the vegetal region of gastrulastage embryos, with punctate staining of GATA4 observed as highlighted in the zoomed panel ( Figure 3A). Expression of both Sox17a and GATA4 are visibly diminished upon X-TSK depletion with 20 ng XMO in 41% and 48% of embryos respectively ( Figure 3A and Table 1). Punctate GATA4 staining is reduced in the presence of XMO, which has enabled quantification of loss-offunction by counting what we have termed as GATA4 positive foci. Figure 3A shows the resulting graphical representation of this quantification, presented as percentages relative to staining in uninjected embryos. GATA4 staining is reduced to only 50% relative to uninjected and CMO injected embryos (p,0.001). In order to confirm specificity of loss-of-function, rescue experiments were performed using human-TSK, unaffected by the XMO that overlaps with the initiation site of X-TSK. Co-injection of 20 ng XMO with 1 ng H-TSK mRNA almost completely rescues X-TSK MO mediated inhibition of endoderm marker expression, Sox17a and GATA4 ( Figure 3A and Table 1) and partially restores numbers of GATA4 positive foci (p,0.001, Figure 3A). Moreover, expression of 50 pg Xnr2, a known endoderm inducer [7] similarly restores expression of endoderm markers (p,0.001, Figure 3A). Interestingly, overexpression of H-TSK and Xnr2 increase levels of GATA4 positive foci. Loss of endoderm upon X-TSK depletion is also clearly demonstrated in later stages where gut development is perturbed, with the gut becoming significantly thinner (21% reduction of measured gut width in comparison to uninjected embryos (p,0.001, Figure 3C). These observations in combination with specific expression of X-TSK in the endoderm indicate that X-TSK is a component of Xenopus endoderm formation.
We continued by targeting XMO to a further X-TSK expressing region, the dorsal mesoderm, site of the organizer. Expression of the organizer gene Gsc is diminished upon X-TSK depletion in 52% of embryos, in agreement with TSK function as a BMP antagonist ( Figure 3B). In order to analyze general mesoderm formation, we examined the effect of XMO on expression of the muscle marker MyoD, localized to ventrolateral mesoderm, and later in somites [42]. XMO was targeted to one side of the embryo, as marked by b-Gal staining ( Figure 3B), thus acting as an internal control by which the proportion of uninjected vs. injected MyoD expression areas was measured using the Image J program. Figure 3B demonstrates a 30% expansion (p,0.001) in the area of MyoD expression, relative to the uninjected side, upon X-TSK depletion. We also analyzed expression of pan-mesoderm marker Xbra, in X-TSK depleted mesoderm. We found that Xbra expression is upregulated in XMO injected DMZ, as measured by RT-PCR (shown later). These effects in general mesoderm are subtle, perhaps due to redundancy with other SLRP family members, such as biglycan [43] or requirement of a positive factor to induce mesoderm. Nevertheless, this data does demonstrate a role for X-TSK in mesoderm formation and dorsal-ventral mesoderm patterning.

X-TSK overexpression expands endoderm and dorsal mesoderm whilst inhibiting ventrolateral mesoderm formation
To support loss-of-function data, expression of germ layer markers was analyzed following overexpression of X-TSK in lateral marginal zone, an area of relatively low X-TSK expression, again with b-Galactosidase to identify the targeted area. Figure 4A shows that embryos injected with 1 ng X-TSK mRNA demonstrate substantial expansion of endoderm markers Sox17a (32% of embryos injected) and GATA4 (37% of embryos injected), into the marginal zone. Although this phenotype is not penetrant, it is observed consistently. This data is also supported in loss-offunction rescues performed above where H-TSK expression increases numbers of GATA4 foci in sectioned embryos by 31% ( Figure 3A).
In contrast to this, pan-mesoderm Xbra expression was inhibited in 70% of embryos injected with 1 ng X-TSK ( Figure 4A). Later in development, these embryos also demonstrate visibly reduced MyoD expression in 33% of embryos. More detailed analysis demonstrates an 18% reduction in area of MyoD expression on the injected side (p,0.001) ( Figure 4B). This is also supported by the observation that gastrula-stage MyoD expression is inhibited by TSK overexpression (data not shown). In contrast to this, X-TSK overexpression in dorsal marginal zone (DMZ) clearly expands Gsc expression of the organizer in 43% of injected embryos ( Figure 4A). Both loss-and gain-of-function analyses indicate that X-TSK acts as a component of endoderm and dorsal mesoderm formation, in addition to negative regulation of ventrolateral mesoderm. Specifically, X-TSK appears to function in Xenopus germ layer formation to induce endoderm and dorsal mesoderm, whilst inhibiting ventrolateral mesoderm formation. This hypothesis is supported by the observation that X-TSK is expressed within endoderm and dorsal mesoderm relative to much lower expression within ventrolateral mesoderm.

X-TSK function in germ layer formation and patterning is not exclusively mediated by BMP inhibition
It was previously reported that TSK inhibits BMP in cooperation with chordin [40], thus raising the possibility that X-TSK functions in germ layer formation and patterning through this mechanism. To test this possibility, germ layer marker expression was analyzed in embryos where BMP signaling was compromised; embryos were injected with 250 pg truncated BMP receptor (tBR) [44] or 125 pg chordin (Chd) with b-Galactosidase as a targeting marker to the lateral or dorsal marginal zone ( Figure 5A). Injection of tBR or Chd into ventral marginal zone (VMZ) at these doses induces secondary axes in 85-100% of embryos (data not shown) and expands expression of organizer marker Gsc in all embryos analyzed when targeted to the DMZ ( Figure 5A). This is expected, as previously reported [44,45] and demonstrates that BMP is indeed effectively inhibited at these doses. X-TSK similarly expands Gsc expression, and this combined with previous knowledge that BMP signals are inhibited by TSK strongly suggests BMP inhibition as the mechanism involved. To confirm this hypothesis, we co-overexpressed X-TSK in DMZ with the constitutively active BMP receptor caALK3 [46]. Indeed, X-TSK mediated expansion of Gsc expression is blocked by hyperactive BMP signaling from caALK3 ( Figure 5B), confirming that X-TSK expands the organizer region through BMP inhibition. In contrast to this, expression of tBR within the lateral marginal zone expands Xbra expression, with no expansion of Sox17a ( Figure 5A). Chd injection did not affect expression of Xbra or Sox17a ( Figure 5A). In addition to this, X-TSK mediated inhibition of Xbra expression is not blocked by activation of BMP signaling by caALK3 (data not shown). These observations strongly indicate that X-TSK does not affect endoderm or ventrolateral mesoderm formation through inhibition of BMP signaling alone, suggesting that other pathways function within this context.

X-TSK inhibits FGF-MAPK and BMP/Smad1 signaling whilst enhancing activin-like/Smad2 Signaling
Since inhibition of BMP signaling alone is unable to explain X-TSK function in germ layer formation, we considered the potential roles of other major signaling pathways involved in early embryogenesis. MAPK, Smad1 and Smad2, downstream of FGF, BMP and activin-like signaling respectively [47] were analysed by Western blotting in X-TSK overexpressing explants. At this point, we did not consider Notch signaling, the function of which is currently complex in germ layer formation, and will be discussed later. As shown in Figure 5C, X-TSK inhibits Smad1 phosphorylation in animal explants in a dose dependent manner, in accordance with its function as a BMP inhibitor. Interestingly, from a lower dose range, X-TSK strongly inhibits phosphorylation of MAPK in animal explants. For analysis of Smad2, X-TSK was overexpressed in DMZ as activin-like ligands are present in this region, in contrast to the animal cap. Figure 5C shows that X-TSK activates Smad2 phosphorylation in DMZ. To address the possibility that X-TSK regulates these pathways through inhibition of BMP signaling, tBR and Chd injected explants were also analyzed ( Figure 5D), resulting in expected inhibition of Smad1 phosphorlylation in animal explants. tBR activates MAPK phosphorylation in animal explants as reported previously [48], whilst having no effect on Smad2 phosphorylation in DMZ, as shown in figure 5D. Chd does not inhibit MAPK phosphorylation in animal explants, which is also supported by the observation that Xbra expression, requiring intact FGF-MAPK signaling [22], is not affected upon Chd overexpression. These results demonstrate that the action of X-TSK upon MAPK and Smad2 is not mediated through BMP antagonism, and suggests potential mechanisms underlying X-TSK function in germ layer formation and patterning, which we have examined in further detail.

X-TSK blocks ventrolateral mesoderm formation through binding and inhibition of FGF8b
A dominant negative FGF receptor (XFD) blocks mesoderm formation, demonstrating that intact FGF-MAPK signaling is required for mesoderm formation in Xenopus [22,49]. However, Smad2 activation is known to enhance formation of mesoderm [50], ruling activin-like signaling out as a candidate in this context since X-TSK enhances Smad2 phosphorylation although inhibits general mesoderm formation. Therefore, we examined FGF-MAPK signaling as a mechanism for X-TSK mediated ventrolateral mesoderm inhibition. Figure 6A shows that MAPK phosphorylation is activated upon X-TSK depletion. Furthermore, upregulation of Xbra expression observed in X-TSK morphants is blocked by FGF inhibition with the dominant negative FGF receptor, XFD ( Figure 6B). This demonstrates a role for endogenous X-TSK in inhibition of FGF signaling in the mesoderm.
We subsequently focused upon the point of X-TSK interaction within the FGF-MAPK pathway. We analysed the effect of V-ras, which functions as a constitutively active component of FGF-MAPK signaling, downstream of the FGF receptor [51]. Coexpression of V-ras with X-TSK and b-Galactosidase as a tracer completely blocks inhibition of Xbra expression by X-TSK and restores MAPK phosphorylation, as analyzed by Western blotting ( Figure 6C-E), suggesting that X-TSK inhibits ventrolateral mesoderm formation through inhibition of FGF-MAPK signaling, upstream of Ras. Moreover, Figure 6G shows that MAPK activation by the potent mesoderm inducer FGF8b [52] is strongly inhibited by X-TSK. In support of this receptor-ligand level interaction, an inducible FGF receptor (iFGFR), which works as a constitutively active FGF receptor in the presence of dimerisation activator, AP20187 [53], rescues X-TSK mediated MAPK inhibition ( Figure 6F).
These observations suggest that X-TSK inhibits FGF-MAPK activity at the extracellular level. FGF8b appeared as a good candidate in the mechanism of X-TSK mediated FGF-MAPK inhibition, hence ventrolateral mesoderm inhibition, thus we analyzed interaction between these proteins in a pulldown assay. Indeed FGF8b-FLAG is pulled down in complex with X-TSK-Myc-His in a reaction with nickel beads ( Figure 6H). This interaction is specific, as FGF8b-FLAG in isolation is unable to bind nickel beads. In addition to this, the interaction with X-TSK is through FGF8b itself and not the FLAG epitope, as we have previously shown that X-TSK does not interact with activin-FLAG, or noggin-FLAG in the equivalent pulldown assay [40]. In combination, this evidence demonstrates that X-TSK inhibits ventrolateral mesoderm formation though binding and inhibition of FGF8b. The next part of the study focused upon the mechanism underlying endoderm induction by X-TSK.

X-TSK induces endoderm through binding and enhancing activity of Xnr2
Data so far has shown that X-TSK functions as an endoderm inducer in Xenopus. Nodal, a member of the TGF-b superfamily, is known to induce endoderm at high concentrations through activation of Smad2 [54]. A total of six nodal related genes have been identified in Xenopus (Xnr1-Xnr6) [7,8,55,56]. With the exception of Xnr3, they all demonstrate conserved functions upon overexpression, although their spatial and temporal expression varies. Of these Xnr family members, Xnr2 expression overlaps most closely with X-TSK expression [7]. As shown in Figure 5C, X-TSK activates Smad2 phosphorylation. These observations suggested Xnr2-Smad2 signaling as a candidate mechanism for X-TSK mediated endoderm induction. Therefore, we analyzed the role of Xnr2 in the context of X-TSK function.
We questioned whether X-TSK requires intact Xnr signaling for endoderm induction. Using a truncated Cerberus mutant (CerS) that specifically inhibits Xnr signaling in Xenopus [8] we studied the effect of Xnr inhibition upon X-TSK mediated induction of endoderm by in situ hybridization. Introduction of CerS inhibits expression of endoderm marker Sox17a in all embryos analyzed. In the presence of CerS, X-TSK mediated endoderm formation is completely blocked (Figure 7A and 7B). Furthermore, 50 ng Xnr2, co-injected with XMO, visibly restores expression of Sox17a and GATA4, and GATA4 positive foci diminished upon loss of X-TSK function ( Figure 3A). These data strongly indicate that intact Xnr signaling is required for X-TSK mediated endoderm induction.
Our previous observations that TSK binds to TGF-b superfamily members [35,40] suggests that X-TSK may regulate Xnr activity through binding to Xnr proteins. As shown in Figure 7C, mature Xnr2-Myc is pulled down in complex with X-TSK-Myc-His in a reaction with nickel beads. This interaction is specific, as Xnr2-Myc in isolation is unable to bind the nickel beads. Moreover, the interaction with X-TSK is through Xnr2 itself and not the Myc epitope, as we have previously shown that X-TSK does not interact with ADMP-Myc, or Follistatin-Myc in equivalent pulldown assays [40]. As shown in Figure 7D, X-TSK enhances Xnr2 mediated Smad2 phosphorylation in animal cap explants; this is further supported by co-overexpression of X-TSK and Xnr2 in whole embryos, where dorsal mesoderm and endoderm formation is enhanced, as marked by Gsc, Sox17a and GATA4 respectively ( Figure 8A). This data, along with rescue of X-TSK loss-of-function by Xnr2, demonstrates that X-TSK binds to and enhances Xnr2, where intact Xnr signaling is required for the function of X-TSK in endoderm induction.
Overexpression of Xnr2 in Xenopus embryos strongly activates expression of pan-mesoderm marker Xbra, in addition to expansion of dorsal marker Gsc, and endoderm markers Sox17a and GATA4 ( Figure 8A). However, in these cases, Xnr2 induces both endoderm and mesoderm markers without forming a clear border between them. Also the induction of Xbra in particular is detected far beyond Xnr2 expressing cells, as identified by b-Galactosidase staining. Very interestingly, co-overexpression of Xnr2 and X-TSK induces Xbra expression only outside the targeted region. Expression of Xbra within injected cells is almost completely inhibited whilst endoderm marker expression is enhanced. These observations indicate that X-TSK exerts its   effects at short-range, while Xnr2 functions over a long range [10]. This idea of X-TSK acting at short range is also supported by our previous observations; TSK is secreted from cultured cells and has no membrane spanning domain or GPI anchoring signal [33], and X-TSK fused with CD2, a membrane linker, showed an identical activity to wild type X-TSK in neural crest development [34].

Combinatorial regulation of BMP, FGF and Xnr signaling by X-TSK potentiates endoderm formation
We have shown that X-TSK binds to and modulates the activity of BMP, FGF8b and Xnr2 in a concentration dependent manner. Although we have demonstrated that X-TSK regulation of Xnr2 is required for endoderm formation, a previous study with chordin and dominant negative FGFR (XFD) has shown that BMP and FGF inhibition also participate in endoderm formation [26]. Thus, we examined the effect of BMP and FGF-MAPK activation upon X-TSK mediated endoderm formation ( Figure 9A and 9B). FGF-MAPK activation by V-ras partially blocks X-TSK mediated endoderm formation (reduced to 15% from 32%, p,0.05). In addition to this, BMP activation by caALK3 also partially blocks X-TSK mediated endoderm formation (reduced to 12% from 32%, p,0.01). This demonstrates that X-TSK activates endoderm formation through extracellular coordination of three pathways: Xnr2, FGF-MAPK and BMP.
To confirm the importance of multiple signal integration, we analyzed the combined effects of BMP and FGF signal inhibition with Xnr2 signal activation upon endoderm induction. Expression of truncated BMP receptor (tBR) in lateral marginal zone did not expand expression of endoderm marker GATA4 (Figure 9C), whereas expression of dominant negative FGF receptor (XFD) or Xnr2 produces only a light, diffuse expansion of GATA4 expression. However, combinations of XFD-tBR, XFD-Xnr2 and tBR-Xnr2 produced a stronger activation of GATA4 expression. Interestingly, combination of tBR, XFD, and Xnr2 produced a much stronger activation of GATA4 expression, indicating that inhibition of both FGF and BMP signaling pathways in combination with Xnr activation is important for endoderm formation, supporting the observed multiple signal regulation by X-TSK.

Zygotic expression of TSK is regulated by FGF-MAPK signaling
Our data indicates that germ layer formation and patterning is influenced by expression of X-TSK, suggesting the importance of TSK transcriptional regulation. As shown in Figure 2, X-TSK possesses a dynamic and unique expression pattern during early embryogenesis. Zygotic transcription of X-TSK is initiated in the dorsal region, followed by expression in the endoderm, with exclusion from ventrolateral mesoderm. This region of exclusion corresponds to an area of high FGF activity [57]. Therefore, the effect of FGF-MAPK activity upon zygotic X-TSK expression was studied by semi-quantitative RT-PCR. As shown in Figure 10A, activation of FGF-MAPK signaling with V-ras or constitutively active FGF receptor (caFGFR) inhibits X-TSK expression, whilst its inhibition with XFD activates X-TSK expression. These findings suggest that X-TSK expression is shaped by FGF signaling, where a feedback loop may be formed by which X-TSK inhibits FGF signaling at the extracellular level.

DISCUSSION
In this paper, we describe three major functions of X-TSK: activation of endoderm formation, inhibition of ventrolateral mesoderm formation, and expansion of dorsal mesoderm. These functions are mediated by multiple signal integration; inhibition of FGF-MAPK and BMP signaling, with enhancement of Xnr2 signaling by X-TSK, regulated in the extracellular space through protein-protein interactions.

TSK in mesoderm formation and patterning
We have shown previously that in chick, TSK functions as an organizer inducer [33]. In Xenopus, X-TSK is expressed in the dorsal blastopore lip where Spemann's organizer is located. Functional analysis presented in this current study shows that X-TSK has activity to expand the region expressing the organizer gene Gsc, mediated through BMP inhibition. In addition to this, we have previously shown that X-TSK directly induces neural tissue in animal caps and expands the neural region in whole embryos [33,34]. These observations indicate that X-TSK is working as with 1ng X-TSK and 500 pg X-TSK, lateral orientation. Xbra expression is not detected in Xnr2-X-TSK expressing cells, as identified by b-Gal staining. Xnr2 mediated expansion of Gsc expression is enhanced by X-TSK. Expression of endoderm markers Sox17a and GATA4 expanded by Xnr2 is enhanced by X-TSK, suggesting that X-TSK changes local cellular response to Xnr2. doi:10.1371/journal.pone.0001004.g008 a component of the organizer. However, X-TSK has two additional functions that differentiate it from other organizer molecules: induction of endoderm and inhibition of ventrolateral mesoderm. X-TSK is excluded from ventrolateral mesoderm, mediated by FGF signaling. Overexpression of X-TSK within the marginal zone inhibits expression of ventrolateral mesoderm markers, mediated through inhibition of FGF8b in the extracellular space by X-TSK.
It is known that pan-mesoderm markers such as Xbra require intact FGF signaling for their expression [22], whilst expression of dorsal organizer markers such as Gsc is not dependent on FGF signal status [23]. Therefore, it is possible to activate organizer formation whilst FGF signals are inhibited, although in vivo signal analysis in Xenopus embryogenesis showed that MAPK is activated in the organizer [57]. Since X-TSK inhibits FGF-MAPK and is expressed within the organizer, it is likely there may be a mechanism in place to interfere with or reduce X-TSK inhibition of FGF-MAPK in this region.

The role of X-TSK in endoderm induction
In this study, we have demonstrated that X-TSK is a component of endoderm induction in Xenopus. Zygotic expression of X-TSK is activated in endoderm, overlapping spatially and temporally with Xnr2 expression [7]. Our functional analysis has shown that X-TSK physically interacts with Xnr2, and enhances Smad2 activity downstream of Xnr2. In addition to this, X-TSK has two more activities: binding and inhibition of FGF8b, in addition to previously reported BMP binding and inhibition. It is known that strong activation of nodal signaling plays a major role in endoderm induction in all vertebrates [58]. In combination with this, FGF and BMP inhibition also function to positively contribute to endoderm induction [26]. Our experiments using Xnr2, tBR, and XFD clearly show that coordination of the three pathways has much stronger endoderm inductive activity relative to manipulation of individual or dual signaling pathway components. We have demonstrated that X-TSK is a key coordinator of these multiple pathways outside the cell through regulation of an extracellular signaling network. In terms of the intracellular pathways involved in endoderm induction downstream of X-TSK activity; it is known that transcription factors such as Sox17, Mixer, Mix1, and Bix2/milk have dual activities that inhibit mesoderm formation and activate endoderm formation [15,20,59]. It is likely that such transcription factors are activated downstream of signaling coordinated by X-TSK, further refining germ layer specific gene expression.
It is important to note that X-TSK overexpression in animal caps produces only a weak induction of Sox17a as measured by in situ hybridization and RT-PCR (data not shown). This induction may be weak as FGF signals are activated upon dissection, in addition to the important point that activin-like ligands are not present in the animal region. Moreover, the demonstrated induction of endoderm markers upon X-TSK overexpression is consistent, although not penetrant. This may be due to the fact that many natural mechanisms are in place to prevent induction of endoderm within the marginal zone.
There are significant differences in mesoderm and endoderm formation between different organisms, although recently, conserved molecular mechanisms have been elucidated [60]. This raises the possibility that TSK may have a conserved function between species. In support of this, we have demonstrated similar roles for TSK in organizer formation in chick and Xenopus. In chick, TSK is expressed in the posterior marginal zone and Koller's sickle, followed by expression in the extending primitive streak and Hensen's node; whilst in zebrafish, TSK is expressed in the blastoderm margin [33], structures composed of both mesoderm and endoderm precursors. Furthermore, in the case of chick, nodal is expressed in Koller's sickle [60], suggesting that TSK may work with nodal to regulate specification of mesoderm and endoderm. More evidence of conservation of mechanism across species can be found in zebrafish. Here we have shown that coordinated modulation of BMP, FGF and Xnr pathways is important in Xenopus endoderm formaton; this is also true for the zebrafish where combined BMP, FGF and nodal signaling have been shown to regulate endoderm formation and segregation of endoderm and mesoderm precursors [61]. Future studies into the potential conservation of TSK function between species may prove to be interesting and may also provide more information on conservation of signaling involved.

Integration of multiple signaling pathways
Temporal and spatial regulation of multiple signaling pathways is essential for tightly controlled regulation of development. Secreted soluble growth factors and their inhibitors have fundamental roles in signal regulation; however, studies about coordination of these pathways are largely restricted to intracellular cross talk [62], yet several extracellular regulators such as follistatin and cerberus are known to interact with multiple signaling pathways. Our analyses have demonstrated that X-TSK binds to and regulates FGF8b, BMP and Xnr2 at the extracellular level, in a concentration dependent manner to function in germ layer formation and patterning. Our previous work suggests that TSK function is not restricted to endoderm and mesoderm; in ectoderm, X-TSK regulates BMP activity, which contributes to a decision between epithelial and neural tissues. In addition to this, X-TSK regulates Delta-Notch signaling during neural crest formation [34]. Currently, the function of Notch in mesoderm formation is complex, where some evidence suggests that activation of Notch contributes to mesoderm inhibition whilst endoderm is induced. Conversely, it has been found that activation of Notch signaling delays loss of mesodermal competence [63,64]. Although we cannot rule out participation of the Notch pathway, which may prove to be interesting in future studies, we did not study Notch signaling in the context of TSK mediated germ layer formation and patterning due to these complexities. The potential participation of Vg1 must also be considered in future studies, as TSK activates Vg1 in chick development [35]. In Xenopus, it had been thought that Vg1 was not processed until a second allele was identified [65], although no effect on endoderm has been reported in loss-of-function studies. Therefore, Xnr2 remained a more attractive candidate in TSK functional mechanism in the endoderm and dorsal mesoderm. Although we have shown here that intact Xnr2 signaling is indeed required for TSK function, future work with Vg1 will be interesting. Even so, we have clearly demonstrated the involvement of FGF, BMP and Xnr pathways. Our analysis shows the importance of extracellular cross talk of these pathways and suggests that extracellular coordination of multiple signaling pathways may have important roles in cell signaling. This is supported by the potential involvement of additional signaling pathways, including Notch and Vg1.

Signal coordination in germ layer formation
Morphogens such as activin-like proteins have been shown create distinct fates depending on their concentrations [66]. It has been reported that cellular response to activin depends on the absolute number of receptors occupied by activin [67], indicating that regulation of morphogen diffusion is critical to create an appropriate concentration gradient. This importance of diffusion control has been demonstrated recently; Drosophila mutants toutvelu and dally, which have defects in the synthesis of extracellular heparan sulfate proteoglycans, demonstate defective Wingless morphogen diffusion [68]. Members of the SLRP family are proteoglycans [69] and thus may regulate morphogen gradients in a similar way.
Here we propose a model for TSK function in Xenopus germ layer formation and patterning. With the combination of X-TSK and Xnr2, there is no overlap between endoderm markers and pan-mesoderm marker Xbra. Thus it is tempting to speculate that TSK may contribute to segregation between endoderm and mesoderm specific gene expression by a combination of four factors. Firstly, Xnr2 and X-TSK have distinct functional ranges with Xnr2 working as a long-range morphogen [70], whilst X-TSK works at short-range [34]. Secondly, Xnr2 and X-TSK regulate distinct sets of signaling pathways. Thirdly, X-TSK functions with several factors to synergistically potentiate or inhibit the activities of these proteins. This interaction may change the effective concentration or diffusion of proteins. Thus, in the case of co-overexpression, X-TSK in proximity to cells expressing the two proteins creates a competent area; here X-TSK makes a complex with Xnr2 and potentiates signaling whilst inhibiting FGF-MAPK and BMP signaling. This coordinated regulation may produce clear regions of gene expression.
We need to consider a fourth important factor: spatial and temporal transcriptional regulation. Zygotic X-TSK is expressed in endoderm, dorsal mesoderm, and ectoderm with exclusion from ventrolateral mesoderm. Zygotic X-TSK expression in ventrolateral mesoderm is inhibited by FGF-MAPK signaling. Within the endoderm, a regulatory loop may be created in which TSK inhibits FGF-MAPK, which in turn promotes TSK expression. Based on our observations, we propose a model for X-TSK mediated embryonic patterning ( Figure 10B). In the absence of X-TSK, Xnr proteins create a pattern of Smad2 activation with a vegetal-animal gradient. Activated X-TSK expression in the endoderm creates a competent area, in which the activity of Xnr proteins is enhanced, possibly by an increase in local effective concentration. X-TSK provides ideal coordination for endoderm formation: Xnr2 activation, FGF-MAPK inhibition, and BMP inhibition. Conversely, MAPK activation inhibits expression of X-TSK in ventrolateral mesoderm, possibly permitting mesodermspecific gene expression in this area. Moreover, X-TSK expression in dorsal mesoderm contributes to organizer formation and function, mainly through BMP inhibition by synergistic ternary complex formation among X-TSK, BMP, and chordin [40] and possibly through Xnr activation. In conclusion, through regulation of these multiple factors, TSK coordinates formation of the endoderm and patterning of mesoderm during early Xenopus embryogenesis.

Future perspectives
It is now becoming clear that future work will focus upon the dissection of an 'extracellular network' of signal regulation, unraveling its importance in cell signaling as a whole. We have demonstrated the importance of extracellular coordination of FGF, Xnr2 and BMP signals in germ layer formation and patterning. Future work will involve the potential role of Notch and other activin-like signaling, which may go to demonstrate the nature of extracellular networks in greater detail. It will also be important to unravel the function of TSK in germ layer formation and patterning in other species such as chick and zebrafish, in order to demonstrate conservation and the importance of extracellular coordination of multiple pathways.

Embryology and in situ hybridization
Xenopus laevis embryos were staged according to Nieuwkoop and Faber [71]. For animal cap assays, mRNA was injected into the animal pole of 2-cell stage embryos. Animal caps were dissected from stage 8-9 embryos in 16 MBS and cultured in 0.76 MBS. For DMZ assays, mRNA was injected into the DMZ of 8-cell stage embryos. DMZ was dissected from stage 10 embryos in 16 MBS and cultured in 0.76MBS. Whole mount in situ hybridization was performed as described [72]. In situ hybridization of sectioned embryos was performed as described [36]. The following probes were used; X-TSK, Xbra [39], MyoD [42], Gsc [38], Sox17a [37], GATA4 [41]. Gut width and areas of MyoD expression were measured using Image J software (NIH). T Tests were performed to evaluate the statistical significance of results.

Semi-quantitative RT-PCR
Embryos, animal caps or DMZ explants at the indicated stages were snap-frozen, followed by RNA isolation with RNeasy kit (Qiagen). cDNAs were generated according to the manufacturers protocol (Taqman RT Reagents, Applied Biosystems). Primers used for the PCR reaction were described previously: ODC [18], Xbra [18], X-TSK [34]. Quantitative ranges were determined before final analysis. All reactions were normalized against ODC gene product.