Tbx6-dependent Sox2 regulation determines neural or mesodermal fate in axial stem cells.

The classical view of neural plate development held that it arises from the ectoderm, after its separation from the mesodermal and endodermal lineages. However, recent cell-lineage-tracing experiments indicate that the caudal neural plate and paraxial mesoderm are generated from common bipotential axial stem cells originating from the caudal lateral epiblast. Tbx6 null mutant mouse embryos which produce ectopic neural tubes at the expense of paraxial mesoderm must provide a clue to the regulatory mechanism underlying this neural versus mesodermal fate choice. Here we demonstrate that Tbx6-dependent regulation of Sox2 determines the fate of axial stem cells. In wild-type embryos, enhancer N1 of the neural primordial gene Sox2 is activated in the caudal lateral epiblast, and the cells staying in the superficial layer sustain N1 activity and activate Sox2 expression in the neural plate. In contrast, the cells destined to become mesoderm activate Tbx6 and turn off enhancer N1 before migrating into the paraxial mesoderm compartment. In Tbx6 mutant embryos, however, enhancer N1 activity persists in the paraxial mesoderm compartment, eliciting ectopic Sox2 activation and transforming the paraxial mesoderm into neural tubes. An enhancer-N1-specific deletion mutation introduced into Tbx6 mutant embryos prevented this Sox2 activation in the mesodermal compartment and subsequent development of ectopic neural tubes, indicating that Tbx6 regulates Sox2 via enhancer N1. Tbx6-dependent repression of Wnt3a in the paraxial mesodermal compartment is implicated in this regulatory process. Paraxial mesoderm-specific misexpression of a Sox2 transgene in wild-type embryos resulted in ectopic neural tube development. Thus, Tbx6 represses Sox2 by inactivating enhancer N1 to inhibit neural development, and this is an essential step for the specification of paraxial mesoderm from the axial stem cells.

Evidence derived from cell marking and lineage tracing in mouse and chicken embryos indicates that the CLE, the region of epiblast flanking the rostral primitive streak, serves as the common precursor pool for the paraxial mesoderm and caudal neural plate which later contributes to the caudal hindbrain and spinal cord1,2,7. The bipotential precursors serve as the pool of “axial stem cells” that contributes to the coordinated elongation of the neural tube, which develops from the cell population remaining in the superficial layer, and paraxial mesoderm, derived from cells that ingress through the primitive streak8-10. The most compelling evidence for this was provided by the single cell lineage analysis reported by Tzouanacou et al.8, who utilized intragenic recombination in a transgene to mark a clone, and demonstrated that a substantial fraction of individual axial stem cells do produce progenies of both cell fates. However, the regulatory mechanism underlying this neural versus mesodermal fate choice remained to be elucidated.

Expression of the transcription factor gene Sox2 is regarded as the hallmark of the neural primordial cell state, and its activation is strongly correlated with the establishment of the embryonic neural plate (Fig.1b; Supplementary Fig.1). Our earlier studies have indicated that among a number of enhancers regulating Sox2, enhancer N1 is responsible for Sox2 activation in the caudally extending neural plate5,6 (Fig.1a-c; Supplementary Fig.1). Followong features of enhancer N1 indicate its involvement in the regulation of CLE-derived cells4: (1) Enhancer N1 is activated precisely in the region of the CLE, and sustained in the zone at the caudal end of neural plate (ZCNP) (Fig.1b,c). Its activation, however, does not immediately lead to Sox2 expression in the CLE, owing to BMP signal-dependent repression of Sox2 in the CLE. Only when the CLE cells become a part of ZCNP located immediately rostral, the cells are relieved from the BMP signal and initiate Sox2 expression. In fact, the inhibition of BMP signals results in precocious Sox2 activation in the entire CLE4(Supplementary Fig.2). (2) Enhancer N1 activity is normally shut off in the mesodermal precursors that have ingressed through the primitive streak, suggestive of the release of this cell population from a neural fate4. (3) Enhancer N1 is activated by the synergistic action of Wnt and Fgf signals4(Fig.1d,e), while the Fgf signal is required for the maintenance of the axial stem cells in the CLE1,11,12. Based on these observations, we hypothesized that regulation of Sox2 through enhancer N1 is an important mechanism to regulate cell fate in the CLE.

Figure 1

Enhancer N1 of the mouse Sox2 gene and its activity in comparison with Sox2 and Tbx6 expression

a. The position of enhancer N1 relative to the Sox2 ORF and the N1 core sequence bearing the conserved Lef1-binding elements and Fgf-responsive element (FgfRE). b. Expression of Sox2 in E8.5 normal embryo in dorsal view. c. Enhancer N1 activity at the same stage, indicated by the expression of enhancer N1-LacZ transgene in primary transgenic embryos. ZCNP, zone at the caudal end of neural plate. d,e. The loss of enhancer activity by the mutations in both Lef1 elements (d, mutWnt) or in FgfRE (e, mutFgf)4 in the transgenic embryos. f. Tbx6 expression in E8.5 mouse embryo detected by in situ hybridization. f’,f”. The cross sections at the node (n)(f’) and primitive streak (ps)(f”) levels. NP, neural plate; CLE, caudal lateral epiblast; PAM, paraxial mesoderm. Scale bars, 200 μm (b-e); 50 μm (f’,f”).

Given the CLE origin of paraxial mesoderm, the phenotype of Tbx6 null mutant embryos is remarkable, where bilateral ectopic neural tubes develop at the expense of the paraxial mesoderm caudal to somite 6 level3. Tbx6, encoding a T-box transcription factor, is expressed in the primitive streak and presomitic paraxial mesoderm during gastrulation13(Fig.1f), and is presumed to have two major regulatory functions: (1) anti-neural fate regulation in the presomitic mesoderm, as indicated by the mutant phenotype, and (2) regulation of later somite segmentation involving Mesp2 and Dll-1 genes14-17.

We investigated the process of ectopic neural tube development from the presumptive paraxial mesoderm in homozygous Tbx6 mutant embryos. In Tbx6 mutant embryos at E8.5, mesoderm development is already defective, as indicated by the absence of the normal pattern of Uncx4.1 expression in the caudal part of segmented somites (Supplementary Fig.3a). The expression of Sox2 in normal embryos at E8.5 (5~6 somites) was confined to the neural plate and neural tube (Fig.2a), whereas in Tbx6 mutant embryos, the cells in the paraxial mesodermal compartment underlying the neural plate fully expressed Sox2, at levels comparable to those in the neural plate (Fig.2b). The ectopic Sox2 expression was followed by the ectopic development of neural tubes in the same compartment at E9.5 (Fig.2c,d), where genes such as Pax6 and Pax3 were expressed, indicating the development of the dorso-ventral patterning of the ectopic neural tubes3 (see also Fig.3).

Figure 2

Ectopic activation of enhancer N1, Sox2 expression and neural development in the paraxial mesoderm compartment of Tbx6−/− embryos

a,b. Sox2 expression in wild type and Tbx6−/− embryos at E8.5, shown by whole mount in situ hybridization. a’,a”,b’,b”. Transverse sections at the axial levels anterior to and through the node (n). NP, neural plate. Arrowheads, Sox2 expression in the paraxial mesodermal compartment in Tbx6−/− embryo. c,d. Sox2 expression in wild type and Tbx6−/− embryos at E9.5. NT, neural tube. Arrowheads, development of Sox2-positive ectopic neural tubes. e,f. and g,h. Sox1 and Sox3 expression, respectively, in wild type and Tbx6−/− embryos at E8.5. i,j. Activity of enhancer N1 at E8.5 in wild type embryos (i) or in Tbx6−/− embryos (j) detected by in situ hybridization of Egfp transcripts. i’,i”,j’,j”. Their transverse sections at the node (n) and primitive streak (ps) levels. In wild type embryos, enhancer N1 activity was absent in the paraxial mesoderm compartments (PAM) (i’,i”), whereas a strong N1 activity was demonstrated in the PAM in Tbx6−/− embryos (j’,j”). Enhancer activity was detected in the node and endoderm (e) in either genotype. k,l. Expression of Wnt3a at E8.5 in wild-type embryos (k) or Tbx6−/− embryos (l) detected by in situ hybridization, in transverse sections at a primitive streak level. Scale bars, 50 μm.

Figure 3

Enhancer N1-dependent paraxial Sox2 expression and ectopic neural tube development in Tbx6−/− embryos, and their suppression in Tbx6−/−;ΔN1/Δ N1 double mutants

a-d. Effects of the enhancer N1 mutation (ΔN1/ΔN1) on Sox2 expression in the neural plate and paraxial tissues in wild type and Tbx6−/− embryos at E8.5. Arrows: the position of node. The CLE, encircled by a broken line, and ZCNP (zone at the caudal end of neural plate) are indicated. Sox2 expression in the ZCNP is lost in the enhancer N1 mutant embryos. c,c’,d,d’. The loss of Sox2 expression in the paraxial mesoderm compartment in Tbx6−/− embryos (c’) by the enhancer N1 mutation (d’), demonstrated by sections of embryos of relevant genotypes (c,d) at the axial levels indicated. Embryo specimens in c and d are overstained to confirm the absence of Sox2 in the paraxial mesoderm compartment. e-g. Expression of Sox2 (e, neural), Pax6 (f, neural), Pax3 (g, both neural and mesodermal) was examined in embryos of respective genotypes at E9.5. Scale bar, 50 μm.

In contrast to Sox2, other group B1 Sox genes, Sox1 and Sox3 that are normally expressed in the neural tube, were not activated in the paraxial mesodermal compartment (Fig.2e-h). This observation strongly suggests that the activation of Sox2 in the paraxial mesodermal compartment is causative of ectopic neural tube development in Tbx6 mutant embryos.

By analogy to the process of caudal neural plate development, we speculated involvement of enhancer N1 activation, leading to Sox2 expression, in ectopic neural tube development in Tbx6 mutant embryos. To test this model, we compared enhancer N1 activity in Tbx6 mutant and wild type embryos (Fig.2i,j; Supplementary Fig.3b) using enhancer N1-driven transgenes for Egfp (for in situ hybridization detection of transcripts) or LacZ (for histochemical enzyme staining). In wild-type embryos, enhancer N1-dependent Egfp expression was clearly detected in the CLE and primitive streak, but not in the paraxial mesodermal compartment (Fig.2i). Analysis of serial sections suggested that enhancer N1 loses its activity after the cells ingress through the primitive streak and when the cells migrate laterally into the mesodermal compartment (Fig.2i’,i”). In contrast to the primitive streak-restricted Egfp transcript in the mesodermal layer, more stable LacZ enzyme activity was detected in more laterally positioned cells. This observation is consistent with the model that the cells turn off enhancer N1 before migrating laterally (Supplementary Fig.3b). An analogous sequence of events occurs in chicken embryos4.

Remarkably, in the Tbx6 mutant embryos, enhancer N1 activation persisted in the paraxial mesodermal compartment, as indicated by strong Egfp and LacZ expression (Fig.2j, Supplementary Fig.3b). Even in Tbx6 mutants, the trans-layer cell ingression pathway appears to be confined to the primitive streak, as the laminin-positive basal lamina separating the epiblastic and mesodermal cells was interrupted only in the primitive streak region (Supplementary Fig.3c). These observations indicate that the persistent activity of enhancer N1 in the paraxial mesodermal compartment caused ectopic Sox2 expression, which in turn leads to the development of ectopic neural tubes. Conversely, this suggests that in normal embryos the activity of Tbx6 expressed in the paraxial mesoderm suppresses enhancer N1 activation, thereby preventing the development of neural tubes from the mesodermal precursors.

To confirm that the Sox2 misexpression in the paraxial mesoderm compartment in Tbx6 mutant embryos, leading to supernumerary neural tube development, is due to the ectopic activation of enhancer N1, we produced an enhancer N1 mutant allele in the mouse, in which only the enhancer N1 sequence was deleted from the Sox2 locus (Supplementary Fig.4). Homozygosity for the enhancer N1 mutation (ΔN1/ΔN1) caused the loss of Sox2 expression in the ZCNP, which is the nascent neural plate positioned immediately rostral to the CLE1,2,12. This loss of Sox2 expression in the ZCNP in the ΔN1/ΔN1 embryos was noticeable after the 5-somite stages (E8.5)(Fig.3b), consistent with the timing of enhancer N1 activation (Fig.1c), and continued to E9.5. Despite this, Sox2 was activated in the closing neural tube, and the development of the spinal cord proceeded without significant morphological defects (Fig.3b). We presume this to be ascribable to the presence of another enhancer to back up Sox2 expression in the neural tube, and/or other group B1 Sox genes expressed in the neural plate to compensate for the role of Sox2.

By crossing Tbx6+/−;+/ΔN1 double heterozygotes, Tbx6−/−;ΔN1/ΔN1 double homozygous mutant embryos were successfully obtained. In Tbx6−/−;ΔN1/ΔN1 embryos at E8.5, the ectopic Sox2 expression in the paraxial mesodermal compartment was totally absent (Fig.3d,d’), in contrast to the case of Tbx6−/− embryos (Fig.3c,c’), providing strong evidence that enhancer N1 determines the ectopic expression of Sox2 in Tbx6 mutants.

In the ectopic neural tubes that develop in Tbx6−/− embryos at E9.5, Sox2 expression continues, while expression of other genes that are associated with neural tube development also initiates, namely Sox1, Pax6, and Pax3, the latter two reflecting the dorso-ventral identity of the neural tube3 (Fig.3, Supplementary Fig.5). In the Tbx6−/−;ΔN1/ΔN1 embryos at E9.5, however, the expression of all these neural genes in the paraxial compartment was extinct, concomitant with the absence of the tubular structure (Fig.3). The paraxial tissue in the Tbx6−/−;ΔN1/ΔN1 embryos failed to express any of the mesoderm or endoderm marker genes examined, namely Pax3 (Fig.3g) and Mox1 for paraxial mesoderm, Pax2 for intermediate mesoderm, Foxf1 for visceral lateral plate mesoderm, and Foxa1 for endoderm (Supplementary Fig.5). These observations indicate that Tbx6 has two distinct functions in the mesodermal precursors: the inhibition of neural development by preventing enhancer N1 activation, and the promotion of mesodermal development.

To confirm that the Sox2 misexpression in the paraxial mesoderm is causative of the ectopic neural tube development, HA-tagged Sox2 was ectopically expressed in the primitive paraxial mesoderm of wild type embryos using the msd enhancer of the Dll1 gene18(Fig.4a). Exogenous Sox2-HA was successfully expressed in the caudal-most mesodermal compartment (Fig.4b), as indicated by HA-tag immunostaining and the mesodermal Sox2 immunostaining (Fig.4c-e). At a more rostral level of the paraxial mesoderm, where the msd enhancer activity was lost and exogenous Sox2-HA expression had ceased, endogenous Sox2 expression persisted, indicating a positive feedback loop, where the former activated the latter. Notably, the clusters of cells with relatively high Sox2 also expressed Pax6 (Fig.4d, open arrowheads), coincident with Pax6 expression in the closing neural tube at the same axial level. At more rostral levels, the cells in the mesodermal compartment with an intensity of Sox2 expression comparable to that in the neural tube formed Pax6-positive miniature neural tubes (Fig.4e, white arrowheads). This result demonstrates that Sox2 expression is sufficient for the initiation of neural tube development from the primitive paraxial mesoderm, even in wild type embryos.

Figure 4

Development of ectopic neural tubes from the wild type paraxial mesoderm by misexpression of exogenous Sox2

a. Structure of the transgene. Msd enhancer of Dll1 gene18 in a dimeric form was used for paraxial mesoderm-specific activation of the transgene coding for Venus and HA3-tagged Sox2 joined with a 2A peptide sequence23. b. Mesodermal precursor-specific activation of Venus in transgenic embryos at E8.5. c-e. Cross sections at axial levels shown in (b). Immunofluorescence for Sox2, Pax6, and HA-tag, and Hoechst33342 staining, and a comparison with Sox2 immunofluorescence of normal embryos. Scale bar, 50 μm. f. Model of enhancer N1 regulation and enhancer N1-dependent Sox2 activation in normal and Tbx6−/− embryos. In the CLE (left), where Wnt signal is above threshold, enhancer N1 is activated, however despite this Sox2 expression itself is repressed by the inhibitory effect of BMP signals. In the ZCNP (right, top), where the Wnt signal stays ON, the relief from the BMP-mediated inhibition effectuates enhancer N1-dependent activation of Sox2 expression in the caudal neural plate. In the paraxial mesoderm (right, bottom), although the BMP signal is downregulated24,25, Wnt signal drops below the threshold required for the activation of enhancer N1 when Tbx6 activity is present, and Sox2 expression is not activated.

Finally, we asked whether the suppression of Sox2 enhancer N1 activation by Tbx6 involves direct interaction of the Tbx6 protein with the enhancer N1 DNA sequence. Various overlapping fragments of the N1 sequence were tested for Tbx6 binding using electrophoretic mobility shift assays (EMSA). None of the N1 subfragments exhibited T-box factor binding capacity, under the condition where all known T-box-binding sequences strongly bound Tbx6 (Supplementary Fig 6), indicating that the suppression of the enhancer N1 activity by Tbx6 is probably not a direct regulation.

As enhancer N1 is activated by synergistic action of Wnt and Fgf signals4 (Fig.1), we investigated the expression patterns of the genes encoding these protein ligands in the CLE and paraxial mesoderm19-22. A remarkable change was observed in the expression pattern of Wnt3a (Fig.2k,l). In normal embryos, Wnt3a expression is strongly expressed in the primitive streak, CLE and caudal neural plate, while its level is very low in the paraxial mesoderm (Fig.2k). In Tbx6 mutant embryos, however, the strong Wnt3a expression extends to the paraxial mesoderm compartment (Fig.2l). On the other hand, the expression patterns of Wnt8a (absent in the mesoderm), Fgf8 and Fgf4 in the CLE and paraxial mesoderm compartment were not significantly altered in the Tbx6 mutant embryos (Supplementary Fig.7). We hypothesize that Wnt3a expression, normally repressed in the paraxial mesoderm is de-repressed in the Tbx6 mutant embryos, and the strong Wnt3a expression in the paraxial mesoderm is causative of the ectopic activation of enhancer N1, as schematized in Fig.4f. The strong spatial correlation between Wnt3a expression and enhancer N1 activity observed at various axial levels in both wild type and Tbx6 mutant embryos lends support for this model (Fig.2i-l).

As discussed above, a significant fraction of the cells in the CLE domain have been shown to serve as the common precursors for the neural plate and paraxial mesoderm8. The present study clarifies the core regulatory circuit underlying the neural versus mesodermal dichotomous fate choice in the axial precursors, where Sox2 expression is regulated by Tbx6 activity. As schematized in Fig.4f, the mechanism to activate Sox2 transcription operates first in CLE, namely enhancer N1 activation, and while this mechanism is still idle under the BMP-dependent Sox2 repression, a fraction of the cells migrate into the mesodermal compartment, where the activity of enhancer N1 is lost, likely due to the very low Wnt activity. Tbx6-dependent mesodermal development is then promoted. Tbx6 appears to have two major functions, inhibition of Sox2 activation via enhancer N1, and activation of the mesoderm specification program. The former function is important for the latter, as Sox2 misexpression in wild type embryos transforms paraxial mesoderm into neural tissue (Fig.4c-e). In the Tbx6 mutant embryos, persistent enhancer N1 activity owing to a high Wnt3a level in the mesodermal compartment causes ectopic Sox2 expression, which is followed by expression of other neural specification genes, indicating that the Sox2 expression provides the ground state for further neural development.

The long-held view that segregation of three germ layers determines cell lineages is challenged by cell lineage analyses1,2,7,8 and this study, which indicate that a pool of bipotential precursor cells in the CLE serves as axial stem cells that concordantly produce neural tube and paraxial mesoderm. This characteristic of CLE is continuous with the chordo-neural hinge at later stages8,10. Thus, the three germ layers describe spatial organization of tissues, but do not indicate the process of tissue derivation. Our evidence reinforces the axial stem cell model by providing the mechanistic basis for the fate choice, namely that Tbx6-dependent regulation of Sox2 determines the neural versus mesodermal fates in the axial stem cells derived from the CLE.

Methods summary

In situ hybridization and immunostaining were performed to analyze expression of relevant gene transcripts and protein products, respectively. Targeted knockout mice were used to analyze the consequence of loss in embryo of Tbx6 and/or Sox2 enhancer N1, and transgenic mice to monitor enhancer N1 activity or to ectopically express Sox2. EMSA analysis was to examine possible interaction of Tbx6 with enhancer N1 sequence.