The Sox3 gene is located on the X-chromosome in mammals, but it is autosomal in other vertebrate species. It is located on chromosome 4 in chicken, on scaffold GL172698.1 (without chromosomal assignment) in Xenopus tropicalis, on chromosome 14 in zebrafish, and on chromosome 10 in medaka (Figure S1). As the majority of enhancer sequences are included in the sequence blocks that are strongly conserved across the species (>60% DNA sequence identity over a length of 100 bp) [27], we compared the distribution of CSBs. As shown in Figure 1 and Table S1, many sequence blocks conserved between chicken and mammals and/or between chicken and Xenopus were identified both upstream and downstream of the Sox3 gene. In fish genomes, a fraction of the blocks were identified, but the degrees of sequence conservation were generally low (data not shown). CSBs present in the chicken genome were numbered from 1 to 47. These blocks were located in the region that was between 134 kb upstream and 97 kb downstream relative to the Sox3 open reading frame (ORF) start site (231 kb total), whereas in mammalian genomes the same conserved sequences were distributed in wider regions (940 kb in the mouse and 1.4 Mb in the human). Upstream CSBs from block 1 to 11 tend to be shorter (average 163 bp) compared to the entire CSBs (average 309 bp, excluding ~1 kb block 23 that included Sox3 ORF). However, downstream of block 47, no CSBs were found between chicken and other animal species. From these observations, the 231-kb span of the chicken genome encompassing the Sox3 gene was subjected to an enhancer survey.

We utilized the tkEGFP vector which by itself does not express EGFP upon electroporation in chicken embryos but expresses EGFP in a tissue-specific manner when an enhancer-bearing genomic fragment is inserted [27,36]. We initially analyzed the Sox3-proximal 50 kb region using bacteriophage clones derived from a chicken genome library. The overlapping DNA fragments C1 to C13 were prepared (Figure 2), inserted into the tkEGFP vector, and assessed for enhancer activities. Two regions of the genomic sequence exhibited enhancer activities: C4/5 in the st. 11 lens, and C12/13 in st. 11 diencephalon and spinal cord (Figure 3 and Figure 4, Table S2). The C12 sequence and the C13 sequence included in C12 exhibited identical enhancer activity. The C4/5 sequence included three CSBs (17, 18 and 19). The cloning of individual CSBs indicated that the CSB19 sequence was responsible for the lens enhancer activity of C4/5. The CSB26 sequence included in the C12 sequence exhibited enhancer activity in the diencephalon and spinal cord, which was identical to the activity of the C12 sequence.

These enhancers accounted for only subdomains of Sox3 expression in embryos. Therefore, we extended the region of the enhancer survey to a 231-kb genomic span. We identified two BAC clones that covered the upstream and downstream regions of the central 50-kb region. We prepared a series of ~5 kb DNA sequences that spanned the upstream and downstream regions with 1 kb terminal overlaps, resulting in 28 upstream sequences (U1–U28) and 17 downstream sequences (D1–D17) that were external to the central 50-kb sequence (Figure 2), and examined their enhancer activities. The number of electroporated specimens for each DNA sequence is indicated in Table S2, where the same specificity of an enhancer was reproducibly demonstrated.

In the upstream region, we identified only one sequence, U3, which showed enhancer activity in the epibranchial placode region at st. 11. However, in the downstream 97 kb region, ten sequences showed enhancer activities of various tissue specificities (Figure 3 and Figure 4, Table S2). The U3, D1, D3, D5 and D8 sequences showed enhancer activities in placode derivatives (Figure 3). The D3 sequence showed an otic placode-specific enhancer activity, while the D8 sequence displayed an epibranchial placode-specific activity that was similar to the aforementioned U3 enhancer. Other placodal enhancers also showed enhancer activity in the CNS as well.The D1, D4, D5, D6, D11, D12, D13 and D15 sequences showed enhancer activities in various portions of the developing CNS (Figure 4). The enhancer activity of the D1 sequence was of particular interest because of its strong activity in the caudal lateral epiblast, which is similar to the Sox2 N1 enhancer [27,29]. The D1 enhancer had additional activity in the posterior neural plate that forms the rhombencephalon and spinal cord. The D1 sequence included CSB27, and the isolated CSB27 sequence alone displayed the same activity (see Figure 5C, below). This CSB enhancer was renamed NP1 and analyzed further. The sequence D5 enhancer showed activity in the CNS anterior to the rhombencephalon and in the lens placode, indicating a resemblance to the N3 enhancer of Sox2.

The tissue territories of activation of the Sox3 neural enhancers are compared in Figure 4 following the time course. Shortly after the electroporation of the tkEGFP reporter vector, the NP1 (CSB27) enhancer was activated at st.5 in the region abutting the anterior primitive streak, which is similar to the Sox2 N1 enhancer [27,29]. Synchronous with primitive streak regression and the posterior migration of Hensen’s node, the peak position of the NP1 enhancer activity moved posteriorly. A moderate level of enhancer activity remained in the region where the enhancer was once activated, namely, in the rhombencephalon and spinal cord. Following the activation of NP1, the D13 and D15 enhancers were activated at st. 8, in the prospective diencephalon, rhombencephalon and anterior spinal cord (D13/15), and in the prospective telencephalon (D15). Later at st. 9, enhancers D4, D5 and D11 were activated in the prospective telencephalon (D4), the prospective di/mesencephalon (D5) and the spinal cord (D11). At st. 11, the CSB26 enhancer was activated in the prospective diencephalon and the medial axial levels of the spinal cord, and the D12 enhancer was activated in the ventral diencephalon. Next, at st. 15, the D6 enhancer was activated in the mes/rhombencephalon and spinal cord. These enhancers produced various overlapping patterns of their activities in the CNS primordium.

In order to examine the NP1 enhancer regulation that was analogous to the Sox2 N1 enhancer activation in axial stem cells, the DNA sequence of the NP1 enhancer was investigated in detail. A comparison of the CSB27 sequence (369 bp in the chicken) in the four vertebrate species indicated the existence of two conspicuous blocks of high sequence conservation, which were positions 52–177 (Block A) and 185–269 (Block B) (Figure 5A).

Starting from a 1681-bp sequence, which included the 369-bp CSB27 sequence, the specific region that accounted for the enhancer sequence was narrowed down in the following steps (Figure 5B). The CSB27 sequence showed activity that was identical to the 1681-bp sequence, whereas the immediate upstream 941-bp sequence showed no activity (Figure 5B(a)). The deletion analysis shown in Figure 5B(b) indicated that the 52 to 269-bp region of the CSB27 sequence was sufficient for full enhancer activity, whereas the sequences bearing only one of the two high-conservation blocks (1–180 bp or 181–369 bp) did not show enhancer activity (Figure 5C). The 5’ step-wise deletions indicated that the 1 to 81-bp region was dispensable, but further deletion to the 92-bp position inactivated the enhancer. The 10-bp 3’ deletion from the 82 to 269-bp sequence inactivated the enhancer. Taken together, these results indicated that the 82 to 269-bp sequence was required for NP1 enhancer activity.

The tissue domain that first activated the Sox3 NP1 enhancer, namely the caudal lateral epiblast abutting the primitive streak, overlapped with that of the Sox2 N1 enhancer as indicated in Figure 6A(a). The N1 enhancer is activated by the combined action of Wnt and Fgf signals [29,33], which is consistent with the requirement of both signals in the maintenance of the axial stem cells that activate the N1 enhancer [32,34]. Thus, whether the NP1 enhancer is also regulated by Wnt and Fgf signals was investigated. The 52-269 sequence of CSB27 contained several potential Lef1 binding sequences and putative Fgf-responsive elements (Figure 6A(b)).

When the expression vector for Dkk1, which is an antagonist of canonical Wnt signaling, was co-electroporated with NP1-tkEGFP, the NP1 enhancer was inactivated (Figure 6B(b)). In contrast, when the expression vector for stabilized β-catenin, which constitutively activates the canonical Wnt signal pathway, was co-electroporated, the NP1 enhancer was activated in the entire embryo (Figure 6B(c)). These results demonstrated the canonical Wnt signal-dependent activation of the NP1 enhancer. Analogously, co-electroporation of a vector to express a soluble form of the Fgf receptor 1 [FGFR1c(ECD)-Fc(IgG2a)] that titrates Fgf molecules, or addition to the culture medium of SU5402, a specific inhibitor of Fgf receptor tyrosine kinase, inactivated the enhancer (Figure 6B(d)-(f)), indicating that NP1 enhancer activation also depended on Fgf signaling. These findings indicated that the Sox3 NP1 enhancer was regulated by Wnt and Fgf signals in a manner similar to that of the Sox2 N1 enhancer, which suggests that both Sox2 and Sox3 genes are regulated in an analogous manner in neural/mesodermal-bipotential axial stem cells. This would allow the regulation of these B1 Sox genes to be coordinated with the maintenance and differentiation of axial stem cells, which are also regulated by Wnt and Fgf signaling.

The basal promoter of the human Sox3 gene has been characterized as having binding sites for several tissue-nonspecific transcription factors [37,38,39], suggesting a limited contribution of the promoter in the tissue-specific regulation of the Sox3 gene. A survey of the mouse genomic sequences within 3 kb of the Sox3 gene with transgenic mouse embryos identified three regulatory regions that dictate transgene expression: the FB sequence of approximately 200 bp long that was located at approximately -1.2 kb in the coordinates given in Figure 1 for expression in the forebrain/midbrain /hindbrain; the V2 sequence that was located between -1.1 and 0.3 kb for expression in the V2 interneuron domain of the spinal cord; and the PNT-R sequence located between 2.8 and 4.1 kb, relative to the Sox3 translational start site for expression in the posterior neural tube, rhombencephalon and otic vesicle [35]. Both the FB and V2 sequences were included in the C7 sequence, and PNT-R was included in the C8 sequence (Figure 1). Generation of the partial expression pattern of Sox3 using a X. laevis Sox3-proximal sequence has also been reported [12].

In our present study, a region of a total of 231 kb in the chicken genome encompassing the Sox3 locus was systematically surveyed for enhancers with the region-scanning ~5-kb DNA sequences with terminal overlaps, and 13 new enhancers that direct gene expression in specific regions of the neural and/or sensory primordia were identified. A combination of these with the mouse-determined enhancer-like elements, account for various features of Sox3 regulation in early neuro-sensory development. However, it should be noted that our survey using chicken embryo electroporation did not detect the aforementioned Sox3-proximal elements that were determined using transgenic mouse embryos. One possible cause of this discrepancy was differences in the assay method. However, it probably reflected that a negatively acting element that was located around -5 kb of the Sox3 gene masked the enhancer activities. In transgenic mouse experiments [35], it has been reported that inclusion of this region inhibited forebrain activity of the FB element. Presence of an analogous enhancer-inhibitory element has also been reported for the X. laevis Sox3 upstream sequence [12]. The chicken DNA sequences (C7 and C8) covering the Sox3-proximal region were assessed as negative and included this -5 kb element, which could have masked the enhancer activities (Figure 2).

The Sox3 expression pattern in the chicken embryos and the activities of the newly identified enhancers are compared in Figure 3 and Figure 4. Sox3 expression at st. 5 covered both the anterior and posterior neural plates, whereas the NP1 enhancer was active in the posterior neural plate and especially in the posterior axial stem cell region. It is likely that the FB element that was determined with the mouse system [35] regulates Sox3 expression in the anterior neural plate. After st. 8, various enhancers that have activities in the subdomains of the forebrain-midbrain region, i.e., D13, D15, D4, D5, D11, CSB26, D12 and D6, were activated in sequence. This resulted in various overlapping patterns of enhancer activities. In addition, at the rhombencephalon level, plural sequences showed enhancer activity, namely NP1, D13, D15, D6 (Figure 4) and PNT-R [35]. In the spinal cord, many enhancers showed non-uniform activity along the antero-posterior axis, with exception of the D15 enhancer. The D13 and D6 enhancers showed strong activity in the anterior half of the spinal cord at st. 11, whereas the D11 and CSB26 enhancers showed activity in the medial portion of the spinal cord (Figure 4). In the posterior region of the spinal cord, the NP1 and PNT-R enhancers exhibited activities. These observations suggested modular regulation of Sox3 expression along the antero-posterior axis of the spinal cord that depended on the differential activities of the enhancers.

In sensory tissue development, the D5 enhancer was activated in the lens placode area at st. 11, and the CSB19 enhancer was activated at st. 12 and continued to be active until the later stages of lens development (Figure 3). Three enhancers exhibiting activity in the otic placode/vesicle have been identified: the otic placode-specific D3 enhancer, NP1 and PNT-R [35]. NP1 and PNT-R showed activity in the posterior neural plate, and the activation of these enhancers in both the otic placode/vesicle and the posterior spinal cord suggested a common signaling input in the development of the two different tissues. The U3 and D8 enhancers showed activities in the epibranchial placode regions.

We have previously shown that during the early stages of neural plate development, Sox2 regulation is divided into non-overlapping anterior and posterior territories that are regulated by the N2 and N1 enhancers, respectively [5,27,32]. This reflects the fact that the anterior and posterior neural plates develop from the epiblast through distinct cellular mechanisms [18,32,33]. Participation of the two enhancers FB and NP1 in Sox3 regulation in the anterior and posterior neural plate development, respectively, indicated that Sox2 and Sox3 are regulated on the basis of the same antero-posterior spatial division of the neural plate.

In a detailed analysis, an important similarity was found in the regulation of the Sox3 NP1 and Sox2 N1 enhancers. Both of these enhancers were activated in the epiblastic tissue abutting the anterior primitive streak in which the axial stem cells reside. Both were activated by the combined action of Wnt and Fgf signals, as inhibition of either of these signals attenuated their enhancer activities (Figure 6B) [29,33]. Multiple Lef1-binding sites are found in the NP1 sequence (Figure 6A(b)), and a pair of functional Lef1-binding sites have been determined in the N1 sequence [29,33]. Putative Fgf-responsive elements were also indicated in both sequences (Figure 6A(b)). These shared features of NP1 and N1 allow for the coordinated regulation of Sox3 and Sox2 when the axial stem cells give rise to the neural plate and paraxial mesoderm. However, their regulation was not identical. The arrangements of the Lef1 sites and putative Fgf responsive elements were very different between NP1 and N1. In contrast to the case of the N1 enhancer, which was strongly downregulated when the neural plate was formed, the NP1 enhancer maintained its activity in the spinal cord and was also activated in the otic vesicle.

Another analogous case had similar enhancer specificities of the Sox3 D5 enhancer and Sox2 N3 enhancer. These enhancers were activated in the forebrain, including the optic vesicle, and in the lens placode in st. 11 chicken embryos. It has been shown that the Sox2 N3 enhancer is activated by the cooperative action of Sox2 and Pax6 in the diencephalon, optic vesicle and lens placode [3]. Considering the similarities in the tissue specificity of the enhancer activity, it is possible that an analogous regulation is involved in the activation of the D5 enhancer. However, their regulation was not identical, while the D5 enhancer was active in the telencephalon at st. 11, the N3 enhancer lacked activity therein. Thus, the case of the D5 enhancer represented an additional example of the analogy between the Sox3 and Sox2 regulatory mechanisms, yet with an appreciable difference.

Regulation of the Sox3 gene in the neural plate progressively became more complex (Figure 4). Following the creation of the antero-posterior subdivisions by the FB and NP1 enhancers in the enhancer territories of the forming neural plate, more Sox3 enhancers with various regional specificities were activated in sequence. The same scenario holds true for Sox2-regulating enhancers [27,28]. However, the activities of individual enhancers differed between Sox3 and Sox2 with respect to their aforementioned specificities.

The genomic arrangements of the enhancers also differed between the Sox3 and Sox2 genes. For instance, the D5 enhancer that was activated in the ocular primordia was located ~50 kb downstream of the Sox3 gene, while the N3 enhancer was positioned ~15 kb upstream of the Sox2 gene. Whereas the major Sox3 enhancers were distributed in the region between -20 kb and 80 kb relative to the Sox3 gene [35] (Figure 2), the Sox2 enhancers were more widely distributed on both sides of the gene over the range of 200 kb [28]. Thus, with respect to their individual activities and genome organizations, the enhancers substantially differed between the Sox3 and Sox2 genes.

The wide expression of Sox3 in the neuro-sensory primordia is common to most vertebrate species, whereas the wide expression of Sox2 is limited to amniote animals [4]. Reflecting this fact, the individual CSBs (potential enhancers) of the Sox3 locus were strongly conserved among the genomes of Xenopus, chicken and mammals (Figure 1), which is in contrast to the high degree of conservation of Sox2-associated CSBs that are confined to amniotes [4].

Along with these general features, species-dependent episodic changes in the enhancers were also observed. The sequence of lens-specific CSB19 enhancer of Sox3 is conserved poorly in the mouse genome, and this may account for the absence of Sox3 expression during mouse lens development [6]. However, the Sox3 D3 sequence, which showed purely otic placode-specific enhancer activity (Figure 3), did not contain CSBs (Figure 2), suggesting that this enhancer was unique to avian genomes.

Conservation of genomic sequences downstream of Sox3, as indicated by the occurrence of CSBs common to chicken, Xenopus and mammals, ends beyond CSB47, which is located at approximately 90 kb in the chicken genome, whereas the conservation among mammalian sequences is extended a further 3’. Both upstream and downstream genomic regions of the Sox3 gene appeared to be expanded in the mammalian genomes compared to the chicken and Xenopus genomes. An interesting issue is whether this genomic expansion is correlated with the location of Sox3 on the X-chromosome in mammalian species. The arrangements of landmark genes in syntenic chromosomal regions indicated extensive rearrangements of the genomic domains around the Sox3 gene in various species (Figure S1), suggesting that the genomic region investigated in this study included the major components of the Sox3 regulatory sequences. Alternatively, the Sox3 gene may be regulated by distant-acting regulatory sequences that withstand extensive genomic rearrangements.

To analyze the enhancers and conserved sequence blocks (CSBs) in the 231-kb region spanning the Sox3 locus in the chicken genome, we used sequence data from Gallus gallus-4.0 assembly (GCA_000002315.2) chromosome 4 genomic scaffold sequence (NW_003763735 GPS_000848988) as a reference. The central region sequence, from -23456 to 29859, covered by our bacteriophage library [27] was determined by us (DDBJ Accession number AB753847). The translational start site of Sox3 was taken as +1 to indicate the genomic location relative to Sox3. This position corresponded to Gallus gallus chromosome 4 NW_003763735 sequence position 10474038, Homo sapiens chromosome X GRCh37 partial sequence position 139587000 (Sox3 is in reverse orientation), Mus musculus chromosome X GRCm38 partial sequence position 60893205 (Sox3 is in reverse orientation), Xenopus tropicalis genome scaffold GL172698.1 position 1000090. The CSBs were screened using VISTA Browser (http://genome.lbl.gov/vista/) using a threshold of >60% identity over a 100-bp length.

The bacteriophage clones that carry the Sox3-proximal region sequences, approximately 50 kb, were used to produce the Sox3-proximal sequence segments C1-C13, and the two BAC clones CH261-112C13 and CH261-98D21 (ENSEMBL) were used to produce a series of ~5 kb DNA sequences with 1-kb terminal overlaps, U1 to U28 (upstream) and D1 to D17 (downstream), by PCR. These isolated sequences were individually inserted into a ptkEGFP vector [27], and electroporated at 2 μg/μL into the dorsal side of st. 4 chicken embryos. The pCMV-mRFP1 vector was also included in order to monitor the region of successful electroporation. The activation of EGFP expression at various developmental stages in New’s culture was scored as enhancer activity.

pNP1[218 bp]-tkEGFP at 2 μg/μL, pCMV/SV2-mRFP1 at 0.6 μg/μL and pCAGGS-based expression vectors [40] at 2 μg/μL for the expression of Dkk1, stabilized β-catenin [29] or rFGFR1(ECD)-rFc(IgG2a) (gift of Claudio Stern) were electroporated in st. 4 embryos in New’s culture. The impact of the expression of these effectors was assessed 8 hours after electroporation.