Interaction of Sox1, Sox2, Sox3 and Oct4 during primary neurogenesis

Abstract

Sox1, Sox2 and Sox3, the three members of the SoxB1 subgroup of transcription factors, have similar sequences, expression patterns and overexpression phenotypes. Thus, it has been suggested that they have redundant roles in the maintenance of neural stem cells in development. However, the long-term effect of overexpression or their function in combination with their putative co-factor Oct4 has not been tested. Here, we show that overexpression of sox1, sox2, sox3 or oct91, the Xenopus homologue of Oct4, results in the same phenotype: an expanded neural plate at the expense of epidermis and delayed neurogenesis. However, each of these proteins induced a unique profile of neural markers and the combination of Oct91 with each SoxB1 protein had different effects, as did continuous misexpression of the proteins. Overexpression studies indicate that Oct91 preferentially cooperates with Sox2 to maintain neural progenitor marker expression, while knockdown of Oct91 inhibits neural induction driven by either Sox2 or Sox3. Continuous expression of Sox1 and Sox2 in transgenic embryos represses neuron differentiation and inhibits anterior development while increasing cell proliferation. Constitutively active Sox3, however, leads to increased apoptosis suggesting it functions as a tumor suppressor. While the SoxB1s have overlapping functions, they are not strictly redundant as they induce different sets of genes and are likely to partner with different proteins to maintain progenitor identity.

Introduction

Neural progenitor identity in vertebrates is regulated in part by the three members of the SoxB1 subgroup of transcription factors, Sox1, Sox2 and Sox3 (Bylund et al., 2003; Graham et al., 2003; Pevny et al., 1998), which maintain proliferation of these multipotent neural progenitors and prevent their differentiation. Overexpression of either Sox1, Sox2 or Sox3 expands the progenitor cell population and inhibits neuronal differentiation in P19 cells (Sox1: Pevny et al., 1998), the chick neural tube (Sox2 and 3: Bylund et al., 2003; Graham et al., 2003) and the Xenopus neural plate (Sox2 and 3: Bylund et al., 2003; Graham et al., 2003; Mizuseki et al., 1998; Pevny et al., 1998; Rogers et al., 2009). Since the SoxB1 proteins have similar expression patterns, protein structures and functions, it has been suggested that they have redundant roles during neural development (Collignon et al., 1996; Pevny and Rao, 2003). However, there are critical differences. The most striking being that only Sox1 has been shown to convert Xenopus ectodermal explants into neuronal tissue (Nitta et al., 2006), whereas Sox2 and Sox3 are not sufficient to induce the formation of neurons, but are necessary to maintain a neural progenitor population (Mizuseki et al., 1998; Rogers et al., 2009). Knockdown of Sox2 or Sox3 prevents both neural specification and neuronal differentiation in the Xenopus neural plate (Kishi et al., 2000), and reduction of Sox2 or Sox3 function in the chick spinal cord increases the number of differentiating neural cells at the expense of proliferating progenitors (Bylund et al., 2003). Furthermore, the earliest of the SoxB1s to be expressed, Sox3, may function as a competency factor, as knockdown of Sox3 prevents Noggin-mediated induction of neural tissue in Xenopus ectodermal explants (Rogers et al., 2009).

The SoxB1 proteins are expressed in distinct spatial and temporal domains during gastrulation, further indicating that each has a unique role during neural induction. In fish, chick, and frog, sox3 is expressed throughout the ectoderm prior to neural induction, and is then restricted to the neuroectoderm during gastrula stages (Koyano et al., 1997; Okuda et al., 2006; Penzel et al., 1997; Rex et al., 1997; Rogers et al., 2008; Wood and Episkopou, 1999). In contrast, sox2 is initiated in early gastrula embryos and expressed only in the neuroectoderm (Mizuseki et al., 1998; Nitta et al., 2006; Okuda et al., 2006; Rex et al., 1997; Wood and Episkopou, 1999), while sox1 is not expressed until the end of gastrulation in the anterior neuroectoderm (Nitta et al., 2006; Rex et al., 1997; Wood and Episkopou, 1999).

SoxB1 proteins are important regulators of neural development but their distinct functions have been difficult to resolve. This is in part due to their ability to compensate for each other’s loss such that in the loss-of-function mutants, early neural development appears relatively normal with only minor consequences to brain structure and function evident in later stages. This compensation appears to occur in vivo as evidenced by the up-regulation of sox2 and sox3 expression in neurospheres lacking sox1 (Kan et al., 2007) and by the up-regulation of sox3 in sox2 conditional knockout mice (Miyagi et al., 2008). Rescue experiments support the idea that their compensatory ability allows for only minor defects in loss-of-functions mutants. For example, addition of Sox1 rescues a decrease in cell proliferation in the chick spinal cord caused by Sox2 inhibition (Graham et al., 2003). Conversely, Sox2 compensates for the loss of Sox1 in telencephalic neurons (Ekonomou et al., 2005). Even though the SoxB1s can compensate for each other’s loss such that early development is relatively normal, they have distinct roles later in development. Sox3-null mice have craniofacial defects and paralysis due to a neural crest migration defect (Rizzoti et al., 2004; Rizzoti and Lovell-Badge, 2007), whereas Sox1-null mice suffer from epilepsy, have small eyes and severe neuronal deficits in the forebrain (Malas et al., 2003). A neural–specific conditional Sox2 knockout was used to study Sox2’s role in neural development since Sox2-null mice do not survive past implantation (Episkopou, 2005), and these mice have more severe phenotypes than the Sox1- or Sox3-null mice (Miyagi et al., 2008). The embryos have enlarged lateral ventricles and prenatal mortality (Miyagi et al., 2008).

The loss of any single SoxB1 in non-mammalian vertebrates also resulted in minor phenotypes, which reinforced the idea that the SoxB1s are redundant. Sox3 knockdown in zebrafish reduced the size and organization of the neural tube (Dee et al., 2008) and Sox3 knockdown in Xenopus did not result in an obvious phenotype but did inhibit Noggin’s induction of neural tissue in ectodermal explants (Rogers et al., 2009). Altogether, the current data indicate that the ability of SoxB1 proteins to maintain neural progenitors is redundant, but that they ultimately have unique roles in different tissues within the brain.

There is increasing evidence that the Sox family of proteins require partner proteins or co-factors to regulate their target genes (Pevny and Lovell-Badge, 1997; Wegner, 1999; Wilson and Koopman, 2002). However, only a few partner proteins have been identified and these are primarily members of the Pou family of transcription factors (Kuhlbrodt et al., 1998; Ma et al., 2000; Tanaka et al., 2004). The best-studied interaction is that between Sox2 and the PouV transcription factor, Oct3/4. The high level of interest in these two proteins is in part due to their role in ES self renewal and in reprogramming differentiated cells into induced pluripotent stem (iPS) cells (Ambrosetti et al., 1997; Avilion et al., 2003; Babaie et al., 2007; Kiskinis and Eggan, 2010; Takahashi and Yamanaka, 2006; Yuan et al., 1995). Both proteins are required for stem cell maintenance. In the absence of Oct4, sox2 is not expressed in ES cells and they differentiate into extraembryonic or trophectoderm tissue (Feldman et al., 1995; Nichols et al., 1998; Yuan et al., 1995). Proper levels of Sox2 are required for stem cell maintenance as both increased and decreased levels of Sox2 cue differentiation (Chew et al., 2005; Kopp et al., 2008) suggesting that high levels of Sox2 interfere with the formation or function of Sox2:Oct4 complexes and low levels of Sox2 result in too few complexes. In support of this, overexpression of Sox2 in ES cells, embryonic carcinoma cells, and the developing eye leads to a downregulation of Sox2:Oct4 target genes such as Nanog and FGF4 and decreased proliferation and premature differentiation (Boer et al., 2007; Kopp et al., 2008; Lin et al., 2009). These data indicate that expression of Sox2 and its target genes requires Oct4, and that the Sox2 overexpression phenotypes without the addition of Oct4 are potentially due to the disruption of the Sox2/Oct4 complex.

Xenopus has the unique characteristic of having three Pou5f1 genes: Oct25, Oct60 and Oct91/Pou91 (Ryan and Rosenfeld, 1997). Of these three, Oct91 is most effective at rescuing an Oct4-depleted mouse ES cell line (Morrison and Brickman, 2006) and thus is the likely Oct4 homologue and Sox2 partner. However, even though Oct3/4 is not expressed in the developing CNS of mammals, the expression pattern and loss-of-function experiments indicate that Oct91 plays a role in neural development in Xenopus. Oct91 is expressed throughout the ectoderm in the mid-blastulae, restricted to the neural plate during gastrulation when neural tissue is induced and remains in the neural tube at the end of neurulation (Morrison and Brickman, 2006). Knockdown of Oct91 in Xenopus inhibits neural induction as indicated by the repression of sox2 and zicr1 expression (Snir et al., 2006). Therefore, Oct91 is expressed at the right time and place to interact with the SoxB1 proteins in Xenopus neural development.

In this report, we have sought: (1) to delineate the overlapping but distinct roles for each SoxB1 protein in neural progenitor formation or maintenance by comparing the target gene expression profile; (2) to determine their level of cooperation with Oct91, and (3) to examine the consequences of constitutively-active SoxB1 proteins. Here we show, through overexpression studies in ectodermal explants, that Sox1 Sox2 and Sox3 are sufficient for neural induction and initiation of neurogenesis as evidenced by the induction of neural markers. However, this induction is insufficient for neuronal differentiation. We also show that Sox2 cooperates with Oct91 to maintain a neural progenitor identity, whereas high levels of Sox1 and Sox3 counteract high levels of Oct91 in ectodermal explants. Additionally, knockdown of Oct91 depresses the neural inductive properties of both Sox2 and Sox3. Continuous expression of Sox1 and Sox2 in transgenic embryos represses neuron differentiation and anterior development while increasing cell proliferation, whereas constitutively active Sox3 leads to increased apoptosis. Hence, we provide evidence that all SoxB1 proteins override BMP signaling and induce neural markers, that the synergy previously seen in mammals between Sox2 and Oct4 is conserved in frogs, and that the SoxB1 proteins are not simply redundant but rather display unique functions that warrant further investigation.

Methods

Embryo culturing and manipulations

Xenopus laevis embryos were obtained using standard methods (Sive et al., 2000) and staged according to Nieuwkoop and Faber (Nieuwkoop and Faber, 1994). Ectodermal explants were isolated from stage 8–9 embryos and cultured in 75% Normal Amphibian Medium (Slack and Forman, 1980). Explants were collected when uninjected sibling embryos reached stages 12 or 22.

Plasmid Construction

The bicistronic TMmCherry:2A:NLSeGFP vector (from M. Parsons) was cloned into pEF6/V5-HIS TOPO TA Expression Plasmid (Invitrogen). To replace mCherry with the soxB1 genes, the BamHI site upstream of 2A was mutated to an EcoRI site via site-directed mutagenesis with the following primer sequence: forward 5′-AGCTGTACAAGCAATTCGGAGCCACG-3′, (mutated nucleotides are underlined). Following mutagenesis, mCherry was removed by double digestion with SpeI and EcoRI. SoxB1s were inserted by PCR amplification that added SpeI to 5′ and EcoRI to 3′. Sox1 was amplified using primers forward 5′- GGACTAGTTGAATGTACAGCATGATGATGGA-3′ and reverse 5′- CGGAATTCGATGTGTGTCAGTGGCATGGT-3′. Sox2 was amplified using primers forward 5′- GGACTAGTATGTACAGCATGATGGAGACCG-3′ and reverse 5′- CGGAATTCCATGTGCGACAGAGGCAGCGT-3′. Sox3 was amplified using primers forward 5′-GGACTAGTATGTATAGCATGTTGGACACC-3′ and reverse 5′- CGGAATTCTTATATGTGAGTGAGCGGTA-3′. SoxB1s were cloned in frame starting at the ATG.

Primary Xenopus cell cultures and transfection

Primary cultures of Xenopus fibroblast cells were derived and cultured from stage 30 embryos using standard procedures (Goswami et al., 2003; Kanamori and Brown, 1993; Rafferty, 1969). The eyes, gut, skin and tail of the embryos were removed and the remaining tissue was surface-sterilized with gentamycin/ampicillin, homogenized with 70% trypsin solution (Gibco), and cultured in 70% Leibovitz’s L-15 (Invitrogen). Transfection with Fugene HD (Roche) was performed following the manufacturer’s directions and optimized to a ratio of 11 μl Fugene HD to 2 μg DNA.

mRNA and morpholino microinjections

Synthetic capped mRNAs were made by in vitro transcription using mMessage mMachine kits (Ambion). Amounts injected for analysis in embryos were as follows: 400 pg of sox3-GR (Kishi et al., 2000), 500 pg of sox1 (gift from T. Grammer and R. Harland), 500 pg of sox2 (Mizuseki et al., 1998), 500 pg of sox3 (Hardcastle and Papalopulu, 2000), 500 or 1000 pg of oct91 (Frank and Harland, 1992), 600 pg of each bicistronic soxB1, 300 pg of eGFP (Clonetech), and 300 pg of lacZ. Amounts injected for explant analysis were as follows: 15 ng Oct91 morpholino (Snir et al., 2006), 25 pg noggin (Knecht et al., 1995), 200 pg sox3-VP16 (Zhang et al., 2003), 500 pg for Low dose, 800 pg for High dose of soxB1 genes, 1000 pg for High dose of oct91, and for co-injections each are injected at the Low dose. All injections of soxB1 and oct91 are done at either a high dose or a low dose but only the data from the high dose injections are shown. The co-injections are done with each mRNA at the low dose for a sum equivalent to the single high dose.

Generation of transgenic embryos

Transgenic embryos were generated as originally described (Kroll and Amaya, 1996) with the following modifications: 250 ng of linearized DNA, 5 μl sperm diluent buffer (SDB) and 3.5 × 10⁴ nuclei/ml were incubated with 5 μl of metaphase oocyte cytoplasm extract until nuclei decondense (approximately 8 min) (for comprehensive methods see (Ishibashi et al., 2008)). This was diluted in SDB to 70 nuclei/μl with 520 μl of SDB and kept at 15 °C until injected into oocytes at a flow rate of 0.6 μl/min over approximately 2 s.

Whole-mount in situ hybridization and β-galactosidase assay

Whole-mount in situ hybridization (WISH) was performed as described (Harland, 1991; Hemmati-Brivanlou et al., 1990) with the following modifications: proteinase K, paraformaldehyde and acetic anhydride treatments were not done. For lineage tracing, β-galactosidase activity was visualized with X-gal (Research Organics). For in situ riboprobes, we used dioxygenin-labeled n-tubulin (Richter et al., 1988), epi-keratin (Jonas et al., 1985) and sox2 (Mizuseki et al., 1998).

Reverse transcription PCR

RT-PCR was performed as described (Wilson and Hemmati-Brivanlou, 1995). Total RNA was extracted from 10 ectodermal explants and cDNA was generated using random hexamer primers and reverse transcriptase (Bioline). In Fig. 6, however, RNA was prepared using RNAqueous Kit (Ambion) with 200 μl of lysis buffer per reaction. Each experiment was performed a minimum of three times. See Table S1 for primer list and PCR conditions.

Fig. 6. Oct91 knockdown depresses neural induction by Sox2 and Sox3.

(A) RT-PCR analysis of uninjected stage 12.5, 16 and 22 whole embryos (WE) or ectodermal explants. (B) RT-PCR of ectodermal explants that are either uninjected (UI), or injected with noggin (Nog), sox1, sox2, or sox3 alone or in combination (+) with Oct91 morpholino (MO). The caps are negative for m act and therefore do not have mesoderm contamination.

Immunohistochemistry

Immunohistochemistry was performed as described (Kyuno et al., 2003) with the following modifications: incubated with primary notochord anti-Tor70 (1:5) (Bolce et al., 1992) or somite anti-12/101 antibodies (1:500) (Kintner and Brockes, 1984); secondary antibody peroxidase-conjugated goat, anti-rabbit IgG (1:300 Upstate/Millipore) as described (Brivanlou and Harland, 1989; Dent et al., 1989); peroxidase visualized with DAB Chromogen tablets (Dako). Proliferating cells were detected with anti-phospho-Histone H3 (Ser10) antibody (Upstate/Millipore). For visualization, dehydrated embryos were cleared by using 2:1 dilution of benzyl benzoate to benzyl alcohol (BB:BA) (Hemmati-Brivanlou et al., 1990).

Terminal transferase assay

Embryos were injected with 25 pg of Sox2:2A:GFP or Sox3:2A:GFP DNA. Whole mount TUNEL assay was performed after (Hensey and Gautier, 1998) with the following modifications: TdT enzyme (NEB) and BMB blocking agent (Roche) were used. Labeled cells were visualized by WISH as described.

Results

Overexpression of Oct91 and SoxB1 expand the neural plate at the expense of epidermis and form posterior protrusions

We demonstrated previously that overexpression of Sox2 or Sox3 expands the neural plate in Xenopus by increasing cell proliferation, inhibiting epidermis formation and delaying neurogenesis (Rogers et al., 2009). To determine if Sox1 and Oct91 have similar effects on neurogenesis, we injected mRNA coding for each protein into one blastomere at the two-cell stage with GFP as a lineage tracer, thereby leaving one half of the embryo as an uninjected control. Like Sox2 and Sox3, Sox1 and Oct91 expanded the neural tube (Fig. 1A-D) and all four proteins inhibited epidermal formation as marked by decreased epi-keratin (epi-k) expression in late neurula embryos (Fig. 1E-H) (Rogers et al., 2009). Furthermore, Oct91, Sox1 and Sox3 repressed epi-k in lateral and ventral tissues discontinuous from the expanded neural tube (data not shown).

Fig. 1. Oct91 and Sox1 expand the neural tube and repress epidermal formation similarly to Sox2 and 3.

(A-D) Bright field and fluorescent microscopy of stage 17 embryos injected with Sox1, Sox2, Sox3 or Oct91 and GFP mRNA into one cell of a two-cell stage embryo. (E-H) WISH for epi-k expression in embryos injected with soxB1 and LacZ, or oct91 and GFP mRNA. Asterisk in E marks the injected side. The injected sides of Sox1 (180/190), Sox2 (135/152), Sox3 (81/81) and Oct91 (49/53) have repressed epi-k expression compared to their uninjected side. All views are dorsal with anterior to the left. Brackets in A-D compare the relative width of the neural tube of the uninjected (top) to the injected side (bottom).

Since ES cell progenitor maintenance is dependent on the cooperation of Sox2 and Oct4 (Tomioka et al., 2002) and the specificity of Sox protein function is dependent on co-regulators (Chew et al., 2005), we tested the effect of overexpression of SoxB1 proteins in combination with Oct91. Previously, we demonstrated that overexpression of Sox2 or Sox3 delays n-tubulin (n-tub; a neuronal marker) expression such that it is repressed at stage 14 and later rebounds from this repression likely due to degradation of the injected mRNA (Rogers et al., 2009). Here we show that in fact, n-tub expression rebounds from inhibition by Sox1, Sox2 and Sox3 by stage 15-16, but is still repressed by Oct91 at this stage (Fig. 2A compared to 2B-D) but rebounds by stage 23 (Fig. 2E-H). At stages 15 and 23, embryos co-injected with oct91 and soxB1 mRNA phenocopied those injected with Oct91 alone; n-tub was inhibited at stage 15 and expanded at stage 23 (Fig. 2A, E compared to 2M-R). However, by stage 30 the embryos co-injected with SoxB1 and Oct91 formed posterior protrusions not found in the single-injected SoxB1 or Oct91 injected embryos (Fig. 2I-L compared to 2T-V). The protrusions contain neural tissue, as demonstrated by n-tub expression but did not contain muscle or notochord tissue and consequently are not secondary axes (Supplemental Fig. 1). To rule out the possibility that doubling the total amount of injected mRNA was the cause of the posterior protrusions, we injected two times the amount of oct91 mRNA (Fig. 2S). A higher concentration of Oct91 did not recapitulate the posterior protrusions, indicating that Oct91 synergized with SoxB1 to form posterior protrusions. These data along with loss-of function studies which demonstrate that Sox2, Sox3 and Oct91 are required for the expression of neural progenitor markers (Kishi et al., 2000; Rogers et al., 2009; Snir et al., 2006), indicate that Oct91 cooperates with the SoxB1s to expand a neural progenitor pool that eventually differentiates into a mass of neurons.

Fig. 2. Oct91 and the SoxB1s synergize to induce posterior protrusions.

(A-L) WISH for n-tub in embryos injected in one blastomere of the four-cell stage embryo with LacZ and oct91, sox1, sox2, or sox3 mRNA. Arrowhead in B marks laterally displaced neurons. Ectopic neurons are seen at both stage 23 and 30 in all treatments. (M-V) WISH for n-tub in embryos injected in one ventral blastomere of a four-cell embryo with oct91 alone or in combination with sox1, sox2 or sox3, and LacZ. Asterisk in top row marks the injected side. In A-H and M-R the injected side is the bottom half of the embryo. In I-L and S-V only the injected side is shown.

SoxB1s and Oct91 are sufficient to induce neural tissue in naïve ectoderm

Previous studies concluded that Sox2 was insufficient to induce neural tissue in ectodermal explants excised from blastula embryos but when combined with FGF did induce neurons in explants excised from gastrula embryos (Mizuseki et al., 1998). One possibility is that the SoxB1 proteins require a partner protein to neuralize ectoderm. To address this possibility, ectodermal explants were isolated from embryos injected with mRNA coding for GFP and either each soxB1 alone (800 pg) or sox2 (400 pg) plus sox3 (400 pg) to determine if they cooperate with each other, or the dominant activator form sox3-VP16 which does not require a partner protein for strong activation of transcription (Emami and Carey, 1992; Wu et al., 1994). These were cultured until late gastrula (Fig. 3) and early tailbud (Fig. 4) and assayed by RT-PCR.

Fig. 3. Sox3 induces neural gene expression in gastrula-staged ectodermal explants.

(A) Diagram of ectodermal explant assay. mRNA is injected into both blastomeres of the two-cell embryo. Explants are excised at stage 8-9. Cultured explants that are ubiquitous for GFP are collected and processed by RT-PCR. (B) RT-PCR analysis of stage 12 ectodermal explants that are uninjected (UI) or injected with noggin, sox2 or sox3, sox2 plus sox3. A whole embryo (WE) control is also shown. In the co-injected caps, the total amount of RNA injected equals that of sox2 or sox3 injected singly. The caps are negative for bra and therefore do not have mesoderm contamination. (C) Diagrams of Sox3 and Sox3-VP16. Twenty amino acids were removed from the C-terminus and replaced with VP16 (Diagram adapted from (Kamachi et al., 2000).

Fig. 4. Sox2, Sox3 and Oct91 induce unique arrays of neural genes in ectodermal explants.

(A) RT-PCR analysis of stage 22 whole embryo (WE) or ectodermal explants from embryos injected with noggin (Nog), sox2 or sox3, sox2 (S2) with sox3 (S3), sox3-VP16 or uninjected (UI). (B) RT-PCR analysis of stage 22 ectodermal explants from embryos injected with oct91, or a low doses of oct91 with sox2 or sox3.

Sox2 and Sox3 overexpression induced different sets of markers in ectodermal explants. In gastrula stage explants, Sox2 induced expression of sox1, slug and the neural progenitor marker musashi (msi1, also know as neural related protein-1 or nrp-1) (Fig. 3B). Sox3 induced these and other markers including the early neural markers geminin (gem) and zicR1, the anterior neural marker eomesodermin (eomes, also known as Tbr2), and the proneural gene neurogeninr-1 (ngnr-1). Neither induced the neuronal marker n-tub and co-injections of Sox2 and Sox3 showed an expression profile similar to that of Sox3 (Figs. 3B, 4A). These data indicate that Sox2 and Sox3 are sufficient to inhibit epidermis formation and induce different subsets of early neural genes but are not sufficient to induce the formation of neurons, as noted by the absence of n-tub expression (Fig. 4A).

We showed previously that in embryos, Sox3 activates sox2 directly (Rogers et al., 2009), however Sox3 is not sufficient to induce sox2 expression in ectodermal explants (Figs. 3B, 4A). One possibility is that Sox3 requires a partner protein to induce sox2. In support of this, Sox3-VP16 is sufficient to induce sox2 and consequent n-tub expression (Fig. 4A). This suggests that in ectodermal explants, Sox3 is missing its partner protein and therefore, unable to activate sox2 expression.

A candidate partner protein for the SoxB1 proteins is Oct91, the Xenopus homologue of Oct4. Oct91 alone is sufficient for neural induction as marked by NCAM expression in ectodermal explants (Snir et al., 2006), and we determined that Oct91 also induced gem, sox3, sox2, sox1, eomes, msi1 and n-tub (Fig. 4B, stage 22 and Supplemental Fig. 2, stage 12). Knockdown of Oct91 in Xenopus embryos has been shown to correspondingly reduce sox2 expression (Snir et al., 2006).To determine the extent to which Sox2 and Sox3 cooperate with Oct91, we co-injected mRNA coding for either Sox2 or Sox3 with Oct91. The addition of Sox2 did not alter the induction of progenitor marker expression by Oct91. In contrast, Sox3 prevented Oct91 from inducing any neural markers even though epi-k was still repressed. These data are consistent with the interpretation that Sox2 cooperates with Oct91 to maintain a neural progenitor identity, whereas Sox3 and Oct91 counteract each other in ectodermal explants. These results were obtained in both gastrula (Supplemental Fig. 2) and neurula (Fig. 4) staged caps.

Sox1, expressed at the end of gastrulation later than both sox2 and sox3, induces the expression of n-tub and the pan-neural marker NCAM in late stage explants (stage 32) (Nitta et al., 2006). To determine the effect of Sox1 on early neural marker expression and if this is altered by the addition of Oct91, ectodermal explants were isolated from embryos injected with sox1, oct91, or oct91 plus sox1 mRNA. Individually, Sox1 or Oct91 induced expression of the progenitor markers sox3, sox2 and msi1 in stage 22 explants; however, together they activated only sox2 (Fig. 5). In summary, Sox1 and Sox3 both induce a broad profile of progenitor markers while Sox2 induces a small subset of these. Even though overexpression of Sox2 induces sox1 expression, the wide array of genes induced by overexpression of Sox1 is not expressed. Taken together, these results suggest that Sox1 and Sox3 are sufficient for neural induction as marked by expression of early neural and neural progenitor markers, and all three SoxB1 proteins counteract neuronal differentiation induced by Oct91 overexpression.

Fig. 5. Sox1 induces early neural markers and counteracts Oct91-mediated neuron formation.

RT-PCR analysis of stage 22 embryos (WE) or ectodermal explants from uninjected (UI) embryos, embryos injected with noggin (Nog), sox1 low, oct91 high, or oct91 plus sox1 (S1). GFP was injected in all treatments as a lineage tracer.

To determine if the SoxB1 proteins require Oct91 to induce neural marker expression, we knocked down the level of Oct91 proteins by injection of an Oct91 morpholino (MO) (Snir et al., 2006). Oct91 was shown previously to be expressed in uninjected ectodermal explants and embryos at stages 9 through 15 (Frank and Harland, 1992) making it a candidate partner for the SoxB1 proteins. Here we confirm this and show that its expression is not maintained in stage 22 explants (Fig. 6A). We also replicated experiments demonstrating that knockdown of Oct91 with a MO reduced expression of neural markers such as sox2 and msi1 (Snir et al., 2006) but not sox3, (data not shown). To test the role of Oct91 in neural tissue formation, we first injected Oct91MO with the neural inducer noggin, which resulted in the reduction of all neural markers assayed and the expression of epi-k (Fig. 6B). Similarly, the Oct91MO reduced neural marker expression and prevented the inhibition of epi-k by Sox2 and Sox3. In contrast, knocking down Oct91 did not alter the Sox1 overexpression phenotype. In summary, Oct91 is expressed in ectodermal explants during neural induction, and the knockdown of Oct91 in the presence of the neural inducer noggin or the Sox 2 and 3, depresses the activation of neural marker gene expression.

Constitutively-active Sox1 and Sox2 retard anterior development and repress neuronal differentiation in the CNS

Overexpression of each SoxB1 delays n-tub expression (Fig. 2) (Rogers et al., 2009), but n-tub expression eventually rebounds and is expanded. This rebound may be due to the depletion and eventual degradation of the exogenous soxB1 mRNA such that the expanded progenitor pool begins to differentiate. To compare the effect of a continuous supply of soxB1 mRNA to the transient overexpression studies, we overexpressed SoxB1 and GFP using a bicistronic construct with expression driven by the constitutively active EF1α promoter. Separating the two genes is the 2A oligopeptide from the picornavirus foot-and-mouth disease virus (FMDV) (de Felipe and Ryan, 2004) that mediates intraribosomal cleavage such that two proteins are generated from a single transcript and the 2A peptide is maintained as a tag on the first protein (Fig. 7A). To ensure the functionality of the system in Xenopus, we transfected Xenopus cells and injected embryos with a fluorescent bicistronic expression vector containing mCherry with a plasma-membrane localization signal, 2A, and eGFP with a nuclear localization signal. In transfected cells and embryos, both red fluorescence and nuclear green fluorescence were detected (Fig. 7B-H). mCherry was replaced by each of the soxB1 genes (Fig. 7I) and injection of the bicistronic mRNA into embryos expands the neural plate and inhibits epidermis formation as shown in Fig. 1.

Fig. 7. A constitutively-active bicistronic vector generates two proteins from a single transcript in Xenopus.

(A) A diagram for the bicistronic 2A vector shows that a single transcript forms two proteins. mCherry (pink) contains a plasma membrane localization signal (PmLS, white) and GFP contains nuclear localization signal (NLS, black). (B-E) Xenopus embryonic fibroblast cell transfected with the bicistronic system with vector in A. mCherry is in the plasma membrane (C) and GFP in the nucleus (D). (F-H) In vitro transcribed 2A vector mRNA injected into half of the two-cell stage embryo is visible under a dissecting fluorescent microscope in a neurula stage 17 embryo. (I) Bicistronic constructs used for the generation of transgenic embryos. Each SoxB1 (purple) is upstream of the 18 amino acid 2A viral sequence (blue) followed by nuclear localizing GFP (green) and poly-A signal (brown) and expression is controlled by the constitutively-active EFlα promoter (orange). Asterisk marks injected side. Diagram not to scale.

To determine the effect of constitutive soxB1 gene expression on embryo development after stage 15, we generated transgenic embryos using the bicistronic vector shown in Fig. 7I. Our average transgenic efficiency (TE) as determined by in situ hybridization for GFP is approximately 50% (Rogers et al., 2008). With these constructs, however, our TE as determined by GFP fluorescence was 8% (50/603 embryos). All of our transgenic embryos had mosaic GFP expression which indicate either late integration of the plasmid DNA, or that those embryos with ubiquitous expression failed during development. Late integration does not explain the extremely low efficiency, and we support the explanation that the ubiquitously expressing embryos did not survive.

Even though a low percentage of transgenic embryos were generated, a consistent phenotype was visible in neurula embryos expressing Sox1:2A:GFP and Sox2:2A:GFP. These embryos developed slower than the non-GFP-expressing sibling embryos generated from sperm nuclei injections, and have a smaller cement gland at neurula stages (data not shown). By tailbud stages, the heads of GFP-positive embryos fail to develop (16/17 for Sox1; 31/33 for Sox2) (Fig. 8).

Fig. 8. Sox1:2A:GFP and Sox2:2A:GFP transgenic embryos have decreased levels of n-tub and Sox2:2A:GFP have increased proliferation.

(A-C) WISH for n-tub. (A) A stage 30 sibling embryo generated from sperm nuclei transfer serves as a staging control. N-tub levels are reduced in the brain and spinal cord of transgenic embryos. (B) Sox1:2A:GFP and (C) Sox2:2A:GFP transgenic embryos. (D-F) WISH for sox2 in stage 30 transgenic Sox1:2A:GFP embryos. Sox2 is not expanded in embryos with small heads or absent heads. (G-I) Immunohistochemistry for phosphorylated histone H3 immunohistochemistry in Sox2:2A:GFP transgenic embryos. Embryos are shown at the same magnification. Hatched box indicates magnified view. (G’-I’) pH3 positive cells have been counted in the magnified views.

To determine the effect of constitutively-active Sox1 and Sox2 on neuronal differentiation, expression of the neuronal marker n-tub was analyzed in Sox1:2A:GFP and Sox2:2A:GFP transgenic embryos (Fig. 8A-C). There are ectopic neurons in Sox1 and Sox2 transgenic embryos, and even though these embryos lack n-tub in the region of the neural tube, they do express n-tub in the brain (10/12). Since neuron formation is reduced and disorganized in these transgenic embryos, we next asked if neural progenitors were expanded by assaying for sox2 expression in the Sox1:2A:GFP transgenic embryos and for phosphorylated histone 3 (pH3) in the Sox2:2A:GFP transgenic embryos. Since Sox1 activated expression of both sox2 and sox3 in explants, we expected it to expand expression of sox2 in the transgenics. However, Sox1:2A:GFP transgenic tailbud embryos had reduced sox2 expression (Fig. 8D-F). In contrast, Sox2:2A:GFP transgenic embryos (n=5) had a higher density of pH3 positive cells (an average of 45 pH3 positive cells per 1.2 mm² area) compared to non-GFP sibling embryos generated from sperm nuclei injections (average of 10 positive cells, n=5) (Fig. 8G-I’). These data indicate that continuous expression of sox2 led to an increased number of proliferating progenitors while continuous expression of sox1 led to ectopic neuron formation. In both cases, anterior neural development was stunted.

Embryos were generated with sperm nuclei transfer using the Sox3:2A:GFP bicistronic expression vector (n=631), however no GFP-positive embryos were generated. We knew that the mRNA was functional because injection of Sox3:2A:GFP mRNA into embryos gave a predictable Sox3 overexpression phenotype (Fig. 2). Therefore, we hypothesized that the transgenic cells that were constitutively expressing Sox3 within the mosaic transgenic embryos did not survive. To determine if cell death was the cause for a lack of GFP-positive embryos, we injected 25 pg of Sox3:2A:GFP DNA into the zygote (Fig. 9). As a control, we also injected 25 pg of Sox2:2A:GFP DNA. Since we injected linearized plasmid DNA into embryos, we expected to see mosaic GFP expression in cells near the injection site. In the Sox2:2A:GFP injected embryos, GFP-positive cells were visible by late gastrula (data not shown), however we did not see any GFP expression in individual cells in the Sox3:2A:GFP injected embryos. Instead, these embryos formed large fluid-filled cavities that were positive for GFP fluorescence while still being optically clear under normal light (data not shown). To determine if these edemas were caused by cell death, we used terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) to stain for apoptotic cells. At stages 11 and 25, there was a five-fold increase in the number of Sox3:2A:GFP embryos that had low (20 TUNEL positive cells per dorsal half) to high (50 positive cells) levels of apoptosis compared to Sox2:2A:GFP embryos. Therefore, we conclude that constitutively-active Sox3 led to increased cell death.

Fig. 9. Continuous expression of sox3 increases cell death.

TUNEL staining in (A) uninjected embryos or those injected with (B, E) Sox2:2A:GFP or (C, F) Sox3:2A:GFP plasmid DNA at the 1 cell stage. (A-C) Stage 11 embryos are animal pole views. (D-F) Stage 20 embryos are lateral views with anterior to the left. Numbers of embryos with the represented phenotype are in the lower right corner. (G) Percentages of embryos with relatively high (more than 50 cells positive), low or no staining.

Discussion

The homologous SoxB1 proteins Sox1, Sox2 and Sox3 are expressed in self-renewing progenitors throughout the development of the CNS, and gain- and loss-of-function studies in vertebrate model organisms indicate that Sox1, 2 and 3 are required for the maintenance of neural stem cells (Bylund et al., 2003; Graham et al., 2003; Pevny et al., 1998). Distinct functions have been difficult to tease apart as the SoxB1 proteins compensate for each other’s loss such that mutants have mild phenotypes indicating that the SoxB1 proteins are functionally redundant (Ekonomou et al., 2005; Graham et al., 2003; Kan et al., 2007). The present study demonstrates that although the overexpression of Sox1, 2 and 3 results in similar phenotypes, each protein activates a different profile of genes and has a different effect when constitutively expressed in explants, transgenic embryos or with a putative partner protein, Oct91.

Sox1, Sox2 and Sox3 have overlapping and distinct roles in neural specification and differentiation

We demonstrated that overexpression of Sox2 or Sox3 expands the neural plate, inhibits epidermis formation, and inhibits neurogenesis embryos (Rogers et al., 2009) and here we show that overexpression of Sox1 or Oct91 exerts the same effect. To determine if these genes use the same mechanism to effect these changes, we compared the profile of neural genes expressed in ectodermal explants in response to Oct91 and each SoxB1 protein. We determined that each induces a unique profile of early neural markers, with Sox1 and Oct91 also inducing expression of the neuronal marker n-tub. These data indicate that they all override BMP signaling to induce neural marker expression in explants but only Sox1 and Oct91 induce the formation of neurons. More importantly, these data indicate that each SoxB1 protein alters the expression of different target genes (Fig. 10).

Fig. 10. Summary of genes induced and cell fates altered by each SoxB1.

Overexpression of each SoxB1 in ectodermal cells inhibits epidermis formation, which is driven by BMP signaling. Sox3 overexpression leads to the formation of neural progenitors by inducing expression of gem, msi1, eomes, sox1, and FGF8. Sox1 and FGF8 are gray because their induction is not maintained past stage 12. For neuronal differentiation, Sox3 requires the presence of a co-factor. Sox2 induces expression of sox1 and eomes in explants. Sox1 induces expression of both sox2 and sox3 and ultimately results in differentiation and the formation of neurons.

Overexpression studies using transgenic embryos indicated that the different expression profiles induced by each SoxB1 lead to different phenotypes. Transgenic embryos constitutively expressing Sox1 or Sox2 were similar in that they had reduced anterior structures and lacked neurons in the neural tube but had ectopic neurons. Sox1 induced many ectopic neurons in non-neural tissue whereas Sox2 increased the population of proliferating cells and induced few ectopic neurons suggesting that Sox2 is more effective than Sox1 at maintaining cells in a progenitor state. In contrast, we were unable to generate any GFP⁺ transgenic embryos using Sox3:2A:GFP, and injection of EFlα:Sox3:2A:GFP reporter DNA or high levels of sox3 mRNA into fertilized embryos caused cell death even though embryos tolerate high levels of Sox1 and Sox2. Thus, Sox3 appears to have a dynamic role in cell proliferation; at high levels, Sox3 causes cell death but at lower levels induces proliferation.

Sox3 can influence either cell proliferation or death, and like Geminin, another early neural protein with a similar expression pattern and function (Kroll et al., 1998), exhibits both oncoprotein and tumor suppressor functions. There are no published reports of Sox3 being up-regulated in any neural cancers even though Sox2 has been identified in brain tumors (Dong et al., 2004; Wang et al., 2006),. However, sox3 is expressed in lung cancer cells (Güre et al., 2000; Vural et al., 2005), upregulated in lymphomas (Kim et al., 2003) and can cause cancerous transformation of chick embryonic fibroblasts (Xia et al., 2000). Furthermore, it has been suggested that sox3 is regulated by both oncogene and tumor suppressor proteins since the human sox3 promoter contains binding sites for the tumor suppressor USF and the oncogene Sp1 (Kovacevic Grujicic et al., 2005). One possibility is that the levels of sox3 are tightly regulated and the cells in the transgenic embryos with high levels of Sox3 apoptose because the constitutively-active promoter overrides the endogenous regulation mechanism that normally represses Sox3, thereby activating a tumor-suppressor pathway to cause cell death.

Gain or loss of Oct91, alters the SoxB1 phenotype and gene expression profile

Since paired homeobox proteins partner with Sox proteins to regulate gene expression (Wilson and Koopman, 2002) and the only SoxB1 partner identified thus far is Oct4 (Yuan et al., 1995), we examined the effect of the overexpression of SoxB1 and the Xenopus Oct4 homolog, Oct91. First we show that Oct91 alone, like Sox1, induces neuron formation in embryos and explants and the expression of neural progenitor, proneural, and neural crest markers in ectodermal explants. However when co-injected with Sox1 or 3, the majority of these neural markers are not expressed even though epidermis formation is repressed, indicating that at high levels they interfere with each other’s function in the CNS. In contrast, the addition of Sox2 does not interfere with or enhance the expression of progenitor markers by Oct91. However, when injected into embryos, all three SoxB1 proteins cooperate with Oct91 to induce the formation of neuron-filled posterior protrusions. Since this is phenocopied by overexpression of Xiro3, Xenopus Iroquois (Bellefroid et al., 1998), one possibility is that Sox3, Oct91 and a third protein (not in explants) induce expression of Xiro3. In support of this, recently is has been shown that Xiro3 and Sox3 have similar functions and are in a shared neural regulatory network (Yan et al., 2009).

Gain-of-function studies in explants indicated that Oct91 cooperated with Sox2 but not Sox1 or 3, however, loss-of-function studies showed that both Sox2 and Sox3 require endogenous levels of Oct91 to induce neural tissue and therefore, partner with Oct91 to activate early neural gene transcription. One possible explanation for the discrepancy between the gain- and loss-of-function data is that Sox3 requires different partner proteins at different points in development. Early in development, high levels of Oct91 interfere with the interaction of Sox3 and a partner/co-activator. Later in development Oct91 is required for Sox3 to activate expression. In Toto, these data indicate that while Oct91 is involved in the same pathway as Sox2 and Sox3, Sox1 is outside the purview of Oct91 and induces neural tissue with limited input from Oct91.

Precise levels of SoxB1s and Oct91 are necessary for proper neural development

Our data indicate that the expression of SoxB1 target genes and therefore SoxB1 protein function is dependent on the level of protein. For example, overexpression of Sox2 induces expression of sox1 but this level of sox1 is not sufficient to induce sox3 expression, which is upregulated in response to Sox1 overexpression. Additionally, both Sox3 and Oct91 are expressed in uninjected ectodermal explants, but are insufficient to repress epidermal formation or induce neural tissue formation in explants, even though they are capable of doing so when overexpressed. Their functional dependence on expression level is supported by the role of SoxB1 proteins levels in human diseases. For example, mutation of a single allele of Sox2 is sufficient to cause severe eye, urogenital, and esophageal defects and seizures (Fantes et al., 2003; Ferri et al., 2004; Williamson et al., 2006), whereas high levels of Sox2 are found in a range of brain tumors (Kong et al., 2008; Phi et al., 2008). Both high and low levels of Sox3 can deleteriously affect normal brain function and physiology such as mental retardation, growth hormone deficiencies and seizures (Bowl et al., 2005; Hamel et al., 1996; Hol et al., 2000; Laumonnier et al., 2002; Rousseau et al., 1991; Solomon et al., 2004; Stevanović et al., 1993; Woods et al., 2005). Haploinsufficiency is a common trait within the Sox family where one mutated allele of Sox8-10 can cause sex reversal or hearing loss and megacolon (Bondurand et al., 2007; Foster et al., 1994; Koopman, 2005; Pingault et al., 1998; Wagner et al., 1994). Together these reports indicate that the levels of one Sox protein may alter other Sox proteins and that normal brain function is sensitive to precise levels of the Sox family proteins even when brain morphology appears normal.

In conclusion, the SoxB1s and Oct91 can override BMP signaling in the ectoderm to induce neural tissue, which parallels their ability to induce iPS cells from mammalian cell lines (Ichida et al., 2009; Takahashi and Yamanaka, 2006). Furthermore, just as Sox2 and Oct4 together regulate pluripotency in mammalian embryonic stem cells, Oct91 cooperates with Sox2 to maintain neural progenitor marker expression in Xenopus. The SoxB1s may have overlapping functions but they are not strictly redundant as they have different roles in maintaining progenitor identity. They reach the same endpoint, progenitor maintenance, via different mechanisms.

Supplementary Material

01

Supplemental Fig. 1 - Posterior protrusions from Oct91 and SoxB1 co-injection do not contain notochord or muscle tissue.

Immunohistochemistry for muscle cells with anti12/101 (A-C) and notochord with antiTor70 (D-F) in embryos injected at 1 of 4 cells with a mixture of lacZ, oct91 and sox1, sox2 or sox3. Embryos stained for Tor70 have been cleared with BB:BA. Asterisk marks injected side.

02

Supplemental Fig. 2 – Sox2, but not Sox3, interacts with Oct91 in gastrula stage explants.

RT-PCR analysis of stage 12 whole embryo (WE) or ectodermal explants from uninjected (UI) embryos or embryos injected with noggin (Nog), sox2, sox, oct91, or oct91 with sox2 or sox3.