|Home | About | Journals | Submit | Contact Us | Français|
foxD5 is expressed in the nascent neural ectoderm concomitant with several other neural-fate specifying transcription factors. We used loss-of-function and gain-of-function approaches to analyze the functional position of foxD5 amongst these other factors. Loss of FoxD5 reduces the expression of sox2, sox11, soxD, zic1, zic3 and Xiro1-3 at the onset of gastrulation, and of geminin, sox3 and zic2, which are maternally expressed, by late gastrulation. At neural plate stages most of these genes remain reduced, but the domains of zic1 and zic3 are expanded. Increased FoxD5 induces geminin and zic2, weakly represses sox11 at early gastrula but later (st12) induces it; weakly represses sox2 and sox3 transiently and strongly represses soxD, zic1, zic3 and Xiro1-3. The foxD5 effects on zic1, zic3 and Xiro1-3 involve transcriptional repression, whereas those on geminin and zic2 involve transcriptional activation. foxD5’s effects on geminin, sox11 and zic2 occur at the onset of gastrulation, whereas the other genes require earlier foxD5 activity. geminin, sox11 and zic2, each of which is up-regulated directly by foxD5, are all required to account for foxD5 phenotypes, indicating that this triad constitutes a transcriptional network rather than linear path that coordinately up-regulates genes that promote an immature neural fate and inhibits genes that promote the onset of neural differentiation. We also show that foxD5 promotes an ectopic neural fate in the epidermis by reducing BMP signaling. Several of the genes that are repressed by foxD5 in turn reduce foxD5 expression, contributing to the medial-lateral patterning of the neural plate.
The neural ectoderm (NE) forms on the dorsal side of the vertebrate embryo in response to factors secreted from the dorsal blastula center (BCNE) and the Organizer that inhibit BMP signaling (De Robertis and Kuroda, 2004; Kuroda et al., 2004; Levine and Brivanlou, 2007). As a result, several transcription factors (TFs) are co-expressed in broad overlapping domains in the nascent NE (Fig. 1A). The mRNAs of several (geminin [gem], sox3, sox11, soxD) are detected throughout the dorsal ectoderm at the onset of gastrulation; soxD also is expressed in the ventral ectoderm until about stage 12. The transcripts of others (foxD5, sox2, zic1, zic2, zic3) are concentrated in a broad band near the blastoporal lip, although faint expression also extends through the gem/sox domain. The transcripts of a third group (Xiro1, Xiro2, Xiro3) are detected in two dorso-lateral bands near the blastoporal lip. Although a few of these NE genes are expressed maternally, most are expressed around the onset of gastrulation and continue to be expressed through neural tube stages (Fig. 1C). As the neural plate begins to form at the end of gastrulation, the expression domains of these genes begin to be confined to distinct domains, presaging their roles in regional identity and neural differentiation (Fig. 1B). foxD5 is strongly expressed along the midline and central region of the anterior neural plate. The sox genes are broadly expressed throughout the neural plate except in the regions expressing high levels of foxD5. gem is expressed similar to the sox genes, but more intensely in the anterior rather than posterior neural plate. The zic genes are expressed in the lateral regions of the neural plate, and the Xiro genes are expressed anteriorly in a lateral bar, and posteriorly in a longitudinal band lateral to the midline. Experiments in Xenopus demonstrated that most of these TFs are induced by reducing BMP signaling, and they all expand the neural plate when their expression levels in dorsal ectoderm are increased by mRNA injection (reviewed in Sasai, 1998; Moody and Je, 2002). In addition, some maintain an immature neural state while others promote the expression of bHLH neural differentiation genes. However, there are very few studies that describe how these genes interact to maintain NE cells as neurogenic, establish the neural plate and initiate neural differentiation. Understanding how these TFs relate to each other is fundamental for understanding the molecular regulation of the progression from the initially-induced NE to a patterned neural plate and then to neural progenitor cells poised to begin their various differentiation programs.
A number of NE genes appear to repress or delay the initiation of neural differentiation. For example, gem and zic2 repress neural differentiation genes and can counteract the formation of ectopic neurons produced by ngnr1 over-expression (Brewster et al., 1998; Kroll et al., 1998). foxD5 expands a marker of immature NE (otx2) and represses neural differentiation genes (Sullivan et al., 2001). In several animals, sox2 and sox3 are expressed throughout neural stem populations and must be down-regulated for neural progenitor differentiation (Misuseki et al., 1998a; Kishi et al., 2000; Bylund et al., 2003; Graham et al., 2003). There are few studies of sox11, but it appears to function in neural stem cells, their transition to neural progenitor cells, and in neuronal progenitors downstream of proneural genes (Uwanogho et al., 1995; Wegner and Stolt, 2005). Other NE genes promote the onset of neural differentiation. Although zic1 is required early for neural competence (Kuo et al., 1998), it also is required for the expression of soxD (a member of the Sox G group; Wegner, 1999), which causes ectopic neural masses that express neural differentiation genes (Mizuseki et al., 1998b); zic3 also induces neural differentiation genes (Nakata et al., 1997). In Drosophila, Iroquois genes are required for the activation of proneural bHLH genes (Gomez-Skarmeta et al., 1996), and in Xenopus, Xiro genes are expressed just prior to the earliest expressed bHLH neural differentiation genes, promote the onset of neural differentiation (Bellefroid et al., 1998; Gomez-Skarmeta et al., 1998), but suppress terminal differentiation into neurons (de la Calle-Mustienes et al., 2002).
It seems likely that together these NE genes coordinately regulate expansion of the NE, formation/patterning of the neural plate, and the onset of neural differentiation. However, our understanding of the functional and transcriptional relationships between them is woefully incomplete. Are the genes in each group and/or in each TF family simply redundant or do they have distinct roles in the transition from neural induction to the onset of neural differentiation? As a first step in answering these questions we analyzed the functional position of foxD5 amongst these 11 other TFs by: 1) reducing endogenous FoxD5 levels in the NE by targeted injection of anti-sense morpholino oligonucleotides directed against foxD5 (foxD5-MOs); 2) elevating foxD5 expression via mRNA injection into 16-cell blastomeres that give rise to either the NE or ventral epidermis; 3) testing the timing of the foxD5 effects on the TFs with a hormone-inducible construct; and 4) elevating the expression of the 11 other TFs by mRNA injection to test whether they alter foxD5 expression. These experiments indicate that foxD5 regulates each TF and a few of them feedback to down-regulate foxD5, contributing to the medial-lateral patterning of the neural plate. Furthermore, we demonstrate that foxD5 regulates these genes by both transcriptional activation and transcriptional repression. We provide evidence that gem, sox11 and zic2 are direct transcriptional targets of FoxD5, and that together they account for most of the effects of foxD5 on the remaining TFs. These experiments reveal that several NE genes coordinately expand the NE during gastrulation, specify neural plate formation and patterning, and regulate the initiation of neural differentiation. foxD5 is a critical upstream component of this transcriptional network and it functions via a transcriptional triad, consisting of gem, sox11 and zic2, to induce/expand those NE genes that maintain an immature, neurally-committed state, and inhibit those NE genes that promote bHLH neural differentiation genes.
Fertilized Xenopus laevis eggs were obtained by gonadotropin-induced natural mating of adult frogs (Moody, 2000). mRNAs were synthesized in vitro and injected at the indicated concentrations: foxD5 (150pg, Sullivan et al., 2001), n-geminin (N-terminal neuralizing domain, 20pg, Kroll et al., 1998), sox2 (200pg, Mizuseki et al., 1998a), sox3 (200pg, Penzel et al., 1997), sox11 (200pg, Hiraoka et al., 1997), soxD (200pg, Mizuseki et al., 1998b), Xiro1 (200pg, Gomez-Skarmeta et al., 2001), Xiro2 (200pg, Gomez-Skarmeta et al., 1998), Xiro3 (200pg, Bellefroid et al., 1998), zic1 (200pg, Mizuseki et al., 1998a), zic2 (100pg, Brewster et al., 1998), zic3 (100pg, Nakata et al., 1997), foxD5VP16 (100pg; Sullivan et al., 2001) and EnRfoxD5 (100pg; Sullivan et al., 2001). Test mRNAs were mixed with lineage tracer mRNA (nuclear and cytoplasmic β-gal, 100pg) and injected into one blastomere of the 16-cell embryo (Moody, 2000). Transcripts were injected into either a dorsal animal blastomere (D1.1) to target expression to the NE or a ventral animal blastomere (V1.1) to target expression to the ventral epidermis (Moody, 1987).
To generate a hormone-inducible foxD5 vector, the ligand-binding domain of the human glucocorticoid receptor (hGR) plus the myc-tag (MT) were released by digesting pCS2+-hGRMT (Hutcheson and Vetter, 2001) with BamHI and NcoI. This fragment was inserted into the BamHI/NcoI site upstream of the XenopusfoxD5 open reading frame replacing the MT in the pCS2+MT-foxD5 plasmid (Sullivan et al., 2001). The pCS2+-hGRMT-foxD5 plasmid was confirmed by sequencing and used as template to generate mRNA (foxD5-hGR). After mRNA injection, cells synthesize the fusion protein, but the hGR domain forms a complex with endogenous heat shock proteins that prevents the transcription factor from entering the nucleus (Mattioni et al., 1994; Kolm and Sive, 1995). To uncouple this complex and allow nuclear translocation, control and injected embryos were incubated in synthetic hormone (10 μM dexamethasone, Dex) according to published protocols (Kolm and Sive, 1995). To ensure that the foxD5-hGR construct functioned as expected, some injected embryos were treated with hormone immediately after mRNA injection (+Dex cleavage; see Fig. 5); these embryos phenocopied those injected with wild-type (wt) foxD5 mRNA (Sullivan et al., 2001). For the experiments reported herein, embryos were treated with hormone starting at several different time points prior to and during gastrulation; hormone was maintained in the medium throughout the culture period. Experiments in tissue culture and in whole embryos with similar hGR-fusion constructs indicate that robust protein activation occurs within 90 minutes after hormone treatment, and is maintained for several days (Hollenberg et al., 1993; Mattioni et al., 1994; Kolm and Sive, 1995; de Graaf et al., 1998). Therefore, we assume that the FoxD5-hGR protein was available to affect downstream targets throughout the culture period of our experiments. Embryos were fixed and analyzed at stages (st) 14/15 (for the 11 NE genes) or st19/20 (for ngnr1, neuroD) when cultured without Dex, or treated with Dex at cleavage or st7–9. Embryos were fixed and analyzed at st15/16 (for sox2, sox3, soxD, Xiro1-3) or st19/20 (for ngnr1, neuroD) when treated with Dex at st11–13. Some embryos were injected with foxD5-hGR mRNA and raised in the absence of hormone (w/o Dex; see Fig. 5); expression patterns of the other NE genes were altered in fewer than 10% of embryos indicating that the hormone-inducible construct has little effect in the absence of hormone, in accord with published accounts (Hollenberg et al., 1993; Mattioni et al., 1994; Kolm and Sive, 1995; de Graaf et al., 1998). In addition, uninjected control embryos were treated with Dex at cleavage stages; expression patterns of the other NE genes were identical to untreated embryos indicating that Dex treatment alone does not affect NE gene expression. To block protein synthesis, embryos were injected at the 16-cell stage with foxD5-hGR, then at st8.5 they were incubated in cycloheximide (Chx, 10 μg/ml; Kurth et al., 2005). After 30–40 minutes the Chx medium was supplemented with Dex and embryos allowed to develop until siblings reached st12.
Two morpholino antisense oligonucleotides (MOs) were synthesized to recognize the translational start site of all three foxD5 paralogues found in Xenopus laevis (5′-CAGACTCCTGGCTAAAGCTCATTGT-3′; 5′-TATACTCTGATGCTGGGTTTGTAGC-3′) (Gene Tools, LLC). An equimolar mixture of the two foxD5-MOs, or a standard control MO (cMO; Gene Tools, LLC) was microinjected (16ng) into blastomere D1.1, a major progenitor of the NE (Moody, 1987). A myc-tagged construct (foxD5-MT) containing the foxD5 wt 5′UTR was generated to assess foxD5-MO knock-down efficacy by immunohistochemical detection of the fusion protein. A myc-tagged, N-terminally truncated foxD5 mRNA (ΔNfoxD5), which has full FoxD5 function (Sullivan et al., 2001) but does not contain the appropriate 5′ sequence to bind the foxD5-MOs, was used to demonstrate that the effects of foxD5-MOs could be rescued specifically by exogenous FoxD5. The mRNAs for foxD5-MT and ΔNfoxD5, with or without foxD5-MOs, were injected, embryos fixed at gastrulation (st10.5–12), sectioned with a cryostat and processed for immunofluorescence as previously reported (Huang and Moody, 1995), using mouse anti-c-myc (Sigma; 1:1000) and Alexa Fluor 488 goat anti-mouse IgG (Invitrogen; 1:200) antibodies. Sections were viewed with epifluorescence and digital images collected using the identical exposure times.
For whole mount in situ hybridization, embryos were fixed, stained for expression of the nuclear βGal lineage tracer and processed using digoxigenin-labeled RNA probes for each of the genes listed above (Sive et al., 2000). Embryos were analyzed for whether the expression domain was expanded or decreased in size or staining intensity, compared to the expression in control embryos and to the uninjected side of the same embryo. For whole-mount immunostaining, foxD5-injected embryos were processed according to Faure et al. (2000) using a rabbit anti-phosphoSMAD1/5/8 antiserum (1:100; Cell Signaling). Stained nuclei in the ventral epidermis were counted within a 1 mm2 grid overlying foxD5-expressing cells (identified by the blue cytoplasmic βGal lineage tracer), cells on the uninjected side of the same embryo, and similarly located cells in control, uninjected embryos. Cell counts from the same embryos (injected versus uninjected sides) were compared by the paired t-test; those from different embryos (injected sides versus control uninjected embryos) were compared by the unpaired t-test.
Both cells of the 2-cell embryo were injected with foxD5 mRNA. Animal cap explants (ACs) were cut at st8.5 and collected at st12–13. Semiquantitative RT-PCR within linear ranges was performed as previously described (Huang et al., 2007). All PCRs were repeated at least 3 times. Some primers were obtained from previous publications (H4: Niehrs et al., 1994; zic1, zic2, zic3: Kato et al., 1999), and others were designed with MIT Primer3 software (Rozen and Skaletsky, 2000). The size of PCR product, cycle numbers and primer sequences for the latter are: geminin (224bp, 25), 5′-GGCAGGCACCAGGTTTAATA-3′ (forward), 5′-TCAGCTGATCATTCCACAGC-3′ (reverse); sox2 (160bp, 25), 5′-AGTCCACCTGTAGTCACCTCTTCTT-3′ (forward), 5′-GCTACTGAGGCACTCTGATAGTGTT-3′ (reverse); sox3 (162bp, 25), 5′-CAAACAGGACTTTTTGTTTTGTTCT-3′ (forward), 5′-TTCATGTCAAAGTCTTTAGAAACCC-3′ (reverse); sox11 (203bp, 25), 5′-GGCTCTGGATGAGAGTGACC-3′ (forward), 5′-TGATGAAGGGGATTTTCTCG-3′ (reverse).
foxD5 and 11 other TFs are expressed at about the same time in the nascent NE (Fig. 1), suggesting that they may coordinately regulate the acquisition of neural fate, formation and patterning of the neural plate and onset of neural differentiation. To determine the functional position of foxD5 among these genes we first tested whether foxD5 is required for their normal expression patterns by reducing endogenous levels of FoxD5 with foxD5-MOs. These effectively blocked translation of a myc-tagged full-length foxD5, but had no detectable effect on a myc-tagged construct (ΔNfoxD5) in which the 5′ sequences recognized by the foxD5-MOs were deleted (Fig. 2A). Embryos were analyzed by comparing the expression domains of NE genes on MO-injected versus uninjected sides. For all NE genes, injection of a control MO (cMO) rarely altered the NE gene domain in the stage 14/15 neural plate on the injected side (Fig. 2B, D); nearly all embryos were indistinguishable from uninjected control embryos. Embryos that were co-injected with ΔNfoxD5 mRNA and foxD5-MOs had normal expression domains on the injected side in the majority of embryos (Fig. 2C, D). Together, these assays demonstrate that the foxD5-MOs are both effective and specific in reducing FoxD5.
Injecting foxD5-MOs into a single 16-cell blastomere on one side of the embryo caused a significant reduction in the stage 14/15 neural plate domains of gem, sox2, sox3, sox11, soxD, zic2, and Xiro1-3; in contrast, the expression domains of zic1 and zic3 were expanded (Table 1; Fig. 2D, E). To determine whether foxD5 is required for the onset of the expression of the NE genes, similarly prepared embryos were analyzed during gastrulation stages. Four different effects were observed (Table 1; Fig. 2D, E). First, the NE domains of sox2, sox11 and soxD were reduced in a significant number of embryos from the onset of gastrulation (st10), indicating that FoxD5 is required for their initial expression. Second, for those NE genes that are expressed maternally (gem, sox3, zic2), very few embryos showed an effect until stage 12, suggesting that either the maternal mRNA is intact at the earlier stages, or that the onset of their zygotic expression does not require foxD5. Third, zic1 and zic3 were initially reduced by foxD5-MOs, but at st12 this effect changed to an expansion of their domains. This result suggests that foxD5 is required for their initial NE expression, but later inhibits their neural plate expression. Fourth, the Xiro genes were reduced starting at early gastrulation, but the frequency of affected embryos gradually increased over developmental time. This result suggests that foxD5 facilitates their initial expression, and is required for their maintenance. Together, these results demonstrate that foxD5 is required for the normal expression patterns of all 11 NE genes, but it has differential effects at different developmental times.
In accord with previous work (Sullivan et al., 2001), increasing foxD5 levels by mRNA injection into a NE precursor blastomere expanded the neural plate on the injected side, regardless of which NE gene was examined (70.9% of embryos, n=457). Expansion was observed in neural plate domains beyond the foxD5-expressing cells, which were identified by a nuclear βgal lineage marker, indicating a non-autonomous effect (see Fig. 3B-D as examples). In addition, foxD5-expressing cells showed several changes in the levels of their expression of the other NE genes (Table 1; Fig. 3). These cells showed markedly increased staining for gem and zic2 compared to adjacent NE cells at both gastrulation and neural plate stages (Fig. 3A, G). The sox genes were differentially affected. sox2 and sox3 were weakly reduced in foxD5-expressing cells at gastrulation, but this effect abated by neural plate stages (Fig. 3B, C). sox11 was weakly reduced at early gastrulation (st10–11), but strongly induced in foxD5-expressing cells from st12 through neural plate stages (Fig. 3D). soxD was unaffected through mid-gastrulation (st11.5), after which it was reduced (Fig. 3E). foxD5 reduced the expression of zic1 and zic3 from early gastrulation through neural plate stages (Fig. 3F, H); Xiro1, Xiro2 and Xiro3 also were repressed at gastrulation (86%, n=36; 90%, n=39; 88%, n=17, respectively) and neural plate (Fig. 3I, J, K) stages. Because zic1, zic3, soxD and Xiro1-3 expression domains overlap with that of foxD5 at early gastrulation (Fig. 1), the repression is likely due to increasing FoxD5 protein above endogenous levels. The effects of foxD5 on some of the NE genes were confirmed by RT-PCR; foxD5-expressing animal caps showed a significant increase in expression over controls for gem, sox11, sox2, sox3 and zic2 and a significant decrease in zic1 and zic3 (Fig. 3L). These data demonstrate that foxD5: 1) expands the neural plate; 2) up-regulates some NE genes; and 3) down-regulates others. They also indicate that the expansion of the NE by foxD5 occurs via cell-to-cell signaling, whereas the up-regulation and down-regulation of the NE genes is either cell autonomous or mediated by very short-range signaling.
To determine whether foxD5 could ectopically induce the expression of the other TFs, mRNA was injected into a blastomere precursor of the ventral epidermis. gem, sox11 and zic2 were ectopically induced in a high percentage of embryos (Fig. 4A, B). A few embryos showed weak ectopic expression of sox2, zic1 and Xiro2 and nearly half showed weak ectopic sox3 expression. The remaining four NE genes (soxD, zic3, Xiro1 and Xiro3) were not ectopically induced in any of the embryos. Because a microarray analysis of potential downstream targets of FoxD5 indicated that both bmp4 and bmp7 are repressed in animal cap explants by foxD5 (Yan, Neilson and Moody, unpublished), we tested whether the ventral ectopic induction of these NE genes resulted from reduced BMP signaling. Two epidermal genes regulated by BMP signaling (AP2 and epidermal keratin; Snape et al., 1991) were strongly repressed by ventral expression of foxD5 (Fig. 4C). Furthermore, szl, a secreted antagonist of BMP signaling that is normally expressed ventrally (Lee et al., 2006), is strongly induced in both the NE and ventral epidermis (Fig. 4D). Finally, the numbers of cells containing nuclear phosphorylated SMAD1/5/8, which indicates active BMP signaling (Faure et al., 2000), were significantly reduced in ventral epidermis expressing foxD5 compared to either adjacent uninjected ventral epidermis in the same embryo or in control, uninjected embryos (Fig. 4E). These data suggest that the non-cell autonomous expansion of the neural plate expression domains of the 11 NE genes is likely accomplished by foxD5 locally down-regulating the BMP pathway. It remains to be determined whether this is accomplished by directly down-regulating bmp gene transcription or up-regulating the expression of secreted BMP antagonists such as Szl.
Endogenous foxD5 is present as both maternal and zygotic mRNAs (Solter et al., 1999; Fetka et al., 2000; Sullivan et al., 2001). Because the above experiments provide foxD5 mRNA during cleavage stages, we used a hormone-inducible foxD5 (foxD5-hGR) to determine the time at which FoxD5 causes the above effects. Dex treatment of uninjected embryos did not alter NE gene expression in the neural plate (st14/15) nor did FoxD5-hGR have significant activity in the absence of Dex (Fig. 5). Adding Dex to the medium when foxD5-hGR injected embryos reached 32–64 cells to activate protein as soon as it is translated demonstrated that the hGR construct caused the same neural plate (st14) phenotypes as wt foxD5 at similar frequencies (cf. Fig. 3). The zygotic expression of foxD5 begins around st8–9 (Sullivan et al., 2001), so we first tested the effects of foxD5-hGR on neural plate phenotypes by adding Dex to the culture just prior to the onset of gastrulation (st9) and fixing embryos at st14/15. gem, sox11 and zic2 were strongly induced in similar intensity and frequency as by wt foxD5 (Fig. 5A). However, expansion of sox2 and sox3 domains and reduction of zic1, zic3, soxD and Xiro1-3 domains occurred much less frequently (Fig. 5B, C). These results indicate that the effects on neural plate expression of gem, sox11 and zic2 are due to zygotic FoxD5 activity at the onset of gastrulation, whereas the effects on the other nine TFs require earlier FoxD5 activity. Culturing foxD5-hGR embryos in Dex at ~st7 (1 hour after 16-cell blastomere injection) or ~st8 (2 hours after injection), however, did not significantly increase the frequencies of the phenotypes. This also was the case for ngnr1 and neuroD (Fig. 5C), two bHLH neural differentiation factors up-regulated by zic1, zic3, soxD and Xiro1-3. It must be kept in mind that it may take up to 90 minutes after Dex treatment for the hGR fusion protein to access its DNA targets (Hollenberg et al., 1993; Mattioni et al., 1994; Kolm and Sive, 1995; de Graaf et al., 1998). Thus, we can not be certain whether maternal or potential pre-MBT transcription (Yang et al., 2002) of foxD5 account for the effects on these TFs. Nonetheless, these data demonstrate that: 1) the effects on gem, sox11 and zic2 expression are mediated by zygotic FoxD5 by gastrula stages; 2) the effects on gem, sox11 and zic2 do not depend on the other NE genes; and 3) the effects on sox2, sox3, soxD, zic1, zic3 and Xiro1-3 partially require FoxD5 activity at least by early blastula stages.
To determine the time after which foxD5 no longer affects the genes involved in promoting neural differentiation, foxD5-hGR embryos were cultured in Dex at gastrulation and fixed at neurulation (st16) or neural tube (st20) stages. When Dex was added at st11 (mid-gastrulation), the foxD5 effects on soxD, Xiro1 and Xiro2 were abolished and those on Xiro3, ngnr1 and neuroD were further reduced. When Dex was added at st13 (end of gastrulation) the foxD5 effects on the latter three genes were nearly abolished. The time course over which the foxD5 effects are abolished suggest that soxD, Xiro1 and Xiro2 act upstream of Xiro3, ngnr1 and neuroD. Furthermore, this time course coincides with the down-regulation of foxD5 transcription at neurulation (Fig. 1; Sullivan et al., 2001).
Different Forkhead TFs are known to regulate transcription by both activation and repression of target genes (Carlsson and Mahlapuu, 2002; Wijchers et al., 2006). FoxD5 contains several regions (an acidic “blob” in the N-terminal portion, and a P/A/Q-rich region, a “region II” and a Groucho interaction motif in the C-terminal portion) that are described in other Forkhead proteins to contribute to both transcriptional activating and repressing activities (reviewed in Sullivan et al., 2001; Yaklichkin et al., 2007). Previously we used constructs that fused the DNA-binding domain to either the VP16 activating or the EnR repressing domains to demonstrate that FoxD5 expansion of sox3 and otx2 and reduction of neural patterning (en2, Krox20) and differentiation (ngnr1, neuroD) genes were mediated via transcriptional repression (Sullivan et al., 2001). Using these constructs we found that gem and zic2 were strongly induced in the NE and ventral epidermis in nearly every embryo by foxD5VP16, whereas only a few induced cells were observed after EnRfoxD5 injection in significantly fewer embryos (Fig. 6A, B). zic1, zic3, and Xiro1-3 were strongly repressed at high frequencies by EnRfoxD5 (Fig. 6A, C). Thus, foxD5 appears to act transcriptionally as both an activator and a repressor.
The effects of these constructs on the sox genes, however, did not clearly replicate the wt foxD5 phenotypes (Fig. 6A). sox2 was more strongly reduced by both constructs compared to wt foxD5 and soxD was more weakly reduced by both constructs. sox3 was reduced only by foxD5VP16. sox11 was weakly reduced at gastrulation stages by foxD5VP16, matching the wt foxD5 phenotype, but was strongly repressed by EnRfoxD5 and neither construct caused its up-regulation in the neural plate. These results indicate that the regulation of sox gene expression by wt FoxD5 involves not only specific DNA binding via the Forkhead domain included in the fusion constructs, but also activities mediated by other domains in the protein (e.g., acidic blob, P/A/Q-rich region, region II, Groucho-interactive motif) that are not included in the fusion constructs. These domains may interact with transcriptional co-factors or provide 3-dimensional structure to the protein that is required for its normal interaction with sox gene regulatory sequences. In summary, these data demonstrate that FoxD5 functions via multiple mechanisms (i. e, transcriptional activation, transcriptional repression, and non-DNA binding domain interactions) to affect NE gene expression.
These results do not indicate whether wt foxD5 or the fusion constructs act directly on the regulatory regions of the NE genes. An in silico analysis of the ~1600 bp upstream of the transcriptional start site of each of the 11 NE genes revealed at least one generic Forkhead consensus binding site (P. Grant and S.A. Moody, unpublished), suggesting that the above effects could be mediated directly. To test this in the whole embryo, we injected foxD5-hGR mRNA into an NE precursor blastomere, and when embryos reached st8.5 incubated them in Chx to prevent further protein synthesis. After 30–40 minutes, Dex was added to the culture medium to allow the previously synthesized FoxD5-hGR protein to access the nucleus. This approach is only possible for the three genes that were affected by foxD5 by Dex treatment at st9 (gem, sox11, zic2); the other 8 NE genes require the presence of FoxD5 during blastula stages at which time Chx treatment prevents further development. As reported in Fig. 5, Dex treatment without Chx induced expression of each gene (Fig. 7). Importantly, each also was induced in the presence of Chx after Dex was added to the medium; they were not induced in the absence of Dex (Fig. 7). These results indicate that gem, sox11 and zic2 are direct transcriptional targets of FoxD5.
As direct targets, gem, sox11 and/or zic2 would be expected to mediate at least some of foxD5’s effects on the other NE genes. To test this, each of these three genes was expressed in the NE precursor blastomere (Fig. 8A). Similar to foxD5, gem induced zic2 (Fig. 8C), and weakly repressed sox11 at gastrula stages but also induced sox11 at neural plate stages (42.6%, n=53; Fig. 8C). gem also caused a reduction in the expression of sox2, soxD, zic1, zic3, Xiro1 and Xiro3, but at low frequencies (Fig. 8A), indicating it does not account for the full spectrum of foxD5 effects on these genes. Similar to foxD5, sox11 induced gem and zic2, and reduced Xiro1-3, but only weakly repressed sox2, soxD, zic1 and zic3 at low frequencies (Fig. 8A). Similar to foxD5, zic2 induced gem, and reduced sox2, sox3, soxD, zic1, zic3, and Xiro1-3. It also reduced sox11 at gastrula stages, but unlike foxD5 and gem it never induced sox11 at neural plate stages (Fig. 8A). When expressed in the ventral epidermis, gem induced zic2 (47%, n=83); sox11 induced gem (94%, n=48), sox2 (45%, n=42), and zic2 (83%, n=53); and zic2 induced gem (94%, n=63), and soxD (28%, n=60). A summary of these effects illustrates that no single target gene accounts for the full range of the foxD5 effects, but together they account for nearly all of them (Fig. 9).
To determine whether gem, sox11 and/or zic2 are required for the foxD5 effects, we co-injected foxD5-MOs with mRNA for each gene (Fig. 8B). gem significantly rescued the effect of foxD5-MOs on sox3, sox11, soxD and zic2 (Fig. 8D); sox11 significantly rescued the effect on gem, sox3, zic2 and Xiro1-3; and zic2 partially rescued the effect on gem, sox3, sox11, zic3 and Xiro2. Thus, gem, sox11 and zic2 can rescue foxD5-MO effects on each other and sox3, but only one or two rescue the effects on soxD, zic3 and Xiro1-3 and none alone can rescue the effects on sox2 (Fig. 8D) or zic1. Together, these data demonstrate that gem, sox11 and zic2: 1) are directly regulated by foxD5; 2) coordinately regulate each other downstream of foxD5, and 3) are all required to differentially accomplish the full range of the foxD5 effects on the other 8 NE genes (Fig. 9).
To test whether the 11 NE genes have a reciprocal influence on foxD5 we injected NE mRNAs into both NE and ventral epidermis precursor blastomeres (Fig. 10). n-gem did not alter foxD5 expression in the NE. sox11, zic1 and zic2 weakly repressed foxD5 in only a small number of cases. Furthermore, none of these TFs induced ectopic foxD5 in the epidermis. These results corroborate the above experiments that these genes act downstream of foxD5 (Fig. 9). The remaining NE genes caused a down-regulation of endogenous foxD5 expression. sox2 and sox3 weakly repressed foxD5 in most embryos, and zic3, soxD and Xiro1-3 strongly repressed foxD5 in the majority of cases (Fig. 10). Interestingly, zic3, soxD, Xiro1 and Xiro3 each strongly repressed foxD5 expression at both early (st10.5) and late (st12) gastrulation, whereas Xiro2-mediated strong repression was only detected at late gastrulation. Furthermore, only sox3, which can down-regulate BMP target genes (Rogers et al., 2008), and Xiro3, which can induce several early neural markers (Bellefroid et al., 1998), caused ectopic expression of foxD5 in the epidermis (Fig. 10 insets), corroborating the placement of foxD5 upstream of this subset of NE genes. These data indicate that a subset of NE genes feedback to down-regulate foxD5 in the neural plate. These effects are consistent with the observations that: 1) foxD5 normally is down-regulated during neurulation when the process of neural differentiation begins, whereas the sox, zic and Xiro genes continue to be expressed throughout neural tube stages (Fig. 1C); and 2) as foxD5 expression becomes restricted to the medial region of the neural plate, sox gene expression becomes excluded from the midline and zic and Xiro genes become restricted to the lateral neural plate (Fig. 1B). Because the effects appear to be cell-autonomous, we postulate that a mutual repression between foxD5 and a subset of NE genes likely contributes to the medial-lateral patterning of the neural plate.
A number of TFs are expressed in the vertebrate NE in response to neural inductive signaling, but their roles in stabilizing neural fate, expanding the NE, patterning the neural plate and initiating neural differentiation have not been studied in detail. Understanding how these TFs relate to each other is fundamental for understanding the molecular regulation of the progression from the initially-induced NE to specified/committed neural stem and progenitor cells that are poised to begin their various neuronal and glial differentiation programs. Developmental events often are controlled by gene regulatory networks of TFs that control temporal and region-specific gene expression (Levine and Davidson 2005). Our studies demonstrate that foxD5 acts upstream of several TFs expressed in the NE that together: 1) stabilize and expand the NE; 2) pattern the nascent neural plate; and 3) regulate the onset of neural differentiation. These are particularly significant findings because although there are homologous genes in mouse (FoxD4/Fkh2; Kaestner et al., 1995) and human (FOXD5 [2q13]; Katoh and Katoh, 2004), no functional information is currently available for them (Tuteja and Kaestner, 2007).
Each of the NE genes studied herein is induced by inhibition of BMP signaling alone or in combination with Wnt, Nodal or FGF signaling at the onset of gastrulation. The zygotic transcription of these genes in the NE may be independently initiated by these common signaling pathways, or they may be sequentially activated in a transcriptional pathway(s). To begin to sort out the relationships between these TFs, we tested the position of foxD5 in the group by loss-of-function (LOF) and gain-of-function (GOF) assays. LOF assays show that foxD5 is required for the NE expression of sox2, sox11, soxD, zic1 and zic3 at the onset of gastrulation, and for gem, sox3, and zic2 beginning at late gastrulation. These data indicate that the first group either is directly transcribed by FoxD5 (as we show for sox11), or is dependent on foxD5’s ability to locally reduce BMP signaling (see next section). We posit that the second group is similarly affected, but the effect is delayed because their maternal mRNAs have not yet been degraded. Alternatively, these genes may only require foxD5 for the maintenance of their expression in the neural plate. LOF assay also show that the requirement for foxD5 by the Xiro genes gradually increases as the NE matures, suggesting that Xiro genes are activated by other factors but require foxD5 for their maintenance. Because increasing foxD5, gem, sox11 and zic2 repress the Xiro genes, we propose that their maintenance may be provided by the other NE genes that require foxD5 at the onset of gastrulation (e.g., sox2, soxD, zic1, zic3). The observation that zic1 and zic3 neural plate domains expand in the absence of foxD5 indicates that while the loss of foxD5 delays the onset of their expression, other factors can later initiate them. Together, the LOF assays indicate that while foxD5 is a critical upstream factor for each of the 11 NE genes, it is not the only factor required for the onset of their expression in the NE.
Assays that increase foxD5 show that this gene alters the expression of all 11 NE genes; some are induced, some are expanded and some are repressed. Conversely, none of the 11 NE genes up-regulates foxD5 expression in the NE. These data indicate that foxD5 functions in an upstream position in the NE. Because the effects on the NE genes were predominantly cell autonomous, either FoxD5 directly interacts with these genes, as the Chx experiments indicate for gem, sox11 and zic2, or very local signaling is involved. The observations that foxD5 is required for the expression of NE genes that are reduced by foxD5 GOF (zic1, zic3, soxD, Xiro1-3) suggest that these genes are indirectly regulated by FoxD5. This is consistent with our transcriptional network model (Fig. 9), but now requires experimental proof.
Previous studies indicate that many NE genes do not directly induce embryonic cells to adopt a neural fate but they are required to stabilize/maintain that fate and to expand the NE (reviewed in Sasai, 1998; Moody and Je, 2002). For example, zic1 causes the NE to be more sensitive to neural induction by Noggin (Kuo et al., 1998), sox11 induces neural markers by antagonizing Wnt signaling (Hyodo-Miura et al., 2002), and gem, sox3 and Xiro1 each antagonize some aspect of the BMP4 pathway (Kroll et al., 1998; Glavic et al., 2001; Gomez-Skarmeta et al., 2001; Rogers et al., 2008). Herein we show that FoxD5 also represses BMP signaling; it reduces the number of phosphorylated SMAD1/5/8-positive cells in the epidermis and reduces the expression of two ventral epidermal genes that normally are up-regulated by BMP4 signaling. These effects could be mediated by a direct repression of bmp transcription or by up-regulating the other NE genes that reduce BMP signaling. For example, gem expression is up-regulated by increased foxD5 and down-regulated by reduced foxD5; sox3 and Xiro1 also are down-regulated by reduced foxD5. Alternatively, the effects could be mediated by up-regulation of secreted signaling factors that antagonize components of the BMP pathway. Support for this latter mechanism includes: 1) the observation that the expansion of the neural plate caused by increased foxD5 occurs in a non-cell autonomous fashion, and 2) the up-regulation of szl, a secreted factor that inhibits tolloid proteinases which in turn degrade Chordin (Lee et al., 2006). Therefore, it is likely that neural fate stabilization and neural plate expansion involve multiple levels of regulation of the BMP pathway. It will be important to precisely define which of the above potential roles of foxD5 are required for this initial step in neural development.
None of the 11 NE genes induces foxD5 expression, indicating that they act downstream. However, a subset (sox2, sox3) weakly and another strongly (zic3, soxD, Xiro1-3) represses foxD5, suggesting that they are involved in a negative feedback loop that causes the loss of foxD5 expression in the neural plate as it matures. This loss may release the neural plate cells from their immature state and thereby allow neural differentiation to begin (see next section). In addition, the locations of the gene expression domains (Fig. 1B) in the neural plate suggest that the negative feedback also contributes to regionalizing medial-lateral domains. foxD5 expression becomes confined to the midline and central anterior regions, whereas the other NE genes are expressed in more lateral domains. The effects are cell-autonomous in nature, suggesting that these genes may be directly regulating foxD5, but this needs to be confirmed experimentally.
Based on the data presented herein, we propose a regulatory network in which FoxD5 acts upstream of the other 11 TFs to control the onset of neural differentiation and thereby promote the maintenance of an immature neural state (Fig. 9). Previous work showed that: 1) foxD5 expands a marker of immature NE (otx2) and represses neural patterning (en2, Krox20) and neural differentiation (ngn1, neuroD, n-tub) genes (Sullivan et al., 2001); 2) gem blocks neural differentiation genes by regulating SWI/SNF chromatin-remodeling proteins, maintains cells in the cell cycle and is down-regulated as neural stem cells differentiate (Luo et al., 2004; Pitulescu et al., 2005; Seo et al 2005; Seo and Kroll 2006; Kroll, 2007; Spella et al., 2007); and 3) zic2 represses bHLH neural differentiation genes (Brewster et al., 1998). Herein we show that: 1) foxD5 directly activates gem, sox11 and zic2; 2) its GOF effects on gem, sox11 and zic2 do not require the expression of the other NE genes; 3) gem, sox11 and zic2 in combination carry out most of the effects of foxD5 on the other NE genes; and 4) gem, sox11 and zic2 regulate each other’s expression. These observations indicate that foxD5, gem, sox11 and zic2 form a network rather than a linear path of transcriptional regulation (Fig. 9). To fully understand the gene interactions that comprise this network, it will be important to determine whether gem, sox11 and zic2 are required for foxD5 expression and what other genes also take part in the network. It should be noted that this early role for sox11 has not previously been reported.
Other NE genes are involved in the transition from immature neural stem to more restricted neural progenitor cells. sox2 and sox3 are necessary for neural differentiation (Kishi et al., 2000; Wegner and Stolt, 2005; Wang et al., 2006). Each maintains neural stem/progenitor cells in a proliferative state upstream of neuronal terminal differentiation genes (Li et al., 1998; Zappone et al., 2000; Bylund et al., 2002; Graham et al 2003; Ellis et al., 2004; Bani-Yoghoub et al., 2006; Wang et al., 2006). Likewise, sox11 is reported to be up-regulated as neural stem cells transition to neural progenitor cells, and later it maintains pan-neural genes in neuronal progenitors downstream of bHLH differentiation factors (Uwanogho et al., 1995; Bergsland et al., 2006). Thus, together these sox genes may function downstream of foxD5, gem and zic2 to promote the initial step from neural stem to neural progenitor cell. Our experiments indicate that foxD5 initially weakly represses sox2, sox3 and sox11, perhaps to delay this transition. By early neural plate, however, sox2 and sox3 are no longer reduced and sox11 is up-regulated by direct transcriptional activation. Thus, foxD5 is required for the stem-to-progenitor transition. Based on mRNA injections, the foxD5 effects on sox2 and sox3 are likely mediated primarily by zic2, but foxD5-MO rescue studies show that gem and sox11 also contribute to sox3 expression (Figs. 8, ,9).9). Interestingly, in chick Gem interacts in a complex at the N2 enhancer to promote sox2 transcription (Papanayotou et al., 2008). However, we found that gem neither ectopically induced sox2 expression nor rescued the foxD5-MO reduction of sox2, and Rogers et al. (2009) showed that sox2 and sox3 directly regulate gem expression. Clearly, there are multiple levels at which these NE genes can interact.
Finally, several NE genes are involved in moving from a transition state to the onset of neural differentiation. soxD, zic1, zic3, and Xiro1-3 also expand the NE when over-expressed, but they additionally promote neural progenitor cells and the initiation of neural differentiation gene expression (Bellefroid et al., 1998; Mizuseki et al., 1998b; Nakata et al., 1998; Aruga et al., 2002; de la Calle-Mustienes et al., 2002; Inoue et al., 2007). Previous work showed that foxD5 reduced the expression of neural differentiation genes, suggesting that one of its functions is to delay the onset of differentiation (Sullivan et al., 2001). Herein we confirm this conclusion by showing that foxD5 reduces the expression of all of the neural differentiation-promoting TFs.
For all six NE genes associated with neural differentiation, FoxD5 activity is required prior to the onset of gastrulation to mimic the full effect of wt foxD5 mRNA injections. We do not know whether the activity of maternal foxD5, early transcription of foxD5 at blastula stages or the input of other genes contributes to the foxD5 effects on these six NE genes. For example, the expansion of zic1 and zic3 in the absence of FoxD5 indicates that they are positively regulated by other factors. We also do not know if the repression of these six NE genes is direct. While this is suggested by the cell-autonomous nature of the effects, our data also show that intermediate genes could be responsible; zic2 represses all of them and gem and sox11 both contribute to their reduction. Because each gene contains a Forkhead binding consensus site, further studies are needed to determine if FoxD5 also binds directly to their regulatory regions.
In conclusion, elucidating the molecules and interactions that comprise the regulatory network that stabilizes and expands the NE, patterns the neural plate and controls the onset of neural differentiation is critical information for understanding how the vertebrate CNS forms. Although a large number of transcription factors have been identified that are expressed during these early developmental stages, there have been very few experiments that relate these genes to each other or define their functions in the NE. We demonstrate that foxD5 is a critical element in this regulatory network. It acts directly via gem, sox11 and zic2 to differentially regulate several downstream genes, the effect of which is to hold the NE in an immature state and to prevent the onset of neural differentiation. Further elucidation of how these different NE genes interact to regulate neural specification and differentiation should ultimately prove useful for regulating the expansion and differentiation of neural stem and progenitor cells.
We thank Eric Bellefroid, Elena Silva Casey, Jose Luis Gomez-Skarmeta, Rob Grainger, Timothy Grammer, Kristen Kroll, Roberto Mayor, Ariel Ruiz i Altaba, Thomas Sargent, Yoshi Sasai and Monica Vetter for providing plasmids. We also thank Rakhee Goel, Himani Majumdar and Lynne Mied for technical assistance. This work was supported by NIH grant NS23158 and the George Washington University School of Medicine.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.