ERF and ETV3L are retinoic acid-inducible repressors required for primary neurogenesis

Primary neurogenesis is a phenomenon associated with anamniote embryos wherein sensorimotor neurons, which are largely transitory in nature, arise from neural-competent tissue and enable the early development of swimming and feeding behaviors (Wullimann et al., 2005). Primary neurogenesis is preceded by neural induction, which requires inhibition of bone morphogenetic protein (BMP) signaling (Wills et al., 2010; Wilson and Hemmati-Brivanlou, 1995) together with active fibroblast growth factor (FGF) signaling (reviewed by Dorey and Amaya, 2010; Marchal et al., 2009; Wills et al., 2010). Neural induction leads to the expression of pro-proliferative and neural-fate stabilizing transcription factors such as Foxd4l1, Geminin, Sox2/3 and Zic-family genes (Branney et al., 2009; Marchal et al., 2009; Yan et al., 2009). The concerted action of these genes promotes proliferation and maintenance of immature neural precursors. Through an incompletely understood mechanism, neural progenitors of the deep neuroectoderm layer of the embryo exit from the cell cycle and differentiate into primary neurons (Chalmers et al., 2002) under the control of proneural and neurogenic transcription factors (reviewed by Rogers et al., 2009).

Retinoic acid (RA) has numerous effects on early development, mostly by acting as a differentiation agent and specifier of position along the body axes (reviewed by Maden, 2007; Niederreither and Dollé, 2008; Rhinn and Dollé, 2012). RA signaling is required for neuronal differentiation (Blumberg et al., 1997; Sharpe and Goldstone, 2000; Sharpe and Goldstone, 1997), inhibiting the expression of pro-proliferation genes while promoting expression of proneural and neurogenic genes (Franco et al., 1999). RA regulates the timing (Papalopulu and Kintner, 1996) and extent (Blumberg et al., 1997; Franco et al., 1999; Sharpe and Goldstone, 1997) of neuronal differentiation, but little is known about underlying molecular mechanisms.

One clue to how RA may be promoting neuronal differentiation came from cell culture systems where RA was shown to regulate expression of genes that facilitate cell cycle exit and differentiation, particularly in cancers (reviewed by Andrews, 1984; Gudas, 1992). FGF and mitogen-activated protein kinase (MAPK) signaling are pro-proliferative, whereas RA inhibits proliferation and promotes differentiation. FGF8 and RA signaling pathways oppose each other’s action in patterning the anteroposterior (A-P) and dorsoventral (D-V), axes and in the differentiation of neurons and somites (Diez del Corral and Storey, 2004; Duester, 2008; Moreno and Kintner, 2004). How RA regulates the molecular components of the FGF signaling pathway to encourage differentiation is poorly understood.

ETS proteins comprise a family of transcription factors targeted by extracellular signaling pathways and modified by MAPK signaling downstream of growth factor receptors, integrins or Ca⁺²/calmodulin-dependent protein kinases (Oikawa and Yamada, 2003; Sharrocks, 2001). ETS proteins can function as transcriptional activators or repressors that interact with other factors to facilitate combinatorial, context-specific regulation of gene expression (reviewed by Hollenhorst et al., 2011; Li et al., 2000; Mavrothalassitis and Ghysdael, 2000; Oikawa and Yamada, 2003; Sharrocks, 2001).

ETS proteins are implicated in early development of the central nervous system, eye and blood, and regulate cell growth, differentiation and apoptosis (reviewed by Hollenhorst et al., 2011; Oikawa and Yamada, 2003; Sharrocks, 2001). ETS repressors ERF (ETS2 Repressor Factor), ETV3 (ETS Variant Protein 3) and ETV3L (ETV3-like) are closely related genes (Hollenhorst et al., 2011; Laudet et al., 1999) that can regulate the switch between proliferation and differentiation in some cell types (Hester et al., 2007; Klappacher et al., 2002; Verykokakis et al., 2007). ERF and ETV3 may displace activating ETS proteins from promoters of cell cycle control genes, while recruiting co-repressor complexes to facilitate cell cycle arrest. ERF and ETV3 are phosphorylated by ERK, which renders them inactive (Carlson et al., 2011; Hester et al., 2007). Phosphorylation of ERF leads to its nuclear export, whereas phosphorylation of ETV3 inhibits its DNA binding (Carlson et al., 2011; Hester et al., 2007). Etv3 promotes cell cycle arrest and differentiation of macrophage progenitors (Klappacher et al., 2002). Erf inhibits proliferation of trophoblast stem cells, and encourages the differentiation of the chorion layer into the labyrinth in the extra-embryonic ectoderm (Papadaki et al., 2007). Thus, ETS repressors are ideal candidates for regulating a proliferation-differentiation switch in primary neurogenesis.

We have previously shown that some Ets genes were responsive to changes in RAR signaling in Xenopus embryos (Arima et al., 2005). Erf was upregulated by increased RA signaling in neurula embryos (Arima et al., 2005). Here, we show that morpholino oligonucleotide (MO)-mediated knockdown of Erf or Etv3 results in loss of primary neurons accompanied by paralysis, phenocopying RAR loss of function. Erf, Etv3/3l, Rarα and Rarγ inhibit early neural progenitor markers, while promoting differentiation of primary neurons. Loss of any of these gene products results in a paucity of primary neurons in the neurula and renders tadpole-stage embryos unresponsive to touch. This loss of neurons is due to increased proliferation in the neural plate and a perpetuation of neural progenitor identity, at the expense of differentiation. Erf and Etv3l are thus key effectors of the RA-mediated switch between proliferation and differentiation during primary neurogenesis.

Xenopus eggs were fertilized in vitro (Janesick et al., 2012) and embryos staged according to Nieuwkoop and Faber (Nieuwkoop and Faber, 1967). Embryos were maintained in 0.1×MBS until appropriate stages or treated at stages 7/8 with 1 μM agonist (TTNPB) and 1 μM antagonist (AGN193109) as described previously (Janesick et al., 2012). Embryos were injected bilaterally or unilaterally at the two- or four-cell stage with combinations of gene specific morpholinos (MO), mRNA and 100 pg/embryo β-galactosidase (β-gal) mRNA or 10 ng/embryo 10 kDa lysine-fixable rhodamine-dextran lineage tracer. MO sequences are listed in supplementary material Table S1. For all MO experiments, control embryos were injected with 10 ng standard control MO (GeneTools).

Embryos processed for whole-mount in situ hybridization were fixed in MEMFA, stained for β-GAL activity with magenta-GAL (Biosynth) and then stored in 100% ethanol (Janesick et al., 2012). For whole-mount immunohistochemistry, embryos were fixed in MEMFA for 1 hour, permeabilized in Dent’s Fixative (80% methanol, 20% DMSO) overnight and stored in 100% methanol.

Whole-mount in situ hybridization was performed as previously described (Janesick et al., 2012; Koide et al., 2001; Sive et al., 1998). Geminin, Foxd4l1, Neogenin, Sox3, Zic1 and Zic3 probes were prepared via PCR amplification of published coding regions from cDNA, adding a bacteriophage T7 promoter to the 3′ end: xGeminin (Kroll et al., 1998; McGarry and Kirschner, 1998), xFoxd4l1 (Sölter et al., 1999; Sullivan et al., 2001), xNeogenin (Anderson and Holt, 2002), xSox3 (Denny et al., 1992; Penzel et al., 1997), xZic1 (Kuo et al., 1998; Mizuseki et al., 1998; Nakata et al., 1998) and xZic3 (Nakata et al., 1997). Erf- and Etv3-coding sequences (Klein et al., 2002) were cloned into pBluescript II SK-. Probes were transcribed with MEGAscript T7 (Ambion) in the presence of digoxigenin-11-UTP as described previously (Janesick et al., 2012). A list of forward primers and reverse primers containing a T7 promoter can be found in supplementary material Table S2. N-tubulin was a gift from Dr Nancy Papalopulu (University of Manchester, UK). Zic2, Ngnr1, Myt1 and Dl1 were a kind gift of Dr Andrés Carrasco (University of Buenos Aires, Argentina) (supplementary material Table S3).

N-tubulin expression was quantitated using Adobe Photoshop and ImageJ. Bright-field images were desaturated of magenta to remove lineage tracer signal. Purple whole-mount in situ hybridization signal was replaced with black and the images converted to binary and cropped such that the injected and uninjected sides were equal in total area. ImageJ was used to quantitate the black pixels as % area fraction using the Analyze → Measure function. The area fractions obtained for each side were normalized to the total area for each embryo. Statistical significance was determined using the Wilcoxon signed rank test in GraphPad Prism v5.0 (GraphPad Software, San Diego, CA). P≤0.05 was considered statistically significant.

Embryos were photographed for fluorescence, then processed separately in individual wells of 96-well plates for whole-mount immunohistochemistry. Embryos were rehydrated in PBS, 0.5% Tween-20 and heated overnight at 65°C with shaking (Lin et al., 2012). After blocking in 2% blocking reagent (Roche), 10% FBS in MAB for 1 hour, embryos were incubated in a 1:1000 dilution of anti-phospho-histone 3 (Cat# 06-570, Millipore) or anti-PCNA (SC-7907, Santa Cruz Biotechnology) in blocking buffer at room temperature for 4 hours. Embryos were washed five times for 1 hour in MABT at 4°C, then blocked in 2% blocking reagent (Roche), 10% goat serum in MAB for 1 hour. AP-conjugated anti-rabbit IgG (Cat# A-3687, Sigma) was diluted 1:10,000 in blocking buffer and embryos incubated at room temperature for 4 hours. After five 1-hour MABT washes, BM purple staining followed by bleaching was performed according to standard methods (Janesick et al., 2012).

Embryos were photographed at 25× magnification in bright-field, with anterior always pointing left. Ovals were selected from each side of the neural plate, then desaturated and converted to binary in MATLAB. Phospho-Histone H3-positive or PCNA-positive staining was quantitated in MATLAB. Statistical significance was determined using the Wilcoxon signed rank test in MATLAB. P≤0.05 was considered statistically significant.

Transient transfections were performed in COS-7 cells as described previously (Chamorro-García et al., 2012). Briefly, COS-7 cells were seeded at 1.8×10⁴ cells/well in BD BioCoat poly-D-lysine eight-well culture slides in 10% CBS. Cells were transfected in Opti-MEM at ∼90% confluency. pCS2-mCherry transfection control plasmid (0.25 μg/well) was co-transfected with 0.25 μg/well of pCDG1-FLAG-hGR-Erf plasmids (or minus-FLAG control) using Lipofectamine 2000 reagent (Invitrogen). After overnight incubation, the medium was replaced with DMEM/10% resin charcoal-stripped FBS (Tabb et al., 2004) plus 1 μM dexamethasone (DEX) or 0.01% DMSO vehicle for 1 hour before fixing cells in 10% formalin in PBS.

Fixed, transfected COS-7 cells were washed with PBS, then permeabilized with 0.25% Triton X-100 in PBS for 10 minutes. Cells were incubated at room temperature with 5% goat serum in PBS for 1 hour to block nonspecific antibody binding, and then incubated for 1 hour with anti-FLAG-M2 (1:50, Sigma). Cells were washed with PBS and then incubated at room temperature with goat anti-mouse Alexa Fluor 488 (1:500, Invitrogen) in 5% goat serum for 1 hour. To visualize nuclei, cells were stained with 1 μg/ml Hoechst 33342 (Invitrogen) for 10 minutes. Fluorescent microscopy images were acquired with Velocity software on a Zeiss Axioplan 2 fluorescence microscope equipped with an ORCA-ER CCD camera (Hamamatsu Photonics). Images were merged with Adobe Photoshop. Images shown are representative of transfected (mCHERRY+) cells observed in each experiment.

Total embryo RNA was DNase treated, LiCl precipitated and reverse transcribed into cDNA as described previously (Janesick et al., 2012). First-strand cDNA was quantitated using SYBR green detection (Roche) in a DNA Engine Opticon Continuous Fluorescence Detection System (Bio-Rad) using primer sets listed in supplementary material Table S4. Each primer set amplified a single band, as determined by gel electrophoresis and melting curve analysis. In Fig. 2A, QPCR data were analyzed using the ΔΔCt method (Livak and Schmittgen, 2001) normalizing to Histone H4 (Janesick et al., 2012). Mann-Whitney statistical analysis was performed in GraphPad Prism v5.0. For Fig. 3L, QPCR data was analyzed by ΔCt normalizing to Histone H4, and correcting for amplification efficiency between Erf, Etv3 and Etv3l (Pfaffl, 2001). Error bars in Fig. 2A, Fig. 3L and supplementary material Fig. S2 represent biological replicates (multiple pools of five embryos from the same female) calculated using standard propagation of error.

The Erf expression construct was made by PCR amplification of the protein-coding regions of the Erf cDNA and cloned into pCDG1 (Janesick et al., 2012). pCDG1-hGR-Erf was constructed by two-fragment PCR using pCDG1-Erf and amino acid residues 512-777 of the hGR (provided by Ron Evans, Salk Institute, San Diego, CA, USA) as templates, and cloned into pCDG1 (supplementary material Table S5). hGR-Erf (Ser^246,251 → Ala^246,251) and hGR-Erf (Ser^246,251 → Glu^246,251) were also constructed by two-fragment PCR and cloned into pCDG1 (supplementary material Table S6). Plasmids were sequence verified and linearized with NotI; mRNA was transcribed using mMessage mMachine T7 (Ambion).

RA signaling is required for neuronal differentiation (Blumberg et al., 1997; Sharpe and Goldstone, 2000; Sharpe and Goldstone, 1997), inhibiting the expression of anti-neurogenic genes (e.g. Zic2) while promoting expression of proneural and neurogenic genes (Franco et al., 1999). As little is known about the underlying molecular mechanisms, we investigated how RA and its nuclear receptors promote primary neurogenesis.

Rarα and Rarγ are localized in the neural plate at stage 14, the correct time and place to regulate primary neurogenesis (supplementary material Fig. S1A,B) (Sharpe, 1992). Overexpression of Rxrβ and Rarα2 produced ectopic neurons, whereas dominant negative Rarα1 or Rarα2 resulted in loss of primary neurons (Blumberg et al., 1997; Sharpe and Goldstone, 1997). Treatment of embryos with the RAR-specific agonist, TTNPB, increased primary neuron formation within the neural plate (Fig. 1B); treatment with the RAR-specific antagonist AGN193109 led to the loss of neurons (Fig. 1C) and subsequent paralysis (embryos did not spontaneously move and were unresponsive to touch). This agrees with previous results using less specific chemicals, including RA, Ro 41-5253 and citral (Franco et al., 1999; Sharpe and Goldstone, 2000). Effects on neurogenesis were observed at 10^-9 M for TTNPB and at 10^-7 M for AGN193109 (supplementary material Fig. S2A,B). The direct RAR target gene HoxA1 (Balmer and Blomhoff, 2002; Sive and Cheng, 1991) was significantly upregulated by TTNPB and downregulated by AGN193109 at 10^-8 M (supplementary material Fig. S2C). This demonstrates that the N-tubulin phenotypes at 10^-6 M TTNPB or 10^-6 M AGN193109 are not off-target non-specific effects.

Published data do not distinguish whether RARα, RARγ or both are required for primary neurogenesis. Using MO-mediated gene knockdown, we found that microinjection of either an Rarα MO (Fig. 1E) or an Rarγ MO (Fig. 1F) greatly diminished the number of primary neurons (revealed by N-tubulin expression) compared with control MO (Fig. 1D). Phenotypes were verified by reproducing them with different MOs (supplementary material Fig. S3). The sequences targeted by each MO show no similarity to each other (supplementary material Fig. S4). We infer that primary neurogenesis requires both RARα and RARγ; therefore, these receptors are not functionally redundant. Loss of the RA metabolizing enzymes RALDH2 or CYP26A1 caused the expected loss or gain of primary neurons, respectively (supplementary material Fig. S5). The phenotypes are weaker and less penetrant than those elicited by the Rar MOs, suggesting that these enzymes are not exclusively responsible for the steady state of RA within the embryo or, alternatively, that eliminating enzymatic activity may require more complete knockdown than could be achieved.

RAR was predicted to be involved in the pre-patterning stage of neurogenesis, with less pivotal roles during earlier stages of neural induction or the later process of lateral inhibition because treatment with RA inhibited expression of Zic2 (Franco et al., 1999), a pre-patterning gene that inhibits differentiation of neural precursors wherever it is expressed (Brewster et al., 1998). To address where RA signaling acts during neurogenesis, we asked whether modulating RAR signaling altered expression of Geminin, Foxd4l1 (also known as FoxD5) and Sox3, all early markers of neural progenitors whose overexpression inhibits differentiation. TTNPB treatment decreased the sizes of Geminin (Fig. 1H) and Foxd4l1 (Fig. 1N) expression domains compared with controls (Fig. 1G,M,), whereas it had little effect on Sox3 (supplementary material Fig. S6A,B). Treatment with AGN193109 markedly expanded Geminin (Fig. 1I), Foxd4l1 (Fig. 1O) and Sox3 expression domains, particularly in the anterior (supplementary material Fig. S6C). Knockdown of either Rarα or Rarγ expanded the Geminin (Fig. 1K,L), Foxd4l1 (Fig. 1Q,R) and Sox3 expression domains (supplementary material Fig. S6E,F); the control MO had no effect (Fig. 1J,P; supplementary material Fig. S6D). These results indicated that RAR activation decreased the expression of early markers, whereas inhibition of RAR action expanded the expression domains.

As RA signaling is required for primary neurogenesis, we asked which genes might mediate the effects of RA on neuronal differentiation. Our previous microarray analysis revealed that Ets2 Repressor Factor (Erf) was upregulated by TTNPB at neurula stages (Arima et al., 2005). Erf is closely related to two other ETS repressors, Ets Variant 3 (Etv3) and Etv3-like, which are linked in mammalian and Xenopus genomes (Hellsten et al., 2010; Muffato et al., 2010). Erf and Etv3 negatively regulate the cell cycle to inhibit proliferation by interfering with the function of ETS activators (e.g. Ets1/2) at the transcriptional level. Initial experiments showed that MO-mediated knockdown of Etv3/3l or Erf rendered microinjected Xenopus embryos unresponsive to touch (not shown), similar to RAR loss of function (Blumberg et al., 1997). We hypothesized that Erf and Etv3/3l might be important downstream effectors of RA action in primary neurogenesis.

Quantitative real time RT-PCR (QPCR) analysis showed that Erf and Etv3l are RA-responsive at the early neurula stage (Fig. 2A). Erf was more strongly induced by RA during early and late neurula stages, compared with Etv3 and Etv3l (Fig. 2A). Whole-mount in situ hybridization revealed that Erf expression was expanded in TTNPB-treated embryos (Fig. 2C): the normally sharp expression of Erf in the neural folds was broadened and was ectopically present in the anterior. Erf expression was blurred by AGN193109 (Fig. 2D), and knockdown of RARα (Fig. 2F) and RARγ (Fig. 2G). Etv3/Etv3l are ubiquitously expressed in the neurula (see below) and expression was not altered detectably by whole-mount in situ hybridization. These results support a role for Erf and Etv3l as potential downstream effectors of RA signaling in primary neurogenesis.

We used whole-mount in situ hybridization to determine whether Erf and Etv3/3l were expressed appropriately to act downstream of RA signaling in primary neurogenesis (Fig. 3). Prior to gastrulation, Etv3/3l (not shown) and Erf (Fig. 3A,B) are present in the animal, but absent from the vegetal hemisphere. At the open neural plate stage, Etv3/3l expression is broad and diffuse in the neural plate (supplementary material Fig. S1D) but absent from the ventral side of the embryo. Erf is expressed throughout the neural plate by stage 13, and concentrated in the neural folds (Fig. 3C-F; supplementary material Fig. S1C). Erf expression later becomes pronounced in the head, particularly the eye, otocyst, forebrain and pharyngeal arches, but is absent from the cement gland (Fig. 3G-K). mRNA encoding the DEAD-box protein DDx20, which interacts with and promotes the repressive function of ERF and ETV3 (Klappacher et al., 2002), is expressed in the neural plate (supplementary material Fig. S1E). Erf, Etv3 and Etv3l are expressed as maternal transcripts (Fig. 3L). Erf mRNA is more abundant than Etv3l, which is much more abundant than Etv3 throughout all stages of development analyzed (Fig. 3L).

As Erf-MO and Etv3/3l-MO-injected embryos were not responsive to touch, we asked whether loss of function altered primary neurogenesis. Microinjection of Erf-MO or Etv3/3l-MO resulted in the loss of N-tubulin expression (Fig. 4B,C). Two different Erf-MOs produced the same phenotype (supplementary material Fig. S7A,B) as did the Etv3 and Etv3l AUG MOs (supplementary material Fig. S7C,D). We combined the Etv3/3l MOs in some figures (Figs 4, 5; supplementary material Fig. S6) because we believed at the time that both Etv3 isoforms might be RA responsive. Extensive experimentation later revealed that only Etv3l is RA responsive. However, since the phenotype of both knockdowns is the same (supplementary material Fig. S7C,D), we mixed the MOs for all early experiments, and this is designated Etv3/3l MO.

As the loss of N-tubulin in Erf-MO and Etv3/3l-MO phenocopied embryos injected with dominant-negative RARα (Blumberg et al., 1997) or with Rar MOs (Fig. 1E,F), or embryos treated with the RAR antagonists Ro 41-5253 (Franco et al., 1999) or AGN193109 (Fig. 1C), we infer that Erf and Etv3/3l act downstream of RA signaling. In support of this, loss of Erf or Etv3l rescued the extra/ectopic neurons phenotype generated by VP16-RARα/γ mRNA (Blumberg et al., 1997) (supplementary material Fig. S8) or by TTNPB (Fig. 6). We next tested where in the neurogenic pathway ETS repressors act, compared with when RARs act, by analyzing the effects of ERF and ETV3/3L knockdown on other genes in the neurogenic pathway. The neural differentiation genes Myt1 (Fig. 4E,F) and Dl1 (Fig. 4H,I), and the proneural gene Ngnr1 (Fig. 4K,L) were all knocked down, suggesting that Erf and Etv3/3l act early in primary neurogenesis.

Microinjection of the Erf MO reduced neural fold elevation (Fig. 4H,K, red arrows) compared with controls. We used Aquaglyceroporin 3 (Aqp3), which marks the tips of the neural folds (Cornish et al., 2009) and Neogenin, which is required for neural fold elevation (Kee et al., 2008), to demonstrate this neural tube defect in Erf MO embryos at stage 14 (supplementary material Fig. S9). This neural tube defect was resolved by stage 22 (supplementary material Fig. S9E), and embryos appeared morphologically normal at the tadpole stage but were unresponsive to touch. Hence, we conclude that Erf knockdown does not simply delay neurogenesis on the injected side. In support of this, stage 22 embryos have diminished N-tubulin expression on the injected side (supplementary material Fig. S10).

Erf and Etv3 are cell cycle inhibitors in certain cell types (Klappacher et al., 2002; Sawka-Verhelle et al., 2004; Sgouras et al., 1995; Verykokakis et al., 2007). Since RARs act early in the neuronal differentiation pathway, we hypothesized that ERF and ETV3/3L might act downstream of RARs to promote cell cycle exit and increase neuronal differentiation. Knockdown of Erf or Etv3/3l should affect early acting genes that increase the proliferation of neural precursors while inhibiting neuronal differentiation.

Neural induction leads to the upregulation of Zic1, Zic3 and Foxd4l1, which allows ectodermal cells to commit to the neural fate and promotes expression of genes that maintain proliferation in the neural plate at the expense of differentiation (reviewed in Aruga and Mikoshiba, 2011; Moody and Je, 2002; Rogers et al., 2009). Erf-MO and Etv3/3l-MO embryos exhibited lateral expansion of Zic1 (Fig. 5B,C,) and Zic3 (Fig. 5E,F,) but posterior reduction of Zic1 (Fig. 5B,C). The characteristic dorsal striped expression pattern of Zic3 (Fig. 5D) was lost (Fig. 5E,F). Erf and Etv3 knockdown expanded Foxd4l1 (Fig. 5H,I), indicating that Erf and Etv3 function early in the pathway to inhibit genes responsible for early neural stabilization and proliferation.

Foxd4l1 upregulates Geminin and Zic2, genes that are known to maintain the neuroectoderm in an immature proliferative state (reviewed by Moody and Je, 2002; Rogers et al., 2009). Erf and Etv3/3l MOs expanded Geminin and Zic2 expression domains (Fig. 5K,L,N,O). These embryos also exhibited blurring of the dorsal striped pattern of Zic2, similar to the effects on Zic3 (Fig. 5E,F) and in embryos treated with an RAR antagonist (Franco et al., 1999).

SoxB1 family genes (e.g. Sox2, Sox3) are downstream of Foxd4l1, Geminin and Zic2 (reviewed by Moody and Je, 2002; Rogers et al., 2009), maintain neural progenitor cells in a proliferative state, and are downregulated during neural differentiation (Archer et al., 2011; Bylund et al., 2003; Miyagi et al., 2009; Wegner and Stolt, 2005). Injection of Rarα-MO, Rarγ-MO, Erf-MO or Etv3/3l-MO led to expanded Sox3 expression (supplementary material Fig. S6D-I). We interpreted the expanded neural plate (as indicated by Sox3) and the broadened expression of Geminin, Zic2 and Foxd4l1 to indicate that the neural plate in these MO-injected embryos remains in an immature proliferative state.

To test this inference, we performed whole-mount immunohistochemisty on injected embryos, detecting proliferating cells with an antibody against phosphorylated Histone H3 (at Ser10) (Hendzel et al., 1997) or an antibody against PCNA (Mathews et al., 1984). Figure 7 shows representative images of embryos taken in bright-field (Fig. 7A,D,G) and fluorescence (Fig. 7B,E,H). Knockdown of Erf or Etv3l produced significantly more phospho-Histone H3+ staining in the neural plate compared with control-MO (Fig. 7C,F,I). This effect is specific to the neural plate as there was no effect on lateral, non-neural expression of phospho-Histone H3 (supplementary material Fig. S11). Knockdown of Erf or Etv3l produced significantly more PCNA staining in the neural plate compared with control-MO (supplementary material Fig. S12).

Erf loss of function phenocopied RAR loss of function (increased proliferation and inhibited neuronal differentiation); thus, overexpression of Erf mRNA should produce more primary neurons in the neural plate. However, unilateral overexpression of Erf mRNA inhibited expression of N-tubulin (not shown). FGF signaling plays at least two roles in neurogenesis: (1) FGF promotes neural induction and posteriorizes the neuroectoderm; and (2) FGF promotes proliferation of neural progenitors, delaying neuronal differentiation. We hypothesized that overexpression of Erf interfered with the first function, inhibiting Ets genes downstream of FGF signaling required for neural induction (Bertrand et al., 2003). This would lead to a paucity of neurogenic precursors and thus fewer primary neurons.

We used hormone-inducible constructs to allow precise temporal control of ERF activity, in order to distinguish the effects of Erf on neural induction from effects on neuronal differentiation. Fusion of ERF to the ligand-binding domain of the human glucocorticoid receptor (hGR) sequesters ERF in the cytoplasm in the absence of dexamethasone (DEX) (Kolm and Sive, 1995). DEX treatment releases ERF, allowing it to enter the nucleus. ERF is exported from the nucleus upon phosphorylation by ERK (extracellular signal-regulated kinase), therefore we designed mutated constructs that manipulate ERK phosphorylation. ERF is phosphorylated at Thr⁵²⁶, Ser¹⁶¹, Ser²⁴⁶ and Ser²⁵¹ in rodent fibroblasts; alanine substitutions at any of these positions decreased nuclear export and increased ERF-mediated repression (Le Gallic et al., 1999). Only Ser²⁴⁶ and Ser²⁵¹ and the surrounding residues required for phosphorylation by ERK are conserved in Xenopus, chick and zebrafish. We designed a ‘constitutively nuclear’ construct to mimic permanently the unphosphorylated state (Ser^246,251 → Ala^246,251), such that ERF is maintained in the nucleus and increases repression. A phosphomimetic (Ser^246,251 → Glu^246,251) mutant that is constitutively exported from the nucleus (inhibiting repression) was also produced (Wagner et al., 2004). These Erf constructs demonstrated the appropriate subcellular localization in the presence and absence of DEX (supplementary material Fig. S13).

mRNAs encoding hGR-ERF fusion proteins were unilaterally microinjected into two- or four-cell embryos, treated with dexamethasone after the beginning of neural induction (stage 11) and fixed at stage 14. Whole-mount in situ hybridization with N-tubulin revealed that overexpression of hGR-Erf (Ser^246,251 → Ala^246,251) after neural induction led to more primary neurons (supplementary material Fig. S14A; Fig. S15B,D). The hGR-Erf (Ser^246,251 → Glu^246,251) mutant did not affect the number of neurons, presumably because it is transported out of the nucleus and is unable to act as a transcriptional repressor (supplementary material Fig. S14B; Fig. S15F,H). DMSO treatment did not alter the number of neurons (supplementary material Fig. S14; Fig. S15A,C,E,G). To demonstrate that Erf acts downstream of RAR, we attempted to rescue the loss of N-tubulin in RARγ-MO embryos with hGR-Erf (Ser^246,251 → Ala^246,251) mRNA. We observed a partial rescue of N-tubulin in dexamethasone-treated embryos (supplementary material Fig. S16).

We examined N-tubulin expression at a slightly earlier stage (stage 13/13.5) to determine whether overexpression of hGR-Erf (Ser^246,251 → Ala^246,251) caused precocious neurogenesis. When embryos injected with hGR-Erf (S→A) were treated with dexamethasone at stage 10.5 or stage 11, we noticed precocious neurogenesis on the injected side in 27-28% of the embryos (Fig. 8E,F,H,I). Overexpression of hGR-Erf (S→A) decreased proliferation on the injected side of dexamethasone-treated embryos (supplementary material Fig. S17D-F). When embryos were treated at stage 9 with dexamethasone, knock down of N-tubulin was observed, as expected, presumably because Erf interfered with the early action of FGF on neural induction (Fig. 8B,C). When embryos were treated with dexamethasone after stage 11, no precocious neurogenesis was observed. We infer that Erf was released to function in the nucleus too late to affect neuronal differentiation (Fig. 8K,L). DMSO vehicle treatment did not alter the number of neurons (Fig. 8A,C,D,F,G,I,J,L) or proliferation (supplementary material Fig. S17A-C).

Developing systems maintain a dynamic balance between cell proliferation and differentiation, yet the molecular mechanisms regulating this equilibrium are poorly understood. Mutually inhibitory interactions between factors promoting proliferation (e.g. FGFs) versus differentiation (e.g. RA), and cell cycle genes involved have been described in a few systems, including neuronal progenitor cells (Chen et al., 2012; Diez del Corral et al., 2003; Seo et al., 2005a). Little is known about the molecular nature of the switch that controls the shift between proliferation of neural precursors and their entry into the neuronal differentiation pathway (Kaldis and Richardson, 2012).

RAR signaling plays a significant role in patterning the neural plate and promoting primary neurogenesis. Knockdown of RARα or RARγ independently led to loss of primary neurons, supporting our previous observation that Xenopus RAR subtypes exhibit temporal and spatial regulation of different target genes (Koide et al., 2001; Shiotsugu et al., 2004). As RA cannot neuralize naïve ectoderm (Blumberg et al., 1997; Sharpe and Goldstone, 1997), we predicted that the decreased number of primary neurons due to RAR loss of function resulted from a failure of neuronal differentiation rather than a loss of neural competence. We found that loss of the differentiation signal from RAR caused expansion of markers of neural progenitors and neural stabilization (Geminin, Foxd4l1 and Sox3). Increasing RA signaling reduced Geminin and Foxd4l1 expression while increasing neuronal differentiation. Hence, we inferred that Rarα and Rarγ play early roles in neural differentiation by inhibiting proliferation of neural progenitors.

Our results showed that ETS repressors are effectors of RA signaling that promote primary neurogenesis. Treatment of embryos with TTNPB or AGN193109 significantly altered the Erf expression domain, and Rarα or Rarγ were required for Erf expression. Knockdown of ERF or ETV3/3L caused embryos to be unresponsive to touch and primary neurons were lost, phenocopying RAR loss of function. By contrast, temporally controlled Erf gain of function increased the number of neurons at stage 14, and led to precocious neurogenesis at slightly earlier stages. We showed that Ets repressors are downstream of RAR in neuronal differentiation. Loss of either ERF or ETV3L blocked the production of excess neurons generated by a constitutively active VP16-RAR or by TTNPB treatment. We predict that the connection between RAR signaling and Ets-repressors will prove to be important in other biological processes in the future.

As RAR loss of function caused expansion of Geminin, Foxd4l1 and Sox3, we hypothesized that Erf and Etv3/3l also act early in the neurogenic pathway. After neural tissue is induced via BMP inhibition and active FGF signaling, Zic1, Zic3 and Foxd4l1 are upregulated (Marchal et al., 2009; Tropepe et al., 2006; Wills et al., 2010). Foxd4l1 and Zic3 are downstream targets of FGF signaling (Branney et al., 2009; Lee et al., 2009; Marchal et al., 2009; Yan et al., 2010), whereas Zic1 is an immediate early gene of BMP inhibition and is driven by a BMP inhibitor-responsive promoter module (Tropepe et al., 2006). Zic1 and Zic3 stabilize the neural fate immediately after neural induction, promoting plasticity of neural progenitors and inhibiting differentiation (Aruga, 2004; Aruga and Mikoshiba, 2011; Aruga et al., 2002; Merzdorf, 2007). Zic1 is highly expressed in human ESC-derived neural rosettes and in proliferating neural stem cell progenitors (Elkabetz et al., 2008; Tabar et al., 2005). Zic3 is a direct target of pluripotency factors in stem cells, and is diminished after differentiation with RA (Lim et al., 2007). Knockdown of ERF or ETV3/3L caused marked expansion of Foxd4l1, Zic1 and Zic3; hence, Erf and Etv3/3l regulate important factors that mediate the early transcriptional response downstream of BMP inhibition and FGF signaling in neural induction.

Knockdown of ERF or ETV3/3L expanded the Geminin and Zic2 expression domain comparable with that observed in Rarα-MO and Rarγ-MO embryos. Expansion of Zic2 leads to the inhibition of Xngnr-1, which normally induces expression of Myt-1 and Delta-1 to promote neurogenesis (Bellefroid et al., 1996; Brewster et al., 1998; Chitnis et al., 1995). Geminin plays a primary role in maintaining neuroectoderm in an ‘immature state’, conferring a reduced response to differentiation signals (Neilson et al., 2012; Yan et al., 2009; Yan et al., 2010). Geminin is highly expressed in proliferating neural progenitors (Spella et al., 2007), where it interacts with Brahma-related gene 1 to inhibit neuronal differentiation and increase proliferation (Seo et al., 2005a; Seo et al., 2005b). Geminin regulates the expression of Sox3 in a positive-feedback loop and is thought to postpone lineage commitment by stabilizing repressive chromatin marks to promote cellular plasticity (Lim et al., 2011). We infer that Erf and Etv3/3l restrict Zic2 and Geminin expression in the neural plate, inhibiting neural progenitor plasticity and proliferation, and defining cell fate and promoting lineage commitment.

RA is known to regulate the expression of genes that facilitate cell cycle exit and differentiation (Andrews, 1984; Rhinn and Dollé, 2012). FGF signaling promotes neural progenitor survival and proliferation in early neural tissue and in the mammalian brain (Chen et al., 2012; Marchal et al., 2009; Mason, 2007), and uses ETS transcription factors as terminal effectors (Bertrand et al., 2003; Sharrocks, 2001; Wasylyk et al., 1998). ERF and ETV3 recognize Ets1/2 consensus sites, which are downstream of FGF signaling in vivo, that are often found in the regulatory regions of positive cell cycle genes such as c-Myc, c-Myb, p54 and Cdc-2 (Carlson et al., 2011; Hester et al., 2007; Klappacher et al., 2002).

We propose that Erf and Etv3/3l play a similar role in the neuroectoderm, inhibiting proliferation of neural progenitors by restricting expression of genes (e.g. Foxd4l1, Geminin, Sox3, Zic1 and Zic3) that stabilize the neural fate and prevent differentiation by maintaining plasticity and/or proliferation of neural progenitors. ERF or ETV3L knockdown elicited a significant increase in cell proliferation in the neural plate, whereas overexpression of ERF produced the opposite result. We conclude that ERF and ETV3/3L play key roles in terminating the cell cycle to facilitate neuronal differentiation.

FGF signaling and BMP inhibition initiate neural induction, leading to the upregulation of genes (such as Foxd4l1, Zic1, Zic2, Zic3, Geminin and Sox3) that promote and maintain neural competence. This early gene network generates a neural progenitor identity within the neuroectoderm characterized by a stem-like fate of plasticity and proliferation. When the neuroectoderm is fully stabilized, it is equally important for this gene network to be downregulated for differentiation to occur. Retinoic acid receptors and their effectors, Erf and Etv3l, are expressed at the correct time and place to interfere with the action of neural progenitor genes and facilitate neuronal differentiation. Loss of ERF or ETV3L prolongs neural progenitor identity, increasing proliferation and preventing the development of mature neurons. ERF overexpression causes the opposite effect, increasing the number of primary neurons.

We infer that ETS repressors, Erf and Etv3l, sit at the intersection of proliferation and differentiation. Fig. 9 summarizes our model for how RA and Erf/Etv3l regulate the proliferation/differentiation switch in primary neurogenesis. RAR action promotes the expression of Erf and Etv3l to inhibit the cell cycle downstream of FGF in the neuroectoderm. Whether Foxd4l1, Zic1, Zic2, Zic3, Geminin and Sox3 are regulated directly by ETS repressors at the transcriptional level remains an unanswered question. Although it is well known that FGF signaling employs ETS proteins in signal transduction, BMP signaling also uses ETS factors that act in synergy with Smad proteins (Koinuma et al., 2009; Morikawa et al., 2011). The BMP inhibitory response module that drives Zic1 expression contains multiple, functional ETS-binding sites (Tropepe et al., 2006). Erf and Etv3/3l could play a direct role, binding ETS sites in the regulatory regions of neural progenitor genes (such as Foxd4l1, Zic2 and Geminin). Alternatively, Erf and Etv3/3l could simply promote cell cycle exit, terminating expression of genes associated with neural progenitor identity and facilitating differentiation. The results presented above demonstrate that RAR negatively influences FGF signaling by upregulating Erf and Etv3l to repress genes that stimulate neural progenitor fate, establishing an important new role for opposing RA and FGF signals in primary neurogenesis.

We thank Dr Tom Schilling (UCI, Irvine, CA, USA) for the use of his fluorescence microscope, and Dr Pierre Le Pabic and Kelly Radtke in his lab for help with photography and Velocity software. We thank Dr Ken Cho (UCI) for the use of his GFP microscope and Dr Ira Blitz in his lab for help with photography. We thank former UCI undergraduate student Jennifer Dean for technical help during the early stages of this study.