The directed differentiation of embryonic stem cells (ESCs) into neural stem cells (NSCs) of specific identities and the identification of endogenous pathways that may mediate expansion of NSCs are fundamental goals for the treatment of degenerative disorders and trauma of the nervous system. We report that timely induction of a Hoxb1 transgene in ESC-derived NSCs resulted in the specification of NSCs toward a hindbrain-specific identity through the activation of a rhombomere 4-specific genetic program and the repression of anterior neural identity. This change was accompanied by changes in signaling pathways that pattern the dorsoventral (DV) axis of the nervous system and concomitant changes in the expression of DV neural progenitor markers. Furthermore, Hoxb1 mediated the maintenance and expansion of posterior neural progenitor cells. Hoxb1+ cells kept proliferating upon mitogen withdrawal and became transiently amplifying progenitors instead of terminally differentiating. This was partially attributed to Hoxb1-dependent activation of the Notch signaling pathway and Notch-dependent STAT3 phosphorylation at Ser 727, thus linking Hox gene function with maintenance of active Notch signaling and the JAK/STAT pathway. Thus, timely expression of specific Hox genes could be used to establish NSCs and neural progenitors of distinct posterior identities. ESC-derived NSCs have a mixed DV identity that is subject to regulation by Hox genes. Finally, these findings set the stage for the elucidation of molecular pathways involved in the expansion of posterior NSCs and neural progenitors.
Disclosure of potential conflicts of interest is found at the end of this article.
Author contributions: M.G.: conception and design, financial support, collection and assembly of data, data analysis and interpretation, A.G.: conception and design, financial support, data analysis and interpretation, manuscript writing, final approval of manuscript.
The application of developmental principles in directed differentiation of embryonic stem cells (ESCs) provides a promising venue for dissecting molecular mechanisms underlying specification of neural progenitor populations that are not otherwise accessible. This line of research could provide insights for generating specific neural subtypes for drug screening and cell therapy and identifying signals to stimulate the expansion of endogenous neural stem cells (NSCs).
NSCs have been derived from ESCs through embryoid body formation and subsequent use of extracellular signals and chemically defined media [1, , –4] or direct induction of neural character using chemically defined or conditioned media [5, , –8]. The evidence obtained so far suggests that these cells have a broad anterior identity [3, 9]. Retinoic acid (RA) has been used as a posteriorizing factor of ESC-derived NSCs, but its effects are pleiotropic, resulting in heterogeneous cultures and precocious terminal differentiation [7, 10]. Broad posterior character has been induced in ESC-derived neural cells in an RA-independent pathway through the use of as yet unidentified neural inducing signals emanating from the PA6 stromal cell line [4, 6, 11]. Furthermore, it remains unclear whether ESC-derived NSCs have a propensity to generate dorsal, ventral, or both types of neural progenitor cells. Given that specific neural subtypes are derived from specific domains in the developing central nervous system (CNS) with precise anteroposterior (AP) and dorsoventral (DV) identities [12, 13], it is important to address the issue of regional specification in ESC-derived NSCs. On the other hand, the potential for mobilization of endogenous NSCs exists, but the adult nervous system has very limited capacity for significant self-regeneration [14, –16]. Identifying the pathways involved in the expansion of posterior NSCs will help to design strategies aiming at mobilizing the endogenous regeneration capacity in cases of degenerative disorders or trauma.
Hox genes play a central role in the specification of AP identity in the posterior CNS and play multiple roles during nervous system specification, cell migration, axon guidance, and finally formation of somatotopic maps [17, , –20]. Despite the extensive analysis of Hox gene function, mainly through genetic analyses, very little is known in terms of molecular targets of their action, particularly in nervous system patterning.
To address the issues outlined above, we used inducible expression of Hoxb1 in ESC-derived NSCs. We chose Hoxb1 because its role in the specification of a single segment of the developing nervous system is well characterized [18, 21], and a few direct target genes that could be used to validate the approach have already been identified.
We report that timely induction of a Hoxb1 transgene in ESC-derived NSCs resulted in the specification of NSCs toward a hindbrain-specific identity through the activation of a rhombomere 4-specific genetic program and the repression of anterior neural identity. We found that ESC-derived NSCs are of mixed DV identity and that Hoxb1 expression results in the redistribution of DV progenitor identities and changes in the expression levels of genes involved in DV signaling pathways. These conclusions were founded on genome-wide expression data obtained by microarray gene expression profiling. Importantly, using this model system, we showed that Hoxb1 mediates the maintenance and expansion of posterior neural progenitor cells. Hoxb1+ cells kept proliferating upon mitogen withdrawal and became transiently amplifying progenitors instead of terminally differentiating. This was partially attributed to Hoxb1-dependent activation of the Notch signaling pathway and Notch-dependent STAT3 phosphorylation at Ser 727, thus linking Hox gene function with maintenance of active Notch signaling and the JAK/STAT pathway.
These findings suggest that timely expression of specific Hox genes could be used to establish NSCs and neural progenitors of distinct posterior identities. Furthermore, they set the stage for the elucidation of molecular pathways involved in the expansion of posterior NSCs and neural progenitors.
Materials and Methods
Generation of Constructs
The generation of targeting vector R26/nls reversed tetracycline transactivator (rtTA) has been described elsewhere . To generate the hypoxanthine phospho ribosyl transferase (Hprt)-pBI-Hoxb1 vector, the Hoxb1 cDNA was cloned from an expression vector (A. Gavalas, unpublished data) as a SalI/NotI fragment and inserted in the same sites of plasmid pBl-2 to generate plasmid pBI-Hoxb1. Plasmid pBI-2 was derived from plasmid pBI (Clontech, Palo Alto, CA, http://www.clontech.com) by the insertion of a PmeI linker into the SapI site and the insertion of an AscI site into the AatII site. The Hprt targeting vector pSKB1  was modified first by BamHI/PmeI digestion and religation to remove the BamHI, AscI, and PmeI sites at the 5′ end of the 5′ Hprt homology region and subsequently by the insertion of a PmeI/AscI linker in the MluI/NotI sites of the cloning polylinker to yield plasmid Hprt2. The AscI/PmeI fragment of plasmid pBI-Hoxb1 was inserted into the same sites of plasmid Hprt2 to generate the Hprt-pBI-Hoxb1 targeting vector.
ESC Culture and Generation of Knock-In Lines
ESC lines were routinely propagated on feeder layers in standard ESC medium supplemented with leukemia inhibitory factor (LIF) (1,000 units/ml). The ESTet−On cell line was generated by gene targeting in the HM-1 cell line as described elsewhere . The ESTet−On cells were electroporated with 30 μg of PvuI linearized Hprt-pBI-Hoxb1 targeting vector. After 8 days of selection in hypoxanthine aminopterin thymidine (HAT; Sigma-Aldrich, St. Louis, http://www.sigmaaldrich.com), medium colonies were analyzed by polymerase chain reaction (PCR)  and genomic Southern blots . The resulting ESTet−On/Hoxb1 clones were analyzed by Western blot analysis and immunofluorescence on ESCs grown on gelatin-coated plates using the Hoxb1 antibody in the presence and in the absence of 1 μg/ml doxycycline (dox; Clontech).
ESC Differentiation Conditions
The differentiation protocol for the generation of nestin+ cells from ESCs was carried out as described elsewhere , with some modifications. Prior to initiation of differentiation, low-passage ESCs were cultured for two passages on feeder layers in standard ES medium containing LIF (1,000 units/ml). To start differentiation, ES cells were dissociated using 0.05% trypsin solution (0.05% trypsin, 1.3 mM EDTA, 0.1% chicken serum) and then were plated onto tissue culture plates for two short successive periods (20–30 minutes) to remove feeder layers. To induce EB formation, the cells were plated onto nonadherent bacterial culture dishes at a density of 2 × 104 cells cm−2 in standard ESC medium. The EBs formed for 4 days, and the medium was changed every 2 days. The EBs was plated onto tissue culture plates precoated with 0.1% gelatin in ES medium for 24 hours. Selection of nestin+ cells was initiated by replacing the ES medium with serum-free medium containing insulin/selenium/transferrin/fibronectin . Nestin+ cells were generated after culture for 8 days under these conditions, and then expansion of nestin+ cells was initiated. Cells were plated as a single-cell suspension on plates coated with Matrigel (BD Biosciences, San Diego, http://www.bdbiosciences.com) at a density of 2 × 105 cells cm−2 in N2 medium (Invitrogen, Carlsbad, CA, http://www.invitrogen.com) supplemented with 10 ng/ml basic fibroblast growth factor (bFGF) for 4 days. To induce differentiation, the cells were plated in four-well dishes precoated with Matrigel in N2 medium with bFGF for 1 day and then were left to differentiate for 4 days in the absence of bFGF (Fig. 1).
Immunoblots, Immunofluorescence, and Antibodies
Immunoblot analysis was performed using standard procedures at the stage indicated in Figure 1. Primary antibodies were as follows: rabbit anti-STAT3, 1:1,000 (Santa Cruz); anti-phospho-Ser 727 STAT3, 1:200 (Cell Signaling Technology, Beverly, MA, http://www.cellsignal.com); and rabbit anti-phospho-Tyr 705 STAT3, 1:1,000 (Cell Signaling Technology). For immunofluorescence, cells were collected at various stages of the differentiation protocol (Fig. 1) and fixed for 10 minutes at 4°C in 4% paraformaldehyde in phosphate-buffered saline (PBS). They were then washed in PBST (PBS with 0.1% Triton X-100) and subsequently permeabilized in 80% methanol for 15 minutes at −20°C. Blocking was for 1 hour in PBST containing 10% normal goat serum (NGS). Primary and secondary antibodies were diluted in PBST containing 1% NGS. Cells were incubated with primary antibodies overnight at 4°C, with secondary antibodies at room temperature for 1 hour, mounted with 4,6-diamidino-2-phenylindole (DAPI) containing Prolong Antifade (Molecular Probes, Eugene, OR, http://probes.invitrogen.com), and fluorescent images were taken using an inverted Leica SP5 confocal microscope. Antibodies were as follows: rabbit anti-Hoxb1, 1:400 (Covance, Princeton, NJ, http://www.covance.com); rabbit anti-pH3, 1:500 (Cell Signaling Technology); mouse anti-nestin, 1:100 (Developmental Studies Hybridoma Bank [DSHB], Iowa City, IA, http://www.uiowa.edu/∼dshbwww); mouse anti-Lhx1/5, 1:100 (DSHB); rabbit anti-GFAP, 1:200 (Sigma-Aldrich); mouse anti-Tuj1, 1:1,000 (Covance); rabbit anti-MAP2, 1:500 (Chemicon, Temecula, CA, http://www.chemicon.com); mouse anti-NeuN, 1:500 (Chemicon); mouse anti-RC2, 1:100 (DSHB); mouse anti-vimentin, 1:200 (DSHB); rabbit anti-Foxg1, 1:200 (a gift from Y. Sasai); rabbit anti-Sox2, 1:250 (Abcam, Cambridge, MA, http://www.abcam.com); mouse anti-Otx-5F5, 1:50 (DSHB); mouse anti-5-bromo-2′-deoxyuridine (BrdU), 1:200 (Dako, Glostrup, Denmark, http://www.dako.com); rabbit anti-Pax2, 1:1,000 (Zymed); mouse anti-pax7, 1:50 (DSHB); mouse anti-Pax7, 1:50 (DSHB); mouse anti-Pax3, 1:250 (DSHB); mouse anti-Nkx6-1, 1:1,000 (DSHB); mouse anti-Msx1/2, 1:50 (DSHB). Secondary antibodies were anti-rabbit and anti-mouse Alexa 488- and Alexa 568-conjugated goat antibodies, 1:500 (Molecular Probes). Quantitation and statistical analyses were carried out as described below.
BrdU Incorporation Assay
BrdU (BD Pharmingen, San Diego, http://www.bdbiosciences.com/index_us.shtml) was added to the cells at a final concentration of 10 μM during the expansion and differentiation stages (Fig. 1) for short (2-hour) and long (24-hour) periods of time. Cells were then fixed in 4% PFA for 15 minutes at 4°C. After three washes in PBS, cells were treated for 10 minutes in 2 N HCl in PBST at room temperature and for 20 minutes in 0.1 M sodium borate, pH 8.5. BrdU-positive cells were detected using a specific mouse anti-BrdU antibody for immunocytochemistry. Quantitation and statistical analyses were carried out as described below.
Inhibition of Notch Activation
The γ-secretase inhibitor N-[N-(3,5-difluorophenacetyl-L-alanyl)]-S-phenylglycine t-butyl ester (DAPT) (Calbiochem, San Diego, http://www.emdbiosciences.com) was added to the cells at a final concentration of 10 μM during the terminal differentiation stage. Fresh inhibitor was added to the culture every 24 hours. The inhibitor was solubilized in dimethyl sulfoxide (Sigma-Aldrich) and diluted in the medium. The Hoxb1+ cells were cultured for 3 days with the inhibitor in the absence of mitogen, at which point they were processed for immunocytochemistry (Fig. 1). Quantitation and statistical analyses were carried out as described below.
Quantitation and Statistical Analysis
Immunocytochemistry and cell count data were derived from a total of three independent experiments. Cells were counted in a uniform random fashion by scoring the number of cells per field of view. The number of cells immunoreactive for a specific antigen was counted in at least three nonoverlapping fields in each experiment. The total number of cells in each field was determined by cell nuclei DAPI counterstaining and was approximately 600–1,000 cells/field. Final values were expressed as percentage of immunoreactive cells and presented as means ± SD. Statistical significance was assessed using the two-sample/unequal variance t test when comparing Hoxb1− with Hoxb1+ cells and one-way analysis of variance when comparing Hoxb1−, Hoxb1+, and Hoxb1+/DAPT-treated cells. The calculated p values are referred to in Figs. 3,4 and supplemental online Table 2.
Reverse Transcription-PCR Analysis
Total RNA was isolated from ESCs and ESC-derived NSCs (Fig. 1) using the RNeasy kit (Qiagen, Hilden, Germany, http://www1.qiagen.com) according to the manufacturer's instructions and digested by RQ1 DNase (Promega, Madison, WI, http://www.promega.com) to remove genomic DNA. First-strand cDNA synthesis was performed with SuperScript II reverse transcriptase (Invitrogen) using random primers. Amplification of cDNAs was performed by PCR using 25–35 cycles with denaturation at 94°C for 30 seconds, annealing at 54°C–62°C for 60 seconds, and extension at 72°C for 60 seconds. With the exception of single-exon genes, all primers were designed to span introns using the MacVector software (MacVector, Inc, Cary, NC, http://www.macvector.com), and conditions were optimized using 10.5-day post coitum mouse embryo total RNA. The primers and temperature used for specific PCRs are shown in supplemental online Table 3. PCR products were analyzed by agarose gel electrophoresis.
Microarray Gene Expression Profiling
Expression of Hoxb1 was induced by addition of dox (1 μg/ml) during the selection period and for 1 additional day after trypsinization (Fig. 1). Gene expression profiling was carried out for biological triplicates (cells taken independently through the differentiation procedure) for both dox-induced (Hoxb1+) and uninduced (Hoxb1−) cells. Total RNA prepared using the RNeasy kit (Qiagen) according to the manufacturer's instructions; it was digested by RQ1 DNase (Promega) to remove genomic DNA and used to generate biotinylated cRNA to probe the Affymetrix Mouse Genome 430 2.0 array (Affymetrix, Santa Clara, CA, http://www.affymetrix.com). The quantity and quality of the total RNA were assessed by UV spectroscopy using the NanoDrop ND-1000 (NanoDrop, Wilmington, DE, http://www.nanodrop.com) and high-performance liquid chromatography (HPLC) using the Agilent 2100 Bioanalyzer (Agilent Technologies, Palo Alto, CA, http://www.agilent.com). Twenty micrograms of RNA were converted to double-stranded cDNA using the Roche Microarray cDNA Synthesis Kit (Roche Diagnostics, Basel, Switzerland, http://www.roche-applied-science.com) and purified using the Roche Microarray Target Purification Kit as recommended. The cDNA was converted to biotinylated cRNA using the GeneChip IVT Labeling Kit as recommended (Affymetrix). The biotin-cRNA was purified using the RNeasy Mini Kit RNA cleanup protocol and analyzed by UV spectroscopy (NanoDrop ND-1000) and HPLC (Agilent 2100 Bioanalyzer). The purified biotin-cRNA was fragmented and hybridized to the GeneChip Mouse Genome 430 2.0 Array as recommended (Affymetrix), and the GeneChip was washed and scanned using the GeneChip Fluidics Station 400 and the GeneArray Scanner (Agilent Technologies) at the Medical Research Council Clinical Science Centre Genomics Core Facility. The robust multiarray average (RMA)  method was used for signal normalization at the probe level and across triplicate chips for each condition. The untreated (−dox) samples were defined as the reference against which the treated (+dox) samples were compared to generate differential expression values. Statistical analysis was performed using statistical analysis for microarrays  and multiple testing correction using Benjamini and Hochberg false discovery rate (FDR) . A list of statistically significant differentially expressed genes (FDR = 0.5) was generated and fold change values were calculated on the basis of the RMA analysis.
Inducible Expression of a Hoxb1 Transgene Activated the Endogenous Hoxb1 Autoregulatory Loop in a Context-Dependent Manner
Function of Hox homeobox transcription factors is largely context-dependent. Therefore, their timely induction in the proper cellular context would be crucial to assess their capacity to drive ESC differentiation and specification to specific fates and to address the molecular underpinnings of their function. To express Hoxb1 in a timely and tightly controlled manner, we generated ESC lines allowing the inducible expression of a Hoxb1 transgene. We used the rtTA system because of its good inducibility characteristics , and we incorporated its components in the genome using knock-ins in constitutive loci of the HM-1 mouse ESC line, which carries a deletion in the HPRT locus . This approach ensured accessibility of the transgenes upon differentiation and maximized phenotypic reproducibility due to isogenicity. Using a knock-in strategy, we incorporated the Tet-On cDNA  in the ROSA26 locus , generating the line ESTet−On . Using a similar approach and HAT selection, a Tet-responsive transgene was incorporated in the HPRT locus of the ESTet−On cells to generate the line ESTet−On/Hoxb1 (supplemental online Fig. 1A, 1B).
Tight dox-mediated inducible expression of Hoxb1 was verified by Western blots in the ESC stage (supplemental online Fig. 1D) and by immunostaining of ESCs and ESC-derived NSCs (supplemental online Fig. 1C). In both ESCs and ESC-derived NSCs, regulation of Hoxb1 expression was tight and expression similarly strong after dox addition, resulting in up to 60% of the cells expressing Hoxb1 (supplemental online Fig. 1C). Using primers specific for the Hoxb1 transgene and the endogenous Hoxb1 gene, we examined their expression by reverse transcription (RT)-PCR in ESCs and ESC-derived NSCs. The Hoxb1 transgene was activated at both stages (ESC stage or during selection for ESC-derived NSCs) upon dox addition, but the generated Hoxb1 protein activated the endogenous Hoxb1 autoregulatory loop only in the ESC-derived NSCs (supplemental online Fig. 1E), thus demonstrating the importance of cellular context for Hox function. Before addressing the effects of inducible Hoxb1 expression in the specification of ESC-derived NSCs, we investigated whether Hoxb1 expression affected the stem cell character and the differentiation potential of the cells. Hoxb1− and Hoxb1+ ESC-derived NSCs retained a similar NSC character, as shown by expression of nestin, RC2, Sox2, vimentin, BLBP, and GLAST (supplemental online Fig. 2A–2H, 2I). Furthermore, Hoxb1− and Hoxb1+ ESC-derived NSCs retained their capacity to differentiate into TujI+ neuronal cells, GFAP+ glial cells, and O4+ and MBP+ immature oligodendrocytes (supplemental online Fig. 2J–2O).
Thus, we generated ESCs affording tight regulation of a Hoxb1 transgene at both the ESC and NSC stages. Expression of the Hoxb1 transgene activated the endogenous autoregulatory loop only at the NSC stage, demonstrating the importance of cellular context in Hox function. Hoxb1 expression in the NSC stage did not alter the stem cell character or differentiation potential of the cells.
Hoxb1 Expression Specified ESC-Derived NSCs by Activating a Hindbrain-Like Genetic Program and Repressing Anterior Genetic Programs
We then assessed to what extent Hoxb1-inducible expression changed the AP specification of ESC-derived NSCs. Induction of the Tet-responsive transgene at the selection stage led to robust expression of the endogenous Hoxb1 gene (Fig. 2B), suggesting that the autoregulatory loop responsible for maintenance of Hoxb1 expression in r4 of the hindbrain  had been activated. Furthermore, the r4 known Hoxb1 target genes Hoxb2 , Hoxa2 , EphA2 , and Phox2b  were also activated in response to Hoxb1 activation, as shown by both microarray gene expression profiling (Fig. 2A) and semiquantitative RT-PCR (Fig. 2B). The Hoxb1 activation indicated from the microarrays (Fig. 2A) corresponded exclusively to induction of the endogenous gene, because the Hoxb1 probe set on the microarrays corresponded to the 3′ untranslated region of the endogenous gene, which was replaced in the transgene with the SV40 pA. Importantly, Krox20 and MafB/kr, genes that are necessary for the specification of rhombomeres r3/r5 and r5/r6, respectively [37, 38], were not expressed in either Hoxb1− or Hoxb1+ ESC-derived NSCs. On the other hand, microarray data showed upregulation at low levels of expression of Gbx2 and Hoxb3. Downregulation of the r1 markers Pax2, En1, and Sprouty1 (Figs. 2A, 3G–GI; supplemental online Table 1)  and upregulation of Olig3, which is expressed exclusively in the neuroepithelium posterior to r1 at early stages , suggested that the modest Gbx2 upregulation reflected its early expression in the posterior hindbrain territory at the 5–10-somite stage . Hoxb3 contributes to the specification of r5 and r6 territories, and its regulatory region contains functional Hox binding sites . Similarly to Gbx2, its modest upregulation may represent an early function of Hoxb1 in the presumptive posterior hindbrain territory . Absence of expression of members of the Hox paralogous group 4 (supplemental online Table 1; data not shown) suggested that hindbrain fates posterior to r6 were not induced. Lhx1 and Lhx5 are LIM homeodomain transcription factors that are expressed in the midbrain (Lhx1 and Lhx5), telencephalon (Lhx5), and spinal cord (Lhx1 and Lhx5), whereas expression of both is downregulated in the hindbrain . Consistent with the notion that Hoxb1+ cells have acquired an r4-like hindbrain identity, microarray data showed strong downregulation of both Lhx1 and Lhx5 (Fig. 2A). This was confirmed by immunofluorescence using an antibody recognizing both transcription factors (Fig. 3J–3L).
Among the 36 Hox genes represented on the arrays, 25 are not expressed at all in either Hoxb1− or Hoxb1+ ESC-derived NSCs (data not shown). Upon Hoxb1 transgene induction, and consistent with the establishment of a narrow hindbrain identity (described above), four Hox genes were upregulated (Hoxb1, Hoxb2, Hoxa2, and Hoxb3; Fig. 2; supplemental online Table 2). The low expression of seven additional Hox genes was essentially the same in both populations (supplemental online Table 1), suggesting that alternative, more posterior programs have not been activated upon Hoxb1 activation. Consistent with this, the microarray results showed that Cdx1, Cdx2, and Cdx4 expression was at background levels in both cell populations. To further confirm that posterior programs were not activated, we examined the expression of Hoxb4 and Hoxb9 by RT-PCR. The results suggested that expression of these genes was nearly undetectable in both cell populations (data not shown).
We then examined whether activation of this genetic program was accompanied by a repression of anterior identities. Microarray gene expression analysis showed that Hoxb1− ESC-derived NSCs strongly expressed a large number of anterior identity genetic markers, suggesting that these cells have a largely anterior identity (supplemental online Table 1). These genes were repressed upon induction of the r4-like genetic program, indicating a general repression of anterior character upon Hoxb1 activation (Fig. 2; supplemental online Table 1).
Homeodomain transcription factors such as Otx1, Otx2, Emx2, and Arx play a central role in the specification of anterior regions of the CNS [44, –46] and are strongly expressed in the Hoxb1− ESC-derived NSCs (supplemental online Table 1). Their downregulation in the Hoxb1+ ESC-derived NSCs was confirmed by semiquantitative RT-PCR for the expression of Emx2, Otx2, and Arx (Fig. 2B) and immunofluorescence for Otx1 (Fig. 3A–3C). Foxg1 is a winged helix transcription factor required for regionalization and growth of the telencephalic and optic vesicles . Its strong expression in Hoxb1− cells was abolished upon Hoxb1 activation (Figs. 2A, 3D–3F). Wnt7b is a member of the Wnt family expressed in the dorsal telencephalon and anterior diencephalon . Its expression in Hoxb1+ cells was similarly downregulated upon Hoxb1 induction (Fig. 2A, 2B). It is worth noting that expression of Otx1, Lhx1/5, and Hoxb1 is mutually exclusive (Figs. 3A, 3B, 3J, 3K), suggesting that repression of anterior identity by Hoxb1 occurs in a cell-autonomous manner.
Taken together, these data suggested that ESC-derived NSCs acquire by default an anterior identity. Induction of the Hoxb1 transgene activated a hindbrain genetic program with an r4-like identity and repressed anterior identities.
Hoxb1 Expression Specified the DV Character of ESC-Derived NSCs by Regulating the Expression of DV Progenitor Markers and Selected Components of DV Signaling Pathways
The issue of DV identity of ESC-derived NSCs has not been addressed. We used microarray gene expression profiling to assess the DV identity of ESC-derived NSCs and ask whether changes in the AP identity of the cells lead to reassignment of DV identities as well. We found that ESC-derived NSCs were of mixed DV identity, as several DV progenitor markers were expressed in both Hoxb1− and Hoxb1+ cells (supplemental online Table 1). Consistent with a change in the AP address, the relative expression of most, albeit not all, of these markers changed upon Hoxb1 induction (supplemental online Table 1; Figs. 3, 4).
The homeodomain transcription factors Msx1, Msx2, and Msx3 are expressed primarily at the posterior CNS during early neuroepithelium development . They are induced by bone morphogenetic protein (BMP) signals at the dorsalmost part of the developing CNS and mediate distinct aspects of BMP signaling in dorsal neuronal progenitor specification . Consistent with posteriorization of Hoxb1+ NSCs, Msx1 and Msx3 were both strongly induced by Hoxb1, whereas Msx2 expression remained low in both Hoxb1− and Hoxb1+ NSCs (Fig. 4A; supplemental online Table 1). Hoxb1-mediated upregulation of Msx1 was confirmed by RT-PCR and immunofluorescence (Figs. 4B, 5J–5L). It is worth noting that no expression of Msx1/2 was detected by immunofluorescence in the Hoxb1− ESC-derived NSCs (Fig. 5J–5L). Members of the Pax family proteins are homeodomain-containing transcription factors and play pivotal roles in various aspects of CNS development. Pax3 and Pax7 are expressed in the dorsal ventricular zone (VZ) of the embryonic spinal cord, whereas Pax6 is expressed in most of the ventral VZ, thus defining broad DV domains of neuronal progenitors [51, –53]. Consistent with the change in AP address of Hoxb1+ ESC-derived NSCs, the expression of all three genes changed in response to Hoxb1 induction, as shown by microarray data (Fig. 4A; supplemental online Table 1), semiquantitative RT-PCR (Fig. 4B), and immunofluorescence (Fig. 5A–5C, 5G–5I). The homeodomain-containing transcription factors Nkx2-2 and Nkx6-1 are expressed in the ventralmost neural progenitor domain during early mouse embryogenesis . The expression of both transcription factors changed upon Hoxb1 induction either at the gene expression level (Nkx2-2; Fig. 4A, 4B) or at the protein level (Nkx6-1; Fig. 5D–5F). Immunofluorescence confirmed that the Hoxb1+ cells contained fewer Nkx2-2+ progenitors, but the scarcity of these cells precluded statistically relevant cell counts (data not shown). It is worth noting that regulation of Nkx6-1 and Pax3 was detected only by immunofluorescence, suggesting that post-transcriptional regulatory mechanisms may also be affected by NSC posteriorization. Dbx1 is another homeodomain transcription factor strongly expressed in the diencephalon, midbrain, and r1 but also in a medial domain in the developing hindbrain and spinal cord . Consistent with a posterior shift in the identity of the ESC-derived NSCs, the expression of Dbx1 was downregulated, as shown by both microarray and RT-PCR data (Fig. 4A, 4B). There was an apparent propensity toward dorsalization considering that regulated dorsal progenitor markers were upregulated whereas regulated ventral progenitor markers were downregulated (Fig. 4A, 4B). Nkx6-1 was a notable exception to this tendency (Fig. 5D–5F). However, shh signaling is still active in the Hoxb1+ population, and Nkx6-1 expression is stronger in the developing hindbrain than in the developing forebrain. Thus, Nkx6-1's upregulation is consistent with hindbrain specification of ESC-derived NSCs. On the other hand, some progenitor markers were expressed in both Hoxb1− and Hoxb1+ ESC-derived NSCs (supplemental online Fig. 1), but they were not regulated to any appreciable extent. Double immunostaining experiments with antibodies for Hoxb1 and either Nkx6-1, Pax3, or Pax7 showed that expression of these markers increased in both Hoxb1− and Hoxb1+ cells upon dox addition. The increase was stronger in Hoxb1+ cells, suggesting the presence of both cell-autonomous and non-cell-autonomous effects (data not shown).
The mixed DV identity of these cells prompted us to investigate the extent to which DV signaling pathways are operational in these cells. Microarray gene expression profiling showed that transforming growth factor-β (TGF-β), Wnt, Shh, and RA signaling pathway components, including the signals themselves, are expressed in both Hoxb1− and Hoxb1+ ESC-derived NSCs (supplemental online Table 1). Therefore, all four DV patterning mechanisms are in place, and cells can generate and respond to such signals in an autocrine/paracrine fashion. Furthermore, expression of selected components of all four signaling pathways changed upon Hoxb1 induction (Fig. 4; supplemental online Table 1). This suggested that Hoxb1-induced changes in these signaling pathways contributed to the redistribution of DV identities in the Hoxb1+ ESC-derived NSCs. This was consistent with the finding that Hoxb1-mediated changes in the DV progenitor identity had a non-cell-autonomous component. Several components of the Wnt signaling pathway were affected by Hoxb1 expression, including Wnt1, Wnt3a, Sfrp1, and Sfrp2 (Fig. 4A, 4B). The most notable regulation of a TGF-β signaling pathway component was that of BMP4 (Fig. 4A, 4B). Expression of Shh itself was also affected (Fig. 4A, 4B), but no other components of Shh signaling pathway were significantly regulated (Fig. 4A, 4B). From the RA signaling pathway, CRABPI and CRABPII, which encode proteins controlling intracellular availability of RA, were strongly upregulated (Fig. 4A, 4B).
Thus, we have established that ESC-derived NSCs assume a mixed DV identity as a result of active DV signaling pathways. Hoxb1-driven specification of the AP identity of ESC-derived NSCs resulted in changes in the expression of DV signaling pathways, as well as changes in the relative distribution of the expression of DV progenitor markers. These data provided, for the first time, strong evidence that Hox transcription factors have a direct and broad impact on DV specification of neuronal progenitors and identified novel potential Hox target genes and processes.
Several phenotypes of Hox compound mutants are linked with loss of entire structures or lineages [56, –58]. We examined whether sustained Hoxb1 expression in ESC-derived NSCs changed the proliferation and differentiation patterns of the cells.
ESC-derived Hoxb1− NSCs responded to mitogen withdrawal with cell cycle exit, formation of neuronal clusters, and extension of axons. In contrast, Hoxb1+ NSCs continued to proliferate, and few cells seemed to differentiate (Fig. 6A, 6B). To quantitate these differences, equal numbers of Hoxb1− and Hoxb1+ cells were replated (day 0) on Matrigel at the end of the selection period in the presence of mitogen to initiate the expansion phase. At day 4, cells were passaged, and the following day (day 5) the mitogen was withdrawn. Cell proliferation was determined by cell counts of Hoxb1− and Hoxb1+ ESC-derived NSCs after successive passages at days 4, 9, and 12. (Fig. 6G; data not shown). Cell cycle kinetics were determined by BrdU incorporation after short (2 hours, to mark cells in the S-phase) and long (20 hours, to mark cells actively dividing) exposure to BrdU at days 2, 7, and 9 (Fig. 6D, 6E; statistics in supplemental online Table 2), as well as PH3 immunofluorescence at days 2 and 9 to mark cells in mitosis (Fig. 6H; statistics in supplemental online Table 2). Terminal deoxynucleotidyl transferase dUTP nick-end labeling (TUNEL) assays were also conducted to determine whether there was any appreciable difference in cell death rates (data not shown).
All parameters were very similar if not identical in Hoxb1− and Hoxb1+ cells during the expansion phase (days 0–4). On the contrary, differences between Hoxb1− and Hoxb1+ cells were dramatic upon mitogen withdrawal. We were unable to passage the Hoxb1− cells after mitogen withdrawal, whereas Hoxb1+ cells kept dividing for up to three passages before slowing down and committing to differentiation (Fig. 6G; data not shown). The proliferation rate of Hoxb1+ cells was transiently higher in the absence of mitogen, as suggested by cell counts at successive passages (Fig. 6G; data not shown) and the number of PH3+ cells after mitogen withdrawal (Fig. 6H; supplemental online Table 2). BrdU incorporation was significantly higher in Hoxb1+ cells at both day 7 and day 9 for both short and long BrdU pulses (Fig. 6D, 6E, 6I–6K; supplemental online Table 2). TUNEL assays at both day 7 and day 9 suggested similar rates of cell death in Hoxb1− and Hoxb1+ cells (data not shown). In the dox-treated cultures proliferation rates of Hoxb1+ and Hoxb1− cells determined by double immunostaining for BrdU incorporation and Hoxb1 expression were virtually identical (data not shown). This suggested that the mitogen-independent proliferative capacity endowed by Hoxb1 expression was a non-cell-autonomous effect.
We then directly compared differentiation rates of Hoxb1− and Hoxb1+ ESC-derived NSCs in the presence (day 2) and absence (day 9) of mitogen using immunofluorescence for the differentiation markers TUJ1, MAP2, and NeuN. Differentiation rates were similarly low in Hoxb1− and Hoxb1+ populations in the presence of mitogen (day 2) (Fig. 7D, 7H, 7L; statistics in supplemental online Table 2). Differentiation rates rose sharply in the Hoxb1− cells in response to mitogen withdrawal (Fig. 7; statistics in supplemental online Table 2). Strikingly, differentiation rates rose only marginally in Hoxb1+ cells, suggesting that sustained expression of Hoxb1+ blocks differentiation. Furthermore, Hoxb1 coexpression with NeuN and Tuj1 was seen in less than 1% of the cells (Fig. 7I; data not shown), suggesting that this effect was cell-autonomous.
To further define the character of Hoxb1+ ESC-derived cells growing in the absence of mitogen, we assessed the expression of NSC markers by immunofluorescence (day 9). Expression of nestin and Sox2 was strongly downregulated, whereas expression of RC2 and vimentin was lost, demonstrating the loss of NSC character (data not shown). Taken together, these data suggest that sustained expression of Hoxb1+ in ESC-derived NSCs confers the cells' transient amplifying progenitor cell character upon mitogen withdrawal.
Hoxb1 Confers Mitogen-Independent Proliferative Capacity Partly by Activation of Notch Signaling
Active Notch signaling has been implicated in the maintenance of NSC identity during the early phases of neural development  and in the differentiation of neural progenitors in the postnatal cortex toward glial fates . Phosphorylated STAT3 at Ser 727 and Tyr 705 mediates some of the effects of Notch signaling in NSCs and neural progenitors. Extracellular signals leading to STAT3 Tyr 705 phosphorylation promoted differentiation of CNS NSCs to a glial fate , whereas Notch-mediated STAT3 Ser 727 phosphorylation correlated with NSC survival and proliferation [60, 61].
The microarray data suggested that Notch signaling is active in both Hoxb1− and Hoxb1+ cells during the expansion phase, as ligands and all necessary components of the pathway were expressed (supplemental online Table 1). To examine whether Notch signaling was implicated in the acquisition of mitogen-independent capacity of Hoxb1+ cells, we blocked Notch signaling by adding the γ-secretase inhibitor DAPT in the culture medium at the time of mitogen withdrawal (day 5). As a result, we observed morphological changes consistent with increased differentiation, which, nonetheless, did not reach the levels of differentiation in Hoxb1− cells. To better define the contribution of Notch signaling in the mitogen-independent proliferation of Hoxb1+ cells, we assessed BrdU incorporation and differentiation rates at day 8. BrdU incorporation rates were consistently lower in Notch-blocked Hoxb1+ cells but remained significantly higher than those observed in Hoxb1− cells (Fig. 6F, 6K; statistics in supplemental online Table 2).
We then assessed rates of differentiation of Hoxb1+ cells in the absence of mitogen and active Notch signaling by immunofluorescence for MAP2, NeuN, and TujI expression. Consistent with the results described above, we found that the differentiation rate of Hoxb1+ cells was accelerated upon blocking of Notch signaling, but it still lagged behind that of Hoxb1− cells (Fig. 5C, 5D, 5G, 5H, 5K, 5L; supplemental online Table 2). It is worth noting that expression of the early differentiation marker TujI was closer to that of Hoxb1− cells than the other late differentiation markers (Fig. 5K, 5L; statistics in supplemental online Table 2). Thus, active Notch signaling contributes to the mitogen-independent proliferative capacity of Hoxb1+ cells, but there are other, so far unknown players, involved as well.
To investigate the mechanism through which active Notch signaling imparts mitogen-independent proliferative capacity in Hoxb1+ cells, we compared the phosphorylation status of STAT3 in Hoxb1− and Hoxb1+ cells in the absence of mitogen (day 8). We found that the balance of phosphorylation was heavily shifted to Ser 727 phosphorylation in Hoxb1+ cells and that it was twice as high as that in Hoxb1− cells, whereas levels of Tyr 705 phosphorylation were two times lower in Hoxb1+ cells than in Hoxb1− cells. STAT3 phosphorylation was generally lower in Hoxb1− cells, with STAT3 Tyr 705 phosphorylation being the predominant form (Fig. 7P). These findings were consistent with the low differentiation and retention of proliferative capacity in Hoxb1+ cells in the absence of mitogen. Inhibition of Notch signaling by DAPT resulted in decrease of both STAT3 phosphorylation forms in both Hoxb1− and Hoxb1+ cells. Importantly, inhibition of Notch signaling by DAPT decreased STAT3 Ser 727 phosphorylation of Hoxb1+ cells to levels similar to those in Hoxb1− cells not treated with DAPT. This suggested that Hoxb1 sustained expression promoted active Notch signaling, which in turn led to increased STAT3 phosphorylation at Ser 727 (Fig. 7P).
Taken together, the data above suggested that sustained Hoxb1 in ESC-derived neural progenitor cells maintained active Notch signaling, which, also acting through STAT3 Ser 727 phosphorylation, conferred the cells' mitogen-independent proliferative capacity. Blocking active Notch signaling did not completely abolish this capacity, suggesting that there are other, as yet unknown, contributing pathways.
Directed differentiation of ESCs into NSCs, neural progenitors, and neural cells has received much attention, particularly for the promise it holds for addressing neurodegenerative disorders and trauma of the CNS. Precise control of the AP and DV identity of ESC-derived NSCs is important for the generation of neural cells with specific phenotypes because NSCs of different AP and DV origins give rise to different types of neurons [12, 13]. However, the issue of spatial specification of ESC-derived NSCs has not been explored. Hox transcription factors are the main effectors of posterior identity in the developing embryo, and we reasoned that they could be used to generate ESC-derived NSCs of specific posterior AP identity.
Inducible expression of a single Hox transgene in ESC-derived NSCs specified cells toward a new AP identity by initiating a genetic program of an r4 hindbrain-like identity and suppressing alternative anterior programs installed by default in these culture conditions. These findings were supported by genome-wide gene expression data through microarray gene expression profiling. The posterior specification took place in the absence of posteriorizing signals that assign posterior cell fates to cells of the neural plate during embryo development [62, , , –66], suggesting that our approach bypassed this early step of regional neural cell fate specification. On the other hand, timely induction of the Hoxb1 transgene was crucial for the activation of a specific posterior program. This observation underlined the importance of cellular context when inducing expression of patterning genes and Hox genes in particular. This cellular context dependence could be due to presence or absence of cofactors and/or coregulators, chromatin accessibility, and epigenetic changes accompanying cell differentiation. Repression of anterior identities occurred in a cell-autonomous manner, suggesting that the system developed recapitulates key aspects of the events taking place in vivo during neural plate regional specification [43, 67]. Incomplete suppression of the anterior programs was attributed to the fact that only up to 60% of the cells were Hoxb1+ after dox-mediated induction of the transgene. This was most likely due to inherent limitations of the system, such as stochastic function of either or both transgenes. We have observed the same effect in similarly generated ESC lines and ESC-derived differentiated cells that inducibly express other transcription factors upon dox addition . The addition of a Tet-inducible resistance transgene could resolve this issue and result in neural cells of a uniform AP identity. Future experiments should concentrate on precise posterior specification of ESC-derived NSCs driven by other Hox genes, particularly Cdx-dependent Hox genes [68, 69]. This will allow the generation of ESC-derived NSCs of specific posterior AP identities and provide us with novel insights into the molecular mechanisms underlying AP specification.
The data presented here suggested for the first time that DV patterning signals are generated in situ by ESC-derived NSCs and act in an autocrine/paracrine fashion to assign DV identities. Specific molecular markers should be used to establish whether these signals emanate from specialized roof or floor plate cells generated in these cultures. Genetic evidence suggested that Hox genes may act as in vivo integrators of AP and DV patterning mechanisms [36, 70, –72]. Accordingly, activation of a Hoxb1-dependent AP genetic program was followed by redistribution of DV progenitor cell identities mimicking the respective in vivo differences along the AP neural axis. This redistribution was accompanied by changes in the expression of selected components of all DV signaling pathways. These changes occurred by both cell-autonomous and non-cell-autonomous mechanisms, suggesting that Hox genes may regulate the DV identity of neural progenitors by directly controlling expression of neural progenitor markers as well as expression and perception of DV signaling pathways. In Drosophila, control of Dpp signaling components by Ubx controls the size of the altere , and our results suggest that similar mechanisms may control the size or DV progenitor pools in the CNS along the AP axis. The ESC-based approach described here will allow us to dissect at the molecular level the interactions of AP and DV patterning mechanisms in controlling neural cell fate by determining commonly regulated genes and gene networks. Given that specific neural subtypes are derived from different DV domains, control of the DV identity of ESC-derived NSCs is an important issue to be addressed. To channel the character of ESC-derived NSCs toward specific DV progenitor subclasses, it may be necessary to simultaneously enhance some DV signals while suppressing others. As these signals have mitogenic activity , the correct combination of signals and inhibitors could selectively promote expansion of the target neuronal progenitor population. The DV expression patterns of known progenitor and postmitotic markers in the hindbrain and r4 in particular will help monitor the success of such an approach.
Several phenotypes of Hox compound mutants are manifested by loss of entire structures or lineages, suggesting that Hox-dependent patterning, lineage specification, and cellular growth are linked [56, –58]. Deregulated Hox expression has been implicated in the expansion of adult hematopoietic stem cells  and associated with malignancies , particularly in the hemopoietic lineage . In this report we demonstrated a direct link of sustained Hox gene expression with increased proliferation of neural progenitors in the absence of a mitogen and provided a mechanistic explanation linking for the first time a Hox gene with neural progenitor cell expansion, maintenance of active Notch signaling, and the JAK/STAT pathway. On the other hand, blocking of Notch signaling did not completely abolished the mitogen-independent proliferative capacity of Hoxb1+ cells. This suggested the presence of other Hoxb1-dependent signals or networks capable of sustaining cell expansion in the absence of a mitogen. It is important to note that this effect was non-cell-autonomous, suggesting that it was mediated by secreted signal(s) produced by the Hoxb1+ cells. Transcripteome comparison of Hoxb1− and Hoxb1+ ESC-derived neural cells in the absence of mitogen may help uncover such factor(s). It is conceivable that they represent mitogenic signals and networks specific for the posterior nervous system. Endogenous posterior NSCs are difficult to expand. Uncovering their expansion requirements will greatly facilitate their study and may help to stimulate the endogenous repair capacities of the CNS in cases of spinal cord degeneration or injury.
Timely inducible expression of a single Hox gene specified ESC-derived NSCs to a restricted AP identity by activating a specific posterior genetic program and repressing alternative anterior programs. This occurred in a cell autonomous manner and in the absence of posteriorising signals such as RA that have pleiotropic effects on cell specification and differentiation. Given that NSCs of different AP identities give rise to different types of neural cells, the ability to precisely control the AP identity of ESC-derived NSCs by timely expression of specific Hox genes will allow the generation of neural progenitors of specific AP identity and this, in turn, will contribute to the in vitro generation of specific neural subtypes of the posterior CNS. The finding that Hoxb1-driven specification of ESC-derived NSCs resulted in changes in the expression of DV signaling pathways as well as changes in the relative distribution of the expression of DV progenitor markers provided strong evidence that Hox transcription factors have a direct and broad impact on the DV specification of neural progenitors. Finally, sustained expression of a single Hox gene in ESC-derived NSCs inhibited differentiation and supported expansion of neural progenitors in the absence of mitogenic stimuli in a non-cell autonomous manner. This was partially attributed to Hoxb1 dependent activation of the Notch signaling pathway and Notch dependent STAT3 phosphorylation at Ser 727 thus forging a link of Hox genes with progenitor cell expansion, maintenance of active Notch signalling and the JAK/STAT pathway. These findings could pave the way to elucidate the requirements to stimulate the expansion of endogenous posterior CNS neural stem and progenitor cells.
Disclosure of Potential Conflicts of Interest
The authors indicate no potential conflicts of interest.
We thank Kerri Reilly for participation in the early stages of this work; Kian-Leong Lee, Alexandros Polyzos, and Thanassis Spathis for help with the microarray experiments; and Robb Krumlauf, James Briscoe, and Vasso Episkopou for discussions and valuable comments on a first draft of the manuscript. A.G. thanks Andrew Lumsden for mentorship during the early stages of the project that were conducted in the Medical Research Council Centre for Developmental Neurobiology at King's College. This work was supported initially by a Wellcome Trust Career Development Fellowship and subsequently by Biomedical Research Foundation of the Academy of Athens (BRFAA) intramural funds (to A.G.) and a BRFAA Graduate Student Fellowship (to M.G.).