Embryonic stem cells neural differentiation qualifies the role of Wnt/β-Catenin signals in human telencephalic specification and regionalization


  • Author contributions: C.N.: conception and design, collection and/or assembly of data, data analysis and interpretation, manuscript writing; C.V., C.B., A.B., P.V., and F.B.-R: collection and/or assembly of data; Y.M.: collection and/or assembly of data, data analysis and interpretation; L.A.: provision of study material; M.P.: conception and design, manuscript writing; A.L.P.: conception and design, data analysis and interpretation, manuscript writing, final approval of manuscript.


Wnt-ligands are among key morphogens that mediate patterning of the anterior territories of the developing brain in mammals. We qualified the role of Wnt-signals in regional specification and subregional organization of the human telencephalon using human pluripotent stem cells (hPSCs). One step neural conversion of hPSCs using SMAD inhibitors leads to progenitors with a default rostral identity. It provides an ideal biological substrate for investigating the role of Wnt signaling in both anteroposterior and dorso-ventral processes. Challenging hPSC-neural derivatives with Wnt-antagonists, alone or combined with sonic hedgehog (Shh), we found that Wnt-inhibition promote both telencephalic specification and ventral patterning of telencephalic neural precursors in a dose-dependent manner. Using optimal Wnt-antagonist and Shh-agonist signals we produced human ventral-telencephalic precursors, committed to differentiation into striatal projection neurons both in vitro and in vivo after homotypic transplantation in quinolinate-lesioned rats. This study indicates that sequentially organized Wnt-signals play a key role in the development of human ventral telencephalic territories from which the striatum arise. In addition, the optimized production of hPSC-derived striatal cells described here offers a relevant biological resource for exploring and curing Huntington disease. Stem Cells 2013;31:1763-1774


In vertebrates, the Wnt/βCatenin canonical pathway is one of the key morphogenetic signals that govern the specification of the telencephalon and later its subregionalization into dorsal (pallium) and ventral (subpallium) structures from which the striatum arises (for review [[1]]). Wnt-agonists or -antagonists are secreted by most, if not all the signaling centers that orchestrate the development of the forebrain [[2]]. As a rule, Wnt caudalizes the developing neural tube [[3]], while DKK1, an inhibitor of Wnt signaling, is necessary to induce anterior brain structures [[4-6]]. In addition, Wnt-signals maintain the identity of the pallium [[7]] while DKK1 is secreted by the floor plate [[4]]. In this context, Wnt-signals counteract sonic hedgehog (Shh) signals that play equally important roles in dorsoventral (D/V) and anteroposterior (A/P) patterning of the mouse forebrain [[8]]. Both signals are integrated at the level of the forkhead transcription factor Foxg1 [[9]].

However, the way Wnt/βCatenin canonical signals impact telencephalic development in human has remained essentially elusive by lack of an appropriate model system. Human pluripotent stem cells (hPSCs) (either human embryonic stem cells [hESCs] or induced pluripotent stem cells [hiPSCs]) provide a unique model system for studying early events of the human brain development [[10]]. They have been shown to be instrumental, for example, in determining the dose requirement for Shh-signals in the D/V patterning of the human neural tube [[11, 12]]. Preliminary data support the hypothesis that blockade of Wnt-signals, alone or in combination with Shh, may play a role in determining either a striatal or a cortical fate [[13-16]]. However, the precise dose and time requirement of the Wnt-signals remain to be determined as are the mutual relationships between Wnt- and Shh-signals during telencephalic specification.

Here, we have taken advantage of a neural conversion system based on the combination of two SMAD inhibitors that induces efficient and synchronized neuralization of hESCs [[17-19]], to characterize with precision the role of Wnt-ligands in human telencephalic specification and subregionalization, with particular reference to their interplay with Shh signaling.

Materials and Methods

Human ES Cell Culture

Human ES cells (H9, XX, passages 40–60; WiCell Research Institute) and I90c17 hiPSC (passages 40–60) are maintained on a layer of mitotically inactivated Mouse Embryonic Fibroblasts (Globalstem, GlobalStem, Rockville, USA, http://www.global stem.com). Human PSCs are cultured in Dulbecco's modified Eagle's medium (DMEM)/F12 Glutamax supplemented with 20% knockout serum replacement, 1 mM nonessential amino acids, 0.55 mM 2-mercaptoethanol, and 10 ng/ml recombinant human FGF2 (all from Invitrogen, Saint Aubin, France, www.invitrogen.com). Cultures are fed daily and manually passaged every 5–7 days. RC9 hESC (XY, passage 20–60; RoslinCells) are maintained on CELLstart (Invitrogen) and grown in STEMPRO medium (Invitrogen). Cultures are fed daily and passaged every 4–5 days using accutase (Invitrogen).

Neural Induction and Differentiation of hPSCs

For neural differentiation, hPSC colonies are manually detached from MEFs (H9 and I90c17) or enzymatically dissociated (RC9) and suspended for 6 hours in neural induction medium consisting of DMEM/F12, neurobasal, N2 and B27 supplement, β-mercapto and penicillin/streptomycin (N2B27 medium) supplemented with 0.1 μM LDN-193189 (Miltenyi Biotec, Bergisch Gladbach, Germany, www.miltenyibiotec.com) or 500 ng/ml Noggin (R&D Systems, Lille, France, www.rndsystems.com), 20 μM SB-431542 (Tocris, Lille, France, www.tocris.com), and 10 μM ROCK inhibitor (Y27632, Calbiochem). At day 1, hPSC-aggregates are transfered on polyornithin laminin-coated dishes without Y27632. At day 10, Noggin or LDN-193189 and SB-431542 are removed and cells are maintained for 10 additional days in N2B27 medium. To study the effect of Wnt/β-catenin signals on A/P and D/V patterning, treatment with mouse recombinant Wnt3a (R&D system), human recombinant DKK-1 (R&D Systems) or XAV-939, a β-catenin-mediated transcription inhibitor, and axin stabilizing agent (Cellagen technology) are started at day 1. To study the effect of Hedgehog pathway on A/P and D/V patterning, treatment with Shh (R&D system) or cyclopamine (Sigma, www.sigmaaldrich.com) are started at days 4 or 10. At day 20, cells are exposed to accutase for 10–20 minutes at 37°C and plated on polyornithin laminin-coated dishes at 75,000 cells per square centimeter. For maintenance and amplification of telencephalic progenitors, cells are cultured for eight additional days. For neuronal differentiation, telencephalic progenitors are exposed to terminal specification medium consisting of N2B27 supplemented with 20 ng/ml Brain-derived neurotrophic factor (BDNF) (R&D Systems), 0.5 mM dbcAMP (Sigma–Aldrich), and 0.5 mM valpromide (Lancaster Synthesis, www.rdchemicals.com) from days 17 to 45.

Quantitative Polymerase Chain Reaction

Total RNA is isolated with the RNeasy Mini Kit (Qiagen) according to the manufacturer's instructions. cDNA is generated from 1 μg of RNA with SuperScript III (Invitrogen) using random primers. Quantitative real time-polymerase chain reactions (QRT-PCRs) are performed with Power SYBR Green PCR Mix (Applied Biosystems, Saint Aubin, France, www.invitrogen.com) and a Chromo 4TM Real-Time system (Bio-Rad). Quantification was performed at a threshold detection line (Ct value). The Ct of each target gene was normalized to the cyclophilin or GAPDH housekeeping gene. For primer sequences see Supporting Information Table 1.

Immunocytochemistry and Quantification

Cells were fixed with 4% paraformaldehyde (PFA) for 10 minutes at room temperature and incubated with primary antibodies overnight at 4°C. Secondary antibodies and DAPI counterstain were applied for 1 hour at room temperature. The primary antibodies used in this study were β-catenin (mouse, BD Bioscience); FOXG1 (rabbit, Neuracell, Rensselaer, NY, www.nsciusa.org or rabbit, Abcam, www.abcam.com); PAX6 (mouse, DHSB); OTX2 (goat, R&D); Nestin (mouse, Chemicon, Billerica, MA, www.millipore.com); SOX1 (rabbit, Chemicon); GSX2 (kind gift from Prof. K. Campbell, Cincinnati, OH); NKX2.1 (rabbit, Biopat, Milan, Italy); CTIP2 (rat, Abcam); HuCD (mouse, molecular probe, France, www.invitrogen.com); MAP2 (mouse, Sigma); DARPP32 (rabbit, Santacruz, Dallas, Texas, www.scbt.com); DARPP32 (mouse, BD Bioscience), Calbindin (rabbit, Swant, Marly, Switzerland, www.swant.com); and Calretinin (rabbit, Chemicon), D2DR (rabbit, Chemicon), Substance P (rat, Chemicon), Tyrosine Hydroxylase (TH) (mouse, Sigma), human specific DARPP32 (rabbit, Abcam), FOXP1 (mouse, Abcam). Quantitative immunocytochemical analysis was performed on randomly selected visual fields from at least two independent differentiation experiments. In each field, images of separate channels were acquired at 10× or 20× magnification on a fluorescence microscope (ImagerZ1; Carl Zeiss) using the Axiovision image capture equipment and software and exported to an imaging software (Adobe Photoshop; Adobe Systems), where separate channel images and the corresponding overlaid images were counted. On average, five visual fields were acquired per 16 mm cover slip, and a total of 100–500 cells were counted per field. The total number of cells expressing the different markers was plotted as a percentage for either all cells 4′,6′-diamidino-2-phénylindole (DAPI), neural precursors (PAX6+) or neurons (MAP2+ ou HuCD+ cells). Data are presented as mean ± SEM or ± SD.

Detection of PAX6 and β-Catenin Nuclear Intensity

Pax6 and β-catenin intensity detection were performed using the Cellomics Array Scan VTI high content imaging system (Thermofisher Scientist, www.thermofisher.com) with the 10× objective to automatically focus the preparation. Images were acquired in high resolution camera mode on two channels: nuclei (channel 1, HOECHST XF93) and PAX6 or β-catenin (channel 2, TRITC (Tetra methyl Rhodamine Iso Thio Cyanate) Sensitive XF32). Based on the Hoechst channel, 10 pictures were acquired and analyzed. Nuclei were identified using an intensity threshold and segmentation parameters and each nuclear region was then reported on the second channel (TRITC) to limit the PAX6 and β-catenin detection to the nuclear area. The entire signal was quantified in each nuclear area applying a low intensity threshold, background correction, and segmentation parameters on the TRITC channel. The average intensity of PAX6 and β-catenin staining was quantified in each nucleus to generate an average intensity value per cell for each image. These average values were then analyzed to calculate the median of the intensity value and SD for each condition. Data are presented as mean ± SEM.

Quinolinic Acid Lesion and Cell Transplantation

All experimental procedures were performed in strict accordance with the recommendations of the European Commission (86/609/EEC) concerning the care and use of laboratory animals. Six adult Nude rats (weight 220–260g at the time of grafting; Charles River Laboratories, Wilmington, MA, www.criver.com) were used. All surgical procedures were achieved under full anesthesia using a mixture of ketamine (15 mg/kg) and xylazine (3 mg/kg; Bayer Healthcare, www.bayerhealthcare.com) and using a stereotaxic frame. Unilateral lesions were made by injecting 1.5 μl of 80 nmol/μl quinolinic acid dissolved in 0.1 M PBS into the right striatum, according to the following coordinates (in millimeters): anteroposterior (A) +0.5; lateral (L) −2.7; ventral (V) −4.2; and tooth bar −3. One week after the lesion, rats received transplants of cells (150,000 in 2 μl of HBSS supplemented with 0.05% DNaseI; Invitrogen) (A+0.5; L-2.7; and V-5.7 and -4.7). Five months after transplantation rats were terminally anesthetized with 1 g/kg intraperitoneally sodium pentobarbital (Ceva Santé Animale, www.ceva.com), and their brain was fixed by transcardial perfusion with 100–150 ml of 0.1 M PBS (pH 7.4), followed by 250 ml of buffered 4% PFA. Brains were removed, postfixed overnight at 4°C in 4% PFA, and then cryoprotected in 30% sucrose solution at 4°C. Coronal brain sections (40 μm) were cut on a freezing microtome, collected serially (interspace, 480 μm), and stored at −20°C in a cryoprotectant solution until analysis.

Grafted cultures were treated as described above with 100 ng/ml DKK1 and 50 ng/ml Shh for 20 days then passaged and maintained at lower density in N2 medium supplemented with 100 ng/ml DKK1 and 50 ng/ml Shh for five extra days before grafting. For fluorescent immunohistochemical analyses, the brain sections were incubated with primary antibodies for 12 hours at 4°C. Secondary antibodies and DAPI counterstain were applied for 3 hours at room temperature. When 3,3′-diaminobenzidine was used in immunohistochemical staining, analyses were performed using a Discovery XT (Ventana, www.ventana.com) staining automate.

Statistical Analysis

The data were processed using Prism5 software. Values are reported as mean ± SEM. Comparisons among values for all groups were performed by one-way analysis of variance, and Dunnett's multiple-comparison test was used to determine the level of significance.


The effect of Wnt-pathway modulation on the specification of human neuroectoderm cells toward a forebrain fate was examined using H9, RC9 hESCs and I90c17 hiPSCs (Supporting Information Fig. S1) neuralized with two SMAD inhibitors, SB-431542 and LDN-193189, in a chemically defined medium (see schematic diagram of the differentiation protocols in Supporting Information Fig. S2). Impact of concurrent dual-SMAD inhibition (SMADi) combined with the activation or inhibition of Wnt-signaling on hESCs or hiPSCs was measured by QRT-PCR and immunostaining after differentiation in vitro for 10 and 20 days. These two time-points were considered representatives for two phases: the initial phase of neural induction and A/P neuraxis specification and the subsequent phase of telencephalic subregionalization. Modulation of Wnt-pathway activity was achieved by addition of Wnt3a and DKK1 recombinant proteins, or the chemical compound XAV-939, a β-catenin-mediated transcription inhibitor and axin stabilizing agent (i.e., Wnt-antagonist) and CHIR-99021 a GSK3 inhibitor (i.e., Wnt-agonist). Addition of Wnt3a, CHIR-99021, DKK1, or XAV-939 to cultures during the first 10 or 20 days of hESC-neural differentiation efficiently modulated canonical Wnt-signaling pathways, as indicated by changes in β-catenin nuclear intensity and the expression levels of two of its transcriptional targets, namely, LEF1 and AXIN2 (Supporting Information Fig. S3A, S3B). As a control, SOX1 and Nestin QRT-PCR analyses and immunostaining showed that the modulation of Wnt-pathway activity did not affect the neural induction mediated by SMADi (Fig. 1A, 1E; Supporting Information Fig. S3C).

Figure 1.

Effect of Wnt signaling on specification of telencephalic progenitors from human embryonic stem cells (hESCs). (A–C): Gene expression in H9-hESCs cultures differentiated in the presence or absence of Wnt3a (40 ng/ml) or the Wnt-antagonist DKK1 (100 ng/ml) from days 1 to 10. (A): Relative mRNA expression for neural marker SOX1, (B) forebrain markers FOXG1, SIX3, and forebrain–midbrain marker OTX2, (C) mesencephalic marker OTX1 and early spinal cord marker HOXB4. (D): Immunolabeling of PAX6, FOXG1, and OTX2 in hESC-derived cultures at day 10. Proportion of PAX6-positive cells expressing FOXG1 and of PAX6 or OTX2-positive cells in cultures generated in the presence or absence of Wnt3a (40 ng/ml), DKK1 (100 ng/ml) or XAV (1 and 4 μM) from days 1 to 10. (E–G) Gene expression in H9-hESCs cultures differentiated after 20 days of treatment with different concentrations of the chemical Wnt-antagonist XAV-939 (0, 0.25, 1, and 4 μM) and 10 days of sonic hedgehog (Shh) (50 ng/ml). Scale bar = 100 μm. (E): Relative mRNA expression for neural marker NESTIN, (F) telencephalic markers FOXG1, SIX3, and diencephalic marker OTX2, (G) mesencephalic markers OTX1 and LMX1A. All expression levels are normalized to levels detected in control condition (no treatment). Data are presented as mean ± SEM, n = 4. (H-I): Immunolabeling of FOXG1 in hESC-derived cultures at day 20. (H): Proportion of cells expressing FOXG1 in cultures differentiated with different concentrations of XAV-939 from days 1 to 20. Scale bar = 50 μm. (I): Proportion of cells expressing FOXG1 in cultures differentiated with XAV1 during 10 days and then maintained in presence of CHIR-99021 (3 μM) or different concentrations of XAV-939 from days 10 to 20. Scale bar = 100 μm. Data are presented as mean ± SEM, n = 2 (G) or mean ± SD, n = 5 for (H). (*, p < .05; **, p < .01; ***, p < .001). Blue, DAPI. See also Supporting Information Figs. S3, S4; Table S1.

Inhibition of Wnt-Pathway Promotes Anterior Central Nervous System Identity of hESC-Neural Derivatives in a Dose-Dependent Manner

Wnt-pathway inhibition or activation produced pronounced and opposite changes in the expression of A/P region-specific markers of the developing brain. QRT-PCR analyses at day 10 showed that addition of DKK1 increased the expression of SIX3 and FOXG1, two markers of the developing forebrain (Fig. 1B), but did not change the expression of more caudal marker genes such as OTX2, OTX1, or HOXB4 (Fig. 1B, 1C). In contrast, Wnt3a treatment increased the expression of the most caudal markers OTX1 and HOXB4, while significantly decreasing the more rostral ones (FOXG1, SIX3, and OTX2) (Fig. 1B, 1C). To specify the role for antagonism of Wnt signaling in the determination and in the long-term maintenance of telencephalic identity in hESC-derived neural cells, we substituted DKK1 by a specific chemical antagonist (XAV-939) of Wnt/β-catenin pathway that allowed discrete dose-dependent studies.

Immunostaining at day 10 confirmed QRT-PCR results (Fig. 1D). In the absence of major changes in the neural markers PAX6, Nestin, and SOX1 (Fig. 1D; Supporting Information Fig. S3D), FOXG1 immunopositive cell counts were significantly increased by DKK1 or XAV-939 treatment. H9 hESCs treated with DKK1 yielded 78 ± 5% of FOXG1+ cells among PAX6+ cells, 85 ± 2% with 1 μM XAV-939 and 90 ± 1% with 4 μM XAV-939. Conversely, Wnt3a treatment totally blocked the production of FOXG1+ cells. Addition of 1 μM of XAV-939 together with Wnt3a partially inhibited the effect of the Wnt-ligand and yielded almost twice as many FOXG1+ cells as in cultures treated with Wnt3a alone (Fig. 1D). Wnt3a treated cultures yielded 43 ± 5% of OTX2 immunopositive cells. In contrast, all cell cultures treated with Wnt-antagonists (XAV-939 or DKK1) induced no change in the proportion of OTX2 immunopositive cells (untreated: 74 ± 1%; DKK1:81 ± 1%; XAV-939 1 μM: 75 ± 3% and 4 μM 81 ± 6%; Wnt3a+XAV-939 1 μM: 70 ± 7%) (Fig. 1D). Similar results were obtained with RC9 hESC and I90 hiPSC (Fig. S4A, S4B)

Dose–response experiments analyzed at day 20 gave results consistent with those obtained at day 10. Maximum increase in the expression of SIX3 and FOXG1 measured by QRT-PCR in H9 culture treated with Shh from days 10 to 20 was achieved with as little as 250 nM XAV-939 (Fig. 1F). OTX2 expression was not changed for any XAV-939 concentration, suggesting that antagonism of Wnt signaling in hESC-derivatives determine only the most rostral part of the forebrain, that is, the telencephalon. A negative and dose-dependent effect of XAV-939 was additionally detected on the expression of OTX1 and LMX1A, markers of brain regions caudal to the diencephalon (Fig. 1G).

These results were confirmed by FOXG1 immunostaining at day 20. A similar increase in proportion of FOXG1+ cells was obtained in H9 cell cultures treated with XAV-939 either with or without Shh from days 10 to 20 (Fig. 1H; Supporting Information Fig. S3E). Proportion of FOXG1 immunopositive H9 cells treated with 0, 0.25, 1, and 4 μM of XAV-939 combined with Shh from days 10 to 20 was 28 ± 4%, 80 ± 2%, 87 ± 4%, and 85 ± 4%, respectively.

To explore Wnt-signals requirement for the maintenance of the telencephalic identity achieved after 10 days of treatment with 1 μM of XAV-939, H9 hESCs were treated from days 10 to 20 with 3 μM CHIR-99021 or with 0, 0.25, 1, and 4 μM of XAV-939 without Shh. Drop to 0 μM or increase to 4 μM of XAV-939 concentration did not change the proportion of FOXG1+ cells (as compared to 1 μM XAV-939 condition). In contrast activation of Wnt-pathway by CHIR-99021 dramatically reduced the proportion of FOXG1+ cells (Fig. 1I). Altogether our results indicate that Wnt-antagonist signals are necessary for the telencephalic specification of hESC neural derivatives and that the maintenance of this rostral identity does not require active inhibition of Wnt-pathway. It can, however, be completely abolished by strong activation of the Wnt/β-catenin pathway.

Shh Pathway Activity Does Not Modulate Forebrain Fate Mediated by Wnt-Inhibitors in hESC-Derivatives

The zona limitans intrathalamica (ZLI), the signaling center that defines the telencephalic–diencephalic boundary in vertebrates, secretes Shh along with Wnt3a and other Wnt-pathway agonists [[20, 21]]. The influence of Shh-pathway activity on the induction and maintenance of telencephalic identity of hESC-derivatives induced by Wnt-inhibiton was, therefore, examined. Modulation of Shh-pathway was achieved by addition of a high-potency form of Shh (SHHC25II) or cyclopamine (CYCLO), a chemical antagonist of Smoothened. Addition of these molecules to neural cell cultures derived from H9 hESCs induced a dose-dependent increase in the expression of Patched-1 (PTC1) and GLI1, two downstream effectors and transcriptional targets of the Hedgehog pathway (Supporting Information Fig. S5A).

In DKK1 treated hESC cultures, the modulation of the activity of Shh-pathway using Shh or cyclopamine did not alter the expression levels of neural (SOX1 or NESTIN), forebrain (SIX3 and FOXG1) or diencephalic (OTX2) markers measured by QRT-PCR at days 10 and 28 (Fig. S5B, S5C). These results were confirmed by immunostaining performed using antibodies against FOXG1 on day 28 cell cultures (Supporting Information Fig. S5D). Similar results were obtained comparing the yield of FOXG1 immunopositive cells in H9 or RC9 hESC cultures treated with XAV-939. At day 20, FOXG1 immunopositive yields were, for example, 87 ± 4% and 82 ± 4% in Shh-treated and untreated cultures, respectively (Fig. 1H; Supporting Information Figs. S3E, S6A). Altogether, a role of Shh-signals on the induction or maintenance of forebrain A/P patterning was not detectable in our in vitro system.

Shh Dose-Dependent Treatment Induces Ventralization of hESC-Derived Telencephalic Cells

Shh is the best known morphogen for ventral patterning of the neural tube throughout the neuraxis including the most anterior part of the developing head [[22]]. The influence of Shh signaling on the D/V identity of DKK1-mediated human telencephalic cells was, therefore examined. QRT-PCR analyses of DKK1-treated neuralized hESCs showed that the level of expression of EMX1 and TBR2, two markers of the developing cortex, was similar in H9 hESC cultures additionally treated with Shh-antagonist cyclopamine (from days 10 to 28) and controls (Fig. 2A). The active suppression of Shh-activity by cyclopamine had consequently no impact on the dorsalization of these progenitors. This result is not consistent with the enhanced dorsal fate observed in mouse ESC-derived telencephalic progenitors treated with cyclopamine [[11]] and suggest that unlike in mouse, in the absence of Shh-signals, human telencephalic cells exhibit a default dorsal identity.

Figure 2.

Dose–response effect of sonic hedgehog (Shh) on dorsal–ventral patterning of human embryonic stem cell-derived telencephalic progenitors. (A): Relative mRNA expression for dorsal (EMX1 and TBR2), lateral ganglionic eminence (GSX2, DLX2, and ASCL1), and medial ganglionic eminence (NKX2.1) markers in cultures generated with DKK1 (100 ng/ml; days 1–28) and cyclopamine (1 μM; days 10–28) or different concentration of Shh (50 and 200 ng/ml; days 10–28). (B): Quantification of GSX2, CTIP2, and NKX2.1-expressing cells among total cells in the presence of cyclopamine (1 μM) or different concentration of Shh (50 and 200 ng/ml) as in (A). Scale bar = 50 μm. Data are presented as mean ± SEM, n = 4. (*, p < .05; **, p < .01; ***, p < .001). Blue, DAPI. See also Supporting Information Figs. S5, S6; Table S1.

In contrast, expression of EMX1 and TBR2 was significantly lower, as compared to controls, in cell cultures treated with Shh. Concomitantly, Shh induced a dose-dependent increase in expression of all tested markers for ventral forebrain regions, namely, DLX2, GSX2, NKX2.1, and ASCL1 (Fig. 2A). Immunostaining for GSX2, NKX2.1, and CITP2 (BCL11B) confirmed QRT-PCR data on pallial and subpallial markers (Fig. 2B). Modulation in Shh-pathway activity changed the yield of cells immunopositive for GSX2, a transcription factor specifically expressed in the lateral ganglionic eminence (LGE) in human [[23]]. Optimal yield of GSX2+ cells were obtained in cell culture treated with intermediate concentration of Shh (50 ng/ml); consistently the lowest yield was observed in cyclopamine treated cultures (Fig. 2B). The yield of cells expressing CTIP2, a marker present both in developing cerebral cortex and in LGE, decreased only in cell cultures treated with the highest dose of Shh (200 ng/ml). With this highest dose of Shh, the yield of cells immunopositive for the medial ganglionic eminence (MGE) marker NKX2.1, the most ventral population of the telencephalon, was maximal (Fig. 2C). Similar results were obtained using RC9 hESCs (Supporting Information Fig. S6). Altogether, telencephalic progenitors obtained using Wnt inhibitors were highly sensitive to D/V patterning cues resulting from the modulation of Shh-pathway activity in agreement with previous results obtained for hESC forebrain derivatives specified by default [[15, 24]].

Wnt-Pathway Inhibition Enhances Shh-Mediated Ventral Patterning of Human Telencephalic Cells in a Dose-Dependent Manner

The role of Wnt-signals inhibition on ventralizing signals mediated by Shh was finally examined using Shh concentrations that gave optimal LGE commitment (50 ng/ml) and increasing doses of XAV-939. In those conditions, QRT-PCR analyses showed that the expression of the cortical markers EMX1 and PAX6 decreased when H9 derived cells were treated with at least 1 μM of XAV-939 between days 1 and 20 (Fig. 3A), while it was stable for a lower concentration. In parallel, expression of the subpallial marker genes DLX2, GSX2, ASCL1, MEIS2, NKX2.1, and LHX6 increased in a XAV-939 dose-dependent manner (Fig. 3B, 3C). Increased inhibition of Wnt-pathway, from 0.25 to 1 μM of XAV-939, proportionally increased the expression of four marker genes expressed in the LGE, namely DLX2, GSX2, ASCL1, and MEIS2 (Fig. 3C). Expression levels of these LGE genes were not further increased with higher concentration of Wnt inhibitor (4 μM). In contrast, marker genes of the MGE, namely, NKX2.1 and LHX6, showed a progressive increase in expression up to the highest concentrations of XAV-939 (Fig. 3B).

Figure 3.

Dose–response effect of XAV-939 on dorsal–ventral identity of human embryonic stem cell (hESC)-derived telencephalic progenitors. (A–D) Gene expression in H9-hESCs derivative cultures generated with different concentrations of XAV-939 (days 1–20) with sonic hedgehog (Shh) (50 ng/ml; days 10–20). (A): Relative mRNA expression for dorsal markers (PAX6 and EMX1), (B) for markers of lateral ganglionic eminence (MEIS2, DLX2, GSX2, and ASCL1), and (C) medial ganglionic eminence markers (NKX2.1 and LHX6). Data are presented as mean ± SEM, n = 4 (D-E) Quantification of NKX2.1-, PAX6-, and GSX2-expressing cells among total cells in hESC derivative cultures generated with different concentrations of XAV-939 and with a 10-day treatment of Shh (50 ng/ml; days 10–20). Scale bar = 50 μm. Data are presented as mean ± SD, n = 5. (F): Quantification of PAX6 intensity in H9-hESCs derivative cultures generated with different concentrations of XAV-939 and with a 10-day treatment of Shh (50 ng/ml). All signal intensity of positive cells are normalized to intensity detected in control condition. Data are presented as mean ± SD, n = 5. Blue, DAPI. (*, p < .05; **, p < .01; ***, p < .001 for comparison with XAV-939 0 μM and °, p < .05; °°, p < .01; °°°, p < .001 for comparison with XAV-939 0.25 μM). See also Supporting Information Figs. S7, S8; Table S1.

GSX2-immunopositive cell counts of culture treated with Shh from days 10 to 20 days depended on the level of Wnt-pathway inhibition (Fig. 3E). Cell cultures treated with 1 μM of XAV-939 from days 1 to 20 yielded the highest number of GSX2+ cells while at a concentration of 4 μM, NKX2.1 immunopositive cells were the most numerous (Fig. 3E). Conversely, while the amount of PAX6 immunopositive cells decreased only at the highest dose of XAV-939 (Fig. 3F), the nuclear intensity of PAX6 immunoreactivity decreased already at the lowest Wnt-inhibitor concentration used (Fig. 3G). These results reflected the dual role of PAX6 in human central nervous system development as an early neuroectodermal cell marker and as a later appearing cortical neuronal marker [[25-27]]. This change in PAX6 intensity was already detectable at day 10 (Fig. S3E). Similar XAV-939 dose-dependent ventral patterning activity was observed using I90c17 hiPSCs (Supporting Information Fig. S7).

The ventralizing activity of Wnt-anatgonist signals in absence of Shh signals was finally examined. H9 hESCs treated for 10 days with 1 μM of XAV-939 to generate telencephalic progenitors were then treated from days 10 to 20 with 3 μM CHIR-99021 or with 0, 0.25, 1, and 4 μM of XAV-939 without Shh. QRT-PCR data at day 20 showed that expression of the cortical markers EMX1 remained unchanged by increasing concentration of XAV-939 while the expression of subpallial marker genes GSX2, NKX2.1 increased with XAV-939 concentration (Supporting Information Fig. S8A). Immunostaining for GSX2, OTX2, NKX2.1, and PAX6 specified more precisely Wnt-antagonist activity in absence of Shh as they revealed that GSX2-immunopositive cell counts of culture treated without Shh from days 10 to 20 days increased with the concentration of Wnt-pathway inhibitor, but never amounted to more than one-third of the cell counts achieved with combined XAV-939 and Shh. Furthermore, NKX2.1 immunopositive cells counts were even lower as hardly any NK2.1+ cells could be observed in most cultures, even when they were treated at the highest concentration of XAV-939 (Supporting Information Fig. S8B). Overall, inhibition of the Wnt signaling pathway coaxed telencephalic cells toward a more ventral fate in a dose-dependent manner. When the Shh-pathway was activated even the most ventral fate could be enhanced by Wnt-inhibition. In the absence of Shh-signals, this fate remained out of reach of Wnt-inhibition and only intermediate telencephalic region were promoted by Wnt-inhibitor.

Defined Levels of Wnt- and Shh-Pathway Modulations Drive Human Neural Progenitors to a LGE-Like Identity with Therapeutic Potential for Huntington Disease

The quality and relevance of the induction of an LGE-like positional identity by moderate Wnt-inhibition and Shh activation in hPSC derivatives largely depend on the strength of this commitment after withdrawal of both signals during the terminal neuronal differentiation and maturation. We, therefore, tested the maintenance of the positional identities mediated by Shh and increasing dose of Wnt-antagonists after terminal differentiation in vitro and in vivo in a phenotypic model of HD-like striatal degeneration in the nude rat.

In vitro, day 20 H9-derived culture were differentiated for 25 extra days in a culture medium that contained BDNF, cAMP and valproic acid (Fig. 4A). The yield of neurons (MAP2 or HuCD-immunopositive cells) expressing DARPP32 or Calbindin, two classical markers of Medium Spiny Neurons (MSN) [[28-31]] increase with XAV-939 concentration, optimal yield being achieved at 1 μM (Fig. 4B). Cell counts of MGE, calretinin-immunoreactive neurons [[32]] significantly increased when cells were treated at a higher concentration of XAV-939 (4 μM). Altogether, ventral telencephalic cells generated using Shh were responsive to D/V patterning cues resulting from the inhibition of the Wnt-pathway. These results confirmed analyses performed on day 20-differentiated cultures that identified a combination of moderate Wnt-inhibition (1 μM of XAV-939 or 100 ng/ml of DKK1) and Shh-activation (50 ng/ml SHHC25II) as the optimal parameters for production of LGE-like neurons.

Figure 4.

Generation of striatal neurons from lateral ganglionic eminence and medial ganglionic eminence progenitors. (A): Immunolabeling at day 45 of MAP2-, HuC/D-, DARPP32-, Calbindin-, and Calretinin-positive cells in cultures differentiated with different concentrations of XAV-939 (days 1–20) and with sonic hedgehog (50 ng/ml; days 10–20). Scale bar = 100 μm. (B): Quantification of GSX2, DLX2, and NKX2.1-expressing cells among total cells. Data are presented as mean ± SEM, n = 2. Blue, DAPI. (*, p < .05; **, p < .01; ***, p < .001) and °, p < .05; °°, p < .01; °°°, p < .001 for comparison with XAV-939 0.25 μM).

Longer term differentiated neuronal cultures (day >60) were characterized by QRT-PCR and immunostaining (Fig. 5A, 5B). Longitudinal differentiation analyses of H9 hESC differentiation showed early expression of FOXG1 starting at day 10 and persisting during terminal differentiation at day >60 (Supporting Information Fig. S9) similarly to what is reported during the development of the telencephalon in mice [[8, 33, 34]]. Significant increases of the expression of marker genes of postmitotic neurons such as MAP2, alpha synuclein (SNCA), or synaptophysin (SYP) were observed only in day >60 cultures (Fig. 5A; Supporting Information Fig. S9). Robust maturation of early telencephalic cells (day 10) into MSN-like cells was illustrated by the increase in expression of key marker genes including: GAD1 (FC = 428 ± 123), CITP2 (FC = 1,064 ± 256) [[35]], ARPP-21, (FC = 6,439 ± 1477) [[36, 37]], STEP (FC = 17 ± 3) [[38]], DARPP32 (FC = 23 ± 6), CALB (FC = 297 ± 55), dopamine receptor D1 (DRD1) (FC = 6 ± 1), and Substance P (Subst P) (FC = 19 ± 1) [[39]]. Immunostaining of these cultures confirmed DARPP32-positive coexpressed CTIP2, FOXP1, and DRD2 (Fig. 5B).

Figure 5.

Striatal differentiation and maturation of human embryonic stem cell-derived lateral ganglionic eminence (LGE)-like culture in vitro and in vivo (A) Relative mRNA expression for MSN markers at day 60 in neuronal cultures obtained from LGE progenitors differentiated with XAV1 μM and sonic hedgehog 50 ng/ml. Data are presented as mean ± SEM, n = 4 (*, p < .05; **, p < .01; ***, p < .001 for comparison with day 10). (B): Immunolabelings at day 60 of DARPP32-, CTIP2, Substance P, and D2DR in neuronal cultures obtained from LGE progenitors. Scale bar = 100 μm. (C): Graft stained for human specific DARPP32 (hDARPP32), TH, CTIP2 FOXP1, and immunolabelings of hDARPP32, FOXP1, CTIP2, FOXG1, HNA, PAX6, and MAP2 in QA-lesioned striatum of nude rats 5 months posttransplantation. Scale bar = 100 μm. Abbreviation: HNA, human nuclear antigen.

The viability and differentiation potential of the same LGE-like cultures was also assessed in vivo. Day-25 differentiated H9 hESCs (150,000 total) were grafted into the striatum of quinolinate-lesioned adult nude rats. Brain tissues were harvested at 5 months posttransplantation and analyzed by immunohistochemistry (Fig. 5C). Graft-derived cells remained clustered and occupied most of the host lesioned striatum. Graft derived cells, positive for human nuclear antigen (HNA) expressed FOXG1 confirming the telencephalic identity of the transplant. Extensive CITP2 and FOXP1 staining was also evident. Human specific DARPP32-immunostaining was very extensive in the graft; most human specific DARRP32-positive cells featured neuronal morphologies, costained for CTIP2 and FOXP1 and extended fibers outside the core graft. Conversely, TH-positive fibers innervated many field within the graft while no TH-positive soma was observed.


The main result of this study is the demonstration that the inhibition of Wnt-signaling promotes human telencephalic specification. Inhibition of Wnt/β-catenin pathway by addition of either DKK1 or XAV-939 in hESC-derived neuroectodermal cell cultures strongly enhanced telencephalic commitment while, conversely, addition of Wnt3a or CHIR99021 prevented a forebrain fate. At a later stage, human telencephalic cells responded in a dose-dependent fashion to ventralizing signals mediated not only by Shh but also by Wnt/β-catenin inhibitors; the combination of both enhancing the effects. Concurrent treatment of neuralized hPSCs with intermediate concentrations of Wnt-antagonist and Shh-agonist produced LGE-like cells mostly committed to differentiation into MSN-like neurons both in vitro and in vivo in a rat model of HD-like striatal degeneration. Overall, our study is consistent with what was found conserved among vertebrates on early neural patterning events (for review see [[40]]). Our data, extending to human the analysis of proteins that direct telencephalic specification and subregionalization, highlight the value of hPSCs as a model system for studying neurodevelopmental biology.

The default positional identity of neuroectodermal cells derived from hPSCs in the absence of patterning signals has most often been described as the primitive anterior fate, independently of the neural conversion protocol [[13, 41-43]]. However, the precise definition of such anterior commitment has not been established, by lack of spatial/temporal correlates. We specifically addressed that issue by assessing the expression of the classical forebrain determinant FOXG1 and OTX2 which are both expressed in the telencephalic vesicle at 11-week of gestation in human [[23]]: Foxg1 is a transcriptional repressor required for the development of the telencephalon [[8, 33, 34]]; OTX2 is expressed during neural induction in the entire anterior neuroectoderm but is later confined to the diencephalon and mesencephalon, with only a minor expression in the basal telencephalon [[44, 45]]. In our system, hESCs defaulted into a mixed population of anterior neural cells including diencephalic–mesencephalic FOXG1-/OTX2+ and telencephalic FOXG1+/OTX2+ cells. Unlike FOXG1, OTX2 expression was not affected by the Wnt-suppression while reduced by Wnt3a as previously shown [[42]]. These results indicate that the fate spontaneously achieved by human neurocetoderm cells, in our culture system is an intermediate anterior fate just posterior to the telencephalon.

In vertebrates, the rostral patterning center (also named anterior neural ridge) has been shown to secrete two classes of Wnt-pathway inhibitors, DKK1 and secreted frizzled-related proteins [[5, 6, 46]]. Direct genetic evidence shows that Wnts inhibit anterior neural plate fates and, to the opposite, caudalize the developing neural tube [[3]]. Conversely, suppression of Wnt-activity is necessary for inducing anterior structures [[4-6]]. The substantial increase in the yield of FOXG1+/OTX2-/Pax6low cells upon treatment with increasing doses of Wnt-inhibitors in our experiments complement those findings. Our work on a human neuraxis in which subregionalization is more amenable to analysis than in previous models indicate that active Wnt-suppression and not just the lack of Wnt-agonists is one of the factors that determine the most-rostral identity. Maintenance of this telencephalic positional identity may however not require sustained Wnt-antagonist signal, at least in absence of high concentration of Wnt-agonist as suggested by our experiments during which the concentration of Wnt-antagonist was increased or reduced in the culture medium after the initial phase of telencephalic specification.

Wnts are also secreted dorsally by the cortical hem (roof plate), the dorsal organizer of the neural tube [[47]], while two classes of Wnt-pathway inhibitors are secreted at the pallial-subpallial boundary by the anti-hem [[48]] and ventrally by the floor plate, the ventral organizer [[49]]. Like Shh-pathway, Wnt-pathway activation or inhibition thus participate in the regulation of (D/V) patterning events in the most rostral part of the brain in vertebrates. The D/V patterning effects of both systems interplay at the levels of FoxG1 and Gli3 [[9]]. Foxg1 is a telencephalon-specific downstream effector of the hedeghog pathway that inhibits Wnt/β-catenin signaling through direct transcriptional repression of Wnt ligands. Conversely, Wnt signals upregulate the truncated form of Gli3 which in turn represses Shh [[15]]. In hPSC-derived telencephalic progenitors, Shh-mediated ventralization of these cells was also subsequently enhanced by Wnt-inhibiton in a dose-dependent manner. This was best illustrated by the generation of either GSX2 and NKX2.1-immunoreactive cells according to Wnt-pathway inhibition levels. In absence of Shh, ventralizing properties of Wnt-antagonist on human telencephalic cells Shh was reduced, high-dose of XAV-939 was not sufficient to promote MGE identity (NKX2.1+ cells) and only partially promoted LGE identity (GSX2+ cells). This might suggest that crosstalk between Wnt and Shh-pathways are specifically relevant for the definition of the D/V positional identity of the LGE.

MSN, are specifically affected in patients with Huntington disease (HD). Unlimited access, in vitro, to authentic MSN population derived from hESCs or hiPSCs open new perspectives to explore and treat HD. MSN culture can, in particular, be used as a substitute source of human striatal cells for current HD cell therapy, as a cellular model of HD for the validation of human-specific gene therapies, for deciphering molecular mechanisms underlying HD and in drug discovery (for review see [[50]]). In this study, while exploring Wnt and Shh requirement for the specification of human telencephalic ventral territories, we defined in vitro conditions for optimal striatal differentiation of hPSCs. Precise concentrations of Wnt-antagonist and Shh, when combined, generate high yields of ventral telencephalic precursors (LGE-like population), capable to mature both in vitro and in vivo into MSN-like neurons coexpressing DARPP32, CTIP2, FOXG1, and FOXP1. In vivo, these LGE-like populations survived transplantation into the quinolinate-lesioned striatum of nude rats, fully maintained their positional identity and extensively matured into MSN-like neurons.

In the specific context of HD cell therapy, preliminary clinical results have been reported using human MSN precursors, directly obtained from fetal tissue [[51, 52]]. However, logistical difficulties in the retrieval of human fetal tissues limit severely the number of eligible patients for this therapy. The precise definition of graft composition (e.g., the proportion of LGE- and MGE-derived projection and interneurons) for an optimal striatal repair is also elusive (for review [[53]]). The protocol presented here may therefore be instrumental to optimize the therapeutic potential of hPSC for HD therapy and may ultimately constitute an effective way of preparing clinically relevant grafts for HD patient.


This study was supported in part by grants from the European Community's seventh Framework Program (Neurostemcell, nr.22943), the French National Agency for Research (ANR-2010-RFCS-003 “HD-SCT”), the laboratoire d'Excellence Revive (Investissement d'Avenir; ANR-10-LABX-73), and by DIM-STEM POLE Region Ile de France fellowship (to F.B.-R.). We thank Prof K. Campbell (Cincinnati Children's, Cincinnati, OH) for his generosity in providing the anti-GSX2 antibody.

Disclosure of Potential Conflicts of Interest

The authors declare no potential conflict of interest.