Zinc finger E-box–binding homeobox 1 (ZEB1) is required for neural differentiation of human embryonic stem cells

Human pluripotent stem cells hold great promise for improving regenerative medicine. However, a risk for tumor formation and difficulties in generating large amounts of subtype derivatives remain the major obstacles for clinical applications of stem cells. Here, we discovered that zinc finger E-box–binding homeobox 1 (ZEB1) is highly expressed upon differentiation of human embryonic stem cells (hESCs) into neuronal precursors. CRISPR/Cas9-mediated ZEB1 depletion did not impede neural fate commitment, but prevented hESC-derived neural precursors from differentiating into neurons, indicating that ZEB1 is required for neuronal differentiation. ZEB1 overexpression not only expedited neural differentiation and neuronal maturation, which ensured safer neural cell transplantation, but also facilitated the generation of excitatory cortical neurons, which were valuable for managing certain neurological disorders, such as Parkinson's disease (PD) and amyotrophic lateral sclerosis (ALS). Our study provides useful information on how human neural cells are generated, which may help in forming strategies for developing and improving replacement therapies for treating patients with neurological diseases.

Human pluripotent stem cells (PSCs), 4 particularly human embryonic stem cells (ESCs) or inducible pluripotent stem cells (iPSCs), because of their ability to differentiate into all cell types, become the most reasonable sources for transplantation therapy (1)(2)(3)(4). In recent years, progress has been made in direct neural differentiation of human PSCs, especially embryonic stem cells, into certain functional cell types (5)(6)(7)(8)(9), which provides an ideal resource for pathogenesis research and clinical therapies of neurodegenerative diseases, such as Parkinson's disease, amyotrophic lateral sclerosis, and other disorders in the nervous system. However, some problems hinder the application of human ESCs, such as time-consuming differentiation, low yield, heterogeneous nature of derived neuronal populations, and a risk of tumor formation upon transplantation in vivo. Therefore, further efforts are required to guarantee both efficiency of differentiation and maturation.
During the differentiation from hPSCs into neurons, hPSCs undergo substantial morphological alterations, first from adherent hESC colonies to cell aggregates, then to neural rosettes, neural stem cells, and neurons or glia, a course very similar to that occurs in epithelial-mesenchymal transition (EMT). It is highly possible that molecules related to EMT may play important roles in neural differentiation. We made a screening and found that ZEB1 was expressed along with neural differentiation.
ZEB1, also named as ZFHX1A or ␦EF1, together with its fellow member ZEB2 (also called ZFHX1B or SIP1), belongs to the zinc finger E-box binding homeobox family (ZFH family). ZEB1 is putatively involved in cell morphological change during EMT. Accordingly, ZEB1 has been extensively studied in various cancers (10 -13). Members of the ZEB family facilitate neural fate decision through tuning TGF␤/BMP signaling or as targets of the miR-200 family (14 -16) and ZEB1 has an essential role in neural stem cell survival in vitro (17). Zeb1 may be functionally redundant to Zeb2, as Zeb1 was ectopically activated in Zeb2 knockout embryos (18). However, in mouse models, Zeb1 has been shown to be strongly expressed in the progenitor cells of the ventricular zone (VZ) and the subventricular zone (SVZ) of the developing cortex and that Zeb1 is necessary for the proliferation of the VZ and SVZ progenitors (19,20). Recently, Singh et al. (21) reported that Zeb1 is required for tangential migration of unpolarized cerebellar granule neuron progenitors, and down-regulation of Zeb1 is necessary for maturation and polarization of the cerebellar granule neurons.
However, whether ZEB1 is involved in human neural differentiation still remains unclear. Here, by using a model system of neural differentiation from hESCs (22,23), we carried out a series of functional assays to explore the role of ZEB1 in human neural differentiation.

ZEB1 was progressively expressed during neural differentiation of hESCs
To inspect the expression of ZEB1 during neural differentiation, we differentiated hESC line WA09 (H9) as shown in Fig.  1A, and harvested cells at the indicated stages for real-time RT-PCR and Western blotting analysis. The mRNA level of ZEB1 was hardly detected in hESC (day 0). Upon neural differentiation, ZEB1 started expression and reached a peak on day 15, and remained at a high level thereafter (Fig. 1B). The expression pattern of ZEB1 was also verified by Western blotting and immunostaining together with cell type-specific markers at different stages of differentiation on samples taken from days 0, 15, 25, and 35 ( Fig. 1, C and D). We also detected the expression of other fellow molecules of EMT in the same set of experiments. Except for N-CADHERIN, which was up-regulated over the progression of differentiation, no other EMT-related molecules, such as ZEB2, SNAIL, SLUG, and TWIST2 were detected (Fig. S1). These results together indicate ZEB1 may play a role during neural differentiation, which was not previously documented.

ZEB1 was required for the intrinsic neuronal differentiation
To examine whether ZEB1 was required for neural and neuronal differentiation, we generated ZEB1-knockout hESCs based on a H9 hESC line that harbored a cassette of inducible Cas9 (iCas9) and a verified sgRNA targeting ZEB1 locus (sgRNA1) (Fig. S2A). We designed three sgRNAs targeting ZEB1 (Table S1), and identified the indel percentages with T7EI assays for each target in HEK293T cells (data not shown), then sgRNA1 (Table S2), which targeted the third exon (CCDS53506.1) was selected as the most efficient one. In H9ESCs, a restriction fragment length polymorphism assay revealed ϳ21% indel rate (Fig. S2B). hESCs were then dissociated into single cells. By picking colonies, using restriction fragment length polymorphism analysis and sequencing (Table S3), we ultimately obtained a hESC line showed the alteration of ZEB1 mRNA levels during neural differentiation of hESCs. The amount of mRNA was normalized to GAPDH levels. Bars represent mean Ϯ S.D. of three experimental replicates. C, Western blotting showed expression of the ZEB1 protein at different stages of neural differentiation from hESCs. ␤-ACTIN was a loading control; n ϭ 3. D, representative images of immunostaining results using indicated antibodies in hESC-derived neural cells. DAPI was included to show the nuclei (here and after); n ϭ 3. Scale bar: 50 m. hESC, human embryonic stem cell; EB, embryoid bodies; NSC, neural stem cell; E8, essential 8 medium; NIM, neural induction medium; NDM, neural differentiation medium.

ZEB1 promotes neural differentiation of hESCs
with a homozygous mutant of ZEB1 (Clone 9, Fig. S2, C and D). Potential off-target sites were identified at MIT, and only one possible off-target with four mismatches fall in gene coding sequences (NM_001130675) (Table S4), indicating the specificity of the selected sgRNA1 locus.
Deletion of ZEB1 was confirmed at the protein level at different stages during neural differentiation ( Fig. 2A). During normal differentiation of hESCs into neurons, cells spread out and formed rossettes at the early stage, and then differentiated into neuron phenotype at the late stage (Fig. 2B, upper panel). ZEB1 knockout cells did not exhibit distinguishable differences from control cells at the hESCs stage and early phase of neural differentiation (Fig. 2B, middle panel). However, with neural differentiation proceeded, most of the ZEB1-depleted cells died at the stage when neuronal differentiation occurred (Fig. 2B, middle panel). We then overexpressed ZEB1 in ZEB1 KO hESCs cells, some cells underwent morphological changes at the early stage of differentiation, and eventually became neuron cells (Fig. 2B, lower panel). We did not observe any significant difference by evaluating cell proliferation with BrdU antibody (Fig. 2C). However, we observed more cells that were undergoing apoptosis in the ZEB1-depleted group compared with the control group using a PE Annexin V/7-AAD Kit (Fig. 2D). Reexpressing of ZEB1 in ZEB1 KO cells rescued the neural differentiation (Fig. 2B, lower panel). In addition to the rescue effects, forced expression of ZEB1 in ZEB1 KO caused discernable dispersing of the differentiated cells at the early stage of neural differentiation (Fig. 2B, right lower panel), which was consistent with the phenomenon detected in ZEB1 overexpression cell lines (as discussed below). Taken together, these data indicated that ZEB1 expression may not be critical for early neural specification, but was essential for the conversion from neural precursors to neurons.

Forced expression of ZEB1 promoted the neural differentiation of hESCs
We next overexpressed ZEB1 (with 3ϫ FLAG fusion to the N terminus) in hESCs with a lentivirus tet-on 3G inducible expression system (Fig. S3A). The insertion of a mNeonGreen-P2A-blasticidin (BSD) fragment facilitated picking of single colony hESCs (Fig. S3B). Selected colonies of H9ESC lines were verified using GFP or antibodies against ZEB1 or FLAG upon doxycycline induction (Fig. S3, C and D and Fig. 3A). Forced expression of GFP itself in hESCs neither affected the morphology of the hESC colonies nor changed the expression of pluripotent marker such as OCT4 (Fig. S3C). However, when overexpressing ZEB1 in hESCs, although the expression patterns of OCT4, NANOG, and SOX2 were not changed, these colonies exhibited unique fragmentation upon long-term culture (Fig. 3B, Figs. S3D and S4). For some hESC colonies, continual overexpression of ZEB1 elicited neurite-like outgrowth, which was immunostained for the neuron-specific antibody TUJ1 (Fig. S3E).

Extra amount of ZEB1 accelerated the neural differentiation of hESCs
To further investigate the effect of ZEB1 in neural differentiation, we induced the expression of ZEB1 once the hESCs were lifted for differentiation. At day 15, most of the WT cell aggregates formed neural rosettes within colonies (Fig. 3C, left panel). However, in the colonies of hESCs with ZEB1 overexpression, only a small number of rosettes were identified. Quite a few cells migrated from the colonies and extended the neurite-like structures surrounding the cells (Fig. 3C, left panel).
Decreasing neural rosettes may reflect a result of inhibited neural differentiation or alteration of the progress of differentiation. We then stained the cells with neuron-specific and dendrites antibodies TUJ1 and MAP2 to address this issue. In the control group, only sporadic TUJ1ϩ cells were detected (ϳ2%) on day 15 (Fig. 3, C and D, p Ͻ 0.001), and the vast majority of TUJ1ϩ neurons did not appear until day 25. In the ZEB1overexpressed group, however, TUJ1ϩ cells appeared as early as day 15 (up to 30%) (Fig. 3D, p Ͻ 0.001), and the number of TUJ1ϩ cells continually increased when neural differentiation proceeded. One week after the cells were attached for neuronal differentiation, only 16% of TUJ1ϩ cells express more mature neuronal marker MAP2 in the control group, whereas approximately 32% among the TUJ1ϩ cells began to express MAP2 in the ZEB1-overexpressed group (Fig. 3, C and E, p Ͻ 0.001). Thus, we deemed that forced expression of ZEB1 promoted neural differentiation by altering the phase of neural differentiation.
The formation of neural rosettes accompanying the precocious TUJ1ϩ and MAP2ϩ cells indicated that neural differentiation was accelerated. Furthermore, we examined the expression of early neural markers such as PAX6, SOX1, as well as neuronal markers TUJ1 and MAP2 by immunofluorescence staining or Western blotting analysis. In the control group, PAX6 was expressed as the earliest transcription factor of neural differentiation from days 6 to 8 on, which was followed by SOX1 expression from days 8 to 10 on (Fig. 3, C and F). The early presence of extra ZEB1 repressed PAX6 but increased the expression of SOX1 on days 10 -15, endowing cells becoming TUJ1ϩ neuronal-like cells directly within 15 days of differentiation ( Fig. 3, C and F). Bypassing the early stage of neural differentiation was also accompanied by down-regulation of various genes, such as ZEB1 itself and PAX6, which likely ensured later smooth transition to an intrinsic neuronal differentiation program (Fig. 3F).

ZEB1 promoted neuronal maturation
Deletion of ZEB1 impeded neuronal differentiation, whereas forced ZEB1 expression expedited this process, which prompted us to examine whether ZEB1 could also promote neuronal maturation. We thus continued the neuronal differentiation experiment by placing the hESCs-derived neural progenitor cells on laminin substrate from day 26 and cultured in neural differentiation medium supplemented with trophic factors such as BDNF, GDNF, IGF1, and cAMP. Four weeks later, the majority of cells in both control and ZEB1-overexpressed groups became elaborately branched and expressed the neuronal marker MAP2 in the cell body and in the neurites (Fig. 4A). Considering that the synaptic connection between neurons was an integral component of functional maturity, we then examined the localization of synaptic proteins such as SYNAPSIN-1

ZEB1 promotes neural differentiation of hESCs
in differentiated neurons by immunofluorescent staining. Although SYNAPSIN-1 was diffusely distributed in the cytoplasm of neurons in the control group, it aggregated as punctu-ated throughout the more complex dendritic arborizations of neurons when ZEB1 was induced during the neuronal differentiation (Fig. 4B).

ZEB1 promotes neural differentiation of hESCs
To further examine whether overexpression of ZEB1 increased the functionality of hESCs-derived neurons, we employed whole cell voltage-clamp to record electrophysiological activities at week 8 of neural differentiation. Cells from both control and ZEB1-overexpressed groups displayed full-blown Na ϩ /K ϩ cur-rents, and these inward currents could be completely blocked by tetrotodoxin (TTX) (1 M) (Fig. 4C, i and ii). However, neurons from the control group failed to fire multiple action potentials (APs) in response to a 70 pA current pulse, or displayed single AP or APs with typically faded amplitude in peak value (Fig. 4D, i),

ZEB1 promotes neural differentiation of hESCs
whereas strains of sequential APs on the similar spikes were detected in neurons from the ZEB1-overexpressed group (Fig. 4D, ii). Neurons in the ZEB1-overexpressed group also showed spontaneous inward synaptic currents at 8 weeks of differentiation, which were almost completely eliminated by the non-N-methyl-D-aspartate antagonist DNQX (20 M) (Fig. 4E). Neurons from the control group did not exhibit spontaneous synaptic currents until 10 weeks of differentiation (Fig. 4F). These data further validated that ZEB1 promoted the functional maturation and synaptogenesis of the hESCs-derived neurons.

ZEB1 promotes neural differentiation of hESCs ZEB1 facilitated the differentiation of excitatory neurons
To identify which specific types of neurons or glia were differentiated when ZEB1 was overexpressed, we labeled the cells with various markers at week 8 of neural differentiation. In the control group, hESCs were randomly differentiated into excitatory, GABAergic, and dopaminergic neurons at a proportion of 52.2 (Ϯ9.0%), 30.5 (Ϯ5.5%), and 8.6% (Ϯ3.7%), according to immunostaining for vGLUT1, GABA, or TH, respectively (Fig.  5, A and B). When ZEB1 was overexpressed during neural differentiation, the majority of the cells were distinctly positive for vGLUT1 (84.9 Ϯ 10.2%) (Fig. 5, A and B, p Ͻ 0.001). Only a small proportion of cells were stained with GABA (4.8 Ϯ 7.9%) and TH (4.8 Ϯ 6.8%) (Fig. 5, A and B). In both groups, we detected a similar proportion of astrocytes (ϳ30%) (Fig. 5A) but no oligodendrocytes (data not shown), as revealed by immuno-staining for S100␤/TUJ1 and myelin basic protein (data not shown), indicating that ZEB1 preferably affected the neuronal subtype specification but not glia fate choice. These results collectively demonstrated that forced expression of ZEB1 specifically promoted hESCs to differentiate into excitatory neuron subtypes.

ZEB1 reduced the overgrowth of hESCs-derived progenitors upon transplantation into mouse brains
Because ZEB1 accelerated neural differentiation and promoted functional maturation of hESCs-derived neurons, it thus became a plausible lever for safety control of hESCs in regenerative medicine. We then examined whether overexpression of ZEB1 in hESCs-derived neural progenitor cells could reduce the chance of tumor formation upon transplanting into brains

ZEB1 promotes neural differentiation of hESCs
of neonatal pups of NOD/SCID mice. Eight weeks after transplantation of hESC-derived neural stem cells (day 15 of differentiation), mice were perfused for brain sections and immunostaining. Monoclonal antibodies that specifically recognized human nuclei were used to identify the transplanted human cells. Graft human cells were found in the cerebral cortex or nearby hippocampus and there was no difference for their distribution in the mouse brain between two groups (Fig. 6A). In the control group, transplanted cells formed substantial rosette-like overgrowth in the resident brain regions (presence in 55.2% of the brain slices) (Fig. 6, B and C). However, in the ZEB1-overexpressed group, this percentage was only 18.3% (Fig. 6C, p Ͻ 0.01), and most of the graft cells dispersed across the brain (Fig. 6, B and C). Consistently, much more of grafts differentiated into neurons, with more TUJ1ϩ cells in ZEB1 group relative to control group (Fig. 6, D and E, p Ͻ 0.05). These data indicate that ZEB1 overexpression promoted the differentiation of hESC-derived neural stem cells and potentially improved safety when used for therapy. In accordance with the in vitro results (Fig. 3, C and F), lack of rosettes in the ZEB1 group in vivo also reflected the loss of a PAX6ϩ fate in the ZEB1-overexpressed cells. For the statistical analysis, 100 -200 brain slices from 9 mice were analyzed, and the brain slices without human cells were discarded directly. **, p Ͻ 0.01; *, p Ͻ 0.05. Bars represent mean Ϯ S.D. of three experimental replicates.

ZEB1 promotes neural differentiation of hESCs ZEB1 knockout prevented neural precursors differentiate into neurons
To elucidate the mechanism of ZEB1 function during the differentiation from hESCs into neurons, we collected RNA samples of different stages and examined genome-wide changes during the process by performing a RNA sequence analysis. Both clustering and principle component analysis indicated that the genome expression pattern occurred with significant changes since day 25 differentiation, as the 25-day samples' genome expression pattern were more close to 30 and 35 days, far from 15-and 10-day samples (Fig. S5, A and B). When we turned to the different expression genes (DEGs) at different stages, we found more DEGs starting at day 25 (Fig.  7B). Meanwhile, the genome expression pattern of the day 25 WT could be easily clustered with days 30 and 35 WT, and the ZEB1 KO cell lines clustered together (Fig. 7A). To further validate the ZEB1 functioning stage, we picked some of the neuronal related genes and found there was no significant difference in expression levels between WT and KO cell lines at the early differentiation stage (Fig. 7C), which were then confirmed by real-time PCR (Fig. 7D). As 25 day seemed to be a coteau, we subjected DEGs at 25 days into gene ontology (GO) analysis, and found most of the genes accumulated to neuronal related items (Fig. 7E), most of these genes were down-regulated after ZEB1 KO, such as FGF8, NKX2.1, SIX3, and DLX1 (Fig. 7F). Besides, we also observed some DEGs subjected to cell adhesion items in GO analysis, and found most of them were downregulated in the ZEB1 KO cell line (Fig. S5C).

Discussion
In human neural differentiation, ZEB transcription factors, particularly ZEB2, as direct targets of miR-200 family members play essential roles in fate specification of the neuroectoderm (14). In this study, we found that when ZEB1 was depleted, although these cells could commence an early neural differentiation, they failed to differentiate further into neurons and died within a few days. Based on our data and previous literature, we speculated that during neural differentiation of hESCs, two members of ZEB family, ZEB1 and ZEB2, synergistically regulated the progression of cell fate conversion, with ZEB2 functioning primarily at the early stage of neural specification, whereas ZEB1 exclusively affecting the neuronal conversion during late neural differentiation.
Although not necessary for early neural specification, forced expression of ZEB1 during the early stage did expedite the progress of neuronal conversion, as evidenced by positive staining of a series of markers of neural differentiation and electrophysiological recordings. Overexpression of ZEB1 in hESCs significantly inhibited the expression of PAX6 but increased the expression of TUJ1, a conventional neuron marker, thus prompting a comprehensive revision of cell fate determination differed from the original opinion, which believed PAX6 pioneered a cohort of neural determinants (24). The expression of SOX1, another neural fate determinant that was usually expressed following PAX6 during neural differentiation of hESCs (24,25), however, did not shift to earlier expression, suggesting the temporary alteration of cell fate did not com-pletely perplex the intrinsic program of neural differentiation. Interestingly, ZEB1 was not detected at the protein level in the following days once it was overexpressed in hESCs at the initiation of neural differentiation. This might be due to an instigation of a quality control system within cells in response to the forced expression of ZEB1 and the sudden shuffle of cell fate. In both groups, ZEB1 was expressed at high levels from days 10 to 15 on, indicating its endogenous expression and intrinsic function. These data collectively supported a notion that cells undergoing cell fate rearrangement were able to gear to the intrinsic neural differentiation program when endogenous ZEB1 start to take over.
ZEB1 may play an important role in regulating the differentiation of ESCs into neurons, particularly in the transition from NPCs to neurons, as ZEB1 KO made no significant difference in the expression of neuronal related genes at the early stage, but after 25 days of differentiation, we detected the remarkable decrease for genes that associated with neuronal differentiation (FGF8, NKX2.1, SIX3, and DLX1). ZEB1 KO also resulted in down-regulation of cell adhesion-related genes, indicating a low adherence ability of ZEB1 KO cells, which may be another reason that blocks NPC cells differentiating into neurons in ZEB1 KO cell line.
We noticed that ZEB1 overexpression promoted neuronal maturation for over 2 weeks in advance, as compared with regular neural differentiation. As it is known that astrocytes promote neuronal maturation both in vivo and in vitro (26,27), we doubt whether ZEB1 accelerates neuronal maturation through an indirect regulation of astrocytes. Whether ZEB1 was overexpressed or not, a similar proportion of S100␤ϩ cells were identified at 8 weeks of neural differentiation. Thus, the presence of the astrocytic lineage in the culture unlikely caused a significant difference of the spontaneous synaptic currents of the hESCs-derived neurons, and the extra ZEB1 expression during neural differentiation autonomously accelerated neuronal maturation and promoted the occurrence of synaptic currents.
ZEB1 may also play a role in neuronal fate determination, as forced expression of ZEB1 almost exclusively induced hESCs into vGLUT1 ϩ glutamatergic excitatory neurons in which ZEB1 may directly interact with key transcription factors involving excitatory fate. Nevertheless, it is also highly possible that the acceleration of neuronal differentiation itself accounts for the augmented generation of excitatory neurons. As noted above, forced expression of ZEB1 granted the neural stem cells for directed neuronal conversion and bypassed certain stages of neural differentiation. This may let neural stem cells miss competence to certain morphogens, such as Sonic Hedgehog (SHH), which is critical for ventral patterning and GABAergic neuron differentiation (22, 28 -30). Thus, ZEB1 appears to be an essential player that ensures appropriate cellular constitution of brains for proper function.
Transplantation of ESC-derived cells even at the late stage of differentiation still bear risks of overgrowth in brains (31). Expedited neuronal differentiation is valuable for efficient differentiation into functional cells as well as for transplantation safety. In a set of proof of concept for neural cell transplantation, we did observe aggregates of cell overgrowth in brains that received transplantation of regular hESC-derived ZEB1 promotes neural differentiation of hESCs neural progenitors, which is in accordance with the previous report (31). Nevertheless, mice transplanted with hESCsderived ZEB1 overexpression neural progenies contained much fewer aggregates, suggesting increased safety from overgrowth with ZEB1 overexpression. Thus, ZEB1 and the relevant signaling pathway could be informative to optimize the procedures of differentiating hESCs into neural cells for therapy.
Overall, we identified an important regulator of neural differentiation that coupled the neural differentiation progression and fate choice. However, the molecular mechanism under it still remains elusive and needs further research.

Ethics statement, animals
All mouse studies were approved by the Ethics Committee of the Institute of Zoology (research license number AEI-06-07-2014), Chines Academy of Sciences. NOD/SCID mice were purchased from Vital River Laboratory Animal Technology Corp. (Beijing, China) and then raised in a sterile room.

ZEB1 promotes neural differentiation of hESCs Culture and differentiation of human ESCs
The H9ESC line (WiCell Institute, Madison, WI, passages 29 -50) was cultured with feeder-free media. Neural differentiation from hESCs was performed as described previously (9). See the Supplemental Experimental Procedures for additional details.

Generation of the ZEB1-knockout hESC line using iCRISPR
To generate loss-of-function mutations of ZEB1 in human ESCs, we first constructed the CRISPR/Cas9-inducible (iCas9) hESC line by inserting a CRISPR/Cas9-inducible expression cassette in the AAVS1 genomic locus with TALEN technology (32,33). For gRNA transfection and transduction in hESCs, we followed the published protocols (34,35). In brief, iCas9 hESCs were treated with doxycycline (2 mg/ml) for 1 or 2 days before and during transfection. Cells were dissociated into single cells using Accutase and transfected the sgRNA using Lipofectamine RNAiMAX (Life Technologies) following the manufacturer's instructions. The details were described in the supporting "Experimental procedures".

Construction of the ZEB1 overexpression hES cell line
To overexpress ZEB1 in H9ESCs and ZEB1-depleted H9ESCs, the tetracycline inducible tet-on lentiviral expression vectors were used. For the plasmid construction please see the supporting "Experimental procedures" Preparation of lentiviruses and transduction were conducted as previously described (24). hESCs were treated with ROCK inhibitor (1 M, Calbiochem, 688001) for 1 day before transduction. Cells were dissociated with 0.5 mM EDTA on the following day (UltraPure TM , 15575-020, Gibco). Cell pellets were incubated with concentrated virus at 37°C for 1 h, then replated onto Matrigel substrate overnight with E8 medium. Two days later, neomycin (50 g/ml, Sigma, G8168) and blasticidin (2 g/ml, Sigma, 15205) were added to the culture medium to select drug-resistant clones. To obtain stable transduced monoclonal lines, hESCs were dissociated into single cells with Accutase and the mNeonGreen-positive cells were sorted by FACS. hESCs were treated with the ROCK inhibitor (1 M) for all the processes. After 1-10 days of cultivation, mNeonGreen-positive colonies were identified and picked.

Real-time quantitative RT-PCR
Total RNA was isolated manually with TRIzol reagent (Invitrogen, 15596-026). cDNA was synthesized using 1 g of RNA using a ReverAid First Strand cDNA Synthesis Kit according to the manufacturer's instruction (Fermentas, K1662). Quantitative RT-PCR was conducted with SYBR Premix ExTaq (Takara, RR430A) and the primers were listed in Table S5.

Proliferation assay
BrdU labeling of proliferating cells was performed according to previously published protocols (36). Differentiated cells were treated with BrdU for 24 h and analyzed using a Cell Proliferation Assay kit (Calbiochem, K306) according to the manufacturer's instructions. Results were analyzed with a CytoFLEX Flow cytometry (Beckman Coulter, Suzhou, China).

Apoptosis assay
Apoptosis analysis was performed using the Annexin V-PE/ 7-ADD Apoptosis Detection Kit (BD Pharmingen 559763) according to the manufacturer's instructions.

Electrophysiology
Whole cell patch clamp recordings of hESCs-derived neurons that were differentiated for 8 -10 weeks were performed as described previously (26). The transmitter receptor blockers, TTX (1 M), 4-aminopyridine (1 mM), bicuculline (20 M), and DNQX (20 M) were used in the bath solution for the detection of action potentials and spontaneous excitatory postsynaptic currents. Data were analyzed using pClamp version 9.2 software (Molecular Devices).

Transplantation of hESCs-derived cells in vivo
Differentiated cells (1 ϫ 10 5 ) from control or ZEB1 groups were transplanted into the lateral ventricles of the brain of P1-2 pups (ice anesthetized), respectively. For the ZEB1 group, 2 mg/ml of doxycycline was administered orally in the drinking water to induce the expression of ZEB1. The mice were perfused (anesthetized with 40 mg/kg of chloral hydrate) with 0.9% saline followed by 4% paraformaldehyde 10 weeks post-transplantation. Consecutive coronal brain sections (30 m) were sliced using a Leica SM 2000R Sliding Microtome and fixed in 4% paraformaldehyde overnight followed by dehydration in 0.1

ZEB1 promotes neural differentiation of hESCs
M PBS containing 30% sucrose for 2 days at 4°C. Immunofluorescence staining for the brain sections were performed as protocols used for cultured cells except using 1% Triton-100 to permeabilize.

Statistical analysis
Statistical analyses were executed with two-tailed Student's t tests. Data are presented as the mean Ϯ S.D., and values were considered statistically significant at p Ͻ 0.05 (*), p Ͻ 0.01 (**), and p Ͻ 0.001 (***), as described in the figure legends. Diagrams were created by GraphPad Prism 5 software.

RNA-seq analysis
For RNA-seq, cells at the indicated time points during differentiation were harvested, and total RNA was extracted with TRIzol reagent (1559618, Ambion, USA). After testing the quality of RNA by an Agilent 2100, 0.2 g of each sample was sent to Annoroad Genomics for RNA-seq analysis. Raw data and processed data were uploaded to the NCBI Gene Expression Omnibus database under accession number GSE118990. RPKMs at the current sequencing depth were evaluated by using the jackets of the total comparison reads in RSeQC, and a relative error rate was utilized to measure the accuracy of the evaluated RPKM. Unsupervised heat map clustering analysis was performed by using Cluster 3.0. The clustering results were visualized and exported by TreeView 1.1.6r4. A principle component analysis image was generated by R(3.3.3)/Bioconductor (3.4) with "edgeR" and "limma" packages. Kyoto encyclopedia of genes and genomes (KEGG) and GO enrichment analyses were executed by R/Bioconductor with the "clusterProfiler" package.