Down-regulation of ATF1 leads to early neuroectoderm differentiation of human embryonic stem cells by increasing the expression level of SOX2

Abstract

We reveal by high-throughput screening that activating transcription factor 1 (ATF1) is a novel pluripotent regulator in human embryonic stem cells (hESCs). The knockdown of ATF1 expression significantly up-regulated neuroectoderm (NE) genes but not mesoderm, endoderm, and trophectoderm genes. Of note, down-regulation or knockout of ATF1 with short hairpin RNA (shRNA), small interfering RNA (siRNA), or clustered regularly interspaced short palindromic repeats (CRISPR)/CRISPR-associated protein 9 (Cas9) was sufficient to up-regulate sex-determining region Y-box (SOX)2 and paired box 6 (PAX6) expression under the undifferentiated or differentiated conditions, whereas overexpression of ATF1 suppressed NE differentiation. Endogenous ATF1 was spontaneously down-regulated after d 1–3 of neural induction. By double-knockdown experiments, up-regulation of SOX2 was critical for the increase of PAX6 and SOX1 expression in shRNA targeting Atf1 hESCs. Using the luciferase reporter assay, we identified ATF1 as a negative transcriptional regulator of Sox2 gene expression. A novel function of ATF1 was discovered, and these findings contribute to a broader understanding of the very first steps in regulating NE differentiation in hESCs.—Yang, S.-C., Liu, J.-J., Wang, C.-K., Lin, Y.-T., Tsai, S.-Y., Chen, W.-J., Huang, W.-K., Tu, P.-W. A., Lin, Y.-C., Chang, C.-F., Cheng, C.-L., Lin, H., Lai, C.-Y., Lin, C.-Y., Lee, Y.-H., Chiu, Y.-C., Hsu, C.-C., Hsu, S.-C., Hsiao, M., Schuyler, S. C., Lu, F. L., Lu, J. Down-regulation of ATF1 leads to early neuroectoderm differentiation of human embryonic stem cells by increasing the expression level of SOX2.

Pluripotent stem cells (PSCs) are characterized by their ability to proliferate almost indefinitely (self-renewal) and differentiate into all cell types except placenta (pluripotency) (1, 2). A type of PSC, human embryonic stem cells (ESCs) (hESCs) are derived from the inner cell mass of blastocysts and initiated a new era in stem cell research and regenerative medicine of human cells (3, 4). Important ESC-like cells, induced PSCs (iPSCs) are reprogrammed from somatic cells via introducing octamer-binding transcription factor 4 (OCT4), sex-determining region Y-box (SOX)2, MYC proto-oncogene (c-MYC), and Kruppel-like factor 4 [or OCT4, SOX2, nanog homeobox (NANOG), and Lin-28 homolog A (LIN28)] and heralded as an unlimited resource for drug screening and disease models (5–8).

The master regulator for maintenance of pluripotency is the regulatory circuit of OCT4, NANOG, and SOX2 (9, 10). These factors regulate not only genes promoting self-renewal and pluripotency in PSCs but also developmental regulators that express during differentiation. The balance of OCT4, NANOG, and SOX2 maintains pluripotency (11, 12), whereas the imbalance of these regulators is sufficient to trigger differentiation of PSCs. OCT4 and NANOG up-regulation induces mesendoderm (ME) and represses neuroectoderm (NE) differentiation (13, 14). Conversely, SOX2 overexpression is sufficient to trigger NE differentiation and inhibit ME fate (14, 15). Recent studies also found some integral components of the stemness regulatory circuit that are important to stabilize the equilibrium (16–19). These studies give rise to a concept of how maintaining the equilibrium is an important issue for the stemness ability in PSCs.

SOX2, a high mobility group box transcription factor, is expressed throughout mouse embryo development in neural progenitors of the CNS (20, 21). In another paper, homozygous knockout of Sox2 in mice disrupted primitive ectoderm formation (22). These studies indicate that SOX2 plays a critical role in pluripotency maintenance and generation of early embryo ectoderm precursor cells during development (23). However, until now, the control mechanisms used to fine-tune SOX2 expression levels from pluripotency to NE determination have not been revealed.

High-throughput screening in PSCs is an efficient way to demonstrate numerous potential candidates that may be involved in stemness and differentiation regulation (24–27). In this study, we engaged in a short hairpin RNA (shRNA) functional screen to identify candidate genes that may be involved in stemness maintenance. We uncovered a novel pluripotent regulator, activating transcription factor 1 (ATF1), which is a basic region-leucine zipper transcription factor that belongs to the cAMP response element-binding protein family. In the early development of mouse embryos, the active form of ATF1 is accumulated in a 2-cell embryo state (28), and double knockout of Atf1 and cAMP response element-binding protein genes led to embryo lethality before embryonic d 4.5 in the mouse model (29). These studies imply that ATF1 may play a crucial role in the preimplantation stage. Our findings are the first demonstration that the knockdown of ATF1 will promote the expression levels of SOX2 in hESCs. Furthermore, the down-regulation of ATF1 promoted NE differentiation, and the overexpression of ATF1 suppressed NE induction. According to the results of the reporter assay, we define ATF1 as a repressor of SOX2. Our observations imply that ATF1 may be significantly associated with stemness equilibrium in hESCs. Overall, we reveal a novel regulatory component, ATF1, that functions as a gatekeeper for neural lineage specification in hESCs and shed light on the very first step from the pluripotent state toward NE lineage.

MATERIALS AND METHODS

All methods followed the relevant guidelines and regulations. All experiments were approved by Human Subject Research Ethics, Academia Sinica (AS-IRB01-14019).

Cell lines and culture conditions

The H9 hESC line was obtained from WiCells (Madison, WI, USA) (3). Another hESC line, HUES-6, was kindly provided by Dr. Douglas A. Melton (Harvard University, Boston, MA, USA) (30). With C57BL6 mouse feeder fibroblasts, hESCs were cultured in DMEM/F12 supplemented with 20% knockout serum replacement, 1% nonessential amino acids, 2 mM l-glutamine, 0.1 mM 2-mercaptoethanol, and 8 ng/ml human basic fibroblast growth factor. Under feeder-free culture conditions, hESCs were seeded on culture plates coated with Matrigel Matrix (Corning, Corning, NY, USA), and the cells were cultured with the undifferentiation medium [conditional medium of mouse embryonic fibroblasts (MEFs)]. HEK293T cells were cultured in DMEM supplemented with 10% fetal bovine serum (Biological Industries, Cromwell, CT, USA). All cells were cultured in a 37°C incubator with a humidified atmosphere containing 5% CO₂.

shRNA high-throughput screening

After microarray analyses of 220 Human Genome U133 Plus 2.0 arrays (Thermo Fisher Scientific) (Supplemental Table S1 and result in Supplemental Table S2) and the Mouse Genome Informatics (MGI) database (http://www.informatics.jax.org/) (Supplemental Table S3), 517 shRNAs targeting 121 genes were picked by the National RNA interference (RNAi) Core Facility (Taipei, Taiwan). Lentiviruses were prepared, and hESCs were infected with a lentivirus (multiplicity of infection = 3) by using a liquid handling system. The alamarBlue (AB) (Biotium, Fremont, CA, USA) and alkaline phosphatase (ALP) assays were sequentially performed on d 6 after virus infection. In brief, 1/10 volume of AB was added to each well. The cells were then incubated overnight at 37°C. The optical density absorbances at 570 and 600 nm were measured and calculated. Cells were then fixed with 4% formaldehyde for 10 min and washed with PBS 3 times, and then 70 μl p-nitrophenyl phosphate liquid substrate was added (N-7653; MilliporeSigma, Burlington, MA, USA). The cells were incubated for 3–5 min at room temperature, and then the ALP activity was examined by measuring the absorption at optical density 405 nm. The image-based ALP staining assay was detected by an ALP Detection Kit (MilliporeSigma).

Plasmids

Control shRNA [shRNA targeting red fluorescent protein (shRFP): TRCN0000072205], shRNA targeting Atf1 (shATF1: TRCN0000273833) with the same target sequence which was used in another study (31), and shRNA targeting Sox2 (shSOX2-1: TRCN000003252 and shSOX2-2: TRCN0000355638) were obtained from the National RNAi Core Facility. The ATF1 overexpression plasmid was cloned by amplifying the ATF1 cDNA from the pCMVATF1 wild-type plasmid (32) (A kind gift from Dr. Yoshiaki Tsuji, North Carolina State University, Raleigh, NC, USA) and subcloned into pLKO AS3w.bsd (National RNAi Core Facility). The shATF1-resistant cDNA clone (mutant at +1094 G to A and +1097 G to A) was also subcloned into pLKO AS3w.bsd. The Sox2 upstream promoter region (−700 to −1) was amplified from the pGL3-SOX2 promoter plasmid (33) (A kind gift from Dr. Cheng-Wen Wu, Academia Sinica) and cloned into the pGL4.20-basic (Promega, Madison, WI, USA) luciferase reporter plasmid. For the clustered regularly interspaced short palindromic repeats (CRISPR)/CRISPR-associated protein 9 (Cas9) gene editing clone, single guide RNAs (GCTATGCTGTCGGATGAGTCC) were selected by an online CRISPR design program (https://zlab.bio/guide-design-resources) and ligated into pX330-U6-Chimeric_BB-CBh-hSpCas9 (42230; Addgene, Watertown, MA, USA).

Lentivirus production and hESC infection

To generate lentiviruses, 293T cells were seeded with 8 × 10⁵ cells per well (6-well plate). The next day, the cells were transfected with 1 μg of the lentivirus vector that contained shRNA or cDNA (Supplemental Tables S2 and S3), 0.9 μg of pCMVR8.91, and 0.1 μg of pMD.G (National RNAi Core Facility) with the Turbofect transfection reagent (Thermo Fisher Scientific, Waltham, MA, USA). At 24 h, the medium was changed to DMEM with 10% fetal bovine serum and 1% bovine serum albumin harvest medium. At 48 h or 72 h after transfection, the supernatants were harvested. For hESC infection, 1.5 × 10⁵ hESCs per well were seeded (6-well plates) and later incubated with supernatants containing lentivirus (lentivirus multiplicity of infection = 10). hESCs were cultured with an MEF-conditioned medium for 16 h with protamine sulfate (8 μg/ml, P3369; MilliporeSigma). Next, hESCs were selected with 2 μg/ml puromycin (P8833; MilliporeSigma) if using shRNA clones and selected with 5 μg/ml of blasticidin (R210-01; Thermo Fisher Scientific) if using cDNA-overexpression clones.

Flow cytometry for cell cycle analysis

For propidium iodide (PI) DNA assays, the cells were fixed with 70% of ethanol for 30 min at 4°C. After washing 3 times with PBS, the cells were stained with PI solution (10 µg/ml) that contain RNAase (5 μg/ml) for 15 min. All stained cells were analyzed by FACSCanto (Becton Dickinson, Franklin Lakes, NJ, USA), and data were processed and calculated by FACSDiva software (BD Biosciences, San Jose, CA, USA) and FlowJo (BD Biosciences).

Knockout Atf1 by CRISPR/Cas9

hESCs were dissociated into single cells by using Accutase (Innovative Cell Technologies, San Diego, CA, USA) at 37°C after 5 to 7 min-incubation and treated with 10 μM Rho-associated protein kinase inhibitor Y-27632 (Y-5301; LC Laboratories, Woburn, MA, USA). On the next day, TransIT-LT1 transfection reagent (MIR2300; Mirus Bio, Madison, WI, USA) was used for CRISPR/Cas9 plasmid delivery according to the manufacturer’s guidelines. In brief, 8 × 10⁵ cells (in 6-well plates) were transfected with 2 μg of sgRNA-Cas9-2A-PAC plasmid and 0.4 μg of pmaxGFP control vector (Lonza, Basel, Switzerland). Green fluorescent protein (GFP)–positive hESCs were sorted after 24–48 h of transfection by using a BD FACSAria III (BD Biosciences; Academia Sinica FACS Core Facility). Cells were sorted into 96-well plates containing 200 μl Stemflex (Thermo Fisher Scientific) and RevitaCell medium (Thermo Fisher Scientific). After 2 d of recovery from flow cytometry, GFP-positive hESCs were switched to E8 medium (Thermo Fisher Scientific) and fed every other day until ∼1–2 wk, when distinct colonies were established.

Differentiation of shRFP/shATF1-expressing hESCs into neurons

The neuron induction protocol was performed based on a previous study by Wen et al. (34). In brief, shRFP/shATF1 was expressed hESCs for 3 d, detached by 1-mg/ml collagenase IV (dissolved in DMEM/F12) (Thermo Fisher Scientific) treatment for 50–70 min, and resuspended in embryoid body (EB) medium consisting of 2 μM dorsomorphin (3093; Tocris, Bristol, United Kingdom) and 2 μM A-83-01 (2939; Tocris) in basic fibroblast growth factor–free hESC culture medium in nontreated polystyrene plates. After 7 d (EB medium changes daily), EBs were reattached on Matrigel-coated 6-well plates, and the EB medium was replaced with neural progenitor cell (NPC) medium consisting of DMEM/F12: neurobasal = 1:1, 1% B27, 1% N₂, 1% nonessential amino acid, 1% GlutaMax, 2 μM cyclopamine (C-8700; LC Laboratories) and 2 μg/ml heparin (H3149; MilliporeSigma). The attached EBs were kept for 14 d with NPC medium, which was changed every other day. On d 22, the NPCs were picked mechanically and then resuspended on low-attachment plates (Corning, Corning, NY, USA) in NPC medium. For neuronal differentiation, suspended neural progenitor spheres were treated with Accutase for 10 min and were placed onto poly-d-lysine (P6407; MilliporeSigma)–coated coverslips in neuronal culture medium, which consists of neurobasal medium supplemented with 1% GlutaMax, 1% B27, 10 ng/ml brain-derived neurotrophic factor (450-02), and 10 ng/ml glial cell-derived neurotrophic factor (450-10) (PeproTech, Rocky Hill, NJ, USA). The medium was changed weekly during the entire culture process.

Electrophysiology studies

After 32 d of neural induction, whole-cell patch-clamp recordings were obtained from visually identified neuronal-like cells. Cells were placed in room-temperature artificial cerebrospinal fluid (119 mM NaCl, 2.5 mM KCl, 26.2 mM NaHCO₃, 1 mM NaH₂PO₄, 1.3 mM MgSO₄, 11 mM glucose, and 2.5 CaCl₂), which was oxygenated with 95% O₂ and 5% CO₂ for 30 min. The internal solution (artificial cerebrospinal fluid) was filled in glass pipettes, and whole-cell patch-clamp recordings were performed. The membrane potentials were held at −70 mV by the voltage clamp. The recorded action potentials and currents were analyzed with Signal v.4.08 and Spike2 v.7.05b (Cambridge Electronic Design, Cambridge, United Kingdom) software.

Mesoderm and endoderm differentiation

For hESC H9 cells to differentiate into mesoderm and endoderm lineages, 1.5 × 10⁵ cells were seeded on Matrigel-coated culture plates overnight. Differentiation conditions were followed according to a previous study (35). To generate mesoderm cells, cells were differentiated in E6 medium (Thermo Fisher Scientific) containing 6 μM CHIR99021 (4423; Tocris) for 4 d. To generate endoderm cells, cells were cultured in E6 medium containing 6 μM CHIR99021 and 100 ng/ml activin A (120-14; PeproTech) for 1 d. Then, cells were switched into E6 medium containing 100 ng/ml activin A for further 3 d.

Primitive streak-like cell differentiation

The primitive streak differentiation protocol was carried out as described in a previous report (36). To generate primitive streak-like cells, 1 × 10⁵ hESCs were seeded on 12-well plates. After 24 h of culture, cells were treated with 50 ng/ml bone morphogenetic protein 4 (BMP4) (120-05; PeproTech) and 50 ng/ml activin A in E5/E6 basal medium for 48 h.

RNA extraction and real-time quantitative RT-PCR

Total RNA was isolated by using the RNeasy Micro Kit (Qiagen, Hilden, Germany) or Trizol LS Reagent (Thermo Fisher Scientific) according to the manufacturer’s instructions. RNA was treated with DNase I to remove contaminated DNA (Promega) and reverse transcribed using Superscript III (Thermo Fisher Scientific). The resulting cDNAs (80 ng/sample) were used as templates for real-time quantitative RT-PCR (qRT-PCR), which was performed using the SYBR Green 2X Master Mix (Kapa Biosystems, Wilmington, MA, USA). The relative cDNA amounts were measured and quantified using the ABI 7900 Real-Time PCR System (Thermo Fisher Scientific). The relative amounts of target genes were normalized against the RNA levels of glyceraldehyde 3-phosphate dehydrogenase (Gapdh). All of the primer sequences are listed in Supplemental Table S6.

Western blot analysis

Western blot analyses were performed as previously described by Wang et al. (37). In brief, 30 μg of proteins were blotted onto PVDF membranes, and then the membranes were incubated with primary antibodies, including anti–β-actin (A5441; MilliporeSigma), anti-SOX2 (2748; Cell Signaling Technology, Danvers, MA, USA), anti-ATF1 (ab134104; Abcam, Cambridge, United Kingdom), anti-SOX1 (AF3369; R&D systems, Minneapolis, MN, USA), anti–paired box 6 (PAX6) (GTX113241; GeneTex, Irvine, CA, USA) and anti-OCT4 (2890; Cell Signaling Technology). Next, the membranes were blotted with secondary antibodies conjugated with horseradish peroxidase. The blots were developed with a chemiluminescent substrate (WBKLS0500; MilliporeSigma). The bands were imaged by the LAS-4000 system (Fujifilm, Tokyo, Japan).

Immunofluorescence assays

Immunofluorescence assays were performed as previously described by Wang et al. (37). In brief, cells were washed with PBS, fixed with 4% formaldehyde solution in PBS for 10 min, and permeabilized by 0.3% Triton X-100 for 10 min. Next, the cells were stained with anti-SOX2 (GTX101507, GTX627405; GeneTex), anti-PAX6 (12323-1-AP; Proteintech, Rosemont, IL, USA), and anti-SOX1 (AF3369; R&D Systems), anti-TUJI (tubulin beta 3 class III) (801202; BioLegend, San Diego, CA, USA), anti–microtubule-associated protein 2 (17490-1-AP; Proteintech), anti-GATA4 (sc-25310; Santa Cruz Biotechnology, Dallas, TX, USA), anti-GATA6 (5851; Cell Signaling Technology), anti-eomesodermin (MAB6166; R&D Systems), and anti-brachyury (sc-20109; Santa Cruz Biotechnology) antibodies in the presence of 2% bovine serum albumin in PBS overnight at 4°C. After washing with PBS, the cells were incubated with Alexa Fluor 555/488 anti-rabbit IgG, anti-mouse IgG, or anti-goat IgG (Thermo Fisher Scientific) for 1 h at room temperature. DAPI dihydrochloride (0.5 μg/ml) (D9542; MilliporeSigma) was used to stain the nuclei. The fluorescence intensity of each image was analyzed by image analysis software (Image-Pro plus v.4.5; Media Cybernetics, Rockville, MD, USA). Each immunofluorescence image quantified >300 cells per field.

Chromatin immunoprecipitation

H9 hESCs were treated with 1% (v/v) formaldehyde for 12 min on ice, and the cross-linking reaction was stopped by 125 mM glycine for 5 min. Fixed cells were suspended in lysis buffer (1 × 10⁷ cells/500 μl), which contained 50 mM Tris-HCl at pH 8.0, 10 mM EDTA, 1% SDS, freshly prepared protease inhibitor cocktail (Roche, Basel, Switzerland), 1 mM PMSF, and 20 mM sodium butyrate. Sonication was then performed to shear the major DNA fragment size to around 300 bp. Immunoprecipitation reaction containing fragmented DNA (1 × 10⁶ cells), precleaned Dynabeads Protein G (Thermo Fisher Scientific), and 10 μg antibody (anti-ATF1) was executed. After washing with RIPA buffer (10 mM Tris-HCl at pH 7.5, 1 mM EDTA, 0.5 mM EGTA, 1% Triton X-100, 0.1% SDS, 0.1% Na-deoxycholate, and 140 mM NaCl), DNA was eluted by elution buffer (20 mM Tris-HCl at pH 7.5, 5 mM EDTA, 20 mM sodium butyrate, 50 mM NaCl, 1% SDS, and 50 μg/ml protease K) at 65°C for 60 min. DNA was purified by the QiaQuick PCR Purification Kit (Qiagen). The ATF1 putative binding region was verified by PCR [primers: pSox2 (promoter of Sox2)-(−228): 5′-AGAGGGGATACAAAGGTTTCTCAGT-3′; pSox2-(+19): 5′-AGTTAATAGACAACCATCCATGTGAC-3′; pSox2-(−289): 5′-TAGTCTTAGTGCTGTTTACCCACTTC-3′; pSox2-(−13): 5′-CTGTCAGGGAATAAATGGGTTTCTA-3′; pSox2-(−233): 5′-CTGCGAGAGGGGATACAAAG-3′; and pSox2-(−88): 5′-CGGGTTTTGCATGAAAGG-3′]. Nonbinding control was verified by using Gapdh promoter forward primer 5′-TACTAGCGGTTTTACGGGCG-3′ and Gapdh promoter reverse primer: 5′-TCGAACAGGAGGAGCAGAGAGCGA-3′.

Luciferase reporter assays

In ATF1-overexpression assays, H9 hESCs (1 × 10⁵ cells per well of a 6-well plate) were cotransfected with the pLKO_AS3w-GFP control or pLKO_AS3w-ATF1 wild-type plasmids (1 μg/well), pRL-TK (100 ng/well), and pGL4.20- pSox2 (852, 452, and 221 bp) (1 μg/well) by TransIT-LT1 Transfection Reagent. All Transfected cells were selected by blasticidin (10 μg/ml) the next day and harvested 48 h later. A luciferase activity assay was performed by using the Dual-Glo Luciferase Assay System (Promega) following the manufacturer’s instructions. Victor3 luminometer (PerkinElmer, Waltham, MA, USA) was used to measure the luminescent signal of luciferase and Renilla. The luciferase activities of each sample were normalized by Renilla luciferase.

Statistical analysis

All of the statistical data used ≥3 biologic replicates performed independently. The data are reported as the mean values ± sd. The significance of the differences was assessed by an unpaired Student’s t test.

RESULTS

Identification of pluripotency-related genes by RNAi screening in hESCs

To pinpoint novel pluripotent regulatory genes of hESCs, we performed an shRNA functional screen. We selected candidate genes from 2 groups. First, we executed differential gene expression analyses from 220 Human Genome U133 Plus 2.0 microarray data (Fig. 1A). By comparing 70 arrays of hESCs and iPSCs with 150 arrays of differentiated cells (oocytes, EBs, fibroblasts, epithelial cells, and human tissues) (Supplemental Table S1), 61 genes increased >5-fold in both hESCs and iPSCs (Fig. 1B). Another 15 genes were up-regulated >5-fold in either hESCs or iPSCs (Fig. 1B). There were also 6 genes with a 2- to 5-fold increase in both hESCs and iPSCs (Fig. 1B). According to differential gene expression analyses from the microarray, 82 genes were selected and all were expressed at a level 2-fold higher in the pluripotent cells (Fig. 1B, C and Supplemental Table S2). Of note, Oct4, Sox2, and Nanog, which are established pluripotent cell genes, were found but excluded from the hit list.

Figure 1.

Strategy to identify critical regulators for hESC renewal by a functional shRNA screen. A) Flowchart of the shRNA functional screen. The ratio of ALP activity to AB assay was used to monitor the degree of pluripotency, and AB assays measure the relative cell number. B) Selection of candidate genes modulating ESC and iPSC renewal by microarray profiles. C) By analyzing 220 microarray data of hESCs, iPSCs, and differentiated cells, 82 genes were significantly enriched in hESCs, and iPSCs were selected (red color). D) Selections of candidate genes by mouse lethal phenotype. Embryonic lethal genes were selected by screening information obtained from the MGI database, where the 39 genes that lead to embryonic lethality at an early stage upon knockout were selected. E, embryonic day; HTP, high throughput.

Second, we selected genes from the MGI database that induced embryonic lethality in mice before embryonic d 4.5. We chose these genes because they may play an important role in early development in ESCs. To select potential candidate genes, we only included genes with human homologs and unknown pluripotency-associated functions and those that displayed high expression levels in the Human Genome U133 Plus 2.0 Arrays (Thermo Fisher Scientific). Finally, 39 genes were chosen from the MGI database mining (Fig. 1D and Supplemental Table S3). After combining the candidate gene lists from the 2 different ways of analysis, 1 gene (chromatin assembly factor 1 subunit A) was found in both sets of genes selected from the microarrays and MGI database analyses. In total, 121 genes were selected to perform further validation experiments (Supplemental Tables S2 and S3).

To determine whether the selected genes were essential for maintaining ESC pluripotency, we used an shRNA functional screen to knock down the candidate genes (121 candidate genes were targeted by 570 shRNA lentiviruses with ∼4–5 shRNAs designed against each gene). We examined the expression levels of the undifferentiation markers of ESCs by measuring ALP activity and detected the relative cell number by the AB assay to broaden the potential candidate genes that are not associated with a particular gene, like Oct4 only (Fig. 2A). The ALP activity was normalized to the relative cell number (AB activity) to exclude the effects of different cell numbers. Candidate genes were defined as having similar or higher knockdown efficiency from ALP:AB or AB activity. In all, we identified 21 genes that might be important for pluripotency based on a decrease in ALP activity (Supplemental Table S4). However, there were 78 genes that might be important for cell expansion because the AB activity was observed to be diminished significantly (Supplemental Table S5). A total of 4 known stemness-related genes, β-catenin (gene name: catenin beta 1), chromatin assembly factor 1 subunit A, Survivin [gene name: Birc5 (Baculoviral IAP Repeat Containing 5)], and checkpoint kinase 1 (38–41), were included in the positive hit list in our screen, suggesting that our screening strategy was valid (Supplemental Table S4).

Figure 2.

Down-regulation of ATF1 decreases the expression level of the hESC-undifferentiated marker ALP but not cell number (AB). A) Schematic flowchart of shRNA functional screening. B) Down-regulation of ATF1 by shRNA decreased the ATF1 mRNA level and protein level. The Atf1 RNA was harvested on d 6 postinfection and analyzed by qRT-PCR. The result was normalized to Gapdh. Western blot analysis was performed with the anti-ATF1 antibody. Actin served as the internal control. C) The decrease of ATF1 down-regulates ALP activity, and ALP activity was normalized against AB activity to exclude the effect of cell number. Right panel: ALP staining on d 6 postinfection. D) The relative cell number was unchanged upon the knockdown of ATF1. AB activity assays were performed. E) The ratios of G₁ phase, S phase, and G₂/M phase cells in shRFP- and shATF1-expressing hESCs were similar; this suggested that cell cycle progression displayed no difference between these 2 groups. FL2-A, FL2 channel (564-606 nm); MOI, multiplicity of infection. Data are presented as means ± sd for triplicate experiments. **P ≤ 0.01, *P ≤ 0.05.

Knockdown of ATF1 decreases ALP activity but not AB activity in hESCs

To select a crucial factor from the screen positive hit list, we considered not only the ALP:AB ratio and AB activity but also the known gene functions of the 21 genes. One gene, Atf1, which functions as a transcription factor, drew our attention as a potential hESC regulator. We verified that shATF1 can efficiently decrease the RNA and protein expression levels of ATF1 with qRT-PCR and Western blot analyses (Fig. 2B). The expression levels of ALP:AB ratios on d 3 and 6 postinfection shATF1 cells were down-regulated (Fig. 2C). However, the expression of shATF1 did not affect the relative cell number or cell cycle of hESCs, which was demonstrated by measuring AB activity and PI DNA staining (Fig. 2D, E). According to previous studies, ATF1 is well known for its role in regulating cell survival and proliferation through Fos proto-oncogene (c-fos), Jun proto-oncogene (c-jun), cyclin D, and cyclin A (42–45). However, it is interesting that ATF1 may only affect pluripotent maintenance but not cell proliferation in hESCs. Based on these results, we decided to investigate the function of ATF1 in hESCs.

Down-regulation of ATF1 specifically induces hESCs toward NE differentiation and excludes an off-target effect of shATF1 by small interfering RNA, a rescue experiment, and Crispr/Cas9 knockout

After 6 d postinfection, the cell morphology of shATF1 was shown (Fig. 3A). To reveal the consequence of ATF1 knockdown in the undifferentiated condition (MEF conditional medium), we examined the mRNA expression of NE, ME, and trophectoderm markers on d 6 postinfection. Interestingly, we did detect that the NE markers Sox2, Pax6, Sox1, tubulin beta 3 class III (Tubb3), and zinc finger protein of cerebellum family member 1 (Zic1) were up-regulated by qRT-PCR (1.5–4-fold) (Fig. 3B). By contrast, the mRNA expression levels of ME (alpha fetoprotein, Gata4, Gata6, activin, T-box 3, and Bmp4) and trophectoderm (caudal type homeobox 2) markers were down-regulated (Fig. 3B). To confirm the protein expression level of NE markers SOX2, PAX6, and SOX1, we further examined the cells by Western blot analyses and immunofluorescence assays on d 6 postinfection (Fig. 3C). The protein expression levels of SOX2, PAX6, and SOX1 were consistently up-regulated, as were the mRNA expression levels. Moreover, we used Smartpool small interfering ATF1 (siATF1) (On-target plus Smartpool; GE Healthcare, Waukesha, WI, USA) containing 4 small interfering RNAs differing from our shRNA target sequence in a pool to knock down the expression of ATF1. The knockdown efficiency of siATF1 and the expression levels of NE markers were confirmed by qRT-PCR and Western blot analyses (Supplemental Fig. S1A). As expected, the mRNA and protein expression levels of NE markers were increased following siATF1 treatment on d 6 (Supplemental Fig. S1A). By using SMARTpool small interfering RNA, our observations rule out the possibility of an off-target effect of shATF1.

Figure 3.

Down-regulation of ATF1 induces hESC NE differentiation in undifferentiation medium. A) Morphology of shATF1 hESCs after 6 d postinfection. Scale bar, 100 μm. B) The mRNA expression levels of ectoderm markers Sox2, Pax6, Sox1, Tubb3, and Zic1 were up-regulated in d 6 shATF1 hESCs, whereas the trophectoderm marker caudal type homeobox 2 (Cdx2) and ME markers alpha fetoprotein (Afp), Gata4, Gata6, activin, T-box 3 (Tbx3), and Bmp4 were down-regulated. Data are presented as means ± sd for triplicate experiments. ****P ≤ 0.0001, ***P ≤ 0.001, **P ≤ 0.01, *P ≤ 0.05. C) The protein expression levels of the neural ectoderm markers SOX2, PAX6, and SOX1 were up-regulated in d 6 shATF1 hESCs. Western blot analysis (left) and immunofluorescence assay (right) were performed. DAPI was used to stain the nuclei. Scale bars, 100 μm. D) The expression levels of SOX2, PAX6, and SOX1 were down-regulated while overexpressing the shATF1-resistant ATF1 gene in shATF1 hESCs. Upper panel: The silent mutation site of the shATF1-resistant ATF1 overexpression clone. E) The protein expression levels of SOX2, PAX6, and SOX1 were up-regulated in homozygous and heterozygous ATF1-knockout hESCs. WT, wild type.

To further exclude off-target effects of shRNA lentiviruses, we performed a rescue assay. The shATF1-resistant ATF1 overexpression clone (silent mutation) was used to rescue the expression level of ATF1 in shATF1- expressing hESCs (Fig. 3D). The pattern of up-regulating NE markers (SOX2, PAX6, and SOX1) in shATF1 hESCs was attenuated while expressing shATF1-resistant ATF1 (Fig. 3D).

Moreover, we decided to knock out the Atf1 gene in hESCs by using the CRISPR/Cas9 genome editing method. The expression levels of SOX2 and PAX6 were up-regulated in homozygous and heterozygous ATF1-knockout hESCs (Fig. 3E), and the expression level of SOX1 was slightly up-regulated in 1 heterozygous ATF1-knockout hESC (Fig. 3E). The marginal expression levels of SOX1 may result from the effect of the pluripotent maintenance culture condition by E8 medium, which is more stringent than the conditional medium that was used in all other experiments. However, the up-regulating pattern of SOX2 and PAX6 in shATF1 hESCs still could be confirmed in homozygous and heterozygous ATF1-knockout hESCs. According to these results, we excluded the possibility of the off-target effect from shATF1. To certify that the up-regulation of key NE markers SOX2 and PAX6 was not only expressed in H9 cells but also in other hESC lines, we knocked down ATF1 in another hESC line, HUES-6. We found that the morphology and the up-regulation of SOX2 and PAX6 proteins on d 6 were consistent with those in the H9 hESC line (Supplemental Fig. S1B). Of note, our results establish that the down-regulation of ATF1 is sufficient to trigger NE differentiation even in undifferentiation conditions.

A decrease in ATF1 accelerates the expression of hESC NE markers SOX2, PAX6, and SOX1 in a neural differentiation condition

To further examine whether down-regulation of ATF1 can speed up the process of early NE differentiation, we treated shRNA-lentivirus–infected cells in neural differentiation medium, which contains 5 μM dorsomorphin and 10 μM SB431542 (46, 47) 2 d postinfection. The expression levels of early NE markers SOX2, PAX6, and SOX1 were analyzed by immunofluorescence assays. We observed that the expression level of NE markers SOX2, PAX6, and SOX1 were higher in shATF1 hESCs than in shRFP controls (Fig. 4A–C). The results were consistent with those of mRNA expression level by qRT-PCR regarding the down-regulation of the ATF1 expression levels and the up-regulation of NE markers (Sox2, Pax6, and Sox1) (Fig. 4D).

Figure 4.

The decrease in ATF1 up-regulates the expression of hESC NE markers SOX2, PAX6, and SOX1 in neural differentiation condition. Cells were first infected with shRFP or shATF1 lentiviruses and selected by 2 μg/ml puromycin for 2 d. The selected cells were then transferred to neural induction medium containing 5 μM dorsomorphin and 10 μM SB431542. The NE markers SOX2 (A), PAX6 (B), and SOX1 (C) were up-regulated in shATF1 cells at the indicated time points. DAPI was used to stain the nuclei. D) The up-regulation of the mRNA of NE markers in shATF1 cells during the neural induction process. The mRNA expression level of Atf1 and NE markers Sox2, Pax6, and Sox1 were analyzed by qRT-PCR. Scale bars, 100 μm. Data are presented as means ± sd for triplicate experiments. ***P ≤ 0.001, **P ≤ 0.01, *P ≤ 0.05.

ATF1–down-regulated cells can differentiate into functional neurons

Although shATF1 hESCs did accelerate the early NE differentiation process, we wondered whether hESCs can differentiate into functional neurons under ATF1–down-regulated conditions. To reveal the neural maturation phenomenon of shATF1 cells, the neural differentiation was performed. After neural differentiation for 7 d, both the control and ATF1-knockdown cells formed neural dendrites (Supplemental Fig. S2A). In addition, both differentiated cells expressed the neuron marker proteins TUJ1 and MAP2 after induction for 14 d (Supplemental Fig. S2B). To confirm mature neural function in shRNA infected cells, we performed the electrophysiology analysis after 32 d of neural induction. The result showed that both shRFP- and shATF1-differentiated cells were capable of generating action potential (Supplemental Fig. S2C). According to the results, this suggests that the function of generating action potentials in neuron cells is not affected by the down-regulation of ATF1.

The ATF1 expression level is down-regulated upon entry into the early stage of NE differentiation but maintained during differentiation into mesoderm, endoderm, and primitive streak cells

To assess whether the expression levels of endogenous ATF1 changed upon NE induction, we decided to harvest NE-induced cells at several time points and examine the expression of ATF1. To mimic the early NE differentiation process from hESCs to neural stem cells, PSC Neural Induction Medium (Thermo Fisher Scientific) was used to induce H9 hESCs into neural stem cells in 7 d. The expression of the endogenous ATF1 protein was decreased upon NE differentiation from d 1 to 7 (Supplemental Fig. S3A). The expression of SOX2 was increased on d 1, and the expression of PAX6 and SOX1 started on d 5 (Supplemental Fig. S3A). The dynamic expression pattern of SOX2 may result from other NE regulatory factors in PSC Neural Induction Medium or complicate regulatory signals during the differentiation process. Of note, the expression of ATF1 was decreased after 24 h of induction. To investigate the expression pattern of ATF1 in other lineages, we differentiated hESCs into mesoderm, endoderm, and primitive streak-like cells and validated differentiated cells by lineage markers (Supplemental Fig. S3B–G). The mRNA and protein expression levels of ATF1 did not show a significant change during mesoderm, endoderm, and primitive streak differentiation (Supplemental Fig. S3H, I). Based on these results, it is implied that endogenous ATF1 might act as a negative regulator during the early stage of NE differentiation.

Overexpression of ATF1 suppresses NE differentiation of hESCs in neural differentiation condition

As described above, the down-regulation of ATF1 could promote early NE differentiation in both undifferentiated conditions and neural induction conditions (Figs. 3 and 4). Accordingly, we asked if the overexpression of ATF1 could block NE differentiation. To address this question, ATF1-overexpressing and control cells were established and cultured in the presence of dorsomorphin and SB431542. The expression of SOX2, PAX6, and SOX1 proteins were significantly down-regulated in ATF1-overexpressing cells compared with control cells during NE differentiation (Fig. 5). These results indicated that ATF1 overexpression could suppress the early NE differentiation in hESCs.

Figure 5.

Overexpression of ATF1 suppresses NE differentiation. Overexpressed ATF1 or GFP cells were incubated with 2 NE-inducing chemicals, 5 μM dorsomorphin, and 10 μM SB431542. The expression of SOX2 (A), PAX6 (B), and SOX1 (C) was down-regulated upon ATF1 overexpression. DAPI was used to stain the nuclei. Quantification results of fluorescence intensities are showed in the right panel. OV, overexpressed. Scale bars, 100 μm. Data are presented as means ± sd for triplicate experiments. ***P ≤ 0.001, **P ≤ 0.01.

ATF1 binds to the Sox2 promoter and negatively regulates Sox2 promoter activity

We studied the kinetics of key NE marker expression after knocking down ATF1 expression. The protein expression levels of the key NE markers SOX2, PAX6, and SOX1 were examined. We found that after ATF1 knockdown, SOX2 and PAX6 were expressed on d 2 and expressed more abundantly on d 3 (Supplemental Fig. S4A, B). SOX1 expression was detectable on d 3 (Supplemental Fig. S4C). SOX2 has been to be present in the pluripotent state and may act as an upstream regulator of PAX6 (48–50). Combined with the data of kinetics assays, this indicates that Sox2 may act as a major downstream gene of ATF1 in hESCs. To investigate whether the transcriptional levels of Sox2 were regulated by ATF1 in hESCs, we analyzed Sox2 promoter activity by luciferase reporter assays. We coexpressed Sox2 promoter [pSox2-852 (−704 to +148), pSox2-452 (−304 to +148), and pSox2-221 (−73 to +148)] and ATF1 or GFP controls in hESCs. The results showed that the promoter activities of pSox2-852 and pSox2-452 were suppressed in ATF1-overexpressing cells (Fig. 6A). These results indicate that the functional region affected by ATF1 was located between −304 to the transcription start site of the Sox2 promoter. In addition, we used chromatin immunoprecipitation–PCR to determine whether ATF1 has potential to bind to the region of the Sox2 promoter and verified with 3 primer sets (−288 to +19, −289 to −13, and −233 to −88) (Fig. 6B). The PCR result showed that ATF1 has the potential to bind to the Sox2 promoter between the −289 and +19 regions (Fig. 6B). To confirm the effect of down-regulation of SOX2 in ATF1-OV hESCs, the protein expression pattern was examined (Fig. 6C). Overall, these results indicated that ATF1 has the potential to bind to the Sox2 promoter and function as a negative transcriptional regulator of Sox2.

Figure 6.

ATF1 binds to the Sox2 promoter and acts as a negative regulator of Sox2 expression. A) Luciferase reporter assay in ATF1-overexpressed hESCs. The result showed that luciferase activity was repressed (pSox2-852, pSox2-452) while ATF1 was overexpressed in hESCs. B) The result of chromatin immunoprecipitation–PCR showed that ATF1 has the potential to bind to the Sox2 promoter region in H9 hESCs. Beads and IgG served as the negative controls, and Gapdh was the nonbinding control. Red line: The amplification region of each primer set. C) The protein expression levels of ATF1 and SOX2 in ATF1–overexpressed (OV) or GFP-OV hESCs. Ns, not significant; TSS, transcription start site. Data are presented as means ± sd for triplicate experiments. ****P ≤ 0.0001.

Double knockdown of ATF1 and SOX2 represses the expression of NE markers

To verify that SOX2 plays critical roles in the tendency of NE in shATF1 cells, we knocked down the expression of SOX2 in shATF1 cells and examined whether the expression levels of NE markers were altered. In brief, shATF1 hESCs were infected with shRFP control, shSOX2-1, and shSOX2-2 lentiviruses. After 3 d of infection, we collected double-knockdown cells to perform the immunofluorescence assays and qRT-PCR. The protein expression levels of SOX2, PAX6, and SOX1 were down-regulated in shATF1/shSOX2-1 and shATF1/shSOX2-2 cells (Fig. 7A–C). The RNA levels of NE markers Zic1 and Tubb3 also exhibited significantly reduced expression levels in the shSOX2 groups (Fig. 7D). These results show that the differentiation phenomenon was abolished when SOX2 is knocked down in shATF1 cells. Altogether, this further indicates that early NE differentiation in shATF1 hESCs is due to the up-regulation of SOX2.

Figure 7.

Down-regulation of ATF1 and SOX2 decreases NE gene expression. The expression of NE genes was examined on d 3 after reseeding shRFP/shATF1 cells. A–C) Down-regulation of the protein levels of SOX2 (A), PAX6 (B), and SOX1 (C) in ATF1/SOX2–double knockdown cells. Quantification results of fluorescence intensities were showed in the right panel. Scale bars, 100 μm. D) Down-regulation of RNA expression levels of Atf1, Zic1, and Tubb3 in ATF1/SOX2–double knockdown cells. Data are presented as means ± sd for triplicate experiments. ***P ≤ 0.001, **P ≤ 0.01, *P ≤ 0.05.

DISCUSSION

In this study, by a high-throughput screen, we have revealed a novel function for ATF1 in maintaining pluripotency without affecting cell renewal or cell cycle progression (Figs. 1 and 2). Knockdown of ATF1 expression in hESCs specifically up-regulates NE in undifferentiation conditions (Fig. 3 and Supplemental Fig. S1). The off-target effect of shATF1 was excluded by siATF1, ATF1 cDNA rescue assays, and CRISPR/Cas9 ATF1 knockout in hESCs (Fig. 3). Under the NE induction process, down-regulation of ATF1 can accelerate early NE differentiation and differentiate into functional neurons (Fig. 4 and Supplemental Fig. S2). The expression of endogenous ATF1 in hESCs is down-regulated after 24 h postinduction (Supplemental Fig. S3). In contrast, overexpression of ATF1 suppresses early NE markers SOX2, PAX6, and SOX1 expression during NE differentiation (Fig. 5). Additionally, ATF1 has the potential to bind to the Sox2 promoter region and function as a negative transcriptional regulator (Fig. 6). Knockdown of SOX2 expression in shATF1 hESCs can abolish the up-regulated phenomena of early NE markers (Fig. 7). Taken together, our results demonstrate that ATF1 is an upstream repressor of the expression of SOX2 and may play an important role in the very first step toward NE differentiation.

SOX2 and PAX6 are both well known for their critical roles in NE fate commitment (51, 52). Similar to our observation in shATF1 hESCs, overexpression of SOX2 protein in the hESC line (H1) enhances NE differentiation in the presence of differentiation medium (14). However, it is unclear which transcription factors restrained SOX2 expression in the undifferentiation conditions to prevent NE differentiation. In this study, we found down-regulation of ATF1 was sufficient to promote early NE differentiation by rapidly up-regulating SOX2 and PAX6 RNA and proteins on d 2 and 3–6 (∼1.5–3-fold) under undifferentiation conditions. Prior studies reported the reciprocal regulation of SOX2 and PAX6 in a mouse model (48–50). Interestingly, our results showed that a double knockdown of ATF1 and SOX2 decreases the expression level of PAX6. This implies that SOX2 acts as an upstream regulator of PAX6 and that the expression levels of SOX2 are important for an early step in NE commitment in hESCs.

Of note, the loss of function and gain of function of PAX6 clearly demonstrates that PAX6 is one of the topmost determinants of human NE fate (52). Unlike SOX2, the upstream regulators of PAX6 expression have been reported (53–56). PAX6 can be directly regulated by SOX2 (48). The miR-96 family suppresses PAX6 expression and blocks neuron differentiation (53). In contrast, a transcriptional factor, LHX2 (LIM homeobox 2), enhances PAX6 expression during NE differentiation in hESCs (54). However, the expression levels of miR-96 and LHX2 are not down-regulated or up-regulated to 50% until 4–6 d post NE induction, which suggests that they might play a role in a later stage. In contrast, we found that ATF1 began to be down-regulated 24 h after neural induction. Although the induction protocols are different from previous studies (53, 54), our observations may account for the expression of ATF1 in the embryonic stage as most likely an important gatekeeper that prevents early neurogenesis in hESCs.

The phenomena in ATF1-knockdown hESCs were different from those in ATF1-null mice. ATF1-null mice were reported to not have a phenotypic defect. Based on previous studies, some essential genes in humans are not essential genes in mice (57). For example, St8sia4 (ST8 alpha-N-acetyl-neuraminide alpha-2,8-sialyltransferase 4) involved in glycosylation (58), TG-interacting factor in holoprosencephaly (59), neurofibromin 1 in neurofibromatosis type 1 genetic disorder (60), and C9orf72 (C9orf72-SMCR8 complex subunit) in familial amyotrophic lateral sclerosis and frontotemporal dementia (61) were all reported to have different functions in humans in comparison with mice. As a result, the different observations in our studies may be caused by interspecies variations.

The function of ATF1 in neurogenesis may also be cell context dependent. In our study, down-regulation of ATF1 facilitates NE differentiation under undifferentiated conditions or differentiated conditions in hESCs. However, in rat PC12D cells, dominant negative ATF1 blocks cAMP-induced neurite outgrowth (62). This inconsistency in the ATF1 function in neurogenesis in our results and the PC12D study might be due to differences in: 1) species (human vs. rat), 2) developmental states (ESC vs. neural crest state), 3) cell states (normal cells vs. tumor cells), and 4) induction signals [dual-SMAD (drosophila mothers against decapentaplegic protein) inhibitor vs. cAMP]. A similar example of such a difference is in CD38–cyclic ADP ribose signaling during neural differentiation, which can be observed in PC12 cells vs. mouse ESCs (63). Thus, the functions of ATF1 in different developmental stages or species will be an interesting subject for future work.

ATF1 is well known for its ability to promote cell survival, proliferation, and tumorigenesis (64–66), but the role of ATF1 in blocking neurogenesis has not been reported. In addition, ATF1 is widely known as a transcriptional activator (67–73). Until now, ATF1 has only been reported to function as a repressor in the regulation of the following 4 genes: ferritin, retinoblastoma, gelsolin, and thrombospondin I (74–77). In our study, ATF1 functions as a repressor of SOX2 expression in hESCs during early development. However, whether the ATF1 functions through traditional ATF1 binding motifs as a repressor still needs to be investigated.

In our study, a 2-dimensional (2D) neural induction protocol was used to examine the influence of ATF1 on neural maturation. Although the action potentials were detected both in shATF1 and control cells by electrophysiology analyses, the firing rate of shATF1 cells was generally lower than shRFP cells. The difference in firing patterns may be caused by the different maturation time of neuron cells (78–80) or different subtypes of neuron cells (81–83). This raises an interesting issue for the functions of ATF1 during the neuron maturation process and needs to be further examined in future work. However, the influence of neural complexity and regulatory circuits may be limited by the 2D induction protocols (84). A 3-dimensional (3D) organoid model has been established recently and may be suitable for analysis of the networks and functions of multiple cell types (85–87). The self-organized property in 3D culture provides a better niche than 2D for neural maturation and subtype neuron differentiation. Therefore, the detailed effects of ATF1 during the late stage of neural differentiation may be uncovered through further analyses using the 3D organoid model.

Recently, nuclear factor erythroid 2-related factor 2 (NRF2) (PAN: primary cilia-autophagy-Nrf2) was established to regulate the NE commitment of hESCs by down-regulating the OCT4-NANOG signal and prolonging the G₁ phase of the cell cycle (47). Both ATF1 and NRF2 are expressed in hESCs; this may suggest they act as the key repressors of the NE lineage. However, they seem to have different regulatory signals in NE commitment. The differences may be due to: 1) different target genes, 2) a different regulatory network in the cell cycle, 3) different regulation mechanisms, or 4) different pathways and kinetics. The differences in target genes mean ATF1 suppresses Sox2, whereas NRF2 reduced the expression of Oct4 and Nanog. In the embryonic stage, down-regulation of ATF1 does not alter the cell cycle, but NRF2 is coupled to cell cycle progression. Different regulatory mechanisms are presented by the transcriptional regulator (ATF1) and the posttranscriptional regulator (NRF2). Different pathways and kinetics are shown by the fact that ATF1–down-regulated cells up-regulated PAX6 expression on d 2–3 in undifferentiated condition, and knocking down NRF2 up-regulated PAX6 expression on d 4 in NE induction medium. Therefore, our findings suggest that ATF1 suppression is a different event compared with NRF2 suppression, and they both occur through different pathways in NE specification.

Here, we provided strong evidence that ATF1 is a gatekeeper of NE commitment for maintaining hESC pluripotency via the balance of SOX2 expression. ATF1 suppresses NE differentiation by repressing the expression of a critical NE regulating gene, SOX2. Our study highlights that ATF1 may play a key role in a very early step of hESC differentiation of the NE fate.

Supplementary Material

This article includes supplemental data. Please visit http://www.fasebj.org to obtain this information.

Glossary

2D

2-dimensional

3D

3-dimensional

AB

alamarBlue

ALP

alkaline phosphatase

ATF1

activating transcription factor 1

BMP4

bone morphogenetic protein 4

Cas9

CRISPR-associated protein 9

CRISPR

clustered regularly interspaced short palindromic repeats

EB

embryoid body

ESC

embryonic stem cell

GAPDH

glyceraldehyde-3-phosphate dehydrogenase

GFP

green fluorescent protein

hESC

human ESC

iPSC

induced PSC

ME

mesendoderm

MEF

mouse embryonic fibroblast

MGI

mouse genome informatics

NANOG

nanog homeobox

NE

neuroectoderm

NPC

neural progenitor cell

NRF2

nuclear factor erythroid 2-related factor 2

OCT4

octamer-binding transcription factor 4

PAX6

paired box 6

PI

propidium iodide

PSC

pluripotent stem cell

qRT-PCR

quantitative RT-PCR

RNAi

RNA interference

shATF1

shRNA targeting Atf1

shRFP

shRNA targeting red fluorescent protein

shRNA

short hairpin RNA

shSOX2

shRNA targeting Sox2

siATF1

small interfering ATF1

SOX

sex-determining region Y-box

Tubb3

tubulin beta 3 class III

Zic1

zinc finger protein of cerebellum family member 1

AUTHOR CONTRIBUTIONS

S.-C. Yang performed most of the research, analyzed the data, and wrote the manuscript; J.-J. Liu generated the CRISPR ATF1-knockout hESCs and wrote the manuscript; S.-C. Yang and C.-K. Wang performed the high-throughput screen; Y.-T. Lin performed mesoderm, endoderm, and primitive streak-like cell differentiation and reporter assay; S.-Y. Tsai supported the feeder-free culture system and CRISPR/Cas9 genome editing method, discussed the data, and contributed to manuscript writing; S.-C. Yang, J.-J. Liu, W.-J. Chen, and W.-K. Huang performed the neural differentiation assay; P.-W. A. Tu, Y.-H. Lee, Y.-C. Chiu, and C.-C. Hsu performed plasmid cloning; Y.-C. Lin performed the surveyor assay; C.-F. Chang and C.-L. Cheng performed cell cycle analysis and statistical data analysis; H. Lin performed immunofluorescence image support; P.-W. A. Tu, C.-L. Cheng, C.-Y. Lai, and C.-Y. Lin performed Western blot assay; S.-C. Hsu discussed the data and contributed to manuscript writing; M. Hsiao provided reagents and discussed the data; S. C. Schuyler and F. L. Lu discussed the data and edited the manuscript; J. Lu proposed and directed the project, analyzed and discussed the data, and wrote the manuscript; and all authors approved the final manuscript.