Genetic screening and multipotency in rhesus monkey haploid neural progenitor cells

Having only one set of chromosomes, haploid cells provide an excellent system with which to uncover gene function in recessive traits because homozygous genotypes are easy to obtain. The derivation of medaka fish haploid embryonic stem cells (haESCs) opened a new era of genetic screening in vertebrates (Yi et al., 2009). Recently, haESCs with pluripotent features were obtained from rodents (Leeb and Wutz, 2011; Li et al., 2014) and primates (Sagi et al., 2016; Yang et al., 2013), which is a promising advance in mammalian forward and reverse genetics (Shuai and Zhou, 2014; Wutz, 2014). haESCs have been practically applied to discover genes that are targeted by toxins (Elling et al., 2011) and pluripotency exit genes (Leeb et al., 2014). haESCs have even served as a mutant bio-bank for tracing unknown genes in many interesting biological processes (Elling et al., 2017). However, haESCs tend to diploidize during proliferation and differentiation, thus necessitating periodic sorting for haploids. This step hinders the development of haploid somatic cell types in mammals and, consequently, lineage-specific genetic screening (Li and Shuai, 2017; Shuai and Zhou, 2014). Unlike mouse haESCs that diploidize quickly, monkey (Yang et al., 2013) and human (Sagi et al., 2016) haESCs seem to more stably maintain haploidy. Whether a primate haploid somatic cell line can be obtained has not yet been investigated. Neural progenitor cells (NPCs) are proliferative adult stem cells that have been extensively studied for their potential in curing neural degenerative diseases and have the capacity to terminally differentiate into neural subtypes (Temple, 2001). In the differentiation of mouse haESCs, although haploid neural precursors (Leeb et al., 2012) and neurons (Xu et al., 2017) have been transiently detected, a haploid NPC (haNPC) line has not, until very recently, been achieved (Gao et al., 2018; He et al., 2017). Rodent haESCs may proliferate and diploidize faster than primate haESCs, but whether monkey haNPCs can be obtained remains unknown.

In this study, we cultured rhesus monkey haESCs in an optimized naïve medium to gain pluripotency. By purifying haploids from early stage NPCs, haNPCs that could maintain long-term haploidy, were generated. These cells are haploid somatic cells that perform excellently in high-throughput genetic screening, thereby serving as a valuable resource for recessive neural gene studies.

To determine whether our monkey haESCs (cell line: PaH-3) were haploid and pluripotent, we compared them with zygote-derived diploid monkey ESCs by assessing various aspects. PaH-3 cells were morphologically similar to diploid ESCs and showed flat compact colonies, indicating a standard, primed pluripotent state (Fig. S1A). After fluorescence-activated cell sorting (FACS) for haploid cells, PaH-3 cells maintained haploidy for more than 2 months without sorting (Fig. 1A), which is in accordance with a previous report (Yang et al., 2013). We further investigated the chromosome numbers in PaH-3 cells and found that most cells were haploid (21 chromosomes) with a few aneuploid cells (Fig. 1B and Fig. S1B). Next, we assessed the pluripotency of PaH-3 cells at molecular and developmental levels. PaH-3 cells had strong alkaline phosphatase (AP) activity (Fig. S1C) and, after immunostaining, expressed pluripotent markers, which included OCT4, NANOG, SOX2, SSEA4, TRA-1-60 and TRA-1-81 (Fig. 1C). This characteristic is similar to the diploid ESCs (Fig. S1D). After the withdrawal of basic fibroblast growth factor (bFGF) and feeder cells, PaH-3 cells formed standard embryoid bodies (EBs) that guaranteed subsequent differentiation experiments (Fig. S1E). Teratomas containing three germ layers formed when ∼1×10⁷ PaH-3 cells were injected into the limbs of severe combined-immune-deficiency (SCID) mice, which indicated that monkey haESCs held pluripotency in vivo as well (Fig. S1F).

Although PaH-3 seemed to be able to stably maintain haploidy during the ESC state, they underwent severe diploidization during differentiation. In preliminary experiments, haNPC derivation of PaH-3 cells from a primed state was difficult. The main reason for this difficulty might be that the haNPCs that were differentiated from primed haESCs were limited and therefore survived poorly. Hence, an advanced haESC culture system was urgently needed. Recently, primate ESCs were converted to a naïve state by optimizing the culture medium (Gafni et al., 2013) or through regulation of transcription (Takashima et al., 2014). These cells were similar to mouse ESCs both morphologically and molecularly. With a few modifications, we therefore optimized the medium of monkey haESCs according to a previous report (Fang et al., 2014) (Fig. S2A). After being cultured in the optimized naïve state medium for three to four passages, the PaH-3 colonies changed from a flattened state to a more compact domed state (Fig. 1D). Optimized PaH-3 cells grew faster and could be single-cell passaged in the presence of the Rho-associated protein kinase inhibitor (ROCKi) Y27632 (Fig. S2A). Newly sorted pure haploid cells were cultured separately in optimized and traditional conditions for over 2 months without sorting. FACS results demonstrated that rhesus monkey haESCs in optimized medium maintained a high percentage of haploid cells, whereas cells in a traditional medium diploidized to some degree (Fig. 1E and Table S1). Accordingly, we investigated the expression levels of pluripotent genes (Oct4 and Nanog) between optimized and traditional haESCs using quantitative PCR (qPCR). The results showed that expression levels of pluripotent genes in naïve haESCs were significantly higher than that in primed haESCs (Fig. S2B). To compare the ability of EB formation between the two conditions, we suspended almost 1×10⁶ cells from each group for aggregation. The number of EBs generated from optimized haESCs was approximately twice that generated from traditionally cultured cells, which indicated that haESCs had a higher clonogenicity ability in an optimized medium (Fig. 1F and Table S1). Moreover, aggregate-generating neural rosettes from optimized haESCs were larger in size and more abundant than those from traditionally cultured cells (Fig. 1G,H and Table S1). During the first sorting of haNPCs and according to DRAQ7 analysis, there were more surviving cells from the optimized group than from the traditional group (Fig. 1I and Table S1). This finding indicated that haESCs in the optimized medium could produce more NPCs with better survival capacity. A similar trend was found in harvested cells, in which we found that fewer cells sorted from the traditional group survived than from the optimized group (Fig. S2C). When NPCs that were derived from primed haESCs were isolated to proliferate until there were sufficient cells for a second sorting, all NPCs had diploidized (Fig. S2D). To assess the differentiation efficiency between the optimized and traditional groups, we performed immunostaining for the NPC-specific (SOX1) and ESC-specific (OCT4) markers in single-cell spreads of neural rosettes that were picked from each of the two groups (Fig. S2E). The statistical analysis of SOX1-positive cells revealed that more NPC cells were gained in the optimized group than in the traditional group (Fig. S2F). To clarify whether optimized conditions warranted pluripotency, we performed differentiation of diploid ESCs cultured in the two conditions separately. In analysing the size of the formed neural rosette, we found that the area of rosettes from the naïve diploid ESCs was much larger than that from the primed ESCs (Fig. S2G). This result indicated that using a naïve medium can benefit pluripotency regardless of ploidy. In total, three haploid NPC lines were derived from the optimized group, whereas no haploid NPC line was successfully obtained from the traditional group (Fig. S2H). Taken together, these observations demonstrate that rhesus monkey haESCs cultured in an optimized naïve culture medium showed mouse ESC-like morphology, were pluripotent and were able to generate proliferative haNPCs.

For neural differentiation, the formation of aggregates in suspension and the adherent culture are the most commonly used methods (Jensen et al., 2013; Zhang et al., 2001). In this study, we developed a protocol according to previous reports with slight modifications to derive haNPCs (Fig. 2A). In the first 5 days, monkey haESCs with good morphology (Fig. S3A) were suspended in embryoid body (EB) medium to allow for aggregation. During the first day, we introduced ROCKi (Y27632) to the EB medium to avoid severe apoptosis. The well-formed EBs (Fig. S3B) were further floating-cultured in a neural-induction medium (NIM) for 2 days. A neuroectoderm structure could be visualized on 7-day-old aggregates (Fig. 2B). All aggregates were seeded onto Matrigel-pre-coated dishes with NIM for further differentiation. Some of the aggregates adhered to the Matrigel and began to form primitive neuroepithelia 4-5 days later (Fig. 2C). Approximately 15 days after differentiation, massive aggregates developed into a definitive neuroepithelial stage with neural rosettes forming in the centre (Fig. 2D). Rosettes were manually picked and expanded in NPC medium for another 14 days (Fig. S3C). HaNPCs were passaged approximately four times before sorting. There were nearly 10% haploid cells at passage 4 (P4) during the first sorting (Fig. 2E). Interestingly, purified haNPCs could be expanded in NPC medium for more than 20 passages without sorting, and presented a high percentage of haploid cells. HaNPCs maintained standard NPC morphology (tiny and bi-polar) before and after sorting (Fig. 2F, Fig. S3D), thereby maintaining the potential to aggregate into neural spheres in untreated dishes (Fig. 2G). We further performed a karyotype analysis of haNPCs and found that their haploid genome remained intact without any aneuploidy being present (Fig. 2H). Next, the cell growth rate between haNPCs and diploid NPCs was compared using the CCK-8 assay. These results showed that there was no significant difference between them (Fig. S3E).

To determine whether haNPCs possessed neural properties, we identified NPC-specific markers by immunostaining, using diploid NPCs as a positive control. The results showed that, similar to the controls (Fig. S3F), haNPCs expressed Nestin, Sox1 and Pax6 during both adherent growth (Fig. 3A) and suspension culture conditions (Fig. 3B). According to a qPCR analysis, haNPCs, which resembled diploid NPCs, expressed only NPC-specific markers (Nestin, Pax6 and Sox1) and did not express any pluripotent markers such as Oct4. This finding indicated that haNPCs had abandoned their ESC state (Fig. 3C,E). Next, we compared the protein levels of NPC-specific markers, SOX1 and PAX6, between haNPCs and diploid NPCs using western blotting. Both haNPCs and diploid NPCs were highly positive for SOX1 and PAX6 (Fig. S3G). Additionally, transcriptome levels of haNPCs were overall close to diploid NPCs and significantly distinct from haESCs and diploid ESCs (Fig. 3D and Fig. S3H). Somatic cells in female individuals undergo X chromosome inactivation (XCI) during development, which is regulated by Xist (Gontan et al., 2012; Grant et al., 2012). However, haESCs have only one X chromosome, which means that they might have a different mechanism for triggering differentiation with one activated X chromosome. We therefore investigated the state of X chromosomes in haNPCs. The results of qPCR demonstrated that haNPCs did not express Xist compared with primed diploid ESCs and their derivative NPCs. This result suggested that haESCs could differentiate into haNPCs without XCI (Fig. 3E).

To assess the differentiation potential of haNPCs, we induced glia and neuron differentiation using specific growth factors. We sorted haploid cells from the 3-week-old differentiated cultures, finding that 13.1% of the cells were still haploid (Fig. 3F). The sorted haploid cells were re-plated and analysed using immunostaining techniques. The results demonstrated that these haploid cells were positive for neuron (Tuj1) and astrocyte (GFAP) markers, indicating that haNPCs potentially differentiate into neural subtypes in a haploid genome (Fig. 3G). In a parallel experiment, diverse neurons (Map2 and NeuN positive) and oligodendrocytes (O4 positive) were generated from haNPCs (Fig. 3H). To address the purity of haNPCs, we randomly picked a single haNPC sphere to perform differentiation (Table S3). The results revealed that each sphere could differentiate into neurons, astrocytes and oligodendrocytes. Therefore, multipotency is an intrinsic feature of haNPCs. To determine the clonal growth of haNPCs from a single cell, we analysed the proliferative ability of haNPCs by plating them at four densities (1, 100, 1000 and 10,000 cells) in a single well of a 96-well plate to expand after sorting. The results demonstrated that a single cell in a well could not survive and expand, suggesting that haNPCs might be a mixture of cells in which some lack self-renewal as stem cells (Fig. S3I and Table S2). Moreover, to examine whether haNPC-derived neurons possessed electrophysiological function, we performed whole-cell recordings of the differentiated cells. Inward, fast inactivating, and outward currents were detected during recording, meaning that the cells opened voltage-dependent Na⁺ and K⁺ channels. The Na⁺ channels could also be blocked in the presence of TTX, which is a standard characteristic of functional neurons (Fig. 3I).

Given that haNPCs are haploid and are proliferative cells with the potential to yield multiple neural subtypes, we performed gene-trapping using a piggyBac transposon system. Briefly, ∼5×10⁵ haNPCs were electroporated with piggyBac (PB)-based trapping vectors (Leeb et al., 2014; Mossine et al., 2013) carrying an eGFP gene (Fig. 4A). After transfection, almost 16.4% of the cells expressed GFP, confirming the insertion of a PB vector (Fig. 4B). An FACS analysis showed that 40.7% of the remaining cells were haploid, whereas the percentage of diploid cells increased slightly (Fig. S4A). According to the FACS analysis, the percentage of haploid cells in GFP-negative cells was much higher than that in GFP-positive cells (Fig. S4B). This result might be because diploid NPCs have a higher tolerance of harsh environments during electroporation than do haNPCs. Eight subclones were randomly picked for analysis of insertion sites by inverse PCR and Sanger sequencing (Fig. S4C). The results showed that all eight clones had multiple insertions (Fig. 4C). According to sequencing data, five insertion sites were integrated into intergenic regions, whereas two were in exonic regions. The two inserted genes in the monkey genome were similar to human MT1JP and APLP2 genes, which are located on chromosomes 16 and 11, respectively (Fig. S4D).

The integration sites of piggyBac vectors in haNPCs were determined using splinkerette PCR followed by deep sequencing. Approximately 180,000 independent insertions were identified, of which 54.1% were derived from the sense orientation (Fig. 4F). In addition, ∼18% of the insertions were located in coding regions, whereas 51% of insertions landed in intergenic regions (Fig. 4G). An enrichment analysis with gene ontology databases showed that insertions preferred genes carrying specific functions for neurons, such as trans-membrane activity and synaptic transmission (Fig. 4H). Taken together, these data from mutant haNPCs were comparable with those derived from other mammalian systems (Elling et al., 2011; Li et al., 2014). To further demonstrate the utility of these cells for genetic screens (Fig. 4D), we sequenced another group of haNPCs with piggyBac insertions that survived A803467 treatment (Fig. 4E). In this analysis, 20,958 genes were detected with 170,000 integration sites after A803467 selection. By comparing the insertion frequencies between this selected library and the control library mentioned above, we identified genes carrying more insertions, as depicted in the top candidates (Table S4). SCN5A, an integral membrane protein, has been proven to be responsible for tetrodotoxin resistance (Gellens et al., 1992) and was listed in the top candidates (Fig. 4I).

To examine whether our candidate genes were target genes for A803467, we performed gene editing in diploid NPCs via the CRIPSR/Cas9 system on the top two genes (B4GALT6 and SCN5A) on the screened list (Table S4). We knocked out B4GALT6 and SCN5A genes separately in diploid NPCs by co-electroporation using the Cas9-expressing vector and two sgRNAs targeting two exons of these genes (Fig. 5A). Approximately 36 h after electroporation, we performed FACS to enrich for GFP-positive cells (for a Cas9 plasmid carrying a GFP gene). An efficiency of 23.9% (B4GALT6) and 37.1% (SCN5A) was indicated (Fig. S5A,B). The gene-edited diploid NPCs and control diploid NPCs were treated with A803467 for 7 days separately. Both B4GALT6 knockout NPCs and SCN5A knockout NPCs survived A803467 treatment, whereas the control diploid NPCs died (Fig. 5B). According to the statistical analysis, 38% of B4GALT6 knockout NPCs and 45% of SCN5A knockout NPCs survived after treatment (Fig. 5C). To confirm the genotypes of these resistant NPCs, we extracted the genomic DNA and performed a T7EN1 cleavage assay. The results demonstrated that the cleavage bands were visible in the B4GALT6- and SCN5A-targeted groups (Fig. 5D). Further sequencing of the targeted fragments confirmed that B4GALT6 and SCN5A were mutated separately in the diploid NPCs (Fig. 5E,F, Fig. S5C,D). Taken together, B4GALT6 and SCN5A were toxicity target genes for A803467.

HaESCs have been derived in many species, including higher mammals, but no studies have addressed the exact mechanism under which diploidization occurs. This issue compromises the development of haploid cells in differentiated somatic cell types. An important issue that has been raised is whether haESCs need to gain their differentiation potential through diploidization. Shuai et al. found that the haploidy of primed-state haploid cells appeared unstable and declared that haploidy maintenance might be irrelevant to the pluripotent state (Shuai et al., 2015). Our results confirmed this idea as rhesus monkey haESCs maintained their haploid state well, irrespective of the state in which they were cultured. Notably, monkey haESCs cultured in optimized naïve medium were quite stable in their haploid states and proliferated quickly, subsequently guaranteeing differentiation by providing large numbers of haploid pluripotent stem cells.

Monkey ESCs could be derived from pre-implantation blastocysts (Thomson et al., 1995) and converted to a presumptive naïve state (Fang et al., 2014), which was proven to be pluripotent through their contribution in chimaeras (Chen et al., 2015). However, whether naïve monkey ESCs show promising differentiation potentials in vitro has not yet been determined. We compared monkey haESCs cultured in optimized and traditional conditions, and found that naïve monkey haESCs had advantages in maintaining haploidy, differentiation viability and neural differentiation potential. This finding suggests that a naïve culture system might be beneficial as a cellular resource for better quality and a higher yield of monkey haESCs.

As described previously, nestin-positive and Oct4-negative cells could be derived from haESCs in vitro but lacked stability, proliferation capacity and neuronal differentiation potential during trials (Elling et al., 2011; Shuai et al., 2015). Leeb et al. obtained mouse haploid extra-embryonic cells by the overexpression of GATA6GR, indicating that haploid cells were sustained beyond a pluripotent state (Leeb et al., 2012). In addition, human haESCs could differentiate into terminal somatic cells representing three germ layers with a haploid genome (Sagi et al., 2016). This finding demonstrated that primate species might have an unknown advantage in maintaining haploidy. Our derivation of monkey haNPCs consequently proved this point because they could expand without sorting over the long term.

In summary, we optimized the culture system, thereby facilitating the derivation of monkey haNPCs. Invaluable to primate neural genetic screening and drug targeting discovery, this is the first report of a multipotent, haploid, monkey somatic cell line with the ability to proliferate into neural cell subtypes. Finally, we screened out two tetrodotoxin-resistance genes using monkey haNPCs and validated their function by gene editing.

Specific pathogen-free (SPF) grade mice were purchased from Beijing Vital River Laboratory Animal Technology and housed at the Nankai University animal centre. All animal procedures were performed according to the guidelines for the Use of Animals in Research issued by Nankai University animal centre.

Monkey haESCs were kindly gifted to us from Dr Tao Tan (Kunming University of Science and Technology, China) and were cultured in optimized ES medium as previously described (Fang et al., 2014). The traditional ES medium consisted of D/F12 (Gibco) supplemented with 20% KOSR (Gibco) and 0.1 mM nonessential amino acids, 0.1 mM 2-mercaptoethanol (Sigma-Aldrich), 100 U/ml penicillin, 0.1 mg/ml streptomycin (Gibco), 2 mM L-glutamine (Sigma-Aldrich) and 5 ng/ml bFGF (Peprotech). The optimized monkey ES medium consisted of D/F12 supplemented with 20% KOSR and 0.1 mM nonessential amino acids, 0.1 mM 2-mercaptoethanol, 100 U/ml penicillin, 0.1 mg/ml streptomycin, 2 mM L-glutamine, 5 ng/ml bFGF, 10 ng/ml human LIF (Peprotech), 0.5 μM PD0325901 (Gene Operation), 3 μM CHIR99021 (Gene Operation), 10 μM SB203580 (Peprotech) and 10 μM SP600125 (Peprotech).

To derive EBs, we digested cells with 0.05% trypsin/EDTA (Gibco) into single cells and cultured as floating aggregations in untreated dishes with EB medium. On the first day, 10 μM Y-27632 was introduced to reduce cell apoptosis. As described previously, the medium was changed to a neural induction medium (NIM) on day 5 (Jensen et al., 2013; Zhang et al., 2001). In short, EBs were plated on day 7 onto culture dishes pre-coated with Matrigel and cultured in NIM. Rosette clumps were generated after culturing in NIM for 8-10 days, then picked and expanded. Briefly, rosette clusters were attached to dishes coated with 30 μg/ml polyornithine and 3 μg/ml laminin (PLO/Lam) (Sigma-Aldrich). HaNPCs were expanded in NPC medium containing human bFGF and EGF (Peprotech).

To purify haploid cells, we dissociated the cells with 0.05% Trypsin/EDTA, incubated them with Hoechst 33342 (5 μg/ml) in a water bath at 37°C for 15 min and purified them on a BD FACS AriaII cell sorter. For DRAQ staining, DRAQ7 (CST) (1:100 diluted) was added to 5×10⁵ cells at a final concentration of 3 μM. The samples were incubated for 10 min on ice and then analysed on a BD LSRII SORP. In clonal haNPCs analysis, four densities (1, 100, 1000 and 10,000 cells) of sorted haploid cells were distributed to a single well of a 96-well plate each on a BD FACS AriaII for further culturing. For DNA content analysis, the cells were trypsinized into single cells and fixed overnight with 75% ethanol at 4°C. Fixed cells were stained with 5 μg/ml propidium iodide (PI) and 2 μg/ml RNase A at 37°C in a water bath for 15 min. Flow cytometry data were recorded using a BD LSRII SORP flow cytometer. All FACS data were analysed using FlowJo software.

To investigate the differentiation potency of haNPCs, we re-plated the haNPC cells onto PLO/Lam-coated plates at a density of 2×10⁴ cells/cm² in DMEM/F12 medium supplemented with N2, B27 (Gibco), GDNF (20 ng/ml, Peprotech), BDNF (20 ng/ml, Peprotech), dibutyryl cyclic AMP (1 mM, Sigma-Aldrich) and ascorbic acid (200 nM, Sigma-Aldrich). For oligodendrocyte and astrocyte differentiation, 1% foetal bovine serum (Biological Industries) was introduced into the differentiation medium (Zhao et al., 2014).

Cells were fixed with 4% paraformaldehyde at 4°C overnight, penetrated with 0.3% Triton X-100 (Sigma) for 1 h at room temperature and then incubated with 3% BSA (Sigma) for 1 h at room temperature. Cells for immunostaining were then incubated overnight at 4°C with primary antibodies against nestin (Abcam, 92391), Pax6 (Abcam, ab5790), Sox1 (Abcam, ab227118), Oct4 (Abcam, ab27985), Nanog (Abcam, ab80892), SSEA4 (Abcam, ab16287), Tuj1 (Abcam, ab18207), Map2 (Abcam, ab32454), GFAP (Abcam, ab7260), NeuN (Abcam, ab104225) and O4 (R&D Systems, MAB1326), all used at 1:400. After the cells were washed three times with PBS, they were incubated with the corresponding secondary antibodies (1:1000; Abcam, ab150080, ab150132, ab150077, ab6724) for 1 h at room temperature. Alkaline phosphatase staining was performed according to the manufacturer's instructions using an alkaline phosphatase staining kit (Beyotime). For western blot analysis, protein samples were extracted from n NPCs and 2n NPCs. Equal amounts of each protein sample were electrophoresed on 10% SDS-polyacrylamide electrophoresis gels and then transferred onto a polyvinylidene fluoride membrane. Blocked membranes were incubated with anti-Sox1 antibody (Santa) and anti-Pax6 antibody (Abcam). After the membranes were washed, a horseradish peroxidase (HRP)-conjugated secondary antibody (1:5000 dilution; Sungene, LK2001) was added to the membranes, which were then incubated at approximately 3°C for 1 h and washed. The immunoreactivity was detected using an ECL detection kit (Beyotime), and the results were quantified using the Quantity One computerized imaging program (Bio-Rad).

ESC and NPC cells were incubated with 0.2 μg/ml nocodazle (Sigma) for 3 and 6 h, respectively. After trypsinization, the resuspended cells were treated with 0.075 mM KCl at 37°C for 30 min, then exposed to a hypotonic solution and fixed in methanol:acetic acid (3:1 in volume) for 20 min at 4°C. Cells were then dropped onto pre-cleaned slides. After being air dried, cells were stained with a Giemsa dye solution for 7 min. More than 20 metaphase spreads were analysed. For teratoma analysis, around 1×10⁷ monkey haESCs were injected subcutaneously into the limbs of 8-week-old male severe combined immunodeficiency (SCID) mice. Fully formed teratomas were dissected 4 weeks later, then fixed in 4% paraformaldehyde, embedded in paraffin, sectioned and stained with Haematoxylin and Eosin for further analysis.

Cells were plated into 96-well plates at a density of 1000 cells per well and cultured in a 90 μl NPC culture medium. Each day, 10 μl cell counting kit-8 (CCK-8) medium (Beyotime) was added into one well. A CCK-8 analysis over 5 days was performed according to previous methods (Yang et al., 2013).

HaNPCs were grown on a PLO/Lam-coated 35 mm dish and differentiated for ∼3 weeks before electrical currents were recorded. Cells were superfused with oxygenated bicarbonate-buffered ACSF. We recorded spontaneous and step-current, injection-evoked activity using current-clamp mode. Whole-cell currents were recorded in voltage-clamp mode with a basal holding potential of −70 mV. Voltage steps that ranged from −60 to +50 mV were delivered in 10 mV increments. TTX was added at a concentration of 500 nM.

Total RNAs from ESCs and ESC-derived NPCs were extracted using Trizol Reagent (Invitrogen), and cDNA was obtained using a Prime Script RT reagent Kit using a gDNA Eraser (Takara). Q-PCR was performed on ABI QuantStudio 6 Flex machine with FS Universal SYBR Green Master (Roche). Relative expression levels were normalized to Gapdh. Data are mean±s.d. from three independent experiments. All the primers used are listed in Table S5.

Total RNAs from 2n ESCs, n ESCs, 2n NPCs and the derived n NPCs were extracted using Trizol Reagent (Invitrogen). The global gene-expression was analysed using BGISEQ-500 according to the manufacturer's instructions. Reads allowing two mismatches were mapped to the genome assembly from UCSC (Rhesus8) (Zimin et al., 2014) using STAR v2.4 (Dobin et al., 2013). Feature Counts (Liao et al., 2014) was used to generate the abundance for each gene on default settings. Genes with summed reads larger than 10 across all samples were kept. Scatterplots were generated using R function smooth Scatter with log2-transformed transcripts per million reads. An R function heatmap was used to generate the correlation matrix and the dendrograms.

To introduce gene mutations into haNPCs, we electroporated 5 μg PBase plasmid (Super PiggyBac Transposase PB200PA-1, ABI) and 15 μg piggyBac plasmid (PiggyBac Dual promoter PB513B-1, ABI) into 5×10⁵ haNPCs. The single-cell mixture was electroporated with an appropriate combination of plasmids in 200 μl of resuspension buffer using the NEON transfection system (Invitrogen) at 1400 V for 13 ms with three pulses. For the screening of the target genes of neural toxicants, we treated haNPCs after trapping with 0.8 μM A803467 for 7 days.

Total DNA was isolated from cells using the Universal Genomic DNA Kit (CWBio) and digested with BstYI (Thermo) at 37°C for 16 h, followed by enzyme inactivation at 80°C for 30 min. The ligation reaction was then performed using T4 DNA Ligase (Takara) at 4°C for 16 h. PCR was conducted under the following conditions: 95°C for 5 min, 25 cycles at 95°C for 30 s, 58°C for 30 s and 72°C for 2 min. Terminal replication (first round) was performed at 72°C for 10 min and 95°C for 5 min. This step was followed by 32 cycles at 95°C for 30 s and 60°C for 30 s. The second round of terminal replication was performed at 72°C for 90 s and 72°C for 10 min (Li et al., 2016). Inverse PCR products were sequenced and aligned to the rhesus monkey genome to identify the integration sites. Additionally, splinkerette PCR (Potter and Luo, 2010) and deep sequencing were performed to identify the integration sites under the treatment of A803467. All the primers used are listed in Table S5.

Reads were mapped to the genome assembly from UCSC (Rhesus8) (Zimin et al., 2014) using bowtie2 (v2.2.3) (Langmead and Salzberg, 2012) with default settings. Reads distributions across various regions were summarized using RSeQC (v2.6.1) (Wang et al., 2012). VISITs (v0.22) (Yu and Ciaudo, 2017) was used to estimate the number of insertions for each gene using the annotation from Rhesus8 (Zimin et al., 2014), excluding duplicated and multiple-hit reads. Data were first filtered by removing genes covered by <2 reads and then normalized using relative-log-expression (RLE) from DESeq2 (v1.10.1). Log2-transformed fold changes were further calculated based on normalized data. The coverage tracks were generated using R package GenomeGraphs (v1.30). For the enrichment analysis, gene sets were retrieved from the gene ontology (Ashburner et al., 2000) and KEGG database (Kanehisa and Goto, 2000). The top 100 genes with the most abundant insertions were used. Fisher's exact test followed by Benjamini-Hochberg correction was used to generate the FDR.

To knock out SCN5A or B4GALT6 in diploid NPCs, we electroporated 10 μg sgRNA-Cas9n (D10A) plasmids into 5×10⁵ haNPCs. Briefly, the single-cell and plasmid mixtures were dissolved in 200 μl of resuspension buffer using the NEON transfection system (Invitrogen) at 1300 V for 13 ms with three pulses. After 36 h, we performed FACS to enrich GFP-positive cells for further culturing.

Total DNA was extracted from alive cells after being treated with A803467, including the SCN5A-sgRNA group and B4GALT6-sgRNA group, using the Universal Genomic DNA Kit (CWBio). Then, targeted fragments were amplified using specific primers with Phanta Max Super-Fidelity DNA Polymerase (Vazyme) and purified with a Gel Extraction Kit (CWBio). Purified PCR product was denatured and re-annealed in NEBuffer 2 (NEB) using a thermocycler to form hybridized DNA. Hybridized PCR products were digested with T7EN1 (NEB) for 37°C for 30 min and separated by 2.5% agarose gel. To sequence the T7EN1 cleavage products, we transformed them into E. coli and randomly picked individual bacterial colonies and sequenced them using M13 primer. The primers for amplifying SCN5A and B4GALT6 targeted fragments are listed in Table S5.

We thank Dr Bingjun He from Nankai University for his kind help with the electrophysiological aspects of the study and Dr Xudong Wu from Tianjin Medical University for FACS service.