BMP inhibition promotes the differentiation of human ESCs and iPSCs into telencephalic dorsal NPCs

Several commercial products are available to induce neural differentiation, and we tried one of these methods, according to the manufacturer’s instructions, to generate human iPSC-derived dorsal telencephalic NPCs (Additional file 4: Figure S1A). To verify the presence of dorsal NPCs, we performed immunocytochemical staining for NPC markers (Nestin, PAX6, and SOX1) and SOX2, which is initially expressed in iPSCs as a pluripotency marker but is also required during neural induction [3133]. The NPCs displayed positive Nestin, SOX1, and SOX2 staining; however, the key dorsal NPC marker PAX6 was absent (Additional file 4: Figure S1B), suggesting that the NPCs generated were not the dorsal subtype. In addition, we performed qRT-PCR to determine the expression levels of several NPC markers and observed low PAX6 expression concordant with our immunostaining (Additional file 4: Figure S1C).

Due to the proprietary nature of the formulations of these commercial reagents, our ability to optimize differentiation was limited. Therefore, we adapted a protocol for differentiating mouse ESCs into dorsal NPCs [26] for human ESCs (H1) and iPSCs (YH10, BJ4, 1323–2). All of the undifferentiated human cell lines were maintained on Vitronectin-coated plates with E8 media and expressed the key pluripotency markers OCT4, SOX2, KLF4, LIN28, and NANOG prior to neural induction (data not shown). To induce NPC differentiation, the hESC and hiPSC colonies were dissociated into single cells and allowed to aggregate into EBs in EB1 medium containing dorsomorphin (Fig. 1a). Dorsomorphin has a potent inhibitory effect on the BMP type I receptors ALK2, ALK3, and ALK6, which has been shown to improve neural differentiation efficiency to drive dorsal specification [20, 23, 24, 34]. The EBs were cultured in suspension for 10 days to allow the formation of self-organized neuroepithelium that will give rise to dorsal NPCs. The EBs were then dissociated into single cells and cultured as an adherent monolayer in the presence of FGF2 without dorsomorphin. A small portion of the neuroepithelial cells organized into neural rosettes, which we were able to select using the STEMdiff™ Neural Rosette Selection Reagent. Based on immunocytochemical staining and qRT-PCR analyses, the hESC and hiPSC-derived NPCs express the dorsal neural progenitor markers Nestin, PAX6, SOX1, and SOX2 (Fig. 1bd, Additional file 5: Figure S2). Dorsal forebrain identification of these NPCs was further confirmed with the expression of the dorsal marker FOXG1, and the absence of the ventral marker DLX2, in the nucleus (Fig. 1b). We observed a high percentage of Nestin-positive cells in all of the NPC lines derived from human ESC and iPSC types during the early stages of differentiation at passage 2 (Fig. 1c), followed by increasing PAX6 and SOX1 expression at passage 5. At passage 2, the percentages of PAX6-positive and SOX1-positive cells varied among the hESC and hiPSC-derived lines, suggesting that there are cell line-specific differences during the early stages of NPC differentiation. However, by passage 5 the average percentages of hiPSC-derived NPCs expressing Nestin, PAX6, and SOX1 were 98.77% ± 3.02%, 80.51% ± 14.37%, and 72.08% ± 16.42%, respectively. We observed similar efficiencies for the hESC-derived NPC line, suggesting that we were able to obtain highly homogeneous populations of dorsal NPCs from all of the human-derived cell lines.

To determine the neural differentiation efficiency for each cell line, we quantified the expression levels of several pluripotency (OCT4, SOX2), mesodermal (Brachyury), endodermal (GATA4), and dorsal NPC (Nestin, PAX6, SOX1, BTG2) genes using qRT-PCR throughout the process of neural differentiation (Fig. 1d, Additional file 5: Figure S2A). We confirmed the downregulation of OCT4 in the EBs by day 4 of differentiation in the hESC and hiPSC-derived lines, and it was undetectable in the NPCs (Fig. 1d, Additional file 5: Figure S2A). Although the expression of SOX2 is associated with pluripotency, it has also been shown to be involved in NPC differentiation [31, 33]. We observed increased SOX2 levels in the EBs followed by a decreased and steady expression pattern after the early differentiation stage (Fig. 1d, Additional file 5: Figure S2A). This expression profile is in accordance with the proliferative and multipotent nature of NPCs and with the cell populations expected to be present at the different time points, as demonstrated by previous studies [32, 35]. The expression of Brachyury was present in the EBs, and after neural rosette selection Brachyury was significantly decreased in the NPCs (Fig. 1d, Additional file 5: Figure S2A). GATA4 expression persisted in the NPCs derived from the hESCs and hiPSCs (Fig. 1d, Additional file 5: Figure S2A), but more so from the hiPSC lines. We observed decreased GATA4 in NPCs generated from the double BMP/SMAD protocol described in the next section, suggesting that inhibition of BMP alone may not be sufficient to fully suppress endodermal GATA4. However, the expression of GATA4 did not appear to interfere with neural differentiation as we observed increases in the NPC-associated genes Nestin, PAX6, SOX1, and BTG2 by passage 5 that persisted at high levels through passage 10 in our analysis (Fig. 1d, Additional file 5: Figure S2A). The overall pattern of Nestin, PAX6, and SOX1 mRNA expression was consistent with previous studies [36] despite our observation of higher initial Nestin protein expression via immunocytochemical staining during the earlier stages of NPC differentiation (Fig. 1c).

Double BMP/SMAD inhibition promotes dorsal NPCs derived from human ESCs and iPSCs

A previously published method to generate NPCs was tested [23] and slightly modified in order to obtain a higher percentage of PAX6-positive cells. The same hESC and hiPSC lines mentioned earlier were used for NPC differentiation. Briefly, the cells were cultured as EBs in EB2 medium, supplemented with two SMAD inhibitors: dorsomorphin (BMP receptor inhibitor) and SB431542 (TGF-β inhibitor). SB431542 blocks phosphorylation of the ALK4, ALK5, and ALK7 receptors, and the synergistic effects of dorsomorphin and SB431542 selectively inhibit the TGF-β/Activin/NODAL signaling pathway to promote neuroectodermal differentiation [37]. Changes to the original protocol are described in Methods, and the protocol timeline is displayed in Fig. 2a. The similarities and differences among the four described protocols are presented in Additional file 1: Table S2. The EBs were seeded and cultured on Geltrex-coated plates, which allowed the cells to organize into a monolayer of neural rosettes (Fig. 2a). To test for neural differentiation efficiency, we performed immunostaining for known NPC markers, and by passage 5 we observed Nestin-positive (97.17% ± 5.17%), PAX6-positive (64.52% ± 19.79%), and SOX-positive (63.62% ± 16.51%) cells among the hiPSC-derived NPC lines (Fig. 2b, c). The percentage of PAX6-positive cells indicates that the population of NPCs was highly homogeneous with a major component of dorsal NPCs. Furthermore, these NPCs express FOXG1, but not DLX2, in the nucleus, which further supports their dorsal forebrain identity (Fig. 2b). In addition, we performed gene expression qRT-PCR analysis on the human ESCs and iPSCs, EBs, and NPCs at passages 2, 5, and 10 (Fig. 2d, Additional file 5: Figure S2B). Similar to our BMP inhibition protocol, we observed downregulation of OCT4 by day 4 in the EBs and total repression by passage 2 in the derived NPCs. Transient expression of Brachyury and GATA4 was present in the EBs but became undetectable in the NPCs by passage 5 for all cell lines. Concurrently, the expression of neuroectodermal genes increased throughout EB formation and throughout NPC differentiation and was comparable for the hESC and hiPSC lines, suggesting that this method, in addition to our single BMP inhibition method, is robust for generating dorsal NPCs (Fig. 2d, Additional file 5: Figure S2B).

Mouse ESCs readily differentiate into Pax6-positive dorsal NPCs

Mouse ESCs were derived from 129/Sv6 wild-type blastocysts, and the expression of pluripotency markers were confirmed (Additional file 6: Figure S3) prior to dorsal neural progenitor differentiation. Adherent mESC colonies were dissociated into single cells and allowed to aggregate to form EBs over a period of 7 days in EB1 medium without dorsomorphin (Fig. 3a). The EBs were then dissociated and plated as a monolayer to allow neural rosettes to organize. The neural rosettes were isolated and enriched over two passages using the STEMdiff™ Neural Rosette Selection Reagent. We confirmed the efficiency of generating dorsal NPCs through the positive expression of the neural progenitor markers (Nestin, PAX6, SOX1) and a dorsal forebrain marker (FOXG1), as well as the absence of nuclear DLX2 expression, which is a ventral forebrain marker (Fig. 3b, c). Similar to the results described for the human-derived NPCs, we confirmed the downregulation of pluripotency and nonectodermal genes during the multiple stages of neural induction along with the upregulation of neuroectodermal genes in the NPCs by qRT-PCR (Fig. 3d). Overall, this method is effective in generating dorsal PAX6-positive NPCs from mouse ESC lines.

Dorsal NPCs generated from single and double BMP inhibition protocols have similar NPC marker signature and proliferation

In order to establish whether the two described protocols for human cell lines are significantly different, we compared the results of our immunofluorescence staining from YH10, BJ4, and 1323–2-derived NPC lines at passage 5 after differentiation (Figs. 1c and 2c, Additional file 1: Table S2 and Additional file 3: Table S3). Overall, there was no significant difference in the expression of NPC markers (F test for variance), except for cell line BJ4 which showed higher PAX6 and SOX1 expression levels using the BMP inhibition protocol (Fig. 1c). These differences may be due to cell line-specific characteristics, but our results suggest that both of the described protocols are suitable for human NPC derivation.

A key feature of NPCs is their ability to rapidly and consistently proliferate. The human and mouse cell lines from all of the described NPC differentiation protocols showed similar expression levels of the proliferative marker Ki67 (Additional file 7: Figure S4A–C). Depending on the NPC subtype, the average population doubling time (PDT) can range from 1.2 to 2.1 days. The NPCs generated with the BMP inhibition and the double BMP/SMAD inhibition protocols displayed an average PDT of 1.61 and 1.65 days, respectively (Additional file 7: Figure S4D), and this difference was not statistically significant. In contrast to these results, the NPCs generated using the commercial protocol showed a significantly different PDT (2.29 days, p < 0.0001), mostly due to a delayed doubling time in later passages (Additional file 7: Figure S4D).

Dorsal NPCs differentiate into a highly homogeneous population of forebrain cortical neurons

The expression levels of dorsal NPCs markers, such as PAX6, SOX1, and FOXG1, confirmed the forebrain identify of our NPCs. However, to demonstrate that the NPCs behaved liked functional progenitors, we differentiated the NPCs into mature cortical neurons. We adopted a recently published method [


], in which a heterogeneous population of NPCs was cultured in neuronal inducing conditions for 4–10 weeks, leading to electrophysiologically mature neuronal networks [


]. We performed the differentiation protocol using human 1323–2 NPCs derived using the BMP/SMAD inhibition protocol (Fig.


) and mouse WT3 NPCs (data not shown), and cultured them for 35 days before performing immunostaining. To characterize the population of mature neurons obtained, we used a forebrain telencephalic marker, FOXG1; a superficial cortical layer (II–IV) marker, BRN2; a deep cortical layer (V and VI) marker, TBR1; a neuronal nuclear marker, NeuN; a dendritic marker, MAP2; and astrocyte markers GFAP, SOX9, and GLT1. The human NPCs successfully differentiated into a population of almost 100% cortical forebrain neurons, as confirmed by the presence of staining for the cortical markers and the absence of staining for all three astrocyte markers (Fig.


). The mouse NPCs, although plated at a lower density, were also able to differentiate into neurons that were TBR1-positive while lacking expression of GFAP and SOX9 (Additional file


: Figure S5).

Fig. 4

Human NPCs differentiate into mature neurons. Immunofluorescence staining of differentiated neurons derived from human dorsal NPCs (1323–2 line, day 35 after differentiation) for mature cortical neuronal markers expressed in the nucleus (BRN2, TBR1, NeuN) and cytoplasm (MAP2), glial markers (SOX9, GFAP, GLT1), and dorsal forebrain marker (FOXG1). Nuclei stained with DAPI, shown as an overlay over brightfield images. The merge is an overlay of the neuronal and glial markers. Scale bar 100 μm. DAPI 4,6′-diamino-2-phenylindole