Rapid, efficient, and simple motor neuron differentiation from human pluripotent stem cells

By modifying our previously established method for deriving neural stem/progenitor cells (NS/PCs) as neurospheres from hPSCs through embryoid body (EB) formation [6, 7], we established a rapid neural differentiation protocol from hPSCs (Fig. 1a). Because the derivation of NS/PCs from EBs using the previously established method takes 1 month, we first utilized dual SMAD inhibition to facilitate the neural differentiation of EBs [8]. KhES1 human embryonic stem cells [9] were detached from the feeder layer en bloc and cultured in suspension to form EBs using a BMP inhibitor (3 μM dorsomorphin) and a TGFβ inhibitor (3 μM SB4315342) during the first three days of differentiation (DS) (from day 1 to day 4). Although this dual SMAD inhibition slightly increased the expression of neural markers (PAX6 and SOX1) compared with untreated control cells, it was not sufficient to derive neural progenitors from EBs within two weeks of differentiation (Fig. 1b). Thus, to further facilitate neural differentiation, we added the GSK3β inhibitor (3 μM (2′Z, 3′E)-6-bromoindirubin-3′-oxime (BIO)) during the same period, in addition to dual SMAD inhibition (DSB). As a result, on day 8, the expression of PAX6 and SOX1 was significantly increased compared with control cells, by 3.7 ± 0.4-fold and 138 ± 34-fold, respectively (Fig. 1b). Moreover, the expression of the proneural gene NGN2 was increased by 68 ± 16-fold after 14 days of differentiation compared with untreated control cells. These results suggest that GSK3β inhibition has a positive effect on the neural differentiation of hESCs in our differentiation system, possibly by activating the canonical Wnt pathway through β-catenin. To drive differentiation into motor neurons, we added retinoic acid (RA) (from day 2 of EB formation) and purmorphamine, which activates the sonic hedgehog (SHH) signaling pathway (from day 4 of EB formation), to confer caudal and ventral regional identities, respectively, upon hESC-derived NS/PCs. On day 14 of differentiation, the EBs expressed not only markers for the neural progenitors PAX6 and SOX1 but also additional transcription factors expressed in motor neuron progenitors, including OLIG2 and NGN2 (Fig. 1c). Subsequently, the EBs were dissociated into single cells and were adherently differentiated into neurons via monolayer culture in motor neuron medium (MNM). Within 1 week of adherent differentiation, HB9 and ISL-1, markers for somatic motor neurons, were significantly upregulated (Fig. 1c). Consistent with these results, immunocytochemical analysis revealed the efficient induction of HB9⁺/ISL-1⁺/βIII-Tubulin⁺ motor neurons on day 7 of monolayer differentiation (HB9, 49.1 ± 3.3 %, ISL-1, 47.9 ± 2.0 %, n = 3, Fig. 1d, e). Thus, we have established a rapid, efficient, and simple culture protocol that achieves motor neuron differentiation in only 2 weeks.

To investigate the effects of the long-term culture of hESC-derived motor neurons, we examined temporal alterations of the expression of several motor neuron markers during 4 weeks of monolayer differentiation in MNM. Based on immunocytochemical analysis, the proportion of HB9⁺ and ISL-1⁺ cells did not change greatly and consistently comprised approximately 50–60 % of the total cells (Fig. 2a, b). By contrast, the proportion of cells positive for ChAT, a marker of mature motor neurons, was 54.6 ± 4.5 % at the end of one week and increased to 68.1 ± 4.1 % by 4 weeks of monolayer differentiation. Moreover, after 4 weeks of monolayer differentiation, the somas of the motor neurons had grown larger and were stained more clearly for ChAT compared with those observed after 1 week of monolayer differentiation (Fig. 2a, insets). Quantitative RT-PCR analysis revealed that the expression of genes known to be expressed in neural progenitor cells (PAX6, SOX1, NGN2, OLIG2, and NKX2.2) decreased during 4 weeks of monolayer differentiation, whereas markers of motor neurons (HB9, ISL-1, and ChAT) were increased during the first 2 weeks of differentiation, followed by a gradual decrease during the subsequent 2 weeks (Fig. 2c). To address the discrepancy in the time course of motor neuron marker expression between the immunocytochemistry and quantitative RT-PCR analyses, we also performed a western blot analysis (Fig. 2d, e). We found that the expression of the ChAT protein significantly increased during 4 weeks of differentiation, whereas the HB9 protein level was not significantly altered during this period (Fig. 2e). The ISL-1 protein was significantly decreased during monolayer differentiation as observed in the quantitative RT-PCR analysis (Fig. 2c). These results suggest that the expression of HB9 and ChAT might be regulated post-transcriptionally during the culture of hESC-derived motor neurons and that 4 weeks of monolayer differentiation facilitates the maturation of hESC-derived motor neurons.

To examine the functional properties of hESC-derived motor neurons, we co-cultured hESC-derived motor neurons with skeletal muscle myotubes derived from the human myoblast cell line Hu5/E18 [10, 11]. KhES1-derived EBs containing motor neuron progenitors were partially dissociated and plated onto human myotubes differentiated from Hu5/E18 in MNM for 3–4 days. hESC-derived motor neurons labeled with βIII-Tubulin contacted myotubes, formed end plate-like structures, and induced clustering of acetylcholine receptors, which was visualized using Alexa 555-conjugated α-Bungarotoxin (Alexa 555-α-BTX) (Fig. 3a, b). These results suggest that motor neurons derived in our system are competent to form functional neuromuscular junctions (NMJs) in vitro.

We further examined whether this motor neuron differentiation system is applicable to hiPSCs. Three hiPSC clones (201B7, TIGE-9, and YFE-16) were differentiated into motor neurons following the same protocol (Fig. 4). 201B7 was kindly provided by Dr. Yamanaka [1], and TIGE-9 and YFE-16 were established in our laboratory from fibroblasts from adult males (36 and 24 years of age, respectively) by introducing OCT4, SOX2, KLF4, L-MYC, LIN28, and shTP53 via episomal vectors; the last two clones were subjected to analysis of pluripotent marker expression via immunocytochemistry and quantitative RT-PCR, silencing of episomal transgenes, karyotype analysis, and teratoma formation capacity testing (Additional file 1: Figure S1). All of these clones efficiently differentiated into HB9⁺ and ISL-1⁺ motor neurons that expressed the mature motor neuron marker ChAT by 4 weeks of monolayer differentiation (Fig. 4a). The proportions of HB9⁺ and ISL-1⁺ cells 1 week after monolayer differentiation of dissociated EBs derived from the three hiPSC clones were similar to those obtained from KhES1 hESCs, at approximately 40-50 % (Fig. 4b). We also examined the time course of motor neuron marker expression via quantitative RT-PCR and western blotting and confirmed similar expression profiles in all of the hiPSC clones to those observed in KhES1 cells (Fig. 4c-e). These results suggest that our differentiation protocol is applicable to hiPSCs, thereby minimizing clonal variation among the hPSC clones.

Although our differentiation protocol achieved rapid and efficient differentiation of motor neurons from hPSCs, we observed significant cell death during EB formation after the addition of 3 compounds (DSB), which may have been caused by the cytotoxicity of the small-molecule compounds (Fig. 5a). To avoid this massive cell death and more efficiently obtain motor neuron progenitors for further analysis, we examined the effects of other compounds that inhibit BMP and GSK3β on the differentiation of 201B7 iPSCs. We first replaced BIO with CHIR99021, a more specific and less toxic inhibitor of GSK3β [12], and observed a significant reduction in cell death during EB formation (Fig. 5a). Treatment of EBs with CHIR99021 in combination with DS (DSC) resulted in a 1.79 ± 0.50-fold increase in the cell number obtained from EBs after 14 days of differentiation compared with those treated with DSB (Fig. 5b). We also examined the effects of LDN193189, a more specific and less toxic inhibitor of BMP signals [13]. When EBs were treated with LDN193189, SB4315342, and CHIR99021 (LSC), we observed a 1.54 ± 0.43-fold increase in the number of cells obtained from EBs after 14 days of differentiation compared with those treated with DSC, but the difference was not statistically significant (Fig. 5c). No significant differences in the differentiation of motor neurons were observed among the populations subjected to the DSB, DSC, and LSC treatments. The proportions of HB9⁺ motor neurons under DSC and LSC conditions were 37.9 ± 3.4 and 39.4 ± 1.1 %, respectively (Fig. 5d, e), and the results of time-course analysis of the expression of motor neuron markers did not differ significantly among the DSB, DSC, and LSC treatments (Fig. 4a, b and 5d, e). Together, these results suggest that both the DSC and LSC protocols are useful in practice for obtaining a greater number of motor neuron progenitors and that the inhibition of BMP and TGFβ signals combined with inhibition of GSK3β, which activates canonical WNT signaling, have positive effects on the induction of neural progenitors, excluding the off-target effects of these compounds.

Although our method achieved efficient motor neuron derivation, the proportion of HB9⁺ motor neurons was approximately 40–50 %, and other cell types persisted in culture, presenting an obstacle for the analysis of motor neuron-specific biological events. To address these issues, we constructed an HB9 ^e438 ::Venus reporter lentivirus, in which the Venus fluorescent protein (an altered yellow fluorescent protein; [14]) is expressed under the control of the 438-base pair HB9 enhancer element (HB9 ^e438 , [15]) and the human β-globin minimal promoter. After 3–4 days of monolayer differentiation, KhES1-derived motor neurons were infected with the HB9 ^e438 ::Venus reporter lentivirus and were observed to extend Venus⁺ neurites 24 h after lentivirus infection (Fig. 6a). At one, two, or four weeks after lentivirus infection, cells were fixed and subjected to immunocytochemistry to detect HB9 and Venus (using an anti-GFP antibody) (Fig. 6b). Most Venus⁺ cells were also positive for HB9 one week after lentiviral infection (HB9/Venus = 84.1 ± 2.4 %, Fig. 6c), suggesting that the expression of the Venus fluorescent protein accurately reflects HB9 expression. Many cells were also positive for another motor neuron marker, ISL-1, after 1 week (ISL-1/Venus = 39.5 ± 13.2 %, Fig. 6c). Moreover, after 4 weeks of monolayer differentiation, most Venus⁺ cells were also positive for ChAT (ChAT/Venus = 83.4 ± 1.7 %, Fig. 6c). These results suggest that the HB9 ^e438 ::Venus lentivirus reporter allows accurate visualization of live motor neurons derived from hPSCs.

To further confirm the relevance of this reporter, we examined the expression of the HB9 ^e438 ::Venus lentivirus through fluorescence-activated cell sorting (FACS) one week after the infection of 201B7-derived motor neurons. We divided the cells into 4 fractions: a negative fraction (gated by the fluorescence intensity of uninfected controls, gray peak) and 3 positive fractions (green peak), where the positive cells were equally subdivided into low-positive, middle-positive, and high-positive fractions (Fig. 6d). The cells in each of the 4 fractions were sorted and subjected to quantitative RT-PCR analysis to detect the expression of Venus mRNA and the motor neuron markers HB9, ISL-1, and ChAT (Fig. 6e). We observed a significant increase in the expression of Venus, HB9, and ISL-1 in the high-positive fraction (p < 0.05, Kruskal-Wallis test) and a tendency for the expression of ChAT to increase. The expression levels of HB9 and ISL-1 in the high-positive fraction were 12.4 ± 5.4-fold and 6.4 ± 2.6-fold higher than in the negative fraction. We also examined the expression of Venus fluorescent protein from a reporter lentivirus containing the same vector backbone, but without the HB9 ^e438 enhancer element (background reporter lentivirus; Yellow peak in Fig. 6d), and found that the fluorescence observed in the high-positive fraction of HB9 ^e438 ::Venus reporter-transduced cells overlapped minimally with the fluorescence generated by the background reporter. Together, these results suggested that the high-positive fraction was a bona fide “motor neuron fraction” and that this approach can be used for the purification of hPSC-derived motor neurons.