Neural Stem Cells Directly Differentiated from Partially Reprogrammed Fibroblasts Rapidly Acquire Gliogenic Competency


  • Author contributions: T.M. and K.Y.: collection and assembly of data, data analysis and interpretation, and manuscript writing; M.T., C.F., S.O., and Y.M.: collection and assembly of data; Y.T.: conception and design; M.N.: conception and design and financial support; H.O. and W.A: conception and design, financial support, administrative support, manuscript writing, and final approval of manuscript. H.O. and W.A. contributed equally to this article.

  • Disclosure of potential conflicts of interest is found at the end of this article.

  • First published online in STEM CELLSEXPRESS August 23, 2011.


Neural stem cells (NSCs) were directly induced from mouse fibroblasts using four reprogramming factors (Oct4, Sox2, Klf4, and cMyc) without the clonal isolation of induced pluripotent stem cells (iPSCs). These NSCs gave rise to both neurons and glial cells even at early passages, while early NSCs derived from clonal embryonic stem cells (ESCs)/iPSCs differentiated mainly into neurons. Epidermal growth factor-dependent neurosphere cultivation efficiently propagated these gliogenic NSCs and eliminated residual pluripotent cells that could form teratomas in vivo. We concluded that these directly induced NSCs were derived from partially reprogrammed cells, because dissociated ESCs/iPSCs did not form neurospheres in this culture condition. These NSCs differentiated into both neurons and glial cells in vivo after being transplanted intracranially into mouse striatum. NSCs could also be directly induced from adult human fibroblasts. The direct differentiation of partially reprogrammed cells may be useful for rapidly preparing NSCs with a strongly reduced propensity for tumorigenesis. STEM CELLS2012;30:1109–1119


The direct reprogramming of somatic cells into other cell lineages facilitates the rapid and efficient production of target cells, which may be used for cell therapies in the future. By introducing tissue-specific transcription factors, neurons [1, 2], cardiomyocytes [3], and cartilage [4] can be generated from skin fibroblasts. However, these transdifferentiated cells have a limited ability to proliferate and differentiate into multiple progenitors. In contrast, partially reprogrammed cells induced using Yamanaka factors (Oct4, Klf4, Sox2, and c-Myc) [5, 6] or Oct4 alone [7] can generate tissue-specific stem/progenitor cells under the appropriate culture conditions.

Although reports claim that these cells are not generated from induced pluripotent stem cells (iPSCs) but from intermediate metastable states, it is difficult to exclude the possibility that some of these cells transiently go through a pluripotent stage during the reprogramming and differentiation processes. Among iPS-derived tissue stem/progenitor cells, residual pluripotent cells that are resistant to forced differentiation give rise to teratomas with three germ layers [8]. We previously showed that safe mouse iPS clones must be selected to achieve efficient functional recovery without tumor formation in spinal cord injury (SCI) model mice by iPS-derived neural stem cell (NSC) transplantation [9]. However, we also showed that more than 80% of iPS clone-derived NSCs generate detectable teratomas after intracranial injection, when the iPS clones are established from adult skin fibroblasts [8]. “Safe” iPS clones must be carefully selected to prevent teratoma formation.

In the differentiation of embryonic stem cells (ESCs)/iPSCs into NSCs, neurogenic NSCs appear first, and they develop into gliogenic NSCs during the recurrent passages in vitro [10]. Such gliogenic NSCs are effective for treating SCI model animals, but neurogenic NSCs are not [9, 11]. For NSCs transplantation to be effective in SCI patients, it must be performed before the chronic phase [12]; therefore, the rapid induction of gliogenic NSCs with reduced tumorigenicity will be required for future treatments involving autograft transplantation.

In this report, we introduce a novel culture system that generates NSCs from adult mouse fibroblasts that were partially reprogrammed by introducing the four Yamanaka factors, and that directs their differentiation into neural lineages. Interestingly, the NSCs derived by this method differentiated into both neuronal and glial cells, even at early passages. We suggest that these gliogenic NSCs, which differentiate more rapidly than ESC/iPSC-derived NSCs, were derived from partially reprogrammed cells, because dissociated ESCs/iPSCs did not form neurospheres under the same culture conditions. Furthermore, we induced human NSCs directly from adult fibroblasts using this method.


Cell Culture

Adult skin fibroblasts were obtained from 8-week-old C57BL/6J mice. To obtain these cells, the skin was peeled from the body of adult mice, minced into 5-mm pieces, placed on culture dishes, and incubated in Dulbecco's modified Eagle's medium (DMEM), containing 10% fetal bovine serum (FBS), 50 U penicillin, and 50 mg/ml streptomycin. Cells that migrated out of the skin pieces were trypsinized and transferred to new plates. We used the adult mouse fibroblasts at passage 3-5 for direct neural induction.

Retroviral Infection

Retroviral infection was performed as described previously [13]. Briefly, the day before transfection, Plat-E cells were seeded at 3.6 × 106 cells per 10-cm dish. The next day, pMX-based retroviral vectors were introduced into the Plat-E cells by the Fugene 6 transfection reagent (Roche, Penzberg, Germany, The virus-containing supernatant was used to infect the fibroblasts, seeded at 8 × 105 cells per 10-cm gelatin-coated dish. The infected cells were collected and subjected to neural induction by suspension culture.

Neurosphere Culture

Mouse fibroblasts transduced with Oct4-, Klf4-, Sox2-, and c-Myc-expressing retroviruses were cultured in DMEM containing 10% FBS for 4 days, then trypsinized and suspended in media hormone mix (MHM) [14, 15] medium with leukemia inhibitory factor (LIF) (1,000 U/ml) and 20 ng/ml basic fibroblast growth factor (bFGF) (PeproTech Inc., Rocky Hill, NJ, for 14 days. In some cases, the neurospheres were dissociated and cultured at 5 × 104 cells per microliter in MHM with FGF2 and/or epidermal growth factor (EGF) (PeproTech). To assay the cells' differentiation, neurospheres were plated on poly(L-ornithine)/fibronectin-coated chambers and allowed to differentiate without growth factors for 7–14 days. A pCPT-cAMP (C3912; Sigma-Aldrich, St. Louis, MO, was applied at 100 μM. For the comparison between two cAMP analogs, 1, 10, 100, 1,000, and 5,000 μM of pCPT-cAMP and 1, 10, 50, and 100 μM of 8-pCPT-2-O-Me-cAMP (C041; Biolog, Bremen, were applied to culture medium. For the direct induction of NSCs from human fibroblasts, fibroblasts were purchased from Cell Applications, Inc. or obtained from skin biopsies only after both the approval of the study protocol by the Ethical committee of the Keio University (No. 20-16-16) and the written informed consent of each patient (Supporting Information Table S4). The induction protocol was principally same as the mouse directly induced NSC (diNSC) induction, except that the adherent culture period was 6 days long.

Time-Lapse Observation of Neurosphere Formation

To evaluate the presence of Nanog-green fluorescent protein (GFP)-positive cells throughout the neurosphere culture period, we analyzed time-lapse videos collected in a Nikon BioStation (Nikon, Tokyo, Japan, CT incubator equipped with a camera for video imaging. For these experiments, adult Nanog-GFP fibroblasts were collected 4 days after retroviral introduction of KOSM reprogramming factors and were subjected to suspension culture. During the culture period, the fibroblasts were filmed for 14 days continuously in the BioStation CT [16].

Bisulfite PCR

The genomic DNA was extracted with a Qiagen DNeasy kit (Qiagen, Venlo, Netherlands, The purified genomic DNA was denatured and converted with a Qiagen Epitect kit. The bisulfite-modified DNA was purified and used as a template for polymerase chain reaction (PCR). The PCR products were subcloned into PT7blue (Novagen, Darmstadt, Germany,, and individual clones were randomly selected for DNA sequencing with U19 primers for each gene. The PCR primers are listed in Supporting Information Table S1.

Immunocytochemical and Immunohistochemical Analyses

Immunocytochemical and immunohistochemical analyses for cultured cells were performed as described previously [8, 9] with the antibodies listed in Supporting Information Table S2. For statistical analysis of the immunocytochemical results, at least 40 colonies were examined. In the immunohistochemical analysis, the phenotypes of the grafted cells were assessed by fluorescent double immunostaining with antibodies against Venus and one of the cell-type-specific markers listed in Supporting Information Table S2. Images were obtained by fluorescence microscopy (Axioplan 2; Carl Zeiss, Thornwood, NY,, and BZ-9000; Keyence, Woodcliff Lake, NJ, or confocal microscopy (LSM700; Carl Zeiss).

Flow Cytometric Analysis

The Nanog-enhanced GFP (EGFP)-positive cells in neurospheres were subjected to flow cytometric analysis on a FACS Calibur. The percentage of GFP-positive cells to the total number of living cells, which were selected by the absence of propidium iodide, is presented. GFP-positive and -negative cells were sorted on a FACS Vantage.

Microarray Analysis

Total RNA isolation was performed with a Qiagen RNeasy Kit ( DNA microarray analysis using Affymetrix Gene-Chip technology was performed as described previously [17–19]. Briefly, 100 ng of total RNA was used as a template for cDNA synthesis, and biotin-labeled cRNA was synthesized with a 3′ IVT Express Kit (Affymetrix, Santa Clara, CA, After generating the hybridization cocktails, hybridization to the DNA microarray (GeneChip Mouse Genome 430 2.0 Array; Affymetrix) [20] and fluorescent labeling were performed. The microarrays were then scanned with a GeneChip Scanner 3000 7G System (Affymetrix). Data analysis was carried out using Expression Console 1.1 (Affymetrix). Signal detection and quantification were performed using the MAS5 algorithm with default settings. Global normalization was performed so that the average signal intensity of all probe sets was equal to 100. For the clustering analysis, the signals were normalized and calculated by Cluster 3.0 [21], and the scores were visualized by Java Treeview [22]. The principal component analysis (PCA) was carried out by Spotfire DecisionSite 9.1.2 using normalized data.

Lentivirus Production and Infection of diNSCs

A self-inactivating HIV-1-based lentivirus vector, pCSII-EF-MCS-IRES2-Venus 4, was used to label NSCs for transplantation into the brain of C57/B6j mice. For lentivirus production, HEK-293T cells were transfected with pCSII-EF-MCS-IRES2-Venus, pCAG-HIVgp, and pCMV-VSV-G-RSV-Rev, and the conditioned medium containing the virus particles was collected. The virus was concentrated by centrifugation at 125,000g for 1.5 hours at 4°C. The concentrated viruses were added to the culture medium in which diNSCs were being formed from fibroblasts.


The transplantation of neurospheres expressing ffLuc-cp156, a fusion protein of a fluorescent protein Venus and firefly luciferase, which had been introduced by lentiviral infection [23] was performed using a Hamilton Syringe with a stereotaxic injector, as described [24]. The needle of the Hamilton Syringe was inserted into the right striatum (2 mm lateral, 1 mm rostral to bregma; depth, 3 mm from dura) of 8-week-old female C57/b6j mice, and 3 μl of NSC suspension (2 × 105 cells) was injected.

Bioluminescent Imaging

A Xenogen-IVIS 100 cooled CCD optical macroscopic imaging system (SC BIoScience, Tokyo, Japan, was used for bioluminescent imaging (BLI), as reported previously [12].

Statistical Analysis

The statistical significance of variations was evaluated by the unpaired two-tailed Student's t test. All the results are presented as the mean ± SEM.


Generation of diNSCs from Skin Fibroblasts

To obtain partially reprogrammed cells, we introduced Oct4, Sox2, c-Myc, and Klf4 (hereafter, KSOM) into adult mouse skin fibroblasts using retroviral vectors [13, 25]. These cells were then dissociated into a single-cell suspension and cultured for 14 days in the presence of LIF and FGF2 in serum-free medium, to propagate LIF-dependent primitive NSCs [26, 27], which were able to differentiate into FGF2-dependent definitive NSCs (Fig. 1A). Spheres that grew to more than 50 μm in diameter were observed 10 days after retroviral infection (Fig. 1B). To examine whether these neurospheres were clonally derived from NSCs, they were subjected to 14 days of adherent culture without growth factors. More than 60% of the spheres generated cells that expressed markers for neurons (β3-tubulin) or astrocytes (glial fibrillary acidic protein [GFAP]) (Fig. 1C). Thus, we concluded that they were neurospheres derived from NSCs, and termed them “directly induced NSCs” (diNSCs).

Figure 1.

Neurosphere formation from partially reprogrammed fibroblasts. (A): Schematic of the procedure for direct neural induction. Adult mouse fibroblasts were infected with Oct4-, Sox2-, Klf4-, and c-Myc-expressing retroviruses, followed by adherent and suspension culture. (B): Process of neurosphere formation. 4F = KSOM. Scale bar = 100 μm. (C, D): Immunocytochemical analysis of neurospheres derived from adult mouse fibroblasts and ESCs. Scale bar = 100 μm. (E--G): Immunocytochemical analysis of neural cell marker proteins in the differentiated neurospheres derived from adult mouse fibroblasts. Scale bar = 50 μm. (H, I): GFP fluorescence in neurospheres generated from adult P0-Cre/LoxP-EGFP mouse fibroblasts. Phase-contrast (H) and fluorescence (I) micrographs are shown. Scale bar = 100 μm. (J): Number of neurospheres obtained under different culture conditions (n = 3; *, p < .001). (K, L): Immunocytochemical analysis of neural cell marker proteins in the neurospheres derived from adult mouse fibroblasts (K) and ESCs (L). Scale bar = 50 μm. (M): Number of neurospheres formed within various periods of adherent culture (n = 3; *, p < .05; **, p < .005; ***, p < .001). (N): Formation of secondary neurospheres. (O): Immunocytochemical analysis of neural cell marker proteins in the differentiated neural cells derived from secondary neurospheres. Scale bar = 100 μm. (P): Differentiation tendency of the primary and secondary neurospheres formed by direct induction. The frequency of colonies consisting of neurons (β3-tubulin) and/or astrocytes (GFAP) was evaluated by immunocytochemistry and presented as the percentage of total colonies. Abbreviations: bFGF, basic fibroblast growth factor; EGFP, enhanced green fluorescent protein; ESCs, embryonic stem cells; FBS, fetal bovine serum; GFAP, glial fibrillary acidic protein; LIF, leukemia inhibitory factor.

Mouse ESCs can differentiate into NSCs and form neurospheres under similar floating culture conditions [26]. We previously reported that the primary neurospheres derived from mouse ESCs via embryoid bodies (EBs) mainly give rise to neurons [10]. We dissociated the same ESC line, EB3, and cultured the cells using the same protocol as for the diNSCs (from day 4 to day 18). We then differentiated these spheres on chamber slides. Neurospheres were formed similarly, but they generated few astrocytes (Fig. 1D) compared to the neurospheres from diNSCs. The neurospheres prepared from diNSCs could differentiate into neurons (Fig. 1E), astrocytes (Fig. 1F), and oligodendrocytes (Fig. 1G), although no oligodendrocytes were observed in the EB3-derived neurospheres. To detect any non-neural cells in differentiated diNSC-derived spheres, we performed triple-label immunostaining with antibodies against β3-tubulin, GFAP, and Nestin. We found that only 1.8% ± 1.8% of the neurosphere-derived cells did not express any of these markers (Supporting Information Fig. S1A). Therefore, this small percentage of cells was considered to be oligodendrocytes or non-neural cells.

Next, we examined whether diNSCs acquired the characteristics of NSCs during reprogramming. To exclude the possibility that the diNSCs were derived from the sphere-forming multipotent neural crest stem cells that are present in the dermal skin [28–30], we made diNSC-derived neurospheres from adult P0-Cre/Floxed-EGFP mouse fibroblasts, which express EGFP in the neural crest lineage [31, 32]. The diNSC-derived neurospheres established from P0-Cre/loxP-EGFP mice showed no EGFP expression (Fig. 1H, 1I), indicating that the diNSCs did not originate from neural crest cells and were distinct from the multipotent stem cells present in dermal skin.

Further detailed characterization showed that the diNSCs formed neurospheres in the presence of LIF and FGF2. The number of neurospheres was significantly reduced in the absence of LIF, and none was formed in serum-free medium that contained neither LIF nor FGF2 (Fig. 1J). When the fibroblasts used were from adult Nestin-second intronic enhancer-EGFP (Nestin-EGFP) transgenic mice [33], all the spheres derived from the diNSCs were positive for EGFP, indicating that they were neurospheres (Supporting Information Fig. S1B, S1C). To characterize the cells present in diNSC-derived spheres, we performed immunocytochemical analysis of diNSC-derived neurospheres using confocal microscopy. In these neurospheres, all of the cells expressed the neural markers Nestin or β3-tubulin (Fig. 1K) to a similar extent as those derived from ESCs (Fig. 1L). These results suggested that only a few fibroblasts or non-neural cells were present in the floating diNSC-derived neurospheres. The number of neurospheres peaked at 4 days in adherent culture (Fig. 1M).

Dissociated diNSC-derived primary neurospheres generated secondary neurospheres (Fig. 1N) that gave rise to neurons and astrocytes (Fig. 1O). These secondary neurospheres were highly astrogenic (Fig. 1P) and resembled the mature NSCs that appear in late embryonic stages in vivo [34]. These data suggest that diNSCs can self-renew, a characteristic of NSCs, and gradually develop into mature multipotent NSCs in vitro.

Cell Type of Origin and Reprogramming Duration Influence the Differentiation Properties of the diNSCs

We next tested whether the somatic cell of origin or reprogramming procedure influenced the properties of the diNSCs. First, we compared the differentiation properties of neurospheres generated from adult cells grown for various periods in adherent cultures after reprogramming factors had been introduced (Fig. 1M). Although longer periods of adherent culture increased the number of neurospheres (Fig. 1M), these neurospheres were neurogenic and not gliogenic (Fig. 2A). This result suggested that the duration of reprogramming affected the maturity of the diNSCs. We then compared the differentiation properties of diNSCs derived from mouse embryonic fibroblasts (MEFs) and adult mouse fibroblasts. Neurospheres derived from MEFs generated mainly neurons and few astrocytes, as did neurospheres derived from dissociated EB3 cells and 38C2-iPSC (Fig. 2B) [35], whereas those generated from adult fibroblasts generated relatively more astrocytes.

Figure 2.

Properties of directly induced NSCs. (A, B): Effect of the adherent-culture duration (A) and cell source (B) on the differentiation tendency of directly induced neurospheres. Differentiated neurospheres were subjected to immunocytochemical analysis for β3-tubulin (neurons) and GFAP (astrocytes). The frequency of colonies consisting of neurons and/or astrocytes is presented as the percentage of total colonies. (C--F): GFP fluorescence in neurospheres generated from Nanog-EGFP adult mouse fibroblasts (C, D) or MEFs (E, F). Phase-contrast (C, E) and fluorescence (D, F) micrographs are shown. Scale bar = 200 μm. (G): Flow cytometric analysis of the Nanog-EGFP-positive cells in neurospheres established from Nanog-EGFP adult mouse fibroblasts or MEFs (n = 3; *, p < .001). (H): Representative fluorescence-activated cell sorting plots for cells in the neurospheres from Nanog-EGFP MEFs or adult fibroblasts. (I, J): Methylation status of the CpGs in the Stat3 recognition sequence of the GFAP promoter (I) (n = 10; *, p < .005) or in the Oct4 promoter region (J) (n = 10; *, p < .001). The proportion of methylated cytosines is shown. Abbreviations: EGFP, enhanced green fluorescent protein; GFAP, glial fibrillary acidic protein; GFP, green fluorescent protein; iPSC, induced pluripotent stem cell; MEFs, mouse embryonic fibroblasts.

Next, to monitor the residual pluripotent cells, we generated diNSC-derived neurospheres from adult and embryonic fibroblasts established from Nanog-GFP transgenic mice (a gift from Dr. Shinya Yamanaka's laboratory) [35]. Although both types of fibroblasts generated neurospheres, only 12.5% of the neurospheres from adult fibroblasts showed visible Nanog-GFP fluorescence, but 93.9% of the ones from MEFs did (Fig. 2C-2F). Quantitative analysis by fluorescence-activated cell sorting (FACS) showed 10 times more Nanog-GFP-positive cells in the MEF-derived neurospheres than in the adult-derived ones (Fig. 2G, 2H). To determine whether the NSCs were generated from Nanog-GFP-positive or -negative cells, we first examined whether Nanog-GFP-positive cells were present on day 4 by using flow cytometric analysis. Consistent with a previous report [35], no GFP+ cells were found on day 4 (Supporting Information Fig. S2), indicating that sphere-forming cells were not derived from fully reprogrammed iPSCs at the start of the neurosphere culture. However, it is possible that a small number of cells were gradually reprogrammed into fully pluripotent GFP+ cells and then committed to neural lineages that led to neurosphere formation during the floating culture period. Alternatively, a small number of cells in the neurospheres may have retained or acquired the pluripotency in this neurosphere culture system. To investigate whether or not the neurospheres were generated from Nanog-GFP-positive cells, we performed a continuous observation using time-lapse microscopy (Supporting Information Movie S1). We observed more than 100 spheres and found none of the observed neurospheres expressed GFP fluorescence by day 10 of sphere formation. However, 13 of 122 neurospheres contained a small number of visible Nanog-GFP-positive cells by day 14. Therefore, we concluded that diNSCs were generated mainly from nonpluripotent cells in this culture system. However, it is possible that a small number of diNSCs were also generated from pluripotent cells or that a small number of cells within the neurospheres were reprogrammed to be pluripotent cells during neurosphere formation.

We next examined the methylation within a STAT3-binding site in the GFAP promoter region using bisulfite PCR and sequencing, because this site is demethylated gradually in NSCs during development [36, 37]. The diNSC-derived neurospheres established from adult mouse fibroblasts had significantly less methylation (49.2%) in the GFAP promoter region than did adult mouse fibroblasts (78.3%) or neurospheres derived from dissociated 38C2-iPSCs in the presence of LIF and FGF (82.0%) (Fig. 2I). The amount of methylation was also lower than in tertiary neurospheres derived from ESCs via EBs [10, 37]. We also examined the methylation in the promoter region of the pluripotent marker gene, Oct4, because it is demethylated in iPSCs [13, 35]. The Oct4 promoter region was highly methylated in the neurospheres derived from adult fibroblasts compared to the LIF- and FGF2-dependent primary neurospheres from dissociated 38C2-iPSCs (Fig. 2J). These results together suggested that the differentiation properties of diNSCs are determined by the cell source and the culture period after the introduction of pluripotent genes, which could affect the state of reprogramming.

Optimization of Culture Conditions to Increase the Efficiency of diNSC Induction

Although we succeeded in making neurospheres derived from diNSCs, less than 0.01% of the dissociated fibroblasts treated with Yamanaka factors generated neurospheres. To increase the efficiency of neurosphere formation, we used cAMP analog, pCPT-cAMP (Sigma; C3912), which increases the survival of LIF-dependent primitive NSCs by inhibiting their apoptosis [27]. We added pCPT-cAMP at various concentrations to the serum-free medium used for the suspension cultivation step and found that 100 μM of pCPT-cAMP most efficiently induced neurosphere formation (to 0.03%) (Fig. 3A--3D; Supporting Information Fig. S3A). Although the diameters of the neurospheres were similar to those that formed without cAMP (Fig. 3E), the frequency of gliogenic neurospheres was significantly lower than in cultures without pCPT-cAMP (Fig. 3F). Thus, we removed pCPT-cAMP from the medium on day 6, and dissociated the neurospheres to enhance the differentiation of NSCs (Fig. 3A). The diNSC-derived neurosphere formation ratio increased to 2.98% in this condition (Fig. 3D). The addition of pCPT-cAMP for 48 hours and additional passages resulted in increased numbers and frequencies of astrogenic neurospheres derived from the diNSCs (Fig. 3F). These frequencies were similar to those of LIF-dependent neurospheres derived from dissociated ESCs and subjected to directed differentiation [26, 27], even though our starting cells were fibroblasts. Different cAMP analogs are known to activate either the Epac or protein kinase A (PKA) pathway [38, 39]. To determine which pathway was crucial for the increase in neurosphere formation efficiency, we tested the effects of 8-pCPT-2-O-Me-cAMP (Biolog C041), a distinct cAMP analog that activates only the Epac pathway. While pCPT-cAMP increased neurosphere formation efficiency at 100 mM (p < .05), 8-pCPT-2-O-Me-cAMP did not increase the number of spheres at any concentration tested (Supporting Information Fig. S3B). This result suggests that the PKA pathway is involved in the survival of diNSCs.

Figure 3.

Transient pCPT-cAMP treatment facilitates mature neurosphere formation. (A): Schematic of culture method using pCPT-cAMP. pCPT-cAMP was added to the serum-free medium and removed after 2 days in groups C and D. In groups C and D, the neurospheres were subjected to one or two dissociations. 4F = KSOM. (B, C): Neurosphere formation after 14 days of suspension culture without (B) or with (C) pCPT-cAMP (scale bar = 100 μm). (D, E): Average sphere formation percentage and diameter of neurospheres in cultures without pCPT-cAMP (n = 17), with pCPT-cAMP (n = 7), with pCPT-cAMP and one passage (n = 3), and with pCPT-cAMP and two passages (n = 3) (*, p < .05). (F): Differentiation potential of neurospheres under various conditions. Differentiated neurospheres were subjected to immunocytochemical analysis for β3-tubulin (neurons) and glial fibrillary acidic protein (astrocytes). The frequency of colonies consisting of neurons and/or astrocytes is presented as the percentage of total colonies. Abbreviations: bFGF, basic fibroblast growth factor; FBS, fetal bovine serum; LIF, leukemia inhibitory factor.

diNSCs Respond to EGF Treatment in Early Passages

To use iPS-derived NSCs for transplantation, it is important to choose safe iPS clones that do not form teratomas in host animals [8, 9]. We previously reported that the safety of iPSCs can be monitored by the expression of pluripotent markers, and that neurospheres containing less than 0.018% Nanog-GFP-positive cells do not form teratomas after their transplantation [8]. To decrease the Nanog-GFP-positive cells in the diNSC-derived neurospheres, we tried to withdraw LIF from the culture medium during the induction of diNSC-derived neurospheres from the fibroblasts of Nanog-GFP mice. Although the frequency of green neurospheres decreased dramatically with LIF withdrawal, the frequency of neurosphere formation in this condition was as low as 0.02%. To solve this problem, we added EGF to the neurosphere culture, because mature NSCs respond to EGF by forming neurospheres [26, 40]. We used EGF instead of LIF on days 11–15 in the suspension culture period (Fig. 4A). In this condition, the frequency of diNSC-derived neurosphere formation increased to 1%, with reduced visible green fluorescence (Fig. 4B--4E).

Figure 4.

Adult fibroblast-derived directly induced neural stem cells respond to EGF. (A): Schematic of EGF introduction and LIF withdrawal in the suspension culture period. 4F = KSOM. (B, C): Neurospheres established in the presence of LIF for 14 days. (D, E): Neurospheres established in the presence of LIF for 10 days and cultured 4 days with EGF and without LIF. Phase-contrast (B, D) and fluorescence (C, E) micrographs are shown. (F): The percentage of Nanog-EGFP+ cells in directly induced neurospheres under various culture conditions. The red bar indicates 0.018%. (G): Flow cytometric analysis of neurospheres established from Nanog-EGFP mouse adult fibroblasts. Nanog-EGFP+ cells were almost completely lost after LIF withdrawal and EGF treatment. (H): Neurosphere formation under the EGF condition. Five kinds of cells were subjected to direct neural induction with EGF treatment and LIF withdrawal. Abbreviations: APC, adenomatous polyposis coli; bFGF, basic fibroblast growth factor; EGF, epidermal growth factor; EGFP, enhanced green fluorescent protein; ESCs, embryonic stem cells; FBS, fetal bovine serum; iPSCs, induced pluripotent stem cells; MEF, mouse embryonic fibroblast; LIF, leukemia inhibitory factor.

Using culture conditions 1-4 presented in Figure 4A, more than 0.15% of the undifferentiated cells in the neurospheres was Nanog-GFP positive, by flow cytometric analysis (Fig. 4F). However, we observed a great reduction in the percentage of Nanog-EGFP-positive cells in neurospheres cultured with EGF (<0.010%) (n = 4) (Fig. 4F, 4G). These neurospheres differentiated into astrocytes and neurons at similar frequencies as seen using the previous conditions (Supporting Information Fig. S4).

Interestingly, in these experiments, EGF-dependent neurospheres were produced only from adult fibroblasts treated with the four Yamanaka factors. Neither dissociated ESCs/iPSCs nor MEFs treated with Yamanaka factors formed neurospheres in suspension culture with EGF (Fig. 4H). This result suggested that both pluripotent cells and NSCs, which did not respond to EGF, were eliminated during the EGF culture period (days 11–15). NSCs derived from adult fibroblasts, but not from MEFs, survived this EGF-treatment period. This result also suggested that the diNSCs were derived from partially reprogrammed cells and not from fully reprogrammed, pluripotent cells.

Microarray Analysis of diNSC-Derived Neurospheres

Next, we characterized the diNSC-derived neurospheres by microarray analysis and examined the similarity of their global gene expression profile to those of mouse ESCs, iPSCs, and their progenies. Twelve kinds of samples, including four different kinds of diNSC-derived neurospheres, were prepared, as shown in Figure 5A. The expression data, except for those of genes with low expression in all samples, were normalized and subjected to hierarchical clustering (Supporting Information Fig. S5B) and PCA (Fig. 5B). The samples were clustered into four distinct groups (ESCs/iPSCs, primary/secondary neurospheres derived from EBs [10], neurospheres derived from E14.5 mouse embryonic striatum, neurospheres derived from diNSCs, and fibroblasts) in both analyses. Given the clustering of these samples, PC1 and PC3 (Fig. 5B for cluster designations) were most likely to reflect pluripotency and the characteristics of fibroblasts, respectively. A three-dimensional scatter plot of the scores for PC1, PC2, and PC3 revealed that the diNSC-derived neurospheres were located between the ESCs/iPSCs and EB-derived neurospheres, suggesting that the diNSCs display characteristics that are intermediate between pluripotent stem cells and fibroblasts but are distinct from fibroblasts (Fig. 5B).

Figure 5.

Global gene expression analysis of directly induced neurospheres. (A): Samples subjected to microarray analysis. (B, C): Principal component analysis of gene expression data. Color key: yellow circles, directly induced neurospheres; green circles, neurospheres derived from embryoid bodies or the ganglionic eminence of an E14 mouse; red circles, ESCs/iPSCs; blue circles, fibroblasts. (D, upper): Cell sources and culture conditions. (D, lower): Hierarchical clustering analysis using selected genes in directly induced neurospheres established under various conditions. (E, F): Expression of representative pluripotent marker genes (E) and neural marker genes (F). Values are the signal intensity detected by microarray analysis. The sample numbers are indicated in D, upper. The data produced in this article have been deposited in NCBI's Gene Expression Omnibus [41] and are accessible through GEO Series accession number GSE31598 ( =GSE31598). Abbreviations: bFGF, basic fibroblast growth factor; EB, embryoid body; EGF, epidermal growth factor; ESCs, embryonic stem cells; iPSCs, induced pluripotent stem cells; MEFs, mouse embryonic fibroblasts; LIF, leukemia inhibitory factor.

The diNSC-derived neurospheres and EB-derived neurospheres were located considerably close to each other in the scatter plot using PC2 and PC4 (Fig. 5C). In this PCA, the contribution ratios of PC1, PC2, PC3, and PC4 were 26.0%, 18.7%, 11.9%, and 9.9%, respectively. The analysis indicated that at least 30% of the genes expressed in diNSC-derived neurospheres had a similar expression profile to that of EB-derived neurospheres. Interestingly, among the various diNSCs, the EGF-dependent diNSC-derived neurospheres from adult fibroblasts were the most similar to the EB-derived secondary neurospheres, indicating that these diNSCs might be very useful for treating SCI [9].

To monitor the possible contamination of non-neural cells in the diNSC-derived neurospheres, we analyzed the expression of fibroblast specific genes (Col1A2 and proline 4-hydroxylase a2) [42, 43]. The expression levels of these fibroblast markers in diNSC-derived spheres grown in EGF were a fraction (1/200th or less) of the levels seen in fibroblasts (Supporting Information Fig. S5C). Furthermore, expression of several transcription factors that are specific for the mesoderm (FoxP3, T-bet, Pax5, and GATA4) and endoderm (Ngn3 and GATA6) were undetectable. Only small amounts of Brachyury, a mesoderm marker, were expressed in diNSC-derived and ES-derived neurospheres (Supporting Information Fig. S5D). In ES-derived neurospheres, the expression level of FOXA2, an endoderm marker, was very low. However, in diNSC-derived neurospheres, the expression level of FOXA2 was only 10% of the expression level seen in ES-derived neurospheres (Supporting Information Fig. S5D). Thus, we concluded that few non-neural cells were present in diNSC-derived neurospheres. This conclusion was consistent with the results of our immunocytochemical analysis (Fig. 1K, 1L).

To evaluate the influence of cell sources and growth factors included in the culture medium, we analyzed the expression of 106 pluripotent and neuronal marker genes (Supporting Information Table S3) [44–48] in diNSC-derived neurospheres established using the various culture conditions shown in Figure 5D, by hierarchical clustering. In this analysis, the MEF-derived neurospheres constituted one group, and the two adult fibroblast-derived neurospheres were clustered separately. Although the diNSC-derived neurospheres still expressed pluripotent marker genes, including Nanog, Sox2, and Oct4, their expression levels in the adult fibroblast-derived neurospheres were significantly lower than in MEF-derived neurospheres. In particular, the adult fibroblast-derived neurospheres cultured with EGF expressed significantly lower levels of the pluripotent marker genes (Fig. 5E). In contrast, the levels of marker genes of neural stem/progenitor cells, such as Nestin, Sox1, Ascl1, Dcx, Zic, Hes1, and EGFR, were significantly higher in the adult fibroblast-derived neurospheres (Fig. 5F). These data support the idea that adult fibroblast-derived diNSCs are mature NSCs compared to MEF-derived diNSCs, not only in their differentiation properties but also in their global gene expressions.

Retroviral Transgenes Are Silenced in diNSCs

Retroviral transgenes are gradually silenced during the induction of iPSC reprogramming [35, 49, 50]. The incomplete silencing of retroviral transgenes in iPS-derived somatic cells makes them resistant to differentiation and increases the number of residual pluripotent cells that form teratomas [51]. Although our microarray data could not distinguish between the endogenous and transgenic expression of Yamanaka factors, the decreased expression of Sox2 and Klf4 in the diNSC-derived neurospheres derived from adult fibroblasts suggested that these transgenes might be silenced in these neurospheres, which are not fully reprogrammed for pluripotency.

To determine whether the retroviral transgenes were silenced in the diNSC-derived neurospheres, adult wild-type mouse fibroblasts were infected with pMX retroviral vectors carrying five genes, KSOM and GFP. Four days later, adherent fibroblasts infected with the five genes (KSOM + GFP) were selected by their fluorescence using a flow cytometer, dissociated, and cultivated in floating culture. At this time, 61.2% of the fibroblasts was GFP-positive. Clonal neurospheres were generated only from the GFP-positive fraction. However, 14.4% of the total cells in these neurospheres was GFP-positive (Supporting Information Fig. S6A, S6B). Quantitative PCR analysis also confirmed that all of the retrovirus-induced transgenes were decreased, but not completely silenced, in diNSC-derived spheres at day 14 of suspension (day 18) (Supporting Information Fig. S6C) [13].

Next, we performed the same analysis in the presence of EGF. In this case, five genes (KSOM + DsRed) were introduced into adult Nanog-GFP fibroblasts. Although no Nanog-GFP-positive cells were observed in the EGF-dependent diNSC-derived neurospheres, a few DsRed-positive cells were observed in them (Supporting Information Fig. S7). These results suggested that the retroviral transgenes are silenced in the diNSC-derived cells, even in the EGF-dependent ones, which should contain more mature NSCs than the LIF, FGF2-dependent cells.

diNSC-Derived Neurospheres Have Successfully Differentiated into Neural Cells In Vivo

To evaluate the in vivo differentiation properties and safety of the diNSCs, we transplanted dissociated diNSC-derived neurospheres, which had been transduced by a lentivirus to express the Luciferase and Venus fluorescent proteins [52], into the striatum of C57/bl6j mice, as previously described [8, 9]. BLI analysis [12], which detects luciferase photon signals only from living cells, showed that the cells were engrafted 28 days after their transplantation (Fig. 6A). We observed the host animals for 4 weeks to monitor engraftment and differentiation of the transplanted cells in vivo. Immunohistological analysis with an anti-GFP antibody was also used for detection of the transplanted cells in the striatum of the host animals (Fig. 6B, 6C). When we transplanted EGF-dependent diNSCs without Nanog-GFP-positive cells, no visible tumor formed. On the other hand, some animals treated with LIF-dependent diNSCs with detectable Nanog-GFP-positive cells (>0.018%) had teratomas (Supporting Information Fig. S8). Finally, EGF-dependent diNSCs successfully differentiated into neurons, astrocytes, and oligodendrocytes in host animals (Fig. 6D).

Figure 6.

In vivo engraftment and differentiation of directly induced neurospheres. (A): IVIS images demonstrating the engraftment of directly induced neurospheres in the striatum. Directly induced neurospheres transduced with luciferase were injected into the striatum and observed at postoperation day 28. (B): Bright-field image of directly induced neural stem cell (diNSC)-derived neurospheres engrafted into the brain of a transplanted mouse, stained with antibodies to GFP, detected with 3,3′-diaminobenzidine (brown). Scale bar = 1 mm. (C): Immunohistochemical analysis showing the engraftment of diNSC-derived neurospheres. Venus-positive transplanted cells were detected by anti-GFP antibody. Scale bar = 50 mm. (D): Immunohistochemical analysis of markers for neurons (Hu), astrocytes (GFAP), and oligodendrocytes (APC) in the C57b6j mouse striatum, 4 weeks after the transplantation of directly induced neurospheres. Scale bar = 10 μm. Abbreviations: APC, adenomatous polyposis coli; GFAP, glial fibrillary acidic protein; GFP, green fluorescent protein.

Establishment of diNSCs from Human Adult Fibroblasts

Finally, we applied our method to the establishment of diNSCs from adult human fibroblasts (Fig. 7A). We introduced human Oct4, Sox2, c-Myc, and Klf4 (KSOM) into adult human fibroblasts (Fig. 7B) using retroviral vectors. These cells were then dissociated into single-cell suspensions and cultured for 14 days in the presence of LIF and FGF2 in serum-free medium. pCPT-cAMP treatment was performed for 10 days. After 14 days of suspension culture, a few spheres were observed (Fig. 7C) and subjected to 14 days of adherent culture without adding exogenous growth factors (Fig. 7D, 7E). Four human fibroblast lines were tested in this experiment. The fibroblasts derived from a 16-year-old female demonstrated the highest sphere formation efficiency (Fig. 7F; Supporting Information Table S4). Formed spheres were allowed to differentiate using adherent culture conditions. These cells expressed β3-tubulin (Fig. 7D) or GFAP (Fig. 7E), indicating that the presence of neurons and astrocytes. The frequency of β3-tubulin-positive cells was 58.0% ± 12.4%, and that of GFAP-positive cells was 8.0% ± 3.74% (Fig. 7G). These data suggested that the majority of cells were neural cells. We also analyzed Oct4 expression to detect pluripotent cells. 2.58% ± 2.56% (Fig. 7G) of the cells was Oct4 positive (Supporting Information Fig. S9).

Figure 7.

Directly induced neural stem cells induced from adult human fibroblasts. (A): Schematic of the method for direct neural induction of human fibroblasts. Adult human fibroblasts were retrovirally transduced with Oct4, Sox2, Klf4, and c-Myc, followed by adherent and suspension culture, with growth factors. 4F = KSOM. (B): Human fibroblasts before the introduction of four reprogramming factors. Scale bar = 100 μm. (C): Neurospheres derived from adult human fibroblasts. Scale bar = 100 μm. (D, E): Immunocytochemical analysis for β3-tubulin and GFAP in the differentiated neurospheres derived from adult human fibroblasts. Scale bar = 50 μm. (F): The efficiency of neurosphere formation from various lines of fibroblasts (n = 3). (G): Immunocytochemical analysis of neural cell or pluripotent cell marker proteins in the differentiated neurospheres derived from adult human fibroblasts. The percentage of cells positive for each indicated marker is shown. Abbreviations: bFGF, basic fibroblast growth factor; FBS, fetal bovine serum; GFAP, glial fibrillary acidic protein; LIF, leukemia inhibitory factor.


One of the major advantages of iPSC technology is that it allows for the creation of cells that are genetically matched to patients. However, in SCI animal models, the transplantation of NSCs is not effective at the chronic stage, after glial scar formation [53], which occurs in humans within a few weeks after the injury. Rather, it is believed that a curative effect may be achieved if the NSC transplantation is done within the subacute phase (approximately 7–14 days after injury) [12], due to the changes in the microenvironment within the injured spinal cord [54, 55], and that glial cells derived from these NSCs can play important roles in restoring neural functions to the host animals [9, 11, 54]. Furthermore, the safety of each iPS clone should be thoroughly evaluated before it is used for cell therapy, because of variations in their propensity to form teratomas [8]. Therefore, it is presently impossible to propagate safe NSCs derived from iPSCs made from the injured patient's own cells, within the subacute phase of SCI.

We found that the diNSCs derived from adult fibroblasts generated astrocytes similar to the NSCs present within the striatum of the midgestation mouse embryo [36, 37, 55, 56]. Gliogenic NSCs are currently considered the best source for cell therapy for SCI [11]. Using our method, fibroblasts were differentiated into diNSC-derived neural cells and amplified more than 133 times within 18 days after their infection with retrovirus. In addition, even though they were generated from adult fibroblasts, the diNSCs contained as few Nanog-GFP-positive pluripotent cells as the NSCs from MEF-derived iPS clones.

When we generated secondary neurospheres from adult tail tip fibroblast-derived iPSCs via EBs, all of the clones contained Nanog-GFP-positive pluripotent cells detectable by FACS analysis. However, the numbers of residual Nanog-GFP-positive cells in diNSC-derived neurospheres grown in EGF alone were less (<0.018%) than seen in NSCs derived from safe iPS clones that is only 10% of total clones from adult fibroblasts [8]. We are still improving the method, and the number of Nanog-GFP-positive cells in the diNSC-derived neurospheres may be made to approach zero by optimizing the culture conditions. Although residual pluripotent cells are present in the cultures used to make diNSC-derived neurospheres, the elimination of persistent, undifferentiated Nanog-positive cells and the propagation of EGF-responsive cells lowered the risk of teratoma formation after intracranial transplantation.

Notably, EGF-responsive NSCs were derived only from adult fibroblasts and not from MEFs or dissociated ESCs/iPSCs (Fig. 4H). In a previous report, NSCs derived from ESCs did not respond to EGF, even as tertiary spheres [57]. We also found that the EGF-responsive diNSCs developed faster than the ESC/iPSC-derived NSCs. We hypothesize that these EGF-responsive diNSCs were induced from partially reprogrammed fibroblasts and not from cells that were fully reprogrammed using the Yamanaka factors.

It is known that primary cultured NSCs generally differentiate faster than ESC/iPSC-derived NSCs [57]. ESCs/iPSCs are established from clonal cells under highly selective culture conditions to maintain their pluripotency. The pluripotency of these ES/iPS clones can be maintained by culturing them in the presence of LIF and bovine serum, but they seem differentiation resistant compared to primary cultured cells when placed in conditions for directed differentiation. Using our method, more gliogenic neurospheres were obtained from adult fibroblasts than from MEFs after the same process of reprogramming and neural induction.

The frequency of iPS generation from adult fibroblasts is reported to be approximately 2% of that from MEFs [13] because of their lower competency to respond to the reprogramming procedure. However, the frequency of diNSC-derived neurospheres from embryonic cells was only approximately threefold higher than that of neurospheres derived from adult fibroblasts (Supporting Information Fig. S10). These data suggest that most of the adult diNSC-derived spheres may have been induced from cells that could not be reprogrammed to iPSCs.

Our method may be less selective than that used for clonal ESCs/iPSCs isolation and maintenance. Therefore, adult fibroblasts, which are mostly incompetent for iPS generation, might be able to receive incomplete reprogramming and then differentiate into diNSCs. The decreased number of Nanog-GFP-positive spheres in the adult-derived diNSCs may support this idea. We propose that these partially reprogrammed cells generated rapidly differentiating diNSCs.

Although ESCs/iPSCs established using the current system easily maintain their pluripotency and are suitable for research on developmental technologies, the differentiation properties of their progeny are significantly different from those of primary cultured tissue stem cells. On the other hand, our method for inducing diNSCs from adult human fibroblasts would be effective for obtaining fast-differentiating NSCs from human cells, and therefore have the potential to be used for autograft transplantations, because of their safety and rapid preparation. However, Oct4-positive cells and residual fibroblasts in human diNSCs-derived neurospheres suggested that the protocol should be modified to purify NSCs and the gene transfer must be performed using integration-free systems [58–60] before our method can be applied to future cell therapies.


Here, we demonstrated that rapidly differentiating NSCs could be derived by the direct differentiation of neural cells from partially reprogrammed fibroblasts. This process required the transduction of the four Yamanaka factors. Moreover, these rapidly differentiating NSCs could be separated from residual pluripotent cells by optimization of the culture conditions, enabling us to prepare NSCs that differentiated into glial cells quickly with no detectable Nanog-GFP-positive cells.


We gratefully thank Dr. Jun Muto for help with the transplantation experiment, Drs. Atsushi Miyawaki and Chikako Hara for the Venus-ffLuc construct, and Hayao Ebise and Dr. Hiroyuki Nakagawa for the microarray analysis. We thank Junko Hayashi for technical assistance, and Manabu Ohyama for providing the human fibroblasts. This work was supported by grants from the Japan Society for the Promotion of Science (JSPS) to T.M. and W.A., the project for the realization of regenerative medicine and support for the core institutes for iPS cell research from the Japanese Ministry of Education, Culture, Sports, Science and Technology (MEXT) to M.N. and H.O., SORST of the Japan Science and Technology Agency to H.O., and Keio Gijuku Academic Development Funds to W.A.


H.O. is the scientific consultant of San Bio, Inc., Eisai Co. Ltd. and Daiichi Sankyo Co. Ltd. K.Y. is employed by Dainippon Sumitomo Pharma Co. Ltd., and C.F. is employed by Lion Co. Ltd., while K.Y. and C.F. worked in Dr. Okano's laboratory in Keio University as a collaborative research fellow. The remaining authors report no conflicts.