Zic3 induces conversion of human fibroblasts to stable neural progenitor-like cells

Dear Editor, Recent advances in cellular reprogramming have opened the door to the generation of patient-specific cells for regenerative medicine and disease modeling (Takahashi and Yamanaka, 2006; Yu et al., 2007; Nakagawa et al., 2008). Apart from OCT4, NANOG, and SOX2, additional transcription factors (TFs) like zinc-finger transcription factor in cerebellum-3 (Zic3) have been shown to play a role in the maintenance of pluripotency (Lim et al., 2007; 2010). Zic3, a member of the GLI superfamily, has been implicated in the maintenance of pluripotency in mouse and human embryonic stem cells (ESCs) by directly controlling the expression of Nanog (Lim et al., 2007; 2010). Zic3 is also an immediate early gene induced by fibroblast growth factor (FGF) signaling during neural specification (Marchal et al., 2009) and, by preventing neuronal differentiation, plays a role in the maintenance of the neural progenitor cell fate (Inoue et al., 2007). These studies demonstrate thus that Zic3 plays important roles in both the maintenance of ESC pluripotency as well as commitment to and the maintenance of neuroprogenitor cells. We hypothesized that co-transduction of Zic3 together with OCT4 (O), SOX2 (S), and KLF4 (K) might lead to improved generation of induced pluripotent stem cells (iPSCs) from human fibroblasts. Because forced expression of SOX17 or SOX7 in human embryonic stem cells (hESCs) creates stable endodermal and primitive endodermal cell fates (Seguin et al., 2008), the alternative outcome of these experiments could be the creation of stable neuroprogenitor lines. To investigate the consequence of Zic3 during reprogramming, human BJ1 fibroblasts were retrovirally transduced with OSK with and without Zic3 (Z). Colonies of cells referred to as Zic3-induced cells (ZiCs) (Supplementary Figure S1A) appeared in the culture plates transduced with OSKZ after 8 days and the frequency increased further by day 14 (Figure 1A). On day 14, ∼50–60 ZiC colonies were obtained from 1.5 × 105 BJ1 cells, with a reprogramming efficiency of ∼0.035%. In contrast, transduction of BJ1 cells with OSK did not yield visible iPSC colonies by day 30, even though iPSC colonies appeared at later times. Omitting either OCT4, SOX2, or KLF4 from the combination completely abrogated the colony formation (Figure 1A). Similar results were obtained when primary human fibroblasts were transduced with OSKZ. Ten to twenty colonies were generated from 1.5 × 105 human fibroblasts, 14 days following transduction with a reprogramming efficiency of 0.009% (Supplementary Figure S2A). ZiC lines generated from BJ1 cells that were further characterized are clones ZiC1 and ZiC2, while ZiC lines generated from primary fibroblasts for further studies are clones pZiC1, pZiC2, and pZiC3. Transduction of Zic3 together with OCT4, SOX2, and KLF4 in human fibroblasts yields neural progenitor-like cells. (A) Quantification of the number of colonies 8 and 14 days after retroviral transduction of BJ1 fibroblasts with different combinations of retroviral vectors encoding for OCT4 (O), SOX2 (S), KLF4 (K), and Zic3 (Z) (n = 3, P= 0.02). No colonies could be observed after transduction with OSK, SKZ, OKZ, and OSZ until after 20 days following transduction. (B) AP staining of BJ1 fibroblasts transduced with OSK or OSKZ encoding retroviral vectors. (C) Immunostaining of the cell line ZiC1 for SSEA1, SSEA4, TRA160, and DAPI (×20, Nikon Eclipse Ti; n= 3). Scale bar = 50 µm. (D) Transcript levels of ectodermal genes in BJ1 fibroblasts, hESCs, hiPSCs, ZiCs, and neurospheres derived from those cells as determined by RT-qPCR (mean ± SEM; n= 3). (E) Transcript levels of pluripotency genes at different time points after retroviral transduction of BJ1 fibroblasts with OSKZ (mean ± SEM; n= 3). (F) Transcript levels of the epiblast marker FGF5 and the neuroprogenitor markers PAX6 and FOXG1 at different time points after retroviral transduction of BJ1 fibroblasts with OSKZ (mean ± SEM; n= 3). (G and H) The ZiC1 cell line was cultured under motor neuron differentiation conditions as described in Supplementary Materials and methods. (G) On day 14 of differentiation, RT-qPCR was used to evaluate the expression of motor neuron-specific marker genes (mean ± SEM; n= 3). Zic3Tg, Zic3 transgene. (H) Differentiated ZiC1 progeny were immunostained for the neuronal proteins NF200 kDa, MAP2 and the motor neuron proteins HB9 and ISL1 (n = 3). Scale bar = 50 µm. (I) ZiCs differentiate towards neuronal and astrocyte lineage cells upon injection in the striatum of adult mice. In total, 150000 green fluorescent protein (GFP)-transduced ZiC1 cells cultured as neurospheres were injected into the striatum of adult mice. The grafts were analyzed 4–5 weeks after transplantation. Colabeling of GFP with antibodies against the neuronal proteins NF200 kDa, NeuN, the neural progenitor protein OTX1/2, and the astrocyte protein GFAP (n = 3). Scale bar = 50 µm. In contrast to hiPSC or hESC lines, ZiCs expressed very low levels of endogenous OCT4 and NANOG mRNA. The expression of endogenous SOX2 was lower compared with hESCs or human induced pluripotent stem cells (hiPSCs), whereas endogenous KLF4 was expressed at levels similar to those in hESCs or hiPSCs (Supplementary Figures S1B and S2B). Endogenous Zic3 was expressed at lower levels in ZiCs compared with hESCs and hiPSCs (Supplementary Figure S1C). The OCT4, SOX2, and KLF4 transgenes were silenced, whereas the Zic3 transgene remained expressed (Supplementary Figure S1C). The cell surface phenotype of ZiCs was similar to that of neuroprogenitor cells, i.e. alkaline phosphatase (AP) (Figure 1B) and SSEA1 positive (Figure 1C). However, ZiCs were negative for SSEA4 and TRA160, which expressed on hESCs (Figure 1C). Altogether, these results indicated that ZiCs were not reprogrammed to an ESC-like fate. RT-qPCR analysis revealed that the ectodermal TF PAX6 was expressed at much higher levels in ZiCs and pZiCs than in hESCs/hiPSCs, whereas the endodermal TFs, FOXA2 and SOX17, and the mesodermal TFs, BRACHYURY and GSC, were not expressed (Supplementary Figures S2C and S3A). The combined expression of SOX2, PAX6, and SSEA1 in strongly AP-positive cells suggested that the new cell lines were of the neuroectoderm fate. We therefore compared the expression levels of additional neuroectodermal TFs in ZiCs and pZiCs with those in hiPSCs or hESC-derived neural stem cells (NSCs), generated by dual homologs of drosophila mothers against decapentaplegic (MAD) and the Caenorhabditis elegans protein SMA (SMAD) inhibition with noggin and SB431542 (Chambers et al., 2009). The neuroectodermal genes, PAX6, OTX1, FOXG1, GLAST, p75, SOX1, and OLIG2, were expressed in hESCs and hiPSCs only upon differentiation towards NSCs, whereas in ZiCs and pZiCs, these genes were highly expressed prior to any cytokine treatment (Figure 1D and Supplementary Figure S2D). Prior to transduction, both BJ1 and primary fibroblasts did not express significant levels of the above-mentioned neuronal genes as well as pluripotent genes. Treatment of ZiCs with the two SMAD inhibitors did not further induce the expression of these neuroectoderm-specific genes (Figure 1D). Near homogenous expression of PAX6, OTX1, FOXG1, and OLIG2 in ZiCs maintained on feeders was confirmed at the protein level by immunostaining (Supplementary Figure S4). To investigate whether the continuous expression of the Zic3 transgene is important for the maintenance of the ZiC fate, we silenced the expression of Zic3 by two independent shRNAs cloned in a pTRIPZ doxycycline inducible lentiviral vector. ZiCs, passage 25, were transduced with shRNAs and selected by puromycin (1 mg/ml) 1 day after transduction. Expression of Zic3-specific shRNA was induced by adding doxycycline (1 mg/ml) to the media. Induction of Zic3-shRNA expression in ZiCs induced an 82%–90% decrease in Zic3 expression. This was associated with a 60%–80% decrease in expression levels for PAX6, OTX1, and OTX2 (Supplementary Figure S5) without concomitant increase in endodermal or mesodermal gene expression (data not shown). To further ascertain the neuroprogenitor-like fate following OSKZ transduction of fibroblasts, the global gene expression profile of ZiCs and hESCs-derived NSCs was compared with microarray analysis. Compared with undifferentiated hESCs, 1617 probes were more highly expressed in hESCs-derived NSCs and 1876 probes in ZiCs, whereas 1550 and 1977 probes were expressed at lower levels in hESCs-derived NSCs and ZiCs, respectively. Of the significantly upregulated probes, 563 were common to hESCs-derived NSCs and ZiCs (Supplementary Figure S3B and Supplementary Data). Gene ontology analysis (http://david.abcc.ncifcrf.gov/) demonstrated that many of these genes are involved in neural differentiation and development (Supplementary Figure S3C). When cells were cultured as embryoid bodies, ensuing clusters resembled neurospheres. RT-qPCR analysis identified only neuroectoderm-specific transcripts (Supplementary Figure S6A). To further confirm the commitment to neuroectodermal lineage, we also injected 1 million ZiCs maintained on mouse embryonic fibroblasts and dissolved in 50% matrigel, subcutaneously in immunodeficient mice. In contrast to teratomas generated by hESCs and hiPSCs, ZiC progeny did not form cells of the three germ layers but only neuroectoderm tumors which were largely positive for β3-tubulin and synaptophysin (Supplementary Figure S6B). This further substantiated the notion that the differentiation capacity of ZiCs is restricted to the (neuro)ectodermal lineage. As colonies were formed as early as 8 days, and very robustly 14 days after transduction of BJ1 and primary fibroblasts (Figure 1 and Supplementary Figure S2A), we hypothesized that the reprogramming to a ZiC fate did not occur via an intermediary more primitive cell type, such as an epiblast stem cell or ground-state embryonic stem cell fate (Nicholas and Smith, 2011). To further exclude the formation of more primitive intermediates, RT-qPCR was performed every 3 days following transduction for 15 days. The endogenous expression of OCT4, REX1, and NANOG was low or absent at any time-point during the reprogramming process, which demonstrates that no transient pluripotent intermediate was present during the reprogramming process (Figure 1E). Low expression levels of SOX2 could be detected from day 9 onwards (Figure 1E), consistent with the expression of PAX6 and FOXG1, two other neuroprogenitor marker genes (Figure 1F). As expression of FGF5 mRNA could not be observed during the reprogramming process, we also concluded that epiblast stem cell intermediates were not generated (Figure 1F). Altogether, these studies demonstrated that no intermediate iPSC or epiblast stem cell was generated during our experiment, but that the neuroprogenitor-like cells were directly reprogrammed from fibroblasts. We next tested whether ZiCs can be directly differentiated into mature neural progeny in vitro. ZiCs could be committed to the astrocyte lineage, as shown by expression of the astrocyte-specific transcript, GFAP (Supplementary Figure S7A) and the astrocyte-specific proteins, S100β and GFAP in 42% ± 7% and 63% ± 11% of cells, respectively (Supplementary Figure S7B). Culture of ZiCs under oligodendrocyte-inducing conditions resulted in the induction of the oligodendrocyte-specific transcripts OLIG2 and Myelin oligodendrocyte glycoprotein (Supplementary Figure S7C), and yielded 11% ± 2% cells that were O4-protein positive (Supplementary Figure S7D). Expression of the Zic3 transgene was decreased by 4- and 9.5-fold following differentiation to the astrocyte and oligodendrocyte lineages, respectively (Supplementary Figure S7A and C). ZiCs could also be committed to motor neuron-like cells. Resulting cells expressed the mature neuronal transcripts NF200 kDa and MAP2 as well as motor neuron-specific transcripts HB9 and ISL1 (Figure 1G). Immunostaining demonstrated 75% ± 8% of cells to be positive for NF200 kDa NF, 83% ± 14% for MAP2, 78% ± 11% for HB9 and 89% ± 13% for ISl1 (Figure 1H). Expression of the Zic3 transgene was decreased by 35-fold following differentiation to the motor neuron lineage (Figure 1G). We finally assessed the in vivo differentiation potential of ZiCs by orthotopic injection of 150000 green fluorescent protein (GFP)-transduced ZiCs into the striatum of 10- to 12-week-old non obese diabetic-sever combined immunodeficiency (NOD-SCID) γC KO mice (n = 9). Analysis of the graft 4–5 weeks later demonstrated the presence of GFP-positive cells within the striatum, and no tumors were detected. Among these GFP-positive cells, some cells expressed the mature neuronal proteins, NF200 kDa, NeuN, the neural progenitor protein OTX1/2, and the astrocyte protein GFAP (Figure 1I). This indicates that ZiCs can mature further to the neuronal and astrocyte lineages in response to in vivo differentiation signals, albeit—in adult brains—at low efficiency. Aside from pluripotent iPSC generation from adult cells, reprogramming of fibroblasts to different mature cell types, including neuronal cells, hepatocytes, and cardiomyocytes, has been described (Ieda et al., 2010; Pang et al., 2011; Sekiya and Suzuki, 2011). For instance, forced combinatorial expression of the neural-lineage-specific TF can rapidly and efficiently convert mouse and human fibroblasts into functional neurons in vitro (Vierbuchen et al., 2010; Pang et al., 2011). Nevertheless, it is difficult to expand mature neurons for many passages. Therefore, the strategy proposed here to generate ZiCs has advantages compared with direct reprogramming of fibroblasts to neurons. These include: (i) ZiCs can be efficiently and easily maintained in neurosphere cultures (Supplementary Figure S8A); (ii) ZiC neurospheres can be expanded for many passages in vitro (at least 40 passages) and still retain the ability to give rise to (motor)neurons, oligodendrocytes, and astrocytes; (iii) compared with hESCs, ZiCs can be directly differentiated to mature neural progeny without prior neural commitment step; (iv) compared with hESCs and hiPSCs, ZiCs can more easily be genetically manipulated as 96% ± 4% of ZiCs were transduced with a GFP expressing lentiviral vector (multiplicity of infection (MOI) of 4), whereas only 9% ± 2% of hESCs were transduced (Supplementary Figure S8B and C). Recently, Kim et al. (2011) also succeeded in reprogramming mouse fibroblasts to neural progenitor cells by short-term expression of the OKSM TFs to induce an unstable pluripotent state that can be quickly differentiated into neural progenitors by maintenance in neural stem cell medium. Whether or not a similar strategy is also possible for human fibroblasts, still needs to be investigated. In this study, direct reprogramming of human fibroblasts to neural progenitor cells using a lineage-specific transcription factor, Zic3, has been demonstrated. Whether or not other neural-lineage-specific TFs can generate neuroprogenitor cells with or without OSK still needs to be further investigated. In conclusion, we demonstrate the very fast generation of neuroprogenitor cells, termed ZiCs, from human fibroblasts by OSKZ transduction. The above-said features of ZiCs may make them a very useful tool for pharmacological compound screening in neural disorders or for assessment of toxicity effects. Furthermore, our data suggest that it may be useful to assess whether other germline-specific TF may assist in the derivation of stable, expandable progenitors of the three germ layers from human adult cells. [Supplementary material is available at Journal of Molecular Cell Biology online. We would like to thank Mrs Savitha Nageshappa for excellent technical assistance and Drs Valerie Roobrouck and Yemiao Chen (Katholieke Universiteit, Leuven) for critical review of all the data, Dr Jyothi Prasanna and Ms Preethi Sheshadhri (MIRM, Bangalore) for their technical assistance and critical reviewing of the manuscript, and Drs Wouter Van Delm and Rudy Van Eijsden (Katholieke Universiteit, Leuven) for their help with the microarray analysis. JD is a postdoctoral fellow of the 'Fonds voor Wetenschappelijk Onderzoek Vlaanderen' (FWO). This work was supported in part by Grants from FWO (G.0832.11N and G.0667.07), and grants from Katholieke Universiteit Leuven (EIW-B4855-EF/05/11, ETH-C1900-PF, and EME-C2161-GOA/11/012) to CMV.]