Functional integration of dopaminergic neurons directly converted from mouse fibroblasts

Parkinson’s disease is one of the most common neurodegenerative disorders and is characterized by a loss of dopaminergic (DA) neurons primarily of the substantia nigra pars compacta (SN), leading to a reduction of dopamine in the striatum (Berke and Hyman, 2000; Huse et al., 2005). While several possible cell sources including fetal brain, ES, and iPS cells are being explored as cell replacement therapies for degenerating DA neurons, ethical and practical barriers to the application of such therapies exist (Kim et al., 2002; Olanow et al., 1996; Wernig et al., 2008). Epigenetic reprogramming to pluripotency has provided critical evidence for the plasticity of the somatic genome (Eggan et al., 2004; Wilmut et al., 1997), and recent studies have demonstrated the feasibility of direct lineage reprogramming from one somatic cell type to another, bypassing a pluripotent intermediate state (Ieda et al., 2010; Szabo et al., 2010; Zhou et al., 2008). Two recent studies reported the generation of DA neurons by direct reprogramming (Caiazzo et al., 2011; Pfisterer et al., 2011). However, in these studies the gene expression profiles of reprogrammed DA neurons differed significantly from primary midbrain DA neurons. Furthermore, induced DA (iDA) neurons were not shown to be functional by in vivo transplantation assays. Thus, to assess the therapeutic potential of lineage converted cells we examined whether alternative induction strategies could generate iDA neurons that resemble midbrain DA neurons more closely and are functional in transplantation assays.

To analyze fibroblast reprogramming to DA neurons, we generated a knock-in (KI) mouse model in which the eGFP coding sequence was targeted to the Pitx3 gene under control of the endogenous promoter (Pitx3-eGFP) (Figure S1a and S1b). Previously, GFP+ cells derived from Pitx3-eGFP KI ES cells and mice have been shown to be functional midbrain DA neurons and to effectively alleviate symptoms in PD animal models (Hedlund et al., 2008; Zhao et al., 2004) while other reporter systems such as Th-GFP were less effective (Hedlund et al., 2007). EGFP expression in DA neurons was assessed after in vitro differentiation of Pitx3-eGFP ES cells and adult mice generated from Pitx3-eGFP ES cells exhibited eGFP+DA neurons in the midbrain that were also positive for tyrosine hydroxylase(TH), a marker of mature DA neurons (Figure S1c and S1d). These data show that the Pitx3-eGFP KI model faithfully induced eGFP expression in midbrain DA neurons thus providing a useful system for the isolation of iDA neurons.

To exclude the possibility of progenitor cell contamination in preparations of embryonic fibroblasts, we prepared tail tip fibroblasts (TTFs) from adult Pitx3-eGFP mice, which were GFP and TH negative (Figure 1a and b, top right). A group of 11 candidate transcription factors was selected based on their known functions in the development and survival of midbrain DA neurons (Table S1) packaged into doxycycline(dox)-inducible lentiviruses and introduced into Pitx3-eGFP TTFs. Twelve days after infection, TH+ cells with neuronal morphology appeared in the cultures (Figure 1b, middle left and bottom panels) with eventually approximately 2% of the cells expressing Pitx3 as indicated by eGFP staining (Figure 1c). We also reprogrammed wild-type TTFs as indicated by the appearance of DA neuron-like cells after viral transduction (Figure 1b, middle right). To determine which factors were critical for this process we divided the 11 factors into three groups and tested the ability of different combinations of the pools to induce Pitx3-eGFP. Using this subtractive approach we found that a pool of 8 factors lacking Pax6, Sox1 and Ngn2 generated Pitx3-eGFP+cells more efficiently than the original pool of 11, while removal of either the Acsl1, Mytl1, Brn2 group or the Lmx1a, Lmx1b, Nurr1, Pitx3, EN1 group failed to produce Pitx3-eGFP+ cells, suggesting that this combination of 8 transcription factors was sufficient for the induction of Pitx3-eGFP+cells (Figure S1e). To identify the key DA-inducing factors, we examined the effects of removing individual factors from the 8 factor pool. Surprisingly, eGFP+cells were not detectable in TTF cultures infected with lentiviral pools lacking Ascl1, and pools lacking Pitx3 induced only a small number of eGFP+ cells (about 0.5%). However, pools lacking any of the remaining 6 factors were able to induce significant numbers of eGFP+cells (Figure S1f). This result suggests that Ascl1 and Pitx3 are necessary for the induction of Pitx3-eGFP+ DA neurons. We also found that these two transcription factors alone were able to induce Pitx3-eGFP+cells in 2-3% of the target cells (Figure 1d). The expression of DA neuronal marker genes was determined using quantitative RT-PCR on FACS-purified Ascl1/Pitx3 transduced GFP+ and GFP-cells. The DA neuron-specific markers examined included genes involved in the biosynthesis of dopamine [TH, aromatic L-amino acid decarboxylase (AADC)], dopamine storage [vesicular monoamine transporter 2 (VMAT2)], and uptake [dopamine transporter (DAT)], all of which were significantly upregulated in Pitx3-eGFP+cells when compared to eGFP-cells (Figure 1e). However, the expression level of these genes was lower than that observed in primary Pitx3-eGFP+DA neurons (Figure S1g) and we were not able to detect dopamine or electrophysiological activity (data not shown). We maintained the eGFP+ cell cultures for up to 4 weeks and did not observe significant maturation of DA neurons (data not shown). These results suggest that Pitx3-eGFP+ cells induced by Ascl1 and Pitx3 are not terminally differentiated DA neurons but may instead represent immature DA neurons. Several extrinsic and intrinsic factors which control maturation of DA neurons are functionally interconnected and cooperate to promote the terminal differentiation of DA neurons during neuronal development (Kim et al., 2002; Lee et al., 2000; Martinat et al., 2006). We therefore tested whether inclusion of additional factors could fully reprogram fibroblasts into functional DA neurons.

Several neurotrophic factors including Sonic hedgehog (Shh) and Fibroblast growth factor 8 (FGF8) appear to be critical for the specification and differentiation of developing midbrain DA neurons (Lee et al., 2000). To examine whether Shh and FGF8 promote DA neuron reprogramming, Ascl1 and Pitx3 transduced fibroblasts were cultured with Shh and FGF8 and the efficiency of iDA neuron generation was determined. Cultures treated with Shh and FGF8 generated about 5% Pitx3-eGFP+ cells, which is 2 fold more than in cultures treated with bFGF alone (Figure S1h). These data suggest that neurotrophic factors are critical components in promoting the generation of iDA neurons from fibroblasts. We also examined the DA neuron-inducing activity of other genes by adding single factors to the Ascl1/Pitx3 infected cells. As shown in Figure S1i, the addition of Lmx1a, Nurr1, Foxa2 and EN1 in combination with Ascl1 and Pitx3 significantly enhanced the efficiency of eGFP+ induction whereas overexpression of Sox1, Pax6 or Lmx1b had an inhibitory or no effect. Because of the positive effects of Lmx1a, Nurr1, Foxa2 and EN1 on the reprogramming process, we tested different combinations of these 4 factors along with Ascl1, Pitx3 and found that combination of all 6 factors gave rise to the highest induction of eGFP+ cells 10 days after infection (Figure S1j). In addition, these 6 factors can induce the expression of DA neuronal marker genes more efficiently than any other combinations including the previously published 3 factor combinations (Figure S1k), suggesting that the induced cells resemble midbrain DA neurons more closely than those generated using other factor combinations. Next, we examined reprogramming efficiency by flow cytometry on days 4, 8, 12, and 18 after viral transduction with six factors in combination with FGF8 and Shh. A significant number of eGFP+ cells were evident as early as day 4, reaching 8% by day 12 and a maximum of 9.1% of total cells by day 18 (Figure 1g). Additionally, figure 1f shows that TH+ iDA neurons are positive for neuron-specific class III beta-tubulin (TuJ1) and microtubule associated protein 2 (MAP2), both mature neuronal makers, and for DAT or AADC, which are markers of mature DA neurons. Furthermore, none of the TH+ cells co-expressed dopamine beta hydrolase (DBH), a marker for noradrenergic neurons. Other types of neurons including 5HT and motor neurons were also not detected in these cultures (data not shown).

We compared the gene expression profiles of 2 and 6 factor reprogrammed iDA neurons, fibroblasts, neural stem cells (NSC), embryonic, and adult midbrain DA neurons by qRT-PCR (Figure 2a). While some variability in marker gene expression was seen, likely as a result of the inconsistent efficiency of viral infection, DA neuronal maker genes were significantly up-regulated in 6 factor induced GFP+ cells and partially upregulated in 2 factor induced eGFP+cells. We also observed induction of NSC makers including Nestin in 2 factor induced Pitx3-eGFP+ cells, but the expression of these genes was significantly lower in 6 factor induced Pitx3-eGFP+ cells. These results again suggest that 2 factor induced Pitx3-eGFP+ cells represent immature DA neurons and that additional factors are required for the maturation of 2 factor induced eGFP+ cells to differentiated DA neurons. Furthermore, we found a significant reduction of fibroblast marker gene expression in 6 factor induced eGFP+ cells to levels indistinguishable from those seen in primary Pitx3-eGFP+ cells. These results indicate that 6 factor iDA neurons are similar at the molecular level to midbrain DA neurons. The dependence of iDA neurons on ectopic transgene expression was examined using a dox-inducible lentviral system. We transduced Pitx3-eGFP TTFs with the dox-inducible factors along with M2rtTA and induced iDA neuron reprogramming with dox for 5 days. After eGFP expression was detected in the reprogramming cultures, dox was withdrawn and the cells were maintained for 7 days to allow for complete silencing of exogenous gene expression (Figure S2a). We observed that Pitx3-eGFP expression was stably maintained in these cultures, demonstrating that the iDA neurons are phenotypically stable in the absence of ectopic factor expression (Figure S2b).

To test whether the reprogrammed cells have functional qualities of DA neurons, we examined dopamine production by reverse-phase high performance liquid chromatography (RP-HPLC). . Significant levels of dopamine and the dopamine derivative, 3,4dihydroxyphenylactic acid (DOPAC) were detected in the iDA neurons (Figure 2b), in contrast to fibroblasts, which were negative (Figure S2c). Importantly, we were able to detect dopamine release in the context of high potassium (56mM)-induced depolarization in the iDA neurons (Figure S2d), as expected for functional neurons that produce and release dopamine. We also examined whether iDA neurons display electrophysiological properties characteristic of DA neurons. iDA neurons exhibiting a differentiated DA neuronal morphology were analyzed using whole-cell patch clamp analysis and the identity of the analyzed cells was retrospectively confirmed by TH immunostaining. A current step protocol (−20pA to +120pA) elicited long-duration action potentials (>2msec) in the majority of reprogrammed iDA neurons (13 out of 20 cells) (Figure 2c) in contrast to control fibroblasts, which were negative (Figure S2e). 15 days after dox treatment, the average resting membrane potential of iDA neurons was 48.23±7.25mV (mean ± s.e.m., n=13), input resistance was 1.13±0.38GΩ (mean± s.e.m., n=13), and action potential amplitude was 84.23±18.69mV (mean± s.e.m., n=13), providing evidence that the electrophysiological properties of the iDA neurons are similar to those of midbrain DA neurons (Figure 2f,g and h). The iDA neurons also exhibited voltage-dependent ionic currents (Figure 2d). In addition, the action potentials evoked by depolarizing current injections in iDA neurons (top panel) were completely abolished by TTX administration (bottom panel) (Figure 2e), indicating the presence of well developed Na channels in iDA neurons. Furthermore, in current-clamp mode we were able to inject prolonged hyperpolarizing current pulses (+20 pA to −140 pA) and observed that the voltage responses started to decline slowly in an iDA neuron (Figure S2f); this type of prominent time-dependent anomalous rectification is characteristic of the functional midbrain DA neurons (Rayport et al., 1992). In sum, these results indicate that the reprogrammed DA-producing neurons acquired functional properties that are highly similar to those of midbrain DA neurons.

Finally, we investigated whether iDA neurons can be effective in a rodent Parkinson’s disease model for cell transplantation therapy. Pitx3-eGFP+ cells were FACS isolated from TTFs 12 days after transduction with 6 factors. The eGFP+ cells were implanted into the striatum of mice that had been lesioned with 6-hydroxydopamine(6OHDA) to mimic the DA loss that occurs in PD. 8 weeks after transplantation, the implanted Pitx3-eGFP+ cells led to a significant reduction in amphetamine-induced rotation scores in 6-OHDA lesioned mice (Figure 2i and j) in contrast to sham transplanted or intact controls, which showed no rescue. The survival of the transplanted iDA neurons was assessed in sections stained for TH (Figure 2k). In control mice, complete loss of TH fibers in the striatum occurred 4 weeks after 6OHDA lesioning (Figure S2g), but mice transplanted Pitx3-eGFP+ cells exhibited integrated grafts containing large numbers of DA neurons, (350-1900 cells, Figure 2m). The grafted cells showed neuronal morphology and extended TH+ fibers into the deinnervated host striatum (Figure 2k). We confirmed that all transplanted eGFP+ cells expressed the DA neuron markers TH and AADC in vivo (Figure 2l) and elevated dopamine levels were detected in transplanted striatum (Figure 2n). These data indicate that transplanted iDA neurons are the major population of DA neurons after transplantation and demonstrate the functional capacity of the iDA neurons, suggesting that this type of strategy may potentially provide a useful therapeutic cell source for cell replacement therapy in PD.

In this study, we have demonstrated that the combined activity of Ascl1 and Pitx3 is sufficient to facilitate the conversion of fibroblasts into an immature DA neuronal cell fate and that ectopic expression of additional transcription factors is required for maturation of 2 factor induced Pitx3-eGFP+cells. Reprogramming to pluripotency is thought to involve numerous rounds of cell division which appear to be critical for the completion of epigenetic changes associated with the acquisition of pluripotency (Hanna et al., 2009). In contrast, the reprogramming into functional neurons does not seem to require multiple cell divisions (Vierbuchen et al., 2010), and differentiated neurons are post-mitotic. Thus, we hypothesized that induction of a functional DA neuron state might require the activity of additional factors. Consistent with this idea, we found that the addition of several other transcription factors to the original 2 factor cocktail resulted in the up-regulation of mature DA neuronal marker genes in Pitx3-eGFP+ cells. However, the expression levels of these DA genes did not reach the same level as is observed in primary Pitx3-GFP+cells. Therefore, future studies need to be performed to identify additional DA neuron-inductive factors, possibly including epigenetic modifiers, for the generation of terminally differentiated iDA neurons indistinguishable from midbrain DA neurons. More importantly, we have shown the functional rescue of iDA neurons in 6OHDA leasioned PD animal models. Although amphetamine-induced rotational tests do provide a functional readout, this assay does not measure a clinical phenotype of PD. Thus, the ability of the iDA neurons to suppress Parkinson-like symptoms in other behavioral tests remains to be examined. Moreover, the number of cells with dopaminergic phenotype in the present grafts(350-1900) is considerably higher than the number estimated to be needed to give compensation following amphetamine administration in grafts of mouse fetal DA neurons to 6OHDA-lesioned mice (25-100 dopaminergic cells; Brundin et al. 1986). Thus, the functional efficacy of the mouse fibroblast-derived iDA neurons described here appears to be relatively low, suggesting that the reprogramming procedure would need to be improved further to be of interest for cell replacement strategies.

Our results have several implications for the potential use of iDA neurons for disease modeling and cell replacement therapy of PD. The ability to generate iDA neurons from abundant somatic cells such as fibroblasts in a short period of time makes this system attractive for autologous cell based approaches. Furthermore, iDA neurons could provide a more homogenous cell source for modeling PD in vitro. Importantly, our approach avoids concerns surrounding the inherent tumorigenicity of ES or iPS cells when transplanted in an undifferentiated state. Therefore, the generation of iDA neurons could provide a reasonable cell source for pharmacological assays or cell replacement therapy for PD.