An important goal in stem cell biology is to develop methods for efficient generation of clinically interesting cell types from relevant stem cell populations. This is particularly challenging for different types of neurons of the central nervous system where hundreds of distinct neuronal cell types are generated during embryonic development. We previously used a strategy based on forced transcription factor expression in embryonic stem cell-derived neural progenitors to generate specific types of neurons, including dopamine and serotonin neurons. Here, we extend these studies and show that noradrenergic neurons can also be generated from pluripotent embryonic stem cells by forced expression of the homeobox transcription factor Phox2b under the signaling influence of fibroblast growth factor 8 (FGF8) and bone morphogenetic proteins. In neural progenitors exposed to FGF8 and sonic hedgehog both Phox2b and the related Phox2a instead promoted the generation of neurons with the characteristics of mid- and hindbrain motor neurons. The efficient generation of these neuron types enabled a comprehensive genome-wide gene expression analysis that provided further validation of the identity of generated cells. Moreover, we also demonstrate that the generated cell types are amenable to drug testing in vitro and we show that variants of the differentiation protocols can be applied to cultures of human pluripotent stem cells for the generation of human noradrenergic and visceral motor neurons. Thus, these studies provide a basis for characterization of yet an additional highly clinically relevant neuronal cell type. Stem Cells2014;32:609–622
Pluripotent embryonic stem cells (ESCs) or induced pluripotent stem (iPS) cells have the capacity to develop into any somatic cell type and for this reason these cells have become a major focus in regenerative medicine. A strong emphasis is focused on generating different types of clinically interesting cell types with the prospect of using these methods for engineering cells for cell replacement therapy or for disease modeling in vitro. It is evident that in addition to carefully characterized ESC and iPS cell lines, these promising applications also require highly effective methods for generating desired cell types from pluripotent stem cells. Major progress has been made in this area and it is now possible to generate a variety of cell types including dopamine neurons (DANs), hepatocytes, and cardiomyocytes from ESCs and iPS cells [1-7]. However, the successful development of methods for cell engineering of specific cell types usually requires extensive knowledge of how cells are generated in normal development. The type of cells that currently can be efficiently generated with a high level of enrichment therefore remains rather limited and there is a need to extend the analyses to additional cell types of importance in disease.
Noradrenergic neurons (NANs) represent one such clinically interesting cell type but methods for efficient NAN generation have not yet been reported. The locus ceruleus is the major site in the brain where NANs are localized and the origin of norepinephrine innervation to many brain areas including the cortex and hippocampus [8-10]. Importantly, disrupted norepinephrine neurotransmission is linked associated to clinical disorders. For example, these neurons degenerate in Parkinson's and Alzheimer's disease causing dementia and depression in these patients [11-19]. Loss of NANs is also a contributing feature of the neurodevelopmental disorder Rett syndrome and is associated with disrupted respiration and cognition deficiencies in affected individuals [20, 21]. It is therefore evident that there exists a need to generate NANs from pluripotent stem cells so that disease mechanisms can be investigated in further detail, and to permit further studies aiming at NAN cell replacement strategies.
In mouse embryogenesis, NANs of the locus ceruleus are born in the hindbrain in dorsal rhombomere 1 . Fibroblast growth factor 8 (FGF8) and bone morphogenetic protein (BMP) 5 and 7 are expressed in the mid-hindbrain border (isthmus) and roof plate, respectively, and are essential for specifying the NAN progenitors early in neural tube development [23-29]. Signaling within the developing neural tube leads to the patterned expression of transcription factors that function as intrinsic determinants in the specification of distinct neural cell types. The two related homeobox transcription factors Phox2a and Phox2b have been shown to play key roles in the specification and differentiation of locus ceruleus NANs [30, 31]. Phox2a is expressed in proliferating NAN progenitors and is required for inducing the expression of Phox2b in cells that have just exited the cell cycle. Phox2b expression is diminished by embryonic (E) day 11.5 while Phox2a expression is maintained in differentiating and mature NANs [30, 31]. Knockout experiments have shown that both Phox2a and Phox2b are essential for the differentiation of locus ceruleus NANs .
Phox2a and Phox2b are also expressed in largely overlapping patterns in additional regions of the developing central nervous system [32, 33]. Extensive characterization of Phox2a and Phox2b functions, including mouse knockout experiments, has revealed that these transcription factors can in part substitute for each other's function in neuron specification, differentiation, and maintenance . Major cell types that are influenced by Phox2 proteins, in addition to locus ceruleus NANs, are motor neurons in the mid- and hindbrain [32, 35]. These neuron populations are generated from progenitors that are specified by the ventralizing activity of the signaling factor sonic hedgehog (Shh) [36, 37]. In mouse midbrain motor nuclei, Phox2a is expressed before Phox2b and is essential for the generation of both oculomotor and trochlear motor neurons as revealed by knockout studies . In contrast, Phox2b is expressed before Phox2a in hindbrain visceral motor neurons (vMNs) and these neurons fail to develop in Phox2b, but not in Phox2a, knockout mice [32, 35]. The predominant effect of Phox2b, as well as the unique requirement for Phox2a in the oculomotor and trochlear nuclei, is further shown in gene replacement experiments in which the coding regions of Phox2a is replaced for Phox2b, and vice versa . Phox2b is able to replace Phox2a in specifying NANs, but not midbrain motor nuclei. Conversely, in the specification of NANs and hindbrain vMNs, Phox2a is unable to take over the requirements of Phox2b. Thus, Phox2a seems predominantly important for motor neurons in the midbrain while Phox2b is more critical for NAN and hindbrain vMN development.
Attempts to use stem cell-derived neurons in regenerative medicine or as disease models have been hampered by difficulties to derive neuron cultures in which specific cell types are highly enriched. We have previously shown that transcription factor determinants can be used in forced expression experiments in mouse ESCs (mESCs) and efficiently promote the generation of specific neuronal cell types. Forced expression of Lmx1a and Phox2b could, in this study, be used to generate highly enriched cultures of DANs and vMNs, respectively [38, 39]. In these experiments, mESCs or human ESCs (hESCs) are cultured in the presence of appropriate regional signaling factors that are normally influencing the specification of these specific neuronal populations in the developing neural tube. Thus, the combined influence of appropriate signaling and forced transcription factor expression was shown to generate enriched neuron cultures by a rational strategy. In this study, we show that forced expression of Phox2b in mESCs and hESCs under appropriate signaling conditions generates enriched cultures of NANs or vMNs. In addition, forced expression of Phox2a in mESCs seems particularly effective in generating vMNs, including the ocular motor neuron (OMN) complex. Our studies also demonstrate that engineered neurons can be used to assay drugs that influence these neuron types.
Extrinsic Signaling Increases the Generation of NANs and vMNs from mESCs
We wished to analyze if we could increase the generation of NANs or vMNs in a predictable way by adding extrinsic signaling factors to differentiating mESCs. For these experiments, mESCs were induced to differentiate in monolayer cultures as previously described [38, 40]. After plating, extrinsic signaling factors were added in different combinations at the neural progenitor stage at the indicated time points (Fig. 1A, 1B) with the aim of inducing regional specification according to how distinct signaling events influence the normal generation of distinct cell types within the developing central nervous system (Fig. 1A). Next to the isthmus, the ventralizing activity of Shh leads to the generation of vMNs and other ventral cell types while the dorsalizing activity of BMPs leads to the generation of cell types such as NANs within the developing hindbrain (Fig. 1A). FGF8 was added to cultures alone or in combination with the Shh antagonist Cyclopamine, BMP5, BMP7, or Shh followed by mRNA extraction after 12 days in culture (Fig. 1B). At this time point, cells had differentiated into Tuj1-positive cells with the morphology of postmitotic neurons [38, 40]; data not shown. Expression of NAN and vMN markers was analyzed by quantitative polymerase chain reaction (qPCR) analysis. The results show that antagonizing the ventralizing activity of Shh results in increase of NAN markers (dopamine β-hydroxylase, Dbh; norepinephrine transporter, NET; and transcription factor, Tlx3) (Fig. 1C). Tyrosine hydroxylase (TH), which is expressed by both NANs, ventral midbrain DANs, and other catecholaminergic neuron types was not increased, probably because of the relatively high expression levels of TH in cultures treated with FGF8 alone. The dorsalizing activity of BMPs should also suppress endogenous Shh signaling and promote the generation of dorsal neural cell types. Indeed, addition of either BMP5 or BMP7 together with FGF8 increased the levels of NAN markers. In contrast, and as predicted (Fig. 1A), Shh combined with FGF8 almost completely diminished the expression of Dbh (Fig. 1C) and TH/Phox2a double positive cells were not generated in this condition (Fig. 1E, 1F). In addition, consistent with how signaling influences the generation of vMNs, analysis of vMN markers Isl1 and Tbx20 was increased by addition of Shh while Cyclopamine or the addition of BMP5 or BMP7 strongly decreased their expression (Fig. 1D).
The generation of NANs and vMNs was also analyzed by immunocytochemistry. The results were consistent with qPCR results and showed that the number of NANs (coexpressing TH and Phox2a) was increased by addition of BMP7 but strongly decreased by the addition of Shh (Fig. 1E, 1F). Gata3 is expressed in NANs in the peripheral nervous system; however, TH-expressing neurons cultured here did not express Gata3 indicating that they are, as expected, cells with the characteristics of NANs from the central nervous system (Supporting Information Fig. S2A). Moreover, immunocytochemistry showed that TH-positive neurons did not express the DAN transcription factor Nurr1 (Supporting Information Fig. S2A). The generation of vMNs (Phox2b/Isl1-positive) was increased by the addition of Shh but strongly decreased when cells were differentiated in the presence of BMP7 (Fig. 1E, 1F). These results indicate that the addition of regionally specifying extrinsic signaling factors Shh and BMPs can predictably increase the number of vMNs and NANs, respectively, when added to differentiating mESCs cultured in the presence of FGF8.
Forced Expression of Transcription Factor Determinants Phox2a and Phox2b Increases the Generation of NANs and vMNs
The role of Phox2a and Phox2b in the specification of NANs and vMNs in vivo prompted us to test if forced expression of these factors could further enrich for these neuron types in differentiating mESC cultures. For these experiments, we generated stably transformed mESC lines harboring transgenes encoding either Phox2a or Phox2b under the control of the Nestin gene enhancer region. This ensures expression of Phox2a and Phox2b specifically in Nestin-positive neural progenitors in which the critical specification events occur [30, 31]. As shown in Figure 2A, Phox2a and Phox2b are robustly expressed in the majority of mESC-derived Nestin-positive progenitors in stably transformed mESC lines (referred to as NesE-Phox2a and NesE-Phox2b lines, respectively). We added Shh, Cyclopamine, BMP5, and BMP7 to NesE-Phox2a and NesE-Phox2b lines and found by qPCR and immunocytochemistry that the response to extrinsic signaling factors was similar as in wild-type mESCs (Supporting Information Fig. S1). Moreover, in mESCs differentiated in the presence of FGF8 and BMP7 forced expression of Phox2b leads to increased generation of NANs as indicated by the robustly increased expression of NAN markers (Fig. 2B and Supporting Information Fig. S1), while we did not observe a similar effect for Phox2a. This increase was particularly striking in NesE-Phox2b cells in which NAN markers were increased up to 10-fold above those achieved by FGF8 and BMP7 in wild-type mESCs (Fig. 2B). Analysis by immunocytochemistry confirmed the robust increase in the number of generated NANs as both the absolute number of TH/Tuj1-positive neurons and TH/Phox2a-positive neurons increased in NesE-Phox2b cells under this differentiation condition (Fig. 2C, 2D). The percentage of TH/Phox2a double positive neurons in relation to all differentiated Tuj1-positive neurons was determined to be 17.9%. This was accompanied by a corresponding reduction of Brn3a/Tuj1-positive neurons in the culture (Supporting Information Fig. S2A, S2B), showing that NANs are generated at the expense of other cell types. The forced expression of Phox2b thus resulted in a remarkable enrichment of neurons with properties of NANs with TH/Phox2a double positive neurons comprising almost 50% of all TH-positive neurons under these conditions (Fig. 2D). Moreover, the number of NET/Phox2a double positive neurons was also strongly increased in these cultures (Fig. 2C).
In contrast, when wild-type and transformed mESC lines were cultured under ventralizing conditions in FGF8 and Shh, forced expression of Phox2a and Phox2b resulted in robust increase of differentiated cells with the characteristic expression profile of vMNs as we have previously shown for Phox2b-expressing mESCs . Accordingly, vMN markers, analyzed by both qPCR and immunocytochemistry, were robustly increased in both NesE-Phox2a (Fig. 2E–2G) and NesE-Phox2b differentiated mESCs (Fig. 2E, Supporting Information Fig. S2C). Analysis by immunocytochemistry showed that the number of Phox2b/Isl1 double positive, Phox2b/Isl1/Prph triple positive, and Phox2a/Nkx6.2/Tuj1 triple positive cells was increased in NesE-Phox2a mESCs (Fig. 2F, 2G). The proportion of vMNs (Phox2b/Isl1 double positive cells in relation to all Tuj1-positive cells) was determined to 5.5% in wt ESCs and 61% in NesE-Phox2a-derived neurons. Virtually no TH-positive neurons were seen in cultures derived from the NesE-Phox2a and NesE-Phox2b expressing cells exposed to FGF8/SHH by immunocytochemistry (data not shown) indicating that the expression of Phox2a or Phox2b leads to the suppression of mDA neurogenesis (data not shown; Fig. 3C below). Interestingly, vMN differentiation of NesE-Phox2a mESCs in the presence of Shh and the caudalizing signal retinoic acid gave rise to considerably fewer vMN-like cells as compared to how these cells differentiated in the presence of Shh and FGF8 (Supporting Information Fig. S2D). This may be consistent with the essential role of Phox2a in midbrain motor neurogenesis (see below) . Taken together, these results show that the forced expression of either Phox2a or Phox2b can robustly and predictably increase the enrichment of specific neuron types in a competent signaling environment. For NAN differentiation, the most robust effect was achieved by forced expression of Phox2b (Fig. 2B) while Phox2a and Phox2b increased the number of vMNs to a similar extent (Fig. 2E).
Genome-Wide Gene Expression Analysis of mESC-Derived NANs and vMNs
To further characterize and verify the cell types derived by forced transcription factor expression, we analyzed the genome-wide gene expression profiles of cultured cell types. Accordingly, NesE-Phox2a or NesE-Phox2b mESCs were differentiated under the conditions that lead to enrichment of either NANs or vMNs (Figs. 1, 2). In addition, wild-type mESCs were used as controls. Neurons were further enriched ([mt]85% based on Tuj1 expression; data not shown) by magnetic sorting using antibodies against polysialylated pan-neural cell adhesion molecule (PSA-NCAM) followed by mRNA isolation and microarray analysis (; Affymetrix Mouse Exon ST1.1). Three to four replicates of each condition were analyzed. The average Pearson coefficient between biological replicates was more than 0.992, thus demonstrating a high degree of reproducibility between experiments.
Gene expression array data were used to generate lists of genes enriched in the differentiated cells generated in the conditions leading to enhanced expression of NAN markers (NesE-Phox2b; FGF8/BMP7) or vMN markers (NesE-Phox2a; FGF8;SHH), respectively (further described in Supporting Information Table S1). In agreement with findings from Figures 1, 2, genome-wide expression analysis shows that NesE-Phox2a and NesE-Phox2b mESCs differentiated in the presence of FGF8/Shh and FGF/BMP7 lead to enrichment of vMNs and NANs, respectively. NesE-Phox2b mESCs differentiated in the presence of FGF8 and BMP7 showed a robust increase in the expression of NAN markers, including Tcfap2a, Dbh, Tlx3, and NET, while NesE-Phox2a cells cultured in the presence of FGF8 and Shh showed increase in vMN markers such as Isl1, Nkx6-2, Tbx2, and Tbx20 (Fig. 3A, 3C). Conversely, DAN markers including Foxa1, Foxa2, Nr4a2 (Nurr1), and Lmx1b were decreased indicating that increased generation of NANs and vMNs was parallelled by the suppression of dopamine neurogenesis.
According to our prediction, the genes that are more highly expressed in NesE-Phox2b cells differentiated in FGF8/BMP7 (NAN-list) should be preferentially expressed in NANs while genes more highly expressed in NesE-Phox2a cells differentiated in FGF8/Shh (vMN-list) should be preferentially expressed in vMNs. To test the prediction and provide further validation of the differentiation protocols, genes encoding primarily transcription factors and gene regulatory proteins were randomly selected from the NAN-list and vMN-list and analyzed by in situ hybridization in sections from E12.5 and E15.5 developing mouse midbrain and hindbrain. A total of 32 genes were tested from the NAN-list. Of these, 23 probes gave detectable signals of which 16 (70%) were clearly detected in NANs overlapping with the expression of Dbh mRNA (Fig. 3B, and Supporting Information Table S2). A total of 16 probes from the vMN-list were tested, 12 gave detectable signals and of these 10 (83%) were clearly detected in the region where vMNs are localized as determined by Phox2b in situ hybridization. We have previously shown that NesE-Phox2b mESCs cultured in the presence of FGF8/Shh lead to a strong enrichment in vMNs . Comparing the vMN-list from this study (using NesE-Phox2a mESCs) with our previous analysis (using NesE-Phox2b mESCs) , we found 28 genes on the current vMN-list that were previously verified by in situ hybridization and shown to be expressed in midbrain oculomotor neurons (Supporting Information Table S3). Thus, unbiased in situ hybridization experiments of genes identified from array experiments using enriched mESC-derived neurons provide additional strong support for the conclusion that NANs and vMNs were strongly enriched in cultures from NesE-Phox2b and NesE-Phox2a cultures, respectively. This is a particularly important conclusion for the conditions leading to the generation of NANs since there are only few antibodies for specific NAN markers that could be used in our immunohistochemical analysis of these cells.
Phox2b is expressed in all vMNs, and in OMNs of the midbrain. It is evident from previous studies that hindbrain and midbrain motor neurons share common developmental pathways and many genes are coexpressed in both populations. However, given the critical importance of Phox2a in midbrain motor neuron populations, as defined by knockout experiments , we were interested in the possibility that genome-wide expression analysis of Phox2a-expressing mESCs differentiated in the presence of FGF8/Shh would identify markers expressed selectively in midbrain motor neurons. Indeed, of the 10 genes with detectable expression in vMNs, 3 (Ace, Grb10, and Slc17a8) were preferentially expressed in midbrain oculomotor neurons but not in hindbrain vMNs (Fig. 3D). In contrast, all genes identified in motor neurons in our previous analysis of NesE-Phox2b cells were either expressed in hindbrain vMN only, or in both hindbrain vMN and midbrain motor neuron populations. Thus, these data suggest that NesE-Phox2a, but not NesE-Phox2b, mESCs can differentiate into motor neurons with specific characteristics of motor neurons of the midbrain (see Discussion).
Engineered mESC-Derived Neurons as Tools in Drug Testing
Of strong interest in stem cell research is the possibility to use in vitro engineered neurons to characterize compounds that are potential new drug candidates. We wished to test if protocols to derive vMNs and NANs used here could in principle be exploited for drug testing. 4-(4-(Dimethylamino)styryl)-N-methylpyridinium (ASP+) is a fluorescent substrate of monoamine transporters including the noradrenaline transporter (NET), dopamine transporter, and serotonin transporter (SERT) [41-47]. ASP+ also binds to acetylcholine receptors (AChR) and has been used to visualize neuromuscular junctions [48-53]. When added to cells expressing the appropriate channels or receptors, the cationic dye binds reversibly and emits a red signal that can be detected by fluorescence microscopy. ASP+ has been used for real-time monitoring of neurotransmitter binding and transport kinetics in both neurons and transfected non-neuronal cell lines [41-47].
As shown, NesE-Phox2b mESCs differentiated in the presence of FGF8/Shh or FGF8/BMP7 will result in robust enrichment for neurons with the characteristics of vMNs and NANs, respectively. NesE-Phox2b were cultured under these conditions and exposed to drugs that bind either to NET or AChR (Fig. 4B). Wild-type mESCs treated with only FGF8 were used to generate a more mixed population of neurons (“mixed neurons”; Fig. 4B). The pharmacological action of tricyclic antidepressants such as imipramine is based on their inhibitory effects on NET and SERT to limit the reuptake of noradrenaline and serotonin, respectively. Differentiated neuronal cell cultures were tested for the ability of imipramine to compete with ASP+ for binding and cellular uptake. ASP+ uptake was measured by an imaging system (ImageExpress) so that the percentage of positively stained cells under each condition could be objectively scored (see Materials and Methods) (Fig. 4A). Quantification of how different drugs affect ASP+ uptake reveals a predictable behavior of each drug in relation to the neuronal cell types that are enriched in different cultures (Fig. 4C, 4D). Thus, 10 µM imipramine reduced ASP+ uptake to the greatest extent in cultures enriched in NANs (Fig. 4C). ASP+ uptake was also reduced, albeit to a lesser extent, in vMN-enriched cultures (Fig. 4C), likely related to the anticholinergic effects of imipramine [54-56]. In contrast, mixed neuron cultures were unaffected, even at the highest concentration (10 µM) of imipramine tested. Hence, we are able to show that imipramine competes with ASP+ in a dose-dependent manner, and that competition is most pronounced in cell populations enriched for NANs and vMNs. In contrast, and as would be predicted, the AChR antagonist atropine significantly blocked the binding of ASP+ more efficiently in vMN-enriched as compared to NAN-enriched cultures (Fig. 4D). These results indicate that the mESCs enriched for NANs and vMNs can be used as tools to screen for drugs with distinct specificities for these two different cell types.
Generation of vMNs and NANs from Human ESCs
We next wished to test if forced expression of Phox2a or Phox2b could promote the generation of NANs and vMNs from hESCs. In these experiments, hESCs were differentiated as described previously (; Materials and Methods) (Fig. 5A). At the stage when neural progenitors had formed, cells were treated with different combinations of signaling factors for 10 days and then allowed to mature into postmitotic neurons in neural differentiation media (DM) (Fig. 5A). When hESC-derived neural progenitors were specified with FGF8 alone, TH/Phox2a coexpressing neurons indicative of NANs were not observed even after 45–50 days in culture (Fig. 5C). Based on our experiments in mESCs, we tested the effects of BMP7 together with FGF8 for the derivation of human NANs. However, we found that this resulted in lower expression of NAN markers (Fig. 5B), presumably due to intrinsic differences in how hESCs and mESCs respond to in vitro culturing conditions. Indeed, in contrast to what was observed in mESC differentiation and consistent with previously published data, BMP7 drastically reduced the generation of Tuj1-postive neurons in the culture, suggesting that the lower expression of NAN markers is a result of decreased neurogenesis in cultures exposed to BMP7 [58-60]. Moreover, the TH-expressing neurons in the BMP7-treated culture did not coexpress Phox2a (Fig. 5B). Since BMP7 did not promote NAN marker expression in hESC differentiation experiments, we next tested how forced transcription factor expression influenced cells cultured in FGF8 alone. Lentiviral vectors for either green fluorescent protein (GFP), Phox2a (L-Phox2a), and Phox2b (L-Phox2b) were added to suspensions of dissociated neuroepithelial rosettes that were then plated in the presence of FGF8. We noted that TH/Phox2a double positive neurons appeared only in L-Phox2b, but not in Lentivirus-GFP (L-GFP) infected cultures (Fig. 5C). qPCR analysis confirmed the upregulation of NAN markers, including DBH, NET, PHOX2A, and PHOX2B, in L-Phox2b infected cultures (Fig. 5D). Thus, forced Phox2b expression can promote the generation of cells with the characteristics of NANs in hESCs cultured in the presence of FGF8.
When hESCs were differentiated and cultured in the presence of FGF8 and Shh, a significant and predictable upregulation of vMN markers was noted as determined by qPCR (Fig. 5E). Forced expression of either Phox2a or Phox2b by transduction with lentivirus vectors (L-Phox2a and L-Phox2b) further increased the expression of vMN markers including ISL1, PHOX2A, and PHOX2B; however, Phox2b was much more effective in promoting vMN neurogenesis (Fig. 5F). Finally, immunocytochemistry was used to demonstrate that cells with the expression profile of vMNs were induced by L-Phox2a and L-Phox2b. Thus, L-Phox2a infected cells contained a large number of differentiated neurons coexpressing Phox2b/Isl1 and PRPH/Isl1 while L-Phox2b infected cells generated numerous Phox2a/Isl1 and PRPH/Isl1 coexpressing neurons (Fig. 5G). Taken together, these data show that forced expression of Phox2 transcription factors can be used to increase the yield of neurons with the characteristics of both NANs and vMNs.
It is evident that one of the major challenges in stem cell biology is to engineer clinically important cell types, either for cell replacement or for drug development. An important goal is therefore to develop reproducible methods to obtain well-defined and pure populations of clinically relevant cell types. However, the complexity of the central nervous system, with hundreds of different neuronal and non-neuronal cell types, provides a particularly difficult challenge. In this study, we have extended our previous work and demonstrated that forced expression of specific transcription factor determinants has the potential to very efficiently promote the generation of specific neuronal cell types.
Many studies have shown that the appropriate signaling factors can specify specific neuron types from stem cells. However, heterogeneity in culturing conditions generally results in protocols that give rise to mixed neuron populations and in many cases the desired neuron type only constitutes a relatively limited fraction of desired neurons. The data presented here, together with our previous studies, show that transcription factor-mediated lineage reprogramming provides a highly efficient method to generate neuronal cell types under the appropriate signaling conditions that render neural progenitors competent for lineage reprogramming by transcription factor expression. A rational strategy, that has been employed here, is therefore to first achieve culturing conditions for a permissive environment in which neural progenitors are competent to be influenced by a given transcription factor determinant. In mESCs, we thus found that the combination of BMP7 and FGF8 induced a regional character resembling the dorsal mid-/hindbrain from which a fraction of cells expressed NAN markers. Importantly, the forced expression of Phox2b under these conditions further promoted NAN differentiation 10-fold.
The conclusion that a stem cell differentiation protocol gives rise to a particular cell type often depends on the analysis of only few markers by immunocytochemistry and/or qPCR. However, the use of stem cell-derived neurons either in cell therapy and/or in vitro analyses will ultimately involve a more stringent verification of cell types. An important outcome of the experiments presented here was the efficient generation of cells expressing the characteristic markers of NANs and vMNs, which made it possible to perform genome-wide gene expression analysis that compared gene expression profiles of distinct differentiated cell populations obtained from the different cell lines. Importantly, data from the analyses resulted in an unbiased verification of cell types by in situ hybridization from enriched gene lists. Thus, 70% and 83% of riboprobes taken from the two gene lists with expression data of NANs and mid-/hindbrain motor neurons showed the predicted expression in in situ staining in NANs and vMNs, respectively. These results further verify that the generated neurons express a large number of genes that normally are expressed in NANs and mid-/hindbrain motor neurons, respectively, and strongly support the conclusion that the mESC differentiation protocols used here generates the expected and correctly differentiated cell types. However, it is important to point out that a full validation of functional NANs should include electrophysiological analyses and grafting experiments to reveal the ability of engineered neurons to integrate into a normal circuitry. Such experiments will be important to pursue in future studies.
The Phox2a/Phox2b transcription factors have key specifying roles for several populations of central neuron types, including mid-/hindbrain motor nuclei and NANs. An interesting question relates to the specific roles of each transcription factors in different cell types. Mammalian homologs Phox2a and Phox2b have identical DNA-binding homeodomains but distinct carboxy-terminal domains . The distinction in their functions seems in part to be related to their temporal expression in each cell type. Thus, in locus ceruleus NANs and in midbrain motor neurons, Phox2a precedes, and is essential, for the later induction of Phox2b. Reciprocally, Phox2b precedes the expression of Phox2a in hindbrain vMNs. Although cell replacement experiments have demonstrated redundant functions they have also suggested that Phox2b has specific instructive roles that cannot be replaced by those of Phox2a . Accordingly, Phox2a knockin into the Phox2b gene locus cannot replace the function of Phox2b in the generation of either NANs or hindbrain motor neurons. Reciprocally, Phox2b is unable to substitute for the function of Phox2a in the generation of midbrain motor nuclei, including oculomotor and trochlear neurons. It is therefore intriguing that the in situ hybridization of genes enriched in NesE-Phox2a mESCs differentiated in the presence of FGF8 expressed several genes that were specifically present in oculomotor neurons but not in hindbrain vMNs. This is in contrast to the genes that we previously identified to be expressed in hindbrain vMNs after forced expression of Phox2b since all those genes expressed in midbrain motor nuclei  were also expressed in the hindbrain vMNs. Thus, it seems as if the Phox2a-specified cells have acquired more of a midbrain character and further emphasizes the specific instructive function of Phox2a for midbrain motor neurons. It is also notable that the three preferentially expressed genes (Ace, Grb10, and Slc17a8), to our knowledge, represent the first identified specific markers for midbrain oculomotor and trochlear neurons, whose gene expression profile otherwise appears to be highly related to that of hindbrain vMNs.
Stem cell-derived neurons are potentially important in future cell replacement therapies, for example, in Parkinson's disease, where the grafting of DANs provides a promising but highly challenging future treatment strategy. However, using cells for studies in vitro is also highly important, for example, in efforts to characterize potential drug candidates, signaling factors, or genes that influence a certain cellular pathway in the studied cell type. The discovery of iPS cells has also opened up entirely new avenues for modeling disease in vitro. Neurodevelopmental disorders may in this regard be particularly relevant since the iPS cell-derived cell engineering recapitulates normal developmental events. Rett syndrome is therefore interesting in relation to the experiments presented here, since a significant abnormality in Rett syndrome patients is associated with noradrenergic neurotransmission, as revealed both from human studies and in Rett syndrome mouse models in which the Rett syndrome-causing mutation in the MeCP2 gene has been introduced [20, 21, 61, 62]. Our data provide a first study reporting a robust protocol for NAN generation and we envisage that further development of these methods directed at NAN generation from human iPS cells will have potential for studies of Rett syndrome-affected cells in vitro. It is notable that several groups have already generated human iPS cell lines from Rett syndrome patients and found neuronal deficiencies in cells derived from these cell lines [63-67]. Thus, further studies focusing specifically on cellular abnormalities in patient-derived NANs should now be feasible.
Using in vitro differentiated cells in drug development will use a growing number of cellular assays and disease models as additional clinically relevant cell types can be engineered from stem cells. Neurons derived from in vitro cultured stem cells can represent a reproducible and biologically relevant source of cells with possible large-scale production for cellular assays. Cell types that we have generated here and in our previous studies include DANs, serotonergic neurons, different types of motor neurons, and NANs. These cell types are clinically relevant in a number of different neurological and psychiatric disorder, and their neurotransmitter signaling machineries are targets of drugs that are already in use. For example, drugs important for the treatment of depression include those that target neurotransmitter transporters found in, for example, DANs, serotonergic neurons, and in NANs. We show here that imipramine, a drug binding to such transporters, can be tested in the context of in vitro engineered neurons using a fluorescent dye binding to several transporters. Such cellular assays often use non-neuronal cells artificially expressing the neurotransmitter transporters, which may lead to different observations as compared to when the appropriate neuronal cell type is instead considered. Imipramine blocks neurotransmitter reuptake but also shows unwanted side effects, including anticholinergic activity, a property that should be amenable to investigation using in vitro differentiated motor neurons. The assay was performed using a high-throughput imaging system (ImageExpress) to minimize bias and increase efficiency. Thus, these studies provide a proof-of-concept for a strategy that could prove important for the analysis of drugs binding norepinephrine transporters in the context of NANs.
The potential to generate specific neuronal cell types in more enriched cultures will prove important both in future cell replacement therapies and in studies using cultured stem cell-derived neurons as experimental platforms. The results provided here extend the panel of clinically important cell types that can now be generated, and in doing so establish a strategy for NAN studies using pluripotent stem cells, including patient-derived iPS cells.
Materials and Methods
Maintenance and Differentiation of ESCs
Mouse embryonic stem (mESC) line (E14) was propagated on Corning® CellBIND® flasks in Dulbecco's modified Eagle's medium (DMEM) high glucose (Invitrogen™, Life Technologies™, Stockholm, Sweden, http://www.lifetechnologies.com) supplemented with 1,000 U/ml ESGRO(r) leukemia inhibitory factor (LIF) (Chemicon (Merck Millipore), Temecula, CA, http://www.millipore.com), 15% ES qualified fetal bovine serum (FBS) (Invitrogen™, Life Technologies™, Stockholm, Sweden, http://www.lifetechnologies.com), 1 mM sodium pyruvate, 0.1 mM nonessential amino acids, and 0.1 µM β-mercaptoethanol (Sigma-Aldrich, Stockholm, Sweden, http://www.sigmaaldrich.com). To differentiate mESCs to neurons, mESCs (1.4 × 104 cells per square centimeter) were seeded as a monolayer as previously reported [40, 68] onto gelatinized Corning® CellBIND® plates in DM (1:1 DMEM/F-12-Neurobasal medium) supplemented with 1× B27 and 1× N2 (cell culture media and supplements were from Invitrogen, Life Technologies™, Stockholm, Sweden, http://www.lifetechnologies.com). Cultures were supplemented with 2 ng/ml bFGF and 100 ng/ml FGF8b (R&D systems, Minneapolis, MN, http://www.rndsystems.com/) till day 6 of differentiation. For noradrenergic differentiation, the following growth factors were added from day 4 to day 6 of differentiation: 10 ng/ml BMP7 (R&D systems), 10 ng/ml BMP5 (R&D systems), or cyclopamine 1 µM (Calbiochem (EMD Millipore Chemicals), Darmstadt, Germany, http://www.emdmillipore.com). For dopaminergic and visceral motor neuronal differentiation, 100 nM Hedgehog agonist Hh-Ag1.3 (Curis, Inc., Curis, Lexington, MA, http://www.curis.com)  was added from day 0 to day 6 of differentiation, as described in .
Human embryonic stem (hESC) lines (H9 WiCell p26–29 and HS293 p29–32) were maintained as previously described . To induce differentiation, hESC colonies were mechanically dissociated and grown as embryoid bodies (EBs) in the differentiation medium (DM) supplemented with 1× B27 and 1× N2 for 5–7 days. EBs were plated on growth-factor reduced extracellular matrix (ECM) (Sigma-Aldrich, Stockholm, Sweden, http://www.sigmaaldrich.com)-coated dishes in DM supplemented with 10 ng/ml bFGF and grown for 7 days until neuroepithelial rosettes appearance. Neuroepithelial rosettes were mechanically isolated, disaggregated into the small clusters (20–100 cells) by pipetting, infected with lentiviral vectors (MOI 10) in suspension, and plated on ECM-coated dishes in DM. Growth factors FGF8b (100 ng/ml), BMP7 (10 ng/ml), and Hh-Ag1.3 (100 nM) were added to the plated cells for 10 days. After which, all factors were withdrawn and cells were replated by mechanical dissociation into small clusters. Replated cells were left to mature in neural DM for 25 days. The two different hESC lines gave similar results.
Generation of Stable ESC Lines and Lentiviral Transduction
We cloned cDNA constructs encoding mouse Phox2aIRESGFP or Phox2b (kind gifts from JF Brunet) into an expression vector driven by the Nestin enhancer, then used these to generate stable ESC lines as described [38, 68]. The same cDNA constructs were cloned into the lentiviral expression vector pRRLSIN.cPPT.PGK-GFP.WPRE (Addgene, Cambridge, MA, http://www.addgene.org/) for transduction in hESCs.
Magnetic Cell Sorting
After 12 days of differentiation, mESC-derived cultures were purified for neurons using MACS technology (Miltenyi Biotec Norden AB, Lund, Sweden, https://www.miltenyibiotec.com/en/) according-to-protocol. Briefly, cells were trypsinized, rinsed in phosphate-buffered saline (PBS), then incubated in PSA-NCAM antibody (1:1,000; Millipore) followed by anti-mouse IgM microbeads (1:10). Magnetically labeled neurons were pulled down by passing them through MACS LS columns using the QuadroMACS separator. Only sortings with a purity of ≥85% Tuj1+ neurons were used in experiments.
Cells were fixed with 2% paraformaldehyde for 30 minutes, then washed three times with PBS. Blocking solution containing 10% FBS and 0.1% Triton X-100 in PBS was added for 60 minutes, and primary antibody was added overnight at 4°C. After further rinses in PBS, 4′,6-diamidino-2-phenylindole (DAPI) and secondary antibodies were added for an hour. Stained cells were visualized and imaged using the Zeiss Axio Imager confocal microscope. Primary antibodies used are as follows: rabbit anti-TH (1:1,000; Pel-Freez Biologicals, Rogers, Arkansas, www.pelfreez-bio.com), mouse anti-TH (1:500; Millipore), mouse anti-NET (1:500; MAb Technologies, MAb Technologies, Stone Mountain, GA, http://mabtechnologies.com), rabbit anti-Phox2a (1:1,000; JF Brunet), guinea-pig anti-Phox2b (1:1,000; J Ericson), mouse/rabbit anti-Tuji (1:2,000; Covance Research Products, Covance, Princeton, New Jersey. http://www.covance.com), and rabbit anti-Prph (1:1,000; Millipore, Chemicon (Millipore), Billerica, MA, http://www.millipore.com).
In Situ Hybridization
Tissue preparation: E12.5 and E15.5 embryos were harvested from wild-type C57BL6 pregnant mice and fixed in 4% paraformaldehyde for 2 and 5 hours, respectively, shaking. Fixed embryos were dehydrated in 30% sucrose overnight at 4°C, then mounted in Tissue-Tek OCT Compound. Cryosections (12 µM) of mouse embryonic brain were made for in situ hybridization.
Probes design and preparation: in situ probes were made with primers designed to label PCR products with a T3 or T7 RNA polymerase promoter in order to carry out in vitro transcription to generate antisense transcripts. E14.5 embryonic mice cDNA was used as the template. Primer sequences were extracted from the Allen Brain Atlas or Genepaint databases if available, or designed using Primer-Blast.
In situ hybridization was done with standard procedures. Briefly, cryosections were fixed, proteinase K treated, acetylated, and probes were hybridized overnight at 70°C. Secondary anti-DIG antibody was added overnight at 4°C, then slides were subjected to washes in decreasing salt concentrations before they are developed in nitro-blue tetrazolium (NBT)/5-bromo-4-chloro-3-indolyl-phosphate (BCIP) solutions.
High-Throughput Imaging Drug Assay
Three days before the assay, 8 × 104cells/well at day 10 of differentiation were seeded evenly onto polyornithine and laminin-coated 96-well imaging plates (BD Falcon™, Belgium, http://www.bdbiosciences.com/cellculture/microplates/, #353219). Imipramine, Desipramine, and Atropine (Sigma) were incubated with the cells for 15 minutes before 1 µM of fluorescent dye ASP+ (4-(4-(dimethylamino)styrl)-N-methyl-pyridinium, Sigma-Aldrich, Stockholm, Sweden, http://www.sigmaaldrich.com)  was added. The drugs and fluorescent dye were left on the cells for 60 minutes before they were rinsed three times with DM (without phenol red). Hoechst 33342 (2 µg/ml) (Invitrogen™, Life Technologies™, Stockholm, Sweden, http://www.lifetechnologies.com) was used to stain the live cell nuclei. The plate was immediately imaged with the ImageXpress Micro high content screening system (Molecular Devices, Sunnyvale, CA, http://www.moleculardevices.com). Images in two fluorescent channels, Cy3 and DAPI, were taken from four to six sites per well and were analyzed using the MetaXpress high content image acquisition and analysis software (Molecular Devices). Cells were scored as “positive” by the software if they meet the following criteria: DAPI; Max width: 20 µM, min width: 6 µM, intensity above background: 50 graylevels; Cy3; max width: 25 µM, min width: 8 µM, intensity above background: 80 graylevels.
RNA Extraction and qPCR
Cells were lysed with buffer RLT (with β-mercaptoethanol) and the lysate was passed through Qiashreddar columns (Qiagen AB, Sollentuna, Sweden, www.qiagen.com) before freezing them down at −80°C. RNA was extracted according to instructions stated in the Qiagen RNeasy mini/micro kit. cDNA was made from extracted RNA using Superscript III (Invitrogen) and OligodT primers (Invitrogen), according to manufacter's recommendations.
QPCR was conducted on the ViiA 7 Real-Time PCR System (Applied Biosystems®, www.lifetechnologies.com) using Taqman assay probes or self-designed primers (Supporting Information Table S4), following default PCR conditions (Tm = 59°C). Gene expression was normalized to Rpl19 expression levels and fold changes in expression were calculated using the 2−ΔΔCT method.
RNA with a RIN (RNA Integrity Number) of ≥9.5 (measured by an Agilent Bioanalyzer) was hybridized onto Affymetrix GeneChip Mouse Exon ST1.1 Array according to manufacter's recommendations. Hybridization, raw data processing, and brief data analysis were conducted in the Bioinformatics and Expression Analysis Core Facility at Novum, Karolinska Institute. CEL files were imported into Agilent's GeneSpring GX12.0 software for analysis. Probe data were summarized using the Exon RMA16 algorithm, then subjected to baseline transformation to median of all samples. Quality control of data was performed by filtering raw intensity signals according to their percentile expression (upper cut-off: 100; lower cut-off: 20) in at least three replicates of all experimental groups. Lists of differentially expressed genes were generated by comparing groups using unpaired t tests and multiple testing correction with Benjamin Hochberg FDR (adjusted p-value [mt].05).
To detect differences that are statistically significant, data from replicates were subjected to t test for independent samples (two-tailed, equal variance). Values are presented as mean ± SEM. Error bars represent SEM. * indicates p-value <.05. ** indicates p-value <.01. *** indicates p-value <.001. For cell counting experiments, two to three replicates of at least five different fields were taken into consideration.
The results presented here demonstrate that NANs can be efficiently derived from ES cells by forced expression of Phox2b under appropriate signaling conditions. By shifting signaling conditions, the competence of cultured stem cells will change so that Phox2b and the related Phox2a instead direct the efficient generation of mid- and hindbrain vMNs. Both NANs and vMNs are interesting from a clinical perspective. Thus, these studies expand the repertoire of clinically relevant neuronal cell types that can be generated from pluripotent stem cells. Such programmed cells may be important as in vitro models of disease, in drug screening or in cell therapy.
We will like to thank members of the Perlmann and Ericson group for helpful discussions; Jean-François Brunet for providing the Phox2a antibody; Florian Salomons and Bhumica Singla for invaluable technical support. This work was supported by funds from European Union FP7 grants (NeuroStemCell and mdDANEURODEV), Vetenskapsrådet and Swedish Foundation for Strategic Research.
J.M.: conception and design, collection and/or assembly of data, data analysis and interpretation, manuscript writing, and final approval of manuscript; L.P.: conception and design, collection and/or assembly of data, data analysis and interpretation, and final approval of manuscript; Z.A. and N.K.: collection and/or assembly of data; L.W.S. and J.E.: final approval of manuscript; T.P.: conception and design, financial support, manuscript writing, and final approval of manuscript. J.M. and L.P. contributed equally to this article.
Disclosure of Potential Conflicts of Interest
The authors indicate no potential conflicts of interest.