Signalling the Future
Neural differentiation of mouse embryonic stem cells M.P. Stavridis and A.G. Smith1 Institute for Stem Cell Research, University of Edinburgh, The King’s Buildings, West Mains Road, Edinburgh EH9 3JQ, U.K.
Abstract Pluripotent embryonic stem cells can give rise to neuroectodermal derivatives in culture. This potential could be harnessed to generate neurons and glia for cell-replacement therapies in the central nervous system and for use in drug discovery. However, current methods of neural differentiation are empirical and relatively innefficient. Here, we review these methodologies and present new tools for quantification, analysis and manipulation of embryonic stem cell neural determination.
Introduction Mouse embryonic stem (ES) cells are pluripotent cell lines derived from the inner cell mass of the blastocyst. Their properties include a stable diploid karyotype, the ability to participate fully in foetal development upon re-introduction into a blastocyst and the capacity to differentiate into a plethora of cell types in vitro. The maintenance of the undifferentiated state of ES cells in vitro is dependent on selfrenewing cell divisions in the presence of cytokines which signal through the gp130 receptor, e.g. leukaemia inhibitory factor (LIF) (reviewed in ). Upon withdrawal of a selfrenewal stimulus, ES cells rapidly lose their undifferentiated phenotype and differentiate. These properties have made ES cells a popular system both to introduce specific genetic alterations (gene targeting) into animals and to study gene function during differentiation in vitro [2,3]. To date, the best-studied mode of ES cell differentiation is the formation in suspension culture of multicellular aggregates called embryoid bodies (EBs) . Within these aggregates, complex interactions between heterologous cell types result in the induction of differentiation of stem cells to derivatives of all three embryonic germ layers . Plating of the embryoid bodies causes further differentiation and outgrowth.
Neural differentiation in EBs The generation of ES cell-derived neural cells, especially neurons [6–8], is a much-studied example of ES cell differentiation. Neural differentiation of ES cells has been achieved by several different protocols, some of which are strikingly different. In the protocols that were published first, EBs were treated with retinoic acid at different time windows and then plated on to laminin , gelatin  or tissue
Key words: in vitro neurogenesis, neuron, neural progenitor. Abbreviations used: BMP, bone morphogenetic protein; EB, embryoid body; ES, embryonic stem; FACS, fluorescence-activated cell sorting; FGF, fibroblast growth factor; GFP, green fluorescent protein; LIF, leukaemia inhibitory factor; Shh, Sonic hedgehog. 1 To whom correspondence should be addressed (e-mail [email protected]
culture plastic . Cells with overt neuronal morphology appeared after plating, and were found to express neuronspecific genes such as neurofilament light chain, microtubuleassociated proteins 2 and 5, synaptophysin and others. These cells were found to respond to a range of neurotransmitters and depolarizing currents, confirming that they were indeed excitable neurons. Glial cell types also appeared in such differentiated cultures, as judged both by morphology and expression of specific glial markers. The majority of glial cells produced were astrocytes, but oligodendrocytes have also been generated and selectively expanded from EB cultures . A variant protocol for neural differentiation of ES cells involved the formation of EBs without retinoic acid treatment, but depended on the subsequent culture of the attached EBs in a selective, serum-free medium to eliminate non-neural cells. Culture of the EB-derived cell pool in this selective medium results in a dramatic decrease in cell number, as the majority of the cells do not survive these culture conditions. The surviving portion is enriched for nestin-positive neural progenitor cells , which can be expanded and induced to differentiate into neurons with high efficiency . The majority of neurons generated from ES cells following differentiation in EBs are GABAergic (where GABA stands for γ -aminobutyric acid), with some glutamatergic neurons also generated at lower frequency. In order to generate dopaminergic neurons (which are potentially useful for treating Parkinson’s disease) and serotonergic neurons, Lee et al.  treated ES cellderived neural precursors with Sonic hedgehog (Shh) and fibroblast growth factor (FGF)-8. Both of these molecules are important for patterning the ventral midbrain and hindbrain , where such neurons are generated in vivo. More recently, another group successfully generated spinal cord motor neurons from ES cells by treating differentiating ES cells with a posteriorizing (retinoic acid) and a ventralizing (an Shh agonist) molecule in order to impart on them the ventral spinal identity of motor neuron precursors . The fact that the neural precursors generated from ES cells can respond to morphogenetic cues suggests that they are (at least to C 2003
Biochemical Society Transactions (2003) Volume 31, part 1
some degree) malleable, and parallels can be drawn between neural differentiation of ES cells and neural development in vivo.
Stromal cells A distinct system for differentiation of ES cells into dopaminergic neurons has been described . This system differs from the other methods for neural differentiation in that the differentiation process happens not in aggregates, but via the co-culture of ES cells on monolayers of bone marrow-derived stromal cells (PA-6 cells). Retinoic acid is not used to induce neural fate and Shh or FGF-8 are not added to the cultures to induce dopaminergic neurons. This has been interpreted in favour of an inductive mechanism for the production of neurons from ES cells, although the molecular nature of the inducing signal(s) remains elusive. It is noteworthy that PA6-mediated neural differentiation is suppressed in the presence of serum or bone morphogenetic protein (BMP)-4 which direct cells to alternative fates.
Defined media The first indication that ES cell neural determination is under inhibitory control and may resemble ‘default’ amphibian neural induction [16,17] came from experiments in serum-free medium. Using EBs generated in chemically defined serumfree medium, Wiles and Johansson  showed that neural markers were expressed unless signals of the transforming growth factor superfamily (including BMPs and activin) were present. In the presence of activin A, an inducer of the mesoderm in the frog, mesodermal as well as neural markers appeared, along with a strong induction of the activin antagonist, follistatin . Treatment with BMP-4 resulted in the loss of neural cell markers, as is seen with frog ectodermal explants . BMP-4 also suppressed neural differentiation of ES cells in a related system . In this case, ES cells were selected in serum-free conditions in the presence of LIF, and at a frequency of 0.1% gave rise to nestin-positive cell aggregates (‘neurospheres’) which could be then expanded in FGF-2 or induced to differentiate into neurons, astrocytes or oligodendrocytes. ES cells that lack a component of the BMP signalling pathway gave rise to 3.5-fold more spheres, suggesting that differentiation in this system is under inhibitory control by BMPs . Sphere formation was inhibited by blocking antibodies to FGF-2, possibly due to a requirement of FGF signalling for expansion of neural cells. The low efficiency of colony formation under these conditions suggests a strongly selective mechanism that precludes the drawing of any conclusions concerning neural determination.
Lineage selection A hurdle in the study of neural cell generation, expansion and differentiation using ES cells as a model system is that, C 2003
even with the best available protocols, differentiation is not uniquely neural. All of the available protocols for neural differentiation result in the generation of multiple cell types (often of more than one germ layer) and the progression from ES cell via a committed neural precursor to a fully differentiated, post-mitotic neural cell is determined post hoc either by antibody staining or gene expression profiling. The presence of alternative differentiation hampers such analyses, as it is not possible to determine if the effect of added factors is direct or via other differentiated cell types in the culture. Furthermore, one potential use of ES-derived neural precursors is as material for transplantation studies for the treatment of neurodegenerative diseases such as Parkinson’s or Alzheimer’s disease, or neural damage following stroke or injury. For such studies it is important that any ES cells are eliminated from the transplanted tissue because they can generate teratocarcinomas [21–23]. We  have used the technique of lineage selection on differentiating ES cells to purify neural precursors (Figure 1a). This technique relies on the introduction of a reporter/selection cassette into a locus with restricted expression in the desired cell type by homologous recombination. The application of a selection drug eliminates cells that do not express the gene, resulting in a pure population of Sox2+ cells. A drawback of our original study was that the gene used for selection (Sox2) is expressed not only by neural precursors, but also by ES cells. Consequently, selection for Sox2+ cells does not eliminate undifferentiated ES cells from the culture. This is significant because persistence of ES cells at some level is a common feature of current differentiation methods. To address this we have engineered an ES cell line (46C) in which one copy of the pan-neural gene Sox1 [25,26] has been replaced by the dual selection/reporter cassette egfpIRESpac, which confers cell-autonomous green fluorescence and puromycin resistance to cells that express Sox1 (Figure 1b). Sox1 belongs to the same family of high-mobility group transcription factors as Sox2, but its expression is restricted to proliferating neuroectodermal cells and the lens . In addition to the restricted expression of the reporter, 46C cells have the advantage of a vital label (green fluorescent protein, GFP) meaning that activation of Sox1 can be monitored in living cells. As Sox1 is not expressed in ES cells, undifferentiated 46C ES cells are not fluorescent. Upon neural differentiation (in EBs treated with retinoic acid, on PA6 or in serum-free medium), Sox1 is activated and the cells produce green fluorescence. This property of 46C cells enables us to purify both neural and non-neural cells generated during differentiation and analyse their phenotype. We used the activation of Sox1–GFP as a marker for neural cells, and measured it by flow cytometry during 46C ES cell differentiation. This enables us to measure the proportion of neural cells at any point during differentiation (Figure 2). Another advantage of the Sox1–GFP reporter is that it enables time-lapse observation of the differentiation process with minimum perturbation of the culture. This method of lineage identification is sensitive enough to detect slight changes in
Signalling the Future
Sox1 lineage selection (a) Diagrammatic representation of the lineage selection scheme. Endodermal and mesodermal cells derived from ES cells are eliminated either by antibiotic selection or flow cytometry, while neurectodermal cells can be amplified and induced to differentiate to neurons, astrocytes and oligodendrocytes. The inset shows ES cell-derived neurons visualized by antibody staining for the neuronal marker βIII-tubulin. Scale bar, 50 µm. (b) Sox1 targeting scheme. The entire Sox1 open reading frame has been replaced by the selection/reporter cassette egfpIRESpac. A cassette for selection in ES cells (Hy-TK) containing its own promoter (arrow) flanked by site-specific recombination sites (Lox; arrowheads) is included, as Sox1 is not expressed by undifferentiated ES cells, but is removed by a transient transfection with a recombinase (Cre)-expressing plasmid.
the efficiency of neural determination or the kinetics by which this process is taking place, making it ideal for the study of the molecular requirements for neural differentiation of ES cells.
Figure 2 Quantification of neural differentiation of 46C ES cells ES cells were allowed to form EBs in the absence of LIF. On day 4, 10−6 M all-trans retinoic acid (RA) was added to half of the plates for a further 4 days. A sample from each culture was analysed for Sox1–GFP fluorescence by flow cytometry every 2 days. Data points are means ± S.D. of three experiments (data courtesy of Dr Meng Li).
Specific regulatory molecules or inhibitors can be introduced into the system and their effect on neural determination can be measured accurately. We have used this system to develop a defined culture system for efficient neural differentiation without aggregation or co-culture . This provides a platform for dissecting the molecular mechanisms underlying neural lineage specification. Furthermore, because the GFP reporter enables the use of cell sorting, it is not necessary to kill off non-neural cell types by drug selection in order to purify the culture. Use of fluorescence-activated cell sorting (FACS) allows for the isolation of both Sox1–GFP-positive and -negative cells for further culture or analysis. Purified populations of Sox1– GFP-positive neural progenitors will allow the study of growth factors, growth factor inhibitors and morphogens on a neural progenitor population, without intermediate effects from non-neural cells. This may enable further study of the factors that direct neural precursors to specific fates, with a view to using such ES cell-derived cell types for treatment of specific neurological conditions. FACS-purified populations can also be used to construct subtractive cDNAs and examine microarray expression profiling in order to C 2003
Biochemical Society Transactions (2003) Volume 31, part 1
Figure 3 Identification of novel genes involved in neural determination After ES cell differentiation, populations of Sox1–GFP+ and Sox1–GFP− cells are isolated by FACS. Following mRNA extraction, subtraction, cDNA synthesis and library generation, array analyses can be used to identify candidate genes. After bioinformatic analysis and obtaining in vivo expression data, candidate genes can be introduced into ES cells for functional analysis of their effect on neural differentiation .
identify genes involved in mammalian neural determination (Figure 3) .
Concluding remarks Generation of differentiated cell lineages from ES cells is a very promising tool for both the study of development and the generation of transplantation material for the treatment of a wide range of degenerative disorders and lesions. In order for this potential to be realized, the differentiation of ES cells must be harnessed to produce suitable cells for the treatment of each condition, eliminate unwanted or harmful cell populations and optimize the conditions for successful graft integration and survival. These objectives highlight the importance of understanding and directing initial lineage choice so that we can make desired cell types with maximum efficiency and reproducibility. Experience to date with generation of neural cell types suggests that addressing all of these issues successfully will involve a combination of both instructive and selective approaches.
References 1 Smith, A.G. (2001) Annu. Rev. Cell. Dev. Biol. 17, 435–462 2 Guan, K., Schmidt, M.M., Ding, Q., Chang, H. and Wobus, A.M. (1999) Altex 16, 135–141 3 O’Shea, K.S. (1999) Anat. Rec. 257, 32–41 4 Doetschman, T.C., Eistetter, H., Katz, M., Schmidt, W. and Kemler, R. (1985) J. Embryol. Exp. Morphol. 87, 27–45 C 2003
5 Martin, G.R., Wiley, L.M. and Damjanov, I. (1977) Dev. Biol. 61, 230–244 6 Bain, G., Kitchens, D., Yao, M., Huettner, J.E. and Gottlieb, D.I. (1995) Dev. Biol. 168, 342–357 7 Strubing, C., Ahnert-Hilger, G., Shan, J., Wiedemann, B., Hescheler, J. and Wobus, A.M. (1995) Mech. Dev. 53, 275–287 8 Fraichard, A., Chassande, O., Bilbaut, G., Dehay, C., Savatier, P. and Samarut, J. (1995) J. Cell Sci. 108, 3181–3188 9 Liu, S., Qu, Y., Stewart, T.J., Howard, M.J., Chakrabortty, S., Holekamp, T.F. and McDonald, J.W. (2000) Proc. Natl. Acad. Sci. U.S.A. 97, 6126–6131 10 Lendahl, U., Zimmerman, L.B. and McKay, R.D.G. (1990) Cell 60, 585–595 11 Okabe, S., Forsberg-Nilsson, K., Spiro, A.C., Segal, M. and McKay, R.D. (1996) Mech. Dev. 59, 89–102 12 Lee, S.H., Lumelsky, N., Studer, L., Auerbach, J.M. and McKay, R.D. (2000) Nat. Biotechnol. 18, 675–679 13 Ye, W., Shimamura, K., Rubenstein, J.L., Hynes, M.A. and Rosenthal, A. (1998) Cell 93, 755–766 14 Wichterle, H., Lieberam, I., Porter, J. and Jessell, T. (2002) Cell 110, 385 15 Kawasaki, H., Mizuseki, K., Nishikawa, S., Kaneko, S., Kuwana, Y., Nakanishi, S., Nishikawa, S.I. and Sasai, Y. (2000) Neuron 28, 31–40 16 Hemmati-Brivanlou, A. and Melton, D. (1997) Cell 88, 13–17 17 Hawley, S.H., Wunnenberg-Stapleton, K., Hashimoto, C., Laurent, M.N., Watabe, T., Blumberg, B.W. and Cho, K.W. (1995) Genes Dev. 9, 2923–2935 18 Wiles, M.V. and Johansson, B.M. (1999) Exp. Cell. Res. 247, 241–248 19 Wilson, P.A. and Hemmati-Brivanlou, A. (1995) Nature (London) 376, 331–333 20 Tropepe, V., Hitoshi, S., Sirard, C., Mak, T.W., Rossant, J. and van der Kooy, D. (2001) Neuron 30, 65–78 21 Bjorklund, L.M., Sanchez-Pernaute, R., Chung, S., Andersson, T., Chen, I.Y., McNaught, K.S., Brownell, A.L., Jenkins, B.G., Wahlestedt, C., Kim, K.S. and Isacson, O. (2002) Proc. Natl. Acad. Sci. U.S.A. 99, 2344–2349
Signalling the Future
22 Zhang, S.C., Wernig, M., Duncan, I.D., Brustle, O. and Thomson, J.A. (2001) Nat. Biotechnol. 19, 1129–1133 23 Deacon, T., Dinsmore, J., Costantini, L.C., Ratliff, J. and Isacson, O. (1998) Exp. Neurol. 149, 28–41 24 Li, M., Pevny, L., Lovell-Badge, R. and Smith, A. (1998) Curr. Biol. 8, 971–974 25 Pevny, L.H., Sockanathan, S., Placzek, M. and Lovell-Badge, R. (1998) Development 125, 1967–1978
26 Wood, H.B. and Episkopou, V. (1999) Mech. Dev. 86, 197–201 27 Ying, Q.-L., Stavridis, M., Griffiths, D., Li, M. and Smith, A. (2003) Nat. Biotechnol., in the press 28 Aubert, J., Dunstan, H., Chambers, I. and Smith, A. (2002) Nat. Biotechnol. 20, 1240–1245 Received 6 September 2002