Merging Fields: Stem Cells in Neurogenesis, Transplantation, and Disease Modeling

Abstract

Traditionally, applied stem cell research has been segregating into strategies aiming at endogenous repair and cell transplantation. Recent advances in both fields have unraveled unexpected potential for synergy between these disparate fields. The increasing dissection of the step‐wise integration of adult‐born neurons into an established brain circuitry provides a highly informative blueprint for the functional incorporation of grafted neurons into a host brain. On the other hand, in vitro recapitulation of developmental differentiation cascades permits the de novo generation of various neural cell types from pluripotent embryonic stem (ES) cells. Advanced tools in stem cell engineering enable not only genetic selection and instruction of disease‐specific donor cells for neural replacement but also the exploitation of stem cells as transgenic cellular model systems for human diseases. In a comparative approach we here illuminate the functional integration of neurons derived from endogenous and transplanted stem cells, the evolving technologies for advanced stem cell engineering and the impact of cloned and mutated stem cells on disease modeling.

INTRODUCTION

Availability and biological physiognomy of central nervous system (CNS) stem and progenitor cells continue to arouse the neuroscience community. A classic research strategy envisions neural stem cells (NSCs, derived from all potential sources) as an ideal tool to replace lost cell populations in brain and spinal cord or to supply “wanted” therapeutic molecules for a variety of CNS disorders. While first clinical trials based on this concept are about to be initiated or already in place [eg, for treating Batten disease (http://www.stemcellsinc.com/clinicaltrials/clinicaltrials.html) or acute spinal cord injury (http://www.geron.com/showpage.asp?code=patiin)], NSC research keeps on sparking and has multiple facets. In this review, we would like to focus on an alternative aspect of current work in our and other laboratories relating to the attempt of modeling essential stages of CNS development and disease. Protocols will be introduced that permit the investigation and modification of NSCs in dispersed cell culture, in three‐dimensional culture environments, and in vivo. Particular attention will be paid to describe functional characteristics of developing neurons and their environmental interactions. Model systems, employing embryonic stem (ES) cells or adult brain NSCs, can certainly be used to generate novel populations of replacement cells, but ultimately may lead to a better understanding of embryonic development, explain limitations of CNS regeneration in the adult, or decipher brain pathologies.

ADULT NEUROGENESIS—NATURE’S BLUEPRINT FOR FUNCTIONAL NEURAL REPLACEMENT

It is well established that the adult brain harbors populations of neurogenic cells in the hippocampus and subventricular zone (SVZ) capable of self‐renewal, proliferation, and differentiation into new neurons and glia throughout life. A variety of life circumstances, such as stress (49), seizures (83), injury (55), exercise 15, 129, and a variety of molecules, such as astrocyte‐derived growth factors [eg, fibroblast growth factor 2 (FGF‐2), ciliary neurotrophic factor (CNTF), and transforming growth factor alpha (TGFα)], neurotransmitters (eg, 5‐HT), hormones (eg, thyroid hormone), and environmental interactions mediated through α6β1 ligands, modulate the process of neurogenesis (reviewed in 51). All these factors are united in the concept of a neurogenic niche where adult neurogenic cells execute highly specialized developmental tasks in defined microenvironments (2). For example, cells in the hippocampal subgranular zone progress through a defined six‐step sequence of differentiation before ultimately turning into mature, excitatory granule cells (58). Most primitive neurogenic cells in the hippocampus express glial fibrillary acidic protein (GFAP) and nestin and demonstrate prominent potassium membrane currents. Their characteristic electrophysiological profile is reminiscent of astrocytes with “passive,” non‐inactivating membrane currents and linear current–voltage relationships (38). Initial developmental progression involves loss of GFAP expression, a dramatic rise in input resistance, as well as the appearance of delayed rectifying potassium currents and voltage‐dependent sodium currents as first signs for neuronal commitment 38, 42. At this stage, developing cells already express functional GABA_A, AMPA, and NMDA neurotransmitter receptors (140), and ambient levels of GABA, potentially provided by local interneurons such as basket cells, seem to control early neurogenic events before phasic and/or synaptic activity becomes evident (44). In striking resemblance to embryonic and early postnatal CNS development (12), integration of newborn neurons into hippocampal circuitries is then accomplished by a sequence of initially exclusive GABAergic innervation followed by glutamate‐mediated synaptic inputs 44, 140. Interestingly, neurogenic events proceed at a significantly slower pace in the adult rodent hippocampus compared with the neonate (82). This is an intriguing finding in light of the fact that depolarizing GABAergic network activity is known to decline in later stages of life (12). Thus, and in perspective, the release of GABA serving as an activity‐dependent regulator of hippocampal neurogenesis 44, 82, 124 could become an exciting target for future studies aiming at control of maturation and synaptic integration of newborn neurons in the adult brain.

In vitro, neurogenic hippocampal cells can generate electrically active neurons with tetrodotoxin (TTX)‐sensitive repetitive action potentials and functional glutamatergic and GABAergic synapses within 14 days. In a pioneering study, Song et al (111) have drawn a compelling comparison between hippocampal NSC‐derived and primary neonatal hippocampal neurons, revealing quantitative differences of cultured cells’ functional properties. Efficacy of synapse formation [ie, percentage of neurons showing spontaneous synaptic currents (SSCs)] and synaptic transmission (amplitude and frequency of SSCs) were significantly reduced in stem cell‐derived newborn neurons. This observation did not seem to represent a result of a different time course for maturation. Based on experimental evidence, Song et al (111) rather speculated that additional factors (eg, brain‐derived neurotrophic factor) might be required to facilitate functional maturation of stem cell‐derived neurons. In vivo, new hippocampal neurons can clearly be distinguished from mature granule cells based on their intrinsic membrane properties. Even 4–8 weeks after birth in the adult hippocampus, newly generated granule cells show lower input resistances, smaller membrane capacitances, and less negative membrane potentials (130). Additional characteristics of these young neurons include their ability for isolated Ca²⁺ spike‐generation via T‐type Ca²⁺ channels that boost fast action potentials, and a high susceptibility for induction of associative long‐term potentiation (104). It may be these distinctive features of newborn hippocampal cells that facilitate synaptic plasticity and represent the cellular basis for the formation of new memories 78, 108.

A second, distinguished neurogenic system in the adult brain is the SVZ. Here, multipotent GFAP‐positive NSCs (type‐B cells) give rise to a transit‐amplifying cell population (type‐C cells) that forms colonies of neuroblasts (type‐A cells) after a series of rapid cell divisions (35). In the rodent, newborn neuroblasts then travel from the SVZ through the rostral migratory stream (RMS) 67, 100, and integrate as specialized interneurons into local circuitries of the olfactory bulb (69). In the past, studies of SVZ cell populations were limited to post hoc identification or dependent on thymidine analog and/or retroviral labeling of precursors during cell division. We have recently introduced a unique cell culture model that allows real‐time analysis of first steps in adult brain SVZ neurogenesis (102). Gliotypic SVZ cells could be expanded as monolayer cultures under the influence of serum and growth factors [epidermal growth factor (EGF), FGF‐2]. Combined withdrawal of serum and mitogens reliably induced the appearance of rapidly dividing cells that differentiated into SVZ‐specific type‐A cells within 4 days (Figure 1A). In this short period of time, SVZ cells accomplished a complete metamorphosis from glial phenotypes to neuroblasts. The observed morphological transitions were paralleled by striking changes in electrophysiological properties (Figure 1B). Gliotypic SVZ cells showed predominant A‐type potassium currents, and while undergoing series of rapid cell divisions, A‐type potassium currents disappeared and delayed rectifying potassium currents became apparent. Sodium currents were present in some of the recorded cells, but action potentials could not be elicited. After 4 days, electrophysiological properties of newborn cells matched those described for neuroblasts in vivo (139). Morphological, phenotypic, as well as ultrastructural analyses suggested that SVZ‐specific neurogenesis could be recapitulated in a culture dish (Figure 1C–F). Surprisingly, in vitro and in vivo (see below), neurogenic SVZ cells have the propensity to generate exclusive sets of inhibitory interneurons (Figure 1G–I). In vivo, newborn type‐A cells can easily be identified in the mouse SVZ before they migrate into the RMS, based on their small size (6–8 µm diameter) and round‐to‐oval morphology. Patch clamp recordings in situ have revealed—comparable to our in vitro findings—high input resistances (about 4 GΩ) and exclusive TEA‐sensitive delayed rectifying potassium currents (139). In an interesting analogy to the adult hippocampus 32, 44, nonsynaptic (SNARE‐independent) GABA signaling in the SVZ may represent an important regulator of neurogenic activity in the subventricular niche. Type‐A cells produce and release GABA spontaneously and upon depolarization; and two consecutive studies 17, 66 have suggested a unique feedback loop between type‐A cells and surrounding GFAP⁺ cells (putative type‐B cells, the parental population of type‐A cells). Because both cell populations express GABA_A receptors, increasing ambient levels of GABA caused by accumulation of newborn type‐A cells could limit cell cycle‐progressions of neighboring GFAP⁺ cells (66), and at the same time increase the migratory velocity of type‐A cells departing from the SVZ (17). Thus, by modifying the environmental GABA equilibrium, SVZ progenitor cells gain access to the control of their own rate of generation.

Figure 1.

Modeling adult subventricular zone (SVZ) neurogenesis in vitro. A. Typical appearance of early SVZ cell glial‐to‐neuron transformation in phase contrast and corresponding electrophysiological membrane profiles (center, voltage clamp; right, current clamp) during the first 4 days after induced neurogenesis (prolif, proliferating cells in culture; 1 day, rapidly dividing transit amplifying cells; 4 days, neuroblasts stage). B. Graphs display characteristic evolutions of passive membrane properties from multipotent glial stem cell stages to the appearance of mature, action potential‐generating interneurons at 28 days (Vm, membrane potential; Cm, membrane capacitance; Rm, input resistance). C. Phase contrast picture of a cluster of newborn type‐A cells 4 days after induction of neurogenesis. D. Similar to findings in vivo, these cells coexpress nestin (green), A2B5 (not shown), βIII tubulin (red), and PSA‐NCAM. E. Inset in (E) shows corresponding cells in phase contrast. F. High magnification of typical type‐A cells at ultrastructural evaluation (compare with ref. 35). G. One example of the mature interneuron morphologies observed after 28 days of SVZ cell differentiation in culture (green, nestin; red, βIII tubulin). H. At that stage, tetrodotoxin (TTX)‐sensitive, repetitive action potentials can be elicited (left, voltage clamp; right, current clamp). I. Surprisingly, spontaneous synaptic activity disappears entirely after application of picrotoxin (PIC), an inhibitor of GABAergic synaptic transmission (6/6 recorded cells). A, B, F (magnified), H, and I: reproduced with permission from (102), [Copyright (2005), National Academy of Sciences, USA]. C, D, E, G: Courtesy of N.M. Walton and B. Scheffler. Scale bars (in µm): C, E (inset), 60; D, 15; E, 40; F, 1; G, 25.

Belluzzi et al (11) and Carleton et al (23) have, in complementing studies, described how physiological properties change when SVZ‐generated neuroblasts establish themselves in the olfactory bulb as periglomerular and granule cells, respectively. A characteristic sequence of events was observed. Comparable to hippocampal neurogenesis, with developmental progression, membrane capacitances increased, input resistances decreased, and resting potentials hyperpolarized. At the same time, neurotransmitter receptor profiles changed. Migrating neuroblasts expressed GABA_A receptors before accumulating AMPA receptors, and NMDA receptors were added latest in this sequence (23). Exclusive delayed rectifying potassium currents were characteristic for migrating cells in the RMS; no sodium or transient potassium currents were observed. Prominent A‐type currents appeared when cells left the RMS (11). With immigration into the olfactory bulb, delayed‐rectifier‐type K⁺ currents disappeared and fast, TTX‐sensitive sodium currents were acquired. Three to four weeks after birth, newborn neurons showed electrophysiological phenotypes comparable to endogenous, neighboring olfactory bulb interneurons. Interestingly, and in contrast to normal CNS development, the ability to fire action potentials was the last property to be acquired. Synaptic connections appeared shortly after arrival at the target sites with early GABAergic input and delayed onset of glutamate‐mediated activity. Thus, newborn olfactory bulb interneurons complete their final stages of functional maturation after integrating into the local synaptic circuitry.

As it is evident from reviewing functional aspects of hippocampal and SVZ neurogenesis, NSCs and their progeny execute highly specialized tasks in their respective environments. In fact, some of the presented findings raise the question of whether there exist limitations for fate choice and functional specification among adult brain neurogenic cells. Clearly, newborn neurons become functionally mature and synaptically integrated, but their generation and purpose in the adult brain is distinct from neurons established during CNS development. Future research is thus challenged to critically evaluate, on a functional level, how endogenous brain resources can be manipulated for desired neurogenesis in CNS disease.

NEURAL REPLACEMENT BY STEM CELL TRANSPLANTATION

Beyond its importance for the maintenance of defined neuronal populations in the hippocampus and olfactory bulb, adult neurogenesis represents the most relevant proof‐of‐concept for the possibility to recruit stem cell‐derived neurons in an existing neuronal circuitry. The functional integration of endogenously generated neurons strongly suggests that the underlying mechanisms of neuronal recruitment might also be exploited for the incorporation of exogenous, transplanted precursors. During the last 30 years, a large number of studies have focused on the question of whether and to what extent neural precursor cell transplantation might be used to compensate for neuronal degeneration and thus restore lost functions in a variety of neurological disorders. One of the most critical parameters of such an approach is an appropriate donor source. Ideally, donor cells for neural transplantation should be expandable in vitro, be easily amenable to genetic modification and exhibit an immature but committed and still migratory phenotype. While precursors derived from fetal brain have traditionally served as a robust donor source, they lack stability and expandability, thus necessitating grafts composed of cells from several donors for an individual application (16). It is only recently that conditions were established which permit the derivation of stably proliferating neural precursors from primary CNS tissue (27). However, it is not yet clear whether these cells may serve as a universal donor source or whether their fate is restricted to distinct neural lineages. While other donor sources such as xenogeneic cells, immortalized precursors and presumptive transdifferentiation‐competent non‐NSCs are being explored (28), ES cells, from a biological point of view, clearly exhibit the broadest differentiation potential, unsurpassed self‐renewal capacity and the most pronounced amenability to genetic engineering. ES cells are derived from the inner cell mass of the preimplantation embryo and have the developmental potential to give rise to cells of all three germ layers as shown in vitro by embryoid body (EB) formation or in vivo by teratoma formation (4). Moreover, mouse ES (mES) cells contribute to all tissues of the adult organism including the germ line after injection into blastocysts. This implies that ES cells not only represent a virtually unlimited source of transplantable somatic precursors for cell replacement but can also serve as a model to study development in vitro.

ENGINEERING ES CELLS FOR NEURAL REPAIR

The successful therapeutic application of ES cell‐derived neural precursors critically depends on cell purity and lineage specification, because undifferentiated ES cells resulting from inchoate differentiation can result in teratomas and/or functional impairment. In general, ES cells induced to differentiate readily acquire neural fates. This has led to the notion that neural differentiation represents a “default” mechanism in ES cell differentiation 75, 110, 125. Yet, further extrinsic guidance cues are necessary to control differentiation into defined neural cell populations. In general there are three strategies of purification, that is, (i) growth factor‐controlled differentiation; (ii) lineage selection; and (iii) overexpression of instructive factors (Figure 2).

Figure 2.

Three strategies to derive differentiated neural progeny from embryonic stem (ES) cells. A. Growth factor‐controlled differentiation (19). Neural precursors can be expanded from EB‐derived cell populations by the application of FGF‐2. Upon growth factor withdrawal these precursors differentiate into neurons and glia. Alternatively, further proliferation in media containing combinations of FGF‐2 and EGF results in a predominant glial fate. EB = embryoid body; FGF‐2 = fibroblast growth factor 2; EGF = epidermal growth factor; pNPC = pan‐neural precursor cell; gNPC = gliogenic neural precursor cell. B. Lineage selection exemplified by antibiotic selection of oligodendrocyte precursors (47). This strategy involves genetic manipulation of ES cells by introducing a β‐geo gene under control of the oligodendrocyte‐specific CNP promoter. In a first step growth factors were used to generate glial progenitors from engineered ES cells. These cells were subjected to antibiotic selection yielding highly purified oligodendrocyte precursors. CNP = 2′3′‐cyclic nucleotide 3′‐phosphodiesterase gene. β‐geo = LacZ/neomycin resistance fusion gene. C. Overexpression of instructive factors is able to determine the fate of differentiating ES cells as exemplified by the derivation of midbrain dopamine neurons (3). Forced overexpression of the homeodomain transcription factor Lmx1a in growth factor‐treated nestin‐positive neural progenitors is sufficient to trigger the differentiation of midbrain dopaminergic neurons. NesE = nestin enhancer; Shh = sonic hedgehog; TH = tyrosine hydroxylase.

Guiding differentiation by extrinsic factors.  The majority of the differentiation protocols published so far are based on the stepwise application of combinations of growth factors, hormones, and other fate‐specifying factors. Retinoic acid (RA) has been most commonly used for neural differentiation of murine ES cells 7, 14, 34, 39, 41. Recently, more complex culture conditions have been employed to enrich proliferating neuroepithelial precursors from differentiating ES cell cultures without the use of RA 19, 20, 27, 33, 45, 80, 125, 151. A robust protocol has been reported that is based on the application of FGF‐2 and EGF on preselected neural precursor cells (Figure 2A). In a first step, ES cells are aggregated to EBs. Upon plating of the EBs in cell culture dishes, neural precursors are selected in ITSFn medium and subsequently expanded in the presence of FGF‐2. Following growth factor withdrawal, these precursors readily differentiate into neurons and glia (81). Alternatively, further proliferation in the growth factor combinations FGF‐2/EGF and FGF‐2/platelet‐derived growth factor (PDGF) can be used to shift the cells toward a predominant glial fate (19). One might speculate whether it is feasible to derive neural cell cultures from ES cells exhibiting stem cell properties, that is, continuous expansion by symmetrical division. In fact there is evidence suggesting that the combination of FGF‐2 and EGF is sufficient for the derivation of a pure population of self‐renewing neural stem (NS) cells (27). After prolonged expansion, NS cells remain able to differentiate into neurons and astrocytes in vitro and in vivo. Clonally expanded NS cells, again, generate neurons and glia, indicating preservation of a self‐renewing multipotent phenotype (27). Recent investigations are employing series of defined media and growth factors to control regionalization and specification of defined cell types such as midbrain dopaminergic neurons 3, 9, 57, 62, spinal motor neurons (148), neural crest‐derived peripheral neurons (73) and telencephalic precursors (142) as discussed comprehensively elsewhere in this issue of the journal (153).

Most of these studies aiming for controlled differentiation of ES cells into neural precursors have been carried out with mouse cells. However, an increasing number of studies indicate that the protocols developed in mES cells can be adapted to human ES (hES) cells. Reubinoff et al reported early stages of neuroectodermal differentiation in hES cells cultured at high density for 3 weeks. Colonies displaying neuroectodermal morphology could be isolated and replated in serum‐free medium, where they formed neurosphere‐like structures. When plated on appropriate substrates these spheres attach and spontaneously differentiate into a heterogeneous population including neural cells (91). Zhang et al reported that neural precursor cells, differentiated via the EB state, formed neural tube‐like structures in the presence of FGF‐2. After enrichment and subsequent growth factor withdrawal hES cell‐derived neural precursors readily differentiated into neurons and glia in vitro and in vivo (154). In fact, thus far most protocols for neural differentiation of hES cells rely on EB formation 24, 90, 154. Some protocols involve mechanical or enzymatic isolation of neural cell clusters from plated EBs or differentiated hES cells 27, 90, 154. Other protocols include the formation of neural cells in suspension cultures of hES cells (106). Moreover, coculture with stromal cell lines has been used to induce neural differentiation both in murine and hES cells without an intermediate EB step 9, 22, 57, 84, 85, 152. In analogy to mES cells, these complex differentiation techniques have been used to generate a variety of human neural subtypes, including mid‐ and hindbrain neurons 22, 62, 84, 85, 152, motoneurons (65), oligodendrocytes (76), neural crest derivatives (88) and others (153).

Lineage selection and transcription factor‐based instruction.  The success of factor‐induced differentiation protocols critically depends on a detailed knowledge of fate‐specifying factors and their corresponding window of activity during neural development in vitro. It is the large number of candidate factors and a lack of understanding of their complex interactions that currently complicates this strategy. A more rational approach for purifying somatic cells from differentiating ES cells in vitro involves the exploitation of cell type‐specific markers for the enrichment of the target populations. Immunoselection techniques such as immunopanning, and fluorescence‐ or magnetic‐activated cell sorting (FACS, MACS) have been used to generate highly enriched populations of ES‐derived neural precursors 24, 56, 74, 94, 105, 127, 146. However, appropriate cell surface markers represent prerequisites for this strategy and are not always available. A few studies have attempted to purify neural precursor cells by genetic selection. This strategy involves genetic manipulation of ES cells or ES cell‐derived precursors by introducing a selection cassette enabling the expression of marker genes under the control of a cell type‐specific promoter that ideally defines the desired cell population (Figure 2B). The Sox2 gene has been used to attempt lineage selection of neural precursors from differentiated ES cells. The Sox2 transcription factor represents an early neuroepithelial marker expressed in the neural plate and proliferating neural progenitors throughout ontogeny (for review see 86). Li et al employed a Sox2‐ßgeo knock‐in line to derive multipotent neural precursors from ES cells. After RA‐induced differentiation, Sox‐2‐expressing neural precursor cells could be efficiently enriched by G418 selection (64). The generation of a highly pure population of ES cell‐derived neuronal cells was accomplished by the combination of both targeted differentiation and genetic lineage selection. Genetically engineered ES cells carrying an eGFP gene targeted into the first exon of the tau gene were subjected to factor‐induced differentiation into pan‐neural precursors. Subsequent cell sorting of neuron‐specific eGFP fluorescence allowed enrichment of postmitotic neurons to purities of more than 90%. Moreover, transplantation of tau‐eGFP ES cell‐derived neural precursors yielded terminally differentiated eGFP‐positive neurons in the host brain (146). A recently reported study demonstrated successful purification of neurons from differentiated wild‐type ES cells (56). Neurons were purified from dissociated embryoid bodies using antibodies directed against the neuronal cell adhesion molecule L1 (89). This procedure yielded a pure population of differentiated neurons, which are excitable and capable to form excitatory, glutamergic, and inhibitory GABAergic synapses (56).

The biomedical application of ES cell‐derived neurons (ESNs) may not only depend on their purity but also on their state of maturation. This is particularly true for cell replacement strategies, which require migratory young neurons to integrate into targeted areas of the host nervous system. Lineage selection paradigms involving promoters of “early” neuronal genes such as T‐apha‐1 tubulin have been applied to purify neurons shortly after initiation of neuronal differentiation (103). Similarly, genetically refined selection strategies are available for the purification of glial cells, for example, ES cell‐derived oligodendrocyte progenitors (47). For this, ES cells were transfected with a βgeo gene under the control of the oligodendrocyte‐specific 2′3′‐cyclic nucleotide 3′‐phosphodiesterase (CNP) promoter. CNP is a myelin‐associated protein, which is already expressed in developing oligodendrocytes 92, 135. CNP‐β‐geo transgenic ES cells were subjected to glial differentiation and G418 selection at the stage of CNP‐positive bipotent glial precursors. Subsequent growth factor withdrawal yielded more than 95% oligodendrocytes whereas only <40% can be found in populations generated without additional genetic selection (47).

Recent progress has been made with the activation of intrinsic key determinants to substantially enrich for neural subtypes (Figure 2C). Overexpression of transcription factors such as Nurr1 26, 60 or Lmx1a (3) in ES cells were reported to result in enhanced differentiation into midbrain dopaminergic neurons and oligodendrocytes, respectively. Oligodendrocyte specification achieved by transient overexpression of Olig1/2 in fetal NSCs 8, 29 might also be applicable to hES cells. A detailed review of these exciting findings is presented by Zhang et al (153, this issue).

Advanced tools for genetic engineering.  Although providing substantial promise, genetic modification of ES cells harbors the risk of unwanted permanent genomic alterations that might hamper any therapeutic use of ES cell‐derived somatic precursors. Chromosomal random integration of foreign DNA is able to disrupt the expression of tumor suppressor genes or activate oncogenes (48) resulting in the malignant transformation of cells as has been shown recently in a clinical gene therapy trial (50). Removal of the selection cassettes by site‐specific recombination may help to circumvent this drawback. The site‐specific recombinase Cre recognizes and recombines specific DNA sequences designated as loxP sites. Cre‐mediated recombination results in deletion, inversion or integration of loxP‐modified DNA sequences, depending on the relative orientation of the loxP sites 18, 43, 63. The Cre/loxP recombination system provides a versatile means to study gene functions during neural specification and development of disease. For that, so‐called conditional alleles, that is, loxP‐modified versions of the gene‐of‐interest have to be constructed and integrated either by random integration or into a specific locus by homologous recombination. Cre‐inducible gain‐of‐function can be achieved by placing a loxP‐flanked transcriptional stop sequence between the promoter and the open reading frame (Figure 3A). Cre‐mediated recombination results in the deletion of the stop cassette, thereby activating transcription of the gene. Vice versa, Cre recombination allows one to conditionally inactivate gene functions by deleting the transgene or removing regions that are essential for gene activity in a knock‐out scenario (Figure 3B). Moreover, the Cre/loxP system can be used to knock down gene functions by Cre‐mediated induction of RNA interference 30, 70, 131. Cre might also be useful to catalyze the integration of transgene cassettes into safe genome locations by site‐specific integration, thereby avoiding insertional mutagenesis 37, 149, 150. Finally, Cre‐mediated deletion could also provide a means to remove engineered selection cassettes from ES cell‐derived somatic precursors after successful purification, thus leaving the isolated cells with a largely unaltered genome (Figure 3C). Such “cleaned” precursor cells should be particularly attractive with respect to therapeutic applications. In a similar scenario, Cre‐mediated recombination has been used to reversibly immortalize mammalian cells by transient overexpression of oncogenes (147).

Figure 3.

Genetic engineering of embryonic stem cells by the site‐specific recombinase Cre. The Cre/loxP recombination system provides a versatile means to genetically manipulate embryonic stem (ES) cells. Cre recognizes and recombines specific DNA sequences referred to as loxP sites (depicted as triangles). Cre‐mediated recombination results in the deletion of the loxP‐flanked sequence. Depending on the experimental set up this reaction can be used to induce A. gain‐of‐function caused by the deletion of a loxP‐flanked transcriptional stop sequence or B. loss‐of‐function upon Cre‐mediated removal of the gene‐of‐interest or essential parts thereof. C. Cre might be instrumental in removing expression cassettes for genetic selection or instruction after successful purification. Because of their largely unaltered genome such “cleaned cells” would be highly attractive for therapeutic applications. pA = poly‐adenylation signal.

Induction of Cre recombinase activity in mammalian cells can be achieved by either transfection or virus‐mediated gene transfer 5, 96, 107. These techniques, however, are often limited by low efficiency and poor cell viability after gene delivery. Recently, our group demonstrated that direct delivery of Cre recombinase by protein transduction allows highly efficient site‐specific recombination in hES cells and hES cell‐derived neural precursor cells without compromising the developmental potential (77). The direct delivery of biologically active Cre recombinase circumvents the need for an additional genetic manipulation. Thus, conditional mutagenesis by Cre protein transduction provides a rapid and efficient means to gain control over genes involved in neural differentiation and molecular pathogenesis of disease.

Functional integration of ESNs.  Because of their high proliferative capacity and pluripotentiality, ES cells are a reliable source for generating a broad spectrum of potential donor cells. Neurons can be derived in high purity and large numbers, yet in contrast to all somatic cell types in the nervous system, ES cell development takes place in a culture dish, remote from the natural environment and stimulations present during embryonic CNS development. The question arises as to how “functional” these cells are and whether they can be synaptically integrated into established or damaged neural circuitries. Analysis of function is generally accepted as a decisive measure for “successful” generation of neurons from potent stem and progenitor cells (59). At a single cell level, neuronal function (ie, excitation and electrical signaling) is essentially based on movements of Na⁺, K⁺, Ca²⁺, and Cl⁻ through membrane‐bound ion channels and the activity of neurotransmitter receptors. Bain et al (6) combined RA as a differentiation stimulus and whole cell patch clamp recordings in dispersed culture to show how cells with neuronal morphologies soon acquire characteristic TTX‐sensitive Na⁺ channels and voltage‐gated K⁺ and Ca²⁺ channels, that is, prerequisites for the generation of action potentials. Sensitivity to applications of excitatory (kainate, NMDA) and inhibitory compounds (GABA, glycine) during the recording procedures furthermore suggested the expression of functional neurotransmitter receptors. In addition to these basic, yet fundamental findings, Strubing et al (118) noted that the densities of voltage‐dependent Na⁺, K⁺, and Ca²⁺ currents (also see 119) increased with time after RA‐induced differentiation. This observation is crucial, because it implies that ES cell‐derived progeny might recapitulate a temporal sequence of voltage‐gated channel expression, which is characteristic and essential for neuronal development and maturation (112). In the same study, Strubing et al (118) provided evidence for synaptic communication between newborn ESNs, another important feature of neuronal function. Interestingly, recordings of evoked and spontaneous postsynaptic currents (PSCs) in these cultures revealed an almost exclusive sensitivity to GABA_A‐receptor antagonists suggesting inhibitory GABA as the main neurotransmitter in this system. In contrast to these findings, Finley et al (40) found a majority of synaptic connections (≈80%) formed between RA‐induced newborn ESNs (in a collagen‐microisland culture system) as excitatory with an abundance of non‐NMDA‐receptor‐mediated excitatory postsynaptic currents (EPSCs). Bibel et al (14) recently made similar observations, describing the vast majority of neurons generated according to their protocol as glutamatergic (for a detailed review see 46). Variations in experimental procedures and ESN‐derivation protocols could explain some of the observed differences in function and fate choice of neuronal progeny in vitro, and should certainly become a focus of future investigations. However, it is apparent from all these studies that hallmarks of intrinsic neuronal function can be identified in dissociated cell culture: ESNs mature, are excitable, and can communicate.

Dissociated cell cultures offer the advantage to monitor developmental progression of differentiating stem cell‐derived progeny in highly defined conditions and in real time, but lack the organization and environmental stimuli of CNS tissue that may be required for immature neurons to ultimately establish themselves. To bridge this gap, we have resorted to long‐term hippocampal slice cultures as host for functional analysis of engrafted donor cell populations (Figure 4A; 101). Our “in vitro‐transplantation” paradigm applies the benefits of classic (monolayer) cell culture systems to an established three‐dimensional brain structure, and it can be used to observe and study functional characteristics of incorporating ESNs for up to 21 days after engraftment (Figure 4B; 13). For analysis of integrating ESNs in hippocampal slice cultures, suitable donor cell populations were derived from an ES cell line that expresses enhanced green fluorescent protein (EGFP) exclusively in neuronal progeny 126, 146. Patch clamp recordings of readily identifiable donor cells (Figure 4B–D) were performed at distinct stages after engraftment and revealed a detailed timeline of developing intrinsic membrane profiles. Similar to data obtained from the developing CNS 87, 141, differentiating ESNs display increasing membrane capacitances, decreasing input resistances, and more negative resting membrane potentials with time of differentiation. Maturation of the ESNs’ intrinsic discharge behavior (ie, increased frequency and amplitude of action potentials) was observed to occur in parallel with increased densities of voltage‐gated Na⁺ and K⁺ channels (Figure 4E,F). Also, AMPA‐ and GABA_A‐mediated spontaneous PSCs increase dramatically with time after engraftment, suggesting a progressive intensification of glutamatergic and GABAergic synaptic input onto ESNs (Figure 4G). Taking advantage of the conserved lamellar organization in the recipient hippocampal tissue, vital entorhinal cortex axons were visualized projecting onto incorporated ESNs within the dentate gyrus, and EPSCs elicited by stimulation of the labeled perforant path revealed marked paired‐pulse facilitation in ESNs. This finding represents compelling evidence for direct synaptic communication between host axons and donor cells, and a form of presynaptic short‐term plasticity characteristic for immature perforant path synapses on newly formed granule cells (141).

Figure 4.

An “in vitro transplantation” assay. A. The 400 µm horizontal sections of postnatal day 9 rat brains containing the hippocampal formation and associated entorhinal and temporal cortex serve as recipient for genetically labeled [eg, enhanced green fluorescent protein (EGFP⁺)] donor cells. These slices can be cultured on porous membranes in interphase conditions (115), permitting donor cell analysis for several weeks in real time. B. Engrafted cells [here ES cell‐derived neurons (ESNs) derived from a tau EGFP knock‐in embryonic stem (ES) cell line 13, 126], migrate into the recipient tissue, incorporate, and acquire mature neuronal morphologies. C. Differential Interference Contrast (DIC) appearance of a slice culture, 21 days after transplantation at the time of whole cell patch clamp analysis (note the tip of the recording electrode located in the hippocampal hilar region). D. Expression of EGFP and the preserved organotypic organization of the hippocampal slice permit reliable identification of donor cells within the recipient tissue (note diffusion of EGFP into the patch electrode). Photograph in (D) represents a magnification of (C), visualizing the recorded donor cell using appropriate GFP filter sets. E. Characteristic series of action potentials (current clamp) and F. membrane currents (voltage clamp) demonstrate functional maturation of ESNs at 21 days after in vitro transplantation. G. Successful synaptic integration is revealed by spontaneous synaptic activity recorded from a 21‐day engrafted donor cell (−80 mV; voltage clamp). B–G: Courtesy of F. Benninger and H. Beck. Scale bars (in µm): B, 60; C, 500; D, 30.

To our surprise, but similar to earlier reports in dispersed cell culture (40), there was a notable paucity of NMDA‐mediated neurotransmission, although immunocytochemistry revealed expression of NR1 NMDA receptor subunits on the cell surface of incorporated ESNs. There are a variety of potential explanations for this phenomenon (extensively discussed in 13); however, the scarcity of NMDA receptor‐mediated synaptic input warrants future investigation as it may affect maturation of engrafted ESNs and/or potential interactions with host cells.

Fundamental characteristics and developmental progression of functional maturation and integration—as observed in culture models—seem universally applicable to transplanted ESNs integrating into the developing, adult, and chronically epileptic CNS 98, 145. Remarkably, functional maturation and synaptic integration of ESN donor cells into the embryonic brain does not seem to require acquisition of a region‐specific transcription factor code (145). In fact, in this study, most incorporated ESNs did not exhibit a regionally appropriate expression of Dlx, Pax6, En1, and Bf1; and overall neurotransmitter expression of engrafted neurons did not reveal regional preferences either. While these findings indicate that adoption of a region‐specific neuronal phenotype is not required for functional integration, they also serve as a note of caution, stressing the need for further studies on the importance of regional “pre‐specification” of donor cells before engraftment.

Future studies on the functional analysis of ES cell‐derived progeny need to reach beyond proof‐of‐principle experimentation. One key challenge is to further determine whether ESNs can become functionally specified (such as seen for ES cell‐derived motoneurons; 72) and integrate with appropriate function into a diversity of region‐specific local circuitries, as well as to then define the impact of newly integrated cells on local network activity and on behavioral levels. A second issue is to translate these findings to hES cells, where initial investigations point to attractive potential applications in drug and toxicology screens, developmental assays, and regenerative medicine 24, 52.

MODELING CNS DISEASES IN ES CELLS

Over the past two decades, the application of mES cells has revolutionized mammalian genetics by providing an invaluable tool for studying developmental processes and modeling human diseases. Due to their superb amenability to genetic manipulation, gain‐ and loss‐of‐function studies, mES cells revealed comprehensive insights into genetic networks underlying physiological functions as well as the analysis of dysregulated function leading to disease. Moreover, cellular models provide a means to screen for novel therapeutic compounds. However, the use of mouse models for human disease is limited because of the genetic, developmental, biochemical, and morphological differences between humans and mice. Thus, animal models do not always fully recapitulate the phenotype of the human disorder. The generation of primary cultures from individual patients represents an alternative to the use of in vivo mouse models; however, only few tissues are accessible and the life span of primary cells is limited. Nowadays, the derivation and culture of pluripotent human cells provides new perspectives for the modeling of human disorders. So far only a few studies have attempted to generate cellular models of CNS diseases in pluripotent human cells 128, 133. These studies involve the generation of disease‐specific pluripotent human cells harboring potentially pathogenic mutations. In principle, there are three ways to derive mutant, pluripotent human cell lines, that is, (i) mutagenesis of hES cells; (ii) isolation of hES cell lines from affected preimplantation embryos; or (iii) cloning from adult patients by nuclear transfer (nt).

Disease modeling in mES cells and ES cell‐derived neural progeny.  Targeted mutagenesis and neural differentiation in vitro represent a promising combination for the pathomolecular analysis of neurological diseases. In this respect, triplet‐repeat disorders have evolved as interesting candidates. Huntington’s disease, spinal and bulbar muscular atrophy or Kennedy disease, dentatorubralpallidoluysian atrophy, and spinocerebellar ataxias are caused by expanded CAG repeats translated into expanded polyglutamine sequences and result in disease‐specific neurodegenerative phenotypes (for review see 97, 143). The underlying molecular mechanisms responsible for disease‐specific neurodegeneration remain elusive. A large number of CAG repeat rodent models have been created; however, they do not fully reproduce the human disease—a fact, which might be caused by inherent species differences. Moreover, there are practical limitations because establishing animal models is very time‐consuming and laborious and the time required to develop neurodegenerative symptoms in mice is long. Thus, creating models of CAG repeat diseases in ES cells and subsequent neural differentiation in vitro represents a promising approach. Lorincz et al constructed a knock‐in model of Huntington’s disease in mES cells by replacing the wild‐type Hdh gene with a gene containing 150 CAG repeats (68). Undifferentiated ES cells appeared normal whereas striking differences between mutant and wild‐type cells became apparent upon neuronal differentiation in vitro. Fewer mutant cells were found to express the neuronal marker TUJ1 and many of these cells displayed dystrophic neurites. Moreover, RE1 silencing factor/neuron restrictive silencer factor (REST/NRSF) expression was elevated in CAG expansion expressing cells (68). Dystrophic neurites and axonal pathology have been widely described in Huntington’s disease and other CAG repeat diseases (95). Although the molecular mechanism of neuritic pathology is unclear so far, transcriptional dysregulation of REST/NRSF might play an important role. REST/NRSF is thought to be involved in neural differentiation and the maintenance of neuronal activity by silencing neuronal genes in nonneuronal tissues. Thus, downregulation of REST/NRSF seems to be critical for neural differentiation. If studies along these lines were to be translated to hES cells, forced expression of CAG repeats may develop into a potent reductionist tool to study the mechanisms leading to human neuronal degeneration.

Another disorder recently targeted by this approach is Alzheimer’s disease (AD) (1). One key limitation of AD research is the inability to comprehensively study postmitotic neurons from patients. Abe et al tried to overcome this limitation by combining targeted mutagenesis in mES cells with factor‐directed neuronal differentiation in vitro. A mutated ES line was generated by knock‐in of the point mutation V642I into the endogenous amyloid precursor protein (APP) locus, a mutation that is linked to familial forms of AD. V642I‐APP cells were differentiated into CNS‐type neurons by a combination of previously published protocols 6, 81, yielding >90% postmitotic neurons. The culture media of these neurons were shown to contain a significantly increased ratio of Aß42 vs. total Aß secretion, indicating that the V642I mutation is a direct cause of elevated Aß42 secretion. However, the V642I mutation neither affected the phosphorylation state of tau nor induced neurofibrillary tangles or increased cell death during the 3‐week time window studied (1). While V642I‐APP mES cell‐derived neurons exhibit only one of several pathognomonic cellular alterations observed in AD, this approach may unravel further phenotypic changes upon translation into a human setting. Further studies on other genes involved in hereditary forms of AD, including presenilin 1 and 2, will enable researchers to gain further insights into molecular pathways underlying AD pathology.

Modeling disease in hES cells by genetic manipulation.  Gene targeting by homologous recombination enables the introduction of well‐defined mutations into specific loci, thereby providing an indispensable tool for functional genomics in mES cells 36, 123. In contrast, random insertions of transgene cassettes are prone to unwanted insertional mutagenesis and positional variegation. Homologous recombination in ES cells involves transfection of gene targeting vectors into ES cells and selection of homologous recombinant clones. Unfortunately, mES and hES cells display marked differences in both transfection and cloning efficiencies. These differences limit the applicability of gene targeting in hES cells (for review see 71). So far, there are only few examples of successfully targeted genes in hES cells. Zwaka et al reported targeting of the Oct‐4 gene, a master regulator of pluripotency in ES cells, and the hypoxanthine phosphoribosyltransferase‐1 (HPRT‐1) gene (155). In this study, hES cells were electroporated in clumps in order to overcome the poor cloning efficiency of single hES cells. Notably, targeting efficiencies for both Oct‐4 and HPRT‐1 appeared comparable to those observed in mES cells indicating that the frequency of homologous recombination as such may be similar. However, inefficient transfection and/or poor cloning efficiency resulted in an overall limited efficiency of homologous recombination in hES cells.

Targeting of the HPRT‐1 locus may serve as a model of Lesch‐Nyhan syndrome as has been recently reported by Urbach et al (128). Lesch‐Nyhan syndrome is a severe genetic disease caused by the malfunction of the HPRT‐1 gene. Lesch‐Nyhan patients suffer from renal failure and a gout‐like syndrome and urinary stones. In addition, the syndrome includes various neurological symptoms such as mental retardation, spastic cerebral palsy, choreoathetosis, and self‐destructive biting of fingers and lips (116). Mutations in the HPRT1 enzyme can cause renal failure because of the accumulation of uric acid; however, it is unclear how HPRT1 deficiency leads to the neurological symptoms. According to recent studies dopaminergic neurons appear to be affected in Lesch‐Nyhan patients 99, 134. Mouse models of the Lesch‐Nyhan syndrome do not recapitulate the human phenotype because uric acid fails to accumulate in mice lacking HPRT activity due to a subtle biochemical difference between rodents and humans (10). In contrast, human HPRT‐deficient ES cells turned out to accumulate uric acid at a higher rate when compared with wild‐type cells. Moreover, Urbach et al report that the drug allopurinol is able to reduce the level of uric acid in mutated cells, indicating that HPRT‐deficient cells may be helpful to search for new medications to treat Lesch‐Nyhan patients (128). It remains to be investigated whether the mutated cells can also help to analyze molecular and cellular pathways underlying the neurological symptoms of affected patients. In vitro differentiation of HPRT‐deficient hES cells into dopaminergic neurons could represent one step into this direction.

Alternative routes to derive disease‐specific pluripotent cells.  Considering the still limited ability for gene targeting in hES cells, the alternative option of deriving hES cells directly from patients has attracted considerable interest. Such patient‐specific stem cell lines would be particularly interesting with respect to their ability to also model complex polygenic traits, whereas genetically engineered lines mainly serve as models of monogenic disorders. In principle, there are two ways to generate disease‐specific hES cell lines, that is, (i) the derivation of ES cell lines from embryos which have been discarded in the context of preimplantation genetic diagnosis (PGD); and (ii) cloning by nt. PGD is a diagnostic procedure employed in conjunction with in vitro fertilization in many countries (for review see 79). PGD is performed to identify genetic defects in embryos before transferring them into the uterus. Because only unaffected embryos are transferred, PGD provides a source for the derivation of hES cell lines with genetic disorders. Recently, Verlinsky and coworkers provided the proof‐of‐concept by generating 18 PGD‐derived hES cell lines exhibiting different genetic abnormalities (133). These mutant hES cell lines represent various human diseases including adrenoleukodystrophy, Duchenne’s and Becker’s muscular dystrophy, Fanconi’s anemia, Huntington’s disease, Marfan syndrome, myotonic dystrophy, neurofibromatosis type I and thalassemia. The preimplantation embryos were obtained from PGD cycles, which were performed either by the first and second polar body removal or embryo biopsy. Unaffected embryos were transferred back to patients, while the mutant ones were used for confirmation of the genotype and derivation of hES cell lines. The established hES lines were maintained in vitro from 10 to 15 passages before freezing (133). For the diseases represented so far, the genetic defects did not affect the efficiency of hES cell derivation. In vitro differentiation of such disease‐specific hES cell lines could provide a valuable and limitless source of human neurons, glia and other somatic cell types for studying the pathogenesis of the respective disorders on a cellular level. However, application of PGD raises grave ethical concerns. Besides being used to predict and prevent defined serious and life‐threatening diseases this technique may also provide a means to select against minor genetic abnormalities and not disease‐related traits such as obesity and the selection of sex (61). These concerns have led to a ban of PGD in various countries including Germany, Ireland, Switzerland, Western Australia, and Austria. France, the Netherlands, Belgium, Italy, Greece, and the United Kingdom have limited the use of PGD to the prevention of severe disorders (109).

Cloning as an avenue toward disease modeling.  Nuclear reprogramming of adult somatic cells represents an alternative approach for the derivation of patient‐specific pluripotent stem cell lines. Somatic cell nuclear transfer (SCNT) experiments in various mammals including sheep, mice, and dogs have shown that a somatic cell nucleus can initiate embryonic development after being transplanted into an enucleated oocyte (for review see 54, 144). Blastocysts derived from these cells can be used to generate ES cells, whose nuclear genome is genetically identical to that of the donor, thus carrying an indistinguishable copy of the genetic repertoire of the patient. A key advantage of such an approach would be the fact that the pathologic changes observed in the somatic progeny of the cloned ES cells could be correlated to the complete disease history of the individual patient—a fact particularly relevant to polygenic diseases exhibiting a large number of different variants. Studies in mice have shown that such nt‐ES cell lines seem to maintain their pluripotency 93, 136, 137. Despite the recently uncovered frauds published by the group of Woo Suk Hwang, it seems likely that the derivation of nt‐hES cells is feasible. Stoijkovic et al reported the successful derivation of a human blastocyst after SCNT into human oocytes (114), albeit with low efficiency. While it remains to be investigated whether the efficiency of SCNT into human oocytes can be significantly improved and whether nt‐hES cells have the full developmental potential to model human disease, the development of alternative reprogramming procedures has become an important focus in this field. SCNT requires oocyte donation—a procedure that raises not only ethical concerns but carries several risks such as complications because of general anesthesia, or the development of an ovarian hyperstimulation syndrome as well as ovarian cancer later in life because of hormonal stimulation (113). The recently reported successful differentiation of mES cells into oocytes (53) may provide a promising alternative to circumvent oocyte donation for nuclear reprogramming. However, it remains to be shown that oocyte differentiation protocols can be adapted to hES cells. Current efforts are aimed at complete circumvention for the need of oocytes in nuclear reprogramming. In principle, dedifferentiation of somatic cells could represent an alternative approach to generate pluripotent cells from differentiated somatic tissues. This strategy dates back to the early 1970s, when Veomett et al fused enucleated oocytes (cytoplasts) and karyoplasts isolated from parental cell lines prior to fusion (132). The enucleation procedure was then omitted, so that the entire parental cell was fused with the cytoplast (138) resulting in a hybrid cell type designated as “cybrid” (cytoplasm hybrid; 21). Recent reports indicate that both mES 121, 122 and hES cells (31) may have the potency to reprogram adult somatic cells similar to that shown for oocytes, thus obviating the need for egg donation. Such hybrid (“stembrid;”117) cells were reported to have the genotype of the donor somatic cells and the pluripotency properties of the recipient ES cells. However, as the resulting hybrid cells contain nuclei of both somatic and ES cells, it remains unclear whether they can fully recapitulate cellular functionality in vitro (120). This limitation may be overcome by enucleation of hES cells prior to cell fusion, thereby allowing a replacement of the ES cell nuclei by nuclei of somatic cells (117). Pluripotent cell lines were also obtained by transfer of human fibroblast nuclei into enucleated rabbit oocytes (25). Although such chimeric embryos failed to develop beyond the eight‐cell stage, they may serve for modeling human disease, because they differentiate into several cell types in vitro. In conclusion, there are several promising routes for the derivation of patient‐specific pluripotent cells. Further progress in the field of human somatic cell reprogramming might facilitate the generation of disease‐specific pluripotent cell lines for a large variety of biomedical applications.

CONCLUSION

A number of recent studies have taught us to distinguish intrinsic biophysical profiles of neurogenic cells in the hippocampus 38, 42, 104 and the SVZ 11, 23, 102, 139 across all developmental stages of maturation in vitro and in vivo. While the field of adult neurogenesis will continue to explore how these data can be used and translated for therapeutic recruitment of endogenous stem cells, they may also serve as a blueprint for the functional integration of transplanted stem cells. However, in appreciating the complexity of the challenges associated with the controlled differentiation and integration of exogenous stem cells into brain circuitry, there is an increasing awareness of other, more tangible stem cell applications. Before their broad use in neural regeneration, the sophisticated tools now being developed for human stem cell engineering may—by analogy to the history of mouse genetics—have their first major biomedical impact in the evolving field of transgenic cellular disease modeling.