Cell therapy is a promising approach to treat neurodegenerative diseases [1–3]. Whereas primary tissue transplants are limited in number and thus not applicable to the numerous individuals suffering from neural disorders, in vitro expandable populations of neural precursors represent an attractive alternative source for neural transplantation . Embryonic stem (ES) cells can proliferate extensively in an undifferentiated state, thus providing an almost unlimited source of neural precursors after predifferentiation [5–7]. However, ES cell-derived neural precursor cell populations have been observed to be tumorigenic, particularly after allogeneic or syngeneic transplantations . Thus, there is a need to predifferentiate ES cells such that their tumorigenic potential is minimized.
We have developed an optimized protocol for predifferentiation of ES cells and isolation of ES cell-derived neural aggregates, resulting in a high yield of neuronally committed cells in vitro and after transplantation in vivo without tumor formation in a syngeneic transplantation paradigm in which predifferentiated ES cell-derived neural aggregates from C57BL/6J mice were injected into the quinolinic acid-lesioned striatum of adult C57BL/6J recipient mice.
Materials and Methods
Generation of an ES Cell Line Expressing Enhanced Green Fluorescent Protein
An ES cell line constitutively expressing enhanced green fluorescent protein (EGFP) was derived from transgenic C57BL/6J mice ubiquitously expressing EGFP under the influence of the chicken β-actin promoter  as described . Pluripotentiality was confirmed by injection of undifferentiated EGFP+ ES cells into the blastocyst at embryonic day 3.5 and determination of EGFP expression in chimeric mice at embryonic day 14, when expression in all three germ layers was found.
Generation of Neurosphere Cultures from Embryonic Day 14
Lateral and medial ganglionic eminences of C57BL/6J mice were removed from 14-day-old embryos and mechanically dissociated with a fire-polished Pasteur pipette in Dulbecco's modified Eagle's medium (DMEM)/F-12 (1:1; Invitrogen, Carlsbad, CA, http://www.invitrogen.com) containing glucose (0.6%; Merck, & Co., Darmstadt, Germany, http://www.merck.com), glutamine (2 mM; Invitrogen), sodium bicarbonate (3 mM, Invitrogen), HEPES buffer (5 mM; Merck), and 20 μl/ml B27 (Invitrogen). For generation and expansion of neurospheres, epidermal growth factor (EGF; PeproTech Inc., Rocky Hill, NJ, http://www.peprotech.com) and fibroblast growth factor-2 (FGF-2; PeproTech) were added to a final concentration of 10 ng/ml each. The initial seeding density was 200,000 cells/ml. After 6 days in vitro cells were passaged for the first time with a seeding density of 50,000 cells/ml. From the first passage onward, neurospheres were passaged every 5th day. Vital cells were determined by 0.5% Trypan blue dye (Invitrogen) exclusion.
Differentiation of cells from EGF/FGF generated neurospheres was allowed to take place between passages 3 and 6. Neurospheres were mechanically dissociated and plated at a density of 50,000 cells/ml onto 15-mm glass coverslips coated with poly-l-lysine. Precursor cells were first maintained in an undifferentiated state for 5 days with EGF/FGF containing serum-free culture medium. After removal of growth factors, precursor cells were allowed to differentiate for another 7 days before they were processed for immunocytochemistry.
Predifferentiation of Neural Precursors from ES Cells
ES cell lines used in this study were R1  and the EGFP+ ES cell line generated in our laboratory. Maintenance of undifferentiated ES cells, embryoid body formation, and selection of neural precursor cells was carried out as described [12, 13] with minor modifications. Briefly, undifferentiated (stage 1) ES cells were expanded in the presence of 1,000 U/ml leukemia inhibitory factor (LIF; Chemicon, Temecula, CA, http://www.chemicon.com). After formation of embryoid bodies (stage 2) in the absence of LIF, selection of neural precursors was initiated with DMEM /F12 supplemented with 5 μg/ml insulin, 50 μg/ml transferrin, 30 nM selenium chloride, and 5 μg/ml fibronectin for 8 days (stage 3). Selected neural precursors were expanded (stage 4) in DMEM/F12 medium supplemented with 2% B-27, 2 mM l-glutamine, 5,000 U/ml penicillin, 5,000 μg/ml streptomycin, and 20 ng/ml FGF-2.
Generation and Isolation of Substrate-Adherent ES Cell-Derived Neural Aggregates
To generate substrate-adherent ES cell-derived neural aggregates (SENAs), predifferentiated ES cell-derived neural precursors, seeded at 100,000 cells/ml medium onto poly-L-ornithine-coated cell culture dishes or coverslips, were maintained for 18 days under the influence of FGF-2. We termed this period “prolonged stage 4” (ps4). Thereafter, FGF-2 was withdrawn for 6 days to induce terminal differentiation of SENA-derived young neurons (stage 5). During ps4 (from 1+ to 18+; the number of days after plating of neural precursors in the presence of FGF-2 will be referred to as n+) and during the subsequent differentiation period in stage 5 without FGF-2 (18+/6−; number of days after stage 4 or ps4 without FGF-2 will be referred to as n−), SENAs developed intermixed with neural precursors in monolayer. These monolayer cells shared features similar to those of “conventional” stage 4 (5+) or stage 5 (5+/5−) cultures.
To obtain highly enriched SENA preparations, cultures were treated with 0.3 μg/ml collagenase XI (Sigma-Aldrich, St. Louis, http://www.sigmaaldrich.com) at 37°C for 10 minutes, thus gently detaching entire SENAs and monolayer precursors from the substrate. Collagenase was preferred to trypsin or accutase because it did not result in dissociation of SENAs. Detached cells were carefully pipetted up and down 10 times with a fire-polished Pasteur pipette to separate monolayer precursors from SENAs, which remained intact during this procedure. SENAs were harvested by sedimentation at 1g for 2 minutes and resuspended in HBSS− to pick SENAs individually using a 10-μl pipette tip.
Three days prior to transplantation, the right striatum of 3-month-old C57BL/6J mice was lesioned by stereotaxic injection of 1 μl of 60 nmol of quinolinic acid (Sigma-Aldrich) at the following coordinates in relation to bregma : 0.4 mm anteroposterior, 1.8 mm mediolateral, and 3.0 mm dorsal. On the day of transplantation, stage 4 precursors (5+) or stage 5 precursors (5+/4−) were dissociated by 0.05% trypsin/0.04% EDTA and resuspended in HBSS− at a density of 100,000 viable cells per μl. Viability (>80%) was determined by Trypan blue exclusion. SENAs were harvested at 18+/3− and resuspended in HBSS− with about 10 SENAs per μl. Cell number per SENA was estimated after dissociation and showed that, on average, one SENA consisted of about 10,000 cells. One μl of stage 4 or stage 5 precursor cell suspensions or SENAs was grafted into the lesioned striatum using the same coordinates as for the lesion. The number of animals studied was: stage 4 precursors (n = 23), stage 5 precursors (n = 18), and SENAs (n = 17). Grafts were analyzed 1, 4, and 16 weeks after transplantation. All animal experiments were approved by the University and State of Hamburg Animal Care Committees.
Immunocytochemistry and Histology
Cultured cells were washed in phosphate-buffered saline (PBS, pH 7.3) and fixed for 15 minutes in 4% paraformaldehyde before incubation with 0.1% bovine serum albumin (BSA; Sigma-Aldrich) in PBS for 40 minutes. Primary antibodies were applied for 1 hour at room temperature. After washing in PBS, appropriate secondary antibodies were applied for 40 minutes at room temperature.
For preparation of tissue sections, animals were perfused with PBS followed by 4% paraformaldehyde in PBS. Brains were postfixed overnight in the same fixative and then cryoprotected with 20% sucrose for 2 days at 4°C. Brains were frozen in liquid nitrogen, and 25-μm serial cryosections were mounted on SuperfrostPlus slides (Menzel, Braunschweig, Germany, http://www.menzel.de) and allowed to dry overnight at room temperature. After incubation in 0.1% BSA in PBS for 40 minutes, primary antibodies were applied for 24 hours at 4°C.
Primary antibodies were monoclonal mouse antibodies to β-tubulin III (Tuj1, 1:400; Sigma-Aldrich), glial fibrillary acidic protein (GFAP, 1:1,000; Dako, Carpinteria, CA, http://www.dako.com), nestin (1:50; Developmental Studies Hybridoma Bank, Iowa City, IA, http://www.uiowa.edu/∼dshbwww), Pax-6 (1:100; Developmental Studies Hybridoma Bank), glutamate decarboxylase (GAD-6, 1:10; Developmental Studies Hybridoma Bank), NeuN (1:1,000; Chemicon), SV-2 (1:50; Developmental Studies Hybridoma Bank), monoclonal rat antibody to myelin basic protein (MBP, 1:200; Chemicon), polyclonal rabbit antibodies to Oct-4 (1:1,000; Santa Cruz Biotechnology, Santa Cruz, CA, http://www.scbt.com), NCAM (1:100; A. Dityatev) , GFAP (1:1,000; Dako), β-tubulin III (1:1,000; Covance Research Products, Princeton, NJ, http://www.covance.com), glutamate decarboxylase (GAD-65/67, 1:500; Sigma), Ki-67 (1:50; Abcam, Cambridge, U.K., http://www.abcom.com), neurofilament-200kD (NF-200, 1:200; Sigma), caspase-3 (1:2,000; R&D Systems, Minneapolis, http://www.rndsystems.com), and a polyclonal goat antibody to ChAT (choline acetyltransferase, 1:100; Chemicon).
For detection of first antibodies, appropriate secondary antibodies, coupled to Cy2 or Cy3 (Dianova, Hamburg, Germany, http://www.dianova.de) were used. Some cell cultures and tissue sections were counterstained for 10 minutes with 50 μg/ml DAPI (Sigma-Aldrich) to visualize cell nuclei. For immunostaining with the monoclonal mouse antibody to bromodeoxyuridine (BrdU, 1:200; Sigma-Aldrich), DNA was denatured with 70% ethanol for 5 minutes at room temperature and with 2.4 M HCl for 10 minutes at 37°C. BrdU was administered overnight at 4°C. For neurofilament 200 staining, tissue sections were incubated in 10 mM sodium citrate (Sigma-Aldrich), pH 9.0, at 80°C for 30 minutes prior to PBS/BSA incubation. Specimens were examined with a fluorescence (Axioplan 2; Carl Zeiss Microimaging, Inc., Jena, Germany, http://www.zeiss.com) or confocal laser-scanning microscope (LSM510; Carl Zeiss Microimaging, Inc.).
Differentiation, Cell Death, and Proliferation
To determine total cell numbers in vitro, cells were counterstained with DAPI, and the ratio of cell type-specific (e.g., NF-200) or functional (BrdU, caspase-3) marker-positive cells of all DAPI+ cells was calculated. To determine total numbers of donor cells in vivo, EGFP+ cells were counted, and the ratio of cell type-specific or functional marker-positive cells of all EGFP+ cells was calculated. To quantify proliferative activity of precursors in vitro, BrdU (10 μM final concentration; Sigma-Aldrich) was administered to the cultures 12 hours before cells were fixed with paraformaldehyde. Cell proliferation of donor cells in vivo was studied using Ki-67 antibodies. At least three independent experiments in duplicates and at least 1,000 cells per marker and experiment were analyzed. Percentages of double-labeled cells were determined and mean values ± standard error of the mean (SEM) were calculated. Student's t test was used for statistical evaluation. All experiments were done in a double-blinded manner using a confocal laser-scanning microscope.
Graft Volume and Density
Unbiased estimates of the total number of grafted cells and graft volume per animal, 1 month after transplantation, were calculated according to the optical disector and Cavalieri methods , respectively. An Axioskop microscope (Carl Zeiss) equipped with a motorized stage and a Neurolucida software-controlled computer system was used for quantitative analysis (MicroBrightField Europe, Magdeburg, Germany, http://www.mbfbioscience.com). Graft volume and cell density of the graft were determined measuring every 10th section of the graft. Transplanted cells were identified by their EGFP signal. Graft areas were outlined on digitized images to calculate volumes considering section thickness and frequency. Using random sampling in the graft core and in the periphery of the graft, cell counts were performed according to the optical disector principle at a magnification of ×40. Nuclei of DAPI+ and EGFP+ grafted cells were counted according to their position in each disector. All counts were performed in a double-blinded manner.
Formation of Neuronal Aggregates from ES Cell-Derived Nestin+ Precursors After Prolonged Exposure to FGF-2
During the first 6 days after plating in the presence of FGF-2, most ES cell-derived neural precursors were NCAM+ (95 ± 3%), Pax-6+ (94 ± 4%), and nestin+ (97 ± 1%) in comparison to immature ES cells (stage 1) that were NCAM−, Pax-6−, and nestin−. Most ES cell-derived neural precursors were also BrdU+ (84 ± 8%), showing that they are in a proliferative state. This stage of cultivation is referred to as stage 4, and the number of days in stage 4 is referred to as x+ (+ indicates the presence of FGF-2 in the culture medium, Fig. 1A). Withdrawal of FGF-2 after 6 days of culture resulted in a maturation and differentiation period that is referred to as stage 5, and the number of days in stage 5 is referred to as x− (− for the absence of FGF-2 from the culture medium, Fig. 1A). At 6+/6− (6 days with FGF-2 in stage 4 and 6 days without FGF-2 in stage 5), the percentage of nestin+ and BrdU+ cells had decreased to ∼42% and ∼16%, respectively, and ∼13% of the cells were identified as β-tubulin+ neurons, ∼25%–35% as GFAP+ astrocytes, and ∼8% as MBP+ oligodendrocytes. At 6+/18−, numbers of GFAP+ astrocytes had further increased at the expense of nestin+ cells. Total percentages of β-tubulin+ neurons and MBP+ oligodendrocytes did not change significantly with time.
When precursors were exposed to FGF-2 for as long as 18 days (18+, in the following referred to as prolonged stage 4, ps4), aggregates of β-tubulin+ neurons formed on top of a confluent monolayer of precursor cells (Fig. 1A). Whereas the percentage of BrdU+ cells in confluent precursor monolayers dropped to 36 ± 6% after 12 days in ps4, circumscribed regions within the precursor monolayers showed an increased incorporation of BrdU (Fig. 1B). Within these proliferative zones, cells started to pile up between 6+ to 12+, resulting in the formation of cellular aggregates (Fig. 1A, 1B). Cells that migrated from the outer parts of these aggregates toward the periphery show characteristics of both chain migration mainly between 6+ and 12+ and radial migration at later time points. As these aggregates and the dissociated and reseeded cells of these aggregates were never observed to form free-floating neurospheres, and as their features of differentiation are fundamentally different from neurospheres generated from the central nervous system (see below), they are referred to as SENAs.
SENAs Consist Mainly of Neuronally Committed Precursors
During formation between 6+ and 12+ in ps4, SENAs consisted almost exclusively of proliferating nestin+ precursors. However, at 18+, the percentage of nestin+ cells had decreased to 15% ± 2%, and 76% ± 6% of the cells now expressed neuron-specific β-tubulin (Fig. 1C). SENAs are oval or round spherical structures with diameters of 100–300 μm. BrdU+ cells in SENAs, which were numerous between 6+ and 12+, decreased in number at subsequent days in culture and were found evenly distributed within the inner and peripheral parts of SENAs (Fig. 1B). To verify the neural character of SENAs, neural cell adhesion molecule (NCAM) immunocytochemistry was performed showing that NCAM+ cells were predominant within SENAs (Fig. 1C). After differentiation of SENAs at 18+ /6−, the percentage of nestin+ or BrdU+ precursors had decreased further, whereas the percentage of β-tubulin+ cells had increased to 92% ± 2.9% demonstrating the neuronal commitment of most of the precursors within SENAs (Fig. 1C). Whereas at 18+ only few cells in SENAs expressed markers for mature neurons, such as neurofilament-200, SV-2, GAD-6, and ChAT, at 18+ /6−, these markers were expressed by 35%–45% of the cells (data not shown). All quantifications of cell types within SENAs were performed in optical slices of 1-μm thickness with a laser scanning microscope. To further verify their cellular composition, SENAs were selectively harvested at 18+ /3−, and dissociated single SENA-cells were processed for immunocytochemistry demonstrating a similarly high percentage of β-tubulin+ neurons (> 90%, Fig. 1D). To directly compare the differentiation potential of SENA-generating nestin+ /Pax-6+ neural precursor cells with those of nestin+ /Pax-6+ stage 4 cells and of nestin+ /Pax-6+ embryonic day 14 central nervous system-derived cells from neurosphere cultures, the yield of neurons, astrocytes, or oligodendrocytes from differentiated SENAs (dissociated), stage 5 cells, and neurosphere cultures (dissociated) was determined (Fig. 2). Whereas the percentages of β-tubulin+ neurons in differentiated SENAs generated from the EGFP+ C57BL/6J or R1 ES cell lines were >90% each, stage 5 cells contained only ∼13% (EGFP+ C57BL/6J) or ∼55% (R1) β-tubulin+ neurons. Differentiated neurosphere cultures gave rise to only 5%–10% β-tubulin+ neurons. Whereas the percentage of GFAP+ astrocytes in differentiated SENAs was < 6% (EGFP+ C57BL/6J or R1 ES cell line), 25%–35% of stage 5 cells (EGFP+ C57BL/6J or R1 ES cell line) and > 70% of differentiated neurosphere cells expressed the astrocyte marker GFAP. Oligodendrocyte-specific MBP was expressed by ∼2% of differentiated SENA cells and ∼8% of stage 5 cells (EGFP+ C57BL/6J or R1 ES cell line). Less than 1% of differentiated neurosphere cells expressed MBP.
Transplanted SENAs Do Not Form Tumors
To induce a unilateral rapid and specific degeneration of striatal medium spiny GABAergic projection neurons, the N-methyl-D-aspartate (NMDA) receptor agonist quinolinic acid was stereotaxically injected into the right mouse striatum. Three days after quinolinic acid injection, stage 4 (6+) or stage 5 (6+ /4−) cells or purified SENAs (18+ /3−) were injected into the center of the lesion site.
Transplantation of nestin+ (>95%) stage 4 cells (6+), 42% ± 1.5% of which were BrdU+ when harvested and transplanted, led to the formation of large, macroscopically visible tumors in 70% of all grafted mice (16 of 23 recipient mice, Fig. 3A). Tumors were defined as EGFP+ cell masses that compressed the ipsilateral ventricle, displaced the midline of the brain to the contralateral side, and/or extended beyond the surface of the cerebral cortex. Mice with tumors showed conspicuous motor abnormalities just 2 weeks after transplantation and were sacrificed to confirm tumor formation. Tumors contained 58.1 ± 5.1% Ki-67+ cells 1 month after grafting, demonstrating extensive proliferation in stage 4 tumor grafts (Fig. 3B, 3C). In contrast, only 2.1% ± 0.4% of the donor cells were Ki-67+ in non-tumor stage 4 grafts at 1 month after grafting (Fig. 3C). Volumes of tumor stage 4 grafts were 25.9 ± 4.0 mm3, whereas volumes of nontumor stage 4 grafts were 2.2 ± 0.3 mm3 (Fig. 3D). Four months after grafting, tumors appeared histomorphologically as mature teratomas.
Compared to stage 4 cells, significantly fewer BrdU+ cells (16% ± 1.1%) and nestin+ cells (42% ± 1%) were identified in stage 5 cells (6+/4−) at the time of harvesting and transplantation (Table 1). Teratoma formation was found in 17% of grafted mice (three of 18 recipient mice, Fig. 3A). Percentages of Ki-67+ cells in stage 5 tumor or nontumor grafts were similar to stage 4 tumor or non-tumor grafts, respectively (Fig. 3C). Stage 5 nontumor grafts tended to be smaller (1.6 ± 0.3 mm ) than stage 4 nontumor grafts (2.2 ± 0.3 mm3), but differences were not statistically significant (Fig. 3D).
We then transplanted purified SENAs (18+/3−). At the time of transplantation, percentages of BrdU+ or nestin+cells in SENAs (Table 1) were significantly lower than those in stage 5 cells. In fact, SENAs consisted of more than 75% of immature, β-tubulin+ neurons. One month after transplantation, SENA grafts were significantly smaller than stage 4 and 5 nontumor grafts (Fig. 3D). Interestingly, no teratoma formation was observed after SENA transplantation in a total of 17 recipient animals after survival periods of 1 month (12 mice) and 4 months (five mice, Fig. 3A). Thus, there was a high correlation between the proliferative activity of grafted cells at the time of transplantation and the frequency of teratoma formation (stage 4, stage 5, SENAs, correlation coefficient: 0.999, Fig. 3E).
Tumor Formation Is Not Imperatively Linked to Immature ES Cells
To investigate whether immature ES cells could have escaped predifferentiation toward stage 4 or 5 and thus could give rise to teratomas, immunocytochemistry for Oct-4, a marker only expressed in embryonic cells, was performed in stages 4 and 5 and SENA-derived cells. Whereas almost all immature ES cells were Oct-4+, very few Oct-4+ cells were found at stages 4 and 5 (Fig. 4). Within ∼3 million stage 4 or 5 cells, fewer than five Oct-4+ immature cells were detected per 100,000 cells. SENAs also contained very low numbers of Oct-4+cells (less than 4 Oct-4+ cells per 100,000 dissociated SENA cells).
SENA-Derived Cells Survive and Differentiate After Transplantation into the Quinolinic Acid-Lesioned Mouse Striatum
One month after transplantation, the total cell number in stage 5 nontumor grafts was about twofold higher than the number of transplanted cells due to proliferation of grafted stage 5 cells within the first weeks after transplantation, as determined by immunohistochemistry for the proliferation-associated antigen Ki-67. There were 7.4% Ki-67+ stage 5 cells at 1 week and 1.1% Ki-67+ stage 5 cells at 4 weeks after transplantation (Fig. 3C). In contrast, cell numbers in SENA grafts 1 month after transplantation were within the range of numbers of grafted cells, indicating low proliferative activity of SENA-derived cells over time (Fig. 3C). Caspase-3+ cells within SENA grafts or stage 5 grafts were only occasionally detected (< 1% at 4 weeks after transplantation), indicating robust survival of donor cells in SENA and stage 5 grafts (Fig. 5).
One month after transplantation, 56.9% ± 5.8% of all SENA-derived cells had differentiated into NF-200+ neurons. The neuronal phenotype in SENA grafts was further confirmed with markers for β-tubulin (60.4% ± 1.5%), NeuN, and SV-2 (Fig. 6A, 6B, 6C). The percentage of GFAP+ astrocytes and MBP+ oligodendrocytes in SENA transplants was 35.4% and 4.4%, respectively. Quantification of GAD-65/67+ cells demonstrated that on average about 47,000 and 59,000 transplanted cells had adopted a GABAergic phenotype 1 month and 4 months after transplantation, respectively (Fig. 6D).
In this study, we developed a protocol for generating and enriching SENAs. In contrast to ES cell-derived neural precursors differentiated with the conventional 5-stage protocol [12, 13], cells within SENAs differentiated into an almost pure (> 90%) population of neurons in vitro.
For transplantation experiments, we harvested SENAs at a stage of differentiation at which SENAs consist predominantly of postmitotic β-tubulin+ neurons (> 75%) that do not yet express mature neuronal markers. When undissociated SENAs of this stage were syngeneically grafted into the quinolinic acid-lesioned striatum, no teratomas were found for up to 4 months after transplantation—the latest time point tested—whereas neural precursors generated with the conventional embryoid body-based, 5-stage differentiation protocol led to the formation of teratomas in 70% or 17% of syngeneic transplantation experiments, depending on their state of maturation.
When SENAs start to form in stage 4 at 6+, they consist mainly of proliferating nestin+, Pax-6+, and NCAM+precursors (from 6+ until 12+), which subsequently (from 12+ until 18+) differentiate into immature, postmitotic, β-tubulin+/NF−/GABA− neurons (about 76% at 18+). After continued differentiation at 18+/6−, the fraction of β-tubulin+ neurons within SENAs, generated from the EGFP+C57BL/6J or R1 ES cell lines, had increased further, to more than 90% with nearly half the cells expressing mature neuronal markers, such as NF-200, GAD-6, SV-2, or ChAT, indicating the predominantly neuronal commitment of SENAs (only ∼8% are glial cells). In comparison, when applying the 5-stage differentiation protocol, the EGFP+ C57BL/6J or R1 ES cell lines, in agreement with previous studies [12, 13], yielded only about 13% or 55% neurons, respectively, whereas 33%–43% glial cells were detected pointing to an approximately equally sized potential of stage 4 cell populations to generate neurons or glial cells. To additionally compare the differentiation potential of SENAs with another type of neural precursor cells in their aggregated state, we generated neurospheres from the central nervous system of 14-day-old embryos and found, also in agreement with previous studies , a predominantly astrocytic commitment of neurospheres (> 70%). Thus, a clear increase in neuronal yield is achieved when applying the SENA differentiation protocol, indicating that the SENA-based differentiation strategy that can be applied to different ES cell lines is able to generate highly enriched, neuronally committed cell populations. In addition, it could be shown that ES cell-derived SENAs are fundamentally different from CNS-derived neurospheres, as SE-NAs cannot be generated in a free-floating fashion and as SENA cells are predominantly neuronally committed.
Purified SENAs at 18+/3− were chosen for transplantation into the acutely lesioned striatum of adult mice. Thus, SENA grafts consisted of numerous committed, immature neurons, but only a few nestin+ proliferating precursors. In comparison, stage 4 grafts, and, to a lesser degree, stage 5 grafts, consisted predominantly of proliferating nestin+ , Pax-6+, and NCAM+ precursors (Table 1). Although neurons are generally regarded to be more vulnerable toward toxic influences at the lesion site in comparison to immature nestin+ cells , most likely due to expression of NMDA receptors, which make them sensitive to excitotoxic insults , survival of cells in SENA grafts was robust, as indicated by the low percentage of caspase+ cells, which was comparable to that of stage 5 grafts. Indeed, determination of the number of NF+ (∼57,000) and β-tubulin+ (∼60,000) cells in SENA grafts revealed neuronal survival of approximately 80% of all grafted neurons. In comparison, neuronal survival of 3%–5% or 22% has been reported in previous experiments [20, 21]. Robust neuronal survival in SENA grafts might be related at least in part to the grafting of nondissociated cellular aggregates. As ∼50,000 transplanted SENA-derived cells had adopted a GABAergic phenotype, as assessed by GAD-65/67 immunohistochemistry (∼83% of all SENA-derived neurons), SENAs provide a particular neuronal subtype after transplantation that can be used for neurodegenerative diseases with a loss of this cell type.
An additional benefit of SENAs was the absence of teratoma formation in the syngeneic transplantation paradigm used in this study. Erdö and colleagues  reported that murine ES cells and ES cell-derived stage 4 neural precursors formed tumors in 86% of recipient mice but not in rats, questioning the applicability of allogeneic or syngeneic transplantation strategies in humans. The reasons for tumor formation after syngeneic transplantation but not after xenogeneic transplantation are not known, but it is likely that immune rejection of the tumor occurs after xenogeneic transplantation. Assessment of possible teratoma formation after xenogeneic transplantation into immunodeficient recipients, for instance into nude rats , should clarify the role of a potentially teratoma-preventing immune response. In agreement with the results of Erdö and colleagues, we found that stage 4 grafts formed teratomas in 70% of recipient mice but not in recipient rats (unpublished observations), suggesting that the therapeutically most relevant allogeneic or syngeneic transplantation paradigms  can reveal the tumorigenic potential of ES cell-derived neural progeny that would not be recognized in xenogenic transplantations.
In our study, teratoma formation after syngeneic transplantation was decreased to 17% when predifferentiated stage 5 cells were transplanted, suggesting a relation between graft maturity and tumor formation. This notion was further supported by the observation that syngeneic SENA grafts did not generate teratomas during a post-transplantation period of up to 4 months. A plausible reason for the reduced teratoma formation of SENAs might be that some immature teratoma-forming ES cells may have escaped pre-differentiation toward stages 4 or 5 but not toward the SENA stage. However, assessment of possible Oct-4+ cells within stages 4 or 5 and SENA cell populations revealed that immature ES cells can be found only rarely in all of these cell populations, with less than 5 Oct-4+ cells/100,000 stage 4, stage 5, or SENA cells. Thus, it is unlikely that the very few immature ES cells in stages 4 and 5 and in SENA populations would give rise to teratomas only after grafting stage 4 or 5 cells.
Alternatively, one may argue that proliferating, immature, nestin+ /Pax-6+ /NCAM+ neural precursors in stage 4 and 5 grafts, induced by cues from the host tissue, regress more easily into teratoma-forming cells than β-tubulin+ neurons within SENA grafts. Our assessment of the proliferation and maturity of stage 4 and 5 cells and cells in SENAs before transplantation revealed a high correlation between proliferation and maturity in vitro and tumor formation in vivo, supporting the possibility that at least some ES cell-derived nestin+ neural precursor cells can be re-programmed in vivo toward an immature, teratoma-forming state. Thus, this study suggests that even purification of ES cell-derived nestin+ /Pax-6+ /NCAM+ neural precursors does not prevent teratoma formation and that ES cells should rather be differentiated toward a more mature state. However, as also SENAs contain some proliferating nestin+ cells, it has to be speculated if they also have a minimal tumorigenic risk.
Our observations are thus noteworthy with regard to long-term safety and efficacy of ES cell therapy in humans. Although most transplantation studies have used immature, nestin+ neural precursors or slightly predifferentiated neural precursors we have grafted ES cell-derived aggregates consisting predominantly of β-tubulin+ neurons. Our study demonstrates that the state of maturity of ES cell-derived transplants critically determines tumorigenicity and provides a platform for further experiments aimed at using ES cell-derived transplants to treat neurological disorders.
M.D. and C.B. contributed equally to this study. We are grateful to Udo Bartsch for help and critical comments on the manuscript, Michael Bösl for assistance with the generation of ES cells, Alexander Nikonenko and Andrey Irintchev for help with statistical evaluations, and Masaru Okabe for providing the EGFP transgenic mice. This work was supported by the Deutsche Forschungsge-meinschaft (Di 881/1-1 and Scha 185/36-1).