SEARCH

SEARCH BY CITATION

Keywords:

  • Fluorescence-activated cell sorting;
  • Embryonic stem cell;
  • Sox1;
  • Teratoma;
  • Transplantation;
  • Dopaminergic neuron

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Disclosures
  9. Acknowledgements
  10. References

The differentiation of dopaminergic (DA) neurons from mouse embryonic stem cells (ESCs) can be efficiently induced, making these neurons a potential source for transplantation as a treatment for Parkinson's disease, a condition characterized by the gradual loss of midbrain DA neurons. One of the major persistent obstacles to the successful implementation of therapeutic ESC transplantation is the propensity of ESC-derived grafts to form tumors in vivo. To address this problem, we used fluorescence-activated cell sorting to purify mouse ESC-derived neural precursors expressing the neural precursor marker Sox1. ESC-derived, Sox1+ cells began to express neuronal cell markers and differentiated into DA neurons upon transplantation into mouse brains but did not generate tumors in this site. In contrast, Sox1 cells that expressed ESC markers frequently formed tumors in vivo. These results indicate that Sox1-based cell sorting of neural precursors prevents graft-derived tumor formation after transplantation, providing a promising strategy for cell transplantation therapy of neurodegenerative disorders.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Disclosures
  9. Acknowledgements
  10. References

Because embryonic stem cells (ESCs) can proliferate extensively in an undifferentiated state, they may serve as an unlimited source of cells for transplantation therapy. Several in vitro systems inducing the differentiation of functional midbrain dopaminergic (DA) neurons from ESCs have provided novel insights into cell transplantation therapy for Parkinson's disease (PD). ESC-derived DA neurons can relieve the symptoms of animal models of PD [1, 2]. For the safe clinical application of ESC-derived DA neurons, however, it is critical to prevent the formation of tumors by these pluripotent cells.

Previous work has demonstrated that transplantation of mouse ESCs can result in the formation of teratomas or teratocarcinomas, even in xenografts [2, 3]. In allograft transplantation, implantation of as few as 400 ESCs resulted in the formation of teratomas in the brain [4], suggesting that contaminating undifferentiated ESCs in ESC-derived neural cell preparations may cause tumor formation. Previous reports examining the transplantation of ventral mesencephalon from rodent embryos determined that the optimal donor age for in vivo survival of the graft was restricted to an early neural differentiation period (E11–15 for rats and E10–13 for mice) [5, 6]. After that period, the survival of the donor cells decreased dramatically, likely because mature neurons are more vulnerable to mechanical damage, inflammatory cytokines, and neurotrophic factor insufficiency, resulting in poor survival after transplantation [7]. To achieve safe and effective transplantation of ESC-derived neural cells, it is necessary to develop a method to purify cells during this early neural differentiation period.

Early neural differentiation of ESCs in adherent monocultures was monitored by Sox1 expression [8]. Sox1 is the earliest known specific marker of neuroectoderm in mouse embryos; the purification of neuroepithelial cells by fluorescence-activated cell sorting (FACS) using this marker has been reported previously [810]. To determine whether the elimination of undifferentiated ESCs prevented tumor formation, we applied this purification technique to neural precursors generated by the stromal cell–derived inducing activity (SDIA) method, in which ESCs effectively differentiate into DA neurons after culture on a mouse PA6 stromal cell feeder layer [11]. We separated SDIA-treated mouse ESCs into two distinct populations, neural and non-neural, using Sox1 expression as a marker of the neural lineage. We then transplanted each population into normal or immunodeficient mice. We demonstrate that Sox1+ cells differentiated into neurons in the brain, but did not form tumors, whereas Sox1 cells formed tumors frequently in vivo after transplantation.

Materials and Methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Disclosures
  9. Acknowledgements
  10. References

Cell Culture

The G4–2 mouse ESC line, which constitutively expresses green fluorescent protein (GFP) under the control of the CAG promoter, was the kind gift of H. Niwa (Center for Developmental Biology, RIKEN, Kobe, Japan). Sox1-GFP knockin (46C) mouse ESCs were generously provided by Dr. Austin Smith (University of Edinburgh, Edinburgh, U.K., http://www.ed.ac.uk). The generation of 46C cells has been described previously [9].

To generate 46C cells expressing β-galactocerebroside (β-gal) as a graft marker, we amplified a β-geo fragment by polymerase chain reaction (PCR) from pGT1.8IRES β-geo [12]; the identity of the fragment was confirmed by DNA sequencing. The PCR product was then inserted into the EcoRI site of pCAGGS [13] to generate pCAG–β-geo. This construct was transfected into 46C cells by electroporation (250 V, 500 μF, suspended in 800 μl phosphate-buffered saline [PBS] in a 0.4-cm cuvette). Clones resistant to G418 (200 μg/ml; Sigma, St. Louis, http://www.sigmaaldrich.com) were selected. One clone (46Cβ14) expressing strong β-gal activity was used for the following transplantation experiments. All recombinant DNA research conformed to National Institutes of Health (NIH) guidelines.

Undifferentiated mouse ESCs (G4–2, 46C, and 46Cβ14) were maintained on gelatin-coated dishes in Glasgow modified Eagle's medium (GMEM; Gibco-Invitrogen, Grand Island, NY, http://www.invitrogen.com) supplemented with 1% fetal calf serum, 5% Knockout Serum Replacement (KSR; Gibco-Invitrogen), 2 mM glutamine, 0.1 mM nonessential amino acids, 1 mM pyruvate, 0.1 mM 2-mercaptoethanol (2-ME), and 2000 units/ml leukemia inhibitory factor (Gibco-Invitrogen). Mouse ESCs were differentiated in SDIA as previously reported [11]. Briefly, ESCs were cultured on a PA6 stromal cell feeder layer in differentiation medium (GMEM supplemented with 5% KSR, 2 mM L-glutamine, 0.1 mM nonessential amino acids, 1 mM pyruvate, and 0.1 mM 2-ME). The day on which ESCs were plated on PA6 monolayers was defined as SDIA day 0.

FACS

ESC colonies differentiated on PA6 cells for 4 days were isolated using Collagenase B (1 mg/ml; Roche, Basel, Switzerland, http://www.roche.com), dissociated into a single-cell suspension with 0.25% trypsin-EDTA (Gibco-Invitrogen), and re-suspended in cold differentiation medium. To separate two distinct cell populations, Sox-GFP+ and Sox1-GFP, cells were sorted using a FACSAria cell sorter and FACSDiva software (Beckton, Dickinson and Company, San Jose, CA, http://www.bd.com). Dead ESCs and PA6 feeder cells were identified and eliminated by propidium iodide staining and forward-side scatter gating, respectively. Gates for each population were set so that the two subsets sorted based on Sox1 staining would not overlap when reanalyzed. Sorted cells were immediately either transplanted or replated onto chamber slides to characterize their behavior in vitro.

To examine the proliferation of the isolated cells, the sorted cells were replated onto chamber slides coated with poly-L-ornithine (Sigma), laminin (Sigma), and fibronectin (Gibco-Invitrogen) (OLF). After culture for 4 days in Alpha Minimum Essential Medium (αMEM; Gibco-Invitrogen), 5-bromo-2′-de-oxyuridine (BrdU; Nacalai Tesque, Kyoto, Japan, http://www.nacalai.co.jp) was added at a final concentration of 5 μg/ml. Twenty-four hours later, cells were fixed, denatured with 2N HCl, and stained with an anti-BrdU antibody (see below). In the differentiation assay, sorted cells were replated onto either OLF-coated slides in αMEM or PA6-coated slides in GMEM. Cells were fixed and immunostained either 5 or 10 days after replating. Nuclei were counterstained with 10 μg/ml Hoechst 33342 (Molecular Probes, Eugene, OR, http://www.probes.com).

Immunohistochemistry, Mediated dUTP Nick-End Labeling, and RT-PCR

After fixation in 4% paraformaldehyde, cells were incubated with the following primary antibodies: rabbit polyclonal antibodies against tyrosine hydroxylase (TH; Chemicon International, Inc., Temecula, CA, http://www.chemicon.com), aromatic acid decarboxylase (AADC; PROTOS Immunoresearch, Burlingame, CA, http://www.protosimmuno.com), or Ki67 (Novocastra, Newcastle upon Tyne, U.K., http://www.novocastra.co.uk), a mouse monoclonal antibody specific for Tuj1 (Covance Research Products, Richmond, CA, http://www.covance.com) and BrdU (Roche), a rat polyclonal antibody against dopamine transporter (DAT; Chemicon International, Inc.), goat polyclonal antibodies that recognize Oct4 (Santa Cruz Biotech, Santa Cruz, CA, http://www.scbt.com) or β-gal (Biogenesis, Poole, U.K., http://www.biogenesis.co.uk), and a sheep polyclonal antibody specific for TH (Chemicon International, Inc.). Appropriate cyanin-3 (Cy3)– and Cy5-labeled secondary antibodies (Jackson Immunoresearch Laboratories, Inc., West Grove, PA, http://www.jacksonimmuno.com) were used to visualize antibody binding. Immunostained cells and brain sections were evaluated using an Olympus DP70 optical microscope or a Fluoview FV300 laser confocal microscope (Olympus Optical Co., Tokyo, http://www.olympus.co.jp). When specified, immunostaining for Ki67 was performed using the avidin-biotin peroxidase method. Briefly, free-floating sections were incubated sequentially in rabbit anti-Ki67 antibody, biotinylated anti-rabbit immunoglobulin G (Vector, Burlingame, CA, http://www.vectorlabs.com), and avidinbiotin-peroxidase complex (Vector). Immunoreactivity was visualized using 3,3′-diaminobenzidine tetrahydrochloride dihydrate (Vector).

Cell death was determined by terminal deoxynucleotidyl transferase-dUTP nick-end labeling (TUNEL) assay using an In Situ Cell Death Detection Kit (Roche). TUNEL staining of both Sox1+ and Sox1 populations was performed 24 hours after plating on poly-d-lysine–coated chamber slides (Beckton, Dickinson and Company).

We extracted total RNA from both ESC colonies detached from PA6 feeder layers and FACS-sorted populations using the RNeasy Minikit (Qiagen, Hilden, Germany, http://www1.qiagen.com). FACS-sorted cells were directly collected into RLT lysis buffer. Total RNA (1 μg) was reverse-transcribed using an oligo dT12–18 primer with a Superscript kit (Gibco-Invitrogen). PCR was performed using 1/20 of the final cDNA volume with Hotstartaq DNA polymerase (Qiagen). For Sox1, Sox2, and CK17 amplification, GC melt polymerase mix (Beckton, Dickinson and Company) was used to facilitate PCR of regions with high GC content. For each amplification reaction, controls without the addition of reverse transcription (RT) were performed to exclude genomic DNA contamination. Reactions were performed at 55°C for 30 cycles, with the exceptions of Oct4 (60°C, 25 cycles) and glyceraldehyde-3-phosphate dehydrogenase (GAPDH) (55°C, 25 cycles). The primer sequences and product lengths were as follows: Sox1, forward 5′-CCTCG-GATCTCTGGTCAAGT and reverse 5′-TACAGAGCCGGC-AGTCATAC, 593 bp; Sox2, forward 5′-CACAGATGCAAC-CGATGCA and reverse 5′-GGTGCCCTGCTGCGAGTA, 121 bp; Nestin, forward 5′-GGAGTGTCGCTTAGAGGTGC and reverse 5′-TCCAGAAAGCCAAGAGAAGC, 327 bp; Engrailed 1 (En1), forward 5′-TGGTCAAGACTGACTCACAGCA and reverse 5′-TCTCGTCTTTGTCCTGAACCGT, 389 bp; Oct4, forward 5′-GGCGTTCTCTTTGGAAAGGTGTTC and reverse 5′-CTCGAACCACATCCTTCTCT, 312 bp; Nanog, forward 5′-AGGGTCTGCTACTGAGATGCTCTG and reverse 5′-CAACCACTGGTTTTTCTGCCACCG, 363 bp; ERas, forward 5′-ACCATGACCCCACTATCCAA and reverse 5′-GTCT-TCTTGCTTGATTCGGC, 433 bp; CK17, forward 5′-TGC-CACCATGACCACCACCATC and reverse 5′-AGAAC-CAGTCTTCGGCATCCTT, 832 bp; GAPDH, forward 5′-GACCACAGTCCATGCCATCACT and reverse 5′-TC-CACCACCCTGTTGCTGTAG, 454 bp.

Transplantation

Animal experiments were performed in accordance with institutional guidelines and with the NIH Guidelines for the Care and Use of Laboratory Animals in Neuroscience Research produced by the Society for Neuroscience. All surgical procedures described below were performed after anesthesia of animals with sodium pentobarbital (30 mg/kg). Male C57BL/6 mice (Japan SLC Inc., Shizuoka, Japan, http://www.jslc.co.jp) weighing 18–22 g, which were not lesioned with either 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP) or 6-hydroxydopamine (6-OHDA), were used for intracranial transplantation. For experiments in which a cell suspension was introduced into mice brains, ESC colonies formed on PA6 monolayers were detached after 4, 6, and 8 days of culture and dissociated by incubation in papain (Worthington, Freehold, NJ, http://www.worthington-biochem.com). We implanted 2 μl of a cell suspension at 105 cells per 1 μl differentiation medium or 2 × 105 FACS-sorted cells (prepared as described above) into the adult mouse striatum.

Mice were provided with drinking water containing 2% ethanol and 200 μg/ml cyclosporine A (CyA) from 3 days prior to intracranial transplantation until they were sacrificed. This treatment maintained CyA blood concentrations (measured by radioimmunoassay) at 297 ± 81 ng/ml, a level comparable to that of patients undergoing liver transplantation at Kyoto University Hospital [14]. Under deep anesthesia, mice were placed in a stereotaxic frame (Narishige, Tokyo, http://www.narishige.co.jp) and given an injection of a 2-μl (1-μl/minutes) cell suspension into the striatum (from the bregma: A +1.0, L +2.0, V +3.0, incisor bar 0) using a Hamilton microsyringe (GL Sciences Inc., Tokyo, http://www.gls.co.jp) fitted with a 26-gauge blunt needle. Injection coordinates were determined according to the Franklin and Paxinos atlas [15]. As a control, an additional group of mice was subjected to sham operation injecting differentiation medium alone.

Eight weeks after transplantation, mice were perfused transcardially first with PBS, then with 4% paraformaldehyde. Brains were removed and sectioned at a thickness of 40 μm. Free-floating sections were immunostained with the indicated primary antibodies and appropriate secondary antibodies as described above. The number of TH+ cells was quantified in every third section for both the graft and the surrounding tissue. These values were corrected using the Abercrombie method [16]. The presence of grafted cells was evaluated by fluorescence of GFP, which was constitutively expressed by the transplanted G4–2 ESCs. During the grafting of cells sorted by FACS, in which GFP fluorescence was not present, hematoxylineosin (HE) staining and β-gal immunoreactivity were used to identify 46Cβ14 ESCs. The observation of a Ki67-positive mass in the brain was defined as positive tumor formation. The graft area, identified by GFP fluorescence (G4–2) or HE staining (46C), was outlined in white and examined using image analysis software (Scion Corporation, Frederick, MD, http://www.scioncorp.com). The graft volume was calculated by summing the graft areas over every sixth section (thickness, 40 μm).

To measure teratoma formation, samples at 106 cells per 10 μl differentiation medium were injected into the abdominal subcutaneous space of female CB17/Icrscid Jcl scid/scid mice (CLEA, Japan Inc., Tokyo, http://www.clea-japan.com) weighing 15–20 g. As a control, 106 naïve ESCs that had passed through the FACS machine (sham-FACS) were also injected. Resultant tumors were removed and analyzed 4 weeks later.

Statistical Analysis

Statistical analyses were performed using a commercially available software package (Statview 5.0; SAS Institute Inc., Cary, NC, http://www.sas.com). Data expressing the number of surviving TH+ cells in vivo were tested by one-factor analysis of variance (ANOVA) and Tukey-Kramer post hoc analysis. TUNEL analyses, in vitro proliferation and differentiation data, and graft volume were tested using the Student's t test. Differences were considered statistically significant when p < .05. Data are presented as the means ± SEM. All in vitro results were derived from at least three independent experiments.

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Disclosures
  9. Acknowledgements
  10. References

Transplantation of Unsorted ESCs

Culture of mouse ESCs on PA6 cell feeder layers induced neuronal differentiation through SDIA, with TH+ cells first becoming detectable at 5–6 days and significantly increasing in proportion after 8 days [11]. Mature TH+ cells, however, exhibit poor survival when grafted into mouse brain [7]. To determine the in vivo survival of the cells displaying early DA differentiation, 2 × 105 ESCs (G4–2 cells that constitutively express GFP) were treated with SDIA for 4, 6, and 8 days (4-, 6-, and 8-day SDIA cells, respectively), dissociated, and grafted into the adult mouse striatum. As a control, undifferentiated ESCs were also transplanted. In vitro characterization of 4-, 6-, and 8-day SDIA cells revealed differing degrees of Tuj1 (neuronal) and TH (DA) expression (Fig. 1A). The extent of tumor formation and DA differentiation within grafts 8 weeks after transplantation is summarized in Table 1 and Figure 1B, respectively. Only mice exhibiting no tumor formation were included in the examination of TH+ cell generation. Immunofluorescence analysis revealed that the number of TH+ cells generated from 4-day SDIA cells (311 ± 98, n = 7) was significantly higher (one-factor ANOVA, p = .0045) than the number observed from 0-day (undifferentiated ESCs: 26 ± 6, n = 2), 6-day (36 ± 10, n = 3), or 8-day SDIA cells (71 ± 29, n = 5) (Fig. 1B). Although the ESCs did not differentiate into TH+ cells within the first 4 days of SDIA treatment in culture (Fig. 1A), these results suggest that 4-day SDIA cells contained a considerable number of DA neuron progenitors suitable for the generation of DA neurons in vivo. The colocalization of TH with GFP confirmed that the TH+ cells were derived from the grafted ESCs (Fig. 2A–C). These cells also expressed Tuj1 (Fig. 2D) and midbrain DA neuronal markers, such as AADC (Fig. 2E) and DAT (Fig. 2F), indicating that the SDIA-treated ESCs differentiated into midbrain DA neurons.

Although 4-day SDIA cells were the most efficient source of DA neurons in vivo, these samples formed tumors in two of the nine animals. The tumors primarily contained cells that expressed the Ki67 proliferation antigen. Immunostaining also revealed that Oct4, an ESC transcriptional organizer expressed in the inner cell mass and epiblast during development [17], colocalized with Ki67 in the tumors, suggesting that tumor formation is due to contamination of the graft with naïve embryonic stem– or epiblast-like (hereafter collectively referred to as ES-like) cells (Fig. 2G–I). For tumors generated from 4-day SDIA cells, the frequency (22%) was lower and the volume smaller (0.56 ± 0.16 mm3) than those generated from undifferentiated ESCs (71%, 5.28 ± 2.00 mm3), likely because of the partial neuronal differentiation of 4-day SDIA cells (Table 1).

Purification of Neural Precursor Cells by FACS

To confirm that the tumors were derived from contaminating ESCs, we separated the 4-day SDIA cells into neural precursors and ES-like cells by FACS. While Sox1, Sox2, and Nestin are all markers of mammalian neural precursors [10, 18, 19], RT-PCR analysis revealed that Sox2 and Nestin were expressed in both naïve ESCs and 4-day SDIA cells. Sox1, however, was specifically expressed in only the 4-day SDIA cells (Fig. 3A). These results are consistent with previous reports indicating that Sox1, but not Sox2 or Nestin, specifically labels neural precursor cells [8, 20, 21]. We therefore subjected Sox1-GFP ESCs (46C cells, generously provided by Dr. Smith), in which the egfp reporter gene had been inserted into the Sox1 locus [8, 9], to SDIA treatment for 4 days. We then marked the ES-like cells by immunofluorescence staining for Oct4. The GFP- and Oct4+-cell populations segregated into distinct, nonoverlapping populations (Fig. 3B). The gating for FACS was strictly set so that the two populations would not overlap when reanalyzed (Fig. 3C).

RT-PCR analysis revealed that the GFP+ population expressed Sox2, Nestin, and Sox1. En1, a transcription factor specifically expressed in the midbrain from the early embryonic to adult stages [22], was also detected. These cells lacked detectable levels of Oct4, Nanog (an ESC marker) [23], ERas (an oncogene regulating the tumorigenicity of ESCs) [24], and CK17 (an epidermal marker) (Fig. 4A). In contrast, the GFP-negative population did not express detectable Sox1, whereas Oct4, Nanog, and ERas transcripts were easily detectable, suggesting that these cells retain an ESC-like character.

Next, to investigate the proliferation and differentiation ability of the isolated cells, we plated either GFP+ or GFP cells onto OLF-coated chamber slides. After proliferation for 2 days, GFP (Sox1+) cells formed small colonies, which then underwent cell cycle arrest and differentiation. In contrast, GFP+ (Sox1+) cells continued to proliferate for at least 10 days. Measurement of BrdU incorporation for 24 hours between day 4 and day 5 revealed that the proliferation ability of GFP cells was significantly higher than that of GFP+ cells (Fig. 4B, C). In addition, GFP+ and GFP cells expressed distinct lineage markers (Fig. 4D). Five days after replating, most GFP+ cells differentiated into Tuj1+ postmitotic neurons; this differentiation was rarely observed for GFP cells (Fig. 4D, E). Approximately half of GFP+ cells retained Oct4 immunoreactivity (an ES-like phenotype), whereas GFP+ cells did not (Fig. 4D, F). Taken together, GFP+ and GFP cells exhibit the characteristics of neural precursor cells and ES-like cells, respectively.

After culture of GFP+ cells on OLF substrate in αMEM medium for 10 days, less than 1% of TH+ cells could be observed (data not shown). When these cells were cultured on a PA6 feeder layer in the differentiation medium for an equivalent period, virtually all surviving cells differentiated into Tuj1+ postmitotic neurons, with 22% ± 6% also exhibiting TH immunoreactivity (Fig. 4G). This ratio of TH-/Tuj1+ cells was comparable to that obtained for mouse ESC cultures plated on PA6 feeder cells for 14 days (25% ± 5%). Together with our data demonstrating the expression of the midbrain-specific marker En1 in GFP+ cells (Fig. 4A), the ability of these cells to differentiate in vitro into TH+ cells suggests their competency to generate DA neurons when stimulated appropriately. In contrast, plating GFP cells on a PA6 feeder layer generated colonies that were indistinguishable from those induced from naïve ESCs after 10 days (Fig. 4H). The rates of Tuj1+ (94%) and TH+ (77%) cells induced from these cultures were comparable to those generated from naïve ESCs after 10 days (97% and 84%, respectively [11]), suggesting that GFP cells retain ES-like properties.

Transplantation of Purified Neural Precursor Cells

The in vitro proliferation assays and the distinct marker expression profile, particularly the downregulation of ERas, suggested that GFP+ cells would not overproliferate in vivo. We examined the tumorigenicity of GFP+ and GFP cells by grafting them into the brains of adult mice (Table 2). Detection of β-gal expression by Sox1-GFP ESCs transfected with β-gal (46Cβ14) was used to identify the grafted cells. Whereas GFP cells survived in 10 of the 29 mice receiving grafts (34%), tumor formation was observed in 9 of these 10 cases (90% of the surviving cases). Tumors derived from GFP cells appeared to be heterogeneous (Fig. 5B); histological studies revealed that tumors contained a variety of cell types (Fig. 5C), including neural (ectodermal; Fig. 5D), cartilaginous (mesodermal; Fig. 5E), and undifferentiated (Fig. 5F) cells. Although endodermal tissues were rare, these findings suggest that the graft-derived tumors exhibited teratoma-like characteristics. The Sox1 grafts varied in size from 0.106 to 39.7 mm3 (mean, 7.50 mm3).

GFP+ cells, however, survived in 23 of 39 mice (59%) but were never observed to form tumors (Figs. 5A, 5K). The size of the graft was less variable, ranging from 0.029 to 0.192 mm3 (mean, 0.090 mm3). These were significantly smaller, however, than the grafts formed by GFP+ cells (p = .0074; Fig. 5G). An average of 568 ± 94 β-gal+ cells survived per graft, 97% of which also expressed Tuj1. These results suggest that the majority of grafted GFP+ cells differentiated into neurons (Fig. 5I). TH+ cell bodies, averaging 14.5 ± 2.4 per graft, were also identified within the graft (Fig. 5J). No TH+ cells could be identified within the striata of control mice injected with culture medium alone (n = 5, data not shown). Within a single 40 μm–thick section, TH+ fibers could be traced within the graft as far as 100 μm from the cell body to the host–graft interface (Fig. 5L).

To determine whether the difference in tumorigenic potential between the two populations resulted from differential cell viability, both GFP+ and GFP cells were plated onto poly-d-lysine–coated chamber slides. Twenty-four hours after plating, cell death was analyzed by TUNEL staining. By FACS, apoptosis was observed in 69.5% and 56.2% of the GFP+ and GFP cells, respectively. Both of these values were significantly higher than that (20.2%) seen in an unsorted cell population dissociated to facilitate FACS analysis (Fig. 5H, p < .05). The rate of apoptosis observed in GFP+ cells was slightly but significantly higher (Fig. 5Hp, = .018) than that seen in GFP cells, suggesting that GFP cells are more resistant to damage during FACS than are GFP+ cells. Cell viability may at least partially account for the difference in tumorigenicity between these two cell populations.

To confirm the ES-like, tumorigenic character of GFP cells and the nontumorigenic character of GFP+ cells, a subset of each population (106 cells each) was grafted subcutaneously into severe combined immunodeficient (SCID) mice. The same number of naïve ESCs taken directly after sham-FACS was also grafted as a positive control. Four weeks later, visible tumors were removed and weighed. GFP+ cells failed to form visible tumors, whereas GFP grafts formed teratomas with an average tumor weight of 0.89 g in four of six mice (Fig. 6A). Although the rate of teratoma formation for these cells was not as high as that seen for naïve ESCs, which formed tumors averaging 3.31 g in weight in 83% of the grafted mice (n = 6), histological analysis revealed that GFP cells did generate teratoma-like heterogenous tumors. Endodermal tissues were rarely observed in tumors derived from either GFP cells or naïve ESCs (Fig. 6B). These results suggest that the GFP population contains a considerable number of ES-like cells with the capacity to form teratoma-like tumors in vivo. In contrast, the GFP+ cell population appears to be free of such cells.

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Disclosures
  9. Acknowledgements
  10. References

ESCs differentiate heterogeneously and fail to behave synchronously. As a result, grafts derived from differentiated ESCs tend to be contaminated by undifferentiated ESCs [8, 11]. The transplantation of undifferentiated mouse ESCs or embryoid bodies into the adult brain results in the formation of teratomas or teratocarcinomas [2, 3]. In naïve mouse ESC allografts, the transplantation of as few as 400 ESCs can result in the formation of teratomas within the brain [4]. In contrast, teratoma formation was not typically observed in the transplantation of differentiated neural grafts derived from mouse or human ESCs [1, 2527]. All of these studies, however, were performed in xeno-graft models (e.g., mouse cells into a rat host), suggesting that the prevention of tumor formation may be due to immunorejection. In fact, allografts of mouse ESC-derived neural precursors still resulted in tumor formation or tumor-like cell growth, despite full (14 days or more) in vitro differentiation [3, 28, 29]. It remains possible, however, that a small fraction of undifferentiated ES-like cells contaminated the graft. Given the efficiency of tumor formation by these cells, a highly stringent method to eliminate undifferentiated ESCs from transplanted grafts is crucial for the safety of ESC-based transplantation therapies. Here, we demonstrate that Sox1-based purification is a promising strategy for preventing tumor formation, completely eliminating the ES-like cells with tumorigenic potential. The differential tumorigenicity of GFP+ and GFP cells appears to be due to their different proliferative potentials rather than to distinct cell viabilities in vivo (Figs. 4B, 4C, 5H). The advantage of this procedure over previously described methods was demonstrated by the absence of tumorigenesis in allografts of Cv129 mouse ES-derived cells into C57B6 mouse hosts with adequate immunosuppression (see Materials and Methods) or immunodeficient hosts.

In our assessment of teratoma formation, GFP cells frequently generated teratoma-like tumors in a manner similar to naïve ESCs, even after being subjected to cell damage by FACS (Fig. 6A). The tumors generated by GFP cells and naïve ESCs were histologically identical, suggesting that the tumorigenic potential of GFP cells is comparable to that of naïve ESCs (Fig. 6B). A previous study suggested that the differentiation of ESCs within a teratoma mimicked normal embryogenesis, in which complex interactions among various embryonic tissues are required for the differentiation of definitive endodermal tissues, but not ectodermal or mesodermal tissues [30]. These results correlate well with the only rare observation of endodermal differentiation in GFP cell–derived teratoma-like tumors and naïve ESC-derived growths.

In this study, FACS purification significantly reduced the number of DA neurons surviving in vivo (311 TH+ cells without sorting in comparison with 14.5 TH+ cells after sorting). This reduction in cell numbers can be attributed to a high rate of apoptosis in the sorted cells. TUNEL staining revealed that 70% of the sorted GFP+ cells were in the process of undergoing cell death, whereas only 20% were apoptotic in the unsorted cell population (Fig. 5H). Thus, for every 200,000 cells prepared, the equivalent of only 60,000 Sox1+ cells was grafted, in comparison with 160,000 cells grafted from unsorted cells. Another possible explanation for the increased numbers of TH+ cells in unsorted grafts is that the undifferentiated ESCs contaminating this cell population may contribute to the generation of additional DA neurons. Suspensions of undifferentiated mouse ESCs or embryoid bodies grafted into rodent striatum generate numerous TH+ neurons [2]. Undifferentiated ESCs may proliferate extensively in the mouse brain, spontaneously giving rise to TH+ neurons.

In this study, the survival rate of GFP+ cells (568/200,000 [0.3%]) was lower than that seen in previous reports, demonstrating a survival rate of 0.8% (4000/500,000) for grafted cells [1]. Given the high rate of apoptosis after cell sorting, administration of anti-apoptotic reagents may prevent the massive cell death caused by damage incurred during FACS [31]. Although 97% of the grafted GFP+ neural precursors differentiated into postmitotic neurons (Fig. 5I), consistent with previous reports that glial cells are rarely generated after SDIA treatment [11], the in vivo differentiation of TH+ neurons from surviving Sox1+ cells (TH/Tuj1 = 14.5/550 [2.6%]) was significantly less efficient than the in vitro differentiation of equivalent cells on PA6 feeder layers (TH/Tuj1 = 22%; Fig. 2E). Despite the expression of the early midbrain marker En1, the sorted Sox1+ cells likely require additional signals that are lacking in vivo to become mature DA neurons. Treatment of Sox1+ cells with exogenous factors such as sonic hedgehog, fibro-blast growth factor-8, and brain-derived neurotrophic factor prior to transplantation might provide a more efficient induction of DA neurons [1, 28, 32]. Another possible reason for the reduction in DA differentiation is that we used normal mice lacking any damage to the DA system. Depletion of host DA neurons within the substantia nigra by neurotoxic reagents, such as MPTP or 6-OHDA, might promote DA neuronal differentiation and survival [33]. Although midbrain DA neurons can be efficiently generated from human ESCs (hESCs) [34], these derivatives are heterogeneous and persistently contain Oct4-positive undifferentiated cells. Because undifferentiated hESCs form teratomas in SCID mice [35], it is necessary to eliminate these cells before transplanting any ES-derived cell population into human patients. Given that hESC-derived neural precursors express Sox1, a marker not found in ESCs [34], a FACS-based strategy similar to that presented in this work would be applicable to clinical trials involving hESCs. A larger sample number (39 in the present study) and longer observation period (8 weeks), however, would be required to rigorously elucidate the safety of this strategy.

Conclusion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Disclosures
  9. Acknowledgements
  10. References

We used FACS to purify neural precursors from ESC-derived grafts which would otherwise form tumors. Tumor formation was completely averted for at least 8 weeks in the group transplanted with purified neural precursors. Given that tumor formation must be prevented in human patients, this result provides a promising strategy for cell transplantation in PD. Although additional studies are required to obtain the increased numbers of functional DA neurons necessary to improve clinical symptoms from purified neural precursors, our results suggest the tumor-free clinical feasibility of ESC transplantation therapy.

Table Table 1.. Tumor formation of embryonic stem cells treated with stromal cell–derived inducing activity
Thumbnail image of
Table Table 2.. Survival and tumorigenicity of the GFP+ or GFP cells in vivo
Thumbnail image of
thumbnail image

Figure Figure 1.. Dopaminergic differentiation of naïve embryonic stem cells (ESCs) or 4-, 6-, and 8-day stromal cell–derived inducingactivity (SDIA) cells. (A): The percentages of ESC colonies exhibiting Tuj1 (neuronal, ○) or TH (dopaminergic, ▪) immunoreactivities show time-dependent sequential differentiation in vitro. Virtually no TH+ cells are observed in 4-day SDIA cell cultures. (B): Number of TH+ cells in grafts 8 weeks after transplantation. Animals with tumor formation were excluded from this analysis. More TH+ cells were observed in the 4-day SDIA grafts than in those from naïve ESCs or 6- or 8-day SDIA grafts (*p < .05; one-factor analysis of variance and Tukey-Kramer post hoc test). Abbreviations: ESC, embryonic stem cell; TH, tyrosine hydroxylase.

Download figure to PowerPoint

thumbnail image

Figure Figure 2.. In vivo differentiation and survival of stromal cell–derived inducing activity (SDIA)–induced dopaminergic neurons 8 weeks after transplantation. (A–C): Immunohistochemistry of 4-day SDIA grafts. TH expression colocalized with the graft marker GFP. Scale bar = 100 μm. (D–F): Immunohistochemistry of 4-day SDIA grafts at a higher magnification. Colocalization of Tuj1, AADC, and DAT with TH suggested that the grafted cells differentiated into midbrain dopaminergic neurons. Scale bar = 50 μm. (G–I): Immunohistochemistry indicated that the embryonic stem cell–like marker Oct4 is abundantly expressed by Ki67-positive cells derived from 4-day SDIA grafts. Scale bar = 100 μm. Abbreviations: AADC, aromatic acid decarboxylase; DAT, dopamine transporter; GFP, green fluorescent protein; TH, tyrosine hydroxylase.

Download figure to PowerPoint

thumbnail image

Figure Figure 3.. Differentiation and separation of Sox1-GFP ESCs in vitro. (A): Reverse transcription–polymerase chain reaction analysis of neural progenitor markers confirmed the neural-specific expression of Sox1. Sox1 expression was limited to the 4-day SDIA cells, and Sox2 and Nestin expression by undifferentiated ESCs was also observed. (B): GFP expression by 46C (Sox1-GFP) ESCs was seen after differentiation for 4 days in SDIA. Immunohistochemistry for the ESC-like marker Oct4 demonstrated that GFP and Oct4 were expressed by distinct cell populations. Scale bar = 100 μm. (C): Flow cytometric profiles of 4-day SDIA cells differentiated from Sox1-GFP ESCs, using the gate settings of M1 for Sox1+ cells and M2 for Sox1 cells. Reanalysis confirmed the lack of overlap between the two populations. Abbreviations: ESC, embryonic stem cell; FITC-A, fluorescein isothiocyanate; GAPDH, glyceraldehyde-3-phosphate dehydrogenase; GFP, green fluorescent protein; SDIA, stromal cell–derived inducing activity.

Download figure to PowerPoint

thumbnail image

Figure Figure 4.. Characterization of GFP+ and GFP populations in vitro. (A): Reverse transcription–polymerase chain reaction analysis of neural, epidermal, and undifferentiated ESC markers revealed differences in marker gene expression profiles between GFP+ and GFP populations. (B, C): BrdU incorporation assay of both GFP+ and GFP cells revealed that GFP cells have a greater proliferative potential than GFP+ cells. BrdU was added for 24 hours to the medium 4 days after replating. Scale bar = 100 μm. In (C), *p < .05; Student's t test. (D–F): GFP+ and GFP cells exhibited distinct marker expression patterns (Tuj1: postmitotic neurons, Oct4: embryonic stem-like cells) 5 days after replating. Virtually no Oct4-positive cells were detectable in the GFP+ cell population. Scale bar = 100 μm. In (E, F), *p < .05; Student's t test. (G): GFP+ cells were plated on PA6 feeder layers and cultured for 10 days. Images in these panels were acquired in the same field of view using different filters. GFP+ cells differentiated into Tuj1-positive cells possessing axonal projections (left panel). A subgroup of these cells simultaneously expressed TH, suggesting that they adopted a dopaminergic neuronal fate (right panel). Each cell terminally differentiates independently without proliferating to form colonies. Scale bar = 100 μm. (H): GFP cells were cultured on PA6 feeder layers for 10 days. GFP cells generated colonies containing Tuj1- (left panel) and TH+ (right panel) neurons indistinguishable from those generated by naïve ESCs cultured on PA6 cells for 14 days. Scale bar = 200 μm. Abbreviations: BrdU, 5-bromo-2′-deoxyuridine; En1, Engrailed 1; ESC, embryonic stem cell; GAPDH, glyceraldehyde-3-phosphate dehydrogenase; GFP, green fluorescent protein; TH, tyrosine hydroxylase.

Download figure to PowerPoint

thumbnail image

Figure Figure 5.. Transplantation of GFP+ and GFP cells into the brain. (A, B): Ki67 immunostaining of GFP+ and GFP grafts. (A): The GFP+ graft (arrowheads) was Ki67-negative. The inset displays a higher magnification of the boxed area, in which proliferating neural precursors lining the ventricular surface serve as a positive control for Ki67 staining. (B): An intrastriatal Ki67-positive tumor derived from a GFP graft appears to be destructive. Scale bar = 500 μm. (C–F): HE staining of a Sox1 graft–derived tumor. The tumor contained a mixture of tissue types (C), including neural tube-like (D), cartilaginous (E), and undifferentiated ESC-like (F) tissues, indicative of teratoma-like growth. Scale bars = 500 μm (C), 50 μm (D–F). (G): The graft volume of GFP (mean, 0.09 mm3; n = 23) and GFP (mean, 7.50 mm3; n = 10) transplants differed significantly (*p = .0074; Student's t test) between the two graft types. (H): TUNEL-positive cells in unsorted, GFP+, and GFP cells 24 hours after replating on poly-L-lysine–coated chamber slides, demonstrating an increased rate of apoptosis in both GFP+ and GFP cell populations in comparison with that observed for unsorted cells (*p and **p < .05; Student's t test in comparison to no FACS population). A statistically significant difference between GFP+ and GFP cell rates of apoptosis (p = .018; Student's t test) suggests tolerance against dissociation and FACS procedure in the GFP population. (I): Immunohistochemical analysis of both Tuj1 and β-gal indicates that GFP+ cells differentiate into postmitotic neurons in vivo. Scale bar = 100 μm. (J): TH+ cells were identified within the graft. Scale bar = 100 μm. (K): β-gal immunostaining indicated a GFP+ graft occupying the maximum allowable space within the host striatum. To facilitate visualization, sections stained with an anti-TH antibody were converted to grayscale. Scale bar = 200 μm. (L): High magnification of a surviving TH+ cell that extended its neurites (arrowheads) to reach the host-graft interface (dashed line). Scale bar = 50 μm. Abbreviations: β-gal, β-galactocerebroside; FACS, fluorescence-activated cell sorting; GFP, green fluorescent protein; HE, hematoxylineosin; TH, tyrosine hydroxylase; TUNEL, terminal deoxynucleotidyl transferase-mediated dUTP nick-end labeling.

Download figure to PowerPoint

thumbnail image

Figure Figure 6.. Teratoma formation assay. Subcutaneous tumors were weighed and analyzed 4 weeks after transplantation. (A): Tumors from GFP+, GFP , and naïve ESCs after sham- fluorescence-activated cell sorting transplanted into SCID mice (n = 6, each group) are shown. Tumor weight is also shown. Teratoma-like tumors were observed in mice subjected to GFP cell grafts as well as in those given naïve ESC grafts, although the frequency and weight of the tumors were reduced in the GFP cell–treated mice. Scale bar = 1 cm. (B): Histological analysis revealed that tumors generated from GFP cells were comparable to those generated by naïve ESCs, both of which contained heterogeneous tissue, including neural tube–like, primitive bone–like, primitive muscle–like, and undifferentiated ESC–like structures with rare definitive endodermal tissues. Scale bar = 200 μm. Abbreviations: ESC, embryonic stem cell; GFP, green fluorescent protein; SCID, severe combined immunodeficient.

Download figure to PowerPoint

Acknowledgements

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Disclosures
  9. Acknowledgements
  10. References

We thank H. Niwa for providing the G4–2 cells and A. Smith for supplying the 46C cells. We also thank K. Minami and M. Kawaguchi for FACS advice; H. Suemori for technical advice on the teratoma assays; A. Nishiyama, N. Murata, and T. Yokota for additional technical assistance; T. Terano for help with the histological analyses; and T. Palmer and D. Yabe for helpful comments on this manuscript. This study was supported by the following grants: grants-in-aid for Scientific Research, grants in Kobe Cluster, and Establishment of International Centers of Excellence for Integration of Transplantation Therapy and Regenerative Medicine from the Ministry of Education, Culture, Sports, Technology and Science, and Health Sciences Research grants in Research on Human Genome, Tissue Engineering, and Food Biotechnology from the Ministry of Health, Labor and Welfare of Japan.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Disclosures
  9. Acknowledgements
  10. References