Neural conversion of ES cells by an inductive activity on human amniotic membrane matrix

Abstract

Here we report a human-derived material with potent inductive activity that selectively converts ES cells into neural tissues. Both mouse and human ES cells efficiently differentiate into neural precursors when cultured on the matrix components of the human amniotic membrane in serum-free medium [amniotic membrane matrix-based ES cell differentiation (AMED)]. AMED-induced neural tissues have regional characteristics (brainstem) similar to those induced by coculture with mouse PA6 stromal cells [a common method called stromal cell-derived inducing activity (SDIA) culture]. Like the SDIA culture, the AMED system is applicable to the in vitro generation of various CNS tissues, including dopaminergic neurons, motor neurons, and retinal pigment epithelium. In contrast to the SDIA method, which uses animal cells, the AMED culture uses a noncellular inductive material derived from an easily available human tissue; therefore, AMED should provide a more suitable and versatile system for generating a variety of neural tissues for clinical applications.

Over the past several years, much progress has been made in the in vitro control of neural differentiation of ES cells. Neural conversion of mouse ES (mES) cells can be induced in vitro by several methods: retinoic acid (RA) treatment of embryoid bodies (EB) (1, 2), multistep-induction/selection culture (3), serum-free adherent monoculture (4), serum-free suspension culture (5), and feeder cell-dependent induction culture (6–9). Each method has its own advantages and disadvantages, depending on the type of neural cells desired. The different methods induce the differentiation of neural tissues with distinct characteristics, particularly with regard to their regional identities in the CNS (2, 5, 8).

We previously reported a feeder cell-dependent induction method that uses a neuralizing activity located on the surface of PA6 stromal cells [stromal cell-derived inducing activity (SDIA)] (6). SDIA induces neural differentiation from mES cells quickly (<5 days) and efficiently (>90%). In response to exogenous patterning signals such as Sonic hedgehog (Shh), bone morphogenetic protein 4 (BMP4), and RA (8, 10), SDIA-induced neural precursors differentiate into a wide range of neural cells of the CNS that correlate with their positions along the dorsal-ventral and rostral-caudal axes (6, 8). SDIA is also applicable to the generation of medically useful neurons such as dopaminergic neurons and motor neurons not only from mES cells but also from primate ES cells (human and nonhuman) (7, 11, 12). In particular, SDIA-treated ES cells efficiently differentiate into midbrain dopaminergic neurons (≈30% of induced neurons) without the addition of exogenous factors such as Shh (6). This method is in contrast to other methods (e.g., the multiple induction/selection method) (3), which require additional treatment with several inducing factors or gene transfer to efficiently generate dopaminergic neurons. Furthermore, when grafted into the striatum of the 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP)-treated Parkinson’s disease model monkey (13), SDIA-induced dopaminergic neurons (from primate ES cells) cause marked improvement in motor function and robustly increase local uptake of the dopamine precursor (Dopa), indicating that the SDIA-induced neurons are functional in the in vivo context.

Because of its simple, speedy procedure and good reproducibility, the SDIA method has become one of the standard methods used in human ES (hES) cell-based therapeutic research for neurological diseases such as Parkinson’s disease (11, 12, 14–17). However, despite its advantages, the SDIA method has a fundamental practical disadvantage for clinical applications of stem cell therapy: the use of xenogenic (mouse) stromal cells as the source of the inductive signals (6). The involvement of xenogenic cells (and any materials derived from them) presumably increases the risk associated with cocultured tissues in transplantation therapy (18). The hES cell-derived neurons induced with PA6 cells might be contaminated with pathogens or unfavorable antigens originating from the cocultured animal cells. For instance, it has been reported that hES cells cultured on mouse feeder cells express an immunogenic non-human sialic acid on their cell surface (19).

In this study, as an alternative inductive material for neural differentiation of ES cells, we introduce the use of the matrix layers of the human amniotic membrane (hAM). The hAM matrix layers possess intriguing biological activities such as the promotion of wound healing and cell growth (20, 21). Here we show that the hAM matrix layers possess an SDIA-like potent neural-inducing activity. Because the matrix of the hAM is a material widely used in surgical practice (20, 22), it serves as a unique, safe source of neural-inducing factors of human origin, which could circumvent problems associated with the use of xenogenic materials.

Results

Efficient Neural Differentiation of ES Cells on the Denuded hAM.

The hAM is composed of an epithelial layer and two matrix layers (the basement membrane and the thick avascular stromal matrix) (23), which underlie the epithelium (illustrated in Fig. 6, which is published as supporting information on the PNAS web site). The matrix layers of hAM (referred to as “denuded hAM” hereafter) have been shown to possess growth-supporting activity on the stem/progenitor cells of various tissues. For instance, corneal limbal progenitors efficiently grow on denuded hAM and eventually form a well organized epithelial tissue, which is suitable for transplantation (21, 22).

We first tested whether the growth of mES cells was supported on the denuded hAM. A detailed protocol is described in the Supporting Material and Methods, which is published as supporting information on the PNAS web site, and illustrated in Fig. 1A and B and Fig. 6. Dissociated mES cells formed large colonies when cultured for a week on the gelatin-coated denuded hAM in serum-free differentiation medium [knockout serum replacement (KSR)-based or chemically defined medium; see Materials and Methods] without leukemia inhibitory factor (Fig. 1C, arrow; the porous filter is seen as the background with halos in this picture). These colonies contained a large proportion of Nestin⁺ and NCAM⁺ neural cells (>90%; Fig. 1 D–F; day 6). We next examined the generation of neural precursors by using an ES cell line in which GFP cDNA was knocked-in at the locus of the early neural gene Sox1 [Sox1-GFP; line 46C; a gift from A. Smith (University of Edinburgh, Edinburgh)] (24). Strong Sox1-GFP expression first appeared in the ES cells between days 3 and 4 (data not shown), and a substantial cell population became Sox1-GFP⁺ on day 5 and after (≈90% of the cells on day 5; Fig. 1G). Most of the Sox1-GFP⁺ cells (>90%) coexpressed the neural precursor markers Musashi (Fig. 1H) and Nestin (data not shown), indicating a preferential generation of neural precursors from ES cells cultured on hAM. In contrast, treatment with BMP4 (0.5 nM, days 0–5), which is a strong inhibitor of early neural differentiation (6), suppressed the Sox1-GFP expression in ES cells cultured on hAM (Fig. 1I; instead, the cells expressed the nonneural epithelial marker E-cadherin; see also Fig. 1 L and M). RT-PCR analysis showed that ES cells cultured on the denuded hAM did not express the mesodermal marker Brachyury or the endodermal markers AFP and Sox17 (Fig. 1J, lane 3) in contrast to EB treated with serum (lane 2).

Fig. 1.

Efficient induction of neural differentiation from mES cells on the denuded hAM. (A) Schematic view of AMED culture. (B) Photograph of a culture insert with denuded hAM and O-ring (black). (C) Phase-contrast view of a mES cell colony (arrow) cultured on hAM for 1 week. Arrowheads indicate neurites extending from the colony. (D–F) Immunostaining of the AMED-treated cells (day 6) with Nestin and NCAM antibodies (D and E) and nuclear staining with DAPI (F). (G and H) Immunostaining of the AMED-treated cells with neural precursor markers. Sox1-GFP signal (green in G and H), Musashi (red in H, day 7), and DAPI (blue in G) are shown. (I) Immunostaining analysis of ES cells treated with AMED and BMP4. Sox1-GFP signal (green) was not detected, and most of the cells expressed the nonneural epithelial marker E-cadherin (red), showing that BMP4 inhibited neural differentiation in AMED culture. (J) RT-PCR analysis of mesodermal (Brachyury) and endodermal (AFP and Sox17) marker genes. (K) Immunostaining of ES cells on hAM for Oct3/4 and Sox1-GFP expression (day 2). Most of the cells expressed Oct3/4 (marker for the undifferentiated state) but not the Sox1-GFP neural precursor marker. (L–N) FACS analysis showing efficient neural differentiation (Sox1-GFP expression on day 6) of mES cells cultured on hAM pretreated with deoxycholate (DOC; N). (L) Negative control for Sox1-GFP expression, ES cells treated with AMED and BMP4 (see also panel I). (M) Positive control, AMED-treated ES cells.

The preferential appearance of neural cells in the culture was unlikely to be caused by the selective adhesion of contaminating neural precursors (ES cell-derived) to the hAM for the following reasons. First, the ES cells used in this study expressed Oct3/4 and Nanog (markers of the undifferentiated state; >95%) in the maintenance culture, whereas no Sox1 expression was observed (data not shown). Second, even 1 or 2 days after plating, ES cells growing on the denuded hAM frequently expressed Oct3/4 (>95%) but not Sox1 (Fig. 1K), indicating that the attached cells were undifferentiated.

The Sox1-GFP⁺ cells on the hAM never expressed the human-specific nuclear antigen (25) (n > 200 colonies; Fig. 7 A–C, which is published as supporting information on the PNAS web site), indicating that these cells were not produced by fusion between the mES cells and contaminating human cells on the denuded hAM. Importantly, both the growth-supporting (data not shown) and neural-inducing (Fig. 1 M and N; shown by the Sox1-GFP⁺ populations in FACS analyses) activities of the denuded hAM remained unaffected even after treatment with 0.5% deoxycholate (12 h at 37°C), which thoroughly removed the cellular components from the hAM (see Supporting Materials and Methods). This finding indicates that the activities are present in the detergent-resistant extracellular matrices.

Taken together, these observations demonstrate that the hAM matrix provides a potent neural-inducing environment for cocultured ES cells. In the present study, “ES cell differentiation on the denuded hAM” is referred to as amniotic membrane matrix-based ES cell differentiation (AMED) hereafter.

On the hAM matrix, as described above, ES cells grew efficiently and selectively differentiated into Sox1⁺ neural precursors (at a 90% or higher frequency) in serum-free medium. We next examined whether such highly selective, efficient generation of Sox1⁺ precursors was seen when other matrix materials were used as a substratum. ES cells were cultured on culture dishes coated with gelatin, collagen IV, laminin, or fibronectin or on dishes with a collagen I gel. In all these cases, the cells grew less robustly than on the hAM matrix, and the efficiency of neural differentiation was substantially lower (60–70%; Fig. 7F) even though the same serum-free differentiation medium was used, demonstrating that the hAM matrix has particularly strong supporting activities for the cell growth and neural differentiation of ES cells.

Regional Characterization of the AMED-Induced Neural Tissues.

To characterize the nature of the AMED-induced neural tissues, we next performed RT-PCR analyses with regional gene markers. AMED-treated ES cells expressed the forebrain-midbrain markers Otx2, TH (tyrosine hydroxylase), Pax2, and En2 and the rostral hindbrain marker Gbx2 at substantial levels (Fig. 2A, lane 3 and data not shown). In contrast, little expression was detected for the spinal cord markers Hoxb4, Hoxb9, and HB9 (Fig. 2A, lane 3).

Fig. 2.

Regional characterization of AMED-induced neural tissues from mES cells. RT-PCR analysis with the rostral-caudal (A and B) and dorsal-ventral (C) marker genes for CNS tissues. (A and C) Lane 1, whole embryo (embryonic day 10.5); lane 2, SDIA-treated ES cells (day 9); lane 3, AMED-treated ES cells (day 9); lane 4, undifferentiated ES cells. (B) Lane 1, whole embryo (embryonic day 10.5); lane 2, AMED-treated ES cells (day 9); lane 3, AMED- and RA-treated ES cells (day 9).

We recently established another in vitro system for neural differentiation of ES cells, namely the SFEB method (serum-free floating culture of EB-like aggregates) (5), which, like AMED, does not use feeder cells as an inducer. SFEB-treated ES cells efficiently generate the rostral-most CNS tissues, particularly Bf1⁺ telencephalic tissues (15–35%), whereas brainstem tissue differentiation (e.g., dopaminergic neurons) is rare (5). Quantitative analysis using immunostaining indicated that the rostral-most CNS marker Bf1 was rarely expressed in AMED-induced neural cells (<1%; data not shown). Thus, the AMED-induced neural tissues show regional characteristics of the rostral-caudal axis, which are similar to those found in the SDIA culture (mainly brainstem regions) rather than in the SFEB culture (mainly the rostral-most CNS).

The rostral-caudal specification of AMED-induced neural cells could be modified by adding the caudalizing factor RA (Fig. 2B). Treatment with RA (0.2 μM, days 4–9) promoted the expression of caudal CNS markers (Gbx2, Hoxb4, Hoxb9, and HB9), whereas the forebrain marker Otx2 was suppressed.

Analyses with dorsal-ventral marker genes showed that the AMED treatment induced both dorsal (Pax7 and Dbx1) and ventral (Irx3 and HNF3β) neural tube markers (Fig. 2C, lane 3). Together with the rostral-caudal marker analysis, these findings show that the regional characteristics of the AMED-induced neural tissues are largely similar to those of the SDIA-induced ones (refs. 5 and 8; lane 2 of Fig. 2 A and C).

AMED-Treated ES Cells Generate a Variety of Neurons Including Dopaminergic Neurons.

Immunostaining showed that all of the colonies on day 9 contained a large number of cells that were positive for the postmitotic neuronal marker TuJ1 (62% of total cells; Fig. 3A). On day 13, the dopaminergic neuron marker TH was expressed in 39% of the postmitotic neurons (26% of total cells) (Fig. 3B). These TH⁺ neurons were negative for the noradrenergic marker dopamine β-hydroxylase (data not shown). AMED-treated ES cells contained other neuron types as well, such as GABAergic (glutamic acid decarboxylase-positive; 22% of the neurons; Fig. 3C) and serotonergic (5HT⁺; 1–3% of the neurons; typically rostral hindbrain; Fig. 3D) neurons.

Fig. 3.

AMED-treated mES cells generate various types of neural cells including dopaminergic neurons. (A) Immunostaining of AMED-treated mES cells with TuJ1 (red) antibody. DAPI, green. (B–D) Immunostaining with neurotransmitter-type markers (red) and TuJ1 (green) antibodies (day 13). (B) TH for dopamine; (C) glutamic acid decarboxylase for GABA; (D) 5HT for serotonin. (E) Expression of Islet1 in mES cells treated with AMED and Shh (30 nM, days 4–9). (F and G) Expression of HNF3β in mES cells treated with AMED alone (F) and with AMED and Shh (300 nM; G). (H and I) Math1 immunostaining of AMED-treated mES cells with (I) or without (H) BMP4 treatment (0.5 nM, days 6–10).

Neural precursors induced in the AMED culture could respond to embryologically relevant patterning signals. Treatment with the ventralizing factor Shh (30 nM, days 4–9) efficiently induced the expression of the motor neuron marker Islet1 (32% of the neurons; ventral CNS marker; 5% without Shh treatment) in AMED-cultured ES cells (Fig. 3E and data not shown). In addition, the differentiation of HNF3β⁺ NCAM⁺ cells (floor plate; the ventral-most CNS tissue) was significantly enhanced by Shh treatment at a high dose (3% and 49% of total cells in the absence and presence of 300 nM Shh during days 4–9, respectively; Fig. 3 F and G). Conversely, treatment with the dorsalizing factor BMP4 (0.5 nM, days 6–10) (8) induced the dorsal CNS marker Math1 (4% of NCAM⁺ neural cells with BMP4, Fig. 3I; <0.1% without BMP4, Fig. 3H) but not Islet1 or HNF3β (data not shown).

These observations demonstrate that AMED-induced neural precursors can generate a variety of CNS tissues in vitro.

AMED Induces the Differentiation of Neural and Eye Tissues also from hES Cells.

Previous studies have shown that the SDIA method is also applicable to neural differentiation of hES cells (11, 12). Therefore, we next examined whether the AMED treatment promoted neural differentiation also in hES cells. Because hES cells generally do not survive and grow well after complete dissociation, we seeded hES cells in small aggregates (clumps of 5–20 cells) onto the denuded hAM (illustrated in Fig. 8, which is published as supporting information on the PNAS web site). In addition, to enhance cellular attachment to the membrane, the hAM was coated with laminin (see the Supporting Materials and Methods). Under these conditions, hES cells reproducibly grew on the hAM and expressed markers for the undifferentiated state (Oct3/4 and Nanog; Fig. 4 A and B) on day 2. On day 15 of AMED culture, hES cells differentiated into Nestin⁺ neural precursors at a high frequency (>85% of the cells; Fig. 4D). A majority of these Nestin⁺ neural precursors formed rosette-like clusters. Oct3/4 expression was substantially down-regulated by day 15 and detected only in a small percentage of cells (<5%; they are Nestin⁻) (Fig. 4E; the Oct3/4-positive population disappeared later by day 25). Pax6 expression first appeared on day 15 and was observed in a subpopulation (≈50%) of cells on day 33 (Fig. 4F). Expression of the regional marker Otx2 (forebrain and midbrain) was also seen (≈30% of total cells on day 33), whereas the Otx2⁺ cells were generally negative for Pax6 (Fig. 4F). Otx2⁺/Pax6⁻ is consistent with the marker expression profile of the midbrain and/or the ventral forebrain, at least in rodent embryos. The telencephalic regional marker Bf1 was not detected in hES cell-derived neural progenitors, at least on and before day 44 (data not shown).

Fig. 4.

Neural differentiation of hES cells treated with AMED. (A–C) Expression Nanog (red in A; Nanog levels were variable as seen in undifferentiated hES cells) and Oct3/4 (green in B) in the majority of AMED-treated hES cells on day 2. DAPI (C) for nuclear staining. (D) Expression of Nestin (green) in AMED-treated hES cells on day 15. DAPI (blue) for nuclear staining. (E) Expression of Nestin (green) and Oct3/4 (red) on day 15. (F) Mutually exclusive expression of Otx2 (red) and Pax6 (green) in AMED-treated hES cells on day 33. (G) Expression of TH (red) in a substantial portion of the TuJ1⁺ neurons (green) induced from hES cells by AMED (day 40). (H) Expression of glutamic acid decarboxylase (red) and TuJ1 (green) on day 42. (I) Expression of Islet1 (red) and TuJ1 (green) in AMED/Shh-treated hES cells on day 42.

The postmitotic neuronal marker TuJ1 was rarely found on or before day 30 and substantially increased during days 35–40 (Fig. 4G and data not shown). On days 40–42, a high percentage of AMED-induced neurons expressed TH (31% of the TuJ1⁺ neurons; 40% of the total cells were positive for TuJ1; Fig. 4G). Expression of the GABAergic neuronal marker glutamic acid decarboxylase was also observed, although less frequently (≈5% of the TuJ1⁺ neurons; Fig. 4H) than TH. The serotonergic neuronal marker (5HT) was detected in <1% neurons (day 44; data not shown). As seen with mES cells (Fig. 3E), Shh treatment (300 nM, days 15–42) induced differentiation of Islet1⁺ neurons in hES cells cultured on the hAM (19% of the TuJ1⁺ neurons with Shh and <1% without Shh) (Fig. 4I).

Our previous studies have shown that SDIA-treated primate ES cells differentiate not only into neural cells but also into eye tissues such as retinal pigment epithelium (7, 26; the retinal pigment epithelium is a CNS tissue derived from the diencephalon during embryogenesis) and lens cells (27). In AMED culture, a number of colonies containing pigmented cells appeared in each well on day 28 (15–40 colonies per well; 200 hES cell clumps were initially seeded per well; Fig. 5A). Importantly, these cells were positive for Pax6 and showed actin bundles (Fig. 5B; phalloidin staining), consistent with the characteristics of pigment epithelial cells. The pigmented cells could be manually isolated with a pipette tip and grown on a collagen I-coated dish. They exhibited a typical retinal pigment epithelial cell-like morphology (hexagonal cells with a cobblestone-like appearance) (26) under light microscopy (Fig. 5C). On day 50 and later, small masses of light-reflecting lentoid tissues were also occasionally found in the culture (Fig. 5D). These tissues were αA-crystallin-positive, consistent with the nature of lens cells (Fig. 5 E and F).

Fig. 5.

Differentiation of pigment epithelia and lentoid tissues from AMED-treated hES cells. (A) On day 28 and after, pigmented colonies were occasionally found in the culture of hES cells on hAM. (B) Phalloidin (red) and anti-Pax6 (green) staining. (C) A high-magnification picture of hES cell-derived retinal pigment epithelial cells (pigmented, polygonal) induced in the AMED culture. The pigmented cells were manually isolated from the hAM by a pipette tip and cultured on a culture dish. (D) A low-magnification picture of an hES cell-derived lentoid tissue. (E) Whole-mount immunostaining with anti-αA-crystallin antibody. (F) A high-magnification confocal picture of αA-crystallin (green) and DAPI (red) staining.

Discussion

Matrix-Associated Neural-Inducing Factors.

This study has shown the presence of matrix-associated activities for inducing selective differentiation of neural and sensory tissues from mES and hES cells in vitro. Our observations clearly demonstrate that many aspects of the controlled differentiation of AMED-treated ES cells are similar to those of SDIA-treated ones (6–8, 26, 27). Thus, the hAM matrix provides a reasonable human-derived candidate material that can substitute for the mouse-derived feeder layer of PA6 cells as a unique, versatile inducer for neural differentiation in ES cell culture.

At the present, the molecular nature of the inducing activity on the hAM matrix remains to be identified, as does that of the activity on the PA6 cells. A particularly intriguing subject for future study is to determine how several activities common to the AMED and SDIA methods (e.g., growth support, induction of neural precursors, and differentiation of dopaminergic neurons) are related at the molecular level.

Another important issue, from the mechanistic viewpoint, is to learn which cells are responsible for the accumulation of the AMED activity on the hAM matrix. In the light of their anatomical relationship, one obvious candidate is the amniotic epithelium, which overlies the hAM matrix layers. However, in a preliminary study, we found that neither the amniotic epithelial cells nor their pericellular matrix promotes neural differentiation of cocultured ES cells (our unpublished observations), indicating that the amniotic epithelium, at least by itself, is unlikely to explain the production of the AMED activity.

AMED Culture as a Versatile Method for Generating Neural Cells.

The AMED system has several advantages for use in ES cell-based regenerative medicine. (i) The AMED culture is remarkably simple; ES cells are plated on the denuded hAM and cultured in a simple serum-free medium. (ii) The AMED method uses a human-derived material (denuded hAM) for which the biological safety has been demonstrated for decades in the clinical practices of dermatology and ophthalmology (20, 28, 29). (iii) hAM is routinely obtainable from Caesarian sections with proper informed consents. (iv) hAM can be easily stored at −80°C for at least for 6 months without losing its AMED activity (see Supporting Materials and Methods). Moreover, the neural-inducing activity on the denuded hAM is retained even after lyophilization/rehydration (Fig. 7G; lyophilization followed by vacuum packaging and γ-ray irradiation and rehydration in culture medium for 1 hour before use) (30). Therefore, the AMED method could be used to prepare ES cell-derived neurons for cell therapy, regardless of the distance between stem cell laboratories and obstetric clinics.

In future technical improvements of the AMED method, solubilization of the AMED activity from the hAM matrix may be useful because a soluble matrix material could be used to coat culture dishes or even three-dimensional polymer scaffolds (which can support tissue formation from RA-induced neural tissues from hES cells; ref. 31) to further simplify the AMED procedure.

In conclusion, this study demonstrates that the AMED method is potentially useful for producing various neurons for cell therapy from ES cells and that AMED provides a practical solution to avoid the use of xenogenic materials, which are used in the SDIA method.

Materials and Methods

ES Cell Culture.

For the maintenance (6), undifferentiated mES cells (EB5 and 46C) were cultured on gelatin-coated dishes at 37°C under 5% CO₂ in Glasgow-MEM (Invitrogen) supplemented with 1% FCS (JRH Biosciences, Lenexa, KS), 10% KSR (Invitrogen), 2 mM glutamine, 0.1 mM nonessential amino acids, 1 mM pyruvate, 0.1 mM 2-mercaptoethanol, and 2,000 units/ml leukemia inhibitory factor (Invitrogen). PA6 cells were maintained in α-MEM with 10% FCS (HyClone) (32).

The hES cells (KhES-1) were a gift from N. Nakatsuji and H. Suemori (Kyoto University) and were used following the hES cell guidelines of the Japanese government. Undifferentiated hES cells were maintained on a feeder layer of mouse embryonic fibroblasts (Invitrogen; inactivated with 10 μg/ml mitomycin C) in DMEM/F12 (Sigma) supplemented with 20% KSR, 2 mM glutamine, 0.1 mM nonessential amino acids (Invitrogen), 5 ng/ml recombinant human bFGF (Upstate), and 0.1 mM 2-mercaptoethanol under 2% CO₂. For passaging, hES cell colonies were detached and recovered en bloc from the feeder layer by treating them with 0.25% trypsin and 0.1 mg/ml collagenase IV in PBS containing 20% KSR and 1 mM CaCl₂ at 37°C for 7 min, followed by tapping the cultures and flushing them with a pipette. After two volumes of culture medium was added, the detached ES cell clumps were broken into smaller pieces (5–20 cells) by gently pipetting several times. The passages were performed at a 1:4 split ratio. For storage, the ES cell colonies were recovered en bloc (without further dissociation) from a 6-cm culture dish, suspended in 1 ml of ice-cold culture medium supplemented with 2 M DMSO, 1 M acetamide, and 3 M polypropylene glycol, and quickly frozen in a 2-ml cryogenic tube (Becton Dickinson Labware) by directly submerging the tube in liquid N₂. The day on which ES cells were seeded on hAM or PA6 cells to start differentiation was defined as day 0.

Preparation of Denuded hAM.

With proper informed consent following the tenets of the Declaration of Helsinki, human AMs were obtained under aseptic conditions at the time of Caesarean section. The method of removing the amniotic epithelium from the amniotic membrane has been reported (refs. 21 and 22; see also Supporting Materials and Methods). The experiments using the hAM were performed according to the institutional guidelines for human-derived materials.

AMED Culture.

The detailed procedures of the AMED culture with mES and hES cells are illustrated in Figs 6 and 8 and described in the Supporting Materials and Methods. For the differentiation medium in this study, we used the KSR-containing G-MEM medium, which was also used in the previous SDIA experiments (6): G-MEM supplemented with 10% KSR, 2 mM glutamine, 1 mM pyruvate, 0.1 mM nonessential amino acids, 0.1 mM 2-mercaptoethanol, 100 units/ml penicillin, and 100 μg/ml streptomycin. In addition to the KSR-based medium used here, we found that chemically defined medium (which does not contain KSR) (33) was suitable for promoting neural differentiation in the AMED culture. The formulation of chemically defined medium is as follows: Iscove’s modified Dulbecco’s medium/Ham’s medium F12 (1:1 ratio) supplemented with glutamine as l-alanyl-l-glutamine or GlutaMAX-I (2 mM; Invitrogen), 5 mg/ml BSA (fraction V), 1× chemically defined lipid concentrate (Invitrogen), 15 μg/ml human apo-transferrin, 0.45 mM monothioglycerol, 7 μg/ml insulin, and 1 unit/ml leukemia inhibitory factor.

Immunostaining, RT-PCR, and FACS Analyses.

The antibodies used for immunostaining are listed in the Supporting Materials and Methods. Cells were fixed with 4% paraformaldehyde at 4°C for 15 min, and immunostaining was performed as described in refs. 5 and 8 and using secondary antibodies conjugated with FITC, cy3, or cy5. For immunostaining cells in large colonies, confocal microscopy (LSM 510; Zeiss) was used to observe the cells inside the colony with good resolution. The total number of cells was counted by staining nuclei with DAPI. For statistical analyses, 100–200 colonies were examined in each experiment. Experiments were performed at least three times. The values shown in the graphs represent the mean ± SD. RT-PCR was performed with ES cell colonies mechanically detached from the hAM or enzymatically detached from feeder cells as described in refs. 8 and 10. FACS analysis using ES cells with the Sox1-GFP reporter was performed as described in refs. 5 and 24.

Abbreviations

mES

mouse ES

hES

human ES

hAM

human amniotic membrane

KSR

knockout serum replacement

AMED

amniotic membrane matrix-based ES cell differentiation

SDIA

stromal cell-derived inducing activity

TH

tyrosine hydroxylase

Shh

Sonic hedgehog

BMP4

bone morphogenetic protein 4

RA

retinoic acid

EB

embryoid bodies.