Multilineage Differentiation of Rhesus Monkey Embryonic Stem Cells in Three-Dimensional Culture Systems

Authors

  • Silvia S. Chen,

    1. NASA/NIH Center for Three Dimensional Tissue Culture, Laboratory of Cellular and Molecular Biophysics, National Institute of Child Health and Human Development (NICHD), NIH, Bethesda, Maryland, USA
    Search for more papers by this author
  • Roberto P. Revoltella,

    1. NASA/NIH Center for Three Dimensional Tissue Culture, Laboratory of Cellular and Molecular Biophysics, National Institute of Child Health and Human Development (NICHD), NIH, Bethesda, Maryland, USA
    2. CNR-Institute of Biomedical Technologies, Unit of Immunobiology and Cell Differentiation, Pisa, Italy
    Search for more papers by this author
  • Sandra Papini,

    1. CNR-Institute of Biomedical Technologies, Unit of Immunobiology and Cell Differentiation, Pisa, Italy
    Search for more papers by this author
  • Monica Michelini,

    1. CNR-Institute of Biomedical Technologies, Unit of Immunobiology and Cell Differentiation, Pisa, Italy
    Search for more papers by this author
  • Wendy Fitzgerald,

    1. NASA/NIH Center for Three Dimensional Tissue Culture, Laboratory of Cellular and Molecular Biophysics, National Institute of Child Health and Human Development (NICHD), NIH, Bethesda, Maryland, USA
    Search for more papers by this author
  • Joshua Zimmerberg,

    1. NASA/NIH Center for Three Dimensional Tissue Culture, Laboratory of Cellular and Molecular Biophysics, National Institute of Child Health and Human Development (NICHD), NIH, Bethesda, Maryland, USA
    Search for more papers by this author
  • Leonid Margolis Ph.D.

    Corresponding author
    1. NASA/NIH Center for Three Dimensional Tissue Culture, Laboratory of Cellular and Molecular Biophysics, National Institute of Child Health and Human Development (NICHD), NIH, Bethesda, Maryland, USA
    • NASA/NIH Center for Three Dimensional Tissue Culture, Laboratory of Cellular and Molecular Biophysics, NICHD, National Institutes of Health, Bethesda, Maryland 20892, USA. Telephone: 301-594-2476; Fax: 301-480-0857
    Search for more papers by this author

Abstract

In the course of normal embryogenesis, embryonic stem (ES) cells differentiate along different lineages in the context of complex three-dimensional (3D) tissue structures. In order to study this phenomenon in vitro under controlled conditions, 3D culture systems are necessary. Here, we studied in vitro differentiation of rhesus monkey ES cells in 3D collagen matrixes (collagen gels and porous collagen sponges). Differentiation of ES cells in these 3D systems was different from that in monolayers. ES cells differentiated in collagen matrixes into neural, epithelial, and endothelial lineages. The abilities of ES cells to form various structures in two chemically similar but topologically different matrixes were different. In particular, in collagen gels ES cells formed gland-like circular structures, whereas in collagen sponges ES cells were scattered through the matrix or formed aggregates. Soluble factors produced by feeder cells or added to the culture medium facilitated ES cell differentiation into particular lineages. Coculture with fibroblasts in collagen gel facilitated ES cell differentiation into cells of a neural lineage expressing nestin, neural cell adhesion molecule, and class III β-tubulin. In collagen sponges, keratinocytes facilitated ES cell differentiation into cells of an endothelial lineage expressing factor VIII. Exogenous granulocyte-macrophage colony-stimulating factor further enhanced endothelial differentiation. Thus, both soluble factors and the type of extracellular matrix seem to be critical in directing differentiation of ES cells and the formation of tissue-like structures. Three-dimensional culture systems are a valuable tool for studying the mechanisms of these phenomena.

Introduction

Embryonic stem (ES) cells are totipotent cells derived from the inner cell mass of preimplantation mammalian embryos. Long-term culture conditions for maintaining undifferentiated ES cells derived from blastocysts of mice, nonhuman primates, and humans have been established [15]. Without these conditions, ES cells differentiate spontaneously into cells of all three embryonic germ layers [4, 68].

Rhesus monkey ES cells injected into muscles of severe combined immunodeficient (SCID) mice formed teratomas that contained elements from all three embryonic germ layers [2, 9, 10]. For example, in this system, ES cells differentiated into structures of ectodermal origin resembling neural tubes, embryonic ganglia, and brain-like gray matter [9], as well as into endoderm-derived tissues including intestinal and ductal epithelium and pancreas [10].

To understand the mechanisms of ES cell differentiation along different lineages, in vitro culture models have been developed [11]. In most of these models, ES cells were induced to form embryoid bodies (EBs), which were then cocultured with a monolayer of stromal cells [12, 13] or plated on various substrates, in the presence or absence of tissue-specific growth factors [7, 1423]. Alternatively, ES cells were transfected with appropriate genes [8, 2426] to identify molecular mechanisms that control ES cell differentiation. All of these experiments were performed in two-dimensional monolayer cultures in which ES cells were induced to differentiate, forming a mixed population of cells of three embryonic germ lines. However, the multicellular structures, which are typically formed in normal embryos and in SCID mouse teratomas, require cells to be able to migrate in three dimensions and to interact with their microenvironment. To study this phenomenon in vitro under controlled conditions, three-dimensional (3D) culture systems are necessary. Here, we studied in vitro differentiation of rhesus ES cells and the formation of tissue-like structures in 3D collagen matrixes, as well as the role of intercellular interactions in this process. We used type I collagen matrixes in two structurally different forms: as a gel and as a sponge. Nonhuman primate ES cells growing on collagen matrixes were cocultured with human dermal fibroblasts or keratinocytes. We found that in this environment multiple cell-cell interactions and soluble factors led to ES cell differentiation along particular lineages. In particular, in these collagen matrixes, complex tubular or spherical glandular-like structures, similar to those in embryos and teratomas, were formed, and these structures ultimately generated differentiated progeny with characteristics of neural, epithelial, or endothelial lineages. We also found that, in 3D collagen matrixes in the presence of feeder cells or exogenous cytokines, ES cell differentiation could be directed into a particular lineage, accompanied by the formation of tissue-like structures.

Materials and Methods

Culture of Undifferentiated ES Cells

Undifferentiated R366.4 rhesus monkey ES cells were cultured as previously described [4]. ES cells were cocultured with mitomycin-C-treated (0.8 mg/ml for 2 hours) (Sigma; St. Louis, MO; http://www.sigmaaldrich.com) murine embryonic fibroblasts (MEFs; Cell Essentials Inc.; Boston, MA; http://www.cell-essentials.com) in gelatin-coated six-well plates (Nalge Nunc Inc.; Naperville, IL; http://www.nalgenunc.com) to prevent spontaneous differentiation. ES cells were maintained in 80% Knock Out Dulbecco's modified Eagle's medium (DMEM; Invitrogen; Rockville, MD; http://www.invitrogen.com) supplemented with 20% defined fetal bovine serum (d-FBS; HyClone Lab Inc.; Logan, UT; http://www.hyclone.com), 1 mM L-glutamine, 0.1 mM 2-mercaptoethanol, and 1% nonessential amino acids (Invitrogen). ES cells were split by releasing them from the culture plate with 0.8 mg/ml collagenase IV (Invitrogen) and seeded onto a new mitomycin-C-treated MEF feeder layer in a gelatin-coated six-well plate.

Controlled Differentiation of ES Cells

ES Cells Cultured as Monolayers on Chamber Slides

In order to verify that the ES cells were still pluripotent, they were released from the six-well plates with collagenase IV (0.8 mg/ml) and suspended in DMEM supplemented with 10% FBS (Gemini Bioproducts; Woodland, CA). Cells from each well of the six-well plates were dissociated by pipetting and seeded on one two-well plastic chamber slide (Nalge Nunc Inc.). Some MEFs were also released and were present among the ES cells. Three different growth media were used: A) Base medium: DMEM supplemented with 10% FBS containing 1% penicillin/streptomycin (Invitrogen). This was used as a simple base medium that relies on serum to support cell proliferation; B) A5RT.1-conditioned medium: a monolayer of high-grade malignant human keratinocyte HaCaT-ras clone A5RT.1 cells [27], which constitutively produce many growth factors including GM-CSF and G-CSF [28], was cultured with fresh base medium until confluence. Two-day conditioned medium was then centrifuged, filtered (0.22-μm pore size filter), diluted 1:3 with fresh base medium, and used; or C) HPI.1-conditioned medium: human dermal HPI.l fibroblasts, which produce many growth factors including fibroblast growth factor, epidermal growth factor, stem cell factor, GM-CSF, vascular endothelial growth factor, and insulin-like growth factor-1 (S. Papini and R.P. Revoltella, unpublished data), were cultured for 2 days with base medium until semiconfluence. The conditioned medium was then centrifuged, filtered, diluted 1:3 with fresh base medium, and used.

All media were changed every 2 days. The morphologies of the ES cells cultured with different media in the chamber slides were observed with a phase-contrast microscope.

On day 8, chamber slides were rinsed with phosphate-buffered saline (PBS; pH 7.4), fixed in 4% formaldehyde for 30 minutes or in a 1:1 mixture of cold methanol and acetone for 10 minutes as recommended by the antibody manufacturers, washed in PBS, and kept in PBS at 4°C until immunohistochemical analysis. Before immunostaining, cells were permeabilized with 0.1% Triton X-100 in 1× PBS for 10 minutes if required, then blocked in Powerblock (Biogenex; San Ramon, CA; http://www.biogenex.com), 5% FBS, or 1%–5% goat serum supplemented with 0.5% bovine serum albumin (BSA; Sigma) for 10–60 minutes, as recommended by the antibody manufacturers.

3D ES Cultures in Collagen Matrixes

Collagen Gel

HPI.1 cells (0.1 × 106 cells/well) were seeded in some wells of six-well plates as a feeder layer 24 hours before ES cell seeding. Meanwhile, each of the six-well inserts was coated with 1 ml of 2.4-mg/ml rat tail type I collagen (Sigma) as described elsewhere [29] with or without 0.1 × 106 HPI.1 cells embedded [30]. After the collagen solution solidified at 37°C overnight, ES cells were released from the six-well plates, as described above, and seeded on top of the collagen gel (with or without HPI.1 cells embedded). The inserts were then transferred to the six-well plates with or without a 24-hour-old feeder layer at the bottom of the wells (Fig. 1). Culture medium supplemented with 50 μg/ml ascorbate (Sigma) and 1% penicillin/streptomycin (Invitrogen) was added to each of the six-well compartments outside the insert so that ES cells were cultured at the air-fluid interface. Media were changed every 2 days. After 7 and 19 days of culture at 37°C in a 5% CO2 in air atmosphere, the inserts were rinsed with PBS, fixed in 4% formaldehyde in PBS at 4°C for 3 days, washed with PBS, and kept in PBS at 4°C until paraffin embedding and sectioning. Sections (∼5 μm) were stained with hematoxylin and eosin (H&E) or subjected to immunohistochemical analysis.

Figure Figure 1..

Experimental procedure.Undifferentiated ES cells cultured on mitomycin-C-treated MEFs were transferred onto two types of collagen I matrixes: collagen gel and collagen sponge. In some experiments, ES cells growing on collagen matrixes were cocultured with human dermal fibroblasts or keratinocytes in the presence or absence of GM-CSF.

Collagen Sponge

HPI.1 cells (0.3 × 106/well) were seeded as monolayers in some of the six-well plates as a feeder layer 24 hours before ES cell seeding. Sterile Gelfoam sponges (Pharmacia; Kalamazoo, MI; http://www.pnu.com), from purified pork skin gelatin, were moisturized with culture medium (DMEM containing 10% FBS and 1% penicillin/streptomycin) just before use. One sponge was cut into four 20 × 15 × 7-mm pieces, and each piece was transferred to a well of a six-well plate with or without a feeder layer. ES cells, released from another six-well plate with 0.8 mg/ml collagenase IV, were suspended in culture medium and placed on the top of the sponge. On each sponge, ES cells were seeded in six distinct spots (20 μl of the ES cell suspension were applied at each spot). Then, 3 ml of culture medium were pipetted into each well to cover the sponge (Fig. 1). The cultures were then incubated at 37°C in a 5% CO2 in air atmosphere for 24 days. The medium was changed every 2 days. On day 24, collagen sponges were fixed in 4% formaldehyde in PBS at 4°C for 3 days, washed with PBS, and kept in PBS at 4°C until paraffin embedding and sectioning. To achieve a better view of cell migration inside and through the sponge, each piece of the sponge was cut vertically into halves along the long axis before paraffin embedding and sectioning. Sections (∼5 μm) were stained with H&E or subjected to immunohistochemical analysis. In order to investigate the effects of different feeder cells and growth factors on the differentiation of ES cells, we used A5RT.1 (0.3 × 106) cells as an alternative feeder layer. In addition, we studied the effects of recombinant human (rh)GM-CSF (a gift from Dr. F. Bertolero; Upjohn; Milan, Italy) on ES cell differentiation. Exogenous rhGM-CSF (20 ng/ml) was added to each of the cultures. The resultant cultures were compared with the ones without exogeneous rhGM-CSF.

Histology and Immunocytochemical Analysis

Target retrieval was performed on all formaldehyde-fixed paraffin sections by means of 0.01 M citrate buffer (pH 6.0) in a microwave oven. The cells were then incubated with primary antibodies. The primary antibodies and their dilutions used are listed in Table 1. ES cells grown in chamber slides and unstained sections from ES cells grown on collagen matrixes were incubated with Vector M.O.M mouse IgG Blocking Reagent (Vector Laboratories; Burlingame, CA; http://vectorlabs.com) to prevent possible cross-reaction between contaminating MEFs and the primary or secondary antibodies. All primary antibodies were diluted in Tris-buffered saline supplemented with 1% BSA. We used the immunoperoxidase system (LSAB+ system; Dako Corporation; Capinteria, CA; http://www.dako.com), which contained biotinylated anti-rabbit, mouse, and goat Ig as a secondary antibody; streptavidin conjugated to horseradish peroxidase as a link agent; and 3,3′-diaminobenzidine (DAB) as a chromogen. Peroxidase-conjugated rabbit anti-rat IgGs were used as the secondary antibody, and DAB was used as a chromogen for anti-stage-specific embryonic antigen 3 (SSEA-3). After the immunostaining procedures were completed, some sections were lightly counterstained with Mayer hematoxylin (Sigma). Positive-stained cells were counted in nine unconnected fields at 100× magnification, and the percentage of positive staining was calculated on the basis of the total number of cells in each view. In addition, using the BCIP/NBT substrate system (Dako), we detected undifferentiated ES cells that expressed alkaline phosphatase.

Table Table 1.. Primary antibodies used in the study
AntibodySpecificitySpeciesClonalityDilutionSource
Anti-SSEA-3 (MC-631)Undifferentiated ES cellsRatMonoclonal1:5Developmental Hybridoma Studies Bank (DHSB), University of Iowa, Iowa City, IA
Anti-SSEA-4 (MC-813-70)Undifferentiated ES cellsMouseMonoclonal1:10DHSB
Anti-class III β-tubulin (TU-20)Neurons at early development stageMouseMonoclonal1:100Chemicon, Temecula, CA
Anti-NCAMNeural/developing skeletal muscle cellsRabbitPolyclonal1:300Chemicon
Anti-nestinNeuroepithelial/neural stem cellsRabbitPolyclonal1:400Chemicon
Anti-chromogranin ANeuronal/neuroectodermal/neural-endocrine cellsRabbitPolyclonal1:150DAKO
Anti-factor VIIIEndothelial cellsRabbitPolyclonal1:200DAKO
Anti-vimentin (V9)Mesenchymal/endothelial/glial/neural stem cellsMouseMonoclonal1:50Vector Laboratories
Anti-cytokeratins (AE1/AE3)Epithelial cellsMouseMonoclonal1:50DAKO
Anti-p63 (4A4)Epithelial stem cellsMouseMonoclonal1:200Santa Cruz Biotechnology, Santa Cruz, CA
Anti-Ki-67 (MIB-1)Proliferating cellsMouseMonoclonal1:50DAKO

Results

Two-Dimensional ES Cells Cultured in Chamber Slides

To evaluate the effect of 3D cultures, we first studied rhesus monkey ES cell differentiation in conventional culture in chamber slides. ES cells seeded in two-well chamber slides were cultured with three different media: base medium, medium conditioned by A5RT.1 human keratinocytes, and medium conditioned by HPI.1 human dermal fibroblasts. Initially, under all conditions, ES cells remained undifferentiated, as evidenced by their morphologies and staining with anti-SSEA-3 and anti-SSEA-4 antibodies (results not shown) [4, 31, 32], and formed compact colonies. Later (after 36–48 hours in culture), some cells began to differentiate and migrated out from the colonies, divided, and formed clusters.

After 8 days under all three tested culture conditions, these cells were predominantly of four different types (Fig. 2): A) undifferentiated ES; B) polygonal, epithelial-like; C) neuronal-like, and D) spread cells with large lamellae. Two types of colonies were found: A) heterogeneous colonies consisting of compact, undifferentiated ES cells in the middle and differentiated cells at the peripheries (Fig. 2A) and B) homogeneous colonies formed by polygonal epithelial-like cells (Figs. 2B and 4A). The sizes of the epithelial colonies in cultures with A5RT.1-conditioned medium were larger than under other conditions. The numbers of cells in the epithelial-like colonies were 14–25, 28–72, and 14–46 cells per cluster in base medium, A5RT.1-conditioned medium, and HPI.1-conditioned medium, respectively. Neuronal-like cells grew on top of MEF cells (transferred to the chamber slides together with ES cells) or adhered directly to the slide; they were unipolar or bipolar and formed networks or aggregates (Fig. 2C). Cells of the fourth type remained scattered and did not form compact colonies (Fig. 2D).

Figure Figure 2..

ES cells in a chamber slide, phase-contrast microscopy.Cells were cultured with HPI.1-conditioned medium for 8 days. Shown are a compact colony of ES-like cells (A), a colony of epithelial cells (arrow) (B), neuronal-like cells with long processes (C), and a cell with a large lamellar cytoplasm (arrow) (D). Scale bars = 50 μm.

Figure Figure 4..

Epithelial differentiation of ES cells in a chamber slide.ES cells were cultured in A5RT.1-conditioned medium for 8 days. Colonies of cells with epithelial morphology (A, phase contrast microscopy, scale bar = 50 μm) were positive for cytokeratins (B, scale bar = 25 μm).

In order to further characterize the cell types, immunohistochemistry was performed. Cells were stained for nestin, neural cell adhesion molecule (NCAM), class III β-tubulin, and cytokeratins (AE1/AE3). These markers are characteristic of neural precursor cells [33], neural cells [34], immature (neuronal) cells [35, 36], and epithelial cells [37], respectively. Cells at the periphery of ES cell colonies appeared to be morphologically differentiated and expressed nestin (Figs. 3A–3BFigure 3.). Epithelial-like cells did not express nestin, NCAM, or class III β-tubulin, but they were positive for cytokeratins (Fig. 4B). Cells of neuronal morphology strongly expressed NCAM (Figs. 3D–3EFigure 3.) and class III β-tubulin (Fig. 3F). The large cells with spread lamellae were also positive for nestin (Fig. 3C). Thus, immunostaining confirmed the morphological evidence of cell differentiation into neuronal and epithelial lineages.

Figure Figure 3..

ES cells in a chamber slide, immunohistochemical staining.Cells were cultured with HPI.1-conditioned medium for 8 days. Cells with neuroepithelial morphology (A and B) as well as large spread cells (C) expressed nestin (A-C). Cells with long sprouting out processes on the top and at the edge of a colony expressed NCAM (D and E) and class III β-tubulin (F). Processes of neuron-like cells aligned in bundles (E) or formed networks (F). Scale bars = 50 μm.

3D ES Cell Cultures in Collagen Matrixes

Collagen Gel

Embryonic stem cells were seeded on a layer of type I collagen gel and cultured in inserts in six-well plates. After 7 and 19 days of culture, H&E staining of the tangential histological sections revealed that some ES cells remained on the top of the collagen gel, whereas a large proportion of cells had penetrated into the gel matrix. Unlike cells in the chamber slide cultures described above, ES cells in or on collagen gel apparently did not form distinguishable clusters of epithelial cells. Neither did we observe large spread cells under these conditions, as observed in chamber slides. Instead, ES cells cultured on 3D collagen gel formed dense clusters as well as tubular or spherical glandular-like structures, which became clearly evident after 19 days in culture. These cells were SSEA-3 and -4 negative. Some structures were formed by one cell layer (Figs. 5A and 5E), whereas others were formed by multiple cell layers surrounding a lumen (Figs. 5B, 5C, and 5F). We call these structures monolayered and multilayered circular structures.

Figure Figure 5..

ES cells on type I collagen gel, H&E staining.ES cells were cultured for 7 days (A, B, and C) or 19 days (D, E, and F). Shown are control cultures (A and D), cultures with HPI.1 cells as feeder layers (B and E), cultures with HPI.1 cells embedded in a collagen gel (C and F), and cells stained with Ki-67 antibody (D, inset). Scale bars = 50 μm.

Since HPI.1-conditioned medium stimulated neural differentiation of ES cells in chamber slides, we cocultured ES cells on collagen gels with HPI.1 cells seeded in two different ways, as shown in Figure 1: A) HPI.1 cells were embedded in the collagen gel placed on a porous membrane of the insert, and ES cells were subsequently seeded on top of that or B) HPI.1 cells were seeded at the bottom of the culture plate as a feeder layer, and subsequently ES cells were plated on top of a collagen gel placed in the insert so that ES and HPI.1 cells were separated by the porous membrane of the insert. As controls, ES cells seeded on collagen gel in the absence of HPl.1 cells were used.

The circular structures both on top of and inside the collagen gel were observed under all three culture conditions (Fig. 5). Also, under all culture conditions, the numbers of these structures on day 19 were greater than those on day 7. At day 7, these circular structures were observed predominantly close to the top surface of the collagen gel, and only a few dispersed cells or small dense clusters had penetrated into the gel (Figs. 5A–5CFigure 5.). In contrast, on day 19, many clusters of cells and a large number of circular structures were found inside the gel (Figs. 5D–5FFigure 5.). The numbers, sizes, and morphologies of these structures also varied at different time points and culture conditions. At day 7, in controls, the structures were predominantly monolayered (Fig. 5A). In contrast, when ES cells were cultured with HPI.1 cells as feeders or embedded in collagen gel, multilayered structures predominated (Figs. 5B–5CFigure 5.). At day 19, the majority of the circular structures were multilayered. The numbers of these structures in the presence of HPI.1 cells were greater than those in the controls. This was especially noted in ES cells cultured with HPI.1 cells embedded in the collagen gel (Fig. 5F). After 7 days and, particularly, after 19 days of culture, many cells within the gel appeared positively stained by anti-Ki-67 antibody (Fig. 5D, inset) and thus were cycling.

To further characterize the cell types, we stained sections with antibodies against cytokeratins and p63 for the epithelial lineage; with antibodies against nestin, class III β-tubulin (TU-20), and NCAM for the neural lineage; with antibodies against chromogranin A for neuroendocrine cells; and with antibodies against factor VIII for endothelial cells. The results of these analyses are summarized in Table 2.

Table Table 2.. Percentages of stained cells in collagen gel
  1. a

    Percentages of cells stained with various antibodies were estimated in nine unconnected fields at 100×. The numbers were rounded. The typical error was about 10%–15% of the average value, except for low average values.

DaysFeederCells in gelCytokeratinp63NestinNCAMClass IIIβ-tubulinChromogranin AFactor VIII
7NoNo004015050
7NoHPI.1028555122
7HPI.1No003525200
19NoNo102204530200
19NoHPI.125406535452
19HPI.1No205203530300
Cytokeratins and p63

After 7 days, in all culture conditions, almost no cells positive for cytokeratins were found within or outside of mono- or multilayered circular structures. However, at day 19, the frequency of cytokeratin-positive cells was greater. In particular, in the presence of HPI.1 feeder cells, the numbers of cytokeratin-positive cells were approximately two times higher than controls (Table 2). In monolayered structures, the cytoplasm of the cells facing the lumen was more intensively stained than the basal site (Fig. 6A). As for multilayered structures, cytokeratin-positive cells were found only in the stratified structures and were predominantly located around the lumen (Fig. 6B). Again, the part of the cytoplasm that faces the lumen was stained more intensively than the basal part (Fig. 6B). The multilayered structures that were negative for cytokeratins seemed to be less organized, without apparent stratification (Fig. 6A).

Figure Figure 6..

Epithelial differentiation of ES cells on type I collagen gel.ES cells were cultured for 19 days. Controls are shown in panels A-C. Panels A and B show cytokeratin (AE1/AE3) staining. A monolayered circular structure is shown in panel A, and a stratified multilayered circular structure and cell clusters that are cytokeratin positive are shown in panel B. Cells in nonstratified circular multilayered structures are shown in panel A (arrow), and scattered cells that are not stained for cytokeratins are shown in panels A and B. Anti-p63 antibody staining is shown in panel C. The p63-positive cells are located in the basal cell layer of the circular multilayered structures (C) and in aggregates without apparent organization (C, inset). Scale bars = 50 μm.

Cells that were positive for cytokeratins were negative for nestin, class III β-tubulin, and NCAM in serial sections. To further confirm results obtained with anti-cytokeratin antibodies and to expand the range of epithelial markers in these samples, we stained sections for p63 antigen, which is expressed by keratinocyte precursor cells [38]. We found that after 19 days, about 5% of cells expressed p63 in monolayered circular structures in cultures with HPI.1 cells either used as feeders or embedded in collagen gel (Table 2). Rarely, p63-positive cells were also found in multilayered structures. These p63-positive cells were located only in the basal cell layer and never near the lumen (Fig. 6C). In 19-day cultures, p63-positive single cells and small aggregates without apparent organization (Fig. 6C, inset) were also detected.

Nestin

The frequency of nestin-positive cells in control cultures on day 7 was on average about 40% (Fig. 7A) of the total, and their frequency decreased to approximately 20% on day 19 (Fig. 8A; Table 2). Nestin-positive cells were found predominantly within various layers of the multilayered structures, and, consistently, few were found outside of these structures. Nestin-positive cells were also found within multilayered structures in cultures with HPl.1 cells either used as feeders (Figs. 7C and 8E) or embedded in collagen gels (Figs. 7E and 8I). However, in the latter case, the numbers of nestin-positive cells was dramatically greater. On day 7, more than 80% of the cells (Fig. 7E) were nestin positive, and about 40% were nestin positive on day 19 (Fig. 8I; Table 2).

Figure Figure 7..

ES cells on type I collagen gel, immunohistochemical staining.ES cells were cultured for 7 days. Shown are control cultures (A and B), cultures with HPI.1 cells as a feeder layer (C and D), and cultures with HPI.1 cells embedded in collagen gel (E and F). Consecutive sections were stained for nestin (A, C, and E) and NCAM (B, D, and F). HPI.1 cells embedded in the collagen gel but not used as a feeder layer facilitated expression of nestin (E). Scale bars = 50 μm.

Figure Figure 8..

ES cells on type I collagen gel, immunohistochemical staining.ES cells were cultured for 19 days. Shown are control cultures (A-D), cultures with HPI.1 cells as a feeder layer (E-H), and cultures with HPI.1 cells embedded in collagen gel (I-L). Consecutive sections were stained for nestin (A, E, and I), NCAM (B, F, and J), class III β-tubulin (C, G, and K), and chromogranin A (D, H, and L). HPI.1 cells embedded in the collagen gel (I-L) but not used as a feeder layer (E-H) facilitated expression of neural markers, nestin, class III β-tubulin, and, especially, NCAM. Scale bars = 50 μm.

NCAM

About 15% of cells in control samples were observed to be NCAM positive on day 7 (Fig. 7B), and their proportion increased to approximately 40% on day 19 (Fig. 8B). For both 7-day (Fig. 7D) and 19-day (Fig. 8F) cultures, coculturing of ES cells with HPI.1 cells as a feeder layer separated by the insert membrane did not notably affect the proportion or the distribution of NCAM-positive cells. In contrast, when HPI.1 cells were embedded in collagen gel, approximately 50% of ES cells were NCAM positive on day 7 (Fig. 7F) and their numbers were greater on day 19 (Fig. 8J; Table 2). Under all these culture conditions, NCAM-positive cells were confined to multilayered structures. Cells intensively stained for NCAM were located in the peripheral cell layers.

Class III β-tubulin

In control samples on day 7 of culture, all cells were negative for class III β-tubulin. Also, only 1%–2% of cells were observed to be class III β-tubulin positive on day 7 when HPI.1 cells were used as a feeder or embedded in collagen gel (Table 2). In all culture conditions, the proportion of class III β-tubulin-expressing cells was greater (∼30%) on day 19, especially when HPI.1 cells were embedded in collagen gel (Table 2). Similar to results obtained with NCAM staining, cells strongly expressing class III β-tubulin were located mainly in multilayered structures, where they were found in the cell layers away from the lumen (Figs. 8C, 8G, and 8K).

Chromogranin A

Chromogranin A was poorly expressed on day 7 both in control, and in HPI.1 cocultures (Table 2). On day 19, the numbers of chromogranin A-positive cells in control cultures were about 20% (Fig. 8D). When HPI.1 cells were used as a feeder (Fig. 8H) or embedded in the collagen gel (Fig. 8L), the numbers of cells expressing chromogranin A were greater (Table 2). These cells were predominantly located in multilayered structures.

Factor VIII

Factor VIII was rarely expressed either in controls or in HPI.1 cocultures at any time (Table 2).

In order to prove that the positively stained cells described above were differentiated ES cells rather than HPI.1 cells or MEFs that could be released and replated when ES cells were transferred to the collagen matrix, we tested whether these feeder cells expressed nestin, NCAM, class III β-tubulin, chromogranin A, factor VIII, cytokeratins, or p63. Immunostaining performed on HPI.1 and MEF monolayers showed that these cells were not stained with any of the antibodies used.

In conclusion, we demonstrated that ES cells grown in 3D collagen gel proliferated and differentiated into neural and epithelial lineages, forming circular structures resembling those evolving during embryogenesis. HPI.1 cells embedded in the collagen gel but not used as a feeder layer facilitated expression of neural markers, nestin, class III β-tubulin, and especially, NCAM.

Collagen Sponges

To determine whether the organization of the collagen matrix affects ES cell differentiation, ES cells were seeded on top of collagen sponges in six-well plates (Fig. 1). In contrast to the collagen gel, the collagen sponge was highly porous, with large pores that allowed cells to migrate. In some experiments, we plated HPI.1 cells on the bottoms of the wells. As a control, we used ES cells seeded on collagen sponges without feeders.

The H&E staining of tangential histological sections of those sponges revealed that after 24 days of culture, only some ES cells remained on the top of the sponge, whereas others penetrated into the pores and migrated downward (Fig. 9A). Within the sponge, we found many cells that were Ki-67 positive (not shown). The cells within the pores tended to adhere to the collagen fibers and formed aggregates mostly in the upper and central parts of the sponge. No aggregates were found near the bottom of the sponge (Fig. 9B). Besides aggregates, single cells were found scattered throughout the entire sponge. The amounts of these cells gradually decreased from the top to the bottom. Similarly to ES cells cultured on a collagen gel, ES cells in collagen sponges formed both monolayered and multilayered circular structures with a central lumen (Fig. 9, inset). However, the numbers of these structures formed within the sponge were consistently lower than those in collagen gel.

Figure Figure 9..

ES cells cultured on a collagen sponge, H&E staining.ES cells were cultured for 24 days (control). Single cells were scattered throughout the entire sponge (A and B). Cell aggregates and circular structures (A, insets) were located predominantly in the upper and central regions (A), whereas no aggregates were found near the bottom of the sponge (B). Scale bars = 50 μm.

To characterize cells in these cultures, we used a set of antibodies against cytokeratins, nestin, NCAM, chromogranin A, factor VIII, and vimentin (Table 3).

Table Table 3.. Percentages of stained cells in collagen sponges
  1. a

    Percentages of cells stained with various antibodies were estimated in nine unconnected fields at 100×. The numbers were rounded. The typical error was about 10%–15% of the average value, except for low average values.

FeederGM-CSF (ng/ml)CytokeratinsNestinNCAMChromogranin AFactor VIIIVimentin
No00500301540
HPI.1053035202580
A5RT.10510553580
No201304052080
HPI.120153550201065
A5RT.1205202016585
Cytokeratins

In control cultures, there were almost no cytokeratin-positive cells. The presence of HPl.1 cells as a feeder layer resulted in slightly greater numbers of such cells (Table 3). Cells immunoreactive for cytokeratins were very rarely found among scattered single cells within the pores but rather in mono- and stratified multilayered structures.

Nestin

A large proportion of nestin-positive cells (more than 50%) was found in cultures without feeder cells. Nestin was strongly expressed in single cells and in cells at the peripheries of multilayered circular structures. Cells in monolayered structures were always nestin negative. Qualitatively, a similar distribution of nestin-positive cells was observed when ES cells were cocultured with HPI.1. However, the amounts of nestin-positive cells, relative to the controls, were notably lower when they were cocultured with HPI.1 (Table 3).

NCAM

No cells expressed NCAM in control cultures in collagen sponges. Positively stained cells were observed when HPI.1 cells were used as a feeder layer (Table 3).

Chromogranin A

In control cultures, more than 30% of cells were chromogranin A positive and were located in multilayered structures. The numbers of positive cells were slightly lower when HPI.1 cells were used as feeders (Table 3).

Factor VIII

Factor VIII was expressed in the cytoplasms of scattered single cells but not in cells within circular structures (Fig. 10A). HPI.1 cell feeders did not result in markedly greater proportions of factor-VIII-positive cells relative to controls (Table 3).

Figure Figure 10..

ES cells cultured on collagen sponges, immunohistochemical staining.ES cells were cultured for 24 days. Shown are staining for factor VIII (A) and vimentin (B). Scale bars = 50 μm.

Vimentin

In control cultures, about 40% of cells were vimentin positive. The majority of such cells were found among scattered single cells inside the sponge (Fig. 10B), whereas few were located in multilayered structures. The fraction of vimentin-positive cells was dramatically greater (more than 80%) when ES cells were cocultured with HPI.1 cells (Table 3).

Thus, HPI.1 feeder layers resulted in dramatically greater NCAM and vimentin expression and lower nestin expression, whereas their effect on cytokeratins, factor VIII, and chromogranin A expression was less pronounced.

Next, we investigated whether the differentiation of ES cells was modified when A5RT.1 cells were used as a feeder layer, as described above for chamber slides. The A5RT.1 feeder cells resulted in noticeably greater factor VIII and vimentin expression and lower nestin and chromogranin A expression, whereas their effect on cytokeratins and NCAM expression was less pronounced (Table 3).

Both HPI.1 and A5RT.1 cells produce numerous factors, many of which are as yet unknown. Since A5RT.1 and HPI.1 cells are known to constitutively produce GM-CSF and express GM-CSF receptors [28, 39] (S. Papini and R.P. Revoltella, unpublished data), we added GM-CSF exogenously in an attempt to mimic the effect of an HPI.1 or A5RT.1 feeder layer. When GM-CSF (20 ng/ml) was added to the control cultures of ES cells, the proportions of cells expressing NCAM and vimentin were greater, the numbers of cells positive for nestin and chromogranin A was markedly lower, and cytokeratin and factor VIII expression was marginally affected (Table 3). Thus, by comparing the proportions of positively stained cells in each of the different culture conditions, we found that GM-CSF mimics the effect of an HPI.1 feeder layer on ES cell differentiation in collagen sponges, with the exception of chromogranin A expression. Also, similar to the A5RT.1 feeder layer, GM-CSF resulted in greater vimentin expression of differentiated ES cells and did not markedly affect expression of cytokeratins and chromogranin A in collagen sponges. Addition of GM-CSF to ES cells cultured with HPI.1 feeders resulted in greater cytokeratin and NCAM expression, slightly greater nestin expression, lower factor VIII and vimentin expression, and unchanged chromogranin A expression, compared with ES cells grown with HPI.1 alone. In contrast, when GM-CSF was added to ES cultures grown with the A5RT.1 feeder, the expression of factor VIII was further stimulated relative to that in cultures grown with A5RT.1 alone (Table 3). Thus, exogenous GM-CSF, or the keratinocyte cell line producing GM-CSF, facilitated ES cell differentiation into an endothelial lineage.

Discussion

In the course of normal embryogenesis, ES cells differentiate along different lineages and form complex 3D tissue and organ structures. This pattern of differentiation is not mimicked in monolayer cultures. Here, we cultured rhesus monkey ES cells in 3D collagen matrixes and induced ES cell differentiation into various cell types that formed several tissue-like structures similar to those in a normal embryo. Coculturing ES cells with keratinocytes or fibroblasts in 3D collagen matrixes enabled us to skew ES cell differentiation into particular lineages. Since the use of human ES cells imposes certain restrictions, we used rhesus monkey ES cells, which are closer to human ES cells both morphologically and in terms of cell surface markers [2, 4] than are mouse ES cells, which are widely used in in vitro studies [40].

The main types of structure that ES cells formed in 3D collagen matrixes were tubular or spherical, each containing a central lumen. Morphologically similar structures have been described in teratomas formed by monkey ES cells in SCID mice [4]. Microscopically, these structures were circular and resembled sections of neural tubes, epithelial tubes, or glands. The physical properties of collagen matrixes seem to affect the formation of these structures, since, in collagen gel, they were formed more readily than in collagen sponges, probably because of the density and the organization of the collagen fibers. It remains to be elucidated how these multicellular structures were formed. We think that collagen gels retain cell aggregates that are plated together with single cells, and from these aggregates, circular structures are formed in the course of ES cell differentiation. In contrast, pores in the collagen sponge favor cell migration out of aggregates, diminishing their sizes and numbers. This speculation is in agreement with the existence of large migratory zones observed at the peripheries of circular structures in collagen sponges, whereas, in collagen gels, such zones were narrower. Also, many more single scattered cells were found in collagen sponges than in collagen gels. Moreover, in the sponges, the tubular or spherical structures were located in the upper part, whereas single cells preferentially migrated to the bottom of the sponge. Thus, the collagen sponge serves not only as a supporting matrix but also as a filter that allows migration of single cells away from clusters. The high mobility of cells in collagen sponges made them a model for testing the invasive ability of tumor cells [41]. Collagen sponges were also used to select highly mobile keratinocyte precursor cells [42]. In addition to the organization of collagen matrixes in which we cultured ES cells, soluble factors and extracellular matrixes produced in these cultures also may skew ES cell differentiation to specific lineages, mimicking the process of embryogenesis.

The circular structures formed in 3D collagen matrixes were of distinct morphological types: a lumen surrounded by one layer or multiple layers of cells, which could be stratified or not stratified. In nonstratified multilayered structures, differentiated ES cells were cytokeratin negative but expressed NCAM, class III β-tubulin, chromogranin A, and nestin. These markers are typically expressed by cells of neural lineage [33, 34, 4348]. In contrast, cells in stratified multilayered and in monolayered structures predominantly expressed cytokeratins [37] and p63 [38], but not nestin, NCAM, class III β-tubulin, or chromogranin A, and thus were of epithelial lineage.

The levels of expression of different markers varied. For example, in nonstratified multilayered structures resembling neural tubes, the proportions of cells expressing nestin were lower, whereas the expression of class III β-tubulin and chromogranin A were greater with time in culture. Moreover, cells expressing different markers were located differently. Cells expressing class III β-tubulin localized mainly in the peripheries of the multilayered structures. The same pattern of temporal and spatial expression of these markers has been reported for neurogenesis in vivo [34, 35, 4649]. This strongly suggests ex vivo maturation of these structures.

To facilitate differentiation of ES cells in 3D matrixes along the neural lineage, we cocultured them with HPI.1 fibroblasts. And indeed, in ES cells cultured with these fibroblasts, differentiation into the neural lineage was facilitated (as evidenced by significantly more multilayered, nonstratified circular structures formed by cells positive for NCAM, class III β-tubulin, chromogranin A, or nestin), but only if HPI.1 fibroblasts were embedded in the collagen gel. In contrast, when HPI.1 fibroblasts were grown as a monolayer separated from ES cells by a membrane, almost no stimulation of ES cell differentiation was observed. One possible explanation for this difference is that fibroblasts directly participated in the formation of circular structures; however, this is highly unlikely, since they would be distinguishable from ES cells both by gross morphology and by the lack of expression of cytokeratins, nestin, class III β-tubulin, and NCAM. Another possibility is that HPI.1 fibroblasts produce soluble factors for ES cell differentiation more efficiently when they are embedded in collagen gel than when they are cultured as a monolayer. It is more likely, however, that short-range ES cell-fibroblast interactions are necessary to stimulate differentiation along the neural lineage. It might be that fibroblasts produce short-lived soluble factors or that they secrete an extracellular matrix or stimulate ES cell differentiation through direct cell-cell contacts.

The importance of cell-cell interactions for ES cell differentiation is emphasized by the fact that single cells in collagen sponges differentiated substantially differently from cells in aggregates. The majority of single cells in sponges expressed factor VIII and thus were apparently of the endothelial type, whereas cells in multicellular structures did not express this marker.

Interactions of cells with each other and with collagen matrixes in the course of 3D structure formation were associated with changes in the morphologies of individual cells. For example, in contrast to 3D systems, cells of the neural type in two-dimensional culture were more differentiated; they displayed classical neuronal morphology with long dendrite-like processes forming a network, and they often assembled in bundles. This network was formed on the top of ES cell colonies and MEFs, and thus cell-cell interaction seems to be an important factor here as well. Also, mouse and monkey ES cells cultured on a stromal cell layer derived from bone marrow differentiated into cells of the neuronal type, with dendrite processes [13, 14]. Another example of how the structures are related to cell morphology is the formation of epithelial-like clusters by cells expressing cytokeratins, but not NCAM, nestin, or chromogranin A. In 3D systems, phenotypically similar cells form circular structures. Thus, cells differentiating along a similar lineage acquire different morphologies and assemble in different structures depending on whether they are restricted to two dimensions or are allowed to interact in 3D matrixes.

Soluble factors significantly modulate ES cell differentiation. One such factor is GM-CSF, a pleiotropic cytokine involved in the proliferation, maturation, and functional activity of hematopoietic [50] and nonhematopoietic progenitor cells [5153]. Addition of soluble GM-CSF in the presence of A5RT.1 cells, which have GM-CSF receptors and also produce GM-CSF, resulted in significantly more factor VIII-positive cells. This observation is consistent with GM-CSF promotion of angiogenesis in an in vivo model [51]. In addition, GM-CSF modulates sensitivity of progenitor cells of different lineages to various cytokines [54, 55]. Recently, it has been shown that embryonic cells express various cytokine receptors [7, 56]. Therefore, it is possible that GM-CSF modulates ES cell sensitivity to A5RT.1-produced cytokines, which are important for ES cell differentiation along the endothelial lineage. In addition to GM-CSF, both HPI.1 and A5RT.1 cells produce numerous factors, many of which are not yet known and may affect ES cell differentiation. In the future, it will be necessary to analyze their effect on ES cell differentiation in the context of 3D systems.

In conclusion, we found that ES cells in 3D collagen matrixes differentiated differently from those in monolayers. Moreover, the abilities of ES cells to form various structures and to differentiate along particular lineages in two chemically similar but topologically different matrixes were different. Cells expressing the same differentiation markers acquired different morphological characteristics in different microenvironments. ES cell differentiation and formation of tissue-like structures could be modulated by already differentiated cells. Both soluble factors and the type of extracellular matrix seemed to be critical in directing differentiation of ES cells and the formation of tissue- and organ-like structures in vitro. Three-dimensional culture systems are a valuable tool for directing ES cell differentiation and formation of organs and tissue for transplantation.

Acknowledgements

The work of Silvia S. Chen, Wendy Fitzgerald, Joshua Zimmerberg, and Leonid Margolis was funded by the NIH intramural program and, in part, by the NASA/NIH Center for Three-Dimensional Tissue Culture. The work of Roberto P. Revoltella, Sandra Papini, and Monica Michelini was supported in part by the Italian Ministry of Health (Project Stem 2001 and FIRB: New Medical Engineering) and in part by Farmigea SpA. We thank Dr. Daniela Campani, Department of Oncology, University of Pisa, for advice and assistance in immunocytochemical studies. We also thank Ms. Emily Graze and Mr. Bennet Porter for their assistance in image analysis and cell counting.

Ancillary