During murine development, the formation of tight junctions and acquisition of polarity are associated with allocation of the blastomeres on the outer surface of the embryo to the trophoblast lineage, whereas the absence of polarization directs cells to the inner cell mass. Here, we report the results of ultrastructural analyses that suggest a similar link between polarization and cell fate in human embryos. In contrast, the five human embryonic stem cell (hESC) lines displayed apical-basal, epithelial-type polarity with electron-dense tight junctions, apical microvilli, and asymmetric distribution of organelles. Consistent with these findings, molecules that are components of tight junctions or play regulatory roles in polarization localized to the apical regions of the hESCs at sites of cell-cell contact. The tight junctions were functional, as shown by the ability of hESC colonies to exclude the pericellular passage of a biotin compound. Depolarization of hESCs produced multilayered aggregates of rapidly proliferating cells that continued to express transcription factors that are required for pluripotency at the same level as control cells. However, during embryoid body formation, depolarized cells differentiated predominantly along mesenchymal lineage and spontaneously produced hematoendothelial precursors more efficiently than control ESC. Our findings have numerous implications with regard to strategies for deriving, propagating, and differentiating hESC.
Disclosure of potential conflicts of interest is found at the end of this article.
The establishment and maintenance of apical-basal (epithelial-type) polarity are crucial to the functional competence of many cell types, prototypically, epithelia  and endothelia . The underlying mechanisms, which are complex, require protein complexes composed of transmembrane and structural elements that regulate the transport of solutes and other molecules . These “fences,” along with the signaling and sorting pathways they regulate, establish the specialized plasma membrane regions of polarized cells: an apical surface (often with microvilli), lateral domains, and a basal area that is attached to the extracellular matrix [4, 5].
Polarization also plays an important role in embryonic development [6, 7]. The first differentiation event in mammalian development is the generation, at the morula stage, of an outer layer of polarized trophectoderm (TE; future placenta) and the nonpolarized inner cell mass (ICM; future embryo) . During compaction, TE polarization is associated with contact asymmetry, assembly of tight junctions, and, consequently, partitioning of the apical from the basolateral region of the plasma membrane. Lineage specification of ICM components is maintained by continued expression of several transcription factors (e.g., Oct 3/4, Nanog, and Sox2) that are downregulated upon TE differentiation. During gastrulation, expression of these factors becomes restricted to cells that are allocated to the germ line. In accord with these findings, pluripotent murine and human embryonic stem cell (hESC) lines also express these transcription factors. Together, these data suggest that pluripotency is associated with the absence of polarization, which is instead evidence of differentiation. This concept is further reinforced by the observation that cultured pluripotent mouse embryonic stem cells are not polarized (data not shown).
Here, we present evidence that, in contrast to the ICM cells from which they are derived, hESC lines exhibit structural and functional characteristics associated with epithelial polarity. To understand the functional consequences of hESC polarization, we depolarized the cells by overlaying hESC colonies with Matrigel. Depolarized hESCs formed multilayered aggregates with morphological similarities to cells of the human ICM that, upon embryoid body formation, preferentially differentiated along the hematoendothelial pathway.
Materials and Methods
hESC Cell Culture
hESCs (H1, H7, H9, Val 1, and Val 2) were grown on either murine (mouse embryo fibroblast [MEF]) or human fibroblast feeders that were plated on gelatin-coated wells of 6-well plates in knockout serum replacement (KSR) medium (80% knockout Dulbecco's modified Eagle's medium [DMEM] and 20% serum-free knockout serum replacement supplemented with 1 mM minimal essential medium nonessential amino acids, 0.1 mM β-mercaptoethanol, and 1 mM l-glutamine) containing 8 ng/ml human basic fibroblast growth factor (Invitrogen, Carlsbad, CA, http://www.invitrogen.com) as previously described . We plated 105 irradiated MEFs from day 13 CF-1 mice (Charles River Laboratories, Wilmington, MA, http://www.criver.com) or human placental fibroblasts per well. Alternatively, hESCs were grown under feeder-free conditions in wells of 6-well plates coated with Matrigel (growth factor reduced; BD Biosciences, San Diego, http://www.bdbiosciences.com) in MEF-conditioned KSR medium as previously described .
Human embryos (n = 4) used in this portion of the study were donated for research purposes under a protocol for human research that was approved by the Spanish National Committee of Assisted Reproduction Technology. Briefly, blastocyst-stage embryos were fixed in 3% glutaraldehyde in 0.1 M phosphate buffer, postfixed in 2% osmium tetroxide, dehydrated through a graded series of ethanol solutions, incubated with a saturated solution of uranyl acetate (30 minutes), and embedded in epoxy resin (Ted Pella Inc., Redding, CA, http://www.tedpella.com). The sections were cut using a diamond knife and examined using a Phillips CM-10 electron microscope.
hESCs were grown as described above. The colonies were scraped off in strips from the wells and washed once in phosphate-buffered saline (PBS) before they were fixed overnight at 4°C in modified Karnovsky's fixative (2% paraformaldehyde, 1% glutaraldehyde, and 1.7 mM CaCl2 in 0.1 M cacodylate buffer [pH 7.4]). Samples were postfixed with 1% osmium tetroxide/0.1 M cacodylate buffer, dehydrated in an ethanol series, and processed using an LR White embedding kit according to the manufacturer's instructions (EMD Chemicals, Darmstadt, Germany, http://www.emdbiosciences.com). Sections were cut with a Reichert-Jung Ultracut E ultramicrotome, mounted on 300-mesh copper grids, and air-dried overnight. After staining with uranyl acetate, sections were examined using a Phillips Tecnai 10 transmission electron microscope.
Human embryos (n = 7), donated under the protocol described above, were fixed in 2% paraformaldehyde for 20 minutes, washed in PBS, permeabilized by incubation for 5 minutes in 0.5% saponin, washed in PBS (three times), incubated with primary antibody for 1 hour, washed in PBS, and incubated for 2 hours in the appropriate species-specific antibody. All incubation and washing steps were performed at room temperature.
hESCs were cultured in 6-well dishes on Matrigel-coated glass coverslips. The cells were fixed in 4% formaldehyde (Sigma-Aldrich, St. Louis, http://www.sigmaaldrich.com) for 20 minutes, washed with PBS, permeabilized by incubation in 0.3% Triton X-100 for 5 minutes, and washed again in PBS. Nonspecific reactivity was blocked by 1-hour incubation in 10% goat serum. Alternatively, before occludin immunolocalization, hESCs were fixed in 1% formaldehyde and permeabilized by incubation for 45 minutes in 0.2% Triton X-100/5% goat serum. Thereafter, the colonies affixed to the coverslips were incubated with the primary antibody for 1 hour at room temperature or overnight at 4°C. The following antibodies were used at the dilutions indicated: fluorescein isothiocyanate (FITC)-conjugated murine anti-ZO-1 (Zymed Laboratories, Carlsbad, CA, http://www.zymed.com) (1:200); rabbit anti-occludin, anti-claudin 1, and anti-claudin 3 (Zymed Laboratories) (1:25); murine anti-E-cadherin (BD Biosciences) (1:100); murine (IgM) anti-Tra 1–60 and anti-Tra 1–81 (Chemicon, Temecula, CA, http://www.chemicon.com) (1:100); rabbit anti-Oct 3/4 (Santa Cruz Biotechnology Inc., Santa Cruz, CA, http://www.scbt.com) (1:200); rabbit anti-Nanog (Dr. Renee Reijo Pera, University of California, San Francisco) (1:100); murine anti-junctional adhesion molecule (JAM)-1 (Cell Sciences, Canton, MA, http://www.cellsciences.com) (1:100); rabbit anti-protein kinase C (PKC) ζ (Santa Cruz Biotechnology) (1:200); murine anti-PKC λ (Becton, Dickinson and Company, Franklin Lakes, NJ, http://www.bd.com) (1:250); and anti-PAR-3 (kind gift from Dr. Tony Pawson, University of Toronto) (1:50). After incubation with the primary antibody, the colonies were washed with PBS and incubated for 45 minutes at room temperature with the appropriate species-specific goat secondary antibody at a dilution of 1:500 (Alexa Fluors; Molecular Probes, Eugene, OR, http://www.probes.invitrogen.com), which contained 1 μg/ml 4,6-diamidino-2-phenylindole (Sigma). After washing, the samples were mounted in Vectashield mounting medium (Vector Laboratories, Burlingame, CA, http://www.vectorlabs.com) and examined under epifluorescence illumination using a Leica microscope and/or by confocal microscopy (Leica TCS NT with Leica software 2.61; Leica, Heerbrugg, Switzerland, http://www.leica.com).
Embryoid bodies were washed in PBS, fixed in 4% paraformaldehyde, washed three times in PBS, passed through graded sucrose solutions (5% and 10%), centrifuged, and the pellet was resuspended in gelatin and polymerized by incubation at 37°C for 20 minutes. The obtained pellet was embedded in optimal cutting temperature compound, snap frozen in liquid nitrogen, and stored at −80°C. Consecutive sections were immunostained according to the protocol described above. The following monoclonal anti-human antibodies were used at 5 μg/ml: vascular endothelial growth factor (VEGF) R2; vascular endothelial (VE)-cadherin (CD144); Nestin; VEGF; CD31 (BD Pharmingen); anti-α-fetoprotein clone 3; anti-β-tubulin isotype III clone SDL.3D10 (Sigma); smooth muscle actin, clone HHF35 (DakoCytomation, Glostrup, Denmark, http://www.dakocytomation.com); TRA 1-81 (Chemicon); CD34, clone 581 (BD Pharmingen); and CD45, clones 2B11 and PD7/26 (DakoCytomation).
Matrigel or Laminin Overlay
hESCs were plated on Matrigel-coated wells or coverslips, allowed to attach overnight, and overlaid with 1% Matrigel (growth factor reduced) in MEF-conditioned medium. Cells were harvested for analysis 24 or 48 hours later. Alternatively, hESCs were overlaid with 1% laminin (from human placenta; Sigma).
Surface Biotinylation to Assess Tight Junction Permeability
hESCs were cultured on Matrigel-coated coverslips (see above). Surface biotinylation was performed as described  with minor modifications. Briefly, 1 mg/ml NHS-LC-Biotin (Pierce, Rockford, IL, http://www.piercenet.com) was added to Hanks' balanced salt solution, which contained 1 mM CaCl2 and 1 mM MgCl2. The hESC colonies were incubated with the freshly prepared biotin solution for 30 minutes, washed with Hanks' balanced salt solution, and fixed for 20 minutes in 4% formaldehyde. After nonspecific reactivity was blocked by incubation in 1% bovine serum albumin/PBS, the coverslips were incubated with FITC-avidin (Pierce) and diluted 1:200 in the same blocking buffer. After 1 hour, the coverslips were washed and mounted in Vectashield mounting medium and examined using a confocal microscope (Leica).
Electrophoretic Separation and Immunoblotting
We used a method that we published previously . Briefly, cell lysates were separated under reducing conditions by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (4% stacking gel and 10% separating gel) and transferred to nitrocellulose membranes. The membrane replicas were incubated in PBS with 0.1% Tween 20 containing 5% nonfat dry milk to prevent nonspecific binding prior to incubation with primary antibodies, which were used at a dilution of 1:1,000. The source of anti-Nanog and anti-Oct 3/4 is given above. Actin expression, used as a loading control, was evaluated using a murine monoclonal antibody (diluted 1:1,000) obtained from Sigma. Immunoreactive bands were detected with horseradish peroxidase-conjugated secondary antibodies and visualized by using chemiluminescence.
Time-Lapse Video Microscopy
hESCs were seeded onto Matrigel-coated wells of a six-well dish in MEF conditioned medium as described above. After 14 hours in a standard tissue culture incubator, the colonies were either overlaid with Matrigel (experimental hESCs) or left untreated (control hESC) and immediately transferred to an environmentally controlled chamber. Then, the chamber was affixed to the stage of a Leica DM IRE2 phase-contrast microscope (Leica) and equipped with a time-lapse video imaging system that was controlled by Improvision Openlab software (Improvision Inc., Lexington, MA, http://www.improvision.com). Images were collected every 10 minutes for 16 hours into a layered file format. Then, layers were compiled into QuickTime movies.
Formation of Embryoid Bodies
Undifferentiated control hESC colonies and colonies that had undergone Matrigel overlay were mechanically dissected, scraped off the Matrigel substrate, and transferred to six-well low attachment plates to allow embryoid body (EB) formation. hESC clumps were washed and replated at high density in serum-containing differentiation medium as previously described  or in serum-free medium (KSR medium without basic fibroblast growth factor: knockout DMEM supplemented with 1% nonessential amino acids, 1 mM l-glutamine, 0.1 mM β-mercaptoethanol, and 20% knockout serum replacement). To account for the faster growth of the depolarized cells, two wells of control colonies were cultured in one well; one well of depolarized cells was split into two wells. EBs were cultured for 3 or 6 days. Medium was replaced every second day. EBs were monitored and photographed with a phase contrast microscope every day and fixed after 3 and 6 days in culture. Experiments were repeated four times, and each experiment was done with H7 and H9 cells, seeded on MEFs or Matrigel. For each condition, we used one full six-well plate. For time course experiments embryoid bodies were harvested from two wells of 6-well plate to get specimens after 3 and 6 days of culture. For cell count embryoid bodies were trypsinized to single cells and three aliquots from each cell suspension were counted.
hESCs Express Epithelial-Like Polarized Phenotype
Initially, we found that hESCs derived and maintained on a variety of substrates (e.g., MEFs, Matrigel, placental fibroblasts) had morphological specializations indicative of polarization. For ultrastructural analysis, we used H1 colonies grown on placental fibroblast feeders that were sectioned in the xz plane in order to observe cell-cell contacts. These cells displayed prominent apical microvilli (Fig. 1A, 1A′) and, at higher magnification (Fig. 1B), tight junctions—discrete sites of apparent fusion among the outer leaflets of the apical-lateral plasma membrane regions of adjacent cells. Together, these data suggested that this hESC line was polarized.
To localize polarized cells in blastocyst-stage human embryos, we used electron microscopy to visualize areas of cell-cell contact. As shown in Figure 1C, the outer TE layer was covered with microvilli, and tight junctions were present near the apical surfaces. In contrast, the ICM lacked these specializations. Figure 1D shows a region in which the plasma membranes of two adjacent cells in the ICM came into close approximation. No areas of electron density were observed in this location or at any other sites of cell-cell contact within the ICM (data not shown).
To determine whether the cells expressed molecules that play a role in polarization, we used fluorescent and confocal microscopy to colocalize Oct 3/4 and ZO-1 in H9 colonies cultured on a Matrigel substrate, with Madin-Darby canine kidney (MDCK) cells as controls. Analysis of the staining pattern visualized by fluorescent microscopy showed that cells with Oct 3/4-positive nuclei were rimmed by linear tracks of ZO-1 immunoreactivity (Fig. 2A). Likewise, E-cadherin expression localized to sites of cell-cell contact, indicating the presence of adherens junctions (Fig. 2B). The exact same staining pattern was observed in MDCK cells (data not shown). Fluorescent microscopy of four other cell lines grown on Matrigel (H1, H7, Val 1, and Val 2) revealed the same Oct 3/4 and ZO-1 staining patterns as the H9 hESC line. In additional confocal microscopy experiments that utilized H1, H7, and H9 cells, we localized, to the region of ZO-1 expression, other components of tight junctions—occludin-1, claudin, and JAM-1 (data not shown). We examined more than 100 colonies in each experiment and performed 3–5 independent experiments for each cell line. We observed uniform expression of polarization markers in all cells within each colony.
In parallel, we localized the expression of a subset of these molecules in blastocyst-stage human embryos. Consistent with the ultrastructural analyses, cells of the TE layer, which lacked Oct 3/4 expression, stained brightly for ZO-1 at sites of cell-cell contact (Fig. 2C). In contrast, cells of the ICM had the opposite staining pattern, with strong nuclear Oct 3/4 expression but largely absent staining for ZO-1 (Fig. 2D).
Matrigel Overlay Induces Depolarization of hESCs
To model an epiblast niche in vivo in which the blastocyst ICM cells are surrounded by basement membrane components produced by the TE or the hypoblast, we coated the apical surfaces of hESC colonies with Matrigel. As a control, MDCK cells were cultured under the same conditions, with no effect in terms of the pattern and intensity of ZO-1 expression as assessed by fluorescent microscopy (Fig. 3; compare panels 3A and 3B) or confocal microscopy in the xz plane (Fig. 3; compare panels 3C and 3D).
In contrast, overlaying H1 (data not shown) or H9 colonies with the same matrix had dramatic effects on the localization and/or expression of several protein components of tight junctions. Control hESCs plated on Matrigel substrates showed tracks of ZO-1 immunoreactivity that localized to the apical-lateral region of the cells (Fig. 3E, 3G). In comparison, the Oct 3/4-positive nuclei had a more basal orientation (Fig. 3G). Within 24 hours of Matrigel overlay, ZO-1 immunoreactivity, which was disorganized, was primarily associated with the cytoplasmic compartment (Fig. 3F, 3H). The occludin-1 staining pattern was similarly disorganized (Fig. 3; compare panels 3I, 3K and 3J, 3L). Likewise, the E-cadherin staining pattern changed, with strong cytoplasmic expression detected (data not shown). Claudin and JAM-1 immunoreactivity was absent in some areas of the colonies and disorganized in others (data not shown). The morphology of the hESC colonies also changed dramatically in cultures overlaid with Matrigel. As recorded by video microscopy (supplemental online data), within 2 hours of Matrigel overlay the cells moved on top of each other, forming multilayered colonies with star-like morphologies (Fig. 4A, 4B).
To test the functional consequences of changes induced by Matrigel overlay, we assessed the barrier function of tight junctions by using a surface biotinylation technique . Consistent with the fact that both H1 (data not shown) and H9 hESCs are polarized, confocal microscopy showed that control colonies excluded the biotin compound, and immunofluorescence was confined to the outer surfaces (Fig. 4C, 4E). Matrigel overlay triggered rapid depolarization (within 24 hours), as evidenced by visualization of biotin penetration among the cells with FITC-avidin (Fig. 4D, 4F). The same results, in terms of ZO-1 and Oct 3/4 expression, were obtained in two separate experiments in which H7 and H9 hESCs were overlaid with purified laminin (not shown).
Next, we assessed the effects of Matrigel “sandwiching” in terms of the partitioning molecule PAR-3, which regulates the development of cell polarity. We also localized the protein kinases with which this molecule associates. The results, as shown by confocal images in the xz plane, showed that hESC depolarization was associated with downregulation of PKC ζ expression (compare Fig. 5A and 5B). PAR-3 and PKC λ, which localized to tight junctions in control hESC colonies (Fig. 5C), displayed diffuse apical immunoreactivity after depolarization by Matrigel overlay (Fig. 5D). Ultrastructural analyses of depolarized (H1) hESCs confirmed that multicellular aggregates lacked tight junctions (Fig. 5F) and microvilli at the apical cell surface (inset, Fig. 5F′). Thus, the depolarized hESCs had a morphological appearance that was similar to that of cells in the ICM of human embryos (Fig. 5G). The expression of transcription factors associated with pluripotency showed very little or no change (relative to actin expression), as demonstrated by immunoblotting with antibodies that specifically reacted with Nanog or Oct 3/4. The same results were obtained in two separate experiments in which H7 and H9 hESCs were overlaid with purified laminin.
Next, we assessed the effects of either Matrigel or laminin overlay in terms of the cells' ability to grow. The same results were obtained under both conditions. With regard to growth, the sandwiched hESC colonies (H1, H7, and H9) contained more than twice the number of cells, evidence of enhanced growth (Fig. 5H). An increase in cell proliferation was also confirmed by 5-bromo-2′-deoxyuridine labeling (data not shown).
Depolarization of hESCs Enhanced the Differentiation into Hematoendothelial Cells
To assess the effect of depolarization on ESC differentiation, we compared several features of EBs produced by H7 or H9 cells cultured on Matrigel under standard control conditions (cEBs) with those of EBs formed from hESCs that were depolarized by Matrigel overlay (dEBs). As described in Materials and Methods, control and depolarized hESC colonies were mechanically dissected, detached, and transferred to six-well low-attachment plates to allow EB formation in serum-supplemented medium, the most widely used culture conditions , or serum-free medium, the most advantageous culture system for studying the initial stages of EB differentiation . To account for the higher growth rates of the depolarized colonies (see above), we plated approximately the same number of cells per clump.
After 24 hours, the number of EBs formed from depolarized hESC colonies in either serum-supplemented or serum-free medium was 2.7 ± 0.6 higher than in control cultures. Additionally, by 36 hours, dEBs grown in serum-supplemented medium were 2.5- and, in serum-free medium, 3- to 5-fold larger than cEBs (p < .0005; Fig. 6A, 6B). The increase in size of dEBs was mainly due to the structural reorganization (see below), as only a slight increase in the proliferation rate was observed (data not shown).
To determine whether depolarization of hESCs impacted cell fate, we first assessed the morphology of experimental and control EBs prepared on days 3 and 6 by staining fixed frozen sections of these structures with hematoxylin. The results showed that dEBs formed vascular-like networks that, by day 2–3, encompassed the entire dEB. In contrast, only such channels were observed in a subset of cEBs by day 3. This initial observation led us to hypothesize that depolarization induced rapid activation of hematoendothelial differentiation.
Accordingly, we immunostained fixed frozen sections of experimental and control EBs prepared on days 3 and 6 for: (a) markers of undifferentiated stem cells (Oct 3/4, Nanog, and TRA 1-81) and (b) markers of the three germ layers (α-fetoprotein [primitive endoderm], βIII tubulin and nestin [ectoderm], and smooth muscle actin and vimentin [mesoderm]). As expected, Oct 3/4 or Nanog expression by cells of cEBs and dEBs was undetectable after 3 days of culture (data not shown). Surprisingly, strong patches of TRA 1-81 immunoreactivity were detected in association with day-3 dEBs, whereas only a few cells in cEBs stained at the same time points (Fig. 6C, 6D). Endoderm and ectoderm markers, which were a feature of cEBs that were cultured up to 6 days, were weakly expressed by cells of dEBs on day 3 (Fig. 6E–6H, supplemental online Table 1) and undetectable by day 6 (supplemental online Table 1). In contrast, most of the cells in the dEBs stained brightly for vimentin at day 3, whereas only a small fraction of cells in cEBs stained in a patchy manner (Fig. 7A, 7B). Together, these findings suggested that depolarization specifically triggered differentiation of mesodermal lineages.
As differentiation progressed, the number of vimentin-positive dEB cells decreased significantly by day 6 (supplemental online Table 1). Given the fact that vessel-like structures formed during this period, we immunostained dEBs cultured for 3 days in the absence of serum for markers that are expressed by endothelial/hematopoietic precursors: CD34, CD45, VE-cadherin (CD144), platelet endothelial cell adhesion molecule (PECAM-1; CD31), and VEGF-R2 (Flk-1, kinase insert domain-containing receptor [KDR]). On day 3, the vimentin-positive cells coexpressed CD34, PECAM-1, VEGF-R2, and VE-cadherin (Fig. 7), but no CD45 immunoreactivity was detected (data not shown). In contrast, cells of cEBs, which displayed faint VEGF-R2 staining, did not express CD34, PECAM-1, VE-cadherin (Fig. 7), or CD45 (data not shown). By day 6, dEB expression of vimentin, CD34, and PECAM-1 decreased, whereas most of the cells retained strong anti-VE-cadherin (supplemental online Table 1) and anti-VEGF-R2 staining (Fig. 7). Interestingly, high VEGF-R2 expression was associated with the dEB vascular-like networks (Fig. 7L). In addition, CD45 positive cells were detected in day 6 dEBs but not cEBs (Fig. 7N vs. 7M). Supplemental online Table 1 summarizes the immunolocalization data presented in Figures 6 and 7.
Finally, we asked whether depolarization of hESCs by Matrigel overlay induced differentiation into hematoendothelial-like cells, that is, whether EB formation was required to redirect the fate of these cells. Therefore, we immunostained tissue sections prepared from undifferentiated control or depolarized hESC colonies with the aforementioned antibodies that recognize markers of endothelial/hematopoietic precursors. In the majority of cases, no staining was detected, but VEGF-R2 was expressed by cells of both groups. Our findings are in accord with the observation that VEGF-R2 mRNA expression, which is reported for hESCs, is maintained during EB formation [14, 15].
Our results show that all the hESC lines we studied exhibit apical-basal polarity. Specifically, we found that the cells have subcellular specializations that are a fundamental characteristic of epithelia, including tight junctions and the asymmetric distribution of organelles and cytoskeletal elements. Our data are in partial disagreement with a study published by Sathananthan et al. , which showed that hESCs in the center of the colonies were not polarized. All of the hESCs in the five lines we examined expressed polarization markers throughout each colony (Fig. 2).
How hESCs become polarized and maintain this state is not clear. They express, in the appropriate patterns, molecules that play important roles in the polarization of other cells: ZO-1, E-cadherin, occludin, PKC ζ, PAR-3, and PKC λ. Currently, the relative contributions of these and other proteins are under intense investigation in a variety of experimental systems. Recent findings suggest that, during Drosophila species development, the asymmetric distribution of PAR-3 to the apical plasma membrane region is required for the initial generation of epithelial cells, and adherens junctions are dispensable during this process. But later in development, these junctions are required for further organization of the epithelia . In worms, epidermal and intestinal cells, which are highly polarized, exhibit apical localization of PAR-3, PAR-6, and PKC-3, the homologue of mammalian atypical PKCs [18, –20]. Thus, the mechanisms that govern the asymmetric expression of these molecules are essential for regulating epithelial polarity. In this regard, protein-protein interactions appear to be the key. For example, in MDCK cells, a complex that contains PAR-3, PAR-6, aPKCs, and cdc-42 is required for maintaining apical-basal polarity . Interestingly, during epithelial development, there are situations in which PAR-3 localizes to the apical cell surface in the absence of other family members, suggesting possible interactions with molecules such as JAM [22, 23]. Whether similar or different mechanisms establish and/or maintain hESC polarization remains to be determined. However, it is clear that the hESC lines we studied constitute an important model system for understanding how apical-basal polarity is established and maintained during human embryonic development.
Our findings also suggest an important role for laminin in regulating hESC polarity [24, 25]. This is in keeping with a great deal of evidence that α1 and α5 laminins contribute to the generation of epithelial polarity, one explanation for why hESCs are able to grow under feeder-free conditions when they are plated on Matrigel, a laminin-rich substrate, or on pure laminin [9, 26]. This multifunctional molecule has different domains that coordinate its molecular self-assembly with cell adhesion and the generation of intracellular signals. In this regard, it is possible that hESC interactions with laminin and/or other extracellular matrix components might trigger their polarization early in the derivation process, with β1 integrins playing a major role. In vivo, blastocyst-stage human embryos contain two complete basement membranes. One separates the trophectoderm, allocated to the extraembryonic lineage, from the ICM, and the other lies within the ICM between the hypoblast and the epiblast. Although the role these structures play in regulating early developmental decisions is not understood, the results of our Matrigel and laminin overlay experiments suggest that a possible effect is delivering signals that prevent polarization in some regions of the ICM.
The results of our experiments show that hESCs are polarized and suggest that depolarization may result in their dedifferentiation to ICM-like cells. The latter theory is supported by a recent report that downregulation of PAR-3 and aPKC function directs cells of preimplantation mouse embryos to the ICM . Judging by the lack of polarized cells in the epiblast of the ICM, there appears to be no hESC equivalent in the ICM of the blastocyst-stage human embryo, although the possibility that this population emerges at the time of implantation cannot be excluded. In support of the latter conjecture, the cells of the epiblast layer of late-stage pig blastocysts become connected by tight junctions, which are maintained when they are placed in culture . Alternatively, components of the human ICM could dedifferentiate, acquiring properties of the polarized cells that form the outer surface of the morula. This hypothesis is consistent with the fact that hESC lines have been derived from disaggregated human morula-stage embryos  and that BMP4 exposure induces hESCs to assume a subset of trophoblast-like properties, most notably hormone secretion .
We think that the switch from the nonpolarized phenotype of ICM cells to the polarized phenotype of ICM-derived hESCs is induced by the artificial niche in which the cells are forced to grow in vitro. The in vivo niche, which maintains ICM cells in an undifferentiated state for approximately 24–48 hours during blastocyst formation, changes immediately after blastocyst implantation. One possibility is that the presence of the zona and the trophectoderm shelters the ICM from various factors that promote differentiation. By necessity, successful derivation of hESCs requires an artificial niche that maintains pluripotency. We theorize that the acquisition of apical-basal polarity is the cells' response to an imperfect artificial niche. Furthermore, polarization as compared with the absence of this specialized state of cell-cell contact is associated with a more complex distribution of signal transduction pathways and, therefore, with more strict control of cell proliferation and differentiation. We note that it appears to take much longer to trigger differentiation of hESCs as compared with their in vivo counterparts. Our results suggest that depolarized hESCs differentiate more rapidly than polarized cells. The practical implications of these findings include the possible importance of hESC polarization in maintaining pluripotency by making them less sensitive to differentiation cues emanating from the matrix, feeder cells, and/or medium. In this context, hESC depolarization may accelerate, stimulate, or channel their differentiation down particular cell lineages.
We were surprised by the structural and antigenic changes that depolarization of hESCs brought about in the context of EB formation. EBs formed from depolarized embryonic stem cells proliferate rapidly and differentiate in a pattern that may replicate aspects of yolk sac development. After 3 days in the absence of serum and growth factors, cells in dEBs spontaneously generated vimentin-positive cells. At a morphological level, the dEBs were composed of vascular-like channels. At a molecular level, the component cells stained for markers of hematoendothelial cells, the common precursors of both hematopoietic and endothelial cells: VE-cadherin, CD31 (PECAM-1), CD34, and VEGF-R2 (Flk-1, KDR). By the end of the 6-day culture period, CD31 (PECAM-1) and CD34 immunoreactivity declined, and a CD45+ population, indicative of blood cell formation, emerged. We note that these morphological and antigenic transitions occurred spontaneously, that is, in the absence of specialized medium. Thus, depolarization of hESCs induced rapid, spontaneous (within first 3 days) commitment to the hematoendothelial lineage.
This is in contrast to recently published work demonstrating that the commitment of hESCs to hematoendothelial lineages during EB formation requires the addition of complex growth factor cocktails [31, 32] or coculture with stromal cells . Additionally, the kinetics of the aforementioned differentiation processes changed with depolarization. Under standard conditions, this process, assayed by fluorescence-activated cell sorting and real-time polymerase chain reaction of CD31 and CD34 expression, required 12–15 days of EB development . Our work shows that if hESCs are depolarized prior to EB formation, this process occurs within 3 days. Finally, we note that published work suggests that, under optimum conditions of growth factor stimulation, ∼10% of cells from EBs express markers of hematopoietic precursors, whereas a majority of dEB components stain for these antigens.
Our findings have several implications. For example, methods for deriving and propagating hESCs in an undifferentiated state should take into account the possibility that triggering polarization is a key to optimizing the culture conditions for both purposes. On the other hand, depolarization might be one way to change or redirect the differentiation capacity of the cells, and protocols for channeling hESCs toward making mesodermal derivatives may be enhanced by depolarization as the first step.
Disclosure of Potential Conflicts of Interest
The authors indicate no potential conflicts of interest.
This work was supported by grants from the UC Discovery Program and Geron Corporation (bio02-10302), The Sandler Family Foundation, The California Breast Cancer Research Program (8KB-0100), The Stem Cell Bank, Prince Felipe Research Center (CIPF), and the IVI Foundation of Valencia, Valencia, Spain. We thank Mathew Gormley and Kristy Red-Horse for assistance with video imaging.