Cell-surface antigens provide invaluable tools for the identification of cells and for the analysis of cell differentiation. In particular, stage-specific embryonic antigens that are developmentally regulated during early embryogenesis are widely used as markers to monitor the differentiation of both mouse and human embryonic stem (ES) cells and their malignant counterparts, embryonic carcinoma (EC) cells. However, there are notable differences in the expression patterns of some such markers between human and mouse ES/EC cells, and hitherto it has been unclear whether this indicates significant differences between human and mouse embryos, or whether ES/EC cells correspond to distinct cell types within the early embryos of each species. We now show that human ES cells are characterized by the expression of the cell-surface antigens, SSEA3, SSEA4, TRA-1-60, and TRA-1-81, and by the lack of SSEA1, and that inner cell mass cells of the human blastocyst express a similar antigen profile, in contrast to the corresponding cells of the mouse embryo.
Permanent lines of pluripotent embryonic stem (ES) cells have now been established in culture from explanted human blastocysts [1,, 2]. They are able to differentiate into a wide range of cell types in vitro and form teratomas in immunosuppressed mice. They also share some features in common with murine embryos and ES cells, for example, high-level expression of alkaline phosphatase and the stem cell transcription factor, Oct4. Nevertheless, these human ES (hES) cells also exhibit marked differences from their murine counterparts, particularly with respect to their expression of several stage-specific embryonic antigens (SSEA). Such differences from the mouse have also been noted before in the case of human embryonal carcinoma (EC) cells, the stem cells of teratocarcinomas and the presumed malignant counterparts of embryo-derived ES cells [3–, 6].
Murine ES cells closely resemble cells of the late inner cell mass (ICM) of the blastocyst in their cell-surface antigen phenotype: SSEA1, a carbohydrate antigen, also known as Lex, is a fucosylated derivative of type 2 polylactosamine and appears during late cleavage stages of mouse embryos [7,, 8]. It is strongly expressed by murine ICM and ES cells, as well as cells of the trophectoderm; it is absent from the primitive endoderm. Two other glycolipid antigens with globoseries carbohydrate core structures, SSEA3 and SSEA4, are expressed by unfertilized eggs and early cleavage embryos, but disappear by the blastocyst stage and are not expressed by cells of the ICM; these antigens are expressed by the primitive endoderm [9,, 10]. Likewise, murine ES cells also do not express either SSEA3 or SSEA4. In culture, the differentiation of murine EC and ES cells is typically characterized by the loss of SSEA1 expression and may be accompanied, in some instances, by the appearance of SSEA3 and SSEA4 [11,, 12].
By contrast, human EC cells typically express SSEA3 and SSEA4 but not SSEA1, while their differentiation is characterized by the downregulation of SSEA3 and SSEA4 and upregulation of SSEA1 [13,, 14]. The initial reports of hES cell lines have indicated that they too express SSEA3 and SSEA4 [1,, 2], as well as the keratan sulphate-associated antigens, TRA-1-60 and TRA-1-81, that are also characteristic of human EC cells [15,, 16]. So far, however, no direct comparison has been made between the surface-antigen phenotype of hES or EC cell lines and the embryonic cells to which they are thought to correspond within the human blastocyst. Thus, hitherto, it has been unclear whether these marked differences in surface-antigen expression indicate significant differences between human and mouse embryos, or whether ES/EC cells correspond to distinct cell types within the early embryos of each species.
We have now shown that hES cells in culture and the ICM cells from human blastocysts share expression of SSEA3, SSEA4, TRA-1-60, and TRA-1-81 and do not express SSEA1. These observations suggest that ES and EC cells do resemble ICM cells in humans as well as in the mouse, and that the differences between human and mouse ES/EC cells reflect differences between human and mouse embryos.
Materials and Methods
The hES cell lines, H7 and H14 , were cultured in Knockout-Dulbecco's modified Eagle's medium (GIBCO-BRL; Grand Island, NY; http://www.invitrogen.com) supplemented by 20% “serum replacement, SR” (GIBCO-BRL) on mouse embryo fibroblast feeder cells that had been inactivated by treatment with mitomycin C under a humidified atmosphere of 5% CO2 in air. For subcultivation, the cells were harvested by treatment with 1 mg/ml collagenase IV and dispersed by scraping, to maintain the cells in small clumps. Differentiation was induced by incorporating all-trans-retinoic acid (Eastman-Kodak; Rochester, NY; http://www.kodak.com) (10−5 M) into the medium as described for human EC cells .
Cell-surface-antigen expression of cultured cells was assessed by indirect immunofluorescence detected by flow cytofluorimetry after harvesting the cultures as single-cell suspensions using trypsin-EDTA, as described previously [14,, 17]. The following primary monoclonal antibodies were used to detect surface-antigen expression: MC631 (anti-SSEA3) , MC813-70 (anti-SSEA4) , MC480 (anti-SSEA1) , TRA-1-60 , TRA-1-81 , TRA-2-54 (anti-liver/kidney/bone alkaline phosphatase) , TRA-1-85 (anti-Ok(a)) , and anti-Thy1 . Fluorescein isothiocyanate (FITC)-labeled goat anti-mouse IgM or anti-IgG was used as the secondary antibody, as appropriate to the isotype of the primary antibody. In some experiments, cells that were either positive or negative for SSEA3 and SSEA1 were isolated by fluorescence-activated cell sorting .
Surface-Antigen Expression by Preimplantation Embryos
Under license from the Human Fertilisation and Embryology Authority (HFEA) and with local ethical approval, frozen-thawed human embryos (n = 52), surplus to clinical needs, were obtained with fully informed consent from patients undergoing assisted conception treatment. Mouse embryos were recovered from superovulated and naturally mated Swiss mice. All embryos were fixed in 4% paraformaldehyde in phosphate-buffered saline (PBS) for 1 hour and washed twice in PBS. Blastocysts were permeabilized with 0.005% triton X-100 for 15 minutes and were washed twice in PBS. Incubations with primary and secondary (anti-mouse FITC-conjugated IgG or IgM) antibodies were for 1 hour, and embryos were mounted in mowiol/Dabco antifade solution before examination by epifluorescent microscopy.
RT-PCR Analysis of Gene Expression
Total RNA was extracted from ES cells, sorted for antigen expression using Tri Reagent (Sigma; Poole, Dorset, UK; http://www.sigmaaldrich.com), and treated with DNase (DNA-Free; Ambion; Austin, TX; http://www.ambion.com). Subsequently, no genomic contamination could be detected by polymerase chain reaction (PCR) (data not shown). One microgram of total RNA was used for each reverse transcription (RT)-PCR reaction as previously described . Equal aliquots of RT product were then subjected to PCR using the primers and conditions summarized in Table 1. In all cases, 35 PCR cycles were used, except for Oct4 (28 cycles), Rex-1 (30 cycles), and β-actin (20 cycles). The PCR primers for alpha-fetoprotein (AFP) were those reported by Schuldiner et al. ; those for NeuroD1 (ND-1) and β-actin were reported by Duran et al. . The remaining primers were designed by using the PRIMERSELECT package from the DNASTAR suite of programs, and the specificity of the RT-PCR products was confirmed by sequencing.
Table Table 1.. PCR primers used to detect gene expression in hES cells
Anneal temp °C
Product size (bp)
The putative human Rex1 sequence was obtained by searching for ESTs homologous to mouse Rex1 (ZFP-42, NM_009556) using the BLAST protocol . One EST (AW665472) mapped to the 3′ end of an open reading frame (ORF) predicted by Fgenesh analysis (CGG Web Server at http://genomic.sanger.ac.uk) on chromosome 4q35.2 (a region syntenic to mouse Rex1 on chromosome 8). The protein translation of this ORF shows very high homology (unpublished data) with the mouse Rex-1 protein . The complete ORF was isolated by PCR from hES clone H7 cDNA and sequenced, in triplicate. The human Rex1 sequence was submitted to Genbank (AF450454).
To confirm the initial reports of surface-antigen expression patterns of hES cells [1,, 2], we examined the two hES cell lines, H7 and H14. As anticipated, many cells in these cultures strongly expressed all four of the antigens, SSEA3, SSEA4, TRA-1-60, and TRA-1-81. They also expressed the liver/bone/kidney isozyme of alkaline phosphatase, detected by antibody TRA-2-54 , and human Thy1  (Fig. 1). Murine EC and ES cells also strongly express alkaline phosphatase , but they do not express Thy1 , although human EC cells are Thy1+ . This expression pattern was repeatedly observed in several assays carried out on cultures maintained by serial passaging over 6 months. When H14 hES cells were cultured in the presence of retinoic acid, marked downregulation of these antigens was seen, consistent with differentiation that is also seen with some human EC cell lines following exposure to retinoic acid (Fig. 1B) [13,, 14]. On the other hand, SSEA1 was expressed on relatively few cells in the H7 and H14 hES cultures and was slightly upregulated after exposure to retinoic acid.
Less than 5% of the cells in the hES stock cultures were mouse fibroblast feeders at the time of assay, as confirmed by reactivity of more than 95% of the cells with TRA-1-85 (data not shown), an antibody that detects the pan-human antigen Ok(a), which is absent from mouse cells . Since none of the other antigens marked all of the hES cells, a proportion of the cultures could consist of spontaneously differentiated derivatives. To test this, we used fluorescence-activated cell sorting to separate H7 hES cells according to their expression of SSEA3 and SSEA1. When analyzed by RT-PCR (Fig. 2), little difference was observed in the levels of Oct4 and Sox2 expression between the SSEA3+ (putatively undifferentiated) cells and the SSEA3− (putatively differentiated) cells. However, Rex1 was expressed more strongly by the SSEA3+ cells and, even more markedly, fibroblast growth factor-4 (FGF4) was almost exclusively expressed by the SSEA3+ subset. By contrast, markers of differentiated cells, AFP, hemoglobin ζ (HBZ), human chorionic gonadotropin (HCG), ND-1, and Sox1, were all most strongly expressed by the SSEA3− subset. These results are consistent with an undifferentiated phenotype for the SSEA3+ cells, whereas they suggest that the SSEA3− cells included variously differentiated derivatives.
Oct4, Sox2, FGF4, and Rex1 have all been reported to be characteristically, though not necessarily exclusively, expressed by murine EC and ES cells and to be generally downregulated upon their differentiation [28–, 30]. Although Oct4, in particular, is restricted to undifferentiated mouse ES and EC cells, its expression appears to be less tightly regulated in primates  and only a slow loss of Oct4 expression has been reported during the differentiation of hES cells . Among markers of the differentiated cells, AFP and HBZ indicate extraembryonic endoderm differentiation and embryonic hematopoietic differentiation [33,, 34], HCG suggests trophectodermal differentiation , and ND-1 and Sox1 suggest neural differentiation [36,, 37]. Indeed, morphologically identifiable neurons were, from time to time, observable in hES stock cultures.
After sorting for SSEA1 expression, Oct4 was similarly detected in the SSEA1+ and SSEA1− subsets. However, Rex1, HCG, and Sox1 were more strongly expressed in the SSEA1+ subsets. The expression of HCG suggests that the SSEA1+ cells include the trophectodermal derivatives, a result consistent with earlier studies of the expression of SSEA1 by choriocarcinoma cell lines  and its expression by trophectoderm of mouse  and human (see below) blastocysts. The significant expression of Rex1 by the SSEA1+ cells is also consistent with trophoblastic differentiation, as previous reports have indicated that Rex1 is expressed by trophectoderm in the mouse . SSEA1 was not significantly discriminatory with respect to the other genes investigated (data not shown).
Together, these results support the notion that hES cells exhibit a cell-surface-antigen phenotype distinct from murine ES cells and that their differentiation into extraembryonic and somatic derivatives is associated with a significant change in surface-antigen expression. To determine whether the surface-antigen phenotype of hES cells does, indeed, correspond to that of the ICM of human blastocysts, early human embryos (n = 52) at cleavage and blastocyst stages, produced after in vitro fertilization treatment of infertile couples, were fixed in 4% paraformaldehyde in PBS and stained by indirect immunofluorescence for the expression of SSEA1, SSEA3, SSEA4, TRA-1-60, and TRA-1-81 (Table 2, Fig. 3). TRA-1-60, TRA-1-81, SSEA3, and SSEA4 were all localized to the ICM in human expanded or hatching blastocysts, although they were absent from mouse embryos at these stages. Conversely, SSEA1 was expressed on the trophectoderm of human blastocysts but not at all on the ICM, though it was expressed on the murine ICM.
Table Table 2.. Cell surface antigen expression by preimplantation stages of human and mouse embryos
Embryos and embryonic stem cells were stained by indirect immunofluorescence for the expression of SSEA1, SSEA3, SSEA4, TRA-1-60, and TRA-1-81. At least three human and 15 mouse embryos were used for each stage of development and for each antibody. Whereas TRA-1-60, TRA-1-81, SSEA3, and SSEA4 were localized mainly to the ICM in human expanded or hatching blastocysts, the same markers were absent from mouse embryos at the same stage but present at earlier cleavage stages. Conversely, SSEA1 was expressed on the trophectoderm of human blastocyst and not at all on ICM while being expressed on murine ICM.
*Absent from cell surface but staining on inner surface of zona pellucida and within perivitelline space.
Cells of preimplantation embryos
At earlier cleavage stages, SSEA3 and SSEA4 were not expressed by human embryos, although they are commonly expressed by corresponding murine embryos [7,, 9,, 10]. However, the cleavage embryos of some mouse strains, notably DBA, also lack expression of SSEA3 and SSEA4, which are evidently associated with maternally derived glycolipids . On the other hand, SSEA1 was expressed by late-cleavage-stage embryos in both species. Curiously, although TRA-1-60 was not expressed by any of the cells of the mouse embryos examined, it was detected in the extracellular zona. The epitope detected by antibody TRA-1-60 is currently not defined, but it is thought to involve a sialylated and sulfated polylactosamine oligosaccharide ; the possibility of cross-reactivity with distinct macromolecules carrying a similar epitope cannot be excluded.
Our results indicate that human ICM and cultured ES cells share a common antigen phenotype that resembles that of human EC cells but is distinct from that of murine ICM, EC, and ES cells. Unfortunately, the limited availability of human embryos precluded a direct study of their expression of TRA-2-54 (alkaline phosphatase) and Thy1. However, given all the other features in common, it seems likely that these antigens too are expressed by human ICM cells.
On balance, the RT-PCR analysis of the expression of various developmentally regulated genes in the SSEA3- and SSEA1-sorted subpopulations of hES cell cultures supports the contention that SSEA3 marks the undifferentiated cells and that SSEA1 is expressed by some of their differentiated derivatives, notably putative trophectoderm. However, our analyses were limited by the semiquantitative character of the RT-PCR analyses possible on small populations of sorted cells, and do not permit a conclusion as to whether or not genes, such as Oct4 and Sox2, which appeared to show little difference between SSEA3+ and SSEA3− subsets, might nevertheless show some downregulation upon differentiation in the SSEA3− cells. Further, in this study, we have not considered the turnover rate of various markers that might permit the persistence of certain transcripts and their corresponding proteins as cells convert from one phenotype to another. Thus, we cannot rule out the presence of “intermediate” transitory cell types that coexpress some features of the undifferentiated stem cells and their differentiated derivatives. This could be particularly the case in the current study in which we have analyzed the subpopulations of cells in asynchronous, spontaneously differentiating cultures of ES cells.
The surface antigens studied in these experiments show marked developmental regulation in both mouse and human embryos. Why, then, the embryos of these two species differ with respect to the expression patterns of these cell surface antigens is unclear. Most of the epitopes studied are associated with oligosaccharides, and one idea is that the core structures of these represent the key functional parts of cell-surface carbohydrates, and that the terminal modifications, which are typically recognized by specific antibodies, are of less functional significance [14,, 40]. In this respect, it is notable that mouse EC and ICM cells also express the Forssman globoseries glycolipid antigen, as well as globoside itself [41,, 42], even though they do not express SSEA3 and SSEA4. Since humans lack the glycosyl transferase necessary for the terminal addition of galactosamine to globoside to yield the Forssman structure, the SSEA3 and SSEA4 antigens in humans might perform the same function as Forssman in the mouse. In this model, the key developmental changes would be the switches between synthesis of lactoseries, globoseries, and ganglioseries oligosaccharide core structures, rather than terminal monosaccharide additions recognized by the antibodies used .
The TRA-1-60 and TRA-1-81 antigens are keratan sulfate proteoglycans, rather than glycolipids [16,, 44]. Again, however, it seems likely that the antigenic epitopes are associated with the oligosaccharide components of these molecules. Their function is also unknown, and except for our current results indicating the presence of cross-reacting antigen in the zona pellucida, there is no evidence of expression of these epitopes associated with cell-surface antigens of mouse embryonic cells. However, the core oligosaccharides of keratan sulfate are polylactosamines, fucosylation of which can produce the SSEA1 epitope. Indeed, high-molecular-weight, SSEA1-reactive glycoproteins have been reported to be expressed by murine EC cells [45–, 47]. It is tempting to consider that the TRA-1-60- and TRA-1-81-reactive proteoglycans of human embryonic cells are members of an equivalent human family of glycoproteins and serve a similar function to the SSEA1-reactive glycoproteins of mouse embryos.
However, the function of these carbohydrate stage-specific embryonic antigens presents a conundrum. There is no doubt that their expression is closely regulated during early development and that changes in expression correlate well with cell differentiation. Further, the SSEA1/Lex epitope has been reported to play a role in cell:cell adhesion between blastomeres at the morula stage in the mouse [48,, 49]. On the other hand, many of these carbohydrates are carried as glycolipids; in one study of embryos of the Medaka fish, it was reported that glycolipid expression can be almost extinguished with no evident effect upon embryonic development .
Nevertheless, whatever their role in normal development, these antigens could play a role in embryo pathology. The SSEA3 and SSEA4 antigens belong to the P-blood-group system, and those small number of individuals who lack the ability to synthesize the extended globoseries structures that form the basis of these antigens (pk and pp individuals) lack expression not only of the P-blood-group antigen but also SSEA3 and SSEA4 on their erythrocytes . It has been reported that women with pp and pk phenotypes display a high rate of spontaneous early abortion, and that abortion in such women is provoked by an immune response to an embryonic antigen . SSEA3 and SSEA4 are clearly obvious targets. Maybe of more significance is the observation that about 1% of Caucasians lack expression of SSEA4 (though not SSEA3 or P antigen), which has been suggested to equate to the Luke blood antigen . This raises the question of whether women with the Luke−/SSEA4− phenotype may also be subject to early abortions due to immune responses to early embryonic cells, like pp and pk women .
The contrast in antigen expression between human EC stem cells derived from teratocarcinomas and murine EC and ES cells is well known. That hES cells express a similar antigen phenotype to human EC cells as well as human ICM cells of the early embryo confirms that pluripotent EC and ES cells can provide complementary tools for investigating the molecular mechanisms that underlie cell differentiation in a way that is pertinent to understanding human development. It also suggests that human teratocarcinomas, one of the most common forms of cancer in young men, do indeed arise because of aberrations in the normal mechanisms that guide cell differentiation during embryogenesis . Whatever the reason for the differences in antigen expression patterns seen in the current studies, the results highlight the potential problems in extrapolating results from mouse to human embryos.
This work was supported in part by a grant from The Wellcome Trust (grant number 016150). Dr. J.K. Henderson was supported by the Infertility Research Trust. We are grateful to Michelle Waknitz, Jessica Antosiewicz, and Christine Pigott for excellent technical assistance. J.K.H. and J.S.D. contributed equally to this work.