Stem Cell Reviews Copyright © 2005 Humana Press Inc. All rights of any nature whatsoever are reserved. ISSN 1550-8943/05/1:111–118/$30.00

Original Article A Molecular Basis for Human Embryonic Stem Cell Pluripotency Scott A. Noggle, Daylon James, and Ali H. Brivanlou* Laboratory of Molecular Vertebrate Embryology,The Rockefeller University, New York, NY 10021-6399, USA

Abstract Embryonic stem cells (ESCs) are able to generate a wide array of differentiated cell fates while maintaining self-renewal. Understanding the biology of these choices may be central to the use of human embryonic stem cells (HESCs), both as a model for early human development as well as a resource for cell based therapies. Efforts to dissect the molecular mechanisms that mediate stem cell identity are underway, and in this review we summarize recent progress in defining the markers and pathways involved in these decisions. We discuss recent efforts to assess the molecular signature of pluripotent HESCs and highlight work demonstrating a set of genes, including representatives from the FGF, TGFβ, and Wnt signaling pathways, that consistently mark the undifferentiated state. In addition, we describe experiments in which signaling of HESCs is augmented by chemical probing with small molecule compounds. Using these compounds, we have demonstrated an important role for Wnt signaling in HESC pluripotency and shown a requirement for TGFβ signaling in the maintenance of the undifferentiated state. These experiments have revealed some molecular aspects of the pluripotent state and demonstrated clear differences between mouse and human ESCs in the maintenance of this identity. Index Entries: Transcriptional profile; FGF; TGFβ; bone morphogenic protein; Wnt; chemical genetics; Xenopus; glycogen synthase kinase-3 (GSK-3); indirubins; 6bromoindirubin-3’-oxine (BIO); leukemia inhibitory factor (LIF); SMAD2/3; SB-431542.

Introduction

*Correspondence and reprint requests to: Ali H. Brivanlou Laboratory of Molecular Vertebrate Embryology, The Rockefeller University, New York, NY 10021–6399, USA. E-mail: [email protected]

Understanding the biology of stem cells may be one of the most important undertakings of this century for science and medicine, for they have the potential to inform our understanding of many aspects of human development as well as disease. Stem cells may also have great potential as a resource for cell-based therapy of degenerative disorders such as Parkinson’s or Huntington’s diseases or diabetes, whether by replacement with cultured cells or by mobilization of a resident tissue stem cell pool. Early success of several therapies using hematapoietic stem cells hint at this potential, however, much remains to be understood if they are to fulfill their promise in other areas of science and in the clinic. From the moment when a fertilized egg divides into two cells and, subsequently, as

111

embryonic development moves forward to generate a group of cells, signaling pathways provide the cues necessary for the proper genesis of independent cell fates within the embryo. The exact integration of these pathways in time and space into networks leads to a hierarchy of events, which ultimately generates a repertoire of tissues and organs. The understanding of these pathways and their integration represents one of the most important goals of modern embryology. Although presently it is still unclear to what degree embryonic stem cells (ESCs) in particular model development in vivo, a fundamental understanding of their biology is a necessary step toward realizing their clinical potential. This review summarizes some of our recent work addressing two basic characteristics of human ESCs (HESCs): their global transcriptional profile, which represents their

112__________________________________________________________________________________________________Noggle et al. molecular signature, and the signaling pathways involved in their maintenance.

Molecular Signature of HESC Pluripotency A defining function of stem cells is their ability to maintain themselves in an undifferentiated state. A stem cell regenerates itself, preserving an array of fate choices identical to that of its parent, so its transcriptome should remain unaltered from generation to generation; and transduction pathways that are enriched in undifferentiated ESCs would be predicted to direct self-renewal, suppressing commitment to any differentiated fate at the same time. In order to define molecular markers of “self-renewal,” we have compiled a list of genes that are expressed consistently and enriched in undifferentiated mouse ESC and HESC lines (1). This list includes markers that are cloned, well defined and unambiguous, as well as those that need more extensive characterization. However, this list is small, and it is becoming clear that there are significant differences between mouse and human ESCs. For example, mouse markers such as SSEA1 and SSEA4 are not equivalently expressed in HESCs. We, among others, have attempted to address this problem by comparing the transcriptome of our HESCs to differentiated cells and other stem cells using global transcriptional profiling and rigorous statistical analysis to find what is unique or enriched in the HESCs. We reported the identification of 918 genes enriched in undifferentiated HESCs compared with their differentiated progeny representing about 9% of expressed transcripts in the undifferentiated HESCs (2). In addition, we identified ligand/ receptor pairs and secreted inhibitors of the fibroblast growth factor (FGF), transforming growth factor-β (TGF-β)/bone morphogenetic protein (BMP), and Wnt signaling pathways. Interestingly, 28% of the enriched genes correspond to expressed sequence tags (ESTs) with no know domain or function. This will undoubtedly be a rich source for the discovery of unique functions in stem cells. This analysis also allowed the comparison of HESC transcriptional profile to the one published in the mouse (3,4). One of these reports has suggested 1703 genes as the signature for stemness in the mouse (4). Surprisingly, after normalizing the two data sets we find only 24% of the genes to be overlapping between the mouse and human. This is unusual in the sense that global comparison between specialized, highly differentiated cell lines in the mouse and human show a more than 90% overlap. Other studies have also found similarities as well as differences (5,6). The remarkably low overlap in ESCs suggests that studies of MESCs alone, although more pragmatic, may not be sufficient for understanding HESCs. Thus, although it may be uselful to extrapolate from MESC studies, understanding stemness in humans will require the study the human cells in parallel with the mouse lines. A note of caution is warranted in the interpretation of comparisons among various lists of genes enriched in HESCs. There appears to be little agreement among microarray studies regarding individual transcripts enriched in the undifferentiated state of HESCs. For example, if one compares the lists of significantly enriched genes in three recent studies (14), only seven genes are found in the intersection of common genes among these studies. This problem is not unique to HESC microarrays; the relatively young science of microarray-based

research has been susceptible to problems with reproducibility since its foundation (7). Moreover, the reproducibility of experiments across labs and platforms has recently become a topic of scrutiny and discussion. Although it has been suggested that discrepancies in platforms, differences in biology, or laboratory variability are to blame (8–13), it has recently been found that it is possible to derive a much larger intersection among lists in these studies. By starting from the raw data and using the same data-processing and statistical criteria on a smaller set of high-coherence genes that are common to all studies, we find that this list can be more than octupled (14). This is despite differences in platform, cell lines, culture conditions, and undifferentiated control samples used among the studies. As many genes, such as Nanog, TDGF1, and Rex1, which have been suggested to be involved in the state of “stemness” (1,15,16), were not included in the universe of genes common to these studies; this list is not comprehensive. However, it does establish a set of markers of undifferentiated HESCs that is reproducible regardless of cell line or culture differences (14). Further analysis also shows that there are a number of genes that, despite being significantly different within the studies, were not reproducible across studies. This suggests that some genes in the global screens arise owing to differences in platform, cell lines, or culture conditions. The majority of these were metabolic or “housekeeping” genes; however, there were some genes involved in developmental decisions, such as MID1 (17) and FGF9 (18,19). However, that a signature set of genes exists in undifferentiated HESCs is also supported by a recent global transcriptional analysis that, in addition to unique genes, identified shared genes among genetically unique undifferentiated HESC lines (5). A common and reproducible set of genes that demarcate the undifferentiated state of HESCs is therefore emerging. Other problems remain in current global transcriptional profiling assays. For instance, negative results with these assays are at best inconclusive. Microarrays suffer from incomplete coverage of the genome. In the studies highlighted above, the number of genes ranged from 16,652 at the low end to more than 44,000 at the high end. As the current UniGene cluster estimate is more than 52,000 and the human genome is estimated to have 30,000 genes, there is likely redundancy on cDNAbased chips as well. Comparing lists based on different versions of an annotation database will likely result in discrepancies. Additionally, as noted above, 20–40% of the hits in the published lists are not annotated. Techniques such as MPSS, SAGE, or EST profiling, which aim to cover the genome in unbiased screens, may alleviate some of these problems. Even if all of the transcripts of the cell were identified, knowing the transcriptional status does not explicitly imply what signaling pathway is necessary or causal to the functions of the stem cell. The transcriptome represents information on the available pathways as well as the output of the transcriptional response to the pathways activated in the cell. For instance, this transcriptional response in a stem cell, in addition to maintaining the factors needed for self-renewal pathways, may also induce factors needed for competence for another fate, such as those genes for signaling pathways required to direct differentiation down a particular pathway. One possible example is the expression of the Lefty genes in both mouse and human ESCs (20,21). Although these genes are expressed in mouse

Stem Cell Reviews ♦ Volume 1, 2005

Molecular Basis for HESC Pluripotency __________________________________________________________________________113 and human ESCs, their expression in the extraembryonic endoderm of mouse embryos as early as E5.5 indicates other early developmental functions, such as the possible involvement in the regulation of anterior–posterior patterning (22). Therefore, it is important to experimentally distinguish the activity of the necessary self-renewal pathways from the transcriptional output of those pathways. Additionally, the question of how many intermediate fates exist between an undifferentiated and a fully differentiated ESC remains to be addressed. Future studies will need to be designed to address the markers expressed in this “transitional” state. These considerations make current transcriptional analyses less than competent to predict pathways causal to the state of stemness. Although they provide rich ground for finding markers of the various stem cell fates, traditional experimental approaches are necessary to understand the causal pathways of pluripotency.

Molecular Dissection of Signaling Pathways Involved in ESCs by Chemical Probing The molecular dissection of the genetics as well as experimental embryology approaches to model systems have, over the past few decades, allowed an appreciation of the complexity of signaling pathways, their tremendous evolutionary conservation, and, not surprisingly, the existence of unique species-specific traits. More recently, a new approach using small molecules that affect development has proven to be a powerful tool toward understanding the output of these pathways, and a complement to the experimental and genetic approaches used in embryogenesis. This approach, sometimes called “chemical genetics,” is based on the study of the biological activity of small compounds in a cellular or organismal context using a set of exogenous ligands, often known as chemical tools (reviewed in ref. 23). This type of approach has been shown to have great advantages when compared to more traditional systems. First, unlike genetic mutations that cause permanent and heritable change in the phenotype, small molecules offer a quantitative and temporal control over gene product modulation. Moreover, although genetic screens are not likely to detect loss of a gene function when more than one copy or a functional isoform of this gene exists, small molecules can disrupt products from all copies of a given gene. Second, this approach eliminates a major limitation in experimental settings in which the assays of the output of pathways are based on injection of DNA, RNA, or proteins: the approach allows for control over the timing of the output. Additionally, whereas soluble proteins can be excluded from their site of action by inability to traverse membranes, small molecules can reach all cellular compartments. By simply adding and washing away the drug, an instantaneous and reversible modulation of the protein function can be achieved. This temporal control can be extremely useful to study development processes in which time plays a crucial role. However, the number of compounds in small-molecule libraries can reach upto millions. Thus, screens must be employed to effectively eliminate the noise of molecules that have no biological activity or that are toxic. Recently, an organismal screen of small compounds, using Xenopus embryos, has been shown to be a powerful approach toward the successful screening of factors with activities in ESCs (Fig. 1). In response

Fig. 1.

to experimental modulation of the eight signaling pathways, shown to be required for early embryogenesis (24), frog embryos elicit phenotypes that can be easily distinguished. For example, ubiquitous activation of the canonical branch of the Wnt pathway generates dorsoanteriorized phenotypes, whereas interference with the sonic hedgehog pathway leads to cyclopia. The decisions governing differentiation toward independent cell types and the maintenance of the pluripotent state in ESCs are under the control of the same eight signaling pathways. Individually purified small natural molecules that elicit a relevant phenotype in the embryos by mimicking those produced by modulation of individual pathways, and induce the expression of cell-type-specific markers in embryonic explants, were tested for their potential influence on ESC maintenance and differentiation. Here we would like to describe our recent work which takes advantage of chemical probing to identify signaling pathways involved in the maintenance of the undifferentiated pluripotent state of HESCs.

Wnt/BIO—A Renewal Pathway Common to Mouse and Human ESCs In canonical Wnt signaling, Wnt ligands convey signals through Frizzled receptors at the cell surface that lead to inhibition of glycogen synthase kinase-3 (GSK-3) (Fig. 2). GSK-3 constitutively marks β-catenin, accumulating in a signaling pool, for degradation by ubiquitination. Inhibition of GSK-3 increases the pool of β-catenin which can travel to the nucleus and, in collaboration with transcription factors (TCFs), activate target gene expression (25,26). Components of the vivid purple dye “Tyrinan purple,” isolated from mollusks in various parts of the world and used by many human societies in classical history and often associated with royalty and aristocracy, include the indirubins. This family of pharmacological agents has recently been shown to be inhibitors of cyclin-dependent kinases (CDKs) and GSK-3 (27,28). A cell-permeable synthetic derivative, 6-bromoindirubin-3′-oxime (BIO), was subsequently shown to be specific for GSK-3. BIO is also an example of a

Stem Cell Reviews ♦ Volume 1, 2005

114__________________________________________________________________________________________________Noggle et al.

Fig. 2.

compound that was characterized originally for its ability to mimic canonical Wnt pathway activation in Xenopus embryos. Subsequently, we found that BIO, but not its kinase-inactive derivative MeBIO, was sufficient in both MESCs and HESCs for the maintenance of the undifferentiated, pluripotent phenotype (29). BIO strongly induced the expression of β-catenin in two HESC lines, again reinforcing the fact that it did influence the canonical Wnt pathway. Examination of endogenous Wnt signaling revealed that the signaling pathway is activated in the undifferentiated MESCs as well as in blastocyst embryos. In MESCs, activity of a reporter driven by the TopFlash promoter module was shown to decrease on withdrawal of leukemia inhibitory factor (LIF) and canonical Wnt pathway activation resulted in increased activity of Rex-1, a MESC pluripotency marker whereas dominant negative TCF-3 abrogated this upregulation. In HESCs, Wnt pathway activation maintained the pluripotency markers Oct3/4 and Nanog, and HESCs cultured in BIO were able to give rise to cellular derivatives representing all three primary germ layers in teratoma assays. In addition to BIO, a number of pharmacological inhibitors of GSK-3 have been recently characterized (30); among them is lithium. However, lithium is only functional at high concentrations and can have toxic side-effects. BIO seems to be the most potent GSK-3 inhibitor available, functioning to inhibit GSK-3 in the low nanomolar range in cell-free kinase assays. In contrast to nonsubstituted indirubins, BIO is over 16-fold more potent for GSK-3 than for the next most susceptible kinase, CDK5, demonstrating good selectivity for GSK-3. BIO is also active in cell culture as it is able to modulate the activity of the

Wnt pathway in the low micromolar range. At these concentrations related CDKs, such as CDK5, are minimally inhibited. However, we also find that BIO-treated HESCs tend to reduce their growth rate and their ability to self-renew after three or four passages under the present conditions. It will be necessary to determine culture conditions that allow us to extend the BIO-mediated self-renewal in HESCs. We believe that developing this culture system is one of the most promising and necessary directions toward therapeutic applications. A fundamental understanding of the mechanisms involved in the maintenance of pluripotency in MESCs and HESCs is only beginning to emerge. Maintenance of self-renewal in HESCs was initially found to be dependent on extrinsic factors that were produced by a feeder layer of mouse embryonic fibroblasts (MEFs). Later studies of the molecular mechanisms that MESCs use to maintain the undifferentiated, pluripotent state showed that signaling through the LIF/gp130 pathway is sufficient for the propagation of undifferentiated MESCs in the presence or absence of MEF feeder layers (31–33). LIF signals through LIFR and gp130 (34,35) to maintain MESC selfrenewal (31–33) by activating Janus kinase/signal transducer and activator of transcription (STAT) pathway (36,37). However, although the activation of the STAT pathway seems to be sufficient for the maintenance of pluripotency in MESCs, a volume of evidence seems to suggest that it is not necessary. For example, loss of function of LIF, gp130, or STAT3 does not affect the maintenance of pluripotency in MESCs. These arguments, taken together, strongly suggest that in MESCs, signals other than STAT3 and LIF could be at play to maintain MESCs in a pluripotent state (38).

Stem Cell Reviews ♦ Volume 1, 2005

Molecular Basis for HESC Pluripotency __________________________________________________________________________115 Regardless of its role for MESC self-renewal and corollaries during in vivo mouse development, LIF cannot sustain the undifferentiated state in HESCs and the STAT pathway has been shown not to be active in undifferentiated HESCs (29). We examined biochemically and functionally the state of activation of the STAT pathway comparatively in MESCs and HESCs, under both pluripotent and differentiating conditions. We demonstrated that although STAT3 signaling is involved in MESC self-renewal, stimulation of this pathway fails to support the undifferentiated state in HESC lines. Neither of the STAT 1, 3, or 5 proteins are phosphorylated under self-renewing conditions in HESCs, whereas STAT3, in agreement with published literature, shows a high level of activation in MESCs grown under the same conditions (29). As STAT3 and STAT5 have shown to have similar output forming their own subgroup in the STAT family, the results suggest that activation of this branch of the pathway is not involved in the maintenance of selfrenewal in HESCs. These observations, in agreement with data published by other groups (39), suggest that the selfrenewal pathway is different in human and mouse or that there is another pathway that can fulfill the same function in both species. Considering the dispensability of LIF/STAT3 signaling for pregastrula mouse embryos and its inability to prevent differentiation of HESCs, there are likely other factors that mediate self-renewal of ESCs. The ability of BIO to maintain self-renewal in MESCs suggests canonical Wnt pathway activation as a candidate for such a factor. However, β-catenin-null mice are defective in anteroposterior axis formation, but not in maintaining pluripotency (40). Given that pluripotency is a fundamental biological function in multicellular organisms, it is certain to be evolutionarily secured by multiple genetic backup systems. Therefore, mouse-null mutants for any one of these systems may be offset by alternative pathways and may only exhibit later developmental phenotypes. Combined gene targeting approaches may be required to uncover the phenotypes that are defective in the maintenance of self-renewal.

The Smad2/3 Branch of TGF-
Signaling/SB-431542 The TGF-β pathway is maintained evolutionarily from worms to humans. In the context of embryogenesis, this pathway is involved in key embryological and cellular decisions which include formation of embryonic germ layers, specification of the body axis, cell migration, and extensive cell fate determination contributing to all organs. From an embryological point of view, stemness can be considered as an embryonic cell fate, and therefore the specification of this unique fate might also be under control of TGF-β signaling. Interestingly, we also know that interruption of ongoing TGF-β signaling has as much of a consequence in determination of cell fate as the instructive induction of the transduction cascade downstream of ligand binding. The formation of the nervous system in the amphibian is a good example of how inhibitory signals can induce and maintain the fate of dorsal ectoderm cells in the earliest stages of neural induction during gastrulation (41). In addition to this bimodal type of specification, another attribute of the TGF-β pathway is that most ligands can act as morphogens, eliciting different cell fates as an

outcome of very small changes in concentration of activity. Because of the conservation of the pathway and its role during not only early development, but also tumorigenesis, this pathway has been the subject of intense scrutiny. Therefore among the core stemness genes in HESCs, expression of components of the TGF-β pathway is particularly intriguing. There are three ligands, GDF3, TGF-β1, and BMP2, expressed in undifferentiated HESCs. There are also a number of membrane-bound as well as secreted inhibitors of the pathway: Cripto, Cerberus, LeftyA, LeftyB, and TMEFF (42), and the expression of several of these is high in undifferentiated HESCs. As suppression of ongoing signals can be as informative for cell fate determination as activating a signal transduction cascade, their expression is also suggestive that TGF-β signaling may be involved in pluripotency. Despite the fact that there are approx 40 independent members of this pathway in the human genome, only a handful of ligands and inhibitors of the pathway are commonly expressed in the undifferentiated state of both MESCs and HESCs. We have shown that Smad2 and/or Smad3 is phosphorylated and enriched in self-renewing H1 HESCs, suggesting activation of the Nodal/Activin branch of the TGF-β pathway, and when the HESCs are pushed to differentiate, Smad2/3 levels gradually decrease as a function of time (20,43). Studying LeftyA levels and utilizing reporter assays showed, in agreement with published literature, that the promoter of LeftyA is under the control of Smad2 and can be activated by Activins in H1 cells. Interestingly, when HESC medium was conditioned by MEFs cultured in the absence of FGF2, which is normally added to the medium used to maintain the undifferentiated state of HESCs, the cells differentiated. This suggests that secreted factors induced by FGF2 in the MEFs are required for the stemness state in H1 cells. Analysis of TGF-β ligands reveals that the expression of inhibins βA and βB, inhibitors of Activin ligands, is decreased when FGF2 is omitted from the medium (Brivanlou and Besser, unpublished data). These data suggest that TGF-β signaling might be involved in the maintenance of the undifferentiated state in HESCs. Recently, the use of another small molecular compound, SB-431542, which elicits a clear Smad2/3 inhibition in Xenopus embryos and embryonic explants, has been instrumental in elucidating a requirement for TGF-β/activin/Nodal input in the maintenance of pluripotency in HESCs (43). SB-431542 is a potent inhibitor of the kinase domain of Alk4/5/7 and precludes phosphorylation of Smad2/3(Fig. 2). Both in the context of MEF conditioned medium (CM) and BIO, HESCs cultured with SB-431542 are unable to maintain the markers of pluripotency, Nanog and Oct3/4. Furthermore, embryoid body formation, which can be used to assay maintenance of the undifferentiated state of HESC, is reduced in the presence of SB-431542, and drastically reduced when the cells are cultured in the presence of BIO and SB-431542. Although canonical Wnt signaling may play a role in the self-renewal of both MESCs and HESCs, the requirement for TGF-β/Activin/Nodal input in the maintenance of self-renewal is only evidenced for HESCs. In fact, whether MESC self-renewal is maintained by LIF or BIO, SB-431542 has no effect in abrogating this maintenance. However, when mouse blastocysts are cultured ex vivo in the presence of SB-431542, Oct3/4, a marker of the pluripotency (44), is significantly reduced in the

Stem Cell Reviews ♦ Volume 1, 2005

116__________________________________________________________________________________________________Noggle et al. stem cell compartment (derived from the inner cell mass) of blastocyst outgrowths. The disparate phenomenology in the character of stem cells depending on species and niche raises many questions. Much of the work done with HESCs uses years of MESC research as a template for its methodology. Indeed, many early efforts to define the character of HESCs extrapolated directly from fundamental aspects of MESC biology. The inability of LIF to maintain HESCs and the lack of any requirement for TGF-β/activin/Nodal input in MESC selfrenewal underscores the need to re-examine these cell types from a different perspective. However, as close as mouse and human may be evolutionarily, the mechanism by which the two species account for and maintain their ESC pool may vary. The disparity in TGF-β/activin/Nodal requirement between ex vivo blastocyst outgrowths and MESC culture further obscures the nature of ESC identity. In vitro culture may augment the properties of ESCs in such a way that they no longer reflect the character of the cells from which they emerged; the establishment of indefinitely self-renewing pools may be a culture artifact in itself. It remains to be seen whether temporal and/or methodological variables will have an effect on the ESC phenotype in both mouse and human. The other branch of TGF-β signaling, the BMP pathway, also has a dynamic influence on the maintenance of self-renewal in HESC, though one that is reciprocal to the TGF-β/activin/Nodal input. In undifferentiated HESCs, Smad1/5 phosphorylation is globally inhibited or reduced and only increases on differentiation. Repression of Smad1/5 phosphorylation may be necessary to prevent differentiation into trophoblast (45), but it is also tempting to speculate that the molecular basis of pluripotent HESC identity may be rooted in an evolutionarily conserved mechanism of primary cell fate specification like that which is evidenced in Xenopus. Indeed, the stabilization of β-catenin, Smad2/3 activation, and Smad1/5 inhibition that occur at the dorsal organizer in Xenopus are all coincident in HESCs, though it remains to be seen whether reduced Smad1/5 phosphorylation in undifferentiated HESCs is maintained by an active or passive process.

Conclusions and Future Directions For ESCs to realize their potential for clinical application or be used as a model for human development, it will first be necessary to address fundamental questions regarding their biology and the molecular nature of stemness. One open question in the field is, “To what degree do ESCs resemble cell types of the tissue from which they were derived?” It is often assumed that ESCs have a counterpart in the embryo; they are most often related to the inner cell mass or the primitive ectoderm due, among other characteristics, to their ease of derivation from these tissues. And recently, early germ cells have been postulated as the actual origin of ESCs based on their gene expression and historical culture conditions (46). It may be dangerous to draw conclusions about the cell of origin based on markers expressed in the derived cell, as derivation of a self-renewing cell line from a non-self-renewing tissue requires adaptation—what you end up with cannot be identical to what you started with. With this in mind, it may be a mistake to assume that ESCs have an in vivo counterpart. There is no reason to assume that all stem cells use the same mechanism to self-renew. Adult stem cells have been defined

based both on form and function within the context of an autopoietic organism (47). A good standard to aim for in the molecular dissection of HESCs is the hematopoietic stem cell, for which there is an ever increasing density of marker resolution and which is a tissue amenable to experimental manipulation and functional assays. These attributes have allowed for the definition of stem cells as distinct from progenitors with different capacities for self-renewal. Conversely, ESCs have been defined more by their general or abstracted functions rather than their form. Although clonal lines of HESCs have been derived from nonclonally derived parent lines, it is possible that heterogeneity existed in these populations, and this has been a common observation in these cultures. For example, populations with different levels of SSEA4 expression are observed (48,49), and Cerberus-expressing cells have also been shown to be sporadically expressed in undifferentiated cultures (50). The assumption of homogeneity in the cultures based on a limited number of markers or that the cultures do not change over time might also be unsafe. For example, we have observed that the expression of FGF4 in early passages of a newly derived HESC line diminishes with passaging and other groups have observed various changes in their populations after prolonged passaging (48,49). It could theoretically be possible, then, to derive unique stem cell lines from any number of early embryonic cell types, all of which have the general property of pluripotency (function) but have different signaling pathways (form) that conduct self-renewal. This is a fundamental problem to be tested. The idea that any cell can become a stem cell has been postulated (51). This may be an extreme view, but the concept that developmental time could be run in reverse is not new. It may be that adaptation in culture allows for ESCs to “rewind” their developmental program back to a point where a self-renewal pathway existed. This rewinding could potentially be a disruptive process, by which the original developmental program is lost; the resulting ESCs may not represent any particular endogenous cell type, but might have similarities to or be remnants of multiple early types. This may be why ESCs have been reported as promiscuous in their gene expression. Another possible outcome in this scenario is the arrival at a karyotypically abnormal state in cells with underlying chromosomal instability. Such instability has been reported to result in selection for advantageous chromosomal acquisitions leading to accelerated growth (52–54). As karyotypically normal lines also exist, this may be the result of one of many self-renewal pathways. Studies to address these issues await the appropriate markers and functional assays. Like MESCs, the first HESCs were derived from blastocysts cultured on MEF feeder cells, but it quickly became clear that the required conditions for maintaining their undifferentiated state were different from those described for MESCs. For example, basic FGF (bFGF) is required for HESCs to maintain their undifferentiated phenotype, while it is not required in MESCs. Additionally, neither human nor nonhuman primate cell lines are sensitive to LIF for the maintenance of pluripotency. Despite these differences, it is clear that the MEF feeder layer provides a major, yet not fully defined, contribution to HESC maintenance. HESCs can be cultured in the absence of direct contact with the feeder cells (55), but require the presence of feederconditioned medium, suggesting that soluble factors are

Stem Cell Reviews ♦ Volume 1, 2005

Molecular Basis for HESC Pluripotency __________________________________________________________________________117 involved. This also implies that maintenance is independent of factors produced by feeders in response to HESC-derived signals, as well as signals passed by direct contact with the feeder layer. Although clinical application of HESCs is not technically prohibitive, the inclusion of xeno-derived culture components presents significant hurdles for products destined for cellbased therapy. Animal pathogens present in the MEF feeder layer could easily be transferred to the HESCs, which in turn may function as a carrier of these pathogens. Transmission to the individual could be efficient, as intended cell replacement therapies would bypass natural immunological barriers to these pathogens. Indeed, a recent study indicates that current culture conditions can result in the transfer of an immunogenic nonhuman sialic acid moiety to the HESCs (56). In addition to demonstrating safety on an individual level, limiting the use of xeno-derived products in cell-based therapies is critical to minimize the risk of transfer of animal pathogens into the general population. Aside from defining components of the conditioned medium that contribute to maintaining the undifferentiated state, there are regulatory issues relevant to therapeutic application, which arise from the use of feeders. The first involves the production of a consistent, predictable product. In terms of cell therapy, it will be necessary to reduce variability in the quality of HESC cultures from lot to lot and from cell line to cell line. Although the quality of feeders is important for maintenance of HESCs, precise indicators of feeder quality have not been found. In addition, feeder layers stimulated by serumderived factors, such as bFGF, might produce factors not found in unstimulated feeder-conditioned medium. As most protocols call for including bFGF in the conditioning medium, this further complication may introduce significant variability in the quality and quantity of factors produced in response to the FGF signal. It is also possible that the strength or quality of the FGF signal might induce different factors or different amounts of active factors. For example, it has been suggested that variation between lots and feeder densities induces variable levels of TGF-β-inhibitory signals that might inhibit extraembryonic endoderm development in HESC cultures or differentially modulate their response to differentiation signals (57). Taken together, these observations indicate that the use of feeders in HESC culture will undermine the consistency of HESCs as a clinically useful product. Achieving controlled application of the active components using purified compounds could reduce variability of cultured HESCs. This is critical for largescale expansion of HESCs to meet clinical needs. Removal of the feeder cells eliminates many of the barriers to their use in the clinic. Replacing their active functions, for example, with small molecule or recombinant protein pathway effectors in HESC culture is a promising task for future development, which will undoubtedly be full of surprises.

References 1. Brivanlou AH, Gage FH, Jaenisch R, Jessell T, Melton D, Rossant J. Science 2003;300:913–916. 2. Sato N, Sanjuan IM, Heke M, Uchida M, Naef F, Brivanlou AH. Dev Biol 2003;260:404–413. 3. Ivanova NB, Dimos JT, Schaniel C, Hackney JA, Moore KA, Lemischka IR. Science 2002;298:601–614.

4. Ramalho-Santos M, Yoon S, Matsuzaki Y, Mulligan RC, Melton DA. Science 2002;298:597–600. 5. Abeyta MJ, Clark AT, Rodriguez RT, Bodnar MS, Pera RA, Firpo MT. Hum Mol Genet 2004;13:601–608. 6. Wei CL, Miura T, Robson P, et al. Stem Cells 2005;23:166–185. 7. Marshall E. Science 2004;306:630, 631. 8. Kothapalli R, Yoder SJ, Mane S, Loughran TP, Jr. BMC. Bioinformatics 2002;3:22. 9. Yuen T, Wurmbach E, Pfeffer RL, Ebersole BJ, Sealfon SC. Nucleic Acids Res 2002;30:e48. 10. Kuo WP, Jenssen TK, Butte AJ, Ohno-Machado L, Kohane IS. Bioinformatics 2002;18:405–412. 11. Li J, Pankratz M, Johnson JA. Toxicol Sci 2002;69:383–390. 12. Barczak A, Rodriguez MW, Hanspers K, et al. Genome Res 2003;13:1775–1785. 13. Yauk CL, Berndt ML, Williams A, Douglas GR. Nucleic Acids Res 2004;32:e124. 14. Suárez-Fariñas M, Noggle S, Heke M, Brivanlou A, Magnasco M. BMC Genomics 2005;6:99. 15. Mitsui K, Tokuzawa Y, Itoh H, et al. Cell 2003;113:631–642. 16. Chambers I, Colby D, Robertson M, et al. Cell 2003;113:643–655. 17. Granata A, Quaderi NA. Dev Biol 2003;258:397–405. 18. Pirvola U, Zhang X, Mantela J, Ornitz DM, Ylikoski J. Dev Biol 2004;273:350–360. 19. Willerton L, Smith RA, Russell D, Mackay S. Int J Dev Biol 2004;48:637–643. 20. Besser D. J Biol Chem 2004;279:45,076–45,084. 21. Hamatani T, Carter MG, Sharov AA, Ko MS. Dev Cell 2004; 6:117–131. 22. Yamamoto M, Saijoh Y, Perea-Gomez A, et al. Nature 2004;428:387–392. 23. Ding S, Schultz PG. Nat Biotechnol 2004;22:833–840. 24. Brivanlou AH, Darnell JE, Jr. Science 2002;295:813–818. 25. van Es JH, Barker N, Clevers H. Curr Opin Genet Dev 2003;13:28–33. 26. Moon RT, Bowerman B, Boutros M, Perrimon N. Science 2002;296:1644–1646. 27. Knockaert M, Greengard P, Meijer L. Trends Pharmacol Sci 2002;23:417–425. 28. Meijer L, Skaltsounis AL, Magiatis P, et al. Chem Biol 2003;10:1255–1266. 29. Sato N, Meijer L, Skaltsounis L, Greengard P, Brivanlou AH. Nat Med 2004;10:55–63. 30. Martinez A, Castro A, Dorronsoro I, Alonso M. Med Res Rev 2002;22:373–384. 31. Smith AG, Hooper ML. Dev Biol 1987;121:1–9. 32. Smith AG, Heath JK, Donaldson DD, et al. Nature 1988;336: 688–690. 33. Williams RL, Hilton DJ, Pease S, et al. Nature 1988;336:684–687. 34. Gearing DP, Bruce AG. New Biol 1992;4:61–65. 35. Gearing DP, Thut CJ, VandeBos T, et al. EMBO J 1991;10: 2839–2848. 36. Niwa H, Burdon T, Chambers I, Smith A. Genes Dev 1998; 12:2048–2060. 37. Burdon T, Stracey C, Chambers I, Nichols J, Smith A. Dev Biol 1999;210:30–43. 38. Dani C, Chambers I, Johnstone S, et al. Dev Biol 1998;203: 149–162. 39. Smith AG. Annu Rev Cell Dev Biol 2001;17:435–462. 40. Huelsken J, Vogel R, Brinkmann V, Erdmann B, Birchmeier C, Birchmeier W. J Cell Biol 2000;148:567–578. 41. Hemmati-Brivanlou A, Melton D. Cell 1997;88:13–17. 42. Chang C, Eggen BJ, Weinstein DC, Brivanlou AH. Dev Biol 2003;255:1–11. 43. James D, Levine A, Besser D, Brivanlou A., in press. 44. Niwa H, Miyazaki J, Smith AG. Nat Genet 2000;24:372–376. 45. Xu RH, Chen X, Li DS, et al. Nat Biotechnol 2002;20:1261–1264. 46. Zwaka TP, Thomson JA. Development 2005;132:227–233. 47. Ramalho-Santos M. BioEssays 2004;26:1013–1016.

Stem Cell Reviews ♦ Volume 1, 2005

118__________________________________________________________________________________________________Noggle et al. 48. 49. 50. 51. 52.

Rosler ES, Fisk GJ, Ares X, et al. Dev Dyn 2004;229:259–274. Carpenter MK, Rosler ES, Fisk GJ, et al. Dev Dyn 2004;229:243–258. Vallier L, Reynolds D, Pedersen RA. Dev Biol 2004;275:403–421. Blau HM, Brazelton TR, Weimann JM. Cell 2001;105:829–841. Mitalipova MM, Rao RR, Hoyer DM, et al. Nat Biotechnol 2005;23:19–20. 53. Draper JS, Smith K, Gokhale P, et al. Nat Biotechnol 2004;22:53, 54.

54. Buzzard JJ, Gough NM, Crook JM, Colman A. Nat Biotechnol 2004;22:381,382; author reply 382. 55. Xu C, Inokuma MS, Denham J, et al. Nat Biotechnol 2001; 19:971–974. 56. Martin MJ, Muotri A, Gage F, Varki A. Nat Med 2005;11:228–232. 57. Pera MF, Andrade J, Houssami S, et al. J Cell Sci 2004; 117:1269–1280.

Stem Cell Reviews ♦ Volume 1, 2005