Role of Signaling Pathways and Epigenetic Factors in Lineage Determination During Human Embryonic Stem Cell Differentiation

Human embryonic stem cells (hESCs) are culture-adapted cells that were originally derived from the inner cell mass (ICM) of the blastocyst-stage embryo [1]. HESCs are pluripotent cells that can be propagated indefinitely in culture, while retaining the in vivo properties of ICM cells; they can give rise to all tissues of the three germ layers (ectoderm, mesoderm and endoderm). Due to their pluripotency, hESCs have been the subject of intense research since they were initially isolated in 1998. HESCs can serve as model systems to study early human development, in addition to providing a potentially unlimited source of functional tissues for use in drug evaluation and regenerative medicine. Nevertheless, despite major advances, the exact molecular mechanisms that govern the self-renewal and differentiation of hESCs remain unclear. Indeed, a mechanistic understanding of the molecular processes regulating hESC fate can elucidate early events in human development and enable the development of protocols for efficient generation of functional tissues. Here we review the molecular mechanisms that regulate hESC fate; specifically, we focus on the role of signaling pathways and factors regulating epigenetic changes, in hESC self-renewal and lineage-specific differentiation. In hESCs, as in embryos, differentiation is triggered by developmental cues such as morphogens or cytokines that are present in the extracellular space. These morphogens or cytokines bind to their cognate plasma membrane-bound receptors and activate specific signaling pathways inside the cell. Activation of signaling pathways involves a sequence of phosphorylation events that eventually result in the regulation of specific transcription factors. These transcription factors, in turn, can recruit other co-factors and directly cause transcription of downstream genes. Furthermore, transcription factors can recruit histone modifying and chromatin remodeling enzymes to reshuffle the epigenetic structure, such that pluripotency genes become inaccessible for transcription and are repressed, whereas lineage-specific genes become accessible and are activated. This sequence of events finally leads to expression of lineage-specific proteins such as transcription factors and structural proteins, causing a morphological change in the cell. Also, pluripotency associated transcription factors and other pluripotency-associated genes are permanently repressed, thereby completing the process of differentiation. Thus, the process of differentiation is a

phosphorylated tyrosines of FGFRs cause recruitment of the GRB2/SOS complex and its subsequent activation. SOS then activates RAS, which triggers the MAPK cascade, finally leading to the activation of extracellular-signal related kinases (ERKs). Activated Erk1/2 can phosphorylate and control the activity of a wide range of proteins [35]. Notably, Erk1/2 can phosphorylate the linker region of SMAD1 and inhibit BMP signaling [36][37]. Additionally, activated FGF receptors can also recruit FRS2/GRB2 and activate them, leading to recruitment of GAB1. GAB1 activates the Phosphatidylinositol 3-kinase (PI3K) pathway. Thus, FGF can also activate the PI3K pathway in a cell-specific context.

Role in lineage determination
The FGF pathway is needed for proper embryonic development in vertebrates. Experiments in mice have shown that FGF4 is required for ICM proliferation and maintenance [38][39], and for trophectoderm and primitive endoderm development [40]. FGF4 is secreted by the ICM and supports trophectoderm maintenance [41]. Interestingly, activation of HRas1 which is a component of the MAPK pathway, in mouse embryonic stem cells (mESCs) caused upregulation of Cdx2, a trophectoderm marker, and trophectoderm stem cells were derived from these mutants [42]. FGF4 secreted by ICM also causes primitive endoderm differentiation [43]. Further along the developmental timeline, FGF signaling is required for the induction of paraxial mesoderm and for maintenance (but not induction) of axial mesoderm [44][45][46]. FGF is required for primitive streak formation and cell migration during gastrulation [45,47]. Inhibition of FGF is required for blood development [48][49][50], whereas activation of FGF signaling is required for neural differentiation [51]. However the mechanism by which FGF aids in embryonic neural differentiation is not fully understood [52][53]. In vitro studies have mimicked the role of FGF signaling in hESC maintenance and differentiation. FGF2 is required for maintaining hESCs in a pluripotent state [24,[54][55]. FGF signaling is also required for maintenance of mouse trophectoderm stem cells which are derived from the trophectoderm tissue of mouse embryos [56]. While there are no specific studies which delineate the inductive and maintenance/proliferative roles of FGF during hESC differentiation, work with mESCs have given ambiguous results, some showing that autocrine FGF2 is essential for neural differentiation [57] while others showing that FGF2 has a role in maintenance rather than induction of neural differentiation [58][59]. However, FGF signaling does seem to be necessary for inducing the posterior nervous system in vertebrate embryos [53,60].

The Wnt pathway
The Wnt pathway is widely implicated during various stages of embryonic development, homeostasis as well as in cancer. This pathway (reviewed in [61][62]) has canonical and noncanonical branches. The canonical Wnt pathway is activated by binding of Wnt ligands to the Frizzled receptors and low-density lipoprotein receptor-related protein 5/6 (LRP5/6) coreceptors leading to the recruitment of Dishevelled to the Frizzled receptor. This causes recruitment of Axin to the receptor complex, causing the subsequent deactivation of Axin. In the absence of Wnt signaling, Axin associates with GSK3, adenomatous polyposis coli (APC), casein kinase 1 (CK1) and -catenin. CK1 and GSK3 phosphorylate -catenin causing it to be degraded. Upon Wnt activation, Axin is inhibited and -catenin becomes derepressed, and subsequently enters the nucleus to function as a transcription factor. Various co-factors associate with -catenin and control its promoter specificity, thus dictating the www.intechopen.com Role of Signaling Pathways and Epigenetic Factors in Lineage Determination During Human Embryonic Stem Cell Differentiation 7 target genes activated or repressed by -catenin [63][64]. The non-canonical Wnt pathway acts independently of -catenin and is also required during embryogenesis. The details of this Ca 2+ -dependent pathway are reviewed in [62].

Role in lineage determination
The canonical Wnt pathway is activated during gastrulation [65] and mutation of Wnt3 blocks primitive streak formation resulting in lack of mesoderm and endoderm [66] (the primitive streak-specific transcription factor Brachyury is a direct target of Wnt3a signaling [67]). Similar defects are seen in Lrp5/6 double mutants and -catenin loss-of-function mutants [68][69]. Interestingly, expression of Cripto, a co-receptor for Nodal signaling, is missing in -catenin loss-of-function mutants [70]. Also, -catenin is indispensable for endoderm formation and loss of -catenin causes definitive endoderm to change into precardiac mesoderm [71]. Although loss of Wnt signaling leads to loss of mesoderm formation, inhibition of Wnt signaling is required for a cardiac fate [65,72], once pre-cardiac mesoderm has been induced. Remarkably, similar reports for the role of Wnt signaling have been obtained through in vitro differentiation studies in hESCs. Over-expression of -catenin in hESC cultures lead to primitive streak formation [73]. Inputs form Activin/Nodal and BMP pathways are necessary for further lineage specification into mesoderm/endoderm. Blocking BMP signaling abolishes mesoderm and leads to endoderm formation, whereas Activin/Nodal is required for endoderm formation [73]. Wnt is required for mesoderm differentiation but must be inhibited thereafter for cardiac mesoderm formation [74].

The Phosphatidylinositol-3Kinase pathway
The Phosphatidylinositol-3Kinase (PI3K) pathway regulates cell survival, apoptosis and has been implicated in cancer. The pathway as well as its role in cancer is reviewed in [75][76][77][78]. It also has been, in select cases, implicated in lineage-specific hESC differentiation [79]. The pathway is activated when PI3K is phosphorylated; this can happen through binding of Insulin to the Insulin receptor or of Insulin-like growth Factor (IGF) to Insulin-like Growth Factor Receptor (IGFR), or as previously discussed, by recruitment and activation of GAB1 by FGFR. Activated PI3K phosphorylates Phosphatidylinositol (4, 5)-biphosphate to Phosphatidylinositol (3, 4, 5)-triphosphate and creates a docking site for proteins with a pleckstrin homology (PH) domain, such as Akt. Once Akt is properly docked, it is phosphorylated and activated by protein-dependent kinase 1 (PDK1). Akt can then dissociate and activate/repress numerous proteins by phosphorylating them [80]. The PI3K pathway has not received much attention during vertebrate embryogenesis, though some recent studies have emerged to show that it is necessary for normal embryo development [81][82]. Homozygous null mutations in the p110 subunit of PI3K cause embryonic lethality before formation of the blastocyst [83]. Thus, in vivo studies do not implicate PI3K signaling in differentiation of cells, but rather in maintenance of cell viability [82]. It has been hypothesized that growth factors maintain PI3K signaling during embryogenesis to guard against ectopic or metastatic growth of cells, since such ectopic/metastatic cells do not receive enough growth factors and enter the default apoptotic pathway [84]. In contrast, in vitro studies with hESCs have shown some supportive role for the PI3K pathway in definitive endoderm differentiation. Inhibition of PI3K signaling enhances definitive endoderm differentiation by Activin [85]. Other conflicting reports show that PI3K signaling stabilizes -catenin during definitive endoderm formation [73]. A major challenge in downstream of receptor activation [9]. Both TGF-and FGF pathways can activate the PI3K pathways directly at the receptor level [9,30]. However, the crosstalk between pathways is cell-specific, since the available pool of interacting proteins depends on the cell-type. Also, the promoter accessibility of downstream genes is dependent on cell-type. Therefore, here we restrict our discussion to crosstalk events identified specifically in hESCs. Activin/BMP signaling induces Wnt ligand expression in hESCs [74]. Also, Activin regulates FGF, Wnt and BMP pathways in hESCs [104]. Inhibition of Activin/Nodal signaling causes downregulation of Wnt3, FGF2, FGF4 and FGF8 expression [26] and upregulation of BMP signaling [26] while activation of Activin signaling causes upregulation of Wnt3 and FGF8 expression [104]. Interestingly, upregulation of Activin signaling also causes upregulation of Nodal and Lefty expression. Cerberus1, an inhibitor of Nodal signaling, is a downstream target of both Wnt and Nodal pathways in hESCs [105]. Expression of Cripto, a co-activator of Nodal signaling, is upregulated by FGF signaling in hESCs [24]. As described earlier, YAP, which is regulated by the Hippo pathway, controls the nuclear localization of Smad2 in hESCs [93]. Thus, it can be seen that hESCs exhibit considerable endogenous signaling wherein, signaling pathways not only control their own ligand expression but also the expression of ligands of other pathways.

Other regulators of differentiation
While morphogens and other developmental cues act as the environmental input to hESCs and trigger the process of differentiation, the molecular mechanisms responsible for carrying out differentiation inside the cell are complex and require many key factors. These factors are required for the following: 1) to bring about a change in gene expression which causes the cell to transition into the new lineage-specific physiology, 2) to reshuffle the epigenetic structure of the genome, and finally 3) to make the new epigenetic structure permanent, lending stability to the newly formed cellular physiology. We will now discuss these intracellular factors that mediate various aspects of the differentiation process.

MicroRNAs
MicroRNAs (miRNAs) have emerged as a new paradigm for regulating gene expression at the post-transcriptional level. The role of miRNAs in embryogenesis, stem cell fate and cancer is reviewed in [106][107][108]. Transcription factors regulate promoter regions of miRNAs, which upon synthesis can target many mRNAs and lead to downregulation of protein synthesis. MiRNAs, which upon transcription are called pri-miRNAs, fold into secondary structures with characteristic hairpin-loops. These are recognized by Drosha, which cleaves the hairpin-loop structures to generate pre-miRNAs. Pre-miRNAs are then exported to the cytoplasm and recognized by Dicer, which cleaves one of the strands and incorporates the other into the RNA-induced silencing complex (RISC). Once incorporated into RISC, the single-stranded miRNA recognizes target mRNAs (usually many different mRNA targets) through partial sequence complimentarily, and causes down-regulation of protein synthesis. It is now known that miRNAs play an important role in embryogenesis [109] and lineagedetermination, and that many lineages have their characteristic miRNA expression patterns, akin to characteristic mRNA expression patterns [110]. The role of miRNAs in embryogenesis is evident from the fact that Dicer mutant mouse embryos die during gastrulation [111], while Dicer deficient zebrafish embryos do not develop beyond day8 [109]. The role of miRNAs in embryonic stem cell pluripotency and differentiation has also been demonstrated recently [107]. Dicer deficient mESCs fail to differentiate in vitro as well as in vivo [112]. Over-expression of miR-302 leads to reprogramming of human hair follicle cells and human skin cancer cells to form iPS cells [113][114]. During mESC differentiation, miR-134, miR-296 and miR-470 target and downregulate the transcription factors Nanog, Sox2 and Oct4 [115]. MiR-200c, miR-203 and miR-183 target and repress Sox2 and Klf4, both of which are involved in maintaining pluripotency in mESCs [116]. Similarly, during hESC differentiation, miR-145 targets and represses OCT4, SOX2 and KLF4 [117]. Sall4, another pluripotency-related transcription factor, is positively regulated by the ESC cell cycle regulating (ESCC) family of miRNAs and negatively regulated by the let7 family [118]. Additionally, miRNAs are also implicated during later stages of differentiation. The musclespecific miR-1 controls cardiomyocyte differentiation and proliferation in mice by targeting the Hand2 transcription factor [119]. miR-181 controls hematopoietic differentiation in mice [120] and miR-143 regulates adipocyte differentiation [121]. MiR-196 is involved in HOX gene regulation [122][123][124] and the miR-200 family regulates olfactory neurogenesis [125]. MiRNAs have also been implicated in skin morphogenesis [126]. Most intriguingly, transfection of muscle-specific miR-1 or brain-specific miR-124 into human HeLa cells shifts the mRNA expression profile towards that of muscle or brain cells, respectively [127]. The expression of miRNAs is regulated by transcription factors which bind promoter regions of genes harboring miRNAs; more than half of known mammalian miRNA genes are within host gene introns and are spliced after transcription [128]. For example, Activin A signaling regulates the expression of ~12 miRNAs in hESC cultures [129]. OCT4, NANOG and SOX2 occupy the promoter regions of ~14 miRNAs in hESCs [130]. Additionally, miRNAs can be regulated directly by signaling pathways. In smooth muscle cells, BMP4 or TGF-signaling causes increased processing of pri-miR-21 and pri-miR-199a [131] and regulates the processing of numerous other miRNAs [132]. The MAPK/ERK pathway can regulate miRNA maturation in the cytosol by controlling phosphorylation of TRBP, which functions with Dicer [133]. However, it is largely accepted that miRNAs do not trigger differentiation but rather, are required for carrying out the process of differentiation [107]. It is hypothesized that miRNAs are required to dampen the stochastic noise in mRNA transcription levels of genes during the process of differentiation. Thus, miRNAs add another layer of complexity to gene regulation during the process of differentiation, by finetuning active mRNA levels of a gene.

Epigenetic factors
Our discussion on signaling pathways focused on how a change in gene expression during differentiation is initiated; while miRNAs are most probably required to stabilize the mRNA levels against stochastic perturbations during differentiation. However, to provide longterm stability to the new gene expression pattern, the epigenetic structure of the genome needs to be changed. Epigenetic factors are responsible for modulating the epigenetic structure of hESCs, while it is pluripotent (reviewed in [134]) as well as while it goes through differentiation. The epigenetic structure of the genome dictates the promoter regions that would be accessible to transcription factors for initiating transcription; the heterochromatin, being densely packed, is inaccessible whereas the euchromatin is loosely packed and readily accessible. The epigenetic structure of hESCs is different from that of differentiated cells. The epigenetic structure also differs across various lineages of differentiation. Thus, epigenetic factors are involved in changing the epigenetic structure of the genome in a lineage-dependent fashion.
Epigenetic factors comprise broadly of histone modifying enzymes and chromatin remodeling complexes. The concerted action of both is needed to bring a stable change in the epigenetic landscape. Histone modifying enzymes are enzymes that modify histones post-translationally and create an epigenetic code of various acetylations, ubiquitinations and methylations throughout the genome (the histone code hypothesis [135][136]). This code is then recognized by chromatin remodeling enzymes, which alter the higher order structure of nucleosomes by creating heterochromatin and euchromatin. By controlling the formation of heterochromatin and euchromatin, these epigenetic factors control promoter accessibility and gene expression during differentiation. Differentiation is thought to proceed through activation of lineage-specific genes and repression of pluripotency genes [137]. This requires epigenetic factors to create repressive histone modifications on pluripotency genes (which were hitherto active) and reciprocally, to create activating histone modifications on lineagespecific genes (which were hitherto repressed). Permanent modification of histones also allows for epigenetic stability of the differentiated cell, which now becomes locked in this lineage. There is also some feedback from chromatin remodeling enzymes back to histone modifying enzymes. This means that certain chromatin remodeling enzymes can recruit back specific histone modifying enzymes for changing the histone code further. This is thought to provide more robustness to this system of epigenetic modification, thus lending further stability to the differentiated phenotype.

Histone Acetyltransferases
Of the various histone modifications, acetylation and methylation are critical for regulating the chromatin structure and gene expression [138]. These histone modifications, which create the genome-wide histone code, are regulated by Histone Acetyltransferases (HATs), Histone Deacetylases (HDACs), Histone Methyltransferases (HMTs) and Histone Demethylases. Histone Acetyltransferases are further classified into five families [139][140]: the Gcn5-related HATs (GNATs); the MYST (MOZ, Ybf2/Sas3, Sas2 and Tip60)-related HATs; p300/CBP HATs; general transcription factor HATs; and nuclear hormone-related HATs. In humans, the identified GNAT-related HAT complexes are PCAF, STAGA and TFTC. All three complexes have the chromatin-binding bromodomain which targets these complexes to chromatin. The bromodomain specifically recognizes and binds acetylated histones [141][142]. In mammals, the identified MYST-related HATs are Moz, Qkf, Mof, Tip60 (homologue of yeast NuA4 [143]) and Hbo1. The TIP60 complex contains the chromatin-binding chromodomain. The chromodomain of yeast SAGA HAT complex has been shown to recognize methylated histones [144] raising the possibility that TIP60 may also be recruited to methylated histones in humans. Together with the case of the bromodomain containing complexes, this implies that HATs may be recruited to specifically tagged histones and may function in a signaling cascade to modify the epigenetic map of the genome [145]. Mof homozygous null mice lack H4K16 acetylation and arrest at blastocyst stage [146]. Homozygous null Tip60 mutant mice also die during blastocyst stage [147]. Tip60 has also been implicated in pluripotency of ESCs [148]. Qkf is required for normal development of neurons of the cerebral cortex [149], whereas Moz is required for normal hematopoietic stem cell development [150][151][152].

Histone Deacetylases
The family of Histone Deacetylases is classified into four groups [153][154][155]: the Class I HDACs (yeast Rpd3-like) comprising of HDAC1/2, HDAC3 and HDAC8; the Class II HDACs (yeast Hda1-like) comprising of HDACs4-7, HDAC9 and HDAC10; the Class III HDACs (Sir2-like) comprising of SIRT1-7; and the Class IV HDACs (HDAC11-like) comprising of HDAC11. Of these, HDAC1 and HDAC2 have been identified in numerous complexes [156], namely: the SIN3 co-repressor complex, the nucleosome remodeling and deacetylase (NuRD) complex, the CoREST complex, the Nanog and Oct4 associated deacetylase (NODE) complex and the SHIP1 containing complex. HDACs complexes become associated with transcription factors through mediator proteins such as Sin3, NCoR, SMRT, CtBP and TLE [157]. Hdac1 and Hdac2 are important for embryonic development, especially during myogenesis, neurogenesis, haematopoiesis and epithelial cell differentiation [156]. The HDAC complexes NuRD and SIN3 are critical during different stages of embryonic development [158]. Mice embryos lacking Mbd3 or p66 , components of the NuRD complex, die during embryonic development [159][160]. Mbd3 null mice show normal segregation of trophoblast and primitive endoderm, but fail to develop embryonic ectoderm and extraembryonic ectoderm [161]. ICM cells of these embryos continue expressing Oct4 and the primitive endoderm marker Gata4 and fail to expand in number. Further, even though the primitive endoderm is present, the visceral endoderm fails to form. Analogously, the ICM cells derived from Mbd3 null mice did not expand ex-vivo and mESCs could not be formed [161]. Mbd3 null mESCs could initiate differentiation but could not commit to the differentiated lineages [162]. Mbd3 was also shown to suppress trophoblast commitment of mESCs [163]. P66 , however, was not required for proper blastocyst formation and implantation, and p66α null mice died later during embryogenesis [160]. Mi-2 , another component of the NuRD complex, is important for haematopoiesis, lymphopoiesis and skin development [164][165][166][167]. Similar to Mbd3 and p66α, Sin3a null mice embryos also die after implantation [168][169]. The ICM derived from these embryos shows severely retarded proliferation ex vivo [168]. Sin3b null embryos show defects in erythrocyte and granulocyte maturation and in skeletal development [170]. The Class III HDACs, known as Sirtuins (SirTs), are also implicated during differentiation and mammalian development [171]. SirT1 is highly expressed in ESCs and decreases during differentiation [172]. During late development, SirT2 modulates skeletal muscle and SirT1 modulates white adipose tissue differentiation [173][174]. Under oxidative stress, SirT1 causes astroglial differentiation in mouse neural progenitor cells [175]. SirtT2 controls gametogenesis in mice embryos [176].

Histone Methyltransferases
Various Histone Methyltransferases (HMTs) exist in the mammalian genome and many putative HMTs are yet to be discovered [177]. The major mammalian HMTs include Ash1l, Dot1l, Ezh1-2, G9a, GLP, Mll1-5, Nsd1, Prdm1-6, Prdm8-16, PrSet7, Setd1-7, Setdb1-2, Setmar, Smyd1-5, Suv39h1-2, Suv4-20h1-2 and Whsc1/l1. Their requirement during specific stages of mammalian development is comprehensively reviewed in [177]. These HMTs are associated with specific histone methylation activities on H3 and H4 histones. Although most of identified methylation marks are promiscuous and need further study, some histone methylation marks correlate well with gene activity. Transcriptionally active genes display H3K4me3 on their promoter region and H3K36me3 across the gene body, while repressed genes are enriched in H3K27me3 over the gene body, with some amount of H3K9me3 and H4K20me3 [178][179][180]. H3K4 methylation, which is associated with gene activation, is induced by Mll1-5, Setd1a/b and Ash1l. Therefore, these HMTs are critical during mammalian development. Mutations in Mll1 lead to embryonic lethality in mice [181][182] www.intechopen.com and cause aberrant regulation of Hox genes. Generation and/or expansion of hematopoietic stem cells, is abrogated in these embryos [183]. Mll2 null mice are capable of blastocyst formation and normal implantation without any lineage-specific growth abnormalities, but die later during embryonic development [184]. Very few genes are misregulated in Mll2 null mESCs, though Mll2 is needed for spermatogenesis [185]. Mll3 mutant mice show impaired differentiation towards the adipocyte lineage [186], while Mll5 mutant mice show impaired hematopoietic development [187]. H3K27 methylation, which is associated with gene repression and is important for embryonic development, is caused by the Ezh1-2 HTMs. Again, Ezh2 knockout causes early embryonic lethality in mouse embryos [188]. These embryos fail to complete gastrulation. Ezh2 is also shown to regulate epidermal and hematopoietic differentiation during embryogenesis [189][190][191]. H3K9 methylation is also associated with gene repression, and is induced by G9a, GLP, Prdm2, Setdb1 and Suv39h1-2. No gene targets for the Suv39h enzymes have been discovered though Suv39h double mutant mice display impaired viability as well as sterility [192]. In contrast, Setdb1 knockout causes early embryonic lethality in mouse embryos due to aberrant blastocyst formation, and mESCs cannot be derived from these mutant blastocysts [193]. Setdb1 also controls the switch between osteoblastogenesis and adipogenesis from bone marrow mesenchymal progenitor cells [194]. Similar to Setdb1, both G9a and GLP null mice also show embryonic lethality, including aberrant somitogenesis and aberrant neural tube formation [195][196]. G9a inactivates ~120 genes during mESC differentiation including Oct4 and Nanog, in concert with DNA Methyltransferases Dnmt3a/b [197][198]. G9a is also implicated for genomic imprinting in the mouse placenta [199].

Histone Demethylases
In humans, the identified histone demethylases include the KDM (Lysine (K) Demethylase) families of demethylases (KDM1-6), the PHF family and the JMJD6 family (reviewed in [200][201]). As with histone methyltransferases, histone demethylases are critical for embryonic development. The KDM1 family comprises of KDM1A and KDM1B. Homozygous deletion mutants of Kdm1a are early embryonic lethal and do not gastrulate [202]. Kdm1a null ES cells are pluripotent but do not form embryoid bodies and do not differentiate [202]. Kdm1b mutant mice embryos are maternal embryonic lethal and defective in imprinting [203]. The KDM2 family comprises of KDM2A and KDM2B, of which, KDM2B is implicated in osteogenesis from mesenchymal stem cells [204]. The KDM3 family consists of KDM3A, KDM3B and JMJD1C. KDM3A is required for spermatogenesis [205][206]. KDM3A is positively regulated by Oct4 and depletion of KDM3A from ES cells leads to differentiation [207]. The KDM4 family consists of KDM4A, KDM4B, KDM4C and KDM4D, of which, KDM4C is also positively regulated by Oct4 [207]. Depletion of KDM4C causes differentiation in ES cells and KDM4C also positively regulates Nanog expression [207]. The KDM5 family comprises of KDM5A, KDM5B, KDM5C and KDM5D, of which, KDM5A has been implicated in differentiation [208]. The KDM6 family consists of KDM6A, UTY and KDM6B. KDM6A and KDM6B are shown to regulate HOX gene expression during development [209][210]. KDM6B also controls neuronal differentiation and epidermal differentiation [211][212][213]. The PHF family includes JHDM1D, PHF2 and PHF8, while the JMJD6 family includes only JMJD6. JHDM1D is required for neural differentiation in mESCs and knockdown of Jhdm1d blocks neural differentiation [214].

Chromatin remodeling enzymes
Chromatin remodeling enzymes are involved in controlling the higher order structure of chromatin by creating heterochromatin and euchromatin, and utilize the energy of ATP to do so. Chromatin remodeling enzymes thereby control cell fate during differentiation of ESCs (reviewed in [229]). During in vivo development, chromatin remodeling enzymes have been shown to be important for myeloid differentiation, erythropoiesis, T-cell development, adipogenesis, neurogenesis and myogenesis (reviewed in [230]). Mammalian chromatin modeling enzymes are categorized into three families: SWI/SNF, ISWI and CHD. The SWI/SNF family is characterized by the presence of either Brg1 or Brm as the catalytic subunit, and is further categorized into two subfamilies: Baf and Pbaf [231]. Brg1 homozygous null mice die at peri-implantation stage and their ICM as well as trophoblast tissues die [232]. Similar phenotypes are observed homozygous mice null for Snf5, another subunit of SWI/SNF complexes [233]. However, downregulation of Baf60c causes late embryonic lethality due to defects in cardiac and skeletal muscle development [234]. Similarly, ablation of Baf180 leads to defects in heart development and placental trophectoderm development, and subsequent embryonic lethality [235] However, homozygous null mutations in Baf155 caused early embryonic lethality due to failure of ICM cells, though trophoblast giant cells were found to be normal [236]. Heterozygotes null for Baf155 showed defective brain development. In mESCs, knockdown of Brg1 leads to loss of self-renewal and impaired ability to differentiate to ectoderm and mesoderm [237]. Inactivation of Baf250b has been associated with reduced self-renewal and increased differentiation [238]. Interestingly, ablation of Baf250a caused failure of mesoderm formation in mouse embryos as well as in ESCbased embryoid body cultures [239]. However, primitive endoderm differentiation and neuronal differentiation could be established in these cells. Brg1, Baf47, Baf155 and Baf57 are www.intechopen.com Role of Signaling Pathways and Epigenetic Factors in Lineage Determination During Human Embryonic Stem Cell Differentiation 15 required to suppress Nanog expression during differentiation and knockdown of Baf155 or Baf57 lead to de-repression of Nanog levels during differentiation [240]. The ISWI family of chromatin remodelers is characterized by the presence of the Snf2h or Snf2l ATPase subunits that interact with unmodified histones. Snf2h is a part of the RSH, WICH, NoRC, CHRAC and ACF chromatin remodeling complexes. Snf2l is a part of the NURF and CERF complexes. Similar to Brg1, Snf2h homozygous null mice die at peri-implantation stage and both ICM and trophectoderm tissues degenerate [241]. Cecr2, a component of CERF, has been implicated in neural tube formation in mice [242]. Human NURF has also been implicated in neuronal development [243]. Interestingly, homozygous mice null for Bptf, a component of NURF, fail to form the visceral endoderm [244]. Bptf is required for ectoderm, mesoderm and both definitive and visceral endoderm development from ESCs in embryoid body cultures. Association between NURF complex and Smad transcription factors are necessary for endoderm formation [244]. The CHD family of chromatin remodelers is characterized by the presence of two chromodomains, with affinity for methylated histones. The CHD family is categorized into three subfamilies: family I with CHD1-2, family II with CHD3-4 and family III with CHD5-9. Downregulation of Chd1 in ESCs leads to impaired pluripotency, such that the cells are incapable of primitive endoderm and cardiac mesoderm differentiation and become prone to neural differentiation [245]. CHD3 and CHD4 are also found in NuRD histone deacetylase complexes described previously. Mutations in Chd7 are embryonic lethal, showing that Chd7 is also important for embryonic development [246].

Cell polarity
Cell polarity is a feature of cellular physiology exhibited by epithelial cells. It refers to uneven spatial distribution of proteins across the cell, causing different parts of the cell to have different morphology and functions. Cell polarity can be classified as apical-basal polarity (epithelial cells), anterior-posterior polarity (neurons) and planar polarity (cochlea). In mammals, the apical-basal polarity of epithelial cells is regulated by three distinct protein complexes: the Crumbs/PALS1/PATJ complex, the Par3/Par6/aPKC complex and the Scribble/Dlg/Lgl complex. A detailed review of these complexes and their function in controlling epithelial architecture, cell migration and tumorigenesis can be found in [272]. The apical-basal polarity is lost during epithelial-to-mesenchymal transformation (EMT), a process that changes epithelial cells to mesenchymal cells with no apical-basal polarity and occurs during embryogenesis, fibrosis and cancer metastasis. Importantly, recent findings have indicated that epithelial cell polarity is required for early mouse embryogenesis and may be the driving factor for differentiation of certain early lineages. For example, the cell polarity regulator Par6 is required for proper trophectoderm formation in mouse embryos [273]. Downregulation of Par3 or aPKC drives blastomeres towards ICM instead of trophectoderm [274]. Interestingly, the Crumbs polarity complex has been found to interact with components of the Hippo signaling pathway [99], and the Hippo has been implicated in trophectoderm formation as discussed in the review. It will therefore be interesting to study whether cell polarity proteins indeed control trophectoderm differentiation during early stages of embryogenesis.

Conclusions
The development and morphogenesis of the embryo is under strict control by a rather small set of signaling pathways. However, the presence of multiple ligands and multiple receptors, numerous transcription factor-binding partners and significant crosstalk between pathways gives rise to vast complexity within this small set of signaling pathways. Further, lineagespecific differentiation is also controlled by complex regulation of various histone acetyltransferases, histone deacetylases, histone methyltransferases, histone demethylases, DNA methyltransferases and chromatin-remodeling enzymes. A detailed and molecular-level understanding of the determinants of lineage specificity of differentiation has only recently begun to emerge through studies of the signaling pathways and their downstream factors. We have summarized the signaling pathways and miRNAs associated with differentiation to various lineages in Table 1. Along similar lines, various histone modifying proteins and chromatin remodeling proteins associated with various lineages are summarized in Table 2. Even though these tables present a concise mechanistic linkage between various regulators of ESC differentiation, questions regarding how lineage-specific transcription factors are regulated and how a balance between opposing factors, such as HATs and HDACs, or HMTs and histone demethylases, is achieved in the cell remain largely unanswered. Further, the possible role of lineage-specific transcription factors in the recruitment of epigenetic factors also remains largely unknown. Thus, our understanding of the lineage-specificity of differentiation is still rudimentary and requires significant additional research.

Acknowledgments
The authors gratefully acknowledge funding support from the National Science Foundation (NSF) Grant CBET-0966859.