The Sox Family of Transcription Factors: Versatile Regulators of Stem and Progenitor Cell Fate

Abstract

The Sox family of transcription factors are well-established regulators of cell fate decisions during development. Accumulating evidence documents that they play additional roles in adult tissue homeostasis and regeneration. Remarkably, forced expression of Sox factors, in combination with other synergistic factors, reprograms differentiated cells into somatic or pluripotent stem cells. Dysregulation of Sox factors has been further implicated in diseases including cancer. Here, we review molecular and functional evidence linking Sox proteins with stem cell biology, cellular reprogramming, and disease with an emphasis on Sox2.

Introduction

Stem cells are characterized by the capacity to continuously self-renew and the potential to differentiate into one or more mature cellular lineages (Simons and Clevers, 2011). They serve to form tissues and organs during mammalian development, and they maintain ongoing cellular turnover and provide regenerative capacity in certain adult tissues. One can distinguish between pluripotent embryonic stem cells (ESCs), which give rise to all embryonic lineages, and somatic stem cells, which give rise to one or more specialized lineages within the tissues they reside in. A stem cell’s decision for self-renewal or differentiation is intrinsically controlled by the interplay of cell type-specific transcription factors and chromatin regulators. Although several such molecules have been implicated in stem cell biology over the last few years, the mechanistic modes of action of these molecules remain incompletely understood.

Research on the Sox gene family began with the seminal discovery of the mammalian testis-determining factor, Sry (Gubbay et al., 1990; Sinclair et al., 1990). Sry carries a characteristic high-mobility-group (HMG) domain that binds DNA in a sequence-specific manner. In general, proteins containing an HMG domain with 50% or higher amino acid similarity to the HMG domain of Sry are referred to as Sox proteins (Sry-related HMG box). So far, twenty different Sox genes have been discovered in mice and humans (Schepers et al., 2002). In addition, two Sox-like genes have been identified in the unicellular choanoflagellate Monosiga brevicollis, suggesting that the origin of Sox proteins predates multicellularity or possibly marks the transition of unicellular to multicellular organisms (Guth and Wegner, 2008; King et al., 2008).

Sox proteins that share an HMG domain with more than 80% sequence identity are divided into different groups termed A to H (Table 1). Individual members within a group share biochemical properties and thus have overlapping functions (Wegner, 2010). In contrast, Sox factors from different groups have acquired distinct biological functions despite recognizing the same DNA consensus motif. Target gene selectivity by different Sox factors can be achieved through differential affinity for particular flanking sequences next to consensus Sox sites, homo- or heterodimerization among Sox proteins, posttranslational modifications of Sox factors, or interaction with other co-factors (Wegner, 2010). This molecular versatility may thus explain why the same Sox factors can play very different molecular and functional roles in distinct biological contexts.

Table 1. Sox factors implicated in stem cell biology.

Note: Only those Sox factors that are linked to stem cells by expression and functional evidence have been highlighted in this table. LT, lineage tracing; LOF, loss of function; GOF, gain of function.

Group	Sox Member	Expression			Function	Citations for functional role
		Pre-implantation Embryo	Fetus	Adult
SoxA	Sry	Founding member of the Sox familly, expression/role in stem cells undefined
SoxB1	Sox1	_	Neural progenitor cells (NPCs)	NPCs	Required for specification and maintenance of undifferentiated stem cells	Bylund et al. (2003); Graham et al. (2003); Pevny et al. (1998); Zhao et al. (2004)
	Sox2	Expression/role in ectoderm, endoderm and mesoderm derivatives, refer to figure 1 and text for details
	Sox3	_	NPCs	_	Same as Sox1	Bylund et al. (2003); Bergsland et al. (2011)
				Spermatogonia	Genetic deletion leads to loss of undifferentiated spermatogonia	Raverot et al. (2003)
SoxB2	Sox14, Sox21	expression/role in stem cells undefined
SoxC	Sox4, Sox11, Sox12	expression/role in stem cells undefined
SoxD	Sox5, Sox6, Sox13	expression/role in stem cells undefined
SoxE	Sox8	_	_	Muscle satellite cells	N/A	Schmidt et al. (2003)
	Sox9	_	Hair follicle stem cells		Deletion leads to loss of specification of early bulge cells needed to form the hair follicle and sebaceous gland	Vidal et al. (2005)
						Nowak et al. (2008)
			Distal tip cells (lung)		N/A	Rawlins et al. (2011)
		_	NPCs	_	Required for specification and maintenance of stem cells (LOF and GOF studies)	Scott et al. (2010)
			Premigratory NCSCs (neural crest stem cells) and migratory cranial NCSCs			Cheng et al. (2003)
		_	Retinal progenitor cells (RPCs)	_	Genetic deletion leads to a loss of stem cell differentiation potential to Muller glial lineage	Poche et al. (2008)
		_	Pancreatic progenitors	Exocrine pancreatic duct cells	Fetal LT marks all pancreatic lineages, adult LT marks all exocrine lineages. Genetic deletion leads to loss of pancreatic progenitors	Seymour et al. (2006); Furuyama et al. (2011)
		_	_	Liver duct cells	LT labels hepatocytes after injury	Furuyama et al. (2011)
		_	_	Intestinal stem/progenitor cells	Embryonic and adult LT labels all intestinal lineages. Genetic deletion depletes stem cells and Paneth cells	Furuyama et al. (2011); Sato et al. (2011)
		_	_	Mammary stem cells (MaSCs)	Knockdown leads to loss of stem cell maintenance	Guo et al. (2012)

Here, we review the biology of Sox factors that are implicated in stem cell biology in the context of development, tissue homeostasis, reprogramming and cancer. We place particular emphasis on the well-studied Sox2 protein with the goal of deriving general molecular and cellular principles by which Sox factors control stem and progenitor cell fates.

Sox factors in pre-implantation development and pluripotency

The formation of the trophectoderm (TE) and inner cell mass (ICM) within the blastocyst is the first lineage specification event in the mammalian embryo (Rossant and Tam, 2009). The ICM contains pluripotent founder cells, which give rise to all embryonic lineages, and a population of extra-embryonic endoderm (ExEn) cells that contribute to the yolk sac. Similarly, the TE contains a population of multipotent stem cells that form the extra-embryonic ectoderm and give rise to the placenta. Sox2 is initially present in both the ICM and the TE but is later confined to the ICM (Avilion et al., 2003). Zygotic deletion of Sox2 results in early embryonic lethality due to a failure to form the pluripotent epiblast but leaves the TE unperturbed (Avilion et al., 2003). Interestingly, subsequent studies showed that maternal Sox2 protein persists in pre-implantation embryos, which might have masked a phenotype in the TE in zygotic Sox2 mutants (Keramari et al., 2010). Indeed, depletion of both maternal and zygotic transcripts by RNAi causes an early arrest of embryos at the morula stage and a failure to form TE, suggesting that Sox2 is required for the segregation of the TE and ICM (Keramari et al., 2010). Consistent with its role in preimplantation development, Sox2-deficient embryos neither support the derivation of ESCs from the ICM, nor the derivation of trophoblast stem cells (TSCs) from the TE (Avilion et al., 2003). Furthermore, deletion of Sox2 in already established ESCs results in their inappropriate differentiation into trophectoderm-like cells, indicating that Sox2 is also critical for the maintenance of ESCs (Masui et al., 2007).

Interestingly, Sox2’s effect on self-renewal and differentiation of ESCs is highly dosage-dependent (Kopp et al., 2008), suggesting that its expression needs to be in equilibrium with other cofactors to maintain pluripotency. Supporting this concept is the observation that Sox2 acts cooperatively with other dosage-sensitive transcription factors, such as Oct4 and Nanog, to maintain the regulatory networks responsible for self-renewal and to repress differentiation programs in ESCs (Boyer et al., 2005; Chen et al., 2008; Kim et al., 2008; Orkin and Hochedlinger, 2011). Co-binding of these factors at targets associated with self-renewal facilitates recruitment of the co-activator p300 and consequently transcriptional activation (Chen et al., 2008), whereas co-binding at developmental target genes causes gene silencing in concert with the repressive polycomb complex (Boyer et al., 2006). Notably, a large fraction of target genes bound by these factors contain composite Oct4/Sox2 consensus binding sites (Masui et al., 2007; Tomioka et al., 2002), suggesting that Sox2 closely collaborates with Oct4 in order to efficiently bind to DNA and recruit other factors important for gene activation. In support of the notion that Oct4 and Sox2 jointly activate many targets is the finding that overexpression of Oct4 can partially compensate for the loss of Sox2 (Masui et al., 2007).

Upon specification of the ICM, the SoxF group member Sox17 becomes detectable in a rare population of cells destined to form the ExEn lineage (Kanai-Azuma et al., 2002; Niakan et al., 2010). Similar to the requirement for Sox2 in ESC and TSC derivation, Sox17 is essential for the establishment of extra-embryonic stem cell lines, termed XEN cells (Kunath et al., 2005; Niakan et al., 2010). At the molecular level, Sox17 has been placed downstream of the master regulator for primitive endoderm, Gata6 (Niakan et al., 2010). Accordingly, forced expression of Sox17 or its related group member Sox7 in ESCs results in a downregulation of the pluripotency gene expression program and an upregulation of the primitive endoderm-associated program, giving rise to endodermal progenitors (Niakan et al., 2010; Seguin et al., 2008). Mechanistically, Sox17 seems to oppose Sox2’s function by repressing pluripotency targets and activating endoderm targets when ectopically expressed in ESCs. Chromatin Immunoprecipitation (ChIP) experiments for Sox17 further suggest that this opposition is in part accomplished by displacing Nanog from silenced Sox2/Nanog targets, resulting in their transcriptional activation (Niakan et al., 2010).

Collectively, these results obtained from in vivo or in vitro studies document that different Sox factors play important and dosage-dependent roles in the establishment of cell lines of the three main cell lineages of the pre-implantation embryo, the ICM, TE and ExEn.

Sox2 in fetal development

After gastrulation of the embryo, Sox2 expression becomes largely restricted to the presumptive neuroectoderm, sensory placodes, brachial arches, gut endoderm and primordial germ cells (Avilion et al., 2003; Wood and Episkopou, 1999; Yabuta et al., 2006). Since Sox2 deficiency causes early post-implantation lethality (Avilion et al., 2003), functional evidence for its role in the fetus has required analyses of hypomorphic and conditional mutants in Xenopus, chick and mouse embryos. These studies have demonstrated the importance of Sox2 in lineage specification, morphogenesis, proliferation and differentiation in a variety of developing tissues of the fetus. In addition, these data have documented that the function of Sox2 is highly dosage and context dependent. In the following paragraphs, we will briefly summarize Sox2’s roles in developing endodermal, ectodermal and mesodermal cell lineages.

Sox2 in ectoderm development

Sox2 is expressed during the earliest stages of ESC differentiation towards the neural lineage in vitro. At the molecular level, Sox2 promotes early neuroectodermal fate by directly suppressing key regulators of the alternative mesendodermal fate such as brachyury (Thomson et al., 2011; Wang et al., 2012). In fact, forced Sox2 expression is sufficient to inhibit mesendodermal differentiation even under mesendoderm-promoting culture conditions, indicating that it acts downstream of environmental cues in this setting (Wang et al., 2012). Sox2 is involved in a similar cell fate decision in vivo during the differentiation of bipotential axial stem cells into either paraxial mesoderm or neural plate (Takemoto et al., 2011). Paraxial mesoderm gives rise to the vertebral column, dermis and skeletal muscle, whereas neural tube develops into the central nervous system (CNS). In the absence of competing factors, Sox2 drives axial stem cells towards a neural plate fate. However, in the presence of Tbx6, a regulator of presomitic mesoderm development, Sox2’s N1 enhancer becomes directly suppressed and axial stem cells are fated towards paraxial mesoderm. In agreement, Tbx6 loss or ectopic Sox2 expression results in the formation of ectopic neural tubes at the expense of paraxial mesoderm (Takemoto et al., 2011). Together, these results emphasize the importance of Sox2 in regulating early neural lineage specification in the embryo and in differentiating ESCs. The antagonism between Sox2 and Tbx6 in axial stem cells further exemplifies a general principle by which Sox factors regulate cell fate decisions during development, and will be discussed later on.

Sox2 continues to play major roles in the developing central (CNS) and peripheral nervous system (PNS) by controlling the proliferation and differentiation of fetal progenitor cells (Pevny and Nicolis, 2010; Wegner and Stolt, 2005). Notably, Sox2 expression overlaps and functions redundantly with that of the other two SoxB1 group factors, Sox1 and Sox3 in the CNS (Bylund et al., 2003; Graham et al., 2003; Wood and Episkopou, 1999)(Table 1). In general, overexpression of any of the SoxB1 factors promotes CNS progenitor cell proliferation, whereas depletion of these factors induces cell cycle exit and onset of differentiation (Bylund et al., 2003; Cavallaro et al., 2008; Ferri et al., 2004; Graham et al., 2003; Kishi et al., 2000; Miyagi et al., 2008). Likewise, Sox2 expression is essential for neural progenitor cell proliferation and differentiation in the retina, in part through its direct activation of the Notch1 gene (Taranova et al., 2006). Comparison of Sox2 hypomorphs of various alleles with Sox2 conditional null mice further suggests that Sox2 function in retinal progenitor cells(RPCs) is highly dosage-dependent. RPCs lacking Sox2 expression lose the competence to proliferate and differentiate, and reductions in Sox2 levels cause variable micropthalmia. In addition to neural progenitors of the brain and eye, Sox2 is transiently expressed in the Schwann cell lineage, which is of neural crest origin and responsible for the myelination of axons of the PNS. Similar to its role in CNS and retinal progenitors, Sox2 prevents terminal differentiation of Schwann cell precursors (Le et al., 2005).

Surprisingly, Sox2 expression has also been reported to be important for the differentiation of subsets of neurons, indicating that its function is not always confined to the maintenance of progenitors and stem cells. For example, Sox2 hypomorphic or knockout mice have reduced GABAergic interneurons in the newborn cortex and adult olfactory bulb (Cavallaro et al., 2008). Consistently, Sox2 mutant NPC cultures generate beta-tubulin-positive neuronal-like cells that are poorly arborized and are negative for markers of mature neurons and GABAergic neurons (Cavallaro et al., 2008; Ferri et al., 2004). In an independent in vitro differentiation paradigm, Sox2 was shown to promote the maturation of migrating neural crest progenitor cells into sensory ganglia (Cimadamore et al., 2011). Collectively, these studies demonstrate that SoxB1 proteins play key roles in the development of the CNS and the PNS by controlling both, the proliferation and differentiation of various progenitor cell populations. It will be important to define the mechanisms by which the same transcription factor regulates progenitor cell maintenance and differentiation within the same lineage (see also Mechanisms section/Pioneer factors). Sox2 is expressed in other developing ectoderm-derived tissues including the inner ear and dental epithelium, which will not be discussed here because of space constraints (see Figure 1 for summary)(Dabdoub et al., 2008; Juuri et al., 2012; Kiernan et al., 2005).

Figure 1. Sox2 expression in pluripotent, fetal and adult progenitor and stem cells.

Sox2 is expressed throughout development, initially in pluripotent founder cells of the blastocyst and subsequently in ectodermal, endodermal and mesodermal progenitors as well as in primordial germ cells. Sox2 expression is maintained in fetal and adult tissues derived from Sox2+ fetal progenitor cells and marks stem/progenitor cells and in some cases also differentiated cells.

Sox2 in endoderm development

Whereas Sox2 counteracts mesoderm specification in vivo and during ESC differentiation, elegant work by Hogan and colleagues showed that it plays multiple additional roles in organ specification of the foregut endoderm (Figure 1). Sox2 is highly expressed in the anterior part of the foregut, giving rise to esophagus and forestomach. However, it is lowly expressed in the future trachea and posterior stomach, respectively (Que et al., 2007). A severe decrease in Sox2 levels in hypomorphic embryos causes a transformation of esophagus into trachea, resulting in a failure to separate future trachea and esophagus (tracheoesophageal fistula) (Que et al., 2007). Interestingly, Sox2 appears to play an independent role in defining the boundary between the keratinized forestomach/esophagus and the glandular hindstomach/intestine based on the observation that Sox2 mutant esophagus and forestomach exhibit histological and molecular signs of glandular stomach and intestine. Experiments regulating Sox2 dosage have further demonstrated that Sox2 is required for patterning and morphogenesis of the embryonic tongue into taste bud sensory cells (Okubo et al., 2006), branching and differentiation of primary lung bud into the lung (Gontan et al., 2008; Ishii et al., 1998) and proper differentiation of the tracheal cartilage (Que et al., 2009).

These experiments document an interesting commonality and difference in how Sox2 controls stem and progenitor cells in distinct developing tissues. A commonality among stem and progenitor cells of the retina, foregut-derived tissues and pluripotent cells is a sensitivity to changes in Sox2 dose. This observation is consistent with the presence of cooperative and/or antagonistic factors whose function depends on finely tuned Sox2 levels and will be discussed below. A notable difference among these tissues is the effect Sox2 deletion has on cell proliferation. While neural progenitors generally exit the cell cycle upon Sox2 deletion, trachea, tongue and esophagus exhibit altered differentiation programs without changes in cell proliferation. Thus, Sox2 seems to control tissue formation in cell proliferation-dependent and independent ways that vary from tissue to tissue. Future studies of Sox2 targets in the respective cell types might give insights into the molecular mechanisms responsible for these different outcomes.

Sox2 in mesoderm development

During skin development, Sox2 is initially expressed in groups of mesenchymal cells called dermal condensates, that precede hair and whisker follicle formation and eventually give rise to the so-called dermal sheath and dermal papilla (DP)(Driskell et al., 2009; Rendl et al., 2005). The DP cyclically provides signals to the surrounding hair follicle to induce hair growth. While all DPs express Sox2 until shortly before birth, only a subset of them continues to be Sox2+ after birth (Driskell et al., 2009). This coincides with the emergence of different types of DP-associated hair follicles during development. In postnatal mice, Sox2+ DPs are associated with so-called guard, auchene and awl follicles, which form earlier in development, whereas Sox2-DPs are associated with zigzag follicles that form late in development. Of note, Sox2+ dermal cells appear to be the cells of origin of multipotent, self-renewing skin derived precursor cells (SKPs)(Fernandes et al., 2004). Both primary Sox2+ dermal cells and clonally derived SKPs induce hair morphogenesis upon transplantation into nude mice, and differentiate into multiple dermal cell types in vivo and neural cells in vitro (Biernaskie et al., 2009; Driskell et al., 2009). These characteristics identify Sox2+ dermal cells as putative dermal stem cells. DP cells and derivative SKPs were originally thought to originate exclusively from the neural crest (ectoderm). However, recent lineage tracing analyses with a somite-specific cre (Myf5-cre) driver has refined this interpretation. Trunk-derived DP cells and SKPs originate from somites (mesoderm), while facial-derived DP cells and SKPs originate from the neural crest (ectoderm)(Jinno et al., 2010). Evidence for a functional role for DP-specific Sox2 expression on hair follicle growth has been provided by the Rendl lab (Clavel et al., 2012). DP-specific Sox2 ablation leads to de-repression of its target Sostdc1, which normally inhibits Bmp signaling. A decrease in Bmp signaling from the DP results in a reduction of hair shaft progenitor cell migration in the adjacent follicle and thus impaired hair growth, which resembles that of zigzag hairs. Whether Sox2 also plays a role in wound repair and SKP self-renewal is an interesting question that remains to be addressed.

Sox2 has also been implicated in the proliferation of osteoblast progenitors in vitro and in vivo. Deletion of Sox2 in cultured osteoblast cell lines leads to a senescence-like phenotype, while its overexpression prevents differentiation (Mansukhani et al., 2005). Similarly, ablation of Sox2 in the osteoblast lineage in vivo using a Collagen (2.3kb)-driven Cre line results in reduced bone mineral density and bone volume (Basu-Roy et al., 2010)(Basu-Roy et al.), while transgenic overexpression inhibits mature osteoblast function (Holmes et al., 2011). Given that osteoblasts, like DP cells, can originate from both, neural crest and paraxial mesoderm, it remains to be formally shown that Sox2 is expressed in osteoblasts derived from both germ layers. Collectively, these experiments extend Sox2 expression and function in stem/progenitor cells from ectoderm and endoderm to that in mesoderm. Another important conclusion from these observations is that Sox2 can influence progenitor cell proliferation either directly by preventing cellular differentiation (e.g., in osteoblasts) or indirectly by suppressing pro-differentiation signals produced from adjacent cells (e.g., in DP cells).

Sox2 specifies cell fate by antagonizing other transcriptional regulators

A common theme emerging from the abovementioned observations is that Sox2 often determines cell fate by antagonizing transcription factors of alternative cell lineages (Figure 2). An example already mentioned in this review is the antagonism between Sox2 and Tbx6 during the specification of bipotential axial stem cells towards either Sox2+ neural tube or Tbx6+ axial mesoderm (Takemoto et al., 2011). Likewise, Sox2 antagonizes the transcription factor Nkx2.1 during foregut development; Sox2 is expressed most anteriorly in the future esophagus and stomach whereas Nkx2.1 is expressed ventrally in the future trachea (Que et al., 2007). Accordingly, embryos deficient for Nkx2.1 exhibit the reciprocal phenotype to Sox2 mutants displaying ectopic Sox2 expression and a transformation of future trachea into esophagus (Que et al., 2007). Furthermore, antagonism between the stomach-specifying Sox2/Barx1/Sfrp pathway and the intestinal fate-promoting Wnt/Cdx2 pathway is responsible for establishing the boundary between the glandular stomach and the intestine (Zorn and Wells, 2009). Lastly, the interaction between Sox2 and Mitf/Egr2 regulates the differentiation of Schwann cell progenitors into either myelinating Schwann cells or melanocytes (Adameyko et al., 2012). Specifically, Sox2 maintains a Schwann cell progenitor state whereas its cross-regulatory interactions with either Mitf or Egr2 consolidates mature Schwann cell or melanocyte fates, respectively (Adameyko et al., 2012). The suppression of Mitf expression by Sox2 may be direct since Sox2 protein was detected at the proximal Mitf-m promoter in ESCs and melanoma cell lines (Adameyko et al., 2012). Whether mutual repression between Sox2 and Nkx2.1 or Sox2 and Cdx2 also involves direct binding to the respective regulatory regions remains to be determined. Development of a Sox2 overexpression mouse model showed that Sox2 activates Sox21, which in turn binds to and represses Cdx2 in ESCs and neural progenitors, thus arguing for an indirect mechanism in this particular context (Kuzmichev et al., 2012). Together, these observations underscore a general principle of how Sox2 drives cell fate decisions during development, namely by directly or indirectly inhibiting regulators of alternative cell fates.

Figure 2. Antagonisms between Sox2 and other lineage transcription factors determines cell fate.

During organogenesis, Sox2 influences cell fate by inhibiting transcription factors that specify alternative cell lineages. Sox2 is expressed in an inverse gradient with the respective other transcription factor, and thus acts in a dosage-dependent manner to establish cellular identities within and boundaries between future tissues.

It is important to recognize, however, that the antagonisms between Sox2 and other transcription factors are highly cell type and developmental stage specific. In fact, transcription factor pairs that are antagonistic in one cell type or developmental stage may cooperate in other cellular or developmental settings. A case in point is the Sox2/Pax6 pair. Ablation of Sox2 in multipotent optic cup progenitors biases them towards a non-neurogenic ciliary body epithelium fate (Matsushima et al., 2011). This phenotype is rescued in a Pax6 heterozygous (haploinsufficient) background (Matsushima et al., 2011), indicating that Sox2 specifies a neurogenic fate whereas Pax6 instructs a non-neurogenic fate. In contrast to this antagonistic relationship during development of the optic cup, Sox2 and Pax6 cooperate during lens development by forming a complex on lens-specific enhancer elements such as that of the delta crystalline gene (Kamachi et al., 2001). In support of the cooperative role of Sox2 and Pax6 in lens specification, combined expression of both factors is sufficient to differentiate embryonic ectoderm into lens ectoderm.

Sox2 in tissue homeostasis and regeneration

Accumulating data indicate that tissues that require Sox2 during development continue to express this factor in some adult stem and progenitor cells derived from that tissue (Figure 1). Below, we will review the expression patterns and, where available, functional data linking Sox2 with adult stem and progenitor cells.

Using Sox2-GFP knock-in mice, Pevny and coworkers first demonstrated that Sox2 is not only expressed in fetal neural progenitors, but also in proliferating cells in the adult CNS, specifically in neurogenic regions, such as the subventricular zone of the lateral ventricle and the subgranular zone of the hippocampus as well as the ependyma of the adult central canal (Ellis et al., 2004). Isolated Sox2+ adult NPCs can be propagated in culture while maintaining their ability to differentiate into neurons, astrocytes and oligodendrocytes, thus documenting their self-renewal and multipotency in vitro (Ellis et al., 2004). The self-renewal and differentiation capacities of Sox2+ adult NPCs were verified in vivo by Fred Gage’s group using lenti- and retroviral mediated fate mapping approaches (Suh et al., 2007). Support for a functional role of Sox2 in NPCs came from knockdown experiments in vitro (Cavallaro et al., 2008) and conditional deletion of Sox2 specifically in the brain (Favaro et al., 2009; Ferri et al., 2004). These experiments revealed that Sox2 depletion in cultured NPCs attenuates their potential to form neurons whereas its absence in vivo causes a rapid loss of GFAP/Nestin-expressing stem/precursor cells and a decline in cell proliferation in the dentate gyrus, indicating that Sox2 marks and maintains NPCs and hence neurogenesis in the adult mouse hippocampus. Together, these studies demonstrate that Sox2 regulates both developmental and adult stem cell populations in the brain.

Sox2 marks stem and progenitor cell populations in other adult tissues that depend on Sox2 expression during development. For example, Sox2+ cells have been detected in progenitors of the adult retina (Taranova et al., 2006), trachea (Que et al., 2009), tongue epithelium (Okubo et al., 2009) and dermal papilla of the hair follicle (Biernaskie et al., 2009; Driskell et al., 2009) as well as in putative progenitors of the pituitary gland (Fauquier et al., 2008). More recently, lineage tracing experiments from our lab and others have demonstrated that immature Sox2+ cells in the adult testes, forestomach, glandular stomach, trachea, anus, cervix, esophagus, lens and dental epithelium give rise to all mature cell types within these tissues (Arnold et al., 2011). Conditional Sox2 deletion in all tracheal cells has further shown that postnatal expression of Sox2 is required to sustain tracheal homeostasis by controlling the number of proliferating epithelial cells as well as the proportion of basal, ciliated and Clara cells. The effect of Sox2 loss on tracheal cell proliferation thus represents an interesting difference compared with Sox2 loss in the embryonic trachea which does not perturb proliferation (Que et al., 2009). Deletion of Sox2 specifically in bronchiolar Clara cells, which serve as facultative stem cells, also causes reduced cell proliferation and a gradual loss of differentiation markers for Clara, ciliated and mucous cells (Tompkins et al., 2009). This loss indicates that Sox2 is required for the self-renewal of Clara cells and their differentiation into ciliated and mucous cells. From a molecular viewpoint, compromised bronchiolar cell proliferation might result from a derepression of the Sox2 target gene Smad3, thus possibly activating the anti-proliferative Tgf-β pathway (Tompkins et al., 2009). An important question that remains to be determined is whether Sox2 expression is required for homeostasis in other Sox2+ adult tissues besides the airways and the brain.

In addition to maintaining tissue homeostasis, Sox2 is involved in tissue repair. For instance, chemically induced damage of the tracheal epithelium in mice is typically repaired within 7-10 days due to the activity of basal stem cells (Que et al., 2009). Sox2-deficient trachea, however, fail to undergo efficient tissue repair with severe reductions in the number of basal, ciliated and Clara cells. Peripheral nerve regeneration is another example for Sox2’s role in tissue repair. Upon injury, mature adult Schwann cells re-express Sox2, shed their myelin sheaths and dedifferentiate to a progenitor cell-like state (Parrinello et al., 2010). Sox2 seems to play a direct role in this process by organizing Schwann cell clustering, a key event during nerve regeneration, through relocating N-Cadherin molecules (Parrinello et al., 2010). This process then enables Schwann cells to form multicellular cords to guide axon re-growth across the site of injury. It should be interesting to determine whether Sox2 is reactivated and plays functional roles in other tissues experiencing cellular damage by promoting dedifferentiation into, or expansion of, resident progenitors.

Sox2 and disease

Sox2 deficiency in developmental disorders

SOX2 mutations have been identified in a number of developmental diseases and cancer. For example, humans carrying a heterozygous mutation for SOX2 develop Anophthalmia-Esophageal-Genital Syndrome (AEG). These patients have abnormalities in ectodermal and endodermal tissues including microphthalmia (small eyes), trachea-esophageal fistula, hearing loss, and brain abnormalities (Kelberman et al., 2006; Williamson et al., 2006). The heterozygous manifestation of disease in patients is consistent with the dose-dependent functions of Sox2 seen in mice. Surprisingly, however, heterozygous mutant mice are comparatively normal although they exhibit reduced pituitary size and hormone production as well as testicular atrophy and infertility with age, possibly from dose-dependent effects on pituitary and germ cell progenitors (Kelberman et al., 2006).

Sox2 dysregulation in cancer

Accumulating evidence suggests that SOX2 acts as an oncogene in some epithelial cancers. The SOX2 locus is amplified in human squamous cell carcinomas of the lung (23%) and esophagus (15%) as well as in 27% of human small cell lung cancers analyzed (Bass et al., 2009; Rudin et al., 2012). Consistently, overexpression of Sox2 in the lungs of mice induces rapid hyperproliferation (Tompkins et al., 2011) and, in some cases, adenocarcinomas (Lu et al., 2010), although SOX2 amplifications have not yet been described in human lung adenocarcinomas. While the molecular function that Sox2 plays in tumorigenesis remains to be determined, recent evidence points towards pro-proliferative, pro-survial and/or anti-differentiation roles. For instance, knockdown of SOX2 in human cell lines, derived from squamous cell caricinomas and small cell lung cancer, compromises growth (Bass et al., 2009; Rudin et al., 2012). Moreover, genetic reduction of Sox2 levels by half in an animal model of pituitary cancer significantly reduces tumor formation(Li et al., 2012) Lastly, Sox2 was shown to be critical for the proliferation and differentiation of human osteosarcoma cell lines in vitro and in an in vivo transplantation model by antagonizing WNT signaling. SOX2 expression has also been suggested to contribute to cellular invasion in tumors of neural and neural crest origin such as glioma (Ikushima et al., 2009), melanoma (Laga et al., 2010), and Merkel cell carcinoma (Laga et al., 2010), where it is overexpressed. Thus, analogous to its multiple roles in development and differentiation, Sox2 appears to function at various levels of carcinogenesis to promote tumor growth.

An important question is whether Sox2 is already expressed in the cell of origin for these tumors or whether it is activated ectopically. While it is plausible that tumors forming within Sox2+ tissues originate from a Sox2+ cell type (e.g., lungs, esophagus, neural cells, Merkel cells), unequivocal (genetic lineage tracing) evidence for this conclusion is lacking. Interestingly, two reports detected ectopic Sox2 expression in rare tumor stem cell-like populations isolated from genetically induced mouse models of squamous cell carcinoma of the skin (Beck et al., 2011; Schober and Fuchs, 2011). Even though skin epidermis (ectoderm) has a similar structure as the Sox2+ squamous epithelia of the gastrointestinal tract (endoderm), Sox2 is not normally expressed in skin keratinocytes. It remains to be tested whether ectopic Sox2 expression has any functional consequences on these tumors. It is further interesting to note in this context that the ectopic expression of the reprogramming gene Oct4 in mice results in rapid but reversible tumor formation in several Sox2-expressing squamous epithelia by expanding adult progenitors and preventing their differentiation (Hochedlinger et al., 2005). It is therefore conceivable that Oct4 and Sox2 cooperate in these tissues, like in pluripotent cells, to induce tissue hyperplasia.

Involvement of other Sox factors in stem cell biology

Like Sox2, the SoxE group member Sox9 is expressed in several endoderm-derived and ectoderm-derived tissues. For example, Sox9 marks stem/progenitor cells in the adult intestine, liver and exocrine pancreas that produce a continuous supply of enterocytes, hepatocytes, and acinar cells, respectively, under both homeostatic and certain injury conditions (Furuyama et al., 2011). In addition to these endodermal tissues, Sox9 functions to maintain stem/progenitor cells in ectodermal tissue stem cells including the hair follicles of the adult skin (Nowak et al., 2008), multi-potent mouse retinal progenitor cells (Poche et al., 2008), NPCs (Scott et al., 2010), neural crest stem cells (Cheung and Briscoe, 2003) and mammary stem cells (Guo et al., 2012). Furthermore, Sox9 is upregulated in a number of neural tumors and basal cell carcinomas (Kordes and Hagel, 2006; Miller et al., 2006; Nowak et al., 2008), and expression of Sox9 promotes the tumorigenic and metastasis-seeding abilities of human breast cancer cells in a transplant model (Guo et al., 2012), raising the interesting possibility that Sox9 confers stem cell-like properties upon tumor cells.

Other Sox factors have been implicated in stem cell maintenance, which will not be covered in detail in this review (Table 1). Briefly, expression of Sox10, which like Sox9 is a member of the SoxE group, ensures stem cell survival, maintains multipotency and suppresses neuronal differentiation in neural crest stem cells (Kim et al., 2003). The SoxF group member Sox17 is required for the maintenance of fetal and neonatal hematopoietic stem cells (HSCs) but is dispensable in adult hematopoiesis (Kim et al., 2007). The SoxE group member Sox8 and the SoxG group member Sox15 mark muscle satellite cells, and their individual overexpression in a myoblast cell line prevents MyoD expression and differentiation into myotubes (Meeson et al., 2007; Schmidt et al., 2003). Individual knockout mice for Sox8 and Sox15 do not have an overt phenotype, suggesting redundancy. However, Sox15 mutant mice exhibit defects in muscle regeneration, indicating a requirement for Sox15 following injury (Meeson et al., 2007). Finally, the SoxB1 group member Sox3 marks undifferentiated spermatogonia, and its depletion leads to loss of spermatogenesis and nearly agametic male mice (Raverot et al., 2005).

A few general conclusions can be drawn from these and other studies examining different Sox genes in stem/progenitor cells. First, most Sox factors are expressed in multiple types of stem/progenitor cell types or tissues. Second, many Sox factors act redundantly in the maintenance of stem cells (e.g., Sox1, Sox2, Sox3, Sox9 in NPCs), which may explain why certain Sox gene knockouts do not exhibit obvious phenotypes due to compensation by other Sox factors. Third, different Sox factors may be expressed at subsequent stages of differentiation within a cell lineage (e.g., Sox2/Sox3-Sox11 during NPC differentiation or Sox9-Sox10 during neural crest stem cell differentiation)(Bergsland et al., 2011; Guth and Wegner, 2008). Lastly, Sox factors may be expressed in complementary patterns within a developing or adult tissue (e.g., Sox9/Sox2 in multipotent distal tip cells/proximal epithelial cells in the developing lung)(Rawlins, 2011). The broad expression patterns and the partial redundancy of many Sox factors are thought to be the consequence of subfunctionalization and neofunctionalization of Sox genes resulting from an expansion of Sox genes during vertebrate evolution (Guth and Wegner, 2008).

Sox factors in cellular reprogramming

Given that Sox factors play critical roles in establishing and maintaining cell types during development and in the adult, it is conceivable that their ectopic expression in heterologous cell types is sufficient to change cell fates. Indeed, Sox2 is one of the key reprogramming factors for the derivation of induced pluripotent stem cells (iPSCs) from somatic cells. Sox2 is required towards the end of reprogramming (Chen et al., 2008), presumably by activating its own transcription as well as hundreds of pluripotency-associated targets to stabilize the pluripotent state. In fact, a recent study by Jaenisch and colleagues suggested that activation of the endogenous Sox2 locus during cellular reprogramming initiates a cascade of transcriptional events that takes place exclusively in cells destined to form iPSCs (Buganim et al., 2012). Notably, as a reprogramming factor, Sox2 can be replaced by the most closely related Sox family members, Sox1 and Sox3, but not by more distant members Sox7, Sox15, Sox17 or Sox18 (Nakagawa et al., 2008). The finding that certain Sox factors cannot replace Sox2 despite similar DNA binding characteristics, might result from the differential abilities of Sox factors to interact with Oct4 to activate common target genes. Indeed, single amino acid substitutions within the Oct4 domain that normally interacts with Sox2 in ESCs can abrogate its ability to generate iPSCs (Jauch et al., 2011). Conversely, introducing Sox17-compatible amino acid changes into this Oct4 domain generates a variant that no longer recognizes Sox2 and instead endows Sox17 with the potential to induce pluripotency in combination with Oct4, Klf4 and c-Myc (Jauch et al., 2011). This experiment corroborates the notion that the binding partners of Sox2 often confer target gene specificity, resulting in the activation of different gene expression programs in cells that express the same Sox factors.

Surprisingly, Sox2 is dispensable for pluripotency gene activation in somatic cells following cell fusion with ESCs. This nonessential role contrasts with a requirement for Oct4 during cell fusion-mediated reprogramming and might suggest that Oct4 can compensate for the loss of Sox2 in this context, similar to what was seen in self-renewing ESCs (Masui et al., 2007). Alternatively, Sox15, which is also expressed in ESCs (Maruyama et al., 2005) might replace Sox2 exclusively during fusion-induced reprogramming.

Sox2 expression alone or in combination with different neural transcription factors has been reported to directly reprogram fibroblasts into neural stem cells (Han et al., 2012; Ring et al., 2012; Thier et al., 2012), suggesting that Sox2 can induce different cell fates depending on the presence of co-factors and environmental cues. This notion is in agreement with an earlier finding by Kondo and Raff, who discovered that exposure of rat oligodendrocyte progenitors to PDGF and bFGF induced their reversion into self-renewing multipotent NPC-like cells capable of giving rise to astrocytes, oligodendriocytes and neurons (Kondo and Raff, 2000; Kondo and Raff, 2004). Interestingly, the authors showed that this growth factor-mediated reversion depends on the reactivation of the Sox2 locus through a mechanism that involves direct recruitment of the chromatin remodeling factor Brahma and the tumor suppressor Brca1 to its promoter region (Kondo/Raff, G&D).

The ability to induce new cell states from heterologous cell types has recently been demonstrated for other Sox family members including Sox9 and Sox17. Specifically, co-expression of Sox9 and Slug in differentiated luminal cells produces induced multipotent cells which have long-term mammary gland reconstituting potential in transplantation assays (Guo et al., 2012). Similarly, forced expression of the fetal hematopoietic stem cell (HSC) transcription factor Sox17 in adult committed progenitors endows them with fetal HSC characteristics including an enhanced self-renewal potential, long-term multilineage reconstitution ability and biased erythroid and myeloid differentiation over lymphoid differentiation, although prolonged overexpression causes leukemia (He et al., 2011). Together, these findings underscore the powerful effects Sox factors have in endowing differentiated cells with immature stem cell-like properties (summarized in Figure 3).

Figure 3. Sox factors as inducers of cellular reprogramming.

Examples of Sox factors whose enforced expression in other cell types induces dedifferentiation. (A) Ectopic expression of Sox2 in combination with Klf4, Oct4 and c-Myc endows somatic cells with pluripotency, giving rise to induced pluripotent stem cells (iPSCs). (B) Sox2 expression alone, or together with other factors reprograms fibroblasts into induced neural stem cells (iNSCs). (C) Sox9 expression in differentiated luminal cells generates luminal progenitors, and in combination with Slug expression, converts them into mammary stem cells capable of generating an entire mammary ductal tree when transplanted into a mammary fat pad. (D) Sox17 expression in adult hematopoietic stem and progenitor cells induces a fetal-like hematopoietic stem cell state. These cells have increased self-renewal potential and express HSC markers. However, long term Sox17 expression in the adult leads to leukemogenesis.

Mechanisms by which Sox2 controls cell fate decisions

Sox2 expression, like that of many other Sox factors, is modulated by extracellular signals and intracellular cofactors. Here, we review examples of how Sox2 expression can be positively or negatively regulated by different extracellular cues in different tissues and discuss intracellular mechanisms by which Sox2 expression is controlled in pluripotent and adult stem cells (Figure 4).

4. Mechanisms by which Sox2 controls self-renewal and differentiation in pluripotent and multipotent stem cells.

(A) Sox2 activates self-renewal genes and represses differentiation genes in a cell type-specific manner by (i) interpreting tissue-specific signals and (ii) interacting with other cell type-specific cofactors. For example, in ESCs Sox2 occupies many targets containing Oct4-Sox2 consensus sequences and partners with downstream effectors of ESC-specific signaling pathways including Stat3 (LIF pathway). In NPCs, Sox2 occupies target genes that also contain binding sites for the brain-specific factors Brn2 and Chd7, thus activating different sets of genes. In addition, Sox2 activates its own transcription and regulates components of the signaling pathways that control self-renewal, thereby promoting maintenance of the undifferentiated state. (B) In addition to activating self-renewal genes and suppressing lineage-specific genes, Sox2 acts as a pioneer factor to prime stem cells for subsequent gene activation. Sox2 occupies silent NPC genes in ESCs, which carry bivalent domains poised for gene activation. Upon differentiation into NPCs, Sox2 and Sox3 cooperate to activate self-renewal genes while keeping neuronal differentiation genes in a silent but bivalent state. When NPCs undergo terminal differentiation, Sox2 and Sox3 disengage from neuronal-specific enhancers and are replaced by Sox11.

Extracellular regulators of Sox2 expression

Sox2 expression is positively and negatively influenced by different extracellular signals in vivo and in vitro. For instance, Fgf signaling from the surrounding ventral mesenchyme negatively regulates Sox2 expression during embryonic foregut patterning, resulting in a separation of esophagus and trachea (Que et al., 2007). In the developing taste buds, Wnt signaling induces Sox2 expression in endodermal progenitors, causing their differentiation into taste bud cells at the expense of keratinocytes (Okubo et al., 2006). In calvarial osteoblast progenitors, however, Sox2 is positively regulated by Fgf signaling. Up regulation of Sox2, in turn inhibits Wnt signaling by means of physical association of Sox2 with beta-catenin (Mansukhani et al., 2005).

In cultured pluripotent ESCs, Sox2 targets are co-occupied by Smad1 and Smad3 proteins, the downstream effectors of Tgf-β signaling that is essential for self-renewal (Chen et al., 2008; Mullen et al., 2011). Notably, one of the genes targeted by Oct4, Sox2 and Smad3 is the Tgf-β inhibitor Lefty1, indicating that tight regulation of this pathway is necessary to maintain pluripotency. Similar to Tgf-β signaling in ESCs, Egf and Shh signaling stimulate Sox2 expression in NPCs (Favaro et al., 2009). Once activated, Sox2 binds to the Egfr and Shh genes amongst many other targets, thus engaging in positive feedback loops that are important for the maintenance of stem/progenitor cells (Engelen et al., 2011; Hu et al., 2010). In agreement with this molecular link, Sox2-deficient NPCs fail to produce sufficient Shh, leading to loss of NPC cultures and dentate gyrus hypoplasia, respectively (Favaro et al., 2009). Remarkably, these phenotypes can be partially restored in vitro and in vivo by supplying recombinant Shh or an Shh agonist (Favaro et al., 2009). A similar connection has been observed between Shh and Sox9 in NPCs (Scott et al., 2010). In contrast to Shh and Egf signaling, which promote Sox2 expression, thyroid hormone signaling induces differentiation of neural progenitors into neuroblasts by suppressing Sox2 expression (Lopez-Juarez et al., 2012). Specifically, thyroid receptor-alpha1 binds to a negative thyroid hormone response element within the Sox2 enhancer, resulting in Sox2 repression in a hormone-dependent fashion. Finally, Ephrin signaling causes Sox2 stabilization during Schwann cell regeneration, leading to N-Cadherin remodeling and subsequent Schwann cell clustering (Parrinello et al., 2010). In summary, these and several other examples (Takemoto et al., Wnt/Fgf Dev 2005, Domyan et al., Bmp Dev, 2010) demonstrate that major signaling pathways can positively or negatively control Sox2 expression levels during embryonic development, stem cell homeostasis and tissue regeneration in a context-dependent manner. Furthermore, Sox2 itself often modulates these signals by directly activating or repressing key regulators of these pathways.

Intracellular modulators of Sox2 expression in pluripotent stem cells

Once Sox2 is activated by extracellular signals, intracellular co-factors ensure that the proper set of target genes is activated in a cell-type specific fashion. One way to achieve this is to collaborate with other cell type-specific transcription factors. As discussed earlier, Sox2 physically associates with and co-occupies targets with other key pluripotency factors including Oct4 and Nanog in ESCs, thus contributing to target gene specificity. Of note, the combination and complexity of these pluripotency transcription factors at individual targets determines whether they will be activated or repressed. That is, targets bound by one or few transcription factors tend to be repressed whereas targets occupied by multiple factors tend to be expressed in ESCs (Kim et al., 2008; Sridharan et al., 2009). To ensure maintenance of the undifferentiated state of ESCs, Sox2 as well as other pluripotency factors engage in auto-regulatory loops to boost their own expression (Boyer et al., 2005).

The observation that ectopic expression of Oct4 and Sox2 alone are insufficient to activate the well-known target gene Nanog in a cell-free system (Fong et al., 2011) motivated efforts to identify additional cofactors. Tjian and colleagues employed an elegant biochemical approach to purify the “stem cell co-activation complex” (SCC) that collaborates with Oct4 and Sox2 to transcriptionally activate the Nanog promoter. SCC components also occupy hundreds of other Oct4/Sox2 targets in ESCs as determined by ChIP-seq analysis (Fong et al., 2011). The SSC complex contains the trimeric XPC-nucleotide excision repair complex, and is thought to act as a molecular link that couples stem cell-specific gene expression programs with genome surveillance and stability in ESCs. Interestingly, the tumor suppressor protein p53 has recently been implicated in a similar role in ESCs. However, unlike the SCC complex, p53 binds to the distal enhancers of ESC specific genes including Sox2, causing their repression upon DNA damage (Li et al., 2012).

In ESCs, Sox2 additionally requires binding of chromatin modifiers to induce expression of pluripotency-associated targets and repression of differentiation-associated targets. For example, Sox2, Oct4 and Nanog cooperate with WD repeat domain 5 (Wdr5), an effector of activating H3K4 methylation, to maintain robust expression of self-renewal genes in ESCs (Ang et al., 2011). Active Sox2 targets are also co-bound by components of the cohesion and mediator complex responsible for bridging enhancer and promoter elements to ensure efficient gene expression (Kagey et al., 2010). Recent evidence suggests that Sox2 might even interact with ESC specific long non-coding RNAs (lncRNAs) (Ng et al., 2012) to silence differentiation-associated genes in self-renewing ESCs.

During ESC differentiation, ESC-associated genes need to be rapidly downregulated, which is again achieved by multiple mechanisms. For example, the H3K4/K9 demethylase Lsd1 and HDACs1/2 silence active Oct4/Sox2-occupied enhancers in ESCs (Whyte et al., 2012). Recent evidence further documents an unanticipated role for cell cycle inhibitors in transcriptional suppression of stem cell genes. The cell cycle dependent kinase inhibitor p27, which is rapidly activated as cells differentiate and thus exit the cell cycle, directly binds to and inhibits Sox2’s SRR2 enhancer (Li et al., 2012). In parallel with these transcriptional and epigenetic mechanisms, negative feedback loops kick in during differentiation that shut down the pluripotency program at the posttranscriptional level. This is exemplified by RNA miR-145, which is normally repressed by OCT4 in ESCs, and becomes activated to target OCT4, SOX2 and KLF4 RNAs for degradation when ESCs differentiate (Xu et al., 2009). Thus, Sox2 interacts at the genic, transcript and protein levels with other core pluripotency factors, DNA repair complexes, cell cycle regulators, miRNAs, activating and repressive chromatin regulators to control specific gene expression programs that balance the decision between self-renewal and differentiation in pluripotent cells.

Intracellular modulators of Sox2 expression in adult stem cells

Similar to ESCs, Sox2 induces the expression of self-renewal pathways and inhibits the expression of differentiation genes in NPCs. Because Oct4 and other pluripotency-associated genes are silenced in NPCs, Sox2 partners with different transcription factors to activate alternative targets. For example, Sox2 has been shown to interact with the brain-specific POU factor Brn2 to activate the neural progenitor-associated Nestin gene in early neural progenitors (Tanaka et al., 2004). More recently, the chromatin remodeling ATPase Chd7, which has been associated with CHARGE syndrome, was shown to physically interact and co-occupy targets with Sox2 in NPCs (Engelen et al., 2011). Sox2 and Chd7 co-regulate a set of target genes of the Notch and Shh signaling pathways important for stem cell self-renewal. The nuclear receptor tailess (TLX) has been identified as another key target of Sox2 in NPCs. TLX functions as a transcriptional repressor that is important for NPC maintenance and neurogenesis in adult mice. Sox2 physically interacts with TLX and forms complexes on DNA, possibly to suppress differentiation genes (Shimozaki et al., 2011).

Sox2 expression itself is maintained in NPCs by direct transcriptional activation through Ars2, a zinc finger protein typically involved in miRNA biogenesis (Andreu-Agullo et al., 2012). Chromatin immunoprecipiation experiments have shown that Ars2, in a miRNA pathway-indepdendent manner, binds to the promoter region of Sox2 and activates its expression. Ars2 deletion leads to a loss of NPC self-renewal and multipotency both in in vitro and in vivo. Importantly, this defect can be rescued by Sox2 overexpression (Andreu-Agullo et al., 2012). Similarly, the transcription factor myeloid Elf-1 like factor (MEF) binds to the Sox2 locus and stimulates its expression in the context of neurospheres and glioma cells (Bazzoli et al., 2012). Forced Sox2 expression also rescues the inability of MEF-/- cells to form neurospheres. In analogy to p27’s inhibition of Sox2 expression during ESC differentiation, the cell cycle dependent kinase inhibitor p21 was shown to suppress Sox2 expression during NPC differentiation(Marques-Torrejon et al., 2013)

Posttranslational modifications, such as acetylation (Baltus et al., 2009; Sikorska et al., 2008), sumoylation (Tsuruzoe et al., 2006), phosphorylation (Jeong et al., 2010) and arginine methylation (Zhao et al., 2011), have also been described to influence the transcriptional activity of Sox2 in ESCs or NPCs. In the case of Sox2, these modifications either cause transcriptional activation (phosphorylation, methylation) or repression (sumoylation, acetylation) by controlling Sox2’s stability, nuclear-cytoplasmic localization or transactivation potential. Collectively, these results demonstrate that Sox2+ adult stem cells utilize some of the same as well as different mechanisms as ESCs to control the balance between self-renewal and differentiation. It is worth mentioning that Sox2 has been shown to collaborate with additional transcription factors in the development of other tissues. We refer to an excellent review exploring the various partners of Sox proteins for greater detail (Kondoh and Kamachi, 2010).

Sox proteins as pioneer factors

Pioneer factors are transcription factors that occupy silenced target genes in progenitor cells and keep them in a poised state for activation at subsequent stages of differentiation (Zaret and Carroll, 2011). A classical example is the transcription factor FoxD3, essential for the maintenance of ESC self-renewal (Hanna et al., 2002). FoxD3 occupies the enhancer of the silent liver-specific Alb1 gene in ESCs, thereby keeping it poised for activation upon differentiation into liver cells, when FoxA1 replaces FoxD3 to activate transcription (Xu, Zaret, G&D, 2009). Recent evidence exploring the genome-wide targets of different Sox factors during neural differentiation from ESCs supports the notion that Sox factors may also function as pioneer factors and thus contribute to differentiated cell fates (Bergsland et al., 2011). In ESCs, Sox2 binds to ESC specific enhancers, which are active and carry H3K4me3 marks, as well as to neural enhancers, which are silent and carry bivalent H3K4me3/H3K27me3 marks. Upon differentiation into NPCs, Sox2 collaborates with Sox3 to relocate from pluripotent to neural specific gene enhancers. These enhancers are either active in NPCs and hence carry the H3K4me3 mark or are inactive and carry bivalent marks. After neuronal differentiation, both types of enhancers exchange their SoxB1 factors for SoxC factors, including Sox11. At the same time, previously active NPC enhancers acquire the repressive H3K27me3 mark, whereas the poised bivalent enhancers convert to a monvalent H3K4-enriched chromatin signature, resulting in gene activation.

Sox2 might also act as a pioneer factor during hematopoeisis. In ESCs, Sox2 binds to the enhancers of the repressed bivalent lambda5-VpreB1 and Pax5 genes important in pro/preB cells (Liber et al., 2010). Specifically, Sox2 has been suggested to mediate deposition of a tightly localized peak of H3K4me2 dimethylation at these enhancers. As development proceeds, Sox2 facilitates activation of these genes by targeting H3K4 di- and trimethylation to these bivalent promoters. Sox2 is subsequently replaced by the SoxC group member Sox4 in hemangioblasts, which are early progenitors for the hematopoietic and endothelial lineages, leading to robust gene activation during B cell differentiation.

Another interesting question raised by these observations is whether Sox2 along with the other transcription factors Oct4, Klf4 and c-Myc may also function as pioneer factors during iPSC generation. An examination of binding patterns of the four factors 48h after their induction in fibroblasts showed that Sox2, Oct4, Klf4 and c-Myc mostly bind to enhancers of early reprogramming genes, which are not yet activated (Soufi et al., 2012). While Sox2, Oct4 and Klf4 expression alone allow access of these targets, c-Myc expression alone does not. Thus, Sox2, Oct4 and Klf4 indeed seem to act as pioneer factors for c-Myc early in reprogramming. In addition, c-Myc expression enhances binding of OKS to their targets, thus facilitating efficient chromatin engagement. Together, these three examples expand the role of Sox2 from a transcriptional activator to a pioneer factor that poises silenced genes for expression during normal development and cellular reprogramming.

Concluding Remarks

The molecular and functional analyses of the Sox family of transcription factors over the past two decades has documented their important roles in various aspects of stem cell biology. Biochemical dissection of protein interaction partners and DNA targets using genome-wide approaches has provided a molecular explanation for the previously observed versatility of individual Sox factors in regulating proliferation and differentiation of progenitor and stem cells in different tissues and at different stages of development. Sox factors respond to different extracellular signals and interact with a host of intracellular co-factors, such as cell type-specific transcription factors and chromatin regulators, to control different sets of genes in distinct cell types. In addition, Sox factors compete with transcription factors of alternative lineages to drive different cell fates during development. At the molecular level, this is often accomplished by directly activating genes that promote their own lineage and repressing genes of alternative lineages. Interestingly, the comparison of genomic binding sites of different Sox proteins along a neural differentiation paradigm demonstrated that Sox factors do not simply serve to activate self renewal genes and repress differentiation genes but also function as pioneer factors to poise genes for activation by a related Sox factor once differentiation ensues. It should be informative to determine whether this principle also applies to Sox factors in other cellular lineages (e.g., Sox17 in hematopoietic cells or Sox9 in hair follicle cells).

Most insights into the biology of Sox factors have come from developmental studies. The finding that Sox factors are also expressed in numerous adult stem and progenitor cell populations raises interesting questions about the molecular and functional roles they play in tissue homeostasis and regeneration compared with their functions during development. The availability of appropriate mouse models and the ability to maintain rare stem cell populations in culture, combined with genome-wide technologies, should now enable researchers to address this fundamental question at the mechanistic level.

Reprogramming experiments have underscored the power of Sox factors in switching cell fates. However, the underlying mechanisms are still poorly understood. It might be possible to predict from available expression and ChIP-seq data which combinations of Sox factors, together with appropriate partners, are sufficient to generate desired cell states in culture from pluripotent or differentiated cells. Given that certain SOX genes are amplified or overexpressed in human cancer, it is intriguing to speculate that Sox factors also contribute to tumorigenesis by endowing differentiated or progenitor cells with a more primitive stem cell-like state. Indeed, studies manipulating Sox17 in hematopoiesis and Sox9 in mammary stem cells support this notion.

In summary, accumulating evidence implicates many Sox factors in pluripotent and multipotent stem cell biology and tissue regeneration. A better understanding of the mechanisms by which Sox factors induce and maintain stem cell populations should provide important insights into how tissue stem cells are generated and maintained and might lead to strategies to treat degenerative diseases or cancer affecting those tissues.