Back to basics: Sox genes


  • Julie C. Kiefer

    Corresponding author
    1. Department of Neurobiology and Anatomy, University of Utah, Salt Lake City, Utah
    • Department of Neurobiology and Anatomy, 20 North 1900 East, 401 MREB, University of Utah, Salt Lake City, UT 84132
    Search for more papers by this author


Sox genes are indispensable for multiple aspects of development. This primer briefly describes shared properties of the Sox gene family, and five well-characterized examples of vertebrate developmental mechanisms governed by Sox gene subgroups: testis development, central nervous system neurogenesis, oligodendrocyte development, chondrogenesis, and neural crest cell development. Also featured is an interview about current issues in the field with experts Jonas Muhr, Ph.D. and Robert Kelsh, Ph.D. Developmental Dynamics 236:2356–2366, 2007. © 2007 Wiley-Liss, Inc.


The male sex determination gene Sry (sex-determining region Y) was the first-identified member of the Sox gene family. It took scientists three decades of careful genotyping of sex-reversed XX men and XY women to find the elusive Y-linked gene. Finally in 1990, Sry was pinpointed to segment 1A1 and identified as the previously characterized testis-determining factor (TDF; Sinclair et al.,1990). Simultaneously, it was reported that Sry belonged to a larger gene family, later dubbed Sox (Sry-related box) genes (Gubbay et al.,1990).

Sox proteins belong to the high mobility group (HMG) superfamily, members of which bear their namesake ∼80 amino acid DNA-binding domain. The superfamily can be broken down into two subfamilies with differing characteristics. Members of the TCF/SOX/MATA family bind specific DNA sequences, whereas generally HMG/UBF proteins bind DNA nonspecifically (Soullier et al.,1999). Within the TCF/SOX/MATA family, the three gene groups are clearly separable based on phylogenetic analysis of HMG domains.

What characteristics do Sox family members share? Sox proteins are animal-specific and can be identified based on a conserved motif within the HMG domain, RPMNAFMVW (Bowles et al.,2000). Of interest, the exception is the first characterized Sox protein, Sry, which bears only a portion of the shared motif, RPMNAF. Sox proteins are transcription factors that bind the DNA sequence (A/T)(A/T)CAA(A/T)G; some are transcriptional activators, others are repressors, while others reportedly lack transactivation domains (Table 1). Sox protein depend on requisite partners for target specificity. Partners vary widely, for example Sox2 and Pax6 together activate δ-crystallin expression during lens induction, and Sox10 and Krox20 activate Connexin32 in glia; for details on Sox partners, the reader is referred to other reviews (Kamachi et al.,2000; Wilson and Koopman,2002). Like its fellow HMG subfamily members, LEF/TCF, Sox bind the minor groove of DNA and bend it at an angle. It has been proposed that this conformational change may bring distal proteins on gene promoters and enhancers closer together, allowing them to functionally interact.

Table 1. Sox Subgroups (Mus musculus)
inline image

Based on phylogenetic analysis of their HMG domains, Sox genes can be separated into subgroups A–J, A–H of which are represented in mouse and humans (Bowles et al.,2000; Table 1). Many genes within each subgroup also share conserved structural domains outside the HMG domain. Thus, subgroup B is further divided, where SoxB1 bear transactivation domains, and SoxB2 repressor domains. Sox genes regulate countless, diverse developmental events, including lens, hair follicle, gut, B-cell, muscle, and blood vessel development, to name just a few. Accordingly Sox genes are expressed in many tissues (Bowles et al.,2000), and, therefore, not surprisingly, they are implicated in the etiology of many diseases (Table 2) and certain cancers (Dong et al.,2004).

Table 2. Genetic Diseases Caused by Sox Mutationsa
SryGonadal dysgenesis, XY female type
Sox2Microphthalmia, syndromic 3 (OMIM 206900) optic nerve hypoplasia and abnormalities of the central nervous system
Sox3Mental retardation, X-linked with isolated growth hormone deficiency (OMIM 300123) infundibular hypoplasia and hypopituitarism
Sox9Campomelic dysplasia with autonomic XY sex reversal (OMIM 114290)
Sox10Waardenburgh-Shah syndrome (OMIM 277580)
 Yemenite deaf-blind hypopigmentation syndrome (OMIM 601706) peripheral demyelinating neuropathy, central dysmyelinating leukodystrophy, Waardenburgh syndrome, and Hirschprung's disease
Sox18Hypotrichosis-lymphedema-telangiectasia syndrome (OMIM 607823)

Below are summarized well-characterized examples of how Sox genes function in five developmental processes: testis, oligodendrocyte, and neural crest development, central nervous system (CNS) neurogenesis, and chondrogenesis. Although mechanisms of Sox function vary among cell types, four themes are reiterated among the examples presented here and may represent generalized functions among members of the Sox gene family: (1) Sox genes regulate specification and differentiation of many cell types, (2) genes within Sox subgroups often share functional roles, (3) genes within one subgroup can counteract the function of genes in another subgroup, and (4) the same gene can mediate different stages of development in one cell type and/or developmental processes in more than one cell type.


The mammalian embryo is initially endowed with a “bipotential gonad,” that is, gonad precursors are equally able to differentiate as testis or ovary (Koopman,2005). In the normal XY gonad primordium, the first male-specific cell to develop from bipotential precursors are Sertoli cells. These cells populate testis cords, which will mature into structures that support sperm maturation, namely seminiferous tubules. Sertoli cells orchestrate the development of many testis cell types, making them central players in male sex determination.

Sry and Sox9 Are Male Sex Determination Factors

As the testis determination factor (see the Introduction section), Sry is necessary for testis development in mammals: in its absence, XY humans develop female genitalia. Moreover, its ectopic expression is sufficient to induce testis development in XX transgenic mice (Berta et al.,1990; Koopman et al.,1991). Transiently expressed in Sertoli cell precursors, Sry jump starts male sex determination by driving their differentiation as Sertoli cells (Polanco and Koopman,2007).

Despite the undeniable significance of Sry in male gonad development, surprisingly little is known about its regulation. Natural mutations that trigger XY sex reversal in humans have helped to identify three biologically relevant mechanisms that regulate Sry function. First, amino acid substitutions in either of Sry's two nuclear localization signal (NLS) motifs impair nuclear accumulation of Sry by cytoplasmic transport factors, and Importin β (Harley et al.,2003). Second, many mutations identified in human XY sex reversal cause failure of Sry to bind and bend DNA, possibly highlighting the importance of Sry's function as an architectural factor (Pontiggia et al.,1994). Finally, in mouse models for Frasier syndrome, mice lacking the +KTS splice isoform of Wilms' tumor 1, Sry expression is drastically reduced (Hammes et al.,2001). The phenotype may be a consequence of WT1+KTS's role in general RNA metabolism. Sry protein partners and targets have not yet been identified; the venture is made difficult by the fact that the protein is expressed in a small number of cells for only a short time in mice (8 hr; Wilhelm et al.,2005).

Sox9 function mimics that of Sry, implicating it as another important player in testis determination. Sox9 is expressed in pre-Sertoli cells shortly after Sry, and like Sry, it is necessary for male sex determination and its ectopic expression induces testis development (Vidal et al.,2001; Chaboissier et al.,2004). These data are evidence that Sry's main function is to activate expression of Sox9.

A recent paper implicates Fgf9 as an important regulator of Sox9 (Kim et al.,2006). Their findings argue that in XX gonad primordia, Wnt4 ensures ovary development by repressing Fgf9, a requisite signaling molecule that maintains Sox9 expression. In XY gonad primordia where FGF9 is expressed, there is a positive feedback loop between Sox9 and Fgf9. Consistent with this idea, mice lacking Fgf9 experience XY sex reversal (Colvin et al.,2001). A model based on these findings states that, in XY cells, Sry up-regulates Sox9 expression, thus establishing an inductive loop between Sox9 and Fgf9, locking in the male determination pathway (Kim et al.,2006).

Sox8 Function Overlaps With Sox9

Sox9's fellow subgroup E member, Sox8, also participates in testis development, but to a lesser degree (Koopman,2005). Experiments where Sox9 is depleted from gonadal cells in Sox8 mutant mice at relatively late steps in gonad development reveal that the two genes have overlapping functions in testis differentiation (Chaboissier et al.,2004). However, the overall role of Sox8 in testis development is minor because mice deficient in Sox8 only, do not exhibit testis defects. These findings, along with the observation that unlike Sry, Sox9's role in testis determination is conserved in vertebrates, suggest that Sox9 wears the pants in the Sox family.


Neural cells, like most cell types, undergo a long journey from progenitor to differentiated cell. Self-renewing progenitor cells are housed in the ventricular zone of the developing spinal cord. During neurogenesis, these cells down-regulate progenitor cell markers, exit the cell cycle, migrate to the marginal zone, and begin to express differentiation markers. Proneural basic helix–loop–helix (bHLH) transcription factors are well-defined regulators of neurogenesis (Kiefer,2005). Ngn1, Ngn2, and Mash1 activate cell cycle exit and a subsequent differentiation program that will lead to expression of pan-neuronal markers. In certain instances, proneural factors can also direct cell type-specific differentiation. The finding that Sox genes are expressed in neural progenitors, neuroblasts, and differentiated neurons suggested that they might also be regulators of neurogenesis.

SoxB1 Maintain Neural Progenitors

Until recently, little was known about how neural progenitors are prevented from differentiating until the appropriate time. Provocatively, SoxB1 genes (Sox1-3) are expressed in cells that are competent to become neurons in the CNS and are generally down-regulated upon differentiation (Bylund et al.,2003). This observation led researchers to believe that SoxB1 may prevent differentiation of the neural progenitor population.

Several key experiments demonstrate that SoxB1 genes maintain the neural progenitor state. First, constitutive expression of Sox1-3 in neural progenitors prevents their differentiation and suspends them in a proliferative, stem cell-like state (Bylund et al.,2003; Graham et al.,2003). In a second set of experiments, the HMG domains of Sox2 and 3 were fused to the repressor domain from the Drosophila protein Engrailed (EnR), converting them to transcriptional repressors. Forced expression of the fusion proteins prompts precocious cell cycle exit and expression of differentiation markers. Third, co-overexpression of Sox3 and the proneural factor Ngn2 blocks Ngn2's ability to promote premature differentiation. Together, these experiments support a model whereby SoxB1 genes maintain neural progenitors in a proliferative state, at least in part by preventing proneural factor activities.

Sox21 Represses SoxB1

Expression of the SoxB2 gene Sox21 overlaps with both the progenitor marker Sox3 and the neuroblast marker Ngn2, implicating a role in progenitor cell differentiation (Sandberg et al.,2005). This idea was tested by determining the consequences of overexpressing and inhibiting Sox21 in neural progenitors within the developing chick neural tube. Ectopic expression of Sox21 in neuroblasts induces down-regulation of progenitor markers, cell cycle exit, and neuronal differentiation (Sandberg et al.,2005). Blocking endogenous Sox21 inhibits differentiation of Ngn2+ progenitors, placing the proneural factor upstream of Sox21. Based on the fact that Sox21 is a known transcriptional repressor, the authors postulated that the protein promotes differentiation by inhibiting progenitor maintenance factors. Indeed, upon reaching a critical expression threshold, Sox21 inhibits Sox3 activity, thus substantiating their hypothesis. Together, this work shows that proneural factors up-regulate expression of Sox21; when a certain expression threshold is reached, Sox21 represses Sox1-3 activity, creating an environment permissive for differentiation. The transition from progenitor to neuron may be controlled by direct repression of Sox1-3 targets by Sox21.

SoxC Endow Neuroblasts With Pan-Neuronal Characteristics

Sox genes regulate yet another aspect of neurogenesis, acquisition of pan-neuronal characteristics. SoxC group genes, Sox4 and Sox11, are expressed in committed, postmitotic neuroblasts, suggesting that they regulate the latter steps of neural development (Bergsland et al.,2006). This notion was confirmed, but their specific function was unexpected. Overexpression of SoxC activates expression of pan-neuronal markers, Tuj1 and MAP2, the result of direct activation, at least in the case of Tubb3 (Tuj1). Surprisingly, despite activation of neuronal genes, the cells do not differentiate. Rather, they continue to proliferate and express the progenitor marker Sox3. These results show that acquisition of pan-neuronal characteristics can be uncoupled from cell cycle exit and differentiation. However, blocking SoxC inhibits Ngn2's ability to induce differentiation. Thus, downstream of proneural factors, SoxC directly activates early genes that endow cells with general neuronal properties, a step that is necessary but not sufficient for cell type-specific differentiation.

SoxB1 Regulate Cell Type-Specific Differentiation

In addition to regulating progression of neurogenesis, Sox genes also function in postmitiotic neurons. In the adult, SoxB1 genes are expressed in mostly nonoverlapping subsets of postmitotic neurons, controlling neuronal subtype-specific differentiation and/or survival. Sox1 is required for terminal differentiation of lens fibers, and functions postmitotically to specify differentiation of ventral striatum neurons (Nishiguchi et al.,1998; Ekonomou et al.,2005). In Sox2 deficient mice, neurons in the thalamus, striatum, and septum that normally express the gene undergo degeneration, suggesting a maintenance role (Ferri et al.,2004). Sox3 is highly expressed in the ventral diencephalon and its absence is thought to result in either cell death, or defects in neuronal activity, possibly accounting for mental retardation observed in patients with Sox3 mutations (Laumonnier et al.,2002; Rizzoti et al.,2004). Although SoxB1 genes play overlapping roles in progenitor cells, they acquire distinct functions upon differentiation.


CNS glia are derived from the ventricular zone (VZ) of the neural tube. The VZ is divided into several domains, which each give rise to different cell types. Initially, VZ progenitors give birth to neurons, then undergo a transition after which time they make glia instead. Motor neurons and V2 interneurons originate from the pMN and p2 regions, respectively (Kessaris et al.,2001). After the neuron to glia fate switch during mouse midgestation stages, pMN precursors yield myelin-forming oligodendrocytes, and p2 make astrocytes, which provide support, ionic homeostasis, survival signals, and nutrients to neurons. Before the discovery of Sox genes, little was understood about the molecular mechanisms behind the neuron to glia fate switch and glia differentiation.

SoxE Regulate Specification and Differentiation

Oligodendrocyte specification and differentiation are dependent on SoxE group genes, most notably, Sox9 and 10. Sox9 expression in glia precursors precedes Sox10 in mice, and whereas Sox9 expression is extinguished, Sox10 continues to be expressed in differentiating oligodendrocytes and astrocytes (Britsch et al.,2001b; Stolt et al.,2002,2003). Examination of mouse mutants confirms that Sox9 functions at earlier stages than Sox10. When Sox9 is ablated from the spinal cord only, numbers of oligodendrocyte progenitors are vastly reduced. Specifically, these mice show defects in the neuron to glia fate switch. This finding is manifested by generation of excess motoneurons and V2 interneurons at the expense of differentiated oligodendrocytes and gray matter astrocytes, respectively. By contrast, Sox10 knockout mice display no significant difference in the number of progenitors, but have far fewer differentiated oligodendrocytes than wild-type (Britsch et al.,2001b; Stolt et al.,2002). Astrocyte differentiation was not assayed. Curiously, reduction of progenitor and/or differentiating oligodendrocytes in Sox9 and 10 mutant mice were transient, and both cell types eventually recovered to near-normal levels. Because expression of the third SoxE gene, Sox8, overlaps with both Sox9 and 10, it was suspected that Sox8 might partially compensate for loss of the other E group genes.

Despite Sox8 expression in progenitors and differentiated oligodendrocytes, Sox8 mutant mice display no apparent glial defects (Stolt et al.,2004). However, when loss of Sox8 was combined with loss of Sox9 in the spinal cord, oligodendrocyte progenitors were almost completely and permanently obliterated (Stolt et al.,2005). Similarly, loss of Sox8 combined with heterozygous loss of Sox10 caused a severe loss of differentiated oligodendrocytes that was neither apparent in the Sox8-deficient mice nor the Sox10 heterozygous mice (Stolt et al.,2004). The phenotypes reveal similar, but nonequivocal roles for Sox8 and 9 in neuron to glia fate transitions, and for Sox8 and 10 in oligodendrocyte differentiation. How might the SoxE genes work together? Like Sox10, Sox8 can directly bind the Mbp promoter and activates its transcription in vitro. Sox8 and 10 also form heterodimers in vitro (physical interactions between Sox8 and 9 were not tested), raising the possibility that oligodendrocyte differentiation may be mediated by cooperative interactions between the two proteins.

SoxD Negatively Regulate SoxE

Another facet of oligodendrocyte development is negative regulation of SoxE group genes by the SoxD genes Sox5 and 6. Expression of the two D group genes initiates slightly before the onset of oligodendrocyte specification, is maintained in oligodendrocyte progenitors, and down-regulated in terminally differentiating cells (Stolt et al.,2006). Mice deficient in Sox5 or Sox6 exhibit precocious specification of VZ cells to oligodendrocyte progenitors at mid-embryogenesis, a phenotype that is even more pronounced in Sox5 and 6 double mutants. Sox5 and Sox6 double mutants additionally display precocious terminal differentiation of oligodendrocyte progenitors at late stages of embryogenesis. Generation of mutants in which Sox6 elimination in the spinal cord was confined to oligodendrocyte progenitors show that SoxD gene effects on oligodendrocyte development are cell-intrinsic. Gel shift assays reveal a potential mechanism of action for SoxD, which are thought to lack transactivation domains (Lefebvre et al.,1998). Sox5 binds to the same sites in myelin gene promoters that are also recognized by Sox10. Results suggest that competition for the same binding sites and SoxD-dependent recruitment of co-repressors to myelin gene promoters prevent transcriptional activation by Sox10. SoxD proteins, thus, regulate the timing of oligodendrocyte development by inhibiting SoxE function.


Whereas flat bones of the skull and clavicle form by intramembranous ossification, most bones of the vertebrate skeleton develop by endochondral ossification, a process that uses a cartilage intermediate (de Crombrugghe et al.,2001). Cartilage and bone progenitors, chondroblasts and osteoblasts, are derived from a common precursor, osteochondroprogenitors. Endochondral ossification initiates when osteochondroprogenitors aggregate to form mesenchymal condensations. Within condensations, precursors differentiate as chondrocytes and deposit a cartilage-specific extracellular matrix. They then proliferate unidirectionally making orderly columns, exit the cell cycle, undergo hypertrophy, and die. Osteoblasts develop from the condensed mesenchymal layer surrounding the cartilaginous matrix and use the matrix as a template to build bone.

Sox9 Is a Master Regulator of Cartilage Development

Sox9 is a master regulator of chondrogenesis and regulates multiple stages of cartilage development. Overexpression of Sox9 drives chondrogenesis in mesenchymal cells, in part by directly activating transcription of genes required for cartilage development including extracellular matrix genes Col11a2 and Col2a1, and Aggrecan (Bridgewater et al.,1998; Sekiya et al.,2000; Kawakami et al.,2005). Strikingly, when Sox9 is eliminated from undifferentiated mesenchyme in limb buds, mesenchymal condensations, osteoblasts, and mature cartilage and bone fail to develop, suggesting that Sox9 is required for specification of osteochondroprogenitors. Conditional elimination of Sox9 from limb buds after mesenchymal condensation shows that the gene is later required for chondrocyte differentiation and inhibiting hypertrophy in prehypotrophic chondrocytes (Akiyama et al.,2002). Consistent with a continuing role in chondrogenesis, Sox9 expression initiates in osteochondroprogenitors and is maintained until chondrocytes experience hypertrophy (Wright et al.,1995; Akiyama et al.,2005).

Sox9 ensures commitment to the chondrocyte lineage by inhibiting osteoblast developmental regulators. One such regulator is β-catenin, a nuclear effector of canonical Wnt signaling (Day et al.,2005). Inactivation of β-catenin in head and limb mesenchyme causes failure of osteoblast differentiation, with some precursors trans-differentiating as chondrocytes (Hill et al.,2005). These results show that β-catenin is required for osteoblast development and that it also suppresses the chondrocyte fate. In line with β-catenin's opposing effects on bone and cartilage development, there is an antagonistic relationship between β-catenin and Sox9. Expression of a stable form of β-catenin inhibits chondrocyte differentiation, mimicking loss of Sox9 (Akiyama et al.,2004). Conversely, a conditional deletion of β-catenin in chondrocytes mimics overexpression of Sox9: endochondral bone formation is delayed. β-catenin also opposes Sox9's role in preventing the transition of chondrocytes to hypertrophy.

There is a mechanistic explanation for Sox9 and β-catenin's mutual antagonism. Sox9 and the β-catenin co-factors, LEF/TCF transcription factors, bind to β-catenin at the same protein domain. Therefore, when Sox9 binds β-catenin, it impedes LEF/TCF binding, prohibiting activation of Wnt/β-catenin target genes. Likewise, β-catenin binds the C-terminal transactivation domain of Sox9, possibly inhibiting Sox9's transactivation abilities, and triggering mutual degradation of both proteins. Relative levels of β-catenin and Sox9 may help control target activation important for bone and cartilage development. Consistent with this idea, chondrogenesis is sensitive to the correct dosage of Sox9. In humans, Sox9 haploinsufficiency causes campomelic dysplasia (OMIM 114290;, patients of which exhibit skeletal abnormalities, as well as other phenotypes (Wagner et al.,1994).

Sox9 inhibits another regulator of osteoblast development, the transcription factor Runx2 (Schroeder et al.,2005). When Sox9 is ectopically induced in osteoblasts, expression of Runx2 targets are lost, and ossification is abnormal (Zhou et al.,2006). In vitro, Sox9 binds Runx2 and decreases its binding affinity to target sequences. By inhibiting Runx2 activity, Sox9 may secure the chondrocyte fate.

Sox9 is subject to several posttranslational modifications, linking its regulation to many signaling pathways. One modification is phosphorylation by cGMP-dependent protein kinase II (cGKII), which prevents Sox9 from entering the nucleus (Chikuda et al.,2004). The Komeda miniature rat Ishikawa has a naturally occurring mutation in CGKII, revealing the importance of its function. These rats exhibit persistent nuclear Sox9 in postmitotic chondrocytes, a defect that inhibits the transition of prehypotrophic chondrocytes to hypertrophy. Another pathway that impinges on Sox9 is PKA signaling, which phosphorylates the protein at two sites, increasing its DNA binding affinity (Huang et al.,2000). More recently, it was found that PIAS1 SUMOylates Sox9, resulting in its stabilization and increased activation of its target, Col2a1 in vitro (Hattori et al.,2006). Cracking a posttranslational modification code may be key to understanding how Sox9 performs its various functions during chondrogenesis.

Downstream of Sox9, SoxD Genes Regulate Differentiation

Echoing a recurring theme observed among many Sox subgroups, SoxD subgroup genes Sox6 and L-Sox5 (a larger splice variant of Sox5) have overlapping functions in chondrogenesis. In double mutants, differentiation is severely affected and mice only display trace expression of extracellular markers (Smits et al.,2001). By comparison, single mutants have relatively mild skeletal defects. Agreeing with a role in differentiation, L-Sox5 and Sox6 are expressed in mesenchymal condensations, and their expression is severely down-regulated in Sox9 conditional mutants (Smits et al.,2001). In vitro data suggest that dimerization may be particularly important for L-Sox5/Sox6 function. L-Sox5 and 6, which reportedly lack transactivation domains, can both homo- and heterodimerize by means of coiled-coil domains (Lefebvre et al.,1998). Furthermore, both proteins bind pairs of HMG binding sites better than a single site, suggesting that being tethered side-by-side may increase their potency. Agreeing with this notion, expressing the two proteins together greatly potentiates transactivation of Col2a1 by Sox9 compared with introduction of L-Sox5 or Sox6 alone. A model consistent with these data is that homo- or heterodimer binding of L-Sox5/Sox6 changes architecture of target promoter/enhancers, facilitating transactivation by Sox9.


Neural crest cells (NCC) are multipotent progenitors that migrate from the dorsal neural tube to various sites in the body where they differentiate as one of several cell types. NCC derivatives include skeletogenic fates such as craniofacial cartilage and bone, and nonskeletogenic fates such as melanocytes, and neurons and glia of the peripheral nervous system (PNS). The SoxE subgroup, Sox8, Sox9, and Sox10, are pivotal in many aspects of NCC development.

There are many species-specific differences in the expression and exact roles of individual SoxE genes in the genesis of NCCs and their derivatives. For example in Xenopus, Sox8 is the first SoxE expressed in NCC progenitors, whereas in mouse the first expressed are Sox9 and Sox10 (Hong and Saint-Jeannet,2005). In zebrafish NCC, Sox8 is not expressed at all (Yan et al.,2005). Likewise, while emergence of neural crest progenitors is delayed in Xenopus injected with Sox8 morpholinos, there is no neural crest phenotype in Sox8 null mice (Sock et al.,2001; O'Donnell et al.,2006). In this case, because loss of mouse Sox8 exacerbates loss of Sox10 phenotypes Sox8 is believed to be a modifier gene that increases penetrance and phenotypic expression of Sox10 (Maka et al.,2005). Regardless of these differences, there is a conserved role for the SoxE subgroup in specification of NCC progenitors and their derivatives.

SoxE Specify and Maintain Neural Crest Progenitors

Defects in expression of SoxE group genes affect multiple NCC lineages, suggesting that they are key regulators of NCC progenitor formation (Kelsh,2006). This notion is bolstered by additional findings. In Xenopus, knockdown of Sox9 by morpholino results in a dramatic loss of cranial neural crest progenitors and expansion of the neural plate, indicating that these cells may instead be incorporated into the neuroepithelium (Spokony et al.,2002). Moreover, overexpression of Sox9 in the chick neural tubes induces expression of neural crest markers in cells that would otherwise become CNS neurons (Cheung and Briscoe,2003). It has not yet been established whether SoxE also regulates NCC induction in mouse. In mice deficient for Sox9, trunk NCCs express the marker Snail, but undergo apoptosis shortly before or after they start to migrate, indicating that it may be particularly important for the epithelial to mesenchymal transition that takes place before delamination (Cheung et al.,2005). It is possible that Sox10, which has overlapping expression with Sox9, compensates for its loss, thus masking an early role in mouse NCC induction.

Like SoxB1 in the CNS (see CNS section), mouse Sox10 supports progenitor qualities of NCCs. Sox10 maintains their multipotency, their capacity to proliferate, and inhibits their differentiation (Kim et al.,2003). Consistent with these functions, Sox10 is expressed in multipotent, premigratory neural crest and is down-regulated upon cellular differentiation.

SoxE Regulate Development of Neural Crest Derivatives

Work in mouse and zebrafish shows that Sox10 controls NCC fate specification of most, if not all nonskeletogenic NCC derivatives (Kelsh,2006). Sox10 specifies NCC fates by activating expression of genes critical for acquisition of melanocyte, Schwann, autonomic, and sensory neuron cell fates: mitfa, erbB3, Phox2b, and Mash1, and neurogenin1, respectively (Britsch et al.,2001b; Elworthy et al.,2003,2005; Kim et al.,2003; Kelsh,2006). In the case of mitfa, Sox10 binds its promoter elements directly. The pivotal role of Sox10 in neural crest cell fate specification is highlighted by the finding that mutations in Sox10 cause Waardenburg-Shah syndrome (OMIM 277580), patients of which exhibit pigmentation and enteric nervous system defects, some with an added demyelinating syndrome (OMIM 609136; Pingault et al.,1998).

Unlike in other NCC derivatives, Sox10 expression persists in glia and functions again to mediate their differentiation (Britsch et al.,2001b; Paratore et al.,2001). Sox10 directly activates differentiation genes involved in peripheral myelination such as Myelin protein zero; it also partners with Krox-20 to activate Connexin-32 (Peirano et al.,2000; Bondurand et al.,2001). These and the data above show that Sox10 performs multiple roles throughout the life of the NCC.

How do SoxE activate different target genes in NCCs and in other cell types? These abilities are at least partly under the control of posttranslational mechanisms. SUMO-1 and the SUMO E2 conjugating enzyme UBC9 bind Sox10 in a yeast 2-hybrid screen (Taylor and Labonne,2005). Changes in SUMOylation status affects outcomes of SoxE function. Overexpression of Xenopus Sox9 or Sox10 with their SUMO sites mutated, triggers ectopic melanocyte development. By contrast, overexpression of a fusion protein that mimics constitutively SUMOylated Sox9 activates inner ear development. Further work shows that differential SUMOylation of SoxE modulates interactions with partner proteins, thereby affecting which target genes are activated (Girard and Goosense,2006).


Developmental Dynamics:

What initially provoked your interest in Sox genes?

Robert Kelsh: I participated in a large-scale forward genetic screen in zebrafish, looking for mutants affecting neural crest-derived pigment cells to get a handle on neural crest development. One class of mutants we were keen to study were those in which pigment cells were essentially missing and which we hypothesized would identify genes crucial for fate specification of pigment cell types from the neural crest. The first we followed up, colourless, turned out to be a zebrafish sox10 orthologue (Dutton et al.,2001).

Jonas Muhr: After I had been working with patterning mechanisms for some time, it became clear that we knew more about the genetic programs that give neurons a certain subtype identity than the mechanisms that regulate the generation of neurons. Sox proteins and especially the SoxB1 proteins have been widely used as markers for neural progenitors, but at the time when we became interested in these genes, their functional roles were still not fully understood. What partly caught my interest was their pattern of expression. Coming from the patterning field where factors (such as homeodomain proteins) have spatially very restricted expression domains, I found Sox proteins' broad expression patterns but tight temporal regulation in differentiating neurons very suggestive.

What is your research focus?

R.K.: In the context of Sox genes, we are working towards assessing the relative importance of Sox10 for maintenance of multipotency versus specification of individual derivative fates from the neural crest. Our data in zebrafish, as well as that of others' groups in mouse, suggest that, for many neural and pigment derivatives, Sox10 has a clear role in fate specification. However, some data from mouse suggest that Sox10 has a role in maintaining neural potential. Yet it is not clear what the mechanism for this is. Neither is it clear if this role can be generalized to other fates. In addition, we have an interest in functional redundancy between related Sox genes, specifically between sox9 and sox10, and in functions in the ear. Our current focus is on using microarray studies to guide these analyses, in combination with careful analysis of sox10 mutant pigment cell phenotypes and of the neuronal phenotype of a unique, neurogenic sox10 allele.

J.M.: The main interest we have in the lab is to characterize the molecular mechanisms regulating the generation of neurons from progenitor cells. To address this question, we have focused on three major issues of neurogenesis: how are progenitor cells maintained in an undifferentiated state? How are progenitor cells instructed to commit to a neuronal differentiation program? How is the neuronal phenotype of postmitotic neurons established? We have previously demonstrated that Sox transcription factors of the B- and C-group have key regulatory roles during these processes. Now we would like to understand how the activity of Sox proteins is regulated; both at the transcriptional and posttranscriptional level. Another aim we have in the lab is to identify and characterize their relevant downstream targets using various array and ChIP techniques.

What papers have most impacted your research?

R.K.: (Anderson,2000) Marked a shift in the mouse neural crest field to a much earlier time of fate specification of neural crest derivatives than previously considered, although the idea had been prominent among chick workers like Jim Weston and Nicole Le Douarin. This finding suggested a strong parallel to the situation for pigment cell derivatives that we and others had observed.

(Southard-Smith et al.,1998) Striking similarity of homozygous loss of function sox10 phenotypes in mouse and in zebrafish colourless phenotypes led to the hypothesis that colourless was a sox10 orthologue. When this proved correct, this then indicated that Sox10 function was highly conserved between fish and mammals.

(Britsch et al.,2001a) This paper suggested that Sox10 had no direct role in mouse sensory neuron development. This challenged the generality of our published model of Sox10 function in fate specification of all neural and pigment cell derivatives and forced us to test directly whether sox10 in zebrafish could be shown to have a direct role in sensory neuron development (Carney et al.,2006). When the result of these studies was positive, this strongly reinforced the model we'd proposed.

J.M.: Within the Sox field, for me the recent papers by Boyer et al. 2005 and Bailey et al. 2006 have been inspiring (Boyer et al.,2005; Bailey et al.,2006). In the paper by Boyer et al. they mapped the genomic wide localization of Sox2, Oct4, and Nanog in ES-cells. It is likely that with this approach they identified several biologically relevant targets of Sox2 in ES-cells. But the paper also suggests that Sox2 (together with Oct4 and Nanog) maintains the pluripotency of ES-cells by regulating several hundred different genes. The paper by Bailey et al. supports this idea. Here, they show that highly evolutionarily conserved noncoding DNA elements are enriched for Sox and POU factor binding sites. They also demonstrate that a large proportion of the hundreds of enhancer elements that were identified are linked to genes expressed in the developing CNS.

There are 20 mammalian Sox genes that presumably evolved from a single, ancient Sox ancestor. Are there any remaining generalities among the gene family?

J.M.: One of the more striking generalities among several Sox proteins is their involvement in regulating differentiation of various cell types. For instance, Sox2 which is expressed in cells of the pregastrulating embryo has been demonstrated to maintain embryonic stem cells in a multipotent state and SoxB1 factors appear to have similar roles in neural progenitors (Avilion et al.,2003; Bylund et al.,2003; Graham et al.,2003). Moreover, Sox10 is vital for the undifferentiated and multipotent state of neural crest cells (Kim et al.,2003), and Sox8 which is expressed in skeletal muscle satellite cells can, when overexpressed, hinder myogenesis (Schmidt et al.,2003). So it is tempting to speculate about the existence of a general downstream program, partly Sox regulated, that is acting in several distinct cell types to control the undifferentiated, and possibly multipotent, state.

However, it is important to note that Sox proteins appear to be equally important for proper cellular differentiation. For instance, Sox21 which is closely related to SoxB1 proteins, but functions as a transcriptional repressor, is necessary for the ability of neural progenitors to commit to differentiation, possibly by antagonizing the function of SoxB1 factors (Sandberg et al.,2005). Furthermore, the SoxC and SoxE factors have been demonstrated to be important during neuronal and glial maturation (for example, Stolt et al.,2005; Bergsland et al.,2006).

R.K.: Aside from the trivial shared property of an ability to bind to DNA, this is a tricky one to answer when so many of the Sox gene family are poorly characterized functionally. I would agree that regulation of multipotency is likely to be a theme, although the extent to which this is a general feature of, for example, Sox10 function remains largely to be seen (Kelsh,2006). In part, this may be because functional redundancy between sox9 and sox10 may be rather prominent here. And clearly sox21 seems to do quite the opposite.

One feature that may be quite widespread is a transience of expression/function in many cell types, but for factors involved in establishing, rather than maintaining, fate this is not so unexpected.

Two or more genes within a Sox subgroup often control the same developmental processes. Why? Are they redundant?

J.M.: This is something that is also true for many patterning genes such as homeodomain transcription factors. There are of course several different possibilities why several Sox genes exist within each subgroup. One possibility is that it simply functions as a “safety mechanism”—if one gene is malfunctioning there is another one that can do the job. Another explanation is that gene duplication may be the most efficient way of crating a dynamic expression pattern. The expression of, for instance, SoxB1 proteins can also be detected in selected sets of differentiated neurons. Additionally in the adult brain, the expression of SoxB1 proteins overlaps to much less of an extent compared with embryonic stages. For example, Sox1 is the only SoxB1 protein expressed in GABAergic neurons of the ventral striatum, and these neurons appear to require Sox1 activity for their proper migration and specification (Ekonomou et al.,2005).

But to address the question whether Sox genes within a certain subgroup are truly redundant, I would answer both yes and no. Sox proteins belonging to the same subgroup have biochemical similarities and gain- and loss-of-function studies suggest that they, to at least some extent, can perform similar functions. But on the other hand, their functions will also diverge to some degree as most subgroup-specific Sox genes do have differences in their expression patterns.

R.K.: I agree. We have identified an example of apparent partial redundancy between sox9b and sox10 in zebrafish sensory neuron development (Carney et al.,2006). This probably reflects a strong overlap of expression of these closely related Sox genes in premigratory neural crest. Nevertheless, functions of these genes in other neural crest derivatives are not redundant, at least in part because the late expression patterns largely do not overlap.

In certain instances subsets of proneural factors and Sox genes have similar roles in neurogenesis. How do their functions differ?

J.M.: It is striking that a link has been found between Sox proteins and bHLH factors in several different cell types, including neural progenitors, neural crest cells, and in myoblasts. My lab has focused on the role of Sox proteins during neural development and has found a close link between Sox proteins and proneural bHLH proteins during neurogenesis. While SoxB1 factors can counteract neurogenesis by blocking the activity of proneural bHLH proteins, the SoxB2 member, Sox21, appears to be required for the ability of proneural proteins to promote neurogenesis (Bylund et al.,2003; Sandberg et al.,2005). Furthermore, the group-C Sox proteins, Sox4 and -11, appear be necessary for the expression of neuronal properties in postmitotic differentiating neurons (Bergsland et al.,2006). So at a first glance, Sox proteins and proneural factors seem to perform similar functions. But I think that the best interpretation of these data is that proneural factors orchestrate neurogenesis by regulating the activity and expression of a set of downstream factors, which partly consists of Sox proteins.

R.K.: But there may be no consistent pattern to this relationship across sox genes in general. Thus, although ngn1 and Mash1 mutant phenotypes share sensory neuron and sympathetic neuron defects, respectively, with sox10 mutants, the evidence suggests that, in both these cases, the proneural gene is the Sox target (Kim et al.,2003; Carney et al.,2006). That said, Kim et al. (2003) do propose that there is negative feedback on Sox10 and this may involve Mash1, but this hypothesis needs to be tested before we can extend to this example the generalization Jonas has noted.

A Sox gene can perform unique functions in different cell types, and regulates different events in the same cell type. Do you think the mechanistic basis of functional specificity mainly hinges on context-specific binding partners?

R.K.: Yes, I think this makes a lot of sense. It must, of course, be influenced by a cell's history/lineage, so that a specific Sox protein has different transcriptional targets in different tissues/cell types; presumably chromatin changes play an important role here. Then the regulation of targets in the one cell might be dependent upon changes in the transcriptional cofactors present at different stages. But there is also growing evidence that posttranslational modification of Sox proteins may be very important; this may work via affects on interactions with transcriptional cofactors, or via independent mechanisms. However, I think we know too little at present about the in vivo transcriptional targets of different Sox genes and the other regulatory factors that act on these promoters to be able to answer the question definitively.

J.M.: I agree with Robert. Emerging data indicate that the activity of Sox proteins is regulated by several different means, including posttranslational modifications, presence of necessary heterodimerizing partner factors, and presence of permissive and antagonizing factors. But the status of the chromatin will most likely also be influencing the activity of Sox proteins.

A striking feature of Sox genes is that they confer different states of cellular maturity (i.e., specification, differentiation). Is this merely a consequence of differential target activation? Might Sox genes somehow confer a global change in the cell, for example by regulating chromatin?

R.K.: I am not sure and, indeed, wonder if it is a little too early to answer this question. I think it still remains to be seen, for some Sox proteins at least, whether all these roles can be generalized to multiple cell types, even among those cell types with a common embryonic origin. So, for example, in the neural crest, whether Sox10 regulates the maintenance of multipotentiality for pigment cell precursors, or just for neural ones, is currently unclear. However, I would argue that a conserved role in specification of derivative cell types looks highly likely. Roles for SoxE factors in differentiation of oligodendrocytes and in Schwann cells could easily be imagined to result simply from overlapping spectra of target genes. In contrast, in other cell types, Sox10 probably does not regulate differentiation per se, although there is some discussion as to how absolutely this is true (Elworthy et al.,2003; Hou et al.,2006).

J.M.: SoxB1 factors are involved in maintaining the progenitor state of ES-cells and neural cells. But Sox3, for instance, can also when misexpressed cause deregulated growth of fibroblasts (Xia et al.,2000), and the same appears to be true when SoxB factors are misexpressed in myoblasts. Whether this ability of SoxB proteins reflects their capacity to regulate similar target genes in different cell types or is due to global changes in chromatin is, however, not understood. It will, therefore, be of importance to examine if a certain Sox protein can regulate the same set of target genes in several distinct cell types.

What are some important questions that remain to be answered?

R.K.: What are the regulatory targets of specific Sox proteins and how do Sox proteins fit into gene regulatory networks?

Are there common regulatory targets for single Sox proteins in multiple cell types? What are the regulatory partners for Sox proteins?

What are the mechanism(s) for regulation of multipotentiality by Sox proteins?

Do Sox proteins show full functional redundancy in any cell type or process?

How divergent are the roles of Sox genes in different vertebrates?

J.M.: I think one could add: How do posttranslational modifications influence the activities of Sox proteins?

What exciting ideas are emerging in the field?

R.K.: Well, one thing that surprised me, given how conserved Sox mutant phenotypes can be even between mouse and fish, was how overexpression/knockdown effects could also be so different in, for example, frogs (Aoki et al.,2003; Honore et al.,2003). This may hint that there are significant differences in the role of Sox genes in gene networks underlying particular developmental processes in different taxa. Also, there are growing hints of post-translational modifications being important in Sox protein function (for example, Taylor and Labonne,2005; Girard and Goossens,2006).

J.M.: Sox proteins bind to the minor groove, which consequently induces a bend in the bound DNA. These data have in part led to the hypothesis that Sox proteins may function as architectural proteins that can organize the chromatin structure. It will be interesting to see whether these ideas hold true and, if so, how such an activity influences the biological functions of different Sox proteins. 1

Figure 1.

Jonas Muhr, Associate Professor, Stem Cell Biology Group, Ludwig Institute for Cancer Research, Karolinska Institute, Sweden (L), and Robert Kelsh, Reader, Department of Biology and Biochemistry, University of Bath, United Kingdom (R).


Many thanks to Maria Bergsland, Tom Carney, Benoit de Chrombrugghe, Peter Koopman, and Michael Wagner for their helpful suggestions. I also thank Jonas Muhr and Robert Kelsh for sharing their insights.