The proliferating field of neural crest stem cells


  • Mariana Delfino-Machín,

    1. Centre for Regenerative Medicine and Department of Biology and Biochemistry, University of Bath, Bath, United Kingdom
    Search for more papers by this author
  • Thomas R. Chipperfield,

    1. Centre for Regenerative Medicine and Department of Biology and Biochemistry, University of Bath, Bath, United Kingdom
    Search for more papers by this author
  • Frederico S.L.M. Rodrigues,

    1. Centre for Regenerative Medicine and Department of Biology and Biochemistry, University of Bath, Bath, United Kingdom
    Search for more papers by this author
  • Robert N. Kelsh

    Corresponding author
    1. Centre for Regenerative Medicine and Department of Biology and Biochemistry, University of Bath, Bath, United Kingdom
    • Centre for Regenerative Medicine and Department of Biology and Biochemistry, University of Bath, Bath, BA2 7AY, UK
    Search for more papers by this author


Neural crest stem cells were first isolated from early embryonic neural crest in the early 1990s, but in the past 5 years, there has been a burst of discoveries of neural crest-derived stem cells from diverse locations. Here, we summarize these data, highlighting the characteristics of each stem cell type. These cells vary widely in the markers they express and the variety of cell types they appear to generate. They occupy diverse locations, but in some cases multiple stem cell types apparently occupy physically proximate niches. To date, few molecular similarities can be identified between these stem cells, although a systematic comparison is required. We note other issues worthy of attention, including aspects of the in vivo behavior of these stem cells, their niches, and their lineage relationships. Together, analysis of these issues will clarify this expanding, but still young, field and contribute to exploration of the important therapeutic potential of these cells. Developmental Dynamics 236:3242–3254, 2007. © 2007 Wiley-Liss, Inc.


The neural crest (NC) is a transient structure, generated by delamination of cells from the dorsal neural tube. Neural crest cells (NCCs) briefly rest in a dorsal or dorsolateral position to the neural tube before migrating along defined pathways to populate various regions of the embryo (Le Douarin and Kalcheim,1999). Remarkably, NCCs contribute to a diverse array of distinct cell types, including multiple skeletal cells (e.g., chondrocytes, osteoblasts, and odontoblasts), pigment cells (e.g., melanocytes, but also four or more other types in anamniotes), and peripheral neurons (e.g., sensory and sympathetic) and glia (e.g., Schwann and satellite cells).

NC development is widely considered to involve a process of progressive fate restriction, whereby pluripotent cells tend to become increasingly limited in their potencies over time (reviewed in Weston,1991; Le Douarin and Dupin,2003; Crane and Trainor,2006). Single cell labelling studies in vivo generally indicate a diversity of cells even in the premigratory neural crest, with some being multipotent and others apparently fate-restricted (Bronner-Fraser et al.,1980; Bronner-Fraser and Fraser,1991; Raible and Eisen,1994; Serbedzija et al.,1994; Dutton et al.,2001). Consistent with this finding, molecular markers may show heterogeneous patterns of expression in premigratory neural crest cells (Ciment and Weston,1982; Girdlestone and Weston,1985; Marusich et al.,1986). The differing potentialities of neural crest cells have been tested more extensively and rigorously in vitro under culture conditions suitable for differentiation of diverse cell types; these studies reinforce a picture of a diversity of cell types within the neural crest, both in premigratory and postmigratory locations (Sieber-Blum and Cohen,1980; Baroffio et al.,1988; Sieber-Blum,1989; Baroffio et al.,1991; Ito and Sieber-Blum,1993; Henion and Weston,1997). Furthermore, the occasional identification in primary neural crest cell culture of single cells apparently able to generate five or more distinct cell types has led to the widespread assumption that fully multipotent neural crest cells, able to generate all neural crest derivatives, exist at least transiently (Baroffio et al.,1991). The huge diversity of neural crest-derived fates and the inevitable focus on the best-studied derivatives means, however, that this attractive assumption remains unproven.

Our understanding of mechanisms of neural crest fate choice has been focused by the concept of stem cells, and this has opened up a vision of a therapeutic approach to congenital and degenerative diseases of the neural crest (“neurocristopathies”; Bolande,1974). However, the stem cell concept is plagued by multiple, partially overlapping definitions. Recently, Smith proposed a unifying stem cell biology vocabulary (Smith,2006). He defined a stem cell as “a cell that can continuously produce unaltered daughters and also has the ability to produce daughter cells that have different, more restricted properties.” Thus, a stem cell is a cell that undergoes self-renewal, but that can generate differentiated progeny. In this article, we will follow this simple definition of a stem cell. There is now abundant evidence that undifferentiated neural crest-derived cells capable of self-renewal can be isolated from postmigratory positions and, indeed, in some cases even from adult tissues.

A further criterion often used is that a stem cell must show multipotentiality (e.g., Merkle and Alvarez-Buylla,2006), but given that progressive restriction in developmental potential is a key feature of neural crest development, it seems logical to consider the possibility that a neural crest-derived stem cell might generate only one fate. Other recent reviews (e.g., Crane and Trainor,2006) have used the term progenitor for cells that are “more …limited in their capacity for self-renewal and differentiation potential,” but as we shall see in many cases the full potential of NC-derived stem cells is still to be explored and the capacity for self-renewal is rarely tested extensively enough to define whether it is “indefinite.” Instead, Smith distinguishes multipotent (can form all lineages constituting a tissue), oligopotent (forms two or more lineages), and unipotent (one lineage only) stem cells (Smith,2006), and we will do so here, although we note that this issue is confounded by our incomplete knowledge of the conditions required to specify each individual fate, so that definitive testing of full potential has thus far not been achieved. We will use the term progenitor as a generic term for precursor cells, whether stem cells or not, principally in discussion of experimental data in which self-renewal capacity was not reported. A further issue regarding cell potency concerns the identification of derivative cell types, especially in vitro. In most cases, a small number of markers, often only one, have been used to identify specific fates. Although understandable from a practical perspective, it means that added caution is required when interpreting cell potencies.

In this review, we will summarize the current knowledge of all stem cell types, both embryonic and adult, that have been shown to be neural crest-derived (Fig. 1; Table 1). Numerous influential studies of the properties and response to various growth factors of neural crest cells have been fundamental in revealing the biology of the NC. However, often these studies have not distinguished between cells of different types. Hence, here, we are restricting our discussion to those studies that examined specific populations of at least partially characterized cells that have been shown to self-renew. We will highlight our understanding of their potencies, degree of self-renewal, and characteristic markers and niche. We will indicate the holes in our understanding and suggest future research directions.

Figure 1.

Origins of neural crest (NC) -derived stem cells. A: Schematic representation of a murine embryo and the locations from which NC-derived stem cells have been isolated. B: Identified stem cells in the hair follicle and dermis. The hair follicle contains a diverse population of neural crest stem cells (NCSCs) including melanocyte stem cells (MSCs, blue), SKin-derived Precursors (SKPs, purple), epidermal neural crest stem cells (EPI-NCSCs, green), and non–NC-derived stem cells such as keratinocyte stem cells (not shown for clarity).

Table 1. Neural Crest-derived Embryonic and Adult Stem Cell Types
Stem cell typeMarkersNC derivation assayAssess- mentSelf- renewalPotencyab
Adipo- cytesChondro- cytesMyofibro- blastsMelano- cytesNeuronsSympathetic neuronsParasym- pathetic neuronsEnteric neuronsDRG sensory neuronsGlia
  • a

    In general, cell identities have been determined using one or, at most, a few established gene markers; hence, it cannot be excluded that these cells are not fully differentiated.

  • b

    In many instances, assessment of potency, neural crest origin, or capacity for self-renewal has only been partial. We therefore identify positive detection with a Y (yes) and use N (no) to denote when a particular potency has been tested but not detected. Blank spaces indicate cases that remain to be tested. Superscript numbers refer to references as follows:

  • 1

    Stemple and Anderson,1992;

  • 2

    Shah et al.,1996;

  • 3

    Morrison et al.,1999;

  • 4

    Sieber-Blum et al.,2004;

  • 5

    Bixby et al.,2002;

  • 6

    Kruger et al.,2002;

  • 7

    Trentin et al.,2004;

  • 8

    Tomita et al.,2005;

  • 9

    Nishimura et al.,2002;

  • 10

    Joseph et al.,2004;

  • 11

    Toma et al.,2001;

  • 12

    Hu et al.,2006;

  • 13

    Osawa et al.,2005;

  • 14

    Lang et al.,2005;

  • 15

    Li et al.,2007;

  • 16

    Yoshida et al.,2006;

  • 17

    Wong et al.,2006;

  • 18

    Youn et al.,2003;

  • 19

    White et al.,2001;

  • 20

    Fernandes et al.,2004;

  • 21

    Mosher et al.,2007;

  • 22

    Lee et al.,2004. CNC-SC, cardiac neural crest-stem cell; NC, neural crest; DRG, dorsal root ganglia; eNCSC, early neural crest stem cell; NCSC, neural crest stem cell; E, embryonic day; MSCs, melanocyte stem cells; EPI-NCSC, epidermal neural crest stem cells; SKPs, SKin-derived Precursors; SP, side population.

eNCSC1222p751222 nestin12Wnt1-Cre22In vitroY1  Y2  Y2  Y22Y1
In vivo      Y1  Y22 
Trunk NCSC7  In vitroY NYYY    Y
Cranial NCSC7  In vitroY YYYY    Y
Sciatic nerve NCSC35101921p7535101921 P0(-)3521 α421 integrin21Wnt1-Cre10 Dhh-Cre10 hPAP-Cre21In vitroY3  Y3510 Y351021    Y351021
In vivoY3  Y10N21Y321Y21Y19N21N1921Y310N21
Gut NCSC56921p75521α4521 integrin9hPAP-Cre21In vitroY5  Y521N21Y521Y6   Y521
In vivo      N21Y21Y21N21 
Postnatal gut NCSC6p75 In vitroY  Y  Y Y Y
In vivo      Y Y Y
DRG-derived15nestin p75 In vitroY  Y Y    Y
MSC91314Dct13 Pax313Dct-lacZ913In vitroY9   Y9      
EPI-NCSC412Sox104 nestin4 Msx212 Myo1012 Tebp12Wnt1-Cre4In vitroY4 Y4Y4Y4Y4    Y4
SKP(Sox10)1120nestin11 Sox920 p75(-)1120 Sox10(-)20Wnt1-Cre20In vitroY1120Y11 Y1120 Y1120    Y1120
In vivo          Y20
SKP(Sox10+)17p75 Sox10Wnt1-CreIn vitroYYYYYY    Y
CNC-SC18  In vitroY YYYY    Y
In vivo          Y
Cardiac NC-derived (cardiac SP)8nestin Musashi-1, MDR-1 GATA-4P0-CreIn vitro  YYYY    Y
In vivoY  Y Y    Y
Corneal precursors (COP)16nestin Notch1 Musashi-1 ABCG2Wnt1-Cre P0-CreIn vitroYYY  Y    Y


Stemple and Anderson first coined the term neural crest stem cell (NCSC) for self-renewing, oligopotent cells isolated from rat neural tube explants by selection for low affinity nerve growth factor (LNGFR or p75LNTR) expression (Stemple and Anderson,1992). These cells expressed nestin, lacked neural differentiation markers, and have been shown to give rise to sensory, sympathetic, and parasympathetic neurons; Schwann cells; and satellite glia and myofibroblasts (Stemple and Anderson,1992; White and Anderson,1999; Lee et al.,2004). These cells are now clearly one of multiple types of neural crest-derived stem cells (Table 1 summarizes all the characterized NC-derived stem cells), and we shall refer to them as early NCSCs (or eNCSCs; Lee,2004). Cells with similar properties to eNCSCs have been isolated from postmigratory phases of NC development, including from rat embryonic sciatic nerve and gut, again by expression of p75. Because the peripheral nervous system (PNS) glia at these stages also show p75 expression, the authors also selected against the peripheral myelin protein, P0 (Morrison et al.,1999). Importantly, upon transplantation back into chicken embryos, these postmigratory NCSCs give rise to cells expressing markers characteristic of sensory, sympathetic, and parasympathetic neurons, and Schwann cells and satellite glia (Morrison et al.,1999; White and Anderson,1999; Mosher et al.,2007). As all these postmigratory NC-derived cells show expression of early NC markers, and also migrate appropriately and contribute to similar structures as eNCSCs, they are regarded here as persistent eNCSCs. However, quantitative and qualitative changes in their fate potential have been demonstrated (Kruger et al.,2002; Kubu et al.,2002; Mosher et al.,2007). For instance, sciatic nerve-derived NCSCs more readily form glia (Kubu et al.,2002). Furthermore, sciatic nerve and gut NCSCs transplanted into chick embryos displayed different migratory behaviors and potencies, with sciatic nerve NCSCs unable to populate the gut or differentiate into enteric neurons, whereas gut NCSCs contributed frequently to enteric neurons (Mosher et al.,2007). Although transplanted gut and sciatic nerve cells both contributed to chick dorsal root ganglia (DRGs), their adoption of sensory neuron fates both in vivo and in vitro is very rare at best, even under culture conditions optimized for sensory neuron survival or stimulation of sensory neurogenesis (Mosher et al.,2007). The concept of progressive changes in NC-derived stem cell properties is further reinforced by analysis of gut-derived NCSCs from postnatal mouse (Kruger et al.,2002). NCSCs could be isolated from the plexus layers of the postnatal gut, even as late as 110 days postparturition. These cells expressed p75 and could be strongly enriched by selection for this marker. Using conditions similar to those for the study of eNCSCs, these cells were shown to be self-renewing, although this ability was reduced in cells from older animals. Furthermore, they were multipotent, generating mixed neuronal/glial/myofibroblast clones, but the spectrum of neuronal subtypes changed with increasing age, becoming biased toward late-developing enteric neuron subtypes, as well as to glial cells. Importantly, the authors showed that the cells' response to neurogenic (bone morphogenetic protein [BMP]) and gliogenic (Delta) factors declined and increased, respectively, compared with cells isolated from earlier stages.

Multipotent cells expressing the eNCSC marker p75 and responding to instructive growth factors in a manner akin to eNCSCs have been isolated from embryonic rat DRG, but their self-renewal capacity was not tested (Hagedorn et al.,1999). Recently adult NC-derived stem cells have been derived from rat DRG (Li et al.,2007). These cells can be grown as spheres in culture, and have been maintained for over 3 months by subcloning. These cells stably express p75, nestin, and ErbB2 and ErbB4. In culture, they can generate neurons, glia, and myofibroblasts. Reverse transcription-polymerase chain reaction (RT-PCR) analysis of various NC, DRG, and neural stem cell marker genes identified many expressed in the spheres, but the authors did not distinguish which were expressed specifically in undifferentiated stem cells. Importantly, the authors provide evidence that these DRG-derived stem cells originate from satellite glial cells.

Analysis of the behavior of eNCSCs from gut has provided an interesting insight into how eNCSC properties change with time, as well as the basis of Hirschsprung's disease, a human neurocristopathy associated with defective RET signalling (Carlomagno et al.,1996; Ivanchuk et al.,1996; Seri et al.,1997). RET and its coreceptor GFRA1 mediate glial-derived neurotrophic factor (GDNF) signalling, and this finding is important for migration of NCCs along the developing gut (Natarajan et al.,2002; Barlow et al.,2003; Yan et al.,2004). Iwashita and colleagues identified expression of Ret and Gfr1 in rat gut eNCSCs; interestingly, RET expression was not found in other eNCSCs (Iwashita et al.,2003). Importantly, they showed that migration, but probably not proliferation, survival, or differentiation, of gut eNCSCs was stimulated in a culture assay by GDNF and that Ret mutant mice show highly decreased numbers of eNCSCs in distal gut regions.

Evidence that NCSCs show self-renewal was paramount to their original classification as stem cells (Stemple and Anderson,1992). This study demonstrated that eNCSCs showed at least a limited capacity for self-renewal, because they generated mostly oligopotent clones after two rounds of subcloning, estimated to correspond to perhaps 6–10 cell generations. Similarly, postmigratory, p75+P0 NCSCs dissociated from embryonic day (E) 14.5 or E16.5 sciatic nerves and cultured for up to 11 days also give rise to oligopotent secondary subclones (Morrison et al.,1999). These authors also presented evidence indicating that p75+P0 sciatic nerve cells incorporate bromodeoxyuridine in vivo, and are therefore mitotic, consistent with persisting NCSCs being self-renewing (Morrison et al.,1999). NCSCs isolated from other postmigratory positions (gut, DRGs, and sympathetic ganglia) have all been shown to generate oligopotent daughter cells after subcloning (Bixby,2002; Kruger,2002). Studies to clarify whether early NCSCs in mammalian and avian systems might be capable of sustained self-renewal remain to be performed (Stemple and Anderson,1992; Bixby et al.,2002; Kruger et al.,2002).

The abundance of NC-derived stem cells at different stages is a key question that remains largely unanswered. This question was addressed directly in a study of gut eNCSCs, which showed that only 1–2% of E14.5 rat gut cells are strongly p75+α4integrin+ (Iwashita et al.,2003). Importantly, when these cells were plated clonally, only some 60% formed colonies, and of these, only 80% generated mixed neuron/glial/myofibroblast colonies. This finding illustrates nicely one complexity of these stem cell studies, namely that even fluorescence-activated cell sorting–purified cells remain a mixture and are simply enriched for stem cells. Studies examining early NCC primary cultures have shown that the majority of NCCs that grow out from, for example, mouse E9 neural tube explants are Sox10+p75+ (Hari et al.,2002), yet in vivo studies of lineage markers suggest that these cells are a mixture of cell types, including, for example, melanocyte precursors (Wilson et al.,2004).

Experiments in culture have identified several key growth factors affecting the fates selected by eNCSCs (reviewed in Basch and Bronner-Fraser,2006). Initially these studies have focused on whether these factors were instructive (driving fate specification of stem cells) or selective (allowing survival of cells specified only to certain fates). In all cases identified to date, they work in an instructive mode, with Wnt/β-catenin, BMP2/4, Neuregulin1 and Notch/Delta, and transforming growth factor-beta (TGFβ) signalling driving sensory neuronal, sympathetic neuronal, glial, and myofibroblast fate, respectively. Importantly, these in vitro studies on isolated eNCSCs have generally been extended convincingly to the in vivo situation, with evidence implicating each of these factors in these same fate choices in the endogenous NC (Shah et al.,1994,1996; Lo et al.,1997; Shah and Anderson,1997; Morrison et al.,2000; Wakamatsu et al.,2000; Hari et al.,2002; Lee et al.,2004). One fate conspicuously absent from current reports on eNCSCs is the melanocyte. Surprisingly, exposure of eNCSCs to Wnt1 is not reported to induce melanocytes in culture (Lee et al.,2004; L. Sommer, personal communication), despite the observation that inactivation of β-catenin in NCCs causes almost total loss of melanocytes, as well as sensory neurons, in vivo (Hari et al.,2002). This finding is more remarkable considering that other oligopotent NC-derived stem cells clearly have melanocyte potential (see below) and given that Wnt signalling is likely to directly regulate melanocyte fate by activation of the master switch gene Mitf (Dorsky et al.,2000; Takeda et al.,2000). The simplest explanation would seem to be that it reflects an inadequacy of the culture conditions used. An alternative, intriguing possibility is that melanocyte fate specification or commitment may be specific to the Wnt used, because Wnt3a, but not Wnt1, biases NCCs to the melanocyte fate (Jin et al.,2001; Dunn et al.,2005). Furthermore, where β-catenin signalling is stabilized in vivo in the NC, sensory neurons are made and other cell types, including melanocytes, are lost; apparently sensory neuron specification is driven to the exclusion of other fates (Lee et al.,2004). This result might originate from altered NCC migration affecting fate choice, but also can be readily reconciled with the β-catenin inactivation data if the sensory neuron fate decision precedes the melanocyte fate decision, or if melanocyte specification requires purely transient, rather than prolonged, signalling. Hence, it will be interesting to see whether eNCSCs and pigment cell precursors segregate from a truly multipotent neural crest stem cell. Such a scenario would be consistent with the identification of segregated melanocyte progenitors in mouse dorsal neural tube (Wilson et al.,2004) and early migrating NCCs from avian neural tube explants (Henion and Weston,1997; Luo et al.,2003).


Extensive characterization of avian NC by primary culture of NCCs has revealed many types of embryonic NC progenitors (Sieber-Blum and Cohen,1980; Baroffio et al.,1988,1991; Sieber-Blum,1989; Ito and Sieber-Blum,1993; Henion and Weston,1997). These influential studies have convincingly demonstrated the diversity of potencies among even early NCCs and have led to a major model of the lineage relationships of progenitor segregation in the NC (Baroffio et al.,1988,1991; Dupin et al.,2007). However, these studies did not address the capacity for self-renewal among these progenitors. Hence, an important development of this work has been to begin to evaluate the variety of stem cells among these progenitors by classifying them according to their resulting cell fates and testing their self-renewal capacity by in vitro serial subcloning (Trentin et al.,2004). These authors found that NCCs from the head and trunk contained different kinds of progenitors, and these progenitors also differ in their ability to self-maintain and to respond to endothelin 3 (ET3), previously shown to be a proliferation and survival-inducing factor for avian NCCs (Lahav et al.,1996,1998). For instance, cranial primary colonies of bipotent cells (glia–melanocyte or GM) could be propagated by subcloning twice, but this was increased to four subcloning rounds in the presence of ET3. Glia–fibroblast (GF) progenitors were able to self-renew for up to three subclonings with or without ET3. In contrast, trunk primary colonies were only successfully subcloned when ET3 was added; one oligopotent (GMF) and two bipotent progenitors (GM and GF) displayed capacity for self-renewal. Further evaluation of the effect of culture conditions on self-renewal in this system will be critical. Likewise, investigation of the self-renewal capacity of such cells in vivo will provide an interesting comparison. Finally, this system presents a unique opportunity to test whether EDNRB signalling might function in melanocyte fate specification from multipotent precursors, because, in contrast to mammals, melanocyte fate specification in quail is accompanied by a switch from EDNRB to EDNRB2 expression (Lecoin et al.,1998). Studies of mouse Ednrb mutant NCCs provide qualitative evidence that, in mammals, Ednrb signalling has no role in melanocyte fate specification from NCCs, but functions in melanocyte differentiation (Lee et al.,2003; Hou et al.,2004), as well as affecting proliferation of some sort of melanocyte progenitor (Reid et al.,1996). The equivalent avian NCC studies are consistent with this finding, but do not rule out EDNRB signalling having an instructive role in melanocyte fate specification.


In mammals, hair pigmentation is derived from melanocytes localized in the hair follicle (Fig. 1B), which secrete pigmented organelles, melanosomes. The cyclical turnover of hair during the moult cycle is accompanied by a similar turnover of melanocytes; hence, a stem cell source of melanocyte renewal was anticipated. Candidate melanocyte stem cells (MSCs) were first identified in the lower permanent portion or bulge of the hair follicle as Dct-LacZ+ cells resistant to treatment with ACK2, a monoclonal c-Kit blocking antibody (Nishimura et al.,2002). In the resting phase (telogen), MSCs are quiescent, but their division is activated during the growth phase of the hair follicle (anagen) to supply proliferating and differentiating progeny to pigment the maturing hair. These cells colonized the bulge of early hair follicles by postnatal day 0.5, were shown to be slow cycling, and lacked melanin granules (Nishimura et al.,2002). Unlike other neural crest-derived stem cells, the potency of MSCs has not been examined; although they generate melanocytes in vivo, it remains to be seen whether they could generate other cell types in culture. Evidence supporting the self-renewal of MSCs in the hair follicle throughout the lifetime of the mouse has been presented (Nishimura et al.,2002), but this finding needs to be tested directly in culture with purified MSCs.

The ability to identify MSCs in their niche has allowed some characterization of their expression profile in vivo (Nishimura et al.,2002; Lang et al.,2005; Osawa et al.,2005). Using both immunofluorescent staining of sectioned mouse skin and single-cell expression profiling of microdissected, genetically labeled MSCs, a molecular signature (Dct+, Pax3+,Tyr, Si, Tyrp1, Kitlow/−, Mitf, Sox10, Lef1, and Mki67) has been proposed as distinguishing MSCs from melanoblasts or melanocytes, which were generally positive for all these markers (Osawa et al.,2005). It is perhaps surprising that MSCs express the melanocyte marker Dct, while not expressing key activators of Dct expression, Mitf and Sox10 (Ludwig et al.,2004). This finding suggests that Dct expression can be induced independently of Mitf and Sox10 by means of another pathway. It has been proposed that MSCs are derived from Sox10+, Mitf+, Dct+ embryonic melanoblasts, and hence, once activated, the Dct chromatin may adopt a configuration in which expression becomes Sox10- and Mitf-independent (Osawa et al.,2005). Alternatively, Dct expression may be activated by another combination of transcription factors in these cells. A study by Lang et al. (2005) indicated that Pax3 is likely to be a key component of this; they identified a set of Pax3+, Dct+ neural crest-derived cells in the bulge as MSCs and suggested that these cells might themselves be derived from Pax3+, Dct cells in response to Wnt signalling. They suggest that Pax3 is a key component of a molecular genetic mechanism maintaining MSCs in an undifferentiated state (Lang et al.,2005). Pax3 is known to act synergistically with Sox10 to activate expression of Mitf (Bondurand et al.,2000; Potterf et al.,2000), thus promoting melanocyte specification. But in addition, at least in cultured cell lines, Pax3 can also bind to the Dct promoter in a manner preventing Mitf-dependent activation; thus, Pax3 may repress Dct transcription and thus melanocyte differentiation (Lang et al.,2005). Hence, Pax3 is proposed to initiate a differentiation cascade in melanocyte precursors, while inhibiting actual differentiation by repressing transcription of key differentiation gene(s). It is unlikely that repression of Dct is sufficient to inhibit melanocyte differentiation, so that studies addressing whether Pax3 binding affects transcription of other Mitf target genes are now required. The authors also show that Wnt signalling in the hair follicle is required for Dct expression in vivo in the MSCs, consistent with their repression model. During Wnt signalling, β-catenin acts to displace Pax3 from the Dct promoter and, therefore, promotes Mitf-dependent activation of Dct expression (Lang et al.,2005). The authors suggest that these mechanisms enable MSCs to prepare for rapid differentiation when required during the hair cycle. This explanation is a fascinating idea, supported by analogy with development of embryonic NCSCs, in which sympathetic neuron specification (activation of MASH-1 and Phox2b expression) is initiated, but differentiation is transiently inhibited, by Sox10 (Kim et al.,2003). However, the recent demonstration of other neural crest-derived stem cells with much broader potential, including melanocytes, in the skin and hair follicles (see below; Sieber-Blum et al.,2004), suggests the possibility that MSCs might in fact be generated by differentiation of other neural crest stem cells in the hair follicle. In this context, it would be of great interest to know whether other NC-derived stem cells in the hair follicle (see below) express Pax3 and how these relate, if at all, to the MSCs. In addition, it now becomes particularly important to test definitively the stem cell status of isolated Pax3+ MSCs to eliminate the alternative idea that these cells are melanocyte progenitors in a very early stage of differentiation, generated from oligopotent stem cells nearby in the bulge.


MSCs are found in the skin and are expected to be unipotent, although this finding has yet to be explored. However, in recent years, several other types of stem cells, of unexpectedly broad potential, have also been identified in the skin. These cells have been labeled epidermal neural crest stem cells (EPI-NCSCs, formerly epidermal NCSCs or even eNCSCs) and SKin-derived Precursors (SKPs; Toma et al.,2001; Sieber-Blum et al.,2004; Wong et al.,2006).

EPI-NCSCs were identified by noting large numbers of neural crest-derived cells in the facial and back skin of adult mouse and were isolated by microdissection from the bulge area of whisker follicles (Sieber-Blum et al.,2004). These cells expressed Sox10 and nestin and were serially cultured. Interestingly, EPI-NCSCs showed very broad potential, generating cells expressing markers appropriate for neurons, glia, myofibroblasts, chondrocytes, and melanocytes. Thus, they have currently the best claim to be truly multipotent neural crest stem cells (see Table 1). Like eNCSCs, EPI-NCSCs respond to neuregulin-1 by generating Schwann cells. However, in contrast to eNCSCs, in response to BMP-2 they form chondrocytes.

The transcriptional profile of EPI-NCSCs in culture has been characterized by long serial analysis of gene expression (Long SAGE) and has been directly compared with that of eNCSCs and a mixture of differentiated neural crest progeny (Hu et al.,2006). As noted by the authors, this study has the limitation that samples were taken from cells of 48-hr primary cultures (EPI-NCSC and eNCSC) and 7-day differentiated primary cultures (neural crest progeny). Nevertheless, the authors identify a small suite of 19 genes shared by EPI-NCSCs and embryonic NCSCs, but absent from epidermal stem cells (Fig. 2). This finding is important at least in demonstrating that the NC-derived stem cells are distinct from epidermal stem cells (which generate keratinocytes), despite sharing a niche in the bulge. Furthermore, it provides the basis for future comparison with other NC-derived stem cell types.

Figure 2.

Markers of melanocyte stem cells (MSCs), epidermal neural crest stem cells (EPI-NCSCs), and SKin-derived Precursors (SKPs). A Venn diagram highlighting the examined and shared markers of the three neural crest stem cell (NCSC) populations identified in the hair follicle. Markers detected are shown in green, markers in plain text have not been detected. References: (1) Nishimura et al.,2002; (2) Osawa et al.,2005; (3) Lang et al.,2005; (4) Toma et al.,2001; (5) Fernandes et al.,2004; (6) Sieber-Blum et al., 2001; (7) Hu et al.,2006; and (8) Wong et al.,2006.


Highly oligopotent SKPs were isolated and characterized from postnatal and adult mouse dermis (Toma et al.,2001). Subsequently, the same group showed they originate from dermal papillae of whisker follicles and are derived from neural crest; similar cells derived from dorsal skin are also, at least in part, likely to be neural crest-derived (Fernandes et al.,2004). Of interest, from a therapeutic point of view, SKPs have also been isolated from both human scalp and foreskin (Toma et al.,2001,2005). These cells have been grown as floating sphere cultures for over 50 generations, but after dissociation and plating on adhesive substrates will generate mixed fate clones containing undefined neurons, glia, adipocytes, and rarely myofibroblasts, but apparently not melanocytes (but see below) or chondrocytes (Toma et al.,2001; Fernandes et al.,2004). In spheres, SKPs express nestin, fibronectin, Sca1, Dermo-1, SHOX2, slug, snail, twist, Pax3, and Sox9, but not Sox10, p75, and PSA–nerve cell adhesion molecule (NCAM; Toma et al.,2001; Fernandes et al.,2004; Fig. 2). The authors distinguish SKPs from mesenchymal stem cells on the basis of marker expression, behavior in floating culture and morphology, and from eNCSCs by lack of p75 (very low level in human SKPs; Toma et al.,2001,2005; Fernandes et al.,2004) and NCAM expression. They can be distinguished from EPI-NCSCs by their expression of SHOX2 and Dermo-1 (Fernandes et al.,2004; Sieber-Blum et al.,2004). However, their relationship to other stem cells, including mesenchymal stem cells, is unknown (Fernandes et al.,2007). SKPs have been tested for their in vivo potential after transplantation into chick embryo (Fernandes et al.,2004). Thus, when yellow fluorescent protein-labeled transgenic SKP spheres were placed in the neural crest migration paths, some cells dispersed along neural crest cell migration routes. Subsequently, they contribute at least to DRG glia and probably also to peripheral nerves. Of interest, they do migrate into the skin, a behavior associated with specified melanoblasts (Erickson and Goins,1995), yet despite the lack of melanocytes in SKP cultures, the fate of these cells was not tested (Fernandes et al.,2004).

A further type of NC-derived precursor capable of forming floating spheres was identified from mouse whisker skin and from both mouse and human adult trunk skin, and these have also been called SKPs (Wong et al.,2006). However, in clear contrast to the SKPs identified by Toma and colleagues, cells of these skin-derived spheres expressed the NCSC markers Sox10 and p75 (Toma et al.,2001; Wong et al.,2006). Self-renewal was demonstrated by maintenance of Sox10/p75 expression in most cells over several months of passages. In common with Sox10-expressing EPI-NCSCs, Sox10+p75+ SKPs displayed broad potency, forming glia, neurons, myofibroblasts, and rare adipocytes in culture; chondrocytes and a few melanocytes were formed with modified culture conditions. Although sharing Sox10 and p75 expression with eNCSCs, when p75+ SKP cells were isolated from spheres and exposed to growth factors, including BMP2 and NRG1, they displayed very different growth factor responsiveness, differentiating into myofibroblasts. This shift in response is similar to, but rather more dramatic than, that noted before in eNCSCs (see, for example, Kruger et al.,2002). The in vivo role of these precursors has yet to be identified.

The apparent niches of Sox10+p75+ SKPs were identified using genetic lineage labelling techniques. This finding suggested that, in the whisker follicles, many structures contained NC-derived cells and could form spheres in culture. The source of SKPs in the trunk was more restricted, but included cells that had expressed Desert Hedgehog and that were likely to be glial-derived and also cells that had expressed Dct and were, thus, likely to be of the melanocyte lineage; both cell types were found in the follicle bulge. There is strong evidence that the Sox10, p75 SKP niche includes dermal papilla, but the isolation of very similar cells from non-hairy skin (e.g., foreskin) suggests that another niche, presumably also dermal, must be suitable (Fernandes et al.,2004; Toma et al.,2005).

Together, these reports identify cells with similar potential from multiple locations in the skin (Toma et al.,2001,2005; Fernandes et al.,2004). However, further work is required to clarify the relationships, if any, between the trunk- and cranial skin-derived precursors identified by these two groups and to reconcile the differences in markers they express.

Both EPI-NCSCs and SKPs are highly oligopotent cells and are readily accessible. Consequently, attempts to test the therapeutic utility of these highly potent cells are becoming an important research focus. Schwann cells derived from Sox10 p75 SKPs can both integrate into and generate myelin sheathing in both the PNS and central nervous system (CNS) of shiverer mice (McKenzie et al.,2006), although less success has yet been obtained with SKP-derived neuronal cells (Fernandes et al.,2006). Likewise, initial attempts to get Sox10+p75+ SKPs to generate cells of the CNS failed even after lesioning, suggesting that these oligopotent cells may not be readily transdifferentiated into CNS cell types (Wong et al.,2006).


Oligopotent cells from mouse cardiac NC generate myofibroblast, neurons, Schwann cells, melanocytes, and chondrocytes in culture (Youn et al.,2003). The authors comment that self-renewal was shown by serial cloning, but they do not give details of, for example, for how many passages they could be maintained. In the same study, other progenitors of more restricted potential (giving myofibroblast and rarely chondrocytes and Schwann cells, but not pigment cells or neurons) were observed as well as committed myofibroblast progenitors. The authors do not comment on whether they tested these latter two progenitors for self-renewal (Youn et al.,2003).

A neural crest-derived stem cell population has also been identified within the adult mice cardiac side population (SP; Tomita et al.,2005). SP cells are dormant, multipotent cells that are found in a variety of tissues and can be identified by their differential ability to efflux the Hoechst 33342 dye. In mice, a subset of cardiac SP cells were able to generate spheres (“cardiospheres”) in vitro (nestin+, Musashi-1+) and to differentiate into neurons, glia, melanocytes, chondrocytes, and myofibroblasts after dissociation and in the absence of endothelial growth factor and fibroblast growth factor-2. When transplanted into chick neural tube, cardiosphere cells behaved as neural crest and migrated into the outflow tract, DRGs and spinal nerve and followed the lateral migration pathway. Together, this evidence suggests that a population of neural crest-derived stem cells remains in the heart and has capacity to differentiate (Tomita et al.,2005).


A further NC-derived and self-renewing SP stem cell has been identified from mouse corneal precursors (Yoshida et al.,2005,2006). These cells form spheres in culture; express markers including nestin, Notch1, Musashi1, Twist, Slug, Snail, and Sox9; and can be clonally expanded for over 18 passages (Yoshida,2006). They generate keratocytes (corneal stroma cells), adipocytes, myofibroblasts, neurons, and glia in culture.

Finally, a population of highly proliferative oligopotent stem cells has been reported from human exfoliated teeth (Miura et al.,2003). Known as SHED, these cells express two early mesenchymal stem cell markers: STRO-1 and CD146. They have been shown to generate odontoblasts, neuronal, glial, and adipocyte fates, and the authors speculate that they may have an NC origin. This speculation remains untested, but is consistent with the demonstration in mouse that NCCs contribute extensively to most cell types in teeth, including the dental pulp, although a cranial mesoderm contribution is likely too (Chai et al.,2000). These and other dental pulp stem cells can generate odontoblasts and dentine and have promise for tooth replacement therapies (Sloan and Smith,2007). Lineage tracing of all these cells is important for a full understanding of NC-derived stem cells.


To date, universal molecular definitions of stem cells have been elusive. A summary of the combination of stem cell and neural crest markers used in the search for self-renewing, oligopotent neural crest cells is presented in Table 1. It would be expected that genes required for the two key processes of self-renewal and multipotentiality might be broadly seen among NC-derived stem cells. Currently, most NC-derived stem cells have been examined in a slightly ad hoc manner for marker genes and this obscures any generalities that may exist. Few studies have explicitly attempted to profile these stem cells and only two have done so in an unbiased way (Molofsky et al.,2003; Hu et al.,2006), but in addition, a wide range of candidate markers have also been examined for MSCs (Osawa et al.,2005). Markers identified by these studies are summarized in Figure 2, but reveal few commonalities. Sox10, p75, and nestin are widespread and may have a relatively general role as markers of NC-derived stem cells. Sox10 has been proposed to be necessary for maintenance of multipotentiality in NCCs (Paratore et al.,2002; Kim et al.,2003) and, thus, might be expected to be a common feature of all multipotent NC stem cells. The apparent discrepancy of Sox10 SKPs (Toma et al.,2001; Fernandes et al.,2004) is, therefore, enigmatic, especially because similar cells expressing Sox10 have also been well-characterized (Wong et al.,2006). MSCs have been characterized as Sox10 (Osawa et al.,2005), which would be consistent with the assumption that these cells are unipotent, but again highlights the significance of a direct test of MSC potential and self-renewal. Interestingly, nestin and p75 are both markers of neural stem cells (Lendahl et al.,1990; Andressen et al.,2001).

An important finding concerns the role for the polycomb family transcriptional repressor Bmi-1 in self-renewal of stem cells in the central and peripheral nervous systems (Molofsky et al.,2003). Bmi-1 appears to act in postnatal stem cells by repressing the cyclin-dependent kinase inhibitors, p16Ink4a and p19Arf, thereby promoting self-renewal, but does not affect the proliferation of restricted progenitors (Molofsky et al.,2003). Whether Bmi-1 or related transcriptional repressors are a common feature of all NC-derived stem cells remains to be examined.

As we have seen, many of the NC-derived stem cells reviewed here respond in a similar way to instructive signals such as BMPs. An elegant study identified a crucial role for combined Wnt and BMP signals in maintenance of eNCSCs (Kleber et al., 2005). Previous work showed that each signal alone drives sensory neuron and autonomic neuron specification and differentiation respectively (Shah and Anderson,1997; Lee et al.,2004), but a combination of Wnt1 and BMP2 represses specification of either neuronal fate (Kleber et al., 2005). Indeed, cells fail to become glia or myofibroblasts either, and instead maintain p75 and Sox10 expression and multipotency. Intriguingly, although multipotentiality was clearly maintained, sensitivity to Wnt was specifically lost, a finding paralleled by eNCSCs prospectively isolated from two postmigratory locations: sciatic nerve and DRGs. Such a shift in the intrinsic properties of eNCSCs during development has also been reported in the context of responsiveness to BMPs and neuregulin (Bixby et al.,2002; Kruger et al.,2002). Thus, it seems that eNCSCs alter their phenotype during embryonic and postnatal development (Bixby et al.,2002; Kruger et al.,2002), a process that may be further exaggerated in SKPs (Wong et al.,2006). This change may, at least in part, explain why markers common to all neural crest-derived stem cells appear so elusive.

Finally, a study of cardiac NCSCs suggests that TrkC function is required for stem cell maintenance, perhaps by driving continuing proliferation, because premature fate specification happens in mutants (Youn et al.,2003). It is currently unclear how widespread is TrkC expression in NC-derived stem cells.


An ongoing issue in the study of any stem cell is the extent to which cells are reprogrammed or modified by culture conditions. This is important for two distinct reasons. On the one hand, from the perspective of understanding normal stem cell biology, avoidance of culture conditions causing such artefactual responses is required. On the other hand, from a therapeutic perspective, controlled reprogramming of stem cells in culture may have great potential in the treatment of disease. Work on avian NC derivatives dramatically reinforces the necessity to take the possibility of reprogramming seriously, because remarkably melanocytes have been generated from cultured sciatic nerve Schwann cells and Schwann cells can be formed by culturing of epidermal melanocytes (Dupin et al.,2000; Nataf and Le Douarin,2000). Furthermore, this reprogramming takes place by means of dedifferentiation to form a precursor expressing NC markers and showing self-renewal and oligopotency, that is, a NC-like stem cell (Dupin et al.,2000,2003; Real et al.,2005,2006). Whether NC-derived stem cells can also be reprogrammed remains to be seen. Although important, determination of whether artefactual modification of stem cell properties has occurred is difficult, especially because a comparison of the in vivo and in vitro derivatives of an oligopotent stem cell need not be identical. Consequently, this issue has been rarely explored and requires more attention in the future. It is perhaps best accomplished by comparison of properties of stem cells freshly ex vivo with those that have been maintained in culture for some time.

There is clearly a close embryological relationship between the neural crest and the neural tube, and this finding may be reflected in the apparent ability of NC-derived stem cells to generate neural stem cell derivatives (Sieber-Blum et al.,2006). Whether this represents true stem cell plasticity or simply the latent potential of stem cells reflecting their shared embryological origin is unclear, but it certainly encourages investigation of the therapeutic potential of adult stem cells.


There has been considerable effort in recent years to identify NC-derived stem cells from multiple sources, both in the embryo and in the adult. A coordinated effort to test definitively the relationship between these different stem cells is now paramount. There is at least some evidence that these diverse stem cell types are derived from other stem cells and, thus, that neural crest stem cells are somehow modified in time. This concept appears to challenge the stem cell definition we are using, because they are not “continuously producing unaltered daughters” (Smith,2006). However, it is important to remember that stem cell behavior in the changing in vivo environment might be expected to be less constant than that under standardized culture conditions; hence, long-term changes in stem cell properties in vivo are not to be considered inconsistent with our stem cell definition. A similar view is increasingly well-supported in the field of neural stem cells (Merkle and Alvarez-Buylla,2006).

Neural stem cells have turned out to be surprisingly morphologically distinctive (i.e., differentiated; Merkle and Alvarez-Buylla,2006), and it will be interesting to see to what extent this is true for neural crest-derived stem cells at different stages. Indeed, the parallel may be even stronger, because neural stem cells in the fetus and adult appear to be specific glial cell types. Of interest, many markers of early neural crest stem cells are also late markers of glia, most notably Sox10, which is functionally associated with stem cell maintenance and glial cell differentiation (Kelsh,2006). Furthermore, some SKPs, at least in the trunk, may actually have the characteristics of glial cells (Wong et al.,2006), and adult DRG-derived stem cells most likely derive from satellite glial cells (Li et al.,2007). Thus, it will be interesting to see if NC-derived stem cells in at least the peripheral nervous system adopt “glial” morphologies.

What is the in vivo role of the NC-derived stem cells? In most cases, this feature has hardly been addressed, but it is important in terms of understanding to what extent stem cell potency is tightly tuned to the in vivo function. For example, it is not clear how neuronal potential is necessary for hair follicle cells, and yet SKPs show this in vitro.

The importance of the stem cell niche in providing a cellular environment maintaining stem cells is becoming increasingly clear. For the NC-derived stem cells discussed here, the precise location of their niche is poorly defined, let alone the details of the signals provided by the niche that promote stem cell self-renewal and prevent differentiation. The hair follicle bulge is apparently the niche for MSCs (Nishimura et al.,2002) along with EPI-NCSCs (Sieber-Blum et al.,2004), SKPs (Wong et al.,2006), as well as the keratinocyte stem cells (Cotsarelis et al.,1990). There is, thus, an excellent opportunity to dissect the specialized cellular microenvironment and its influences on these diverse stem cell types (Nishikawa and Osawa,2007). Stem cell populations tend to become weakened with physiological aging, and in the case of MSCs, this weakening may result, at least in part, from ectopic differentiation within the niche and result in hair graying (Nishimura et al.,2005). Identification of the changes in the niche that occur with ageing will, thus, help elucidate some of the key factors conferring stem cell characteristics. Conversely, it will be exciting to see what factors activate stem cell proliferation and differentiation and, in the follicle bulge, how is this controlled with respect to the hair cycle.

The current list of growth factors and their receptors that instruct NC-derived stem cells to adopt specific fates is certainly inadequate to explain the full set of NC fate specification events. Furthermore, the response to the same signal or combinations of signals varies with different NC-derived stem cells. We are mostly ignorant of the receptors mediating growth factor signalling in a NC context. It is likely that these signals act combinatorially to drive specification to individual fates. Likewise, the cellular context of these signals is also crucial for the effects they have (Paratore et al.,2001; Kruger et al.,2002; Lewis et al.,2004). A concerted effort to analyse, for each class of stem cell, the response to all the known instructive factors is required. Certainly, there are other factors in the NC literature, including growth differentiation factor-7 and brain-derived neurotrophic factor (Sieber-Blum,1991; Lo et al.,2005), that should be tested on defined NC-derived stem cells, but this problem deserves a comprehensive and methodical analysis of all candidates. The development of NCSC-like immortalised cell lines, MONC-1 and especially the more stable JoMa1, expressing p75 and Sox10 and with very broad potency, including autonomic neurons, glia, smooth muscle, melanocytes and chondrocytes, promise to be invaluable tools in this regard (Rao and Anderson,1997; Maurer et al.,2007).

The lineage relationship of the various NC-derived stem cells is a crucial issue that needs to be resolved. Whereas their origins at some stage from early NC is generally not in doubt, the idea that some may share a closer relationship is certainly plausible. At the level of cell potential, a nested hierarchy with EPI-NCSCs and SKPs showing broadest potency, eNCSCs a more restricted one, and MSCs an untested, but assumed unipotency, would be consistent with this idea. The issue is particularly poignant in the case of the three stem cells described from the hair follicle bulge. Are there really three distinct stem cell types in such close proximity (four, if the non-NC derived keratinocyte stem cell is included)? We note that both SKPs and MSCs have been identified as being labeled by lineage tracing using the Dct promoter and both reside in the bulge (Nishimura et al.,2002; Osawa et al.,2005; Wong et al.,2006). Furthermore, SKPs can, at least under some culture conditions, generate melanocytes (Wong et al.,2006). Indeed, they have been proposed to be identical (Wong et al.,2006), but discrepancies in the described marker characteristics (especially Sox10) of these two cell types need to be resolved. It is highly plausible that they are lineally related. Direct tests of the potency, self-renewal, and molecular characteristics of SKPs and MSCs during late phases of the hair cycle when the stem cell-containing bulge region and the melanocyte-containing hair papilla are physically most discrete should address this issue. Currently, understanding of the in vivo lineage relationships between different types of NCCs is limited, but a detailed model based on extensive characterization of pluripotentiality of cultured NCCs has been proposed by the Le Douarin lab (Le Douarin and Dupin,2003; Le Douarin et al.,2004; Dupin et al.,2007). As noted above, several of the intermediates they proposed show at least some capacity for self-renewal (Trentin et al.,2004); as in vivo lineage relationships and progenitor types become clearer, it will be important to establish their self-renewal capacity.

A related issue is that of whether there is a truly multipotent NC-derived stem cell, as has been proposed (Dupin et al.,2007). Potentiality of NC-derived stem cells has rarely been evaluated systematically. eNCSCs do produce multiple neuronal cell types, but not, apparently, chondrocytes or melanocytes, so the relationship between these cells, if any, needs to be evaluated; EPI-NCSCs and SKPs might be multipotent cells, but this finding remains to be tested definitively. A further question relates to how potential changes with development; it is perhaps natural to tend to assume that potential decreases with developmental age, but the initial studies on eNCSCs argue more for a shift in responsiveness (White et al.,2001; Bixby et al.,2002; Mosher et al.,2007), while EPI-NCSCs appear to have a broader potential than eNCSCs. Again, understanding the lineage relationship between these stem cell types will help make sense of the varied stem cell potencies.

Part of the problem is that we remain ignorant about how many stem cells of any specific type are present at any stage in vivo. For example, the avian studies show that early NCCs consist of a mixture of stem cells of different potencies and other progenitors. Conceivably, therefore, cells akin to EPI-NCSCs are mixed in with more restricted eNCSCs in the embryonic NC. Differences in their relative abundance and in experimental technique will then affect our ability to culture them from different source material. Stem cell numbers are likely to be very variable for each stem cell type; for example, MSCs are described as being at very low abundance in the hair follicle (Nishimura et al.,2002; Mak et al.,2006), whereas SKPs in the whisker follicles are more abundant (Fernandes et al.,2007). Identification of both definitive markers and the endogenous niche for each stem cell will be necessary to assess this question, one which has great significance for the therapeutic use of these stem cells. Therapeutic applications of these stem cells are only just being explored and consideration of all of the above issues are likely to influence the success of these endeavors in the coming years.


We thank Sarah Colanesi, Marios Stavridis, and two anonymous reviewers for comments on a draft of this manuscript.