An Epidermal Neural Crest Stem Cell (EPI-NCSC) Molecular Signature


  • Yao Fei Hu,

    1. Department of Cell Biology, Neurobiology and Anatomy, Medical College of Wisconsin, Milwaukee, Wisconsin, USA
    Search for more papers by this author
  • Zhi-Jian Zhang,

    1. Department of Cell Biology, Neurobiology and Anatomy, Medical College of Wisconsin, Milwaukee, Wisconsin, USA
    Search for more papers by this author
  • Maya Sieber-Blum Ph.D.

    Corresponding author
    1. Department of Cell Biology, Neurobiology and Anatomy, Medical College of Wisconsin, Milwaukee, Wisconsin, USA
    • Department of Cell Biology, Neurobiology and Anatomy, Medical College of Wisconsin, 8701 Watertown Plank Road, Milwaukee, Wisconsin 53226, USA. Telephone: 414-456-8465; Fax: 414-456-6517
    Search for more papers by this author


Here, we report the first transcriptome for mouse epidermal neural crest stem cells (EPI-NCSC, formerly eNCSCs). In addition, our study resolves conflicting opinions in the literature by showing that EPI-NCSC are distinct from other types of skin-resident stem cells/progenitors. Finally, with the three gene profiles, we have established a foundation and provide a valuable resource for future mouse NCSC research. EPI-NCSC represent a novel type of multipotent adult stem cell that originates from the embryonic neural crest and resides in the bulge of hair follicles. We performed gene profiling by LongSAGE (long serial analysis of gene expression) with mRNA from EPI-NCSC, embryonic NCSC, and in vitro differentiated embryonic neural crest progeny. We have identified important differentially expressed genes, including novel genes and disease genes. Furthermore, using stringent criteria, we have defined an NCSC molecular signature that consists of a panel of 19 genes and is representative of both EPI-NCSC and NCSC. EPI-NCSC have characteristics that combine advantages of embryonic and adult stem cells. Similar to embryonic stem cells, EPI-NCSC have a high degree of innate plasticity, they can be isolated at high levels of purity, and they can be expanded in vitro. Similar to other types of adult stem cell, EPI-NCSC are readily accessible by minimal invasive procedure. Multipotent adult mammalian stem cells are of great interest because of their potential value in future cell replacement therapy by autologous transplantation, which avoids graft rejection.


The aims of this study were to obtain comprehensive information on the epidermal neural crest stem cell (EPI-NCSC) transcriptome and to distinguish EPI-NCSC from other skin-resident stem cells. EPI-NCSC are multipotent stem cells, which are derived from the embryonic neural crest and which reside in the bulge of the outer root sheath of adult murine hair follicles. EPI-NCSC can be isolated as a highly pure population, can generate all major neural crest derivatives, and can be expanded in vitro into millions of cells [1, [2]–3]. In a spinal cord injury model, EPI-NCSC grafts express markers for GABAergic neurons and myelinating oligodendroyctes [3]. These data indicate that human EPI-NCSC candidates are of potential interest for future stem cell replacement therapy.

The neural crest is of importance because it generates a multitude of cell types and tissues in the adult vertebrate organism, including the autonomic and enteric nervous systems, primary sensory neurons, endocrine cells, smooth musculature of the cardiac outflow tract and the great vessels, pigment cells, and the cranial mesenchyme. The cranial mesenchyme generates the dermis of the face and ventral neck, craniofacial bone/cartilage, tooth papillae, meninges, striated musculature of the eye and the corneal stroma, among other structures [4]. NCSC/progenitors are also present in the avian embryonic ectoderm [5], dorsal root ganglia [6], sympathetic ganglia [6], and the cardiac outflow tract [7]. In mammalians, NCSC/progenitors have been identified in the fetal sciatic nerve [8] and in the adult gut [9]. Thus, there are precedents for the continued existence of NCSC/progenitors in the periphery during embryonic development and postnatally.

The bulge is a multilayered part of the outer root sheath of the hair follicle (Fig. 1A; [10]). While using the bulge region in Wnt1-cre/R26R mouse whisker follicles as an experimental location to show that mammalian Merkel cells are of neural crest origin [11], we noticed neural crest-derived cells in the inner layers of the bulge [1, 2]. Given that the bulge is a niche for epidermal stem cells [12, [13], [14], [15]–16], we reasoned that neural crest-derived cells in the inner layers of the bulge might be stem cells, which turned out to be the case [1, 2].

Figure Figure 1..

Characteristics of EPI-NCSC and embryonic neural crest cultures. (A): EPI-NCSC in the bulge of hair follicles. (B): Bulge (b) explant 18 hours after onset of EPI-NCSC emigration. Migratory cells are present on the collagen substratum. (B′): Higher magnification of boxed area in (B). (C): β-Galactosidase immunoreactivity in a section through the neural tube (nt) of a Wnt1-lacZ embryo [18], in which neural crest cells (arrow) and dorsal neural tube cells (∗) are marked by the expression of β-galactosidase. (D): Neural tube (nt) explant at 48 hours after explantation. Numerous emigrated neural crest cells form a halo around the explant. (D′): Higher magnification of boxed area in (D). (D″): All cells are nestin-immunoreactive. (D‴): DAPI (blue) and nestin (green) merged images of same area as in (D″). (E): Neural crest cells at culture day 5. A monolayer of cells is interspersed with cell aggregates. (F–F′): Neuron-specific β-III tubulin (F) (Texas red fluorescence) and Sox10 (F′) (fluorescein fluorescence). Cells in aggregates are neurons that also express Sox10. (G–G″): Brn3a and c-kit double stain. Whereas neural crest cells are c-kit-immunonegative, c-kit is expressed in a subset of cells at culture day 5. c-kit is a marker for a subset of small-diameter sensory neurons and for melanocytes/progenitors. A subset of c-kit-positive cells also express Brn3a, a marker for sensory neurons. (G): Brn3a immunoreactivity. (G′): c-kit immunoreactivity. (G″): Merged images in pseudocolor. (H): Chondrocyte/osteoblast-specific collagen type II immunoreactivity (merged FITC/DAPI images). (J): Myofibroblast-specific smooth muscle actin immunoreactivity (FITC/DAPI merged images). Scale bars = 50 μm (B, B′, C, F, F′, J), 200 μm (D, E), 20 μm (D′–D‴, G–G″), 10 μm (H). Abbreviations: bv, blood vessel; DAPI, 4,6-dia-midino-2-phenylindole; dp, dermal papilla; e-bl, basal layer of the surface epidermis; EPI-NCSC, epidermal neural crest stem cell; FITC, fluorescein isothiocyanate; h, hair; m, matrix; NCP, neural crest progeny; NCSC, neural crest stem cell; ors, outer root sheath.

There are no known markers that are unique to murine NCSC in the sense that they are not also expressed in some crest-derived cell lineages later in development or in noncrest cells. Furthermore, no single marker, but a panel of markers only, can reliably identify a particular cell type. For these reasons, and for the purpose of this report, we have prepared three long serial analysis of gene expression (LongSAGE) libraries with RNA from (a) culture day-2 EPI-NCSC (48 hours after onset of emigration of cells from the bulge explant), (b) culture day-2 NCSC (48 hours after explantation of the neural tube), and (c) culture day-7 in vitro differentiated neural crest progeny (NCP) (at 7 days after explantation of the neural tube). Comparison analyses provided comprehensive information on differentially expressed genes and an NCSC molecular signature.

SAGE has been developed by Velculescu et al. [17] as a tool to quantify the transcriptome. It is based on the isolation of unique sequences (tags) from defined positions at the 3′-end of each mRNA molecule. The abundance of these tags is determined by large-scale sequencing, and it reflects the abundance of expression of the corresponding gene. We have chosen gene profiling by LongSAGE for four reasons. First, LongSAGE gene profiling can be performed with less starting material than profiling on microarrays. This was an important feature because both the embryonic neural tube and the bulge of whisker follicles initially yield relatively few cells, and we sought to avoid in vitro expansion to mimic the in vivo situation as closely as possible. Second, LongSAGE has the capability of detecting novel genes. Accordingly, several novel differentially expressed genes were indeed identified in this study. Third, LongSAGE is conducive to quantification of gene expression, which was an important aspect of our study. Fourth, LongSAGE can identify isoforms of genes (within the tag sequence). LongSAGE has the advantage over classical SAGE in that it yields 17-nucleotide rather than 10-nucleotide sequences, and thus provides a higher yield of tag-to-gene annotation.

We here provide the first transcriptome for mouse EPI-NCSC and transcriptomes for embryonic NCSC and in vitro differentiated NCP. The three longSAGE libraries are a valuable resource for future research on embryonic and adult mouse neural crest cells.

Materials and Methods


C57BL/6J mice (The Jackson Laboratory, Bar Harbor, ME, were used for all experiments. The image in Figure 1C was taken from a Wnt1-LacZ mouse (The Jackson Laboratory; [18]). The animals were housed in the transgenic facility of the Medical College of Wisconsin, and the experiments were conducted in accordance with the Guidelines for the Care and Use of Animals and approved by the Medical College of Wisconsin Animal Care and Use Committee. Embryos were obtained from timed-pregnant females and used for preparing neural crest cultures. The morning vaginal plugs were observed, was considered day 0.5 of gestation. Whisker follicles were dissected from 10-week-old C57BL/6J mice.

Cell Culture

Neural crest cell primary explant cultures were prepared from embryos at day 9.5 of gestation, as described previously [19, [20]–21]. Forty-eight hours after explantation, the neural tubes were carefully detached from the plates and discarded. The emigrated neural crest cells, which remained in the culture plate, were incubated at 37°C in a humidified atmosphere of 5% CO2 and 10% O2, as we have described previously [22]. The culture medium consisted of 75% α-minimum essential medium, 10% fetal bovine serum, and 5% day-11 chicken embryo extract. The culture medium was supplemented with basic fibroblast growth factor (FGF-2) (Upstate Biotechnology, Lake Placid, NY, at 2.5 ng/ml, neurotrophin-3 (Promega Corporation, Madison, WI, at 10 ng/ml, and mouse stem cell factor (R&D Systems, Inc., Minneapolis, at 100 ng/ml. One-half of the culture medium and supplements was exchanged daily with fresh medium and supplements. mRNA was isolated from 48-hour and 7-day-old cultures.

EPI-NCSC were obtained from anagen-phase ([10]; 6 days after depilation) whisker follicles exactly as we have described previously [1, 2]. Briefly, whisker follicles were dissected from the whisker pad, and the dermal tissue removed mechanically. A superficial longitudinal cut was made first, and the follicle was then transected first below and then above the bulge region. The bulge was isolated and placed in a collagen-coated 35-mm culture plate.

LongSAGE Library Construction and Data Analysis

Total RNA was isolated according to manufacturer's recommendations with TRIzol reagent (Invitrogen, Carlsbad, CA, from approximately 90 day-2 EPI-NCSC, 90 day-2 NCSC, and 60 day-7 NCP cultures. To avoid potential contamination with genomic DNA, the total RNA preparation was treated with DNase I (Invitrogen) as recommended in the manufacturer's protocol. The LongSAGE libraries were constructed using the I-SAGE long kit (Invitrogen) according to manufacturer's instructions. In brief, mRNA was bound to Dynal oligo(dT) magnetic beads using the cDNA synthesis module of the kit. Subsequently, mRNA transcripts were converted to cDNA with biotinylated oligo(dT)18 as the primer and Superscript III reverse transcriptase from the cDNA synthesis module of the kit. The cDNA was digested with NlaIII, and the 3′-ends were recovered. The cDNA pool was then divided in half and ligated to LS-adapters 1 and 2. Subsequently, the restriction enzyme, MmeI, was used to release the tags, which were then pooled and ligated to form ditags. Ditags were amplified by polymerase chain reaction (PCR) and subsequently isolated by polyacrylamide gel electrophoresis (PAGE) (12%) and digested again with NlaIII to release the 34-bp LongSAGE ditags. The ditags were then purified by PAGE (12%). Subsequently, the ditags were concatemerized at their NlaIII overhangs with T4 DNA ligase. Concatemers with a minimum size of 500 bp were obtained by gel purification, ligated into the cloning vector pZEro-1, and finally transformed into TOP10 bacteria by electroporation. Automated fast throughput sequencing was performed by Agencourt Bioscience Corporation (Beverly, MA, Additional information about the SAGE and LongSAGE techniques can be found at

LongSAGE data were analyzed with SAGE2000 version 4.5 software ( Tags corresponding to linker sequences were discarded, and duplicate dimers were counted once. Both 17-bp LongSAGE tags and corresponding 10-bp SAGE tags were extracted for further analysis. All tags were mapped to their corresponding genes using SAGEmap data from the National Center for Biotechnology Information (NCBI) (

Comparison between the two LongSAGE libraries was carried out with the SAGE2000 software version 4.5 (Invitrogen). p values were determined according to Audic and Claverie ([23]; Tags with multiple matches were excluded. Different tags that matched the same Unigene cluster were combined. A difference with a p value of ≤.05 was considered significant.

Green fluorescent protein (GFP)high raw Affymetrix chip data files were downloaded from Microarray Suite (MAS 5.0) software was used to read the raw data. NetAffx annotation file (December 10, 2004) was used to annotate the data. Genes marked as present (“P”) at least once were used to compare with our LongSAGE data.

Reverse Transcription (RT)-PCR and Quantitative RT-PCR

Primer sets for the following genes were purchased from SuperArray Bioscience Corporation (Frederick, MD, Rex3 (Mm.14768; PPM25806A; position 595–613; band size 162 bp); Vdac1 (Mm.3555; PPM04115A; position 582–604; band size 118 bp); Ets1 (Mm.292415; PPM03600A; position 509–529; band size 151 bp); Pcbp4 (Mm.286394; PPM37494A, position 623–642; band size 90 bp); Myo10 (Mm.60590; PPM29651A, position 5,298–5,316; band size 179 bp); H1fx (Mm.33796; PPM28101A, position 128–147; band size 129 bp); Thop1 (Mm.26995; PPM26958, position 2,128–2,147; band size 120 bp); Msx2 (Mm.1763; PPM03170A, position 903–925; band size 149 bp); Cryab (Mm.178, PPM03570A, position 503–523, band size 192 bp); Vars2 (Mm.28420, PPM03327A, position 3,665–3,683; band size 191 bp); Peg10 (Mm.320575, PPM24997A, position 608–628, band size 203 bp); Calr (Mm.1971, PPM05020, position 1,527–1,547, band size 169 bp); Crmp1 (Mm.290995, PPM37854A, position 1,804–1,822, band size 144 bp); Ube4b (Mm.288924, PPM37643A, position 4,160–4,179, band size 81 bp); Pygo2 (Mm.22521, PPM26308A, position 1,428–1,448, band size 160 bp); 5730449L18Rik (Mm.21065, PPM26127A, position 445–467, band size 136 bp); AU041707 (Mm.200898, PPM32982A, position 1,890–1,911, band size 83 bp); Nes (nestin; Mm.331129; PPM04735; position 915–936; band size 100 bp); Tebp (telomerase binding protein, p23; Mm. 305,816; position 881–903; band size 145 bp). Primers sets for the following genes (according to Rendl et al. [24]; their supplemental Table S14) were purchased from Operon Biotechnologies, Inc. (Huntsville, AL, Akp2: 5′-TGCGC-TCCTTAGGGCTGCCG-3′, 5′-GGTGTACCCTGAGATTCGTCC-3′, band size 160 bp; Hoxa9: forward: 5′-GTTCTCCGGGATGCATAGATTCA-3′, reverse: 5′-GAACCCCAAA-TTCA-CCAAGGATAC-3′, band size 346 bp; Zic1: forward: 5′-GC-GGCCGAAAGCCAACT-3′, reverse: 5′-TGCCAAAAGCAA-TGGACAGC-3′, band size 315 bp. Cd34 primers were used according to Drew et al. [25]: forward 5′-CTCTAGATCACAAGTTCTGTGTCAGC-3′, reverse 5′-TAGCACAGAACTTCCCAGCAAAC-3′; hypoxanthine guanine phosphoribosyl transferase (Hprt) primers with the following sequences were used for normalization: 5′-CCTGCTGGATTACATTAAAGCACTG-3′ and 5′-CCTGAAGTACTCATTATAGTCAAGG-3′ (band size 350 bp).

For RT, total RNA was extracted with TRIzol reagent and treated with DNase (Invitrogen) to remove traces of genomic DNA. First-strand cDNA was synthesized using the SuperScript III First-strand synthesis system for RT-PCR (Invitrogen) and primed with oligo(dT) according to manufacturer's instructions. For quantitative PCR, each 25-μl of PCR mix consisted of 12.5 μl of 2× RT2 Real-Time SYBR Green/Fluorescein PCR Master Mix, 1.0 μl of First Strand cDNA template, and 1.0 μl of RT2 PCR Primer Set and was brought to a final volume of 25 μl using ddH2O. Thermocycling was performed as follows: 95°C, 15 minutes and 38 cycles of 95°C, 30 seconds; 55°C, 30 seconds; and 72°C, 30 seconds. The amplification rate of each target was evaluated from the cycle threshold (Ct) numbers obtained for serial cDNA dilution. For a given target, the Ct difference was calculated and then expressed as relative mRNA level. For each PCR product, a single narrow peak was obtained by melting curve analysis at the specific melting temperature and only a single band of the predicted size was observed by agarose gel electrophoresis. For conventional PCR, 12.5 μl of 2× ReactionReady HotStart “Sweet” PCR Master Mix, 1.0 μl of First Strand cDNA template, and 1.0 μl of RT2 PCR Primer Set were combined and brought to a final volume of 25 μl using ddH2O. Thermocycling was performed as follows: 95°C, 15 minutes and 30 cycles of (95°C, 30 seconds; 55°C, 30 seconds; and 72°C, 30 seconds). Ten-microliter PCR product was separated on a 1.8% agarose gel containing ethidium bromide.


The following antibodies were used: mouse monoclonal β-III-tubulin (1:200; Chemicon International, Temecula, CA,, rabbit polyclonal Sox10 (1:100; Chemicon International), rabbit polyclonal Brn3a (1:300; Chemicon International), c-kit (ACK2; 1:500; Chemicon International), mouse monoclonal nestin (1:400; BD Biosciences, San Jose, CA,, rabbit polyclonal Msx2 (1:400; CeMines, Inc., Golden, CO,, goat polyclonal Myo10 (1:200; Santa Cruz Biotechnology, Inc., Santa Cruz, CA,, rabbit polyclonal β-galactosidase (gift of J. Sanes; 1:10), mouse monoclonal smooth muscle actin (1:800; Sigma-Aldrich, St. Louis,, and mouse monoclonal collagen type II (CIIC1; undiluted culture supernatant; Developmental Studies Hybridoma Bank, Iowa City, IA,∼dshbwww). The cultures/tissue was fixed with 4% paraformaldehyde for 30 minutes, rinsed three times, blocked with normal host serum, and incubated with the primary antibody overnight in the cold. The cultures/tissue was then rinsed again exhaustively and incubated with the secondary antibody (1:200; fluorescein- or Texas red-conjugated; Jackson ImmunoResearch Laboratories, Inc., West Grove, PA, After a final cycle of rinsing, they were incubated for 15 minutes with 4,6-diamidino-2-phenylindole (DAPI) (Invitrogen) and finally mounted with Vectashield (Vector Laboratories, Burlingame, CA, and a coverslip.

Results and Discussion

Cultures of EPI-NCSC, NCSC, and NCP

Within 2 to 3 days after explantation, whisker bulges released highly migratory and proliferative NCSC (Fig. 1B, 1B′; [1, 2]). Embryonic NCSC originate from the dorsal aspect of the neural tube (Fig. 1C). They started to emigrate from neural tube explants within several hours (Fig. 1D, 1D′). All NCSC were intensely nestin-immunofluorescent (Fig. 1D″, 1D‴). NCSC proliferated rapidly between culture days 2 and 7 and differentiated into the major neural crest derivatives. After day 7, there was a decline in proliferation, but some dividing cells persisted for the next several weeks. By day 5, the cultures consisted of a monolayer interspersed with multilayered cell aggregates (Fig. 1E). The aggregates consisted of β-III-tubulin immunoreactive neurons (Fig. 1F) that also expressed Sox10 (Fig. 1F′). Some cells were immunoreactive for Brn3a, which is a marker for sensory neurons (Fig. 1G–1G″). Unlike at the onset of emigration, when all cells were c-kit-negative, by culture day 5 many cells expressed c-kit (Fig. 1G′), a marker for a subset of small-diameter sensory neurons [26] and melanocytes [27]. Cells in the flat areas of the culture included smooth muscle actin-immunoreactive myofibroblasts and nestin-immunoreactive putative stem cells. As we have described previously, these cells also express markers for autonomic neurons, Schwann cells, and melanocytes [20]. In addition, they differentiate into bone/cartilage cells (Fig. 1H; [28]) and myofibroblasts (Fig. 1J). Thus, by day 7, NCSC cultures contained all major neural crest derivatives. Although it would have been preferable to isolate pure cell populations from the embryo, this was impossible. Premigratory NCSC share the dorsal neural tube with neuroepidermal progenitors (Fig. 1C). Migratory NCSC rapidly lose some of their developmental potentials [7, 29], and they intermingle with sclerotome cells. NCPs are distributed throughout the organism. Therefore, isolating individual types of tissue, such as dorsal root or sympathetic ganglia, would not have provided a comprehensive NCP transcriptome. Although not ideal, the current NCP transcriptome nevertheless contains all major NCP. Day-7 and day-14 NCSC cultures were still growing in size and contained nestin-expressing cells (e.g., Fig. 2D–2D‴), indicating that 7-day-old cultures used for the NCP library still contained stem cells/progenitors. This has likely led to an underestimate of the fold difference of differentially expressed genes in the NCSC-to-NCP comparison.

Figure Figure 2..

Msx2 and Myo10 immunoreactivity in EPI-NCSC, NCSC, and NCP. (A–A″): At day 2 after onset of emigration from the bulge explant, all EPI-NCSC are intensely Msx2-immunofluorescent. (A): Msx2 fluorescein fluorescence. (A′): Corresponding DAPI nuclear stain. (A″): Merged images. Immunofluorescence is predominantly nuclear. (B-B″): All NCSC expressed Msx2 at day 2 in culture. (B): Msx2 fluorescein fluorescence. (B′): Corresponding DAPI nuclear stain. (B″): Merged images. (C–C″): Most day-2 NCP have lost Msx2 immunofluorescence. Some cells in clusters (neuronal cells) still express Msx2. (C): Msx2 immunofluorescence. (C′): Corresponding DAPI nuclear stain. (C″): Merged images. (D–D″): Msx2 immunofluorescence in culture day-14 NCSC. Most cells, including neuronal cells in aggregates, have lost Msx2 immunofluorescence. Rare cells show asymmetric Msx2 immunofluorescence (D) (arrows). A subset expresses nestin (D′) (arrowhead). (D″): Merged Msx2 and nestin images. (D‴): Merged Msx2 (fluorescein), nestin (Texas red), and DAPI (blue) stains. Nestin and Msx2 are asymmetrically expressed, on the same side of the nucleus, but segregated from each other. (E–E″): Myosin X (Myo10) immunofluorescence. (E): All day-2 EPI-NCSC are intensely Myo10-immunoreactive. (E′): All day-2 NCSC are intensely Myo10-immunoreactive. (E″): By culture day 7, all NCP show low levels of Myo10-immunoreactivity. Scale bars = 20 μm (A–C″, E–E″), 10 μm (D–D‴). Abbreviations: DAPI, 4,6-diamidino-2-phenylindole; EPI-NCSC, epidermal neural crest stem cells; NCP, neural crest progeny; NCSC, neural crest stem cells.

Quality of LongSAGE Libraries

Three LongSAGE libraries were prepared with mRNA from approximately 90 48-hour EPI-NCSC, 90 48-hour NCSC, and 60 7-day-old NCP explants. A combined total of 103,858 LongSAGE tags were sequenced from EPI-NCSC (38,404), NCSC (31,064), and NCP (34,390). This translated into a total of 44,296 unique transcripts from the EPI-NCSC (15,190), NCSC (13,054), and NCP (16,052) libraries. A total of 16,821 tags could be annotated unambiguously for EPI-NCSC (5,696), NCSC (5,128), and NCP (6,097). Ambiguous tags that could be annotated to more than one gene (EPI-NCSC, 3,626; NCSC, 3,490; NCP, 3,852), and tags that could not be annotated (EPI-NCSC, 5,360; D2-NCSC, 4,438; D7-NCC, 5,516) were excluded from analysis. The latter could represent novel sequences and/or sequencing errors.

Several lines of evidence show the high quality and equivalence of the three LongSAGE libraries. First, as expected, housekeeping genes such as GAPDH (glyceraldehyde-3-phosphate dehydrogenase), HPRT, and ubiquitin C were expressed equally in the three LongSAGE libraries (p ≥ 1; Table 1). Second, as expected, tag distribution between the three libraries was similar (Table 2). Third, most genes expressed in the three libraries were expressed at similar abundance due to the similarity of the cells of origin and the prevalence of housekeeping genes (supplemental Tables S1, S2, and S3). Forty-three of 50 tags were the same between the NCSC and NCP libraries (supplemental Table S1). Forty of 50 tags were the same between the NCSC and EPI-NCSC libraries (supplemental Table S2), and 40 of 50 tags were common between the NCP and EPI-NCSC libraries (supplemental Table S3). Unambiguously annotated tags only were used for further analysis. For normalizing purposes, the tag numbers were expressed as “tags per million.” GC content bias [30] was unlikely to be a problem, because ditags were kept on ice at all times. Moreover, the increased length of LongSAGE tags (17-bp tags) compared with classical SAGE (10-bp tags) further minimized the likelihood of GC-related bias. The purity of the EPI-NCSC starting material is also reflected in the complete absence of cytokeratins, which are expressed by epidermal stem cells and keratinocytes.

Table Table 1.. Quality assessment of LongSAGE libraries
original image
Table Table 2.. LongSAGE tag distribution in NCSC, NCP, and EPI-NCSC LongSAGE libraries
original image

We have generated three high-quality LongSAGE libraries of medium size. The libraries have been deposited in Gene Expression Omnibus at; series number GSE4680.

Differentially Expressed Sequences Abundant in Both NCSC and EPI-NCSC

Comparison in silico of the NCSC and NCP libraries (Table 3) eliminated housekeeping genes and identified 7,755 unambiguous sequences that were expressed in NCSC, in NCP, or in both (Table 3). Of these, 119 sequences were significantly more abundant (≥twofold and p ≤ .05) in the NCSC library than in the NCP library, indicating that they are characteristic for NCSC. When compared in silico to the EPI-NCSC library, 91 of 119 mRNA sequences had a p value of ≥.07, indicating that they were expressed to a similarly high degree in both NCSC and EPI-NCSC. Genes with known biological process are listed in Table 4.

Table Table 3.. Library comparisons
original image
Table Table 4.. NCSC and EPI-NCSC-characteristic transcripts sorted according to known biological process
original image

Divergent EPI-NCSC and Bulge Epidermal Stem Cell Gene Profiles

Bulge epidermal stem cells and EPI-NCSC share the bulge as stem cell niche. To differentiate between these two types of colocalized stem cell, we next made a qualitative comparison between our 91 NCSC/EPI-NCSC-characteristic genes and the gene profile of infrequently cycling epidermal stem cells in the bulge by Tumbar et al. ([16]; GFPhigh minus GFPlow/β-4 integrin+ population; their supporting Table S2). Only one of 91 genes, myosin 1b (Myo1b; Mm.3390), was in common. This result confirms that at high stringency, bulge epidermal stem cell markers and NCSC/EPI-NCSC markers are distinct.

To further increase the stringency of our NCSC/EPI-NCSC marker genes, we subsequently compared the 91 NCSC/EPI-NCSC-characteristic genes with the 4,671 genes in the GFPhigh population by Tumbar et al. [16] (, which may also include basal layer and outer root sheath progeny. Nineteen of 91 genes were absent, 51 of 91 were common, and 21 of 91 genes could not be compared because they are not present on the MG-U74Av2 Affymetrix array used by Tumbar et al. [16]. The data indicated that, at this level of scrutiny, approximately 50% of genes are differentially expressed by all four types of stem cell: bulge epidermal stem cells, nonbulge epidermal stem cells, NCSC, and EPI-NCSC. They are listed in supplemental online Table S4. Collectively, these analyses confirmed, as expected, that EPI-NCSC and bulge epidermal stem cells have divergent gene profiles and that the different types of stem cell can be characterized according to well-defined and stringent criteria.

An NCSC Molecular Signature

According to the above in silico comparisons, we have defined an NCSC molecular signature that consists of a panel of 19 genes, which are expressed more abundantly by NCSC than by in vitro differentiated NCP (≥twofold and p ≤ .05), equally abundant in day-2 EPI-NCSC (p ≥ .07), and absent in bulge epidermal stem cells and nonbulge epidermal stem cells/progenitors (Table 5). Differential expression of signature genes was validated by quantitative RT-PCR (Table 5). Expression at the protein level was monitored by immunocytochemistry for two genes, Msx2 and myosin10 (Myo10).

Table Table 5.. List of genes comprising the neural crest crem cell molecular signature
original image

All day-2 EPI-NCSC were intensely Msx2-immunoreactive (100% ± 0% of total ± SD). Localization was primarily nuclear, but also cytoplasmic (Fig. 2A–2A″). Likewise, all day-2 NCSC were intensely Msx2-immunoreactive with predominantly nuclear localization (Fig. 2B–2B″; 100% ± 0%). In contrast, most day-7 NCP had lost Msx2 expression (9.6% ± 1.5%). Cells in clusters of neurons that were still intensely Msx2 immunofluorescent were a notable exception (Figs. 1E and 2C–2C″). By culture day 14, cells in neuronal aggregates were Msx2-negative and only rare Msx2-positive cells could be found within the culture (2.0% ± 0.7%), in particular at its periphery, where proliferating cells were still present. Some of these cells coexpressed Msx2 and nestin (Fig. 2D–Fig. 2D‴), the latter confirming the continued presence of NCSC/progenitors in day-14 cultures. Other Msx2-positive cells showed low levels of nestin expression or none (Fig. 2D–2D‴). In these cells at culture day 14, Msx2 and nestin immunoreactivity was localized asymmetrically in the cytoplasm. Msx2 immunoreactivity was segregated from nestin immunoreactivity, but both were localized asymmetrically at the same side of the nucleus (Fig. 2D–2D‴). This asymmetric distribution is reminiscent of the asymmetric distribution of Numb [31] and epidermal growth factor receptor [32] in neural stem cells and asymmetric protein distribution in epidermal cells [33]. Asymmetric subcellular distribution of immunoreactivity is of considerable interest, given that it appears to be an underlying mechanism for cell fate choices during development. It is conceivable that cells with nuclear and homogeneous cytoplasmic localization of Msx2 (Fig. 2A, 2B) and nestin (Fig. 1D″) in early cultures represent self-renewing stem cells, whereas cells with asymmetric Msx2/nestin distribution at day 14 (Fig. 2D–2D‴) represent more mature progenitors capable of giving rise to both a stem/progenitor daughter cell and a differentiating daughter cell.

Myo10 is a vertebrate-specific unconventional myosin, an actin-based motor to which a growing number of functions are attributed. They include relocalization of β-integrins, formation of adhesive structures, promotion of filopodial extension, cell-cell adhesion-associated signaling, neuronal development, and migration [34, [35]–36]. All day-2 EPI-NCSC (Fig. 2E) and NCSC (Fig. 2E′) were intensely Myo10-immunoreactive (100% ± 0% of total), whereas Myo10 was significantly downregulated in all day-7 NCP (Fig. 2 E″). In the center of day-7 NCP cultures, 0% ± 0% of cells were brightly Myo10-immunoractive, whereas intensely Myo10 immunofluorescent cells were present at the periphery of the cultures. Due to the high cell density and progressing cell differentiation, cells within day-7 cultures are no longer migratory, whereas cells at the periphery still migrate. Myo10 may thus be associated with neural crest cell migration.

Several signature genes are crucial for embryonic development and are known disease genes. Msx2 (Msh-like 2) has key functions in neural crest development. It is a homeobox gene immediately downstream of Pax3 [37] which is expressed at sites of epithelial-mesenchymal inductive interactions and which acts as a master gene in regulating smooth muscle differentiation [38], neurogenesis [38], and osteogenesis [39, [40]–41]. Loss of Msx2 function causes cranial tissue defects, whereas Msx2 overexpression causes aortico-pulmonary-septation defects (reviewed by Ramos and Robert [39]). Msx2 mutants also show alopecia and have curly whiskers [42]. Mutations in human MSX2 are associated with craniosynostosis type 2 [43], hypodontia [44], parietal foramina [45], and Saethre-Chotzen syndrome [46, 47].

Other known disease genes in the panel of signature genes include Calr, Ube4b, and Adam12. Targeted mutations in calretinin (Calr) cause neural tube defects, increased apoptosis in cardiomyocytes, and mid- to late-gestational lethality [48, 49]. Mice homozygous for a disruption in ubiquitination factor E4B (Ube4b) die by mid-gestation, whereas heterozygotes exhibit deficiencies in the heart and nervous system [50]. Homozygous Adam12 mouse mutants die postnatally [51], and mutations in the human ADAM12 gene cause cardiomyopathies [52].

As judged by genes that have a known function, the NCSC molecular signature represents a rounded picture of neural crest cell characteristics, because it includes mechanisms that direct lineage choice (Msx2), Wnt signaling (Pygo2), cell migration (Myo10, Vars2, Ets1), cell invasiveness (Adam12, Ets1), cell proliferation (Vars2), prevention of apoptosis (Ets1, Vdac1, Peg10), prevention of premature neuronal differentiation (Thop1, Rex3), regulation of growth factor function (Thop1, Adam12), early nervous system development (Crmp1), oncogenic activity (Peg10), and gene regulation (Pcbp4, Msx2, H1fx).

EPI-NCSC Are Distinct from Stem Cells/Progenitors in the Dermal Papilla

Several types of multipotent stem cell/progenitor have been identified in the skin, both in the epidermis and dermis. Applying criteria that distinguish between them is therefore of high importance to obtain better clarity with regard to stem cell identities in the skin.

Toma and collaborators have suggested that dermal skin-derived precursors localize to the dermal papilla and are of neural crest origin as they express the neural crest markers Snail/Slug, Twist, and Pax3 [53, 54]. The dermal papilla is a dermal structure outside the epidermal boundary of the hair follicle (Fig. 1A). It has potent follicle-inductive activity and is therefore an essential tissue in hair follicle morphogenesis [55]. Of note, the craniofacial dermis, including dermal papilla, is of neural crest origin, whereas all other dermis is not neural crest-derived [56]. Rendl et al. [24] used a genetic approach to isolate and purify various anagen hair follicle cell populations, including dermal papilla cells. They also found marker genes that are expressed by neurons and/or neural crest cells but came to the conclusion that overall the dermal papilla signature is unlikely to represent NCSC. Our data are in agreement with those of Rendl et al. [24]. We have compared our NCSC molecular signature to data in the above reports (Fig. 3) and to a third dermal papilla signature by O'Shaughenessy et al. [57]. None of the NCSC signature genes occurred in the three dermal papilla signatures (Fig. 3). Conversely, whereas some dermal papilla-expressed genes were present also in one or more of our libraries, other transcripts were not present in any of our three libraries. We confirmed by RT-PCR the absence of expression in day-2 EPI-NCSC of the dermal papilla-characteristic genes, alkaline phosphatase (Akp2), Zic1, and Hoxa9 (Fig. 3C). In contrast, the stem cell markers Tebp and nestin were abundantly expressed (Fig. 3C). We conclude that neither the facial (crest-derived) nor the back skin (noncrest) dermal papilla contains multipotent NCSC as defined by our stringent criteria. This does not preclude the presence of more mature neural crest-derived progenitors within the dermal papilla, in particular in the neural crest-derived dermal papilla of the whisker follicle.

Figure Figure 3..

Absence of dermal papilla (DP) markers in EPI-NCSC. (A): DP markers from Toma et al. [53] and Fernandes et al. [54] compared with the NCSC signature (Table 5). None of our signature genes is present. Pax 3, Sox 9, alkaline phosphatase (Akp2), and versican were not present in the EPI-NCSC library. Akp2 was not present in any of our three libraries. (B): None of our signature genes is part of the DP signature by Rendl et al. [24]. Akp2, Alx4, Hoxa9, Zic1, Cntn1, and Enpp2 were not present in any of our three libraries. (C): Verification by RT-PCR of the absence of DP signature genes in EPI-NCSC. Whereas Hprt, Tebp, and nestin were abundantly expressed in day-2 EPI-NCSCs, the DP signature genes Zic1, Hoxa9, and Akp2 were not expressed. CD34 was not expressed in any of our libraries. The absence of Cd34 transcripts in day-2 EPI-NCSC is confirmed here by RT-PCR (lane 7). Abbreviations: EPI-NCSC, epidermal neural crest stem cells; LongSAGE, long serial analysis of gene expression; NCP, neural crest progeny; NCSC, neural crest stem cells; RT-PCR, reverse transcription-polymerase chain reaction; SKP, skin-derived precursor.

EPI-NCSC Are Distinct from CD34+ Skin-Resident Stem Cells/Precursors

Another marker, CD34, is of importance because it is expressed by murine hematopoietic stem cells [58], adipocyte stem cells [59] and by a population of multipotent progenitors from the hair follicle [60]. Blood vessels are abundant in the skin, including in the dermal papilla. Furthermore, the lower part of anagen hair follicles extends into the lower dermis and is surrounded by adipose tissue. No Cd34 sequences were present in any of our three libraries. Furthermore, we have verified the absence of Cd34 transcripts in EPI-NCSC by RT-PCR (Fig. 3C). The lack of Cd34 expression sets EPI-NCSC apart from the nestin+/CD34+ follicular cells described by Amoh et al. [60], from hematopoietic stem cells, and from adipocyte stem cells.


In this study, we provide the first transcriptome for EPI-NCSC. Furthermore, we have prepared gene profiles of embryonic mouse neural crest cells and in vitro differentiated NCP. Together, the three LongSAGE libraries establish a foundation and provide a valuable resource for future research on embryonic and adult mouse neural crest cells.

According to the three gene profiles and published reports, we have defined by stringent criteria a molecular signature that is shared by embryonic NCSC and bulge-derived EPI-NCSC. By applying this panel of signature genes to markers/signatures for other types of skin-resident cells, we have provided unequivocal evidence that EPI-NCSC from anagen whisker follicles have a unique gene profile, which sets them apart from all other known skin-resident stem cells/progenitors.

In the present study, we have also identified novel genes, some of which are likely to be of importance in morphogenetic mechanisms. New knowledge about growth factor and receptor expression provided useful information for the expansion of EPI-NCSC into millions of cells so that they could be used in animal models of human disease, such as in a spinal cord injury model [3].


The authors indicate no potential conflicts of interest.


We thank Dr. M. Battle for comments on the manuscript. This work was supported by USPHS (United States Public Health Service) Grant NS38500 from the National Institute of Neurological Disorders and Stroke, National Institutes of Health (M.S-B.) and Charles E. Culpeper Biomedical Pilot Initiative Grant 03-128 from the Rockefeller Brothers Fund, New York (M.S-B.). The collagen type II (CIIC1) monoclonal antibody developed by F. Holmdahl and K. Rubin was purchased from the Developmental Studies Hybridoma Bank, University of Iowa. Z-J.Z. is currently affiliated with the Stem Cell Research Program, University of Wisconsin-Madison, Madison, Wisconsin, USA.