The C8/144B monoclonal antibody recognizes ... - Semantic Scholar

3179

Journal of Cell Science 111, 3179-3188 (1998) Printed in Great Britain © The Company of Biologists Limited 1998 JCS4622

The C8/144B monoclonal antibody recognizes cytokeratin 15 and defines the location of human hair follicle stem cells Stephen Lyle1,3, Melpo Christofidou-Solomidou2, Yaping Liu3, David E. Elder1, Steven Albelda2 and George Cotsarelis3,* University of Pennsylvania Medical Center, 1Departments of Pathology and Laboratory Medicine, 2Medicine and 3Dermatology, Philadelphia, PA 19104, USA *Author for correspondence (e-mail: [email protected])

Accepted 23 September; published on WWW 14 October 1998

SUMMARY Stem cells are vital for the homeostasis of self-renewing tissues such as the hair follicle. Epithelial stem cells have been implicated in tumorigenesis and wound healing, and their manipulation may have wide ranging applications including gene therapy and tissue transplantation. Rodent hair follicle stem cells have been localized to an area of the follicle called the bulge, however, the identification and characterization of human hair follicle stem cells has been hampered by a lack of cellular markers for this area. We have determined that the C8/144B monoclonal antibody, originally generated against a short intracytoplasmic peptide of CD8, preferentially immunostains hair follicle

bulge keratinocytes without staining the remaining hair follicle. Using expression cloning, we identified cytokeratin 15 as the keratinocyte protein recognized by the C8/144B monoclonal antibody. By delineating the bulge using this antibody, we demonstrated that bulge cells possess a stem cell phenotype characterized by their slowly-cycling nature, preferential proliferation at the onset of new hair follicle growth, high level of β1 integrin expression, and expression of cytokeratin 19.

INTRODUCTION

differentiating cells similar to other self-renewing tissues, such as bone marrow, gastrointestinal tract and epidermis. Unlike these other tissues, however, hair follicle growth (anagen) is interrupted by periods of regression (catagen) and rest (telogen), with subsequent regeneration (Cotsarelis, 1997; Paus, 1996). Based on LRC studies in mice, we proposed that stem cells responsible for this cyclical regeneration of the lower follicle reside in an area called the hair follicle bulge (Cotsarelis et al., 1990). Supporting this concept, Kobayashi et al. (1993) demonstrated that bulge cells from microdissected rat follicles possess in vitro characteristics of stem cells because they have a high colony forming efficiency. However, similar types of in vitro analyses by the same investigators and others on human hair follicles suggested that human hair follicle stem cells were located in the lower hair follicle well below the level of the bulge (Moll, 1995; Rochat et al., 1994). Therefore, the exact location of human hair follicle stem cells remains unclear, and this has hindered the characterization of these cells, as well as our understanding of their role in hair follicle biology, wound healing and tumor formation. To better define the location of human hair follicle stem cells, we performed LRC studies on human scalp grafted to immunodeficient mice. Our results demonstrated that long-lived LRCs were located in the bulge area, not in the lower follicle. Furthermore, using immunohistochemistry we demonstrated that the C8/144B monoclonal antibody, originally raised against a

The homeostasis of all self-renewing tissues, including the epidermis and hair follicle, is thought to be dependent on stem cells (Quesenberry and Levitt, 1979; Wright and Alison, 1984). As undifferentiated stem cells divide, they generate daughter cells that retain the stem cell phenotype, and daughter cells (called transit-amplifying (TA) cells) that undergo rapid proliferation and terminal differentiation to repopulate the tissue (Jones and Watt, 1993; Quesenberry and Levitt, 1979). Stem cells are generally slowly cycling, but they have a high proliferative potential and proliferate at times of tissue expansion such as during fetal development and wound healing (Potten, 1974; Potten et al., 1979; Cotsarelis et al., 1990; Morris, 1997). On the basis of these characteristics, epithelial stem cells have been identified in the epidermis and hair follicle as keratinocytes with a high in vitro proliferative potential or as long-lived, slowly-cycling ‘label-retaining cells’ (LRCs) in vivo (Barrandon and Green, 1987; Bickenbach and Mackenzie, 1984; Cotsarelis et al., 1990). Although no specific markers for epithelial stem cells are known, stem cells within the epidermis express higher levels of β1 integrin compared to surrounding cells, and cytokeratin 19 (K19) is present, though not exclusively, in hair follicle stem cells (Heid et al., 1988; Jones et al., 1995; Jones and Watt, 1993; Lane et al., 1991; Michel et al., 1996). The hair follicle contains rapidly proliferating and

Key words: Stem cell, Hair follicle, Integrin, Cytokeratin 19, Labelretaining cell, Cytokeratin 15

3180 S. Lyle and others CD8 peptide, preferentially identified human adult hair follicle bulge cells. Through expression cloning and immunoprecipitation, we determined that this antibody recognizes cytokeratin 15 (K15), not CD8, in follicular keratinocytes. Using this antibody to detect the bulge, we further analyzed the stem cell characteristics of these cells. We determined that bulge cells are generally slowly-cycling, longlived cells that proliferate at the onset of anagen. Bulge keratinocytes expressed high levels of β1 integrins on their surface, similar to epidermal keratinocytes that have a high colony forming efficiency (Jones and Watt, 1993). Collectively, our results support the concepts that human hair follicle stem cells localize to the bulge and that K15 is a marker for these cells. MATERIALS AND METHODS Immunostaining Adult scalp skin (obtained from the Cooperative Human Tissue Network) was either frozen in OCT compound or fixed in formalin or ethanol (70%, 4°C) and paraffin embedded. Frozen or paraffin sections were immunostained with mouse monoclonal antibodies (DAKO clones C8/144B and DK25; Novocastra clones 4B11 and UCH-T4; Coulter clone T8; anti-BrdU antibody from Boehringer-Mannheim). Ethanol or formalin-fixed tissue sections were steamed in citrate buffer (10 mM sodium citrate, pH 6.77) for 15 minutes prior to incubation. The AvidinBiotin complex technique was used for immunohistochemistry and Texas-Red or FITC conjugated anti-mouse antibodies were used for immunofluorescence. Double immunostaining was performed using the EnVision Doublestain System (DAKO) with anti-Ki-67 (Immunotech) as the first primary antibody and horseradish peroxidase-labeled secondary antibody followed by C8/144B and alkaline phosphataselabeled secondary antibody. Diaminobenzidine (Sigma) and Fast Red (DAKO) were the chromogens. Competition experiments were performed using a synthetic peptide with a sequence corresponding to the 14 carboxy-terminal amino acids of CD8 used to generate the C8/144B antibody (Mason et al., 1992): Cys-Lys-Ser-Gly-Asp-Lys-Pro-Ser-Leu-Ser-Ala-Arg-Tyr-Val, obtained from Biosynthesis, Inc. A molar excess of this peptide or an unrelated control peptide was incubated with the antibody overnight at 4°C and then used for immunohistochemistry of formalin-fixed tissue as above. Immunofluorescence staining of whole mounted hair follicles Hair follicle whole mounts were obtained by treating fresh human scalp skin with Dispase (Sigma, 12.5 mg/ml in DMEM) for 1 hour at 37°C followed by plucking and fixation in acetone for 10 minutes. The follicles were immunofluorescently stained for C8/144B, K19 (Am. Res. Products clone Ks 19.1) or β1 integrin. Double-labeling of follicles was adapted from the method of Jones et al. (1995) using C8/144B antibody followed by Texas Red-conjugated anti-mouse antibody and then FITC-conjugated anti-β1 integrin antibody (DAKO). All follicles were also stained with the Hoechst nuclear stain for 5 minutes and mounted with Vectashield (Vector). The whole mounts were imaged using a Leica confocal microscope. 2 µm thick optical sections were obtained through the mid-sagittal plane of the follicles. For composite views (Fig. 5E,F), ten serial 2 µm thick optical sections through the follicle were used to construct the images. The directly FITC-conjugated anti-β1 integrin antibody allows for the quantification of fluorescence levels and thus integrin expression by measuring the intensity across the basal pole of individual cells imaged through the confocal microscope as previously described (Jones et al., 1995). Label-retaining cells Slowly-cycling cells were identified as LRCs by long-term labeling techniques as previously described (Cotsarelis et al., 1990) with minor

modifications using adult (40- to 70-year-old) human scalp grafted onto adult 5- to 7-week-old CB 17/Icr-scid/scid mice (Charles River). Grafting was done according to previously described techniques (Christofidou-Solomidou et al., 1997). Three weeks after grafting, 5bromo-2′-deoxy-uridine (BrdU) (Sigma) 0.4 mg/day was administered continuously for 2 weeks (total dose=5.6 mg) by 2 Alzet osmotic minipumps (model 1001, Alza corp., Palos Alto, CA) implanted intraperitoneally following the manufacturer’s instructions. Pumps were surgically removed after 2 weeks of continuous labeling, and mice were killed after chase periods of 0-4 months. Sections of ethanol or formalin-fixed, paraffin embedded skin were stained with anti-BrdU and/or the C8/144B monoclonal antibodies after steaming in citrate buffer for 15 minutes as described above. cDNA library construction and expression cloning mRNA was extracted from 1.5 g of flash-frozen, 18-week human fetal scalp (Advanced Bioscience Resources) using oligo(dT) cellulose and the FastTrack mRNA isolation protocol (Invitrogen). cDNA was synthesized from 7.5 µg of mRNA according to the ZAP Express cDNA Synthesis protocol (Stratagene). The cDNA was fractionated over a Sepharose CL-2B column and fractions containing cDNA sizes from 0.5 to >4.5 kb were pooled. The pooled cDNA was ligated into the ZAP Express vector and packaged into XL-1 Blue MRF′ cells. The resulting 1.06×106 clones were amplified to a final titer of 3×106 pfu/µl. The clones were screened according to the plaque screening method (Stratagene picoBlue Immunoscreening Kit). Briefly, a phage mixture representing 1×106 clones was combined with E. coli and plated onto twenty 150 mm plates (50,000 clones/plate). After ~3.5 hours, IPTG (10 mM)-impregnated nitrocellulose membranes were overlayed on the plates and incubated for 3.5 hours at 37°C. The membranes were washed and incubated with the C8/144B antibody (1:1,000 in TBST: 20 mM Tris-HCl, 150 mM NaCl, 0.05% Tween-20, pH 7.5) overnight at 4°C. The membranes were then washed and incubated with sheep anti-mouse-Ig-AP, FAB fragments (BMB) for 1 hour at 25°C. Thirtytwo positive clones were visualized with NBT-BCIP color development solution (BMB). During plaque purification, all positive clones were screened with the secondary anti-mouse antibody to ensure specificity of the C8/144B monoclonal antibody. The positive clones were also screened with a mixture of anti-CD8 antibodies (Coulter clone T8 and Novocastra clone UCH-T4) and were found to be negative. The purified clones were excised from the phage into plasmid form according to the In Vivo Excision protocol (Stratagene). Synthesis of follicle-specific cDNA probe Total RNA was isolated from 100 hair follicles plucked from Dispasetreated (as above) fresh human scalp using guanidine isothiocyanate and purified over Glassmax spin columns (Gibco). An aliquot of the total RNA (3 µg) was treated with DNase 1(Gibco) for 15 minutes at 65°C and precipitated with phenol-chloroform. Approximately 1 µg of this follicle-specific RNA was used to synthesize and amplify cDNA with the SMART PCR cDNA Synthesis Kit (Clontech). A portion of the amplified cDNA (100 ng) was used as a template to produce 32P-labeled probe with the Rediprime random primer labeling kit (Amsersham) and [32P]dCTP (50 µCi). The probe was purified over Microspin S-300 HR columns (Pharmacia). Southern dot blots/clone analysis The purified plasmids from the thirty-two positive clones were blotted onto nylon and hybridized with the follicle-specific 32P-labeled probe for 4 hours at 65°C. The blot was washed with 2× SSC/0.1% SDS at 65°C for 30 minutes and then 0.1× SSC/1% SDS for 15 minutes at room temp. The blot was developed on Kodak XAR film overnight. Positive clones were sequenced from the 5′ and 3′ ends with T3 and T7 primers using Dye Terminator automated cycle sequencing (Applied Biosystems) on an ABI 377 sequencer. Nucleotide sequences were compared to Genbank/EMBL databases using the BLAST algorithm (Altschul et al., 1990).

Human hair follicle stem cells 3181 Immunoprecipitation of in vitro translated K15 protein The 5′ end of one of the six clones representing K15 cDNA (clone 1C) started at nucleotide 30 of the published K15 nucleotide sequence (Leube et al., 1988) and included the N-terminal methionine start codon. This clone was used as a template for in vitro transcription/translation. The K15 cDNA clone (clone 1C in Fig. 7A) was translated in vitro, in the presence of [35S]methionine, using the TNT Coupled Reticulocyte Lysate SystemTM (Promega) according to the manufacturer’s protocol. After preclearing with streptavidinSepharose (Gibco) beads, the reaction was immunoprecipitated with the C8/144B monoclonal antibody or an unrelated, isotype-matched (IgG1) monoclonal antibody in 1× PBS buffer containing 0.5% NP40, 0.1% SDS and 1% BSA at 4°C for three hours, followed by overnight incubation with anti-mouse biotinylated IgG and streptavidinSepharose (Gibco) beads. The precleared beads and immunoprecipitated beads were washed and resuspended in 15 µl of 1× SDS loading dye. After boiling, samples were separated on 10% SDS-PAGE, and subjected to autoradiography.

RESULTS C8/144B monoclonal antibody staining delineates the human hair follicle bulge Although the bulge is readily apparent morphologically in adult human telogen follicles, anagen follicles often possess no obvious bulge area. Without a cellular marker for bulge keratinocytes investigators have used the arrector pili muscle to approximate the location of these cells, thus, making direct comparisons between studies difficult (Moll, 1995, 1996; Rochat et al., 1994; Yang et al., 1993). During unrelated experiments, we surprisingly discovered that the C8/144B monoclonal antibody, originally raised against the carboxy-terminal peptide of the T cell protein CD8 (Mason et al., 1992), selectively immunostained bulge keratinocytes of hair follicles, as well as a subset of lymphocytes, in tissue sections of adult human skin (Fig. 1). In addition to the morphologically evident bulges occasionally seen in tissue sections (Fig. 1D), the C8/144B monoclonal antibody immunostained a discrete area of the outer root sheath basal layer below the sebaceous gland duct at the attachment site of the arrector pili muscle of anagen, catagen and telogen follicles (Fig. 1A-C). The staining extended variably above and below the exact arrector pili muscle attachment site, but was usually sharply demarcated, while the upper portion of the follicle (infundibulum) and overlying epidermis were generally negative. The C8/144B antibody also stained the basal layer of keratinocytes at the bottom of the telogen follicle traditionally referred to as the ‘secondary germ’ or telogen germinal unit (Fig. 1C) (Headington, 1984). C8/144B + cells were present in the bulge area of follicles from a variety of anatomic sites such as scalp, eyebrow and extremity (data not shown). In approximately one-third (53/150) of scalp anagen follicles in which the bulb area of the follicle was apparent, we observed a few isolated weakly positive cells in the immediate suprabulbar outer root sheath of the lower follicle as well. Nonetheless, by immunostaining, the keratinocyte protein detected by C8/144B antibody appears preferentially located in the hair follicle bulge, and serves as a marker for these cells. Slowly-cycling LRCs are located in the human hair follicle bulge area defined by C8/144B staining By taking advantage of the human skin/scid mouse model, we studied the distribution of LRCs within human hair follicles. Hair

Fig. 1. Human hair follicles. (A) Anagen follicle stained with the C8/144B antibody. The C8/144B+ bulge cells (brown) form a cylinder of basal cells in the outer root sheath of the lower isthmus. Bar, 50 µm. (B) Catagen follicle with C8/144B+ bulge cells at the lower portion of the isthmus, above the degenerating lower follicle. Bar, 50 µm. (C) Telogen follicle with C8/144B+ bulge cells in the basal layer of the outer root sheath surrounding the club hair as well as in the secondary germ. Bar, 50 µm. (D) Early anagen follicle of adult skin with morphologic bulges showing positive staining. Bar, 20 µm. The epidermis is negative for staining. Bar, 20 µm. Brackets denote bulge; epi, epidermis; dlf, degenerating lower follicle; sg, secondary germ; irs, inner root sheath; ors, outer root sheath.

follicles that retain a normal morphologic and functional appearance have been successfully transplanted to immunocompromised mice in the past (Gilhar et al., 1988). These transplanted follicles produce a normal appearing hair shaft, and transit through the different phases of the hair follicle cycle (Van Neste et al., 1996). Because hair normally is shed from grafts 2 to 4 weeks after grafting in a well-described phenomenon known as ‘telogen effluvium,’ and then subsequently regrows, we reasoned that hair follicle stem cells must be proliferating during this transition from telogen to anagen to generate a new lower follicle that will produce a new hair. To label these proliferating stem cells, we administered BrdU continuously for 2 weeks from

3182 S. Lyle and others Table 1. Proliferative characteristics of hair follicle subpopulations during follicle cycling Anagen Stem cells Matrix cells

Quiescent Proliferating

Catagen Quiescent Degenerating

Telogen Quiescent Absent

Anagen onset*

Anagen

Proliferating Regenerating/proliferating

Quiescent Proliferating

Based on this and previous work (Commo and Bernard, 1997; Cotsarelis et al., 1990; Moll, 1995; Wilson et al., 1994), the proliferative activity within the hair follicle epithelium is summarized. To identify hair follicle stem cells as label-retaining cells, the proliferative behavior of the hair follicle throughout the hair follicle cycle must be considered. The nucleoside analog, BrdU, is infused during anagen onset (asterisk) to label the stem cells as they proliferate. The stem cells retain the label as they remain quiescent during the remainder of the hair cycle, while the proliferating matrix cells dilute out their label.

3-5 weeks after grafting. We labeled for this prolonged period to increase the likelihood of ‘capturing’ stem cell proliferation, because these cells are thought to proliferate only briefly to produce TA cells that repopulate and regenerate the follicle (Cotsarelis et al., 1990). The proliferative behavior of the hair follicle epithelium throughout the hair cycle is summarized in Table 1. In addition to cells within the hair follicle bulge, nearly every basal epidermal cell was labeled after 2 weeks of continuous labeling, as seen previously in mouse skin (Cotsarelis et al., 1990; Morris et al., 1985). To identify slowly-cycling LRCs, the grafts were chased for 4 months after the labeling period. During this time, rapidly proliferating TA cells of the lower follicle dilute their label, and only slowly-cycling LRCs remain labeled (Table 1). After the 4 month chase period, LRCs were present in the bulge areas of hair follicles (Fig. 2B,C). No LRCs were present in the bulb area or lower outer root sheath of over 60 human hair follicles examined (Fig. 2D,E). Occasional LRCs were present in the upper follicle (infundibulum) and in the epidermis as well. In addition, other normally slowly-cycling cells such as fibroblasts and eccrine ductal epithelia contained LRCs. However, within the hair follicle, LRCs were concentrated in bulge areas delineated by C8/144B staining (Fig. 3). C8/144B+ keratinocytes possess the proliferative behavior of hair follicle stem cells Although others have reported that the bulge area possesses the lowest percentage of actively proliferating cells relative to other portions of the follicle (Commo and Bernard, 1997; Moll, 1995), we have established that bulge cells are slowlycycling because they retain BrdU for at least 4 months. To further understand the proliferative behavior of the bulge cells, we examined their expression of the Ki-67 proliferation antigen (Gerdes et al., 1983) throughout the hair follicle cycle. During anagen onset, many C8/144B+ cells expressed the Ki67 proliferation antigen indicating their entrance into the cell cycle (Fig. 4A). In later anagen, a column of proliferating Ki67+, C8/144B-negative cells formed beneath the bulge area (Fig. 4B). During mid-anagen, when bulge cells form a relatively flattened, concentric cylinder in the outer root sheath, the proliferating cells were predominantly in the basal and suprabasal cells of the lower third of the follicle (Fig. 4C,D), as well as in the rapidly dividing matrix cells which produce the hair shaft (not shown), as has been previously reported (Moll, 1995). However, no mid-anagen, catagen or telogen follicles contained C8/144B+ keratinocytes that were immunoreactive for Ki-67, indicating that these cells are quiescent during the long anagen growth phase as well as catagen and telogen. They only briefly proliferate at anagen onset to generate rapidly proliferating progeny that regenerate a new follicle. These findings support the concept that the

lower ‘transient’ portion of the follicle is composed of TA cells which are important for hair growth during anagen but are dispensable for hair follicle cycling (Cotsarelis et al., 1990). C8/144B+ cells express high levels of β1 integrins In addition to their slowly-cycling nature, another important characteristic of stem cells is their high proliferative capacity. In keratinocytes, this proliferative capability has been studied in vitro by examining the clonogenicity of individual cells through serial passage. Approximately 5% of adult epidermal basal cells possess a ‘holoclone’ phenotype characterized by high reproductive capacity and low level of terminal differentiation, and these cells are thought to represent stem cells (Barrandon and Green, 1987). Another indicator of proliferative capability is colony forming efficiency (colonies per number of cells plated) that is thought to correlate to the number of stem cells in a tissue (Jones and Watt, 1993; Kobayashi et al., 1993). Epidermal keratinocytes with high colony forming efficiency express high levels of β1 integrin on their surface, and have been termed ‘integrin bright’ (Jones et al., 1995). Integrin bright epidermal keratinocytes possess fewer S-phase cells on average than integrin dull cells, suggesting that these cells divide infrequently under steady state conditions. High surface integrin expression is therefore a characteristic of epidermal stem cells, and these integrin bright cells have been localized to specific areas of the epidermis, such as the tips of the rete ridges in glabrous skin, where stem cells were previously described based on morphologic and kinetic criteria (Lavker and Sun, 1982). The exact localization of β1 integrin bright cells in the hair follicle has not been reported, although Jones et al. (1995) described a subpopulation of integrin bright cells within K19-positive follicular keratinocytes. In addition, Moll (1995) found intense β1 integrin immunostaining of the basal cells of the bulge area and of the bulb matrix keratinocytes, although quantitative analysis of integrin levels using confocal microscopy was not performed. To accurately identify integrin bright outer root sheath cells and analyze their relationship to C8/144B+ cells, we examined Table 2. β1 integrin expression of hair follicle epithelium Fluorescence* Ratio‡

Bulge cells

Nonbulge ORS

Matrix cells

161±6

44±2 3.7:1

66±2 2.4:1

*Average fluorescence measured in arbitrary units of pixel intensity on a linear scale from 0-255 (n=25). ‡Ratio of fluorescence intensity (integrin bright bulge cells:integrin dull cells). ORS, outer root sheath.

Human hair follicle stem cells 3183 immunofluorescently stained hair follicle whole mounts using confocal microscopy. The relative levels of β1 integrin expression were determined using a FITC-conjugated monoclonal primary antibody as previously described (Jones et al., 1995). Integrin bright cells co-distributed precisely with C8/144B+ cells in the bulge area of anagen follicles (Fig. 5). Outer root sheath cells

above and below the C8/144B+ bulge area showed only weak β1 integrin positivity. The C8/144B+ cells expressed approximately four-fold higher levels of β1 integrins than neighboring outer root sheath cells (Table 2). The hair matrix cells in the bulb area of the follicle expressed an intermediate level of β1 integrin and the ratio between C8/144B+ bulge cells and the C8/144B negative matrix cells was approximately 2.5:1. In telogen follicles, C8/144B expression co-localized with intense β1 integrin staining in the basal layer of the outer root sheath surrounding the club hair, while the C8/144B negative suprabasal cells demonstrated an intermediate level of β1 integrin positivity similar to the matrix cells (Fig. 5E,F). Serial sections of frozen scalp skin, stained individually for C8/144B or β1 integrin, also demonstrated that the C8/144B+ bulge was qualitatively ‘integrin bright’ (data not shown). In addition, the few discrete C8/144B+ cells present in the lower portion of several follicles in frozen sections showed intense β1 integrin immunofluorescence. Because β1 integrins are important for adhesion of keratinocytes to substrates in culture, the high level of integrins in the bulge area may explain why follicles in explant culture develop outgrowths more frequently from this area of the follicle (Moll, 1996; Yang et al., 1993). C8/144B-positive bulge cells express K19 Primarily because K19 is present in label-retaining bulge cells in mice (Michel et al., 1996), K19 is purported to be a stem cell marker. We therefore examined K19 expression in wholemounted human follicles and frozen tissue sections using immunofluorescence and immunohistochemistry. Although K19 positive cells co-localized with C8/144B+ cells in the basal layer of the outer root sheath within the bulge area (Fig. 6), K19+ cells extended downward throughout the entire basal layer of the outer root sheath from the bulge to the bulb in whole-mounted follicles (Fig. 6). Thus, C8/144B+ cells comprised a subset of K19+ cells in human hair follicles. Frozen section immunohistochemistry also showed K19 reactive cells throughout the basal outer root sheath, while C8/144B immunoreactivity was limited to the basal layer of the follicle bulge. These findings are in line with previous reports of K19 expression in outer root sheath cells throughout the full length of scalp hair follicles including the bulb (Heid et al., 1988; Lane et al., 1991). C8/144B monoclonal antibody recognizes K15 Several additional anti-CD8 monoclonal antibodies, developed by immunizing mice with human thymocytes or peripheral blood lymphocytes, did not stain the hair follicle bulge on Fig. 2. Slowly-cycling, label-retaining cells of human hair follicles. Human scalp grafted onto SCID mice was labeled continuously with BrdU for 2 weeks and then collected after a 4-month chase period. (A) Section of graft showing two follicles in human skin adjacent to mouse skin. LRCs, detected by immunohistochemistry for BrdU, are present in the bulge area at the lower portion of the isthmus. (B and C) High power views of bulges 1 and 2, respectively, demonstrating LRCs in the basal layer of the outer root sheath at the attachment of the arrector pili muscle. LRCs are also commonly seen in eccrine ducts, another slowly-cycling epithelium. (D and E) High power views of lower follicle and bulb, respectively. No LRCs are present in lower follicle or bulb. Bars: 100 µm (A); 20 µm (B-E). hf, human hair follicle; mf, mouse hair follicle; apm, arrector pili muscle; ecc, eccrine duct.

3184 S. Lyle and others

Fig. 3. C8/144B+ LRCs. Human skin/SCID mouse chimera labeled with BrdU. After a four-month chase period, serial sections were immunofluorescently stained for C8/144B (A) and BrdU (B). This high power image of the outer root sheath demonstrates that the long-term label-retaining cells (arrowheads) are present within the C8/144B+ bulge cells. Bar, 20 µm.

frozen sections or formalin-fixed paraffin-embedded tissue (data not shown), suggesting that the C8/144B monoclonal antibody was not recognizing CD8 on bulge keratinocytes. In addition, bulge keratinocytes had a granular cytoplasmic staining pattern in contrast to the delicate membrane staining of lymphocytes. Competition experiments with a synthetic peptide corresponding to the sequence used to produce the antibody, abolished the immunostaining, while an unrelated peptide did not affect the staining, thus, confirming the specificity of the antibody for epitope(s) present on bulge cells and lymphocytes (data not shown). The sequence of the CD8 peptide used to generate the antibody shared no homology to other proteins, also suggesting that this antibody cross-reacts with a keratinocyte protein. Although this type of crossreactivity of monoclonal antibodies has been previously described, Kramer (1997) and Keitel (1997) recently demonstrated that monoclonal antibodies commonly have cross-reactivity for homologous proteins and often demonstrate polyspecificity for unrelated proteins with completely different primary amino acid sequences. Other examples of monoclonal antibodies displaying polyspecificity to keratin proteins also have been reported (Weinberg and Yuspa, 1997). To determine the keratinocyte protein recognized by the C8/144B monoclonal antibody, we constructed and screened a human fetal skin cDNA expression library, since bulges are prominent structures in fetal hair follicles (Holbrook and Minami, 1991). We screened one million clones and found 32 positives. Each clone was plaque-purified and the phage was excised into the plasmid form (see Materials and Methods). We further screened these clones with Southern dot blot hybridization using 32P-labeled amplified cDNA from plucked hair follicles (see Materials and Methods; Fig. 7A). Sequence analysis of the 5′ and 3′ ends of the resulting nine positive clones revealed that six were identical to human K15. The remaining three clones were nearly identical to each other and represented a genomic Alu-sequence with a 600 bp open reading frame and 3′ poly(A) tail that resulted in reverse transcription during cDNA synthesis. To confirm that the C8/144B monoclonal antibody recognized human K15, we initially attempted to immunoblot protein from human fetal skin and a keratinocyte cell line (HaCaT), but no specific band was evident, suggesting that the antibody does not

Fig. 4. Human scalp double-stained for Ki-67 (brown/black nuclear staining) and C8/144B (red cytoplasmic staining). (A) Follicle at anagen onset showing C8/144B+ bulge cells co-expressing Ki-67 proliferation antigen (arrows indicate double-labeled cells). Bar, 25 µm. (B) Early anagen follicle showing C8/144B+, Ki-67-negative bulge and a column of C8/144B-negative, Ki-67+ cells above dermal papilla. Asterisk, dermal papilla. Bar, 10 µm. Mid-anagen follicle has no proliferating cells in C8/144B+ bulge area (C) and numerous proliferating cells in suprabasal and basal layers of the lower portion of follicle (D). Bars, 25 µm.

recognize denatured K15 protein. We therefore used immunoprecipitation to determine whether this antibody would recognize the in vitro translated product of a full length K15 cDNA. The resulting band of the expected 50 kDa size confirmed that the C8/144B monoclonal antibody recognized human K15 (Fig. 7B). The faint bands in lanes 1 and 2 can be accounted for by nonspecific binding of K15 to Sepharose beads. DISCUSSION Stem cells and hair follicle cycling During anagen, the hair is produced by rapidly proliferating matrix keratinocytes in the hair bulb (Fig. 8). The duration of

Human hair follicle stem cells 3185

Fig. 5. Co-distribution of C8/144B immunostaining and β1 integrin expression in whole-mounted follicles examined by confocal microscopy. Plucked follicles were simultaneously stained with antiβ1 integrin and with the C8/144B antibodies. The basal layer of the outer root sheath in the upper portion of the anagen follicle shows intense β1 integrin (green, A) and C8/144B staining (red, B). Bars, 50 µm. (C and D) High power 2 µm confocal optical views of the upper bulge area of the follicle (rectangles in A and B, respectively). C8/144B and β1 integrin immunostaining co-distributes precisely to the same (bulge) keratinocytes within the follicle. Bars, 20 µm. Outer root sheath cells above the bulge area show very weak β1 integrin expression. In telogen follicles (E, phase contrast), C8/144B+ cells and intense β1 integrin immunofluorescence co-distribute to the basal layer of the outer root sheath surrounding the club hair (F). (E and F) Composite confocal images of ten serial 2 µm thick optical views. Bars, 10 µm. The double exposure of green β1 integrin and red C8/144B immunofluorescence produces orange in double-labeled cells. The C8/144B negative suprabasal cells surrounding the club hair are weakly β1 integrin positive. Asterisk, dermal papilla.

anagen determines the final length of the hair, and human scalp follicles may stay in anagen continuously for over 8 years. Primarily because of this lengthy proliferative period, which suggests that matrix cells have a high in vivo proliferative potential, investigators have assumed that human hair follicle stem cells reside in the matrix. It was thought that stem cells migrate downward from the base of the telogen follicle during anagen onset to the matrix cells, and then back upward to the lower telogen follicle during catagen (Kligman, 1959). Identifying the exact location of human hair follicle stem cells is important for understanding hair follicle biology, and ultimately these results should also facilitate the elucidation of the role of bulge cells in maintaining the homeostasis of the epidermis, since the follicle outer root sheath serves as a reservoir for epidermal keratinocytes (Al-Barwari and Potten, 1976). To localize human hair follicle stem cells, Rochat et al. (1994) examined the in vitro proliferative behavior of dissociated keratinocytes from microdissected portions of human hair follicles. They suggested that an area of the hair follicle well below the bulge, but above the bulb, in the lower

outer root sheath contained the highest percentage of hair follicle stem cells (Rochat et al., 1994). Because the lower outer root sheath is present only during anagen, and undergoes degeneration during catagen, they hypothesized that hair follicle stem cells must migrate during the hair cycle, although they did not examine follicles in catagen or telogen to demonstrate that the remaining, permanent portion of these follicles possessed an expected high colony forming efficiency. In addition, the location of stem cells in an area present only transiently is not in line with evidence from other epithelial systems, such as the palm and corneal epithelium, where stem cells are strongly attached to the basement membrane, and are located in permanent, well-protected areas (Cotsarelis et al., 1989; Lavker and Sun, 1982). Our results clearly show that the slowest-cycling keratinocytes, identified as LRCs, were concentrated in the bulge area of the human hair follicle, not in the lower outer root sheath or bulb. We show that the stem cell-rich bulge area can be defined by the C8/144B monoclonal antibody which recognized K15. The discrepancy between the LRC results and the colony forming efficiency analysis suggests that the latter may not accurately reflect the location of bona fide epithelial stem cells in human follicles. Current cell culture systems used to determine colony forming efficiency apparently select for populations of clonogenic keratinocytes that include stem cells as well as more differentiated TA cells (Kobayashi et al., 1993; Li et al., 1998), suggesting that the lower outer root sheath cells may contain ‘early’ TA cells resulting in a high colony forming efficiency. The future use of the C8/144B antibody as a marker for epithelial stem cells should lead to the development of culture conditions that selectively support stem cell growth and can differentiate between stem cells and early TA cells that still possess a high proliferative potential. Our findings that human bulge cells selectively express K15 throughout all stages of the hair cycle in different types of follicles indicate that the bulge is composed of a well-defined, permanent population of cells within the hair follicle outer root sheath. C8/144B immunostaining in the follicle was found in the basal cell layer of the outer root sheath of the lower isthmus and the secondary germ, the traditional site of follicular stem cells in telogen follicles (Silver and Chase, 1970). Our operational definition of the bulge, therefore, is not simply limited to the morphologic bulges that attach to the arrector pili muscle, but also includes these other areas delineated by C8/144B monoclonal antibody staining. Although historically the secondary germ cells were thought to travel downward to the matrix keratinocytes during anagen onset, we found no evidence of LRCs in anagen bulbs. This suggests that matrix keratinocytes in the human follicle are rapidly proliferating keratinocytes derived from the bulge cells. Thus, despite the overt differences between mouse and human follicle size and proliferative behavior, their stem cell populations seem to be located in analogous bulge areas. Bulge cells possess an epithelial stem cell phenotype Bulge cells preferentially proliferate at the onset of anagen Within murine skin, hair follicle stem cells are generally slowly cycling but they proliferate at the onset of anagen, or in response to population depletion caused by wounding or other proliferative stimuli (Silver et al., 1969; Wilson et al., 1994).

3186 S. Lyle and others

Fig. 6. C8/144B+ cells comprise a subset of K19+ cells in wholemounted follicles. Whole-mounted anagen follicle stained with C8/144B monoclonal antibody (A). (B and C) High power, 2 µm confocal optical views of the bulge area and bulb (rectangles), respectively. C8/144B+ keratinocytes are present in the bulge area but not the bulb. (C8/144B, red; Hoechst nuclear stain, blue). Whole-mounted anagen follicle stained for K19 (D) shows immunofluorescence throughout the outer root sheath. (E and F) 2 µm confocal optical views of the bulge and bulb (rectangles), respectively. Both bulge and bulb areas are K19+ (K19, red; Hoechst nuclear stain, blue). Bars: 100 µm (A and D); 25 µm (B,C,E,F).

The in vivo proliferative behavior of human hair follicle bulge cells during the hair cycle has not been addressed, primarily because of the lack of a marker for these cells. Our results parallel our findings in mice (Cotsarelis et al., 1990; Wilson et al., 1994), and clearly show that bulge cells are extremely slowly cycling, retaining BrdU for over 4 months, but proliferate at the onset of anagen to produce a new lower follicle. Based on our results, we propose that the bulge area of the human hair follicle, defined by C8/144B immunostaining, serves as a reservoir of stem cells which cyclically regenerate the hair follicle (Fig. 8) An analysis of proliferative activity using the combination of C8/144B and Ki-67 immunostaining could potentially serve as a bioassay to study the effectiveness of an agent’s ability to stimulate follicular stem cells. Bulge cells are β1 integrin bright Populations of epidermal keratinocytes selected for high levels of β1 integrin expression (integrin bright) demonstrate a higher colony forming efficiency than epidermal keratinocytes with low integrin expression (integrin dull) (Jones et al., 1995). Because approximately 40% of basal epidermal keratinocytes are integrin bright, and the percentage of stem cells has been estimated at 5 to 10% (Barrandon and Green, 1987; Potten and Morris, 1988), integrin bright cells probably encompass both stem cells and TA cells (Jones et al., 1995). Within the hair follicle, however, β1 integrin bright cells co-distribute precisely with C8/144B+ cells that comprise only a minority of basal follicular cells (approximately 15% of follicular basal cells in an anagen follicle) (Fig. 5). In addition, integrin bright cells in the bulge carry an almost 4-fold higher level of integrin immunofluorescence compared to surrounding keratinocytes. In contrast, integrin bright cells in the epidermis express 2-3 times more β1 integrin than surrounding cells (Jones et al., 1995). This suggests that the highest β1 integrin expressing keratinocytes are located in the hair follicle bulge, and that a spectrum of β1 integrin expression exists throughout even the integrin bright

cell population. Therefore, by further quantitating degrees of β1 integrin on keratinocytes, at least in vivo, it may be possible to further subdivide integrin bright keratinocytes into stem cells and TA cells. Our evidence that bulb matrix keratinocytes express an intermediate level of integrin intensity supports this concept, because these cells have a very high but finite proliferative potential and can be considered TA cells. These findings, together with our LRC data, provide strong evidence that human bulge cells possess perhaps two of the most important criteria for defining an epithelial stem cell: slowly-cycling nature and high proliferative potential indicated by intense β1 integrin expression. Use of the C8/144B marker for follicular bulge cells should expedite the development of methods to specifically isolate hair follicle stem cells and the identification of additional stem cell markers. Cytokeratin 15 as a marker for hair follicle stem cells Within stratified squamous epithelia, such as the epidermis, hair follicle and cornea, cytokeratin expression is generally restricted to well-defined populations of cells in similar states of differentiation (Lane et al., 1991; Schermer et al., 1986; Schirren et al., 1997). For example, within the epidermis, the keratin pair K5/K14 is expressed predominantly in the basal layer of the epidermis which is composed of proliferating, relatively undifferentiated keratinocytes (Coulombe et al., 1989), while the more differentiated, postmitotic cells in the suprabasal layer express K1/K10. Similarly, in the cornea, the more differentiated central corneal epithelium expresses K3/K12 throughout the basal and suprabasal layers, while the stem cell rich peripheral cornea (limbus) only expresses these keratins in the more differentiated suprabasal cells (Cotsarelis et al., 1989; Schermer et al., 1986). Our data suggest that both K15 and K19 are markers of relatively undifferentiated keratinocytes in the bulge, however K15 appears restricted to the permanent portion of the follicle containing β1 integrin bright LRCs, while K19 is present in more differentiated TA cells in the lower follicle as well.

Human hair follicle stem cells 3187

Fig. 7. C8/144B antibody recognized K15. (A) Southern dot blot of 32 clones isolated from expression cloning and probed with 32Plabeled, amplified hair follicle cDNA. Six (1A, 1C, 12A, 12B, 13B, 14C) of nine positive clones represented K15. (B) The keratin 15 in vitro translated protein was immunoprecipitated by the C8/144B antibody giving the expected 50 kDa product (lane 3). Lanes 1 and 2 represent immunoprecipitations with no antibody and an unrelated mouse monoclonal IgG, respectively.

Jones et al. (1995) determined that K19+ follicular keratinocytes express higher levels of integrins than K19 negative cells, but the ratio of integrin fluorescence intensity for K19+ to K19 negative cells is lower than the ratio of integrin-bright to integrin-dull cells in the epidermis. They suggested that only a subset of the follicular K19+ cells represent true stem cells. Based on our data, it is likely that the K15 positive keratinocytes comprise this subset. These studies support the notion that loss of K15 expression may be one of the earliest signs of the transition from stem to TA cell, and that K15 negative/K19 + cells may be an indication that these cells are ‘early’ TA1 cells. We identified low levels of C8/144B immunostaining in a small but discrete cluster of cells in the lower outer root sheath of approximately one-third of anagen follicles, in an area where Rochat found 24% of the total colony forming cells present within dissected adult follicles (Rochat et al., 1994). Because these few discrete cells were also qualitatively integrin bright, they may represent ‘early’ TA cells that have migrated downward during anagen. The presence or absence of these C8/144B+ cells in the lower outer root sheath could indicate the position within the hair follicle cycle (e.g. early versus late anagen), although future studies of synchronized anagen follicles are needed to address this issue. Integrins have been shown to mediate the adhesion of freshly isolated keratinocytes (Adams and Watt, 1991) and there may be a direct correlation between integrin expression, cell adhesion and cell proliferation (Clark and Brugge, 1995). The striking co-distribution of C8/144B staining and intense β1 integrin expression within bulge cells suggests that K15 and β1 integrin expression may be coregulated. Future studies to address this issue may lead to a better understanding of the function of integrins in the regulation of hair follicle stem cells within the bulge as well as other epithelial tissues. In mice, both K14 and K15 can pair with K5 in the skin and other stratified epithelia (Lloyd et al., 1995). In neonatal mice, both K14 and K15 are limited to the basal layers of epidermis, outer root sheath of hair follicles and cornea as well as internal

Fig. 8. The hair follicle cycle. The shaded area represents slowlycycling, K15+ bulge cells. These keratinocytes function as hair follicle stem cells which proliferate at anagen onset to regenerate the new lower follicle. The bulge cells then become quiescent and remain so for the remainder of the hair cycle. In anagen, the matrix cells, which express an intermediate level of β1 integrin, proliferate to produce the hair shaft. Proliferating cells are denoted by black dots. APM, arrector pili muscle; DLF, degenerating lower follicle; DP, dermal papilla; epi, epidermis; M, matrix; ors, outer root sheath.

stratified epithelia such as tongue and forestomach. However, K15 appears to be a minor component in the skin of neonatal mice which shows a high K14/K15 ratio, while K15 demonstrates a higher expression than K14 in esophagus of neonatal mice. Importantly, the high ratio of K15/K14 in neonatal mouse esophagus dramatically reverses in older mice as K14 becomes the predominant keratin. Age related changes of K15 levels in the skin have not yet been reported, but similar types of alterations in K15 expression may account for changes in the physical properties of skin during aging. In addition to studies on aging, the use of the C8/144B antibody to detect K15 and bulge-related cells will include examining the role of the bulge in carcinogenesis, wound healing and alopecias. The possibility that K15 is a general epithelial stem cell marker will also be explored. We thank Dr Leonard Jarett, Simon Flexner Professor and Chair of the Department of Pathology and Laboratory Medicine at the University of Pennsylvania for his support and encouragement, Dr Peter Bannerman, Director of the Confocal Microscopy Core of the Children’s Hospital of Philadelphia and the Cancer Center of the University of Pennsylvania, for performing confocal microscopy, Motha Kalyani, Shelly Roberts, Dorothy Campbell and the histology staff of the Departments of Pathology and Dermatology at the University of Pennsylvania for technical assistance. This project was supported by NIH grant #1-R29-AR-44038-01 and grants from the National Alopecia

3188 S. Lyle and others Areata Foundation, as well as by the Cooperative Human Tissue Network, which is funded by the National Cancer Institute.

REFERENCES Adams, J. C. and Watt, F. M. (1991). Expression of β1, β3, β4 and β5 integrins by human epidermal keratinocytes and non-differentiating keratinocytes. J. Cell Biol. 115, 829-841. Al-Barwari, S. E. and Potten, C. S. (1976). Regeneration and dose-response characteristics of irradiated mouse dorsal epidermal cells. J. Radiat. Biol. 30, 201-216. Altschul, S., Gish, W., Miller, W., Myers, E. and Lipman, D. (1990). Basic local alignment search tool. J. Mol. Biol. 215, 403-410. Barrandon, Y. and Green, H. (1987). Three clonal types of keratinocyte with different capacities for multiplication. Proc. Nat. Acad. Sci. USA 84, 23022306. Bickenbach, J. and Mackenzie, I. (1984). Identification and localisation of label retaining cells in hamster epithelium. J. Invest. Dermatol. 82, 618-622. Clark, E. A. and Brugge, J. S. (1995). Integrins and signal transduction pathways: the road taken. Science 268, 233-239. Christofidou-Solomidou, M., Longley, B. J., Whitacker-Menezes, D., Albelda, S. M. and Murphy, G. F. (1997). Human skin/SCID mouse chimeras as an in vivo model for human cutaneous mast cell hyperplasia. J. Invest. Dermatol. 109, 102-107. Commo, S. and Bernard, B. A. (1997). Immunohistochemical analysis of tissue remodelling during the anagen-catagen transition of the human hair follicle. Br. J. Dermatol. 137, 31-38. Cotsarelis, G. (1997). The hair follicle: dying for attention. Am. J. Pathol. 151, 1505-1509. Cotsarelis, G., Cheng, S. Z., Dong, G., Sun, T. T. and Lavker, R. M. (1989). Existence of slow-cycling limbal epithelial basal cells that can be preferentially stimulated to proliferate: implications on epithelial stem cells. Cell 57, 201-209. Cotsarelis, G., Sun, T. T. and Lavker, R. M. (1990). Label-retaining cells reside in the bulge area of pilosebaceous unit: implications for follicular stem cells, hair cycle, and skin carcinogenesis. Cell 61, 1329-1337. Coulombe, P. A., Kopan, R. and Fuchs, E. (1989). Expression of keratin K14 in the epidermis and hair follicle: insights into complex programs of differentiation. J. Cell Biol. 109, 2295-2312. Gerdes, J., Schwab, U., Lemke, H. and Stein, H. (1983). Production of a mouse monoclonal antibody reactive with a human nuclear antigen associated with cell proliferation. Int. J. Cancer 31, 13-20. Gilhar, A., Pillar, T. and Etzioni, A. (1988). The effect of topical cyclosporin on the immediate shedding of human scalp hair grafted onto nude mice. Br. J. Dermatol. 119, 767-770. Headington, J. T. (1984). Transverse microscopic anatomy of the human scalp. Arch. Dermatol. 120, 449-456. Heid, H. W., Moll, I. and Franke, W. W. (1988). Patterns of expression of tricocytic and epithelial cytokeratins in mammalian tissues. I. Human and bovine hair follicles. Differentiation 37, 137-157. Holbrook, K. A. and Minami, S. I. (1991). Hair follicle embryogenesis in the human: characterization of events in vivo and in vitro. Ann. NY Acad. Sci. 642, 167-196. Jones, P. H. and Watt, F. M. (1993). Separation of human epidermal stem cells from transit amplifying cells on the basis of differences in integrin function and expression. Cell 73, 713-724. Jones, P. H., Harper, S. and Watt, F. M. (1995). Stem cell patterning and fate in human epidermis. Cell 80, 83-93. Kligman, A. M. (1959). The human hair cycle. J. Invest. Dermatol. 33, 307-316. Keitel, T., Kramer, A., Wessner, H., Scholz, C., Schneider-Mergener, J. and Hohne, W. (1997). Crystallographic analysis of anti-p24 (HIV-1) monoclonal antibody cross-reactivity and polyspecificity. Cell 91, 811-820. Kobayashi, K., Rochat, A. and Barrandon, Y. (1993). Segregation of keratinocyte colony forming cells in the bulge of the rat vibrissa. Proc. Nat. Acad. Sci. USA 90, 7391-7395. Kramer, A., Keitel, T., Winkler, K., Stocklein, W., Hohne, W. and Schneider-Mergener, J. (1997). Molecular basis for the binding promiscuity of an anti-p24 (HIV-1) monoclonal antibody. Cell 91, 799-809. Lane, E. B., Wilson, C. A., Hughes, B. R. and Leigh, I. M. (1991). Stem cells in hair follicles: cytoskeletal studies. Ann. NY Acad. Sci. 642, 197-213. Lavker, R. and Sun, T.-T. (1982). Heterogeneity in epidermal basal keratinocytes: morphological and functional correlations. Science 215, 1239-1241.

Leube, R. E., Bader, B. L., Bosch, F. X., Zimbelmann, R., Achtstaetter, T. and Franke, W. W. (1988). Molecular characterization and expression of the stratification-related cytokeratins 4 and 15. J. Cell Biol. 106, 1249-1261. Li, A., Simmons, P. J. and Kaur, P. (1998). Identification and isolation of candidate human keratinocyte stem cells based on cell surface phenotype. Proc. Nat. Acad. Sci. USA 95, 3902-3907. Lloyd, C., Yu, Q.-C., Cheng, J., Turksen, K., Degenstein, L., Hutton, E. and Fuchs, E. (1995). The basal keratin network of stratified squamous epithelia: defining K15 function in the absence of K14. J. Cell Biol. 129, 1329-1344. Mason, D. Y., Cordell, J. L., Gaulard, P., Tse, A. G. D. and Brown, M. H. (1992). Immunohistological detection of human cytotoxic/suppressor T cells using antibodies to a CD8 peptide sequence. J. Clin. Pathol. 45, 1084-1088. Michel, M., Torok, N., Godbout, M.-J., Lussier, M., Pierrete, G., Royal, A. and Germain, L. (1996). Keratin 19 as a biochemical marker of skin stem cells in vivo and in vitro: keratin 19 expressing cells are differentially localized in function of anatomic sites, and their number varies with donor age and culture stage. J. Cell Sci. 109, 1017-1028. Moll, I. (1995). Proliferative potential of different keratinocytes of plucked human hair follicles. J. Invest. Dermatol. 105, 14-21. Moll, I. (1996). Differential epithelial outgrowth of plucked and microdissected human hair follicles in explant culture. Arch. Dermatol. Res. 288, 604-610. Morris, R., Fischer, S. and Slaga, T. (1985). Evidence that the centrally and peripherally located cells in the murine epidermal proliferative unit are two distinct cell populations. J. Invest. Dermatol. 84, 277-281. Morris, R. J., Coulter, K., Tryson, K. and Steinberg, S. R. (1997). Evidence that cutaneous carcinogen-initiated epithelial cells from mice are quiescent rather than actively cycling. Cancer Res. 57, 3436-3443 Paus, R. (1996). Control of the hair cycle and hair diseases as cycling disorders. Curr. Opin. Dermatol. 3, 248-258. Potten, C. (1974). The epidermal proliferative unit: the possible role of the central basal cell. Cell Tissue Kinet. 7, 77-88. Potten, C., Schofield, R. and Lajtha, L. (1979). A comparison of cell replacement in bone marrow, testis and three regions of surface epithelium. Biochem Biophys Acta 560, 281-299. Potten, C. and Morris, R. (1988). Epithelial stem cells in vivo. J. Cell Sci. Suppl. 10, 45-62. Quesenberry, P. and Levitt, L. (1979). Hematopoietic stem cells. N. Eng. J. Med. 301, 819-823. Rochat, A., Kobayashi, K. and Barrandon, Y. (1994). Location of stem-cells of human hair-follicles by clonal analysis. Cell 0076, 1063-1073. Schermer, A., Galvin, S. and Sun, T.-T. (1986). Differentiation-related expression of a major 64K corneal keratin in vivo and in culture suggests limbal location of corneal epithelial stem cells. J. Cell Biol. 103, 49-62. Schirren, C. G., Burgdorf, W. H. C., Sander, C. A. and Plewig, G. (1997). Fetal and adult hair follicle: an immunohistochemical study fo anticytokeratin antibodies in formalin-fixed and paraffin-embedded tissue. Am. J. Dermatopathol. 19, 334-340. Silver, A., Chase, H. and Arsenault, C. (1969). Early anagen initiated by plucking compared with early spontaneous anagen. Germany: Pergamon Press. Silver, A. F. and Chase, H. B. (1970). DNA synthesis in the adult hair germ during dormancy (telogen) and activation (early anagen). Dev. Biol. 21, 440451. Van Neste, D., de Brouwer, B., Tetelin, C. and Bonfils, A. (1996). Testosterone conditioned nude mice: an improved model for experimental monitoring of human hair production by androgen dependent balding scalp grafts. In Hair Research for the Next Millenium (ed. D. Van Neste and V. A. Randall), pp. 319-326. New York: Elsevior Science B.V. Weinberg, W. and Yuspa, S. (1997). An antibody to p53 recognizes soluble keratins in epidermal keratinocyte cultures under differentiating, but not proliferating, conditions. J. Invest. Dermatol. 109, 611-612. Wilson, C., Cotsarelis, G., Wei, Z. G., Fryer, E., Margolis-Fryer, J., Ostead, M., Tokarek, R., Sun, T. T. and Lavker, R. M. (1994). Cells within the bulge region of mouse hair follicle transiently proliferate during early anagen: heterogeneity and functional differences of various hair cycles. Differentiation 55, 127-136. Wright, N. and Alison, M. (1984). The biology of epithelial cell populations. Oxford: Clarendon Press. Yang, J. S., Lavker, R. M. and Sun, T. T. (1993). Upper human hair follicle contains a subpopulation of keratinocytes with superior in vitro proliferative potential. J. Invest. Dermatol. 101, 652-659.