Structure and properties of the hair follicle

The mature hair follicle is a complex structure, composed of several concentric cylinders of epithelial cells, known as root sheaths, which surround the hair shaft (^{Sperling, 1991}) (Figure 1). Although it is largely epithelial in origin, the follicle contains at its base a ball of specialized dermal cells, the dermal papilla, which play a crucial part in the regulation of successive cycles of postnatal hair growth (Figure 2). At the onset of phases of hair growth, signals from the dermal papilla are thought to instruct epithelial stem cells residing in the bulge region of the follicle to divide transiently (^{Oliver and Jahoda, 1988;Cotsarelis et al, 1990;Wilson et al, 1994;Lyle et al, 1998}). Stem cell progeny migrate to the base of the follicle, where they surround the dermal papilla, forming the hair matrix (^{Oshima et al, 2001;Taylor et al, 2000}). In response to further signals from the dermal papilla, matrix cells proliferate and begin the process of terminal differentiation, moving upward in the follicle and forming the hair shaft and inner root sheath (^{Oliver and Jahoda, 1988;Taylor et al, 2000;Oshima et al, 2001}). Pigmentation of the hair results from the activity of melanocytes, which reside in the hair follicle bulb and deposit pigment granules into the hair shaft as it forms. Periods of hair growth are followed by a regression phase, when the lower part of the follicle undergoes programmed cell death (^{Cotsarelis, 1997}), and a resting phase, before onset of a new growth phase (Figure 2). Cyclical growth of hair continues throughout postnatal life, and allows the follicle to remodel itself. This is particularly evident in the response of hair follicles to androgens, which cause enlargement of beard hair follicles in adolescent boys, and miniaturization of scalp hair follicles in men with androgenetic alopecia (^{Hamilton, 1942}).

Figure 1.

Comparison of the structure of the hair bulb in human scalp and mouse pelage. Paraffin sections were stained with hematoxylin and eosin and photographed at magnifications of 10 (human) and 20 (mouse). IRS, inner root sheath; ORS, outer root sheath; CTS, connective tissue sheath; HS, hair shaft.

Full figure and legend (191K)

Figure 2.

Schematic depiction of the hair growth cycle. (i) During anagen, matrix cells proliferate and differentiate to form the inner root sheath and hair shaft. (ii) During catagen the lower two-thirds of the follicle undergo programmed cell death. (iii) The follicle enters a resting phase, telogen. (iv) At the onset of a new cycle of hair growth, follicular epithelial stem cells are stimulated to divide by signals from the dermal papilla (green arrow). (v) During anagen signals from the dermal papilla stimulate the division of matrix cells and the inductive properties of the dermal papilla are maintained by signals from the follicular epithelium (green arrows). (vi) Differentiation of matrix cells into hair shaft and inner root sheath may involve lateral signaling between epithelial cells (green arrows). IRS, inner root sheath; E, epidermis; HS, hair shaft; ORS, outer root sheath; DP, dermal papilla; M, matrix; CTS, connective tissue sheath; B, bulge; S, sebaceous gland.

Full figure and legend (35K)

Communication between cells of different types is required for hair follicle formation

The formation of hair follicles occurs during embryogenesis and relies on a series of signals sent between dermal cells and overlying surface epithelial cells that cause fate changes in both cell populations, ultimately resulting in differentiation of the hair shaft, root sheaths, and dermal papilla (^{Hardy, 1992}). The existence of these signals was revealed by experiments performed as long ago as the 1950s, in which dermis and epidermis of different origins were recombined at different embryologic stages. These experiments utilized mouse and chick skin, and took advantage of similarities in the mechanisms regulating the early steps of hair and feather development. It was found that an initial signal arising in the dermis (the "first dermal signal") causes the formation of regularly spaced thickenings in the epidermis, known as placodes (^{Hardy, 1992}) (Figure 3). An "epithelial signal" from the placode causes the clustering of a group of underlying cells in the mesenchyme, forming a "dermal condensate". In response to a "second dermal message" from the dermal condensate, the epithelial placode cells proliferate and invade the dermis, eventually surrounding the dermal condensate, which develops into the hair follicle dermal papilla (^{Hardy, 1992}). Further proliferation and differentiation of the epithelial cells results in the formation of the inner root sheath and hair shaft of the mature follicle, processes that are likely to require lateral communication between epithelial cells (^{Millar et al, 1999;Lin et al, 2000}).

Figure 3.

The events of hair follicle morphogenesis. The upper panels show paraffin sections of primary mouse hair follicles at embryonic days 14.5 (undifferentiated epithelium and placode), 15.5 (germ and peg), and 18.5 (bulbous peg), stained with hematoxylin and eosin. The middle panels show these same developmental stages represented schematically, with intercellular signals indicated by colored arrows, and candidate molecules important for each stage listed below (see text for references). (i) The initial signal directing hair follicle formation arises in the mesenchyme (gray dots) and instructs the overlying epithelium (pink stripe) to thicken, forming a placode. (ii) Placode formation is facilitated by promoting signals (+), shown as green arrows, and prevented in neighboring epithelial cells by inhibitory signals (–), shown as red arrows. (iii) Signals from the epithelium induce the clustering of mesenchymal cells to form a dermal condensate (purple dots). (iv) The dermal condensate signals to the follicular epithelium to proliferate and grow down into the dermis. (v) The dermal condensate becomes enveloped by follicular epithelial cells to form the dermal papilla. Differentiation of the inner root sheath (blue) may be regulated in part by lateral signaling between epithelial cells (green arrows). Photographs in each of the upper panels were taken at the same magnification. Scale bar: 20 m.

Full figure and legend (61K)

The finding that communication between different types of cells is critical for hair follicle development implies that intercellular signaling molecules play key roles in this process. A variety of approaches have been taken to identify these molecules, their effectors, and their downstream targets, including surveys of the expression patterns of families of candidate signaling factors, the analysis of transgenic and knockout mice carrying mutations in candidate genes, and analysis of the effects of introducing candidate molecules through bead implantation into the skin. Most of these studies have been carried out in animal systems, particularly mouse and chick. These are not ideal systems for studying all aspects of human hair follicle biology, for instance mouse and human follicles show differing responses to androgens. Descriptive analyses of gene expression patterns in human skin, where available, however, suggest that the molecules regulating the basic mechanisms of follicle development are similar in human, mouse, and chick (^{Holbrook et al, 1993;Kaplan and Holbrook, 1994}). Importantly, human families carrying mutations in genes controlling hair follicle development show similar phenotypes to the corresponding mouse mutants (^{Cachon-Gonzalez et al, 1994;Nehls et al, 1994;Segre et al, 1995;Brissette et al, 1996;Hahn et al, 1996;Johnson et al, 1996;Kere et al, 1996;Unden et al, 1996;Ferguson et al, 1997;Srivastava et al, 1997;Ahmad et al, 1998;Gat et al, 1998;Aszterbaum et al, 1999;Chan et al, 1999;Frank et al, 1999;Headon and Overbeek, 1999;Monreal et al, 1999}). Although several key signals regulating hair follicle formation remain unknown or partially characterized, recent studies have led to an explosion of information about signaling in developing hair follicles, providing exciting opportunities for therapeutic innovation.

The first dermal signal

It is well established from experiments in the mouse and chick, that dermis from body regions that will eventually develop hair or feathers when combined with epidermis from non-hair bearing regions, will direct the formation of appendages of the size and spacing characteristic of the region from which the dermis was derived (^{Hardy, 1992}). Thus, in humans, one would expect that scalp dermis is able to induce the formation of large hair follicles that produce thick, long hair, whereas dermis from the forearm directs the formation of smaller follicles. Consistent with this, dermal sheath dissected from a male human scalp was able to induce the formation of thick, darkly pigmented hair when implanted into the forearm of a female recipient (^{Reynolds et al, 1999}). In developing molecular strategies for creating new hair follicles, for instance in patients suffering from scarring alopecia or severe burns, determining the identity of the "first dermal signal" is likely to be of key importance. Currently, however, this signal has not been characterized, and its mode of action remains a subject for speculation.

In one model to explain how the dermal signal operates, the spacing and size of placodes are regulated by an intrinsically periodic dermal signal, which varies in character in different body regions. To test this hypothesis,^{Jiang et al (1999)} marked dermal condensates in chick skin by injection of the lipophilic dye, DiI, and then dissociated the dermis and epidermis into single cell suspensions. When these cells were recombined, a regular array of feather placodes and dermal condensates was established. Labeled cells were distributed randomly between the dermal condensates and interfollicular dermis, suggesting that the initial position of dermal cells in the skin does not determine their ability to induce placodes, and arguing against the idea of a periodic dermal signal. In an alternative model, the dermal signal occurs uniformly within each body region and triggers the activation of promoters and repressors of follicle fate that then compete with one another, resulting in the establishment of a regular array of follicles (^{Slack, 1991;Barsh, 1999}). Differences in the levels of promoter and repressor activation might then account for regional differences in the size and spacing of follicles. Consistent with this model, several positive and negative regulators of hair follicle fate are initially expressed uniformly in the epidermis and subsequently become localized to placodes (see below).

A first clue to the molecular nature of the dermal signal has come from studies of the subcellular localization of -catenin. This molecule is usually rapidly degraded in the cytoplasm. In response to the action of WNT intercellular signaling molecules, which act over short distances to control cell fate in many organ systems, degradation of cytoplasmic -catenin is inhibited. In these circumstances -catenin accumulates in the cytoplasm and translocates to the nucleus where it forms transcriptionally active complexes with members of the lymphoid enhancer-binding factor/T cell factor (LEF/TCF) family of DNA binding factors (Figure 4a) (reviewed in^{Wodarz and Nusse, 1998}). In the chick, nuclear -catenin is found transiently in the dense dermis underlying the feather tract 2 d before the appearance of molecular and morphologic signs of placode development (^{Noramly et al, 1999}). Consistent with this, Lef1 is expressed in the mesenchyme of the mouse vibrissa pad prior to vibrissa follicle development, and initiation of vibrissa follicle development is dependent on this expression (^{Kratochwil et al, 1996}). Interestingly, development of dorsal feather inducing dermis has recently been found to be dependent on a signal from the dorsal neural tube, which can be substituted by Wnt1 (^{Olivera-Martinez et al, 2001}). These findings suggest that activation of the WNT signaling pathway in the dermis may be involved in establishing the first dermal signal. The appearance of nuclear -catenin is uniform within the dense dermis (^{Noramly et al, 1999}), consistent with the hypothesis that the "first dermal signal" is uniform rather than periodically localized.

Figure 4.

Schematic depictions of the WNT and Sonic hedgehog (SHH) signaling pathways. (A) The WNT signaling pathway. In the absence of a WNT signal, cytoplasmic -catenin is phosphorylated and targeted for degradation by a complex of proteins, including axin, adenomatous polyposis coli tumor suppressor protein (APC) and glycogen synthase 3- (GSK3-). In the presence of a WNT signal the degradation machinery is inhibited and -catenin translocates to the nucleus where it forms a transcription complex with DNA binding factors of the lymphoid enhancer-binding factor/T cell factor (LEF/TCF) family, and activates transcription of target genes. (B) The SHH signaling pathway. In the absence of an SHH signal, activity of the smoothened (SMO) protein is inhibited by the actions of the SHH receptor Patched 1 (PTC1). In the presence of SHH, the repression of SMO is lifted, and target genes, including PTC1 and GLI1, are transcribed through the actions of the transcription factors GLI1 and GLI2.

Full figure and legend (71K)

Formation of placodes: a competition between placode promoters and placode repressors

Descriptive and functional studies in mouse and chick have revealed that in response to the first dermal signal, members of several classes of signaling molecules are expressed in the follicular epithelium. These include both promoters and repressors of placode fate. Several lines of evidence suggest that WNT paracrine signaling molecules act early in the process of follicle formation to promote placode fate. In the mouse embryo, Wnt10b is initially expressed uniformly in the epidermis and is markedly upregulated in placodes (^{St-Jacques et al, 1998;Reddy et al, 2001}) and Wnt7a shows a similar pattern of expression in embryonic chick skin (^{Widelitz et al, 1999}). In the chick, 1 d after the appearance of nuclear -catenin in the dense dermis, nuclear -catenin is detected in the overlying epithelium, indicating that the WNT signaling pathway is activated in the epithelium (^{Noramly et al, 1999}). Like expression of mouse Wnt10b and chick Wnt7a, nuclear -catenin becomes gradually elevated in regions destined to form placodes (^{Noramly et al, 1999}).^{DasGupta and Fuchs (1999)} have generated transgenic mice bearing a TOPGAL reporter gene that is responsive to LEF/-catenin signaling and provides a means of identifying cells in which the WNT signaling pathway is active. In TOPGAL transgenic skin, expression of the reporter is detected in placodes and dermal condensates, providing an indication that the WNT pathway is active in both epithelial and mesenchymal components of developing follicles (^{DasGupta and Fuchs, 1999}). Expression of a stabilized form of -catenin protein in chick or mouse epidermis causes the formation of ectopic follicles, indicating that activation of the WNT signaling pathway in the epithelium is sufficient to direct follicle development (^{Gat et al, 1998;Noramly et al, 1999;Widelitz et al, 2000}). Conversely, loss of function of the -catenin gene in mouse epidermis results in a failure of placode development (^{Huelsken et al, 2001}), and a null mutation of the mouse gene encoding the WNT effector protein LEF1 causes failure of formation of vibrissae and two-thirds of the body hair follicles (^{van Genderen et al, 1994;Kratochwil et al, 1996}).

Elegant tissue recombination experiments performed by^{Kratochwil et al (1996)} demonstrated that expression of Lef1 in the mesenchyme is necessary for the initiation of mouse vibrissa follicle development, and that transient expression of Lef1 in the epithelium is required for the completion of morphogenesis. Pelage hair follicles developed normally in the absence of mesenchymal Lef1, but formed in reduced numbers and with less pronounced keratinization in the absence of epithelial Lef1. These results suggest that WNT signaling is required at several different stages in the development of vibrissae and pelage hair follicles, and that there may be partial functional redundancy of LEF1 with other LEF/TCF family members. Three other members of this gene family have been identified: Tcf1, Tcf3, and Tcf4. Tcf1 is reportedly not expressed in keratinocytes (^{Zhou et al, 1995}); however, its expression in the dermis preceding hair follicle development has not been examined. Tcf3 is known to be expressed in the skin and hair follicles, although the TCF3 protein is generally associated with repression of the canonical WNT signaling pathway (^{Barker et al, 1999;Merrill et al, 2001}). TCF4 protein is apparently not expressed in human keratinocytes or hair follicles in the second trimester (^{Barker et al, 1999}); however, expression of TCF4 in the dermis and epithelium at earlier stages of human hair follicle development has not been carefully analyzed. Close inspection of published in situ hybridization data reveals expression of Tcf4 in the skin and vibrissa plate of the mouse embryo at embryonic day (E)13.5 (^{Korinek et al, 1998}) and in vibrissa follicles at E15.5 and E18 (^{Lee et al, 1999}). It is likely that further analysis of Tcf1 and Tcf4 expression patterns in skin, vibrissa, and hair follicles will reveal which of these genes might substitute for Lef1 at certain stages of follicle development.

In addition to ectopic follicles, expression of stabilized -catenin in mouse or chick causes follicle-based tumors (^{Gat et al, 1998;Noramly et al, 1999;Widelitz et al, 2000}), and stabilizing mutations in the -catenin gene are found in human pilomatricoma, a tumor of hair follicle matrix cells (^{Chan et al, 1999}), establishing a molecular link between the early events of follicle formation and tumorigenesis.

Fibroblast growth factors (Fs) and FGF receptor genes are expressed at early stages of follicle development and also appear to promote follicle formation, as exogenous FGF molecules induce ectopic follicles in wild-type chick embryos, and feather buds in scaleless (sc/sc) embryos that otherwise fail to develop feathers as a result of an ectodermal defect (^{Chuong et al, 1996;Song et al, 1996;Widelitz et al, 1996;Jung et al, 1998}). Conversely, a loss of function mutation in the mouse gene encoding fibroblast growth receptor 2-IIIB causes defective development of the skin and hair follicles (^{Revest et al, 2001}). Overexpression of FGF4 in embryonic chick skin alters the pattern of -catenin mRNA expression suggesting a possible role for FGF in elevating -catenin expression in placodes (^{Widelitz et al, 2000}).

The gene encoding transforming growth factor- (TGF-2) is expressed in both the placode and the dermal condensate (^{Ting-Berreth and Chuong, 1996a;Ting-Berreth and Chuong, 1996b}), and TGF-2-soaked beads can induce the formation of dermal papillae in chick embryo mesenchyme that has been stripped of epithelium, and hair follicles in mouse embryo skin explants (^{Ting-Berreth and Chuong, 1996b;Foitzik et al, 1999}), suggesting TGF-2 as another promoter of follicle fate. Consistent with this, mice lacking a functional Tgf2 gene have reduced numbers of hair follicles and a delay in follicular morphogenesis (^{Foitzik et al, 1999}). The homeobox-containing genes Msx1 and Msx2 are expressed in placodes (^{Noveen et al, 1995}), and mutant mice lacking both of these genes have reduced numbers of hair follicles (^{Satokata et al, 2000}), indicating that the MSX1 and MSX2 transcription factors also play important parts in promoting placode fate.

Recently, ectodysplasin (EDA), a molecule related to tumor necrosis factor, and its receptor ectodysplasia receptor (EDAR) have been identified as essential factors for initiating the development of hair follicles on the basis of their mutant phenotypes in humans and mice (reviewed in^{Barsh, 1999}). The EDA gene is mutated in human X-linked anhidrotic ectodermal dysplasia, and in the Tabby mouse, causing decreased numbers of hair follicles, and defects of the teeth and sweat glands (^{Kere et al, 1996;Ferguson et al, 1997;Srivastava et al, 1997}). The EDAR gene is mutated in human autosomal recessive and dominant hypohidrotic ectodermal dysplasias (^{Monreal et al, 1999}) and in the downless mouse (^{Headon and Overbeek, 1999}), causing identical phenotypes to those resulting from EDA mutations. The mouse Edar mRNA is expressed ubiquitously in the epithelium prior to placode formation, and then becomes restricted to placodes (^{Headon and Overbeek, 1999}), whereas the Eda mRNA is ubiquitously expressed even after placode formation (^{Srivastava et al, 1997;Mikkola et al, 1999}). Based on its similarity to the tumor necrosis factor family, EDA is likely to be cleaved from the membrane and to be capable of diffusing to sites of EDAR expression (^{Headon and Overbeek, 1999}). Loss of EDA/EDAR signaling in mutant mice affects the formation only of specific subtypes of hair follicles (^{Headon and Overbeek, 1999}). Two other EDAR family members, TROY and X-linked ectodysplasin-A2 receptor (XEDAR), are expressed in developing hair follicles (^{Kojima et al, 2000;Yan et al, 2000}) and may substitute for EDAR in unaffected hair types; alternatively, a different signaling pathway may be utilized.

In contrast to EDA and EDAR, members of the bone morphogenic protein (BMP) family of secreted signaling molecules appear to act as inhibitors of follicle formation. Bmp2 is expressed diffusely in the ectoderm, but then locates at an early stage to the preplacode epithelium and underlying mesenchyme (^{Noramly and Morgan, 1998}). Bmp7 is expressed in placodes, and Bmp4 is expressed in the prefollicle mesenchyme (^{Jung et al, 1998;Noramly and Morgan, 1998;Huelsken et al, 2001}). Ectopic expression of Bmp2 or Bmp4 suppresses the formation of feather buds in chick embryos (^{Jung et al, 1998;Noramly and Morgan, 1998}). In the mouse, the Edar gene is required for the expression of Bmp4, as well as Sonic hedgehog (Shh) (see below) and Edar itself, indicating that EDAR acts very early in follicular morphogenesis, and is required both for promoting the placode and for lateral inhibition of placode fate in surrounding cells (^{Barsh, 1999;Headon and Overbeek, 1999}). Expression of -catenin in the epithelium is also necessary for the expression of Bmp4 and Shh, as well as Bmp2 and Bmp7, indicating that WNT signaling in the epithelium is also required at a very early stage of morphogenesis (^{Huelsken et al, 2001}). Edar is expressed in the absence of epithelial -catenin (^{Huelsken et al, 2001}), suggesting that Edar expression is either regulated independently of epithelial -catenin, or lies upstream of activation of WNT signaling in the epithelium.

The genes encoding several secreted molecules capable of inhibiting BMP action, including Noggin, Follistatin (FS), and Gremlin, are expressed in developing follicles (^{Feijen et al, 1994;Roberts and Barth, 1994;Noramly and Morgan, 1998;Jiang et al, 1999;Merino et al, 1999;Patel et al, 1999;Ohyama et al, 2001}). These are thought to negate BMP action within the follicle, but, in contrast to BMP, may not diffuse into the interfollicular regions. Support for this hypothesis comes from the phenotype of mice lacking Noggin, which have fewer hair follicles than normal and retarded follicular development (^{Botchkarev et al, 1999}). Expression of Lef1 is reduced in Noggin-null mice, suggesting that Lef1 expression may be repressed by BMP (^{Botchkarev et al, 1999}). Conversely, ectopic expression of Noggin in chick or mouse embryonic skin causes enlarged and ectopic follicles (^{Noramly and Morgan, 1998;Botchkarev et al, 1999;Jiang et al, 1999}). The Notch pathway also appears to play a part in determining the follicular pattern. The Notch ligand Delta1 is normally expressed in the mesenchyme underlying the placode (^{Crowe et al, 1998;Crowe and Niswander, 1998;Powell et al, 1998;Viallet et al, 1998}), and when misexpressed in a small area of epithelium promotes expression of Notch1 and accelerates placode formation, while suppressing placode formation in surrounding cells (^{Crowe et al, 1998;Viallet et al, 1998}). Restriction of Delta-1 expression to the prefollicle may be controlled by FGF, as in scaleless chick mutants Delta1 is ubiquitously expressed in the mesenchyme, but becomes localized after addition of FGF (^{Viallet et al, 1998}).

In summary, the formation of placodes in response to the first dermal signal involves activation of EDA/EDAR signaling in the epithelium, followed by epithelial WNT signaling, and subsequent activation of BMP signaling (Figure 3). The actions of EDA/EDAR and WNT promote placode formation, whereas BMP signaling represses placode fate in adjacent skin. FGF signaling promotes placode fate, and regulates expression of Delta1. TGF-2 and MSX factors also promote piacorse fate, but their interactions with other factors regulating hair follicle development have not yet been established.

Induction of the dermal condensate: the first epithelial signal

WNT signaling is likely to be required for induction of the dermal condensate. Evidence for this is that the WNT-responsive TOPGAL reporter gene is expressed in the dermal condensate as well as in follicular epithelium (^{DasGupta and Fuchs, 1999}), and the dermal condensate fails to develop in the absence of epithelial -catenin (^{Huelsken et al, 2001}).

Platelet-derived growth factor-A is expressed in the placode, whereas its receptor is expressed in the dermal condensate (^{Karlsson et al, 1999}). Mice lacking platelet-derived growth factor-A have small dermal papillae, dermal sheath abnormalities, and thin hair, compared with wild-type siblings, suggesting that platelet-derived growth factor-A is also required for normal cross-talk between the follicle epithelium and its mesenchyme (^{Karlsson et al, 1999}).

Another secreted protein present in the follicular placode that plays a major part in epithelial–mesenchymal signaling is Sonic hedgehog (SHH) (Figure 4b) (^{Bitgood and McMahon, 1995;Iseki et al, 1996}). In mice lacking SHH, hair follicle formation is initiated and the dermal condensate is formed, but mature hair follicles fail to develop. Analysis of SHH-null skin indicates that SHH is not a component of the first epithelial signal, but is required for subsequent signaling from the epithelium to both epithelial and mesenchymal cells, regulating proliferation and further downgrowth of the follicular epithelium and development of the dermal papilla (^{St-Jacques et al, 1998;Chiang et al, 1999;Karlsson et al, 1999}). Shh expression is absent from the follicles of mice lacking epithelial -catenin, indicating that Shh lies downstream of WNT signaling in hair follicle development (^{Huelsken et al, 2001}). Hair keratin genes are expressed in Shh^–/– hair follicles, indicating that SHH is not required for the initiation of hair shaft differentiation (^{St-Jacques et al, 1998;Chiang et al, 1999}). The genes encoding Patched1 (PTC1), a receptor for SHH, and GLI1, a transcriptional effector of SHH signaling, are expressed in follicular epithelium and in the dermal condensate, consistent with the idea that SHH signals are required for the development of both components of the follicle (^{Dahmane et al, 1997;Platt et al, 1997;Ghali et al, 1999}).

Mutations in PTC1 play a causative role in the etiology of the majority of human basal cell carcinomas (BCC) (^{Hahn et al, 1996;Johnson et al, 1996;Unden et al, 1996}), tumors that have several characteristics in common with immature hair follicles, including similar histology, ultrastructure, and patterns of keratin gene expression (^{Kumakiri and Hashimoto, 1978;Markey et al, 1992;Jih et al, 1999}). These findings have led to intense interest in the regulation of the SHH signaling pathway in skin and hair, and recognition of the central part played by hair follicle signaling pathways in skin tumorigenesis (^{Gailani et al, 1996;Johnson et al, 1996}). Identification of the downstream genes regulated by SHH signaling in hair follicles and BCC is of particular interest, as these are potential targets for novel BCC therapies. A Wnt family member, Wnt5a, is expressed in the dermal condensate of developing hair follicles, and is absent from these cells in Shh^–/– embryos, suggesting it as a target of SHH signaling in hair follicles (^{Reddy et al, 2001}). Interestingly, Wnt5a message is upregulated in a Xenopus model for BCC, and in human BCC samples (^{Bonifas et al, 2001;Mullor et al, 2001}).

Tgf2 has been suggested to be a target of SHH signaling in follicles as its expression domain is widened in response to ectopic expression of SHH in embryonic chick skin (^{Ting-Berreth and Chuong, 1996a}); however, normal levels of Tgf2 message were detected in the hair follicles of mice lacking SHH (^{St-Jacques et al, 1998}). Loss of function mutations in the Tgf2 gene result in the development of reduced numbers of hair follicles (^{Foitzik et al, 1999}) (see above), indicating that, whatever its relationship to Shh, this gene plays critical roles in hair follicle morphogenesis. The neurotrophin receptors TrkC and p75 neurotrophin receptor (p75NTR) are expressed in developing hair follicles in the placode and dermal condensate, respectively, and loss of function mutations in the TrkC and p75NTR genes, as well as in the gene encoding neurotrophin 3 (NT-3), affect the rate of hair follicle morphogenesis, suggesting that neurotrophins may also be involved at early steps of this process (^{Botchkarev et al, 1998, 1999a, b}).

In summary, WNT and platelet-derived growth factor-A molecules are strong candidates as components of the first epithelial signal inducing formation of the dermal condensate. SHH acts later in follicular morphogenesis, is dependent on WNT signaling and is required for proliferation of follicular epithelium and development of the dermal condensate into a dermal papilla (Figure 3).

Epithelial proliferation and downgrowth: the second dermal signal

The "second dermal signal" regulating proliferation and downgrowth of the follicular epithelium is likely to be activated by SHH, as significant downgrowth fails to occur in SHH-null mice (^{St-Jacques et al, 1998;Chiang et al, 1999}); however the nature of this signal is not known. One possible candidate is activin A (ActA), a secreted signaling molecule expressed in the dermal condensate (^{Feijen et al, 1994}) (^{Roberts and Barth, 1994}). Mice lacking ActA display defective morphogenesis of vibrissa follicles (^{Matzuk et al, 1995a}). The secreted protein FS binds and inhibits activins as well as BMP (see above). FS is expressed in the follicular epithelium in a pattern complementary to that of ActA (^{Feijen et al, 1994;Roberts and Barth, 1994}), and mice lacking FS have thin, inappropriately oriented vibrissae due to defects in vibrissa follicle development (^{Matzuk et al, 1995b}). Hepatocyte growth factor/scatter factor is expressed in the dermal condensate and its receptor, Met, is expressed in the follicular epithelium (^{Lindner et al, 2000}). Mice overexpressing hepatocyte growth factor have increased numbers of hair follicles and accelerated hair follicle development, suggesting a possible role for this molecule in signaling between the follicular mesenchyme and epithelium (^{Lindner et al, 2000}).

A member of the Sry-type high mobility group box (SOX) transcription factor family, SOX18, is expressed in the dermal condensate, and mice homozygous for a semidominant mutation in this gene lack several subtypes of hair, suggesting a role for SOX family members in the development or function of the dermal condensate (^{Pennisi et al, 2000a, b}). Analyses of conditional mutant mice have revealed that expression of 1 integrin in the epithelium is required for remodeling the basement membrane and follicular downgrowth (^{Brakebusch et al, 2000;Raghavan et al, 2000}), and that -catenin is also required for normal follicular morphogenesis (^{Vasioukhin et al, 2001}), revealing essential roles for adhesion molecules in this process.

Differentiation of the inner root sheath and hair shaft

As the hair follicle bulb appears (bulbous peg stage; Figure 1, Figure 3), at least seven different epithelial cell layers constituting the components of the mature hair follicle are formed (^{Sperling, 1991}). Recent studies have begun to reveal some information about the genes responsible for determining and maintaining the phenotypes of particular layers within the highly organized hair follicle structure. The genes encoding Notch1, a membrane protein involved in determining cell fate through cell–cell interactions and intracellular signal transduction, and its ligands Serrate1 and Serrate2, are expressed in matrix cells destined to form the inner root sheath and hair shaft (^{Kopan and Weintraub, 1993;Powell et al, 1998;Favier et al, 2000}), and expression of activated Notch1 in hair shaft precursor cells under the control of a mouse hair keratin A1 promoter causes failure of differentiation of the hair shaft medulla (^{Lin et al, 2000}); this suggests that Notch1 is one of the factors responsible for controlling the phenotype of keratinocytes as they leave the bulb matrix and differentiate into specific cell types of the follicle.

In mature hair follicles, Bmp4 is expressed in the dermal papilla and both Bmp2 and Bmp4 are expressed in hair shaft precursor cells (^{Wilson et al, 1999;Kulessa et al, 2000}). Ectopic expression of Bmp4 in the hair follicle outer root sheath in transgenic mice inhibits the proliferation of matrix cells and activates hair keratin gene expression in the outer root sheath (^{Blessing et al, 1993}), suggesting that BMP signaling is important for hair shaft differentiation. This hypothesis is supported by the phenotype of transgenic mice expressing the BMP inhibitor Noggin in matrix cells of the hair bulb under the control of an Msx2 promoter. Expression of Noggin causes severe defects in differentiation of the hair shaft cortex and cuticle, loss of hair shaft differentiation markers, including trichohyalin and acidic hair keratins, and inhibition of the expression of several transcription factors normally expressed in mature follicles (see below) (^{Kulessa et al, 2000}).

Several lines of evidence suggest a role for WNT signaling in regulating differentiation of the hair shaft. First, expression of the WNT-responsive TOPGAL reporter gene appears in hair shaft precursor cells as they begin the process of terminal differentiation (^{DasGupta and Fuchs, 1999}); these cells also express LEF1 protein (^{DasGupta and Fuchs, 1999}), and Dishevelled 2 (^{Millar et al, 1999}), a component of the WNT signaling pathway, whereas adjacent cells express Wnt3 (^{Millar et al, 1999}). Ectopic expression of Wnt3 in the hair follicle outer root sheath causes hair shaft fragility and elevated expression of several nonintermediate filament proteins in the hair shaft (^{Millar et al, 1999}), and in mice lacking LEF1 the few hair shafts that are made appear poorly keratinized (^{Kratochwil et al, 1996}). The regulatory regions of hair shaft keratin genes contain binding sites for LEF1 (^{Zhou et al, 1995}). Mutation of the LEF1 site in the promoter of the wool keratin intermediate filament gene K2.10 decreased promoter activity in the hair follicles of transgenic mice, suggesting that WNT signaling might regulate expression of hair shaft intermediate filament genes (^{Dunn et al, 1998}); however, this has yet to be proven. Mice lacking the putative transcription factor MOVO1 also show hair shaft defects, including kinks and intercellular splits (^{Dai et al, 1998}). Interestingly, the Drosophila homolog of mOvo1 controls epidermal differentiation, and is transcriptionally repressed by a dTCF/armadillo complex (analogous to LEF/-catenin) in response to WNT signaling (^{Payre et al, 1999}).

Analyses of naturally occurring or induced mouse mutants have identified roles for several other transcription factors in the control of hair shaft differentiation, including a homeobox protein HOXC13 (^{Godwin and Capecchi, 1998}), which may regulate expression of certain keratin-associated protein genes (^{Tkatchenko et al, 2001}), and a winged-helix/forkhead transcription factor FOXN1 (formerly WHN) (^{Nehls et al, 1994;Segre et al, 1995;Brissette et al, 1996}), which is mutated in nude mice and regulates expression of an acidic hair keratin gene (^{Meier et al, 1999}). A case of two sisters carrying a mutated FOXN1 gene and displaying hair, nail, and immune defects has been reported, indicating conservation of the role of FOXN1 in mouse and humans (^{Frank et al, 1999}). The genes encoding the MSX1 and MSX2 transcription factors are coexpressed in hair follicle matrix cells during anagen, suggesting they may play overlapping roles (^{Satokata et al, 2000}). In mice ectopically expressing Noggin in the hair matrix, expression of the HoxC13, Foxn1, Msx1, and Msx2 genes is inhibited, indicating that activation of these genes requires BMP signaling (^{Kulessa et al, 2000}). A more detailed analysis of how these various factors interact with each other and with other growth and transcription factors expressed in follicles to produce the normal variety of hair types in humans and other mammals remains a fascinating area for future study.

Polarity of the hair follicle

Hair follicles grow at an angle to the skin, pointing from anterior to posterior, and regulation of this polarity may be controlled in part by Shh, which shows an asymmetric pattern of expression in hair and feather follicles (^{Millar, 1997;Gat et al, 1998;Morgan et al, 1998}). In support of this idea, overexpression of Shh in embryonic chick skin causes the formation of enlarged feather buds that have lost their normal orientation (^{Ting-Berreth and Chuong, 1996a}). Perturbations of the WNT signaling pathway, including overexpression of Wnt7a or stabilized -catenin in embryonic chick skin and overexpression of Lef1 or stabilized -catenin in transgenic mouse skin result in altered follicular polarity (^{Zhou et al, 1995;Gat et al, 1998;Noramly et al, 1999;Widelitz et al, 1999, 2000}), and can induce symmetrical expression of Shh in the follicle (^{Gat et al, 1998}), suggesting that WNT signals may lie upstream of Shh in controlling polarity.

Control of hair follicle shape

Mutations in the mouse genes encoding TGF-, the TGF- receptor (epidermal growth factor receptor), and the transcription factor ETS2, cause altered hair follicle architecture and wavy hair, indicating that these factors play critical roles in governing the shape of hair follicles (^{Luetteke et al, 1993;Mann et al, 1993;Luetteke et al, 1994;Yamamoto et al, 1998}). Several other mutations, e.g., waved coat (Wc) (^{Kojima et al, 2000}), cause wavy hair in mice. It is tempting to speculate that alterations in the functions or regulation of some of these genes might be responsible for the many variations in hair texture seen in human populations.

The hair growth cycle: similarities to morphogenesis

During postnatal life, hair follicles undergo successive cycles of growth (anagen), regression (catagen), and rest (telogen) (^{Dry, 1926}). Like hair follicle morphogenesis, the initiation of a new growth phase and the subsequent downgrowth, proliferation, and differentiation of the follicle require signaling between follicular dermal and epithelial cells (^{Oliver and Jahoda, 1988}), and recent molecular analyses suggest that signaling pathways active during hair follicle morphogenesis are reutilized in postnatal, cycling hair. At anagen onset Wnt10b mRNA localizes to epithelial cells adjacent to the dermal papilla (^{Reddy et al, 2001}) and the WNT-responsive TOPGAL reporter gene is activated in the hair follicle bulge, the location of follicle stem cells (^{DasGupta and Fuchs, 1999}). Onset of the first postnatal anagen fails to occur in mice that progressively lose -catenin from the epidermis and follicular epithelium, indicating that activation of WNT signaling in the epithelium is necessary for this process (^{Huelsken et al, 2001}). Isolated dermal papilla cells cultured in the presence of certain WNT proteins maintain their hair follicle-inducing properties, which otherwise are lost after several passages in culture (^{Kishimoto et al, 2000}), suggesting that, as in morphogenesis, WNTs play an important part in conveying inductive signals between the follicular epithelium and mesenchyme of postnatal follicles. SHH may also recapitulate some of its morphogenetic functions during anagen. While not required for anagen onset, SHH is necessary for subsequent events of anagen, including proliferation of epithelial cells and downgrowth of the follicle into the dermis (^{Wang et al, 2000}), and ectopic expression of Shh is capable of inducing resting follicles to enter a growth phase (^{Sato et al, 1999}). A plethora of other molecules have been either suggested or proven to play roles in controlling postnatal hair growth cycles, some of which, such as FGF5 and hairless (^{Cachon-Gonzalez et al, 1994;Hebert et al, 1994;Ahmad et al, 1998}), are not required for morphogenesis (reviewed in^{Cotsarelis and Millar, 2001;Stenn and Paus, 2001}).

Clinical implications: what does the future hold?

Knowledge of the molecules and pathways that regulate hair follicle formation and hair growth will be essential for achieving therapeutic goals for hair loss conditions, including the ability: (i) to create new hair follicles; (ii) to change the characteristics (such as size or shape) of existing follicles; and (iii) to alter hair growth in existing follicles, as well as aiding in the search for target antigens important in the etiology of alopecia areata and scarring alopecia (reviewed in^{Cotsarelis and Millar, 2001}). Inhibiting the activities of molecules important for hair follicle formation and cyclical growth may also ultimately provide us with means for treating hirsutism. While we are far from achieving these goals, the identification of molecules such as -catenin and SHH that are capable of inducing the formation of new hair follicles provides us with potential strategies for treating conditions in which follicles have been completely destroyed. SHH is also able to induce anagen (^{Sato et al, 1999}), a property that may be useful for treating a variety of conditions. In designing therapeutic approaches, however, it must be borne in mind that these molecules can also cause the formation of tumors such as pilomatricoma and BCC (^{Oro and Scott, 1998;Chan et al, 1999}). It will therefore be important to ensure, by controlling the dose, or modifying the properties of these molecules, that one can induce follicle formation without producing harmful side-effects. The development of methods for delivering genes to hair follicles is an area of active research that will clearly be critical for achieving therapeutic goals (^{Alexeev et al, 2000;Li and Hoffman, 1995;Fan et al, 1999;Domashenko et al, 2000;Cotsarelis and Millar, 2001}).

Although striking advances have been made recently in our understanding of hair follicle development, several areas remain mysterious. The nature of the first dermal signal regulating hair follicle development is not known, and we cannot yet explain why hair follicles in different regions of the body have different properties, including size, duration of the anagen growth phase, and sensitivity to androgens. The mechanisms by which testosterone impacts on signaling in the hair follicle to achieve its effects are also not understood, an area of importance for the development of novel therapies for androgenetic alopecia. The next few years are likely to bring us answers to some of these outstanding questions, and perhaps the beginnings of therapeutic applications for our recently acquired knowledge.

References

References

Acknowledgments

The author thanks Dr Andrzej Dhigosz for the SHH pathway diagram, Dr George Cotsarelis for comments on the manuscript, Drs Alla Domashenko and George Cotsarelis for samples of human scalp skin, and Dorothy Campbell for histologic preparation. The author is a recipient of a Dermatology Foundation/Dermik Laboratories Career Development Award, and this work was also supported by the National Alopecia Areata Foundation and the National Institute of Arthritis and Musculoskeletal and Skin Diseases.