Hair follicle predetermination

An essential feature of mammalian cells is the inability to self-renew. Mammalian embryonic stem cells appear to function only during the limited period of early ontogenesis (Pera et al., 2000), and thus most mammalian organs inevitably undergo a period of senescence and consequent death. Several systems, however, possess their own stem cell lineages, including male germ cells, the hematopoietic system, neural tissue and a number of epithelial structures. A few, such as the female endometrium, undergo cycles of development and regression in response to the action of specific regulatory factors. Perhaps uniquely, the hair follicle (HF) both can renew itself through an intrinsic stem cell population and has the ability to cycle. Its structural complexity, potential immortality, and rapid growth rate, together with its ability to return periodically to its incipient state, make the HF one of the most dynamic mammalian structures.

Recent studies have provided insight into molecular control of HF induction and early morphogenesis (Kratochwil et al., 1996; Gat et al., 1998; St-Jacques et al., 1998; Botchkarev et al., 1999); however, the mechanisms governing adult HF growth (anagen), regression (catagen) and quiescence (telogen) – that is, the hair cycle (Fig. 1) – remain a mystery (Chase, 1954; Hardy, 1992; Stenn and Paus, 2001).

In the past few years, gene expression profiles for many different cytokines and regulatory factors during defined stages of the HF cycle have been elucidated. More than one hundred regulatory and structural proteins have been localized to the epithelial and mesenchymal compartments of the HF (Stenn et al., 1994a; Stenn et al., 1996; Stenn and Paus, 2001), producing a complicated spatio-temporal expression pattern that exhibits dramatic cycle-associated changes. Until now, no single systematic attempt to unify these data has been undertaken. Consequently, a substantial body of gene expression and knockout and transgenic mouse data that could potentially delineate the basic regulatory pathways in the HF cycle is still largely unorganized and unexploitable.

This lack of progress could be due to the extreme complexity of the HF structure and our rudimentary understanding of the basic cellular rearrangements and interactions that maintain HF cyclic transformations. In contrast, our understanding of the molecular mechanisms of HF induction and early morphogenesis is much more advanced than our understanding of hair cycling (Philpott and Paus, 1998; McElwee and Hoffmann, 2000) – perhaps because the former have much simpler cellular kinetics (Chase, 1951; Holbrook and Minami, 1991), which facilitate interpretation of expression patterns and knockout/transgenic data.

Several pioneering studies have attempted to define the cellular dynamics of the HF cycle (Starile, 1965; Chapman, 1971; Ito, 1986; Tezuka et al., 1990; Reynolds and Jahoda, 1991; Jahoda et al., 1992; Reynolds and Jahoda, 1993; Kulessa et al., 2000; Oshima et al., 2001). However, most have focused on discrete HF structures or limited aspects of HF biology and do not provide a basis for a unified model of HF function. In 1991, Sun and co-authors published the bulge activation hypothesis (Sun et al., 1991) – the first and only model of the cellular kinetics in the HF^*, which raised many new and interesting questions.

Here, we present a further attempt to understand the interactions between different HF cell populations and the major cellular processes that take place during the HF cycle: the hypothesis of hair follicle predetermination. Our model is a refinement of preexisting hypotheses, with some critical modifications, which include the timing of stem cell recruitment and the dual origin of the cycling portion of the HF.

The hypothesis of hair follicle predetermination is based on conclusions made from an extensive analysis of the literature, as well as our own experimental data. Here, we present a compilation of evidence in support of these assertions.

Much recent work has focused on the idea that stem cells are localized in the bulge region of the HF and that this is the source of cells for HF renewal (Cotsarelis et al., 1989; Lavker et al., 1993; Lyle et al., 1999; Akiyama et al., 2000). Several reports have clearly indicated the high proliferative potential of the upper portion of the HF, which includes the bulge region. For example, slow cycling, [³H]thymidine-label-retaining cells (LRCs) have been observed in the middle third of the HF immediately beneath the sebaceous gland near the attachment of the arrector pili muscle 10 weeks after labeling (Cotsarelis et al., 1990; Morris and Potten, 1994). Furthermore, cytokeratins 15 and 19, putative markers for epithelial progenitor cells (Bartek et al., 1986; Lyle et al., 1998) localize to the HF bulge (Lane et al., 1991; Lyle et al., 1999), and cells in this region actively proliferate during the early stages of anagen (Silver et al., 1969; Wilson et al., 1994; Lyle et al., 1999).

At the same time, the telogen HF contains a developmentally distinct population of small, tightly compacted cells at its base, where it contacts the follicular papilla (FP; Fig. 2). In 1926, on the basis of a morphological study, Dry suggested that this cell population plays the key role in the production of the next generation of HFs and designated these cells the HF ‘germ’^‡ (Dry, 1926). Chase et al. (Chase et al., 1951), Montagna (Montagna, 1962) and Silver et al. (Silver et al., 1969) later reiterated this idea. Recent studies, however, failed to detect slow cycling LRCs in this region of the HF (Cotsarelis et al., 1990; Morris and Potten, 1999) and hence became a sort of death sentence for the role of these cells in HF renewal, although the term hair germ is still widely used. Currently, these cells are largely ignored or considered to be a “left-over of some ORS cells of the previous lower follicle” (Wilson et al., 1994). The direct involvement of HF germ cells in early anagen proliferation – obvious from morphology studies and mitotic-specific labeling (Chase et al., 1951; Silver et al., 1969) – if mentioned at all, is explained as a result of migration of transient amplifying cells from the bulge region to the hair germ during anagen induction (Cotsarelis et al., 1990) or simply by inclusion of the hair germ region into the operational definition of the bulge (Cotsarelis, 1997).

Despite these views, several lines of evidence support the suggestion that HF germ cells have an intrinsic (bulge-region-independent) proliferative potential and are directly involved in the earliest events of HF anagen induction.

First, calculation of mitotic figures (6 h after colchicine treatment) in HFs during a spontaneous telogen-anagen transition (Silver et al., 1969) revealed that in late-telogen follicles, some germ cells display characteristics of ‘activated’ cells about to divide, according to definitions used to describe regenerating tissue (Hay, 1966) or differentiating epithelial tissue (Stockdale et al., 1966). During the telogen-anagen transition, the cytoplasm of the germ cells becomes strongly basophilic, but no changes occur elsewhere in the follicle, including the bulge region. Furthermore, in anagen I, numerous mitotic figures were seen in the hair germ (Silver et al., 1969). Given that no cell migration to the hair germ from the bulge region has been reported (Morris and Potten, 1999), hair germ cells must therefore possess an intrinsic ability to proliferate.

Second, BrdU-incorporation studies revealed two positive zones in mouse HFs during the early anagen stages (Tezuka et al., 1991): (1) the bottom of the HF epithelium (the hair germ); and (2) the outermost layer of the outer root sheath (ORS) surrounding the club hair at the level of arrector pili attachment (the bulge region). As anagen progresses, these two zones retain their individual identities: numerous BrdU-positive cells were visible around the FP (hair matrix precursors) and along the entire length of the ORS (Tezuka et al., 1991). The authors have suggested that these two populations are defined very early in anagen (Tezuka et al., 1991). Finally, the same parceling of proliferating cell populations was recently shown in human HFs in a study of K19 expression patterns (Commo et al., 2000). We propose that these two cell populations are incipiently discrete and are present as separate structures in telogen HF before induction of anagen.

Collectively, these data indicate that the telogen HF contains two different cell populations that have independent proliferative potentials, reconciling the opposing views on the cellular source of HF regrowth (Dry, 1926; Montagna, 1962; Silver et al., 1969; Cotsarelis et al., 1990). These populations have distinct fates and functions, whose precise coordination in anagen dictates the program of HF cellular dynamics (see below).

According to the bulge activation hypothesis, upon induction of anagen in response to messages from the FP, LRCs in the bulge region divide and the daughter transient amplifying cells migrate to the hair germ to initiate a new cycle of hair growth (Cotsarelis et al., 1990; Sun et al., 1991) (Fig. 3A). However, several lines of evidence challenge this suggestion and point to the hair germ as the site of anagen induction.

Surprisingly, 3D reconstruction of distribution patterns of slow cycling LRCs in the HFs of BDF₁ mice and studies of proliferative activity of these cells (Morris and Potten, 1999) revealed that the first mitotic LRCs in the bulge region appear two days after plucking, when most HFs are already into the next follicle cycle. Thus, LRCs in the bulge region do not even begin to divide until many of the hair germ cells are already well into this process. Furthermore, no migration of bulge LRCs or their labeled daughters into the hair germ region during the initial anagen stages has been observed (Morris and Potten, 1999). Thus, the hair germ alone appears to be the source of early anagen transformations and is independent of bulge cell activity.

This idea is further supported by data from Silver and colleagues (Silver et al., 1969). They showed that, in anagen I, numerous mitotic figures are present in the hair germ, whereas mitotic activity in the bulge region appears no earlier than anagen II and reaches significant levels only in anagen III. In the hair germ, it achieved its maximal level during anagen I and anagen II. Studies of BrdU incorporation during the initial stage of anagen in mouse HFs produced similar results, upregulation of proliferation in the bulge region being observed later than that in the hair germ (Tezuka et al., 1991).

It is widely accepted that mesenchymal (FP)-epithelial intersignaling is essential in the induction of HF morphogenesis (Hardy, 1992; Botchkarev et al., 1999). Despite the lack of direct evidence (Stenn and Paus, 2001), it is assumed that a similar mechanism is recapitulated during the induction of anagen (Paus et al., 1999). An important requirement for all mesenchymal-epithelial signaling is that the interacting cells are in close proximity (Gilbert, 1997). In this context, the hair germ cell population is ideally situated (Fig. 2) to participate in the FP-epithelial crosstalk believed to trigger anagen induction (Hardy, 1992). In contrast, the bulge region is located at a significant distance from the FP (5-7 cells in mouse telogen HFs). The fact that both the hair germ and the FP upregulate BM-CSPG (basement-membrane-specific chondroitin sulfate proteoglycan or bamacan) in anagen I (Couchman et al., 1991; Couchman and du Cross, 1995) further indicates that hair germ cells are an immediate target of FP signaling. BM-CSPG is a component of the anagen FP cellular matrix and probably plays an essential role in organizing mesenchymal-epithelial crosstalk (Couchman and du Cros, 1995). Interestingly, in culture, FP cells gradually lose BM-CSPG expression along with the ability to induce HFs (Jahoda et al., 1984), which confirms the association of BM-CSPG with HF induction.

We therefore suggest that the hair germ, and not the bulge cell population, has the primary role in anagen induction^* (Fig. 3B).

The diversity of cell populations in the anagen HF represents a long-standing mystery. There is a strong belief that all layers of the HF, including the ORS, are products of the hair matrix germinative cells (Birbeck and Mercer, 1957; De Weert, 1989; Orwin, 1989; Moore, 1989). Nevertheless, this belief is not consistent with the enormous structural and functional complexity of the lower follicle. Such complexity might be more easily explained if the diverse anagen HF cell lineages originate from several sources rather than just one.

Using hair reconstitution assays (Weinberg et al., 1993), Kamimura and co-authors (Kamimura et al., 1997) demonstrated that HFs derive from at least two unrelated cells, in contrast to proliferating columnar units of interfollicular epidermis, which always derive from a single progenitor. Follicular epithelial cells transduced in culture with a retrovirus carrying the human alkaline phosphatase gene were recombined with FP cells in the nude mouse model (Lichti et al., 1993), such that only 10-20% of cells expressed the marker. Instead of finding labeled cells at random throughout the newly formed follicles, Kamimura et al. found that the majority of the reconstituted follicles were AP-positive in their external regions (the ORS) and AP-negative in their central portions (the hair shaft/IRS) or vice versa (Kamimura et al., 1997). These data indicated that all HF epithelial layers originate from at least two progenitors.

Ghazizadeh and Taichman recently provided compelling data in support of a dual origin of the HF following in vivo transduction of mouse skin with β-gal-encoding retroviral vectors (Ghazizadeh and Taichman, 2001). The authors analyzed the distribution pattern of transduced cells and their progeny thirty-seven weeks later, after five cycles of depilation-induced hair regrowth. In most labeled follicles (70%), labeled cells were confined to the hair shaft/IRS, the ORS, or the sebaceous gland, indicating that regeneration of the HF involves multiple progenitor cells, at least one for each compartment (Ghazizadeh and Taichman, 2001).

The above data, together with the studies showing that two proliferative units are present in the mature anagen HF (Tezuka et al., 1991; Commo et al., 2000), suggest that cells in the hair germ and in the bulge region contribute to the development of different compartments of the anagen HF. We propose that the hair germ cells give rise exclusively to the ascending part of the growing HF, including the hair shaft and IRS, whereas the descending ORS is a product of the bulge region progeny (Fig. 4). Studies of presumptive stem cell markers in human embryonic HFs by Akiyama et al. support such a model (Akiyama et al., 2000). Their results (summarized in Table 1) clearly indicate that the expression patterns of proteins in the bulge area and the outermost layer of the ORS (basal cells with high proliferative activity) are identical, whereas the hair matrix and all resultant layers have a different expression profile. Moreover, ultrastructural studies using transmission electron microscopy revealed that the cells of the bulge region and outermost layer of the ORS both possess strikingly similar morphological features (Akiyama et al., 2000).

A large body of evidence supports the notion that β1-integrin-rich, E-cadherin- and β/γ-catenin-poor keratinocytes represent a stem cell population (Moles and Watt, 1997, Watt, 1998). Cells of the bulge region and ORS exhibit such a phenotype (Akiyama et al., 1996; Akiyama et al., 2000), as well as the expression of putative stem cell markers K19 and EGF (Lane et al., 1991; Michel et al., 1996; Akiyama et al., 1996; Commo et al., 2000). These data support our idea that the cells of the outermost ORS layer, but not the matrix cells, represent the progeny of bulge-located stem cells.

Earlier pulse-labeling experiments (Epstein and Maibach, 1969; Chapman, 1971) failed to identify any upward movement of ORS cells but instead revealed movement of label from the outermost (basal) layer of the ORS horizontally inward. Therefore, the ORS is produced not by the hair bulb but instead arises through inward proliferation of its own outermost cell layer, which exhibits many specific features of the bulge cells (Akiyama et al., 2000).

Recent work has confirmed the existence of a follicular stem cell population that contributes to both the interfollicular epidermis (upward movement) and the HF (downward movement) (Taylor et al., 2000). In the context of the bulge activation hypothesis, this HF stem cell activity appears to be inexplicably asymmetrical, producing excessive diversity of daughter cell lineages in the downward direction (at least seven cell types) and just one in the upward direction. Our proposition that the bulge cells contribute only to the ORS, but not the ascending HF layers, resolves this discrepancy. We suggest that HF stem cells give rise to only one cell population in both the upward and downward direction: the epidermis and the ORS, respectively. The ORS and epidermis have similar structures (a basal lamina, a basal layer of proliferating cells and a differentiating multi-layer compartment), similar pathways of differentiation (Reynolds and Jahoda, 1991) and can reciprocally replace one another (Limat and Noser, 1986; Lenoir et al., 1988). Thus, the immediate progeny of HF stem cells are relatively uniform in both the upward and downward directions (Fig. 5). The formation of the ORS in early-mid anagen might therefore be analogous to the re-epithelialization of wounded epidermis that can be initiated from the bulge-containing region of the HF (Lenoir et al., 1988). Note that both processes can be induced by similar signaling pathways. The accelerated healing of porcine wounds during anagen (W. Eaglstein and R. Paus, personal communication) and the high incidence of irregular expansions of the bulge region after epidermal injury, which suggests locally increased proliferation of ORS cells (Lane et al., 1991), both support this notion.

Cell proliferation in the bulge region in pathological human skin conditions usually does not produce hair-shaft-like structures (Lever and Schaumburg-Lever, 1990), and even keratinizing bulge-derived tumors undergo trichilemmal keratinization specific for the ORS, but not for the hair shaft (Pinkus, 1969; Poiares Baptista et al, 1983; Reichel and Heilman, 1999). Again, this supports the idea that bulge-located stem cells form the ORS but do not contribute to the internal layers of the HF, which are instead produced by the hair germ.

Another argument in support of different cellular origins of the ORS and hair matrix derives from the unlikely probability that cells of the same progeny can undergo spatially directed proliferation in completely opposite directions. There is a belief that upward growth of internal HF layers (shaft and IRS) starts only after the completion of downward HF expansion by a prompt reversion in the direction of growth (Stenn and Paus, 2001). In fact, in anagen III (active formation of the hair shaft and IRS), the HF has not yet reached its final depth. At this time, a substantial part of the HF epithelium represented by the ORS is actively growing downward, thus driving the rapid penetration of the HF into the deep dermis. Simultaneously, the developing hair matrix cells undergo an enormous burst of upwardly directed proliferation (Fig. 4D). In addition, the downward expansion of the HF is due simply to multiplication of a homogenous population of keratinocytes, whereas the upward-directed growth is quite diverse and complex. Given a common origin of IRS and ORS precursors during early anagen, it is difficult to envision the instant reprogramming of some cells localized in a spatially restricted zone from the downward production of a uniform mass of epithelial cells to sudden, upward, complex and diverse differentiation. Our conclusion that the simultaneous downward and upward growth in the early anagen HF is provided by cells of different origins that have undergone different types of pre-anagen commitment provides a more plausible model of cellular kinetics in the early anagen HF. This model is further supported by the fact that the only significant movement of undifferentiated ORS cells is in a downward direction (Reynolds and Jahoda, 1991; Oshima et al., 2001). This finding indicates that the ORS does not originate from the hair matrix and supports our idea that the ORS pushes the bulb portion of the HF downward (Fig. 4C,D).

Recently, Taylor et al. proposed that presumptive stem cells located in the bulge region undergo two distinct pathways of migration/specialization: ‘the bulge-epidermal’ (upward) pathway and ‘the bulge-hair’ (downward) pathway (Taylor et al., 2000). We accept the idea of bi-directional bulge cell migration/specialization but prefer to designate these pathways as ‘bulge-epidermal’ and ‘bulge-ORS’, thus emphasizing their significant similarity and regulation by parallel, if not identical, molecular mechanisms (Fig. 5).

The dual origin of the cycling portion of the HF is also supported by our recent finding of downward-directed bulge proliferation in hairless mouse skin, in which the upper HF portion, including the presumptive bulge cells, is physically separated from the FP, which remains in the deep dermis (Fig. 6A). These bulge outgrowths do not form structures resembling ascending portions of the HF, such as an IRS or hair shaft (Fig. 6B,D), but instead form a long downward-growing string of undifferentiated cells similar to the developing ORS of the normal HF (Panteleyev et al., 1998). The FP-associated epithelial cells stranded in the deep dermis are also able to proliferate (Fig. 6C) and produce structures resembling anagen HFs that contain hair-shaft-, cuticle-, and IRS-like components. However, because these cells are physically separated from the bulge region, these ‘second wave’ HFs lack the ORS (Panteleyev et al., 1999b). Culturing of the lower portion of hair bulbs microdissected from rat vibrissa follicle (which contain the FP, matrix cells and the lowermost portion of the ORS) results in complete reconstitution of the ascending vibrissa follicle structures, including the shaft, cuticle and IRS, but the residual ORS does not normally grow beyond the level of amputation (C.A.B.J., unpublished). These facts further support the idea that the ORS originates from the bulge region and thus cannot be reconstituted solely by the lower bulb, which produces only the ascending layers of the HF.

During mature anagen (stage VI), when all HF structures, including the ORS, are fully developed, mitotic activity is evident in the bulge region (Chase, 1954; Lane et al., 1991). At that time, the ORS-driven downward growth of the HF is complete, and the structure of the ORS is supported by the inward growth of its basal cells (Tezuka et al., 1991). Moreover, despite the marked mitotic activity in the bulge region, no outgrowths or other changes in the morphology of this region during middle anagen have been reported. So, why do these cells proliferate and what is the fate of their progeny? We propose that they give rise to a specific cell population with high proliferative potential that migrates downward along the previously formed ORS (Fig. 7A).

Several elegant studies have localized cells that have high colony-forming potential in different portions of human scalp hair and rat vibrissae follicles (Yang et al., 1993; Kobayashi et al., 1993; Rochat et al., 1994). The results are quite consistent in that they found some colony-forming cells in the hair bulb (always a very low percentage). However, the data regarding the localization of colony-forming cells in the middle-upper portions of HF were quite different. Yang and co-authors placed whole isolated human scalp HFs in explant culture, and the bulge area yielded the most outgrowths (Yang et al., 1993). They then microdissected subpopulations of keratinocytes from different regions of the HF and tested their proliferative potential and life span in the presence of 3T3 feeder cells. The keratinocytes of the upper follicle, including the bulge area, had the longest life span in culture and the highest growth rates. Kobayashi et al., obtained similar results for rat vibrissa follicles, in which ∼95% of colony-forming cells were located in the bulge region (Kobayashi et al., 1993). In contrast, Rochat and co-authors found that most colony-forming keratinocytes are concentrated at a region further down the ORS, around and below the midpoint of the anagen human HF, much deeper than the bulge region (Rochat et al., 1994). Downward migration of bulge-derived cells that have high proliferative potential in middle-late anagen might explain this discrepancy, since such a model supposes the presence of migrating cells in different portions of the HF at different time points of anagen. Also, the migration may be not an intermittent event but instead a permanent downward flow of bulge-derived cells along the ORS. This possibility is supported by the observation that, in the human HF, cells that have clonogenic abilities have been found along the length of the ORS (Jahoda and Reynolds, 2000).

Barrandon and co-workers’ findings provide strong support for active downward migration of ORS cells during mature anagen (Oshima et al., 2001). They dissected the bulge regions from the vibrissa follicle of Rosa 26 mice expressing β-galactosidase, transplanted them into vibrissae follicles of wild-type mice and then implanted the chimeric follicles under the kidney capsule of athymic mice. Harvesting at regular intervals during the whisker regrowth allowed them to track the fate of the implanted bulge cells. The results clearly showed gradual movement of β-galactosidase-positive cells down the ORS and their eventual concentration in the hair matrix periphery (Oshima et al., 2001).

While the process of stem cell recruitment is clearly different in vibrissae and pelage follicles, the core aspects of vibrissae and pelage follicle function are obviously similar. Thus, the findings of Barrandon’s group provide unequivocal evidence that bulge-derived cells are able to migrate down the ORS to the lowermost follicle portion during active anagen not only in vibrissae but in pelage follicles as well (Oshima et al., 2001).

Kinetic studies of mouse skin either transduced with β-gal-encoding retroviral vectors or after BrdU/[³H]TdR injections provide clear evidence that bulge-derived keratinocytes are capable of active and prolonged migration through the upper HF ORS and interfollicular epidermis. These cells repopulate the interfollicular epidermis and become a source of epithelial proliferative units (Taylor at al., 2000; Ghazizadeh and Taichman, 2001). If upward and horizontal migration of bulge-derived cells in the pelage follicle is possible, then the downward migration of bulge-derived cells along the ORS during late anagen proposed earlier (Reynolds and Jahoda, 1991; Oshima et al., 2001) is also quite likely (Fig. 5; Fig. 7A). Furthermore, this downward migration of bulge-derived cells through ORS might be governed by mechanisms similar to those operating in the interfollicular epidermis, since these epithelial structures are very similar.

Recent studies have revealed a region of asymmetrical gene expression in the hair bulb of mouse follicles, demarcating the presence of a specific cell population that differs from the rest of the matrix cells (Fig. 7B). This structure consistently resides on one side of the FP, usually the side of the bulb closer to the epidermis (Panteleyev et al., 2000b; McGowan and Coulombe, 2000), although some authors report that the position of this specific zone in the hair bulb is random (Gambardella et al., 2000). The cells of this zone do not exhibit any structural differences when compared with the surrounding matrix cells and thus are not detectable by routine histological staining. Several proteins are specifically expressed in this ‘lateral disc’, including Shh (Gat et al., 1998), ICAM-1 (Muller-Rover et al., 2000), Ap-2 (Panteleyev et al., 2000b), K-17 (McGowan and Coulombe, 2000), Krox-20 (Gambardella et al., 2000), TGFβ RII (Paus et al., 1997), IL1-R1, α2 integrin and K14 (V. Botchkarev, S. Muller-Rover, E. Peters and R. Paus, personal communication). Only 5-19% of anagen mouse HFs display a lateral disc on 6-μm histology sections (Panteleyev et al., 2000b; Muller-Rover et al., 2000), which suggests that these cells occupy a very small part of the hair bulb volume and/or are characteristic of a short period of mature anagen, perhaps the very late anagen phase.

What is the origin and the function of lateral disc cells? We suggest that ORS envelops the hair bulb asymmetrically by forming a leading edge at the front of its downward growth (see above; Fig. 7A). This lowermost extension of the ORS may represent the final destination of bulge-derived cells migrating downward, where they gradually concentrate during mature anagen and form the lateral disc. The pressure of the perifollicular connective tissue sheath would cause the gradual submersion of the enlarging lateral disc into the hair matrix (Fig. 7B). Thus, we hypothesize that the lateral disc is formed by a concentration of bulge-derived cells at the lowermost tip of the ORS.

Currently, no direct information about the lateral disc in human HFs exists. The substantial thickness of the lower ORS portion in the human HF (up to 10 cell layers; Moll, 1995) might be sufficient to harbor a distinct cell population, and thus the formation of a separate lateral disc might be not necessary (Fig. 7C). This suggestion is supported by the recent finding of a K19-positive zone (a putative marker of clonogenic cells) in the lowermost portion of the ORS (embracing the upper bulb) in human anagen HFs (Commo et al., 2000). These K19-positive cells remain inactive in the lower ORS during anagen, and only a few undergo mitosis (Commo et al., 2000). They might therefore be quiescent – perhaps awaiting a new hair cycle.

The first change in HF morphology at the onset of catagen (along with termination of melanin formation) (Straile et al., 1961) is the reorganization of hair bulb architecture. The matrix cells cease proliferation and undergo extensive apoptosis (Lindner et al., 1997), which significantly reduces the hair bulb volume. Despite the widespread opinion that all cells of the lower hair bulb are completely eliminated during catagen progression (Stenn et al., 1994b; Seiberg et al., 1995), we favor the previously proposed possibility that some hair bulb cells survive apoptosis and contribute to the formation of the hair germ (Reynolds and Jahoda, 1993).

Ultrastructural studies of the lower portion of the HF during catagen using TEM revealed that the autolysis of peripheral bulb cells and the cells of the adjacent ORS is initiated only in the terminal stage of catagen, after the complete autolysis of the central cells of the hair matrix (De Weert et al., 1982). Thus, the loss of central bulb cells by apoptosis in early catagen would bring the lateral disc (the lower ORS in human HFs) into direct contact with the FP (Fig. 8B).

Our studies of the hairless phenotype in mice revealed that cells of the bulge region possess a certain level of resistance to apoptosis (Fig. 6B) and survive the total disintegration of the HF in hr/hr mutants (Panteleyev et al., 1998; Panteleyev et al., 1999a). Lateral disc cells might therefore also resist catagen-associated apoptosis, since they are presumably bulge region progeny (see above). Indeed, this suggestion is further supported by the fact that early catagen-associated detachment of the hair bulb in vibrissae is usually asymmetrical, perhaps sparing the lateral disc (Reynolds and Jahoda, 1993).

In middle catagen, the connection between the FP and neighboring apoptosis-resistant epithelial cells is significantly reinforced by the formation of hemidesmosomes along the keratinocytes apposed to the basal lamina at the FP-epithelium junction. In addition, a strong network of specific, compact fine fibrils is found throughout the dilated intercellular spaces in the FP. These fibrils are contiguous with the papilla cell membranes and penetrate the basal lamina of the adjacent HF epithelium (Sugiyama et al., 1976). This reinforcement of the papilla-epithelium junction and its consequent persistence suggests that these peripapillar keratinocytes play a key role in upward movement of the FP during catagen and provide essential contact between FP cells and the epithelial follicular compartment in telogen as well. De Weert et al. have suggested that the peripapillar keratinocytes originate from the ORS (De Weert et al., 1982). Although this may be true for the human HF, in the mouse HF these FP-associated keratinocytes most likely represent the apoptosis-resistant cells of the lateral disc. We propose that the lateral disc cells travel upward in close association with FP cells (Fig. 8C) and eventually adjoin the overlying epithelial compartment of the telogen HF (Fig. 8D). This suggestion is supported by our observation of continuous expression of hairless mRNA in a small population of FP-associated keratinocytes during catagen and the eventual positioning of these hr-positive cells at the bottom of the HF, just above FP cells during telogen. Note that the loss of hr gene activity in hr/hr mutants results in anchoring of the FP in the deep dermis (Panteleyev et al., 2000a).

Considering the extensive transformations the lateral disc cells must undergo, it is difficult to envision that they retain their undifferentiated features and high proliferative potential – assuming the conventional model of stem cell activity, in which there is a progressive and irreversible transition from stem cells to transient amplifying cells to differentiated cells (Lajtha, 1979). So, how do lateral disc cells reacquire (or retain) their ‘stem-cell-like’ features and transform into a hair germ that can form a new hair? We believe that the formation of intimate contacts between the lateral disc cells and the FP is the determining event in their final transformation into the hair germ, which forms as catagen progresses (Dry, 1926).

Recent work indicates that adult nerve stem cells can be completely reprogrammed and produce hematopoietic cells (Bjornson et al., 1999) and an unexpected plasticity between cell types in general (Morrison, 2001). Collectively, these findings suggest that progenitor cells exhibit extraordinary plasticity and that their differentiation program depends on the surrounding tissue, underscoring the importance of the concept of the epithelial stem cell niche (Watt and Hogan, 2000). According to this idea, the microenvironment, probably determined by the neighboring mesenchymal cells, supports the maintenance and self-renewal of stem cells and greatly affects their ability to select particular fates (Fuchs and Segre, 2000).

Recently, Ferraris et al. showed that the mesenchymal microenvironment can change the fate not only of epithelial stem cells but also of cells at advanced stages of differentiation (Ferraris et al., 2000). They created epithelial-mesenchymal recombinants from rabbit adult central corneal epithelium and embryonic dorsal and plantar mouse dermis. This experiment clearly showed that signals from embryonic mouse dermis can induce transdifferentiation of central corneal epithelium into normal mouse epidermis and pilosebaceous units. The dissection of corneal transplants from the central portion of the cornea excludes the possibility that limbal stem cells are present in the cornea-dermal recombinant (Ferraris et al., 2000). Thus, the study by Ferraris et al. clearly indicates that a distinct transient amplifying cell population can reacquire some stem cell characteristics under the influence of the underlying mesenchyme (Ferraris et al., 2000).

Taking into account the above findings, we suggest that the intimate contact established between lateral disc cells and the FP during catagen (Sugiyama et al., 1976) causes the lateral disc cells to revert from their transient amplifying state back to a stem-cell-like phenotype or at least significantly renew their proliferative potential (Fig. 8B-D). This process probably involves a certain degree of de-differentiation (Fuchs and Segre, 2000) of lateral disc cells, which is initiated and regulated by an FP-derived signal. We believe this process of specific ‘education’ or ‘maturation’ of lateral disc cells under the influence of the FP results in the formation of the functional hair germ (Fig. 8).

The presence of an apoptosis-resistant epithelial cell population with the potential to produce a new hair in the lower portion of the catagen follicle is confirmed by our studies of HRS/J hairless mice (Panteleyev et al., 1999a). In these mutants, after the first catagen-associated HF disintegration, the few epithelial cells that remain in contact with the detached FP remnants (presumably the lateral disc cells) (Fig. 6C). Around week 4 post partum (the time of anagen in wild-type mouse skin), the proliferation of these FP-associated keratinocytes results in the gradual formation of HF-like structures (the “second wave” of hair growth in hr skin) (Panteleyev et al., 1999b). These findings indicate that the lower portion of the catagen HF contains epithelial cells that are apoptosis resistant and can produce a new HF without any connection to the bulge region, independently of the stem cell population located in the upper follicle.

The proliferative activity of the cells in the bulge region of the ORS can be induced by many nonspecific internal and external factors, including wounding and TPA (Holecek and Ackerman, 1993), and in culture these cells have a prominent proliferative capacity (Yang et al., 1993; Rochat et al., 1994). The proliferative activity of the hair matrix cells, by contrast, is strictly dependent on the associated mesenchyme, and these cells cannot proliferate in culture (Jones et al., 1988; Limat et al., 1991). When lower matrix cells of the hair bulb were cultured in combination with FP cells, however, they actively proliferated (Reynolds and Jahoda, 1993). Thus, the bulge (ORS) and hair matrix cells (which we believe originate from hair germ cell precursors) have prominent differences in their responses: hair matrix cell activity is strictly dependent on the associated mesenchyme (the FP), whereas the inductive mechanisms of the bulge-ORS cell lineage are not specific. Our suggestion that the cells of the bulge region produce an essentially uniform cell lineage in both the upward (epidermis) and downward (ORS) directions, while the hair-germ-derived matrix cells produce a multitude of different layers, makes the demarcation between cells of the bulge region and hair matrix even more distinct.

The complexity of germ-derived anagen HF structures and the very short period of initial (predifferentiation) anagen means that some preparatory preprogramming of progenitor cells located in the hair germ must occur. By contrast, cells of the bulge region, producing relatively uniform and non-complex progeny, do not appear to require any such morphogenetic education. Given the key role of mesenchymal cells in determination of cell fate and the intrinsic abilities of transient amplifying epithelial cells (Ferraris et al., 2000), we further suggest that the intimate contact between the lateral disc or hair germ cells and FP fibroblasts during catagen-telogen phases specifically primes these cells with high receptivity to FP-derived morphogenetic signals (Fig. 8). Thus, we hypothesize that the specific receptivity of the hair germ cells to FP signaling and their commitment to produce ascending layers of HF are predetermined by the previous hair cycle during a priming phase of the hair germ cells derived originally from the bulge region of the HF.

The essential elements of the hypothesis of HF predetermination put forth here are the dual source of HF renewal (hair germ and cells of the bulge region) and the recruitment of HF stem cells capable of new HF production during the previous hair cycle. These two basic propositions provide a framework for understanding of some long-standing discrepancies in HF biology.

One of these discrepancies is a consequence of the odd combination of the essential autonomy of the hair cycle and its strong dependence on systemic and local factors (Stenn and Paus, 2001). When isolated or transplanted, HFs retain their intrinsic patterns of cycling and growth, which appear to be independent of the recipient tissue and other systemic influences (Durward and Rudall, 1949; Nagorcka and Mooney, 1985). The best proof of this characteristic was provided by Ebling and Johnson (Ebling and Johnson, 1959). They rotated a small portion of telogen rat flank skin up to 180° by plastic surgery. At the proper time, the HFs in the translocated portion of skin underwent wave-like initiation of anagen but in the opposite direction to the surrounding intact skin. Therefore, all hair follicles in the translocated skin portion retained their intrinsic (pre-surgical) cycling schedule during the first post-operational hair cycle.

Although the hair cycle is intrinsic and essentially autonomous, it cannot maintain this ‘donor pattern’ over the next cycle and subsequently acquires the features of the surrounding tissue. In parabiotic rats that have surgically joined vascular systems, the first hair cycle goes on independently, whereas subsequent waves of hair growth tend to become more and more synchronized (Ebling and Harvey, 1964). Thus, the inherent rhythm is strictly retained during only one hair cycle, and then cycling is influenced by shared systemic stimuli. If we assume predetermination of hair growth patterns from the previous hair cycle, this phenomenon of HF biology can finally be explained. Given such a one-cycle delay in HF reactivity, the results of experimental hair cycle modulation should be assessed not only during the cycle immediately following treatment but also during the second cycle. In routine experimental studies of HFs, this is usually not the case.

HF predetermination also explains the elevated sensitivity of the HF to hormone action in the catagen–early-telogen stages, after the termination of the active growth phase (Ebling and Johnson, 1964). According to our hypothesis, this is the time of hair germ formation that determines the characteristics of the next hair cycle and thus should represent the stage of the hair cycle most sensitive to modulation by systemic factors.

HF predetermination also explains the second wave of hair growth that occurs in hairless mouse skin despite the complete separation of FP cells from the bulge region. Interestingly, if abnormal and disoriented follicles of the second wave manage to establish contact with the persistent upper HF portion and the bulge cells, they acquire the ability to undergo a third cycle. Our hypothesis suggests that this contact with bulge-located (stem) cells provides the source of cells for the next cycle.

In the hypothesis of HF predetermination, we have sought to provide a unified view of the mechanisms of epithelial stem cell recruitment during HF cycling. The emerging revolutionary data about the exceptional plasticity of stem cells, their mobility and ability to be reprogrammed by surrounding tissues (Bjornson et al., 1999; Ferraris et al., 2000; Fuchs and Segre, 2000; Seale and Rudniki, 2000) prompt an urgent re-evaluation of our current thinking about the entire stem cell concept in the HF. The data and ideas presented here further suggest that stem cell recruitment is a complex, relatively protracted, multi-stage process that requires specific interactions with the microenvironment. Our hypothesis further supports the notion that the seminal question is not where stem cells are located or what their specific markers might be but, simply, what they are.

Despite the converging lines of evidence in favor of our hypothesis and the fact that it answers some critical questions in HF biology, it remains speculative and raises as many questions as it answers. Nevertheless, our intention was not to present a final and indisputable model of HF cycle progression but rather to take a fresh look at some of the most critical but poorly understood aspects of HF biology. We hope that our hypothesis inspires future studies of the cellular dynamics that underlie HF progression through the periods of growth, regression and quiescence.

The authors are grateful to Andrey Panteleyev Jr and Dmitry Panteleyev for their help with artwork. We also thank Vladimir Botchkarev, Boston University, and the members of the Skin Disease Research Center of the Department of Dermatology, Columbia University for stimulating discussion and for critical reading of the manuscript. This study was supported in part by grants from National Alopecia Areata Foundation and the NIH USPHS Skin Disease Research Center of the Department of Dermatology, Columbia University (P30-44534 and RO1-47338 to A.M.C.).