I. INTRODUCTION

Top Previous Next References

In these pages nearly a half-century ago Chase (63) published a review on the growth of hair that has become a classic in the field. In his discussion of the state of current hair biology, ". ... which only recently [has] come under serious investigation," Chase carefully defined the development, structure, and cycling of the follicle. His emphasis throughout was the central place and the dynamics of hair growth in relation to the biology of skin itself. Implicit in his discussion were the basic themes in the biology of his day (63, 64).

In this review we revisit the controls of mature hair follicle cycling. At the outset it is obvious to the reader that a lot has changed since Chase; above all, more scientists are considering the hair follicle as an attractive system for studying major biological phenomena (75, 167, 410, 534). Moreover, today we enjoy powerful analytical tools, which he did not have, and we study hair with the confidence that therapeutic intervention can indeed influence hair growth (e.g., Refs. 248, 616).

Our focus in this review is on the mature hair cycle and its controls. We attempt here to integrate the morphology, which was pretty well known by Chase, with the physiology, molecular biology, and genetics, about which we have learned much since then. Except for a broad outline, we do not cover hair morphogenesis (384, 422, 438) or pigmentation (522, 563) because these are major subjects deserving complete attention in and of themselves. Moreover, because of space restrictions, we do not cover the formation of the shaft (455), the hair follicle immune system (223, 406), and the fine details of hair follicle neurobiology (41, 155, 423). Our goal is to put into perspective our understanding of the cyclic growth of this deceptively simple structure, to signal hiatuses in our understanding, and to project where we think we need to go in the future. Although we try for a balanced presentation, our essay is neither meant to be encyclopedic nor a tabulation of relevant factors (see Refs. 92, 250, 300, 405, 410, 530). We have tried to refer to the literature liberally so that the interested student will have a confident starting point.

The most obvious function of the hair follicle is to produce a hair shaft, or fiber. The latter serves many more functions than are usually appreciated (Table 1). Although this is also true for other mammals, for humans the most important function of the shaft is as a physical medium of social communication; in fact, scalp, facial, and body hairs are essentially the only body parts an individual can shape to influence social intercourse. This point underscores the importance of hair and the psycho-social consequences of its pathology. To effect its function, follicles differ from site to site producing shafts differing in size, shape, curl, and color. Coarse body hair is referred to as terminal hair, and fine, short, nonpigmented hair as vellus hair (see sect. XI).

The hair follicle develops from the embryonic epidermis as an epithelial finger (Figs. 1 and 2). This peg differentiates into three enclosed epithelial cylinders. The central most cylinder forms the shaft (fiber). The outermost cylinder forms the outer root sheath (ORS) that separates the whole structure from the dermis. The middle cylinder, the inner root sheath (IRS), molds and guides the shaft in its passage outward. The shaft and the IRS move outward together. All mature follicles undergo a growth cycle consisting of phases of growth (anagen), regression (catagen), rest (telogen), and shedding (exogen). The anagen follicle consists of a so-called "permanent" portion above the muscle insertion and a cycling portion below. Because the hair shaft may not shed before the next anagen starts, it is important to appreciate that the hair cycle with respect to the shaft is different from the hair cycle with respect to the follicle; in other words, the fully formed telogen shaft ("club hair"), adherent to the pilary canal wall, may rest completely independently of the cycle expressed in the underlying follicle (we will return to this concept in sect. VII).

View larger version (29K): [in this window] [in a new window]   Fig. 1. Follicular morphogenesis and cycling. The hair cycle is envisioned here as a number 6, where the limb of the number represents the morphogenetic phases and the circular base represents the repeating cycle after the morphogenesis. Although the morphogenetic limb occurs just once in the lifetime of a follicle, the cycle repeats many times. Although morphogenesis and cycling are distinct phenomena, they share morphologic and molecular features (see text). DP, follicular papilla; SG, sebaceous gland; APM, arrector pili muscle; HS, hair shaft; mel, melanin; BM, basement membrane; POD, programmed organ deletion; HF, hair follicle; ORS, outer root sheath; IRS, internal root sheath. [Adapted from Paus et al. (414) and Slominski et al. (521).]

View larger version (91K): [in this window] [in a new window]   Fig. 2. Follicular histomorphology: human skin. The 8 cellular lineages of the mature follicle are illustrated by these histological preparations. The defined lineages are the outer root sheath (ORS), the companion layer (CL), the internal root sheath Henle's layer (He), internal root sheath Huxley's layer (Hu), the cuticle of the internal root sheath (Csth), the cuticle of the hair shaft (Csft), the cortex of the shaft (CTX), and the medulla of the shaft (Med). It is notable that the shaft and the IRS move up the follicle as a unit, with the IRS falling from the shaft just below the sebaceous gland duct. The slippage plane for the outward-moving shaft sheath is in the companion layer. A: vertical section of proximal anagen follicle, low power. B: vertical section of proximal anagen follicle, high power. C: cross section of proximal anagen follicle. D: telogen shaft with anagen follicle below. Note the secondary hair germ epithelium at the base of the telogen follicle (arrow). CTS, connective tissue sheath; SG, sebaceous gland; Club, telogen shaft base; Anag, proximal anagen follicle; FP, follicular papilla; GE, germinative epithelium. (Histology provided by Dr Leonard Sperling.)

As manifest by its cycle, the hair follicle is a regenerating system (512); the inferior hair follicle dramatically reforms itself over the cycle but the upper, permanent, follicle undergoes substantial remodeling (297). Hair follicle cycling is a developmental process, which occurs over the total lifetime of a mammal, well beyond the organogenesis of other systems and the cycling lifetime of the ovary or endometrium. This cyclic regeneration is thought to require many of the cellular signals integral to other morphogenetic (e.g., salivary glands, kidney, breast, and tooth) and regenerating systems (e.g., the amphibian limb). In all these systems, there is an intercommunicating epithelium and mesenchyme. For example, the regenerating amphibian limb needs a very special mesenchyme, the blastema, as well as an intact overlying epithelium to orchestrate the regenerative process (566). It is not surprising then to recognize that to grow and cycle, the normal follicle also requires intimate epithelial-mesenchymal interactions, receptive follicular epithelium, and follicular connective tissues. The regeneration theme is also upheld on a molecular level in that growth factor families found active in the regenerating amphibian limb are also expressed in the cycling hair follicle (384, 438).

As we will use the terms in this discussion, follicular regeneration is considered to be an integral part of the follicle growth cycle. On the other hand, follicular morphogenesis and the follicle cycle are, at least in part, considered to be different phenomena (Fig. 1). Although these two processes share many features, they differ from one another in at least one fundamental way: in the one a whole structure generates from a primitive epidermis (morphogenesis), and in the other, a partial structure generates from adult tissues (regeneration) (for recent reviews, see Refs. 82, 336, 384, 438).

For most mammals during the prenatal period, body follicles (pelage follicles) form in certain regions and then extend as a wave over the skin surface (63, 167). Follicles form from the primitive epidermis as a result of signals arising in both the primitive epithelium and the underlying mesoderm. The elements necessary for follicle formation are inherent to embryonic epithelium and mesenchyme (e.g., they do not need intact hormonal or neural circuits) since follicles can form de novo from organ-culture fragments of embryonic skin (164, 166). Early in development, specific foci of the primitive epidermis become competent to produce hair. Whether that signal is primary to epidermis or secondary to an inductive stimulus from the dermis is debated (38, 167, 266, 384, 438, 618). The epithelium grows down into the dermis as a plug that joins at its proximal end a mesenchymal condensation, referred to as the follicular, or dermal, papilla.¹ As the isolated papilla has inductive properties in the adult, it is thought to play a significant role in the morphogenesis of the follicle (214, 379, 380).

Follicle formation occurs but once in the lifetime of an individual, so a mammal is born with a fixed number of follicles, which does not normally increase thereafter. However, postnatal folliculoneogenesis is probably allowable in biology under certain unique circumstances. For example, new follicles are formed in the skin growing over newly formed (regenerating) antlers in mature deer (146, 534), and it has been proposed that follicle neogenesis can be associated with wound healing in rabbit skin (25). Relative to this phenomenon is that experimentally when -catenin is constitutively overexpressed in the skin of a transgenic mouse, new follicular structures form (K14 promoter, Ref. 132).

At the end of its morphogenetic phase, the follicle has a cycling inferior (proximal) region and a so-called permanent superficial (distal) region. Once fully formed, the follicle enters its first genuine cycle.²

	II. ASSAYS USED TO ASSESS HAIR GROWTH

Top Previous Next References

To measure hair growth, relevant, easy, and inexpensive experimental models are essential. To be effective, these models, in vivo or in vitro, have to reflect the major regulatory processes. The phenomena which the assays must describe include 1) hair follicle morphogenesis; 2) hair follicle cell differentiation leading to shaft and sheath formation; 3) hair follicle cycling including anagen, catagen, telogen, and exogen; 4) hair follicle heterogeneity; 5) hair follicle switch from the vellus to the terminal state; and 6) hair shaft pigmentation. In this section we discuss laboratory models currently used for assessing hair growth. It is important to recognize at the outset that the assays we describe are limited vis-à-vis the foregoing list.

In constructing assays for hair growth, the challenge has been to ensure that the system used actually measures hair growth and not a nonspecific or irrelevant biochemical/physiological pathway. Because we have not yet defined the cellular and molecular pathways, which uniquely control the cycle, the use of pure molecules or cells for measuring the cycle is limited. For all hair follicle assays, the parameters of donor animal type, donor animal age, and site of follicle origin are all very important to any experimental interpretation.

Although relevant as an assay, hair follicle organ cultures are limited by their difficulty of preparation, variability (the cycle phase during which a follicle is harvested for study may influence its experimental response in culture, Ref. 457), and viability (it has not yet been convincingly shown if dissected follicles can traverse the full cycle in vitro).

Whole animal systems are the most relevant but also the most difficult to control, quantify, and analyze. Animals commonly used include mice (63, 177), rats (231), sheep (203), and monkeys (569), but studies have been conducted on other mammals including the cat (181), horse (595), rabbit (553), opossum (305), guinea pig (55), prairie vole (527), and hamster (333). In this regard it is notable that there is no evidence that the most basic controls of hair follicle cycling among mammalian species are different. In assessing hair growth for human disorders, the most relevant model is still the human, but the second best is considered to be the macaque (569). Although most difficult to use routinely because of its rarity, housing, handling, expense, and ethical implications (of primate research), the macaque has, nonetheless, been effectively used as a quantifiable model (569). In this system both female and male adults show patterned scalp alopecia.

The laboratory mouse has been a favorite subject for hair studies, and the pigmented C57BL/6 (63, 426) and C3H (177) mice are the most commonly used strains. The rationale for choosing these mice is that their truncal pigmentation is entirely dependent on their follicular melanocytes; their truncal epidermis lacks melanin-producting melanocytes. Because pigment production is active only during the follicle growth (anagen) phase, the only time skin is dark is when the hair is growing. Therefore, by assessing the skin color one can also assess follicle growth phase. Another feature of the mouse system is that the growth phase of its follicles can be synchronized, allowing the investigator to isolate and analyze follicles of certain phases after hair growth induction by plucking (63, 513). Growth can also be stimulated by chemicals (464), including depilatory creams. Because active hair growth can be induced by minoxidil (54) or cyclosporin A (426) in these mice (as in humans), it is argued that the mouse model has some use in a drug discovery environment. Finally, the mouse system is particularly attractive because of the existing genetic databases, the availability of specific mutants (544), and the possibility of generating desired hair mutations by transgenic manipulation. It is also notable that gene delivery to mouse hair follicles has been successfully achieved (6, 95a, 121, 136, 188, 615). Because the C57BL/6 strain has been most extensively studied, we recommend to the researcher new to the field to start with this strain. We are indebted to Chase for this model.

Ex vivo systems combine in vitro and in vivo approaches. Workers have dissected out follicular structures, grew follicular cells and tissues in culture, and then transplanted these tissues back to the skin of immunoincompetent animals (292, 477, 585) or under the kidney capsule of a syngeneic living animal (e.g., Ref. 257).

In the very incisive Lichti system (292, 477), newborn follicular epithelium and mixed or cloned (457) follicular papilla cells are first grown in culture and then transplanted to immunodeficient mice. This system can be used to study not only folliculoneogenesis and skin organ regeneration but also follicular cell lineages (241) and the effect of specific cellular and genetic manipulations on follicle growth (253).

Reminiscent of the rotation-mediated method of Moscona (355), Takeda et al. (552) aggregated cells in a rotating in vitro culture system and then transplanted the aggregates to a receptive animal. Similar results have been found by growing follicular fragments beneath the kidney capsule of syngeneic animals (257). These cultures, like the Lichti system, also readily regenerate complete, cycling, mature hair follicles.

Whole skin explants have been used to study hair growth. Early attempts to grow embryonic skin in vitro demonstrated that such explants retain the ability to form hair (independent of vicinal influences such as intact blood vessels or nerves) (164, 194). More recently, mature whole skin from mice and humans has been used to study the hair cycle (286, 287, 416, 417). Whole skin on a gelatin sponge at the air/liquid interface can be sustained for up to 40 days in culture (286). Variations of this whole skin explant technique on collagen sponges ("histoculture") allows one to follow and pharmacologically manipulate hair follicle morphogenesis (36, 49, 124, 538a), anagen I-VI development, anagen-catagen transformation in vitro (36, 49, 124), and hair follicle pigmentation (287).

It has been demonstrated that dissected mature human anagen follicles can be successfully grown in culture (435). In this approach, human scalp/facial growing skin follicles are truncated below the dermis, dissected free of dermal/subcutis tissue, and placed free-floating in serum-free culture medium. With the use of this method, successful hair follicle organ cultures from multiple other species have been demonstrated (e.g., rat, Ref. 436; sheep, Refs. 34, 594; horse, Ref. 595). Although these follicles continue to grow for ~9 days in culture and are thus useful for studying the anagen phase and the onset of catagen, they do not traverse the full hair cycle and do not generate new follicles. It is important to note that this system, unlike a whole skin organ culture, lacks potentially important influences from the more distal pilosebaceous apparatus, namely, the epithelial stem cell region, the sebaceous gland, the perifollicular connective tissue sheath, and the epidermis (427, 595). A method for preserving whole follicles in a skin equivalent has recently been described (334).

Although growing dissociated cells, both mesenchymal (216, 217, 328, 578, 594) and epithelial (295), from mature hair follicles has been successfully achieved, generating follicles in vitro from these cells has been less successful. Jahoda and Reynolds (220) found that dissociated follicle matrix epithelial cells and follicular papilla fibroblasts placed within the connective tissue sheath of a vibrissae follicle would, in turn, generate a follicle. Limat et al. (294) found that growing ORS cells in an extracellular matrix in vitro supported the reorganization of hair follicle-like structures. Although in these studies aggregates suggestive of sebaceous gland and infundibulum were seen, clear-cut hair follicle formations were not.

A major current challenge for the laboratory assessment of hair follicle growth is to have inexpensive, biologically relevant systems for assessing hair follicle cycling. No such system is now available. In fact, we currently have no good in vitro system for assessing the vellus to terminal switch, for measuring anagen (growth phase) induction, or exogen (shedding) induction. For this reason, we are unfortunately dependent on the living, nonhuman animal assays for our initial assessment.

	III. NATURE AND THEORIES OF FOLLICLE CYCLING: A CYCLE ON CYCLES

Top Previous Next References

Whatever the evolutionary pressure, we recognize that the most unique feature of hair growth is its cycle. So, why does a hair follicle cycle? The answer is not obvious, but biologists have suggested several notions recognizing that skin molting (and the shedding of its appendageal products) is inherent to the integument of all organisms. It is suggested that hair follicle cycling may have arisen because it offers a mechanism whereby animals could 1) expand and grow (488); 2) control the length of body hair uniquely from site to site; 3) shed fur periodically to cleanse the body surface; 4) adapt and change its body cover in response to changing environmental (e.g., winter to summer), and perhaps social, conditions; 5) protect from the improper formation of the follicle; or 6) protect against malignant degeneration that might occur in this rapidly dividing tissue.

The hair growth cycle describes the changing morphology of the shaft, grossly, and the follicle, histologically, over time (96) (see Figs. 1 and 2). Starting with anagen, the follicle and its shaft pass through catagen, telogen, and finally exogen. All body hairs manifest this cycle, although the duration of that cycle, the duration of the individual phases, and the length of the individual shafts vary dramatically from site to site (490, 565). In the human and guinea pig, each follicle has its own inherent rhythm, and thus the cycles are asynchronous (63). In most rodents, however, large collections of follicles cycle together; in this situation, synchronous follicle growth occurs in waves that sweep posteriorly and dorsally.³ As the mouse ages, the waves become less frequent so that in the mature and senile mouse synchronous hair growth occurs only in relatively small patches. How the cycle spreads in waves is not clear, but experimental studies suggest that growth waves are controlled by factors intrinsic to the hair follicle groups in a manner that has been pictured as a "reaction-diffusion" system (103, 105, 365). This inherent rhythm, however, is influenced by neighboring follicles (229) and/or systemic (e.g., endocrine) stimuli, since in parabiotic rats the waves of hair growth tend to become synchronized with time (107). So, although the cycle is intrinsic and essentially autonomous, it is influenced by environmental systemic and local factors (entrained).

Although circannual rhythms are obvious in animal fur follicles, seen as shedding, and as a change of coat character (e.g., seasonal molting), their expression is subtler in humans (90, 382, 461). In the end, though, the long hair follicle cycles observed in human scalp hair, sheep wool, and horse mane represent very special cycles in biology: these follicle cycles are independent of the sun, moon, or constellations over a period of 2-6 years, a biological clock not yet characterized. The scalp follicle clock, then, being supra-circannual and independent of seasons (light) and temperature, demands its own paradigm in the field of chronobiology (102, 149).

Where the rhythm center rests in the follicle is unknown, and no experimental data indicate whether this site resides in the epithelium (e.g., bulb or bulge), mesenchyme (papilla), or a resonance between factors in the environment of the follicle (420, 535). Recent work suggests that many cell types have the ability to generate unique rhythms, independent of other tissue rhythms (18). It is important to recognize and distinguish the fact that the hair growth cycle is imposed on the cycle of growing cells that make up the follicle and its fiber; these cells undergo many cell divisions during the follicle growth cycle. The follicle cycle, then, is a cycle imposed on many cellular cycles.

From where is hair growth cycling initiated? What is the initiating stimulus? How does it get to its target? How does it affect hair growth? These are the most fundamental unanswered questions in contemporary hair biology (discussed in a recent forum, Ref. 535). The first signal for hair follicle formation (i.e., morphogenesis) is generally held to come from the mesenchyme (38, 167, 422). However, what initiates anagen from telogen in the mature follicle and from where that stimulus comes is debated. The signal could arise in the resting papilla, the resting epithelial germ, the adjacent epidermis, or, in theory, even the supportive vessels, nerves, lymphatics, and resident dermal hematopoietic cells of the region. Tissue culture and in vivo studies suggest that the signal arises independent of central organized neural elements and vascular or endocrine signals (166, 323).

The cells initiating anagen growth must have certain features: they must have stem cell-like characteristics (see below), they must have the machinery to maintain a unique rhythm (see below), and they must be able to send its signal to the surrounding epithelium and/or mesenchyme. That many soluble growth factors have been associated with the anagen follicle (530) suggests that a paracrine mechanism is plausible (365). How the signal spreads is unanswered. If the signal is spread from site to site by means of the epithelium, then it might, besides classical paracrine and juxtacrine signaling forms, also be transduced by means of an electrical pulse or ion flow through gap junctions. The epidermis and the hair follicle are richly endowed with such junctions (474). That these gap junctions might serve as the conduit for message transmission has been tested. Although communication by means of these junctions is seen in the developing follicle and in the germinative matrix, such interactions between the epithelial cells, between the mesenchymal and epithelial cells of the differentiated follicle (71, 240), or between epithelial cells of neighboring hair follicle via the interfollicular epidermis have not yet been found.

Theories proposing a mechanism for hair follicle cycling must include the characteristics of the cycle (its periodicity, persistence, and autonomy), epithelial-mesenchymal interactions, the variation from site to site, and the exquisite sensitivity to numerous extrafollicular growth-modulating signals, such as hormones, growth factors, and dietary changes. Six theories on the regulation of hair cycling have been recently presented and discussed (535) (see Table 2). In the past, arguments in favor of both a stimulatory and inhibitory control of spontaneous hair growth induction have been presented (231, 421). Chase (63) and others (12, 471) held that anagen initiation was due to the loss of an inhibitor, namely, an inhibitor release mechanism (the inhibition-disinhibition theory). Concrete evidence for the theory is scant, although telogen epidermis has been reported to contain an inhibitor to hair growth induction while anagen epidermis does not (427). If a parallel can be drawn, the recent demonstration that follicle formation is at least in part controlled by an inhibitor release mechanism (bone morphogenic protein-4/noggin complex, Ref. 38) encourages one to reassess Chase's inhibitor-release hypothesis for mature cycle initiation.

	IV. ANAGEN

Top Previous Next References

Anagen initiation is studied in the laboratory either as it arises, spontaneously, from telogen or, artificially, by experimental induction. As discussed above, we still do not know what spontaneous initiation means in terms of cells or molecules, except that its occurrence is, generally, quite predictable. In the laboratory, anagen can be induced under controlled conditions by taking advantage of the fact that hair growth is initiated by trauma/wounding (10, 11, 63, 135, 289). Trauma/wounding may mean hair plucking, vigorous shaving, or chemical exposure (e.g., caustic materials, depilatory agents). In this respect, we do not know what "trauma/wounding" actually means, although we do know that clipping the hair shaft without injury to the skin surface and follicle epithelium does not initiate growth (306). That trauma may have effects (cell necrosis, inflammation) besides the induction of growth has been documented (513). We assume trauma/wounding causes the release of proinflammatory cytokines that directly initiate anagen. Although we doubt that this form of initiation accurately reflects the spontaneous event, we assume that at least some aspects of the pathway(s) activated are common to the two situations.

It has been found that the traumatic stimulus has a certain threshold below which synchronized growth will not occur. Chase and Eaton (65) found that at least 1,000 hair shafts have to be plucked to initiate hair growth; plucking a single hair in the mouse does not initiate anagen activity in that follicle (65). This observation suggests that the pathway requires a quantifiably minimal threshold stimulus. It is also notable that the traumatic signal does not appear to spread beyond the area of wounding (63).

Although they are believed to have a specific mechanistic effect, selected pharmaceuticals, such as minoxidil (54), cyclosporin A (322, 420, 425), FK506 (213, 322, 425), norepinephrine-depleting agent (433), estrogen receptor antagonist (mouse, Ref. 376), tretinoin (596), tumor promoter agent (TPA) (375, 596), and various growth factors and neural mediators, such as keratinocyte growth factor (KGF) (92), hepatocyte growth factor (HGF) (228), sonic hedgehog (492), substance P (419), capsaisin (419), the antagonist parathyroid hormone (PTH)-(7---34) (195, 500), ACTH, and mast cell degranulation (409) can all induce anagen. The participation of any of these components, or the pathway they implicate, in the spontaneous signal for anagen induction is not yet clear.

It has long been known that anagen is suppressed by glucocorticoids (375). What was not appreciated is that the anagen stimulus, for spontaneously and traumatically induced signals, is actually only suspended, not abrogated, by glucocorticoids (534). The overriding aspect of the glucocorticoid block is that the trauma-induced signal is preserved as long as the steroid is applied. It is as if, paradoxically, the activated follicles were frozen in morphological telogen. Once the steroid applications cease, anagen begins. The implication is that the steroid block occurs downstream of a common pathway, but which mediator pathway it affects is unclear. The fact that the cyclosporin A signal for anagen induction is completely suppressed by betamethasone (Paus et al., unpublished data) while the morphogenetic signal cannot be blocked by corticosteroids (596) suggests that the latter two signal pathways differ somewhat from the spontaneous and trauma-induced signals. The implicated role of KGF and HGF in hair growth (see below) and the fact that the production of KGF and HGF are blocked by glucocorticoids may suggest a role for KGF and HGF in the glucocorticoid effect (68, 297). In any case, the steroid block would appear to be an ideal tool for dissecting the signals of anagen initiation before any morphological features of anagen are apparent. In the mouse, anagen induction is also blocked by 17-estradiol (207, 375, 376) and, reportedly, by androgens (375), but the relationship of the sex hormone steroid block to the glucocorticoid block has not been determined.

Anagen is that phase of hair follicle growth extending from the termination of the quiescent phase, telogen, to the beginning of the regressing phase, catagen. Although morphologically it is comparatively simple to define when anagen and catagen begin and end (64), the molecular sign posts are less clear. Anagen involves the complete regrowth or regeneration of the lower, cycling portion of the follicle, i.e., the hair shaft factory. To that end, the cells regenerating the lower follicle (i.e., the cells of the secondary hair germ; Fig. 2D) must receive a signal to proliferate, grow down into the dermis, form the epithelial lineages making up the essential cylindrical layers, and differentiate to produce that shaft which is characteristic of the skin region. Because there is a limit to the time a follicle stays in anagen, there is also a limit to the length of its product, the hair shaft. The anagen phase has been divided into six subphases (64, 359). Except for the last subphase, anagen VI (the duration of which dictates the shaft length), the length of the anagen subphases I-V does not differ substantially between follicles from different regions (490, 565).

Transcriptional activation, the earliest observed changes of anagen, occurs in the cells of the papilla and the secondary hair germ (a cluster of epithelial cells at the base of the telogen follicle) (511, 513). Before that time, the lower follicle is widely thought to be fundamentally "quiescent," mitotically and transcriptionally (however, see sect. VI).

In early anagen, the epithelial cells of the secondary hair germ grow down into the dermis as an epidermal finger. Once they reach their destined depth, the cells in the central cylinder reverse their growth direction and now progress distally (outward), forming the IRS and the hair shaft. Most of the cell divisions occurring in the bulb appear below a horizontal line drawn across the widest portion of the papilla (Auber's critical level, Ref. 6). This relative restriction of substantial cell division to the lowest region of the anagen follicle appears to reflect the profound structural changes the central portion of the follicle must undergo in its differentiation toward a highly rigid structure. The cell kinetics of the matrix during the cycle have been reviewed (313, 573). Such studies indicate that the matrix cell cycle time is 12-13 h in mice and 23 h in humans. In anagen, the cells of the bulb show a mitotic index of 2.4% in mice (4.3% in humans) with a labeling index of 32% and a growth fraction for the proliferative compartment of ~60%.

Although their function is not yet clear, a group of molecular markers of the anagen follicle bulb/matrix has been described (93, 250, 299, 300, 427, 531, 535). For example, actively dividing cells of the lowest portion of the bulb express telomerase, but as anagen approaches catagen, this expression ceases (374, 460). The same pattern is found for the zinc finger protein basonuclein, which is expressed in the basal layer of the ORS as far distal as the bulge and in the peripapilla cells of the bulb (586). Proliferating hair matrix cells are distinguished from other epithelial cells of the skin by their high level of LEF1 expression (60, 132, 618).

In contrast to the epithelial compartment of the follicle and the endothelial cells in the papilla (444), the fibroblastic cells of the papilla, in the normally cycling follicle, have been reported to show no division, thymidine uptake, or mitotic figures (589). However, in organ culture of human hair follicles, fibroblasts of the papilla incorporate thymidine apparently exclusively during anagen (20). In addition, it has been found that ferret follicles show a burst of proliferative activity in their papillary fibroblasts in very early anagen (535), as do murine hair papilla fibroblasts (M. Magerl, D. Tobin, and R. Paus, unpublished data). The significance of this proliferative activity to the hair cycle, as an effect of or a regulatory element of hair cycle control, remains to be established.

Any cellular structure that periodically renews itself depends on stem cells, cells that retain the ability to divide over the lifetime of the animal and regenerate that structure. Characteristics of stem cells include their paucity; their infrequency of division (slow cycling cells); their ability to generate transient amplifying (TA) cells in response to stimuli; their location in protected, well-vascularized, well-defined areas; their undifferentiated properties; and their colony-forming ability (82). However, what exactly defines a stem cell and what markers reliably delineate stem cell properties are debated (see Ref. 82); moreover, to make the concept even more complex, there may be some plasticity, or interconversion, between TA cells and stem cells (82).

Because the hair follicle is, above all, a regenerating system, workers have puzzled over the location and property of the cells that support this regenerative property. Because the bulb of the follicle shows significant cell division, it was implicit in the older literature that the site for important cell division and new anagen formation was in the bulb (e.g., Refs. 6, 573). Although that rationale is less obvious today, it appears stem cells were thought to reside among the cells that have the largest capacity to divide. Chase (63) was not so sure. In his review he states that "there is an equipotentiality of the epithelium ... of the upper, `permanent' external sheath" that can "refurnish lost epidermis in wound healing, ... give rise to a new sebaceous gland ... [and]... serve as the 'germ' in the presence of the dermal papilla."

In fact, when slow-cycling label-retaining cells were initially looked for, they were not found in the bulb but were found in the area of the follicle at the level of the muscle insertion site, the bulge/isthmus region (83, 353). These bulge cells are slow cycling (over 14 months in the mouse, Ref. 354), relatively undifferentiated, located in a well-protected well-nourished environment, and, finally, are indispensable to follicle cycling. The label-retaining cells have convoluted nuclei reflecting their proliferative inactivity (83). In addition, in the region of the bulge, and somewhat below it, there are colony-forming epithelial cells that have the greatest ability, compared with cells of other follicular levels, to form colonies in culture (holoclone or meroclone) (258, 476, 611). Although in young mice these special cells are found in the subsebaceous/muscle insertion site, the bulge area, in adult mice the bulge itself is not very apparent (83, 353). In humans, these cells are found deep to the muscle insertion site (258, 476).

There is reason to believe that the nature and location of stem cells may differ between follicle types. The vibrissal follicle in this respect differs in many ways from the pelage follicle. Specifically, "germinative" epithelial cells of vibrissae have been identified at the extreme base of the follicle bulb and display many of the characteristics of stem cells; they have unspecialized, primitive features and are located in a well-vascularized and sheltered region. This distinct group of epithelial cells remains behind when the rest of the matrix regresses; they appear to remain attached to the telogen papilla (468, 470) (Fig. 2D).

Biochemical studies also support the notion that cells in the bulge segment and its surrounding mesenchyme are special. For example, 1) the bulge cells are rich in keratin 15 (304), keratin 19 (K19) (273), epidermal growth factor (EGF) receptor, ₂₁- and ₃₁-integrins (236), high levels of ₆-integrin and low levels of the proliferation marker 10G7 (283), and platelet-derived growth factor (PDGF)-A/PDGF-B ligand chains. 2) In contrast to the surrounding bulge cells, the label-retaining cells are CD24 negative (310). 3) Recently, it has been found that among the K19-positive cells in the bulge is a subset that lacks a specific differentiation marker, a gap junction protein, connexin (Cx) 43; it was proposed that these cells may represent the actual stem cells (317). 4) That certain papillomavirus expression is limited to the epithelial cells of the hair bulge region suggests a unique follicular stem cell surface receptor that the papillomavirus exploits (499, 503). Potentially, this marker might have use in follicle stem cell studies. 5) Finally, the surrounding connective tissue of this region is also unique. The follicular connective tissue sheath cells in this area stain for both PDGF- and PDGF- receptors and versican (5, 101). Despite considerable effort, no definitive follicular stem cell marker has yet been found, although a combination of the above parameters provides a reasonable cellular signature for the stem cell region of the hair follicle.

It has been observed that as cells age the telomeric ends of their chromosomes shorten. This phenomenon is inversely related to the activity of telomerase, a ribonucleoprotein complex, which is able to reconstitute chromosome ends. Telomerase activity is high in proliferating cells and low in differentiating cells (169). Recent work suggests that telomerase is an important component of cells undergoing growth. Although quiescent stem cells normally would not be expected to express this enzyme (23), the transient amplifying cells, which they generate, would. A working definition of a stem cell, in this regard, would be a cell in which telomerase is inducible (while in differentiated cells telomerase would not be expressed, in transient amplifying and cancer cells it would be constitutively expressed) (C. Harley, personal communication). In the hair follicle, telomerase is only weakly expressed in the area of the bulge and strongly expressed in the lower bulb cells (460). It is found that telomerase expression ceases with the onset of catagen, a change that may reflect no more than the dramatic decline in cell proliferation associated with this phase of the cycle.

Although most workers in cutaneous biology have focused on epithelial stem cells, in other fields there is a concept of mesenchymal stem cells as well (58). It is possible that cells with stem cell-like properties also reside in the dermis. In fact, slow-cycling cells are found in the dermis (354). That the dermal sheath cells may regenerate a papilla suggests that it houses cells with stem cell-like properties (216, 377-379, 472). Papilla cells have the property of orchestrating the regeneration of the whole skin organ from dissociated epithelial-mesenchymal cells (292, 457, 473). In one set of experiments, complete hair follicle and dermal regeneration was demonstrated starting with newborn mouse epidermal cells and cloned papilla cells (457, 458). These experiments show that hair follicles house mesenchymal (papilla) cells with stem cell-like properties.

Evidence suggesting that there are precursor cells for other resident cells in the follicle, such as melanocytes (563) and Langerhans cells (140), has been presented.

Central to hair follicle growth are powerful epithelial-mesenchymal (E-M) interactions. Because we are focusing in this review on the mature cycling follicle, we will not discuss in detail those E-M interactions important to folliculomorphogenesis, which has been reviewed elsewhere (84, 85-87, 217, 330, 337, 384, 442).

Mature hair follicle mesenchyme is placed in two communicating compartments: the surrounding connective tissue sheath (CTS) and the follicular papilla (FP). The CTS embeds the follicle in the dermis and subcutis. The FP, a nubbin of connective tissue and cells, is separated from the proximal follicle in telogen but is embraced by the lower follicle matrix or bulb portion of the follicle during anagen.

The character of both mesenchymal regions changes dramatically over the growth cycle (212). In early anagen, the CTS consists of a thin basal lamina surrounded by collagen and stromal cells. With the development of anagen, the connective tissue (outside of the basal lamina) thickens into three separate layers. The innermost layer lacks cells and consists of collagen fibers running parallel to the long axis of the follicle. The middle layer consists of spindle-shaped fibroblasts and collagen fibers running transversely to the long axis of the follicle. The outermost layer contains cells and collagen fibers that run in various directions parallel to the outer surface of the follicle. In late anagen/catagen, the basal lamina of the connective tissue sheath thickens. Later, it and the other connective tissue layers become, by light microscopy, hyalinized and corrugated. Fibroblasts within and surrounding the CTS actively produce the collagen fibers of the sheath, which appear to fill the spaces left by the retracting catagen follicle (212). In telogen, the CTS at the level of the bulge stains intensely for versican, but as the follicle enters anagen, versican expression in this region disappears and reappears within the papilla (101, 253).

In anagen, the papilla is composed of a group of fibroblasts, embedded in a loose connective tissue stroma. It is encapped by the epithelial cells of the bulb but separated from the epithelium by a well-defined trilaminar basement membrane (167, 372, 398). During morphogenesis, in culture and after retinoid treatment, the papilla-matrix BMZ becomes fenestrated, allowing processes from papilla cells to contact epithelial hair matrix cells, an important relationship that appears to play a role in E-M signal transmission (143, 167). In the earliest phases of catagen, the papillary stroma decreases, and the bulb epithelium withdraws to release the papilla. In telogen, the papilla rests at the proximal base of the follicle as a tight cluster of cells containing virtually no ultrastructurally recognizable extracellular matrix (ECM) (398).

The morphological changes of the papilla over the cycle reflect primarily changes in its ECM. As anagen progresses, the extracellular matrix of the papilla enriches in mucins (347). In catagen, total glycosaminoglycan content of the papilla decreases; in telogen, it is scant (85, 220).

This unique ECM is characterized by its similarity to the basement membrane zone ECM (86). It contains typical basement membrane components such as type IV collagen, laminin, fibronectin, chondroitin sulfate, heparan sulfate, and versican (85-87, 101, 591).

Embedded in the papilla are very unique fibroblastic cells. These cells not only produce but also require their unique stoma to facilitate E-M communication (298, 404, 413). An excellent marker of mouse papilla cells is alkaline phosphatase, which, in contrast to earlier studies (67, 165), has been found to be expressed in the FP throughout the cycle (161). With respect to the regeneration concept and the cycling hair follicle mentioned above, it is of interest that the blastema of regenerating newt limb is also rich in alkaline phosphatase (566), a property thus shared by these two inductive mesenchymal tissues. Typical of papilla cells in vivo and in vitro (214, 329) is that they have a tendency to aggregate. In that regard, it is of interest that throughout the cycle papilla cells express neural cell adhesion molecule (NCAM) (80, 360), a molecule mediating cell-to-cell and cell-to-matrix adhesion. The presence of this molecule suggests that it plays a role in the integrity of the papilla over the cycle. The role of NCAM in follicle cycling is also demonstrated in the hairless (hr/hr) mouse where the papilla disintegrates shortly after hair follicle morphogenesis is complete (393-395). It is of interest that the papilla of hairless mice (hr/hr) express an abnormally low amount of NCAM. So, the notion is that the hr/hr gene plays a role in the control of NCAM expression, and NCAM plays a role in holding the papilla cells together.

During anagen, and not telogen, papilla cells express a potent protease inhibitor of the serpin family, nexin-1 (614), which has been implicated in normal organogenesis (315). Because the expression of nexin-1 correlates directly with the inductive ability of papilla cell lines (like versican, Ref. 253), it is believed this molecule is linked to a very important pathway in papilla-mediated follicle growth induction (614). In various cell strains, the expression of nexin-1 was shown to be influenced by molecules to which hair follicles are exposed; for example, it is upregulated by interleukin (IL)-1 (fibroblasts, Ref. 153), up- or downregulated by androgens (seminal vesicle, Ref. 574; papilla cells, Ref. 528), and downregulated by dexamethasone (fibroblasts, Ref. 153). The mechanism by which this protease inhibitor influences the cycle is yet to be shown, but it could be related to follicle extension into the dermis or to the activation of hair growth modulating factors, such as HGF/scatter factor (SF) (297).

Except for the substantial changes in its extracellular matrix, the FP stays remarkably constant over the hair cycle as a cluster of cells. Although the cells of the papilla continue to show RNA synthesis over the cycle, it is generally held that papilla cells themselves do not undergo cell division or take up tritiated thymidine (342, 589). The notion is that the number of cells for any specific papilla remains constant over the cycle, and thus over the lifetime of the mammal. As mentioned in section IVB, this dogma may be shaken as we analyze the papilla more carefully. In the ferret (496), sheep (537), and mouse (308), for example, papilla cells show new DNA synthesis in the early stages of hair growth initiation. Preliminary evidence from mouse pelage hair follicles suggests that the number of papilla fibroblasts changes significantly throughout the hair cycle (308). In the vellus-to-terminal follicle switch during adolescence (where the papilla increases in size) and, conversely, in the terminal-to-vellus switch in the pathogenesis of male pattern balding (where the papilla decreases in size), there is a change in the number of cells making up the papilla (115).

The papilla is an inductive structure that sends and receives signals. Its effect depends on continuous and intimate interaction with the hair matrix epithelium via native extracellular matrix (298). Epithelial stimulatory signals produced by human scalp papilla cells in culture have been demonstrated in tissue culture where ORS cells grow in the direction of dissected, contiguously placed papillae (8). Early workers demonstrated the hair-inductive properties of the papilla and its passaged cells (77, 199, 216, 379, 380, 451). Anagen papillae dissected free of the epithelial follicle and inserted into nonhair bearing skin can be shown to induce hair follicle formation from the resident epithelium (470). The inductive properties can also be demonstrated in skin equivalents (580). For effective follicular induction, continuous and close papilla contact with the receptive epithelium appears to be needed; in fact, if the papilla is separated from a growing follicle experimentally (298) or developmentally (hr/hr mouse) (393), follicle growth ceases. The importance of the papilla-epithelium interactions in the mature cycle is illustrated by the hairless (hr/hr) mouse and its human counterpart where disintegration of the papilla due to a transcription factor defect irreversibly abrogates the follicle's capacity to cycle, and ultimately leads to hair follicle destruction and alopecia (mouse, Refs. 56, 57, 392, 393, 395; human Ref. 2). Within the theme of E-M interactions, it is of interest that keratinocytes may act in turn on the mesenchyme; for example, keratinocytes produce a specific factor, which stimulates the growth of human scalp papilla cells in vitro (579).

As hair follicles differ from site to site, so do their papillae (e.g., Refs. 209, 463). Moreover, it is the papilla that appears to establish the character and size of the follicle and its shaft (204, 215, 572, 573). So, changes in the size of the hair follicle (important to folliculogenesis and to the vellus-terminal switch, see sect. XI) involves parallel changes in the volumes of both the epithelial and dermal parts of the lower follicle, namely, the FP dictates the size of the bulb (541, 572).

Many E-M interacting systems, such as the developing or regenerating limb bud, tooth, and feather (75, 79, 390, 557), express a set of patterning gene families (517). Because of the developmental/regenerative properties of the follicle, we are not surprised to find that many, if not all, of these same morphogenetic molecules are also used by the cycling follicle. Six major morphogenetic molecular family systems are now recognized to be important in this context: fibroblast growth factor (FGF), transforming growth factor (TGF)-, sonic hedgehog (shh), Wingless or wnt pathway, neurotrophins, and homeobox (hox) gene families (82, 384, 438).

In Table 3, we list these and other gene family molecules that appear to impact the control of the normal hair cycle. The published descriptions of the actions of these factors are in the main phenomenological, although new laboratory techniques, such as transgenic (gene addition and knockout) models, are offering new mechanistic insights. At the outset, it is important to appreciate that there are four themes regarding what we know about these molecular systems. First, the key signaling molecule families provide the metaphorical letters for the cross-talk occurring between the populations of the hair follicle mesenchymal and epithelial cells. Second, these gene families consist of cognate members at several levels: at the level of the ligand, the cell surface receptor, cytoplasmic transmitter, and nuclear transcription factor (e.g., Ref. 267). Third, these gene families have been found to interact extensively with one another and with other gene families (517). Although we have little direct data, we have every reason to believe that since these gene families are present, they are also interacting with many signaling pathways of the cycling follicle. Fourth, nature has endowed the hair follicle with substantial redundant pathways with which to function. The role of FGF5 is one example where, in its absence, catagen is only temporarily, but not permanently, delayed (179); this observation suggests, for example, that there are other catagen-inducing factors in the follicle. A second example deals with follicle morphogenesis where it has been found that no gene is absolutely crucial to hair follicle morphogenesis (providing that the animal develops to the stage where follicles first appear). The themes described are not dissimilar from those seen in other developmental systems (138). Undoubtedly, the organizing themes that will surface as we sort out the apparent welter of factors influencing hair growth and cycling will offer insights and new paradigms to the biologist, the pharmacologist, and the clinician.

For complete hair growth, the proliferative epithelium of the resting follicle must grow down into the deep dermis and, for large follicles, the subcutis. The downgrowths occur along a dermal trail, the fibrous stele (or streamer), which is established in the dermis by the first mature follicle (255). In contrast to the pelage follicle, cycling of the vibrissae follicle is fundamentally different in that the lower follicle does not regress upward; nevertheless, there is a cyclical change in the diameter of the vibrissae follicle with apparent movement of cells upward from the deep follicle bulb (613).

Although the downward growth of the early anagen follicle occurs by growth pressure (i.e., primarily by epithelial cell proliferation rather than migration, Refs. 308, 422), in order for this finger of cells to penetrate the dermis, the fibers of the fibrous stele must loosen. That the early anagen follicle is associated with proteolytic enzyme production (416, 584) suggests that the stroma is conditioned by enzymes produced by the down-growing follicular epithelium. Gene expression (4) and immunolocalization studies of matrix metalloproteinases (MMPs) show that while the papilla is negative for these enzymes, the epithelial cells of the down-growing bulb and ORS are positive. Follicles placed in vitro have been shown to have the ability to lyse collagen gels (584) and to synthesize and secrete various MMPs including interstitial collagenase, stromelysin-1, gelatinases, collagenase, and matrilysin (MMP-7) (145, 246, 324, 416, 584). EGF and TGF- which impact follicle growth also stimulate follicles to release various MMPs and lyse collagen gels (an action synergized by TGF-1 and TGF-2; Ref. 584).

Controlled degradation of ECM in the cycle requires a fine balance between the MMPs and their inhibitors. There is cyclic expression of the tissue inhibitor of metalloproteinases (TIMP) in the adult human follicle with localization, unexpectedly, to Henle's layer (249). TIMP is also found in the sebaceous gland and the proximal papilla (249). It is relevant that TIMP-3 is expressed by the anagen hair follicle, infiltrating basal cell carcinoma cells, and the stromal cells of squamous cell carcinoma (4). The similarity of the early anagen finger invading into the dermis to invasive basal or squamous carcinomas suggests a common mechanism between the down-growing follicle and invasive skin tumors (239). In contrast to its disruption in carcinomas, the BMZ of the follicle is never disrupted during anagen, but rather undergoes a highly coordinated process of constant remodeling so as to fully and continuously ensheath the growing follicle epithelium (398, 399).

At the end of anagen, terminal pelage hair follicles come to reside in the subcutis. Because the additional downward movement of the anagen follicle into the subcutis occurs at the expense of much energy and nutrient consumption, it would appear that follicular placement in the subcutis must be important; however, why the follicle base comes to be surrounded by adipocyes is unknown. Because the actual length of a hair follicle does not appear to dictate the length of its shaft (rather the duration of anagen determines hair shaft length), it is tempting to speculate that the subcutis offers optimal growth conditions (e.g., nutrients, temperature, hormones, neural stimulus) for the hair shaft factory and protects it from environmental insults by "buffering" it in fat cells as far removed from the skin surface as possible.

Excepting bilateral symmetry, it is probably true that no two follicles on the body are the same (e.g., Refs. 490, 504, 565). This point cannot be overemphasized because in experimental and clinical situations follicles from different regions are quite dissimilar. Follicles, and their shafts, differ grossly in length, thickness, curl, color, cross-section pattern, hormone sensitivity, innervation, vascularity, and the average time periods they spend in each of the phases of the cycle (203, 463, 490). The character of a follicle and its shaft is established by its papilla (214, 358, 438). Direct evidence for this conclusion is found experimentally where follicles induced by transplanted papillae are found to reflect the follicle from which the papilla originated (214).

Because the follicle regenerates itself in the course of each cycle, those factors affecting the character and placement of the follicle and its shaft must be active throughout the cycle. It is believed that follicle patterning occurs during follicle morphogenesis and that, in general, the character of the follicle is retained for the lifetime of the individual. Support for the unique and conserved character of a given follicle is most powerfully demonstrated by transplantation (381, 382, 568). A transplanted follicle retains the characteristics of its skin of origin. Thus eyebrow hair growing on the scalp would not be of value and scalp hair in the eyebrow region could be downright dangerous. Although one assumes that the patterning of a follicle is inherent to that follicle, in fact, with aging and in some disease states (e.g., acquired immunodeficiency syndrome, trichopathy, hair regrowth after chemotherapy, Ref. 94), the pattern of hair growth can change. The basis for the latter newly programmed structural changes is not yet clear, although its elucidation will be of great practical value.

The molecular basis for follicle pattern formation is only now being addressed. The regulatory gene families mentioned above, which impact hair morphogenesis, such as FGF, Wnt, TGF-, and hedgehog, undoubtedly play a critical role in the patterning of numerous developmental systems (138, 604). Because homeobox gene transcription factors play a pivotal role in body and limb development (138, 604) and are expressed in the follicle (24, 243), it is reasonable to assume that they also impact hair character. When one looks for the expression of these genes in the follicle, they are present. Hoxc8 gene, for example, is expressed in the papilla of follicles in a caudal to cephalic gradient; expression is most prominent in the papillae of the dorsal posterior pelage hairs (24, 243). Hoxd9 and Hoxd11 are expressed in the differentiating matrix of the anagen follicle. Hoxd11 is also strongly expressed in the basal cells of the ORS. Hoxd13 is restricted to the matrix cells of the follicular bulb, and its expression ceases in catagen (243). Msx1 and Msx2 are both restricted to the lower epithelial layers of the anagen follicle (465), and Alx-4 to the papilla (201). Because many patterning genes are expressed in the follicle, the situation is complex. It may be that the variable expression of patterning genes gives each follicle a unique address and thus a unique morphology. Arguing by analogy to the Drosophila (138, 604) perhaps it is the gradients of these gene products that color the actual heterogeneity of hair morphology. We assume these genes influence those aspects of the follicle and shaft that vary from site to site, and they act by orchestrating the expression of many downstream genes.

The first example of a defective homeobox gene with a corresponding hair phenotype was found with the Hoxc13 transgenic null mouse. It was found that Hoxc13 is essential to normal hair shaft development. In the absence of functional Hoxc13, hair shafts form but are friable (142). Paradoxically, this patterning gene plays a structural function in the mature follicle in addition to its patterning role in early development (97).

Without any idea as to its molecular basis, the embryologic origin of the dermis is also postulated to influence the pattern of scalp hair follicles. Chicken (88) and mouse (389) embryologic studies have shown that the mesenchyme of the scalp crown, face, and anterior neck are derived from the neural crest while the temporal and occipital scalp regions are derived from cephalic or somitic mesoderm (138). It is notable that in clinical hair disorders such as male pattern baldness and alopecia areata (ophiasis type) the respective regions are dramatically circumscribed and differ substantially in their clinical course and prognosis (94); in the former, the bald area corresponds to the neural crest mesoderm, whereas in the latter, the bald area corresponds to the somitic-derived mesoderm. The basis for hair follicle heterogeneity, then, may result from signals arising very early in development, which are expressed by a complex of patterning genes.

It is increasingly appreciated that the controls of patterning during anagen development are likely to involve the participation of direct cell-cell communication via adhesion molecules such as cadherins, cell adhesion molecules (CAMs), and integrins (358, 359). These serve as key elements in translating the one-dimensional genetic code into a three-dimensional tissue architecture (74, 75, 108, 138). A body of information is slowly emerging on the expression patterns and functional significance of adhesion molecules such as E- and P-cadherins, specific integrin pairs, NCAM, and intercellular adhesion molecule (ICAM)-1, during hair follicle morphogenesis (80, 127, 185, 244, 356-358, 360). The expression patterns of NCAM, ICAM-1, and E- and P-cadherin during murine anagen development have recently been characterized (80, 127, 185, 244, 356, 358, 359, 361). Although we now know their expression pattern, we know very little about how any of these adhesion molecules act in anagen initiation, development, or hair cycle control.

Generating from stem cell-like precursors in the "resting" telogen bulb epithelium (secondary hair germ, Fig. 2D), the earliest phase of anagen downgrowth (anagen I) shows no cylindrical layer differentiation. By very early anagen III (357), the IRS/ORS are identifiable. At the point that the finger of epithelium reaches its deepest level and perhaps somewhat before (anagen III), the layers of the follicle begin to form. Morphologically, there are at least eight cell lineages in the anagen follicle: ORS, companion layer, Henle's layer, Huxley's layer, cuticle of the IRS, cuticle of the shaft, shaft cortex, and shaft medulla (Fig. 2). In this section we focus on the cell lineages making up the cycling portion of the follicle. We would like to know from where each of the follicular layers arises and what controls their differentiation. Although the answers we have are still unsatisfactory, some basic principles are clear.

As mentioned above, the cycling portion of the anagen follicle is a solid epithelial cylinder made of embedded, concentric, unique cylindrical layers. The outermost cylinder, the ORS, separates the whole hair shaft factory from the dermis and subcutis. The ORS appears to be established during the early stages of anagen by the downward migration of the regenerating epithelium and then maintains itself (in contrast to the IRS and shaft), independent of the bulbar matrix, by basal cell growth (469). The thickness and cellularity of the ORS vary with the level of the follicle: it is single-layered just about the bulb, higher up it is composed of multilayered cuboidal cells which accumulate glycogen, and at the level of, and distal to, the sebaceous gland it becomes multilayered and is structurally similar to the epidermis (397). Although its role is thought to serve predominately as a support for the outward-growing shaft and IRS, the ORS is not quiescent; in fact, the ORS may well play an active role in hair cycle control. During anagen, the basal cells of the ORS below the sebaceous duct divide and replenish this layer (540); in addition, the innermost cells, but not the outermost cells, migrate distally and, ultimately, slough into the pilary canal (61, 62). That the ORS produces catagen-producing growth factors, like FGF5 and neurotrophins (40, 46), supports the idea that the ORS plays an important role in regulating the cycle (179, 432).

The middle cylinders of the follicle make up the IRS (541). The IRS molds and holds the shaft on its way to the surface. The IRS consists of three layers: the cuticle of the IRS, Huxley's layer, and Henle's layer. The cuticle layer of the IRS is made of scales that point distally and interlock with similar but opposing scales making up the cuticle of the hair shaft surface. The hair shaft scales point proximally. The combined, interlocked, cuticle structure allows the hair shaft and IRS to move together during the period of growth. Henle's layer, the first layer of the anagen follicle to keratinize (63, 357), encloses the shaft-sheath structure and interfaces the stationary ORS. As a supporting structure, Henle's layer appears to be inherently strong and tightly attached to the ORS. Its durability is appreciated by the rare separation that occurs within this plane after hair shaft pull (e.g., during preparation of a trichogram, Ref. 94) or histological preparation (540); in contrast, there is often disruption of Huxley's layer in such preparations. The IRS may move distally somewhat ahead of the shaft (human, Ref. 119; sheep, Ref. 62).

Huxley's layer is the major component of the IRS. It varies eccentrically in its thickness and thus molds the shaft to have round, oval, or flat cross-sectional morphology. The hair shaft and its IRS are complementary in shape so that together they form a solid core of hardened tissue with a nearly circular cross-section (456, 541). When the IRS is distorted, the shaft formed by that follicle is also distorted. This was observed in a mouse transgenic model in which ORS cell division was enhanced leading to malformed IRS and abnormal hair shafts (339). In addition, mice that have no TGF-, or EGF receptor (EGFR) expression (normally found in the ORS, Ref. 150), form irregular and curled hair shafts (302, 314); such findings suggest a role for the EGFR pathway in normal hair curl (shaft crimp is briefly discussed elsewhere, Ref. 455).

The molecular structure of the IRS is discussed in detail elsewhere (455). It is interesting and puzzling that many molecules of apparently disparate function are found within the IRS, such as FGF-1, FGFR4, TGF-, sonic hedgehog, dishevelled, -catenin (337), TIMP (249), and clusterin (506). Although we have no reason to doubt the latter observations, we recognize that the IRS takes up marker antibodies nonspecifically in many instances.

In addition to the shaft and its cuticle, at least five products normally pass from the pilary canal to the surface: sebum, sloughed IRS, sloughed ORS, apocrine secretions (in certain regional follicles), and organisms that inhabit the pilary canal (422). The egress of these materials is assisted by the cuticle of the outwardly moving hair shaft, which functions not unlike a "conveyor belt" made up of cuticular shovels that transport these materials to the skin surface. Thus the growth phase of the shaft, anagen VI, not only serves as the main production period for the fiber but also as a system for transporting skin secretions, debris, and parasites/microorganisms from the pilary canal to the skin surface.

In a study testing the origin of the hair follicle cell lineages, Kamimura et al. (241) labeled follicular epithelial cells in culture and recombined them with papilla cells using the nude mouse model of Lichti et al. (292) (see sect. II). They found that, instead of finding labeled cells at random throughout the newly formed follicles, the majority of the follicles appeared to derive from a minimum of two or three progenitor cells: one for the shaft, one for the IRS, and one for the ORS. The results suggest that only three precursor cells are needed to generate the eight epithelial cell lineages seen in the mature anagen follicle.

One conceivable molecular pathway for the control of the epithelial cell lineages of the hair follicle was suggested by studies of the Notch gene family. The Notch genes encode transmembrane proteins which, by facilitating local cellular interactions, serve to signal various cell fate decisions during development (13,