Hair follicle stem cells? A distinct germinative epidermal cell population is activated in vitro by the presence of hair dermal papilla cells

The hair follicle is a mammalian skin appendage that contains specialized dermal and epidermal components whose complex interactions result in epidermal hair fibre production. In contrast to interfollicular skin, which exists in a state of continual self-renewal, hair follicles display cyclical behaviour whereby periods of high epidermal mitotic activity and complex differentiation (anagen), are interspersed with apparent mitotic quiescence (telogen) (Dry, 1926; Chase, 1965). A group of germinative cells at the base of the follicle end bulb region have long been considered to be an epidermal stem cell population with a pluripotential nature, since during each period of active hair fibre growth they give rise to a number of distinct cell lines as they differentiate, including hair medulla, cortex, cuticles, and Huxley’s and Henle’s layers (Bullough and Laurence, 1958; Montagna and Van Scott, 1958; Epstein and Maibach, 1969; Montagna and Parakkal, 1974; Chapman, 1986; Powell et al. 1989).

The nature of hair germinative cells is poorly understood, and their relationship with hair outer root sheath (ORS) epidermal cells (which are more widely distributed along the length of the follicle) remains to be established. During telogen, a morphologically distinct germinative cell population cannot be identified, while ORS cells are permanently recognisable. Indeed recently, Cotsarelis et al. (1990) proposed that the ORS cells residing in the bulge area of the pilosebaceous unit are the only true follicular epidermal stem cells. On the other hand, the dermal papilla component of the follicle can induce the formation of hair-producing follicles containing germinative and ORS epidermis when associated with adult skin epidermis (Oliver, 1970; Reynolds and Jahoda, 1990).

Many aspects of hair follicle germinative epidermal cell behaviour also warrant investigation in their own right. For instance, in the whisker follicle the very rapid rate of epidermal cell division can produce 1.5 mm of vibrissa fibre growth in 24 h (Jahoda, 1982). This mitotic rate could be due to a stimulatory influence provided by the follicular dermis (Reynolds et al. 1991), an intrinsically high mitotic potential within the follicular epidermal cells (Malkinson and Keane, 1978), or both. By culturing germinative cells in isolation from papilla cells, two questions could be asked. (1) Were papilla cells required for continued rapid germinative cell mitotic activity? (2) Would the germinative cells retain their distinct small rounded morphology in the absence of contact with dermal papilla cells? In other words, do germinative cells only stay as such under papilla influence?

Previous hair follicle epidermal cell culture has almost exclusively used ORS cells prepared from plucked or whole follicles (Ward, 1976; Wells, 1982; Weterings et al. 1982; Vermorken and Bloemendal, 1986; Lenoir et al. 1988; Schaart et al. 1990). A single other report documents an attempt to grow a population of hair germinative epidermal cells in vitro. Jones et al. (1988) placed human hair matrix tissue (whole pieces or dispersed cells) into culture over dermal feeder-layers of a non-follicular origin. Cell numbers may have increased, but the authors state that their observations were inconclusive. Perhaps crucially, this work was done with material prepared from the bases of plucked hairs, while any cells left within the follicle after plucking were ignored.

The large size of the rat vibrissa (or whisker) follicle facilitates the manipulation and isolation of its component parts (Fig. 1). It has repeatedly and consistently been demonstrated (Oliver, 1965; Young, 1977; Jahoda, 1982) that when a growing vibrissa is plucked, although most of its lower matrix detaches from around the papilla and is removed with the fibre, a very small band of cells from the lowermost internal apex of the epidermal matrix horn remains attached around the basal stalk (see Fig. 2). This small amount of tissue is what we consider to be the hair germinative epidermal material, and differs in cellular composition from the larger amount of tissue that Jones et al. (1988) cultured from the bases of their plucked fibres.

Before comparing the in vitro morphology and behaviour of isolated germinative cells with other follicle bulb matrix (ORS) and non-follicular (skin basal) epidermal cells, we showed that the germinative material was compositionally different from these other epidermal cell types, using SDS-PAGE. We also demonstrated that our isolated germinative material was ultrastructurally identical to the appropriate region in situ. Finally, the hair follicle dermal papilla is postulated to induce, maintain and direct epidermal fibre production in situ. When, the three epidermal populations under surveilance were cultured in recombination with papilla cells, the papilla and germinative cells were shown to participate in profound and complex interactions with each other.

Male and female PVG/C rate (mbred colony, Department of Biological Sciences, Dundee University) were killed by ether anaesthesia as tissue donors for the procedures described below.

Prior to the isolation of follicular tissue, the relative lengths of club and growing fibres and the appearance of end bulb were noted so as to give an indication of the stage of the hair cycle from which material was being dissected.

Newborn rat skin was dissected free under sterile conditions and placed in Eagle’s minimal essential medium (MEM, Gibco) with antibiotics (penicillin 50i.u.ml^-1, streptomycin 50i.u.ml^-1, kanamycin 150 μgml^-1 and fungizone 2.5 μgml^-1; Gibco) for around 5h. Individual pieces of skin were scraped and cut into 5 mm² sections before incubation in a solution of 0.25 % dispase (Boehringer) in MEM at 4°C for 45 min. The epidermis was then peeled from the dermis and placed in 0.2 % crude trypsin (Sigma) in phosphate-buffered saline (Dulbecco A, Oxoid) containing 0.2 mg ml^-1 of ethylene diamine tetraacetic acid (PBS/EDTA) at 37 °C for 3 min, before an equal volume of 10 % foetal calf serum (FCS, Gibco) was added and the samples agitated to facilitate the release of a maximum number of skin basal epidermal cells. Cornified, suprabasal epidermal sheets were then removed and the solution was filtered through nylon gauze (12 pm aperture) and spun at 1800 revs min^-1 to obtain a pellet consisting predominantly of basal epidermal cells.

Mid-anagen hair fibres (and adhering layers) were plucked from adult rat mystacial (whisker) pad follicles and digested in 0.025 % dispase for 1 h at 4 °C. ORS material was then dissected from each ‘fibre’ (Fig. 3) and either placed into culture immediately, or following further dispase treatment (0.2%, 30 min, 4°C).

Fat and connective tissue were cleared from the undersurface of killed, adult rat mystacial skin flaps to expose the embedded follicle end bulbs. The most proximal tip of each (less than Jth of its entire length) was removed and transferred to a dish containing MEM with antibiotics at 4 °C. Alternatively, specimens were put aside to be processed for transmission electron microscopy (TEM). Using fine forceps, the outer collagen capsule and dermal sheath of this element were inverted. The exposed vibrissa matrix was then eased from the papilla and manipulated with forceps to release the lowest germinative band of cells (Fig. 4). Sometimes on separating the hair matrix from the papilla, germinative cells would remain attached to the latter, from which they could be retrieved (Fig. 4).

Germinative epidermal cells, in their natural position in unmanipulated hair follicle end bulbs, were compared with the cell material that had been isolated in the above procedure.

Tissue was fixed for 30 min in Sorenson’s 0.1M phosphate buffer (pH 7.2) containing 6 % glutaraldehyde and then post-fixed for 1 h in 1 % osmium tetroxide, followed by 0.3 % tannic acid for 10 min and a 1 % solution of sodium sulphate for 10 min. After dehydration in a graded ethanol series and Epon embedding, sections were cut on a Reichert ultramicrotome and stained with 2 % uranyl acetate and Reynold’s lead citrate.

Hair fibres were plucked from the mystacial follicles of dead animals. Any visible ORS material was peeled away from the growing fibres, prior to the removal of the lower portion of their matrices (including any attached VG cells). This amputated tissue was then placed in culture, either immediately, or following subjection to enzymic digestion (0.025 or 0.25 % dispase, 30 min, 4°C).

The material obtained from this method of preparation was similar to that isolated by Jones et al. (1988).

According to Lynch et al. (1986), the proteins in desquaminating tissues predominantly contribute towards ‘soft keratins’, while those in many appendages (e.g. hairs, feathers and scales) have lower sulphur contents and contribute towards ‘hard keratins’. We employed electrophoretic analysis to confirm that the three main epidermal types being investigated, and portions of more differentiated hair matrix, all represented discrete epidermal material. However, because pilot studies revealed that hard keratins were either absent, or only present in extremely small quantities in the three main epidermal preparations, only their soft keratin and first extraction fluid protein profiles were considered.

Soft keratins were extracted from freshly isolated populations of the three epidermal cell types described above (germinative, ORS, skin basal) and portions of hair matrix from which the germinative cells had been removed, using a method modified from Lynch et al. (1986). In brief, the material was homogenised for 20min at 4°C in 50 μl of extraction buffer (0.6M KC1, 25mM Tris-base, 1.0% Triton X-100, ImM EDTA, ImM EGTA, 5 μg ml^-1 pepstatin and 10 μg ml^-1 antipain, at pH 7.4) for every 2 μg (wet weight). An exception to this was germinative material, where the quantités available were so small that samples contained approximately Ipg (wet weight) in 50 pl of buffer. Samples were pelleted at 10 000 g for 20 min (4 °C) and the supernatants (termed the first extraction fluids) were retained. The procedure was repeated for the pellets and the second supernatants discarded: the second pellets of soft keratin and the first extraction fluids were then analysed by one dimensional gel electrophoresis.

Soft keratin and first extraction fluid samples were processed and then silver stained by the method of Heukeshoven and Dernick (1988), using PhastSystem gel electrophoresis (Pharmacia).

Dermal papillae were isolated during the manipulations involved in the simple dissection of germinative epidermal material (Fig. 4), fragmented and then forced to adhere to the bottom of Nunc Petri dishes. Approximately 30 vibrissa papillae were initiated in each dish.

Adult rat skin fibroblasts were cultured by explant outgrowth from small fragments of dorsal dermal tissue obtained during the skin basal epidermal cell preparations (see above).

All primary cultures were provided with 20% FCS in MEM with antibiotics, and cultivated at 37 °C with 5% CO₂.

The growth and survival of cells isolated from all three epidermal sources were initially examined under standard conditions of 37°C, pH7.3, and 5% CO₂ in Primaria dishes (Falcon laboratories) with epidermal medium. A favourable epidermal medium was based on published information relating to mouse and human epidermal cell culture, further supplemented by a preparation of foetal rat pituitary extract. It consisted of 80 % MEM containing 20% foetal calf serum (FCS, Gibco), 1% L-glutamine at a final concentration of 2 mM and antibiotics, in addition to epidermal supplements of; 140μgml^-1 foetal rat pituitary extract (technique modified from Tsao et al. (1982)), 10/igml^-1 insulin (Sigma), 5μgml^-1 transferrin (Sigma), 0.4μgml^-1 hydrocortisone (Sigma), 0.01 pgml^-1 epidermal growth factor (Sigma) and 10⁻⁹M cholera toxin (Sigma).

Comparative morphological observations were made using a Nikon inverted microscope (model: Diaphot-TMD).

Germinative cell responses to hair papilla or skin fibroblast dermal support were compared with those of the other follicular epidermal population (ORS), and basal epidermal cells from newborn skin. Germinative epidermal tissue pieces (each containing a few hundred cells) from around 50 follicles, or equivalent quantities of ORS or skin basal epidermal cells, were seeded over primary populations of either hair follicle dermal papilla cells or newborn skin fibroblasts. Twelve repetitions were conducted for each variation of cell recombination, each of which was provided with epidermal medium as described above under standard conditions.

Inverted microscope observations and photographic recordings of the cultured cells were made at regular intervals. Selected material was also processed for TEM as described above, except that filtered Ruthenium Red (Gurr) was included in both the glutaraldehyde-based fixative and the osmium-based post-fixative to give a final concentration of 0.05 % (Meyer et al. 1981), as an aid to the visualisation of extracellular materials and of proteoglycans m particular.

Unpigmented bands of material containing small, round cells, were isolated from the appropriate regions of anagen follicles (Fig. 5). The dissection of VG cells from fibres that were near to the end of their growing phase was less clear cut, as material was patchy and unevenly distributed around the papilla so that smaller amounts were obtained. Tissue was not isolated from telogen (apparently quiescent) hair follicles.

The hair germinative epidermal matrix cells that we had isolated by microdissection (Fig. 7) were observed to be morphologically very similar to hair germinative epidermal cells in situ (Fig. 6). In both populations the cells were observed to have 4–7 μm diameters, and be surrounded by a small amount of homogeneous cytoplasm that contained very few organelles.

Three alternative methods of obtaining germinative epidermal cells were then compared with the above method for ease of dissection and consistency of material isolated.

Dispase pretreatment of end bulbs (0.025, 0.25 or 2.50% dispase for 10–100 min at 4 °C) did not facilitate dissection, since the VG cells became more sticky, were less easily delineated and more difficult to manipulate.

Less germinative matrix tissue was obtained when hair fibres were plucked immediately prior to dissection. Histology demonstrated that the plucking of growing fibres left most of the germinative material behind (Fig. 2), but loss and disruption of this fragile component occurred during bulb manipulation, making cell retrieval much more difficult.

Dissection of VG cells at progressive stages of hair fibre regeneration (days 1–10) revealed that most material (greater than that achieved by any of the previous methods) was present 5 days post-plucking. However, the papilla-epidermal junction was so tightly associated that attempts to separate components were largely unsuccessful. In contrast to unplucked material, treatment with 0.025% dispase for 1–2h at 4°C was found to greatly facilitate separation, whereupon only the lowest matrix cells were broken away using forceps.

In stark contrast to the skin basal cells, and consistent with the results obtained using larger gels (data not shown), no significant amounts of soft keratin could be detected in germinative cells (Figs 8, 9). Although the skin basal cell preparations contained the most complex set of polypeptides, the simpler ORS and matrix material profiles were nevertheless clearly variable from each other and that of the skin basal cells (Fig. 9).

The most intense band (which appeared to have a M_T value of approximately 56 × 10³) was from the germinative cells (Figs 8, 10), while a pair of faint, lower M_r bands (approx. 43 ×10^s and 44×10³) was unique to the ORS lanes (Figs 8,10). The matrix material produced no obvious bands (Fig. 10) and the basal cells contained considerably smaller quantities of the polypeptide band observed in the germinative tissue (Figs 8, 10). All of the epidermal extracts contained three very low M_r bands (around 13×l0³, 15.5×10^s and 16×l0³), and additional very faint bands were similarly distributed throughout all but the germinative cell profiles (Figs 8, 10).

Most of the dissociated epidermal cells (initially added to their dishes in suspension) were attached individually, or in clumps,, after approximately 5h. They subsequently spread and proliferated into ever larger colonies of polygonal (mainly hexagonal) cells, until a confluent monolayer was formed about one week after plating. At this stage most of the dish contained exposed basal cells, but some areas were more densely covered, possibly owing to greater initial attachment or subsequent stratification (Fig. 11). Irregular patches of stratifying cells gradually enlarged and formed morphologically variable layers in various states of differentiation. Dead cells were large, flat, striated and enucleate (Fig-12).

Initial cell outgrowth from isolated material was more rapid in the dispase treated (cf. untreated) preparations, but the final cell numbers were similar. Colonies were usually well-formed around their central expiants within a few days and appeared morphologically similar to the vast majority of the skin basal epidermal cells (Fig. 13), although there was less stratification and peripheral cells appeared larger. After about two weeks in culture, areas consisting of dead/differentiating, anucleate cells spread out from the centre of each colony (adjacent to the expiants), as in the skin derived cultures (Fig. 14).

The few cells (derived from mid-anagen follicles) that initially attached in culture were small, round and frequently detached within a few days. They displayed no indication of growth, division or flattening into an ORS-type morphology, either in isolation, or alongside ORS cells (Fig. 15). Material from late anagen dissections was sparser than the above and cells were even more reluctant to attach. Finally, the very early anagen material (Fig. 16) was different in that a proportion of cells were larger, more compactly arranged and morphologically variable (Fig. 17).

When entire matrices were placed in culture, so that the lowest regions of the expiants remained in close contact with the substratum, cells with a typically ORS-like morphology often grew out from the sides and established colonies (Fig. 13).

Three of the twelve germinative cell cultures initiated directly over dermal papilla cells remained more or less inactive. The germinative cells appeared small and round, and never formed large pavement arrangements typical of other epidermal types (Fig. 18). Whereas in isolation they had been observed to detach over time, in combination they remained attached to the papilla cells for as long as the cultures were maintained. In the remaining nine associations the epidermal cells divided and developed in a variety of ways, while still retaining their distinctive small, round morphology. They either formed tight clumped associations with the underlying papilla populations, or remained as patches of cells that spread and rearranged themselves above them; in the latter case this sometimes resulted in the creation of extended horizontal outgrowths of germinative cells above the papilla monolayers (Figs 19, 20). On at least four occasions, dermal-epidermal interaction resulted in the formation of even more complex, structured arrangements. Sometimes these appeared as compact, localized cell associations, which developed small vertical mounds at their centres and subsequently grew into elevated domes of tissue (Figs 21, 22). Alternatively, organotypic-like structures developed in a predominantly horizontal direction from thickened regions of interacting cells. In these, a definite internal organisation was visible macroscopically, with most appearing to have two or three lobes around a central projection (Fig. 23). Semi-thin sections and TEM revealed that dermal papilla cells had completely surrounded an outgrowing epidermal extension at the apex of one of these structures (Fig. 24). A basement membrane zone, with a recognisable basal lamina, was present at the dermal-epidermal interface (Figs 25, 26) and hemidesmosome junctions were apparent between this and the epidermal cells. Gap (Fig. 27) and desmosome (Fig. 28) junctions were abundant between the closely packed, central germinative epidermal cells (Fig. 29) indicative of cell-cell communication (see Fig. 27). The germinative cells possessed a high nuclear to cytoplasmic ratio; the nucleus was spherical with patchy heterochromatin and the cytoplasm was scant with few organelles (Fig. 29). Frequently mitochondria were the only organelles observed, although the cytoplasm contained numerous free ribosomes (Fig. 29). This contrasts, for example, with skin basal epidermal cells that have been cultured over hair dermal papilla cells; these are larger, have a much lower nuclear to cytoplasmic ratio, and more varied cytoplasmic contents (Fig. 30).

Dermal papilla cell membranes in the upper regions of the organotypic structures displayed particularly intense Ruthenium Red staining (Fig. 31), which may have obscured the presence of intermediate junctions in these regions. Cell-cell communication and gap junctions were especially common throughout the centrally located papilla cells, but were observed throughout the structure. Some extracellular fibrous material, intermediate in size between keratin and mature collagen fibrils, was evident between the more central papilla cells, and this may have represented elastin or immature collagen. The more peripherally located dermal cells, either side of the germinative epidermal outgrowth, had produced much larger quantities of fibrous collagen (Fig. 32). Both dermal and epidermal cells contained many vesicles filled with multi-laminated membranous structures (Figs 25, 31, 33), the significance of which was unknown.

Germinative epidermal cell growth over skin fibroblasts was very poor (Fig. 34), but they survived for several days longer than in isolation. There was no indication of epidermal cell division, differentiation, or the formation of any kind of organised structures. Conversely, the fibroblast cells’ growth and behaviour did not appear to be significantly affected by this epidermal presence.

The ORS cells grew well both directly over, and on exposed substratum between, the papilla cells. However, they did not form the tight aggregations or compact cell outgrowths seen in the germinative/papilla recombinations. Where they were attached to the substratum they formed pavements, but when they settled directly over the dermal cells they remained small and round (Fig. 35). As ORS cell division proceeded, most of the newly formed cells were also small and round (Fig. 36), but when placed into cultures containing germinative cells, the size difference between the small, round cells from each population was readily apparent (Fig. 37). ORS cell numbers did not increase significantly after 2 weeks in culture and after 4 weeks they often died or differentiated (but the cells remained smaller and more loosely arranged than in isolation; compare Figs 14 and 38). Between weeks 2 and 4 cells changed little. Although the ORS/papilla cell interface was irregular and contained numerous long extensions of epidermal cells making contact with dermal cells (Figs 39, 40), there was no sign of any more complex developments.

There appeared to be a consistent tendency for a higher proportion of the ORS cells to attach on the substratum between the fibroblasts, than had been observed in the dishes containing papilla cells. Less interaction and visible cell-cell contact were obvious between the dermal and epidermal cells at their interface, and the duration of epidermal cell survival was reduced by about a third.

The extent to which the basal cells spread varied; those that attached directly above papilla cells tended to remain rounded, whereas those attaching to the uncovered substratum often formed pavements (Fig. 41), as was the case with the ORS cells. Subsequent cell behaviour and survival were also very similar to those observed for the ORS cells, except that the dermal-epidermal cell interfaces were generally much more regular.

A small proportion of the basal epidermal cells attached directly over the dermal fibroblasts, but the majority settled on the substratum between them. As during the recombination with papilla cells, their morphology and behaviour was generally quite distinct from the germinative cells and similar to that of the ORS populations. Identifiable structures were never formed, and when these epidermal cells eventually died/differentiated their appearance was similar to that when cultured in isolation (Fig. 42).

Certain inferences can be made from the fact that germinative cell dissection at different stages of the hair cycle produced quantitative variations. In general, the amounts of material obtained from different sizes of midanagen end bulb reflected the direct relationship between dermal papilla and germinative matrix volumes in vibrissa follicles (Ibrahim and Wright, 1982). However, it was particularly noteworthy that less material was obtained from the maximum sized bulbs at late anagen/early catagen; apparently due to an asymmetrical distribution of the germinative population. This finding could be explained by assuming that the germinative cells are uniformly active and supply the whole matrix circumference during mid-anagen, but then switch to an asymmetrical pattern of proliferative activity towards the end of the growing phase. It also supports the idea that during whisker follicle regression, one side of the hair matrix switches off before the other, a phenomenon that was related to other rat whisker follicle asymmetries by Jahoda (1982). The presence of fewer germinative cells during late anagen might indicate that these cells are lost at the end of each hair cycle.

The fact that careful dissection of untreated follicles provided a better yield of germinative cells, than when enzymic pretreatments were involved, reflects the fragility and difficulty of handling these cells. Although plucking of fibres produced relatively large batches of material after a few days, these regenerating germinative cells displayed certain characteristics that suggested that they might not be totally analogous to the equivalent cells from a normal cycle.

Transmission electron microscopy of rat vibrissa follicle germinative material immediately after dissection, was an important verification of the finding that the cells that had been isolated were morphologically identical to those in situ. It also confirmed that the cell population obtained was essentially homogeneous, different from, and not obviously contaminated by, differentiating matrix cells positioned just above in the hair bulb, or ORS cells to the side. In this respect, our germinative cells differed from the epidermal populations, which are derived by cutting the matrix material from the base of a plucked fibre, as used by Jones et al. (1988).

Hair germinative, matrix, ORS and skin basal cells were also demonstrated to have variable biochemical compositions, by comparing their respective protein profiles after a soft keratin extraction procedure. The hair matrix, ORS and skin basal cells all displayed variable combinations of soft keratin bands, while the germinative cells contained none. Most studies investigating hair epidermal keratins have used sheep, human or mouse tissue, rather than rat (Gillespie, 1983; Lynch et al. 1986; Delorme et al. 1987; Franke and Heid, 1989; Schweizer, 1989), but to our knowledge no one has previously analysed any population of carefully isolated germinative cells. However, Heid et al. (1988) have shown that hair matrix germinative cells do not express trichocytic (hairtype) cytokeratins, so that their findings are consistent with our observations.

Although the hair germinative cells did not appear to contain any keratins, a single first extraction fluid (FEF) polypeptide band (at a M_r of about 56 × 10³) stood out by its isolation and intensity. There is potential for further investigation into its possible role as a marker for the cells, and it may represent a molecule of some functional importance. Interestingly, hyperproliferative epidermal cells express a keratin with an equivalent M_r (Moll et al. 1982, 1984). As support for the hypothesis that skin and hair follicle germinative cells could be similar, skin basal epidermal cell FEF extracts also contained an equivalent protein, but in lower amounts.

The fact that isolated germinative cells retained a very different morphology from ORS, matrix or skin basal cells, and were reluctant to attach, spread or divide under conditions suited to epidermal growth, reinforced the supposition that they constituted a distinct population. As well as the method described here, a comprehensive range of conditions, with variation in pH, temperature, substratum and media (including the use of unusual additives such as foetal rat serum), has been used in attempts to incite activity in isolated germinative cells. All were unsuccessful (Reynolds, 1989). When the growth and behaviour of plucked ORS and skin epidermal cells were monitored under these multiple variations (Reynolds, 1989), they displayed properties generally comparable with previous investigations (Weterings et al. 1981; Wells, 1982; Limat and Noser, 1986; Imcke et al. 1987; Limât et al. 1989).

In the present work, it was particularly significant that when the environment was identical for both ORS and germinative cells (when both were established in the same Petri dish), each retained its identity, with ORS cells becoming flattened and the germinative cells remaining small (approximately 4–7 μm) and round (Fig. 15). If the isolated germinative material had been ‘contaminated’ with ORS or more committed matrix cells, then a proportion of the germinative cells would have developed ORS-typical morphology and behaviour, or displayed some evidence of hair cell differentiation. Two pieces of information could be gleaned from these observations. First, the very high mitotic activity of germinative cells seems to be under extrinsic control, so that its initiation requires the presence of certain active or permissive environmental influences. Second, the germinative cells do not seem to have a unique morphology merely as a consequence of their position in the hair follicle. For example, they are not ORS cells that take on temporary, reversible germinative cell properties as a result of the influence of lower follicle mesenchyme. If this were the case, they would have reverted to ORS-typical morphology and behaviour when removed from the in situ influences. Regardless of their origin, hair germinative epidermal cells that have been removed from the follicle retain distinct and unusual properties.

Our findings are not comparable with those of Jones et al. (1988) because their human germinative material was obtained from the bases of plucked fibres, while ours was equivalent to the material that is left attached to the dermal-epidermal junction after plucking. Our attempts to grow germinative cells from the bases of plucked vibrissa (or human hair) fibres (i.e. matrix regions), usually resulted in no obvious germinative cell activity, although ORS cell outgrowth was sometimes seen. Moreover, isolated human germinative cells (dissected from follicle end bulbs) displayed similar general patterns of behaviour to those from vibrissa follicles (Reynolds, 1989).

The direct association of germinative epidermal and dermal papilla cells resulted in enhanced germinative cell attachment and mitosis for the first time in vitro. Since outgrowths of compactly arranged, multi-layered germinative cells (with an apparently undifferentiated phenotype) were observed, their previous inactivity in isolation could not have been due to poor viability. The possibility that the germinaitve material consisted of postmitotic cells committed along the hair differentiation pathway, but arrested by the isolation process, was also disproved. The fact that even active germinative cells retained a morphology that was distinct from ORS cells, once again reinforced their separate identity. Moreover, the absence of significant division in germinative cells recombined with skin fibroblasts, inferred that the activating influenced) were the result of specific stimuli from follicular dermal cells.

Dermal papilla cell support of ORS and skin basal epidermal cell growth, reinforced the finding that papilla cells are capable of exerting a general positive influence on epidermal cell maintenance and division (Reynolds et al. 1991). Adult hair papilla cells produce a unique extracellular matrix that includes basement membrane components, and a much lower concentration of interstitial collagens than is normally found in dermis (Couchman, 1986; Messenger, 1988). Epidermal cells display preferential adherence to, and subsequent growth on, basement membrane components, and a number of reports have shown that type IV collagen-coated surfaces promote keratinocytes adhesion (Murray et al. 1975; Kubo et al. 1987; Wilke and Furcht, 1990). Moreover, the epidermal growth factor-like sequences in the laminin molecule have been shown to elicit growth factor activity, raising the possibility that laminin might provide growth factor influences by enzymic release of the appropriate regions of the molecule (Panayotou et al. 1989). Hence, the enhanced epidermal cell growth could have resulted from the effect of attachment and growth promoting extracellular materials (and possibly medium-borne growth factors) produced by papilla-epidermal cell interactions.

Neither ORS, nor skin basal epidermal cells, participated in the development of complex structures made up of dermal and epidermal components. However, since germinative epidermal and dermal papilla cells are normally intimately associated in vivo, it seems logical that they might be more suited than other cell types to interact with each other in vitro. The nature of their interaction was also informative, and to some extent surprising, because it did not result in epidermal cell differentiation but in morphogenetic activity. The formation of organotypic structures in the germinative/papilla recombinations involved the co-ordination of specific cell arrangements, as well as cell division. It even appeared as if the recombined cells were trying to recreate an arrangement similar to that in the base of the hair follicle, because the papilla cells migrated over and totally enclosed the germinative epidermal cells to form a comparable arrangement. The inference is that the overriding or first signal between these two cellular elements when in direct contact is to ‘make a follicle’, rather than ‘make hair’. The formation of a distinct basal lamina between the papilla and germinative populations suggested that there was complex communication in process, with the extracellular matrix basement membrane components being organized to facilitate both structural and functional roles (Briggaman, 1982; Stanley et al. 1982; Martin and Timpl, 1987). While the presence of extracellular matrix molecules that are common to basement membranes has been recorded many times during in vitro dermal-epidermal recombination (Timpl and Dziadek, 1986; Hornung et al. 1987; Simon-Assmann et al. 1988), the presence of a visible basal lamina in the form of a continuous sheet has not to our knowledge previously been cited in skin-derived, cultured cell populations. In fact, Hornung et al. (1987) state that the structural organization of a basement membrane (basal lamina) requires epithelial functions that are not expressed in vitro. Other authors, including Bohnert et at. (1986) and Breitkreutz et al. (1986), suggest that normal or transformed keratinocytes are only able to form a structurally organised basement membrane in vivo. In contradiction, our results suggest that this is not the case, at least for hair follicle-derived cells. Moreover, the simplicity of the recombination, which was made on a basic plastic substratum, infers that a favourable cell combination appears to be paramount, rather than complex culture conditions. In other words, it shows that papilla and germinative cells can between them produce all the extracellular consituents for basal lamina formation, and provide any additional biochemical or topographical elements required for normal structural assembly. In this context, it is interesting that Grant et al. (1989) have shown that lamina densa-like sructures can be produced simply by incubating the three most universal basement membrane components together at 35 °C. At any rate, papilla-germinative associations may turn out to be a useful model system in this field of in vitro basal lamina formation.

Concerning the nature and distribution of extracellular materials in the organotypic structures, the size and organization of the collagen fibres was impressive, since they are not observed in isolated dermal papilla cell cultures (Jahoda, unpublished). The uneven pattern of distribution may be a reflection of morphogenetic activity, because fibrous collagen is known to accumulate in histogenetically stabilised zones in the development of cutaneous appendages Sengel (1986). Ruthenium Red marking indicated that proteoglycans, which have recently been the subject of considerable interest with regard to hair growth (Couchman, 1987; Couchman et al. 1990), were prevalent among upper (possibly more active) regions of the structures.

A pertinent dilemma, which is highlighted by this study, relates to the title that germinative epidermal cells should be given in accordance with their behaviour and the position they occupy within the hair follicle. That is, do they represent pluripotential stem cells; stem cells from the bottom of the hair differentiation lineage, a ‘transit’ amplifying population, some combination of these states, or a category of cells for which there is at present no accurate definition.

Stem cells are generally defined as slow-cycling subpopulations at the origin of any ordered cellular arrangement, while those at the next level are said to undergo a limited number of more rapid (or amplifying) ‘transit’ divisions before giving rise to non-proliferative, terminally differentiating cells (Potten, 1974; Caimie et al. 1976; Lord and Dexter, 1988).

The unknown inter-relationship between follicular germinative and ORS epidermal populations during the hair follicle cycle, is probably the main factor preventing any absolute classification in this area (discussed by Reynolds, 1989; Reynolds and Jahoda, unpublished data). In brief, either: (1) germinative cells successively give rise to a new hair fibre during each anagen and remain inactive during each telogen, or (2) they are lost (or destroyed) towards the end of each cycle and then replaced by ORS cells at the initiation of the next.

Whichever scenario proves to be the case, it would appear to us that both germinative cells and a subpopulation of ORS cells represent stem cell-type populations in their own right, but with differing characteristics. Both have unspecialised/primitive/embryonic-like features, such as small size, few organelles and an abundance of free ribosomes, which are stem cell traits (Leblond and Cheng, 1975). Also, both occupy specific, protected (that is, not destroyed by plucking) locations, and are low in number.

Recently, Cotsarelis et al. (1990) suggested that a certain part of the upper ORS contained the only true follicular stem cells, and proposed that hair matrix germinative cells are a ‘transit’ amplifying population with a limited proliferative potential and a set path of differentiation. However, our results with vibrissa follicle material suggest a less rigid classification, since in combination with dermal papilla cells we found that germinative cells were capable of extended proliferation and participated in the initiation of profound morphogenetic arrangements. In other words, they are certainly not limited to proliferation followed by hair-type differentiation and thus are not easily categorized as ‘transit’ amplifying cells (as the definition stands).

In our opinion, at least some ORS cells are accurately designated as pluripotent stem cells, since they can provide a new population of germinative cells after follicle end bulb amputation (Oliver, 1967a,b) and contribute towards new skin epidermis during wound repair (Eisen et al. 1955; Sanford et al. 1965). As further support for this proposition, if (2) is true, all the epidermal cell layers of the hair follicle would ultimately derive from the ORS cells. On the other hand, dermal papillae can switch-on hair follicle epidermal differentiation! in scrotal sac skin epidermis (Oliver, 1970), and dermal papilla cells have recently been shown to induce glabrous footpad epidermal cells to become fully functional follicular ORS or germinative epidermis (Reynolds, 1989; Reynolds and Jahoda, 1990). Therefore, the stem cell attributes of ORS cells must be considered alongside, and balanced against, the capabilities of other epidermal populations to become apparently ORS cells when subjected to specific dermal influences. Furthermore, because we found that follicular germinative epidermal cells have inductive and interactive capabilities that ORS cells do not possess (Reynolds, 1989; Reynolds and Jahoda, unpublished), the former would appear to be uniquely primed in some way. The significance of their unusual state of activity is therefore elevated - particularly with reference to potential experimentation.

In summary, we have shown that epidermal cells from the adult hair follicle germinative region are a discrete epidermal population, whose activities, both in isolation and in recombination with appendgeal dermal cells, distinguish them from other follicle epidermis. Since multiple cell-cell and cell-matrix interactions are involved in the complex hair cycle in vivo (with reciprocal interactions between the dermal and epidermal populations) it is likely that study of many follicular activities in vitro will be improved by the presence of both dermal and epidermal types of known origin. To this end, our isolation and cultivation of germinative cells, provides a novel tool. Moreover, because papilla-germinative cell recombinations proliferate and undergo various forms of complex activity in vitro, they represent a potential model system with which to identify factors that influence hair growth and development.

We thank the Wellcome Trust for support, and Dr J. Riley for helpful work on the manuscript. The technical assistance of Bruce Pert and Sean Earnshaw is gratefully acknowledged. A.J.R. was supported by a University of Dundee scholarship. C.A.B.J. was a Royal Society 1983 University Research Fellow.