Clinical Science (1996)
91, 141-146 (Printed in G r e a t Britain)
141
GlaxolMRS Young Investigator Prize
Isolation and characterization of human epidermal stem cells
P.
H. JONES
Keratinocyte laboratory, Imperial Cancer Research Fund, London, U.K.
1. The keratinocytes in human epidermis are constantly turned over and replaced by a population of
stem cells located in the basal epidermal layer. Until
recently there were no markers allowing the isolation
of viable epidermal stem cells. However, it has now
been shown that epidermal stem cells can be isolated
both in vitro and direct from the epidermis as they
express high levels of functional
integrin family
receptors for extracellular matrix proteins.
2. The evidence for integrins as stem cell markers
and the insights that have been gained into stem cell
behaviour are reviewed.
INTRODUCTION
The outermost layer of human skin, the epidermis, consists of layers of keratinocytes. Proliferation
is mostly confined to cells in the basal layer. As
basal cells differentiate, they leave the basal layer
and migrate towards the surface of the epidermis.
As they progress they undergo a series of biochemical and morphological changes which culminate in
formation of the anucleate cornified cells that make
up the outermost epidermal layer (Fig. 1).
Keratinocytes are continually shed from the surface of the cornified layer; in adult hair-bearing skin
almost all the cells lying at the epidermal surface
are lost every 24h [l]. The cells.that are lost are
replaced by proliferation of cells in the basal layer.
The ability of the epidermis to maintain this constant turnover of cells led to the suggestion that
among the keratinocytes in the basal layer was a
population of stem cells (Fig. 1). These retain the
ability to proliferate throughout life, both generating
cells that become terminally differentiated keratinocytes and self renewing to maintain the stem cell
population [2, 31. By analogy with haemopoiesis it
has been suggested that stem cells generate postmitotic terminally differentiating keratinocytes via
an intermediate population known as transit amplifying (TA) cells [3]. After four to five rounds of cell
division, all of the daughters of TA cells withdraw
permanently from the cell cycle and undergo terminal differentiation.
Evidence for the existence of stem cells comes
from two sources. It is argued that because each
stem cell division results in 32 terminally differentiating cells, in the steady state, when the rate of
production of terminally differentiated cells is constant, stem cells divide infrequently compared with
TA cells [4]. Basal cells that appear to divide
relatively infrequently were identified in mouse epidermis using tritiated thymidine labelling techniques
[ S ] . In monkeys injected with pulses of tritiated
thymidine, fewer basal cells are labelled in the rete
ridges of the palm, where the epidermis projects
down into the underlying dermis, than the basal
cells in other areas (compare with Fig. 3b [6]).
These observations are consistent with the stem
model suggesting that stem cells may lie in the rete
ridges in the palm [6].
The second line of evidence comes from studies in
vitro. Cultures of normal human keratinocytes contain colonies that differ in size. Some colonies are
large and circular, containing thousands of keratinocytes, whereas others are small and irregularly
shaped [7]. In a series of experiments involving
subcloning the colonies produced by single cells, it
has been shown that when the large, circular keratinocyte colonies are subcloned they produce further
large colonies as well as some small colonies [7].
The subcloning of small colonies produces either a
few small colonies or no colonies at all. No case of
a small colony generating a large colony when it
was subcloned was observed. The cells that found
large colonies behave like stem cells in that they are
capable of extensive self renewal and generate many
terminally differentiating cells [7]. The cells that
found the small colonies are like TA cells, with a
restricted ability to proliferate and little or no self
renewal.
Clinical studies involving the grafting of sheets of
cultured keratinocytes onto patients with extensive
burns also provide powerful evidence for the existence of stem cells. The autografts give rise to a near
Key wordr: cell adhesion, cell differentiation, cell division, integrins, keratinocytes.
Abbreviations: CFE, colony forming efficiency; ECM, extracellular matrix; FITC, fluorescein isothiocyanate; TA, transit amplifying.
Correspondence: D r P. H. Jones, Keratinocyte Laboratory, Imperial Cancer Research Fund, 44 Lincoln’s Inn Fields, London W C U 3PX, U.K.
P. H. Jones
I42
a+
Shedding
Table I: lntegrinr expressed by keratinocytes and their ligands
lntegrin
1
S
t
x Tryit
y
Committed
--.
Fig. I. Model of epidermal stem cells. Among the proliferating cells in
the basal layer of the epidermis are stem cells, which can extensively self
renew and retain their proliferative potential throughout adult life. Stem
cells generate transit cells, which are only capable of four t o five rounds of
cell division. All of the daughters of transit cells become post-mitotic basal
cells, committed t o terminal differentiation, which leave the basal layer and
are eventually shed from the epidermal surface. Only basal cells are able to
adhere to matrix proteins in the underlying basement membrane via
integrin receptors which are only expressed in the basal layer.
Ligand
Collagen, laminin I
Laminin 5, laminin I
Fibronectin
Unknown
Unknown
Laminin 5, laminin I
Vitronectin
*aspIis readily detectable in culture, but is either undetectable or expressed at
very low levels in normal epidermis.
process, at the stage when the cell leaves the basal
layer, all the integrins are lost from its surface [14161. Thus, differentiation in keratinocytes is accompanied by a loss of integrin mediated adhesiveness
to ECM proteins. The hypothesis was that the least
differentiated cells in the basal layer, the stem cells,
might be the basal cells that had the highest levels
of integrin mediated adhesion.
normal epidermis that persists for many years
[8-1 01.
The major stumbling block in the study of epidermal stem cells has been the lack of molecular
markers to allow the isolation of viable stem cells
from epidermis. The goal of the experiments described here was to find a stem cell marker, isolate
stem cells, localize them in the epidermis and begin
to characterize their behaviour.
STEM CELLS AND ADHESION
The strategy used to isolate stem cells exploited
what was known about changes in the adhesiveness
of keratinocytes to extracellular matrix (ECM) proteins during their differentiation. An intrinsic part of
keratinocyte differentiation is a loss of adhesion to
the basement membrane, essential for stratification
[l 11. Basal keratinocytes adhere to ECM proteins
in the underlying basement membrane via receptors
on their surface. These receptors are known as
integrins. They are very widely expressed in different
cell types and have a key role in numerous developmental and pathological processes [121. Integrins
are dimeric proteins comprising an a and a B
subunit. There are several different a and p subunits
and in a given cell the different combinations of CI
and p subunits form receptors for different ligands.
Expression of a given integrin does not necessarily
mean a cell will adhere to the relevant ligand as
integrins can be functionally active or inactive. The
integrins expressed by keratinocytes in culture and
in uiuo are shown in Table 1 [13].
In experiments in which cultured keratinocytes
are induced to differentiate by placing them in
suspension it was found that as basal cells became
irreversibly committed to terminal differentiation,
their integrin receptors became inactivated and
unable to bind to ligand. Later in the differentiation
ISOLATING STEM CELLS IN VITRO
A convenient source of keratinocytes was epidermis cultured in uitro. The culture method used was
that of Rheinwald and Green [17, 181, in which the
integrin expression and adhesive properties of keratinocytes have been well characterized [191.
The first problem was how to define a stem cell.
Clonogenicity is the characteristic of stem cells that
is simplest to assay in culture, so a working definition of a stem cell based on clonogenicity was
adopted: a stem cell was a cell that founded a
keratinocyte colony visible under a dissecting microscope after 14 days in culture. In practice this
means a colony of at least 32 cells, although most of
the colonies were much larger, containing over 5000
cells.
Two methods were used to fractionate basal
keratinocytes on the basis of their adhesiveness to
ECM. The first was to sort cells according to the
level of surface expression of integrin receptors using
flow cytometry. The basal cells in a single cell
suspension of cultured keratinocytes were identified
on the basis of their light scattering properties. The
keratinocytes were then stained with anti-integrin
antibodies conjugated to fluorescein and sorted into
the cells with highest, mid and lowest fluorescence.
The sorted cells were cultured for 2 weeks and their
colony forming efficiency (CFE) was scored. The
results of sorting cells on the basis of C I ~integrin
subunit expression are shown in Fig. 2. Using
analysis of covariance it was found that CFE was
proportional to the log of integrin fluorescence in
basal cells for the u2 subunit and the
integrin
subunit that partners a2 [20]. The differences in
fluorescence between the brightest and the dullest
cells were only 2-3-fold, although these cells differed
3 4 fold in CFE. Three-colour flow cytometry
Epidermal stem cells
50 I
-
i
Loweit 20%
Mid"%
Higheit 20%
Sorted groups
Control
I43
that do not form visible colonies? By following the
fate of slowly adherent cells in a series of identical
dishes stained after 1-14 days of culture, it was
shown that they have two different fates. Some form
colonies of under 32 cells in which all the individual
cells terminally differentiate, and others terminally
differentiate and detach from the dish without dividing. The former behave as predicted for the TA cell
population, and the latter resemble committed cells.
These observations support the model of basal
epidermis shown in Fig. 1 and demonstrate that
viable subpopulations of stem and transit and
committed cells can be isolated from basal epidermis, because they differ in integrin expression and
function.
Fig. 2 lntegrin expression and colony forming ability of keratinocytes in vitm Basal keratinocytes were labelled with a FITC-conjugated
STEM CELLS IN WVO
antibody against the a1 integrin subunit and then sorted by flow cytometry
[20]. The cells were sorted into three groups, the 2G% with highest
fluorescence, the mid 40% of the population and the 20% of cells with the
lowest level of fluorescence, cultured for 14 days and their CFE scored. The
graph shows the mean CFE for each group, with unfractionated basal cells
as a control. Error bars show the SEM.
If the relationship between adhesiveness and proliferative potential that operates in oitro also existed
in v i m , it might allow stem cells to be localized and
isolated directly from the epidermis. One prediction
from the observations in uitro was that basal cells in
the epidermis would differ in their expression of p1
integrins. Published studies of integrin expression in
epidermal cells using indirect immunofluorescence
had not looked for quantitative differences in integrin expression between basal cells in the epidermis
[2 1, 223. Indirect immunofluorescence techniques,
involving a primary antibody and a secondary,
fluorochrome-conjugated antibody, are designed to
maximize the signal from the primary antibody. The
difference in the level of integrin expression between
stem cells and other basal cells in culture is only 23-fold; such small differences in integrin levels may
be lost in the amplification of the primary antibody
signal. Frozen sections of skin were therefore
labelled with anti-integrin antibodies which had
been directly conjugated to fluorescein isothiocyanate (FITC), so that the fluorescence seen was in
direct proportion to the number of antibody molecules bound at the cell surface. Detecting small
differences in fluorescence also required sections of
epidermis of uniform thickness and an accurate
method to quantify fluorescence. Both of these
requirements were met by the use of a confocal
microscope. This instrument can make thin (< 1 pm)
optical sections through tissue, thereby eliminating
errors due to variations in thickness of histological
sections.
When epidermis from the palm, scalp and foreskin was examined it became clear that at all sites
basal cells differed in expression of the a, and a j
integrin subunits. Integrin-bright and integrin-dull
basal cells were seen. The bright cells were not
randomly distributed but clustered together, separated by cells with a lower level of integrin expression (Fig. 3). The bright patches were not randomly
distributed but had a patterned distribution that
varied with body site. In the scalp and foreskin the
bright patches were found to overlie the dermal
demonstrated that the cells which had the highest
level of the a, integrin subunit also had the highest
expression of the other PI associated subunits, a j
and a5. This meant that no further enrichment in
clonogenic cells could be obtained by sorting the
cells on the basis of expression of more than one
integrin subunit. In contrast to the results seen with
the a integrin subunits that partner the p1 subunit,
there is no relationship between integrin expression
and CFE when cells are sorted after staining for the
as subunit that partners the p4 integrin.
An alternative method of sorting cells by integrin
mediated adhesiveness is to plate them on plates
coated in the ECM protein ligands of the integrins
for different lengths of time. The cells that adhered
rapidly or more slowly were removed from the
dishes, cultured and their CFE scored. The cells that
adhered most rapidly to type IV collagen, fibronectin and a preparation of the ECM made by keratinocytes themselves had up to two to five times the
CFE of unfractionated basal cells. The best enrichment for clonogenic cells was obtained with the
keratinocytes that adhered to type IV collagen
within 20 min (colZ0cells),with which CFEs of over
90% were seen, with 90% of the colonies being over
5000 cells in size. Col'Ocells had a higher than
average level of
integrin expression assayed by
fluorescence-activated cell sorting, in keeping with
the cell sorting results and suggesting that they
represented a distinct subpopulation of basal cells
[20]. Additionally, the col'Ocells were able to self
renew and generate the whole range of keratinocytes, including cells that were themselves colZocells
c201.
Col'Ocells have stem cell characteristics, but what
is the fate of the more slowly adherent keratinocytes
144
P. H. Jones
Palm
papillae, whereas in the palm they lie at the tips of
the rete ridges (Fig. 3). In contrast, the a6 subunit
shows no variation in expression between basal
cells. This correlates with the lack of relationship
between a6 expression and CFE in the cell sorting
experiments [23].
The intensity of fluorescence in the cells in bright
patches was approximately 2-fold that of cells in the
integrin-dull areas. Differences of this order reflect
significant differences in CFE in vitro, but to confirm that cells that are clonogenic in vitro reside
within the integrin-bright regions, the properties of
rapidly and slowly adhering keratinocytes isolated
directly from neonatal foreskins were investigated.
As with the cultured keratinocytes, the rapidly
adhering cells had the highest level of integrin
expression as determined by fluorescence-activated
cell sorting, and also had a high CFE. When the
rapidly adherent cells were cultured and grafted
onto nude mice they were able to reconstitute a
complete epidermis with almost normal histology
[23]. The slowly adhering cells had lower integrin
expression and were enriched for committed cells
and cells forming transit-type colonies [23].
To verify the cell kinetic prediction that stem cells
divide infrequently compared with transit cells, the
proportion of cells in S-phase in the integrin-bright
regions compared with the integrin-dull regions was
determined. I n situ hybridization to mRNA for
histones 2B, 3 and 4 was used to identify S-phase
cells [23]. In each site examined, the proportion of
histone-positive cells varied as predicted, with significantly fewer cycling cells in the integrin-bright
regions (Fig. 3).
Microdissection experiments on human hair follicles had established that the region of the hair
follicle in the vicinity of the bulge contained cells
that were able to generate colonies of keratinocytes
in vitro [24, 251. A molecular marker of the bulge
region is keratin 19 [26, 271; using two-colour
confocal microscopy it was shown that keratin 19positive cells in hair follicles have a significantly
higher level of a2 and a3 integrin subunit expression
than the adjoining integrin-negative cells [23].
STEM CELL PAlTERNlNG
Fig. 3. lntegrin expression and cell kinetics in epidermis. (a) Confocal
micrograph of human scalp epidermis stained for the a* integrin subunit
with the same FITC-conjugated antibody used for the cell sorting experk
ment in Fig. 2. The epidermis (e) is uppermost and the dermis (d) below.
All the basal cells stain for integrin, but the intensity of staining varies. A
patch of integrin-bright basal cells can be seen overlying the dermal papilla
(dp), whereas the cells in the rete ridge (rr) are integridull. Scale bar,
15pm. (b) lntegrin expression and cell kinetics in scalp and palm. In the
scalp, patches of integrin-bright basal cells (solid line) are located overlying
each dermal papilla, with integrin-dull cells (hatched line) in the rete ridges;
in the palm this pattern is reversed. The mean percentages of integrib
bright and integrindull basal cells that were positive for histone 2B, 3 and 4
mRNA, indicating that they were in S-phase of the cell cycle, are shown. In
each case the integrin-bright cells were less likely t o be in S-phase. The
differences were highly statistically significant PI.
It has been argued in the haemopoietic literature
that stem cells reside in a special microenvironment,
termed a niche [28]. The patterned distribution of
stem cells in the epidermis may arise from their
residing in a particular microenvironment, located
in the deep rete ridges in palmar skin or overlying
the dermal papillae in thin skin. Alternatively, stem
cell patterning may be established by interactions
set up between keratinocytes themselves.
Confluent stratified sheets of keratinocytes cultured on plastic in the absence of dermis were
examined to see whether or not patches of integrinbright and integrin-dull cells formed in the absence
Epidermal stem cells
I45
of dermis. Sheets of keratinocytes were prepared
from c01~~cells
and unfractionated keratinocytes,
enzymically detached from the plastic and either
sectioned before staining with FITC-conjugated
anti-integrin antibodies or stained as whole mounts.
Labelling with antibodies to the a2 or ag integrin
subunits revealed patches of integrin-bright cells
whether the sheets were generated by co120cells or
by unselected cells (Fig. 4).The variation in fluorescence intensity was similar to that observed in uiuo.
Thus the presence of dermis is not required for
pattern formation [23].
Is it possible to vary the pattern of bright and
dull patches by using different keratinocyte subpopulations to establish the culture? Using selected
col'Ocells or unfractionated cells over a 10-fold
difference in plating density made no difference to
the size of the integrin-bright patches in the resulting sheet. The effect of varying the number of cells
used to found the cultures was then examined in
more detail, using the co120cells. Over a 500-fold
range of seeding density, from 6 cells/cm2to 3000
cells/cm2, the percentage of integrin-bright cells and
the mean patch size remained constant. Patterning
must therefore be autoregulated and established by
keratinocyte-keratinocyte interactions (Fig. 4) [23].
DISCUSSION
Stem cells can be isolated in uitro and directly
from epidermis as they have higher levels of surface
integrin expression and adhere more rapidly than
other basal cells to ECM proteins. These cells are
clonogenic in uitro, can reconstitute an epidermis in
uiuo, exhibit self renewal and are less likely to be in
the cell cycle than other basal keratinocytes in
steady-state epidermis.
The biochemical and cell biological characteristics
of stem cells can now be analysed for the first time.
As well as giving insights into the earliest events in
keratinocyte differentiation this may also improve
our understanding of epidermal disease. For example, it has been argued that the stem cells must be
the cells in the epidermis that undergo neoplastic
transformation [29, 301. However, there is good
evidence that, at least in the case of a viral oncogene, transformation can reverse the normal differentiation process, causing transit-type keratinocytes
to revert to stem cells [31]. The ability to separate
stem and transit populations may help us to discover which epidermal cells are susceptible to neoplastic transformation.
Although integrin expression is a useful first stemcell marker, additional markers are needed as only
one in four cells in an integrin-bright patch in
foreskin is clonogenic in uitro. Also, the use of
integrins to study stem cells in disease is limited by
the fact that integrin expression is greatly perturbed
in wound healing, psoriasis and cancer [32, 331. It
cannot be assumed that the relationship between
Fig. 4. Stem cell patterning in vitra (a) Cryorection of a confluent
keratinocyte culture stained for the a, integrin subunit with an FITC-conjugated antibody and viewed on the confocal microscope. As with epidermis,
the basal cells stain for integrin but a cluster of integrin-bright cells can be
seen. Scale bar, 25pm. (b) Stem cell patterning is mediated by cell-cell
interaction and is autoregulated. Over a 5Wfold range of seeding density of
COP
cells the pattern of bright patches (shaded circles) generated in the
resulting confluent epidermal sheets remains the same. This implies that
cell-cell interaction is involved in patch generation and that the patterning
process is autoregulated. (c) Possible mechanisms of stem cell patterning. In
developing Drasophila and in other species, cells can specify the fate of their
immediate neighbours, using the Notch-Delta signalling system [31] (solid
arrows), or send short-range signals, such as the wingless gene product, to
specify cell fate [35] (shaded arrows).
146
P. H.Jones
integrin expression and proliferative potential that
operates in normal epidermis will be preserved.
Understanding stem cell behaviour in these important disease states also requires the development of
new stem cell markers.
One way forward that may enable the identification of such novel markers and also enrich our
understanding of the behaviour of stem cells within
tissues rather than under clonal growth conditions
in uitro is the observation that stem cell patterning
is mediated by keratinocyte-keratinocyte interaction. This is reminiscent of processes that are known
to occur in development, in particular ‘lateral specification’ [34, 351, in which adjacent cells communicate to ensure that they adopt different fates (Fig. 4).
The study of genes involved in cell fate patterning in
developmental models may be the next step in
understanding human epidermal stem cells.
ACKNOWLEDGMENTS
This work was only possible through the inspiration and encouragement of Dr F. M. Watt and
the other members of the keratinocyte laboratory,
particularly Dr D. Hudson and Dr E. Rytina. I am
also very grateful to Dr S . Harper of Lecicester
General Hospital for carrying out the in situ
hybridizations.
REFERENCES
I. Roberts D, Marks R. The determination of regional and age variations in the
rate of desquamation: a comparison of four techniques. J Invest Dermatol 1980
74: 13-16.
2. Adami JG. The causation of cancerous and other new growths. Br Med J 1901;
i: 621-8.
3. Potten CS, Morris RJ. Epithelial stem cells in vivo. J Cell Sci 1988 10 (Suppl.):
45-52.
4. Potten CS. Cell replacement in the epidermis (keratopoiesis) via discrete units
of proliferation. Int Rev Cytol 1981; 6 9 271-318.
5. Bickenbach ]R. Identification and behaviour of label retaining cells in oral
mucosa and skin. J Dental Res 1981; 60: 162C-2.
6. Lavker RM, Sun T-T. Epidermal stem cells. J Invest Dermatol 1983; 81: 121-7s.
7. Barrandon Y, Green H. Three clonal types of keratinocyte with different
capacities for multiplication. Proc Natl Acad Sci 1987; 84: 2302-6.
8. Compton CC. Current concepts in pediatric burn care - the biology of
cultured epithelial autografts - an 8 year study in pediatric burns patients. Eur
J Pediatr Surg 1992; 2 216-22.
9. Compton CC, Gill JM, Bradford DA, Regauer S, Gallico GG, O’Connor NE.
Skin regenerated from cultured epithelial autografts on full-thickness burn
wounds from 6 days t o 5 years after grafting. A light, electron microscopic
and immunohirtochemical study. Lab Invest 1989; 60: 600-12.
10. Gallico Ill GG, O’Connor NE, Compton CC, Kehinde 0, Green H. Permanent
coverage of large skin wounds with autologous cultured epithelium. N Engl J
Med 1984; 311: 448-51.
II. Watt FM, Green H. Stratification and terminal differentiation of cultured
human cells. Nature (London) 1982; 295 434-6.
12. Hynes RO. Integrins: versatility, modulation, and signalling in cell adhesion.
Cell 1992; 69: 11-25.
13. Watt FM, Jones PH. Expression and function of the keratinocyte integrins.
Development 1993; (Suppl.): 16-92,
14. Adams JC, Watt FM. Changes in keratinocyte adhesion during terminal
differentiation: reduction in fibronectin binding precedes sfi integrin loss
from the cell surface. Cell 1990; 63: 425-35.
15. Hotchin NA, Kovach NL, Watt FM. Functional down regulation of integrins is
reversible but commitment t o terminal differentiation is not. J Cell Sci 1993;
Iw: 1131-8.
16. Hotchin NA, Watt FM. Transcriptional and post-translational regulation of fi
integrin expression during keratinocyte terminal differentiation. J Biol Chem
1992; 267: 14852-8.
17. Rheinwald RG, Green H. Serial cultivation of strains of human epidermal
keratinocytes: the formation of keratinizing colonies from single cells. Cell
1975; 6 331-43.
18. Rheinwald R. Methods for clonal growth and serial cultivation of normal
human epidermal keratinocytes and mesothelial cells. In: R Baserga, ed. Cell
on. A practical approach. 1st edn. Oxford: IRL Press, 1989
8 1-94,
19. Adams JC, Watt FM. Expression of fi, b, fi and fi integrins by human
epidermal keratinocytes and nondifferentiating keratinocytes. J Cell Biol 1991;
1 1 5 829-41.
20. Jones PH, Watt FM. Separation of human epidermal stem cells from transit
amplifying cells on the basis of integrin function and expression. Cell 1993; 73:
7 I3-24.
21. De Luca M, Tamura RN, Kajiji S, et al. Polarised integrin mediates human
keratinocyte adhesion t o basal lamina. Proc Natl Acad Sci U.S.A. 1990; 87:
6888-92.
22. Hertle MD, Adams JC, Watt FM. lntegrin expression during human epidermal
development in vivo and in vitro. Development 1991; 112: 193-206.
23. Jones PH, Harper S, Watt FM. Stem cell fate and patterning in human
epidermis. Cell 1995; 80. 83-93.
24. Yang YS, Lavker RM, Sun T-T. Upper human hair follicle contains a
subpopulation of keratinocytes with superior in vitro proliferative potential. J
Invest Dermatol 1993; 101: 652-9.
25. Rochat A, Kobayashio K, Barrandon Y. Location of stem cells of human hair
follicles by clonal analysis. Cell 1994; 76 1063-73.
26. Stasiak PC, Purkis PE, Leigh IM, Lane EB. Keratin 19 predicted amino acid
sequence and broad tissue distribution suggests it evolved from keratinocyte
keratins. J Invest Dermatol 1989 91: 707-16.
27. Lane EB, Wilson CA, Hughes BR, Leigh IM. Stem cells in hair follicles.
s
Cytoskeletal studies. Annals NY Acad Sci 1991; 642: 197-213.
28. Trentin SL, Curry JL, Wolf N, Cheng V. Factors controlling stem cell
differentiation and proliferation: the haematopoietic microenvironment. In: The
proliferation and spread of neoplastic cells. Baltimore: Williams and Wilkins,
1968: 713-31.
29. Pierce GB, Speers WC. Tumours as caricatures of the process of tissue
renewal: prospects for therapy by directing differentiation. Cancer Res 1988;
48: 1996-2004.
30. Cairns I. Mutational selection and the natural history of cancer. Nature
(London) 1975; 255 197-200.
31. Barrandon Y, Morgan JR, Mulligan RC, Green H. Restoration of growth
potential in paraclones of human keratinocytes by a viral oncogene. Proc Natl
Acad Sci U.S.A. 1989; 86: 4102-6.
32. Hertle MD, Kubler MD. Leigh IM, Watt FM. Aberrant integrin expression
during epidermal wound healing and in psoriatic epidermis. J Clin Invest 1992;
89: 1892-901.
33. Jones J, Sugiyama M, Watt FM, Speight PM. lntegrin expression in normal,
hyperplastic, dysplastic, and malignant oral epithelium. J Pathol 1993; 169
23543.
34. Artavanis-Tsakonas S, Matsuno K. Fortini ME. Notch signalling. Science
(Washington DC) 1995; 268: 225-32.
35. Peifer M, Bejsovec A. Knowing your neighbours: cell interactions determine
intrasegmental patterning in Drosophilo. Trends Genet 1992; 8: 243-9.