Oncogene (1999) 18, 397 ± 406
ã 1999 Stockton Press All rights reserved 0950 ± 9232/99 $12.00
http://www.stockton-press.co.uk/onc
Coordinated changes in cell cycle machinery occur during keratinocyte
terminal di€erentiation
Luis A Martinez, Yian Chen, Susan M Fischer and Claudio J Conti
The University of Texas MD Anderson Cancer Center, Science Park Research Division, Smithville, Texas, USA
Di€erentiation of cells is typically marked by a cessation
of proliferation with a concurrent entrance into a distinct
metabolic state marked by tissue speci®c gene expression. The mechanism by which the cell exits the cell cycle
in this process is poorly understood. To determine the
potential roles of the cell cycle machinery in the
regulation of the terminal di€erentiation process of
epidermal cells, we selected a well characterized in vitro
model in which primary mouse keratinocytes are induced
to di€erentiate in response to a raised calcium ion
concentration in the medium. The withdrawal from the
cell cycle correlates very well with a number of changes
in the cell cycle machinery. Changes in the phosphorylation status of the Rb family of proteins occurs
coordinately with an increased association of p21, p27
and p57 with cdk2. Furthermore, we ®nd that inhibition
of cdk2 activity is not sucient to elicit changes that
occur during keratinocyte di€erentiation. Finally, the
previously described v-Ha-ras block of keratinocyte
di€erentiation correlates with altered regulation of both
cyclin D1 and cdk2 suggesting that these genes may play
a role in the Ha-ras transformation of a keratinocyte.
Keywords: cell cycle; cyclins; CDK; epidermis, terminal
di€erentiation; keratinocytes
Introduction
The cyclins are proteins that in general exhibit cell
cycle dependent expression and form complexes with
their catalytic partners, the cyclin dependent kinases
(cdks), to promote the transition of the cell through
distinct restriction points of the cell cycle. The
coordinated regulation of the cyclin/cdk holoenzyme's
activity occurs at various levels including periodic
synthesis of the cyclins, post-translational modifications of the cdks, and subunit composition of these
complexes (reviewed by Morgan, 1995). The cyclins
that have been shown to promote the organization of
events in the rate limiting phase of the cell cycle, G1,
are the three D-type cyclins and cyclin E (Ewen et al.,
1993; Ko€ et al., 1992; Lew et al., 1991; Matsushime et
al., 1991; Xiong et al., 1991). Regulation of cell cycle
progression in somatic cells appears to involve the
integration of extracellular signals into a program
culminating in the modulation of the activity of the Rb
family of proteins (reviewed by Sherr, 1994).
Correspondence: CJ Conti
Received 7 June 1998; revised 20 July 1998; accepted 20 July 1998
The catalytic activity of the cyclin/cdk holoenzyme
can be regulated by a number of cdk inhibitors (CKI)
such as p21, p27, p57 and the members of the Ink4
family (reviewed by Sherr and Roberts, 1995). The
expression of each cdk inhibitor appears to be
regulated by various mechanisms speci®c for each
inhibitor (reviewed by Sherr and Roberts, 1995). The
cdk inhibitor p21 was initially identi®ed because of its
induction in response to various growth arrest stimuli
through p53 dependent and independent mechanisms
(reviewed by Sherr and Roberts, 1995). Recently an
increase in p21 expression has been described during
di€erentiation of various cell lineages including
keratinocytes, indicating that this gene might be an
important component of the di€erentiation process
(Deng et al., 1995; Halevy, et al., 1995; Jiang et al.,
1994; Missero et al., 1995; SteinMan et al., 1994;
Weinberg et al., 1995).
The balance between proliferation, di€erentiation
and programmed cell death is a ®nely tuned process in
the epidermis. The basal layer of the epidermis is made
up of proliferative cells. The suprabasal layers have
exited the cell cycle and have begun the process of
terminal di€erentiation. The role of the cell cycle
regulatory machinery during this process is not well
understood. Studies on this process in other cell types
have revealed that cyclin expression does not always
correlate with proliferation. In myoblasts induced to
di€erentiate, an increase in cyclin D3 protein is
observed in cells that are no longer dividing (Kiess et
al., 1995). The increased expression of the cyclins
during and after these cells have di€erentiated suggests
that these cyclins might play a role in either the
initiation or maintenance of the di€erentiated state.
Previous work from our lab has shown that cyclin
D1 is overexpressed in early neoplastic lesions resulting
from the two-stage mouse skin carcinogenesis protocol
(Bianchi et al., 1993; Robles and Conti, 1995). This
prompted us to study the potential involvement of the
deregulation of cell cycle genes during the development
of tumors. To study the molecular mechanisms of cell
cycle withdrawal and di€erentiation of epidermal cells
we selected a well characterized in vitro model in which
primary mouse keratinocytes are induced to differentiate by raising the extracellular calcium ion concentration (Hennings et al., 1980). Primary mouse
keratinocytes can be maintained in culture with
phenotypic properties similar to those of the basal
cells of the epidermis. Once the calcium ion concentration is adjusted to 1.2 mM, these cells become
irreversibly committed to terminal di€erentiation and
exhibit the phenotypic properties of the suprabasal
layers of the epidermis (Hennings et al., 1980). Within
24 h after the Ca2+ switch, there is a dramatic
reduction in DNA synthesis. Previous work on this
Cell cycle regulation in terminal differentiation
LA Martinez et al
398
model has implicated a role for the cdk inhibitor p21
during di€erentiation (Missero et al., 1995). Although
p21 expression has been associated with di€erentiation
of various cell types, p21 nullizygous mice appear to
develop normally, suggesting either redundant or
overlapping roles in CKI functions (Deng et al.,
1995). To gain an understanding of the leading events
of the di€erentiation process, we have systematically
analysed the changes that occur in expression,
interactions and associated kinase activities of the cell
cycle regulatory machinery and compared it to the
timing of withdrawal from the cell cycle. In addition,
we ®nd that expression of an activated v-Ha-ras in
primary keratinocytes induces multiple changes in the
cell cycle machinery which may play a role in the v-Haras-mediated block of the di€erentiation process.
Results
Induction of di€erentiation
The expression of two epidermal di€erentiation
markers and changes in proliferative activity were
examined in keratinocyte cultures induced to differentiate by raising the calcium to 1.2 mM. The purpose
of these experiments was to generate a time frame for
cell cycle withdrawal and expression of key differentiation genes under our experimental conditions. The
transcription factor mad-1 which is expressed in
di€erentiating keratinocytes (Chin et al., 1995; Hurlin
et al., 1995; Vastrik et al., 1995) and keratin 1 (K1)
which is primarily expressed in the suprabasal
di€erentiating cells of the epidermis were upregulated
in cultured keratinocytes in response to the Ca2+ switch
indicating that the cells had entered the di€erentiation
process (Figure 1). To determine the proliferative
activity of the keratinocytes, DNA synthesis was
assessed by measuring the level of incorporation of
[3H]thymidine. Within 24 h after the Ca2+ switch, there
was a dramatic reduction in incorporated label to
approximately 10% of the control (Figure 2). These
changes are consistent with previously published results
on this model (Hennings et al., 1980).
Figure 1 Expression of di€erentiation markers after Ca2+
switch: (a) Total RNA isolated at di€erent time points after the
Ca2+ concentration was adjusted to 1.2 mM, was used for
Northern blot analysis to determine mad-1 expression; (b)
Western blot analysis of keratin 1 protein levels
Expression of cell cycle machinery during di€erentiation
Exit from the cell cycle would presumably be re¯ected
by changes in expression or activity of the cell cycle
regulating genes. Analysis of the G1 cyclins indicated
distinct mechanisms of regulation. At the mRNA level,
the three D-type cyclins underwent signi®cant downregulation within the ®rst 6 h, with no appreciable
change thereafter (data not shown). Fifteen hours after
the Ca2+ switch, cyclin D1 protein levels were reduced
approximately 70% while cyclin D2 protein was
undetectable at this time (Figure 3). Cyclin D3 protein
exhibited a 60% reduction in its level by 15 h, although
it remained detectable after the cells had ceased cycling
(Figure 3). Two isoforms of cyclin E were detectable in
the keratinocytes (Figure 3). The faster migrating form
decreased along with DNA synthesis whereas the
slower migrating form was essentially invariable. Two
di€erent mRNA transcripts have been isolated from
both human and mouse cells that can give rise to cyclin
E isoforms (Botz et al., 1996; Ohtsubo et al., 1995).
Figure 2 DNA synthesis during keratinocyte di€erentiation.
Primary mouse keratinocytes were grown for 3 days as described
in Materials and methods. To induce di€erentiation, the medium
Ca2+ concentration was adjusted to 1.2 mM and DNA synthesis
was monitored by measuring the level of [3H]-thymidine
incorporation
The appearance and disappearance of a smaller form
of cyclin E has also been observed during the transition
from G1 to S in other studies (Muller et al., 1994).
It has been proposed that inhibition of cdk activity
is a prerequisite for exit from the cell cycle and
entrance into the di€erentiation pathway of various cell
types. To determine the regulation of the cdks during
di€erentiation we ®rst assessed their protein levels.
Ca2+-induced di€erentiation results in downregulation
of both cdk2 isoforms whereas the protein levels of
Cell cycle regulation in terminal differentiation
LA Martinez et al
399
Figure 4 Western blot analysis of the CIP and Rb families.
Equal amounts of protein taken at the indicated times after the
Ca2+ switch were analysed by SDS ± PAGE and Western blot
Figure 3 Western blot analysis of cell cycle machinery during
terminal di€erentiation. Equal amounts of protein taken at the
indicated times after the Ca2+ switch were analysed by SDS ±
PAGE and Western blot with the indicated antibodies
cdk4 and cdk6 remain unchanged (Figure 3). Reduction of the steady-state level of cdk2 lags behind the
drop in DNA synthesis.
Regulation of cdk activity occurs at various levels
and involves multiple interactions with regulatory
proteins (reviewed by Sherr and Roberts, 1995). We
observed marked di€erences in the expression of the
di€erent members of the CIP family of cdk inhibitors.
Both p21 and p57Kip2 were expressed at low levels in
cycling keratinocytes but were clearly induced in
response to the Ca2+ switch. In contrast, p27Kip1, a
CIP family member, underwent a reduction in the
protein steady-state level (Figure 4 and data not
shown).
Various studies on di€erentiation have indicated that
pRb plays an essential role during this process
(reviewed by Kranenburg et al., 1995). In contrast to
those studies in which pRb levels become elevated in
response to di€erentiation stimuli, the steady-state
levels of pRb did not increase during keratinocyte
di€erentiation. In actively proliferating cells, pRb was
detected as two bands, presumably the hyper- and
hypo-phosphorylated forms of the protein (Figure 4).
Within 6 h after the Ca2+ switch the level of Rb
protein decreased and there was an apparent prevalence of the hypophosphorylated (active) form. Also,
we observed that the Rb family member, p107,
underwent similar changes in its phosphorylation
status along with a decrease in steady-state protein
levels. In contrast, p130 was upregulated during
keratinocyte di€erentiation suggesting that this member of the Rb family plays a major role in the
regulation of this process.
Functional interactions between cyclins, Cdks and CKIs
Although the observed changes in expression of the cell
cycle machinery provided a clue as to the potential
regulatory mechanisms involved in di€erentiation, this
data did not de®nitively implicate an important
target(s) for cell cycle withdrawal. Therefore, we
extended our analysis to the interactions of the cell
cycle machinery. Immunoblot analysis of cdk associated proteins revealed di€erential binding of the
cyclins to the cdks. Cyclin D1 associated primarily with
cdk4 and to a much lesser extent with cdk6 and cdk2
(Figure 5). Changes in the amount of cyclin D1 bound
to the cdks became evident within the ®rst 4 h of the
Ca2+ switch with decreased association occurring by
24 h. The initial changes in the binding of this cyclin to
the cdks may be in part due to its reduced synthesis.
Cyclin D2 was only detectable in cdk4 immunoprecipitates and exhibited decreased association with this
cdk after the Ca2+ switch (Figure 5). Cyclin D3 was
also found associated with cdk2, cdk4 and cdk6
(Figure 5). At 24 h after the Ca2+ switch, cyclin D3
is present at high levels in cdk4 immunoprecipitates.
The only apparent changes in association of cyclin D3
with the cdks is in its reduced presence in cdk2
immunoprecipitates which might be a result of the
decrease in the steady-state level of both proteins.
Both p27 and p21 appear to be bound to the cdks
but possibly to di€erent extents (Figure 5). Within 4 h
of the Ca2+ switch, the levels of p21 and p27 associated
with all three of the cdks increased. At later time
points, both p21 and p27 were associated at higher
levels with cdk4 than with cdk2 and cdk6. We further
explored whether the induction of p57 would be
re¯ected in changes in its association with the different
cdks. We observed a temporal increase in the
association of p57 with cdk2 and cdk4, 6 (Figure 6
Cell cycle regulation in terminal differentiation
LA Martinez et al
400
Figure 5 Analysis of cdk-associated proteins during keratinocyte di€erentiation. Equal amounts of lysate were immunoprecipitated
with antibodies against either cdk2, cdk4 or cdk6. Co-immunoprecipitated proteins were visualized by Western blot analysis with
the antibodies indicated on the left of the ®gure
and data not shown). Therefore all three of the CIP
family of cdk inhibitors respond to the Ca2+ trigger
and associate to a higher degree with all three of the
cdks.
Various studies have suggested that the cdk
inhibitors p21 and p27 can bind to cyclins in a
manner that is independent of the presence of a cdk
(Hall et al., 1995; Lin et al., 1996). If increased
association of these two inhibitors with the cyclins was
a mechanism of functionally sequestering the cyclins,
one might predict that changes in the level of the
cyclins associated with p21 and p27 might arise during
di€erentiation. To examine this possibility we
immunoprecipitated these two inhibitors at di€erent
times after the Ca2+ switch and determined the level
of associated D-type cyclins by Western blot (Figure
6). All three of the D-type cyclins are detectable in
p21 and p27 immunoprecipitates. A slight increase in
the levels of the D-type cyclins bound to these
inhibitors occurred within 4 h of the Ca2+ switch,
although reduced coimmunoprecipitation was apparent at 12 h. This data suggests that functional
sequestration of the cyclins does not appear to be a
major mechanism for the downregulation of the
cyclin-cdk complexes. Of the D-type cyclins, we
could only detect cyclin D1 co-immunoprecipitating
with p57, in a similar pattern to that observed with
p21 and p27 (data not shown).
Western blot analysis of p21, p27 and cyclin E
immunocomplexes revealed that all three of these
proteins bound both of the cdk2 isoforms. Cdk2 is
found in cyclin E immunocomplexes up to 24 h after
the Ca2+ switch (Figure 6).
Since we observed an increase in the association of
CKIs with the cdks within 4 h, one would predict that
the stoichiometric changes in the CKIs bound to the
complexes might result in the modulation of cyclin-cdk
associated kinase activities. Cdk2-associated kinase
activity was measured in an in vitro kinase assay
using either an H1 or an Rb substrate. Within 8 h of
the Ca2+ switch, there was a 90% reduction in cdk2
kinase activity towards an H1 substrate and an 80%
reduction in cdk2 kinase activity towards an Rb
substrate (Figure 7). Within 4 h of Ca2+ treatment,
the level of cyclin E-associated kinase activity towards
H1 substrate was reduced by approximately 90%.
In contrast the kinase activities of cdk4 and cdk6 did
not exhibit any appreciable changes over the 24 h
analysed (Figure 7 and data not shown).
By using densitometric analysis of the major changes
in cyclin-cdk-CKI interactions and activities, we
plotted this data to obtain a correlation between these
Cell cycle regulation in terminal differentiation
LA Martinez et al
401
Figure 6 Changes in p21, p27, cyclin E and cdk2 complexes during di€erentiation. After immunoprecipitation with the antibodies
indicated on the top of the ®gure, co-immunoprecipitating proteins were visualized with the antibodies indicated on the left of the
®gure
changes and cell cycle withdrawal. Figure 8 shows that
there is an increase in the association of both p21 and
p27 with cdk2 within 4 h of the Ca2+ switch. Analysis
of cyclin E-associated kinase activity reveals immediate
downregulation within 4 h of the di€erentiation
stimuli. Similarly, cdk2-associated kinase activities are
downregulated within 8 h. Since cdk2 can form
complexes with various partners (D-type cyclins,
cyclin E and cyclin A), the cdk2 kinase activity that
we measured in vitro is likely to re¯ect the combined
activities of the various cyclin-cdk2 combinations.
Hence, the reduction in cdk2 associated kinase
activity, regardless of the cyclin that is bound to it, is
a critical point in the mechanism of cell cycle
withdrawal that accompanies terminal di€erentiation.
To assess whether the inhibition of cdk2 kinase
activity could trigger entrance into the di€erentiation
process, we used a very potent and speci®c inhibitor of
cdk2, roscovitine (Meijer et al., 1997). This inhibitor
can suppress the growth of both primary keratinocytes
and established cell lines derived from tumors (data not
shown). Roscovitine suppresses DNA synthesis in
primary keratinocytes more rapidly than the calcium
treatment. Within 2 h after treatment with roscovitine,
there is a 25% reduction in DNA synthesis and by 4 h
of treatment there is a 40% reduction (data not
shown). In comparison, high calcium treatment of
keratinocytes reduces DNA synthesis by 25% at 4 h
(Figure 2). However, at a concentration of roscovitine
(30 mM) which potently suppressed cell growth, we did
not observe the changes that Ca2+ induced (Figure 9).
Treatment of the cells for up to 8 h did not induce
morphological di€erentiation, expression of differentiation markers, nor di€erentiation-associated changes in
expression of the cell cycle machinery (Figure 9). This
leads us to conclude that the inhibition of cdk2 activity
is an early event during keratinocyte di€erentiation
which appears to be mechanistically involved in cell
cycle withdrawal during the di€erentiation process.
However, inhibition of cdk2 is not sucient for the
induction of di€erentiation markers in these cells other
than the inhibition of proliferation.
Previous work from Dr Stuart Yuspa's laboratory
has shown that the expression of an activated v-Ha-ras
in primary mouse keratinocytes can alter the differentiation program induced by Ca2+ (Yuspa et al.,
1985). Speci®cally, these studies showed that although
the v-Ha-ras expressing keratinocytes undergo
Cell cycle regulation in terminal differentiation
LA Martinez et al
402
Figure 8 Correlation of changes in Cdk associated CIP proteins
and cell cycle withdrawal. Densitometric analysis of the
association and kinase activities of the cell cycle machinery. The
quantitation of the levels of p21 and p27 co-immunoprecipitating
with cdk2 were performed with a BioImager densitometer.
Densitometry analysis of cdk2 and cyclin E associated kinase
activities are presented as percentage of control (0 h). The
readings were superimposed on DNA synthesis results from
Figure 2 to show the timing of changes that occur during
keratinocyte di€erentiation. This data is from one experiment
which was repeated at least twice yielding virtually identical
results
Figure 7 Analysis of associated kinase activities during
di€erentiation. Immunoprecipitations were performed on equal
amounts of lysate taken at the indicated time after the Ca2+
switch. Associated kinase activities were then assayed in an in
vitro kinase reaction with either (a) Histone 1 substrate or (b)
GST-Rb. The antibodies used were all rabbit polyclonal
purchased from Santa Cruz
morphological di€erentiation and cell cycle withdrawal,
they do not express certain di€erentiation markers.
Furthermore, they can revert to the basal cell
phenotype if the Ca2+ concentration in the medium is
reduced to 0.05 mM, suggesting that they are blocked
at an early and reversible stage of di€erentiation. This
prompted us to determine if Ca2+ causes any
alterations in the regulation of the cell cycle machinery
in v-Ha-ras expressing keratinocytes. We infected
primary keratinocytes with a retrovirus carrying an
activated v-Ha-ras and analysed their regulation of the
cell cycle machinery in response to Ca2+. We found
that two characteristic changes that occur during
keratinocyte di€erentiation were signi®cantly altered
(Figure 10). Cyclin D1 and cyclin D2 steady-state levels
were slightly elevated in the v-Ras expressing cells and
surprisingly they were not suppressed during differentiation. Analysis of cdk2 kinase activity also revealed
an altered response to Ca2+. Whereas in normal cells
Figure 9 Selective inhibition of cdk2 activity does not induce
di€erentiation. Primary keratinocyte cultures were treated with
either DMSO, calcium (1.2 mM) or the cdk2 inhibitor, roscovitine
(30 mM ®nal concentration) for the indicated times (0, 4, and 8 h).
Lysates were then analysed by Western blot with polyclonal cyclin
D1 and keratin 1 antibodies
Figure 10 The v-Ha-ras-induced block in keratinocyte differentiation correlates with altered regulation of cyclin D1 and cdk2.
Primary mouse keratinocyte cultures were infected as described in
Materials and methods. The medium was changed 24 h prior to
treatment with Ca2+. (a) Cyclin D1 Western blot analysis and (b)
cdk2 kinase activity were assayed at the indicated times after
treatment
Cell cycle regulation in terminal differentiation
LA Martinez et al
cdk2 kinase activity was completely suppressed within
24 h, in v-Ha-ras expressing cells cdk2 activity
persisted up to that point. We are presently addressing
whether these alterations in the cell cycle machinery are
required for the v-Ha-ras mediated di€erentiation
block.
Discussion
We have systematically analysed the cell cycle
regulatory machinery to gain an understanding of the
changes that might drive permanent cell cycle withdrawal during di€erentiation of mouse keratinoctyes in
vitro. We show that as the keratinocytes underwent this
process, there was an early reduction in the protein
levels of the D-type cyclins, cyclin E and cdk2 and
associated kinase activities.
Phosphorylation of pRb has been proposed as the
mechanism of functional inactivation of this protein to
allow cell cycle progression (reviewed by Weinberg,
1995). Also, pRb appears to have a critical role in the
di€erentiation of various cell types. Rb nullizygous
mice die in utero and exhibit speci®c defects in the
di€erentiation process of certain cell lineages (Clarke
et al., 1992; Jacks et al., 1992; Lee et al., 1992;
Schneider et al., 1994). In our study as the
asynchronous population of cells began to differentiate, the hypophosphorylated form of the protein
became predominant. Changes in the associated
kinase activities of cdk2 correlate well with the
decreased phosphorylation of pRb and the decrease
in DNA synthesis. Of particular interest to us is the
observation that pRb steady-state levels also decrease
during di€erentiation. At this time we do not know
whether the decreasing levels of pRb also results in
decreased pRb function therefore indicating that Rb
activity is not required for keratinocyte di€erentiation.
There is a possibility that the level of Rb present in the
di€erentiating cells does not re¯ect its functional
activity as has been shown for p53 during keratinocyte di€erentiation (Weinberg et al., 1995). Furthermore, we observed di€erences in the expression
patterns of the other Rb family members. p107 clearly
followed the same pattern as Rb whereas p130 was
induced during di€erentiation. These di€erences
suggest very distinct roles for the Rb family members
in the regulation of growth and di€erentiation of
keratinocytes. Recently with the generation of knockout mice lacking one or more of the di€erent members
of the Rb family, it has become apparent that these
genes may play redundant roles during development,
but their loss can also have a very de®ned consequence
in certain tissues (Cobrinik et al., 1996; Hurford et al.,
1997). This is the ®rst report on keratinocyte
di€erentiation chronicling the potentially distinct roles
of the Rb family in this process. It would be of interest
to determine whether mice nullizygous for the di€erent
Rb family members display a phenotype in the
epidermis. Furthermore, if indeed p130 plays a role
in keratinocyte di€erentiation, then this might be a
gene whose function is altered during mouse two-stage
carcinogenesis. Future studies will need to address
whether changes in the Rb family members are also
re¯ected in the functions of di€erent interacting
proteins such as the E2F family.
Of the various changes in expression and activities of
the cell cycle regulatory proteins, some of the most
striking are the changes in the cdk inhibitors. Although
only p21 and p57 are induced, all three of the cdk
inhibitors interacted with the cdks to a higher degree in
response to the Ca2+ stimulus. The observed
`mobilization' of the three inhibitors by Ca2+ suggests
that there may be a common Ca2+-regulated pathway
coordinating their interaction with the cdks.
Given the observed increase in expression of both
p21 and p57, and the continued expression of these
inhibitors well after cell cycle withdrawal, it would be
of interest to determine if they might play a role in
regulating other aspects of di€erentiation. Recently,
p21 has been shown to associate and modulate the
activity of the stress-activated protein kinases (SAPKs)
and protein kinase CK2 (Gotz et al., 1996; Shim et al.,
1996). Furthermore, during the preparation of this
manuscript two groups published reports suggesting a
potential role for p21 and p27 in regulating the
expression of di€erentiation markers (Missero et al.,
1996; Hauser et al., 1997). These studies showed that
the absence of one of the inhibitors resulted in an
altered expression of di€erentiation markers, yet the
cells still suppressed cdk activity and withdrew from
the cell cycle. Our data suggests that all three of the
CIP family of cdk inhibitors may be involved in
coordinating cell cycle withdrawal with di€erentiation.
Therefore, there is likely to be sucient functional
redundancy between the CKIs that allows cell cycle
withdrawal in the absence of one of the inhibitors.
p27 protein levels appear to be regulated by various
stimuli including growth inhibitory signals (reviewed by
Sherr and Roberts, 1995) and di€erentiation in postmitotic neurons (Lee et al., 1996). In this study, p27
protein is detectable as a doublet in cycling cells and
then changes to one band after the Ca2+ switch
suggesting alterations in covalent modi®cation. Also,
the steady-state protein level of p27 is reduced after the
keratinocytes are induced to di€erentiate. Both p21
and p27 appear to be able to block cdk activity in part
by inhibiting activation by phosphorylation of the
cyclin-cdk holoenzyme by the cdk-activating kinase
(Kato et al., 1994a,b). Increased binding of p27 to the
cyclin-cdk holoenzyme would then lead to reduced
levels of active holoenzymes and subsequently exit
from the cell cycle. Within 4 h of the Ca2+ switch there
is an increase in cdk bound p21 and p27. Worth noting
is that in p27 immunocomplexes, both phosphorylated
and unphosphorylated cdk2 are present. It has been
reported that p21 and p27 can bind cyclins independently of the presence of a cdk but the binding of the
cdks by these two inhibitors appears to require the
presence of a cyclin (Hall et al., 1995; Lin et al., 1996).
Therefore the initial increase of these two CKIs in the
cdk immunoprecipitations might mediate the inhibition
of the cyclin-cdk holoenzymes and subsequently the
drop in the levels of the cyclins results in fewer
holoenzymes and therefore less CKIs bound to the
cdks.
We could not detect the expression of either p15 or
p16 by Western blot (data not shown). Given that cdk4
and cdk6 did not show any appreciable changes in
their associated kinase activities during di€erentiation,
one might predict that neither p15 nor p16 play a role
in this process.
403
Cell cycle regulation in terminal differentiation
LA Martinez et al
404
Previous studies on the di€erentiation of other cell
types have shown a requirement for the downregulation of either both cdk4 and cdk2 or only cdk4 (Lee et
al., 1996; Lukas et al., 1995; Muller et al., 1994). In
contrast to cdk2, we do not see any reduction in the
expression of cdk4 and cdk6 protein nor do we see
appreciable changes in their associated kinase activities,
suggesting a critical role for inhibition of cdk2associated kinase activity during keratinocyte differentiation. Although in this study cdk2-associated
kinase activities towards di€erent substrates di€er in
the timing of their attenuation, by 8 h after the Ca2+
switch, there is approximately an 80% reduction in
both. Cyclin E associated kinase activity is downregulated at an earlier time point suggesting that
inhibition of cyclin E-cdk2 activity is the critical step
in the regulation of the cell cycle machinery during
keratinocyte di€erentiation. Since there was no
apparent correlation between the attenuation of cyclin
E-cdk2-associated kinase activities and the levels of this
holoenzyme (Figure 6), this suggests that downregulation of the kinase activity of this holoenzyme is
occurring at another level. The increased association
of p21, p27 and p57 to cdk2-containing complexes is
likely to mediate the inhibition of the associated kinase
activities of these complexes resulting in cell cycle
withdrawal.
Of particular interest to us is the ability of an
activated Ha-ras to block certain changes in the cell
cycle machinery during keratinocyte di€erentiation. As
mentioned above topical treatment of mouse epidermis
with the chemical carcinogen, dimethylbenzanthracene
(DMBA), results in the activation of the Ha-ras gene.
Furthermore, cyclin D1 overexpression occurs in
virtually all lesions resulting from this treatment. Our
group and others have previously shown that an
activated Ha-ras can induce cyclin D1 expression in
di€erent cells (Albanese et al., 1995; Filmus et al.,
1994; Liu et al., 1995; Robles and Conti, manuscript in
prep.). In this study we consistently observed higher
levels of cyclin D1 and cyclin D2 in v-Ha-ras carrying
cells (especially under starvation conditions). Our data
provide a potential link between an activated Ha-ras
and the observed deregulation of cyclin D1 in mouse
skin tumors. The inability of keratinocytes carrying an
activated Ha-ras to suppress cdk2 activity also implies
that Ha-ras can a€ect multiple targets in the cell cycle
machinery. We have found signi®cant alterations in the
expression of genes that regulate both cyclin D1 and
cdk2, and we are addressing their requirement for the
Ha-ras block of di€erentiation (Martinez and Conti,
manuscript in preparation).
It is apparent from this study that highly
coordinated mechanisms are involved in the regulation
of keratinocyte di€erentiation. The concurrent changes
in the expression patterns of the cell cycle regulatory
proteins might indicate successive changes in the
activities of these proteins. For instance, upregulation
of CKI activity might lead to inactivation of the cyclincdk complexes subsequently allowing the differentiation process to occur. Alternatively, changes in the
expression of genes required for the maintenance of the
proliferative state might be downregulated resulting in
a cascade of changes that ultimately converges on the
inactivation of the cyclin-cdk holoenzymes and cell
cycle withdrawal. At present, the inactivation of cyclin/
cdk complexes appears to be a result of the
di€erentiation process rather than the driving event in
keratinocyte di€erentiation. Further dissection of the
upstream cell cycle regulatory components is required
to determine the link between the induction of
di€erentiation and cell cycle withdrawal.
Materials and methods
Cell culture
Primary mouse keratinocytes were cultured from epidermis
of 1 ± 2 day old newborn SSIN mice. Brie¯y, the epidermis
was incubated with trypsin overnight at 48C. After removal
of the dermis, keratinocytes were isolated and plated in
EMEM (Sigma) with high Ca2+ (1.2 mM) for approximately 3 h to allow the cells to attach. The cultures were
then rinsed with warm PBS and maintained in low Ca2+
medium (EMEM) supplemented with 8% chelex treated
serum, EGF (5 ng/ml), insulin, phosphoethanolamine,
transferrin and hydrocortisone. The media was changed
the second day after plating and the cells were used on the
third day. Cultures produced as described above were
con¯uent by the third day. The keratinocytes were induced
to di€erentiate by increasing the medium Ca2+ concentration to 1.2 mM.
For the cdk2 inhibition experiments, roscovitine (30 mM
®nal concentration) was dissolved in DMSO and added to
the media directly. Control cultures received DMSO at a ®nal
concentration (0.06%) treatment only.
Keratinocytes were infected with a retrovirus carrying an
activated Ha-ras as has been previously described (Yuspa et
al., 1985). Brie¯y, on the third day after plating, cells were
washed with warm PBS and then medium harvested from the
retrovirus packaging cells was applied to the cultures in the
presence of 4 mg/ml polybrene. The retroviral infection was
allowed to proceed for 2 h after which fresh medium was
added. These cultures were then used 3 days after the
infection.
Analysis of DNA synthesis
[3H]thymidine (5 uCi/ml) was added to the keratinocyte
cultures 1 h before harvest. Cells were ®rst washed with
ice-cold PBS and then 10% TCA. After lysis in 0.2 N
NaOH, the cell lysates were then analysed for label
incorporation.
RNA analysis
Cells harvested at di€erent time points were lysed in TRIreagent (Molecular Research Center, Ohio, USA) and
RNA was prepared as described by the manufacturer.
Northern blot analysis was performed by conventional
methods according to Sambrook et al., (1989). D-type
cyclin probes were a generous gift from Dr Charles Sherr,
St Jude Children's Hospital.
Protein analysis
Protein was extracted from the cell cultures at di€erent
times after the Ca2+ switch. Cells were scraped on ice into
ice-cold NP40 bu€er (50 mM Tris, 0.5% NP-40, 150 mM
NaCl, 5 mM EDTA) containing protease inhibitors PMSF
(1 mM), DTT (1 mM) and aprotinin (20 u/ml). Lysates were
then sonicated and clari®ed by centrifugation. Protein
concentration was determined using the Bio-Rad Dc
protein assay kit. Approximately 100 mg of protein were
subjected to SDS ± PAGE (in 8 ± 12% gels), after which the
Cell cycle regulation in terminal differentiation
LA Martinez et al
protein was then transferred to nitrocellulose membranes.
The blots were blocked with 5% milk in TBS. All
antibodies used for Western blot analysis were from Santa
Cruz and Neomarkers (D-type cyclins). For Western blot
analysis, all primary antibodies were used at a 1 : 1000
dilution, 1 : 2000 secondary HRP conjugated antibodies
and visualized with the ECL detection kit (Amersham).
For immunoprecipitations, cells were lysed in NP-40 bu€er
as described above. Equal amounts of protein extracts were
then reacted with antibody coupled to protein G agarose
(GIBCO) overnight and washed ®ve times in lysis bu€er. The
samples were then boiled in loading bu€er and analysed as
described.
Conditions for kinase assays were as described in
Matsushime et al. (1994) except that the Rb substrate was
purchased from Santa Cruz. Kinase reactions were carried
out at 308C for 30 min. The reactions were stopped by the
addition of SDS-loading bu€er and analysed by gel
electrophoresis/autoradiography.
Densitometry analysis
Densitometry analysis was performed with a BioImager as
described in ®gure legend. The data is based on one
representative experiment. Similar experiments yielded
identical results.
Acknowledgements
We would like to thank Melissa Bracher for helping with
the preparation of this manuscript, Judy G Ing and the Art
Department for the art work, Marilyn Lee for culture
setups, Dr Charles Sherr for providing antibodies and
cDNA probes, Dr Dennis Roop for providing Keratin
antibodies, Dr Stuart Yuspa for providing the retrovirus
packaging cells and David Nors for his technical advice.
We also want to thank Dr Laurent Meijer for generously
providing roscovitine and for providing us with unpublished results. This work was supported by grants from the
National Cancer Institute CA 42157 and CA 57596.
References
Albanese C, Johnson J, Watanabe G, Eklund N, Vu D,
Arnold A and Pestell RG. (1995). J. Biol. Chem., 270,
23589 ± 23597.
Bianchi AB, Fischer SM, Robles AI, Rinchik EM and Conti
CJ. (1993). Oncogene, 8, 1127 ± 1133.
Botz J, Zerfass-Thome K, Spitkovsky D, Delius H, Vogt B,
Eilers M, Hatzigeorgiou A and Jansen-Durr P. (1996).
Mol. Cell. Biol., 16, 3401 ± 3409.
Chin L, Schreiber-Agus N, Pellicer I, Chen K, Lee HW,
Dudast M, Cordon-Cardo C and De Pinho RA. (1995).
Proc. Natl. Acad. Sci. USA, 92, 8488 ± 8492.
Clarke AR, Maandag ER, Van Roon M, Van der Lugt
NMT, Van der Valk M, Hooper ML, Berns A and Riele H.
(1992). Nature, 359, 328 ± 330.
Cobrinik D, Lee M, Hannon G, Mulligan G, Bronson R,
Dyson N, Harlow E, Beach D, Weinberg R and Jacks T.
(1996). Genes Dev., 10, 1633 ± 1644.
Deng C, Zhang P, Harper JW, Elledge SJ and Leder P.
(1995). Cell, 82, 675 ± 684.
Ewen ME, Sluss HK, Sherr CJ, Matsushime H, Kato J and
Livingston DM. (1993). Cell, 73, 487 ± 497.
Filmus J, Robles AI, Shi W, Wong MJ, Colombo LL and
Conti CJ. (1994). Oncogene, 12, 3627 ± 3633.
Gotz C, Wagner P, Issinger O-G and Montenarh M. (1996).
Oncogene, 13, 391 ± 398.
Halevy O, Novitch BG, Spicer DB, Skapek SX, Rhee J,
Hannon G, Beach D and Lassar AB. (1995). Science, 267,
1018 ± 1021.
Hall M, Bates S and Peters G. (1995). Oncogene, 11, 1581 ±
1588.
Hauser P, Agrawal D, Flanagan M and Pledger W. (1997).
Cell Growth Di€er., 8, 203 ± 211.
Hennings H, Michael D, Cheng C, Steiner P, Holbrook K
and Yuspa SH. (1980). Cell, 19, 245 ± 254.
Hurlin PJ, Foley KP, Ayer DE, Eisenman RN, Hanahan D
and Arbeit JM. (1995). Oncogene, 11, 2487 ± 2501.
Hurford RJ, Cobrinik D, Lee M and Dyson N. (1997). Genes
Dev., 11, 1447 ± 1463.
Jacks T, Fazeli A, Schmitt EM, Bronson RT, Goodell MA
and Weinberg RA. (1992). Nature, 359, 295 ± 300.
Jiang H, Lin J, Su SS, Collart FR, Huberman E and Fisher
PB. (1994). Oncogene, 9, 3397 ± 3406.
Kato JY, Matsuoka M, Polyak K, Massague J and Sherr C.
(1994a). Cell, 79, 487 ± 496.
Kato JY, Matsuoka M, Strom DK and Sherr C. (1994b).
Mol. Cell. Biol., 14, 2713 ± 2721.
Kiess M, Gill MR and Hamel PA. (1995). Oncogene, 10,
159 ± 166.
Ko€ A, Giordano D, Desai D, Yamashita K, Harper JW,
Elledge S, Nishimoto T, Morgan DO, Franza BR and
Roberts JM. (1992). Science, 257, 1689 ± 1694.
Kranenburg O, van der Eb AJ and Zantema A. (1995). FEBS
Letts, 367, 103 ± 106.
Lee EYHP, Chang CY, Hu N, Wang YCJ, Lai CC, Herrup
K, Lee WH and Bradley A. (1992). Nature, 359, 288 ± 294.
Lee MH, Nikolic M, Baptista CA, Lai E, Tsai L-H and
Massague J. (1996). Proc. Natl. Acad. Sci. USA, 93, 3259 ±
3263.
Lew DJ, Dulic V and Reed SI. (1991). Cell, 66, 1197 ± 1206.
Lin J, Reichner C, Wu X and Levine AJ. (1996). Mol. Cell.
Biol., 16, 1786 ± 1793.
Liu JJ, Chao JR, Jiang MC, Ng SY, Yen JJ and Yang-Yen
HF. (1995). Mol. Cell. Biol., 15, 3654 ± 3663.
Lukas J, Bartkova J, Rohde M, Strauss M and Bartek J.
(1995). Mol. Cell. Biol., 15, 2600 ± 2611.
Matsushime H, Roussel MF, Ashmun RA and Sherr CJ.
(1991). Cell, 65, 701 ± 713.
Matsushime H, Quelle DE, Shurtle€ SA, Shibuya M, Sherr
CJ and Kato JA. (1994). Mol. Cell. Biol., 14, 2066 ± 2076.
Meijer L, Borgne A, Mulner O, Chong J, Blow J, Inagaki N,
Inagaki M, Delcros J and Moulinoux J. (1997). Eur. J.
Biochem., 243, 527 ± 536.
Missero C, Calautti E, Eckner R, Chin J, Tsai LH,
Livingston DM and Dotto GP. (1995). Proc. Natl. Acad.
Sci. USA, 92, 5451 ± 5455.
Missero C, Di Cunto F, Kiyokawa H, Ko€ A and Dotto GP.
(1996). Genes Dev., 10, 3065 ± 3075.
Morgan DO. (1995). Nature, 374, 131 ± 134.
Muller H, Lukas J, Scheider A, Warthoe P, Bartek J, Eilers
M and Strauss M. (1994). Proc. Natl. Acad. Sci. USA, 91,
2945 ± 2949.
Ohtsubo M, Theodoras AM, Shumacker J, Roberts JM and
Pagano M. (1995). Mol. Cell. Biol., 15, 2612 ± 2627.
Robles AI and Conti CJ. (1995). Carcinogenesis, 16, 781 ±
786.
Sambrook J, Fritsch EF and Maniatis T. (1989). Cold
Springs Harbor Lab. Press.
Schneider JW, Gu W, Zhu L, Mahdavi V and Nadal-Ginard
B. (1994). Science, 264, 1467 ± 1471.
Sherr CJ. (1994). Cell, 79, 557 ± 562.
Sherr CJ and Roberts JM. (1995). Genes Dev., 9, 1149 ± 1163.
Shim J, Lee H, Park J, Kim H and Choi E-J. (1996). Nature,
381, 804 ± 807.
SteinMan RA, Ho€man B, Iro A, Guillouf C, Liebermann
DA and El-Houseini ME. (1994). Oncogene, 9, 3389 ±
3396.
405
Cell cycle regulation in terminal differentiation
LA Martinez et al
406
Vastrik I, Kaipainen A, Penttila TL, Lymboussakis A,
Alitalo R, Parvinen M and Alitalo K. (1995). J. Cell Biol.,
128, 1197 ± 1208.
Weinberg RA. (1995). Cell, 81, 323 ± 330.
Weinberg WC, Azzoli CG, Chapman K, Levine AJ and
Yuspa SH. (1995). Oncogene, 10, 2271 ± 2279.
Xiong Y, Connolly T, Futcher B and Beach D. (1991). Cell,
65, 691 ± 699.
Yuspa SH, Kilkenny AE, Stanley J and Lichti U. (1985).
Nature, 314, 459 ± 462.