Induction of p21(WAF1/CIP1) and inhibition of Cdk2 mediated by the tumor suppressor p16(INK4a).

Abstract

The tumor suppressor p16(INK4a) inhibits cyclin-dependent kinases 4 and 6. This activates the retinoblastoma protein (pRB) and, through incompletely understood events, arrests the cell division cycle. To permit biochemical analysis of the arrest, we generated U2-OS osteogenic sarcoma cell clones in which p16 transcription could be induced. In these clones, binding of p16 to cdk4 and cdk6 abrogated binding of cyclin D1, p27(KIP1), and p21(WAF1/CIP1). Concomitantly, the total cellular level of p21 increased severalfold via a posttranscriptional mechanism. Most cyclin E-cdk2 complexes associated with p21 and became inactive, expression of cyclin A was curtailed, and DNA synthesis was strongly inhibited. Induction of p21 alone, in a sibling clone, to the level observed during p16 induction substantially reproduced these effects. Overexpression of either cyclin E or A prevented p16 from mediating arrest. We then extended these studies to HCT 116 colorectal carcinoma cells and a p21-null clone derived by homologous recombination. In the parental cells, p16 expression also augmented total cellular and cdk2-bound p21. Moreover, p16 strongly inhibited DNA synthesis in the parental cells but not in the p21-null derivative. These findings indicate that p21-mediated inhibition of cdk2 contributes to the cell cycle arrest imposed by p16 and is a potential point of cooperation between the p16/pRB and p14(ARF)/p53 tumor suppressor pathways.

Full text

Page 1

MOLECULAR AND CELLULAR BIOLOGY,
0270-7306/99/$04.00?0
Copyright © 1999, American Society for Microbiology. All Rights Reserved.
May 1999, p. 3916–3928Vol. 19, No. 5
Induction of p21WAF1/CIP1and Inhibition of Cdk2 Mediated by
the Tumor Suppressor p16INK4a
JAYASHREE MITRA,1,2,3CHARLOTTE Y. DAI,1,2,3KUMARAVEL SOMASUNDARAM,1,2,3,4
WAFIK S. EL-DEIRY,1,2,3,4KAPAETTU SATYAMOORTHY,5MEENHARD HERLYN,5
AND GREG H. ENDERS1,2,3*
Departments of Medicine1and Genetics,2Cancer Center,3and Howard Hughes Medical Institute,4University of
Pennsylvania School of Medicine, and The Wistar Institute,5Philadelphia, Pennsylvania 19104
Received 15 October 1998/Returned for modification 14 November 1998/Accepted 22 February 1999
The tumor suppressor p16INK4ainhibits cyclin-dependent kinases 4 and 6. This activates the retinoblastoma
protein (pRB) and, through incompletely understood events, arrests the cell division cycle. To permit bio-
chemical analysis of the arrest, we generated U2-OS osteogenic sarcoma cell clones in which p16 transcription
could be induced. In these clones, binding of p16 to cdk4 and cdk6 abrogated binding of cyclin D1, p27KIP1, and
p21WAF1/CIP1. Concomitantly, the total cellular level of p21 increased severalfold via a posttranscriptional
mechanism. Most cyclin E-cdk2 complexes associated with p21 and became inactive, expression of cyclin A was
curtailed, and DNA synthesis was strongly inhibited. Induction of p21 alone, in a sibling clone, to the level
observed during p16 induction substantially reproduced these effects. Overexpression of either cyclin E or A
prevented p16 from mediating arrest. We then extended these studies to HCT 116 colorectal carcinoma cells
and a p21-null clone derived by homologous recombination. In the parental cells, p16 expression also aug-
mented total cellular and cdk2-bound p21. Moreover, p16 strongly inhibited DNA synthesis in the parental
cells but not in the p21-null derivative. These findings indicate that p21-mediated inhibition of cdk2 contrib-
utes to the cell cycle arrest imposed by p16 and is a potential point of cooperation between the p16/pRB and
p14ARF/p53 tumor suppressor pathways.
p16INK4a/CDKN2A/MTS1is a specific inhibitor of the closely
related cyclin-dependent kinases (cdk’s) 4 and 6 (96, 109).
Mutations at the INK4a locus segregate with tumor formation
in familial melanoma pedigrees, and the locus is within a small
region of chromosome 9 that is deleted in multiple human
tumor cell lines (45, 46, 73). The locus encodes a second
protein, p14ARF(ARF), from an alternative reading frame (19,
63, 81, 100). ARF is unrelated in sequence to p16 but, surpris-
ingly, can also inhibit cell cycle progression (56, 81). Although
p16 is preferentially inactivated by point mutation or promoter
methylation in a number of tumors (35, 80), recent data sug-
gest that ARF is a tumor suppressor protein in its own right,
potentiating effects of p53 (14, 47, 78, 111). Thus, the INK4a
locus regulates the two most pervasive tumor suppressor path-
ways.
p16 is a potent mediator of G1cell cycle arrest in tissue
culture cells (52, 58, 67, 95). The only known effector of the
arrest is the retinoblastoma protein (pRB), consistent with
evidence that a major role of cdk4/6 is to phosphorylate and
inactivate pRB (98). Prominent among pRB’s properties is
that it binds E2F transcription factors and represses the activity
of promoters with the corresponding response elements (20,
39, 106). These elements are found in several genes required
for cell replication, including the cdk2 partners cyclins E and
A, cdk1 (cdc2), ribonucleotide reductase, and DNA polymer-
ase alpha (15). Overexpression of a dominant-negative form of
DP-1, the major heterodimeric partner of E2Fs, can inhibit G1
progression in mammalian cell lines (107). However, sub-
stantial cell replication occurs in Drosophila melanogaster
embryos with mutations in E2F or DP-1 after apparent ex-
haustion of the maternally derived proteins (88). Hence, it is
unclear whether pRB-mediated repression of E2F-responsive
genes is sufficient to account for the ability of pRB and p16 to
rapidly impose a cell cycle arrest.
There is increasing evidence that inhibition of cdk2 com-
plexes may be sufficient and even necessary for effective cell
cycle inhibition in many settings. Inhibitors of the Cdk-inter-
acting protein (CIP)/Clk-inhibitin, protein (KIP) class, al-
though capable of binding many cdk’s, appear to primarily act
through inhibition of cdk2 (8, 11, 12, 60, 99). Efficient G1arrest
in response to DNA damage is brought about by the ARF/p53
pathway and requires induction of p21WAF1/CIP1/sdi1/CDKN1
transcription (10, 16, 22). Inhibition of the cell cycle by glu-
cocorticoids also appears to involve p21 induction and can
occur in pRB-deficient cells (87). p27KIP1contributes to cell
cycle arrests mediated by transforming growth factor ? and
contact inhibition (76). Null alleles of p27 in mice and the
structurally related inhibitor dacapo in Drosophila melano-
gaster enhance cell cycling in diverse tissues (17, 27, 50, 54, 72).
Overexpression studies suggest that dysregulation of cyclin E
can bypass normal G1controls. Forced expression of cyclin E
can induce DNA synthesis in most postmitotic cells in the
Drosophila embryo (51). In mammalian cells, ectopic expres-
sion of cyclin E-cdk2 can rescue an arrest mediated by repres-
sion of c-Myc-responsive genes (7). Overexpression of cyclin E
has recently been found to rescue an arrest mediated by dom-
inant-negative DP-1, a phosphorylation-deficient mutant of
pRB, or p16 in rodent fibroblasts (2, 57). This occurs without
reversing pRB hyperphosphorylation or its association with
E2F transcription factors. Although overexpression studies
must be interpreted cautiously, p16-mediated activation of
pRB and sequestration of E2F appear to be insufficient to
maintain a cell cycle arrest in these settings if cyclin E-cdk2 is
ectopically hyperactivated. On the other hand, these studies
have not addressed whether cell cycle inhibition by p16 nor-
* Corresponding author. Mailing address: Penn/GI Division, 600
CRB, 415 Curie Blvd., Philadelphia, PA 19104-6144. Phone: (215)
898-0159. Fax: (215) 573-2024. E-mail: endersgh@mail.med.upenn
.edu.
3916

Page 2

mally involves and/or requires inhibition of endogenous cdk2.
Studies of p15INK4b, a cdk inhibitor structurally related to p16,
indicate that its binding to cdk4 and cdk6 results in redistribu-
tion of p27 and p21 to cdk2, potentially contributing to cell
cycle arrest (83).
To further explore the mechanism of action of p16, we
performed a biochemical analysis of events that occur during
p16-mediated cell cycle inhibition. We then manipulated these
events through targeted genetic experiments. We generated
clones of the human osteogenic sarcoma cell line U2-OS which
could be induced to express p16. U2-OS is the cell line in which
the p16’s function has been most extensively studied (52, 58,
59, 67). p16 is not expressed endogenously in these cells, and
its exogenous expression strongly inhibits DNA synthesis. This
inhibition appears to be specific, because it is abrogated by
coexpression of cdk4 or cdk6 or by introducing tumor-associ-
ated mutations into p16 (52, 58, 67).
Using such clones, we found that cell cycle inhibition by p16
is associated with a posttranscriptional induction of p21 and a
strong inhibition of cyclin E-cdk2 kinase activity. In a sibling
clone, induction of p21 alone to the level induced by p16 was
sufficient to significantly inhibit cyclin E-cdk2 activity and
DNA synthesis. We then extended our studies to the HCT 116
human colorectal carcinoma cell line and a p21-null clone
derived by two rounds of homologous recombination (104).
p16 expression in the parental HCT 116 cells also caused an
increase in total cellular and cdk2-bound p21. Moreover, p16
inhibited DNA synthesis in these cells much more effectively
than in the p21-null derivative. These results indicate that
p21-mediated inhibition of cdk2 is a key means by which p16
achieves cell cycle arrest and a common element of the ARF/
p53 and p16/pRB response pathways.
MATERIALS AND METHODS
Cell culture and generation of inducible cell lines. Stable p16-inducible clones
were generated by transfection. A U2-OS osteogenic sarcoma cell clone (U24)
containing an integrated pUHD15 vector (31), which encodes a tetracycline
(TET)-repressible transcriptional activator, was kindly provided by Liang Zhu
(Albert Einstein College of Medicine) (43). The p16 cDNA was cloned into the
pUHD10-3 vector (31) and cotransfected into U24 with a puromycin resistance-
encoding plasmid. Puromycin-resistant colonies were maintained in the presence
of TET. There were no gross differences in the numbers or sizes of puromycin-
resistant colonies obtained with the p16 construct and those obtained with the
empty target vector. This suggests that in the presence of TET, the level of
expression of p16 was sufficiently low in most transfected clones that their growth
was not inhibited; it also indicates that there was little selection pressure for
lesions in the p16 response pathway.
We randomly picked 34 well-separated clones obtained by transfection of
approximately 5 ? 105cells with the p16 construct. We also picked two vector-
transfected control clones and pooled the rest as OSvec.pool. p16-inducible
clones were identified by immunofluorescence, using the JC6 monoclonal anti-
body provided by Jim Koh (University of Vermont Cancer Center) (24). Four
clones showed levels of staining distinctly above background in the absence of
TET and were amplified for further analysis. p21-inducible cells were generated
in a similar fashion, except that an anti-p21 antibody (WAF1; Oncogene Re-
search) was used to screen for expression.
Cells were maintained in Dulbecco’s modified Eagle’s medium (DMEM; Life
Technologies) containing 10% fetal bovine serum (Life Technologies), penicillin
(100,000 U/liter), and streptomycin sulfate (100,000 U/liter). TET was added to
the culture medium at 1 to 2 ?g/ml to suppress expression of the inducible
proteins. To allow induction, cells were trypsinized, washed with DMEM, pel-
leted at 300 ? g for 5 min, and plated in DMEM without or with appropriate
concentrations of TET. Cells were generally maintained at no more than 50 to
70% confluence. At greater cell densities, U2-OS cells display moderate contact
inhibition of growth, associated with increased p21 levels (70a).
Parental and p21?/?HCT 116 cells were obtained from Bert Vogelstein
(Howard Hughes Medical Institute, The Johns Hopkins Oncology Center) and
maintained in McCoy’s 5A medium supplemented with 10% fetal calf serum
(FCS) and penicillin-streptomycin. The p21?/?cells were periodically treated
with neomycin and puromycin to prevent outgrowth of potential residual paren-
tal cells.
Cell cycle synchronization. Cells were synchronized with lovastatin (Merck
Research Laboratories, West Point, Pa.) by the following method (33, 49, 108).
Twenty-four to 36 h after seeding of cells, fresh medium containing 40 ?M
lovastatin was added, and the cells were then grown for 40 h with or without
TET. At the end of 40 h (0 h time point; see Results), the cells were released
from lovastatin-induced G1arrest by addition of fresh medium containing 4 mM
mevalonic acid (Sigma). Cells were harvested at various times and processed
either for biochemical analysis or for cell cycle position analysis by flow cytom-
etry.
Flow cytometry. For flow cytometry, cells were harvested by trypsinization,
centrifuged at 300 ? g for 5 min, and resuspended in 100 ?l of phosphate-
buffered saline (PBS). The cells were fixed by dropwise addition of 1 ml of 100%
ethanol and stained for 30 min at 37°C with a solution consisting of 0.001%
propidium iodide and 250 ?g of RNase A/ml. Total cellular DNA content was
determined on a Becton-Dickinson flow cytometer, using ModFit and Synch-
Wizard software. Small debris (less than about one-quarter of the diameter of a
cell) and cell aggregates were gated out of the analysis, and each profile was
compiled from approximately 5,000 gated events.
Analysis of CD20-stained cells was performed as described previously (103).
Tritiated-thymidine incorporation. Cell proliferation was also measured by
[3H]thymidine uptake as follows. Cells were pulse-labeled with 2.0 ?Ci of
[3H]thymidine/ml for 45 min prior to being harvested. After two washes with
PBS, the cells were trypsinized and retrieved by centrifugation as described
above. The cell pellet was resuspended in 0.3 N NaOH for 30 min on ice prior
to the addition of 20% cold trichloroacetic acid (TCA). The TCA precipitate was
filtered through glass fiber filters (Whatman GF/C) by the use of a benchtop
apparatus (Millipore model 1225). The filter was washed with 5 ml of cold 10%
TCA and then with 5 ml of 95% ethanol. After the filter was air dried, the
radioactivity associated with it was counted by the use of a fluor scintillant
(Beckman model LS6000IC counter).
Preparation of cell extracts. Subconfluent cells on 10-cm2plates were washed
twice with PBS and lysed by addition of 0.8 ml of cold Ela lysis buffer (38) with
protease inhibitors (50 mM HEPES [pH 7.0], 250 mM NaCl, 0.1% Nonidet P-40,
5 mM EDTA, 0.5 mM dithiothreitol [DTT], 1 mM phenylmethylsulfonyl fluo-
ride, and 0.1 mM sodium vanadate). Lysed cells were scraped off the plates,
incubated on ice for 15 min, and subjected to centrifugation at 14,000 ? g for 15
min at 4°C. The protein contents of the supernatants were assayed by the
Bradford method and confirmed by Coomassie blue staining following sodium
dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE).
Abs. All antibodies (Abs) were from Santa Cruz Biotechnology unless other-
wise indicated. For immunoblotting of p16, we used a mouse monoclonal Ab,
JC6, raised against human p16, at a 1:4 dilution (24). Immunoprecipitation (IP)
of p16 was performed with a polyclonal rabbit anti-p16 Ab from PharMingen
(San Diego, Calif.), raised against full-length human recombinant glutathione
S-transferase (GST) fusion protein, at 2 ?l/200 ?g of total cell extract. For
immunoblotting of p21, an affinity-purified mouse monoclonal Ab, WAF1 (On-
cogene Research Products, Cambridge, Mass.), raised against full-length recom-
binant human p21 WAF1, was used at a 1:200 dilution (with overnight incuba-
tion). p21 was immunoprecipitated with a rabbit polyclonal Ab (C19) at 0.5
?g/200 ?g of total cell extract. Immunoblotting and IP of cdk2 was performed
with a rabbit polyclonal Ab (M2) at a 1:500 dilution. A mouse monoclonal
anti-cdk2 Ab (D12) was used to confirm the levels of cdk2 immunoprecipitated
with the rabbit Ab. D12 was raised against a peptide corresponding to amino
acids (aa) 1 to 290 of full-length human cdk2 and was used at a 1:100 dilution.
For both IP and immunoblotting of cdk4, C22, a rabbit polyclonal Ab raised
against a peptide corresponding to aa 282 to 303 C at the carboxy terminus of
mouse cdk4, was used at a 1:100 dilution. The anti-cdk6 Ab used for both IP and
immunoblotting (1:500 dilution) was a rabbit polyclonal Ab raised against the
carboxy-terminal peptide. A mouse monoclonal immunoglobulin G1 (IgG1) Ab
(HD11) raised against recombinant human cyclin D1 was used for both IP and
immunoblotting of cyclin D. We used two cyclin E antibodies. A rabbit poly-
clonal Ab (C19) raised against a peptide corresponding to aa 377 to 395 of
human cyclin E was used for IP of cyclin E from extracts (1.5 ?g/200 ?g of total
cell extract). A mouse monoclonal IgG2b Ab raised against recombinant human
cyclin E protein (HE12) was used at a 1:100 dilution for immunoblotting of the
protein in immunoprecipitates. The 12CA5 Ab, directed against the influenza
virus hemagglutinin epitope, was used as a control in some precipitations. For IP
as well as immunoblotting of pRB, a rabbit polyclonal Ab (C15) raised against a
peptide corresponding to aa 914 to 928 at the carboxy terminus of human pRB
(110 kDa) was used at a 1:1,000 dilution. The anti-rabbit and anti-mouse sec-
ondary antibodies (Amersham) were used at a 1:5,000 dilution.
Immunoblotting and IP. Cell extracts were normalized for protein concentra-
tion and resolved by SDS-PAGE. The proteins were transferred to polyvinyli-
dene difluoride membranes (Immobilon P or Hybond P; Amersham Life Sci-
ence) either for 2 h or overnight. The membrane was blocked in 5% nonfat milk
in TBS-Tween (10 mM Tris-HCl [pH 8.0], 100 mM NaCl, 1% Tween 20) for 1 h.
The membrane was then washed for 1 h in TBS-Tween with three changes prior
to incubation with the primary Ab for 1 h or overnight. After a second 1-h wash
in TBS-Tween, the membrane was incubated with the appropriate secondary Ab
for 1 h. The membrane was washed in TBS-Tween for 1 h, and the bound antigen
was detected by enhanced chemiluminescence (ECL; Amersham). For IP, cell
extracts were precleared with a 20-?l packed volume of protein G-agarose beads
(Life Technologies) for 1 h at 4°C. The precleared extracts were incubated for 2 h
with fresh, washed protein G-agarose and an appropriate Ab in a total volume of
VOL. 19, 1999 p21 IS AN EFFECTOR FOR p16 3917

Page 3

300 ?l of ELB. After IP, the beads were washed three times with ELB, heated
at 90°C for 3 min with 2? sample buffer (100 mM Tris-HCl [pH 6.8], 200 mM
DTT, 4% SDS, 0.2% bromophenol blue, and 20% glycerol), and the supernatant
proteins were resolved by SDS-PAGE. Immunoblotting of the precipitated an-
tigen was performed by enhanced chemiluminescence as described above. All
images were scanned with a PhosphorImager (Molecular Dynamics Storm).
IP-kinase assays. (i) Histone H1 as substrate. Precleared cell extracts were
subjected to IP with the appropriate antibody (anti-cyclin E or anti-cdk2) as
described above. After the immunoprecipitation, the beads were washed three
times with ELB and once with 1? kinase assay buffer (25 mM HEPES buffer [pH
7.4], 1 mM phenylmethylsulfonyl fluoride, 0.1 mM sodium vanadate, 10 ?g of
aprotinin/ml, and 10 ?g of leupeptin/ml) and then incubated with a kinase
reaction mix (5 ?g of histone H1 [Sigma], 0.1 mM ATP, and 0.2 ?Ci of
[?-32P]ATP in a total volume of 10 ?l of 1? kinase buffer) for 30 min at 30°C
with shaking. The reaction was stopped by addition of 2? sample buffer to the
suspension; the reaction mixture was then heated at 90°C for 3 min, and the
products were resolved by SDS-PAGE (10% gels).
(ii) GST-pRB as substrate. GST-tagged carboxy-terminal fragment of pRB
was prepared as described elsewhere (68). The cdk4 kinase assay method fol-
lowed was as described elsewhere with slight modifications (53). Cell extracts
were precleared with a 20-?l packed volume of protein G-agarose for 30 min at
4°C. The precleared extracts were immunoprecipitated with anti-cdk4 antibody
for 2 h at 4°C in a total volume of 300 ?l of ELB buffer. The immunoprecipitated
complexes were washed three times with 300 ?l of ELB buffer and once with
pRB kinase buffer (50 mM HEPES [pH 7.5] containing 10 mm MgCl2and 5 mM
MnCl2). The beads were resuspended in kinase reaction mix (0.125 mM EGTA,
10 ?g of bovine serum albumin/ml, 2.5 ?M cold ATP, 5 ?Ci of [32P]ATP, and 2
?g of GST-pRB/reaction) and incubated for 30 min at 30°C. The reaction was
quenched by addition of 2? sample buffer; then the reaction mixture was heated
for 3 min at 90°C, and the products were resolved by SDS-PAGE (10% gels). The
signal intensities of the bands were quantified by using a PhosphorImager fitted
with ImageQuant version 1.1 software.
Northern blotting. Total RNA was isolated from cultured cells by using Trizol
reagent (Life Technologies), and 20 ?g of each preparation was subjected to
electrophoresis in a 1% agarose gel containing formaldehyde (91). RNA was
blotted onto a nylon membrane and detected with32P-labeled probes generated
from the entire 2.1-kb p21WAF1/CIP1cDNA. As a loading control, the same blot
was rehybridized with an RNase P probe (a gift from Robert P. Ricciardi
[University of Pennsylvania Dental School]).
Metabolic labeling of cells and IP of p21. Cells were preincubated for 30 min
(37°C; 5% CO2environment) in methionine-deficient DMEM containing 10%
dialyzed FCS (Life Technologies). Fresh medium containing 25 ?Ci of [35S]me-
thionine-[35S]cysteine (?1,000 Ci/mmol) (Translabel; Amersham)/ml was added
for either 1 h or 30 min. The cells were washed two times with PBS and once with
complete DMEM containing 10% FCS and 100 ?g of unlabeled L-methionine
andL-cysteine per ml and then incubated in this medium. After another 1, 2, and
4 h (or 30 min, 1 h, and 2 h for cells exposed to the label for 30 min), extracts
were prepared as described above. Protein content was normalized by the Brad-
ford reaction and confirmed by Coomassie blue staining of proteins separated by
SDS-PAGE. We found that this also effectively normalized incorporation of the
radiolabel in extracts prepared from the pulse phase, with or without p16 induc-
tion. p21 was immunoprecipitated as described above, using the C19 antibody
(Santa Cruz). Proteins were resolved by SDS-PAGE (12% gels), and the gels
were fixed for 30 min in a 50% methanol–10% acetic acid solution, dried, and
applied to a PhosphorImager. Parental and p21-null HCT 116 cells were labeled
and processed in the same way to confirm the identity of the p21 band. Other
bands in the lanes derived from cells subjected to p16-induction typically exhib-
ited either threefold-higher intensity than in immunoprecipitates from cells with-
out p16 induction (data not shown), suggesting co-IP with p21, or similar inten-
sity, suggesting cross-reactivity of the antibody with unrelated cellular proteins.
One of the latter is shown in Fig. 2D, as a loading control.
Transient transfection. Cells (2 ? 105) plated in 35-mm-diameter dishes were
transiently transfected for 6 h by using 2 ?l of Lipofectin reagent (Life Tech-
nologies/Gibco BRL) and 2 to 4 ?g of DNA diluted in 100 ?l of reduced-serum
medium (OPTIMEM; Gibco BRL) per the manufacturer’s instructions. HCT
116 p21?/?cells are more transfectable than the parental p21?/?cells; somewhat
less Lipofectin was required to achieve similar fractions of transfected cells and
similar levels of protein expression with the same amounts of DNA.
Adenoviruses. A recombinant adenovirus expressing p16 (Ad5CMV/p16, or
Ad5p16) was generated by established protocols (32). Briefly, p16 cDNA (pro-
vided by David Beach, Cold Spring Harbor Laboratories) was cloned into shuttle
vector pAD/CMV (Vector Core, Institute for Human Gene Therapy, University
of Pennsylvania, Philadelphia) and cotransfected with E3-deleted human ade-
novirus dl70001 into 293 cells, using calcium phosphate precipitation. Positive
plaques were identified by Southern blotting, repurified three times, and prop-
agated in 293 cells. Subsequently, the recombinant virus was produced on a large
scale and purified twice by cesium chloride density gradient centrifugation (93,
94). Preparation of Ad5CMV/lacZ (Ad5lacZ) has been described previously
(94).
HCT 116 cells were infected at 10% confluence 2 days after being replated.
The cells were washed once with PBS and incubated with ca. 20 PFU (deter-
mined on 293 cells) of virus/cell in serum-free McCoy’s 5A medium for 3 h. The
medium was removed, the cells were washed once with PBS, and growth medium
was added. Cell extracts were prepared extracts 24 h after infection.
BrdU labeling and immunofluorescence. Bromodeoxyuridine (BrdU) labeling
was performed as described previously (24) with the following modifications. For
labeling of U2-OS cells, BrdU was added in aliquots 24 to 32 h after TET
withdrawal. For labeling of HCT 116 cells, BrdU was added in aliquots 24 to 48 h
after the end of the transfection period. Immunofluorescence staining was per-
formed as described in reference 24, except that an incubation time of 1 h was
routinely used. Nuclear DNA was stained, with bisbenzimide (1 ?g/ml) (Hoechst
33258; Sigma) being included in the final wash. The results were scored by direct
microscopic examination of randomly chosen medium-power fields by an obser-
vor blinded to the treatment conditions. For photography, images of each fluo-
rescence wavelength were independently and quantitatively captured in greyscale
using a cooled charge-coupled device camera (Sensys model KF 1400) mounted
to a Leica DM IRB inverted microscope and controlled by IP Lab Spectrum
software (Scanalytics, Fairfax, Va.). These images were recombined and pseudo-
colored by using the Multiprobe software (Scanalytics). Each field presented was
treated with the same refinements of contrast and color.
Polyribosome preparation and RT-PCR. OSp16.1 cells were lysed in a solution
containing 10 mM Tris (pH 8.0), 30 mM MgCl2, 150 mM KCl, 1 mM DTT, 1%
Nonidet P-40, and 20 mM ribonucleoside-vanadate complexes/ml (23). Nuclei
and cell debris were removed by centrifugation at 10,000 ? g for 10 min. Equal
amounts of total RNA from the cytoplasmic extracts (estimated from samples
prepared by Trizol extraction) in 1.5-ml volumes were layered onto continuous
10 to 50% sucrose gradients in a solution containing 20 mM Tris (pH 8.0), 50
mM KCl, and 5 mM MgCl2. Centrifugation was carried out at 28,000 rpm for 4 h
in an SW41 rotor. Preliminary experiments employing Northern blotting had
demonstrated that the p21 mRNA was largely confined to the lower half of a
gradient. After the run, the upper halves of all gradient were therefore pooled as
fraction 1 and the lower halves was separated into four fractions (2 to 5). tRNA
(50 ?g; Gibco BRL) was added to each of the fractions as a carrier to facilitate
RNA recovery. RNA in the fractions was precipitated with 2.5 volumes of 100%
ethanol and with 3 M sodium acetate (10% of the total volume), treated with
RNase-free DNase (Gibco BRL), and extracted with Trizol. Equal portions (by
volume) were reverse transcribed, using 10 U of avian myeloblastosis virus
reverse transcriptase (Promega), 40 U of RNasin, 70 ng of random primer, and
0.4 mM each dATP, dCTP, dGTP and dTTP in a total volume of 20 ?l. Reverse
transcription (RT) was conducted in a model 9600 thermal cycler (Perkin-Elmer)
for 1 h at 42°C. An appropriate volume of the RT reaction product was used in
PCR reaction mixtures containing 10 mM Tris (pH 8.3), 50 mM KCl, 1.45 mM
MgCl2, 0.4 mM each deoxynucleoside triphosphate, 0.8 U of Taq DNA poly-
merase (Boehringer Mannheim), and 125 ng each of the 5? and 3? primers in a
total volume of 20 ?l. The WAF1/p21 primers used were as follows: 5?-AGGA
TCCATGTCAGAACCGGCTGG-3? and 5?-CAGGATCCTGTGGGCGGATT
AGGGCT-3? (37). A three-step cycling PCR program was used, consisting of 40
cycles of 94°C for 1 min, 54°C for 1 min, and 72°C for 1 min, with a final 5-min
strand extension at 72°C. Glyceraldehyde 3-phosphate dehydrogenase used as
internal standard was amplified for 24 cycles with the following primers: 5?-TG
GTATCGTGGAAGGACTCATGAC-3? and 5?-ATGCCAGTGAGCTTCCCG
TTCAGC-3? (73).
RESULTS
p16 disrupts cyclin D1-cdk4 binding. We generated U2-OS
cell clones in which p16 expression is regulated by the “TET-
off” transcription system (31). Induction of p16 and its effects
on DNA synthesis over various time courses in these clones
will be presented in detail in the future (14a). In a represen-
tative clone, OSp16.1, TET withdrawal yielded an eightfold
induction of p16 at 24 h, to a level moderately higher than that
in SaOS osteogenic sarcoma cells (Table 1). DNA synthesis
was inhibited by 75% during this period, as determined by
monitoring tritiated-thymidine incorporation or by flow cytom-
etry. By day 3 following TET removal, the p16 level was 25-fold
above the baseline and DNA synthesis was inhibited by more
TABLE 1. Inhibition of [3H]thymidine incorporation by induction
of p16 in OSp16.1
DayFold p16 induction% Inhibitiona
1
2
3
8 75 ? 9
85 ? 7
92 ? 2
18
25
aValues, normalized to percent [3H]thymidine incorporation in uninduced
cells, are means ? standard deviations of data from three determinations.
3918 MITRA ET AL.MOL. CELL. BIOL.

Page 4

than 90% compared to that of uninduced cells (Table 1).
Strong inhibition of DNA synthesis was observed on TET
withdrawal in three other clones with inducible p16 expression,
but not in two control clones transfected with the empty vector
(data not shown). Within 24 h of p16 induction in OSp16.1,
most pRB shifted from a slowly migrating form to a more
rapidly migrating form on SDS-PAGE, suggesting that it was
predominantly hypophosphorylated (data not shown) (see ref-
erences 26, 62, 92, and 102). Thereafter there was little change
in the ratio of the faster-migrating and slower-migrating pRB
forms, but the overall levels of the protein declined, as re-
ported by others (92).
cdk4 (Fig. 1A) and cdk6 (data not shown) activities, assayed
by IP and in vitro phosphorylation of a GST-pRB fusion pro-
tein (gstRB), were largely inhibited by p16 induction. This
coincided with p16 binding and with the loss of cdk4 from
cyclin D1 complexes (Fig. 1B). Similarly, most cdk6 was lost
from cyclin D1 immunoprecipitates and most cyclin D1 was
lost from cdk6 immunoprecipitates at days 1 and 3 of p16
induction (data not shown). These results differ from those
reported for p15INK4b, in which evidence for trimeric p15-
cyclin D-cdk complexes was obtained (82). Whether this can be
explained by differences in the cellular background or by in-
trinsic differences between the two inhibitors remains to be
determined.
p16 induction yields an inhibition of cdk2 complex activity.
We asked whether cdk2 activity was affected by the p16-medi-
ated arrest. We prepared extracts from cells which were or
were not subjected to 1 to 3 days of p16 induction and nor-
malized them for protein content. We immunoprecipitated
cdk2 and cyclin E complexes and assayed their kinase activity
in vitro, using histone H1 as a substrate. cdk2- and cyclin
E-associated kinase activities were inhibited in all lysates of
cells subjected to p16 induction (data not shown). To distin-
guish a decrease in the specific activity of the complexes from
a decrease in either their abundance or the ability of the
antibodies to precipitate them, we assayed kinase activity in the
immunoprecipitates after normalizing for cdk2 content. p16
induction resulted in a marked decrease in the specific activity
of cdk2 complexes (data not shown) and cyclin E complexes
(Fig. 1C).
Because p16 induction causes a cell cycle arrest in late G1,
we asked whether cell cycle position alone could account for
the observed decrease in the specific activity of cdk2 com-
plexes. We induced p16 during an arrest mediated by the
cholesterol biosynthesis inhibitor lovastatin. This drug has
been shown to effectively synchronize U2-OS cells in early G1
phase and is thought to interfere with critical membrane sig-
naling events (49, 79, 108). Following lovastatin treatment, we
restored cell cycling by addition of the downstream metabolite
mevalonic acid and assayed cell cycle position by flow cytom-
etry. Lovastatin caused a strong, predominantly G1arrest in a
pool of vector-transfected control clones (data not shown) and
in OSp16.1 (Fig. 1D). Mevalonic acid treatment restored cell
cycling in about half the control cells within 24 h, in the pres-
ence or absence of TET (data not shown). About half of the
OSp16.1 cells also resumed cycling in the presence of TET, but
entry into S phase was blocked by p16 induction in the absence
of TET (Fig. 1D).
We prepared lysates from OSp16.1 cells at successive 12-h
intervals following mevalonic acid rescue, with or without p16
induction, and normalized these samples for protein content.
Total cdk2 activity (data not shown) and cyclin E-associated
kinase activity were largely inhibited by p16 induction (Fig.
1E). Note that this inhibition occurred even at the 0-h time
point of mevalonic acid addition, a point at which the two
FIG. 1. p16 mediates inhibition of both cdk4 and cdk2 in OSp16.1 cells. (A)
Inhibition of cdk4 kinase activity. Extracts were prepared from cells cultured with
(?) or without (?) TET (1 ?g/ml) for 1 to 3 days and were normalized for
protein content. cdk4 was immunoprecipitated, and its kinase activity was as-
sayed using gstRB as a substrate. (B) (Left panel) Binding of p16 to cdk4.
Extracts were prepared from cells cultured with or without TET for 3 days and
were normalized for protein content. p16 complexes were immunoprecipitated
and immunoblotted for cdk4 (K4). (Right panel) Inhibition of cyclin D1 binding
to cdk4 on p16 induction. Extracts were prepared from cells that were or were
not subjected to p16 induction for 3 days. Cyclin D1 complexes were immuno-
precipitated (D1). These samples were normalized for cyclin D1 content by
immunoblotting (below) and immunoblotted for cdk4 (K4). (C) Inhibition of
cyclin E-associated kinase activity. Extracts were prepared from cells that were or
were not subjected to p16 induction for 1 to 3 days. Cyclin E complexes were
immunoprecipitated and normalized for cdk2 content by immunoblotting (K2)
(shown below). Kinase activities were then assayed using histone H1 (HH1) as a
substrate. (D) p16 induction prevents S phase entry following cell cycle synchro-
nization with lovastatin and rescue with mevalonic acid. OSp16.1 cells were
treated with lovastatin for 40 h in the presence (Uninduced) and absence (In-
duced) of TET. At the end of this period, a portion of the cells was harvested and
assayed for DNA content by flow cytometry (0 h; x axis, DNA content; y axis, cell
number, normalized to each G1peak). The lovastatin block was rescued in the
remaining cells by addition of mevalonic acid for another 36 h. Cells were
removed at 12-h intervals for analysis. The S phase fraction at 24 h was 43% in
the uninduced cells versus 5% in the induced cells. (E) Inhibition of cyclin
E-associated kinase activity by p16 induction in synchronized cells. Extracts were
prepared from cells treated as described for panel D and normalized for protein
content. Cyclin E complexes were immunoprecipitated, and their kinase activi-
ties were assayed using histone H1 (HH1) as a substrate. C, control (employing
nonspecific rabbit IgG). The relative signal intensity is given below each lane.
VOL. 19, 1999 p21 IS AN EFFECTOR FOR p163919

Page 5

cultures have similar flow cytometry profiles. Thus, inhibition
of cyclin E-associated kinase activity mediated by p16 was
observed in a comparison of cells synchronously arrested in
early G1and synchronously released from this block.
Induction of p21 by p16. A variety of experiments have
demonstrated that p16 does not associate with cdk2 (52, 57,
96), and we were unable to detect such an association in
OSp16.1 cells (data not shown). We asked whether the inhibi-
tion of cyclin E-associated kinase activity mediated by p16
could be explained by a decrease in expression of cyclin E or
cdk2 in the synchronized cells. Cyclin D1 and cdk4/6 levels
were not affected by the treatments (data not shown). Cyclin E
expression was also unaffected, and cdk2 levels dropped only
slightly in the cultures subjected to p16 induction (Fig. 2A and
data not shown). Expression of cyclin A was strongly inhibited
by p16 induction, although this effect was less pronounced in
unsynchronized cells (data not shown). The decrease in cyclin
A levels following addition of mevalonic acid illustrates the
point that the culture in which p16 was induced was arrested at
0 h by the effects of lovastatin but at later points by the effects
of p16.
We asked whether the observed decrease in cdk2-associated
kinase activity could be accounted for by a change in the total
cellular level of p27 or p21. p27 expression was not affected,
but p21 levels were more than eightfold higher in the cells
subjected to p16 induction (Fig. 2A). Note that like the de-
crease in cyclin E-associated kinase activity, induction of p21
occurred by the 0-h time point. The two cultures have similar
cell cycle profiles at this point (Fig. 1D), with the vast majority
of cells being in G1, before the normal arrest point mediated by
p16. In fact, synchronization of these cells in early G1by
lovastatin treatment results in a slight decrease in p21 levels
compared to those of untreated cells, accentuating p16’s effect
(data not shown) (see below). Furthermore, passage through
late G1did not result in a prominent increase in the p21 level.
We conclude that p16-mediated induction of p21 is not simply
a secondary effect of arresting cells in G1.
We then asked whether induction of p21 is seen in cultures
under standard passage conditions. Figure 2B shows that p21
was induced within 1 day of p16 induction in unsynchronized
cells (passage of U2-OS cells causes a moderate, transient
synchronization in G1phase [data not shown], but we refer to
them as unsynchronized for simplicity). Over a series of more
than 10 experiments with unsynchronized cells, we found that
p21 was typically induced three- to fourfold after 3 days of p16
induction (Fig. 2B [see also Fig. 4A below] and data not
shown). Induction of p21 was also seen on TET withdrawal in
three other clones with inducible p16 expression, but not in two
control clones (data not shown). We also analyzed p21 expres-
sion by immunohistochemistry. There was a greater degree of
p21 staining in cells subjected to p16 induction but no detect-
able change in the subcellular distribution, which remained
preferentially nuclear (data not shown).
We performed initial studies to characterize the level at
which p21 expression is regulated. We prepared RNA from
OSp16.1 cells maintained in the presence or absence of TET
for 1 and 3 days. In Northern hybridizations using the p21
cDNA as a probe, a single major species, of the expected size
(2.1 kb), was present at similar levels in all samples (Fig. 2C).
Metabolic pulse-labeling with [35S]methionine-[35S]cysteine
showed that the half-life of p21 remained about 1 h in the
presence or absence of p16 induction for 3 days (Fig. 2D and
E). In contrast, threefold more label was incorporated into p21
during the pulse-labeling phase in the setting of p16 induction
than in the controls. This ratio was unaffected by further re-
ducing the pulse duration from 1 h to 30 min (3.3- and 2.8-fold
in two independent experiments) (data not shown). The in-
creased metabolic labeling of p21 accounts well for the three-
to fourfold increase in the steady-state level of the protein
detected by immunoblotting following p16 induction and
points to regulation of p21 expression at the level of transla-
tion. Most translation appears to be controlled at the initiation
step, correlating with an increased number of ribosomes per
mRNA (64). Consistent with this, we observed increased sed-
imentation of p21 mRNA in polyribosome gradients on p16
induction (Fig. 2F). Further work will be required to elucidate
more fully the mechanism of regulation.
Inhibition of cdk2 complex activity by p21. We examined the
physical association of p27 and p21 with cdk’s by IP followed by
immunoblotting. We noted an increase in p27 associated with
cdk2 after 1 day of p16 induction, but this increase was lost by
day 3 (Fig. 3A). Association of p21 with cdk2 was also in-
creased by day 1 and had increased further by day 3. The
displacement of cyclin D1 from cdk4/6 complexes mediated by
p16 binding (Fig. 1B) would be expected to prevent the binding
of CIP/KIP inhibitors, since they require cyclins for efficient
binding to cdk’s (12, 90). Consistent with this, we found that
binding of p21 to cdk4 (Fig. 3A) and cdk6 (data not shown)
was disrupted by induction of p16 at both the 1-day and 3-day
time points. Thus, binding of p16 to cdk4/6 prevents the bind-
ing of p21, even as p21 levels increase.
cdk complexes associated with p21 can retain catalytic activ-
ity, typically at low inhibitor stoichiometries (11, 53, 110). We
therefore asked whether p21-bound cdk2 complexes were in-
deed inhibited. We immunoprecipitated cdk2 and p21 com-
plexes and normalized these precipitates for cdk2 content. The
p21-bound complexes retained far less catalytic activity for
histone H1 or gstRB (Fig. 3B and data not shown).
We next asked whether p21 or p27 complexes bound a major
fraction of cyclin E complexes in extracts prepared from cells
that were or were not subjected to 3 days of p16 induction. We
performed successive immunodepletions with antibodies di-
rected against p27, p21, or a negative-control antigen and
measured cyclin E in the remaining supernatants by immuno-
blotting. We observed the loss of the majority of cyclin E from
the supernatant only after immunodepletion with anti-p21 an-
tibodies and induction of p16 (Fig. 3C). These results indicate
that the majority of cyclin E in the extract from cells subjected
to p16 induction is bound to p21. Since p21 cannot bind effi-
ciently to free cyclins (12, 90) and cdk2 is the only major cdk
partner for cyclin E (18, 55, 75, 101), we infer that most of the
cyclin E-cdk2 complexes in these extracts are bound by p21.
These results do not exclude the possibility that p27 binds a
significant fraction of cyclin E complexes in the initial stages of
the arrest.
We asked whether overexpression of cyclin E or A could
prevent a p16-mediated arrest of these cells. We transfected
OSp16.1 cells with an empty cytomegalovirus (CMV) expres-
sion vector or one expressing cyclin E or A, induced p16 for
24 h, and assayed DNA synthesis by measuring BrdU incorpo-
ration. Successfully transfected cells were identified by cotrans-
fection with a vector expressing ?-galactosidase (?-gal) (see
Fig. 6A for examples of the technique). Expression of either
cyclin effectively prevented p16 from inhibiting DNA synthesis
(Fig. 3D). Immunoblotting demonstrated about a twofold in-
crease in cyclin levels in the respective transfected cultures
(data not shown). Immunofluorescence staining demonstrated
that about 25% of the cells were successfully transfected (data
not shown). We therefore estimate that cyclins E and A were
expressed at about eight times the endogenous level in the
successfully transfected cells.
3920MITRA ET AL.MOL. CELL. BIOL.