Oncogene (2001) 20, 7453 ± 7463
ã 2001 Nature Publishing Group All rights reserved 0950 ± 9232/01 $15.00
www.nature.com/onc
ORIGINAL PAPERS
Inhibition of Chk1-dependent G2 DNA damage checkpoint radiosensitizes
p53 mutant human cells
Kalli Koniaras1,3, Andrew R Cuddihy1,3, Helen Christopoulos1, Annette Hogg1 and
Matthew J O'Connell*,1,2
1
Trescowthick Research Laboratories, Peter MacCallum Cancer Institute, Locked Bag 1, A'Beckett St, Melbourne, VIC 8006
Australia; 2Department of Genetics, University of Melbourne, Parkville VIC 3052, Australia
Cell cycle checkpoints are surveillance mechanisms that
monitor and coordinate the order and ®delity of cell
cycle events. When defects in the division program of a
cell are detected, checkpoints prevent the pursuant cell
cycle transition through regulation of the relevant cyclincdk complex(es). Checkpoints that respond to DNA
damage have been described for the G1, S and G2 phases
of the cell cycle. The p53 tumour suppressor is a key
regulator of G1/S checkpoints, and can promote cell
cycle delay or apoptosis in response to DNA damage.
The importance of these events to cellular physiology is
highlighted by the fact that tumours, in which p53 is
frequently mutated, have widespread defects in the G1/S
DNA damage checkpoints and a heightened level of
genomic instability. G2/M DNA damage checkpoints
have been de®ned by yeast genetics, though the genes in
this response are conserved in mammals. We show here
using biochemical and physiological assays that p53 is
dispensable for a DNA damage checkpoint activated in
the G2 phase of the cell cycle. Moreover, upregulation of
p53 through serine 20 phosphorylation, does not occur in
G2. Conversely, we show that the Chk1 protein kinase is
essential for the human G2 DNA damage checkpoint.
Importantly, inhibition of Chk1 in p53 de®cient cells
greatly sensitizes them to radiation, validating the
hypothesis of targeting Chk1 in rational drug design
and development for anti-cancer therapies. Oncogene
(2001) 20, 7453 ± 7463.
Keywords: checkpoint; Chk1; p53; Cds1; DNA damage
Introduction
Cells replicate themselves with remarkable ®delity.
Signal transduction pathways that control the major
events in the cell cycle are well understood, and centre
on the activation of cyclin-dependent kinases (Cdks).
*Correspondence: MJ O'Connell, Trescowthick Research Laboratories, Peter MacCallum Cancer Institute, St Andrew's Place,
East Melbourne, VIC 3002, Australia;
E-mail: [email protected]
3
These authors contributed equally to the work
Received 25 June 2001; revised 15 August 2001; accepted 30
August 2001
Whilst Cyclin-Cdk complexes accumulate in interphase,
they are kept in an inactive state by either inhibitory
phosphorylations or by binding to Cdk inhibitor
proteins (CKIs) (Dunphy, 1994; Hunter and Pines,
1994; Sherr and Roberts, 1999). Checkpoints are in
place to ensure both the interdependency and ®delity of
S phase and mitosis. When defects in the division
program of a cell are detected, checkpoints are evoked
to inhibit or to prevent the activation of Cdks until
DNA replication or mitosis are completed, or when
perturbed, recti®ed to ensure stable genomic inheritance (Elledge, 1996; O'Connell et al., 2000).
In contrast to the ®delity of normal cell division,
cancer cells exhibit a high degree of genomic
instability, widely thought to promote a heightened
rate of mutation, allowing cancer cells to evolve
(Lengauer et al., 1998). The loss of checkpoint controls
operative in the G1 and S phases are essentially
universally seen in human tumours. A prime example
is the high rate of inactivation of the p53 tumour
suppressor. DNA damage-activated p53 can promote
cell cycle arrest, in order to allow time for DNA repair,
through transactivation of the CKI p21cip1 gene (ElDeiry et al., 1993), which inhibits the Cdk2/CyclinE
and Cdk2/CyclinA complexes ultimately required for
entry into and passage through S phase. In other cases,
either when DNA repair cannot be completed successfully or if cells are not programmed to respond to
stresses by viable growth arrest, p53 may destine the
cells for apoptosis (White, 1996). Inactivation of p53
thus promotes the evolution of cancer in two ways:
genomic instability and the absence of apoptotic
programs that would normally be triggered in response
to genetic abnormalities. From the point of view of
clinical management of many tumour types, the
absence of p53 function thwarts the ecacy of
genotoxic anti-cancer therapies.
The archetypal cdk, Cdc2, together with A and Btype cyclins, controls the transition from G2 into
mitosis (Dunphy, 1994). Unlike the G1/S Cdk2
complexes, these are not inhibited by CKIs, but rather
are under the control of inhibitory phosphorylations
on T14 and Y15 (Poon et al., 1996). Phosphorylation
on Y15 is a highly conserved event, with identical
regulation spanning back to ®ssion yeast (Nurse, 1990).
Studies of G2/M DNA damage checkpoints in this
Human G2 DNA damage checkpoint
K Koniaras et al
7454
Oncogene
yeast have identi®ed that Y15 phosphorylation of Cdc2
is the mechanism of checkpoint-mediated arrest
(O'Connell et al., 1997; Rhind et al., 1997). In human
cells, expression of a non-phosphorylatable mutant of
Cdc2 (Y15F) greatly abrogates the ecacy of the
checkpoint (Blasina et al., 1997; Jin et al., 1996). The
upstream checkpoint proteins are also highly conserved
between yeast and humans. In yeast these proteins send
a signal that results in the activation of a protein
kinase, Chk1, which in turn regulates Y15 phosphorylation through the activation of the Wee1 kinase and
inactivation of the Cdc25 phosphatase that regulate
Y15 phosphorylation (O'Connell et al., 2000; Raleigh
and O'Connell, 2000). Studies with relatively nonspeci®c protein kinase inhibitors in cell lines (Graves et
al., 2000), combined with in vitro biochemical experiments (Sanchez et al., 1997) have suggested that human
Chk1 behaves in an analogous way to yeast chk1,
however, loss-of-function experiments have yet to be
reported. Chk17/7 mice display early embryonic
lethality (Liu et al., 2000), which may be due to a
prevalence of G2/M checkpoints over the relative
absence of G1/S checkpoints early in development as
suggested by studies in Xenopus (Masui and Wang,
1998) and murine embryonic stem cells (SchmidtKastner et al., 1998). The PI3-K related protein
kinases, Ataxia Telangiectasia Mutated ATM, and a
related protein, ATR, are thought to activate Chk1 by
phosphorylation (Guo et al., 2000; Liu et al., 2000;
O'Connell et al., 2000; Zhao and Piwinica-Worms,
2001). While ATR7/7 mice are also embryonic lethal
(Brown and Baltimore, 2000), expression of dominant
negative ATR in somatic cells is not lethal, although it
does abrogate G2/M DNA damage checkpoints (Cliby
et al., 1998). This suggests that the essential nature of
ATR and Chk1 may be con®ned to the early
embryonic cell cycles.
Does p53 play a role in G2/M checkpoints? The
answer to this question is surprisingly unclear. Many
studies have shown that p53 mutant cells are capable
of G2 arrest in response to DNA damage (Kastan et
al., 1991; Taylor and Stark, 2001). However, such
studies have looked at cells in which DNA damage is
in¯icted either when growing asynchronously or
synchronized in S phase. Moreover, many of these
studies have used high doses of DNA damage from
which the cells cannot recover. Under these conditions,
studies with human cell lines in which p53 has been
deleted by homologous recombination suggest that p53
plays a role in maintaining a sustained G2 arrest (Bunz
et al., 1999). However, under these regimes, it may well
be that inhibition of CyclinA/Cdk2 complexes in late S
phase to early G2 are halting cell cycle progression.
Hence, the relationship between p53 and the checkpoint that signals through the ATM/R-Chk1 pathway
to maintain Y15 phosphorylation on Cdc2 is unclear.
There are few reports of mutations in regulators and
checkpoint genes involved in G2/M in tumours. With
the abundance of G1/S abnormalities in tumours, G2/
M checkpoint genes may become essential in somatic
tumour cells as seen in murine embyrogenesis. Should
this be the case, one would predict that abrogation of
G2/M checkpoints in tumour cells lacking G1/S
checkpoint control should greatly sensitize these cells
to genotoxic stress. We show here that p53 is
dispensable for a DNA damage mediated checkpoint
arrest when the checkpoint is evoked in G2 cells.
Conversely, interference with the ATM/R-Chk1 pathway by expression of a dominant negative Chk1
mutant signi®cantly abrogates the ability of cells to
undergo G2/M arrest. The combined e€ect of interfering with Chk1 signalling in the absence of p53 is a
profound radiosensitization that indicates the utility in
targeting Chk1 during genotoxic stress to kill p53
mutant tumour cells.
Results
p53 is dispensable for DNA damage-induced cell cycle
arrest in G2
Instability of the genome is seen in the vast majority, if
not all, of human tumours. Several mechanisms have
been shown to contribute to this, including failure of
DNA repair mechanisms and perturbations to cell
cycle checkpoints, both in interphase and mitosis. In
the G1 phase of the cell cycle, the tumour suppressor
protein p53 is a key checkpoint protein that is upregulated in response to DNA damage, and signals cell
cycle arrest through transactivation of cell cycle
inhibitory genes such as p21.
The transition from S phase into G2 and through to
mitosis in human cells sees a `passing of the baton' of
various cyclin/cdk complexes. Early on in the transition, cdk2/cyclinA complexes predominate, and act
upstream of the cdc2/cyclinA and B complexes that
ultimately are required for mitotic entry. Cdk2
complexes are inhibited by p21, and ternary complexes
consisting of p21, cdk2 and cyclinA are found
following DNA damage. Conversely, Cdc2 containing
complexes are neither inhibited by p21, nor does Cdc2
interact with p21, either during G2 or a checkpoint
arrest (Poon et al., 1996). As p21 is a major
transcriptional target of p53, the phase of the cell
cycle at which a checkpoint is activated may be
important in discerning the relevant target cyclin-cdk
complex. Based on this, we reasoned that some of the
discrepancy in the conclusions regarding the role of
p53 in G2 arrest could be due to the point in the cell
cycle at which the checkpoint is activated.
We therefore assessed whether p53 mutant cell lines
that had been synchronized in G2 prior to irradiation
still retained G2 checkpoint function. A signi®cant
enrichment of G2 cells in all experiments was assayed
by FACS DNA pro®les, the appearance of inactive,
Y15 phosphorylated, cyclinB1/Cdc2 complexes and the
absence of Cdc2 activity. Initially, we looked at DNA
damage-induced checkpoint control in G2 synchronized HeLa cells, which lack p53 function due to the
presence of E6 and E7 oncoproteins from the human
papilloma virus. These cells have been shown by many
Human G2 DNA damage checkpoint
K Koniaras et al
groups to be capable of G2 arrest when irradiated
during S phase or when growing asynchronously. In
G2 synchronized HeLa cells, pilot experiments using a
range of doses showed that 6 Gy delayed entry into
mitosis for 6 ± 8 h, and cells maintained *30%
viability in clonogenic assays (data not shown). Figure
1 shows an analysis of G2 cell cycle proteins from cells
prepared by this protocol. Under these conditions,
inactive, Y15 phosphorylated Cdc2 accumulated to
high levels. However, this accumulation can be
accounted for by concomitant accumulation of cyclins
A and B1, which would provide a larger substrate pool
for the Y15 kinases Wee1 and Myt1, which only
phosphorylate Cdc2 when bound to a cyclin (Booher et
al., 1997; Fattaey and Booher, 1997; Liu et al., 1997;
Parker et al., 1991). This was notable because when
cells are irradiated at higher doses, from which fewer
cells will recover, the transcription of cyclin B1 is
repressed and the proteins levels are unable to
accumulate or be maintained (Muschel et al., 1993).
Altered sub-cellular localization of cell cycle proteins
has been proposed to be a contributing factor to
checkpoint control in yeast and human cells, though
follow-up experiments have questioned the initial
conclusions (Lopez-Girona et al., 2001). Wee1 is a
nuclear protein, and its localization has been proposed
to protect the nucleus from precocious mitotic events
(Heald et al., 1993). However, one study found that
Wee1 co-immunoprecipitated with the cytoplasmic and
p53-regulated 14-3-3s protein in cells treated with
Adriamycin (Chan et al., 1999). We investigated the
e€ects of ionizing radiation on the localization of Wee1
by indirect immuno¯uorescence, and found its nuclear
location to be una€ected by irradiation (Figure 2) and
Figure 1 Accumulation of G2 cell cycle proteins in checkpoint
arrested cells. Protein extracts were prepared from HeLa cells that
were growing asynchronously (A), synchronized in S phase (S) or
G2 (G2). The G2 cells were either mock irradiated (U) or
irradiated with 6 Gy ionizing radiation, and protein preparation
made 8 h later, whence the mock irradiated cells had cycled
through mitosis and the irradiated cells were still arrested in G2.
The levels of the indicated proteins were then assayed by Western
blotting
so it is still in position to protect the nucleus from Cdk
activity.
Whilst these experiments show biochemically and
physiologically that G2 DNA damage checkpoint
arrest occurs in G2 synchronized cells in the absence
of p53 function, the lack of isogenic wildtype p53
controls makes it dicult to assess the overall validity
of these experiments. Recently, cell lines based on an
HCT116 colonic carcinoma background have been
generated in which p53 and some target genes such as
p21 and 14-3-3s have been deleted by homologous
recombination (Bunz et al., 1999; Chan et al., 1999).
We again performed pilot experiments to de®ne
appropriate doses at which HCT116 cells arrest and
recover. We found a dose of 8 Gy delayed mitotic
entry by *4 h, though greater doses lead to signi®cant
death and failure to recover over several days, as has
previously been reported when these cells were
irradiated with 12 Gy. The checkpoint kinetics of
wildtype and p537/7 in response to irradiation were
indistinguishable: both cell lines remained in G2 for an
additional *4 h following irradiation (Figure 3), then
passed through morphologically normal mitoses (data
not shown) and exited into G1 (Figure 3). We next
examined the levels of cyclins A and B1, together with
the Y15 phosphorylation status of Cdc2. In the
presence or absence of p53, Y15 phosphorylated
Cdc2 increased over the duration of the arrest.
CyclinB1 continued to accumulate in the checkpointarrested cells, which may account for the increase in
Cdc2 Y15 phosphorylation, though cyclin A levels did
not change to a signi®cant extent. The more modest
changes in protein levels compared to HeLa cells most
probably re¯ected the shorter duration of the arrest.
We also found a concomitant delay to Cdc2 kinase
activation (Figure 3c).
These observations unequivocally show that, under
conditions where the checkpoint is activated in G2
7455
Figure 2 Wee1 is nuclear, and remains nuclear following
irradiation. HCT116 cells were grown on coverslips, mock
irradiated (7IR) or irradiated with 8 Gy ionizing radiation
(+IR), and Wee1 localized by indirect immuno¯uorescence. Note
with the exception of the mitotic cell in the mock-irradiated cells
(arrowed), Wee1 is a nuclear protein that is not relocalized by
DNA damage
Oncogene
Human G2 DNA damage checkpoint
K Koniaras et al
7456
Figure 3 p53 is dispensable for the G2 DNA damage checkpoint in human cells. (a) Parental HCT116 cells (p53+/+, shaded
histograms), and a derivate harbouring targeted alleles of p53 (p537/7, open histograms) were synchronized by a double-thymidine
block and release protocol. DNA FACS pro®les of asynchronously growing cells (Asyn) and G2 synchronized cells (Oh) were
irradiated with 8 Gy of ionizing radiation (+IR) or mock irradiated (7IR). Cells were then cultured for the indicated times, and
DNA pro®les assayed. Both populations of mock irradiated cells progressed through mitosis into G1 after *4 h in culture. Both
populations of irradiated cells were delayed in G2 until *6 h in culture, (i.e. an *2 h checkpoint delay) and thereafter entered
mitosis. (b) Proteins were extracted from the cells in a, and the levels of the indicated proteins were assayed by Western blotting.
The radiation induced checkpoint in both cell lines correlated with an accumulation of inactive, Y15 phosphorylated Cdc2,
accumulation of cyclin B1, and maintenance of cyclin A. The small di€erences in FACS pro®les of mock-irradiated cells represent
subtle di€erences between individual synchronizations. (c) Cdc2 kinase activity was assayed at the indicated timepoints following
irradiation in an independent timecourse (FACS pro®les shown) that had slightly di€erent kinetics to those in a. Both lines show a
radiation induced delay to Cdc2 kinase activation
Oncogene
Human G2 DNA damage checkpoint
K Koniaras et al
cells, p53 is entirely dispensable for checkpoint delay,
leading to the possibility that inhibiting the mechanisms that control G2 checkpoints may aid in killing
p53 mutant cells.
S20 phosphorylation and up-regulation of p53 does not
occur in G2 cells
A key event in the p53-dependent signalling of a G1
DNA damage checkpoint is the accumulation of p53
protein levels in response to DNA damage (Oren,
1999). Interestingly, during the course of our experiments we observed that, while asynchronously growing
cells (predominately G1 and S) showed a radiationdependent increase in p53 protein levels, such an
increase was not observed in G2 synchronized cells
(Figure 4).
Amongst the multitude of post-translational modi®cations that a€ect the activation and stabilization of
p53, two N-terminal phosphorylation events are under
the control of known checkpoint kinases. Serine 15
(S15) is phosphorylated by the ATM and ATR kinases
(Canman et al., 1998; Khanna et al., 1998; Tibbetts et
al., 1999), and although initial data implicated this
event in the stabilization of p53, it has subsequently
been shown that phosphorylation of S20 by Cds1 (also
known as Chk2) is the critical event to inhibit the
interaction of p53 with the E3 ubiquitin ligase Mdm2
(Chehab et al., 2000). We reasoned that the inability of
p53 to accumulate in response to DNA damage in G2
synchronized cells might result from the absence of one
or more N-terminal phosphorylation events.
Using phospho-speci®c antibodies in extracts from
independently synchronized cells, we observed that S15
Figure 4 Radiation-induced stabilization of p53 does not occur
in G2 cells. Asynchronously growing (A) or G2 synchronized
(G2) HCT116 cells were irradiated with 8 Gy of ionizing
radiation at t=0 h, and p53 levels assayed by Western blotting
at the indicated timepoints
phosphorylation was not cell cycle speci®c and
increased in both asynchronous and G2 cells in
response to irradiation. This is consistent with the
widespread roles for ATM and ATR in DNA damage
responses. Conversely, while an increase in S20
phosphorylation was observed in asynchronously
growing cells following irradiation, no such e€ect was
seen in G2 synchronized cells (Figure 5). The lack of
upregulation of p53 in G2 is consistent with the
observations that p53 is, minimally, dispensable for
this arrest.
7457
Abrogation of G2 checkpoint arrest by dominant-negative
chk1
Pathways involving p53 are essentially universally
mutated in human cancer. The independence of p53mediated G1 checkpoints inhibiting Cdk2 through p21,
and G2 checkpoints resulting from inhibition of Cdc2
by Y15 phosphorylation, makes the G2 checkpoint an
attractive target for drug development to treat p53
mutant tumours. Data in ®ssion yeast indicate that the
only function for the Chk1 protein kinase is in
signalling a DNA damage-mediated checkpoint arrest,
whereas other proteins that act upstream from Chk1
are implicated in other cellular processes including
telomere maintenance and DNA replication (O'Connell
Figure 5 Radiation-induced phosphorylation of p53 on S20 does
not occur in G2 cells. Asynchronously growing (A), or G2
synchronized (G2) HCT116 cells were irradiated with 8 Gy of
ionizing radiation and t=0 h, and protein samples were prepared
from one aliquot of each immediately. Two hours later, irradiated
asynchronous cells (A+IR), G2 mock irradiated cells (G2) and
G2 irradiated cells (G2+IR) were harvested and protein extracts
made. The levels of total p53, and S15 or S20 phosphorylated p53
were assayed by Western blotting. Note that S15 phosphorylation
does not vary between irradiated samples (when corrected for
total p53 levels), though only residual S20 phosphorylation,
probably catalyzed by Cds1/Chk2, is seen in the G2+IR sample,
possibly due to the residual G1/S peak shown in the FACS
pro®les below, and correlates with the small degree of p53
stabilization (n.b. Figure 4)
Oncogene
Human G2 DNA damage checkpoint
K Koniaras et al
7458
Oncogene
et al., 2000). This unique function of Chk1, together
with the amenity of serine-threonine protein kinases to
pharmacological inhibition makes Chk1 a key molecule
for such drug development programs.
Is human Chk1 required for G2 checkpoints in a
similar manner to its homologue in ®ssion yeast? The
broad-spectrum kinase inhibitor UCN-01 (7-hydroxystuarosporine) has been shown to inhibit Chk1 in vitro
(Graves et al., 2000), and whilst it also abrogates
checkpoint responses of cells in culture, it is dicult to
attribute all its e€ects simply to Chk1 inhibition.
Another related inhibitor of Chk1, SB-218078, has
been developed (Jackson et al., 2000), though similar
arguments regarding in vivo speci®city apply, and
knock-out mice have not been informative due to
embryonic lethality (Liu et al., 2000). We therefore
chose to more speci®cally address the requirement for
Chk1 in human G2 checkpoint signalling using overexpression of a dominant negative mutant (D148E),
which we had shown to phenocopy deletion of chk1 in
®ssion yeast (K Koniaras and M O'Connell, unpublished). A similar approach has been used to show G2
checkpoint function for ATR.
Stable HeLa-based cell lines were selected that
overexpressed dominant-negative Chk1, and two were
retained for further analysis (Figure 6). Although both
lines grew adequately in culture, a higher proportion of
spontaneous apoptosis was observed (data not shown).
However, attempts to synchronize these cells with a
double-thymidine block and release protocol greatly
increased cell death, possibly due to an S-M coordination defect as seen in Xenopus (Kumagai et al., 1998).
We therefore, by necessity, resorted to nocodazole
trapping experiments in asynchronously growing cells
to assess the checkpoint status of these clones. In the
presence of the topoisomerase II poison Etoposide,
which causes G2 arrest in HeLa cells, both clones
expressing dominant negative Chk1 continued to enter
mitosis (Figure 6b). In similar experiments using
ionizing radiation (Figure 6c), over the ®rst 6 h
following irradiation, again a signi®cantly higher
proportion of cells expressing dominant negative
Chk1 continued to enter mitosis compared to control
parental cells. These timepoints correspond to the cells
that were in G2 at the point of irradiation, and was
concomitant with a lesser degree of inhibition of Cdc2
kinase activity (Figure 6d). In latter timepoints, the
di€erence was less signi®cant, perhaps due to Chk1independent mechanisms of delay in G1/S, or to
completion of DNA repair. These data are consistent
with Chk1 being required for the G2 DNA damage
checkpoint in human cells, as was suggested by the
inhibitor studies, albeit without the complications
caused by the lack of in vivo inhibitor speci®city.
On the basis of these data, we then tested the e€ect
of dominant negative chk1 expression on clonogenic
survival following irradiation. Both clones were
profoundly radiosensitive compared to the HeLa
parental cells (Figure 6e), and colonies that did grow
generally contained less cells. Thus, the checkpoint
defects caused by inhibition of Chk1 manifest as a
signi®cantly radiosensitizing event. These data con®rm
the model that Chk1 may be a prime anti-cancer drug
target in cells that lack p53 function, perhaps in
combination with radiation or genotoxic chemotherapy.
Discussion
The last decade has seen an enormous increase in our
understanding of the molecular mechanisms that
regulate cell growth and division. In more recent years
there has been a concerted attempt to exploit this
knowledge to design more e€ective anti-cancer treatments. The next generation of anti-cancer treatments
will speci®cally target individually characterized cancer
cells by exploiting a molecular `Achilles heel'. Such a
strategy would ideally leave normal somatic cells
relatively unscathed, which would in turn reduce the
severity of side e€ects encountered using adjuvant
chemotherapy and/or radiotherapy.
Inactivation of the p53 tumour suppressor, leading
to the loss of G1/S checkpoints, is a key ingredient for
neoplastic transformation in the majority of human
cancers. Unfortunately, another key ingredient in the
ability of cancer cells to proliferate even in the face of
genomic instability is the continued presence of a G2/
M checkpoint. This prevents cancer cells from
accumulating the types of destabilizing chromosomal
aberrations that would eventually prevent cells from
replicating successfully (Levine, 1997). Hence, an
attractive strategy for anti-cancer therapy would both
exploit the absence of a G1/S checkpoint and
abrogate the remaining G2/M checkpoint to kill
cancer cells.
Indeed, several reports have identi®ed novel compounds that are capable of inhibiting the ability of cells
lacking functional p53 to undergo G2/M arrest (Curman et al., 2001; Jackson et al., 2000; Roberge et al.,
1998). Such studies have demonstrated that a much
lower dose of radiation or a chemotherapeutic agent is
required to e€ectively kill the cells. In contrast, normal
cells with functional p53 are not a€ected as they retain
the ability to undergo a transient G1 arrest in response
to the treatments. Often, however, the pathways
a€ected by the compounds are not known, and may
not necessarily target any one molecule in particular.
This study has taken a twofold approach to
addressing the molecular mechanisms behind a DNA
damage-induced G2/M arrest in order to more
accurately determine which molecule(s) involved would
present a logical target for rational drug design aimed
at inhibiting G2/M phase arrest. First, we wanted to
investigate more closely the molecular events that occur
in cells that lack functional p53, but are known to
retain the ability to undergo G2/M arrest in response
to DNA damage. Speci®cally, unlike previous studies,
we were interested in what would happen to such cells
if they were synchronized to the G2 phase of the cell
cycle when subject to DNA damage. This protocol was
employed so that the regulation of a predominant Cdk,
Human G2 DNA damage checkpoint
K Koniaras et al
7459
Figure 6 Abrogation of the G2 DNA damage checkpoint by expression of dominant negative Chk1. HeLa cells were stably
transfected with plasmids expressing a dominant negative, HA-tagged Chk1 cDNA. Two clones (Clone 1 and Clone 2) were selected
for further analysis. (a) Western blot analysis using anti-Chk1 antibodies (upper panel) showed overexpression of Chk1, and with
12CA5 (lower panel) showed expression of the dominant negative allele. Reactivity to a non-speci®c band (*) is included as a
loading control. (b) Cells were incubated in the presence of 2 mM Etoposide for 3 h, and then nocodazole was added to trap cells
entering mitosis. After a further 24 h, mitotic cells were assayed by FACS using MPM-2 antibodies, and the per cent MPM-2
positive normalized against cultures without Etoposide. Data is mean+s.d. (c) The indicated lines were irradiated with 6 Gy
ionizing radiation at time = 0, or mock irradiated, and then cultured in the presence of nocodazole. Cells were then assayed by
FACS for the percentage of MPM-2 positive cells. Data represents the percentage of MPM-2 positive cells in the mock-irradiated
population, divided by that for the irradiated cultures. Data points are mean+s.d. (d) The indicated lines were irradiated with 6 Gy
ionizing radiation at time = 0, or mock irradiated, and then cultured in the presence of nocodazole. Cdc2 kinase activity was
assayed, and is displayed as a percentage of that in mock irradiated cultures. (e) Clonogenic survival assays for the indicated strains
were performed as described in Materials and methods. Data points are mean+s.d.
Cdc2, could be studied without the confusing e€ects of
upstream regulation mediated through Cdk2. Initially,
we used HeLa cells, which not only lack p53 due to
enforced degradation due to the presence of the HPVE6 oncoprotein, but are known to retain the ability to
undergo G2/M arrest. An increase in Y15-phosphorylated Cdc2, one of the de®nitive markers indicating
G2/M arrest, along with an increase in cyclins A and
B1, was observed in G2 irradiated cells. This indicates
that the upstream signalling pathways leading to
phosphorylation of Cdc2 were intact in these cells.
Even though p53 is continually degraded in HeLa
cells by HPV-E6, there is still a small amount of the
protein that is expressed. We could not discount the
possibility that this small amount of p53 could be
sucient to a€ect the signalling pathways culminating
in Y15 phosphorylation of Cdc2. Also, the e€ect of
this small amount of p53 leading to the induction of
Oncogene
Human G2 DNA damage checkpoint
K Koniaras et al
7460
Oncogene
G2 arrest independently of Cdc2 phosphorylation had
to be taken into account. We thus sought another
model system that would provide a clearer picture of
whether p53 is involved in some aspect of G2 arrest.
To this end we used a HCT116 human colon
carcinoma cell line in which both p53 alleles had
sequentially been knocked out by homologous recombination (Bunz et al., 1999). In addition, the wildtype
parental HCT116 cell line provided an isogenic control
for these experiments. Both cell lines, when irradiated
in G2, arrested for the same period of time, showed
similar patterns of Cdc2 Y15 phosphorylation, concomitant with a similar increase in cyclin B1 expression. Therefore, under these conditions, the presence or
absence of p53 does not appear to either enhance or
hinder G2 arrest, nor the molecular events associated
with it. Furthermore, we observed that, in wildtype
HCT116 cells, p53 levels did not increase in response to
DNA damage whilst in G2; this appears to be
attributed to the inability of p53 to undergo
phosphorylation at S20 in G2, which promotes
dissociation of the p53-Mdm2 complex and allows
for the stabilization of p53 (Chehab et al., 2000; Hirao
et al., 2000; Shieh et al., 2000). The lack of signalling
to p53 in G2, together with its dispensable nature for
the G2 checkpoint arrest, suggests p53 is not a major
player in human G2 DNA damage checkpoints.
Previous reports have suggested that one of the
mechanisms by which p53 might play a role in G2
arrest is through its ability to down-regulate genes such
as cyclin B1 (Muschel et al., 1993). These reports have
relied on enforced overexpression of p53 or high-level
irradiation, each leading to a sustained arrest of up to
several days. Under such conditions of sustained arrest,
studies with the HCT116-based cell lines implicate p53
in its maintenance (Bunz et al., 1999). While not
discounting the possibility that such an event may
occur in a physiological system, we have shown here
that under conditions where cells arrest transiently and
recover to pass through mitosis, the p53-independent
arrest was accompanied by cyclins A and B1
accumulation. The increased pool of inactive mitotic
cyclin/cdk complexes, otherwise known as pre-MPF,
which accumulate during this G2 delay results in the
accumulation of Y15 phosphorylated Cdc2. Coupling
these observations with the checkpoint-abrogating
e€ects of expressing non-phosphorylatable mutants of
Cdc2 (Blasina et al., 1997; Jin et al., 1996) highlights
the importance of the checkpoint to ensure inactivation
of accumulating pre-MPF. Many studies have focused
on inactivation of Cdc25C by Chk1 phosphorylation as
a mechanism of checkpoint arrest. However, the
absence of the relevant site in murine Cdc25C (Ogg
et al., 1994), and the lack of phenotype of Cdc25C
nullizygous mice (Chen et al., 2001) suggest that this is
insucient to mediate a G2 arrest. In ®ssion and
budding yeasts, Wee1 protein levels are upregulated
due to inhibition of proteolysis following DNA
damage checkpoint activation (Leu and Roeder, 1999;
Raleigh and O'Connell, 2000). In our experiments,
however, Wee1 levels were upregulated by the
synchronization protocol, which is based on evoking
the DNA replication checkpoint. Studies in Xenopus
have shown the replication checkpoint leads to Wee1
accumulation through inhibition of Cdc34-dependent
ubiquitination (Michael and Newport, 1998); it is
probable that similar mechanisms may occur in human
cells. Cells arrested in G2 with Etoposide also showed
high levels of Wee1 despite not having been through
the synchronization protocol (data not shown). The
nuclear pools of Wee1, which we have shown to
remain nuclear following irradiation, may be another
important e€ector, as seen in ®ssion yeast. We have
attempted to test this with constitutive and tetregulated expression of kinase dead Wee1 mutants,
but have been unable to gain evidence of dominant
negative e€ect over endogenous Wee1 (K Koniaras and
M O'Connell, unpublished). Recent studies with
tyrosine kinase inhibitors show that Wee1 activity is
required for G2 arrest following DNA damage (Alan J
Kraker, personal communication; Kraker et al., 2000),
and so maintaining or increasing pools of nuclear
Wee1 may be an important aspect of human G2
checkpoints. Cdc25B, which is also phosphorylated by
Chk1 in vivo (Sanchez et al., 1997), may also be an in
vitro target, and although extensive screens in yeast
have not found other G2 checkpoint e€ectors, other
yet to be identi®ed human proteins may contribute to
the p53-independent G2 arrest in response to DNA
damage.
Targeting core cell cycle regulators such as Wee1 or
Cdc25 to modulate a G2 checkpoint arrest, however,
would be likely to have profound side e€ects in a
clinical setting. Using a dominant negative Chk1
mutant, however, we have shown here both checkpoint
abrogation, and most importantly, clonogenic radiosensitization of HeLa cells, without dramatic e€ects on
cell proliferation in the absence of extrinsic DNA
damage. These observations validate Chk1 as a target
for drug development, particularly when p53 function
is absent or diminished by viral oncoproteins.
We propose that human DNA damage checkpoints,
both of which are essential in maintaining genomic
stability, consist of at least two independent and nonoverlapping signalling pathways (Figure 7). In G1 and
S phase, p53 plays a crucial role to ensure inhibition of
cyclin/Cdk2 complexes, which may extend into early
G2 phase. At least one important e€ector of p53
function is p21. In later G2, where Cdc2 is the
predominant Cdk, Chk1 dependent signalling to ensure
maintenance of the G2 state through Y15 phosphorylation of Cdc2 is the key event. Although the adjacent
threonine residue, T14, is also phosphorylated in
inactive cyclin A, B/Cdc2 complexes, the localization
of the kinase responsible, Myt1, to the cytoplasm
(Booher et al., 1997; Fattaey and Booher, 1997; Liu et
al., 1997) makes it an unlikely e€ector of DNA damage
signalling, though there is no data to disprove this. The
almost universal inactivation of G1/S checkpoints in
human tumours is contrasted by the apparently
universal retention of active G2 checkpoints. These
observations, together with the observed e€ects
Human G2 DNA damage checkpoint
K Koniaras et al
7461
Figure 7 Model for major DNA damage checkpoint pathways in human cells. In G1 and S phase, p53, through p21 transactivation, is the major e€ector of cell cycle arrest. p53 is upregulated by phosphorylation on S15 and S20, the latter catalyzed by
Cds1 and prevents binding to Mdm2. The upregulated p21 prevents cell cycle progression by binding to and inhibiting CyclinE/
Cdk2 (E/K2) and CyclinA/Cdk2 (A/K2) complexes. The dashed line represents that the inhibitory e€ects over CyclinA/Cdk2 may
continue to manifest after S phase completion into early G2. In G2 cells, Cds1-mediated upregulation of p53 does not occur, and
p53 is dispensable for cell cycle arrest. In this phase of the cell cycle, ATM and ATR activate Chk1, probably through
phosphorylation of S345. Chk1 can then down regulate Cdc25 family members or ensure they remain inactive, and may upregulate
Wee1, or alternatively Wee1 activity is maintained throughout the arrest. The net e€ect of this maintenance of G2 is inhibition of
accumulating CyclinA/Cdc2 (A/C2) and CyclinB/Cdc2 (B/C2) complexes through inhibitory phosphorylation on Y15. Inactivation
of both pathways has a synergistic e€ect, greatly radiosensitizing the cells
reported here of interfering with Chk1 function,
strongly implicate the necessity of the G2 checkpoint
for the variability of p53-dependent checkpoint de®cient tumours. We eagerly await the development of
speci®c Chk1 inhibitors, which we would predict to
have enormous ecacy in clinical oncology.
Materials and methods
DNA manipulations
Human Chk1 cDNA was tagged with its 5' end with six
histidines and one haemagglutinin tag by PCR with Deep
Vent polymerase (NEB) with the following oligonucleotide:
5'-TACTG CAGACGC GTACCAT GGCACAT CACCATCATCA CCACTATCCCTA TGACGTCCCGGACTATGCAGTGCCCTTTGTGGAAGAGCTG-3'. A 3' primer of 5'CCTGCCATGAGTTGATGGAAGAATCTCTGA-3'
was
used for these reactions, and the resulting PCR products
were cloned into pLitmus28 (NEB) and the ®delity of PCR
con®rmed by sequencing. Aspartic acid 148 was mutated to
glutamic acid with the following primer 5'-CTCAAAATCTCAGAATTCGGCTTGGCAACA-3', using previously described methods (Kunkel et al., 1987), and the ®delity of
the reaction again con®rmed by sequencing. This mutation
phenocopies the deletion of chk1 in ®ssion yeast (K Koniaras
and M O'Connell, unpublished). The remaining portion of
Chk1 was added to this construct, and the inserts were then
subcloned into pIRES-Puro (Hobbs et al., 1998).
Cell culture
HeLa cells, and stably transfected derivatives, were propagated in DMEM (Sigma) supplemented with 10% FCS (CSL)
and 100 units/ml of penicillin/streptomycin in 5% CO2.
HCT116 and HCT116-p537/7 cells (a kind gift of Dr B
Vogelstein) were propagated in McCoys medium (Sigma)
supplemented with 10% FCS (CSL) and 100 units/ml of
penicillin/streptomycin in 5% CO2. For synchronizations, a
double-thymidine block and release protocol was employed.
Brie¯y, sub-con¯uent cells were incubated in the presence of
2 mM thymidine (Sigma) for 17 h, allowing the majority of
cells to accumulate in early S phase. Cells were then washed
with PBS and incubated in medium without thymidine for 9 h
to allow release into the cell cycle. A second S phase block was
imposed by a further 14-h incubation in the presence of 2 mM
thymidine. These S phase synchronized cells were again
washed with PBS, incubated with fresh medium, allowed to
progress into G2 (approximately 5 h following release, as
assessed by FACS, appearance of cyclinB1/Cdc2 complexes,
the absence of mitotic ®gures and the absence of Cdc2 kinase
activity, analysis prior to irradiation). Stably transfected HeLa
cell lines were obtained by electroporating cells with pIRESPuro based plasmids described above followed by sequential
selection in 0.4, 1, 2, 5 and 10 mg/ml Puromycin (Sigma) as
described (Hobbs et al., 1998).
Irradiation and survival assays
For irradiation timecourse experiments, subcon¯uent synchronized or asynchronously growing cells were irradiated in
medium with a 137Cs source at a dose rate of 0.75 Gy/minute.
Oncogene
Human G2 DNA damage checkpoint
K Koniaras et al
7462
For nocodazole trapping experiments, following irradiation
cells were incubated in medium containing 150 ng/ml
nocodazole. For survival assays, cells were irradiated,
harvested and replated at varying densities depending on
dose. Colonies were allowed to form over 4 weeks, whence
surviving colonies were ®xed and stained in neutral formalin/
0.01% Crystal violet. Colonies were counted from 6 ± 12
dishes, and the data normalized to unirradiated cells, which
corrects for varying plating eciencies between lines.
FACS analysis
For analysis of DNA content, cells were harvested with
trypsin, washed in PBS and ®xed with 70% ethanol. Fixed
cells were rehydrated with PBS, and stained with Propidium
Iodide (PI) as described. To assay the fraction of cells in
mitosis, cells were ®xed in 0.5% formaldehyde/PBS for
15 min, post-®xed in 70% ethanol and stained with the
MPM-2 monoclonal antibody (1 : 2000, a kind gift of Dr P
Rao, University of Texas, USA), and with FITC-conjugated
anti-mouse IgG secondary antibodies (Sigma) diluted 1 : 100.
Cells were counter stained with PI. FACS analysis was
carried out with a Becton Dickinson FACSCalibur, and data
analysed with CellQuest software. Estimates of mitotic
fractions in MPM-2 stained cells were veri®ed by direct
counting on an epi¯uorescence microscope.
Protein extractions, kinase assays and Western blotting
For protein analysis cells were harvested with trypsin, washed
in PBS and snap frozen in liquid nitrogen. Extracts were
prepared in 50 mM Tris pH 8, 150 mM NaCl, 1 mM EDTA,
0.5% NP-40. Electrophoresis and Western blotting was
performed as previously described (O'Connell et al., 1997).
Antigens were detected with horseradish peroxidase coupled
secondary antibodies and ECL detection reagents according
to the manufacturer's speci®cations (Amersham). The following primary antibodies were obtained from the following
providers and used at the indicated concentrations: Cdc2
(Santa Cruz no SC-954) 1 : 1000, Cdc2-Y15 (Cell Signalling
Technologies no 9111) 1 : 100, CyclinA (Santa Cruz no SC751) 1 : 500, CyclinB1 (Santa Cruz no SC-752) 1 : 100, Wee1
(Santa Cruz no SC-325) 1 : 100, Chk1 (Santa Cruz no SC7234), 12CA5 (a gift of Dr R Pearson, PMCI) 1 : 500, p53
(Ab-6, Calbiochem no OP43) 1 : 2000, p53-S15 (Cell Signalling Technologies no 9284) 1 : 100, p53-S20 (Cell Signalling
Technologies no 9287) 1 : 100. Cdc2 activity was assayed from
Cdc2 (SC-954) immunoprecipitates as described (O'Connell et
al., 1994).
Indirect immunofluorescence
Cells were grown on poly-L-lysine coated cover slips,
irradiated with 6 Gy or mock irradiated, and ®xed in 4%
formaldehyde/PBS 1 h following irradiation. Cells were then
processed for indirect immuno¯uorescence using anti-Wee1
antibodies at 1 : 100, counter-stained with 10 mg/ml DAPI
and mounted in 50% glycerol, 1 mg/ml p-Phenylenediamine.
Cells were viewed on a Zeiss Axiophot microscope, and
images captured with a Spot2 camera.
Acknowledgements
We thank Drs B Vogelstein and P Rao for the prevision of
reagents, Alan J Kraker for communication of results prior
to their publication, and Doris Germain for critical reading
of the manuscript. This work was supported by grants
from the NHMRC. A Cuddihy is a Fellow, and M
O'Connell a Scholar, of the Leukemia and Lymphoma
Society.
References
Blasina A, Paegle ES and McGowan CH. (1997). Mol. Cell.
Biol., 8, 1013 ± 1023.
Booher RN, Holman PS and Fattaey A. (1997). J. Biol.
Chem., 272, 22300 ± 22306.
Brown EJ and Baltimore D. (2000). Genes Dev., 14, 397 ±
402.
Bunz F, Dutriaux A, Lengauer C, Waldman T, Zhou S,
Brown JP, Sedivy JM, Kinzler KW and Vogelstein B.
(1999). Science, 282, 1497 ± 1501.
Canman CE, Lim DS, Cimprich KA, Taya Y, Tamai K,
Sakaguchi K, Appella E, Kastan MB and Siliciano JD.
(1998). Science, 281, 1677 ± 1679.
Chan TA, Hermeking H, Lengauer C, Kinzler KW and
Vogelstein B. (1999). Nature, 401, 616 ± 620.
Chehab NH, Malikzay A, Appel M and Halazonetis TD.
(2000). Genes Dev., 14, 278 ± 288.
Chen MS, Hurov J, White LS, Woodford-Thomas T and
Piwnica-Worms H. (2001). Mol. Cell. Biol., 21, 3853 ±
3861.
Cliby WA, Roberts CJ, Cimprich KA, Stringer CM, Lamb
JR, Schreiber SL and Friend SH. (1998). EMBO J., 17,
159 ± 169.
Curman D, Cinel B, Williams DE, Rundle N, Block WD,
Goodarzi AA, Hutchins JR, Clarke PR, Zhou BB, LeesMiller SP, Andersen RJ and Roberge M. (2001). J. Biol.
Chem., 276, 17914 ± 17919.
Oncogene
Dunphy WG. (1994). Trends Cell Biol., 4, 202 ± 207.
El-Deiry W, Tokino T, Velculescu VE, Levy DB, Parsons R,
Trent JM, Lin D, Mercer WE, Kinzler KW and Vogelstein
B. (1993). Cell, 75, 817 ± 825.
Elledge SJ. (1996). Science, 274, 1664 ± 1672.
Fattaey A and Booher RN. (1997). Prog. Cell Cycle Res., 3,
233 ± 240.
Graves PR, Yu L, Schwarz JK, Gales J, Sausville EA,
O'Connor PM and Piwnica-Worms H. (2000). J. Biol.
Chem., 275, 5600 ± 5605.
Guo Z, Kumagai A, Wang SX and Dunphy WG. (2000).
Genes Dev., 14, 2745 ± 2756.
Heald R, McLoughlin M and McKeon F. (1993). Cell, 74,
463 ± 474.
Hirao A, Kong YY, Matsuoka S, Wakeham A, Ruland J,
Yoshida H, Liu D, Elledge SJ and Mak TW. (2000).
Science, 287, 1824 ± 1827.
Hobbs S, Jitrapakdee S and Wallace JC. (1998). Biochem.
Biophys. Res. Comm., 252, 368 ± 372.
Hunter T and Pines J. (1994). Cell, 79, 573 ± 582.
Jackson JR, Gilmartin A, Imbrugia C, Winkler JD, Marshall
LA and Roshak A. (2000). Cancer Res., 60, 566 ± 572.
Jin P, Gu Y and Morgan DO. (1996). J. Cell Biol., 134, 963 ±
970.
Kastan MB, Onyekwere O, Sidransky D, Vogelstein B and
Craig RW. (1991). Cancer Res., 51, 6304 ± 6311.
Human G2 DNA damage checkpoint
K Koniaras et al
Khanna KK, Keating KE, Kozlov S, Scott S, Gatei M,
Hobson K, Taya Y, Gabrielli B, Chan D, Lees-Miller SP
and Lavin MF. (1998). Nat. Genet., 20, 398 ± 400.
Kraker AJ, Choy DF, Boushelle SL, Moore CW, Garrett
MD and Booher RN. (2000). Proc. Am. Assoc. Cancer
Res., 41, 468.
Kumagai A, Guo Z, Emani KH, Wang SX and Dunphy WG.
(1998). J. Cell Biol., 142, 1559 ± 1569.
Kunkel TA, Roberts JD and Zakour RA. (1987). Methods
Enzymol, 154, 367 ± 382.
Lengauer C, Kinzler KW and Vogelstein B. (1998). Nature,
396, 643 ± 649.
Leu JY and Roeder GS. (1999). Mol. Cell, 4, 805 ± 814.
Levine AJ. (1997). Cell, 88, 323 ± 331.
Liu F, Stanton JJ, Wu Z and Piwnica-Worms H. (1997). Mol.
Cell. Biol., 17, 571 ± 583.
Liu Q, Guntuku S, Cui XS, Matsuoka S, Cortez D, Tamai K,
Luo G, Carattini-Rivera S, DeMayo F, Bradley A,
Donehower LA and Elledge SJ. (2000). Genes Dev., 14,
1448 ± 1459.
Lopez-Girona A, Kanoh J and Russell P. (2001). Curr. Biol.,
11, 50 ± 54.
Masui Y and Wang P. (1998). Biol. Cell, 90, 537 ± 548.
Michael WM and Newport J. (1998). Science, 282, 1886 ±
1889.
Muschel RJ, Zhang HB and McKenna WG. (1993). Cancer
Res., 53, 1128 ± 1135.
Nurse P. (1990). Nature, 344, 503 ± 508.
O'Connell MJ, Norbury C and Nurse P. (1994). EMBO J.,
13, 4926 ± 4937.
O'Connell MJ, Raleigh JM, Verkade HD and Nurse P.
(1997). EMBO J., 16, 545 ± 554.
O'Connell MJ, Walworth NC and Carr AM. (2000). Trends
Cell Biol., 10, 296 ± 303.
Ogg S, Gabrielli B and Piwnica-Worms H. (1994). J. Biol.
Chem., 269, 30461 ± 30469.
Oren M. (1999). J. Biol. Chem., 274, 36031 ± 36034.
Parker LL, Atherton-Fessler S, Lee MS, Ogg S, Falk JL,
Swenson KI and Piwnica-Worms H. (1991). EMBO J., 10,
1255 ± 1263.
Poon RYC, Jiang W, Toyoshima H and Hunter T. (1996). J.
Biol. Chem., 271, 13283 ± 13291.
Raleigh JM and O'Connell MJ. (2000). J. Cell Sci., 113,
1727 ± 1736.
Rhind N, Furnari B and Russell P. (1997) Genes Dev., 11,
504 ± 511.
Roberge M, Berlinck RG, Xu L, Anderson HJ, Lim LY,
Curman D, Stringer CM, Friend SH, Davies P, Vincent I,
Haggarty SJ, Kelly MT, Britton R, Piers E and Andersen
RJ. (1998). Cancer Res., 58, 5701 ± 5706.
Sanchez Y, Wong C, Thoma RS, Richman R, Wu Z,
Piwnica-Worms H and Elledge SJ. (1997). Science, 277,
1497 ± 1501.
Schmidt-Kastner PK, Jardine K, Cormier M and McBurney
MW. (1998). Oncogene, 16, 3003 ± 3011.
Sherr CJ and Roberts JM. (1999). Genes Dev., 13, 1501 ±
1512.
Shieh SY, Ahn J, Tamai K, Taya Y and Prives C. (2000).
Genes Dev., 14, 289 ± 300.
Taylor WR and Stark GR. (2001). Oncogene, 20, 1803 ± 1815.
Tibbetts RS, Brumbaugh KM, Williams JM, Sarkaria JN,
Cliby WA, Shieh SY, Taya Y, Prives C and Abraham RT.
(1999). Genes Dev., 13, 152 ± 157.
White E. (1996). Genes Dev., 10, 1 ± 15.
Zhao H and Piwnica-Worms H. (2001). Mol. Cell Biol., 21,
4129 ± 4139.
7463
Oncogene