Download Disorders in cell circuitry during multistage

Carcinogenesis vol.21 no.5 pp.857–864, 2000
COMMENTARY
Disorders in cell circuitry during multistage carcinogenesis: the
role of homeostasis
I.Bernard Weinstein
Herbert Irving Comprehensive Cancer Center and Departments of Medicine,
Genetics and Development and Public Health, Columbia University College
of Physicians and Surgeons, 701 West 168th Street, New York, NY 10032,
USA
The multistage process of carcinogenesis involves the
progressive acquisition of mutations, and epigenetic
abnormalities in the expression, of multiple genes that have
highly diverse functions. An important group of these genes
are involved in cell cycle control. Thus, cyclin D1 is
frequently overexpressed in a varety of human cancers.
Cylin D1 plays a critical role in carcinogenesis because (i)
overexpression enhances cell transformation and tumorigenesis, and enhances the amplification of other genes, and
(ii) an antisense cyclin D1 cDNA reverts the malignant
phenotype of carcinoma cells. Therefore, cyclin D1 may be
a useful biomarker in molecular epidemiology studies, and
inhibitors of its function may be useful in both cancer
chemoprevention and therapy. We discovered a paradoxical
increase in the cell cycle inhibitors protein p27Kip1 in
a subset of human cancers, and obtained evidence for
homeostatic feedback loops between cyclins D1 or E and
p27Kip1. Furthermore, derivatives of HT29 colon cancer
cells with increased levels of p27Kip1 showed increased
sensitivity to induction of differentiation. This may explain
why decreased p27Kip1 in a subset of human cancers is
associated with a high grade (poorly differentiated) histology and poor prognosis. Agents that increase cellular levels
of p27Kip1 may, therefore, also be useful in cancer therapy.
Using an antisense Rb oligonucleotide we obtained
evidence that the paradoxical increase in pRb often seen
in human colon cancers protects these cells from growth
inhibition and apopotosis. On the basis of these, and
other findings, we hypothesize that homeostatic feedback
mechanisms play a critical role in multistage carcinogenesis. Furthermore, because of their bizarre circuitry,
cancer cells suffer from ‘gene addiction’ and ‘gene hypersensitivity’ disorders that might be exploited in both cancer
prevention and chemotherapy.
The current paradigm of multistage carcinogenesis
A variety of experimental and clinical studies carried out
during the past century established the principle that cancers
develop through a multistage process which can encompass
an appreciable fraction of the lifespan of the species (1–3).
Within the past few decades astounding progress has been
made in our understanding of the cellular, biochemical and
molecular genetic events that occur during this multistage
process. A current paradigm is that this stepwise process
Abbreviations: CDI, CDK inhibitor protein; CDK, cyclin-dependent kinase.
© Oxford University Press
reflects the progressive acquisition of activating mutations in
dominant acting growth enhancing genes (oncogenes) and
inactivating recessive mutations in growth inhibitory genes
(tumor suppressor genes) (4). It is also apparent that epigenetic
abnormalities in the expression of these genes also play an
important role in carcinogenesis (1–3).
Following the initial discovery of oncogenes, ~20 years ago,
it appeared that there might be a small number of such genes.
However, since then over 100 oncogenes and at least 12 tumor
suppressor genes have been identified, and the list keeps
growing (5–7). Moreover, a single colon cancer cell frequently
contains defined mutations in multiple genes (four or more)
plus numerous less well defined mutant and/or aberrantly
expressed genes, as well as gross chromosomal abnormalities.
Indeed, in human colon tumors ~25% of all loci show loss of
heterozygosity, and in cancer cells with defects in DNA
mismatch repair thousands of loci can be mutated in a single
cancer cell (6,8,9). Presumably, these widespread changes
reflect the deleterious effects of mutagens, both exogenous
and endogenous, as well as various types of genomic instability
acquired during tumor development. In the subset of familial
cancers some of these mutations, usually in tumor suppressor
genes, are inherited, thus enhancing the multistage process of
carcinogenesis.
Because of the large number and diverse functions of
the known oncogenes and tumor suppressor genes we have
developed a classification scheme which is based on their
specific biochemical functions (Table I) (10). The genes are
divided into two broad functional categories, those that control
intracellular regulatory circuitry and those that influence cell
surface and extracellular functions. The first category is
further divided into four subcategories: (1) genes that play a
role in the responses of cells to external growth stimuli
(i.e. genes that encode growth factors, cellular receptors,
coupling proteins, and protein kinases that transduce information across the cytoplasm to the nucleus) and nuclear
transcription factors that then increase or repress the expression
of specific genes; (2) genes involved in DNA replication and
repair; (3) genes involved in cell cycle control, including
checkpoint functions; and (4) genes that determine cell fate,
i.e. cellular differentiation, senescence and programmed cell
death (apoptosis). Many of the oncogenes, for example ras,
are in subcategory 1. Subcategory 2 includes the DNA excision
and mismatch repair genes. Subcategory 3 includes the tumor
suppressor genes Rb and p53. Recent studies on cyclins and
cyclin-related genes and their abnormalities in cancer have
rapidly expanded this subcategory (11,12). This subject is
discussed in greater detail below. Subcategory 4 includes the
bcl-2 family of proteins that regulate apoptosis. This category
is of considerable importance since it is now apparent that the
increased proliferation of cancer cells reflects a disturbance in
the balance between de novo cell replication and terminal
857
I.B.Weinstein
Table I. Categories of genes targeted during multistage carcinogenesis
Intracellular circuitry
1. Agonist-induced signal transduction
2. DNA replication and repair
3. Cell cycle control
4. Cell fate: survival, differentiation, senescence and apoptosis
Cell surface and extracellular functions
Adhesion molecules; proteases; angiogenic factors, etc.
For additional details see text.
differentiation, senescence and apoptosis, rather than simply
increased cell replication. The second category includes
genes that influence how cells interact with the extracellular
matrix and/or neighboring cells. This category includes various
cell surface proteins, cell adhesion molecules, extracellular
proteases and angiogenesis factors. Alterations in these genes
are especially relevant to tumor cell invasion and metastasis.
There are several caveats related to this classification scheme.
Thus, some of the above-mentioned genes perform multiple
functions that extend across these categories (i.e. p53); there
is cross-talk between components in each category and between
categories, and the biological effects of some of these genes
are dependent upon the context of the specific cell type in
which it is expressed, a theme which will be further developed
later in this paper. Therefore, the classification scheme shown
in Table I is an oversimplification, nevertheless it may provide
a useful framework and highlights the diverse functions that
are perturbed during carcinogenesis.
Abnormalities in cell cycle control proteins in cancer
Subcategory 3 in Table I represents a recent set of oncogenes
and tumor suppressor genes, discovered as a result of the
recent elucidation of the specific proteins that normally regulate
the cell cycle in a variety of eukaryotic cells. As shown in
Figure 1, the orderly progression of dividing mammalian cells
through the G1, S, G2 and M phases of the cell cycle is
governed by a series of proteins called cyclins, which exert
their effects by binding to and activating a series of specific
cyclin-dependent kinases (CDKs). This process is further
modulated by the phosphorylation and dephosphorylation of
CDK proteins by specific protein kinases and phosphatases
and by a series of specific CDK inhibitor proteins (CDIs),
including p16INK4a, p21Waf1 and p27Kip1 (Figure 1B) (11,12).
In recent years it has become apparent that carcinogenesis
is frequently associated with mutations or abnormalities in the
expression of various cyclins, CDKs and CDIs, in several
types of human cancers (for review see refs 11,12). Thus, the
cyclin D1 gene, which acts at the mid-portion of the G1–S
transition, is often overexpressed in human breast, colon and
squamous carcinomas, and several other types of cancer, and
the cyclin E gene, which acts in late G1 is also overexpressed
and dysregulated in a variety of human cancers (11,12). Indeed,
increased expression of cyclin D1 is one of the most frequent
abnormalities in human cancer since it occurs in ~60% of
breast cancers, 40% of colorectal cancers, 40% of squamous
carcinomas of the head and neck and 20% of prostate cancers
(10–13). Furthermore, increased expression of cyclin D1 can
be an early event in carcinogenesis, since it is also seen in
precursor lesions of the colon, esophagous and breast (14,15
and unpublished data). It may, therefore, be a useful biomarker
in molecular epidemiology and chemoprevention studies. It is
858
Fig. 1. (A) Simplified model of the cell cycle indicating the G0 phase of
non-dividing cells, the G1 phase when cells enter the cell cycle and prepare
for DNA synthesis, which occurs in the S phase, and the G2 and M phases
in which cells prepare for and then undergo mitosis. The major cyclin–CDK
complexes acting at each phase of the cell cycle are also shown. There are
two checkpoints during the cell cycle: the G1/S checkpoint (also termed
restriction point) at which Rb and p53 exert inhibition on the G1/S
transition, and the G2/M checkpoint at which cells are also prevented from
progressing through the cell cycle until errors are corrected. (For additional
details see text and refs 11,12.) (B) Multiple mechanisms regulate cyclin/
CDK activities. This figure indicates the regulation of cyclin D1–CDK4.
Phosphorylation on a conserved Thr (Thr-172 in CDK4) and dephosphorylation
on Thr-14 and Tyr-15 are required for activation of the complex. A group of
CDIs designated p21CIP1, p27Kip1, p16INK4, and other related proteins bind
to the cyclin–CDK complex and inhibit kinase activity. Various external
factors acting through the CDIs can cause cell cycle arrest. Phosphorylation
of the Rb protein by the active cyclin D–CDK4 complex can release E2F
transcription factors, which enhances the G1–S transition and the onset of
DNA synthesis in the S phase. p27Kip1 plays a critical role by binding to
and inhibiting cyclin E/CDK2. (For additional details see text and refs
11,12.)
of interest that the APC/β-catenin pathway regulates the
expression of cyclin D1 (16,17), which may explain why cyclin
D1 is often overexpressed in colorectal cancers (11,12,14).
Amplification and overexpression of CDK4 is also seen in
human cancers (11,12). Abnormalities in the expression of
CDIs and in the retinoblastoma (Rb) gene, which plays a
crucial role in controlling the G1–S transition, are described
below. The proteins CDC25A and CDC25B, which are phos-
Disorders in cell circuitry
phatases that activate cyclin/CDK complexes, are frequently
overexpressed in human breast cancers (18), non-small-cell
cancers (19) and head and neck cancers (20). Using an MMTVCDC25B construct we have developed transgenic mice that
overexpress CDC25B in the mammary gland and found that
this increases susceptibility to induction of mammary tumors
by dimethylbenz[a]anthracene (21), thus demonstating a
synergistic interaction between a carcinogen and a cell cycle
control protein in carcinogenesis.
In mechanistic studies we demonstrated that overexpression of cyclin D1 can play a critical role in carcinogenesis
since introduction of an antisense cDNA to cyclin D1 into
esophageal or colon cancer cells reverts their phenotype
towards normal and inhibits tumorigenicity (22,23). This
finding has been extended by other investigators, to pancreatic
(24) and squamous carcinomas (25). Inhibition of cyclin D1
expression in human pancreatic cancer cells is associated with
increased sensitivity to chemotherapeutic agents (24).
We also found that overexpression of cyclin D1 can enhance
the process of gene amplification (26), and this finding has
been confirmed by others (27). Therefore, increased expression
of cyclin D1 could enhance the process of tumor progression
during multistage carcinogenesis, by causing genomic
instability. Taken together, these findings suggest that drugs
that inhibit cyclin D1 expression or activity, or the activity of
CDK4, may be useful in cancer chemoprevention or treatment
(10,22,23).
Paradoxical overexpression of tumor suppressor genes
Studies on p27Kip1
As discussed above, according to the current paradigm carcinogenesis is associated with activation of oncogenes and
decreased expression of tumor suppressor genes. Therefore,
we were surprised to find relatively high levels of expression
of the protein p27Kip1 in a series of human esophageal
cancer cell lines (28). This protein inhibits cell cycle progression by binding to and inhibiting the activities of cyclin/
CDK complexes (Figure 1B), especially cyclin E/CDK2 and,
therefore, it is a putative tumor suppressor gene (29). In
addition, genetically engineered mice with reduced expression
of p27Kip1 display increased sensitivity to carcinogen-induced
tumor formation, even if only one allele is inactivated (30).
We found that several human colon and breast cancer cell
lines also express high levels of p27K1p1, even during exponential growth, but this protein is expressed at low levels in three
normal human mammary cell lines (31–35).
Furthermore, we found that whereas in normal mammary
epithelial cells the level of the p27Kip1 protein varies during
the cell cycle, in breast cancer cell lines the level can remain
high throughout the cell cycle (34). The high level of p27Kip1
in these cancer cells is not simply an artefact of cell culture,
since we and other investigators have found that p27Kip1 is
also expressed at relatively high levels in a subset of primary
human breast and colon cancers (34–38). It is also overexpressed in small-cell carcinomas of the lung, despite their
high degree of malignancy (39).
The increased expression of p27Kip1 in cancer cells seems
paradoxical, especially because mutations in this gene have
not been found or are extremely rare in various cancers (29).
A possible explanation for the paradoxical increase in p27Kip1
in some cancer cells is that they have become refractory to
the inhibitory effects of this protein. To address this question,
Fig. 2. Schematic diagram indicating that cyclin D1 and cyclin E, when
bound to CDKs, stimulate the G1–S transition of the cell cycle. At elevated
levels they also can induce (through unknown mechanisms) an increase in
cellular levels of the p27Kip1 protein, thus providing feedback inhibition of
the activities of these cyclins.
we transfected the MCF7 human breast cancer cell line and
the MCF10F human non-tumorigenic mammary epithelial cell
line with a vector containing the p27Kip1 cDNA to obtain
derivatives that express increased levels of p27Kip1 (40). The
increased expression of p27Kip1 in derivatives of both of these
cell lines was associated with lengthening of the G1 phase, an
increase in the doubling time, a decreased saturation density and
a decreased plating efficiency. In the MCF7 cells, anchorageindependent growth and in vivo tumorigenicity were also
suppressed. These effects were associated with decreased
cyclin E-associated in vitro kinase activity, in both cell lines.
Thus, breast cancer cells are still responsive to p27Kip1mediated inhibition of cell growth despite the high basal level
of this protein. These results suggest that therapeutic strategies
that further increase the level of expression of p27Kip1 or
mimic its activity might be useful in cancer therapy (40).
Curiously, we found that cancer cell lines and tumors that
had high levels of p27Kip1 also frequently had high levels of
cyclin D1 (28, 31–35). Furthermore, ectopic overexpression
of cyclin D1 in esophageal (28) or mammary epithelial cell
lines (31) was associated with increased expression of p27Kip1,
and when an antisense cyclin D1 cDNA was introduced into
either esophageal or colon cancer cells to reduce the expression
of cyclin D1, this led to reduced levels of the p27Kip1
protein (22,23 and unpublished data). We also found that
overexpression of cyclin E in mammary epithelial cells is
associated with increased expression of p27Kip1 (33,41). Taken
together, these findings suggest the existence of a feedback
loop between cyclin D1 or cyclin E and p27Kip1, the purpose
of which is to maintain a homeostatic balance between positive
and negative regulators of the G1–S transition in the cell cycle
(Figure 2). The increased levels of p27Kip1 in cancer cells
might protect these cells from potentially toxic effects of
increased expression of cyclin D1 and or cyclin E (10,29).
This regulation of p27Kip1 appears to occur at a posttranslational level which is consistent with the fact that the
regulation of its expression is usually regulated at this level
by a ubiquitin-protesome mediated mechanism (29). There is
evidence that one mechanism by which cancer cells are
protected from the inhibitory effects of p27Kip1 is to sequester
this protein in the cytoplasm to prevent it from inhibiting
859
I.B.Weinstein
cyclin E/CDK2 in the nucleus. It should be mentioned that at
low levels of expression p27Kip1 enhances the formation of
cyclin–CDK complexes by acting as an assembly factor (29).
Although, as discussed above, a subset of human cancers
display relatively high levels of the p27Kip1 protein, including
esophageal, breast, colon and small-cell lung cancers, recent
studies indicate that another subset of human cancers displays
decreased expression of this protein, and that this decrease is
associated with high grade (poorly differentiated) tumors and
an unfavorable prognosis. This association is remarkable since
it has now been seen in a variety of human cancers including
carcinomas of the breast, colon, stomach, prostate and oral
cavity; as well as non-small cell lung carcinomas, gliomas,
endocrine tumors and lymphomas (for review see ref. 29).
Because we found a correlation between the subset of colon
cancers with high levels of p27Kip1 with well and moderately
differentiated carcinomas (35) we wondered if p27Kip1 played
a role in the differentiation of these cancers. Therefore, we
examined the effects of stably overexpressing high levels of
p27Kip1 in the human colon cancer cell line HT29, which can
be induced to undergo differentiation in response to treatment
with sodium butyrate (42). We found that the p27Kip1 overexpressor clones displayed an increase in the amount of the
p27Kip1 protein in cyclin E/CDK2 immunoprecipitates and a
corresponding decrease in cyclin E-associated kinase activity,
when compared with vector control clones, providing evidence
that the overexpressed protein was functional. Clones with a
high level of p27Kip1 displayed partial growth inhibition in
monolayer culture and a decrease in plating efficiency, even
though they expressed increased levels of the cyclin D1 protein.
Using alkaline phosphatase expression as a marker, we found
that the p27Kip1 overexpressor clones displayed a 2–3-fold
increase in sensitivity to induction of differentiation by 2 mM
sodium butyrate. In contrast with these results, derivatives of
HT29 cells that stably overexpressed p21Cip1/Waf1 displayed
decreased sensitivity to the induction of differentiation (42).
These results may explain why decreased levels of p27Kip1 in
certain human cancers are associated with high grade tumors.
They also provide further evidence that therapeutic strategies
that cause an increase in the level of p27Kip1 may be useful in
cancer therapy. Since there already exist several agents that
can increase the expression of p27Kip1 in specific cell types,
including TGFβ, IFN-β, IFN-γ, cAMP agonists and rapamycin
(29), in specific cell systems, this approach may be clinically
feasible. Furthermore, adenoviral p27Kip1 gene transfer can
induce apoptosis in several types of cancer cell lines (for
review see ref. 29).
Paradoxical increase in the Rb protein in colorectal cancer
The protein encoded by the Rb gene, pRb, normally plays a
key role as a negative regulator of the G1–S transition in the cell
cycle by binding the transcription factor E2F and preventing it
from activating the transcription of genes required for the S
phase (11,12). The Rb gene is inactivated in a variety of
human cancers, but in colorectal carcinomas there is frequently
increased expression of this gene (43,44). This is paradoxical
in view of the known role of Rb as a tumor suppressor gene.
In a recent study we compared the levels of expression
of pRb in normal human colorectal mucosa, adenomatous
polyps, and carcinomas by immunohistochemistry (44). We
found that there was a progressive increase in the expression
of pRb during the multistage process of colon carcino860
genesis. Thus, during the transition from normal mucosa to
adenomatous polyps to carcinomas there was a progressive
increase in pRb expression. In vitro studies were also done to
examine the phenotypic effects of an antisense oligodeoxynucleotide (AS-Rb) targeted to Rb mRNA in the HCT116
colon carcinoma cell line that expresses a relatively high level
of pRb (44). Treatment of HC116 cells with AS-Rb decreased
the level of pRb by ~70% and also decreased the levels of the
cyclin D1 protein and cyclin D1-associated kinase activity.
This finding is consistent with other evidence of the existence
of a feedback regulatory loop between Rb and cyclin D1
(44,45). A striking finding was that AS-Rb inhibited the growth
of HCT116 cells and induced apoptosis. Reporter assays
indicated an ~17-fold increase in E2F activity. Furthermore,
we could mimic the growth-inhibitory and apoptosis-inducing
effects of AS-Rb by simply ectopically overexpressing E2F in
HCT116 cells. These findings suggest that the increased
expression of pRb in colorectal carcinoma cells may provide
a homeostatic mechanism that protects them from growth
inhibition and apoptosis, perhaps by counterbalancing the
potentially toxic effects of excessive E2F. There is, indeed,
evidence that the majority of colon tumors have high E2F
activity (46). This could reflect the effects of activating
mutations in the k-ras oncogene, increased expression of cyclin
D1, or other factors that affect E2F levels and/or activity.
We found that transfection of our As-Rb into WI38 human
lung fibroblasts stimulated rather then inhibited growth (44),
which is consistent with previous evidence that in several cell
types pRb acts as a growth inhibitor. The seemingly paradoxical
effects found in colon cancer are not unique since there is
evidence that subsets of human bladder and breast cancers and
leukemias can also display increased expression of pRb, and
that pRb can protect bladder cancer cells, osteosarcoma cells
and hepatic carcinoma cells from apoptosis induced by various
agents (44,47). It is also apparent that, whereas E2F can act
as an oncogene in some cell systems, in others it can induce
apoptosis (44,48). Thus, the effects of altered expression of
pRb and E2F, like that of numerous other oncogenes or tumor
suppressor genes, is highly context dependent. This subject is
discussed in greater detail below. The mechanism by which
pRb expression is upregulated in some human cancers is not
known. It has been suggested that in some cases this may be
due to loss of expression of p16INK4a, since there appears to
be a homeostatic feedback regulatory loop between pRb and
p16INK4a (44,49,50).
Paradoxical increases in other inhibitors of the cell cycle
As described above, the CDI p27Kip1 is often expressed at
relatively high levels in human cancer cells. High levels of
expression of another CDI, p21WAF1, have also been seen in
some human tumors, including glial tumors (51), non-small
cell lung carcinomas (52), leiomyosarcomas (53) and breast
carcinomas (54). Curiously, in breast cancers high p21WAF1
expression was associated with high tumor grade and a poor
prognosis (54). In pancreatic cancer cells there was a correlation
between high expression of cyclin D1 and p21WAF1 (55). In
addition, cyclin D1 can induce increased expression of p21WAF1
through an E2F mechanism (56). This provides another
example of a homeostatic feedback mechanism that is retained
in many tumors.
Human tumors often display loss of expression of the CDI
and tumor suppressor p16INK4a, either because of mutations in
the gene or transcriptional silencing due to hypermethylation
Disorders in cell circuitry
Fig. 3. Examples in which a factor which enhances growth (‘Yang’) leads to
an increase in the expression of a factor that inhibits growth (‘Yin’) and
vice versa. These effects are often cell type specific. For additional details
see text and related references. For the effect of p53 on cyclin D1 see Chen
et al. (59).
(57). However, recent studies indicate that subsets of gastric
and colon cancer can display increased expression of the
p16INK4a protein (H.Yamamoto, personal communication).
High levels of expression of p16INK4 have also been seen in
human neuroblastoma cell lines (58). The significance of this
finding remains to be determined. It is of interest that loss of
expression of pRb is often associated with increased expression
of p16INK4a, suggesting the existence of a homeostatic feedback
loop between these two proteins, as discussed above (44,49,50).
Overview: the role of homeostatsis in carcinogenesis, when
Yin meets Yang
The above studies indicate that in several types of human
cancer there can be an increase in the expression of the tumor
suppressor genes p27Kip1, p21WAF1, p16INK4 or Rb. As discussed
above, this may reflect, at least in part, the existence of
homeostatic feedback loops in cell circuitry that maintain an
appropriate balance between growth promoting and growth
inhibitory factors. Figure 3 lists additional examples of what
appear to be homeostatic feedback loops in pathways of signed
transduction, i.e. examples of a ‘Yin/Yang’ phenomenon in
which a growth enhancing factor induces a growth inhibitory
factor or visa versa. In addition, it is now apparent that the
biologic effects of oncogenes and tumor suppressor genes are
highly context dependent (10,60). Examples include the ability
of an activated ras gene or the transcription factor E2F to either
enhance apoptosis or induce malignant cell transformation,
depending on the cell system (60). Furthermore, the biologic
effects of several oncogenes depend on their level of expression.
For example, moderate overexpression of cyclin D1 can
enhance cell growth but a high level of expression can be
toxic to cells (61). The biologic effects of various protein
kinases are also dependent on the cell context and their level
of expression. We encountered this phenomenon in our studies
on the biologic effects of specific isoforms of protein kinase
C (62) and their roles in signal transduction (63).
Table I illustrates the remarkable diversity in function of
the known oncogenes and tumor suppressor genes. It should
also be emphasized that many of the respective proteins interact
with each other in complex networks, rather than simple linear
pathways, that display cross-talk and negative or positive
feedback loops, analogous to electronic circuits (10,64–66),
and the various pathways of signal transduction can be thought
of as interconnecting ‘modules’ (66). Therefore, the accumulated effects of the multiple mutations in cancer cells leads to
bizarre types of circuitry, i.e. circuits which were not present
in the original parental cell, or in any other normal cell type.
As a consequence, certain proteins in a cancer cell function
within a novel context, since they are linked, either upstream
or downstream, to proteins they are not linked to in normal
cells. For similar reasons, an increase, decrease or loss of a
given protein in a tumor cell might have a different ‘meaning’
to a cancer cell than to a normal cell, and the experimental
re-introduction of a deleted protein into a cancer cell might
exert effects different from those that occur when the same
protein is present and expressed in normal cells.
This formulation can help to explain certain otherwise
unexpected experimental results, and may also provide reason
for optimism in the design of cancer-specific therapeutic
agents. It has always seemed puzzling why the introduction
of a single wild-type tumor suppressor gene, like p53, Rb or
APC, into malignant tumor cells that carry multiple mutations
can profoundly inhibit growth or induce apoptosis and/or
inhibit tumorigenicity (67–69). If, according to the current
paradigm, these cells originally evolved into a malignant tumor
through the stepwise acquisition of several mutations, then the
correction of one of these mutations should have only a small
inhibitory effect. We believe that these results reflect the
altered or bizarre circuitry of cancer cells, and refer to this
phenomenon as ‘gene hypersensitivity’ (10). In our studies on
cyclin D1 we encountered another effect which also seemed
puzzling. As mentioned above, we found that stable expression
of an antisense cyclin D1 cDNA in a human esophageal cancer
cell line, which carries an amplified cyclin D1 gene and
expresses high levels of cyclin D1, depressed cyclin D1
expression, and this was associated with a dramatic reversion
of the cells towards a more normal phenotype (22). Nevertheless, the residual level of cyclin D1 protein expression in the
reverted cells was considerably higher than in other rapidly
growing and highly tumorigenic cells in which cyclin D1 was
not amplified or overexpressed. These findings suggest that
during the original evolution of these cancer cells they became
‘addicted’ to cyclin D1 (10) and, therefore, require high levels
of this protein to maintain their cancer phenotype. A possible
explanation is that these cancer cells express relatively high
levels of proteins that counteract the effects of cyclin D1, for
example Rb or one of the CDIs. Thus, even only a partial
decrease in cyclin D1 in these cells would alter the stoichiometry between it and the respective inhibitory proteins, thus
resulting in net inhibition of cell growth (10). If this explanation
is correct, then cancer cells that are addicted to cyclin D1
might be unusually susceptible to drugs that block the action
861
I.B.Weinstein
of cyclin D1. This general model might also apply to other
genes that are amplified and/or overexpressed, or constitutively activated, in cancer cells. Indeed, it has been shown that
pancreatic cancer cells that carry a mutated and activated k-ras
gene are more dependent on the function of the k-ras gene
for growth than pancreatic cancer cells that do not carry this
mutation (70). Another example of gene addiction is the finding
that erB-2 antisense oligonucleotides inhibit the proliferation of
breast carcinomas cells with erB-2 amplification but have no
specific effect on breast cancer lines that do not have amplification of erB-2 (71). The bizarre circuitry of cancer cells and
the phenomena of gene hypersensitivity and gene addition
could be the long sought Achilles’ heal of cancer cells. Indeed,
they might explain why tumor cells are often more susceptible
to the induction of apoptosis than normal cells by some of the
currently employed cancer chemotherapy agents.
The concept of homeostasis is pervasive in biologic systems
and dates back to the 19th century physiologist Claude Bernard
who emphasized the constancy of the ‘interior milieu’ of the
body in the face of an ever-changing exterior environment.
The term itself was first used by Walter B.Cannon in the 1930s
who emphasized the role of the autonomic nervous system in
maintaining steady states within the body (72). Subsequent
studies of the endocrine system provided further examples.
With the more recent elucidation of biochemical pathways of
biosynthesis and energy metabolism and current studies on
pathways of signal transduction, the concept of homeostasis
has been extended to intracellular mechanisms. It seems likely
that the principal of homeostasis is also maintained during the
process of multistage carcinogenesis, as discussed above. This
seems reasonable since, despite its numerous abnormalities,
the cancer cell must coordinate highly complex functions in
order to survive and replicate. Therefore, the clonal evolution
theory of cancer, proposed by Nowell (73), and the current
paradigm of oncogenes and tumor suppressor genes (4),
requires modification. The multistage process of carcinogenesis
does not simply involve the step-wise activation of growthpromoting oncogenes and inactivation of growth inhibitory
tumor suppressor genes. The regulatory circuitry of the
evolving population of tumor cells must adapt to the stochastic
occurrence of these mutations, some of which might on their
own inhibit growth or cause apoptosis. Presumably this occurs
through homeostatic feedback mechanisms, like those
described above, and/or cell selection, thus maintaining a
homeostatic balance that favors optimal growth and viability.
This concept could help to explain the long latent period in
carcinogenesis and the complex and heterogenous phenotypes
of cancer cells. It also has implications with respect to novel
approaches to cancer chemoprevention and therapy, because
of the bizarre circuitry that results from these alterations and
the phenomena of gene hypersensitivity and gene addition, as
discussed above.
The concept of cancer as a global disturbance of the network
of regulatory circuitry within cells also has implications with
respect to the limitations of the current approaches used
for characterizing the genotypes and phenotypes of specific
cancers. Currently, this is often done by analyzing a few genes,
transcripts or proteins. The recent development of microarray
methods (74) and proteomics markedly expands our ability to
assess complex profiles of gene expression in cancer cells and,
therefore, are major advances. However, these methods do not
provide a dynamic view of the actual circuitry of cancer cells.
A challenging future goal is to develop novel methods to
862
assess this circuitry in living cells, and also to develop
mathematical models (for example, see refs 64–66) for
analyzing the complex networks and their interactions. Hopefully, the insights obtained from this new level of analysis will
provide even more powerful approaches to cancer prevention
and treatment.
Acknowledgements
This paper is dedicated to Anthony Dipple, an outstanding leader in carcinogenesis research and wonderful colleague. I am indebted to several members
of my laboratory who made invaluable contributions to this research. For
brevity, I have often cited previous review articles or representative research
papers. I apologize to other investigators for omitting numerous additional
pertinent references. This research was supported by NIH Grant CA63467,
AIBS grant DAMRD 17-94-J-4101, and awards from the National Foundation
for Cancer Research, the T.J.Martell Foundation and the Alma Toorock
Memorial for Cancer Research.
References
1. Weinstein,I.B. (1988) The origins of human cancer: molecular mechanisms
of carcinogenesis and their implications for cancer prevention and
treatment. Twenty-seventh G.H.A. Clowes memorial lecture. Cancer Res.,
48, 4135–4143.
2. Weinstein,I.B. (1991) Cancer prevention: recent progress and future
opportunities. Cancer Res., 51, 5080S–5085S.
3. Weinstein,I.B., Carothers,A.M., Santella,R.M. and Perera,F.P. (1995)
Molecular mechanisms of mutagenesis and multistage carcinogenesis.
In Mendelsohn,J., Howley,P.M., Israel,M.A. and Liotta,L.A. (eds) The
Molecular Basis of Cancer. W.B. Saunders, Philadelphia, PA, pp. 59–85.
4. Fearon,E.R. and Vogelstein,B. (1990) A genetic model for colorectal
tumorigenesis. Cell, 61, 759–767.
5. Park,M. (1998) Oncogenes. In Vogelstein,B. and Kinzler,K.W. (eds) The
Genetic Basis of Human Cancer. McGraw-Hill, New York, pp. 205–228.
6. Fearon,E.R. (1998) Tumor suppressor genes. In Vogelstein,B. and
Kinzler,K.W. (eds) The Genetic Basis of Human Cancer. McGraw-Hill,
New York, pp. 229–236.
7. Hanahan D. and Weinberg,R.A. (2000) The hallmarks of cancer. Cell, 100,
57–70.
8. Lengauer,C., Kinzler,K.W. and Vogelstein,B. (1998) Genetic instabilities
in human cancers. Nature, 396, 643–649.
9. Malkhosyan,S., Yasuda,J., Soto,J.L., Sekiya,T., Yokato,J. and Perucho,M.
(1998) Molecular karyotype (amplotype) of metastatic colorectal cancer
by unbiased arbitrarily primed PCR DNA fingerprinting. Proc. Natl Acad.
Sci. USA, 17, 10170–10175.
10. Weinstein,I.B., Begemann,M., Zhou,P., Han,E.K.-H., Sgambato,A.,
Doki,Y., Arber,N. and Ciaparrone,M. and Yamamoto,H. (1997) Disorders
in cell circuitry associated with multistage carcinogenesis: exploitable
targets for cancer prevention and therapy. Clin. Cancer Res., 3, 2696–2702.
11. Weinstein,I.B. and Zhou,P. (1997) Defects in cell cycle control genes in
human cancer. In Bertino,J.R. (ed.) Encyclopedia of Cancer. Academic
Press, New York, Vol. 1, pp. 256–267.
12. Sgambato,A., Flamini,G., Cittadini,A. and Weinstein,I.B. (1998)
Abnormalities in cell cycle control in cancer and their clinical implications.
Tumori, 84, 421–433.
13. Han,E.K.-H., Rubin,M.A., Lim,J.T., Arber,N., Xing,W.-Q. and
Weinstein,I.B. (1998) Cyclin D1 expression in human prostate cell lines
and primary tumors. The Prostate, 35, 95–101.
14. Arber,N., Hibshoosh,H., Moss,S.F., Sutter,T., Zhang,Y., Begg,M., Wang,S.,
Weinstein,I.B. and Holt,P. (1996) Increased expression of cyclin D1 is an
early event in multistage colorectal carcinogenesis. Gastroenterology, 110,
669–676.
15. Arber,N., Lightdale,C., Rotterdam,H., Finegold,J., Ahsan,H., Han,E.K.H., Sgambato,A., Stevens,P.D., Green,P.H.R., Neugut,A.I., Holt,P.R. and
Weinstein,I.B. (1996) Increased expression of the cyclin D1 gene in
Barrett’s Esophagus. Cancer Epidemiol. Biomarkers Prev., 3, 457–459.
16. Tetsu,O. and McCormick,F. (1999) β-Catenin regulates expression of
cyclin D1 in colon carcinoma cells. Nature, 144, 422–426.
17. Shutman,M., Zhurinsky,J., Simcha,I., Albanese,C., D’Amico,M., Pestell,R.
and Ben-Zeev,A. (1999) The cyclin D1 gene is a target of the β-catenin/
Lef-1 pathway. Proc. Natl Acad. Sci. USA, 96, 5522–5527.
18. Galaktionov,K., Lee,A.K., Eckstein,J., Draetta,G., Meckler,J., Massimo L.
and Beach,D. (1995) cdc25 phosphatases as potential human oncogenes.
Science, 269, 1575–1577.
Disorders in cell circuitry
19. Wu.,W., Fan,Y.-H., Kemp,B.L., Walsh,G. and Mao,L. (1998) Overexpresion
of cdc25A and cdc25B is frequent in primary non-small cell lung cancer
but is not associated with overexpression of c-myc. Cancer Res., 58,
4082–4085.
20. Gaspparotto,D., Maestro,R., Piccinin,S., Vukosavljevic,T., Barzan,L.,
Sulfaro,S. and Boiocchi,M. (1997) Overexpression of cdc25A and cdc25B
in head and neck cancers. Cancer Res., 57, 2366–2368.
21. Yao,Y., Slosberg,E., Wang,L., Hibshoosh,H., Zhang,Y.-J., Xing,W.-Q.,
Santella,R.M. and Weinstein,I.B. (1999) Increased susceptibility to
carcinogen-induced mammary tumors in MMTV-Cdc25B transgenic mice.
Oncogene, 18, 5159–5166.
22. Zhou,P., Jiang,W., Zhang,Y.-J., Kahn,S.M., Schieren,I., Santella,R.M. and
Weinstein,I.B. (1995) Antisense to cyclin D1 inhibits growth and reverses
the transformed phenotype of human esophageal cancer cells. Oncogene,
11, 571–580.
23. Arber,N., Doki,Y., Han,E.K.-H., Sgambato,A., Zhou,P., Kim,N.,
Delohery,T., Klein,M.G., Holt,P.R. and Weinstein,I.B. (1997) Antisense to
cyclin D1 inhibits the growth and tumorigenicity of human colon cancer
cells. Cancer Res., 57, 1569–1574.
24. Kornmann,M., Danenberg,K.D., Arber,N., Beger,H.G., Danenberg,P.V. and
Kore,M. (1999) Inhibition of cyclin D1 expression in human pancreatic
cancer cells is associated with increased chemosensitivity and decreased
expression of multiple chemoresistance genes. Cancer Res., 59, 3505–3511.
25. Sauter,E.R., Nesbit,M., Litwin,S., Klein-Szanto,J.-P., Cheffetz,S. and
Herlyn,M. (1999) Antisense cyclin D1 induces apoptosis and tumor
shrinkage in human squamous carcinomas. Cancer Res., 59, 4876–4881.
26. Zhou,P., Jiang,W., Weghorst,C.M. and Weinstein,I.B. (1996)
Overexpression of cyclin D1 enhances gene amplification. Cancer Res.,
56, 36–39.
27. Asano,K., Sakamoto,H., Sasaki,H., Ochiya,T., Yoshida,T., Ohishi,Y.,
Machida,T., Kakizoe,T., Sugimura,T. and Terada,M. (1995) Tumorigenicity
and gene amplification potentials of cyclin D1-overexpressing NIH3T3
cells. Biochem. Biophys. Res. Commun., 217, 1169–1176.
28. Doki,Y., Imoto,M., Han,E.K.-H., Sgambato,A. and Weinstein,I.B. (1997)
Increased expression of the p27kip1 protein in human esophageal cancer
cell lines that over-express cyclin D1. Carcinogenesis, 18, 1139–1148.
29. Sgambato,A.A., Cittadini,A. and Weinstein,I.B. (2000) Multiple functions
of p27Kip1 and its alterations in tumor cells. A review. J. Cell. Physiol.,
in press.
30. Fero,M.L., Randel,E., Gurley,K.E., Roberts,J.M. and Kemp,C.J. (1998)
The murine gene p27Kip1 is haplo-insufficient for tumour suppression.
Nature, 396, 177–180.
31. Han,E.K.-H., Sgambato,A., Jiang,W., Zhang,Y.-J., Santella,R.M., Doki,Y.,
Cacace,A., Schieren,I. and Weinstein,I.B. (1995) Stable overexpression of
cyclin D1 in a human mammary epithelial cell line prolongs the S-phase
and inhibits growth. Oncogene, 10, 953–961.
32. Sgambato,A., Han,E.K.-H., Zhou,P., Schieren,I. and Weinstein,I.B. (1996)
Overexpression of cyclin E in the HC11 mouse mammary epithelial cell
line inhibits growth and induces expression of the p27Kip1 inhibitor protein.
Cancer Res., 56, 1389–1399.
33. Sgambato,A., Doki,Y., Schieren,I. and Weinstein,I.B. (1997) Effects of
cyclin E Overexpression on cell growth and response to TGFβ depend on
cell context and p27KIP1 expression. Cell Growth Differ., 8, 393–405.
34. Sgambato,A., Zhang,Y.-J., Arber,N., Hibshoosh,H., Doki,Y., Ciaparrone,M.,
Santella,R.M. and Weinstein,I.B. (1997) Deregulated expression of p27Kip1
in human breast cancers. Clin. Cancer Res., 3, 1879–1887.
35. Ciaparrone,M., Yamamoto,H., Yao,Y., Sgambato,A., Cattoretti,G.,
Tomita,N., Monden,T., Rotterdam,H. and Weinstein,I.B. (1998)
Localization and expression of p27Kip1 in multistage colorectal
carcinogenesis. Cancer Res., 58, 114–122.
36. Fredersdorf,S., Burns,J., Milne,A.M., Packham,G., Fallis,L., Gillet,C.E.,
Royds,J.A., Peston,D., Hall,P.A., Hanby,A.M., Barnes,D.M., Shousha,S.,
O’Hare,M.J. and Lu,X. (1997) High level expression of p27(kip1) and
cyclin D1 in come human breast cancer cells: inverse correlation between
the expression of p27(kip1) and degree of malignancy in human breast
and colorectal cancers. Proc. Natl Acad. Sci. USA, 94, 6380–6385.
37. Porter,P.L., Malone,K.E., Heagerty,P.J., Alexander,G.M., Gatti,L.A.,
Firpo,E.J., Daling,J.R. and Robert,J.M. (1997) Expression of cell-cycle
regulators p27Kip1 and cyclin E, alone or in combination, correlate with
survival in young breast cancer patients. Nature Med., 3, 222–225.
38. Catzavelos,C., Bhattacharya,N., Ung,Y.C., Wilson,J.A., Roncari,L.,
Sandhu,C., Shaw,P., Yeger,H., Morave-Protzner,I., Kapusta,L., Franssen,E.,
Pritchard,K.I. and Slingeland,J.M. (1997) Decreased levels of the cellcycle inhibitor p27Kip1 protein: prognostic implications in primary breast
cancer. Nature Med., 3, 227–230.
39. Yatabe,Y., Masuda,A., Koshikawa,T. et al. (1998) p27Kip1 in human lung
cancers: differential changes in small cell and non-small cell carcinomas.
Cancer Res., 58, 1042–1047.
40. Sgambato,A., Zhang,Y.-J., Ciaparrone,M., Soh,J.-W., Cittadini,A.,
Santella,R.M. and Weinstein,I.B. (1998) Overexpression of p27Kip1 inhibits
the growth of both normal and transformed human mammary epithelial
cells. Cancer Res., 58, 3448–3454.
41. Sgambato,A., Han,E.K.-H., Zhou,P., Schieren,I. and Weinstein,I.B. (1996)
Overexpression of cyclin E in the HC11 mouse mammary epithelial cell
line inhibits growth and induces expression of the p27Kip1 inhibitor protein.
Cancer Res., 56, 1389–1399.
42. Yamamoto,H., Soh,J.-W., Shirin,H., Xing,W.-Q., Ciaparrone,M.,
Sgambato,A., Schieren,I., Yao,Y., Slosberg,E., Tomita,N. and
Weinstein,I.B. (1999) Comparative effects of overexpression of p27Kip1
and p21Cip1/Waf1 on growth and differentiation in human colon carcinoma
cells. Oncogene, 18, 103–115.
43. Wildrick,D.M. and Boman,B.M. (1994) Does the human retinoblastoma
gene have a role in colon cancer? Mol. Carcinog., 10, 1–7.
44. Yamamoto,H., Soh,J.-W., Klein,M.G., Zhang,L.-M., Schieren,I., Stein,C.A.
and Weinstein,I.B. (1999) Paradoxical increase in retinoblastoma protein
in colorectal carcinomas may protect cells from apoptosis. Clin. Cancer
Res., 5, 1805–1815.
45. Muller,H., Lucas,J., Schneider,A., Warthtoe,P., Bartek,J., Eilers,M. and
Strauss,M. (1994) Cyclin D1 expression is regulated by the retinoblastoma
protein. Proc. Natl Acad. Sci. USA, 91, 2945–2949.
46. Lemass,H., Ryan,E., Mathuna,P.M., Crowe,J. and O’Keane,J.C. (1998)
Elevated pRB is associated with elevated E2F-1 but not cyclin D1
or cyclin dependent kinase 4 in the colorectal carcinomas cell cycle.
Gastroenterology (Am. Gastroenterol. Assoc. Abstr.), 114, G2612.
47. Haas-Kogan,D.A., Kogan,S.C., Levi,D., Dazin,P., T’Ang,A., Fung,Y.K.
and Israel,M.A. (1995) Inhibition of apoptosis by the retinoblastoma gene
product. EMBO J., 14, 461–472.
48. Zwicker,J., Liu,N., England,K., Lucibello,F.C. and Muller,R. (1996) Cell
cycle regulation of E2F site occupation in vivo. Science, 271, 1595–1597.
49. Li.,Y., Michael,A., Nichols,J., Shay,J.W. and Xiong,Y. (1994)
Transcriptional repression of the D-type cyclin-dependent kinase inhibitor
p16 by the retinoblastoma susceptibility gene produce pRb. Cancer Res.,
54, 6078–6082.
50. Otterson,G.A., Kratzke,R.A., Coxon,A., Kim,Y.W. and Kaye,F.J. (1994)
Absence of p16INK4 protein is restricted to the subset of lung cancer
lines that retains wild-type R.B. Oncogene, 9, 3375–3378.
51. Jung,J.M., Bruner,J.M., Rusan,S., Langford,L.A., Kyritsis,A.P.,
Kjobayashi,T., Levin,V.A. and Zhang,W. (1995) Increased levels of
p21WAF1/CIP1 in human brain tumors. Oncogene, 11, 2021–2028.
52. Marchetti,A., Doglioni,C., Barbareschi,M., Buttitta,F., Pellegrini,S.,
Bertacca,G., Chella,A., Merlo,G., Angeletti,C.A., Palma,P.D. and
Bevilacqua,G. (1996) p21 RNA and protein expression in non-small cell
lung carcinomas: evidence of p53-independent expression and association
with tumoral differentiation. Oncogene, 12, 1319–1324.
53. Palazzo,J.P., Mercer,E., Kovatich,A.J. and McHugh,M. (1997)
Immunohistochemical localization of p21WAF1/CIP1 in normal, hyperplastic
and neoplastic uterine tissues. Hum. Pathol., 28, 60–66.
54. Barbarescho,M., Caffo,O., Doglioni,C., Fina,P., Marchetti,A., Buttitta,F.,
Leek,R., Morelli,L., Leonardi,E., Bevilacqua,G., Dalla Palma,P. and
Harris,A.I. (1996) p21WAF1 immunohistochemical expression in breast
carcinoma: correlations with clinicopathological data, oestrogen receptor
status. MIB1 expression, p53 gene and protein alterations and relapse-free
survival. Br. J. Cancer, 74, 208–215.
55. Gansauge,S., Gansauge,F., Ramadani,M., Stobbe,H., Rau,B., Harada,N.
and Beger,H.G. (1997) Overexpression of cyclin D1 in human pancreatic
carcinoma is associated with poor prognosis. Cancer Res., 57, 1634–1637.
56. Himaya,H., Iavarone,A., LaBaer,J. and Reeves,S. (1997) Regulated ectopic
expression of cyclin D1 induces transcriptional activation of the cdk
inhibitor p21 gene without altering cell cycle progression. Oncogene, 14,
2533–2542.
57. Herman,J.G., Merlo,J.J. and Baylin,S.B. (1996) Hypermethylationassociated inactivation indicates a tumor suppressor role for p15INK4B1.
Cancer Res., 56, 722–727.
58. Easton,J., Wei,T., Lahti,J.M. and Kidd,V.J. (1998) Disruption of the
cyclin D/cyclin-dependent kinase/INK4/retinoblastoma protein regulatory
pathway in human neuroblastoma. Cancer Res., 58, 2624–2632.
59. Chen,X., Bargpmetto,J. and Prives,C. (1995) p53 through p21 (WAF1/
CIPI), induces cyclin D1 synthesis. Cancer Res., 55, 4257–4263.
60. Evan,G. and Littwood,T. (1998) A matter of life and death. Science, 281,
1317–1326.
863
I.B.Weinstein
61. Han,E.K.-H., Ng,S.C., Arber,N., Begemann,M. and Weinstein,I.B. (1999)
Roles of cyclin D1 and related genes in growth inhibition, senescence and
apoptosis. Apoptosis, 4, 213–219.
62. Cacace,A., Ueffing,M., Phillip,A., Han,E.K.-H., Kolch,W. and
Weinstein,I.B. (1996) PKC epsilon functions as an oncogene by activating
the RAF kinase. Oncogene, 13, 2517–2526.
63. Soh,J.-W., Lee,E.H., Prywes,R. and Weinstein,I.B. (1999) Novel roles of
specific isoforms of protein kinase C in activation of the c-fos serum
response element. Mol. Cell. Biol., 19, 1313–1325.
64. Bhalla,S. and Iyengar,R. (1999) Emergent properties of networks of
biological signaling pathways. Science, 283, 381–390.
65. Kohn,K.W. (1998) Functional capabilities of molecular network
components controlling the mammalian cell cycle phase transition.
Oncogene, 16, 1065–1075.
66. Hartwell,L.H., Hopfield,J.J., Leibler,S. and Murray,A.W. (1999) From
molecular to modular cell biology. Nature, 402, C47–C52.
67. Takahashi,T., Carbone,D., Takahashi,T., Nau,M.M., Hida,T., Linnoila,I.,
Ueda,R. and Minna,J.D. (1992) Wild-type but not mutant p53 suppresses
the growth of human lung cancer cells bearing multiple genetic lesions.
Cancer Res., 52, 2340–2343.
68. Li,J., Hu,S.-X., Perng,G.-S., Zhou,Y., Xu,K., Zhang,C., Seigne,J.,
864
Benedict,W.F. and Xu,H-J. (1996) Expression of the retinoblastoma (RB)
tumor suppressor gene inhibits tumor cell invasion in vitro. Oncogene, 13,
2379–2386.
69. Morin,P.J., Vogelstein,B. and Kinzler,K. (1996) Apoptosis and APC in
colorectal tumorigenesis. Proc. Natl Acad. Sci. USA, 93, 7950–7954.
70. Aoki,K., Yoshida,T., Matsumoto,N., Ide,H., Sugimura,T. and Terada,M.
(1997) Suppression of Ki-ras p21 levels leading to growth inhibition of
pancreatic cancer cell lines with Ki-ras mutation but not those without
Ki-ras mutation. Mol. Carcinog., 20, 251–258.
71. Colomer,R., Lupu,R., Bacus,S.S. and Gelmann,E.P. (1994) erb-B2 antisense
oligonucleotides inhibit the proliferation of breast carcinoma cells with
erB-2 oncogene amplification. Br. J. Cancer, 70, 819–825.
72. Cannon,W.B. (1939) The Wisdom of the Body. W.W. Norton and Co. Inc.,
New York.
73. Nowell,P. (1976) The clonal evolution of tumor populations. Science, 194,
23–28.
74. Vishwanath,R., Iyer,V.R., Elsen,M.B., Ross,D.T., Schuler,G., Moore,T.,
Fee,J.C.F., Trent,J.M., Staude,L.M., Hudson,J., Bogusti,M.S., Lashkari,D.,
Shalon,D., Botstein,D. and Brown,P.O. (1999) The transcriptional program
in the response of human fibroblasts to serum. Science, 283, 83–87.
Received and accepted February 8, 2000