Abstract
Cell senescence is broadly defined as the
physiological program of terminal growth arrest, which can be triggered
by alterations
of telomeres or by different forms of stress.
Neoplastic transformation involves events that inhibit the program of
senescence,
and tumor cells were believed until recently to
have lost the ability to senesce. It has now become apparent, however,
that
tumor cells can be readily induced to undergo
senescence by genetic manipulations or by treatment with
chemotherapeutic drugs,
radiation, or differentiating agents.
Treatment-induced senescence, which has both similarities with, and
differences from,
replicative senescence of normal cells, was shown
to be one of the key determinants of tumor response to therapy in vitro and in vivo.
Although senescent cells do not proliferate, they remain metabolically
active and produce secreted proteins with both tumor-suppressing
and tumor-promoting activities. Expression of
tumor-promoting factors by senescent cells is mediated, at least in
part, by
senescence-associated cyclin-dependent kinase
inhibitors such as p21Waf1/Cip1/Sdi1. Clinical and
preclinical studies indicate that expression of different biological
classes of senescence-associated growth-regulatory
genes in tumor cells has significant prognostic
implications. Elucidation of the genes and regulatory mechanisms that
determine
different aspects of tumor senescence makes it
possible to design new therapeutic approaches to improving the efficacy
and
to decreasing the side effects of cancer therapy.
Senescence as an Anticarcinogenic Program of Normal Cells
Cell senescence, originally defined as
proliferative arrest that occurs in normal cells after a limited number
of cell divisions,
has now become regarded more broadly as a general
biological program of terminal growth arrest, a definition that is used
throughout this review. Cells that underwent
senescence cannot divide even if stimulated by mitogens, but they remain
metabolically
and synthetically active and show characteristic
changes in morphology, such as enlarged and flattened cell shape and
increased
granularity
(1)
. The most widely used surrogate marker with
considerable (but not absolute) specificity to senescent cells is the
SA-β-gal,
3
which is detectable by X-gal staining at pH 6.0
(2)
. SA-β-gal appears to reflect increased lysosomal mass of senescent cells
(3)
. Senescent cells also produce many ECM components
and secreted factors that affect the growth of their neighboring cells
as well as tissue organization. In particular,
paracrine factors produced by senescent cells have major effects on the
growth
and survival of tumor cells in vitro and in vivo
(4
, 5)
. Hence, senescence should not be viewed as merely
an end point in a cell’s life cycle but rather as a physiological state
determined by the homeostatic programs of a
multicellular organism.
Senescence (“growing old”) was originally described in normal human cells explanted in culture; such cells undergo a finite
number of divisions before permanent growth arrest
(6)
. This gradual process of “replicative senescence”
in human cells results primarily from the shortening and other
structural
changes of telomeres at the ends of the chromosomes
(7)
. Telomeric changes in cells undergoing replicative
senescence show similarities with DNA damage or may even directly
involve
such damage
(8
, 9)
. It is not surprising, therefore, that DNA damage
was also found to induce rapid cell growth arrest, which was
characterized
as phenotypically indistinguishable from
replicative senescence
(7
, 10)
. This “accelerated senescence,” which does not
involve telomere shortening, is also triggered in normal cells by the
expression
of mutant Ras or Raf
(11
, 12)
and by some other forms of supraphysiological mitogenic signaling
(1)
.
The key events in replicative and accelerated senescence of normal fibroblasts (the best-studied cellular system of senescence)
are schematized in Fig. 1
⇓
. Growth arrest of senescent cells is initiated
with the activation of p53. In the case of replicative senescence, p53
protein
is stabilized through the involvement of p14ARF, a tumor suppressor that sequesters the Mdm2 protein, which promotes p53 degradation
(13)
. Another protein that stimulates p53 under the
conditions of replicative and RAS-induced accelerated senescence is
promyelocytic
leukemia (PML) tumor suppressor, which regulates
p53 acetylation
(14
, 15)
. The activated p53 has multiple effects on gene
expression, the most relevant of which in regard to senescence is
transcriptional
activation of p21Waf1/Cip1/Sdi1, a pleiotropic inhibitor of different cyclin/CDK complexes
(16)
. p21 induction causes cell cycle arrest in senescent cells
(17
, 18)
. The activation of p53 and p21 in senescent cells
is only transient, and protein levels of p53 and p21 decrease after the
establishment of growth arrest. Whereas p21
expression decreases, another CDK inhibitor, p16Ink4A becomes constitutively up-regulated, suggesting that p16 may be responsible for the maintenance of growth arrest in senescent
cells
(18
, 19)
. Recent studies have implicated several positive
and negative transcription regulatory factors (ETS, ID-1, BMI-1) in the
transcriptional activation of p16 in senescent
cells
(20
, 21)
. Other CDK inhibitors, p27Kip1
(22
, 23)
and p15Ink4b
(24)
, were also shown to play a role in fibroblast
senescence. The best known (but by no means the only) mechanism for
growth
arrest induced by CDK inhibitors is the blockage of
CDK-mediated inhibitory phosphorylation of tumor suppressor protein Rb.
At the onset of senescence, Rb is converted to its
active hypophosphorylated form, which sequesters and inhibits E2F
transcription
factors that are necessary for cell proliferation.
The cellular levels of Rb decrease, however, after the onset of
senescence
in normal fibroblasts
(19)
, suggesting that Rb may not play an active role in the maintenance of the senescent phenotype.
Fig. 1.
Key events in replicative and accelerated senescence of normal fibroblasts. PML, promyelocytic leukemia.
Both replicative and accelerated
senescence are believed to be essential anticarcinogenic programs in
normal cells. Replicative
senescence imposes a limit on the total number of
divisions a cell can undergo, and it should be expected, therefore, to
interfere
with tumor growth. However, studies with mice
deficient for the enzyme telomerase, which counteracts the shortening of
telomeres,
have yielded a more complicated picture. Telomere
shortening, which occurs in telomerase-deficient mice after several
generations,
was found paradoxically to promote the rate of
spontaneous carcinogenesis, most probably because telomeric aberrations
destabilize
the genome
(7
, 25)
. On the other hand, telomerase-deficient mice were
more resistant to carcinogenesis under several conditions that increase
the rate of tumor initiation, in agreement with the
anticarcinogenic role of replicative senescence
(7)
. The tumor-suppressive function may be more
central to the program of accelerated senescence, which prevents the
outgrowth
of cells that have experienced oncogenic mutations
(such as RAS or RAF mutations) or that underwent genome-destabilizing DNA damage.
In agreement with a role of senescence in
cancer prevention, the process of carcinogenesis almost inevitably
involves one
or more events that inhibit senescence. Tumor cells
avoid replicative senescence through the up-regulation of telomerase,
or (less frequently) by using alternative
mechanisms of telomere maintenance (ALT; Ref.
7
). Telomerase expression does not prevent accelerated senescence induced by DNA damage
(26)
, but both replicative and accelerated forms of
senescence are inhibited by the inactivation of p53 or p16, two of the
most
commonly disabled tumor suppressors in different
types of cancer. As a result, most tumor cells have both
senescence-promoting
changes (short telomeres, RAS mutations) and
senescence-inhibiting adaptations (activation of telomerase,
inactivation of
p53 and/or p16). Until a few years ago, it had been
a common assumption that neoplastically transformed cells are no longer
capable of senescence. Today we know, however, that
tumor cells can undergo senescence and can be forced into this process
by various genetic manipulations and by epigenetic
factors, including conventional anticancer drugs, radiation, and
differentiating
agents.
Inducing Senescence in Tumor Cells by Genetic Modifications
The earliest means used to induce senescence in “immortal” tumor cell lines was somatic cell fusion with normal cells or with
other tumor cell lines. These studies (reviewed in Ref.
27
) have demonstrated that senescence is dominant
over immortality, and they have identified four senescence-determining
complementation
groups. A senescence-determining gene of one
complementation group has been isolated from chromosome 4 and termed MORF4. Transfection of MORF4, which encodes a transcription factor-like protein, induces cessation of proliferation and the senescent phenotype after
18–35 population doublings in different tumor cell lines that belong to the same complementation group
(28)
.
In the past 5 years, the transfection of
many growth-inhibitory genes into tumor cell lines was shown to produce
stable growth
arrest, followed by the appearance of
senescence-associated phenotypic changes and SA-β-gal expression.
Growth-inhibitory
genes that induce senescence-like growth arrest in
tumor cells are listed in Table 1
⇓
; most of these genes are known to play a role in
the program of senescence in normal cells. These genes include RB, which induces senescence through a pathway that appears to depend on the induction of p27
(29)
, p53, and two p53-related proteins (p63 and p73), several CDK inhibitors (p21, p16, p57Kip2, and p15Ink4b),
and IGFBP-rP1, a member of the IGF binding protein family, which is
often up-regulated in normal senescent cells. Constitutively
active mutants of two genes stimulating the MAPK
pathway, including RAF-1 and MAPK kinase MKK6 (which specifically activates p38HOG), also induced tumor cell senescence.
View this table:
Table 1
Genes that induce senescence-like growth arrest in tumor cells
Is the senescence-like growth arrest
induced in tumor cells by the overexpression of growth-inhibitory genes
irreversible?
The ability of tumor cells to recover once the
expression of a growth-inhibitory gene has been turned off has been
investigated
for p53, p21, and p16 tumor suppressors, which were expressed in tumor cells from regulated promoters
(30,
31,
32,
33)
. In all of these cases, the ability of the cells
to grow and form colonies after the promoter shutoff was inversely
related
to the duration of expression of the tumor
suppressor, with very few cells recovering after prolonged induction
(4–5 days).
The failure to recover after the release from
p21-induced growth inhibition was also shown to depend on the level to
which
p21 expression was induced
(34)
. The latter study, in which p21 was expressed from
an inducible promoter in HT1080 fibrosarcoma cells, has also addressed
the mechanism of the failure of the cells to
recover after the shutoff of p21. All of the cells, despite their
senescent phenotype
and regardless of the duration or the magnitude of
p21 induction, were found to reenter the cycle and replicate their DNA.
On entering mitosis, however, most of the cells
that were released after prolonged arrest developed grossly abnormal
mitotic
figures and either died through mitotic catastrophe
or underwent senescence-like growth arrest in a subsequent cell cycle
(34)
. It remains to be determined whether the failure
to recover from senescence-like growth arrest induced by any other
growth-inhibitory
genes is attributable to a genuinely permanent cell
cycle arrest that can be maintained without the inducing protein.
Another type of genetic manipulation that
induces senescence in tumor cells is based on inhibiting the tumor
proteins that
counteract senescence. Somewhat surprisingly, the
inhibition of telomerase by a dominant-negative mutant was found to
induce
cell death rather than senescence in tumor cell
lines
(35)
. This result can be understood in light of the
antiapoptotic function of telomerase, which may be independent of its
effect
on telomeres
(36)
, and which may also reflect the induction of
mitotic catastrophe by abnormal telomeric structures. In contrast to the
outcome
of telomerase inhibition, senescence was readily
induced in cervical carcinoma cells by inhibiting papillomavirus
oncoproteins
E6 and E7, which inhibit p53 and Rb
tumor suppressors, respectively. Introduction of bovine papillomavirus
protein E2, a negative regulator of both E6 and E7,
into several human cervical carcinoma cell lines
induced accelerated senescence in almost 100% of tumor cells
(37
, 38)
. The effect of E2 was not accompanied by telomere shortening
(37)
, and it was not prevented by constitutive overexpression of telomerase
(39)
. Induction of senescence by E2 was associated with
p53 stabilization and with strong induction of p21, and it was
prevented
by using p21-inhibiting antisense oligonucleotides
or by increasing the expression of E6 or E7
(38)
. These results demonstrate that tumor cells are
“primed” to undergo accelerated senescence once senescence-restraining
mechanisms
that inhibit the p53 and Rb pathways are removed.
Enhancement of the extant program of accelerated senescence in tumor
cells
can be viewed, therefore, as a biologically
justified approach to cancer therapy.
Induction of Senescence in Tumor Cell Lines by Chemotherapy, Radiation, and Retinoids
The propensity of tumor cells to undergo senescence in response to damage was demonstrated by the analysis of the effects
of chemotherapeutic drugs and radiation on cell lines derived from different types of human solid tumors
(40)
. A wide variety of anticancer agents induced
senescence-like morphological changes and SA-β-gal expression in tumor
cells.
When equitoxic (ID85) doses of
different agents were applied to HT1080 fibrosarcoma cells, the
strongest induction of the senescent phenotype
was observed with DNA-interactive agents
doxorubicin, aphidicolin, and cisplatin; a somewhat weaker response was
seen with
ionizing radiation, cytarabine, and etoposide; and
the weakest effect was seen with microtubule-targeting drugs (Taxol and
vincristine). Induction of senescence by the drugs
was dose-dependent, and it was detectable even at the lowest drug doses
that had a measurable growth-inhibitory effect.
Moderate doses of doxorubicin induced the senescent phenotype in 11 of
14
cell lines derived from different types of human
solid tumors
(40)
. Other investigators have demonstrated the
induction of the senescent phenotype in different tumor cell lines
treated with
cisplatin
(41)
, hydroxyurea
(42
, 43)
, doxorubicin
(44
, 45)
, camptothecin
(46)
, or bromodeoxyuridine
(47
, 48)
. Drug-induced senescent phenotype in tumor cells
was not associated with telomere shortening and was not prevented by the
overexpression of telomerase
(45)
. Notably, in some of the cell lines, the senescent
phenotype was observed in 10–20% of the cells even without drug
treatment
(40)
, suggesting that tumor cell senescence could
develop spontaneously, possibly in response to subtle changes in the
cell environment.
Drug-induced senescent phenotype was specifically associated with the tumor cells that underwent terminal growth arrest in
response to treatment
(31
, 40
, 49)
. The most conclusive evidence for this came from
the analysis of growth-arrested and proliferating cells that were
separated
after release from the drug. This FACS-based
separation procedure involves labeling cells with a lipophilic
fluorophore PKH2,
which stably incorporates into the plasma membrane
and distributes evenly between daughter cells, resulting in gradual
decrease
in PKH2 fluorescence with increasing numbers of
cell divisions
(50)
. In the experiment shown in Fig. 2
⇓
(from Ref.
49
), HCT116 cells were exposed to 200 μm
doxorubicin for 24 h and then were labeled with PKH2. Changes in PKH2
fluorescence were monitored on subsequent days. Drug-treated
HCT116 cells remained growth arrested (PKH2hi) for 2–3 days after the removal of doxorubicin, but a proliferating cell population (PKH2lo) emerged starting from day 4. A large fraction of cells, however, remained PKH2hi and did not change their fluorescence throughout the experiment, indicating that these cells did not divide even after release
from the drug (Fig. 2A)
⇓
. Six to 9 days after treatment, the cells were separated by FACS into PKH2hi (growth-arrested) and PKH2lo (proliferating) fractions. The PKH2hi cells were large, flat, and SA-β-gal+, but PKH2lo cells did not express the markers of senescence and were otherwise indistinguishable from the untreated cells (Fig. 2B)
⇓
. The PKH2lo (proliferating) but not the PKH2hi (growth-arrested) cells gave rise to colonies (Fig. 2C)
⇓
, indicating that the senescent phenotype of
doxorubicin-treated HCT116 cells is associated with terminal growth
arrest
(49)
. On the other hand, when doxorubicin-treated
HT1080 fibrosarcoma cells were analyzed by a similar procedure, most of
the
cells in the PKH2hi population divided
once or twice after release from doxorubicin before undergoing terminal
growth arrest, indicating that
drug-induced senescent phenotype can be associated
with both an immediate and a delayed terminal growth arrest
(40)
.
Fig. 2.
Doxorubicin-induced senescence in HCT116 colon carcinoma cells (from Ref.
49
). A, FACS profiles of PKH2 fluorescence of HCT116 cells on the indicated days after release from doxorubicin. B, SA-β-gal staining of PKH2hi (right) and PKH2lo (left) populations, separated 6 days after release from the drug, photographed at ×200. C, colony formation by PKH2hi and PKH2lo populations, separated 9 days after drug treatment and plated at 10,000 live cells per 10-cm plate.
The role of senescence as a determinant of treatment response was also indicated by two other in vitro
studies. One of these studies found that expression of the MDR1
P-glycoprotein, which acts both as a drug efflux pump and
as an inhibitor of apoptosis, protects a HeLa
derivative and NIH 3T3 cells from radiation-induced apoptosis, but it
does not
increase their clonogenic survival. This apparent
paradox was resolved by finding that a decrease in the fraction of
apoptotic
cells was accompanied by a commensurate increase in
the fraction of cells undergoing either senescence or mitotic
catastrophe
(the principal nonapoptotic form of cell death),
indicating that the latter responses, without apoptosis, are sufficient
to
stop the proliferation of tumor cells
(51)
. The second study identified a novel phenotype of
resistance to multiple DNA-interactive drugs, and showed that this form
of resistance was associated with a decreased
senescence response
(52)
.
Aside from cytotoxic drugs and radiation, tumor cell senescence was also found to be induced by TGF-β
(53)
and by compounds that are usually referred to as “differentiating agents,” including sodium butyrate
(54)
and retinoids. The induction of senescence has been
analyzed in most detail for retinoids, natural and synthetic
derivatives
of vitamin A, which regulate cell growth and
differentiation through their effects on gene expression. The latter
effects
are mediated by the binding of retinoids to
retinoid receptors that act as regulators of transcription.
Retinoid-induced growth
arrest of tumor cells is commonly assumed to result
from the induction of differentiation, and, in some cases, this
assumption
has been corroborated by the appearance of
differentiation-specific markers in retinoid-treated cells. It has now
become apparent,
however, that retinoid treatment can also induce
senescence rather than differentiation
(55)
. In particular, exposure of MCF-7 breast carcinoma cells to a low noncytotoxic dose of all-trans RA induced growth arrest and the senescent phenotype
(40)
. cDNA microarray analysis showed that this
response involves the up-regulation of several senescence-associated
genes but
not of any markers of epithelial differentiation
(56)
. In another example
(57)
, two sublines of the same neuroblastoma cell line
SK-N-SH showed different morphological responses to RA. In one subline,
RA induced neuronal differentiation, as defined by
morphological and antigenic markers. In the other subline, RA treatment
produced morphological features of senescence and
SA-β gal expression. Significantly, RA treatment increased the levels of
p21 in the differentiating SK-N-SH cells, but it decreased p21 expression in the subline undergoing senescence, suggesting that p21 could act as a switch between differentiation and
senescence in retinoid-treated cells
(57)
. RA was also shown to decrease p21 levels in MCF-7 cells
(58)
. In contrast to RA-induced senescence, p21 levels
are increased under the conditions of senescence induced by DNA-damaging
chemotherapeutic drugs
(31
, 49)
. As discussed in a later section, these
differences in p21 expression are paralleled by major differences in the
spectra
of genes that are up-regulated in senescent tumor
cells.
Senescence as a Determinant of in Vivo Treatment Response
The induction of tumor senescence by
anticancer agents is not limited to cell culture. SA-β-gal staining
showed that the senescent
phenotype develops in human tumor xenografts grown
in nude mice and treated in vivo with a retinoid (fenretinide; Ref.
40
) or with doxorubicin
(59)
. Recently te Poele et al.
(44)
became the first to investigate the correlation
between the senescence response and chemotherapeutic treatment in
clinical
cancer. This study used newly sectioned material
from frozen archival breast tumors of patients who had or had not
received
neoadjuvant chemotherapy (the CAF regimen:
cyclophosphamide, doxorubicin, and 5-fluorouracil). The senescent
phenotype was
detected by SA-β-gal activity, and the tumors were
also stained for p53 and p16. Although SA-β-gal enzymatic activity is
unstable
even in freshly frozen tissue samples, te Poele et al.
have succeeded in demonstrating SA-β-gal in 15 (41%) of 36 treated
tumors. Remarkably, SA-β-gal staining was confined to
tumor cells, whereas normal tissue was completely
negative, suggesting that chemotherapy-induced senescence is a specific
response of tumor cells. Tumor sections of patients
who had not received chemotherapy showed SA-β-gal staining of isolated
tumor cells in 2 of 20 cases, suggesting that
spontaneous senescence also occurs in clinical cancer. SA-β-gal staining
in
breast cancer was associated with low p53 staining,
indicative of the lack of mutant p53, and with high staining for p16
(44)
.
In another recent study
(60)
, both senescence and apoptosis were shown to determine an in vivo response to chemotherapy in Eμ-myc lymphoma, a transgenic mouse model of B-cell lymphoma. When the program of apoptosis in
Eμ-myc lymphoma was inhibited by transduction with a retrovirus that expresses the antiapoptotic gene BCL-2,
senescence became the principal tumor response to cyclophosphamide, as
indicated by apparently complete cessation of DNA
replication and mitosis and by drastic induction of
the SA-β-gal marker. The senescence response (along with apoptosis) was
also observed in the absence of BCL-2.
Treatment-induced senescence became undetectable on the knockout of
either p53 (which also abolished the apoptotic response)
or p16 (which had no effect on apoptosis).
Inhibition of either apoptosis or senescence in Eμ-myc lymphoma made
these tumors
significantly more sensitive to chemotherapy,
indicating that both of these physiological programs contribute to
treatment
success
(60)
.
It should be noted, however, that
senescence is not the only antiproliferative response that determines
treatment response
in the absence of apoptosis. Another principal
effect of anticancer agents is mitotic catastrophe, cell death resulting
from
abnormal mitosis, which usually ends in the
formation of large cells with multiple micronuclei and uncondensed
chromatin (reviewed
in Ref.
59
). Etoposide-induced mitotic catastrophe in HeLa
cells was greatly increased when apoptosis was inhibited by BCL-2
(61)
, and, as mentioned above, both mitotic catastrophe
and senescence were augmented in irradiated tumor cells with
MDR1-suppressed
apoptotic response
(51)
. There are as yet no in vivo studies in
which all three responses have been analyzed at the same time. Such
analysis should elucidate the relative contribution
of apoptosis, senescence, and mitotic catastrophe
to the overall outcome of cancer treatment.
p53, p21, and p16 in Tumor Cell Senescence
Some observations indicate that p53, p21, and p16, which regulate replicative and accelerated senescence in normal cells (see Fig. 1
⇓
), also play a role in treatment-induced senescence
of tumor cells. Treatment-induced senescence in murine Eμ-myc lymphoma
required wild-type p53 and p16
(60)
, and p16 expression correlated with SA-β-gal staining in treated breast cancers
(44)
. As mentioned above, p53 and p16
are frequently inactivated in cancers. If these genes were required for
treatment-induced senescence, one could expect that
the majority of tumors, which are deficient in one
or both of these genes, would be unable to undergo senescence. This,
however,
is not the case. Chemotherapeutic drugs readily
induced senescence in vitro and in vivo in p16-deficient tumor cell lines, such as HT1080 and HCT116
(40)
. Furthermore, moderate doses of doxorubicin induced the senescent phenotype in p53-null Saos-2 cell line, in SW480 and U251 cells carrying mutant p53, and in HeLa and Hep-2 cell lines, in which p53 function has been inhibited by papillomavirus protein E6
(40)
. In the breast cancer study of te Poele et al.
(44)
, 20% of the SA-β-gal+ tumors showed high p53 staining (suggestive of p53 mutations) and 13% of the SA-β-gal+ tumors did not stain for p16, indicating that wild-type p53 and p16 induction are not necessary for senescence in clinical breast cancer.
The role of p53 and p21 in
treatment-induced senescence was analyzed in HT1080 fibrosarcoma cells
in which p53 function and
p21 expression were blocked by a p53-derived
genetic suppressor element and in HCT116 colon carcinoma cells with
homozygous
knockout of p53 or p21. In both cell lines, the
inhibition or knockout of p53 or p21 strongly decreased but did not
abolish
drug- or radiation-induced senescence, as
determined by PKH2 analysis of cell division and by SA-β-gal staining
(31)
. Hence, p53 and p21 act as positive regulators of
accelerated senescence in tumor cells, but they are not absolutely
required
for this response. In addition, as described above,
retinoid-induced senescence involves a decrease rather than an increase
in p21 expression. These observations suggest that
some genes other than p53, p21, or p16 are likely to play a role in accelerated senescence of tumor cells.
Inhibition of Cell Cycle Progression Genes and Induction of Intracellular and Secreted Growth Inhibitors in Senescent Tumor Cells
Additional determinants of drug-induced
senescence in tumor cells were identified by gene expression profiling
of doxorubicin-induced
senescence in HCT116 colon carcinoma cells, which
are p16-deficient and wild-type for p53
(49)
. The proliferating and senescent fractions of
HCT116 cells were separated by PKH2 labeling and flow sorting 6–9 days
after
1-day doxorubicin treatment (Fig. 2)
⇓
and were used for RNA extraction. cDNA microarray
hybridization followed by reverse transcription-PCR analysis of
individual
genes revealed major biological clusters of genes
that were either down-regulated or up-regulated in senescent cells. More
than one-half of all of the genes that were
strongly inhibited in senescent cells are known to play a role in cell
cycle progression,
with the largest groups of genes involved in
mitosis and DNA replication. The inhibition of these genes became
apparent 1–2
days after doxorubicin treatment and was likely to
contribute to the maintenance of drug-induced growth arrest. Analysis of
HCT116 cell lines with homozygous disruption of
either p53 or p21 demonstrated that doxorubicin-induced inhibition of cell cycle progression genes was fully dependent on p21
(49)
. Furthermore, ectopic expression of p21 in HT1080 fibrosarcoma cells was sufficient to inhibit the transcription of the same set of genes that are down-regulated
in doxorubicin-treated cells
(4)
. This effect of p21 is exerted at the
level of transcription, and it is mediated at least in part by negative
regulatory elements in the corresponding
promoters, such as CDE/CHR (cell cycle dependence
element/cell cycle gene homology region; Ref.
62
).
In addition to the down-regulation of genes required for cell cycle progression, senescent HCT116 cells were also found to
up-regulate multiple genes with growth-inhibitory activities (Table 2
⇓
). Concerted and sustained induction of such genes
explains the growth arrest of senescent cells despite the lack of p16
and
suggests that this arrest is maintained by many
apparently redundant mechanisms. Several of these genes are known or
putative
tumor suppressors that are silenced in the course
of neoplastic transformation but become reactivated with the onset of
senescence.
The up-regulated growth inhibitors include p21, tumor suppressor BTG1, a related gene BTG2, and candidate tumor suppressor EPLIN.
Of special interest, senescent HCT116 cells also overexpress several
secreted growth inhibitors, including serine protease
inhibitor Maspin, a tumor suppressor shown to
inhibit the invasion and angiogenesis of breast and prostate cancers, as
well
as MIC-1/pTGF-β (a member of the TGF-β family),
IGF-binding protein 6 (IGFBP-6), and amphiregulin, an epidermal growth factor (EGF)-related factor that inhibits the growth of several carcinoma cell lines
although promoting the growth of normal epithelial cells
(49)
. Exposure to chemotherapeutic drugs and radiation
was previously shown to induce a similar set of secreted
tumor-suppressing
factors, and paracrine growth-inhibitory activities
of the damaged cells have been documented by conditioned media and
coculture
assays
(63)
. The finding that the same factors are stably
overexpressed by senescent cells suggests that such cells may provide a
reservoir
of tumor-suppressing factors that may contribute to
the long-term success of chemotherapy.
View this table:
Table 2
Tumor-inhibiting and tumor-promoting genes induced in senescent tumor cells
Induction of secreted tumor-suppressing factors was previously found to be mediated by p53
(63)
. Analysis of p53-deficient HCT116 cells showed, however, that induction of senescence-associated growth inhibitors (intracellular or secreted)
showed either no dependence on p53 (BTG1, IGFBP-6) or limited p53 dependence (BTG2, EPLIN, Maspin, MIC-1, amphiregulin). The latter genes were still induced in the absence of p53, albeit their induction was delayed or diminished relative to the cells with the wild-type p53
(49)
. These results help to explain why p53 deficiency diminishes but does not abolish drug-induced senescence
(31)
. p21 knockout in HCT116 cells had no effect on the induction of senescence-associated growth inhibitors (except for EPLIN; Ref.
49
), explaining why p21-deficient cells can still undergo senescence
(31)
. On the other hand, the reduction in the senescence response of p21-deficient cells can be readily explained by a failure to inhibit the transcription of cell cycle progression genes and by
the lack of p21 itself.
Concerted induction of several
growth-inhibitory genes was also observed in retinoid-induced senescence
of MCF-7 breast carcinoma
cells
(56)
. cDNA microarray hybridization and reverse
transcription-PCR analysis showed that RA-induced senescent phenotype of
MCF-7
cells is associated with the strong induction of
four genes with growth-inhibitory activity. These genes (Table 2)
⇓
encode intracellular inhibitors EPLIN and FAT10 (a
ubiquitin family member), as well as a secreted growth-inhibitor
IGFBP-3,
which is related to IGFBP-6 up-regulated in
senescent HCT116 cells), and a tumorigenicity-suppressing cell adhesion
protein
βIG-H3. Interestingly, a survey of the published
genes that are inducible by retinoids in different types of tumor cells
showed
that retinoids induce many other intracellular and
secreted growth inhibitors, most of which are also known to be
up-regulated
in senescent cells
(55)
. βIG-H3, IGFBP-3, and a related inhibitor IGFBP-4
were also up-regulated in HeLa cells by bromodeoxyuridine treatment,
under
the conditions that induce senescence in this cell
line
(48)
. Induction of multiple growth-inhibitory proteins
appears, therefore, to be a general phenomenon in treatment-induced
senescence
of tumor cells.
Tumor Senescence Is Associated with the Induction of Tumor-Promoting Genes
Inhibition of cell proliferation,
however, is not the only aspect of tumor senescence with potential
clinical implications.
Replicative senescence of normal fibroblasts is
characterized by changes in the expression of multiple proteins
(1)
. Some of the proteins that are highly expressed in
senescent cells have long-range pathogenic effects, including βAPP
(64)
, as well as degradative enzymes, inflammatory
cytokines, and growth factors, which may contribute to carcinogenesis
and tumor
progression
(1)
. Indeed, coculture and conditioned media
experiments showed that normal human fibroblasts undergoing either
replicative or
accelerated senescence stimulate the growth of
transformed epithelial cells in vitro and in vivo
(5)
. These paracrine effects of senescent fibroblasts
closely resemble the cancer-promoting activities of tumor-associated
stromal
fibroblasts
(65)
. The procarcinogenic function of normal senescent cells in vivo is also supported by the findings that SA-β-gal expression in normal human hepatocytes is strongly correlated with the presence
of hepatocellular carcinoma in the surrounding liver
(66)
, and that prostate enlargement correlates with SA-β-gal expression in prostate epithelial cells
(67)
.
In agreement with these observations in
normal senescent cells, doxorubicin-treated senescent HCT116 carcinoma
cells also
showed increased expression of genes for many
proteins with diverse paracrine activities. In fact, secreted factors,
ECM components,
ECM receptors and other integral membrane proteins
make up 33 of 68 genes with known functions that are strongly induced in
senescent HCT116 cells (in contrast, only 2 of 64
known genes that are down-regulated in such cells belong to this
category).
The secreted proteins up-regulated in senescent
HCT116 cells include not only the above-described growth inhibitors but
also
several proteins with mitogenic, antiapoptotic, and
angiogenic activities (Table 2
⇓
; Ref.
49
). Some of these proteins are an ECM component
Cyr61 with mitogenic and angiogenic functions, an antiapoptotic and
mitogenic
ECM factor prosaposin, transforming growth factor
TGFα, and several proteases that may potentially contribute to
metastatic
growth. Several other genes induced in senescent
cells encode cell adhesion and cell-cell contact proteins and ECM
receptors,
including several integrins and syndecan-4,
involved in angiogenesis. Other transmembrane proteins induced in
senescent cells
are βAPP, which has mitogenic activity
(68)
, another amyloid precursor, BRI, associated with
an Alzheimer’s-like disease, and growth-regulatory proteins CD44 and
Jagged-1
(49)
. Thus, senescence-associated changes in gene
expression involve the induction of both tumor-suppressive and
tumor-promoting
proteins, as well as proteins involved in
age-related diseases other than cancer. Relative expression of different
biological
classes of senescence-associated genes is likely,
therefore, to determine whether tumor senescence would have a mostly
positive
or a mostly negative effect on the outcome of
treatment.
Role of CDK Inhibitors in Senescence-associated Changes in Gene Expression: Implications for Tumor-Promoting Stromal Fibroblasts
About one-third of senescence-associated
genes that were induced by doxorubicin in HCT116 cells showed decreased
or delayed
induction in a p21−/− derivative of this cell line,
indicating that p21 plays a role not only in the inhibition but also in
the induction of gene expression in senescent
cells. Some of the genes that showed p21 dependence encode secreted
mitogenic/antiapoptotic
proteins, such as prosaposin, TGFα, and βAPP
(49)
. These findings were in accord with the results of
cDNA microarray analysis of the effects of p21 on gene expression in
HT1080
fibrosarcoma cells
(4)
. p21 induction in the latter cells produces growth
arrest and the senescent phenotype, inhibits transcription of multiple
genes, most of which are involved in cell cycle
progression, and also leads to the induction of a set of genes with
important
paracrine activities. Altogether, 40% of
p21-inducible genes encode secreted proteins, ECM components, or ECM
receptors. Most
of the genes induced by p21 in HT1080 cells were
also induced in WI-38 normal human fibroblasts infected with a
p21-expressing
adenoviral vector
4
and in a human melanoma cell line treated with a
polyamine-depleting regimen that induces strong p21 expression
(69)
. Furthermore, the effects of p21 on the induction
of gene expression in HT1080 cells can be largely reproduced by another
senescence-associated CDK inhibitor, p16.
5
Many p21-induced genes are known to be up-regulated during replicative senescence
(4)
. Furthermore, products of many genes that are
induced by p21 have been linked to age-related diseases, including
Alzheimer’s
disease, amyloidosis, atherosclerosis, and
arthritis. Some examples are βAPP, serum amyloid A, tissue
transglutaminase, connective
tissue growth factor (CTGF), and p66Shc, a positive mediator of oxidative stress, knockout of which increases toxin resistance and the life span in mice
(70)
. Another group of p21-induced genes encode
secreted proteins with known mitogenic, antiapoptotic, or angiogenic
activities
(Table 2)
⇓
, such as prosaposin, epithelin/granulin,
galectin-3, CTGF, or VEGF-C. The induction of such genes produces
paracrine growth-promoting
activities, as demonstrated by the fact that
conditioned media from p21-arrested HT1080 cells has mitogenic and
antiapoptotic
effects
(4)
. These paracrine effects of p21 induction mimic
the tumor-promoting activities that were demonstrated in different types
of senescent fibroblasts
(5)
and in tumor-associated stromal fibroblasts
(65)
. As discussed elsewhere
(71)
, all of the treatments that are known to induce
the tumor-promoting functions of stromal fibroblasts also result in p21
induction,
suggesting that p21 or related proteins could be
responsible for the paracrine tumor-promoting functions of stromal
fibroblasts.
Induction of gene expression by p21 occurs at the level of transcription, because p21 stimulated the activity of all six of
the tested promoters of p21-inducible genes.
6
Although p21 is best known as an inhibitor of
cyclin/CDK complexes, it also interacts with many transcription factors
and
cofactors and regulators of signal transduction
(16)
, which can account for its pleiotropic effects on
gene expression. One of the effects of p21 is the augmentation of
transcription
factor NFκB
(72)
. This effect is mediated through the activation of
transcription cofactors/histone deacetylases p300 and CBP, which
augment
not only NFκB but also many other inducible
transcription factors
(72)
. The stimulation of p300 and CBP by p21 is mediated through a repressor domain of p300/CBP termed CRD1
(73)
. Recently, the ability of p21 to stimulate
p300-mediated transactivation of different genes was investigated in U-2
OS osteosarcoma
cells cotransfected with p21, p300 fused to yeast
Gal4 DNA-binding protein domain, and reporter constructs containing core
promoters of different genes linked to Gal4
DNA-binding sites
(74)
. p21 enhanced the effect of p300 on the core
promoters of five genes that are known to be induced by p21 and on three
strong
promoters of viral origin, but p21 did not
stimulate the transactivation of four core promoters from genes that are
not induced
by p21. The effect of p21 on promoter
transactivation by p300 appears, therefore, to be an important
determinant of the selectivity
of the induction of gene expression by p21. The
ability of a core promoter to respond to p21 was found to be determined
primarily
by the sequences flanking the TATA box
(74)
. The effect of p21 in this system does not appear
to be mediated by cyclin/CDK binding, because two p21 mutants deficient
in such binding were as active in stimulating the
effect of p300 as was the wild-type p21
(74)
. On the other hand, the same mutants showed little
or no effect on the transcription of p21-inducible genes or complete
promoters
of such genes in HT1080 cells,
6
suggesting that there may be cell type-specific
differences in the induction of disease-associated genes by p21.
The involvement of p21 and other CDK
inhibitors in the expression of genes associated with the undesirable
effects of senescence
suggests that treatments that induce senescence
without activating p21 or p16 may be more beneficial in the long term.
In
agreement with this hypothesis, none of the genes
that were found by cDNA microarray analysis to be up-regulated in
retinoid-induced
senescence of MCF-7 breast carcinoma cells encode
secreted factors with tumor-promoting activities, or other
disease-promoting
factors (such as amyloid proteins; Ref.
56
). This result is likely to reflect the lack of p21
or p16 induction in RA-induced senescence of MCF-7 cells. Thus, the
positive
effects of tumor senescence (permanent growth
arrest of tumor cells and secretion of tumor-suppressing factors) can be
separated
from the disease-promoting activities of senescent
cells.
Prognostic Implications of Tumor Senescence: Examples from Prostate Cancer
Senescent cells appear within the tumor
either as a consequence of treatment or, spontaneously, as a result of
environmental
stress or sporadic inactivation of
senescence-restraining mechanisms in an individual cell. Senescent cells
are generally
resistant to apoptosis, and senescent fibroblasts
in culture are known to survive for more than a year. The persistence of
senescent cells in the tumor is a double-edged
sword. On one hand, the senescent tumor cells do not proliferate and,
furthermore,
serve as a reservoir of secreted factors that
inhibit tumor growth. On the other hand, the same cells can also produce
secreted
factors with mitogenic, antiapoptotic, and
angiogenic activities. These tumor-promoting functions of senescent
cells are determined
to a large extent by the expression of p21 and p16.
The presence of senescent cells in the tumor and the relative abundance
of different proteins produced by the senescent
cells are important biological factors that should have significant
prognostic
implications for the disease outcome. One would
expect that more aggressive tumors might contain few or no senescent
cells.
Alternatively, such tumors may have a substantial
fraction of senescent cells that express CDK inhibitors (p16 or p21), as
well as senescence-associated tumor-promoting
factors that are up-regulated by CDK inhibitors. On the other hand,
tumors containing
senescent cells that express high levels of
secreted growth inhibitors but little p16 or p21 should have a more
favorable
prognosis. This concept is illustrated in Fig. 3
⇓
.
Fig. 3.
Tumor senescence and its consequences.
Although the breast carcinoma study of te Poele et al.
(44)
is the only one thus far that has directly
addressed the development of tumor senescence in clinical cancer, there
is abundant
evidence in the literature in support of various
aspects of the model shown in Fig. 3
⇓
. This can be illustrated especially well by the
results of clinical and biological studies in prostate cancer as follows
(similar examples can be found for many other tumor
types):
(a) a prostate carcinoma cell
line LNCaP growing in culture contains 10–15% SA-β-gal+ cells,
suggesting that prostate cancer
cells can undergo spontaneous senescence. The
senescent fraction in this cell line is increased by doxorubicin
(40)
or by transfection with an activatable form of c-Raf
(75)
, thus demonstrating the susceptibility of prostate cancer cells to accelerated senescence;
(b) indirect observations in
radiation therapy of prostate cancer suggest that the induction of
senescence may be a primary
mode of treatment response. In particular, complete
regression of prostate cancers was reported in some patients to take
more
than a year after radiation treatment
(76)
. This slow course of tumor disappearance seems
most consistent with radiation-induced senescence. In an example from
another
tumor type, regression of desmoid tumors took up to
2 years after radiation treatment
(77)
;
(c) p16, the CDK inhibitor primarily associated with senescence, is a tumor suppressor, which is infrequently mutated in prostate
cancer
(78)
. p16 expression was not detectable by immunohistochemistry in the normal prostate
(79)
, but it is observed in close to one-half of prostate carcinomas
(79)
. In all of these studies, p16 was found to be an unfavorable prognostic marker for prostate cancer
(80
, 81)
. In particular, p16 expression was an independent
indicator of early relapse after radical prostatectomy, and p16 was
elevated
in cancers relative to benign prostatic hyperplasia
(BPH). These paradoxical adverse correlations of the tumor suppressor
p16 are readily explained by the ability of p16 to
up-regulate secreted tumor-promoting factors. Notably, higher p16 levels
in tumor cells have also been associated with the
history of androgen ablation treatment
(79)
, a result that probably reflects the induction of
prostate cancer senescence by this treatment. Strong adverse
correlations
for p16 expression have also been reported in
breast cancer
(82,
83,
84)
;
(d) similar unfavorable
prognostic correlations were found for another CDK inhibitor, p21, in
the majority of studies that analyzed
this protein in prostate cancer
(85,
86,
87,
88,
89,
90)
. p21 expression in prostate cancer showed no
correlation with p53 status, suggesting that p21 is induced in this
tumor primarily
by p53-independent mechanisms
(85)
. Like p16, p21 was found to be an independent
marker of early relapse after prostatectomy, and it was associated with
high
pathological grade and high Ki-67 index. p21 was
also shown to be a highly significant marker of progression from
androgen-dependent
to androgen-independent cancer
(90)
. The adverse prognostic role of p21 is disputed in some reports
(91
, 92)
, but a potential explanation for such discrepancies is suggested by the findings of Sarkar et al.
(88)
. The latter study noted a dependence of p21
correlations on the racial background (p21 is a strong independent
marker of
negative prognosis in Caucasians but not in African
Americans) and suggested that some as yet unknown genetic factors may
affect the role of p21 in prostate cancer. Studies
of p21 expression in other tumor types produced both favorable
prognostic
correlations (reflecting the role of p21 as a
marker of wild-type p53 function) and unfavorable correlations, similar
to those
in prostate cancer
(71)
;
(e) prostate cancers also
express senescence-associated growth inhibitors with good prognostic
correlations. One of the most
commonly used markers in prostate cancer is
IGFBP-3, a secreted protein that induces growth arrest and apoptosis.
Low IGFBP-3
levels in the plasma, alone or in combination with
high IGF-1, have been associated with the presence of advanced prostate
cancer
(93
, 94)
, and an increase in serum IGFBP-3 has been used as an indicator of treatment response
(95)
. IGFBP-3 is induced at senescence in different types of normal and tumor cells (Table 2
⇓
; Ref.
55
). In particular, IGFBP-3 is consistently
up-regulated in senescent prostate epithelial cells and silenced in
prostate cancer
cell lines and tumors
(96)
. This senescence-associated growth inhibitor is
induced in prostate cancer cells by many antiproliferative agents
(97
, 98)
. Another IGFBP family member, IGFBP-rP1 increases
during senescence of normal prostate epithelial cells and is
down-regulated
in prostate cancer
(99)
. Overexpression of IGFBP-rP1 induces growth arrest
and the senescent phenotype in the M12 prostate cancer cell line
(100)
; and
(f) the strongest correlations
with good prognosis in prostate cancer have been reported thus far for
another senescence-associated
tumor suppressor, serine protease inhibitor Maspin.
Maspin is induced to a very high level by DNA damage in several tumor
cell lines, including LNCaP prostate carcinoma
(101)
. The absence or low levels of Maspin in prostate
cancer samples have been correlated to higher tumor stages, histological
dedifferentiation and early relapse
(102)
. Conversely, Maspin expression was strongly
elevated in the cancers of patients treated with neoadjuvant androgen
ablation
therapy, and treatment-induced Maspin was
specifically associated with tumor cells that showed morphological
effects of the
treatment
(103)
.
The above observations in prostate cancer seem to be in excellent agreement with the model depicted in Fig. 3
⇓
. As predicted by the model, markers of senescence
are observed in untreated tumors, but their expression is elevated after
treatment. Furthermore, senescence-associated
growth inhibitors (Maspin and IGFBP-3) correlate with good prognosis,
whereas
p16 and p21 correlate with bad prognosis. With the
identification of multiple senescence-associated growth regulators
(Table
2)
⇓
, it should now be possible to investigate their
expression and coexpression in tumor tissues by conventional
immunocytochemical
techniques. Such analyses should allow us to
determine the relationship between the expression of these proteins,
tumor staging,
and treatment outcome. It should also be possible
to determine whether radiation or chemotherapy-induced tumor senescence
leads to sustained secretion of
senescence-associated growth inhibitors into bodily fluids. Production
of such proteins could
be an important factor in preventing the tumor
growth. New diagnostic approaches that will arise from understanding the
biology
of tumor senescence may be of considerable benefit
in the management of cancer patients.
Potential for Developing Senescence-based Anticancer Drugs
Activating the program of senescence in
tumor cells seems an attractive approach to cancer treatment. This
response to chemotherapy
is induced by a wide variety of anticancer agents,
even under the conditions of minimal cytotoxicity. Even if not all of
the
tumor cells are rendered senescent as a result of
treatment, such cells may provide a reservoir of secreted
tumor-suppressing
factors that will inhibit the growth of
nonsenescent cells. On the other hand, senescent cells can overexpress
secreted tumor-promoting
factors, as well as proteins associated with
various pathological conditions. The side effects of senescence, which
are mediated
at least in part by CDK inhibitors, may have
potential adversarial effects in the short term (growth stimulation of
nonsenescent
tumor cells) and in the long term (increased
likelihood of de novo carcinogenesis and the development of
age-related diseases). On the basis of these considerations,
senescence-oriented therapeutic
strategies may include two general strategies. The
first direction is to develop the agents that will interfere with the
induction
of disease-promoting genes by CDK inhibitors. The
second strategy is to develop drugs that will induce tumor cell
senescence
without up-regulating p21 (which, unlike p16, is
almost never inactivated in tumors) or p21-inducible disease-promoting
genes.
These approaches are schematized in Fig. 4
⇓
.
Fig. 4.
Strategies for senescence-based therapy in cancer treatment.
Elucidation of the mechanisms that
mediate the induction of transcription by p21 or by other CDK inhibitors
should provide
the essential information for developing compounds
that will prevent such induction. Such agents may include inhibitors of
p21-stimulated transcription factors and cofactors,
such as NFκB or p300. Alternatively, these compounds may be identified
by de novo screening of chemical
libraries, based on their effect on the expression of p21-inducible
genes or promoter-reporter constructs.
Agents that prevent the induction of gene
expression by p21 are likely to interfere with the tumor-promoting
paracrine activities
of senescent cells that arise spontaneously or as a
result of conventional chemotherapy or radiation therapy. These
compounds
may also block the tumor-promoting activities of
stromal fibroblasts and may potentially be useful in the chemoprevention
of age-related diseases, such as Alzheimer’s
disease or atherosclerosis (Fig. 4)
⇓
.
The second strategy in Fig. 4
⇓
is based on the development of agents that would
induce tumor cell senescence without its associated side effects. The
feasibility
of this strategy is suggested by the finding that
retinoid-induced senescence of breast carcinoma cells involves the
induction
of the tumor-suppressing but not of the
disease-promoting genes. The therapeutic use of retinoids is limited by
the fact that
these drugs act though retinoid receptors, which
are readily lost in tumor cells. Senescence-associated growth-inhibitory
genes, however, contain no discernible retinoid
receptor-binding sites in their promoters, and they appear to be induced
by
retinoids through an indirect mechanism
(55)
. Furthermore, the same proteins are induced by
retinoids and by nonretinoid drugs, such as doxorubicin or
bromodeoxyuridine
(Table 2)
⇓
. It seems likely, therefore, that the mechanisms
that produce concerted up-regulation of retinoid-inducible growth
inhibitors
may also be stimulated by other inducers of
senescence. Identification of senescence-associated growth-inhibitory
genes makes
it possible to develop high-throughput screening
systems for agents that induce such genes. Thus, the elucidation of the
biological
aspects of tumor cell senescence offers plausible
approaches to the development of novel therapeutic strategies to stop
the
growth of tumor cells.
No comments:
Post a Comment