1. Introduction
Our cells are continuously exposed to potentially deleterious genotoxic events from both endogenous and exogenous sources that jeopardize genome integrity. The plethora of DNA lesions include DNA strand breaks and base alterations induced by ionizing radiation and chemical agents that generate reactive oxygen species, DNA alkylation and formation of abasic sites induced by alkylating agents, bulky DNA lesions induced by ultraviolet light (UV), DNA interstrand crosslinks induced by bifunctional alkylating agents and platinum drugs, and DNA-protein crosslinks arising from a wide range of chemicals, such as chemotherapeutic drugs and formaldehyde [
1,
2,
3,
4]. Constitutively available DNA repair processes deal with low levels of genomic injury and assist in ameliorating the detrimental effects of such agents. An increase in DNA damage above a threshold level activates the DNA damage surveillance network, which involves multiple signaling pathways that protect against genomic instability and restrict aberrant cell growth in response to genotoxic stress [
5]. The wild-type p53 tumor suppressor functions at the hub of this network [
6,
7].
In the mid 1990s it was proposed that the principal role of p53 in determining cell fate following genotoxic stress is to either promote survival by activating cell cycle checkpoints and facilitating DNA repair or induce apoptotic cell death. This two-armed model of the DNA damage surveillance network—namely, repair and survive, or die through apoptosis—provided the impetus for extensive research directed towards modulating p53 in an attempt to improve the outcome of conventional cancer therapies. However, it soon became clear that p53’s function extends beyond canonical cell cycle and apoptotic signaling, and impacts additional diverse biological processes including senescence and metabolism [
8,
9]. Murine cancer models have been employed to investigate the impact of p53 activation in the response of oncogene-driven cancers. As pointed out by Stegh [
9], “
confirming important roles of p53 in cancer suppression, these studies showed that reactivation of p53 in established tumors can temporarily stop tumor growth; the precise cellular mechanism is cancer type-specific, as lymphomas die by apoptosis, whereas p53 restoration in sarcomas and liver carcinomas leads to growth arrest and senescence. p53-driven apoptosis and senescence responses associated with temporary p53 reactivation led to prolonged survival. Although cancer remission was not permanent, and p53-resistant tumors emerged…” The promises, challenges and perils of targeting p53 in cancer therapy have been extensively discussed [
8,
9,
10].
A growing body of evidence suggests that the primary response triggered by moderate, clinically relevant doses of cancer therapeutic agents is a sustained proliferation block and not apoptosis in most human cell types (e.g., dermal fibroblasts, solid tumor-derived cells) [
6,
11], with activation of p53 signaling suppressing (rather than promoting) apoptosis [
12,
13,
14]. Such growth-arrested cells remain viable for long times (months) post-treatment, secrete a myriad of biologically active factors, and can give rise to progeny exhibiting stem-cell-like properties.
Herein we briefly review the mechanisms by which wild-type p53 suppresses apoptosis following genotoxic stress, focusing on the roles played by DNAJ homolog subfamily B member 9 (DNAJB9) and p21WAF1 (p21; also called CDKN1A). In addition, we discuss the significance of p53-mediated protection against apoptosis under physiological conditions, and the dark side of this function of p53 in the context of cancer chemotherapy.
3. Activation of Apoptotic Signaling Does Not Always Lead to Cell Death: Impact on Chemosensitivity Assessment
It is now widely accepted that transcriptional activation of pro-apoptotic proteins (e.g., PUMA, NOXA, BAX) might not inevitably lead to cell death as a result of concomitant activation of a host of anti-apoptotic proteins that maintain p53 under the apoptotic threshold level (e.g., MDM2, p21, WIP1, DNAJB9) [
6,
47], sequester pro-apoptotic factors such as BAX (e.g., 14-3-3δ) [
62,
63,
64,
65,
66], and inhibit ASK1 and the caspase cascade (e.g., p21) [
6]. Similarly, while caspase 3 functions as a key apoptosis executioner under some conditions, such as in the development and maintenance of the hematopoietic system, under other conditions it reveals its dark side by promoting tumor growth [
67,
68,
69,
70,
71,
72,
73,
74,
75,
76,
77]. For these and several other reasons, the Nomenclature Committee on Cell Death (NCCD) has cautioned the scientific community about the use/misuse of terminologies and concepts in the area of cell death research. Notably, in their 2009 article, the NCCD pointed out that bona fide “dead cells” would be different from “dying cells” that have not crossed the point of no return and have not concluded their demise [
78]. It is worth noting that radiosensitivity and chemosensitivity, assessed by the widely-used multi-well plate colorimetric assays, which determine the inhibition of cell growth (resulting from the combined impact of checkpoint activation, growth inhibition and cytotoxicity), have often been misinterpreted to reflect loss of viability and hence cell death.
Recently we reviewed the current knowledge on responses induced by ionizing radiation that can lead to cancer cell death or survival depending on the context [
11]. These include activation of caspases (e.g., caspase 3), growth arrest through SIPS, and creation of polyploid/multinucleated giant cells (hereafter called MNGCs) (also see
Figure 2). Such potentially pro-survival responses are triggered not only by ionizing radiation, but also by chemotherapeutic drugs [
74,
79,
80,
81,
82,
83,
84] and hypoxia [
72,
85,
86,
87,
88,
89].
Caspase 3 is extensively studied for its role in the execution phase of apoptosis [
90]. Accordingly, the activated (cleaved) form of caspase 3 has often been used as a molecular marker of apoptosis. Paradoxically, in recent years caspase 3 has also been demonstrated to function as a survival factor, promoting the growth of tumor-repopulating cells [
68,
69,
70,
71,
72,
73,
74,
75,
76,
77]. This pro-survival effect of caspase 3 has been attributed to secretion of prostaglandin E
2 (PGE
2) [
70,
72,
74]. The caspase 3-PGE
2 survival pathway is triggered by various stimuli, including ionizing radiation [
70,
74], chemotherapeutic drugs [
72,
74] and hypoxia [
72]. Interestingly, the biological outcome associated with caspase 3 activation is in part dependent on p21 (reviewed in [
32]). Thus, p21-mediated inhibition of caspase 3 activity results in suppression of apoptosis in response to genotoxic stress. On the other hand, caspase 3-mediated cleavage of p21 generates a 15 kDa fragment of p21 that appears to positively regulate apoptosis by forming a complex with active caspase 3. It is currently unknown whether p21 might play a role in the regulation of the caspase 3-PGE
2 survival pathway.
Whether caspase 3 plays a role in growth-arrested cancer cells also remains to be elucidated. However, it is well known that cancer cells undergoing SIPS remain viable and acquire the ability to secrete factors that can promote proliferation and invasiveness in cell culture models and tumor development in vivo [
91,
92]. This so-called “senescence-associated secretory phenotype” (SASP) includes several families of soluble and insoluble factors that can affect surrounding cells by activating various cell surface receptors and corresponding signal transduction pathways [
91,
92].
While some authors consider SASP to be the “dark” side of senescence [
91,
92,
93,
94,
95,
96], others have proposed that induction of senescence (SIPS) might be advantageous for cancer treatment [
97,
98,
99,
100,
101]. As pointed out by Maier et al. [
101], although “
accumulation of senescent cancer cells leads to an increased secretion of inflammatory cytokines, which might cause age-related pathologies, like secondary cancers, in the long term, the primary aim of cancer treatment leading to a longer overall survival should always take preference. Thus, the hypothetical possibility that senescent cells may be dormant with an intrinsic capability to reawaken years after the treatment is of secondary concern, similar to the risk of inducing second cancers.” However, aside from SASP, there is now compelling evidence that cancer cells undergoing SIPS can themselves give rise to stem-cell-like progeny, thereby contributing to cancer relapse following therapy [
11,
102,
103].
Like cells undergoing SIPS, MNGCs also remain viable and secrete cell-growth promoting factors [
11]. This property of MNGCs was first reported over 60 years ago for HeLa cervical carcinoma cells exposed to ionizing radiation [
104,
105,
106]. HeLa cells harbor wild-type alleles of
TP53, but are infected with human papillomavirus (HPV) 18, the E6 protein of which disables the p53–p21 axis [
107]. This observation prompted Puck and Marcus to develop the feeder layer clonogenic assay, in which a “lawn” of heavily-irradiated feeder cells (which encompass MNGCs) is inoculated into a culture dish to promote the growth of test cells given graded doses of genotoxic agents [
105]. Recently, we demonstrated that exposure of a panel of p53-deficient or p21-deficient solid tumor-derived cell lines to moderate doses of ionizing radiation (e.g., 8 Gy) results in the development of MNGCs that remain adherent to the culture dish, retain viability, metabolize 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl-tetrazolium bromide (MTT), and exhibit DNA synthesis for long times (e.g., three weeks) post-irradiation [
108].
Collectively, these observations underscore the importance of distinguishing between dead cells and growth arrested cells that might be mistakenly scored as “dead” in the colony formation and other cell-based radiosensitivity/chemosensitivity assays. As pointed out recently [
108], the creation of viable growth arrested cells (e.g., MNGCs) complicates the interpretation of data obtained with multi-well plate colorimetric tests routinely used in anti-cancer drug-screening endeavors.
4. Extrapolating Results Obtained in Overexpression Studies to Clinically Relevant Conditions
The preceding discussion raises a fundamental question with respect to p53 regulation and function. As discussed by Uversky [
7], p53 undergoes extensive post-translational modifications (e.g., phosphorylation, acetylation) that are critical for its stabilization and activation. Such modifications result in accumulation of p53 in the nucleus and the formation of p53 tetramers, which then bind to the promoters of target genes and trigger their expression. Genotoxic stress activates factors such as ATM and ATR that initiate the DNA damage surveillance network by mediating p53 posttranslational modifications (also see
Figure 1). Despite the wealth of knowledge regarding the importance of genotoxic stress (e.g., DNA damage) in activating the p53-mediated transcriptional program, this same transcriptional program has also been reported to be activated by ectopic expression of wild-type p53 without exposure to exogenous stress [
41]. Does this indicate that stress-triggered p53 posttranslational modifications are not needed for activation of its transcriptional program, which appears to be highly unlikely, or does p53 overexpression by itself create a non-physiological condition that it is sufficient to trigger the stress response?
It is important to note that many reports suggesting a positive role for wild-type p53 in triggering apoptosis, either with or without exposure to genotoxic agents, involved overexpression experiments with a variety of transformed/malignant cell types (e.g., T-cell leukemia cell lines). In addition, many authors did not follow the NCCD recommendations to distinguish between “dying cells” (i.e., exhibiting transient activation of a death-related biochemical pathway) and cells that are irreversibly committed to die. Taken together with the observations of Kracikova et al. [
41] described above (
Section 2.5.), demonstrating that different expression levels of exogenous p53 yield different outcomes (G1 arrest versus apoptotic signaling), caution should be exercised in extrapolating results obtained in overexpression studies to clinically relevant conditions (e.g., cancer chemotherapy) particularly when it pertains to p53-directed cancer cell death.
6. Targeting Growth-Arrested Cancer Cells as a Potential Therapeutic Strategy
The importance of MNGCs in the failure of cancer therapy has been largely overlooked. In part, this may be because the creation of such cells has been considered to be rare and also multinucleation has often been assumed to reflect death through “mitotic catastrophe” or other mechanisms. As discussed in this article, when administered at clinically relevant doses, cancer chemotherapeutic agents trigger a high proportion of MNGCs in solid tumor-derived cell lines (especially those lacking wild-type p53 function) that remain viable and can give rise to tumor repopulating progeny. Shockingly, only a single multinucleated giant cancer cell has been shown to be sufficient to cause metastatic disease when grafted under the skin of an animal [
115]. Cancer cells undergoing SIPS in response to chemotherapeutic agents can also escape from the growth-arrested state and give rise to tumor-repopulating progeny.
Accordingly, targeting growth-arrested cancer cells might represent an effective therapeutic strategy. To this end, Crescenzi et al. [
61] reported that downregulating either ATM or p21 in cancer cells that have undergone SIPS in response to chemotherapeutic drugs results in their demise. For targeting MNGCs for destruction, different approaches have been reported to be effective. These include viral infection [
104] and treatment with pharmacological inhibitors of different members of the BCL-XL/BCL-2 pathway [
130]. In addition, we have recently demonstrated that the apoptosis activators sodium salicylate (an inhibitor of the p38 MAPK) or dichloroacetate (a modulator of glucose metabolism) also kill MNGCs under conditions that have little or no effect on parental (mono-nucleated) cells [
108]. The results of these proof-of-principle in vitro experiments are encouraging and warrant further studies with animal models.
7. Mutational Signatures in Human Cancers
The aforementioned reports concluding that the creation of MNGCs following chemotherapy might represent a survival mechanism for cancer cells involved studies not only with cultured cells and animal models, but also with patient specimens. Other studies, however, also reporting extensive experimental and clinical data, have concluded that this response might reflect a favorable therapeutic outcome (e.g., [
131]). Similarly, the conclusions that SIPS might represent a favorable [
97,
98,
99,
100,
101] or unfavorable [
97,
98,
99,
100,
101,
102,
103] therapeutic outcomes have also been based on extensive experimental/clinical data. Such apparently conflicting observations might not be entirely unexpected when considering the distinct mutational types in aging and cancer.
The advent of next-generation sequencing technologies has enabled large-scale sequencing of all protein-coding exons (whole-exome sequencing) or even whole cancer genomes (whole-genome sequencing) in a single experiment [
132,
133,
134,
135,
136,
137,
138]. These sequencing efforts have enabled the identification of many thousands of mutations per cancer which provided sufficient power to detect different mutational patterns or “signatures.” Each biological perturbation or “mutational process” (e.g., tobacco smoke, sunlight exposure, deamination of DNA bases) is shown to leave a characteristic “mark” or mutational signature on the cancer genome (reviewed in [
135]) (
Figure 3).
Each mutational signature is defined by: (i) the type of genomic injury that has occurred as a result of a diversity of exogenous and endogenous genotoxic stresses; (ii) the integrity of DNA repair and other aspects of the DNA damage surveillance network that were successively activated; and (iii) the strength and duration of exposure to each mutational process. Additionally, as pointed out by Helleday et al. [
135], “
cancers are likely to comprise different cell populations (that is, subclonal populations), which can be variably exposed to each mutational process; this promotes the complexity of the final landscape of somatic mutations in a cancer genome. The final “mutational portrait,” which is obtained after a cancer has been removed by surgery and then sequenced, is therefore a composite of multiple mutational signatures.”
Thus, as previously anticipated, these large-scale sequencing technologies coupled with bioinformatic and computational tools for deciphering the “scars” (signatures) of mutational processes have demonstrated significant variability in the mutation landscape in cancers of the same histological type. Application of such approaches might similarly unfold the molecular basis for the fate of growth-arrested cancer cells in terms of death versus survival. This might in turn set the stage for designing novel therapeutic strategies for specifically targeting growth-arrested cancer cells before they will have the opportunity to generate tumor-repopulating progeny.