Dormant cancer cells: programmed quiescence, senescence, or both?

The most intuitive case of the involvement of senescence (a form of proliferative arrest) in cancer (a disease of hyperproliferation) is its role as an opposing force to tumor growth. Several groups have reported roles for senescence in protecting against transformation and tumor initiation. Studies in various tumor types have documented the presence of senescent cells in premalignant lesions, which are then rare once the lesions have progressed to malignancy [51‐55], implying that senescent cells present a barrier for tumorigenesis. Although these studies are correlational, they are largely in agreement with the data on oncogene-induced senescence in normal or premalignant cells in which the senescence response halts the proliferation of cells at risk of transformation, preventing tumor formation. This response was first described in the context of HRAS^V12 expression and has since been shown to occur in response to gain-of-function mutations in other oncogenes, including BRAF, AKT, and the cell cycle regulators E2F1 and cyclin E. Similarly, loss-of-function mutations in the tumor suppressors PTEN and NF1 can trigger the same response in vitro and in vivo [28, 51‐57]. Moreover, senescent cells are characterized by the SASP, which contains a host of immune-modulating cytokines and chemokines. Similar to its roles in tissue development and patterning, the SASP may induce immune-mediated clearance of precancerous cells, which may act to further prevent tumor initiation (Fig. 1) [58‐60]. Cells that avoid senescence, successfully transform, and go on to form detectable tumors will inevitably be treated with chemotherapy or radiation, both of which have been shown to drive TIS, halting tumor growth and promoting regression through immune-mediated clearance [10, 60, 61]. Although these mechanisms of senescence are intuitively tumor antagonistic, the same mechanisms may actually be tumor promoting depending on the context.

Although the role of cellular senescence in organismal aging is not entirely clear, it is now well understood that aged organisms accumulate senescent cells in a variety of tissues. As cancer is often described as a ‘disease of aging,’ with incidence rates increasing dramatically with age, it seems likely that senescence and cancer are related. In fact, Rudolf Virchow touched on this over 100 years before the discovery of cellular senescence with his ‘irritation theory’ from 1858 [62], where he proposed that chronic irritation and inflammation are drivers of tumorigenesis. Harold Dvorak subsequently characterized tumors as ‘wounds that do not heal’. Senescent cells, known to play a role in wound healing and inflammation through the action of the SASP, induce remodeling of the microenvironment that, while beneficial for wound healing, may contribute to the formation of neoplasia [63, 64]. In this way, the same process that halts proliferation of cells at risk for transformation may paradoxically drive tumor initiation. Halazonetis and colleagues proposed that oncogene-induced replication stress and subsequent DNA damage may drive not only OIS but also the genomic instability critical for the accumulation of mutations in cells that promote cancer development [47]. While it is unclear under which circumstances cells would be induced to undergo senesce versus undergoing transformation and tumor formation, it is likely that the same upstream events can promote both fates, further linking senescence to cancer development.

Much of the research on senescence in cancer has focused on elucidating the roles of aging stromal components in cancer progression. Senescent fibroblasts and other stromal components have been shown to induce proliferation, motility, and angiogenesis in cancer cells, largely through the autocrine and paracrine actions of SASP components (Fig. 1) [65‐68]. In the same way, however, components of the SASP secreted from a cancer cell that has undergone OIS or TIS may also impact other cancer cells in a paracrine fashion and impact tumor progression and response to therapy. In vitro, media conditioned by cells that have undergone TIS can stimulate proliferation, induction of hallmark epithelial-mesenchymal transition (EMT) markers, and increased invasiveness [69]. Additionally, responses to radiation and chemotherapy can be affected by cancer cells undergoing TIS. Senescence in response to these therapies can inhibit apoptosis and even lead to faster relapse than in tumors unable to undergo senescence [60, 70, 71].

The identity of dormant cancer cells remains elusive in general, and therefore, studies directly identifying dormant cancer cells as senescent are scarce. The definition of a ‘dormant cancer cell’ requires that the cell be able to reawaken and proliferate to repopulate a tumor. Therefore, by the classical definition of senescence as permanent proliferative arrest, cells that have undergone senescence could not contribute to delayed relapse as a dormant seed. For a senescent cancer cell and a dormant cancer cell to be one in the same hinges on the controversial claim that senescence is reversible. Data on the reversibility of senescence do exist, although none are definitive, and all are context dependent (Fig. 1) [9, 10, 72‐78].

Some of the studies linking senescence to tumor initiation could also be relevant to the reawakening and colonization of dormant cells. This may be especially true in TIS, as dormant tumor cells are relevant to disease progression during and after treatment. Senescent cells persisting after therapy would certainly be a component contributing to minimal residual disease (MRD). One study, for example, directly links a treatment-induced ‘senescence-like state’ to dormancy in the context of EGFR targeted therapy in EGFR-mutant non-small-cell lung cancer [79].

If senescence is truly reversible in certain contexts, then senescent cells could plausibly be related to dormant cells based on several shared characteristics. First, the duration of dormancy seen in patients and in preclinical models lends itself to the idea that these dormant cells have undergone a durable arrest, much like senescent cells. Other types of arrested cells, such as those in quiescence, could presumably resume proliferation more quickly than what is seen in dormancy models, although progression of a solitary dormant cell to so-called ‘tumor-mass dormancy’ controlled by the immune system could account for this time before detectable relapse. Senescent cells, like dormant cancer cells, are also refractory to apoptosis in response to therapies that target proliferative cells. This resistance is critical for dormant cells to survive long enough in a patient undergoing treatment to go on to seed relapse. Induction of a reversible senescence program could represent an adaptive mechanism for cells in response to chemotherapeutic stress.

In addition to the senescent cell playing the role of the dormant seed for delayed metastasis, a possibly less controversial role for senescent cells in dormancy and delayed relapse could be due to paracrine signaling from the senescent cells’ SASP. As has been proposed in the initial transformation and formation of a primary tumor, perhaps signals from a senescent tumor cell at the metastatic site remodel the microenvironment to drive proliferation, angiogenesis, and successful colonization of other tumor cells that have arrived at the site. Indeed, inflammation has been shown to push dormant cells toward re-entry into the cell cycle [18, 80‐82], as has age-related senescence in the stromal microenvironment [68].

A subpopulation of stem cell-like quiescent or slow-cycling cells have been identified in multiple cancer types and termed cancer stem cells (CSCs). CSCs from tumors of different tissue origins are phenotypically diverse and are identified by different molecular markers [84], but all display the unifying feature of slow or absent cell division and the presence of tumor-formation capacity [85‐87]. Their role(s) in tumor progression are still controversial, but several studies indicate that they are critical drivers of tumor growth [84]. For example, slow-cycling CSCs in melanoma have been demonstrated to be required for tumor maintenance, as when this population is disrupted, serial xenografts fail to proliferate and metastatic outgrowth is hampered [88]. In addition to slow proliferation and tumor formation capacity, CSCs display features of EMT and are therefore of interest in the context of metastatic dormancy and delayed colonization [8‐11, 89]. One study demonstrated that ~ 65% of disseminated breast cancer cells in the bone marrow display a CSC phenotype [90]. When compared to the < 10% of these CSCs that are found in the primary tumors, this is strong evidence for a potential role for CSCs in disseminated disease. Regarding delayed relapse, Meng et al. showed that circulating CSCs can be found in the blood of patients who have been disease-free for 20 years, indicating that CSCs may act as a latent reservoir of cells with tumor-forming capacity for late colonization [91].

Another hallmark that is shared by dormant cancer cells and CSCs is the ability to avoid destruction by the immune system [8, 10, 92]. CSCs appear to adopt programs employed by tissue-resident stem cells to avoid immune recognition. One such mechanism is quiescence itself, which promotes the downregulation of antigen presentation, preventing recognition and killing by T cells and natural killer cells [92, 93].

Despite these connections, debate remains over whether CSCs represent a truly dormant population, as populations can be heterogeneous in their proliferative state [10] and have been shown to be proliferative, albeit at slow rates. Therefore, CSCs may more accurately described as ‘slow-cycling’ vs. quiescent [8]. More evidence is needed to determine whether this debate is merely semantic and whether CSCs can play a role in cancer dormancy regardless.

In cancer, “dormancy” and “quiescence” have overlapping, broad definitions — both describe cancer cells that are held in a reversible state of cell cycle arrest — and these terms are therefore commonly used interchangeably. Similar to cancer dormancy, however, the quiescent state is not a monolith. Entry into the quiescent state can be driven by intrinsic cell cycle–related programs or in response to new or newly absent signals from the microenvironment [83]. Induction of G0 arrest is commonly attributed to the actions of cyclin-dependent kinase inhibitors (CDKis), especially p27. Increased p27 activity reduces the ability of CDKs to phosphorylate retinoblastoma protein (Rb), and hypophosphorylated Rb inhibits E2F1 from transcribing target genes necessary for cell cycle progression [83]. Increased p27 expression is commonly used as a marker of quiescence induction, and fluorescent p27 reporters can be used to track live cell entry to and exit from quiescence [94]. Especially in vivo, environmental cues are critical in the entry into, maintenance of, and exit from quiescence. Tissue-resident stem cells, for example, reside in specialized niches that contain signals from other cells in the tissue, the extracellular matrix, and soluble factors from local vasculature that drive proliferation or quiescence in a context-dependent manner. A cancer cell also interacts with and responds to its surroundings, which change dramatically upon entrance into the metastatic cascade and eventual deposition in a foreign tissue. Niche factors- both cellular and acellular- in the secondary site may be unfamiliar, and newly present or absent signals from this microenvironment can drive the cell to enter quiescence or promote the maintenance of quiescence if the disseminated cell was previously dormant. Changes to the secondary microenvironment, some driven by the cancer cell itself, may then release the cell into a proliferative state toward eventual colonization and disease relapse [8]. Various recent studies have begun to illustrate just how complex these communication networks are, describing how interactions with immune cells, extracellular matrix components, tissue-specific stromal cells, and environmental factors like oxidative stress can all impact a DTC’s entry into or exit from the dormant state [68, 95‐104]. With the increasing understanding that every facet of the cancer cell’s environment impacts its capability to actively divide, it becomes clear that even cells from the same tumor, once disseminated, will have much different fates. The majority will die, but those that survive circulation and extravasation at secondary sites and are destined for a period of dormancy will enter quiescence through different mechanisms depending on the integration of signals from the microenvironment. Even in vitro, where quiescence is commonly induced by either contact inhibition, serum deprivation, or loss of adhesion, each of these methods induces G0 arrest through different mechanisms, and the resulting quiescent cells display different gene expression signatures [83, 105‐107].

In addition to p27, the CDKis p21, p16, and p57 also play a role in quiescence. Interestingly, these are all also implicated in driving senescence. p21, for instance, is a major player in the maintenance of hematopoietic stem cell quiescence and is also considered a canonical marker of senescent cells [35, 81]. Moreover, the transcriptional signatures of quiescent and senescent cells overlap (Table 1) [108]. These common molecular pathways between quiescence and senescence point to the fact that the two states are not wholly unrelated. In fact, cells that have been growth arrested for long periods differ from those arrested for shorter periods in the ‘depth’ of their quiescence. Long-term quiescent cells require more intense growth stimulation and take longer to re-enter the cell cycle [83, 109], and deep quiescence appears to be a transitional state to senescence. The quiescence/deep quiescence/senescence transition is controlled by lysosomal function, which acts as a dimmer switch [110]. These overlaps in phenotype and molecular control indicate that quiescent and senescent states may not be completely binary but lie along a graded spectrum (Fig. 2).