Polyploid giant cancer cells and ovarian cancer: new insights into mitotic regulators and polyploidy

Abstract

Current first-line treatment of patients with high-grade serous ovarian cancer (HGSOC) involves the use of cytotoxic drugs that frequently lead to recurrent tumors exhibiting increased resistance to the drugs and poor patient survival. Strong evidence is accumulating to show that HGSOC tumors and cell lines contain a subset of cells called polyploidy giant cancer cells (PGCCs) that act as stem-like, self-renewing cells. These PGCCs appear to play a key role in tumor progression by generating drug-resistant progeny produced, in part, as a consequence of utilizing a modified form of mitosis known as endoreplication. Thus, developing drugs to target PGCCs and endoreplication may be an important approach for reducing the appearance of drug-resistant progeny. In the review, we discuss newly identified regulatory factors that impact mitosis and which may be altered or repurposed during endoreplication in PGCCs. We also review recent papers showing that a single PGCC can give rise to tumors in vivo and spheroids in culture. To illustrate some of the specific features of PGCCs and factors that may impact their function and endoreplication compared to mitosis, we have included immunofluorescent images co-localizing p53 and specific mitotic regulatory, phosphoproteins in xenografts derived from commonly used HGSOC cell lines.

Cytotoxic drugs frequently used in ovarian cancer treatment impact tumor cell cycle progression and the emergence of polyploid cells where genetic and epigenetic events ultimately lead to drug-resistant diploid progeny.

Introduction

High-grade serous ovarian cancer (HGSOC) is a highly heterogeneous disease with respect to morphology, genetic mutations, and sites of origin [1, 2]. However, one unifying feature of HGSOC is the presence of polyploid giant cancer cells (PGCCs) [3]. Although the ovarian surface epithelium has long been considered the source of HGSOC, recent studies provide convincing evidence that the fallopian tube epithelium is also a source of the primary malignant lesions in response to specific mutations [1]. For example, depletion of Pten selectively in mouse fallopian tube epithelium leads to the formation of multicellular tumor spheroids (MTS) that eventually detach, exfoliate from the fallopian tube, and attach to the ovary and mesentery [4–6]. These MTS contain a population of heterogeneous large cells, designated as cancer stem-like cells (CSCs), that are morphologically similar to PGCCs, express different levels of stem cell markers, and grow aggressively in Matrigel and generate tumors in mouse xenografts [6]. Diverse stress factors derived from follicular fluid may also contribute to tumor growth and metastasis [1, 7]. Because HGSOC is most often not detected until advanced stages when it has spread throughout the peritoneal cavity, it is difficult to manage. Hence HGSOC remains a deadly disease for which currently there are no cures with promising outcomes.

Although cancer chemotherapy medications such as cisplatin, carboplatin, and paclitaxel are most often used as first round treatments, overall patient survival remains poor. Most patients exhibit recurrent, drug-resistant tumor progression that is more severe than that of the primary tumors and involves the generation of PGCCs and the release of drug-resistant, morphologically distinct diploid daughter cells [8–10] (Figure 1A–D). This lack of remission is related, in part, to the low doses of cytotoxic drugs that are tolerated in patients compared to the higher doses that can lead to the death (apoptosis) of tumor cells in culture [9]. Lack of remission is also related to many changes in HGSOC cell genetic, epigenetic, and metabolic regulatory mechanisms [3, 11–14]. The very high mutation rates (>90%) in the tumor protein 53 (TP53; p53) also contribute to ovarian cancer tumor malignancy. While some p53 mutations are null, others exhibit gain of function (GOF) activity, the most prevalent of which are p53-R248Q, p53-R175H, and p53-R273H [15]. These p53 GOF mutants are expressed in the HGSOC cell lines OVCAR3, TYK-NU, and OVCA420, respectively. These cell lines exhibit distinct morphologies and metastatic progression in vivo and contain multinucleated PGCCs [3, 9, 10, 12] (Figure 2A and B).

Figure 1.

PGCCs in the HGSOC HEY cells (WTp53) and in patient tumor tissue. (A) Longitudinal profiling of PGCCs in HGSOC cells exposed to the cytotoxic drug paclitaxel (PTX). Numbers of diploid tumor cells decrease by apoptosis between days 1 and 9, whereas the number of PGCCs (>4C) increases to 80% of the total cell population during the same time interval. These cells have escaped apoptosis and undergo endoreplication to spawn diploid daughter cells after ~day 21. (B) Morphological changes of HEY cells labeled with H2BGFP imaged by conventional microscopy. Thick black arrows, mononuclear and polynuclear PGCCs; thin black arrows, diploid daughter cells; a thick white arrow at D13 designates a PGCC undergoing multipolar mitosis. (C) Immunofluorescent images of HEY cells: upper panels, subcellular localization of Aurora A kinase (green), α-tubulin (red), and DAPI (blue) and lower panels, γ tubulin (green) and DAPI (blue) during mitosis and in budding PGCCs. Mitosis: white arrows, centrosomes; PGCCs with budding: (c) yellow arrow, multiple budding nuclei; multiple budding nuclei; (d) white arrow designates a tubulin bridge that connects to a daughter cell (yellow arrow). (e) White arrow, multiple fragmented nuclei. (f) Diffuse γ tubulin in PGCC (Adapted from Niu et al., Oncogenesis 5:e281, 2016; Figure 1 and 4) Creative Commons License (http://creativecommons.org/licenses/by/4.0). (D) H&E-stained images of HGSOC tumor tissue from a patient before chemotherapy (left) and following 6 cycles of paclitaxel- and carboplatin-based chemotherapy (right). Note the prominent presence of a PGCC and heterogeneous cell morphology in the tumor tissue after chemotherapy compared to the lack of PGCCs and more uniform morphology in tissue before chemotherapy. (Adapted from Niu et al. [13]; Figure 7A; Creative Commons License (http://creativecommons.org/licenses/by-nc-sa/4.0).

Figure 2.

Multinucleated PGCCs are present in human HGSOC cell lines. (A) H&E-stained sections of xenografts of HGSOC cell lines OVCAR3 (p53-R248Q), TYK-NU (p53-R175H) and OVCA420 (p53-R273H) grown on the omentum of Foxn1^−/− immune compromised mice. The tumors exhibit distinct morphologies: OVCAR3, solid epithelial/mesenchymal-like; TYK-Nu, vascular-like; OVCA420, exclusively epithelial. The distribution of immune cells is also distinct for each tumor type: They are excluded from the solid OVCAR3 tumors, whereas in TYK-Nu and OVCA420 cells, scattered patches of immune cells are observed around the tumors and within the omental fat pad. t: tumor; i: immune cells; s: stromal cells. The inserts in (A) show multinucleated PGCCs. (B) Immunofluorescent images of sections of OVCAR3 xenografts showing large multinucleated PGCCs that express high levels of mutant p53 (green). Several cells also exhibit signs of stress as noted by the presence of γH2AX (red), a marker of DNA damage. (C) Immunofluorescent image of OVCAR3 cells in culture showing that cells are at different stages of the cell cycle. P53 (green) is nuclear in all cells that are not dividing, including PGCCs that are multinuclear (a) and (b). PGCC during abnormal cytokinesis (c). P53 is redistributed in cells undergoing normal mitosis where the condensed chromosomes are either at the midbody (d) or are undergoing cytokinesis (e). See Appendix for Supplemental Information describing the experimental procedures and histological/immunofluorescent imaging procedures for Figures 2–5).

Accumulating evidence indicates that PGCCs lie at the heart of cancer cell immortality: Amend et al. [16] argue convincingly that although the number of PGCCs is limited in tumors, these cells are the “keystone actuators” of cancer by regulating tumorigenesis, metastasis, and therapy resistance. In tumors of ovarian cancer patients, PGCCs are induced in response to clinically relevant concentrations of cytotoxic drugs such as cisplatin and paclitaxel [13]. These PGCCs persist in recurrent tumors and act as self-renewing CSCs by utilizing alternative asymmetric mitotic, endoreplication mechanisms that lead to the generation of morphologically distinct drug-resistant diploid daughter cells [10, 14] (Figure 1A′). Although they may enter a senescent phase at some point, the PGCCs are not strictly senescent cells but express many stem cell markers and inflammatory molecules [8–10, 13, 14, 16, 17]. Remarkably, a single PGCC can generate a spheroid in culture or lead to tumor formation in vivo [10, 13, 14].

Thus, understanding the factors that control mitosis and alternative endoreplication mechanisms utilized by tumor cells, and PGCCs in particular, is of critical relevance for designing novel therapeutic strategies for HGSOC patients undergoing chemotherapy. This mini review summarizes some of the recent studies of PGCCs and the potential impact of these cells in ovarian cancer progression and in response to therapeutic drugs. In particular, the main focus of the review is on pathways that regulate mitosis in normal cells and cancer cells and how these pathways are altered by cytotoxic stress, the appearance of PGCCs, and the alternative endoreplication pathway of cell division that is operative in these PGCCs.

Regulators of mitosis

Mitosis is a complex, highly regulated process during which transcription is largely suppressed and levels of many cellular kinases are increased to regulate the proper transmission of the replicated chromosomes [18, 19]. Mitosis can be disrupted by stress factors leading to the emergence of malignant cells, PGCCs, and altered forms of replication. During mitosis key cellular components, notably α-/γ-tubulin and microtubules undergo redistribution and repurposing (Figure 1C). Therefore to understand the inception of PGCCs and their persistence, it is critical to first understand the factors that regulate the S- and M-phases of the cell cycle, which rely on processes of phosphorylation and hence are controlled by specialized kinases. These kinases include the cyclins B1 (CCNB1) and E1 (CCNE1) and the cyclin-dependent kinases 1 (CDK1) and 2 (CDK2), which regulate the initiation of mitosis and Aurora kinase B (AURKB), a mitotic checkpoint kinase that ensures correct chromosome segregation. Although the regulation of mitotic cell cycle progression has been analyzed extensively, recent studies have identified new regulatory mechanisms that control mitosis and may also be relevant to our understanding of cell replication in cancer cells and PGCCs present in HGSOC, in particular.

Mitotic progression requires that several checkpoints become silenced. It is well known that WT p53 is critical for controlling cell cycle arrest in response to cytotoxic stress that impacts DNA repair mechanisms and has been justly called “the guardian of the genome” [20]. One of the key targets of p53 transactivation is the kinase inhibitor p21 (p21CIP1; CDK1 interacting protein 1; CDKN1A) that suppresses the activity of CCNB1/CDK1 required for mitotic entry. As a consequence, most cells then undergo apoptosis. Recent studies have shown that increased activation of p53 and CDKN1A1 indirectly regulate CCNB1/CDK1 activities by inducing a multifactor complex designated DREAM [21]. DREAM (downstream regulatory element antagonist modulator) is a transcriptional repressor that currently is considered a master regulator of cell cycle progression and together with p53 and CDKN1A, and it controls the expression and functions of over 100 proteins that regulate the cell cycle [21]. One of the factors that promotes cell cycle progression and is negatively regulated by DREAM is the microtubule-associated serine/threonine kinase-like protein (MASTL) [21].

MASTL also known as Greatwall kinase (GWL; Drosophila) acts as a master regulator of the mitotic (M-phase) phase of the cell cycle and is overexpressed in a high percentage of HGSOC patient tumors as well as in breast cancer [22]. MASTL controls the activity of the protein phosphatase PP2A-B55 indirectly by regulating endosulfines that include endosulfine alpha (ENSA) and cAMP-regulated phosphatase (ARPP19) [22]. PP2A-B55 in turn controls the phosphorylation status of CCNB1 and hence the activity of the CCNB1/CDK1 complex and mitosis initiation. High levels of MASTL and loss of PP2A-B55 activity enhance proliferation and alter mitotic functions that lead to DNA instability and genetic aberrations [22]. HGSOC patient tumor samples express high levels of MASTL, ARPP19, and CCNB1 [22].

Mitosis is also controlled by the proper activation and localization of the kinases Aurora kinase A (AURKA) and AURKB [14]. AURKA localizes to centrosomes (Figure 2), whereas AURKB is a checkpoint kinase known to regulate the attachment of the spindle microtubules to the kinetochores and is overexpressed in many cancers [18]. AURKB also interacts with p53 during different phases of the cell cycle. During interphase, AURKB is nuclear where it phosphorylates p53 at Ser183, Thr211, and Ser215. This facilitates the degradation of p53 and hence blocks p53 regulation of genes that impair mitosis and lead to apoptosis, such as CDKN1A [18]. During mitosis, AURKB interacts with p53 at the centromeres. Although the function of p53 at this site is not yet known, it may contribute to the error-correction role of AURKB that is known to ensure proper kinetochore functions [23]. Importantly, chromosome instability can be corrected by enhanced expression of AURKB [23]. Conversely, pharmacological inhibition of AURKB in cancer cells expressing WT p53 enhances expression of p53-regulated genes and growth inhibition [18].

High levels of a specific form of phosphorylated β-catenin (p-CΤNNΒ1-Ser33, Ser37, Thr41) are also elevated in HGSOC cells [24]. Although it is well known that when p-CΤNNΒ1 is phosphorylated by CDK1 at Serine 45 and subsequently by glycogen synthase kinase 3 (GSK3) at S33, S37, and T41, it is targeted primarily for ubiquitination and degradation. However, recent elegant studies provide strong evidence that p-CΤNNΒ1–S33, S37, and T41 play a specific role in centrosomal functions of mitotic cells (Figure 3A and B) [24]. Moreover, inhibitor studies show that this phosphorylation is totally independent of GSK3 [24]. Rather, it is sustained by the NIMA-related protein kinase 2 (NEK2) that is downstream of Polo-Like Kinase 1 (PLK1) [24, 25]. Of note, both NEK2 and PLK1 are also targets of the DREAM-MASTL pathways [21]. Specifically, there is evidence that in tumor cells certain phosphorylated forms of CΤNNΒ1 are stable and distinctly nuclear [24, 25]. Moreover, NEK2-mediated p-CΤNNΒ1–S33, S37, and T41 appear to be required for centrosome disjunction at mitosis [25]. Our results indicate that p-CΤNNΒ1 levels are elevated selectively in PGCCs in HGSOC cells (Figure 3B), indicating that the degradation pathway is bypassed in these cells and is likely mediated by the NEK2 kinase. This may explain, in part, the potent role of distinct phosphorylated forms of p-CΤNNΒ1 in the progression of ovarian cancer and other cancer cells [25].

Figure 3.

Localization of factors to centrosomes during mitosis and endoreplication. (A) Immunofluorescent images showing subcellular localization of p-CTNNB1 and its co-localization with factors regulating replication in normal cells during mitosis and in PGCCs undergoing endoreplication. (A) p-CTNNB1 (S33, S37, T41) co-localizes with α-tubulin and γ-tubulin at the centrosomes of normal mitotic cells. (Adapted from Mbom et al. [25] (Figure 4A); Creative Commons License (http://creativecommons.org/licenses/by-nc-sa/3.0). (B) Localization of p-CTNNB1-S33, S37, T41 (green), CTNNB1 (red), and KRT8 (blue) in xenografts of OVCAR3 cells on the omentum of Foxn1^−/− mice. KRT8 (blue) and nonphosphorylated CTNNB1 (red) mark the plasma membranes. White box in nonmitotic OVCAR3 cells p-CTNNB1 appears as distinct small dots in the cytoplasm. White box surrounding a OVCAR3 PGCC shows p-CTNNB1 (green) is distinctly more abundant in an where the condensed chromosomes (white) are not tightly aligned.

It is important to note that CΤNNΒ1 “wears many hats” and that its functions during mitosis are distinct from the roles of CΤNNΒ1 as a downstream mediator of the canonical Wnt signaling pathway. In response to Wnt signaling, CΤNNΒ1 regulates the transcription of target genes controlling cell proliferation and differentiation. In cancer cells, elevated Wnt signaling can impact cell adhesion, metastasis, stem cell self-renewal as well as drug resistance and is related to poor prognosis [26–28]. Analyses of the TCGA [27] datasets and recent studies show that many HGSOC tumor patient specimens and HGSOC cells express high levels of WNT7A and WNT7B [29]. Moreover, deletion as well as overexpression of WNT7A impacts tumor cell proliferation in culture and tumor progression in vivo [29]. Wnt signaling also regulates the tumor microenvironment as shown by recent studies in which WNT5A regulated the functions of stromal cells and immune cells leading to changes in the expression of specific cytokines that then impacted tumor growth [30]. In spheroid co-cultures comprised M2 macrophages and ALDH+, CD133+ cancer stem cells derived from HGSOC OVCAR3 cells, macrophage-derived Wnt ligands (WNT5b) enhanced the stemness of the OVCAR3-derived cancer stem cells by increasing Wnt ligands (WNT2) that in turn drive M2 macrophage activation. This positive feedback loop leads to tumor chemoresistance, invasiveness, and immunosuppression [31]. Thus, Wnt signaling and CΤNNΒ1 impact many aspects of ovarian tumor cell function, the totality of which is too vast to incorporate into this mini review but is covered in other reviews [26, 27].

MASTL not only controls the entry of cells into mitosis but also can activate GSK3 by converting inactive phospho-GSK3 (p-GSK3) into a nonphosphorylated “active” form. Although the precise pathway by which this occurs is not yet known, it does not involve ENSA/ARPP19 and is independent of PP2A-B55 [22]. Active GSK3 not only phosphorylates a subset of cellular CTNNB1 to target it for degradation but also the phosphatase PHLPP that is subsequently ubiquitinated and degraded [22]. Because PHLPP dephosphorylates and inactivates p-AKT-S473, the phosphorylation and degradation of PHLPP removes an inhibitory mechanism and leads to increased levels of p-AKT-S473. Persistent activation of AKT can further promote proliferation by many pathways including its ability to phosphorylate and inactivate GSK3, and thereby promoting nuclear mitotic actions of specific phosphorylated forms of CΤNNΒ1 [22].

The mitogen and stress-activated kinase MSK1 (RPS6KA5) is a chromatin-associated kinase involved in many cellular processes [32, 33]. In mitotic cells, the levels of MSK1 phosphorylated at T581 (p-MSK1-T581) are markedly increased and impact chromosome condensation and segregation, in part by phosphorylating histone H3 Ser10 (p-H3S10) that co-localizes with condensed chromosomes and is critical for their formation [33]. Phospho-MSK1-T581 is also elevated in OVCAR3 PGCCs undergoing endoreplication and co-localizes with mutant p53-R248Q, notably at the multiple centrosomes present in the PGCCs (Figure 4A and B). The levels of p-MSK1 are also elevated in nonmalignant mitotic cells such as ovarian granulosa and uterine epithelia cells (data not shown), and thus, p-MSK1 serves as a general marker of mitotic cells.

Figure 4.

Localization of p53, p-MSK1-T581, and KRT8 in xenografts of OVCAR3 cells on the omentum of Foxn^−/− mice. (A) p53 (red) is distinctly nuclear in nondividing tumor cells, whereas in PGCCs undergoing endoreplication p53 (red) and p-MSK1 (green) are elevated and associated with the mitotic spindle region but do not co-localize with the condensed chromosomes (white) or the cell membrane (blue). These cells exhibit a morphology typically observed in PGCCs where the cells are large and chromosomes are not tightly aligned. (B) p53 (red) and p-MSK1 (green) co-localize (yellow) to multiple centrosomes (white arrows), a defining feature of endoreplication in PGCCs, but do not co-localize to the condensed chromosomes (blue).

In nonmitotic cells, in marked contrast to cells in mitosis, p-MSK1-T581 and histone H3 exert different functions related to transcriptional activation of specific genes [34]. Specifically, activation of stress-induced DNA damage pathways leads to the activation of p38 MAPK. Activated p38 MAPK mediates the phosphorylation of both p53 at S15 and MSK1 at T581, leading to histone H3 S10 phosphorylation and the transcription of the cell cycle inhibitor CDKN1A as well as the other cell cycle inhibitors GADD45 and PUMA [32]. While increased levels of CDKN1A would suppress rather than enhance proliferation by intersecting with the WT p53-p21-DREAM complex, this mechanism would not be operative in p53 null cells or in cancer cells expressing p53 GOF mutants. For example, levels of p-MSK1 T581 are highly elevated in replicating cancer cells [33], including HGSOC OVCAR3-R248Q cells present in xenografts in Foxn1^−/− mice (Figure 4A and B). However, it is important to note that p-MSK1 is also present but at much lower levels in the nonmitotic cells suggesting that p-MSK1-T581 has multiple functions in cancer cells (Figure 4A and B).

Steroid hormones impact ovarian cancer [1, 7, 35]. Estradiol is associated with more rapid tumor progression, whereas progesterone is more likely to suppress tumor growth [36–38]. Androgens also have a demonstrated impact on ovarian cancer [7, 39]. Although the androgen receptor is present in many HGSOC patient samples (The Human Tissue Atlas), the mechanisms by which androgens act remain unclear. Of relevance to this review, because estradiol can rapidly activate (phosphorylate) MSK1 leading to increased breast cancer cell proliferation [40], it is possible that steroid-dependent activation of MSK1 may also impact HGSOC cell proliferation. Thus, p-MSK1 and steroids may provide a potent auxiliary pathway to regulate proliferation and activate mitosis in hormone responsive HGSOC tumors. The effects of steroids may also depend on the mutant status of p53. For example, studies in Kras/Pten mutant mice have documented that estradiol can potently stimulate the metastasis of tumors that are also null for p53 (Kras/Pten/p53−/−) but is less effective in Kras/Pten tumors expressing WT p53 [41, 42]. Moreover, estradiol accelerates tumor progression of in HGSOC ALST cells that express WT p53 but only after they have been engineered to also express the GOF mutant p53-R273H [43]. Likewise, SKOV3 cells that are null for p53 exhibit rapid tumor progression in response to estradiol but only when they are engineered to also express p53-R273H. Importantly, steroids also regulate functional changes in tumor-associated stromal cells and immune cells, such as the release of potent cytokines that impact tumor cell functions [44–49].

Lastly, it is important to note that specific cellular regulators of metabolism are selectively increased during mitosis. The alpha subunit of 5′-adenosine monophosphate-activated protein kinase (AMPKα) not only impacts metabolic functions in cells but also exerts specific functions in mitosis where it is expressed at high levels and is autophosphorylated at threonine 172 [50]. Moreover, phospho-AMPKα–T172, like p53 and p-CTNNB1, specifically localizes to the centrosome in cells entering mitosis and then relocates to other parts of the mitotic apparatus during prophase, metaphase, telophase, and at cytokinesis [50, 51]. Likewise, levels of phospho-mTOR-Ser2481 (mammalian target of rapamycin) are elevated in mitotic cells and distributed to similar areas of the mitotic apparatus as phospho-AMPKα-T172 [52]. mTOR phosphorylated on S2448 is also present during mitosis, but its localization is more diffuse during each stage of mitosis [52]. In OVCAR3 cells undergoing replication, mTOR-2448 co-localizes with the GOF p53-R248Q on mitotic spindles but not to condensed chromosomes (Figure 5).

Figure 5.

Co-localization of p53 and p-mTOR-S2448 in OVCAR3 PGCCs. In nondividing tumor cells, p53 (red) is distinctly nuclear, whereas levels of p-mTOR (green) are low. However, in endoreplicating PGCCs, p53 and p-mTOR-S2448 levels are elevated but do not co-localize to the condensed chromosomes (white) or to KRT8-positive cell membranes (blue). Rather, they appear to be associated with the spindle region of the cells where the condensed chromosomes are not tightly aligned.

Regulators of PGCC asymmetric replication: endoreplication

Amazingly, polyploid cells are common throughout the biological kingdom [53]. The underlying mechanism that leads to the formation of polyploid cells involves an altered form of genome replication called endoreplication in which DNA is replicated but mitosis fails to occur [53]. It is associated with diverse developmental processes and growth, and also with ancestral survival mechanisms in response to environmental stress, leading to adaptation and a form of “immortality” [10, 11, 54]. Numerous studies indicate that polyploid cells (and maybe MTS [4]) exhibit enhanced resistance to stress, such as DNA damage and accommodate gene alterations that are not tolerated in diploid cells [11, 53]. This resistance to stress appears to involve mechanisms that buffer the DNA damage response and may be related to reduced p53 damage activity or the presence of p53 mutants that are associated with increased proliferation and with polyploidy [11, 53].

Polyploidy can be irreversible or reversible. Irreversible polyploidy is observed in nurse cells and follicle cells of Drosophila melanogaster and is essential for oogenesis and early development in the insect embryo: disruption of the endoreplication processes causes sterility [53]. Irreversible polyploidy and endoreplication also occur in cells in healthy mammalian tissues such as in over 80% of adult mouse liver cells, in 64 N megakaryocytes that produce thrombocytes, or in trophoblast giant cells of the placenta where this process is essential for embryo development [11, 53]. By contrast, tumor cells can respond to cytotoxic stressors by undergoing reversible polyploidy (Figure 1A–C). This process involves the establishment of an endoreplicative state during which the cells enlarge to form PGCCs in which DNA repair and rearrangement of the genome can occur [3, 8–10]. Ultimately, the PGCCs transition to a mitotic stage by undergoing depolyploidization and release large numbers of diploid progeny cells that may have acquired an altered genomic karyotype and/or epigenetic program underlying their increased drug resistance, stemness, and inflammatory secretory activity [8, 10, 13, 14, 53–55].

Remarkably, recent studies have shown that the stem-like (ALDH+) PGCCs can form spheroids when grown in stem cell medium in culture and can generate tumors in vivo [2, 13, 14]. The appearance of these PGCCs can be induced in HGSOCs in culture (HEY WT p53; SKOV3 p53 null; IGROV-1 WT p53; and OVCA433 p53 WT/mutant) by exposure to cytotoxic drugs [10, 14]. Because cells expressing WT p53 as well as those null or heterozygous for p53 can become PGCCs, it appears that their emergence is not strictly dependent on p53 status. Rather, their appearance may be related to the suppression of p53 activity associated with genomic instability and may underlie the frequent mutations of p53 observed in ovarian cancer.

In addition, several recent studies have documented that the emergence of PGCCs in response to drugs is nonlinear with respect to drug dosage and time of exposure [2, 9, 10]. Specifically, in response to clinical doses of carboplatin, cisplatin, or paclitaxel (<10 μM), the number of PGCCs (>4 N) rapidly increased between days 1 and 14 of culture, while the number of diploid (2 N) cancer cells decreased by apoptosis during the same time interval [2, 10, 14] (Figure 1A and B). At Day 14, the PGCCs exhibit elevated levels of CDKN1A and specific cytokines such as IL6, IL8, and CCL20, the expression of which appears to be p53 independent. Thus, what factors and pathways regulate the increases in these factors remain to be identified and what roles they play in endoreplication are not yet understood. By days 27–30 of culture, endoreplication within the PGCCs leads to the formation of diploid 2N daughter cells (Figure 1A and B) by processes of budding or bursting. What regulates these processes is not entirely clear but they may be related, in part, to changes in AURKA and AURKB [14, 53].

The 1–14–27-day intervals in culture during which PGCCs increase in number in response to cytotoxic stressors and then spawn diploid progeny have been documented in several laboratories using different ovarian cancer cell lines. The 21-day burst of PGCCs remarkably also coincides with the time period during which many xenograft tumors switch from a slow-growing state to a rampant, exponential phase of growth [6, 49]. Thus, it is tempting to speculate that xenograft tumors formed after injecting HGSOC cells are derived, at least in part, from “dormant” PGCCs or stem cells that were present in the population of injected cells, which then underwent endoreplication to produce diploid daughter cells. Clearly, this needs to be documented experimentally. However, that a single PGCC can give rise to drug-resistant tumors in vivo strongly suggests that PGCCs play critical functions in tumor formation and metastatic progression [2, 14].

Based on molecular studies of SKOV3 cells exposed to carboplatin, the daughter cells derived from the PGCCs are genetically and functionally distinct from both the parental SKOV3 cells and the SKOV3 PGCCs. The daughter cells are more drug resistant and express increased levels of drug transporters [10]. They also express higher levels of cytokines that can promote tumor growth and regulate the tumor microenvironment. The daughter cells also show lower levels of CDH1. Likewise, PGCCs isolated from HGSOC HEY cells that express WT p53 differ from parental HEY ovarian cancer cells: HEY PGCCs exhibit elevated levels of CDK4, CCND1, CCNE1, and CCNB1 compared to control tumor cells [2, 14]. OVCAR3 cells are known to express elevated levels of CCNE1 and exhibit chromosome instability, the latter of which can be corrected by expressing AURKB [23] or disrupting CDK2 [56]. OVCAR3 cells are highly heterogeneous and contain numerous sets of chromosomes in culture and in xenograft samples on the omentum (Figures 2A and B, 3B, 4A and B, and 5). The presence and functions of PGCCs need to be more carefully analyzed in all tumors and in response to clinically relevant doses of drugs [9].

Because most studies of PGCCs have been done on a limited number of HGSOC cell lines that are available and primarily derived from ascites, it is also critical to determine when PGCCs are induced in primary ovarian tumors and what factors (steroid, cytokines, hypoxia) respond to cytotoxic stressors to initiate the appearance of malignant cells (Figure 3). As indicated above, recent studies [4, 6] showed that MTS can be induced in primary fallopian tube epithelial cells by depletion of Pten, or by increased expression of a p53 GOF mutant, or by both and that these MTS contain a population of “large” cancer stem cells (PGCCs) that form tumors in Matrigel and in vivo. These results provide intriguing evidence that the formation of MTS in primary cells may occur by mechanisms that impose stress on cell survival and that lead to the emergence of PGCCs. More studies are clearly needed to determine specific stress-mediated factors and events that transform normal cells to a malignant state.

As mentioned above, the endoreplication cycle that is functional in PGCCs has been highly conserved throughout evolution and appears in many different cell types [10, 12, 14, 53]. Specifically, in PGCCs endoreplication involves successive rounds of DNA replication in S-phase but cytokinesis does not occur until daughter cells are released [53]. Endoreplication allows for epigenetic and genetic changes to occur for adaptation, and specifically in PGCCs exposed to chemotherapy medication, to compensate for cytotoxic stress and provide drug resistance. In comparison with the known pathways and factors that regulate mitosis in normal and some cancer cells (Figures 6 and 7), less is known about which cell cycle regulators are present and are operative during the process of endoreplication. The Notch signaling pathway appears to be involved and may regulate the marked increase in the S-phase cyclin, CCNE1 and the decrease in CCNB1 that is a critical driver of the M-phase [53]. Recently, the PGCC endoreplication cycle has been likened to embryonic blastomere division where, amazingly, early mammalian embryonic development is associated with considerable aneuploidy and genetic instability [10, 54, 57, 58]. Many fertilized eggs are genetically impaired and are described as in “chaos” [57, 58]. Also remarkable, genes that are expressed in meiosis and during blastomere formation are often increased in the PGCCs during endoreplication and daughter cell formation, including MOS, RAS, and MYC [10, 13, 14, 54, 55]. Thus, PGCCs appear to represent a dedifferentiated state similar to that observed in embryonic blastomeres.

Figure 6.

Schematic of signaling pathways that regulate mitosis and are altered in response to mutations, cytotoxic stressors, changes in p53 status, and steroids. Activating interactions are denoted by the solid black arrows; possible activating interactions by dashed black arrows; repressive interactions by the solid red lines. Stress-activated default pathways can lead to cell death by apoptosis and senescence, whereas stress-induced cell survival is mediated by endoreplication in polyploid cells. Activation of the default pathways is dependent on the dose and duration of stress factors such as cytotoxic drugs. Stress-induced polyploidy occurs in cancer cells expressing WT, null, or mutant p53 and may involve p-MSK1-regulated induction of p21 that can occur independently of p53 and may involve phosphorylation of H3S10. (Model based on references cited in the review.)

Figure 7.

Schematics of mitosis and endoreplication: redistribution and repurposing of key phosphorylated regulatory factors. Mitosis: Mitosis is initiated when the single centriole of an interphase cell undergoes splitting and duplication. This is coincident with DNA replication and eventually centrosome maturation. These processes involve different cyclins and CDKs as well as AURKA, PLK1, NEK2 and p53. Mitosis is completed with centrosome separation and spindle formation during metaphase, chromosome separation at telophase/anaphase, and ultimately cytokinesis. These processes are highly dependent on specific kinases, the regulation of microtubule functions, and the redistribution and phosphorylation of p53. Note: (1) the specific localization of p-p53-S15 and pCTNNB1-S33, S37, T41, NEK2, and γ-tubulin on the centrosomes along with p-AMPKα-T172, p-mTOR-S2481, and AURKB; (2) the specific localization of pMSK1-T581 and pH3S10 on the condensed chromosomes; and (3) the general localization of these factors to the mitotic spindles. Green, centriole; blue, centrosome; black, spindle microtubules; multiple colors, condensed chromosomes; central lines, midbody and cytokinesis. Endoreplication: When cancer cells are stressed by cytotoxic drugs, they can enter a process called endoreplication that involves the formation and function of PGCCs. Ultimately, the PGCCs release (via budding) diploid daughter cells that are highly drug resistant. Note that (1) P53 along with pp-CTNNB1, pp-MTOR, and pp-MSK1 is associated with the multiple centrosomes, whereas AURKA is notably low in PGCCs; (2) many kinases associated with spindle fibers in mitotic cells also co-localize with the randomly aligned spindle fibers in the PGCCs; and (3) less is known about how the PGCCs ultimately undergo budding and cytokinesis that give rise to diploid daughter cells. (Model based on references cited in the review.)

Relatively little is known about the expression and subcellular localization of the key mitotic regulators in endoreplication, such as the DREAM complex, MASTL, p-MSK1-T581, p-CTNNB1-Ser33, 37, Thr41, p-mTOR-S2481 and p-mTOR-S2448, and p-AMPKα-T172 (Figures 6 and 7). Omental tumors derived in vivo from xenografts of OVCAR3 cells (p53-R248Q), TYK-Nu (p53-R175H), or OVCA420 (R273H) cells contain p53-positive PGCCs that are multinucleated (Figure 2A–C). The OVCAR3 tumors also have cells with multiple centrosomes to which p-MSK1 T581 and p53-R248Q are associated (Figure 4A and B) and contain replicating metaphase cells in which p53-R248Q overlaps with p-MSK1-T581 and p-mTOR-2448 (Figures 4A and B and 5). Whether p53-R248Q is phosphorylated on serine15 by MSK1 as is WT p53 is not known. Because PGCCs are generated in cancer cells that lack functional p53, it is also possible that p-MSK1-T581 leads to increased levels of CDKN1A in PGCCs by p53-independent mechanisms that may also involve H3S10 [10, 32–34]. p-CTNNB1-S33, T37, and S41 appears to be localized in clusters throughout the nucleus of mitotic OVCAR3 cells (Figure 3B). By contrast, the chromosomal passenger proteins, AURKA and AURKB, that regulate cytokinesis are notably low/absent in PGCCs in HEY, SKOV3, and OVCAR433 cells compared to the parental cells and daughter cells [13, 14]. The different expression of these kinases may underlie a major functional difference between the normal mitotic cycle and the endoreplicative cycle in PGCCs. In addition, the distinct centrosomal localization of γ-tubulin in normal mitotic cells is absent in the PGCCs. Rather γ-tubulin is randomly localized in each multinucleated PGCCs [14] (Figure 1C). The localization of alpha tubulin is also distinctly different [14].

Conclusions and questions

The regulation of mitosis is composed of exquisite and intertwined, conserved biological processes of phosphorylation and proteolysis that require precise repurposing of key regulatory factors and pathways in a cell [19]. Successful mitosis requires correct spatiotemporal regulation of microtubules, the end results being the faithful generation of identical daughter cells. The nuclear membrane breaks down, specific factors such as p53-S15 and p-CΤNNΒ1-S33, T37, S41 attach to the centrosome leading to its disjunction. Microtubules are organized and attach to condensed chromosomes at the kinetochore where AURKB plays a critical role in spindle integrity. Phospho-MSK1 phosphorylates histone H3-S10 at the metaphase plate to stabilize the structure. P-MSK1 and H3-S10 appear to have additional functions in nonmitotic cells. P-p53-S15 may play a key role at the centromere along with many other phosphorylated factors (Figures 6 and 7).

However, it is clear that stress factors, such as cytotoxic insults, can trigger the response of cells that leads to the remarkable phenomenon of polyploidy that appears to be an ancestral mechanism for protecting cells and the genome. By providing a time interval for genomic repair, epigenetic events, and dedifferentiation, the endoreplication process in polyploid cells ultimately leads to the spawning of differentiated diploid progeny that is more resistant to toxic insults than the parental cells (Figures 1 and 7). Mechanistically, how the endoreplication process repurposes the mitotic machinery remains to be clearly understood, calling for further studies. How can we ever reconcile current drug treatments with the knowledge that the outcomes are often worse? New approaches are needed to prevent or disrupt the polyploidy state and endoreplication in early tumorigenesis or to terminally differentiate polyploid cells. The most important question that still awaits answers is: will PGCCs be targeted to control cancer cell proliferation and tumor progression?

Authors’ contributions

All authors contributed to the writing of the review and design of the figures (JSR, NRC, RBL); H&E and immunofluorescent images of HGSOC cells in vivo and in culture were generated by NRC.

Data availability statement

The data underlying this article will be shared on reasonable request to the corresponding author.

Conflict of interest: None.