Primary high-grade serous ovarian cancer cells are sensitive to senescence induced by carboplatin and paclitaxel in vitro

Various types of normal and cancer cells undergo senescence in response to carboplatin and paclitaxel, which are considered the gold standard treatments in ovarian cancer management. Surprisingly, the effect of these drugs on ovarian cancer cell senescence remained unknown. The experiments were conducted on primary high-grade serous ovarian cancer cells. Molecular markers of senescence were evaluated using cytochemistry and immunofluorescence. Cell cycle distribution was analyzed using flow cytometry. Expression of cyclins and signaling pathways was tested using western blot. Telomere length and telomerase activity were measured using qPCR, and the colocalization of telomeres with DNA damage foci using immuno-FISH. Oxidative stress-related parameters were quantified using appropriate fluorescence probes. Production of cancerogenic agents was analyzed using qPCR and ELISA. Carboplatin applied with paclitaxel induces senescence of ovarian cancer cells in vitro. This activity was reflected by permanent G2/M growth arrest, a high fraction of cells expressing senescence biomarkers (SA-β-Gal and γ-H2A.X), upregulated expression of p16, p21, and p53 cell cycle inhibitors, and decreased expression of cyclin B1. Neither telomere length nor telomerase activity changed in the senescent cells, and the majority of DNA damage was localized outside telomeres. Moreover, drug-treated cancer cells exhibited increased production of STAT3 protein, overproduced superoxide and peroxides, and increased mitochondrial mass. They were also characterized by upregulated ANG1, CCL11, IL-6, PDGF-D, TIMP-3, TSP-1, and TGF-β1 at the mRNA and/or protein level. Our findings imply that conventional chemotherapy may elicit senescence in ovarian cancer cells, which may translate to the development of a cancer-promoting phenotype, despite the inability of these cells to divide.


Background
Although cellular senescence is a postmitotic state in which cells irreversibly lose the ability to divide and is classically considered the ultimate fate of normal cells bearing extensive and irreparably deteriorated DNA [1,2], in recent years, the view that senescence may also be elicited in cancer cells subjected to radio-and chemotherapy has been increasingly supported [3]. The list of agents capable of triggering drug-induced premature senescence (DIPS) of cancer cells is long and includes aphidicolin, bleomycin, cisplatin, doxorubicin, etoposide, mitoxantrone, retinols, hydroxyurea, carboplatin combined with docetaxel, and many others [4]. The senescence-promoting activity of conventional chemotherapeutics has been documented in several tumors, such as breast, lung, prostate, and colon carcinoma [4], irrespective of their p53 profile [5]. Among various female genital tract malignancies, epithelial ovarian cancer is the most lethal [6]. Because the majority of ovarian cancer cases are diagnosed at advanced stages of the disease, patients are recommended to undergo maximal surgical debulking followed by systemic administration of carboplatin (CPT) and paclitaxel (PCT), which is considered the best chemotherapeutic option [7]. Despite different modes of action, both drugs have been found to jeopardize normal cell functioning, leaving a mark on their energetic metabolism, driving their inflammatory responses and transition to the reactive stroma, generating DNA injury, and inducing apoptosis and senescence [8,9]. Surprisingly, the potential impact of CPT and/or PCT on primary ovarian cancer cell senescence has never been systematically tested either in vitro or in vivo. Taking this into account, we designed a study in which the effect of a CPT-PCT mixture applied at clinically relevant doses [10] to a wide range of molecular senescence-associated parameters in cultured primary high-grade serous ovarian cancer cells (HGSOCs) was examined and discussed.

Cell cultures and experimental conditions
Primary high-grade serous ovarian cancer cells (HGSOCs) were isolated from tumors obtained from 14 different patients during cytoreductive surgery. The tissues were cut into small fragments of similar weight and then placed in 0.05% trypsin-0.02% EDTA for 20 min at 37 °C with shaking. The cells were grown in RPMI 1640 supplemented with l-glutamine (2 mM) and 20% FBS. Their cancerous nature was confirmed according to the expression of epithelium-related antigen (MOC-31) and CA125 antigen. The study was approved by an institutional ethics committee (consent number 578/18), and all patients gave their informed consent.
Cells from the first passage (young cells) were incubated with 50 µM carboplatin (CPT) combined with 25 nM paclitaxel (PCT) (both from Cayman Chemical, Ann Arbor, MI, USA) in standard growth medium for 72 h at 37 °C. Then, the cells were rinsed with Hanks' balanced salt solution and incubated in standard growth medium for 24 h to recover. Then, the culture was passaged and incubated in standard conditions for an additional 7-10 days. After incubation, the cells were considered senescent, which was confirmed by their morphology, inability to proliferate, and the presence of senescence biomarkers. The concentrations of drugs were established in pilot experiments on the basis of clinical observations [10] and experimental data [11]. The time of exposure was selected according to our pilot studies which allowed us to identify the length of treatment which does not induce massive cell death but results instead in certain metabolic effects associated with cellular senescence.

Analysis of cell cycle distribution and inhibitory proteins
The distribution of cells (10 000) in the cell cycle was determined using flow cytometry as described previously [14]. The fraction of cells expressing p16, p21, and p53 cell cycle inhibitors was quantified using immunofluorescence. Cells were fixed with 4% paraformaldehyde and 100% methanol, permeabilized with 0.1% Triton-X in PBS for 10 min, and blocked with a solution containing 1% bovine serum albumin, 22.5 mg/mL glycine, and 0.1% Triton-X in PBS for 30 min. Afterwards, specimens were incubated with antibodies directed against p16 (#ab108349; Abcam), p21 (#2947, Cell Signaling Technology, Danvers, MA, USA), and p53 (#2527, Cell Signaling), prepared at 1:200 dilution, for 24 h, at 4 °C. Then, the cells were extensively washed and subjected to DyLight 488 IgG (#ab96899, Abcam; 1:500) for 1.5 h at room temperature. After the incubation, the specimens were preserved with Fluoroshield medium with DAPI (Abcam) and inspected under a fluorescence microscope. By moving from left to right and from top to bottom, 500 cells were analyzed.

Quantification of telomeres and telomerase
Telomeres were quantified using an Absolute Human Telomere Length Quantification qPCR Assay Kit (ScienCell, Carlsbad, CA, USA), whereas telomerase activity was measured using a TRAPEZE XL Telomerase Detection Kit (Merck). Both the assays were performed strictly according to manufacturers' instructions.

Analysis of the senescence-associated secretory phenotype (SASP)
Total RNA was isolated using TRI Reagent (Merck) and then purified using a Cleanup RNA Concentrator kit (A&A Biotechnology, Gdynia, Poland). The concentration of RNA was estimated according to an optical density at 260 nm. Its purity was determined by a 260/280 nm absorption ratio (> 1.  Table 1. To evaluate the cell secretome, conditioned media were collected from young and senescent HGSOCs, as described previously [16]. Proteins were quantified using the appropriate DuoSet Immunoassay Development kits (R&D Systems, Abingdon, UK).

Statistics
Statistical tests were conducted using GraphPad Prism 6.00 (GraphPad Software, San Diego, USA). The means were compared using the Wilcoxon test. The data are shown as the means ± SEM. Differences with a p-value less than 0.05 were considered significant.

CPT-PCT combination induces senescence in primary HGSOCs
Primary HGSOCs isolated from tumors from chemotherapy-naïve patients were exposed to CPT (50 µM) combined with PCT (25 nM) to verify their sensitivity to druginduced premature senescence (DIPS). Two kinds of senescent biomarkers were analyzed: the activity of SA-β-Gal and the activation of the DNA damage response marked by the presence of γ-H2A.X and 53BP1 foci. The experiments showed that first passage cell cultures undergoing DIPS permanently lost the ability to proliferate, which was manifested by their inability to reinitiate division for three weeks. At the same time, they developed hypertrophic morphology (Fig. 1c), and the majority of them displayed activity of SA-β-Gal and the presence of γ-H2A.X foci. Notably, the fraction of SA-β-Gal-and γ-H2A.X-positive cells increased significantly during senescence, as in the young, earlypassage cells the percentage of cells bearing the senescence signature was far below 20%. Conversely, the fraction of 53BP1-positive cells was unexpectedly high in young cells and did not change any further upon treatment with CPT/PCT (Fig. 1a-c).

Characteristics of DIPS-associated cell cycle arrest
Flow cytometry analysis of cells stained with propidium iodide revealed that CPT-PCTtreated HGSOCs showed arrested growth in the G2/M phase of the cell cycle, which was reflected by a twofold increase in the fraction of cells identified in this phase with the simultaneous reduction by a half of the fraction of the DNA-replicating cells identified in the S phase (Fig. 2a). This coincided with upregulated expression of p16, p21, and p53 proteins, as evidenced by immunofluorescence. In young cultures, less than 10% of cells displayed expression of the tested cell cycle inhibitors, indicating vigorous proliferation of these cells. When the cells became senescent, the fraction of p16-, p21-and p53-positive cells increased several-fold, in line with the reduced proliferative capacity of these cells (Fig. 2b). The magnitude of this increase in these cell cycle inhibitors was comparable. With changes in cell cycle-promoting cyclins, DIPS cells displayed decreased expression of cyclin B1 and unaltered expression of cyclin D1, as evidenced by western blot analysis. Remarkably, the pattern of the cyclin expression was highly heterogenous,  ANG1  GCC TAC ACT TTC ATT CTT CCAGA  TCT TCC TTG TGT TTT CCT TCCAT   CCL11  CCC TTC AGC GAC TAG AGA GC  CAG CTT TCT GGG GAC ATT TG   CXCL12  GTG CCC TTC AGA TTG TAG  which was plausibly a result of the primary nature of tested cells and several patientspecific and genetic variables (Fig. 2c, d and Additional file 1).

Analysis of the engagement of telomeres and telomerase
Three parameters were quantified to establish the role of telomeres in DIPS of HGSOCs. These were telomere length, telomerase activity, and colocalization of telomeres with γ-H2A.X foci followed by a Pearson's correlation analysis. Comparative measurements of young cells with DIPS cells showed that telomeres in young cells are relatively short (~ 3 kbp) and that they do not shorten after exposure to CPT + PCT (Fig. 3a). At the same time, a catalytic subunit of telomerase, hTERT, has unaltered activity (Fig. 3b). A search for colocalization of γ-H2A.X foci with telomeres showed that the level of DNA damage within these structures is relatively low in young cells (far below 10%) and does not increase significantly after cell exposure to the drugs (Fig. 3c, e). This observation was consistent with Pearson's assessment of the correlation between γ-H2A.X foci and telomeres in deconvoluted pictures, which yielded coefficients of 0.15 ± 0.01 and 0.16 ± 0.02 for young and DIPS cells, respectively (Fig. 3d).

DIPS-associated changes in signaling molecules
Six signaling molecules known to be engaged in cellular senescence, that is, AKT, AP-1, ERK1/2, NF-κB, p38 MAPK, and STAT3, were tested using immunoblotting in young and DIPS cancer cells. The protein levels of the five molecules did not change during senescence, except for STAT3; in the case of this transcription factor, the level of STAT3α markedly increased along with the appearance of the second band representing the STAT3β isoform (Fig. 4a, b and Additional file 1).

Formation of reactive oxygen species and mitochondrial metabolism during DIPS
DIPS cancer cells produce higher amounts of mitochondrial superoxide and cellular peroxides than young cells, as shown by fluorescent staining of cells with MitoSOX Red (Fig. 5a) and DHR (Fig. 5b), respectively. This effect was accompanied by unaltered values of ΔΨ m potential (JC-1 fluorescence; Fig. 5c) and elevated mitochondrial mass (NAO fluorescence; Fig. 5d).

Discussion
In this study, we provide the first comprehensive evidence that the CPT-PCT mixture, applied to primary HGSOCs in vitro using an experimentally established algorithm and clinically relevant doses [10], triggers the development of a phenotype similar to that of cellular senescence. This finding agrees with the study by Li et al., in which they observed the prosenescence impact of cisplatin. Notably, however, that study was performed on an established ovarian cancer cell line, A2780 [17]. The dynamics of senescence development in primary HGSOCs was similar to that reported in other tumor types [4]. Ovarian cancer cells undergoing drug-induced senescence (DIPS) displayed classical biomarkers of this phenomenon, including the activity of SA-β-Gal [12] and the presence of γ-H2A.X foci [18] (Fig. 1a, b). The recruitment of the second element of the DNA damage response, p53 binding protein 1 (53BP1) [19], did not change in line with γ-H2A.X (Fig. 1a, b), and one may theorize that senescence persistence was enabled by the ineffective repair of DNA double-strand breaks, plausibly due to insufficient 53BP1controlled nonhomologous end-joining (NHEJ) [20]. Taking into account the knowledge about the effects of platins, both carboplatin [21] and cisplatin [22], on DNA integrity, one cannot exclude that senescence-associated DNA damage in CPT-PCT-treated cells may also include single-strand breaks. At least theoretically, the lack of 53BP1 elevation could also be associated with the fact that 94% of HGSOCs display mutated P53, which might translate into decreased reactivity of its binding protein [23]. G2/M growth arrest of HGSOCs, whose senescence was evoked by CPT-PCT (Fig. 2a), is a typical hallmark of cancer cells undergoing DIPS [24] and probably reflects interference of CPT-generated modifications of purines (covalent binding to the N7 sites) with cell replication machinery [25]. The activation of the G2/M checkpoint was also reported in breast cancer cells, whose senescence was provoked by paclitaxel [26], as well as in lung [27] and mammary gland cancer cells [28]. In the latter two cases, this pattern of cell distribution coincided with depressed cyclin B1, which agrees with our findings (Fig. 2c). Intriguingly, G2/M exit from the cell cycle distinguishes DIPS ovarian cancer cells from their counterparts undergoing replicative senescence, in which senescence-associated growth arrest occurred in the G1 phase and involved elevated cyclin D1, whose level in DIPS cells [29] and other cellular models undergoing G2/M phase block [30] was unchanged. Representative bands corresponding to certain signaling molecules analyzed using western blot (a) and their densitometric analysis with GAPDH used as the reference protein (b). In order to reduce the risk of incorrect observations due to cellular hypertrophy in senescent cells, SDS-PAGE was performed using samples corresponding to the experimentally optimized number of cells (3-5 × 10 4  This difference may result from the fact that the replication-dependent senescence of cancer cells involves telomere attrition, which is causatively linked with G1 exit [31]. In the case of DIPS cells, the process is clearly telomere-independent, as it suggests an unchanged telomere length, stable hTERT, and predominantly non-telomeric localization of DNA damage foci (Fig. 3e). Interestingly, the pattern of changes in cell cycle inhibitors, that is, the upregulated expression of p16, p21, and p53 (Fig. 2b), is similar in both types of ovarian cancer cell senescence and is consistent with observations of other cancer [32] and normal cell types rendered senescent by CPT and PCT [33]. The elevation of these inhibitory proteins revealed in DIPS ovarian cancer cells is also in line with the G2/M growth arrest of these cells; recently, similar coincidence has been described in fibroblasts, whose premature senescence was induced by ultraviolet A irradiation [34].
Regarding signaling molecules potentially controlling DIPS in HGSOCs, only signal transducer and activator of transcription 3α (STAT3α [35]) protein increased significantly (Fig. 4a). Notably, in replicatively senescent ovarian cancer cells, the STAT3 level was among the pathways whose level declined. Our observation corresponds, however, with reports in which hyperphosphorylation of STAT3 was found in fibroblasts treated with cisplatin [36] and PCT [37]. Furthermore, the levels of AKT, AP-1, and ERK1/2 in DIPS cells remained unchanged, which also distinguishes them from replicatively senescent ovarian cancer cells, in which the levels of these molecules was elevated [29]. In addition to plausible STAT3 induction by CPT-PCT, some role in this process, additionally facilitated by partial mitochondrial distribution of this factor [38], may also involve reactive oxygen species (ROS) [39]. Indeed, DIPS ovarian cancer cells generate more mitochondrial superoxide (Fig. 5a) and cellular peroxides (Fig. 5b) than young cells, which is generally in line with the pro-oxidative effects of these drugs [28,40]. At the same time, in contrast to replicatively senescent ovarian cancer cells, this intensified ROS release was not associated with retrograde, compensatory biogenesis of mitochondria, driven by declining values of inner membrane potential (ΔΨ m ) (Fig. 5c) [29,41]. Because DIPS cancer cells seem to maintain their ability to generate ATP (unaltered ΔΨ m ), the more probable scenario is that their elevated mitochondrial content (Fig. 5d) is not the direct reason for increased ROS production but rather a byproduct of CPT-PCT activity. This view supports research on A549 cancer cells in which PCT strongly induces peroxisome proliferator-activated receptor-gamma coactivator (PGC)-1α, the master regulator of mitochondrial biogenesis [42].
The increased level of STAT3 in the presence of stable NF-κB and p38 MAPK, i.e., the leading molecules governing the senescence-associated secretory phenotype (SASP) [43], may also contribute to changes in the DIPS cell transcriptome and secretome. In fibroblasts senesced upon exposure to PCT, the activation of STAT3 was linked with hypersecretion of IL-6 [33,44], i.e., one of the agents whose mRNA and protein were upregulated in DIPS ovarian cancer cells. According to a recent report, the IL-6-JAK/ STAT3 interplay was found to correlate with increased ovarian tumor cell growth and resistance to chemotherapy [45].

Conclusions
The findings presented in this report improve the understanding of the effect exerted by conventional chemotherapeutics (here, CPT and PCT) on the tumor microenvironment. More precisely, they show that drugs treated as the gold standard in ovarian cancer therapy affect not only the metabolism and behavior of normal (peritoneal) cells [8] but may also, paradoxically, contribute to the development of cancer-promoting features in ovarian cancer cells by provoking their premature senescence. Because knowledge about the biological and clinical outcomes of ovarian cancer cell senescence is still very limited, a detailed comparative analysis of the plausible cancer-promoting effects of replicatively senescent and DIPS cancer cells using both in vitro and in vivo models is urgently required. To make the whole picture of the discussed processes clearer and more reliable, a separate group of these tests should include cells from patients in whom CPT and PCT were administered in vivo as elements of their therapy. Another critical issue that has to be additionally verified is the concentration of drugs and time of their administration. The results of this study were based on doses matching as much as possible the in vivo conditions. The time of drug activity is however more complex because in vivo it is a sum of their actual delivery and further activity in tissues associated with patientspecific pharmacokinetics, whereas in vitro it was fixed at the value of 72 h.